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.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/seccomp.h>
44#include <linux/rcupdate.h>
45#include <linux/rculist.h>
46#include <linux/rtmutex.h>
47
48#include <linux/time.h>
49#include <linux/param.h>
50#include <linux/resource.h>
51#include <linux/timer.h>
52#include <linux/hrtimer.h>
53#include <linux/kcov.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#include <linux/magic.h>
61#include <linux/cgroup-defs.h>
62
63#include <asm/processor.h>
64
65#define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
66
67/*
68 * Extended scheduling parameters data structure.
69 *
70 * This is needed because the original struct sched_param can not be
71 * altered without introducing ABI issues with legacy applications
72 * (e.g., in sched_getparam()).
73 *
74 * However, the possibility of specifying more than just a priority for
75 * the tasks may be useful for a wide variety of application fields, e.g.,
76 * multimedia, streaming, automation and control, and many others.
77 *
78 * This variant (sched_attr) is meant at describing a so-called
79 * sporadic time-constrained task. In such model a task is specified by:
80 * - the activation period or minimum instance inter-arrival time;
81 * - the maximum (or average, depending on the actual scheduling
82 * discipline) computation time of all instances, a.k.a. runtime;
83 * - the deadline (relative to the actual activation time) of each
84 * instance.
85 * Very briefly, a periodic (sporadic) task asks for the execution of
86 * some specific computation --which is typically called an instance--
87 * (at most) every period. Moreover, each instance typically lasts no more
88 * than the runtime and must be completed by time instant t equal to
89 * the instance activation time + the deadline.
90 *
91 * This is reflected by the actual fields of the sched_attr structure:
92 *
93 * @size size of the structure, for fwd/bwd compat.
94 *
95 * @sched_policy task's scheduling policy
96 * @sched_flags for customizing the scheduler behaviour
97 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
98 * @sched_priority task's static priority (SCHED_FIFO/RR)
99 * @sched_deadline representative of the task's deadline
100 * @sched_runtime representative of the task's runtime
101 * @sched_period representative of the task's period
102 *
103 * Given this task model, there are a multiplicity of scheduling algorithms
104 * and policies, that can be used to ensure all the tasks will make their
105 * timing constraints.
106 *
107 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
108 * only user of this new interface. More information about the algorithm
109 * available in the scheduling class file or in Documentation/.
110 */
111struct sched_attr {
112 u32 size;
113
114 u32 sched_policy;
115 u64 sched_flags;
116
117 /* SCHED_NORMAL, SCHED_BATCH */
118 s32 sched_nice;
119
120 /* SCHED_FIFO, SCHED_RR */
121 u32 sched_priority;
122
123 /* SCHED_DEADLINE */
124 u64 sched_runtime;
125 u64 sched_deadline;
126 u64 sched_period;
127};
128
129struct futex_pi_state;
130struct robust_list_head;
131struct bio_list;
132struct fs_struct;
133struct perf_event_context;
134struct blk_plug;
135struct filename;
136struct nameidata;
137
138#define VMACACHE_BITS 2
139#define VMACACHE_SIZE (1U << VMACACHE_BITS)
140#define VMACACHE_MASK (VMACACHE_SIZE - 1)
141
142/*
143 * These are the constant used to fake the fixed-point load-average
144 * counting. Some notes:
145 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
146 * a load-average precision of 10 bits integer + 11 bits fractional
147 * - if you want to count load-averages more often, you need more
148 * precision, or rounding will get you. With 2-second counting freq,
149 * the EXP_n values would be 1981, 2034 and 2043 if still using only
150 * 11 bit fractions.
151 */
152extern unsigned long avenrun[]; /* Load averages */
153extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
154
155#define FSHIFT 11 /* nr of bits of precision */
156#define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
157#define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
158#define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
159#define EXP_5 2014 /* 1/exp(5sec/5min) */
160#define EXP_15 2037 /* 1/exp(5sec/15min) */
161
162#define CALC_LOAD(load,exp,n) \
163 load *= exp; \
164 load += n*(FIXED_1-exp); \
165 load >>= FSHIFT;
166
167extern unsigned long total_forks;
168extern int nr_threads;
169DECLARE_PER_CPU(unsigned long, process_counts);
170extern int nr_processes(void);
171extern unsigned long nr_running(void);
172extern bool single_task_running(void);
173extern unsigned long nr_iowait(void);
174extern unsigned long nr_iowait_cpu(int cpu);
175extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
176
177extern void calc_global_load(unsigned long ticks);
178
179#if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
180extern void cpu_load_update_nohz_start(void);
181extern void cpu_load_update_nohz_stop(void);
182#else
183static inline void cpu_load_update_nohz_start(void) { }
184static inline void cpu_load_update_nohz_stop(void) { }
185#endif
186
187extern void dump_cpu_task(int cpu);
188
189struct seq_file;
190struct cfs_rq;
191struct task_group;
192#ifdef CONFIG_SCHED_DEBUG
193extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
194extern void proc_sched_set_task(struct task_struct *p);
195#endif
196
197/*
198 * Task state bitmask. NOTE! These bits are also
199 * encoded in fs/proc/array.c: get_task_state().
200 *
201 * We have two separate sets of flags: task->state
202 * is about runnability, while task->exit_state are
203 * about the task exiting. Confusing, but this way
204 * modifying one set can't modify the other one by
205 * mistake.
206 */
207#define TASK_RUNNING 0
208#define TASK_INTERRUPTIBLE 1
209#define TASK_UNINTERRUPTIBLE 2
210#define __TASK_STOPPED 4
211#define __TASK_TRACED 8
212/* in tsk->exit_state */
213#define EXIT_DEAD 16
214#define EXIT_ZOMBIE 32
215#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
216/* in tsk->state again */
217#define TASK_DEAD 64
218#define TASK_WAKEKILL 128
219#define TASK_WAKING 256
220#define TASK_PARKED 512
221#define TASK_NOLOAD 1024
222#define TASK_NEW 2048
223#define TASK_STATE_MAX 4096
224
225#define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
226
227extern char ___assert_task_state[1 - 2*!!(
228 sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
229
230/* Convenience macros for the sake of set_task_state */
231#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
232#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
233#define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
234
235#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
236
237/* Convenience macros for the sake of wake_up */
238#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
239#define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
240
241/* get_task_state() */
242#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
243 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
244 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
245
246#define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
247#define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
248#define task_is_stopped_or_traced(task) \
249 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
250#define task_contributes_to_load(task) \
251 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
252 (task->flags & PF_FROZEN) == 0 && \
253 (task->state & TASK_NOLOAD) == 0)
254
255#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
256
257#define __set_task_state(tsk, state_value) \
258 do { \
259 (tsk)->task_state_change = _THIS_IP_; \
260 (tsk)->state = (state_value); \
261 } while (0)
262#define set_task_state(tsk, state_value) \
263 do { \
264 (tsk)->task_state_change = _THIS_IP_; \
265 smp_store_mb((tsk)->state, (state_value)); \
266 } while (0)
267
268#define __set_current_state(state_value) \
269 do { \
270 current->task_state_change = _THIS_IP_; \
271 current->state = (state_value); \
272 } while (0)
273#define set_current_state(state_value) \
274 do { \
275 current->task_state_change = _THIS_IP_; \
276 smp_store_mb(current->state, (state_value)); \
277 } while (0)
278
279#else
280
281/*
282 * @tsk had better be current, or you get to keep the pieces.
283 *
284 * The only reason is that computing current can be more expensive than
285 * using a pointer that's already available.
286 *
287 * Therefore, see set_current_state().
288 */
289#define __set_task_state(tsk, state_value) \
290 do { (tsk)->state = (state_value); } while (0)
291#define set_task_state(tsk, state_value) \
292 smp_store_mb((tsk)->state, (state_value))
293
294/*
295 * set_current_state() includes a barrier so that the write of current->state
296 * is correctly serialised wrt the caller's subsequent test of whether to
297 * actually sleep:
298 *
299 * for (;;) {
300 * set_current_state(TASK_UNINTERRUPTIBLE);
301 * if (!need_sleep)
302 * break;
303 *
304 * schedule();
305 * }
306 * __set_current_state(TASK_RUNNING);
307 *
308 * If the caller does not need such serialisation (because, for instance, the
309 * condition test and condition change and wakeup are under the same lock) then
310 * use __set_current_state().
311 *
312 * The above is typically ordered against the wakeup, which does:
313 *
314 * need_sleep = false;
315 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
316 *
317 * Where wake_up_state() (and all other wakeup primitives) imply enough
318 * barriers to order the store of the variable against wakeup.
319 *
320 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
321 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
322 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
323 *
324 * This is obviously fine, since they both store the exact same value.
325 *
326 * Also see the comments of try_to_wake_up().
327 */
328#define __set_current_state(state_value) \
329 do { current->state = (state_value); } while (0)
330#define set_current_state(state_value) \
331 smp_store_mb(current->state, (state_value))
332
333#endif
334
335/* Task command name length */
336#define TASK_COMM_LEN 16
337
338#include <linux/spinlock.h>
339
340/*
341 * This serializes "schedule()" and also protects
342 * the run-queue from deletions/modifications (but
343 * _adding_ to the beginning of the run-queue has
344 * a separate lock).
345 */
346extern rwlock_t tasklist_lock;
347extern spinlock_t mmlist_lock;
348
349struct task_struct;
350
351#ifdef CONFIG_PROVE_RCU
352extern int lockdep_tasklist_lock_is_held(void);
353#endif /* #ifdef CONFIG_PROVE_RCU */
354
355extern void sched_init(void);
356extern void sched_init_smp(void);
357extern asmlinkage void schedule_tail(struct task_struct *prev);
358extern void init_idle(struct task_struct *idle, int cpu);
359extern void init_idle_bootup_task(struct task_struct *idle);
360
361extern cpumask_var_t cpu_isolated_map;
362
363extern int runqueue_is_locked(int cpu);
364
365#if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
366extern void nohz_balance_enter_idle(int cpu);
367extern void set_cpu_sd_state_idle(void);
368extern int get_nohz_timer_target(void);
369#else
370static inline void nohz_balance_enter_idle(int cpu) { }
371static inline void set_cpu_sd_state_idle(void) { }
372#endif
373
374/*
375 * Only dump TASK_* tasks. (0 for all tasks)
376 */
377extern void show_state_filter(unsigned long state_filter);
378
379static inline void show_state(void)
380{
381 show_state_filter(0);
382}
383
384extern void show_regs(struct pt_regs *);
385
386/*
387 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
388 * task), SP is the stack pointer of the first frame that should be shown in the back
389 * trace (or NULL if the entire call-chain of the task should be shown).
390 */
391extern void show_stack(struct task_struct *task, unsigned long *sp);
392
393extern void cpu_init (void);
394extern void trap_init(void);
395extern void update_process_times(int user);
396extern void scheduler_tick(void);
397extern int sched_cpu_starting(unsigned int cpu);
398extern int sched_cpu_activate(unsigned int cpu);
399extern int sched_cpu_deactivate(unsigned int cpu);
400
401#ifdef CONFIG_HOTPLUG_CPU
402extern int sched_cpu_dying(unsigned int cpu);
403#else
404# define sched_cpu_dying NULL
405#endif
406
407extern void sched_show_task(struct task_struct *p);
408
409#ifdef CONFIG_LOCKUP_DETECTOR
410extern void touch_softlockup_watchdog_sched(void);
411extern void touch_softlockup_watchdog(void);
412extern void touch_softlockup_watchdog_sync(void);
413extern void touch_all_softlockup_watchdogs(void);
414extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
415 void __user *buffer,
416 size_t *lenp, loff_t *ppos);
417extern unsigned int softlockup_panic;
418extern unsigned int hardlockup_panic;
419void lockup_detector_init(void);
420#else
421static inline void touch_softlockup_watchdog_sched(void)
422{
423}
424static inline void touch_softlockup_watchdog(void)
425{
426}
427static inline void touch_softlockup_watchdog_sync(void)
428{
429}
430static inline void touch_all_softlockup_watchdogs(void)
431{
432}
433static inline void lockup_detector_init(void)
434{
435}
436#endif
437
438#ifdef CONFIG_DETECT_HUNG_TASK
439void reset_hung_task_detector(void);
440#else
441static inline void reset_hung_task_detector(void)
442{
443}
444#endif
445
446/* Attach to any functions which should be ignored in wchan output. */
447#define __sched __attribute__((__section__(".sched.text")))
448
449/* Linker adds these: start and end of __sched functions */
450extern char __sched_text_start[], __sched_text_end[];
451
452/* Is this address in the __sched functions? */
453extern int in_sched_functions(unsigned long addr);
454
455#define MAX_SCHEDULE_TIMEOUT LONG_MAX
456extern signed long schedule_timeout(signed long timeout);
457extern signed long schedule_timeout_interruptible(signed long timeout);
458extern signed long schedule_timeout_killable(signed long timeout);
459extern signed long schedule_timeout_uninterruptible(signed long timeout);
460extern signed long schedule_timeout_idle(signed long timeout);
461asmlinkage void schedule(void);
462extern void schedule_preempt_disabled(void);
463
464extern long io_schedule_timeout(long timeout);
465
466static inline void io_schedule(void)
467{
468 io_schedule_timeout(MAX_SCHEDULE_TIMEOUT);
469}
470
471void __noreturn do_task_dead(void);
472
473struct nsproxy;
474struct user_namespace;
475
476#ifdef CONFIG_MMU
477extern void arch_pick_mmap_layout(struct mm_struct *mm);
478extern unsigned long
479arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
480 unsigned long, unsigned long);
481extern unsigned long
482arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
483 unsigned long len, unsigned long pgoff,
484 unsigned long flags);
485#else
486static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
487#endif
488
489#define SUID_DUMP_DISABLE 0 /* No setuid dumping */
490#define SUID_DUMP_USER 1 /* Dump as user of process */
491#define SUID_DUMP_ROOT 2 /* Dump as root */
492
493/* mm flags */
494
495/* for SUID_DUMP_* above */
496#define MMF_DUMPABLE_BITS 2
497#define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
498
499extern void set_dumpable(struct mm_struct *mm, int value);
500/*
501 * This returns the actual value of the suid_dumpable flag. For things
502 * that are using this for checking for privilege transitions, it must
503 * test against SUID_DUMP_USER rather than treating it as a boolean
504 * value.
505 */
506static inline int __get_dumpable(unsigned long mm_flags)
507{
508 return mm_flags & MMF_DUMPABLE_MASK;
509}
510
511static inline int get_dumpable(struct mm_struct *mm)
512{
513 return __get_dumpable(mm->flags);
514}
515
516/* coredump filter bits */
517#define MMF_DUMP_ANON_PRIVATE 2
518#define MMF_DUMP_ANON_SHARED 3
519#define MMF_DUMP_MAPPED_PRIVATE 4
520#define MMF_DUMP_MAPPED_SHARED 5
521#define MMF_DUMP_ELF_HEADERS 6
522#define MMF_DUMP_HUGETLB_PRIVATE 7
523#define MMF_DUMP_HUGETLB_SHARED 8
524#define MMF_DUMP_DAX_PRIVATE 9
525#define MMF_DUMP_DAX_SHARED 10
526
527#define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
528#define MMF_DUMP_FILTER_BITS 9
529#define MMF_DUMP_FILTER_MASK \
530 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
531#define MMF_DUMP_FILTER_DEFAULT \
532 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
533 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
534
535#ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
536# define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
537#else
538# define MMF_DUMP_MASK_DEFAULT_ELF 0
539#endif
540 /* leave room for more dump flags */
541#define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
542#define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
543/*
544 * This one-shot flag is dropped due to necessity of changing exe once again
545 * on NFS restore
546 */
547//#define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
548
549#define MMF_HAS_UPROBES 19 /* has uprobes */
550#define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
551#define MMF_OOM_SKIP 21 /* mm is of no interest for the OOM killer */
552#define MMF_UNSTABLE 22 /* mm is unstable for copy_from_user */
553#define MMF_HUGE_ZERO_PAGE 23 /* mm has ever used the global huge zero page */
554
555#define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
556
557struct sighand_struct {
558 atomic_t count;
559 struct k_sigaction action[_NSIG];
560 spinlock_t siglock;
561 wait_queue_head_t signalfd_wqh;
562};
563
564struct pacct_struct {
565 int ac_flag;
566 long ac_exitcode;
567 unsigned long ac_mem;
568 cputime_t ac_utime, ac_stime;
569 unsigned long ac_minflt, ac_majflt;
570};
571
572struct cpu_itimer {
573 cputime_t expires;
574 cputime_t incr;
575 u32 error;
576 u32 incr_error;
577};
578
579/**
580 * struct prev_cputime - snaphsot of system and user cputime
581 * @utime: time spent in user mode
582 * @stime: time spent in system mode
583 * @lock: protects the above two fields
584 *
585 * Stores previous user/system time values such that we can guarantee
586 * monotonicity.
587 */
588struct prev_cputime {
589#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
590 cputime_t utime;
591 cputime_t stime;
592 raw_spinlock_t lock;
593#endif
594};
595
596static inline void prev_cputime_init(struct prev_cputime *prev)
597{
598#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
599 prev->utime = prev->stime = 0;
600 raw_spin_lock_init(&prev->lock);
601#endif
602}
603
604/**
605 * struct task_cputime - collected CPU time counts
606 * @utime: time spent in user mode, in &cputime_t units
607 * @stime: time spent in kernel mode, in &cputime_t units
608 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
609 *
610 * This structure groups together three kinds of CPU time that are tracked for
611 * threads and thread groups. Most things considering CPU time want to group
612 * these counts together and treat all three of them in parallel.
613 */
614struct task_cputime {
615 cputime_t utime;
616 cputime_t stime;
617 unsigned long long sum_exec_runtime;
618};
619
620/* Alternate field names when used to cache expirations. */
621#define virt_exp utime
622#define prof_exp stime
623#define sched_exp sum_exec_runtime
624
625#define INIT_CPUTIME \
626 (struct task_cputime) { \
627 .utime = 0, \
628 .stime = 0, \
629 .sum_exec_runtime = 0, \
630 }
631
632/*
633 * This is the atomic variant of task_cputime, which can be used for
634 * storing and updating task_cputime statistics without locking.
635 */
636struct task_cputime_atomic {
637 atomic64_t utime;
638 atomic64_t stime;
639 atomic64_t sum_exec_runtime;
640};
641
642#define INIT_CPUTIME_ATOMIC \
643 (struct task_cputime_atomic) { \
644 .utime = ATOMIC64_INIT(0), \
645 .stime = ATOMIC64_INIT(0), \
646 .sum_exec_runtime = ATOMIC64_INIT(0), \
647 }
648
649#define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
650
651/*
652 * Disable preemption until the scheduler is running -- use an unconditional
653 * value so that it also works on !PREEMPT_COUNT kernels.
654 *
655 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
656 */
657#define INIT_PREEMPT_COUNT PREEMPT_OFFSET
658
659/*
660 * Initial preempt_count value; reflects the preempt_count schedule invariant
661 * which states that during context switches:
662 *
663 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
664 *
665 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
666 * Note: See finish_task_switch().
667 */
668#define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
669
670/**
671 * struct thread_group_cputimer - thread group interval timer counts
672 * @cputime_atomic: atomic thread group interval timers.
673 * @running: true when there are timers running and
674 * @cputime_atomic receives updates.
675 * @checking_timer: true when a thread in the group is in the
676 * process of checking for thread group timers.
677 *
678 * This structure contains the version of task_cputime, above, that is
679 * used for thread group CPU timer calculations.
680 */
681struct thread_group_cputimer {
682 struct task_cputime_atomic cputime_atomic;
683 bool running;
684 bool checking_timer;
685};
686
687#include <linux/rwsem.h>
688struct autogroup;
689
690/*
691 * NOTE! "signal_struct" does not have its own
692 * locking, because a shared signal_struct always
693 * implies a shared sighand_struct, so locking
694 * sighand_struct is always a proper superset of
695 * the locking of signal_struct.
696 */
697struct signal_struct {
698 atomic_t sigcnt;
699 atomic_t live;
700 int nr_threads;
701 struct list_head thread_head;
702
703 wait_queue_head_t wait_chldexit; /* for wait4() */
704
705 /* current thread group signal load-balancing target: */
706 struct task_struct *curr_target;
707
708 /* shared signal handling: */
709 struct sigpending shared_pending;
710
711 /* thread group exit support */
712 int group_exit_code;
713 /* overloaded:
714 * - notify group_exit_task when ->count is equal to notify_count
715 * - everyone except group_exit_task is stopped during signal delivery
716 * of fatal signals, group_exit_task processes the signal.
717 */
718 int notify_count;
719 struct task_struct *group_exit_task;
720
721 /* thread group stop support, overloads group_exit_code too */
722 int group_stop_count;
723 unsigned int flags; /* see SIGNAL_* flags below */
724
725 /*
726 * PR_SET_CHILD_SUBREAPER marks a process, like a service
727 * manager, to re-parent orphan (double-forking) child processes
728 * to this process instead of 'init'. The service manager is
729 * able to receive SIGCHLD signals and is able to investigate
730 * the process until it calls wait(). All children of this
731 * process will inherit a flag if they should look for a
732 * child_subreaper process at exit.
733 */
734 unsigned int is_child_subreaper:1;
735 unsigned int has_child_subreaper:1;
736
737 /* POSIX.1b Interval Timers */
738 int posix_timer_id;
739 struct list_head posix_timers;
740
741 /* ITIMER_REAL timer for the process */
742 struct hrtimer real_timer;
743 struct pid *leader_pid;
744 ktime_t it_real_incr;
745
746 /*
747 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
748 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
749 * values are defined to 0 and 1 respectively
750 */
751 struct cpu_itimer it[2];
752
753 /*
754 * Thread group totals for process CPU timers.
755 * See thread_group_cputimer(), et al, for details.
756 */
757 struct thread_group_cputimer cputimer;
758
759 /* Earliest-expiration cache. */
760 struct task_cputime cputime_expires;
761
762#ifdef CONFIG_NO_HZ_FULL
763 atomic_t tick_dep_mask;
764#endif
765
766 struct list_head cpu_timers[3];
767
768 struct pid *tty_old_pgrp;
769
770 /* boolean value for session group leader */
771 int leader;
772
773 struct tty_struct *tty; /* NULL if no tty */
774
775#ifdef CONFIG_SCHED_AUTOGROUP
776 struct autogroup *autogroup;
777#endif
778 /*
779 * Cumulative resource counters for dead threads in the group,
780 * and for reaped dead child processes forked by this group.
781 * Live threads maintain their own counters and add to these
782 * in __exit_signal, except for the group leader.
783 */
784 seqlock_t stats_lock;
785 cputime_t utime, stime, cutime, cstime;
786 cputime_t gtime;
787 cputime_t cgtime;
788 struct prev_cputime prev_cputime;
789 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
790 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
791 unsigned long inblock, oublock, cinblock, coublock;
792 unsigned long maxrss, cmaxrss;
793 struct task_io_accounting ioac;
794
795 /*
796 * Cumulative ns of schedule CPU time fo dead threads in the
797 * group, not including a zombie group leader, (This only differs
798 * from jiffies_to_ns(utime + stime) if sched_clock uses something
799 * other than jiffies.)
800 */
801 unsigned long long sum_sched_runtime;
802
803 /*
804 * We don't bother to synchronize most readers of this at all,
805 * because there is no reader checking a limit that actually needs
806 * to get both rlim_cur and rlim_max atomically, and either one
807 * alone is a single word that can safely be read normally.
808 * getrlimit/setrlimit use task_lock(current->group_leader) to
809 * protect this instead of the siglock, because they really
810 * have no need to disable irqs.
811 */
812 struct rlimit rlim[RLIM_NLIMITS];
813
814#ifdef CONFIG_BSD_PROCESS_ACCT
815 struct pacct_struct pacct; /* per-process accounting information */
816#endif
817#ifdef CONFIG_TASKSTATS
818 struct taskstats *stats;
819#endif
820#ifdef CONFIG_AUDIT
821 unsigned audit_tty;
822 struct tty_audit_buf *tty_audit_buf;
823#endif
824
825 /*
826 * Thread is the potential origin of an oom condition; kill first on
827 * oom
828 */
829 bool oom_flag_origin;
830 short oom_score_adj; /* OOM kill score adjustment */
831 short oom_score_adj_min; /* OOM kill score adjustment min value.
832 * Only settable by CAP_SYS_RESOURCE. */
833 struct mm_struct *oom_mm; /* recorded mm when the thread group got
834 * killed by the oom killer */
835
836 struct mutex cred_guard_mutex; /* guard against foreign influences on
837 * credential calculations
838 * (notably. ptrace) */
839};
840
841/*
842 * Bits in flags field of signal_struct.
843 */
844#define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
845#define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
846#define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
847#define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
848/*
849 * Pending notifications to parent.
850 */
851#define SIGNAL_CLD_STOPPED 0x00000010
852#define SIGNAL_CLD_CONTINUED 0x00000020
853#define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
854
855#define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
856
857#define SIGNAL_STOP_MASK (SIGNAL_CLD_MASK | SIGNAL_STOP_STOPPED | \
858 SIGNAL_STOP_CONTINUED)
859
860static inline void signal_set_stop_flags(struct signal_struct *sig,
861 unsigned int flags)
862{
863 WARN_ON(sig->flags & (SIGNAL_GROUP_EXIT|SIGNAL_GROUP_COREDUMP));
864 sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags;
865}
866
867/* If true, all threads except ->group_exit_task have pending SIGKILL */
868static inline int signal_group_exit(const struct signal_struct *sig)
869{
870 return (sig->flags & SIGNAL_GROUP_EXIT) ||
871 (sig->group_exit_task != NULL);
872}
873
874/*
875 * Some day this will be a full-fledged user tracking system..
876 */
877struct user_struct {
878 atomic_t __count; /* reference count */
879 atomic_t processes; /* How many processes does this user have? */
880 atomic_t sigpending; /* How many pending signals does this user have? */
881#ifdef CONFIG_INOTIFY_USER
882 atomic_t inotify_watches; /* How many inotify watches does this user have? */
883 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */
884#endif
885#ifdef CONFIG_FANOTIFY
886 atomic_t fanotify_listeners;
887#endif
888#ifdef CONFIG_EPOLL
889 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
890#endif
891#ifdef CONFIG_POSIX_MQUEUE
892 /* protected by mq_lock */
893 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
894#endif
895 unsigned long locked_shm; /* How many pages of mlocked shm ? */
896 unsigned long unix_inflight; /* How many files in flight in unix sockets */
897 atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */
898
899#ifdef CONFIG_KEYS
900 struct key *uid_keyring; /* UID specific keyring */
901 struct key *session_keyring; /* UID's default session keyring */
902#endif
903
904 /* Hash table maintenance information */
905 struct hlist_node uidhash_node;
906 kuid_t uid;
907
908#if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
909 atomic_long_t locked_vm;
910#endif
911};
912
913extern int uids_sysfs_init(void);
914
915extern struct user_struct *find_user(kuid_t);
916
917extern struct user_struct root_user;
918#define INIT_USER (&root_user)
919
920
921struct backing_dev_info;
922struct reclaim_state;
923
924#ifdef CONFIG_SCHED_INFO
925struct sched_info {
926 /* cumulative counters */
927 unsigned long pcount; /* # of times run on this cpu */
928 unsigned long long run_delay; /* time spent waiting on a runqueue */
929
930 /* timestamps */
931 unsigned long long last_arrival,/* when we last ran on a cpu */
932 last_queued; /* when we were last queued to run */
933};
934#endif /* CONFIG_SCHED_INFO */
935
936#ifdef CONFIG_TASK_DELAY_ACCT
937struct task_delay_info {
938 spinlock_t lock;
939 unsigned int flags; /* Private per-task flags */
940
941 /* For each stat XXX, add following, aligned appropriately
942 *
943 * struct timespec XXX_start, XXX_end;
944 * u64 XXX_delay;
945 * u32 XXX_count;
946 *
947 * Atomicity of updates to XXX_delay, XXX_count protected by
948 * single lock above (split into XXX_lock if contention is an issue).
949 */
950
951 /*
952 * XXX_count is incremented on every XXX operation, the delay
953 * associated with the operation is added to XXX_delay.
954 * XXX_delay contains the accumulated delay time in nanoseconds.
955 */
956 u64 blkio_start; /* Shared by blkio, swapin */
957 u64 blkio_delay; /* wait for sync block io completion */
958 u64 swapin_delay; /* wait for swapin block io completion */
959 u32 blkio_count; /* total count of the number of sync block */
960 /* io operations performed */
961 u32 swapin_count; /* total count of the number of swapin block */
962 /* io operations performed */
963
964 u64 freepages_start;
965 u64 freepages_delay; /* wait for memory reclaim */
966 u32 freepages_count; /* total count of memory reclaim */
967};
968#endif /* CONFIG_TASK_DELAY_ACCT */
969
970static inline int sched_info_on(void)
971{
972#ifdef CONFIG_SCHEDSTATS
973 return 1;
974#elif defined(CONFIG_TASK_DELAY_ACCT)
975 extern int delayacct_on;
976 return delayacct_on;
977#else
978 return 0;
979#endif
980}
981
982#ifdef CONFIG_SCHEDSTATS
983void force_schedstat_enabled(void);
984#endif
985
986enum cpu_idle_type {
987 CPU_IDLE,
988 CPU_NOT_IDLE,
989 CPU_NEWLY_IDLE,
990 CPU_MAX_IDLE_TYPES
991};
992
993/*
994 * Integer metrics need fixed point arithmetic, e.g., sched/fair
995 * has a few: load, load_avg, util_avg, freq, and capacity.
996 *
997 * We define a basic fixed point arithmetic range, and then formalize
998 * all these metrics based on that basic range.
999 */
1000# define SCHED_FIXEDPOINT_SHIFT 10
1001# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
1002
1003/*
1004 * Increase resolution of cpu_capacity calculations
1005 */
1006#define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
1007#define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
1008
1009/*
1010 * Wake-queues are lists of tasks with a pending wakeup, whose
1011 * callers have already marked the task as woken internally,
1012 * and can thus carry on. A common use case is being able to
1013 * do the wakeups once the corresponding user lock as been
1014 * released.
1015 *
1016 * We hold reference to each task in the list across the wakeup,
1017 * thus guaranteeing that the memory is still valid by the time
1018 * the actual wakeups are performed in wake_up_q().
1019 *
1020 * One per task suffices, because there's never a need for a task to be
1021 * in two wake queues simultaneously; it is forbidden to abandon a task
1022 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
1023 * already in a wake queue, the wakeup will happen soon and the second
1024 * waker can just skip it.
1025 *
1026 * The DEFINE_WAKE_Q macro declares and initializes the list head.
1027 * wake_up_q() does NOT reinitialize the list; it's expected to be
1028 * called near the end of a function, where the fact that the queue is
1029 * not used again will be easy to see by inspection.
1030 *
1031 * Note that this can cause spurious wakeups. schedule() callers
1032 * must ensure the call is done inside a loop, confirming that the
1033 * wakeup condition has in fact occurred.
1034 */
1035struct wake_q_node {
1036 struct wake_q_node *next;
1037};
1038
1039struct wake_q_head {
1040 struct wake_q_node *first;
1041 struct wake_q_node **lastp;
1042};
1043
1044#define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1045
1046#define DEFINE_WAKE_Q(name) \
1047 struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1048
1049extern void wake_q_add(struct wake_q_head *head,
1050 struct task_struct *task);
1051extern void wake_up_q(struct wake_q_head *head);
1052
1053/*
1054 * sched-domains (multiprocessor balancing) declarations:
1055 */
1056#ifdef CONFIG_SMP
1057#define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
1058#define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
1059#define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
1060#define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
1061#define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
1062#define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
1063#define SD_ASYM_CPUCAPACITY 0x0040 /* Groups have different max cpu capacities */
1064#define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu capacity */
1065#define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
1066#define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
1067#define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
1068#define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
1069#define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
1070#define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
1071#define SD_NUMA 0x4000 /* cross-node balancing */
1072
1073#ifdef CONFIG_SCHED_SMT
1074static inline int cpu_smt_flags(void)
1075{
1076 return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1077}
1078#endif
1079
1080#ifdef CONFIG_SCHED_MC
1081static inline int cpu_core_flags(void)
1082{
1083 return SD_SHARE_PKG_RESOURCES;
1084}
1085#endif
1086
1087#ifdef CONFIG_NUMA
1088static inline int cpu_numa_flags(void)
1089{
1090 return SD_NUMA;
1091}
1092#endif
1093
1094extern int arch_asym_cpu_priority(int cpu);
1095
1096struct sched_domain_attr {
1097 int relax_domain_level;
1098};
1099
1100#define SD_ATTR_INIT (struct sched_domain_attr) { \
1101 .relax_domain_level = -1, \
1102}
1103
1104extern int sched_domain_level_max;
1105
1106struct sched_group;
1107
1108struct sched_domain_shared {
1109 atomic_t ref;
1110 atomic_t nr_busy_cpus;
1111 int has_idle_cores;
1112};
1113
1114struct sched_domain {
1115 /* These fields must be setup */
1116 struct sched_domain *parent; /* top domain must be null terminated */
1117 struct sched_domain *child; /* bottom domain must be null terminated */
1118 struct sched_group *groups; /* the balancing groups of the domain */
1119 unsigned long min_interval; /* Minimum balance interval ms */
1120 unsigned long max_interval; /* Maximum balance interval ms */
1121 unsigned int busy_factor; /* less balancing by factor if busy */
1122 unsigned int imbalance_pct; /* No balance until over watermark */
1123 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */
1124 unsigned int busy_idx;
1125 unsigned int idle_idx;
1126 unsigned int newidle_idx;
1127 unsigned int wake_idx;
1128 unsigned int forkexec_idx;
1129 unsigned int smt_gain;
1130
1131 int nohz_idle; /* NOHZ IDLE status */
1132 int flags; /* See SD_* */
1133 int level;
1134
1135 /* Runtime fields. */
1136 unsigned long last_balance; /* init to jiffies. units in jiffies */
1137 unsigned int balance_interval; /* initialise to 1. units in ms. */
1138 unsigned int nr_balance_failed; /* initialise to 0 */
1139
1140 /* idle_balance() stats */
1141 u64 max_newidle_lb_cost;
1142 unsigned long next_decay_max_lb_cost;
1143
1144 u64 avg_scan_cost; /* select_idle_sibling */
1145
1146#ifdef CONFIG_SCHEDSTATS
1147 /* load_balance() stats */
1148 unsigned int lb_count[CPU_MAX_IDLE_TYPES];
1149 unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
1150 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
1151 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
1152 unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
1153 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
1154 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
1155 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
1156
1157 /* Active load balancing */
1158 unsigned int alb_count;
1159 unsigned int alb_failed;
1160 unsigned int alb_pushed;
1161
1162 /* SD_BALANCE_EXEC stats */
1163 unsigned int sbe_count;
1164 unsigned int sbe_balanced;
1165 unsigned int sbe_pushed;
1166
1167 /* SD_BALANCE_FORK stats */
1168 unsigned int sbf_count;
1169 unsigned int sbf_balanced;
1170 unsigned int sbf_pushed;
1171
1172 /* try_to_wake_up() stats */
1173 unsigned int ttwu_wake_remote;
1174 unsigned int ttwu_move_affine;
1175 unsigned int ttwu_move_balance;
1176#endif
1177#ifdef CONFIG_SCHED_DEBUG
1178 char *name;
1179#endif
1180 union {
1181 void *private; /* used during construction */
1182 struct rcu_head rcu; /* used during destruction */
1183 };
1184 struct sched_domain_shared *shared;
1185
1186 unsigned int span_weight;
1187 /*
1188 * Span of all CPUs in this domain.
1189 *
1190 * NOTE: this field is variable length. (Allocated dynamically
1191 * by attaching extra space to the end of the structure,
1192 * depending on how many CPUs the kernel has booted up with)
1193 */
1194 unsigned long span[0];
1195};
1196
1197static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
1198{
1199 return to_cpumask(sd->span);
1200}
1201
1202extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1203 struct sched_domain_attr *dattr_new);
1204
1205/* Allocate an array of sched domains, for partition_sched_domains(). */
1206cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
1207void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
1208
1209bool cpus_share_cache(int this_cpu, int that_cpu);
1210
1211typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1212typedef int (*sched_domain_flags_f)(void);
1213
1214#define SDTL_OVERLAP 0x01
1215
1216struct sd_data {
1217 struct sched_domain **__percpu sd;
1218 struct sched_domain_shared **__percpu sds;
1219 struct sched_group **__percpu sg;
1220 struct sched_group_capacity **__percpu sgc;
1221};
1222
1223struct sched_domain_topology_level {
1224 sched_domain_mask_f mask;
1225 sched_domain_flags_f sd_flags;
1226 int flags;
1227 int numa_level;
1228 struct sd_data data;
1229#ifdef CONFIG_SCHED_DEBUG
1230 char *name;
1231#endif
1232};
1233
1234extern void set_sched_topology(struct sched_domain_topology_level *tl);
1235extern void wake_up_if_idle(int cpu);
1236
1237#ifdef CONFIG_SCHED_DEBUG
1238# define SD_INIT_NAME(type) .name = #type
1239#else
1240# define SD_INIT_NAME(type)
1241#endif
1242
1243#else /* CONFIG_SMP */
1244
1245struct sched_domain_attr;
1246
1247static inline void
1248partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1249 struct sched_domain_attr *dattr_new)
1250{
1251}
1252
1253static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1254{
1255 return true;
1256}
1257
1258#endif /* !CONFIG_SMP */
1259
1260
1261struct io_context; /* See blkdev.h */
1262
1263
1264#ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1265extern void prefetch_stack(struct task_struct *t);
1266#else
1267static inline void prefetch_stack(struct task_struct *t) { }
1268#endif
1269
1270struct audit_context; /* See audit.c */
1271struct mempolicy;
1272struct pipe_inode_info;
1273struct uts_namespace;
1274
1275struct load_weight {
1276 unsigned long weight;
1277 u32 inv_weight;
1278};
1279
1280/*
1281 * The load_avg/util_avg accumulates an infinite geometric series
1282 * (see __update_load_avg() in kernel/sched/fair.c).
1283 *
1284 * [load_avg definition]
1285 *
1286 * load_avg = runnable% * scale_load_down(load)
1287 *
1288 * where runnable% is the time ratio that a sched_entity is runnable.
1289 * For cfs_rq, it is the aggregated load_avg of all runnable and
1290 * blocked sched_entities.
1291 *
1292 * load_avg may also take frequency scaling into account:
1293 *
1294 * load_avg = runnable% * scale_load_down(load) * freq%
1295 *
1296 * where freq% is the CPU frequency normalized to the highest frequency.
1297 *
1298 * [util_avg definition]
1299 *
1300 * util_avg = running% * SCHED_CAPACITY_SCALE
1301 *
1302 * where running% is the time ratio that a sched_entity is running on
1303 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1304 * and blocked sched_entities.
1305 *
1306 * util_avg may also factor frequency scaling and CPU capacity scaling:
1307 *
1308 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1309 *
1310 * where freq% is the same as above, and capacity% is the CPU capacity
1311 * normalized to the greatest capacity (due to uarch differences, etc).
1312 *
1313 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1314 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1315 * we therefore scale them to as large a range as necessary. This is for
1316 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1317 *
1318 * [Overflow issue]
1319 *
1320 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1321 * with the highest load (=88761), always runnable on a single cfs_rq,
1322 * and should not overflow as the number already hits PID_MAX_LIMIT.
1323 *
1324 * For all other cases (including 32-bit kernels), struct load_weight's
1325 * weight will overflow first before we do, because:
1326 *
1327 * Max(load_avg) <= Max(load.weight)
1328 *
1329 * Then it is the load_weight's responsibility to consider overflow
1330 * issues.
1331 */
1332struct sched_avg {
1333 u64 last_update_time, load_sum;
1334 u32 util_sum, period_contrib;
1335 unsigned long load_avg, util_avg;
1336};
1337
1338#ifdef CONFIG_SCHEDSTATS
1339struct sched_statistics {
1340 u64 wait_start;
1341 u64 wait_max;
1342 u64 wait_count;
1343 u64 wait_sum;
1344 u64 iowait_count;
1345 u64 iowait_sum;
1346
1347 u64 sleep_start;
1348 u64 sleep_max;
1349 s64 sum_sleep_runtime;
1350
1351 u64 block_start;
1352 u64 block_max;
1353 u64 exec_max;
1354 u64 slice_max;
1355
1356 u64 nr_migrations_cold;
1357 u64 nr_failed_migrations_affine;
1358 u64 nr_failed_migrations_running;
1359 u64 nr_failed_migrations_hot;
1360 u64 nr_forced_migrations;
1361
1362 u64 nr_wakeups;
1363 u64 nr_wakeups_sync;
1364 u64 nr_wakeups_migrate;
1365 u64 nr_wakeups_local;
1366 u64 nr_wakeups_remote;
1367 u64 nr_wakeups_affine;
1368 u64 nr_wakeups_affine_attempts;
1369 u64 nr_wakeups_passive;
1370 u64 nr_wakeups_idle;
1371};
1372#endif
1373
1374struct sched_entity {
1375 struct load_weight load; /* for load-balancing */
1376 struct rb_node run_node;
1377 struct list_head group_node;
1378 unsigned int on_rq;
1379
1380 u64 exec_start;
1381 u64 sum_exec_runtime;
1382 u64 vruntime;
1383 u64 prev_sum_exec_runtime;
1384
1385 u64 nr_migrations;
1386
1387#ifdef CONFIG_SCHEDSTATS
1388 struct sched_statistics statistics;
1389#endif
1390
1391#ifdef CONFIG_FAIR_GROUP_SCHED
1392 int depth;
1393 struct sched_entity *parent;
1394 /* rq on which this entity is (to be) queued: */
1395 struct cfs_rq *cfs_rq;
1396 /* rq "owned" by this entity/group: */
1397 struct cfs_rq *my_q;
1398#endif
1399
1400#ifdef CONFIG_SMP
1401 /*
1402 * Per entity load average tracking.
1403 *
1404 * Put into separate cache line so it does not
1405 * collide with read-mostly values above.
1406 */
1407 struct sched_avg avg ____cacheline_aligned_in_smp;
1408#endif
1409};
1410
1411struct sched_rt_entity {
1412 struct list_head run_list;
1413 unsigned long timeout;
1414 unsigned long watchdog_stamp;
1415 unsigned int time_slice;
1416 unsigned short on_rq;
1417 unsigned short on_list;
1418
1419 struct sched_rt_entity *back;
1420#ifdef CONFIG_RT_GROUP_SCHED
1421 struct sched_rt_entity *parent;
1422 /* rq on which this entity is (to be) queued: */
1423 struct rt_rq *rt_rq;
1424 /* rq "owned" by this entity/group: */
1425 struct rt_rq *my_q;
1426#endif
1427};
1428
1429struct sched_dl_entity {
1430 struct rb_node rb_node;
1431
1432 /*
1433 * Original scheduling parameters. Copied here from sched_attr
1434 * during sched_setattr(), they will remain the same until
1435 * the next sched_setattr().
1436 */
1437 u64 dl_runtime; /* maximum runtime for each instance */
1438 u64 dl_deadline; /* relative deadline of each instance */
1439 u64 dl_period; /* separation of two instances (period) */
1440 u64 dl_bw; /* dl_runtime / dl_deadline */
1441
1442 /*
1443 * Actual scheduling parameters. Initialized with the values above,
1444 * they are continously updated during task execution. Note that
1445 * the remaining runtime could be < 0 in case we are in overrun.
1446 */
1447 s64 runtime; /* remaining runtime for this instance */
1448 u64 deadline; /* absolute deadline for this instance */
1449 unsigned int flags; /* specifying the scheduler behaviour */
1450
1451 /*
1452 * Some bool flags:
1453 *
1454 * @dl_throttled tells if we exhausted the runtime. If so, the
1455 * task has to wait for a replenishment to be performed at the
1456 * next firing of dl_timer.
1457 *
1458 * @dl_boosted tells if we are boosted due to DI. If so we are
1459 * outside bandwidth enforcement mechanism (but only until we
1460 * exit the critical section);
1461 *
1462 * @dl_yielded tells if task gave up the cpu before consuming
1463 * all its available runtime during the last job.
1464 */
1465 int dl_throttled, dl_boosted, dl_yielded;
1466
1467 /*
1468 * Bandwidth enforcement timer. Each -deadline task has its
1469 * own bandwidth to be enforced, thus we need one timer per task.
1470 */
1471 struct hrtimer dl_timer;
1472};
1473
1474union rcu_special {
1475 struct {
1476 u8 blocked;
1477 u8 need_qs;
1478 u8 exp_need_qs;
1479 u8 pad; /* Otherwise the compiler can store garbage here. */
1480 } b; /* Bits. */
1481 u32 s; /* Set of bits. */
1482};
1483struct rcu_node;
1484
1485enum perf_event_task_context {
1486 perf_invalid_context = -1,
1487 perf_hw_context = 0,
1488 perf_sw_context,
1489 perf_nr_task_contexts,
1490};
1491
1492/* Track pages that require TLB flushes */
1493struct tlbflush_unmap_batch {
1494 /*
1495 * Each bit set is a CPU that potentially has a TLB entry for one of
1496 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1497 */
1498 struct cpumask cpumask;
1499
1500 /* True if any bit in cpumask is set */
1501 bool flush_required;
1502
1503 /*
1504 * If true then the PTE was dirty when unmapped. The entry must be
1505 * flushed before IO is initiated or a stale TLB entry potentially
1506 * allows an update without redirtying the page.
1507 */
1508 bool writable;
1509};
1510
1511struct task_struct {
1512#ifdef CONFIG_THREAD_INFO_IN_TASK
1513 /*
1514 * For reasons of header soup (see current_thread_info()), this
1515 * must be the first element of task_struct.
1516 */
1517 struct thread_info thread_info;
1518#endif
1519 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1520 void *stack;
1521 atomic_t usage;
1522 unsigned int flags; /* per process flags, defined below */
1523 unsigned int ptrace;
1524
1525#ifdef CONFIG_SMP
1526 struct llist_node wake_entry;
1527 int on_cpu;
1528#ifdef CONFIG_THREAD_INFO_IN_TASK
1529 unsigned int cpu; /* current CPU */
1530#endif
1531 unsigned int wakee_flips;
1532 unsigned long wakee_flip_decay_ts;
1533 struct task_struct *last_wakee;
1534
1535 int wake_cpu;
1536#endif
1537 int on_rq;
1538
1539 int prio, static_prio, normal_prio;
1540 unsigned int rt_priority;
1541 const struct sched_class *sched_class;
1542 struct sched_entity se;
1543 struct sched_rt_entity rt;
1544#ifdef CONFIG_CGROUP_SCHED
1545 struct task_group *sched_task_group;
1546#endif
1547 struct sched_dl_entity dl;
1548
1549#ifdef CONFIG_PREEMPT_NOTIFIERS
1550 /* list of struct preempt_notifier: */
1551 struct hlist_head preempt_notifiers;
1552#endif
1553
1554#ifdef CONFIG_BLK_DEV_IO_TRACE
1555 unsigned int btrace_seq;
1556#endif
1557
1558 unsigned int policy;
1559 int nr_cpus_allowed;
1560 cpumask_t cpus_allowed;
1561
1562#ifdef CONFIG_PREEMPT_RCU
1563 int rcu_read_lock_nesting;
1564 union rcu_special rcu_read_unlock_special;
1565 struct list_head rcu_node_entry;
1566 struct rcu_node *rcu_blocked_node;
1567#endif /* #ifdef CONFIG_PREEMPT_RCU */
1568#ifdef CONFIG_TASKS_RCU
1569 unsigned long rcu_tasks_nvcsw;
1570 bool rcu_tasks_holdout;
1571 struct list_head rcu_tasks_holdout_list;
1572 int rcu_tasks_idle_cpu;
1573#endif /* #ifdef CONFIG_TASKS_RCU */
1574
1575#ifdef CONFIG_SCHED_INFO
1576 struct sched_info sched_info;
1577#endif
1578
1579 struct list_head tasks;
1580#ifdef CONFIG_SMP
1581 struct plist_node pushable_tasks;
1582 struct rb_node pushable_dl_tasks;
1583#endif
1584
1585 struct mm_struct *mm, *active_mm;
1586 /* per-thread vma caching */
1587 u32 vmacache_seqnum;
1588 struct vm_area_struct *vmacache[VMACACHE_SIZE];
1589#if defined(SPLIT_RSS_COUNTING)
1590 struct task_rss_stat rss_stat;
1591#endif
1592/* task state */
1593 int exit_state;
1594 int exit_code, exit_signal;
1595 int pdeath_signal; /* The signal sent when the parent dies */
1596 unsigned long jobctl; /* JOBCTL_*, siglock protected */
1597
1598 /* Used for emulating ABI behavior of previous Linux versions */
1599 unsigned int personality;
1600
1601 /* scheduler bits, serialized by scheduler locks */
1602 unsigned sched_reset_on_fork:1;
1603 unsigned sched_contributes_to_load:1;
1604 unsigned sched_migrated:1;
1605 unsigned sched_remote_wakeup:1;
1606 unsigned :0; /* force alignment to the next boundary */
1607
1608 /* unserialized, strictly 'current' */
1609 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1610 unsigned in_iowait:1;
1611#if !defined(TIF_RESTORE_SIGMASK)
1612 unsigned restore_sigmask:1;
1613#endif
1614#ifdef CONFIG_MEMCG
1615 unsigned memcg_may_oom:1;
1616#ifndef CONFIG_SLOB
1617 unsigned memcg_kmem_skip_account:1;
1618#endif
1619#endif
1620#ifdef CONFIG_COMPAT_BRK
1621 unsigned brk_randomized:1;
1622#endif
1623
1624 unsigned long atomic_flags; /* Flags needing atomic access. */
1625
1626 struct restart_block restart_block;
1627
1628 pid_t pid;
1629 pid_t tgid;
1630
1631#ifdef CONFIG_CC_STACKPROTECTOR
1632 /* Canary value for the -fstack-protector gcc feature */
1633 unsigned long stack_canary;
1634#endif
1635 /*
1636 * pointers to (original) parent process, youngest child, younger sibling,
1637 * older sibling, respectively. (p->father can be replaced with
1638 * p->real_parent->pid)
1639 */
1640 struct task_struct __rcu *real_parent; /* real parent process */
1641 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1642 /*
1643 * children/sibling forms the list of my natural children
1644 */
1645 struct list_head children; /* list of my children */
1646 struct list_head sibling; /* linkage in my parent's children list */
1647 struct task_struct *group_leader; /* threadgroup leader */
1648
1649 /*
1650 * ptraced is the list of tasks this task is using ptrace on.
1651 * This includes both natural children and PTRACE_ATTACH targets.
1652 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1653 */
1654 struct list_head ptraced;
1655 struct list_head ptrace_entry;
1656
1657 /* PID/PID hash table linkage. */
1658 struct pid_link pids[PIDTYPE_MAX];
1659 struct list_head thread_group;
1660 struct list_head thread_node;
1661
1662 struct completion *vfork_done; /* for vfork() */
1663 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1664 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1665
1666 cputime_t utime, stime;
1667#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1668 cputime_t utimescaled, stimescaled;
1669#endif
1670 cputime_t gtime;
1671 struct prev_cputime prev_cputime;
1672#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1673 seqcount_t vtime_seqcount;
1674 unsigned long long vtime_snap;
1675 enum {
1676 /* Task is sleeping or running in a CPU with VTIME inactive */
1677 VTIME_INACTIVE = 0,
1678 /* Task runs in userspace in a CPU with VTIME active */
1679 VTIME_USER,
1680 /* Task runs in kernelspace in a CPU with VTIME active */
1681 VTIME_SYS,
1682 } vtime_snap_whence;
1683#endif
1684
1685#ifdef CONFIG_NO_HZ_FULL
1686 atomic_t tick_dep_mask;
1687#endif
1688 unsigned long nvcsw, nivcsw; /* context switch counts */
1689 u64 start_time; /* monotonic time in nsec */
1690 u64 real_start_time; /* boot based time in nsec */
1691/* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1692 unsigned long min_flt, maj_flt;
1693
1694 struct task_cputime cputime_expires;
1695 struct list_head cpu_timers[3];
1696
1697/* process credentials */
1698 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
1699 const struct cred __rcu *real_cred; /* objective and real subjective task
1700 * credentials (COW) */
1701 const struct cred __rcu *cred; /* effective (overridable) subjective task
1702 * credentials (COW) */
1703 char comm[TASK_COMM_LEN]; /* executable name excluding path
1704 - access with [gs]et_task_comm (which lock
1705 it with task_lock())
1706 - initialized normally by setup_new_exec */
1707/* file system info */
1708 struct nameidata *nameidata;
1709#ifdef CONFIG_SYSVIPC
1710/* ipc stuff */
1711 struct sysv_sem sysvsem;
1712 struct sysv_shm sysvshm;
1713#endif
1714#ifdef CONFIG_DETECT_HUNG_TASK
1715/* hung task detection */
1716 unsigned long last_switch_count;
1717#endif
1718/* filesystem information */
1719 struct fs_struct *fs;
1720/* open file information */
1721 struct files_struct *files;
1722/* namespaces */
1723 struct nsproxy *nsproxy;
1724/* signal handlers */
1725 struct signal_struct *signal;
1726 struct sighand_struct *sighand;
1727
1728 sigset_t blocked, real_blocked;
1729 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1730 struct sigpending pending;
1731
1732 unsigned long sas_ss_sp;
1733 size_t sas_ss_size;
1734 unsigned sas_ss_flags;
1735
1736 struct callback_head *task_works;
1737
1738 struct audit_context *audit_context;
1739#ifdef CONFIG_AUDITSYSCALL
1740 kuid_t loginuid;
1741 unsigned int sessionid;
1742#endif
1743 struct seccomp seccomp;
1744
1745/* Thread group tracking */
1746 u32 parent_exec_id;
1747 u32 self_exec_id;
1748/* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1749 * mempolicy */
1750 spinlock_t alloc_lock;
1751
1752 /* Protection of the PI data structures: */
1753 raw_spinlock_t pi_lock;
1754
1755 struct wake_q_node wake_q;
1756
1757#ifdef CONFIG_RT_MUTEXES
1758 /* PI waiters blocked on a rt_mutex held by this task */
1759 struct rb_root pi_waiters;
1760 struct rb_node *pi_waiters_leftmost;
1761 /* Deadlock detection and priority inheritance handling */
1762 struct rt_mutex_waiter *pi_blocked_on;
1763#endif
1764
1765#ifdef CONFIG_DEBUG_MUTEXES
1766 /* mutex deadlock detection */
1767 struct mutex_waiter *blocked_on;
1768#endif
1769#ifdef CONFIG_TRACE_IRQFLAGS
1770 unsigned int irq_events;
1771 unsigned long hardirq_enable_ip;
1772 unsigned long hardirq_disable_ip;
1773 unsigned int hardirq_enable_event;
1774 unsigned int hardirq_disable_event;
1775 int hardirqs_enabled;
1776 int hardirq_context;
1777 unsigned long softirq_disable_ip;
1778 unsigned long softirq_enable_ip;
1779 unsigned int softirq_disable_event;
1780 unsigned int softirq_enable_event;
1781 int softirqs_enabled;
1782 int softirq_context;
1783#endif
1784#ifdef CONFIG_LOCKDEP
1785# define MAX_LOCK_DEPTH 48UL
1786 u64 curr_chain_key;
1787 int lockdep_depth;
1788 unsigned int lockdep_recursion;
1789 struct held_lock held_locks[MAX_LOCK_DEPTH];
1790 gfp_t lockdep_reclaim_gfp;
1791#endif
1792#ifdef CONFIG_UBSAN
1793 unsigned int in_ubsan;
1794#endif
1795
1796/* journalling filesystem info */
1797 void *journal_info;
1798
1799/* stacked block device info */
1800 struct bio_list *bio_list;
1801
1802#ifdef CONFIG_BLOCK
1803/* stack plugging */
1804 struct blk_plug *plug;
1805#endif
1806
1807/* VM state */
1808 struct reclaim_state *reclaim_state;
1809
1810 struct backing_dev_info *backing_dev_info;
1811
1812 struct io_context *io_context;
1813
1814 unsigned long ptrace_message;
1815 siginfo_t *last_siginfo; /* For ptrace use. */
1816 struct task_io_accounting ioac;
1817#if defined(CONFIG_TASK_XACCT)
1818 u64 acct_rss_mem1; /* accumulated rss usage */
1819 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1820 cputime_t acct_timexpd; /* stime + utime since last update */
1821#endif
1822#ifdef CONFIG_CPUSETS
1823 nodemask_t mems_allowed; /* Protected by alloc_lock */
1824 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1825 int cpuset_mem_spread_rotor;
1826 int cpuset_slab_spread_rotor;
1827#endif
1828#ifdef CONFIG_CGROUPS
1829 /* Control Group info protected by css_set_lock */
1830 struct css_set __rcu *cgroups;
1831 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1832 struct list_head cg_list;
1833#endif
1834#ifdef CONFIG_INTEL_RDT_A
1835 int closid;
1836#endif
1837#ifdef CONFIG_FUTEX
1838 struct robust_list_head __user *robust_list;
1839#ifdef CONFIG_COMPAT
1840 struct compat_robust_list_head __user *compat_robust_list;
1841#endif
1842 struct list_head pi_state_list;
1843 struct futex_pi_state *pi_state_cache;
1844#endif
1845#ifdef CONFIG_PERF_EVENTS
1846 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1847 struct mutex perf_event_mutex;
1848 struct list_head perf_event_list;
1849#endif
1850#ifdef CONFIG_DEBUG_PREEMPT
1851 unsigned long preempt_disable_ip;
1852#endif
1853#ifdef CONFIG_NUMA
1854 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1855 short il_next;
1856 short pref_node_fork;
1857#endif
1858#ifdef CONFIG_NUMA_BALANCING
1859 int numa_scan_seq;
1860 unsigned int numa_scan_period;
1861 unsigned int numa_scan_period_max;
1862 int numa_preferred_nid;
1863 unsigned long numa_migrate_retry;
1864 u64 node_stamp; /* migration stamp */
1865 u64 last_task_numa_placement;
1866 u64 last_sum_exec_runtime;
1867 struct callback_head numa_work;
1868
1869 struct list_head numa_entry;
1870 struct numa_group *numa_group;
1871
1872 /*
1873 * numa_faults is an array split into four regions:
1874 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1875 * in this precise order.
1876 *
1877 * faults_memory: Exponential decaying average of faults on a per-node
1878 * basis. Scheduling placement decisions are made based on these
1879 * counts. The values remain static for the duration of a PTE scan.
1880 * faults_cpu: Track the nodes the process was running on when a NUMA
1881 * hinting fault was incurred.
1882 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1883 * during the current scan window. When the scan completes, the counts
1884 * in faults_memory and faults_cpu decay and these values are copied.
1885 */
1886 unsigned long *numa_faults;
1887 unsigned long total_numa_faults;
1888
1889 /*
1890 * numa_faults_locality tracks if faults recorded during the last
1891 * scan window were remote/local or failed to migrate. The task scan
1892 * period is adapted based on the locality of the faults with different
1893 * weights depending on whether they were shared or private faults
1894 */
1895 unsigned long numa_faults_locality[3];
1896
1897 unsigned long numa_pages_migrated;
1898#endif /* CONFIG_NUMA_BALANCING */
1899
1900#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1901 struct tlbflush_unmap_batch tlb_ubc;
1902#endif
1903
1904 struct rcu_head rcu;
1905
1906 /*
1907 * cache last used pipe for splice
1908 */
1909 struct pipe_inode_info *splice_pipe;
1910
1911 struct page_frag task_frag;
1912
1913#ifdef CONFIG_TASK_DELAY_ACCT
1914 struct task_delay_info *delays;
1915#endif
1916#ifdef CONFIG_FAULT_INJECTION
1917 int make_it_fail;
1918#endif
1919 /*
1920 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1921 * balance_dirty_pages() for some dirty throttling pause
1922 */
1923 int nr_dirtied;
1924 int nr_dirtied_pause;
1925 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1926
1927#ifdef CONFIG_LATENCYTOP
1928 int latency_record_count;
1929 struct latency_record latency_record[LT_SAVECOUNT];
1930#endif
1931 /*
1932 * time slack values; these are used to round up poll() and
1933 * select() etc timeout values. These are in nanoseconds.
1934 */
1935 u64 timer_slack_ns;
1936 u64 default_timer_slack_ns;
1937
1938#ifdef CONFIG_KASAN
1939 unsigned int kasan_depth;
1940#endif
1941#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1942 /* Index of current stored address in ret_stack */
1943 int curr_ret_stack;
1944 /* Stack of return addresses for return function tracing */
1945 struct ftrace_ret_stack *ret_stack;
1946 /* time stamp for last schedule */
1947 unsigned long long ftrace_timestamp;
1948 /*
1949 * Number of functions that haven't been traced
1950 * because of depth overrun.
1951 */
1952 atomic_t trace_overrun;
1953 /* Pause for the tracing */
1954 atomic_t tracing_graph_pause;
1955#endif
1956#ifdef CONFIG_TRACING
1957 /* state flags for use by tracers */
1958 unsigned long trace;
1959 /* bitmask and counter of trace recursion */
1960 unsigned long trace_recursion;
1961#endif /* CONFIG_TRACING */
1962#ifdef CONFIG_KCOV
1963 /* Coverage collection mode enabled for this task (0 if disabled). */
1964 enum kcov_mode kcov_mode;
1965 /* Size of the kcov_area. */
1966 unsigned kcov_size;
1967 /* Buffer for coverage collection. */
1968 void *kcov_area;
1969 /* kcov desciptor wired with this task or NULL. */
1970 struct kcov *kcov;
1971#endif
1972#ifdef CONFIG_MEMCG
1973 struct mem_cgroup *memcg_in_oom;
1974 gfp_t memcg_oom_gfp_mask;
1975 int memcg_oom_order;
1976
1977 /* number of pages to reclaim on returning to userland */
1978 unsigned int memcg_nr_pages_over_high;
1979#endif
1980#ifdef CONFIG_UPROBES
1981 struct uprobe_task *utask;
1982#endif
1983#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1984 unsigned int sequential_io;
1985 unsigned int sequential_io_avg;
1986#endif
1987#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1988 unsigned long task_state_change;
1989#endif
1990 int pagefault_disabled;
1991#ifdef CONFIG_MMU
1992 struct task_struct *oom_reaper_list;
1993#endif
1994#ifdef CONFIG_VMAP_STACK
1995 struct vm_struct *stack_vm_area;
1996#endif
1997#ifdef CONFIG_THREAD_INFO_IN_TASK
1998 /* A live task holds one reference. */
1999 atomic_t stack_refcount;
2000#endif
2001/* CPU-specific state of this task */
2002 struct thread_struct thread;
2003/*
2004 * WARNING: on x86, 'thread_struct' contains a variable-sized
2005 * structure. It *MUST* be at the end of 'task_struct'.
2006 *
2007 * Do not put anything below here!
2008 */
2009};
2010
2011#ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
2012extern int arch_task_struct_size __read_mostly;
2013#else
2014# define arch_task_struct_size (sizeof(struct task_struct))
2015#endif
2016
2017#ifdef CONFIG_VMAP_STACK
2018static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
2019{
2020 return t->stack_vm_area;
2021}
2022#else
2023static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
2024{
2025 return NULL;
2026}
2027#endif
2028
2029/* Future-safe accessor for struct task_struct's cpus_allowed. */
2030#define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
2031
2032static inline int tsk_nr_cpus_allowed(struct task_struct *p)
2033{
2034 return p->nr_cpus_allowed;
2035}
2036
2037#define TNF_MIGRATED 0x01
2038#define TNF_NO_GROUP 0x02
2039#define TNF_SHARED 0x04
2040#define TNF_FAULT_LOCAL 0x08
2041#define TNF_MIGRATE_FAIL 0x10
2042
2043static inline bool in_vfork(struct task_struct *tsk)
2044{
2045 bool ret;
2046
2047 /*
2048 * need RCU to access ->real_parent if CLONE_VM was used along with
2049 * CLONE_PARENT.
2050 *
2051 * We check real_parent->mm == tsk->mm because CLONE_VFORK does not
2052 * imply CLONE_VM
2053 *
2054 * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus
2055 * ->real_parent is not necessarily the task doing vfork(), so in
2056 * theory we can't rely on task_lock() if we want to dereference it.
2057 *
2058 * And in this case we can't trust the real_parent->mm == tsk->mm
2059 * check, it can be false negative. But we do not care, if init or
2060 * another oom-unkillable task does this it should blame itself.
2061 */
2062 rcu_read_lock();
2063 ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
2064 rcu_read_unlock();
2065
2066 return ret;
2067}
2068
2069#ifdef CONFIG_NUMA_BALANCING
2070extern void task_numa_fault(int last_node, int node, int pages, int flags);
2071extern pid_t task_numa_group_id(struct task_struct *p);
2072extern void set_numabalancing_state(bool enabled);
2073extern void task_numa_free(struct task_struct *p);
2074extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
2075 int src_nid, int dst_cpu);
2076#else
2077static inline void task_numa_fault(int last_node, int node, int pages,
2078 int flags)
2079{
2080}
2081static inline pid_t task_numa_group_id(struct task_struct *p)
2082{
2083 return 0;
2084}
2085static inline void set_numabalancing_state(bool enabled)
2086{
2087}
2088static inline void task_numa_free(struct task_struct *p)
2089{
2090}
2091static inline bool should_numa_migrate_memory(struct task_struct *p,
2092 struct page *page, int src_nid, int dst_cpu)
2093{
2094 return true;
2095}
2096#endif
2097
2098static inline struct pid *task_pid(struct task_struct *task)
2099{
2100 return task->pids[PIDTYPE_PID].pid;
2101}
2102
2103static inline struct pid *task_tgid(struct task_struct *task)
2104{
2105 return task->group_leader->pids[PIDTYPE_PID].pid;
2106}
2107
2108/*
2109 * Without tasklist or rcu lock it is not safe to dereference
2110 * the result of task_pgrp/task_session even if task == current,
2111 * we can race with another thread doing sys_setsid/sys_setpgid.
2112 */
2113static inline struct pid *task_pgrp(struct task_struct *task)
2114{
2115 return task->group_leader->pids[PIDTYPE_PGID].pid;
2116}
2117
2118static inline struct pid *task_session(struct task_struct *task)
2119{
2120 return task->group_leader->pids[PIDTYPE_SID].pid;
2121}
2122
2123struct pid_namespace;
2124
2125/*
2126 * the helpers to get the task's different pids as they are seen
2127 * from various namespaces
2128 *
2129 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
2130 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
2131 * current.
2132 * task_xid_nr_ns() : id seen from the ns specified;
2133 *
2134 * set_task_vxid() : assigns a virtual id to a task;
2135 *
2136 * see also pid_nr() etc in include/linux/pid.h
2137 */
2138pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2139 struct pid_namespace *ns);
2140
2141static inline pid_t task_pid_nr(struct task_struct *tsk)
2142{
2143 return tsk->pid;
2144}
2145
2146static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2147 struct pid_namespace *ns)
2148{
2149 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2150}
2151
2152static inline pid_t task_pid_vnr(struct task_struct *tsk)
2153{
2154 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2155}
2156
2157
2158static inline pid_t task_tgid_nr(struct task_struct *tsk)
2159{
2160 return tsk->tgid;
2161}
2162
2163pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2164
2165static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2166{
2167 return pid_vnr(task_tgid(tsk));
2168}
2169
2170
2171static inline int pid_alive(const struct task_struct *p);
2172static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2173{
2174 pid_t pid = 0;
2175
2176 rcu_read_lock();
2177 if (pid_alive(tsk))
2178 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2179 rcu_read_unlock();
2180
2181 return pid;
2182}
2183
2184static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2185{
2186 return task_ppid_nr_ns(tsk, &init_pid_ns);
2187}
2188
2189static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2190 struct pid_namespace *ns)
2191{
2192 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2193}
2194
2195static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2196{
2197 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2198}
2199
2200
2201static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2202 struct pid_namespace *ns)
2203{
2204 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2205}
2206
2207static inline pid_t task_session_vnr(struct task_struct *tsk)
2208{
2209 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2210}
2211
2212/* obsolete, do not use */
2213static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2214{
2215 return task_pgrp_nr_ns(tsk, &init_pid_ns);
2216}
2217
2218/**
2219 * pid_alive - check that a task structure is not stale
2220 * @p: Task structure to be checked.
2221 *
2222 * Test if a process is not yet dead (at most zombie state)
2223 * If pid_alive fails, then pointers within the task structure
2224 * can be stale and must not be dereferenced.
2225 *
2226 * Return: 1 if the process is alive. 0 otherwise.
2227 */
2228static inline int pid_alive(const struct task_struct *p)
2229{
2230 return p->pids[PIDTYPE_PID].pid != NULL;
2231}
2232
2233/**
2234 * is_global_init - check if a task structure is init. Since init
2235 * is free to have sub-threads we need to check tgid.
2236 * @tsk: Task structure to be checked.
2237 *
2238 * Check if a task structure is the first user space task the kernel created.
2239 *
2240 * Return: 1 if the task structure is init. 0 otherwise.
2241 */
2242static inline int is_global_init(struct task_struct *tsk)
2243{
2244 return task_tgid_nr(tsk) == 1;
2245}
2246
2247extern struct pid *cad_pid;
2248
2249extern void free_task(struct task_struct *tsk);
2250#define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2251
2252extern void __put_task_struct(struct task_struct *t);
2253
2254static inline void put_task_struct(struct task_struct *t)
2255{
2256 if (atomic_dec_and_test(&t->usage))
2257 __put_task_struct(t);
2258}
2259
2260struct task_struct *task_rcu_dereference(struct task_struct **ptask);
2261struct task_struct *try_get_task_struct(struct task_struct **ptask);
2262
2263#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2264extern void task_cputime(struct task_struct *t,
2265 cputime_t *utime, cputime_t *stime);
2266extern cputime_t task_gtime(struct task_struct *t);
2267#else
2268static inline void task_cputime(struct task_struct *t,
2269 cputime_t *utime, cputime_t *stime)
2270{
2271 *utime = t->utime;
2272 *stime = t->stime;
2273}
2274
2275static inline cputime_t task_gtime(struct task_struct *t)
2276{
2277 return t->gtime;
2278}
2279#endif
2280
2281#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
2282static inline void task_cputime_scaled(struct task_struct *t,
2283 cputime_t *utimescaled,
2284 cputime_t *stimescaled)
2285{
2286 *utimescaled = t->utimescaled;
2287 *stimescaled = t->stimescaled;
2288}
2289#else
2290static inline void task_cputime_scaled(struct task_struct *t,
2291 cputime_t *utimescaled,
2292 cputime_t *stimescaled)
2293{
2294 task_cputime(t, utimescaled, stimescaled);
2295}
2296#endif
2297
2298extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2299extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2300
2301/*
2302 * Per process flags
2303 */
2304#define PF_IDLE 0x00000002 /* I am an IDLE thread */
2305#define PF_EXITING 0x00000004 /* getting shut down */
2306#define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2307#define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2308#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2309#define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2310#define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2311#define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2312#define PF_DUMPCORE 0x00000200 /* dumped core */
2313#define PF_SIGNALED 0x00000400 /* killed by a signal */
2314#define PF_MEMALLOC 0x00000800 /* Allocating memory */
2315#define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2316#define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2317#define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2318#define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2319#define PF_FROZEN 0x00010000 /* frozen for system suspend */
2320#define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2321#define PF_KSWAPD 0x00040000 /* I am kswapd */
2322#define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2323#define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2324#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2325#define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2326#define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2327#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2328#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2329#define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2330#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2331#define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2332
2333/*
2334 * Only the _current_ task can read/write to tsk->flags, but other
2335 * tasks can access tsk->flags in readonly mode for example
2336 * with tsk_used_math (like during threaded core dumping).
2337 * There is however an exception to this rule during ptrace
2338 * or during fork: the ptracer task is allowed to write to the
2339 * child->flags of its traced child (same goes for fork, the parent
2340 * can write to the child->flags), because we're guaranteed the
2341 * child is not running and in turn not changing child->flags
2342 * at the same time the parent does it.
2343 */
2344#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2345#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2346#define clear_used_math() clear_stopped_child_used_math(current)
2347#define set_used_math() set_stopped_child_used_math(current)
2348#define conditional_stopped_child_used_math(condition, child) \
2349 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2350#define conditional_used_math(condition) \
2351 conditional_stopped_child_used_math(condition, current)
2352#define copy_to_stopped_child_used_math(child) \
2353 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2354/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2355#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2356#define used_math() tsk_used_math(current)
2357
2358/* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2359 * __GFP_FS is also cleared as it implies __GFP_IO.
2360 */
2361static inline gfp_t memalloc_noio_flags(gfp_t flags)
2362{
2363 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2364 flags &= ~(__GFP_IO | __GFP_FS);
2365 return flags;
2366}
2367
2368static inline unsigned int memalloc_noio_save(void)
2369{
2370 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2371 current->flags |= PF_MEMALLOC_NOIO;
2372 return flags;
2373}
2374
2375static inline void memalloc_noio_restore(unsigned int flags)
2376{
2377 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2378}
2379
2380/* Per-process atomic flags. */
2381#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2382#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2383#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2384#define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
2385
2386
2387#define TASK_PFA_TEST(name, func) \
2388 static inline bool task_##func(struct task_struct *p) \
2389 { return test_bit(PFA_##name, &p->atomic_flags); }
2390#define TASK_PFA_SET(name, func) \
2391 static inline void task_set_##func(struct task_struct *p) \
2392 { set_bit(PFA_##name, &p->atomic_flags); }
2393#define TASK_PFA_CLEAR(name, func) \
2394 static inline void task_clear_##func(struct task_struct *p) \
2395 { clear_bit(PFA_##name, &p->atomic_flags); }
2396
2397TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2398TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2399
2400TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2401TASK_PFA_SET(SPREAD_PAGE, spread_page)
2402TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2403
2404TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2405TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2406TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2407
2408TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
2409TASK_PFA_SET(LMK_WAITING, lmk_waiting)
2410
2411/*
2412 * task->jobctl flags
2413 */
2414#define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2415
2416#define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2417#define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2418#define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2419#define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2420#define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2421#define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2422#define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2423
2424#define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2425#define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2426#define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2427#define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2428#define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2429#define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2430#define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2431
2432#define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2433#define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2434
2435extern bool task_set_jobctl_pending(struct task_struct *task,
2436 unsigned long mask);
2437extern void task_clear_jobctl_trapping(struct task_struct *task);
2438extern void task_clear_jobctl_pending(struct task_struct *task,
2439 unsigned long mask);
2440
2441static inline void rcu_copy_process(struct task_struct *p)
2442{
2443#ifdef CONFIG_PREEMPT_RCU
2444 p->rcu_read_lock_nesting = 0;
2445 p->rcu_read_unlock_special.s = 0;
2446 p->rcu_blocked_node = NULL;
2447 INIT_LIST_HEAD(&p->rcu_node_entry);
2448#endif /* #ifdef CONFIG_PREEMPT_RCU */
2449#ifdef CONFIG_TASKS_RCU
2450 p->rcu_tasks_holdout = false;
2451 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2452 p->rcu_tasks_idle_cpu = -1;
2453#endif /* #ifdef CONFIG_TASKS_RCU */
2454}
2455
2456static inline void tsk_restore_flags(struct task_struct *task,
2457 unsigned long orig_flags, unsigned long flags)
2458{
2459 task->flags &= ~flags;
2460 task->flags |= orig_flags & flags;
2461}
2462
2463extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2464 const struct cpumask *trial);
2465extern int task_can_attach(struct task_struct *p,
2466 const struct cpumask *cs_cpus_allowed);
2467#ifdef CONFIG_SMP
2468extern void do_set_cpus_allowed(struct task_struct *p,
2469 const struct cpumask *new_mask);
2470
2471extern int set_cpus_allowed_ptr(struct task_struct *p,
2472 const struct cpumask *new_mask);
2473#else
2474static inline void do_set_cpus_allowed(struct task_struct *p,
2475 const struct cpumask *new_mask)
2476{
2477}
2478static inline int set_cpus_allowed_ptr(struct task_struct *p,
2479 const struct cpumask *new_mask)
2480{
2481 if (!cpumask_test_cpu(0, new_mask))
2482 return -EINVAL;
2483 return 0;
2484}
2485#endif
2486
2487#ifdef CONFIG_NO_HZ_COMMON
2488void calc_load_enter_idle(void);
2489void calc_load_exit_idle(void);
2490#else
2491static inline void calc_load_enter_idle(void) { }
2492static inline void calc_load_exit_idle(void) { }
2493#endif /* CONFIG_NO_HZ_COMMON */
2494
2495#ifndef cpu_relax_yield
2496#define cpu_relax_yield() cpu_relax()
2497#endif
2498
2499/*
2500 * Do not use outside of architecture code which knows its limitations.
2501 *
2502 * sched_clock() has no promise of monotonicity or bounded drift between
2503 * CPUs, use (which you should not) requires disabling IRQs.
2504 *
2505 * Please use one of the three interfaces below.
2506 */
2507extern unsigned long long notrace sched_clock(void);
2508/*
2509 * See the comment in kernel/sched/clock.c
2510 */
2511extern u64 running_clock(void);
2512extern u64 sched_clock_cpu(int cpu);
2513
2514
2515extern void sched_clock_init(void);
2516
2517#ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2518static inline void sched_clock_tick(void)
2519{
2520}
2521
2522static inline void sched_clock_idle_sleep_event(void)
2523{
2524}
2525
2526static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2527{
2528}
2529
2530static inline u64 cpu_clock(int cpu)
2531{
2532 return sched_clock();
2533}
2534
2535static inline u64 local_clock(void)
2536{
2537 return sched_clock();
2538}
2539#else
2540/*
2541 * Architectures can set this to 1 if they have specified
2542 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2543 * but then during bootup it turns out that sched_clock()
2544 * is reliable after all:
2545 */
2546extern int sched_clock_stable(void);
2547extern void set_sched_clock_stable(void);
2548extern void clear_sched_clock_stable(void);
2549
2550extern void sched_clock_tick(void);
2551extern void sched_clock_idle_sleep_event(void);
2552extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2553
2554/*
2555 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2556 * time source that is monotonic per cpu argument and has bounded drift
2557 * between cpus.
2558 *
2559 * ######################### BIG FAT WARNING ##########################
2560 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2561 * # go backwards !! #
2562 * ####################################################################
2563 */
2564static inline u64 cpu_clock(int cpu)
2565{
2566 return sched_clock_cpu(cpu);
2567}
2568
2569static inline u64 local_clock(void)
2570{
2571 return sched_clock_cpu(raw_smp_processor_id());
2572}
2573#endif
2574
2575#ifdef CONFIG_IRQ_TIME_ACCOUNTING
2576/*
2577 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2578 * The reason for this explicit opt-in is not to have perf penalty with
2579 * slow sched_clocks.
2580 */
2581extern void enable_sched_clock_irqtime(void);
2582extern void disable_sched_clock_irqtime(void);
2583#else
2584static inline void enable_sched_clock_irqtime(void) {}
2585static inline void disable_sched_clock_irqtime(void) {}
2586#endif
2587
2588extern unsigned long long
2589task_sched_runtime(struct task_struct *task);
2590
2591/* sched_exec is called by processes performing an exec */
2592#ifdef CONFIG_SMP
2593extern void sched_exec(void);
2594#else
2595#define sched_exec() {}
2596#endif
2597
2598extern void sched_clock_idle_sleep_event(void);
2599extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2600
2601#ifdef CONFIG_HOTPLUG_CPU
2602extern void idle_task_exit(void);
2603#else
2604static inline void idle_task_exit(void) {}
2605#endif
2606
2607#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2608extern void wake_up_nohz_cpu(int cpu);
2609#else
2610static inline void wake_up_nohz_cpu(int cpu) { }
2611#endif
2612
2613#ifdef CONFIG_NO_HZ_FULL
2614extern u64 scheduler_tick_max_deferment(void);
2615#endif
2616
2617#ifdef CONFIG_SCHED_AUTOGROUP
2618extern void sched_autogroup_create_attach(struct task_struct *p);
2619extern void sched_autogroup_detach(struct task_struct *p);
2620extern void sched_autogroup_fork(struct signal_struct *sig);
2621extern void sched_autogroup_exit(struct signal_struct *sig);
2622extern void sched_autogroup_exit_task(struct task_struct *p);
2623#ifdef CONFIG_PROC_FS
2624extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2625extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2626#endif
2627#else
2628static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2629static inline void sched_autogroup_detach(struct task_struct *p) { }
2630static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2631static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2632static inline void sched_autogroup_exit_task(struct task_struct *p) { }
2633#endif
2634
2635extern int yield_to(struct task_struct *p, bool preempt);
2636extern void set_user_nice(struct task_struct *p, long nice);
2637extern int task_prio(const struct task_struct *p);
2638/**
2639 * task_nice - return the nice value of a given task.
2640 * @p: the task in question.
2641 *
2642 * Return: The nice value [ -20 ... 0 ... 19 ].
2643 */
2644static inline int task_nice(const struct task_struct *p)
2645{
2646 return PRIO_TO_NICE((p)->static_prio);
2647}
2648extern int can_nice(const struct task_struct *p, const int nice);
2649extern int task_curr(const struct task_struct *p);
2650extern int idle_cpu(int cpu);
2651extern int sched_setscheduler(struct task_struct *, int,
2652 const struct sched_param *);
2653extern int sched_setscheduler_nocheck(struct task_struct *, int,
2654 const struct sched_param *);
2655extern int sched_setattr(struct task_struct *,
2656 const struct sched_attr *);
2657extern struct task_struct *idle_task(int cpu);
2658/**
2659 * is_idle_task - is the specified task an idle task?
2660 * @p: the task in question.
2661 *
2662 * Return: 1 if @p is an idle task. 0 otherwise.
2663 */
2664static inline bool is_idle_task(const struct task_struct *p)
2665{
2666 return !!(p->flags & PF_IDLE);
2667}
2668extern struct task_struct *curr_task(int cpu);
2669extern void ia64_set_curr_task(int cpu, struct task_struct *p);
2670
2671void yield(void);
2672
2673union thread_union {
2674#ifndef CONFIG_THREAD_INFO_IN_TASK
2675 struct thread_info thread_info;
2676#endif
2677 unsigned long stack[THREAD_SIZE/sizeof(long)];
2678};
2679
2680#ifndef __HAVE_ARCH_KSTACK_END
2681static inline int kstack_end(void *addr)
2682{
2683 /* Reliable end of stack detection:
2684 * Some APM bios versions misalign the stack
2685 */
2686 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2687}
2688#endif
2689
2690extern union thread_union init_thread_union;
2691extern struct task_struct init_task;
2692
2693extern struct mm_struct init_mm;
2694
2695extern struct pid_namespace init_pid_ns;
2696
2697/*
2698 * find a task by one of its numerical ids
2699 *
2700 * find_task_by_pid_ns():
2701 * finds a task by its pid in the specified namespace
2702 * find_task_by_vpid():
2703 * finds a task by its virtual pid
2704 *
2705 * see also find_vpid() etc in include/linux/pid.h
2706 */
2707
2708extern struct task_struct *find_task_by_vpid(pid_t nr);
2709extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2710 struct pid_namespace *ns);
2711
2712/* per-UID process charging. */
2713extern struct user_struct * alloc_uid(kuid_t);
2714static inline struct user_struct *get_uid(struct user_struct *u)
2715{
2716 atomic_inc(&u->__count);
2717 return u;
2718}
2719extern void free_uid(struct user_struct *);
2720
2721#include <asm/current.h>
2722
2723extern void xtime_update(unsigned long ticks);
2724
2725extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2726extern int wake_up_process(struct task_struct *tsk);
2727extern void wake_up_new_task(struct task_struct *tsk);
2728#ifdef CONFIG_SMP
2729 extern void kick_process(struct task_struct *tsk);
2730#else
2731 static inline void kick_process(struct task_struct *tsk) { }
2732#endif
2733extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2734extern void sched_dead(struct task_struct *p);
2735
2736extern void proc_caches_init(void);
2737extern void flush_signals(struct task_struct *);
2738extern void ignore_signals(struct task_struct *);
2739extern void flush_signal_handlers(struct task_struct *, int force_default);
2740extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2741
2742static inline int kernel_dequeue_signal(siginfo_t *info)
2743{
2744 struct task_struct *tsk = current;
2745 siginfo_t __info;
2746 int ret;
2747
2748 spin_lock_irq(&tsk->sighand->siglock);
2749 ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2750 spin_unlock_irq(&tsk->sighand->siglock);
2751
2752 return ret;
2753}
2754
2755static inline void kernel_signal_stop(void)
2756{
2757 spin_lock_irq(&current->sighand->siglock);
2758 if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2759 __set_current_state(TASK_STOPPED);
2760 spin_unlock_irq(&current->sighand->siglock);
2761
2762 schedule();
2763}
2764
2765extern void release_task(struct task_struct * p);
2766extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2767extern int force_sigsegv(int, struct task_struct *);
2768extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2769extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2770extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2771extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2772 const struct cred *, u32);
2773extern int kill_pgrp(struct pid *pid, int sig, int priv);
2774extern int kill_pid(struct pid *pid, int sig, int priv);
2775extern int kill_proc_info(int, struct siginfo *, pid_t);
2776extern __must_check bool do_notify_parent(struct task_struct *, int);
2777extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2778extern void force_sig(int, struct task_struct *);
2779extern int send_sig(int, struct task_struct *, int);
2780extern int zap_other_threads(struct task_struct *p);
2781extern struct sigqueue *sigqueue_alloc(void);
2782extern void sigqueue_free(struct sigqueue *);
2783extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2784extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2785
2786#ifdef TIF_RESTORE_SIGMASK
2787/*
2788 * Legacy restore_sigmask accessors. These are inefficient on
2789 * SMP architectures because they require atomic operations.
2790 */
2791
2792/**
2793 * set_restore_sigmask() - make sure saved_sigmask processing gets done
2794 *
2795 * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code
2796 * will run before returning to user mode, to process the flag. For
2797 * all callers, TIF_SIGPENDING is already set or it's no harm to set
2798 * it. TIF_RESTORE_SIGMASK need not be in the set of bits that the
2799 * arch code will notice on return to user mode, in case those bits
2800 * are scarce. We set TIF_SIGPENDING here to ensure that the arch
2801 * signal code always gets run when TIF_RESTORE_SIGMASK is set.
2802 */
2803static inline void set_restore_sigmask(void)
2804{
2805 set_thread_flag(TIF_RESTORE_SIGMASK);
2806 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2807}
2808static inline void clear_restore_sigmask(void)
2809{
2810 clear_thread_flag(TIF_RESTORE_SIGMASK);
2811}
2812static inline bool test_restore_sigmask(void)
2813{
2814 return test_thread_flag(TIF_RESTORE_SIGMASK);
2815}
2816static inline bool test_and_clear_restore_sigmask(void)
2817{
2818 return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK);
2819}
2820
2821#else /* TIF_RESTORE_SIGMASK */
2822
2823/* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */
2824static inline void set_restore_sigmask(void)
2825{
2826 current->restore_sigmask = true;
2827 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2828}
2829static inline void clear_restore_sigmask(void)
2830{
2831 current->restore_sigmask = false;
2832}
2833static inline bool test_restore_sigmask(void)
2834{
2835 return current->restore_sigmask;
2836}
2837static inline bool test_and_clear_restore_sigmask(void)
2838{
2839 if (!current->restore_sigmask)
2840 return false;
2841 current->restore_sigmask = false;
2842 return true;
2843}
2844#endif
2845
2846static inline void restore_saved_sigmask(void)
2847{
2848 if (test_and_clear_restore_sigmask())
2849 __set_current_blocked(&current->saved_sigmask);
2850}
2851
2852static inline sigset_t *sigmask_to_save(void)
2853{
2854 sigset_t *res = &current->blocked;
2855 if (unlikely(test_restore_sigmask()))
2856 res = &current->saved_sigmask;
2857 return res;
2858}
2859
2860static inline int kill_cad_pid(int sig, int priv)
2861{
2862 return kill_pid(cad_pid, sig, priv);
2863}
2864
2865/* These can be the second arg to send_sig_info/send_group_sig_info. */
2866#define SEND_SIG_NOINFO ((struct siginfo *) 0)
2867#define SEND_SIG_PRIV ((struct siginfo *) 1)
2868#define SEND_SIG_FORCED ((struct siginfo *) 2)
2869
2870/*
2871 * True if we are on the alternate signal stack.
2872 */
2873static inline int on_sig_stack(unsigned long sp)
2874{
2875 /*
2876 * If the signal stack is SS_AUTODISARM then, by construction, we
2877 * can't be on the signal stack unless user code deliberately set
2878 * SS_AUTODISARM when we were already on it.
2879 *
2880 * This improves reliability: if user state gets corrupted such that
2881 * the stack pointer points very close to the end of the signal stack,
2882 * then this check will enable the signal to be handled anyway.
2883 */
2884 if (current->sas_ss_flags & SS_AUTODISARM)
2885 return 0;
2886
2887#ifdef CONFIG_STACK_GROWSUP
2888 return sp >= current->sas_ss_sp &&
2889 sp - current->sas_ss_sp < current->sas_ss_size;
2890#else
2891 return sp > current->sas_ss_sp &&
2892 sp - current->sas_ss_sp <= current->sas_ss_size;
2893#endif
2894}
2895
2896static inline int sas_ss_flags(unsigned long sp)
2897{
2898 if (!current->sas_ss_size)
2899 return SS_DISABLE;
2900
2901 return on_sig_stack(sp) ? SS_ONSTACK : 0;
2902}
2903
2904static inline void sas_ss_reset(struct task_struct *p)
2905{
2906 p->sas_ss_sp = 0;
2907 p->sas_ss_size = 0;
2908 p->sas_ss_flags = SS_DISABLE;
2909}
2910
2911static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2912{
2913 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2914#ifdef CONFIG_STACK_GROWSUP
2915 return current->sas_ss_sp;
2916#else
2917 return current->sas_ss_sp + current->sas_ss_size;
2918#endif
2919 return sp;
2920}
2921
2922/*
2923 * Routines for handling mm_structs
2924 */
2925extern struct mm_struct * mm_alloc(void);
2926
2927/* mmdrop drops the mm and the page tables */
2928extern void __mmdrop(struct mm_struct *);
2929static inline void mmdrop(struct mm_struct *mm)
2930{
2931 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2932 __mmdrop(mm);
2933}
2934
2935static inline void mmdrop_async_fn(struct work_struct *work)
2936{
2937 struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
2938 __mmdrop(mm);
2939}
2940
2941static inline void mmdrop_async(struct mm_struct *mm)
2942{
2943 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
2944 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
2945 schedule_work(&mm->async_put_work);
2946 }
2947}
2948
2949static inline bool mmget_not_zero(struct mm_struct *mm)
2950{
2951 return atomic_inc_not_zero(&mm->mm_users);
2952}
2953
2954/* mmput gets rid of the mappings and all user-space */
2955extern void mmput(struct mm_struct *);
2956#ifdef CONFIG_MMU
2957/* same as above but performs the slow path from the async context. Can
2958 * be called from the atomic context as well
2959 */
2960extern void mmput_async(struct mm_struct *);
2961#endif
2962
2963/* Grab a reference to a task's mm, if it is not already going away */
2964extern struct mm_struct *get_task_mm(struct task_struct *task);
2965/*
2966 * Grab a reference to a task's mm, if it is not already going away
2967 * and ptrace_may_access with the mode parameter passed to it
2968 * succeeds.
2969 */
2970extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2971/* Remove the current tasks stale references to the old mm_struct */
2972extern void mm_release(struct task_struct *, struct mm_struct *);
2973
2974#ifdef CONFIG_HAVE_COPY_THREAD_TLS
2975extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2976 struct task_struct *, unsigned long);
2977#else
2978extern int copy_thread(unsigned long, unsigned long, unsigned long,
2979 struct task_struct *);
2980
2981/* Architectures that haven't opted into copy_thread_tls get the tls argument
2982 * via pt_regs, so ignore the tls argument passed via C. */
2983static inline int copy_thread_tls(
2984 unsigned long clone_flags, unsigned long sp, unsigned long arg,
2985 struct task_struct *p, unsigned long tls)
2986{
2987 return copy_thread(clone_flags, sp, arg, p);
2988}
2989#endif
2990extern void flush_thread(void);
2991
2992#ifdef CONFIG_HAVE_EXIT_THREAD
2993extern void exit_thread(struct task_struct *tsk);
2994#else
2995static inline void exit_thread(struct task_struct *tsk)
2996{
2997}
2998#endif
2999
3000extern void exit_files(struct task_struct *);
3001extern void __cleanup_sighand(struct sighand_struct *);
3002
3003extern void exit_itimers(struct signal_struct *);
3004extern void flush_itimer_signals(void);
3005
3006extern void do_group_exit(int);
3007
3008extern int do_execve(struct filename *,
3009 const char __user * const __user *,
3010 const char __user * const __user *);
3011extern int do_execveat(int, struct filename *,
3012 const char __user * const __user *,
3013 const char __user * const __user *,
3014 int);
3015extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
3016extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
3017struct task_struct *fork_idle(int);
3018extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
3019
3020extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
3021static inline void set_task_comm(struct task_struct *tsk, const char *from)
3022{
3023 __set_task_comm(tsk, from, false);
3024}
3025extern char *get_task_comm(char *to, struct task_struct *tsk);
3026
3027#ifdef CONFIG_SMP
3028void scheduler_ipi(void);
3029extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
3030#else
3031static inline void scheduler_ipi(void) { }
3032static inline unsigned long wait_task_inactive(struct task_struct *p,
3033 long match_state)
3034{
3035 return 1;
3036}
3037#endif
3038
3039#define tasklist_empty() \
3040 list_empty(&init_task.tasks)
3041
3042#define next_task(p) \
3043 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
3044
3045#define for_each_process(p) \
3046 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
3047
3048extern bool current_is_single_threaded(void);
3049
3050/*
3051 * Careful: do_each_thread/while_each_thread is a double loop so
3052 * 'break' will not work as expected - use goto instead.
3053 */
3054#define do_each_thread(g, t) \
3055 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
3056
3057#define while_each_thread(g, t) \
3058 while ((t = next_thread(t)) != g)
3059
3060#define __for_each_thread(signal, t) \
3061 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
3062
3063#define for_each_thread(p, t) \
3064 __for_each_thread((p)->signal, t)
3065
3066/* Careful: this is a double loop, 'break' won't work as expected. */
3067#define for_each_process_thread(p, t) \
3068 for_each_process(p) for_each_thread(p, t)
3069
3070static inline int get_nr_threads(struct task_struct *tsk)
3071{
3072 return tsk->signal->nr_threads;
3073}
3074
3075static inline bool thread_group_leader(struct task_struct *p)
3076{
3077 return p->exit_signal >= 0;
3078}
3079
3080/* Do to the insanities of de_thread it is possible for a process
3081 * to have the pid of the thread group leader without actually being
3082 * the thread group leader. For iteration through the pids in proc
3083 * all we care about is that we have a task with the appropriate
3084 * pid, we don't actually care if we have the right task.
3085 */
3086static inline bool has_group_leader_pid(struct task_struct *p)
3087{
3088 return task_pid(p) == p->signal->leader_pid;
3089}
3090
3091static inline
3092bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
3093{
3094 return p1->signal == p2->signal;
3095}
3096
3097static inline struct task_struct *next_thread(const struct task_struct *p)
3098{
3099 return list_entry_rcu(p->thread_group.next,
3100 struct task_struct, thread_group);
3101}
3102
3103static inline int thread_group_empty(struct task_struct *p)
3104{
3105 return list_empty(&p->thread_group);
3106}
3107
3108#define delay_group_leader(p) \
3109 (thread_group_leader(p) && !thread_group_empty(p))
3110
3111/*
3112 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
3113 * subscriptions and synchronises with wait4(). Also used in procfs. Also
3114 * pins the final release of task.io_context. Also protects ->cpuset and
3115 * ->cgroup.subsys[]. And ->vfork_done.
3116 *
3117 * Nests both inside and outside of read_lock(&tasklist_lock).
3118 * It must not be nested with write_lock_irq(&tasklist_lock),
3119 * neither inside nor outside.
3120 */
3121static inline void task_lock(struct task_struct *p)
3122{
3123 spin_lock(&p->alloc_lock);
3124}
3125
3126static inline void task_unlock(struct task_struct *p)
3127{
3128 spin_unlock(&p->alloc_lock);
3129}
3130
3131extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
3132 unsigned long *flags);
3133
3134static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
3135 unsigned long *flags)
3136{
3137 struct sighand_struct *ret;
3138
3139 ret = __lock_task_sighand(tsk, flags);
3140 (void)__cond_lock(&tsk->sighand->siglock, ret);
3141 return ret;
3142}
3143
3144static inline void unlock_task_sighand(struct task_struct *tsk,
3145 unsigned long *flags)
3146{
3147 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
3148}
3149
3150/**
3151 * threadgroup_change_begin - mark the beginning of changes to a threadgroup
3152 * @tsk: task causing the changes
3153 *
3154 * All operations which modify a threadgroup - a new thread joining the
3155 * group, death of a member thread (the assertion of PF_EXITING) and
3156 * exec(2) dethreading the process and replacing the leader - are wrapped
3157 * by threadgroup_change_{begin|end}(). This is to provide a place which
3158 * subsystems needing threadgroup stability can hook into for
3159 * synchronization.
3160 */
3161static inline void threadgroup_change_begin(struct task_struct *tsk)
3162{
3163 might_sleep();
3164 cgroup_threadgroup_change_begin(tsk);
3165}
3166
3167/**
3168 * threadgroup_change_end - mark the end of changes to a threadgroup
3169 * @tsk: task causing the changes
3170 *
3171 * See threadgroup_change_begin().
3172 */
3173static inline void threadgroup_change_end(struct task_struct *tsk)
3174{
3175 cgroup_threadgroup_change_end(tsk);
3176}
3177
3178#ifdef CONFIG_THREAD_INFO_IN_TASK
3179
3180static inline struct thread_info *task_thread_info(struct task_struct *task)
3181{
3182 return &task->thread_info;
3183}
3184
3185/*
3186 * When accessing the stack of a non-current task that might exit, use
3187 * try_get_task_stack() instead. task_stack_page will return a pointer
3188 * that could get freed out from under you.
3189 */
3190static inline void *task_stack_page(const struct task_struct *task)
3191{
3192 return task->stack;
3193}
3194
3195#define setup_thread_stack(new,old) do { } while(0)
3196
3197static inline unsigned long *end_of_stack(const struct task_struct *task)
3198{
3199 return task->stack;
3200}
3201
3202#elif !defined(__HAVE_THREAD_FUNCTIONS)
3203
3204#define task_thread_info(task) ((struct thread_info *)(task)->stack)
3205#define task_stack_page(task) ((void *)(task)->stack)
3206
3207static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
3208{
3209 *task_thread_info(p) = *task_thread_info(org);
3210 task_thread_info(p)->task = p;
3211}
3212
3213/*
3214 * Return the address of the last usable long on the stack.
3215 *
3216 * When the stack grows down, this is just above the thread
3217 * info struct. Going any lower will corrupt the threadinfo.
3218 *
3219 * When the stack grows up, this is the highest address.
3220 * Beyond that position, we corrupt data on the next page.
3221 */
3222static inline unsigned long *end_of_stack(struct task_struct *p)
3223{
3224#ifdef CONFIG_STACK_GROWSUP
3225 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
3226#else
3227 return (unsigned long *)(task_thread_info(p) + 1);
3228#endif
3229}
3230
3231#endif
3232
3233#ifdef CONFIG_THREAD_INFO_IN_TASK
3234static inline void *try_get_task_stack(struct task_struct *tsk)
3235{
3236 return atomic_inc_not_zero(&tsk->stack_refcount) ?
3237 task_stack_page(tsk) : NULL;
3238}
3239
3240extern void put_task_stack(struct task_struct *tsk);
3241#else
3242static inline void *try_get_task_stack(struct task_struct *tsk)
3243{
3244 return task_stack_page(tsk);
3245}
3246
3247static inline void put_task_stack(struct task_struct *tsk) {}
3248#endif
3249
3250#define task_stack_end_corrupted(task) \
3251 (*(end_of_stack(task)) != STACK_END_MAGIC)
3252
3253static inline int object_is_on_stack(void *obj)
3254{
3255 void *stack = task_stack_page(current);
3256
3257 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3258}
3259
3260extern void thread_stack_cache_init(void);
3261
3262#ifdef CONFIG_DEBUG_STACK_USAGE
3263static inline unsigned long stack_not_used(struct task_struct *p)
3264{
3265 unsigned long *n = end_of_stack(p);
3266
3267 do { /* Skip over canary */
3268# ifdef CONFIG_STACK_GROWSUP
3269 n--;
3270# else
3271 n++;
3272# endif
3273 } while (!*n);
3274
3275# ifdef CONFIG_STACK_GROWSUP
3276 return (unsigned long)end_of_stack(p) - (unsigned long)n;
3277# else
3278 return (unsigned long)n - (unsigned long)end_of_stack(p);
3279# endif
3280}
3281#endif
3282extern void set_task_stack_end_magic(struct task_struct *tsk);
3283
3284/* set thread flags in other task's structures
3285 * - see asm/thread_info.h for TIF_xxxx flags available
3286 */
3287static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3288{
3289 set_ti_thread_flag(task_thread_info(tsk), flag);
3290}
3291
3292static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3293{
3294 clear_ti_thread_flag(task_thread_info(tsk), flag);
3295}
3296
3297static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3298{
3299 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3300}
3301
3302static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3303{
3304 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3305}
3306
3307static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3308{
3309 return test_ti_thread_flag(task_thread_info(tsk), flag);
3310}
3311
3312static inline void set_tsk_need_resched(struct task_struct *tsk)
3313{
3314 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3315}
3316
3317static inline void clear_tsk_need_resched(struct task_struct *tsk)
3318{
3319 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3320}
3321
3322static inline int test_tsk_need_resched(struct task_struct *tsk)
3323{
3324 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3325}
3326
3327static inline int restart_syscall(void)
3328{
3329 set_tsk_thread_flag(current, TIF_SIGPENDING);
3330 return -ERESTARTNOINTR;
3331}
3332
3333static inline int signal_pending(struct task_struct *p)
3334{
3335 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3336}
3337
3338static inline int __fatal_signal_pending(struct task_struct *p)
3339{
3340 return unlikely(sigismember(&p->pending.signal, SIGKILL));
3341}
3342
3343static inline int fatal_signal_pending(struct task_struct *p)
3344{
3345 return signal_pending(p) && __fatal_signal_pending(p);
3346}
3347
3348static inline int signal_pending_state(long state, struct task_struct *p)
3349{
3350 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3351 return 0;
3352 if (!signal_pending(p))
3353 return 0;
3354
3355 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3356}
3357
3358/*
3359 * cond_resched() and cond_resched_lock(): latency reduction via
3360 * explicit rescheduling in places that are safe. The return
3361 * value indicates whether a reschedule was done in fact.
3362 * cond_resched_lock() will drop the spinlock before scheduling,
3363 * cond_resched_softirq() will enable bhs before scheduling.
3364 */
3365#ifndef CONFIG_PREEMPT
3366extern int _cond_resched(void);
3367#else
3368static inline int _cond_resched(void) { return 0; }
3369#endif
3370
3371#define cond_resched() ({ \
3372 ___might_sleep(__FILE__, __LINE__, 0); \
3373 _cond_resched(); \
3374})
3375
3376extern int __cond_resched_lock(spinlock_t *lock);
3377
3378#define cond_resched_lock(lock) ({ \
3379 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3380 __cond_resched_lock(lock); \
3381})
3382
3383extern int __cond_resched_softirq(void);
3384
3385#define cond_resched_softirq() ({ \
3386 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3387 __cond_resched_softirq(); \
3388})
3389
3390static inline void cond_resched_rcu(void)
3391{
3392#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3393 rcu_read_unlock();
3394 cond_resched();
3395 rcu_read_lock();
3396#endif
3397}
3398
3399static inline unsigned long get_preempt_disable_ip(struct task_struct *p)
3400{
3401#ifdef CONFIG_DEBUG_PREEMPT
3402 return p->preempt_disable_ip;
3403#else
3404 return 0;
3405#endif
3406}
3407
3408/*
3409 * Does a critical section need to be broken due to another
3410 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3411 * but a general need for low latency)
3412 */
3413static inline int spin_needbreak(spinlock_t *lock)
3414{
3415#ifdef CONFIG_PREEMPT
3416 return spin_is_contended(lock);
3417#else
3418 return 0;
3419#endif
3420}
3421
3422/*
3423 * Idle thread specific functions to determine the need_resched
3424 * polling state.
3425 */
3426#ifdef TIF_POLLING_NRFLAG
3427static inline int tsk_is_polling(struct task_struct *p)
3428{
3429 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3430}
3431
3432static inline void __current_set_polling(void)
3433{
3434 set_thread_flag(TIF_POLLING_NRFLAG);
3435}
3436
3437static inline bool __must_check current_set_polling_and_test(void)
3438{
3439 __current_set_polling();
3440
3441 /*
3442 * Polling state must be visible before we test NEED_RESCHED,
3443 * paired by resched_curr()
3444 */
3445 smp_mb__after_atomic();
3446
3447 return unlikely(tif_need_resched());
3448}
3449
3450static inline void __current_clr_polling(void)
3451{
3452 clear_thread_flag(TIF_POLLING_NRFLAG);
3453}
3454
3455static inline bool __must_check current_clr_polling_and_test(void)
3456{
3457 __current_clr_polling();
3458
3459 /*
3460 * Polling state must be visible before we test NEED_RESCHED,
3461 * paired by resched_curr()
3462 */
3463 smp_mb__after_atomic();
3464
3465 return unlikely(tif_need_resched());
3466}
3467
3468#else
3469static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3470static inline void __current_set_polling(void) { }
3471static inline void __current_clr_polling(void) { }
3472
3473static inline bool __must_check current_set_polling_and_test(void)
3474{
3475 return unlikely(tif_need_resched());
3476}
3477static inline bool __must_check current_clr_polling_and_test(void)
3478{
3479 return unlikely(tif_need_resched());
3480}
3481#endif
3482
3483static inline void current_clr_polling(void)
3484{
3485 __current_clr_polling();
3486
3487 /*
3488 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3489 * Once the bit is cleared, we'll get IPIs with every new
3490 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3491 * fold.
3492 */
3493 smp_mb(); /* paired with resched_curr() */
3494
3495 preempt_fold_need_resched();
3496}
3497
3498static __always_inline bool need_resched(void)
3499{
3500 return unlikely(tif_need_resched());
3501}
3502
3503/*
3504 * Thread group CPU time accounting.
3505 */
3506void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3507void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3508
3509/*
3510 * Reevaluate whether the task has signals pending delivery.
3511 * Wake the task if so.
3512 * This is required every time the blocked sigset_t changes.
3513 * callers must hold sighand->siglock.
3514 */
3515extern void recalc_sigpending_and_wake(struct task_struct *t);
3516extern void recalc_sigpending(void);
3517
3518extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3519
3520static inline void signal_wake_up(struct task_struct *t, bool resume)
3521{
3522 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3523}
3524static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3525{
3526 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3527}
3528
3529/*
3530 * Wrappers for p->thread_info->cpu access. No-op on UP.
3531 */
3532#ifdef CONFIG_SMP
3533
3534static inline unsigned int task_cpu(const struct task_struct *p)
3535{
3536#ifdef CONFIG_THREAD_INFO_IN_TASK
3537 return p->cpu;
3538#else
3539 return task_thread_info(p)->cpu;
3540#endif
3541}
3542
3543static inline int task_node(const struct task_struct *p)
3544{
3545 return cpu_to_node(task_cpu(p));
3546}
3547
3548extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3549
3550#else
3551
3552static inline unsigned int task_cpu(const struct task_struct *p)
3553{
3554 return 0;
3555}
3556
3557static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3558{
3559}
3560
3561#endif /* CONFIG_SMP */
3562
3563/*
3564 * In order to reduce various lock holder preemption latencies provide an
3565 * interface to see if a vCPU is currently running or not.
3566 *
3567 * This allows us to terminate optimistic spin loops and block, analogous to
3568 * the native optimistic spin heuristic of testing if the lock owner task is
3569 * running or not.
3570 */
3571#ifndef vcpu_is_preempted
3572# define vcpu_is_preempted(cpu) false
3573#endif
3574
3575extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3576extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3577
3578#ifdef CONFIG_CGROUP_SCHED
3579extern struct task_group root_task_group;
3580#endif /* CONFIG_CGROUP_SCHED */
3581
3582extern int task_can_switch_user(struct user_struct *up,
3583 struct task_struct *tsk);
3584
3585#ifdef CONFIG_TASK_XACCT
3586static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3587{
3588 tsk->ioac.rchar += amt;
3589}
3590
3591static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3592{
3593 tsk->ioac.wchar += amt;
3594}
3595
3596static inline void inc_syscr(struct task_struct *tsk)
3597{
3598 tsk->ioac.syscr++;
3599}
3600
3601static inline void inc_syscw(struct task_struct *tsk)
3602{
3603 tsk->ioac.syscw++;
3604}
3605#else
3606static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3607{
3608}
3609
3610static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3611{
3612}
3613
3614static inline void inc_syscr(struct task_struct *tsk)
3615{
3616}
3617
3618static inline void inc_syscw(struct task_struct *tsk)
3619{
3620}
3621#endif
3622
3623#ifndef TASK_SIZE_OF
3624#define TASK_SIZE_OF(tsk) TASK_SIZE
3625#endif
3626
3627#ifdef CONFIG_MEMCG
3628extern void mm_update_next_owner(struct mm_struct *mm);
3629#else
3630static inline void mm_update_next_owner(struct mm_struct *mm)
3631{
3632}
3633#endif /* CONFIG_MEMCG */
3634
3635static inline unsigned long task_rlimit(const struct task_struct *tsk,
3636 unsigned int limit)
3637{
3638 return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3639}
3640
3641static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3642 unsigned int limit)
3643{
3644 return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3645}
3646
3647static inline unsigned long rlimit(unsigned int limit)
3648{
3649 return task_rlimit(current, limit);
3650}
3651
3652static inline unsigned long rlimit_max(unsigned int limit)
3653{
3654 return task_rlimit_max(current, limit);
3655}
3656
3657#define SCHED_CPUFREQ_RT (1U << 0)
3658#define SCHED_CPUFREQ_DL (1U << 1)
3659#define SCHED_CPUFREQ_IOWAIT (1U << 2)
3660
3661#define SCHED_CPUFREQ_RT_DL (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL)
3662
3663#ifdef CONFIG_CPU_FREQ
3664struct update_util_data {
3665 void (*func)(struct update_util_data *data, u64 time, unsigned int flags);
3666};
3667
3668void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3669 void (*func)(struct update_util_data *data, u64 time,
3670 unsigned int flags));
3671void cpufreq_remove_update_util_hook(int cpu);
3672#endif /* CONFIG_CPU_FREQ */
3673
3674#endif
3675