xref: /linux/include/linux/sched.h (revision bb9707077b4ee5f77bc9939b057ff8a0d410296f)
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 
9 struct 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  */
111 struct 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 
129 struct futex_pi_state;
130 struct robust_list_head;
131 struct bio_list;
132 struct fs_struct;
133 struct perf_event_context;
134 struct blk_plug;
135 struct filename;
136 struct 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  */
152 extern unsigned long avenrun[];		/* Load averages */
153 extern 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 
167 extern unsigned long total_forks;
168 extern int nr_threads;
169 DECLARE_PER_CPU(unsigned long, process_counts);
170 extern int nr_processes(void);
171 extern unsigned long nr_running(void);
172 extern bool single_task_running(void);
173 extern unsigned long nr_iowait(void);
174 extern unsigned long nr_iowait_cpu(int cpu);
175 extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
176 
177 extern void calc_global_load(unsigned long ticks);
178 
179 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
180 extern void cpu_load_update_nohz_start(void);
181 extern void cpu_load_update_nohz_stop(void);
182 #else
183 static inline void cpu_load_update_nohz_start(void) { }
184 static inline void cpu_load_update_nohz_stop(void) { }
185 #endif
186 
187 extern void dump_cpu_task(int cpu);
188 
189 struct seq_file;
190 struct cfs_rq;
191 struct task_group;
192 #ifdef CONFIG_SCHED_DEBUG
193 extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
194 extern 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_STATE_MAX		2048
223 
224 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPN"
225 
226 extern char ___assert_task_state[1 - 2*!!(
227 		sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
228 
229 /* Convenience macros for the sake of set_task_state */
230 #define TASK_KILLABLE		(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
231 #define TASK_STOPPED		(TASK_WAKEKILL | __TASK_STOPPED)
232 #define TASK_TRACED		(TASK_WAKEKILL | __TASK_TRACED)
233 
234 #define TASK_IDLE		(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
235 
236 /* Convenience macros for the sake of wake_up */
237 #define TASK_NORMAL		(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
238 #define TASK_ALL		(TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
239 
240 /* get_task_state() */
241 #define TASK_REPORT		(TASK_RUNNING | TASK_INTERRUPTIBLE | \
242 				 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
243 				 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
244 
245 #define task_is_traced(task)	((task->state & __TASK_TRACED) != 0)
246 #define task_is_stopped(task)	((task->state & __TASK_STOPPED) != 0)
247 #define task_is_stopped_or_traced(task)	\
248 			((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
249 #define task_contributes_to_load(task)	\
250 				((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
251 				 (task->flags & PF_FROZEN) == 0 && \
252 				 (task->state & TASK_NOLOAD) == 0)
253 
254 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
255 
256 #define __set_task_state(tsk, state_value)			\
257 	do {							\
258 		(tsk)->task_state_change = _THIS_IP_;		\
259 		(tsk)->state = (state_value);			\
260 	} while (0)
261 #define set_task_state(tsk, state_value)			\
262 	do {							\
263 		(tsk)->task_state_change = _THIS_IP_;		\
264 		smp_store_mb((tsk)->state, (state_value));		\
265 	} while (0)
266 
267 /*
268  * set_current_state() includes a barrier so that the write of current->state
269  * is correctly serialised wrt the caller's subsequent test of whether to
270  * actually sleep:
271  *
272  *	set_current_state(TASK_UNINTERRUPTIBLE);
273  *	if (do_i_need_to_sleep())
274  *		schedule();
275  *
276  * If the caller does not need such serialisation then use __set_current_state()
277  */
278 #define __set_current_state(state_value)			\
279 	do {							\
280 		current->task_state_change = _THIS_IP_;		\
281 		current->state = (state_value);			\
282 	} while (0)
283 #define set_current_state(state_value)				\
284 	do {							\
285 		current->task_state_change = _THIS_IP_;		\
286 		smp_store_mb(current->state, (state_value));		\
287 	} while (0)
288 
289 #else
290 
291 #define __set_task_state(tsk, state_value)		\
292 	do { (tsk)->state = (state_value); } while (0)
293 #define set_task_state(tsk, state_value)		\
294 	smp_store_mb((tsk)->state, (state_value))
295 
296 /*
297  * set_current_state() includes a barrier so that the write of current->state
298  * is correctly serialised wrt the caller's subsequent test of whether to
299  * actually sleep:
300  *
301  *	set_current_state(TASK_UNINTERRUPTIBLE);
302  *	if (do_i_need_to_sleep())
303  *		schedule();
304  *
305  * If the caller does not need such serialisation then use __set_current_state()
306  */
307 #define __set_current_state(state_value)		\
308 	do { current->state = (state_value); } while (0)
309 #define set_current_state(state_value)			\
310 	smp_store_mb(current->state, (state_value))
311 
312 #endif
313 
314 /* Task command name length */
315 #define TASK_COMM_LEN 16
316 
317 #include <linux/spinlock.h>
318 
319 /*
320  * This serializes "schedule()" and also protects
321  * the run-queue from deletions/modifications (but
322  * _adding_ to the beginning of the run-queue has
323  * a separate lock).
324  */
325 extern rwlock_t tasklist_lock;
326 extern spinlock_t mmlist_lock;
327 
328 struct task_struct;
329 
330 #ifdef CONFIG_PROVE_RCU
331 extern int lockdep_tasklist_lock_is_held(void);
332 #endif /* #ifdef CONFIG_PROVE_RCU */
333 
334 extern void sched_init(void);
335 extern void sched_init_smp(void);
336 extern asmlinkage void schedule_tail(struct task_struct *prev);
337 extern void init_idle(struct task_struct *idle, int cpu);
338 extern void init_idle_bootup_task(struct task_struct *idle);
339 
340 extern cpumask_var_t cpu_isolated_map;
341 
342 extern int runqueue_is_locked(int cpu);
343 
344 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
345 extern void nohz_balance_enter_idle(int cpu);
346 extern void set_cpu_sd_state_idle(void);
347 extern int get_nohz_timer_target(void);
348 #else
349 static inline void nohz_balance_enter_idle(int cpu) { }
350 static inline void set_cpu_sd_state_idle(void) { }
351 #endif
352 
353 /*
354  * Only dump TASK_* tasks. (0 for all tasks)
355  */
356 extern void show_state_filter(unsigned long state_filter);
357 
358 static inline void show_state(void)
359 {
360 	show_state_filter(0);
361 }
362 
363 extern void show_regs(struct pt_regs *);
364 
365 /*
366  * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
367  * task), SP is the stack pointer of the first frame that should be shown in the back
368  * trace (or NULL if the entire call-chain of the task should be shown).
369  */
370 extern void show_stack(struct task_struct *task, unsigned long *sp);
371 
372 extern void cpu_init (void);
373 extern void trap_init(void);
374 extern void update_process_times(int user);
375 extern void scheduler_tick(void);
376 extern int sched_cpu_starting(unsigned int cpu);
377 extern int sched_cpu_activate(unsigned int cpu);
378 extern int sched_cpu_deactivate(unsigned int cpu);
379 
380 #ifdef CONFIG_HOTPLUG_CPU
381 extern int sched_cpu_dying(unsigned int cpu);
382 #else
383 # define sched_cpu_dying	NULL
384 #endif
385 
386 extern void sched_show_task(struct task_struct *p);
387 
388 #ifdef CONFIG_LOCKUP_DETECTOR
389 extern void touch_softlockup_watchdog_sched(void);
390 extern void touch_softlockup_watchdog(void);
391 extern void touch_softlockup_watchdog_sync(void);
392 extern void touch_all_softlockup_watchdogs(void);
393 extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
394 				  void __user *buffer,
395 				  size_t *lenp, loff_t *ppos);
396 extern unsigned int  softlockup_panic;
397 extern unsigned int  hardlockup_panic;
398 void lockup_detector_init(void);
399 #else
400 static inline void touch_softlockup_watchdog_sched(void)
401 {
402 }
403 static inline void touch_softlockup_watchdog(void)
404 {
405 }
406 static inline void touch_softlockup_watchdog_sync(void)
407 {
408 }
409 static inline void touch_all_softlockup_watchdogs(void)
410 {
411 }
412 static inline void lockup_detector_init(void)
413 {
414 }
415 #endif
416 
417 #ifdef CONFIG_DETECT_HUNG_TASK
418 void reset_hung_task_detector(void);
419 #else
420 static inline void reset_hung_task_detector(void)
421 {
422 }
423 #endif
424 
425 /* Attach to any functions which should be ignored in wchan output. */
426 #define __sched		__attribute__((__section__(".sched.text")))
427 
428 /* Linker adds these: start and end of __sched functions */
429 extern char __sched_text_start[], __sched_text_end[];
430 
431 /* Is this address in the __sched functions? */
432 extern int in_sched_functions(unsigned long addr);
433 
434 #define	MAX_SCHEDULE_TIMEOUT	LONG_MAX
435 extern signed long schedule_timeout(signed long timeout);
436 extern signed long schedule_timeout_interruptible(signed long timeout);
437 extern signed long schedule_timeout_killable(signed long timeout);
438 extern signed long schedule_timeout_uninterruptible(signed long timeout);
439 extern signed long schedule_timeout_idle(signed long timeout);
440 asmlinkage void schedule(void);
441 extern void schedule_preempt_disabled(void);
442 
443 extern long io_schedule_timeout(long timeout);
444 
445 static inline void io_schedule(void)
446 {
447 	io_schedule_timeout(MAX_SCHEDULE_TIMEOUT);
448 }
449 
450 struct nsproxy;
451 struct user_namespace;
452 
453 #ifdef CONFIG_MMU
454 extern void arch_pick_mmap_layout(struct mm_struct *mm);
455 extern unsigned long
456 arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
457 		       unsigned long, unsigned long);
458 extern unsigned long
459 arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
460 			  unsigned long len, unsigned long pgoff,
461 			  unsigned long flags);
462 #else
463 static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
464 #endif
465 
466 #define SUID_DUMP_DISABLE	0	/* No setuid dumping */
467 #define SUID_DUMP_USER		1	/* Dump as user of process */
468 #define SUID_DUMP_ROOT		2	/* Dump as root */
469 
470 /* mm flags */
471 
472 /* for SUID_DUMP_* above */
473 #define MMF_DUMPABLE_BITS 2
474 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
475 
476 extern void set_dumpable(struct mm_struct *mm, int value);
477 /*
478  * This returns the actual value of the suid_dumpable flag. For things
479  * that are using this for checking for privilege transitions, it must
480  * test against SUID_DUMP_USER rather than treating it as a boolean
481  * value.
482  */
483 static inline int __get_dumpable(unsigned long mm_flags)
484 {
485 	return mm_flags & MMF_DUMPABLE_MASK;
486 }
487 
488 static inline int get_dumpable(struct mm_struct *mm)
489 {
490 	return __get_dumpable(mm->flags);
491 }
492 
493 /* coredump filter bits */
494 #define MMF_DUMP_ANON_PRIVATE	2
495 #define MMF_DUMP_ANON_SHARED	3
496 #define MMF_DUMP_MAPPED_PRIVATE	4
497 #define MMF_DUMP_MAPPED_SHARED	5
498 #define MMF_DUMP_ELF_HEADERS	6
499 #define MMF_DUMP_HUGETLB_PRIVATE 7
500 #define MMF_DUMP_HUGETLB_SHARED  8
501 #define MMF_DUMP_DAX_PRIVATE	9
502 #define MMF_DUMP_DAX_SHARED	10
503 
504 #define MMF_DUMP_FILTER_SHIFT	MMF_DUMPABLE_BITS
505 #define MMF_DUMP_FILTER_BITS	9
506 #define MMF_DUMP_FILTER_MASK \
507 	(((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
508 #define MMF_DUMP_FILTER_DEFAULT \
509 	((1 << MMF_DUMP_ANON_PRIVATE) |	(1 << MMF_DUMP_ANON_SHARED) |\
510 	 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
511 
512 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
513 # define MMF_DUMP_MASK_DEFAULT_ELF	(1 << MMF_DUMP_ELF_HEADERS)
514 #else
515 # define MMF_DUMP_MASK_DEFAULT_ELF	0
516 #endif
517 					/* leave room for more dump flags */
518 #define MMF_VM_MERGEABLE	16	/* KSM may merge identical pages */
519 #define MMF_VM_HUGEPAGE		17	/* set when VM_HUGEPAGE is set on vma */
520 #define MMF_EXE_FILE_CHANGED	18	/* see prctl_set_mm_exe_file() */
521 
522 #define MMF_HAS_UPROBES		19	/* has uprobes */
523 #define MMF_RECALC_UPROBES	20	/* MMF_HAS_UPROBES can be wrong */
524 #define MMF_OOM_REAPED		21	/* mm has been already reaped */
525 
526 #define MMF_INIT_MASK		(MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
527 
528 struct sighand_struct {
529 	atomic_t		count;
530 	struct k_sigaction	action[_NSIG];
531 	spinlock_t		siglock;
532 	wait_queue_head_t	signalfd_wqh;
533 };
534 
535 struct pacct_struct {
536 	int			ac_flag;
537 	long			ac_exitcode;
538 	unsigned long		ac_mem;
539 	cputime_t		ac_utime, ac_stime;
540 	unsigned long		ac_minflt, ac_majflt;
541 };
542 
543 struct cpu_itimer {
544 	cputime_t expires;
545 	cputime_t incr;
546 	u32 error;
547 	u32 incr_error;
548 };
549 
550 /**
551  * struct prev_cputime - snaphsot of system and user cputime
552  * @utime: time spent in user mode
553  * @stime: time spent in system mode
554  * @lock: protects the above two fields
555  *
556  * Stores previous user/system time values such that we can guarantee
557  * monotonicity.
558  */
559 struct prev_cputime {
560 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
561 	cputime_t utime;
562 	cputime_t stime;
563 	raw_spinlock_t lock;
564 #endif
565 };
566 
567 static inline void prev_cputime_init(struct prev_cputime *prev)
568 {
569 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
570 	prev->utime = prev->stime = 0;
571 	raw_spin_lock_init(&prev->lock);
572 #endif
573 }
574 
575 /**
576  * struct task_cputime - collected CPU time counts
577  * @utime:		time spent in user mode, in &cputime_t units
578  * @stime:		time spent in kernel mode, in &cputime_t units
579  * @sum_exec_runtime:	total time spent on the CPU, in nanoseconds
580  *
581  * This structure groups together three kinds of CPU time that are tracked for
582  * threads and thread groups.  Most things considering CPU time want to group
583  * these counts together and treat all three of them in parallel.
584  */
585 struct task_cputime {
586 	cputime_t utime;
587 	cputime_t stime;
588 	unsigned long long sum_exec_runtime;
589 };
590 
591 /* Alternate field names when used to cache expirations. */
592 #define virt_exp	utime
593 #define prof_exp	stime
594 #define sched_exp	sum_exec_runtime
595 
596 #define INIT_CPUTIME	\
597 	(struct task_cputime) {					\
598 		.utime = 0,					\
599 		.stime = 0,					\
600 		.sum_exec_runtime = 0,				\
601 	}
602 
603 /*
604  * This is the atomic variant of task_cputime, which can be used for
605  * storing and updating task_cputime statistics without locking.
606  */
607 struct task_cputime_atomic {
608 	atomic64_t utime;
609 	atomic64_t stime;
610 	atomic64_t sum_exec_runtime;
611 };
612 
613 #define INIT_CPUTIME_ATOMIC \
614 	(struct task_cputime_atomic) {				\
615 		.utime = ATOMIC64_INIT(0),			\
616 		.stime = ATOMIC64_INIT(0),			\
617 		.sum_exec_runtime = ATOMIC64_INIT(0),		\
618 	}
619 
620 #define PREEMPT_DISABLED	(PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
621 
622 /*
623  * Disable preemption until the scheduler is running -- use an unconditional
624  * value so that it also works on !PREEMPT_COUNT kernels.
625  *
626  * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
627  */
628 #define INIT_PREEMPT_COUNT	PREEMPT_OFFSET
629 
630 /*
631  * Initial preempt_count value; reflects the preempt_count schedule invariant
632  * which states that during context switches:
633  *
634  *    preempt_count() == 2*PREEMPT_DISABLE_OFFSET
635  *
636  * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
637  * Note: See finish_task_switch().
638  */
639 #define FORK_PREEMPT_COUNT	(2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
640 
641 /**
642  * struct thread_group_cputimer - thread group interval timer counts
643  * @cputime_atomic:	atomic thread group interval timers.
644  * @running:		true when there are timers running and
645  *			@cputime_atomic receives updates.
646  * @checking_timer:	true when a thread in the group is in the
647  *			process of checking for thread group timers.
648  *
649  * This structure contains the version of task_cputime, above, that is
650  * used for thread group CPU timer calculations.
651  */
652 struct thread_group_cputimer {
653 	struct task_cputime_atomic cputime_atomic;
654 	bool running;
655 	bool checking_timer;
656 };
657 
658 #include <linux/rwsem.h>
659 struct autogroup;
660 
661 /*
662  * NOTE! "signal_struct" does not have its own
663  * locking, because a shared signal_struct always
664  * implies a shared sighand_struct, so locking
665  * sighand_struct is always a proper superset of
666  * the locking of signal_struct.
667  */
668 struct signal_struct {
669 	atomic_t		sigcnt;
670 	atomic_t		live;
671 	int			nr_threads;
672 	atomic_t oom_victims; /* # of TIF_MEDIE threads in this thread group */
673 	struct list_head	thread_head;
674 
675 	wait_queue_head_t	wait_chldexit;	/* for wait4() */
676 
677 	/* current thread group signal load-balancing target: */
678 	struct task_struct	*curr_target;
679 
680 	/* shared signal handling: */
681 	struct sigpending	shared_pending;
682 
683 	/* thread group exit support */
684 	int			group_exit_code;
685 	/* overloaded:
686 	 * - notify group_exit_task when ->count is equal to notify_count
687 	 * - everyone except group_exit_task is stopped during signal delivery
688 	 *   of fatal signals, group_exit_task processes the signal.
689 	 */
690 	int			notify_count;
691 	struct task_struct	*group_exit_task;
692 
693 	/* thread group stop support, overloads group_exit_code too */
694 	int			group_stop_count;
695 	unsigned int		flags; /* see SIGNAL_* flags below */
696 
697 	/*
698 	 * PR_SET_CHILD_SUBREAPER marks a process, like a service
699 	 * manager, to re-parent orphan (double-forking) child processes
700 	 * to this process instead of 'init'. The service manager is
701 	 * able to receive SIGCHLD signals and is able to investigate
702 	 * the process until it calls wait(). All children of this
703 	 * process will inherit a flag if they should look for a
704 	 * child_subreaper process at exit.
705 	 */
706 	unsigned int		is_child_subreaper:1;
707 	unsigned int		has_child_subreaper:1;
708 
709 	/* POSIX.1b Interval Timers */
710 	int			posix_timer_id;
711 	struct list_head	posix_timers;
712 
713 	/* ITIMER_REAL timer for the process */
714 	struct hrtimer real_timer;
715 	struct pid *leader_pid;
716 	ktime_t it_real_incr;
717 
718 	/*
719 	 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
720 	 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
721 	 * values are defined to 0 and 1 respectively
722 	 */
723 	struct cpu_itimer it[2];
724 
725 	/*
726 	 * Thread group totals for process CPU timers.
727 	 * See thread_group_cputimer(), et al, for details.
728 	 */
729 	struct thread_group_cputimer cputimer;
730 
731 	/* Earliest-expiration cache. */
732 	struct task_cputime cputime_expires;
733 
734 #ifdef CONFIG_NO_HZ_FULL
735 	atomic_t tick_dep_mask;
736 #endif
737 
738 	struct list_head cpu_timers[3];
739 
740 	struct pid *tty_old_pgrp;
741 
742 	/* boolean value for session group leader */
743 	int leader;
744 
745 	struct tty_struct *tty; /* NULL if no tty */
746 
747 #ifdef CONFIG_SCHED_AUTOGROUP
748 	struct autogroup *autogroup;
749 #endif
750 	/*
751 	 * Cumulative resource counters for dead threads in the group,
752 	 * and for reaped dead child processes forked by this group.
753 	 * Live threads maintain their own counters and add to these
754 	 * in __exit_signal, except for the group leader.
755 	 */
756 	seqlock_t stats_lock;
757 	cputime_t utime, stime, cutime, cstime;
758 	cputime_t gtime;
759 	cputime_t cgtime;
760 	struct prev_cputime prev_cputime;
761 	unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
762 	unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
763 	unsigned long inblock, oublock, cinblock, coublock;
764 	unsigned long maxrss, cmaxrss;
765 	struct task_io_accounting ioac;
766 
767 	/*
768 	 * Cumulative ns of schedule CPU time fo dead threads in the
769 	 * group, not including a zombie group leader, (This only differs
770 	 * from jiffies_to_ns(utime + stime) if sched_clock uses something
771 	 * other than jiffies.)
772 	 */
773 	unsigned long long sum_sched_runtime;
774 
775 	/*
776 	 * We don't bother to synchronize most readers of this at all,
777 	 * because there is no reader checking a limit that actually needs
778 	 * to get both rlim_cur and rlim_max atomically, and either one
779 	 * alone is a single word that can safely be read normally.
780 	 * getrlimit/setrlimit use task_lock(current->group_leader) to
781 	 * protect this instead of the siglock, because they really
782 	 * have no need to disable irqs.
783 	 */
784 	struct rlimit rlim[RLIM_NLIMITS];
785 
786 #ifdef CONFIG_BSD_PROCESS_ACCT
787 	struct pacct_struct pacct;	/* per-process accounting information */
788 #endif
789 #ifdef CONFIG_TASKSTATS
790 	struct taskstats *stats;
791 #endif
792 #ifdef CONFIG_AUDIT
793 	unsigned audit_tty;
794 	struct tty_audit_buf *tty_audit_buf;
795 #endif
796 
797 	/*
798 	 * Thread is the potential origin of an oom condition; kill first on
799 	 * oom
800 	 */
801 	bool oom_flag_origin;
802 	short oom_score_adj;		/* OOM kill score adjustment */
803 	short oom_score_adj_min;	/* OOM kill score adjustment min value.
804 					 * Only settable by CAP_SYS_RESOURCE. */
805 
806 	struct mutex cred_guard_mutex;	/* guard against foreign influences on
807 					 * credential calculations
808 					 * (notably. ptrace) */
809 };
810 
811 /*
812  * Bits in flags field of signal_struct.
813  */
814 #define SIGNAL_STOP_STOPPED	0x00000001 /* job control stop in effect */
815 #define SIGNAL_STOP_CONTINUED	0x00000002 /* SIGCONT since WCONTINUED reap */
816 #define SIGNAL_GROUP_EXIT	0x00000004 /* group exit in progress */
817 #define SIGNAL_GROUP_COREDUMP	0x00000008 /* coredump in progress */
818 /*
819  * Pending notifications to parent.
820  */
821 #define SIGNAL_CLD_STOPPED	0x00000010
822 #define SIGNAL_CLD_CONTINUED	0x00000020
823 #define SIGNAL_CLD_MASK		(SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
824 
825 #define SIGNAL_UNKILLABLE	0x00000040 /* for init: ignore fatal signals */
826 
827 /* If true, all threads except ->group_exit_task have pending SIGKILL */
828 static inline int signal_group_exit(const struct signal_struct *sig)
829 {
830 	return	(sig->flags & SIGNAL_GROUP_EXIT) ||
831 		(sig->group_exit_task != NULL);
832 }
833 
834 /*
835  * Some day this will be a full-fledged user tracking system..
836  */
837 struct user_struct {
838 	atomic_t __count;	/* reference count */
839 	atomic_t processes;	/* How many processes does this user have? */
840 	atomic_t sigpending;	/* How many pending signals does this user have? */
841 #ifdef CONFIG_INOTIFY_USER
842 	atomic_t inotify_watches; /* How many inotify watches does this user have? */
843 	atomic_t inotify_devs;	/* How many inotify devs does this user have opened? */
844 #endif
845 #ifdef CONFIG_FANOTIFY
846 	atomic_t fanotify_listeners;
847 #endif
848 #ifdef CONFIG_EPOLL
849 	atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
850 #endif
851 #ifdef CONFIG_POSIX_MQUEUE
852 	/* protected by mq_lock	*/
853 	unsigned long mq_bytes;	/* How many bytes can be allocated to mqueue? */
854 #endif
855 	unsigned long locked_shm; /* How many pages of mlocked shm ? */
856 	unsigned long unix_inflight;	/* How many files in flight in unix sockets */
857 	atomic_long_t pipe_bufs;  /* how many pages are allocated in pipe buffers */
858 
859 #ifdef CONFIG_KEYS
860 	struct key *uid_keyring;	/* UID specific keyring */
861 	struct key *session_keyring;	/* UID's default session keyring */
862 #endif
863 
864 	/* Hash table maintenance information */
865 	struct hlist_node uidhash_node;
866 	kuid_t uid;
867 
868 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
869 	atomic_long_t locked_vm;
870 #endif
871 };
872 
873 extern int uids_sysfs_init(void);
874 
875 extern struct user_struct *find_user(kuid_t);
876 
877 extern struct user_struct root_user;
878 #define INIT_USER (&root_user)
879 
880 
881 struct backing_dev_info;
882 struct reclaim_state;
883 
884 #ifdef CONFIG_SCHED_INFO
885 struct sched_info {
886 	/* cumulative counters */
887 	unsigned long pcount;	      /* # of times run on this cpu */
888 	unsigned long long run_delay; /* time spent waiting on a runqueue */
889 
890 	/* timestamps */
891 	unsigned long long last_arrival,/* when we last ran on a cpu */
892 			   last_queued;	/* when we were last queued to run */
893 };
894 #endif /* CONFIG_SCHED_INFO */
895 
896 #ifdef CONFIG_TASK_DELAY_ACCT
897 struct task_delay_info {
898 	spinlock_t	lock;
899 	unsigned int	flags;	/* Private per-task flags */
900 
901 	/* For each stat XXX, add following, aligned appropriately
902 	 *
903 	 * struct timespec XXX_start, XXX_end;
904 	 * u64 XXX_delay;
905 	 * u32 XXX_count;
906 	 *
907 	 * Atomicity of updates to XXX_delay, XXX_count protected by
908 	 * single lock above (split into XXX_lock if contention is an issue).
909 	 */
910 
911 	/*
912 	 * XXX_count is incremented on every XXX operation, the delay
913 	 * associated with the operation is added to XXX_delay.
914 	 * XXX_delay contains the accumulated delay time in nanoseconds.
915 	 */
916 	u64 blkio_start;	/* Shared by blkio, swapin */
917 	u64 blkio_delay;	/* wait for sync block io completion */
918 	u64 swapin_delay;	/* wait for swapin block io completion */
919 	u32 blkio_count;	/* total count of the number of sync block */
920 				/* io operations performed */
921 	u32 swapin_count;	/* total count of the number of swapin block */
922 				/* io operations performed */
923 
924 	u64 freepages_start;
925 	u64 freepages_delay;	/* wait for memory reclaim */
926 	u32 freepages_count;	/* total count of memory reclaim */
927 };
928 #endif	/* CONFIG_TASK_DELAY_ACCT */
929 
930 static inline int sched_info_on(void)
931 {
932 #ifdef CONFIG_SCHEDSTATS
933 	return 1;
934 #elif defined(CONFIG_TASK_DELAY_ACCT)
935 	extern int delayacct_on;
936 	return delayacct_on;
937 #else
938 	return 0;
939 #endif
940 }
941 
942 #ifdef CONFIG_SCHEDSTATS
943 void force_schedstat_enabled(void);
944 #endif
945 
946 enum cpu_idle_type {
947 	CPU_IDLE,
948 	CPU_NOT_IDLE,
949 	CPU_NEWLY_IDLE,
950 	CPU_MAX_IDLE_TYPES
951 };
952 
953 /*
954  * Integer metrics need fixed point arithmetic, e.g., sched/fair
955  * has a few: load, load_avg, util_avg, freq, and capacity.
956  *
957  * We define a basic fixed point arithmetic range, and then formalize
958  * all these metrics based on that basic range.
959  */
960 # define SCHED_FIXEDPOINT_SHIFT	10
961 # define SCHED_FIXEDPOINT_SCALE	(1L << SCHED_FIXEDPOINT_SHIFT)
962 
963 /*
964  * Increase resolution of cpu_capacity calculations
965  */
966 #define SCHED_CAPACITY_SHIFT	SCHED_FIXEDPOINT_SHIFT
967 #define SCHED_CAPACITY_SCALE	(1L << SCHED_CAPACITY_SHIFT)
968 
969 /*
970  * Wake-queues are lists of tasks with a pending wakeup, whose
971  * callers have already marked the task as woken internally,
972  * and can thus carry on. A common use case is being able to
973  * do the wakeups once the corresponding user lock as been
974  * released.
975  *
976  * We hold reference to each task in the list across the wakeup,
977  * thus guaranteeing that the memory is still valid by the time
978  * the actual wakeups are performed in wake_up_q().
979  *
980  * One per task suffices, because there's never a need for a task to be
981  * in two wake queues simultaneously; it is forbidden to abandon a task
982  * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
983  * already in a wake queue, the wakeup will happen soon and the second
984  * waker can just skip it.
985  *
986  * The WAKE_Q macro declares and initializes the list head.
987  * wake_up_q() does NOT reinitialize the list; it's expected to be
988  * called near the end of a function, where the fact that the queue is
989  * not used again will be easy to see by inspection.
990  *
991  * Note that this can cause spurious wakeups. schedule() callers
992  * must ensure the call is done inside a loop, confirming that the
993  * wakeup condition has in fact occurred.
994  */
995 struct wake_q_node {
996 	struct wake_q_node *next;
997 };
998 
999 struct wake_q_head {
1000 	struct wake_q_node *first;
1001 	struct wake_q_node **lastp;
1002 };
1003 
1004 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1005 
1006 #define WAKE_Q(name)					\
1007 	struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1008 
1009 extern void wake_q_add(struct wake_q_head *head,
1010 		       struct task_struct *task);
1011 extern void wake_up_q(struct wake_q_head *head);
1012 
1013 /*
1014  * sched-domains (multiprocessor balancing) declarations:
1015  */
1016 #ifdef CONFIG_SMP
1017 #define SD_LOAD_BALANCE		0x0001	/* Do load balancing on this domain. */
1018 #define SD_BALANCE_NEWIDLE	0x0002	/* Balance when about to become idle */
1019 #define SD_BALANCE_EXEC		0x0004	/* Balance on exec */
1020 #define SD_BALANCE_FORK		0x0008	/* Balance on fork, clone */
1021 #define SD_BALANCE_WAKE		0x0010  /* Balance on wakeup */
1022 #define SD_WAKE_AFFINE		0x0020	/* Wake task to waking CPU */
1023 #define SD_SHARE_CPUCAPACITY	0x0080	/* Domain members share cpu power */
1024 #define SD_SHARE_POWERDOMAIN	0x0100	/* Domain members share power domain */
1025 #define SD_SHARE_PKG_RESOURCES	0x0200	/* Domain members share cpu pkg resources */
1026 #define SD_SERIALIZE		0x0400	/* Only a single load balancing instance */
1027 #define SD_ASYM_PACKING		0x0800  /* Place busy groups earlier in the domain */
1028 #define SD_PREFER_SIBLING	0x1000	/* Prefer to place tasks in a sibling domain */
1029 #define SD_OVERLAP		0x2000	/* sched_domains of this level overlap */
1030 #define SD_NUMA			0x4000	/* cross-node balancing */
1031 
1032 #ifdef CONFIG_SCHED_SMT
1033 static inline int cpu_smt_flags(void)
1034 {
1035 	return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1036 }
1037 #endif
1038 
1039 #ifdef CONFIG_SCHED_MC
1040 static inline int cpu_core_flags(void)
1041 {
1042 	return SD_SHARE_PKG_RESOURCES;
1043 }
1044 #endif
1045 
1046 #ifdef CONFIG_NUMA
1047 static inline int cpu_numa_flags(void)
1048 {
1049 	return SD_NUMA;
1050 }
1051 #endif
1052 
1053 struct sched_domain_attr {
1054 	int relax_domain_level;
1055 };
1056 
1057 #define SD_ATTR_INIT	(struct sched_domain_attr) {	\
1058 	.relax_domain_level = -1,			\
1059 }
1060 
1061 extern int sched_domain_level_max;
1062 
1063 struct sched_group;
1064 
1065 struct sched_domain {
1066 	/* These fields must be setup */
1067 	struct sched_domain *parent;	/* top domain must be null terminated */
1068 	struct sched_domain *child;	/* bottom domain must be null terminated */
1069 	struct sched_group *groups;	/* the balancing groups of the domain */
1070 	unsigned long min_interval;	/* Minimum balance interval ms */
1071 	unsigned long max_interval;	/* Maximum balance interval ms */
1072 	unsigned int busy_factor;	/* less balancing by factor if busy */
1073 	unsigned int imbalance_pct;	/* No balance until over watermark */
1074 	unsigned int cache_nice_tries;	/* Leave cache hot tasks for # tries */
1075 	unsigned int busy_idx;
1076 	unsigned int idle_idx;
1077 	unsigned int newidle_idx;
1078 	unsigned int wake_idx;
1079 	unsigned int forkexec_idx;
1080 	unsigned int smt_gain;
1081 
1082 	int nohz_idle;			/* NOHZ IDLE status */
1083 	int flags;			/* See SD_* */
1084 	int level;
1085 
1086 	/* Runtime fields. */
1087 	unsigned long last_balance;	/* init to jiffies. units in jiffies */
1088 	unsigned int balance_interval;	/* initialise to 1. units in ms. */
1089 	unsigned int nr_balance_failed; /* initialise to 0 */
1090 
1091 	/* idle_balance() stats */
1092 	u64 max_newidle_lb_cost;
1093 	unsigned long next_decay_max_lb_cost;
1094 
1095 #ifdef CONFIG_SCHEDSTATS
1096 	/* load_balance() stats */
1097 	unsigned int lb_count[CPU_MAX_IDLE_TYPES];
1098 	unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
1099 	unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
1100 	unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
1101 	unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
1102 	unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
1103 	unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
1104 	unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
1105 
1106 	/* Active load balancing */
1107 	unsigned int alb_count;
1108 	unsigned int alb_failed;
1109 	unsigned int alb_pushed;
1110 
1111 	/* SD_BALANCE_EXEC stats */
1112 	unsigned int sbe_count;
1113 	unsigned int sbe_balanced;
1114 	unsigned int sbe_pushed;
1115 
1116 	/* SD_BALANCE_FORK stats */
1117 	unsigned int sbf_count;
1118 	unsigned int sbf_balanced;
1119 	unsigned int sbf_pushed;
1120 
1121 	/* try_to_wake_up() stats */
1122 	unsigned int ttwu_wake_remote;
1123 	unsigned int ttwu_move_affine;
1124 	unsigned int ttwu_move_balance;
1125 #endif
1126 #ifdef CONFIG_SCHED_DEBUG
1127 	char *name;
1128 #endif
1129 	union {
1130 		void *private;		/* used during construction */
1131 		struct rcu_head rcu;	/* used during destruction */
1132 	};
1133 
1134 	unsigned int span_weight;
1135 	/*
1136 	 * Span of all CPUs in this domain.
1137 	 *
1138 	 * NOTE: this field is variable length. (Allocated dynamically
1139 	 * by attaching extra space to the end of the structure,
1140 	 * depending on how many CPUs the kernel has booted up with)
1141 	 */
1142 	unsigned long span[0];
1143 };
1144 
1145 static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
1146 {
1147 	return to_cpumask(sd->span);
1148 }
1149 
1150 extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1151 				    struct sched_domain_attr *dattr_new);
1152 
1153 /* Allocate an array of sched domains, for partition_sched_domains(). */
1154 cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
1155 void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
1156 
1157 bool cpus_share_cache(int this_cpu, int that_cpu);
1158 
1159 typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1160 typedef int (*sched_domain_flags_f)(void);
1161 
1162 #define SDTL_OVERLAP	0x01
1163 
1164 struct sd_data {
1165 	struct sched_domain **__percpu sd;
1166 	struct sched_group **__percpu sg;
1167 	struct sched_group_capacity **__percpu sgc;
1168 };
1169 
1170 struct sched_domain_topology_level {
1171 	sched_domain_mask_f mask;
1172 	sched_domain_flags_f sd_flags;
1173 	int		    flags;
1174 	int		    numa_level;
1175 	struct sd_data      data;
1176 #ifdef CONFIG_SCHED_DEBUG
1177 	char                *name;
1178 #endif
1179 };
1180 
1181 extern void set_sched_topology(struct sched_domain_topology_level *tl);
1182 extern void wake_up_if_idle(int cpu);
1183 
1184 #ifdef CONFIG_SCHED_DEBUG
1185 # define SD_INIT_NAME(type)		.name = #type
1186 #else
1187 # define SD_INIT_NAME(type)
1188 #endif
1189 
1190 #else /* CONFIG_SMP */
1191 
1192 struct sched_domain_attr;
1193 
1194 static inline void
1195 partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1196 			struct sched_domain_attr *dattr_new)
1197 {
1198 }
1199 
1200 static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1201 {
1202 	return true;
1203 }
1204 
1205 #endif	/* !CONFIG_SMP */
1206 
1207 
1208 struct io_context;			/* See blkdev.h */
1209 
1210 
1211 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1212 extern void prefetch_stack(struct task_struct *t);
1213 #else
1214 static inline void prefetch_stack(struct task_struct *t) { }
1215 #endif
1216 
1217 struct audit_context;		/* See audit.c */
1218 struct mempolicy;
1219 struct pipe_inode_info;
1220 struct uts_namespace;
1221 
1222 struct load_weight {
1223 	unsigned long weight;
1224 	u32 inv_weight;
1225 };
1226 
1227 /*
1228  * The load_avg/util_avg accumulates an infinite geometric series
1229  * (see __update_load_avg() in kernel/sched/fair.c).
1230  *
1231  * [load_avg definition]
1232  *
1233  *   load_avg = runnable% * scale_load_down(load)
1234  *
1235  * where runnable% is the time ratio that a sched_entity is runnable.
1236  * For cfs_rq, it is the aggregated load_avg of all runnable and
1237  * blocked sched_entities.
1238  *
1239  * load_avg may also take frequency scaling into account:
1240  *
1241  *   load_avg = runnable% * scale_load_down(load) * freq%
1242  *
1243  * where freq% is the CPU frequency normalized to the highest frequency.
1244  *
1245  * [util_avg definition]
1246  *
1247  *   util_avg = running% * SCHED_CAPACITY_SCALE
1248  *
1249  * where running% is the time ratio that a sched_entity is running on
1250  * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1251  * and blocked sched_entities.
1252  *
1253  * util_avg may also factor frequency scaling and CPU capacity scaling:
1254  *
1255  *   util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1256  *
1257  * where freq% is the same as above, and capacity% is the CPU capacity
1258  * normalized to the greatest capacity (due to uarch differences, etc).
1259  *
1260  * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1261  * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1262  * we therefore scale them to as large a range as necessary. This is for
1263  * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1264  *
1265  * [Overflow issue]
1266  *
1267  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1268  * with the highest load (=88761), always runnable on a single cfs_rq,
1269  * and should not overflow as the number already hits PID_MAX_LIMIT.
1270  *
1271  * For all other cases (including 32-bit kernels), struct load_weight's
1272  * weight will overflow first before we do, because:
1273  *
1274  *    Max(load_avg) <= Max(load.weight)
1275  *
1276  * Then it is the load_weight's responsibility to consider overflow
1277  * issues.
1278  */
1279 struct sched_avg {
1280 	u64 last_update_time, load_sum;
1281 	u32 util_sum, period_contrib;
1282 	unsigned long load_avg, util_avg;
1283 };
1284 
1285 #ifdef CONFIG_SCHEDSTATS
1286 struct sched_statistics {
1287 	u64			wait_start;
1288 	u64			wait_max;
1289 	u64			wait_count;
1290 	u64			wait_sum;
1291 	u64			iowait_count;
1292 	u64			iowait_sum;
1293 
1294 	u64			sleep_start;
1295 	u64			sleep_max;
1296 	s64			sum_sleep_runtime;
1297 
1298 	u64			block_start;
1299 	u64			block_max;
1300 	u64			exec_max;
1301 	u64			slice_max;
1302 
1303 	u64			nr_migrations_cold;
1304 	u64			nr_failed_migrations_affine;
1305 	u64			nr_failed_migrations_running;
1306 	u64			nr_failed_migrations_hot;
1307 	u64			nr_forced_migrations;
1308 
1309 	u64			nr_wakeups;
1310 	u64			nr_wakeups_sync;
1311 	u64			nr_wakeups_migrate;
1312 	u64			nr_wakeups_local;
1313 	u64			nr_wakeups_remote;
1314 	u64			nr_wakeups_affine;
1315 	u64			nr_wakeups_affine_attempts;
1316 	u64			nr_wakeups_passive;
1317 	u64			nr_wakeups_idle;
1318 };
1319 #endif
1320 
1321 struct sched_entity {
1322 	struct load_weight	load;		/* for load-balancing */
1323 	struct rb_node		run_node;
1324 	struct list_head	group_node;
1325 	unsigned int		on_rq;
1326 
1327 	u64			exec_start;
1328 	u64			sum_exec_runtime;
1329 	u64			vruntime;
1330 	u64			prev_sum_exec_runtime;
1331 
1332 	u64			nr_migrations;
1333 
1334 #ifdef CONFIG_SCHEDSTATS
1335 	struct sched_statistics statistics;
1336 #endif
1337 
1338 #ifdef CONFIG_FAIR_GROUP_SCHED
1339 	int			depth;
1340 	struct sched_entity	*parent;
1341 	/* rq on which this entity is (to be) queued: */
1342 	struct cfs_rq		*cfs_rq;
1343 	/* rq "owned" by this entity/group: */
1344 	struct cfs_rq		*my_q;
1345 #endif
1346 
1347 #ifdef CONFIG_SMP
1348 	/*
1349 	 * Per entity load average tracking.
1350 	 *
1351 	 * Put into separate cache line so it does not
1352 	 * collide with read-mostly values above.
1353 	 */
1354 	struct sched_avg	avg ____cacheline_aligned_in_smp;
1355 #endif
1356 };
1357 
1358 struct sched_rt_entity {
1359 	struct list_head run_list;
1360 	unsigned long timeout;
1361 	unsigned long watchdog_stamp;
1362 	unsigned int time_slice;
1363 	unsigned short on_rq;
1364 	unsigned short on_list;
1365 
1366 	struct sched_rt_entity *back;
1367 #ifdef CONFIG_RT_GROUP_SCHED
1368 	struct sched_rt_entity	*parent;
1369 	/* rq on which this entity is (to be) queued: */
1370 	struct rt_rq		*rt_rq;
1371 	/* rq "owned" by this entity/group: */
1372 	struct rt_rq		*my_q;
1373 #endif
1374 };
1375 
1376 struct sched_dl_entity {
1377 	struct rb_node	rb_node;
1378 
1379 	/*
1380 	 * Original scheduling parameters. Copied here from sched_attr
1381 	 * during sched_setattr(), they will remain the same until
1382 	 * the next sched_setattr().
1383 	 */
1384 	u64 dl_runtime;		/* maximum runtime for each instance	*/
1385 	u64 dl_deadline;	/* relative deadline of each instance	*/
1386 	u64 dl_period;		/* separation of two instances (period) */
1387 	u64 dl_bw;		/* dl_runtime / dl_deadline		*/
1388 
1389 	/*
1390 	 * Actual scheduling parameters. Initialized with the values above,
1391 	 * they are continously updated during task execution. Note that
1392 	 * the remaining runtime could be < 0 in case we are in overrun.
1393 	 */
1394 	s64 runtime;		/* remaining runtime for this instance	*/
1395 	u64 deadline;		/* absolute deadline for this instance	*/
1396 	unsigned int flags;	/* specifying the scheduler behaviour	*/
1397 
1398 	/*
1399 	 * Some bool flags:
1400 	 *
1401 	 * @dl_throttled tells if we exhausted the runtime. If so, the
1402 	 * task has to wait for a replenishment to be performed at the
1403 	 * next firing of dl_timer.
1404 	 *
1405 	 * @dl_boosted tells if we are boosted due to DI. If so we are
1406 	 * outside bandwidth enforcement mechanism (but only until we
1407 	 * exit the critical section);
1408 	 *
1409 	 * @dl_yielded tells if task gave up the cpu before consuming
1410 	 * all its available runtime during the last job.
1411 	 */
1412 	int dl_throttled, dl_boosted, dl_yielded;
1413 
1414 	/*
1415 	 * Bandwidth enforcement timer. Each -deadline task has its
1416 	 * own bandwidth to be enforced, thus we need one timer per task.
1417 	 */
1418 	struct hrtimer dl_timer;
1419 };
1420 
1421 union rcu_special {
1422 	struct {
1423 		u8 blocked;
1424 		u8 need_qs;
1425 		u8 exp_need_qs;
1426 		u8 pad;	/* Otherwise the compiler can store garbage here. */
1427 	} b; /* Bits. */
1428 	u32 s; /* Set of bits. */
1429 };
1430 struct rcu_node;
1431 
1432 enum perf_event_task_context {
1433 	perf_invalid_context = -1,
1434 	perf_hw_context = 0,
1435 	perf_sw_context,
1436 	perf_nr_task_contexts,
1437 };
1438 
1439 /* Track pages that require TLB flushes */
1440 struct tlbflush_unmap_batch {
1441 	/*
1442 	 * Each bit set is a CPU that potentially has a TLB entry for one of
1443 	 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1444 	 */
1445 	struct cpumask cpumask;
1446 
1447 	/* True if any bit in cpumask is set */
1448 	bool flush_required;
1449 
1450 	/*
1451 	 * If true then the PTE was dirty when unmapped. The entry must be
1452 	 * flushed before IO is initiated or a stale TLB entry potentially
1453 	 * allows an update without redirtying the page.
1454 	 */
1455 	bool writable;
1456 };
1457 
1458 struct task_struct {
1459 	volatile long state;	/* -1 unrunnable, 0 runnable, >0 stopped */
1460 	void *stack;
1461 	atomic_t usage;
1462 	unsigned int flags;	/* per process flags, defined below */
1463 	unsigned int ptrace;
1464 
1465 #ifdef CONFIG_SMP
1466 	struct llist_node wake_entry;
1467 	int on_cpu;
1468 	unsigned int wakee_flips;
1469 	unsigned long wakee_flip_decay_ts;
1470 	struct task_struct *last_wakee;
1471 
1472 	int wake_cpu;
1473 #endif
1474 	int on_rq;
1475 
1476 	int prio, static_prio, normal_prio;
1477 	unsigned int rt_priority;
1478 	const struct sched_class *sched_class;
1479 	struct sched_entity se;
1480 	struct sched_rt_entity rt;
1481 #ifdef CONFIG_CGROUP_SCHED
1482 	struct task_group *sched_task_group;
1483 #endif
1484 	struct sched_dl_entity dl;
1485 
1486 #ifdef CONFIG_PREEMPT_NOTIFIERS
1487 	/* list of struct preempt_notifier: */
1488 	struct hlist_head preempt_notifiers;
1489 #endif
1490 
1491 #ifdef CONFIG_BLK_DEV_IO_TRACE
1492 	unsigned int btrace_seq;
1493 #endif
1494 
1495 	unsigned int policy;
1496 	int nr_cpus_allowed;
1497 	cpumask_t cpus_allowed;
1498 
1499 #ifdef CONFIG_PREEMPT_RCU
1500 	int rcu_read_lock_nesting;
1501 	union rcu_special rcu_read_unlock_special;
1502 	struct list_head rcu_node_entry;
1503 	struct rcu_node *rcu_blocked_node;
1504 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1505 #ifdef CONFIG_TASKS_RCU
1506 	unsigned long rcu_tasks_nvcsw;
1507 	bool rcu_tasks_holdout;
1508 	struct list_head rcu_tasks_holdout_list;
1509 	int rcu_tasks_idle_cpu;
1510 #endif /* #ifdef CONFIG_TASKS_RCU */
1511 
1512 #ifdef CONFIG_SCHED_INFO
1513 	struct sched_info sched_info;
1514 #endif
1515 
1516 	struct list_head tasks;
1517 #ifdef CONFIG_SMP
1518 	struct plist_node pushable_tasks;
1519 	struct rb_node pushable_dl_tasks;
1520 #endif
1521 
1522 	struct mm_struct *mm, *active_mm;
1523 	/* per-thread vma caching */
1524 	u32 vmacache_seqnum;
1525 	struct vm_area_struct *vmacache[VMACACHE_SIZE];
1526 #if defined(SPLIT_RSS_COUNTING)
1527 	struct task_rss_stat	rss_stat;
1528 #endif
1529 /* task state */
1530 	int exit_state;
1531 	int exit_code, exit_signal;
1532 	int pdeath_signal;  /*  The signal sent when the parent dies  */
1533 	unsigned long jobctl;	/* JOBCTL_*, siglock protected */
1534 
1535 	/* Used for emulating ABI behavior of previous Linux versions */
1536 	unsigned int personality;
1537 
1538 	/* scheduler bits, serialized by scheduler locks */
1539 	unsigned sched_reset_on_fork:1;
1540 	unsigned sched_contributes_to_load:1;
1541 	unsigned sched_migrated:1;
1542 	unsigned sched_remote_wakeup:1;
1543 	unsigned :0; /* force alignment to the next boundary */
1544 
1545 	/* unserialized, strictly 'current' */
1546 	unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1547 	unsigned in_iowait:1;
1548 #ifdef CONFIG_MEMCG
1549 	unsigned memcg_may_oom:1;
1550 #ifndef CONFIG_SLOB
1551 	unsigned memcg_kmem_skip_account:1;
1552 #endif
1553 #endif
1554 #ifdef CONFIG_COMPAT_BRK
1555 	unsigned brk_randomized:1;
1556 #endif
1557 
1558 	unsigned long atomic_flags; /* Flags needing atomic access. */
1559 
1560 	struct restart_block restart_block;
1561 
1562 	pid_t pid;
1563 	pid_t tgid;
1564 
1565 #ifdef CONFIG_CC_STACKPROTECTOR
1566 	/* Canary value for the -fstack-protector gcc feature */
1567 	unsigned long stack_canary;
1568 #endif
1569 	/*
1570 	 * pointers to (original) parent process, youngest child, younger sibling,
1571 	 * older sibling, respectively.  (p->father can be replaced with
1572 	 * p->real_parent->pid)
1573 	 */
1574 	struct task_struct __rcu *real_parent; /* real parent process */
1575 	struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1576 	/*
1577 	 * children/sibling forms the list of my natural children
1578 	 */
1579 	struct list_head children;	/* list of my children */
1580 	struct list_head sibling;	/* linkage in my parent's children list */
1581 	struct task_struct *group_leader;	/* threadgroup leader */
1582 
1583 	/*
1584 	 * ptraced is the list of tasks this task is using ptrace on.
1585 	 * This includes both natural children and PTRACE_ATTACH targets.
1586 	 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1587 	 */
1588 	struct list_head ptraced;
1589 	struct list_head ptrace_entry;
1590 
1591 	/* PID/PID hash table linkage. */
1592 	struct pid_link pids[PIDTYPE_MAX];
1593 	struct list_head thread_group;
1594 	struct list_head thread_node;
1595 
1596 	struct completion *vfork_done;		/* for vfork() */
1597 	int __user *set_child_tid;		/* CLONE_CHILD_SETTID */
1598 	int __user *clear_child_tid;		/* CLONE_CHILD_CLEARTID */
1599 
1600 	cputime_t utime, stime, utimescaled, stimescaled;
1601 	cputime_t gtime;
1602 	struct prev_cputime prev_cputime;
1603 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1604 	seqcount_t vtime_seqcount;
1605 	unsigned long long vtime_snap;
1606 	enum {
1607 		/* Task is sleeping or running in a CPU with VTIME inactive */
1608 		VTIME_INACTIVE = 0,
1609 		/* Task runs in userspace in a CPU with VTIME active */
1610 		VTIME_USER,
1611 		/* Task runs in kernelspace in a CPU with VTIME active */
1612 		VTIME_SYS,
1613 	} vtime_snap_whence;
1614 #endif
1615 
1616 #ifdef CONFIG_NO_HZ_FULL
1617 	atomic_t tick_dep_mask;
1618 #endif
1619 	unsigned long nvcsw, nivcsw; /* context switch counts */
1620 	u64 start_time;		/* monotonic time in nsec */
1621 	u64 real_start_time;	/* boot based time in nsec */
1622 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1623 	unsigned long min_flt, maj_flt;
1624 
1625 	struct task_cputime cputime_expires;
1626 	struct list_head cpu_timers[3];
1627 
1628 /* process credentials */
1629 	const struct cred __rcu *real_cred; /* objective and real subjective task
1630 					 * credentials (COW) */
1631 	const struct cred __rcu *cred;	/* effective (overridable) subjective task
1632 					 * credentials (COW) */
1633 	char comm[TASK_COMM_LEN]; /* executable name excluding path
1634 				     - access with [gs]et_task_comm (which lock
1635 				       it with task_lock())
1636 				     - initialized normally by setup_new_exec */
1637 /* file system info */
1638 	struct nameidata *nameidata;
1639 #ifdef CONFIG_SYSVIPC
1640 /* ipc stuff */
1641 	struct sysv_sem sysvsem;
1642 	struct sysv_shm sysvshm;
1643 #endif
1644 #ifdef CONFIG_DETECT_HUNG_TASK
1645 /* hung task detection */
1646 	unsigned long last_switch_count;
1647 #endif
1648 /* filesystem information */
1649 	struct fs_struct *fs;
1650 /* open file information */
1651 	struct files_struct *files;
1652 /* namespaces */
1653 	struct nsproxy *nsproxy;
1654 /* signal handlers */
1655 	struct signal_struct *signal;
1656 	struct sighand_struct *sighand;
1657 
1658 	sigset_t blocked, real_blocked;
1659 	sigset_t saved_sigmask;	/* restored if set_restore_sigmask() was used */
1660 	struct sigpending pending;
1661 
1662 	unsigned long sas_ss_sp;
1663 	size_t sas_ss_size;
1664 	unsigned sas_ss_flags;
1665 
1666 	struct callback_head *task_works;
1667 
1668 	struct audit_context *audit_context;
1669 #ifdef CONFIG_AUDITSYSCALL
1670 	kuid_t loginuid;
1671 	unsigned int sessionid;
1672 #endif
1673 	struct seccomp seccomp;
1674 
1675 /* Thread group tracking */
1676    	u32 parent_exec_id;
1677    	u32 self_exec_id;
1678 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1679  * mempolicy */
1680 	spinlock_t alloc_lock;
1681 
1682 	/* Protection of the PI data structures: */
1683 	raw_spinlock_t pi_lock;
1684 
1685 	struct wake_q_node wake_q;
1686 
1687 #ifdef CONFIG_RT_MUTEXES
1688 	/* PI waiters blocked on a rt_mutex held by this task */
1689 	struct rb_root pi_waiters;
1690 	struct rb_node *pi_waiters_leftmost;
1691 	/* Deadlock detection and priority inheritance handling */
1692 	struct rt_mutex_waiter *pi_blocked_on;
1693 #endif
1694 
1695 #ifdef CONFIG_DEBUG_MUTEXES
1696 	/* mutex deadlock detection */
1697 	struct mutex_waiter *blocked_on;
1698 #endif
1699 #ifdef CONFIG_TRACE_IRQFLAGS
1700 	unsigned int irq_events;
1701 	unsigned long hardirq_enable_ip;
1702 	unsigned long hardirq_disable_ip;
1703 	unsigned int hardirq_enable_event;
1704 	unsigned int hardirq_disable_event;
1705 	int hardirqs_enabled;
1706 	int hardirq_context;
1707 	unsigned long softirq_disable_ip;
1708 	unsigned long softirq_enable_ip;
1709 	unsigned int softirq_disable_event;
1710 	unsigned int softirq_enable_event;
1711 	int softirqs_enabled;
1712 	int softirq_context;
1713 #endif
1714 #ifdef CONFIG_LOCKDEP
1715 # define MAX_LOCK_DEPTH 48UL
1716 	u64 curr_chain_key;
1717 	int lockdep_depth;
1718 	unsigned int lockdep_recursion;
1719 	struct held_lock held_locks[MAX_LOCK_DEPTH];
1720 	gfp_t lockdep_reclaim_gfp;
1721 #endif
1722 #ifdef CONFIG_UBSAN
1723 	unsigned int in_ubsan;
1724 #endif
1725 
1726 /* journalling filesystem info */
1727 	void *journal_info;
1728 
1729 /* stacked block device info */
1730 	struct bio_list *bio_list;
1731 
1732 #ifdef CONFIG_BLOCK
1733 /* stack plugging */
1734 	struct blk_plug *plug;
1735 #endif
1736 
1737 /* VM state */
1738 	struct reclaim_state *reclaim_state;
1739 
1740 	struct backing_dev_info *backing_dev_info;
1741 
1742 	struct io_context *io_context;
1743 
1744 	unsigned long ptrace_message;
1745 	siginfo_t *last_siginfo; /* For ptrace use.  */
1746 	struct task_io_accounting ioac;
1747 #if defined(CONFIG_TASK_XACCT)
1748 	u64 acct_rss_mem1;	/* accumulated rss usage */
1749 	u64 acct_vm_mem1;	/* accumulated virtual memory usage */
1750 	cputime_t acct_timexpd;	/* stime + utime since last update */
1751 #endif
1752 #ifdef CONFIG_CPUSETS
1753 	nodemask_t mems_allowed;	/* Protected by alloc_lock */
1754 	seqcount_t mems_allowed_seq;	/* Seqence no to catch updates */
1755 	int cpuset_mem_spread_rotor;
1756 	int cpuset_slab_spread_rotor;
1757 #endif
1758 #ifdef CONFIG_CGROUPS
1759 	/* Control Group info protected by css_set_lock */
1760 	struct css_set __rcu *cgroups;
1761 	/* cg_list protected by css_set_lock and tsk->alloc_lock */
1762 	struct list_head cg_list;
1763 #endif
1764 #ifdef CONFIG_FUTEX
1765 	struct robust_list_head __user *robust_list;
1766 #ifdef CONFIG_COMPAT
1767 	struct compat_robust_list_head __user *compat_robust_list;
1768 #endif
1769 	struct list_head pi_state_list;
1770 	struct futex_pi_state *pi_state_cache;
1771 #endif
1772 #ifdef CONFIG_PERF_EVENTS
1773 	struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1774 	struct mutex perf_event_mutex;
1775 	struct list_head perf_event_list;
1776 #endif
1777 #ifdef CONFIG_DEBUG_PREEMPT
1778 	unsigned long preempt_disable_ip;
1779 #endif
1780 #ifdef CONFIG_NUMA
1781 	struct mempolicy *mempolicy;	/* Protected by alloc_lock */
1782 	short il_next;
1783 	short pref_node_fork;
1784 #endif
1785 #ifdef CONFIG_NUMA_BALANCING
1786 	int numa_scan_seq;
1787 	unsigned int numa_scan_period;
1788 	unsigned int numa_scan_period_max;
1789 	int numa_preferred_nid;
1790 	unsigned long numa_migrate_retry;
1791 	u64 node_stamp;			/* migration stamp  */
1792 	u64 last_task_numa_placement;
1793 	u64 last_sum_exec_runtime;
1794 	struct callback_head numa_work;
1795 
1796 	struct list_head numa_entry;
1797 	struct numa_group *numa_group;
1798 
1799 	/*
1800 	 * numa_faults is an array split into four regions:
1801 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1802 	 * in this precise order.
1803 	 *
1804 	 * faults_memory: Exponential decaying average of faults on a per-node
1805 	 * basis. Scheduling placement decisions are made based on these
1806 	 * counts. The values remain static for the duration of a PTE scan.
1807 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1808 	 * hinting fault was incurred.
1809 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1810 	 * during the current scan window. When the scan completes, the counts
1811 	 * in faults_memory and faults_cpu decay and these values are copied.
1812 	 */
1813 	unsigned long *numa_faults;
1814 	unsigned long total_numa_faults;
1815 
1816 	/*
1817 	 * numa_faults_locality tracks if faults recorded during the last
1818 	 * scan window were remote/local or failed to migrate. The task scan
1819 	 * period is adapted based on the locality of the faults with different
1820 	 * weights depending on whether they were shared or private faults
1821 	 */
1822 	unsigned long numa_faults_locality[3];
1823 
1824 	unsigned long numa_pages_migrated;
1825 #endif /* CONFIG_NUMA_BALANCING */
1826 
1827 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1828 	struct tlbflush_unmap_batch tlb_ubc;
1829 #endif
1830 
1831 	struct rcu_head rcu;
1832 
1833 	/*
1834 	 * cache last used pipe for splice
1835 	 */
1836 	struct pipe_inode_info *splice_pipe;
1837 
1838 	struct page_frag task_frag;
1839 
1840 #ifdef	CONFIG_TASK_DELAY_ACCT
1841 	struct task_delay_info *delays;
1842 #endif
1843 #ifdef CONFIG_FAULT_INJECTION
1844 	int make_it_fail;
1845 #endif
1846 	/*
1847 	 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1848 	 * balance_dirty_pages() for some dirty throttling pause
1849 	 */
1850 	int nr_dirtied;
1851 	int nr_dirtied_pause;
1852 	unsigned long dirty_paused_when; /* start of a write-and-pause period */
1853 
1854 #ifdef CONFIG_LATENCYTOP
1855 	int latency_record_count;
1856 	struct latency_record latency_record[LT_SAVECOUNT];
1857 #endif
1858 	/*
1859 	 * time slack values; these are used to round up poll() and
1860 	 * select() etc timeout values. These are in nanoseconds.
1861 	 */
1862 	u64 timer_slack_ns;
1863 	u64 default_timer_slack_ns;
1864 
1865 #ifdef CONFIG_KASAN
1866 	unsigned int kasan_depth;
1867 #endif
1868 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1869 	/* Index of current stored address in ret_stack */
1870 	int curr_ret_stack;
1871 	/* Stack of return addresses for return function tracing */
1872 	struct ftrace_ret_stack	*ret_stack;
1873 	/* time stamp for last schedule */
1874 	unsigned long long ftrace_timestamp;
1875 	/*
1876 	 * Number of functions that haven't been traced
1877 	 * because of depth overrun.
1878 	 */
1879 	atomic_t trace_overrun;
1880 	/* Pause for the tracing */
1881 	atomic_t tracing_graph_pause;
1882 #endif
1883 #ifdef CONFIG_TRACING
1884 	/* state flags for use by tracers */
1885 	unsigned long trace;
1886 	/* bitmask and counter of trace recursion */
1887 	unsigned long trace_recursion;
1888 #endif /* CONFIG_TRACING */
1889 #ifdef CONFIG_KCOV
1890 	/* Coverage collection mode enabled for this task (0 if disabled). */
1891 	enum kcov_mode kcov_mode;
1892 	/* Size of the kcov_area. */
1893 	unsigned	kcov_size;
1894 	/* Buffer for coverage collection. */
1895 	void		*kcov_area;
1896 	/* kcov desciptor wired with this task or NULL. */
1897 	struct kcov	*kcov;
1898 #endif
1899 #ifdef CONFIG_MEMCG
1900 	struct mem_cgroup *memcg_in_oom;
1901 	gfp_t memcg_oom_gfp_mask;
1902 	int memcg_oom_order;
1903 
1904 	/* number of pages to reclaim on returning to userland */
1905 	unsigned int memcg_nr_pages_over_high;
1906 #endif
1907 #ifdef CONFIG_UPROBES
1908 	struct uprobe_task *utask;
1909 #endif
1910 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1911 	unsigned int	sequential_io;
1912 	unsigned int	sequential_io_avg;
1913 #endif
1914 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1915 	unsigned long	task_state_change;
1916 #endif
1917 	int pagefault_disabled;
1918 #ifdef CONFIG_MMU
1919 	struct task_struct *oom_reaper_list;
1920 #endif
1921 /* CPU-specific state of this task */
1922 	struct thread_struct thread;
1923 /*
1924  * WARNING: on x86, 'thread_struct' contains a variable-sized
1925  * structure.  It *MUST* be at the end of 'task_struct'.
1926  *
1927  * Do not put anything below here!
1928  */
1929 };
1930 
1931 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
1932 extern int arch_task_struct_size __read_mostly;
1933 #else
1934 # define arch_task_struct_size (sizeof(struct task_struct))
1935 #endif
1936 
1937 /* Future-safe accessor for struct task_struct's cpus_allowed. */
1938 #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
1939 
1940 static inline int tsk_nr_cpus_allowed(struct task_struct *p)
1941 {
1942 	return p->nr_cpus_allowed;
1943 }
1944 
1945 #define TNF_MIGRATED	0x01
1946 #define TNF_NO_GROUP	0x02
1947 #define TNF_SHARED	0x04
1948 #define TNF_FAULT_LOCAL	0x08
1949 #define TNF_MIGRATE_FAIL 0x10
1950 
1951 #ifdef CONFIG_NUMA_BALANCING
1952 extern void task_numa_fault(int last_node, int node, int pages, int flags);
1953 extern pid_t task_numa_group_id(struct task_struct *p);
1954 extern void set_numabalancing_state(bool enabled);
1955 extern void task_numa_free(struct task_struct *p);
1956 extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
1957 					int src_nid, int dst_cpu);
1958 #else
1959 static inline void task_numa_fault(int last_node, int node, int pages,
1960 				   int flags)
1961 {
1962 }
1963 static inline pid_t task_numa_group_id(struct task_struct *p)
1964 {
1965 	return 0;
1966 }
1967 static inline void set_numabalancing_state(bool enabled)
1968 {
1969 }
1970 static inline void task_numa_free(struct task_struct *p)
1971 {
1972 }
1973 static inline bool should_numa_migrate_memory(struct task_struct *p,
1974 				struct page *page, int src_nid, int dst_cpu)
1975 {
1976 	return true;
1977 }
1978 #endif
1979 
1980 static inline struct pid *task_pid(struct task_struct *task)
1981 {
1982 	return task->pids[PIDTYPE_PID].pid;
1983 }
1984 
1985 static inline struct pid *task_tgid(struct task_struct *task)
1986 {
1987 	return task->group_leader->pids[PIDTYPE_PID].pid;
1988 }
1989 
1990 /*
1991  * Without tasklist or rcu lock it is not safe to dereference
1992  * the result of task_pgrp/task_session even if task == current,
1993  * we can race with another thread doing sys_setsid/sys_setpgid.
1994  */
1995 static inline struct pid *task_pgrp(struct task_struct *task)
1996 {
1997 	return task->group_leader->pids[PIDTYPE_PGID].pid;
1998 }
1999 
2000 static inline struct pid *task_session(struct task_struct *task)
2001 {
2002 	return task->group_leader->pids[PIDTYPE_SID].pid;
2003 }
2004 
2005 struct pid_namespace;
2006 
2007 /*
2008  * the helpers to get the task's different pids as they are seen
2009  * from various namespaces
2010  *
2011  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
2012  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
2013  *                     current.
2014  * task_xid_nr_ns()  : id seen from the ns specified;
2015  *
2016  * set_task_vxid()   : assigns a virtual id to a task;
2017  *
2018  * see also pid_nr() etc in include/linux/pid.h
2019  */
2020 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2021 			struct pid_namespace *ns);
2022 
2023 static inline pid_t task_pid_nr(struct task_struct *tsk)
2024 {
2025 	return tsk->pid;
2026 }
2027 
2028 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2029 					struct pid_namespace *ns)
2030 {
2031 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2032 }
2033 
2034 static inline pid_t task_pid_vnr(struct task_struct *tsk)
2035 {
2036 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2037 }
2038 
2039 
2040 static inline pid_t task_tgid_nr(struct task_struct *tsk)
2041 {
2042 	return tsk->tgid;
2043 }
2044 
2045 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2046 
2047 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2048 {
2049 	return pid_vnr(task_tgid(tsk));
2050 }
2051 
2052 
2053 static inline int pid_alive(const struct task_struct *p);
2054 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2055 {
2056 	pid_t pid = 0;
2057 
2058 	rcu_read_lock();
2059 	if (pid_alive(tsk))
2060 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2061 	rcu_read_unlock();
2062 
2063 	return pid;
2064 }
2065 
2066 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2067 {
2068 	return task_ppid_nr_ns(tsk, &init_pid_ns);
2069 }
2070 
2071 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2072 					struct pid_namespace *ns)
2073 {
2074 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2075 }
2076 
2077 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2078 {
2079 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2080 }
2081 
2082 
2083 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2084 					struct pid_namespace *ns)
2085 {
2086 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2087 }
2088 
2089 static inline pid_t task_session_vnr(struct task_struct *tsk)
2090 {
2091 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2092 }
2093 
2094 /* obsolete, do not use */
2095 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2096 {
2097 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
2098 }
2099 
2100 /**
2101  * pid_alive - check that a task structure is not stale
2102  * @p: Task structure to be checked.
2103  *
2104  * Test if a process is not yet dead (at most zombie state)
2105  * If pid_alive fails, then pointers within the task structure
2106  * can be stale and must not be dereferenced.
2107  *
2108  * Return: 1 if the process is alive. 0 otherwise.
2109  */
2110 static inline int pid_alive(const struct task_struct *p)
2111 {
2112 	return p->pids[PIDTYPE_PID].pid != NULL;
2113 }
2114 
2115 /**
2116  * is_global_init - check if a task structure is init. Since init
2117  * is free to have sub-threads we need to check tgid.
2118  * @tsk: Task structure to be checked.
2119  *
2120  * Check if a task structure is the first user space task the kernel created.
2121  *
2122  * Return: 1 if the task structure is init. 0 otherwise.
2123  */
2124 static inline int is_global_init(struct task_struct *tsk)
2125 {
2126 	return task_tgid_nr(tsk) == 1;
2127 }
2128 
2129 extern struct pid *cad_pid;
2130 
2131 extern void free_task(struct task_struct *tsk);
2132 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2133 
2134 extern void __put_task_struct(struct task_struct *t);
2135 
2136 static inline void put_task_struct(struct task_struct *t)
2137 {
2138 	if (atomic_dec_and_test(&t->usage))
2139 		__put_task_struct(t);
2140 }
2141 
2142 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2143 extern void task_cputime(struct task_struct *t,
2144 			 cputime_t *utime, cputime_t *stime);
2145 extern void task_cputime_scaled(struct task_struct *t,
2146 				cputime_t *utimescaled, cputime_t *stimescaled);
2147 extern cputime_t task_gtime(struct task_struct *t);
2148 #else
2149 static inline void task_cputime(struct task_struct *t,
2150 				cputime_t *utime, cputime_t *stime)
2151 {
2152 	if (utime)
2153 		*utime = t->utime;
2154 	if (stime)
2155 		*stime = t->stime;
2156 }
2157 
2158 static inline void task_cputime_scaled(struct task_struct *t,
2159 				       cputime_t *utimescaled,
2160 				       cputime_t *stimescaled)
2161 {
2162 	if (utimescaled)
2163 		*utimescaled = t->utimescaled;
2164 	if (stimescaled)
2165 		*stimescaled = t->stimescaled;
2166 }
2167 
2168 static inline cputime_t task_gtime(struct task_struct *t)
2169 {
2170 	return t->gtime;
2171 }
2172 #endif
2173 extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2174 extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2175 
2176 /*
2177  * Per process flags
2178  */
2179 #define PF_EXITING	0x00000004	/* getting shut down */
2180 #define PF_EXITPIDONE	0x00000008	/* pi exit done on shut down */
2181 #define PF_VCPU		0x00000010	/* I'm a virtual CPU */
2182 #define PF_WQ_WORKER	0x00000020	/* I'm a workqueue worker */
2183 #define PF_FORKNOEXEC	0x00000040	/* forked but didn't exec */
2184 #define PF_MCE_PROCESS  0x00000080      /* process policy on mce errors */
2185 #define PF_SUPERPRIV	0x00000100	/* used super-user privileges */
2186 #define PF_DUMPCORE	0x00000200	/* dumped core */
2187 #define PF_SIGNALED	0x00000400	/* killed by a signal */
2188 #define PF_MEMALLOC	0x00000800	/* Allocating memory */
2189 #define PF_NPROC_EXCEEDED 0x00001000	/* set_user noticed that RLIMIT_NPROC was exceeded */
2190 #define PF_USED_MATH	0x00002000	/* if unset the fpu must be initialized before use */
2191 #define PF_USED_ASYNC	0x00004000	/* used async_schedule*(), used by module init */
2192 #define PF_NOFREEZE	0x00008000	/* this thread should not be frozen */
2193 #define PF_FROZEN	0x00010000	/* frozen for system suspend */
2194 #define PF_FSTRANS	0x00020000	/* inside a filesystem transaction */
2195 #define PF_KSWAPD	0x00040000	/* I am kswapd */
2196 #define PF_MEMALLOC_NOIO 0x00080000	/* Allocating memory without IO involved */
2197 #define PF_LESS_THROTTLE 0x00100000	/* Throttle me less: I clean memory */
2198 #define PF_KTHREAD	0x00200000	/* I am a kernel thread */
2199 #define PF_RANDOMIZE	0x00400000	/* randomize virtual address space */
2200 #define PF_SWAPWRITE	0x00800000	/* Allowed to write to swap */
2201 #define PF_NO_SETAFFINITY 0x04000000	/* Userland is not allowed to meddle with cpus_allowed */
2202 #define PF_MCE_EARLY    0x08000000      /* Early kill for mce process policy */
2203 #define PF_MUTEX_TESTER	0x20000000	/* Thread belongs to the rt mutex tester */
2204 #define PF_FREEZER_SKIP	0x40000000	/* Freezer should not count it as freezable */
2205 #define PF_SUSPEND_TASK 0x80000000      /* this thread called freeze_processes and should not be frozen */
2206 
2207 /*
2208  * Only the _current_ task can read/write to tsk->flags, but other
2209  * tasks can access tsk->flags in readonly mode for example
2210  * with tsk_used_math (like during threaded core dumping).
2211  * There is however an exception to this rule during ptrace
2212  * or during fork: the ptracer task is allowed to write to the
2213  * child->flags of its traced child (same goes for fork, the parent
2214  * can write to the child->flags), because we're guaranteed the
2215  * child is not running and in turn not changing child->flags
2216  * at the same time the parent does it.
2217  */
2218 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2219 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2220 #define clear_used_math() clear_stopped_child_used_math(current)
2221 #define set_used_math() set_stopped_child_used_math(current)
2222 #define conditional_stopped_child_used_math(condition, child) \
2223 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2224 #define conditional_used_math(condition) \
2225 	conditional_stopped_child_used_math(condition, current)
2226 #define copy_to_stopped_child_used_math(child) \
2227 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2228 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2229 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2230 #define used_math() tsk_used_math(current)
2231 
2232 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2233  * __GFP_FS is also cleared as it implies __GFP_IO.
2234  */
2235 static inline gfp_t memalloc_noio_flags(gfp_t flags)
2236 {
2237 	if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2238 		flags &= ~(__GFP_IO | __GFP_FS);
2239 	return flags;
2240 }
2241 
2242 static inline unsigned int memalloc_noio_save(void)
2243 {
2244 	unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2245 	current->flags |= PF_MEMALLOC_NOIO;
2246 	return flags;
2247 }
2248 
2249 static inline void memalloc_noio_restore(unsigned int flags)
2250 {
2251 	current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2252 }
2253 
2254 /* Per-process atomic flags. */
2255 #define PFA_NO_NEW_PRIVS 0	/* May not gain new privileges. */
2256 #define PFA_SPREAD_PAGE  1      /* Spread page cache over cpuset */
2257 #define PFA_SPREAD_SLAB  2      /* Spread some slab caches over cpuset */
2258 #define PFA_LMK_WAITING  3      /* Lowmemorykiller is waiting */
2259 
2260 
2261 #define TASK_PFA_TEST(name, func)					\
2262 	static inline bool task_##func(struct task_struct *p)		\
2263 	{ return test_bit(PFA_##name, &p->atomic_flags); }
2264 #define TASK_PFA_SET(name, func)					\
2265 	static inline void task_set_##func(struct task_struct *p)	\
2266 	{ set_bit(PFA_##name, &p->atomic_flags); }
2267 #define TASK_PFA_CLEAR(name, func)					\
2268 	static inline void task_clear_##func(struct task_struct *p)	\
2269 	{ clear_bit(PFA_##name, &p->atomic_flags); }
2270 
2271 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2272 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2273 
2274 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2275 TASK_PFA_SET(SPREAD_PAGE, spread_page)
2276 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2277 
2278 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2279 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2280 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2281 
2282 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
2283 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
2284 
2285 /*
2286  * task->jobctl flags
2287  */
2288 #define JOBCTL_STOP_SIGMASK	0xffff	/* signr of the last group stop */
2289 
2290 #define JOBCTL_STOP_DEQUEUED_BIT 16	/* stop signal dequeued */
2291 #define JOBCTL_STOP_PENDING_BIT	17	/* task should stop for group stop */
2292 #define JOBCTL_STOP_CONSUME_BIT	18	/* consume group stop count */
2293 #define JOBCTL_TRAP_STOP_BIT	19	/* trap for STOP */
2294 #define JOBCTL_TRAP_NOTIFY_BIT	20	/* trap for NOTIFY */
2295 #define JOBCTL_TRAPPING_BIT	21	/* switching to TRACED */
2296 #define JOBCTL_LISTENING_BIT	22	/* ptracer is listening for events */
2297 
2298 #define JOBCTL_STOP_DEQUEUED	(1UL << JOBCTL_STOP_DEQUEUED_BIT)
2299 #define JOBCTL_STOP_PENDING	(1UL << JOBCTL_STOP_PENDING_BIT)
2300 #define JOBCTL_STOP_CONSUME	(1UL << JOBCTL_STOP_CONSUME_BIT)
2301 #define JOBCTL_TRAP_STOP	(1UL << JOBCTL_TRAP_STOP_BIT)
2302 #define JOBCTL_TRAP_NOTIFY	(1UL << JOBCTL_TRAP_NOTIFY_BIT)
2303 #define JOBCTL_TRAPPING		(1UL << JOBCTL_TRAPPING_BIT)
2304 #define JOBCTL_LISTENING	(1UL << JOBCTL_LISTENING_BIT)
2305 
2306 #define JOBCTL_TRAP_MASK	(JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2307 #define JOBCTL_PENDING_MASK	(JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2308 
2309 extern bool task_set_jobctl_pending(struct task_struct *task,
2310 				    unsigned long mask);
2311 extern void task_clear_jobctl_trapping(struct task_struct *task);
2312 extern void task_clear_jobctl_pending(struct task_struct *task,
2313 				      unsigned long mask);
2314 
2315 static inline void rcu_copy_process(struct task_struct *p)
2316 {
2317 #ifdef CONFIG_PREEMPT_RCU
2318 	p->rcu_read_lock_nesting = 0;
2319 	p->rcu_read_unlock_special.s = 0;
2320 	p->rcu_blocked_node = NULL;
2321 	INIT_LIST_HEAD(&p->rcu_node_entry);
2322 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2323 #ifdef CONFIG_TASKS_RCU
2324 	p->rcu_tasks_holdout = false;
2325 	INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2326 	p->rcu_tasks_idle_cpu = -1;
2327 #endif /* #ifdef CONFIG_TASKS_RCU */
2328 }
2329 
2330 static inline void tsk_restore_flags(struct task_struct *task,
2331 				unsigned long orig_flags, unsigned long flags)
2332 {
2333 	task->flags &= ~flags;
2334 	task->flags |= orig_flags & flags;
2335 }
2336 
2337 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2338 				     const struct cpumask *trial);
2339 extern int task_can_attach(struct task_struct *p,
2340 			   const struct cpumask *cs_cpus_allowed);
2341 #ifdef CONFIG_SMP
2342 extern void do_set_cpus_allowed(struct task_struct *p,
2343 			       const struct cpumask *new_mask);
2344 
2345 extern int set_cpus_allowed_ptr(struct task_struct *p,
2346 				const struct cpumask *new_mask);
2347 #else
2348 static inline void do_set_cpus_allowed(struct task_struct *p,
2349 				      const struct cpumask *new_mask)
2350 {
2351 }
2352 static inline int set_cpus_allowed_ptr(struct task_struct *p,
2353 				       const struct cpumask *new_mask)
2354 {
2355 	if (!cpumask_test_cpu(0, new_mask))
2356 		return -EINVAL;
2357 	return 0;
2358 }
2359 #endif
2360 
2361 #ifdef CONFIG_NO_HZ_COMMON
2362 void calc_load_enter_idle(void);
2363 void calc_load_exit_idle(void);
2364 #else
2365 static inline void calc_load_enter_idle(void) { }
2366 static inline void calc_load_exit_idle(void) { }
2367 #endif /* CONFIG_NO_HZ_COMMON */
2368 
2369 /*
2370  * Do not use outside of architecture code which knows its limitations.
2371  *
2372  * sched_clock() has no promise of monotonicity or bounded drift between
2373  * CPUs, use (which you should not) requires disabling IRQs.
2374  *
2375  * Please use one of the three interfaces below.
2376  */
2377 extern unsigned long long notrace sched_clock(void);
2378 /*
2379  * See the comment in kernel/sched/clock.c
2380  */
2381 extern u64 running_clock(void);
2382 extern u64 sched_clock_cpu(int cpu);
2383 
2384 
2385 extern void sched_clock_init(void);
2386 
2387 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2388 static inline void sched_clock_tick(void)
2389 {
2390 }
2391 
2392 static inline void sched_clock_idle_sleep_event(void)
2393 {
2394 }
2395 
2396 static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2397 {
2398 }
2399 
2400 static inline u64 cpu_clock(int cpu)
2401 {
2402 	return sched_clock();
2403 }
2404 
2405 static inline u64 local_clock(void)
2406 {
2407 	return sched_clock();
2408 }
2409 #else
2410 /*
2411  * Architectures can set this to 1 if they have specified
2412  * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2413  * but then during bootup it turns out that sched_clock()
2414  * is reliable after all:
2415  */
2416 extern int sched_clock_stable(void);
2417 extern void set_sched_clock_stable(void);
2418 extern void clear_sched_clock_stable(void);
2419 
2420 extern void sched_clock_tick(void);
2421 extern void sched_clock_idle_sleep_event(void);
2422 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2423 
2424 /*
2425  * As outlined in clock.c, provides a fast, high resolution, nanosecond
2426  * time source that is monotonic per cpu argument and has bounded drift
2427  * between cpus.
2428  *
2429  * ######################### BIG FAT WARNING ##########################
2430  * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2431  * # go backwards !!                                                  #
2432  * ####################################################################
2433  */
2434 static inline u64 cpu_clock(int cpu)
2435 {
2436 	return sched_clock_cpu(cpu);
2437 }
2438 
2439 static inline u64 local_clock(void)
2440 {
2441 	return sched_clock_cpu(raw_smp_processor_id());
2442 }
2443 #endif
2444 
2445 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2446 /*
2447  * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2448  * The reason for this explicit opt-in is not to have perf penalty with
2449  * slow sched_clocks.
2450  */
2451 extern void enable_sched_clock_irqtime(void);
2452 extern void disable_sched_clock_irqtime(void);
2453 #else
2454 static inline void enable_sched_clock_irqtime(void) {}
2455 static inline void disable_sched_clock_irqtime(void) {}
2456 #endif
2457 
2458 extern unsigned long long
2459 task_sched_runtime(struct task_struct *task);
2460 
2461 /* sched_exec is called by processes performing an exec */
2462 #ifdef CONFIG_SMP
2463 extern void sched_exec(void);
2464 #else
2465 #define sched_exec()   {}
2466 #endif
2467 
2468 extern void sched_clock_idle_sleep_event(void);
2469 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2470 
2471 #ifdef CONFIG_HOTPLUG_CPU
2472 extern void idle_task_exit(void);
2473 #else
2474 static inline void idle_task_exit(void) {}
2475 #endif
2476 
2477 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2478 extern void wake_up_nohz_cpu(int cpu);
2479 #else
2480 static inline void wake_up_nohz_cpu(int cpu) { }
2481 #endif
2482 
2483 #ifdef CONFIG_NO_HZ_FULL
2484 extern u64 scheduler_tick_max_deferment(void);
2485 #endif
2486 
2487 #ifdef CONFIG_SCHED_AUTOGROUP
2488 extern void sched_autogroup_create_attach(struct task_struct *p);
2489 extern void sched_autogroup_detach(struct task_struct *p);
2490 extern void sched_autogroup_fork(struct signal_struct *sig);
2491 extern void sched_autogroup_exit(struct signal_struct *sig);
2492 #ifdef CONFIG_PROC_FS
2493 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2494 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2495 #endif
2496 #else
2497 static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2498 static inline void sched_autogroup_detach(struct task_struct *p) { }
2499 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2500 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2501 #endif
2502 
2503 extern int yield_to(struct task_struct *p, bool preempt);
2504 extern void set_user_nice(struct task_struct *p, long nice);
2505 extern int task_prio(const struct task_struct *p);
2506 /**
2507  * task_nice - return the nice value of a given task.
2508  * @p: the task in question.
2509  *
2510  * Return: The nice value [ -20 ... 0 ... 19 ].
2511  */
2512 static inline int task_nice(const struct task_struct *p)
2513 {
2514 	return PRIO_TO_NICE((p)->static_prio);
2515 }
2516 extern int can_nice(const struct task_struct *p, const int nice);
2517 extern int task_curr(const struct task_struct *p);
2518 extern int idle_cpu(int cpu);
2519 extern int sched_setscheduler(struct task_struct *, int,
2520 			      const struct sched_param *);
2521 extern int sched_setscheduler_nocheck(struct task_struct *, int,
2522 				      const struct sched_param *);
2523 extern int sched_setattr(struct task_struct *,
2524 			 const struct sched_attr *);
2525 extern struct task_struct *idle_task(int cpu);
2526 /**
2527  * is_idle_task - is the specified task an idle task?
2528  * @p: the task in question.
2529  *
2530  * Return: 1 if @p is an idle task. 0 otherwise.
2531  */
2532 static inline bool is_idle_task(const struct task_struct *p)
2533 {
2534 	return p->pid == 0;
2535 }
2536 extern struct task_struct *curr_task(int cpu);
2537 extern void set_curr_task(int cpu, struct task_struct *p);
2538 
2539 void yield(void);
2540 
2541 union thread_union {
2542 	struct thread_info thread_info;
2543 	unsigned long stack[THREAD_SIZE/sizeof(long)];
2544 };
2545 
2546 #ifndef __HAVE_ARCH_KSTACK_END
2547 static inline int kstack_end(void *addr)
2548 {
2549 	/* Reliable end of stack detection:
2550 	 * Some APM bios versions misalign the stack
2551 	 */
2552 	return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2553 }
2554 #endif
2555 
2556 extern union thread_union init_thread_union;
2557 extern struct task_struct init_task;
2558 
2559 extern struct   mm_struct init_mm;
2560 
2561 extern struct pid_namespace init_pid_ns;
2562 
2563 /*
2564  * find a task by one of its numerical ids
2565  *
2566  * find_task_by_pid_ns():
2567  *      finds a task by its pid in the specified namespace
2568  * find_task_by_vpid():
2569  *      finds a task by its virtual pid
2570  *
2571  * see also find_vpid() etc in include/linux/pid.h
2572  */
2573 
2574 extern struct task_struct *find_task_by_vpid(pid_t nr);
2575 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2576 		struct pid_namespace *ns);
2577 
2578 /* per-UID process charging. */
2579 extern struct user_struct * alloc_uid(kuid_t);
2580 static inline struct user_struct *get_uid(struct user_struct *u)
2581 {
2582 	atomic_inc(&u->__count);
2583 	return u;
2584 }
2585 extern void free_uid(struct user_struct *);
2586 
2587 #include <asm/current.h>
2588 
2589 extern void xtime_update(unsigned long ticks);
2590 
2591 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2592 extern int wake_up_process(struct task_struct *tsk);
2593 extern void wake_up_new_task(struct task_struct *tsk);
2594 #ifdef CONFIG_SMP
2595  extern void kick_process(struct task_struct *tsk);
2596 #else
2597  static inline void kick_process(struct task_struct *tsk) { }
2598 #endif
2599 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2600 extern void sched_dead(struct task_struct *p);
2601 
2602 extern void proc_caches_init(void);
2603 extern void flush_signals(struct task_struct *);
2604 extern void ignore_signals(struct task_struct *);
2605 extern void flush_signal_handlers(struct task_struct *, int force_default);
2606 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2607 
2608 static inline int kernel_dequeue_signal(siginfo_t *info)
2609 {
2610 	struct task_struct *tsk = current;
2611 	siginfo_t __info;
2612 	int ret;
2613 
2614 	spin_lock_irq(&tsk->sighand->siglock);
2615 	ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2616 	spin_unlock_irq(&tsk->sighand->siglock);
2617 
2618 	return ret;
2619 }
2620 
2621 static inline void kernel_signal_stop(void)
2622 {
2623 	spin_lock_irq(&current->sighand->siglock);
2624 	if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2625 		__set_current_state(TASK_STOPPED);
2626 	spin_unlock_irq(&current->sighand->siglock);
2627 
2628 	schedule();
2629 }
2630 
2631 extern void release_task(struct task_struct * p);
2632 extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2633 extern int force_sigsegv(int, struct task_struct *);
2634 extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2635 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2636 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2637 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2638 				const struct cred *, u32);
2639 extern int kill_pgrp(struct pid *pid, int sig, int priv);
2640 extern int kill_pid(struct pid *pid, int sig, int priv);
2641 extern int kill_proc_info(int, struct siginfo *, pid_t);
2642 extern __must_check bool do_notify_parent(struct task_struct *, int);
2643 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2644 extern void force_sig(int, struct task_struct *);
2645 extern int send_sig(int, struct task_struct *, int);
2646 extern int zap_other_threads(struct task_struct *p);
2647 extern struct sigqueue *sigqueue_alloc(void);
2648 extern void sigqueue_free(struct sigqueue *);
2649 extern int send_sigqueue(struct sigqueue *,  struct task_struct *, int group);
2650 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2651 
2652 static inline void restore_saved_sigmask(void)
2653 {
2654 	if (test_and_clear_restore_sigmask())
2655 		__set_current_blocked(&current->saved_sigmask);
2656 }
2657 
2658 static inline sigset_t *sigmask_to_save(void)
2659 {
2660 	sigset_t *res = &current->blocked;
2661 	if (unlikely(test_restore_sigmask()))
2662 		res = &current->saved_sigmask;
2663 	return res;
2664 }
2665 
2666 static inline int kill_cad_pid(int sig, int priv)
2667 {
2668 	return kill_pid(cad_pid, sig, priv);
2669 }
2670 
2671 /* These can be the second arg to send_sig_info/send_group_sig_info.  */
2672 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2673 #define SEND_SIG_PRIV	((struct siginfo *) 1)
2674 #define SEND_SIG_FORCED	((struct siginfo *) 2)
2675 
2676 /*
2677  * True if we are on the alternate signal stack.
2678  */
2679 static inline int on_sig_stack(unsigned long sp)
2680 {
2681 	/*
2682 	 * If the signal stack is SS_AUTODISARM then, by construction, we
2683 	 * can't be on the signal stack unless user code deliberately set
2684 	 * SS_AUTODISARM when we were already on it.
2685 	 *
2686 	 * This improves reliability: if user state gets corrupted such that
2687 	 * the stack pointer points very close to the end of the signal stack,
2688 	 * then this check will enable the signal to be handled anyway.
2689 	 */
2690 	if (current->sas_ss_flags & SS_AUTODISARM)
2691 		return 0;
2692 
2693 #ifdef CONFIG_STACK_GROWSUP
2694 	return sp >= current->sas_ss_sp &&
2695 		sp - current->sas_ss_sp < current->sas_ss_size;
2696 #else
2697 	return sp > current->sas_ss_sp &&
2698 		sp - current->sas_ss_sp <= current->sas_ss_size;
2699 #endif
2700 }
2701 
2702 static inline int sas_ss_flags(unsigned long sp)
2703 {
2704 	if (!current->sas_ss_size)
2705 		return SS_DISABLE;
2706 
2707 	return on_sig_stack(sp) ? SS_ONSTACK : 0;
2708 }
2709 
2710 static inline void sas_ss_reset(struct task_struct *p)
2711 {
2712 	p->sas_ss_sp = 0;
2713 	p->sas_ss_size = 0;
2714 	p->sas_ss_flags = SS_DISABLE;
2715 }
2716 
2717 static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2718 {
2719 	if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2720 #ifdef CONFIG_STACK_GROWSUP
2721 		return current->sas_ss_sp;
2722 #else
2723 		return current->sas_ss_sp + current->sas_ss_size;
2724 #endif
2725 	return sp;
2726 }
2727 
2728 /*
2729  * Routines for handling mm_structs
2730  */
2731 extern struct mm_struct * mm_alloc(void);
2732 
2733 /* mmdrop drops the mm and the page tables */
2734 extern void __mmdrop(struct mm_struct *);
2735 static inline void mmdrop(struct mm_struct *mm)
2736 {
2737 	if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2738 		__mmdrop(mm);
2739 }
2740 
2741 static inline bool mmget_not_zero(struct mm_struct *mm)
2742 {
2743 	return atomic_inc_not_zero(&mm->mm_users);
2744 }
2745 
2746 /* mmput gets rid of the mappings and all user-space */
2747 extern void mmput(struct mm_struct *);
2748 #ifdef CONFIG_MMU
2749 /* same as above but performs the slow path from the async context. Can
2750  * be called from the atomic context as well
2751  */
2752 extern void mmput_async(struct mm_struct *);
2753 #endif
2754 
2755 /* Grab a reference to a task's mm, if it is not already going away */
2756 extern struct mm_struct *get_task_mm(struct task_struct *task);
2757 /*
2758  * Grab a reference to a task's mm, if it is not already going away
2759  * and ptrace_may_access with the mode parameter passed to it
2760  * succeeds.
2761  */
2762 extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2763 /* Remove the current tasks stale references to the old mm_struct */
2764 extern void mm_release(struct task_struct *, struct mm_struct *);
2765 
2766 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
2767 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2768 			struct task_struct *, unsigned long);
2769 #else
2770 extern int copy_thread(unsigned long, unsigned long, unsigned long,
2771 			struct task_struct *);
2772 
2773 /* Architectures that haven't opted into copy_thread_tls get the tls argument
2774  * via pt_regs, so ignore the tls argument passed via C. */
2775 static inline int copy_thread_tls(
2776 		unsigned long clone_flags, unsigned long sp, unsigned long arg,
2777 		struct task_struct *p, unsigned long tls)
2778 {
2779 	return copy_thread(clone_flags, sp, arg, p);
2780 }
2781 #endif
2782 extern void flush_thread(void);
2783 
2784 #ifdef CONFIG_HAVE_EXIT_THREAD
2785 extern void exit_thread(struct task_struct *tsk);
2786 #else
2787 static inline void exit_thread(struct task_struct *tsk)
2788 {
2789 }
2790 #endif
2791 
2792 extern void exit_files(struct task_struct *);
2793 extern void __cleanup_sighand(struct sighand_struct *);
2794 
2795 extern void exit_itimers(struct signal_struct *);
2796 extern void flush_itimer_signals(void);
2797 
2798 extern void do_group_exit(int);
2799 
2800 extern int do_execve(struct filename *,
2801 		     const char __user * const __user *,
2802 		     const char __user * const __user *);
2803 extern int do_execveat(int, struct filename *,
2804 		       const char __user * const __user *,
2805 		       const char __user * const __user *,
2806 		       int);
2807 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
2808 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
2809 struct task_struct *fork_idle(int);
2810 extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
2811 
2812 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
2813 static inline void set_task_comm(struct task_struct *tsk, const char *from)
2814 {
2815 	__set_task_comm(tsk, from, false);
2816 }
2817 extern char *get_task_comm(char *to, struct task_struct *tsk);
2818 
2819 #ifdef CONFIG_SMP
2820 void scheduler_ipi(void);
2821 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
2822 #else
2823 static inline void scheduler_ipi(void) { }
2824 static inline unsigned long wait_task_inactive(struct task_struct *p,
2825 					       long match_state)
2826 {
2827 	return 1;
2828 }
2829 #endif
2830 
2831 #define tasklist_empty() \
2832 	list_empty(&init_task.tasks)
2833 
2834 #define next_task(p) \
2835 	list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
2836 
2837 #define for_each_process(p) \
2838 	for (p = &init_task ; (p = next_task(p)) != &init_task ; )
2839 
2840 extern bool current_is_single_threaded(void);
2841 
2842 /*
2843  * Careful: do_each_thread/while_each_thread is a double loop so
2844  *          'break' will not work as expected - use goto instead.
2845  */
2846 #define do_each_thread(g, t) \
2847 	for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
2848 
2849 #define while_each_thread(g, t) \
2850 	while ((t = next_thread(t)) != g)
2851 
2852 #define __for_each_thread(signal, t)	\
2853 	list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
2854 
2855 #define for_each_thread(p, t)		\
2856 	__for_each_thread((p)->signal, t)
2857 
2858 /* Careful: this is a double loop, 'break' won't work as expected. */
2859 #define for_each_process_thread(p, t)	\
2860 	for_each_process(p) for_each_thread(p, t)
2861 
2862 static inline int get_nr_threads(struct task_struct *tsk)
2863 {
2864 	return tsk->signal->nr_threads;
2865 }
2866 
2867 static inline bool thread_group_leader(struct task_struct *p)
2868 {
2869 	return p->exit_signal >= 0;
2870 }
2871 
2872 /* Do to the insanities of de_thread it is possible for a process
2873  * to have the pid of the thread group leader without actually being
2874  * the thread group leader.  For iteration through the pids in proc
2875  * all we care about is that we have a task with the appropriate
2876  * pid, we don't actually care if we have the right task.
2877  */
2878 static inline bool has_group_leader_pid(struct task_struct *p)
2879 {
2880 	return task_pid(p) == p->signal->leader_pid;
2881 }
2882 
2883 static inline
2884 bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
2885 {
2886 	return p1->signal == p2->signal;
2887 }
2888 
2889 static inline struct task_struct *next_thread(const struct task_struct *p)
2890 {
2891 	return list_entry_rcu(p->thread_group.next,
2892 			      struct task_struct, thread_group);
2893 }
2894 
2895 static inline int thread_group_empty(struct task_struct *p)
2896 {
2897 	return list_empty(&p->thread_group);
2898 }
2899 
2900 #define delay_group_leader(p) \
2901 		(thread_group_leader(p) && !thread_group_empty(p))
2902 
2903 /*
2904  * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
2905  * subscriptions and synchronises with wait4().  Also used in procfs.  Also
2906  * pins the final release of task.io_context.  Also protects ->cpuset and
2907  * ->cgroup.subsys[]. And ->vfork_done.
2908  *
2909  * Nests both inside and outside of read_lock(&tasklist_lock).
2910  * It must not be nested with write_lock_irq(&tasklist_lock),
2911  * neither inside nor outside.
2912  */
2913 static inline void task_lock(struct task_struct *p)
2914 {
2915 	spin_lock(&p->alloc_lock);
2916 }
2917 
2918 static inline void task_unlock(struct task_struct *p)
2919 {
2920 	spin_unlock(&p->alloc_lock);
2921 }
2922 
2923 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
2924 							unsigned long *flags);
2925 
2926 static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
2927 						       unsigned long *flags)
2928 {
2929 	struct sighand_struct *ret;
2930 
2931 	ret = __lock_task_sighand(tsk, flags);
2932 	(void)__cond_lock(&tsk->sighand->siglock, ret);
2933 	return ret;
2934 }
2935 
2936 static inline void unlock_task_sighand(struct task_struct *tsk,
2937 						unsigned long *flags)
2938 {
2939 	spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
2940 }
2941 
2942 /**
2943  * threadgroup_change_begin - mark the beginning of changes to a threadgroup
2944  * @tsk: task causing the changes
2945  *
2946  * All operations which modify a threadgroup - a new thread joining the
2947  * group, death of a member thread (the assertion of PF_EXITING) and
2948  * exec(2) dethreading the process and replacing the leader - are wrapped
2949  * by threadgroup_change_{begin|end}().  This is to provide a place which
2950  * subsystems needing threadgroup stability can hook into for
2951  * synchronization.
2952  */
2953 static inline void threadgroup_change_begin(struct task_struct *tsk)
2954 {
2955 	might_sleep();
2956 	cgroup_threadgroup_change_begin(tsk);
2957 }
2958 
2959 /**
2960  * threadgroup_change_end - mark the end of changes to a threadgroup
2961  * @tsk: task causing the changes
2962  *
2963  * See threadgroup_change_begin().
2964  */
2965 static inline void threadgroup_change_end(struct task_struct *tsk)
2966 {
2967 	cgroup_threadgroup_change_end(tsk);
2968 }
2969 
2970 #ifndef __HAVE_THREAD_FUNCTIONS
2971 
2972 #define task_thread_info(task)	((struct thread_info *)(task)->stack)
2973 #define task_stack_page(task)	((task)->stack)
2974 
2975 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
2976 {
2977 	*task_thread_info(p) = *task_thread_info(org);
2978 	task_thread_info(p)->task = p;
2979 }
2980 
2981 /*
2982  * Return the address of the last usable long on the stack.
2983  *
2984  * When the stack grows down, this is just above the thread
2985  * info struct. Going any lower will corrupt the threadinfo.
2986  *
2987  * When the stack grows up, this is the highest address.
2988  * Beyond that position, we corrupt data on the next page.
2989  */
2990 static inline unsigned long *end_of_stack(struct task_struct *p)
2991 {
2992 #ifdef CONFIG_STACK_GROWSUP
2993 	return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
2994 #else
2995 	return (unsigned long *)(task_thread_info(p) + 1);
2996 #endif
2997 }
2998 
2999 #endif
3000 #define task_stack_end_corrupted(task) \
3001 		(*(end_of_stack(task)) != STACK_END_MAGIC)
3002 
3003 static inline int object_is_on_stack(void *obj)
3004 {
3005 	void *stack = task_stack_page(current);
3006 
3007 	return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3008 }
3009 
3010 extern void thread_stack_cache_init(void);
3011 
3012 #ifdef CONFIG_DEBUG_STACK_USAGE
3013 static inline unsigned long stack_not_used(struct task_struct *p)
3014 {
3015 	unsigned long *n = end_of_stack(p);
3016 
3017 	do { 	/* Skip over canary */
3018 # ifdef CONFIG_STACK_GROWSUP
3019 		n--;
3020 # else
3021 		n++;
3022 # endif
3023 	} while (!*n);
3024 
3025 # ifdef CONFIG_STACK_GROWSUP
3026 	return (unsigned long)end_of_stack(p) - (unsigned long)n;
3027 # else
3028 	return (unsigned long)n - (unsigned long)end_of_stack(p);
3029 # endif
3030 }
3031 #endif
3032 extern void set_task_stack_end_magic(struct task_struct *tsk);
3033 
3034 /* set thread flags in other task's structures
3035  * - see asm/thread_info.h for TIF_xxxx flags available
3036  */
3037 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3038 {
3039 	set_ti_thread_flag(task_thread_info(tsk), flag);
3040 }
3041 
3042 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3043 {
3044 	clear_ti_thread_flag(task_thread_info(tsk), flag);
3045 }
3046 
3047 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3048 {
3049 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3050 }
3051 
3052 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3053 {
3054 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3055 }
3056 
3057 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3058 {
3059 	return test_ti_thread_flag(task_thread_info(tsk), flag);
3060 }
3061 
3062 static inline void set_tsk_need_resched(struct task_struct *tsk)
3063 {
3064 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3065 }
3066 
3067 static inline void clear_tsk_need_resched(struct task_struct *tsk)
3068 {
3069 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3070 }
3071 
3072 static inline int test_tsk_need_resched(struct task_struct *tsk)
3073 {
3074 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3075 }
3076 
3077 static inline int restart_syscall(void)
3078 {
3079 	set_tsk_thread_flag(current, TIF_SIGPENDING);
3080 	return -ERESTARTNOINTR;
3081 }
3082 
3083 static inline int signal_pending(struct task_struct *p)
3084 {
3085 	return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3086 }
3087 
3088 static inline int __fatal_signal_pending(struct task_struct *p)
3089 {
3090 	return unlikely(sigismember(&p->pending.signal, SIGKILL));
3091 }
3092 
3093 static inline int fatal_signal_pending(struct task_struct *p)
3094 {
3095 	return signal_pending(p) && __fatal_signal_pending(p);
3096 }
3097 
3098 static inline int signal_pending_state(long state, struct task_struct *p)
3099 {
3100 	if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3101 		return 0;
3102 	if (!signal_pending(p))
3103 		return 0;
3104 
3105 	return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3106 }
3107 
3108 /*
3109  * cond_resched() and cond_resched_lock(): latency reduction via
3110  * explicit rescheduling in places that are safe. The return
3111  * value indicates whether a reschedule was done in fact.
3112  * cond_resched_lock() will drop the spinlock before scheduling,
3113  * cond_resched_softirq() will enable bhs before scheduling.
3114  */
3115 extern int _cond_resched(void);
3116 
3117 #define cond_resched() ({			\
3118 	___might_sleep(__FILE__, __LINE__, 0);	\
3119 	_cond_resched();			\
3120 })
3121 
3122 extern int __cond_resched_lock(spinlock_t *lock);
3123 
3124 #define cond_resched_lock(lock) ({				\
3125 	___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3126 	__cond_resched_lock(lock);				\
3127 })
3128 
3129 extern int __cond_resched_softirq(void);
3130 
3131 #define cond_resched_softirq() ({					\
3132 	___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET);	\
3133 	__cond_resched_softirq();					\
3134 })
3135 
3136 static inline void cond_resched_rcu(void)
3137 {
3138 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3139 	rcu_read_unlock();
3140 	cond_resched();
3141 	rcu_read_lock();
3142 #endif
3143 }
3144 
3145 /*
3146  * Does a critical section need to be broken due to another
3147  * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3148  * but a general need for low latency)
3149  */
3150 static inline int spin_needbreak(spinlock_t *lock)
3151 {
3152 #ifdef CONFIG_PREEMPT
3153 	return spin_is_contended(lock);
3154 #else
3155 	return 0;
3156 #endif
3157 }
3158 
3159 /*
3160  * Idle thread specific functions to determine the need_resched
3161  * polling state.
3162  */
3163 #ifdef TIF_POLLING_NRFLAG
3164 static inline int tsk_is_polling(struct task_struct *p)
3165 {
3166 	return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3167 }
3168 
3169 static inline void __current_set_polling(void)
3170 {
3171 	set_thread_flag(TIF_POLLING_NRFLAG);
3172 }
3173 
3174 static inline bool __must_check current_set_polling_and_test(void)
3175 {
3176 	__current_set_polling();
3177 
3178 	/*
3179 	 * Polling state must be visible before we test NEED_RESCHED,
3180 	 * paired by resched_curr()
3181 	 */
3182 	smp_mb__after_atomic();
3183 
3184 	return unlikely(tif_need_resched());
3185 }
3186 
3187 static inline void __current_clr_polling(void)
3188 {
3189 	clear_thread_flag(TIF_POLLING_NRFLAG);
3190 }
3191 
3192 static inline bool __must_check current_clr_polling_and_test(void)
3193 {
3194 	__current_clr_polling();
3195 
3196 	/*
3197 	 * Polling state must be visible before we test NEED_RESCHED,
3198 	 * paired by resched_curr()
3199 	 */
3200 	smp_mb__after_atomic();
3201 
3202 	return unlikely(tif_need_resched());
3203 }
3204 
3205 #else
3206 static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3207 static inline void __current_set_polling(void) { }
3208 static inline void __current_clr_polling(void) { }
3209 
3210 static inline bool __must_check current_set_polling_and_test(void)
3211 {
3212 	return unlikely(tif_need_resched());
3213 }
3214 static inline bool __must_check current_clr_polling_and_test(void)
3215 {
3216 	return unlikely(tif_need_resched());
3217 }
3218 #endif
3219 
3220 static inline void current_clr_polling(void)
3221 {
3222 	__current_clr_polling();
3223 
3224 	/*
3225 	 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3226 	 * Once the bit is cleared, we'll get IPIs with every new
3227 	 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3228 	 * fold.
3229 	 */
3230 	smp_mb(); /* paired with resched_curr() */
3231 
3232 	preempt_fold_need_resched();
3233 }
3234 
3235 static __always_inline bool need_resched(void)
3236 {
3237 	return unlikely(tif_need_resched());
3238 }
3239 
3240 /*
3241  * Thread group CPU time accounting.
3242  */
3243 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3244 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3245 
3246 /*
3247  * Reevaluate whether the task has signals pending delivery.
3248  * Wake the task if so.
3249  * This is required every time the blocked sigset_t changes.
3250  * callers must hold sighand->siglock.
3251  */
3252 extern void recalc_sigpending_and_wake(struct task_struct *t);
3253 extern void recalc_sigpending(void);
3254 
3255 extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3256 
3257 static inline void signal_wake_up(struct task_struct *t, bool resume)
3258 {
3259 	signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3260 }
3261 static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3262 {
3263 	signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3264 }
3265 
3266 /*
3267  * Wrappers for p->thread_info->cpu access. No-op on UP.
3268  */
3269 #ifdef CONFIG_SMP
3270 
3271 static inline unsigned int task_cpu(const struct task_struct *p)
3272 {
3273 	return task_thread_info(p)->cpu;
3274 }
3275 
3276 static inline int task_node(const struct task_struct *p)
3277 {
3278 	return cpu_to_node(task_cpu(p));
3279 }
3280 
3281 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3282 
3283 #else
3284 
3285 static inline unsigned int task_cpu(const struct task_struct *p)
3286 {
3287 	return 0;
3288 }
3289 
3290 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3291 {
3292 }
3293 
3294 #endif /* CONFIG_SMP */
3295 
3296 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3297 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3298 
3299 #ifdef CONFIG_CGROUP_SCHED
3300 extern struct task_group root_task_group;
3301 #endif /* CONFIG_CGROUP_SCHED */
3302 
3303 extern int task_can_switch_user(struct user_struct *up,
3304 					struct task_struct *tsk);
3305 
3306 #ifdef CONFIG_TASK_XACCT
3307 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3308 {
3309 	tsk->ioac.rchar += amt;
3310 }
3311 
3312 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3313 {
3314 	tsk->ioac.wchar += amt;
3315 }
3316 
3317 static inline void inc_syscr(struct task_struct *tsk)
3318 {
3319 	tsk->ioac.syscr++;
3320 }
3321 
3322 static inline void inc_syscw(struct task_struct *tsk)
3323 {
3324 	tsk->ioac.syscw++;
3325 }
3326 #else
3327 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3328 {
3329 }
3330 
3331 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3332 {
3333 }
3334 
3335 static inline void inc_syscr(struct task_struct *tsk)
3336 {
3337 }
3338 
3339 static inline void inc_syscw(struct task_struct *tsk)
3340 {
3341 }
3342 #endif
3343 
3344 #ifndef TASK_SIZE_OF
3345 #define TASK_SIZE_OF(tsk)	TASK_SIZE
3346 #endif
3347 
3348 #ifdef CONFIG_MEMCG
3349 extern void mm_update_next_owner(struct mm_struct *mm);
3350 #else
3351 static inline void mm_update_next_owner(struct mm_struct *mm)
3352 {
3353 }
3354 #endif /* CONFIG_MEMCG */
3355 
3356 static inline unsigned long task_rlimit(const struct task_struct *tsk,
3357 		unsigned int limit)
3358 {
3359 	return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3360 }
3361 
3362 static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3363 		unsigned int limit)
3364 {
3365 	return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3366 }
3367 
3368 static inline unsigned long rlimit(unsigned int limit)
3369 {
3370 	return task_rlimit(current, limit);
3371 }
3372 
3373 static inline unsigned long rlimit_max(unsigned int limit)
3374 {
3375 	return task_rlimit_max(current, limit);
3376 }
3377 
3378 #ifdef CONFIG_CPU_FREQ
3379 struct update_util_data {
3380 	void (*func)(struct update_util_data *data,
3381 		     u64 time, unsigned long util, unsigned long max);
3382 };
3383 
3384 void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3385 			void (*func)(struct update_util_data *data, u64 time,
3386 				     unsigned long util, unsigned long max));
3387 void cpufreq_remove_update_util_hook(int cpu);
3388 #endif /* CONFIG_CPU_FREQ */
3389 
3390 #endif
3391