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