]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blob - include/linux/sched.h
sched/core: Convert ___assert_task_state() link time assert to BUILD_BUG_ON()
[mirror_ubuntu-hirsute-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
33 #include <linux/smp.h>
34 #include <linux/sem.h>
35 #include <linux/shm.h>
36 #include <linux/signal.h>
37 #include <linux/compiler.h>
38 #include <linux/completion.h>
39 #include <linux/pid.h>
40 #include <linux/percpu.h>
41 #include <linux/topology.h>
42 #include <linux/seccomp.h>
43 #include <linux/rcupdate.h>
44 #include <linux/rculist.h>
45 #include <linux/rtmutex.h>
46
47 #include <linux/time.h>
48 #include <linux/param.h>
49 #include <linux/resource.h>
50 #include <linux/timer.h>
51 #include <linux/hrtimer.h>
52 #include <linux/kcov.h>
53 #include <linux/task_io_accounting.h>
54 #include <linux/latencytop.h>
55 #include <linux/cred.h>
56 #include <linux/llist.h>
57 #include <linux/uidgid.h>
58 #include <linux/gfp.h>
59 #include <linux/magic.h>
60 #include <linux/cgroup-defs.h>
61
62 #include <asm/processor.h>
63
64 #define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
65
66 /*
67 * Extended scheduling parameters data structure.
68 *
69 * This is needed because the original struct sched_param can not be
70 * altered without introducing ABI issues with legacy applications
71 * (e.g., in sched_getparam()).
72 *
73 * However, the possibility of specifying more than just a priority for
74 * the tasks may be useful for a wide variety of application fields, e.g.,
75 * multimedia, streaming, automation and control, and many others.
76 *
77 * This variant (sched_attr) is meant at describing a so-called
78 * sporadic time-constrained task. In such model a task is specified by:
79 * - the activation period or minimum instance inter-arrival time;
80 * - the maximum (or average, depending on the actual scheduling
81 * discipline) computation time of all instances, a.k.a. runtime;
82 * - the deadline (relative to the actual activation time) of each
83 * instance.
84 * Very briefly, a periodic (sporadic) task asks for the execution of
85 * some specific computation --which is typically called an instance--
86 * (at most) every period. Moreover, each instance typically lasts no more
87 * than the runtime and must be completed by time instant t equal to
88 * the instance activation time + the deadline.
89 *
90 * This is reflected by the actual fields of the sched_attr structure:
91 *
92 * @size size of the structure, for fwd/bwd compat.
93 *
94 * @sched_policy task's scheduling policy
95 * @sched_flags for customizing the scheduler behaviour
96 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
97 * @sched_priority task's static priority (SCHED_FIFO/RR)
98 * @sched_deadline representative of the task's deadline
99 * @sched_runtime representative of the task's runtime
100 * @sched_period representative of the task's period
101 *
102 * Given this task model, there are a multiplicity of scheduling algorithms
103 * and policies, that can be used to ensure all the tasks will make their
104 * timing constraints.
105 *
106 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
107 * only user of this new interface. More information about the algorithm
108 * available in the scheduling class file or in Documentation/.
109 */
110 struct sched_attr {
111 u32 size;
112
113 u32 sched_policy;
114 u64 sched_flags;
115
116 /* SCHED_NORMAL, SCHED_BATCH */
117 s32 sched_nice;
118
119 /* SCHED_FIFO, SCHED_RR */
120 u32 sched_priority;
121
122 /* SCHED_DEADLINE */
123 u64 sched_runtime;
124 u64 sched_deadline;
125 u64 sched_period;
126 };
127
128 struct futex_pi_state;
129 struct robust_list_head;
130 struct bio_list;
131 struct fs_struct;
132 struct perf_event_context;
133 struct blk_plug;
134 struct filename;
135 struct nameidata;
136
137 #define VMACACHE_BITS 2
138 #define VMACACHE_SIZE (1U << VMACACHE_BITS)
139 #define VMACACHE_MASK (VMACACHE_SIZE - 1)
140
141 /*
142 * These are the constant used to fake the fixed-point load-average
143 * counting. Some notes:
144 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
145 * a load-average precision of 10 bits integer + 11 bits fractional
146 * - if you want to count load-averages more often, you need more
147 * precision, or rounding will get you. With 2-second counting freq,
148 * the EXP_n values would be 1981, 2034 and 2043 if still using only
149 * 11 bit fractions.
150 */
151 extern unsigned long avenrun[]; /* Load averages */
152 extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
153
154 #define FSHIFT 11 /* nr of bits of precision */
155 #define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
156 #define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
157 #define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
158 #define EXP_5 2014 /* 1/exp(5sec/5min) */
159 #define EXP_15 2037 /* 1/exp(5sec/15min) */
160
161 #define CALC_LOAD(load,exp,n) \
162 load *= exp; \
163 load += n*(FIXED_1-exp); \
164 load >>= FSHIFT;
165
166 extern unsigned long total_forks;
167 extern int nr_threads;
168 DECLARE_PER_CPU(unsigned long, process_counts);
169 extern int nr_processes(void);
170 extern unsigned long nr_running(void);
171 extern bool single_task_running(void);
172 extern unsigned long nr_iowait(void);
173 extern unsigned long nr_iowait_cpu(int cpu);
174 extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
175
176 extern void calc_global_load(unsigned long ticks);
177
178 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
179 extern void cpu_load_update_nohz_start(void);
180 extern void cpu_load_update_nohz_stop(void);
181 #else
182 static inline void cpu_load_update_nohz_start(void) { }
183 static inline void cpu_load_update_nohz_stop(void) { }
184 #endif
185
186 extern void dump_cpu_task(int cpu);
187
188 struct seq_file;
189 struct cfs_rq;
190 struct task_group;
191 #ifdef CONFIG_SCHED_DEBUG
192 extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
193 extern void proc_sched_set_task(struct task_struct *p);
194 #endif
195
196 /*
197 * Task state bitmask. NOTE! These bits are also
198 * encoded in fs/proc/array.c: get_task_state().
199 *
200 * We have two separate sets of flags: task->state
201 * is about runnability, while task->exit_state are
202 * about the task exiting. Confusing, but this way
203 * modifying one set can't modify the other one by
204 * mistake.
205 */
206 #define TASK_RUNNING 0
207 #define TASK_INTERRUPTIBLE 1
208 #define TASK_UNINTERRUPTIBLE 2
209 #define __TASK_STOPPED 4
210 #define __TASK_TRACED 8
211 /* in tsk->exit_state */
212 #define EXIT_DEAD 16
213 #define EXIT_ZOMBIE 32
214 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
215 /* in tsk->state again */
216 #define TASK_DEAD 64
217 #define TASK_WAKEKILL 128
218 #define TASK_WAKING 256
219 #define TASK_PARKED 512
220 #define TASK_NOLOAD 1024
221 #define TASK_NEW 2048
222 #define TASK_STATE_MAX 4096
223
224 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
225
226 /* Convenience macros for the sake of set_current_state */
227 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
228 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
229 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
230
231 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
232
233 /* Convenience macros for the sake of wake_up */
234 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
235 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
236
237 /* get_task_state() */
238 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
239 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
240 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
241
242 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
243 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
244 #define task_is_stopped_or_traced(task) \
245 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
246 #define task_contributes_to_load(task) \
247 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
248 (task->flags & PF_FROZEN) == 0 && \
249 (task->state & TASK_NOLOAD) == 0)
250
251 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
252
253 #define __set_current_state(state_value) \
254 do { \
255 current->task_state_change = _THIS_IP_; \
256 current->state = (state_value); \
257 } while (0)
258 #define set_current_state(state_value) \
259 do { \
260 current->task_state_change = _THIS_IP_; \
261 smp_store_mb(current->state, (state_value)); \
262 } while (0)
263
264 #else
265 /*
266 * set_current_state() includes a barrier so that the write of current->state
267 * is correctly serialised wrt the caller's subsequent test of whether to
268 * actually sleep:
269 *
270 * for (;;) {
271 * set_current_state(TASK_UNINTERRUPTIBLE);
272 * if (!need_sleep)
273 * break;
274 *
275 * schedule();
276 * }
277 * __set_current_state(TASK_RUNNING);
278 *
279 * If the caller does not need such serialisation (because, for instance, the
280 * condition test and condition change and wakeup are under the same lock) then
281 * use __set_current_state().
282 *
283 * The above is typically ordered against the wakeup, which does:
284 *
285 * need_sleep = false;
286 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
287 *
288 * Where wake_up_state() (and all other wakeup primitives) imply enough
289 * barriers to order the store of the variable against wakeup.
290 *
291 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
292 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
293 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
294 *
295 * This is obviously fine, since they both store the exact same value.
296 *
297 * Also see the comments of try_to_wake_up().
298 */
299 #define __set_current_state(state_value) \
300 do { current->state = (state_value); } while (0)
301 #define set_current_state(state_value) \
302 smp_store_mb(current->state, (state_value))
303
304 #endif
305
306 /* Task command name length */
307 #define TASK_COMM_LEN 16
308
309 #include <linux/spinlock.h>
310
311 /*
312 * This serializes "schedule()" and also protects
313 * the run-queue from deletions/modifications (but
314 * _adding_ to the beginning of the run-queue has
315 * a separate lock).
316 */
317 extern rwlock_t tasklist_lock;
318 extern spinlock_t mmlist_lock;
319
320 struct task_struct;
321
322 #ifdef CONFIG_PROVE_RCU
323 extern int lockdep_tasklist_lock_is_held(void);
324 #endif /* #ifdef CONFIG_PROVE_RCU */
325
326 extern void sched_init(void);
327 extern void sched_init_smp(void);
328 extern asmlinkage void schedule_tail(struct task_struct *prev);
329 extern void init_idle(struct task_struct *idle, int cpu);
330 extern void init_idle_bootup_task(struct task_struct *idle);
331
332 extern cpumask_var_t cpu_isolated_map;
333
334 extern int runqueue_is_locked(int cpu);
335
336 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
337 extern void nohz_balance_enter_idle(int cpu);
338 extern void set_cpu_sd_state_idle(void);
339 extern int get_nohz_timer_target(void);
340 #else
341 static inline void nohz_balance_enter_idle(int cpu) { }
342 static inline void set_cpu_sd_state_idle(void) { }
343 #endif
344
345 /*
346 * Only dump TASK_* tasks. (0 for all tasks)
347 */
348 extern void show_state_filter(unsigned long state_filter);
349
350 static inline void show_state(void)
351 {
352 show_state_filter(0);
353 }
354
355 extern void show_regs(struct pt_regs *);
356
357 /*
358 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
359 * task), SP is the stack pointer of the first frame that should be shown in the back
360 * trace (or NULL if the entire call-chain of the task should be shown).
361 */
362 extern void show_stack(struct task_struct *task, unsigned long *sp);
363
364 extern void cpu_init (void);
365 extern void trap_init(void);
366 extern void update_process_times(int user);
367 extern void scheduler_tick(void);
368 extern int sched_cpu_starting(unsigned int cpu);
369 extern int sched_cpu_activate(unsigned int cpu);
370 extern int sched_cpu_deactivate(unsigned int cpu);
371
372 #ifdef CONFIG_HOTPLUG_CPU
373 extern int sched_cpu_dying(unsigned int cpu);
374 #else
375 # define sched_cpu_dying NULL
376 #endif
377
378 extern void sched_show_task(struct task_struct *p);
379
380 #ifdef CONFIG_LOCKUP_DETECTOR
381 extern void touch_softlockup_watchdog_sched(void);
382 extern void touch_softlockup_watchdog(void);
383 extern void touch_softlockup_watchdog_sync(void);
384 extern void touch_all_softlockup_watchdogs(void);
385 extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
386 void __user *buffer,
387 size_t *lenp, loff_t *ppos);
388 extern unsigned int softlockup_panic;
389 extern unsigned int hardlockup_panic;
390 void lockup_detector_init(void);
391 #else
392 static inline void touch_softlockup_watchdog_sched(void)
393 {
394 }
395 static inline void touch_softlockup_watchdog(void)
396 {
397 }
398 static inline void touch_softlockup_watchdog_sync(void)
399 {
400 }
401 static inline void touch_all_softlockup_watchdogs(void)
402 {
403 }
404 static inline void lockup_detector_init(void)
405 {
406 }
407 #endif
408
409 #ifdef CONFIG_DETECT_HUNG_TASK
410 void reset_hung_task_detector(void);
411 #else
412 static inline void reset_hung_task_detector(void)
413 {
414 }
415 #endif
416
417 /* Attach to any functions which should be ignored in wchan output. */
418 #define __sched __attribute__((__section__(".sched.text")))
419
420 /* Linker adds these: start and end of __sched functions */
421 extern char __sched_text_start[], __sched_text_end[];
422
423 /* Is this address in the __sched functions? */
424 extern int in_sched_functions(unsigned long addr);
425
426 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
427 extern signed long schedule_timeout(signed long timeout);
428 extern signed long schedule_timeout_interruptible(signed long timeout);
429 extern signed long schedule_timeout_killable(signed long timeout);
430 extern signed long schedule_timeout_uninterruptible(signed long timeout);
431 extern signed long schedule_timeout_idle(signed long timeout);
432 asmlinkage void schedule(void);
433 extern void schedule_preempt_disabled(void);
434
435 extern int __must_check io_schedule_prepare(void);
436 extern void io_schedule_finish(int token);
437 extern long io_schedule_timeout(long timeout);
438 extern void io_schedule(void);
439
440 void __noreturn do_task_dead(void);
441
442 struct nsproxy;
443 struct user_namespace;
444
445 #ifdef CONFIG_MMU
446 extern void arch_pick_mmap_layout(struct mm_struct *mm);
447 extern unsigned long
448 arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
449 unsigned long, unsigned long);
450 extern unsigned long
451 arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
452 unsigned long len, unsigned long pgoff,
453 unsigned long flags);
454 #else
455 static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
456 #endif
457
458 #define SUID_DUMP_DISABLE 0 /* No setuid dumping */
459 #define SUID_DUMP_USER 1 /* Dump as user of process */
460 #define SUID_DUMP_ROOT 2 /* Dump as root */
461
462 /* mm flags */
463
464 /* for SUID_DUMP_* above */
465 #define MMF_DUMPABLE_BITS 2
466 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
467
468 extern void set_dumpable(struct mm_struct *mm, int value);
469 /*
470 * This returns the actual value of the suid_dumpable flag. For things
471 * that are using this for checking for privilege transitions, it must
472 * test against SUID_DUMP_USER rather than treating it as a boolean
473 * value.
474 */
475 static inline int __get_dumpable(unsigned long mm_flags)
476 {
477 return mm_flags & MMF_DUMPABLE_MASK;
478 }
479
480 static inline int get_dumpable(struct mm_struct *mm)
481 {
482 return __get_dumpable(mm->flags);
483 }
484
485 /* coredump filter bits */
486 #define MMF_DUMP_ANON_PRIVATE 2
487 #define MMF_DUMP_ANON_SHARED 3
488 #define MMF_DUMP_MAPPED_PRIVATE 4
489 #define MMF_DUMP_MAPPED_SHARED 5
490 #define MMF_DUMP_ELF_HEADERS 6
491 #define MMF_DUMP_HUGETLB_PRIVATE 7
492 #define MMF_DUMP_HUGETLB_SHARED 8
493 #define MMF_DUMP_DAX_PRIVATE 9
494 #define MMF_DUMP_DAX_SHARED 10
495
496 #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
497 #define MMF_DUMP_FILTER_BITS 9
498 #define MMF_DUMP_FILTER_MASK \
499 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
500 #define MMF_DUMP_FILTER_DEFAULT \
501 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
502 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
503
504 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
505 # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
506 #else
507 # define MMF_DUMP_MASK_DEFAULT_ELF 0
508 #endif
509 /* leave room for more dump flags */
510 #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
511 #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
512 /*
513 * This one-shot flag is dropped due to necessity of changing exe once again
514 * on NFS restore
515 */
516 //#define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
517
518 #define MMF_HAS_UPROBES 19 /* has uprobes */
519 #define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
520 #define MMF_OOM_SKIP 21 /* mm is of no interest for the OOM killer */
521 #define MMF_UNSTABLE 22 /* mm is unstable for copy_from_user */
522 #define MMF_HUGE_ZERO_PAGE 23 /* mm has ever used the global huge zero page */
523
524 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
525
526 struct sighand_struct {
527 atomic_t count;
528 struct k_sigaction action[_NSIG];
529 spinlock_t siglock;
530 wait_queue_head_t signalfd_wqh;
531 };
532
533 struct pacct_struct {
534 int ac_flag;
535 long ac_exitcode;
536 unsigned long ac_mem;
537 u64 ac_utime, ac_stime;
538 unsigned long ac_minflt, ac_majflt;
539 };
540
541 struct cpu_itimer {
542 u64 expires;
543 u64 incr;
544 };
545
546 /**
547 * struct prev_cputime - snaphsot of system and user cputime
548 * @utime: time spent in user mode
549 * @stime: time spent in system mode
550 * @lock: protects the above two fields
551 *
552 * Stores previous user/system time values such that we can guarantee
553 * monotonicity.
554 */
555 struct prev_cputime {
556 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
557 u64 utime;
558 u64 stime;
559 raw_spinlock_t lock;
560 #endif
561 };
562
563 static inline void prev_cputime_init(struct prev_cputime *prev)
564 {
565 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
566 prev->utime = prev->stime = 0;
567 raw_spin_lock_init(&prev->lock);
568 #endif
569 }
570
571 /**
572 * struct task_cputime - collected CPU time counts
573 * @utime: time spent in user mode, in nanoseconds
574 * @stime: time spent in kernel mode, in nanoseconds
575 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
576 *
577 * This structure groups together three kinds of CPU time that are tracked for
578 * threads and thread groups. Most things considering CPU time want to group
579 * these counts together and treat all three of them in parallel.
580 */
581 struct task_cputime {
582 u64 utime;
583 u64 stime;
584 unsigned long long sum_exec_runtime;
585 };
586
587 /* Alternate field names when used to cache expirations. */
588 #define virt_exp utime
589 #define prof_exp stime
590 #define sched_exp sum_exec_runtime
591
592 /*
593 * This is the atomic variant of task_cputime, which can be used for
594 * storing and updating task_cputime statistics without locking.
595 */
596 struct task_cputime_atomic {
597 atomic64_t utime;
598 atomic64_t stime;
599 atomic64_t sum_exec_runtime;
600 };
601
602 #define INIT_CPUTIME_ATOMIC \
603 (struct task_cputime_atomic) { \
604 .utime = ATOMIC64_INIT(0), \
605 .stime = ATOMIC64_INIT(0), \
606 .sum_exec_runtime = ATOMIC64_INIT(0), \
607 }
608
609 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
610
611 /*
612 * Disable preemption until the scheduler is running -- use an unconditional
613 * value so that it also works on !PREEMPT_COUNT kernels.
614 *
615 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
616 */
617 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
618
619 /*
620 * Initial preempt_count value; reflects the preempt_count schedule invariant
621 * which states that during context switches:
622 *
623 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
624 *
625 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
626 * Note: See finish_task_switch().
627 */
628 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
629
630 /**
631 * struct thread_group_cputimer - thread group interval timer counts
632 * @cputime_atomic: atomic thread group interval timers.
633 * @running: true when there are timers running and
634 * @cputime_atomic receives updates.
635 * @checking_timer: true when a thread in the group is in the
636 * process of checking for thread group timers.
637 *
638 * This structure contains the version of task_cputime, above, that is
639 * used for thread group CPU timer calculations.
640 */
641 struct thread_group_cputimer {
642 struct task_cputime_atomic cputime_atomic;
643 bool running;
644 bool checking_timer;
645 };
646
647 #include <linux/rwsem.h>
648 struct autogroup;
649
650 /*
651 * NOTE! "signal_struct" does not have its own
652 * locking, because a shared signal_struct always
653 * implies a shared sighand_struct, so locking
654 * sighand_struct is always a proper superset of
655 * the locking of signal_struct.
656 */
657 struct signal_struct {
658 atomic_t sigcnt;
659 atomic_t live;
660 int nr_threads;
661 struct list_head thread_head;
662
663 wait_queue_head_t wait_chldexit; /* for wait4() */
664
665 /* current thread group signal load-balancing target: */
666 struct task_struct *curr_target;
667
668 /* shared signal handling: */
669 struct sigpending shared_pending;
670
671 /* thread group exit support */
672 int group_exit_code;
673 /* overloaded:
674 * - notify group_exit_task when ->count is equal to notify_count
675 * - everyone except group_exit_task is stopped during signal delivery
676 * of fatal signals, group_exit_task processes the signal.
677 */
678 int notify_count;
679 struct task_struct *group_exit_task;
680
681 /* thread group stop support, overloads group_exit_code too */
682 int group_stop_count;
683 unsigned int flags; /* see SIGNAL_* flags below */
684
685 /*
686 * PR_SET_CHILD_SUBREAPER marks a process, like a service
687 * manager, to re-parent orphan (double-forking) child processes
688 * to this process instead of 'init'. The service manager is
689 * able to receive SIGCHLD signals and is able to investigate
690 * the process until it calls wait(). All children of this
691 * process will inherit a flag if they should look for a
692 * child_subreaper process at exit.
693 */
694 unsigned int is_child_subreaper:1;
695 unsigned int has_child_subreaper:1;
696
697 #ifdef CONFIG_POSIX_TIMERS
698
699 /* POSIX.1b Interval Timers */
700 int posix_timer_id;
701 struct list_head posix_timers;
702
703 /* ITIMER_REAL timer for the process */
704 struct hrtimer real_timer;
705 ktime_t it_real_incr;
706
707 /*
708 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
709 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
710 * values are defined to 0 and 1 respectively
711 */
712 struct cpu_itimer it[2];
713
714 /*
715 * Thread group totals for process CPU timers.
716 * See thread_group_cputimer(), et al, for details.
717 */
718 struct thread_group_cputimer cputimer;
719
720 /* Earliest-expiration cache. */
721 struct task_cputime cputime_expires;
722
723 struct list_head cpu_timers[3];
724
725 #endif
726
727 struct pid *leader_pid;
728
729 #ifdef CONFIG_NO_HZ_FULL
730 atomic_t tick_dep_mask;
731 #endif
732
733 struct pid *tty_old_pgrp;
734
735 /* boolean value for session group leader */
736 int leader;
737
738 struct tty_struct *tty; /* NULL if no tty */
739
740 #ifdef CONFIG_SCHED_AUTOGROUP
741 struct autogroup *autogroup;
742 #endif
743 /*
744 * Cumulative resource counters for dead threads in the group,
745 * and for reaped dead child processes forked by this group.
746 * Live threads maintain their own counters and add to these
747 * in __exit_signal, except for the group leader.
748 */
749 seqlock_t stats_lock;
750 u64 utime, stime, cutime, cstime;
751 u64 gtime;
752 u64 cgtime;
753 struct prev_cputime prev_cputime;
754 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
755 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
756 unsigned long inblock, oublock, cinblock, coublock;
757 unsigned long maxrss, cmaxrss;
758 struct task_io_accounting ioac;
759
760 /*
761 * Cumulative ns of schedule CPU time fo dead threads in the
762 * group, not including a zombie group leader, (This only differs
763 * from jiffies_to_ns(utime + stime) if sched_clock uses something
764 * other than jiffies.)
765 */
766 unsigned long long sum_sched_runtime;
767
768 /*
769 * We don't bother to synchronize most readers of this at all,
770 * because there is no reader checking a limit that actually needs
771 * to get both rlim_cur and rlim_max atomically, and either one
772 * alone is a single word that can safely be read normally.
773 * getrlimit/setrlimit use task_lock(current->group_leader) to
774 * protect this instead of the siglock, because they really
775 * have no need to disable irqs.
776 */
777 struct rlimit rlim[RLIM_NLIMITS];
778
779 #ifdef CONFIG_BSD_PROCESS_ACCT
780 struct pacct_struct pacct; /* per-process accounting information */
781 #endif
782 #ifdef CONFIG_TASKSTATS
783 struct taskstats *stats;
784 #endif
785 #ifdef CONFIG_AUDIT
786 unsigned audit_tty;
787 struct tty_audit_buf *tty_audit_buf;
788 #endif
789
790 /*
791 * Thread is the potential origin of an oom condition; kill first on
792 * oom
793 */
794 bool oom_flag_origin;
795 short oom_score_adj; /* OOM kill score adjustment */
796 short oom_score_adj_min; /* OOM kill score adjustment min value.
797 * Only settable by CAP_SYS_RESOURCE. */
798 struct mm_struct *oom_mm; /* recorded mm when the thread group got
799 * killed by the oom killer */
800
801 struct mutex cred_guard_mutex; /* guard against foreign influences on
802 * credential calculations
803 * (notably. ptrace) */
804 };
805
806 /*
807 * Bits in flags field of signal_struct.
808 */
809 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
810 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
811 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
812 #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
813 /*
814 * Pending notifications to parent.
815 */
816 #define SIGNAL_CLD_STOPPED 0x00000010
817 #define SIGNAL_CLD_CONTINUED 0x00000020
818 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
819
820 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
821
822 #define SIGNAL_STOP_MASK (SIGNAL_CLD_MASK | SIGNAL_STOP_STOPPED | \
823 SIGNAL_STOP_CONTINUED)
824
825 static inline void signal_set_stop_flags(struct signal_struct *sig,
826 unsigned int flags)
827 {
828 WARN_ON(sig->flags & (SIGNAL_GROUP_EXIT|SIGNAL_GROUP_COREDUMP));
829 sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags;
830 }
831
832 /* If true, all threads except ->group_exit_task have pending SIGKILL */
833 static inline int signal_group_exit(const struct signal_struct *sig)
834 {
835 return (sig->flags & SIGNAL_GROUP_EXIT) ||
836 (sig->group_exit_task != NULL);
837 }
838
839 /*
840 * Some day this will be a full-fledged user tracking system..
841 */
842 struct user_struct {
843 atomic_t __count; /* reference count */
844 atomic_t processes; /* How many processes does this user have? */
845 atomic_t sigpending; /* How many pending signals does this user have? */
846 #ifdef CONFIG_FANOTIFY
847 atomic_t fanotify_listeners;
848 #endif
849 #ifdef CONFIG_EPOLL
850 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
851 #endif
852 #ifdef CONFIG_POSIX_MQUEUE
853 /* protected by mq_lock */
854 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
855 #endif
856 unsigned long locked_shm; /* How many pages of mlocked shm ? */
857 unsigned long unix_inflight; /* How many files in flight in unix sockets */
858 atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */
859
860 #ifdef CONFIG_KEYS
861 struct key *uid_keyring; /* UID specific keyring */
862 struct key *session_keyring; /* UID's default session keyring */
863 #endif
864
865 /* Hash table maintenance information */
866 struct hlist_node uidhash_node;
867 kuid_t uid;
868
869 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
870 atomic_long_t locked_vm;
871 #endif
872 };
873
874 extern int uids_sysfs_init(void);
875
876 extern struct user_struct *find_user(kuid_t);
877
878 extern struct user_struct root_user;
879 #define INIT_USER (&root_user)
880
881
882 struct backing_dev_info;
883 struct reclaim_state;
884
885 #ifdef CONFIG_SCHED_INFO
886 struct sched_info {
887 /* cumulative counters */
888 unsigned long pcount; /* # of times run on this cpu */
889 unsigned long long run_delay; /* time spent waiting on a runqueue */
890
891 /* timestamps */
892 unsigned long long last_arrival,/* when we last ran on a cpu */
893 last_queued; /* when we were last queued to run */
894 };
895 #endif /* CONFIG_SCHED_INFO */
896
897 #ifdef CONFIG_TASK_DELAY_ACCT
898 struct task_delay_info {
899 spinlock_t lock;
900 unsigned int flags; /* Private per-task flags */
901
902 /* For each stat XXX, add following, aligned appropriately
903 *
904 * struct timespec XXX_start, XXX_end;
905 * u64 XXX_delay;
906 * u32 XXX_count;
907 *
908 * Atomicity of updates to XXX_delay, XXX_count protected by
909 * single lock above (split into XXX_lock if contention is an issue).
910 */
911
912 /*
913 * XXX_count is incremented on every XXX operation, the delay
914 * associated with the operation is added to XXX_delay.
915 * XXX_delay contains the accumulated delay time in nanoseconds.
916 */
917 u64 blkio_start; /* Shared by blkio, swapin */
918 u64 blkio_delay; /* wait for sync block io completion */
919 u64 swapin_delay; /* wait for swapin block io completion */
920 u32 blkio_count; /* total count of the number of sync block */
921 /* io operations performed */
922 u32 swapin_count; /* total count of the number of swapin block */
923 /* io operations performed */
924
925 u64 freepages_start;
926 u64 freepages_delay; /* wait for memory reclaim */
927 u32 freepages_count; /* total count of memory reclaim */
928 };
929 #endif /* CONFIG_TASK_DELAY_ACCT */
930
931 static inline int sched_info_on(void)
932 {
933 #ifdef CONFIG_SCHEDSTATS
934 return 1;
935 #elif defined(CONFIG_TASK_DELAY_ACCT)
936 extern int delayacct_on;
937 return delayacct_on;
938 #else
939 return 0;
940 #endif
941 }
942
943 #ifdef CONFIG_SCHEDSTATS
944 void force_schedstat_enabled(void);
945 #endif
946
947 enum cpu_idle_type {
948 CPU_IDLE,
949 CPU_NOT_IDLE,
950 CPU_NEWLY_IDLE,
951 CPU_MAX_IDLE_TYPES
952 };
953
954 /*
955 * Integer metrics need fixed point arithmetic, e.g., sched/fair
956 * has a few: load, load_avg, util_avg, freq, and capacity.
957 *
958 * We define a basic fixed point arithmetic range, and then formalize
959 * all these metrics based on that basic range.
960 */
961 # define SCHED_FIXEDPOINT_SHIFT 10
962 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
963
964 /*
965 * Increase resolution of cpu_capacity calculations
966 */
967 #define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
968 #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
969
970 /*
971 * Wake-queues are lists of tasks with a pending wakeup, whose
972 * callers have already marked the task as woken internally,
973 * and can thus carry on. A common use case is being able to
974 * do the wakeups once the corresponding user lock as been
975 * released.
976 *
977 * We hold reference to each task in the list across the wakeup,
978 * thus guaranteeing that the memory is still valid by the time
979 * the actual wakeups are performed in wake_up_q().
980 *
981 * One per task suffices, because there's never a need for a task to be
982 * in two wake queues simultaneously; it is forbidden to abandon a task
983 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
984 * already in a wake queue, the wakeup will happen soon and the second
985 * waker can just skip it.
986 *
987 * The DEFINE_WAKE_Q macro declares and initializes the list head.
988 * wake_up_q() does NOT reinitialize the list; it's expected to be
989 * called near the end of a function. Otherwise, the list can be
990 * re-initialized for later re-use by wake_q_init().
991 *
992 * Note that this can cause spurious wakeups. schedule() callers
993 * must ensure the call is done inside a loop, confirming that the
994 * wakeup condition has in fact occurred.
995 */
996 struct wake_q_node {
997 struct wake_q_node *next;
998 };
999
1000 struct wake_q_head {
1001 struct wake_q_node *first;
1002 struct wake_q_node **lastp;
1003 };
1004
1005 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1006
1007 #define DEFINE_WAKE_Q(name) \
1008 struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1009
1010 static inline void wake_q_init(struct wake_q_head *head)
1011 {
1012 head->first = WAKE_Q_TAIL;
1013 head->lastp = &head->first;
1014 }
1015
1016 extern void wake_q_add(struct wake_q_head *head,
1017 struct task_struct *task);
1018 extern void wake_up_q(struct wake_q_head *head);
1019
1020 /*
1021 * sched-domains (multiprocessor balancing) declarations:
1022 */
1023 #ifdef CONFIG_SMP
1024 #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
1025 #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
1026 #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
1027 #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
1028 #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
1029 #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
1030 #define SD_ASYM_CPUCAPACITY 0x0040 /* Groups have different max cpu capacities */
1031 #define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu capacity */
1032 #define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
1033 #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
1034 #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
1035 #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
1036 #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
1037 #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
1038 #define SD_NUMA 0x4000 /* cross-node balancing */
1039
1040 #ifdef CONFIG_SCHED_SMT
1041 static inline int cpu_smt_flags(void)
1042 {
1043 return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1044 }
1045 #endif
1046
1047 #ifdef CONFIG_SCHED_MC
1048 static inline int cpu_core_flags(void)
1049 {
1050 return SD_SHARE_PKG_RESOURCES;
1051 }
1052 #endif
1053
1054 #ifdef CONFIG_NUMA
1055 static inline int cpu_numa_flags(void)
1056 {
1057 return SD_NUMA;
1058 }
1059 #endif
1060
1061 extern int arch_asym_cpu_priority(int cpu);
1062
1063 struct sched_domain_attr {
1064 int relax_domain_level;
1065 };
1066
1067 #define SD_ATTR_INIT (struct sched_domain_attr) { \
1068 .relax_domain_level = -1, \
1069 }
1070
1071 extern int sched_domain_level_max;
1072
1073 struct sched_group;
1074
1075 struct sched_domain_shared {
1076 atomic_t ref;
1077 atomic_t nr_busy_cpus;
1078 int has_idle_cores;
1079 };
1080
1081 struct sched_domain {
1082 /* These fields must be setup */
1083 struct sched_domain *parent; /* top domain must be null terminated */
1084 struct sched_domain *child; /* bottom domain must be null terminated */
1085 struct sched_group *groups; /* the balancing groups of the domain */
1086 unsigned long min_interval; /* Minimum balance interval ms */
1087 unsigned long max_interval; /* Maximum balance interval ms */
1088 unsigned int busy_factor; /* less balancing by factor if busy */
1089 unsigned int imbalance_pct; /* No balance until over watermark */
1090 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */
1091 unsigned int busy_idx;
1092 unsigned int idle_idx;
1093 unsigned int newidle_idx;
1094 unsigned int wake_idx;
1095 unsigned int forkexec_idx;
1096 unsigned int smt_gain;
1097
1098 int nohz_idle; /* NOHZ IDLE status */
1099 int flags; /* See SD_* */
1100 int level;
1101
1102 /* Runtime fields. */
1103 unsigned long last_balance; /* init to jiffies. units in jiffies */
1104 unsigned int balance_interval; /* initialise to 1. units in ms. */
1105 unsigned int nr_balance_failed; /* initialise to 0 */
1106
1107 /* idle_balance() stats */
1108 u64 max_newidle_lb_cost;
1109 unsigned long next_decay_max_lb_cost;
1110
1111 u64 avg_scan_cost; /* select_idle_sibling */
1112
1113 #ifdef CONFIG_SCHEDSTATS
1114 /* load_balance() stats */
1115 unsigned int lb_count[CPU_MAX_IDLE_TYPES];
1116 unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
1117 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
1118 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
1119 unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
1120 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
1121 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
1122 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
1123
1124 /* Active load balancing */
1125 unsigned int alb_count;
1126 unsigned int alb_failed;
1127 unsigned int alb_pushed;
1128
1129 /* SD_BALANCE_EXEC stats */
1130 unsigned int sbe_count;
1131 unsigned int sbe_balanced;
1132 unsigned int sbe_pushed;
1133
1134 /* SD_BALANCE_FORK stats */
1135 unsigned int sbf_count;
1136 unsigned int sbf_balanced;
1137 unsigned int sbf_pushed;
1138
1139 /* try_to_wake_up() stats */
1140 unsigned int ttwu_wake_remote;
1141 unsigned int ttwu_move_affine;
1142 unsigned int ttwu_move_balance;
1143 #endif
1144 #ifdef CONFIG_SCHED_DEBUG
1145 char *name;
1146 #endif
1147 union {
1148 void *private; /* used during construction */
1149 struct rcu_head rcu; /* used during destruction */
1150 };
1151 struct sched_domain_shared *shared;
1152
1153 unsigned int span_weight;
1154 /*
1155 * Span of all CPUs in this domain.
1156 *
1157 * NOTE: this field is variable length. (Allocated dynamically
1158 * by attaching extra space to the end of the structure,
1159 * depending on how many CPUs the kernel has booted up with)
1160 */
1161 unsigned long span[0];
1162 };
1163
1164 static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
1165 {
1166 return to_cpumask(sd->span);
1167 }
1168
1169 extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1170 struct sched_domain_attr *dattr_new);
1171
1172 /* Allocate an array of sched domains, for partition_sched_domains(). */
1173 cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
1174 void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
1175
1176 bool cpus_share_cache(int this_cpu, int that_cpu);
1177
1178 typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1179 typedef int (*sched_domain_flags_f)(void);
1180
1181 #define SDTL_OVERLAP 0x01
1182
1183 struct sd_data {
1184 struct sched_domain **__percpu sd;
1185 struct sched_domain_shared **__percpu sds;
1186 struct sched_group **__percpu sg;
1187 struct sched_group_capacity **__percpu sgc;
1188 };
1189
1190 struct sched_domain_topology_level {
1191 sched_domain_mask_f mask;
1192 sched_domain_flags_f sd_flags;
1193 int flags;
1194 int numa_level;
1195 struct sd_data data;
1196 #ifdef CONFIG_SCHED_DEBUG
1197 char *name;
1198 #endif
1199 };
1200
1201 extern void set_sched_topology(struct sched_domain_topology_level *tl);
1202 extern void wake_up_if_idle(int cpu);
1203
1204 #ifdef CONFIG_SCHED_DEBUG
1205 # define SD_INIT_NAME(type) .name = #type
1206 #else
1207 # define SD_INIT_NAME(type)
1208 #endif
1209
1210 #else /* CONFIG_SMP */
1211
1212 struct sched_domain_attr;
1213
1214 static inline void
1215 partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1216 struct sched_domain_attr *dattr_new)
1217 {
1218 }
1219
1220 static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1221 {
1222 return true;
1223 }
1224
1225 #endif /* !CONFIG_SMP */
1226
1227
1228 struct io_context; /* See blkdev.h */
1229
1230
1231 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1232 extern void prefetch_stack(struct task_struct *t);
1233 #else
1234 static inline void prefetch_stack(struct task_struct *t) { }
1235 #endif
1236
1237 struct audit_context; /* See audit.c */
1238 struct mempolicy;
1239 struct pipe_inode_info;
1240 struct uts_namespace;
1241
1242 struct load_weight {
1243 unsigned long weight;
1244 u32 inv_weight;
1245 };
1246
1247 /*
1248 * The load_avg/util_avg accumulates an infinite geometric series
1249 * (see __update_load_avg() in kernel/sched/fair.c).
1250 *
1251 * [load_avg definition]
1252 *
1253 * load_avg = runnable% * scale_load_down(load)
1254 *
1255 * where runnable% is the time ratio that a sched_entity is runnable.
1256 * For cfs_rq, it is the aggregated load_avg of all runnable and
1257 * blocked sched_entities.
1258 *
1259 * load_avg may also take frequency scaling into account:
1260 *
1261 * load_avg = runnable% * scale_load_down(load) * freq%
1262 *
1263 * where freq% is the CPU frequency normalized to the highest frequency.
1264 *
1265 * [util_avg definition]
1266 *
1267 * util_avg = running% * SCHED_CAPACITY_SCALE
1268 *
1269 * where running% is the time ratio that a sched_entity is running on
1270 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1271 * and blocked sched_entities.
1272 *
1273 * util_avg may also factor frequency scaling and CPU capacity scaling:
1274 *
1275 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1276 *
1277 * where freq% is the same as above, and capacity% is the CPU capacity
1278 * normalized to the greatest capacity (due to uarch differences, etc).
1279 *
1280 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1281 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1282 * we therefore scale them to as large a range as necessary. This is for
1283 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1284 *
1285 * [Overflow issue]
1286 *
1287 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1288 * with the highest load (=88761), always runnable on a single cfs_rq,
1289 * and should not overflow as the number already hits PID_MAX_LIMIT.
1290 *
1291 * For all other cases (including 32-bit kernels), struct load_weight's
1292 * weight will overflow first before we do, because:
1293 *
1294 * Max(load_avg) <= Max(load.weight)
1295 *
1296 * Then it is the load_weight's responsibility to consider overflow
1297 * issues.
1298 */
1299 struct sched_avg {
1300 u64 last_update_time, load_sum;
1301 u32 util_sum, period_contrib;
1302 unsigned long load_avg, util_avg;
1303 };
1304
1305 #ifdef CONFIG_SCHEDSTATS
1306 struct sched_statistics {
1307 u64 wait_start;
1308 u64 wait_max;
1309 u64 wait_count;
1310 u64 wait_sum;
1311 u64 iowait_count;
1312 u64 iowait_sum;
1313
1314 u64 sleep_start;
1315 u64 sleep_max;
1316 s64 sum_sleep_runtime;
1317
1318 u64 block_start;
1319 u64 block_max;
1320 u64 exec_max;
1321 u64 slice_max;
1322
1323 u64 nr_migrations_cold;
1324 u64 nr_failed_migrations_affine;
1325 u64 nr_failed_migrations_running;
1326 u64 nr_failed_migrations_hot;
1327 u64 nr_forced_migrations;
1328
1329 u64 nr_wakeups;
1330 u64 nr_wakeups_sync;
1331 u64 nr_wakeups_migrate;
1332 u64 nr_wakeups_local;
1333 u64 nr_wakeups_remote;
1334 u64 nr_wakeups_affine;
1335 u64 nr_wakeups_affine_attempts;
1336 u64 nr_wakeups_passive;
1337 u64 nr_wakeups_idle;
1338 };
1339 #endif
1340
1341 struct sched_entity {
1342 struct load_weight load; /* for load-balancing */
1343 struct rb_node run_node;
1344 struct list_head group_node;
1345 unsigned int on_rq;
1346
1347 u64 exec_start;
1348 u64 sum_exec_runtime;
1349 u64 vruntime;
1350 u64 prev_sum_exec_runtime;
1351
1352 u64 nr_migrations;
1353
1354 #ifdef CONFIG_SCHEDSTATS
1355 struct sched_statistics statistics;
1356 #endif
1357
1358 #ifdef CONFIG_FAIR_GROUP_SCHED
1359 int depth;
1360 struct sched_entity *parent;
1361 /* rq on which this entity is (to be) queued: */
1362 struct cfs_rq *cfs_rq;
1363 /* rq "owned" by this entity/group: */
1364 struct cfs_rq *my_q;
1365 #endif
1366
1367 #ifdef CONFIG_SMP
1368 /*
1369 * Per entity load average tracking.
1370 *
1371 * Put into separate cache line so it does not
1372 * collide with read-mostly values above.
1373 */
1374 struct sched_avg avg ____cacheline_aligned_in_smp;
1375 #endif
1376 };
1377
1378 struct sched_rt_entity {
1379 struct list_head run_list;
1380 unsigned long timeout;
1381 unsigned long watchdog_stamp;
1382 unsigned int time_slice;
1383 unsigned short on_rq;
1384 unsigned short on_list;
1385
1386 struct sched_rt_entity *back;
1387 #ifdef CONFIG_RT_GROUP_SCHED
1388 struct sched_rt_entity *parent;
1389 /* rq on which this entity is (to be) queued: */
1390 struct rt_rq *rt_rq;
1391 /* rq "owned" by this entity/group: */
1392 struct rt_rq *my_q;
1393 #endif
1394 };
1395
1396 struct sched_dl_entity {
1397 struct rb_node rb_node;
1398
1399 /*
1400 * Original scheduling parameters. Copied here from sched_attr
1401 * during sched_setattr(), they will remain the same until
1402 * the next sched_setattr().
1403 */
1404 u64 dl_runtime; /* maximum runtime for each instance */
1405 u64 dl_deadline; /* relative deadline of each instance */
1406 u64 dl_period; /* separation of two instances (period) */
1407 u64 dl_bw; /* dl_runtime / dl_deadline */
1408
1409 /*
1410 * Actual scheduling parameters. Initialized with the values above,
1411 * they are continously updated during task execution. Note that
1412 * the remaining runtime could be < 0 in case we are in overrun.
1413 */
1414 s64 runtime; /* remaining runtime for this instance */
1415 u64 deadline; /* absolute deadline for this instance */
1416 unsigned int flags; /* specifying the scheduler behaviour */
1417
1418 /*
1419 * Some bool flags:
1420 *
1421 * @dl_throttled tells if we exhausted the runtime. If so, the
1422 * task has to wait for a replenishment to be performed at the
1423 * next firing of dl_timer.
1424 *
1425 * @dl_boosted tells if we are boosted due to DI. If so we are
1426 * outside bandwidth enforcement mechanism (but only until we
1427 * exit the critical section);
1428 *
1429 * @dl_yielded tells if task gave up the cpu before consuming
1430 * all its available runtime during the last job.
1431 */
1432 int dl_throttled, dl_boosted, dl_yielded;
1433
1434 /*
1435 * Bandwidth enforcement timer. Each -deadline task has its
1436 * own bandwidth to be enforced, thus we need one timer per task.
1437 */
1438 struct hrtimer dl_timer;
1439 };
1440
1441 union rcu_special {
1442 struct {
1443 u8 blocked;
1444 u8 need_qs;
1445 u8 exp_need_qs;
1446 u8 pad; /* Otherwise the compiler can store garbage here. */
1447 } b; /* Bits. */
1448 u32 s; /* Set of bits. */
1449 };
1450 struct rcu_node;
1451
1452 enum perf_event_task_context {
1453 perf_invalid_context = -1,
1454 perf_hw_context = 0,
1455 perf_sw_context,
1456 perf_nr_task_contexts,
1457 };
1458
1459 /* Track pages that require TLB flushes */
1460 struct tlbflush_unmap_batch {
1461 /*
1462 * Each bit set is a CPU that potentially has a TLB entry for one of
1463 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1464 */
1465 struct cpumask cpumask;
1466
1467 /* True if any bit in cpumask is set */
1468 bool flush_required;
1469
1470 /*
1471 * If true then the PTE was dirty when unmapped. The entry must be
1472 * flushed before IO is initiated or a stale TLB entry potentially
1473 * allows an update without redirtying the page.
1474 */
1475 bool writable;
1476 };
1477
1478 struct task_struct {
1479 #ifdef CONFIG_THREAD_INFO_IN_TASK
1480 /*
1481 * For reasons of header soup (see current_thread_info()), this
1482 * must be the first element of task_struct.
1483 */
1484 struct thread_info thread_info;
1485 #endif
1486 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1487 void *stack;
1488 atomic_t usage;
1489 unsigned int flags; /* per process flags, defined below */
1490 unsigned int ptrace;
1491
1492 #ifdef CONFIG_SMP
1493 struct llist_node wake_entry;
1494 int on_cpu;
1495 #ifdef CONFIG_THREAD_INFO_IN_TASK
1496 unsigned int cpu; /* current CPU */
1497 #endif
1498 unsigned int wakee_flips;
1499 unsigned long wakee_flip_decay_ts;
1500 struct task_struct *last_wakee;
1501
1502 int wake_cpu;
1503 #endif
1504 int on_rq;
1505
1506 int prio, static_prio, normal_prio;
1507 unsigned int rt_priority;
1508 const struct sched_class *sched_class;
1509 struct sched_entity se;
1510 struct sched_rt_entity rt;
1511 #ifdef CONFIG_CGROUP_SCHED
1512 struct task_group *sched_task_group;
1513 #endif
1514 struct sched_dl_entity dl;
1515
1516 #ifdef CONFIG_PREEMPT_NOTIFIERS
1517 /* list of struct preempt_notifier: */
1518 struct hlist_head preempt_notifiers;
1519 #endif
1520
1521 #ifdef CONFIG_BLK_DEV_IO_TRACE
1522 unsigned int btrace_seq;
1523 #endif
1524
1525 unsigned int policy;
1526 int nr_cpus_allowed;
1527 cpumask_t cpus_allowed;
1528
1529 #ifdef CONFIG_PREEMPT_RCU
1530 int rcu_read_lock_nesting;
1531 union rcu_special rcu_read_unlock_special;
1532 struct list_head rcu_node_entry;
1533 struct rcu_node *rcu_blocked_node;
1534 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1535 #ifdef CONFIG_TASKS_RCU
1536 unsigned long rcu_tasks_nvcsw;
1537 bool rcu_tasks_holdout;
1538 struct list_head rcu_tasks_holdout_list;
1539 int rcu_tasks_idle_cpu;
1540 #endif /* #ifdef CONFIG_TASKS_RCU */
1541
1542 #ifdef CONFIG_SCHED_INFO
1543 struct sched_info sched_info;
1544 #endif
1545
1546 struct list_head tasks;
1547 #ifdef CONFIG_SMP
1548 struct plist_node pushable_tasks;
1549 struct rb_node pushable_dl_tasks;
1550 #endif
1551
1552 struct mm_struct *mm, *active_mm;
1553 /* per-thread vma caching */
1554 u32 vmacache_seqnum;
1555 struct vm_area_struct *vmacache[VMACACHE_SIZE];
1556 #if defined(SPLIT_RSS_COUNTING)
1557 struct task_rss_stat rss_stat;
1558 #endif
1559 /* task state */
1560 int exit_state;
1561 int exit_code, exit_signal;
1562 int pdeath_signal; /* The signal sent when the parent dies */
1563 unsigned long jobctl; /* JOBCTL_*, siglock protected */
1564
1565 /* Used for emulating ABI behavior of previous Linux versions */
1566 unsigned int personality;
1567
1568 /* scheduler bits, serialized by scheduler locks */
1569 unsigned sched_reset_on_fork:1;
1570 unsigned sched_contributes_to_load:1;
1571 unsigned sched_migrated:1;
1572 unsigned sched_remote_wakeup:1;
1573 unsigned :0; /* force alignment to the next boundary */
1574
1575 /* unserialized, strictly 'current' */
1576 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1577 unsigned in_iowait:1;
1578 #if !defined(TIF_RESTORE_SIGMASK)
1579 unsigned restore_sigmask:1;
1580 #endif
1581 #ifdef CONFIG_MEMCG
1582 unsigned memcg_may_oom:1;
1583 #ifndef CONFIG_SLOB
1584 unsigned memcg_kmem_skip_account:1;
1585 #endif
1586 #endif
1587 #ifdef CONFIG_COMPAT_BRK
1588 unsigned brk_randomized:1;
1589 #endif
1590
1591 unsigned long atomic_flags; /* Flags needing atomic access. */
1592
1593 struct restart_block restart_block;
1594
1595 pid_t pid;
1596 pid_t tgid;
1597
1598 #ifdef CONFIG_CC_STACKPROTECTOR
1599 /* Canary value for the -fstack-protector gcc feature */
1600 unsigned long stack_canary;
1601 #endif
1602 /*
1603 * pointers to (original) parent process, youngest child, younger sibling,
1604 * older sibling, respectively. (p->father can be replaced with
1605 * p->real_parent->pid)
1606 */
1607 struct task_struct __rcu *real_parent; /* real parent process */
1608 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1609 /*
1610 * children/sibling forms the list of my natural children
1611 */
1612 struct list_head children; /* list of my children */
1613 struct list_head sibling; /* linkage in my parent's children list */
1614 struct task_struct *group_leader; /* threadgroup leader */
1615
1616 /*
1617 * ptraced is the list of tasks this task is using ptrace on.
1618 * This includes both natural children and PTRACE_ATTACH targets.
1619 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1620 */
1621 struct list_head ptraced;
1622 struct list_head ptrace_entry;
1623
1624 /* PID/PID hash table linkage. */
1625 struct pid_link pids[PIDTYPE_MAX];
1626 struct list_head thread_group;
1627 struct list_head thread_node;
1628
1629 struct completion *vfork_done; /* for vfork() */
1630 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1631 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1632
1633 u64 utime, stime;
1634 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1635 u64 utimescaled, stimescaled;
1636 #endif
1637 u64 gtime;
1638 struct prev_cputime prev_cputime;
1639 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1640 seqcount_t vtime_seqcount;
1641 unsigned long long vtime_snap;
1642 enum {
1643 /* Task is sleeping or running in a CPU with VTIME inactive */
1644 VTIME_INACTIVE = 0,
1645 /* Task runs in userspace in a CPU with VTIME active */
1646 VTIME_USER,
1647 /* Task runs in kernelspace in a CPU with VTIME active */
1648 VTIME_SYS,
1649 } vtime_snap_whence;
1650 #endif
1651
1652 #ifdef CONFIG_NO_HZ_FULL
1653 atomic_t tick_dep_mask;
1654 #endif
1655 unsigned long nvcsw, nivcsw; /* context switch counts */
1656 u64 start_time; /* monotonic time in nsec */
1657 u64 real_start_time; /* boot based time in nsec */
1658 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1659 unsigned long min_flt, maj_flt;
1660
1661 #ifdef CONFIG_POSIX_TIMERS
1662 struct task_cputime cputime_expires;
1663 struct list_head cpu_timers[3];
1664 #endif
1665
1666 /* process credentials */
1667 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
1668 const struct cred __rcu *real_cred; /* objective and real subjective task
1669 * credentials (COW) */
1670 const struct cred __rcu *cred; /* effective (overridable) subjective task
1671 * credentials (COW) */
1672 char comm[TASK_COMM_LEN]; /* executable name excluding path
1673 - access with [gs]et_task_comm (which lock
1674 it with task_lock())
1675 - initialized normally by setup_new_exec */
1676 /* file system info */
1677 struct nameidata *nameidata;
1678 #ifdef CONFIG_SYSVIPC
1679 /* ipc stuff */
1680 struct sysv_sem sysvsem;
1681 struct sysv_shm sysvshm;
1682 #endif
1683 #ifdef CONFIG_DETECT_HUNG_TASK
1684 /* hung task detection */
1685 unsigned long last_switch_count;
1686 #endif
1687 /* filesystem information */
1688 struct fs_struct *fs;
1689 /* open file information */
1690 struct files_struct *files;
1691 /* namespaces */
1692 struct nsproxy *nsproxy;
1693 /* signal handlers */
1694 struct signal_struct *signal;
1695 struct sighand_struct *sighand;
1696
1697 sigset_t blocked, real_blocked;
1698 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1699 struct sigpending pending;
1700
1701 unsigned long sas_ss_sp;
1702 size_t sas_ss_size;
1703 unsigned sas_ss_flags;
1704
1705 struct callback_head *task_works;
1706
1707 struct audit_context *audit_context;
1708 #ifdef CONFIG_AUDITSYSCALL
1709 kuid_t loginuid;
1710 unsigned int sessionid;
1711 #endif
1712 struct seccomp seccomp;
1713
1714 /* Thread group tracking */
1715 u32 parent_exec_id;
1716 u32 self_exec_id;
1717 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1718 * mempolicy */
1719 spinlock_t alloc_lock;
1720
1721 /* Protection of the PI data structures: */
1722 raw_spinlock_t pi_lock;
1723
1724 struct wake_q_node wake_q;
1725
1726 #ifdef CONFIG_RT_MUTEXES
1727 /* PI waiters blocked on a rt_mutex held by this task */
1728 struct rb_root pi_waiters;
1729 struct rb_node *pi_waiters_leftmost;
1730 /* Deadlock detection and priority inheritance handling */
1731 struct rt_mutex_waiter *pi_blocked_on;
1732 #endif
1733
1734 #ifdef CONFIG_DEBUG_MUTEXES
1735 /* mutex deadlock detection */
1736 struct mutex_waiter *blocked_on;
1737 #endif
1738 #ifdef CONFIG_TRACE_IRQFLAGS
1739 unsigned int irq_events;
1740 unsigned long hardirq_enable_ip;
1741 unsigned long hardirq_disable_ip;
1742 unsigned int hardirq_enable_event;
1743 unsigned int hardirq_disable_event;
1744 int hardirqs_enabled;
1745 int hardirq_context;
1746 unsigned long softirq_disable_ip;
1747 unsigned long softirq_enable_ip;
1748 unsigned int softirq_disable_event;
1749 unsigned int softirq_enable_event;
1750 int softirqs_enabled;
1751 int softirq_context;
1752 #endif
1753 #ifdef CONFIG_LOCKDEP
1754 # define MAX_LOCK_DEPTH 48UL
1755 u64 curr_chain_key;
1756 int lockdep_depth;
1757 unsigned int lockdep_recursion;
1758 struct held_lock held_locks[MAX_LOCK_DEPTH];
1759 gfp_t lockdep_reclaim_gfp;
1760 #endif
1761 #ifdef CONFIG_UBSAN
1762 unsigned int in_ubsan;
1763 #endif
1764
1765 /* journalling filesystem info */
1766 void *journal_info;
1767
1768 /* stacked block device info */
1769 struct bio_list *bio_list;
1770
1771 #ifdef CONFIG_BLOCK
1772 /* stack plugging */
1773 struct blk_plug *plug;
1774 #endif
1775
1776 /* VM state */
1777 struct reclaim_state *reclaim_state;
1778
1779 struct backing_dev_info *backing_dev_info;
1780
1781 struct io_context *io_context;
1782
1783 unsigned long ptrace_message;
1784 siginfo_t *last_siginfo; /* For ptrace use. */
1785 struct task_io_accounting ioac;
1786 #if defined(CONFIG_TASK_XACCT)
1787 u64 acct_rss_mem1; /* accumulated rss usage */
1788 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1789 u64 acct_timexpd; /* stime + utime since last update */
1790 #endif
1791 #ifdef CONFIG_CPUSETS
1792 nodemask_t mems_allowed; /* Protected by alloc_lock */
1793 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1794 int cpuset_mem_spread_rotor;
1795 int cpuset_slab_spread_rotor;
1796 #endif
1797 #ifdef CONFIG_CGROUPS
1798 /* Control Group info protected by css_set_lock */
1799 struct css_set __rcu *cgroups;
1800 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1801 struct list_head cg_list;
1802 #endif
1803 #ifdef CONFIG_INTEL_RDT_A
1804 int closid;
1805 #endif
1806 #ifdef CONFIG_FUTEX
1807 struct robust_list_head __user *robust_list;
1808 #ifdef CONFIG_COMPAT
1809 struct compat_robust_list_head __user *compat_robust_list;
1810 #endif
1811 struct list_head pi_state_list;
1812 struct futex_pi_state *pi_state_cache;
1813 #endif
1814 #ifdef CONFIG_PERF_EVENTS
1815 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1816 struct mutex perf_event_mutex;
1817 struct list_head perf_event_list;
1818 #endif
1819 #ifdef CONFIG_DEBUG_PREEMPT
1820 unsigned long preempt_disable_ip;
1821 #endif
1822 #ifdef CONFIG_NUMA
1823 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1824 short il_next;
1825 short pref_node_fork;
1826 #endif
1827 #ifdef CONFIG_NUMA_BALANCING
1828 int numa_scan_seq;
1829 unsigned int numa_scan_period;
1830 unsigned int numa_scan_period_max;
1831 int numa_preferred_nid;
1832 unsigned long numa_migrate_retry;
1833 u64 node_stamp; /* migration stamp */
1834 u64 last_task_numa_placement;
1835 u64 last_sum_exec_runtime;
1836 struct callback_head numa_work;
1837
1838 struct list_head numa_entry;
1839 struct numa_group *numa_group;
1840
1841 /*
1842 * numa_faults is an array split into four regions:
1843 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1844 * in this precise order.
1845 *
1846 * faults_memory: Exponential decaying average of faults on a per-node
1847 * basis. Scheduling placement decisions are made based on these
1848 * counts. The values remain static for the duration of a PTE scan.
1849 * faults_cpu: Track the nodes the process was running on when a NUMA
1850 * hinting fault was incurred.
1851 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1852 * during the current scan window. When the scan completes, the counts
1853 * in faults_memory and faults_cpu decay and these values are copied.
1854 */
1855 unsigned long *numa_faults;
1856 unsigned long total_numa_faults;
1857
1858 /*
1859 * numa_faults_locality tracks if faults recorded during the last
1860 * scan window were remote/local or failed to migrate. The task scan
1861 * period is adapted based on the locality of the faults with different
1862 * weights depending on whether they were shared or private faults
1863 */
1864 unsigned long numa_faults_locality[3];
1865
1866 unsigned long numa_pages_migrated;
1867 #endif /* CONFIG_NUMA_BALANCING */
1868
1869 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1870 struct tlbflush_unmap_batch tlb_ubc;
1871 #endif
1872
1873 struct rcu_head rcu;
1874
1875 /*
1876 * cache last used pipe for splice
1877 */
1878 struct pipe_inode_info *splice_pipe;
1879
1880 struct page_frag task_frag;
1881
1882 #ifdef CONFIG_TASK_DELAY_ACCT
1883 struct task_delay_info *delays;
1884 #endif
1885 #ifdef CONFIG_FAULT_INJECTION
1886 int make_it_fail;
1887 #endif
1888 /*
1889 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1890 * balance_dirty_pages() for some dirty throttling pause
1891 */
1892 int nr_dirtied;
1893 int nr_dirtied_pause;
1894 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1895
1896 #ifdef CONFIG_LATENCYTOP
1897 int latency_record_count;
1898 struct latency_record latency_record[LT_SAVECOUNT];
1899 #endif
1900 /*
1901 * time slack values; these are used to round up poll() and
1902 * select() etc timeout values. These are in nanoseconds.
1903 */
1904 u64 timer_slack_ns;
1905 u64 default_timer_slack_ns;
1906
1907 #ifdef CONFIG_KASAN
1908 unsigned int kasan_depth;
1909 #endif
1910 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1911 /* Index of current stored address in ret_stack */
1912 int curr_ret_stack;
1913 /* Stack of return addresses for return function tracing */
1914 struct ftrace_ret_stack *ret_stack;
1915 /* time stamp for last schedule */
1916 unsigned long long ftrace_timestamp;
1917 /*
1918 * Number of functions that haven't been traced
1919 * because of depth overrun.
1920 */
1921 atomic_t trace_overrun;
1922 /* Pause for the tracing */
1923 atomic_t tracing_graph_pause;
1924 #endif
1925 #ifdef CONFIG_TRACING
1926 /* state flags for use by tracers */
1927 unsigned long trace;
1928 /* bitmask and counter of trace recursion */
1929 unsigned long trace_recursion;
1930 #endif /* CONFIG_TRACING */
1931 #ifdef CONFIG_KCOV
1932 /* Coverage collection mode enabled for this task (0 if disabled). */
1933 enum kcov_mode kcov_mode;
1934 /* Size of the kcov_area. */
1935 unsigned kcov_size;
1936 /* Buffer for coverage collection. */
1937 void *kcov_area;
1938 /* kcov desciptor wired with this task or NULL. */
1939 struct kcov *kcov;
1940 #endif
1941 #ifdef CONFIG_MEMCG
1942 struct mem_cgroup *memcg_in_oom;
1943 gfp_t memcg_oom_gfp_mask;
1944 int memcg_oom_order;
1945
1946 /* number of pages to reclaim on returning to userland */
1947 unsigned int memcg_nr_pages_over_high;
1948 #endif
1949 #ifdef CONFIG_UPROBES
1950 struct uprobe_task *utask;
1951 #endif
1952 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1953 unsigned int sequential_io;
1954 unsigned int sequential_io_avg;
1955 #endif
1956 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1957 unsigned long task_state_change;
1958 #endif
1959 int pagefault_disabled;
1960 #ifdef CONFIG_MMU
1961 struct task_struct *oom_reaper_list;
1962 #endif
1963 #ifdef CONFIG_VMAP_STACK
1964 struct vm_struct *stack_vm_area;
1965 #endif
1966 #ifdef CONFIG_THREAD_INFO_IN_TASK
1967 /* A live task holds one reference. */
1968 atomic_t stack_refcount;
1969 #endif
1970 /* CPU-specific state of this task */
1971 struct thread_struct thread;
1972 /*
1973 * WARNING: on x86, 'thread_struct' contains a variable-sized
1974 * structure. It *MUST* be at the end of 'task_struct'.
1975 *
1976 * Do not put anything below here!
1977 */
1978 };
1979
1980 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
1981 extern int arch_task_struct_size __read_mostly;
1982 #else
1983 # define arch_task_struct_size (sizeof(struct task_struct))
1984 #endif
1985
1986 #ifdef CONFIG_VMAP_STACK
1987 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1988 {
1989 return t->stack_vm_area;
1990 }
1991 #else
1992 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1993 {
1994 return NULL;
1995 }
1996 #endif
1997
1998 /* Future-safe accessor for struct task_struct's cpus_allowed. */
1999 #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
2000
2001 static inline int tsk_nr_cpus_allowed(struct task_struct *p)
2002 {
2003 return p->nr_cpus_allowed;
2004 }
2005
2006 #define TNF_MIGRATED 0x01
2007 #define TNF_NO_GROUP 0x02
2008 #define TNF_SHARED 0x04
2009 #define TNF_FAULT_LOCAL 0x08
2010 #define TNF_MIGRATE_FAIL 0x10
2011
2012 static inline bool in_vfork(struct task_struct *tsk)
2013 {
2014 bool ret;
2015
2016 /*
2017 * need RCU to access ->real_parent if CLONE_VM was used along with
2018 * CLONE_PARENT.
2019 *
2020 * We check real_parent->mm == tsk->mm because CLONE_VFORK does not
2021 * imply CLONE_VM
2022 *
2023 * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus
2024 * ->real_parent is not necessarily the task doing vfork(), so in
2025 * theory we can't rely on task_lock() if we want to dereference it.
2026 *
2027 * And in this case we can't trust the real_parent->mm == tsk->mm
2028 * check, it can be false negative. But we do not care, if init or
2029 * another oom-unkillable task does this it should blame itself.
2030 */
2031 rcu_read_lock();
2032 ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
2033 rcu_read_unlock();
2034
2035 return ret;
2036 }
2037
2038 #ifdef CONFIG_NUMA_BALANCING
2039 extern void task_numa_fault(int last_node, int node, int pages, int flags);
2040 extern pid_t task_numa_group_id(struct task_struct *p);
2041 extern void set_numabalancing_state(bool enabled);
2042 extern void task_numa_free(struct task_struct *p);
2043 extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
2044 int src_nid, int dst_cpu);
2045 #else
2046 static inline void task_numa_fault(int last_node, int node, int pages,
2047 int flags)
2048 {
2049 }
2050 static inline pid_t task_numa_group_id(struct task_struct *p)
2051 {
2052 return 0;
2053 }
2054 static inline void set_numabalancing_state(bool enabled)
2055 {
2056 }
2057 static inline void task_numa_free(struct task_struct *p)
2058 {
2059 }
2060 static inline bool should_numa_migrate_memory(struct task_struct *p,
2061 struct page *page, int src_nid, int dst_cpu)
2062 {
2063 return true;
2064 }
2065 #endif
2066
2067 static inline struct pid *task_pid(struct task_struct *task)
2068 {
2069 return task->pids[PIDTYPE_PID].pid;
2070 }
2071
2072 static inline struct pid *task_tgid(struct task_struct *task)
2073 {
2074 return task->group_leader->pids[PIDTYPE_PID].pid;
2075 }
2076
2077 /*
2078 * Without tasklist or rcu lock it is not safe to dereference
2079 * the result of task_pgrp/task_session even if task == current,
2080 * we can race with another thread doing sys_setsid/sys_setpgid.
2081 */
2082 static inline struct pid *task_pgrp(struct task_struct *task)
2083 {
2084 return task->group_leader->pids[PIDTYPE_PGID].pid;
2085 }
2086
2087 static inline struct pid *task_session(struct task_struct *task)
2088 {
2089 return task->group_leader->pids[PIDTYPE_SID].pid;
2090 }
2091
2092 struct pid_namespace;
2093
2094 /*
2095 * the helpers to get the task's different pids as they are seen
2096 * from various namespaces
2097 *
2098 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
2099 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
2100 * current.
2101 * task_xid_nr_ns() : id seen from the ns specified;
2102 *
2103 * set_task_vxid() : assigns a virtual id to a task;
2104 *
2105 * see also pid_nr() etc in include/linux/pid.h
2106 */
2107 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2108 struct pid_namespace *ns);
2109
2110 static inline pid_t task_pid_nr(struct task_struct *tsk)
2111 {
2112 return tsk->pid;
2113 }
2114
2115 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2116 struct pid_namespace *ns)
2117 {
2118 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2119 }
2120
2121 static inline pid_t task_pid_vnr(struct task_struct *tsk)
2122 {
2123 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2124 }
2125
2126
2127 static inline pid_t task_tgid_nr(struct task_struct *tsk)
2128 {
2129 return tsk->tgid;
2130 }
2131
2132 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2133
2134 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2135 {
2136 return pid_vnr(task_tgid(tsk));
2137 }
2138
2139
2140 static inline int pid_alive(const struct task_struct *p);
2141 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2142 {
2143 pid_t pid = 0;
2144
2145 rcu_read_lock();
2146 if (pid_alive(tsk))
2147 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2148 rcu_read_unlock();
2149
2150 return pid;
2151 }
2152
2153 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2154 {
2155 return task_ppid_nr_ns(tsk, &init_pid_ns);
2156 }
2157
2158 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2159 struct pid_namespace *ns)
2160 {
2161 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2162 }
2163
2164 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2165 {
2166 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2167 }
2168
2169
2170 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2171 struct pid_namespace *ns)
2172 {
2173 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2174 }
2175
2176 static inline pid_t task_session_vnr(struct task_struct *tsk)
2177 {
2178 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2179 }
2180
2181 /* obsolete, do not use */
2182 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2183 {
2184 return task_pgrp_nr_ns(tsk, &init_pid_ns);
2185 }
2186
2187 /**
2188 * pid_alive - check that a task structure is not stale
2189 * @p: Task structure to be checked.
2190 *
2191 * Test if a process is not yet dead (at most zombie state)
2192 * If pid_alive fails, then pointers within the task structure
2193 * can be stale and must not be dereferenced.
2194 *
2195 * Return: 1 if the process is alive. 0 otherwise.
2196 */
2197 static inline int pid_alive(const struct task_struct *p)
2198 {
2199 return p->pids[PIDTYPE_PID].pid != NULL;
2200 }
2201
2202 /**
2203 * is_global_init - check if a task structure is init. Since init
2204 * is free to have sub-threads we need to check tgid.
2205 * @tsk: Task structure to be checked.
2206 *
2207 * Check if a task structure is the first user space task the kernel created.
2208 *
2209 * Return: 1 if the task structure is init. 0 otherwise.
2210 */
2211 static inline int is_global_init(struct task_struct *tsk)
2212 {
2213 return task_tgid_nr(tsk) == 1;
2214 }
2215
2216 extern struct pid *cad_pid;
2217
2218 extern void free_task(struct task_struct *tsk);
2219 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2220
2221 extern void __put_task_struct(struct task_struct *t);
2222
2223 static inline void put_task_struct(struct task_struct *t)
2224 {
2225 if (atomic_dec_and_test(&t->usage))
2226 __put_task_struct(t);
2227 }
2228
2229 struct task_struct *task_rcu_dereference(struct task_struct **ptask);
2230 struct task_struct *try_get_task_struct(struct task_struct **ptask);
2231
2232 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2233 extern void task_cputime(struct task_struct *t,
2234 u64 *utime, u64 *stime);
2235 extern u64 task_gtime(struct task_struct *t);
2236 #else
2237 static inline void task_cputime(struct task_struct *t,
2238 u64 *utime, u64 *stime)
2239 {
2240 *utime = t->utime;
2241 *stime = t->stime;
2242 }
2243
2244 static inline u64 task_gtime(struct task_struct *t)
2245 {
2246 return t->gtime;
2247 }
2248 #endif
2249
2250 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
2251 static inline void task_cputime_scaled(struct task_struct *t,
2252 u64 *utimescaled,
2253 u64 *stimescaled)
2254 {
2255 *utimescaled = t->utimescaled;
2256 *stimescaled = t->stimescaled;
2257 }
2258 #else
2259 static inline void task_cputime_scaled(struct task_struct *t,
2260 u64 *utimescaled,
2261 u64 *stimescaled)
2262 {
2263 task_cputime(t, utimescaled, stimescaled);
2264 }
2265 #endif
2266
2267 extern void task_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
2268 extern void thread_group_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
2269
2270 /*
2271 * Per process flags
2272 */
2273 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
2274 #define PF_EXITING 0x00000004 /* getting shut down */
2275 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2276 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2277 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2278 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2279 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2280 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2281 #define PF_DUMPCORE 0x00000200 /* dumped core */
2282 #define PF_SIGNALED 0x00000400 /* killed by a signal */
2283 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
2284 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2285 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2286 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2287 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2288 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
2289 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2290 #define PF_KSWAPD 0x00040000 /* I am kswapd */
2291 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2292 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2293 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2294 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2295 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2296 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2297 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2298 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2299 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2300 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2301
2302 /*
2303 * Only the _current_ task can read/write to tsk->flags, but other
2304 * tasks can access tsk->flags in readonly mode for example
2305 * with tsk_used_math (like during threaded core dumping).
2306 * There is however an exception to this rule during ptrace
2307 * or during fork: the ptracer task is allowed to write to the
2308 * child->flags of its traced child (same goes for fork, the parent
2309 * can write to the child->flags), because we're guaranteed the
2310 * child is not running and in turn not changing child->flags
2311 * at the same time the parent does it.
2312 */
2313 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2314 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2315 #define clear_used_math() clear_stopped_child_used_math(current)
2316 #define set_used_math() set_stopped_child_used_math(current)
2317 #define conditional_stopped_child_used_math(condition, child) \
2318 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2319 #define conditional_used_math(condition) \
2320 conditional_stopped_child_used_math(condition, current)
2321 #define copy_to_stopped_child_used_math(child) \
2322 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2323 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2324 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2325 #define used_math() tsk_used_math(current)
2326
2327 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2328 * __GFP_FS is also cleared as it implies __GFP_IO.
2329 */
2330 static inline gfp_t memalloc_noio_flags(gfp_t flags)
2331 {
2332 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2333 flags &= ~(__GFP_IO | __GFP_FS);
2334 return flags;
2335 }
2336
2337 static inline unsigned int memalloc_noio_save(void)
2338 {
2339 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2340 current->flags |= PF_MEMALLOC_NOIO;
2341 return flags;
2342 }
2343
2344 static inline void memalloc_noio_restore(unsigned int flags)
2345 {
2346 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2347 }
2348
2349 /* Per-process atomic flags. */
2350 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2351 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2352 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2353 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
2354
2355
2356 #define TASK_PFA_TEST(name, func) \
2357 static inline bool task_##func(struct task_struct *p) \
2358 { return test_bit(PFA_##name, &p->atomic_flags); }
2359 #define TASK_PFA_SET(name, func) \
2360 static inline void task_set_##func(struct task_struct *p) \
2361 { set_bit(PFA_##name, &p->atomic_flags); }
2362 #define TASK_PFA_CLEAR(name, func) \
2363 static inline void task_clear_##func(struct task_struct *p) \
2364 { clear_bit(PFA_##name, &p->atomic_flags); }
2365
2366 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2367 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2368
2369 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2370 TASK_PFA_SET(SPREAD_PAGE, spread_page)
2371 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2372
2373 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2374 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2375 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2376
2377 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
2378 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
2379
2380 /*
2381 * task->jobctl flags
2382 */
2383 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2384
2385 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2386 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2387 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2388 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2389 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2390 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2391 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2392
2393 #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2394 #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2395 #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2396 #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2397 #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2398 #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2399 #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2400
2401 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2402 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2403
2404 extern bool task_set_jobctl_pending(struct task_struct *task,
2405 unsigned long mask);
2406 extern void task_clear_jobctl_trapping(struct task_struct *task);
2407 extern void task_clear_jobctl_pending(struct task_struct *task,
2408 unsigned long mask);
2409
2410 static inline void rcu_copy_process(struct task_struct *p)
2411 {
2412 #ifdef CONFIG_PREEMPT_RCU
2413 p->rcu_read_lock_nesting = 0;
2414 p->rcu_read_unlock_special.s = 0;
2415 p->rcu_blocked_node = NULL;
2416 INIT_LIST_HEAD(&p->rcu_node_entry);
2417 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2418 #ifdef CONFIG_TASKS_RCU
2419 p->rcu_tasks_holdout = false;
2420 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2421 p->rcu_tasks_idle_cpu = -1;
2422 #endif /* #ifdef CONFIG_TASKS_RCU */
2423 }
2424
2425 static inline void tsk_restore_flags(struct task_struct *task,
2426 unsigned long orig_flags, unsigned long flags)
2427 {
2428 task->flags &= ~flags;
2429 task->flags |= orig_flags & flags;
2430 }
2431
2432 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2433 const struct cpumask *trial);
2434 extern int task_can_attach(struct task_struct *p,
2435 const struct cpumask *cs_cpus_allowed);
2436 #ifdef CONFIG_SMP
2437 extern void do_set_cpus_allowed(struct task_struct *p,
2438 const struct cpumask *new_mask);
2439
2440 extern int set_cpus_allowed_ptr(struct task_struct *p,
2441 const struct cpumask *new_mask);
2442 #else
2443 static inline void do_set_cpus_allowed(struct task_struct *p,
2444 const struct cpumask *new_mask)
2445 {
2446 }
2447 static inline int set_cpus_allowed_ptr(struct task_struct *p,
2448 const struct cpumask *new_mask)
2449 {
2450 if (!cpumask_test_cpu(0, new_mask))
2451 return -EINVAL;
2452 return 0;
2453 }
2454 #endif
2455
2456 #ifdef CONFIG_NO_HZ_COMMON
2457 void calc_load_enter_idle(void);
2458 void calc_load_exit_idle(void);
2459 #else
2460 static inline void calc_load_enter_idle(void) { }
2461 static inline void calc_load_exit_idle(void) { }
2462 #endif /* CONFIG_NO_HZ_COMMON */
2463
2464 #ifndef cpu_relax_yield
2465 #define cpu_relax_yield() cpu_relax()
2466 #endif
2467
2468 /*
2469 * Do not use outside of architecture code which knows its limitations.
2470 *
2471 * sched_clock() has no promise of monotonicity or bounded drift between
2472 * CPUs, use (which you should not) requires disabling IRQs.
2473 *
2474 * Please use one of the three interfaces below.
2475 */
2476 extern unsigned long long notrace sched_clock(void);
2477 /*
2478 * See the comment in kernel/sched/clock.c
2479 */
2480 extern u64 running_clock(void);
2481 extern u64 sched_clock_cpu(int cpu);
2482
2483
2484 extern void sched_clock_init(void);
2485
2486 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2487 static inline void sched_clock_init_late(void)
2488 {
2489 }
2490
2491 static inline void sched_clock_tick(void)
2492 {
2493 }
2494
2495 static inline void clear_sched_clock_stable(void)
2496 {
2497 }
2498
2499 static inline void sched_clock_idle_sleep_event(void)
2500 {
2501 }
2502
2503 static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2504 {
2505 }
2506
2507 static inline u64 cpu_clock(int cpu)
2508 {
2509 return sched_clock();
2510 }
2511
2512 static inline u64 local_clock(void)
2513 {
2514 return sched_clock();
2515 }
2516 #else
2517 extern void sched_clock_init_late(void);
2518 /*
2519 * Architectures can set this to 1 if they have specified
2520 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2521 * but then during bootup it turns out that sched_clock()
2522 * is reliable after all:
2523 */
2524 extern int sched_clock_stable(void);
2525 extern void clear_sched_clock_stable(void);
2526
2527 extern void sched_clock_tick(void);
2528 extern void sched_clock_idle_sleep_event(void);
2529 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2530
2531 /*
2532 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2533 * time source that is monotonic per cpu argument and has bounded drift
2534 * between cpus.
2535 *
2536 * ######################### BIG FAT WARNING ##########################
2537 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2538 * # go backwards !! #
2539 * ####################################################################
2540 */
2541 static inline u64 cpu_clock(int cpu)
2542 {
2543 return sched_clock_cpu(cpu);
2544 }
2545
2546 static inline u64 local_clock(void)
2547 {
2548 return sched_clock_cpu(raw_smp_processor_id());
2549 }
2550 #endif
2551
2552 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2553 /*
2554 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2555 * The reason for this explicit opt-in is not to have perf penalty with
2556 * slow sched_clocks.
2557 */
2558 extern void enable_sched_clock_irqtime(void);
2559 extern void disable_sched_clock_irqtime(void);
2560 #else
2561 static inline void enable_sched_clock_irqtime(void) {}
2562 static inline void disable_sched_clock_irqtime(void) {}
2563 #endif
2564
2565 extern unsigned long long
2566 task_sched_runtime(struct task_struct *task);
2567
2568 /* sched_exec is called by processes performing an exec */
2569 #ifdef CONFIG_SMP
2570 extern void sched_exec(void);
2571 #else
2572 #define sched_exec() {}
2573 #endif
2574
2575 extern void sched_clock_idle_sleep_event(void);
2576 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2577
2578 #ifdef CONFIG_HOTPLUG_CPU
2579 extern void idle_task_exit(void);
2580 #else
2581 static inline void idle_task_exit(void) {}
2582 #endif
2583
2584 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2585 extern void wake_up_nohz_cpu(int cpu);
2586 #else
2587 static inline void wake_up_nohz_cpu(int cpu) { }
2588 #endif
2589
2590 #ifdef CONFIG_NO_HZ_FULL
2591 extern u64 scheduler_tick_max_deferment(void);
2592 #endif
2593
2594 #ifdef CONFIG_SCHED_AUTOGROUP
2595 extern void sched_autogroup_create_attach(struct task_struct *p);
2596 extern void sched_autogroup_detach(struct task_struct *p);
2597 extern void sched_autogroup_fork(struct signal_struct *sig);
2598 extern void sched_autogroup_exit(struct signal_struct *sig);
2599 extern void sched_autogroup_exit_task(struct task_struct *p);
2600 #ifdef CONFIG_PROC_FS
2601 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2602 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2603 #endif
2604 #else
2605 static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2606 static inline void sched_autogroup_detach(struct task_struct *p) { }
2607 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2608 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2609 static inline void sched_autogroup_exit_task(struct task_struct *p) { }
2610 #endif
2611
2612 extern int yield_to(struct task_struct *p, bool preempt);
2613 extern void set_user_nice(struct task_struct *p, long nice);
2614 extern int task_prio(const struct task_struct *p);
2615 /**
2616 * task_nice - return the nice value of a given task.
2617 * @p: the task in question.
2618 *
2619 * Return: The nice value [ -20 ... 0 ... 19 ].
2620 */
2621 static inline int task_nice(const struct task_struct *p)
2622 {
2623 return PRIO_TO_NICE((p)->static_prio);
2624 }
2625 extern int can_nice(const struct task_struct *p, const int nice);
2626 extern int task_curr(const struct task_struct *p);
2627 extern int idle_cpu(int cpu);
2628 extern int sched_setscheduler(struct task_struct *, int,
2629 const struct sched_param *);
2630 extern int sched_setscheduler_nocheck(struct task_struct *, int,
2631 const struct sched_param *);
2632 extern int sched_setattr(struct task_struct *,
2633 const struct sched_attr *);
2634 extern struct task_struct *idle_task(int cpu);
2635 /**
2636 * is_idle_task - is the specified task an idle task?
2637 * @p: the task in question.
2638 *
2639 * Return: 1 if @p is an idle task. 0 otherwise.
2640 */
2641 static inline bool is_idle_task(const struct task_struct *p)
2642 {
2643 return !!(p->flags & PF_IDLE);
2644 }
2645 extern struct task_struct *curr_task(int cpu);
2646 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
2647
2648 void yield(void);
2649
2650 union thread_union {
2651 #ifndef CONFIG_THREAD_INFO_IN_TASK
2652 struct thread_info thread_info;
2653 #endif
2654 unsigned long stack[THREAD_SIZE/sizeof(long)];
2655 };
2656
2657 #ifndef __HAVE_ARCH_KSTACK_END
2658 static inline int kstack_end(void *addr)
2659 {
2660 /* Reliable end of stack detection:
2661 * Some APM bios versions misalign the stack
2662 */
2663 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2664 }
2665 #endif
2666
2667 extern union thread_union init_thread_union;
2668 extern struct task_struct init_task;
2669
2670 extern struct mm_struct init_mm;
2671
2672 extern struct pid_namespace init_pid_ns;
2673
2674 /*
2675 * find a task by one of its numerical ids
2676 *
2677 * find_task_by_pid_ns():
2678 * finds a task by its pid in the specified namespace
2679 * find_task_by_vpid():
2680 * finds a task by its virtual pid
2681 *
2682 * see also find_vpid() etc in include/linux/pid.h
2683 */
2684
2685 extern struct task_struct *find_task_by_vpid(pid_t nr);
2686 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2687 struct pid_namespace *ns);
2688
2689 /* per-UID process charging. */
2690 extern struct user_struct * alloc_uid(kuid_t);
2691 static inline struct user_struct *get_uid(struct user_struct *u)
2692 {
2693 atomic_inc(&u->__count);
2694 return u;
2695 }
2696 extern void free_uid(struct user_struct *);
2697
2698 #include <asm/current.h>
2699
2700 extern void xtime_update(unsigned long ticks);
2701
2702 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2703 extern int wake_up_process(struct task_struct *tsk);
2704 extern void wake_up_new_task(struct task_struct *tsk);
2705 #ifdef CONFIG_SMP
2706 extern void kick_process(struct task_struct *tsk);
2707 #else
2708 static inline void kick_process(struct task_struct *tsk) { }
2709 #endif
2710 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2711 extern void sched_dead(struct task_struct *p);
2712
2713 extern void proc_caches_init(void);
2714 extern void flush_signals(struct task_struct *);
2715 extern void ignore_signals(struct task_struct *);
2716 extern void flush_signal_handlers(struct task_struct *, int force_default);
2717 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2718
2719 static inline int kernel_dequeue_signal(siginfo_t *info)
2720 {
2721 struct task_struct *tsk = current;
2722 siginfo_t __info;
2723 int ret;
2724
2725 spin_lock_irq(&tsk->sighand->siglock);
2726 ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2727 spin_unlock_irq(&tsk->sighand->siglock);
2728
2729 return ret;
2730 }
2731
2732 static inline void kernel_signal_stop(void)
2733 {
2734 spin_lock_irq(&current->sighand->siglock);
2735 if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2736 __set_current_state(TASK_STOPPED);
2737 spin_unlock_irq(&current->sighand->siglock);
2738
2739 schedule();
2740 }
2741
2742 extern void release_task(struct task_struct * p);
2743 extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2744 extern int force_sigsegv(int, struct task_struct *);
2745 extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2746 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2747 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2748 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2749 const struct cred *, u32);
2750 extern int kill_pgrp(struct pid *pid, int sig, int priv);
2751 extern int kill_pid(struct pid *pid, int sig, int priv);
2752 extern int kill_proc_info(int, struct siginfo *, pid_t);
2753 extern __must_check bool do_notify_parent(struct task_struct *, int);
2754 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2755 extern void force_sig(int, struct task_struct *);
2756 extern int send_sig(int, struct task_struct *, int);
2757 extern int zap_other_threads(struct task_struct *p);
2758 extern struct sigqueue *sigqueue_alloc(void);
2759 extern void sigqueue_free(struct sigqueue *);
2760 extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2761 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2762
2763 #ifdef TIF_RESTORE_SIGMASK
2764 /*
2765 * Legacy restore_sigmask accessors. These are inefficient on
2766 * SMP architectures because they require atomic operations.
2767 */
2768
2769 /**
2770 * set_restore_sigmask() - make sure saved_sigmask processing gets done
2771 *
2772 * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code
2773 * will run before returning to user mode, to process the flag. For
2774 * all callers, TIF_SIGPENDING is already set or it's no harm to set
2775 * it. TIF_RESTORE_SIGMASK need not be in the set of bits that the
2776 * arch code will notice on return to user mode, in case those bits
2777 * are scarce. We set TIF_SIGPENDING here to ensure that the arch
2778 * signal code always gets run when TIF_RESTORE_SIGMASK is set.
2779 */
2780 static inline void set_restore_sigmask(void)
2781 {
2782 set_thread_flag(TIF_RESTORE_SIGMASK);
2783 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2784 }
2785 static inline void clear_restore_sigmask(void)
2786 {
2787 clear_thread_flag(TIF_RESTORE_SIGMASK);
2788 }
2789 static inline bool test_restore_sigmask(void)
2790 {
2791 return test_thread_flag(TIF_RESTORE_SIGMASK);
2792 }
2793 static inline bool test_and_clear_restore_sigmask(void)
2794 {
2795 return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK);
2796 }
2797
2798 #else /* TIF_RESTORE_SIGMASK */
2799
2800 /* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */
2801 static inline void set_restore_sigmask(void)
2802 {
2803 current->restore_sigmask = true;
2804 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2805 }
2806 static inline void clear_restore_sigmask(void)
2807 {
2808 current->restore_sigmask = false;
2809 }
2810 static inline bool test_restore_sigmask(void)
2811 {
2812 return current->restore_sigmask;
2813 }
2814 static inline bool test_and_clear_restore_sigmask(void)
2815 {
2816 if (!current->restore_sigmask)
2817 return false;
2818 current->restore_sigmask = false;
2819 return true;
2820 }
2821 #endif
2822
2823 static inline void restore_saved_sigmask(void)
2824 {
2825 if (test_and_clear_restore_sigmask())
2826 __set_current_blocked(&current->saved_sigmask);
2827 }
2828
2829 static inline sigset_t *sigmask_to_save(void)
2830 {
2831 sigset_t *res = &current->blocked;
2832 if (unlikely(test_restore_sigmask()))
2833 res = &current->saved_sigmask;
2834 return res;
2835 }
2836
2837 static inline int kill_cad_pid(int sig, int priv)
2838 {
2839 return kill_pid(cad_pid, sig, priv);
2840 }
2841
2842 /* These can be the second arg to send_sig_info/send_group_sig_info. */
2843 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2844 #define SEND_SIG_PRIV ((struct siginfo *) 1)
2845 #define SEND_SIG_FORCED ((struct siginfo *) 2)
2846
2847 /*
2848 * True if we are on the alternate signal stack.
2849 */
2850 static inline int on_sig_stack(unsigned long sp)
2851 {
2852 /*
2853 * If the signal stack is SS_AUTODISARM then, by construction, we
2854 * can't be on the signal stack unless user code deliberately set
2855 * SS_AUTODISARM when we were already on it.
2856 *
2857 * This improves reliability: if user state gets corrupted such that
2858 * the stack pointer points very close to the end of the signal stack,
2859 * then this check will enable the signal to be handled anyway.
2860 */
2861 if (current->sas_ss_flags & SS_AUTODISARM)
2862 return 0;
2863
2864 #ifdef CONFIG_STACK_GROWSUP
2865 return sp >= current->sas_ss_sp &&
2866 sp - current->sas_ss_sp < current->sas_ss_size;
2867 #else
2868 return sp > current->sas_ss_sp &&
2869 sp - current->sas_ss_sp <= current->sas_ss_size;
2870 #endif
2871 }
2872
2873 static inline int sas_ss_flags(unsigned long sp)
2874 {
2875 if (!current->sas_ss_size)
2876 return SS_DISABLE;
2877
2878 return on_sig_stack(sp) ? SS_ONSTACK : 0;
2879 }
2880
2881 static inline void sas_ss_reset(struct task_struct *p)
2882 {
2883 p->sas_ss_sp = 0;
2884 p->sas_ss_size = 0;
2885 p->sas_ss_flags = SS_DISABLE;
2886 }
2887
2888 static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2889 {
2890 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2891 #ifdef CONFIG_STACK_GROWSUP
2892 return current->sas_ss_sp;
2893 #else
2894 return current->sas_ss_sp + current->sas_ss_size;
2895 #endif
2896 return sp;
2897 }
2898
2899 /*
2900 * Routines for handling mm_structs
2901 */
2902 extern struct mm_struct * mm_alloc(void);
2903
2904 /**
2905 * mmgrab() - Pin a &struct mm_struct.
2906 * @mm: The &struct mm_struct to pin.
2907 *
2908 * Make sure that @mm will not get freed even after the owning task
2909 * exits. This doesn't guarantee that the associated address space
2910 * will still exist later on and mmget_not_zero() has to be used before
2911 * accessing it.
2912 *
2913 * This is a preferred way to to pin @mm for a longer/unbounded amount
2914 * of time.
2915 *
2916 * Use mmdrop() to release the reference acquired by mmgrab().
2917 *
2918 * See also <Documentation/vm/active_mm.txt> for an in-depth explanation
2919 * of &mm_struct.mm_count vs &mm_struct.mm_users.
2920 */
2921 static inline void mmgrab(struct mm_struct *mm)
2922 {
2923 atomic_inc(&mm->mm_count);
2924 }
2925
2926 /* mmdrop drops the mm and the page tables */
2927 extern void __mmdrop(struct mm_struct *);
2928 static inline void mmdrop(struct mm_struct *mm)
2929 {
2930 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2931 __mmdrop(mm);
2932 }
2933
2934 static inline void mmdrop_async_fn(struct work_struct *work)
2935 {
2936 struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
2937 __mmdrop(mm);
2938 }
2939
2940 static inline void mmdrop_async(struct mm_struct *mm)
2941 {
2942 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
2943 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
2944 schedule_work(&mm->async_put_work);
2945 }
2946 }
2947
2948 /**
2949 * mmget() - Pin the address space associated with a &struct mm_struct.
2950 * @mm: The address space to pin.
2951 *
2952 * Make sure that the address space of the given &struct mm_struct doesn't
2953 * go away. This does not protect against parts of the address space being
2954 * modified or freed, however.
2955 *
2956 * Never use this function to pin this address space for an
2957 * unbounded/indefinite amount of time.
2958 *
2959 * Use mmput() to release the reference acquired by mmget().
2960 *
2961 * See also <Documentation/vm/active_mm.txt> for an in-depth explanation
2962 * of &mm_struct.mm_count vs &mm_struct.mm_users.
2963 */
2964 static inline void mmget(struct mm_struct *mm)
2965 {
2966 atomic_inc(&mm->mm_users);
2967 }
2968
2969 static inline bool mmget_not_zero(struct mm_struct *mm)
2970 {
2971 return atomic_inc_not_zero(&mm->mm_users);
2972 }
2973
2974 /* mmput gets rid of the mappings and all user-space */
2975 extern void mmput(struct mm_struct *);
2976 #ifdef CONFIG_MMU
2977 /* same as above but performs the slow path from the async context. Can
2978 * be called from the atomic context as well
2979 */
2980 extern void mmput_async(struct mm_struct *);
2981 #endif
2982
2983 /* Grab a reference to a task's mm, if it is not already going away */
2984 extern struct mm_struct *get_task_mm(struct task_struct *task);
2985 /*
2986 * Grab a reference to a task's mm, if it is not already going away
2987 * and ptrace_may_access with the mode parameter passed to it
2988 * succeeds.
2989 */
2990 extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2991 /* Remove the current tasks stale references to the old mm_struct */
2992 extern void mm_release(struct task_struct *, struct mm_struct *);
2993
2994 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
2995 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2996 struct task_struct *, unsigned long);
2997 #else
2998 extern int copy_thread(unsigned long, unsigned long, unsigned long,
2999 struct task_struct *);
3000
3001 /* Architectures that haven't opted into copy_thread_tls get the tls argument
3002 * via pt_regs, so ignore the tls argument passed via C. */
3003 static inline int copy_thread_tls(
3004 unsigned long clone_flags, unsigned long sp, unsigned long arg,
3005 struct task_struct *p, unsigned long tls)
3006 {
3007 return copy_thread(clone_flags, sp, arg, p);
3008 }
3009 #endif
3010 extern void flush_thread(void);
3011
3012 #ifdef CONFIG_HAVE_EXIT_THREAD
3013 extern void exit_thread(struct task_struct *tsk);
3014 #else
3015 static inline void exit_thread(struct task_struct *tsk)
3016 {
3017 }
3018 #endif
3019
3020 extern void exit_files(struct task_struct *);
3021 extern void __cleanup_sighand(struct sighand_struct *);
3022
3023 extern void exit_itimers(struct signal_struct *);
3024 extern void flush_itimer_signals(void);
3025
3026 extern void do_group_exit(int);
3027
3028 extern int do_execve(struct filename *,
3029 const char __user * const __user *,
3030 const char __user * const __user *);
3031 extern int do_execveat(int, struct filename *,
3032 const char __user * const __user *,
3033 const char __user * const __user *,
3034 int);
3035 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
3036 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
3037 struct task_struct *fork_idle(int);
3038 extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
3039
3040 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
3041 static inline void set_task_comm(struct task_struct *tsk, const char *from)
3042 {
3043 __set_task_comm(tsk, from, false);
3044 }
3045 extern char *get_task_comm(char *to, struct task_struct *tsk);
3046
3047 #ifdef CONFIG_SMP
3048 void scheduler_ipi(void);
3049 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
3050 #else
3051 static inline void scheduler_ipi(void) { }
3052 static inline unsigned long wait_task_inactive(struct task_struct *p,
3053 long match_state)
3054 {
3055 return 1;
3056 }
3057 #endif
3058
3059 #define tasklist_empty() \
3060 list_empty(&init_task.tasks)
3061
3062 #define next_task(p) \
3063 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
3064
3065 #define for_each_process(p) \
3066 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
3067
3068 extern bool current_is_single_threaded(void);
3069
3070 /*
3071 * Careful: do_each_thread/while_each_thread is a double loop so
3072 * 'break' will not work as expected - use goto instead.
3073 */
3074 #define do_each_thread(g, t) \
3075 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
3076
3077 #define while_each_thread(g, t) \
3078 while ((t = next_thread(t)) != g)
3079
3080 #define __for_each_thread(signal, t) \
3081 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
3082
3083 #define for_each_thread(p, t) \
3084 __for_each_thread((p)->signal, t)
3085
3086 /* Careful: this is a double loop, 'break' won't work as expected. */
3087 #define for_each_process_thread(p, t) \
3088 for_each_process(p) for_each_thread(p, t)
3089
3090 typedef int (*proc_visitor)(struct task_struct *p, void *data);
3091 void walk_process_tree(struct task_struct *top, proc_visitor, void *);
3092
3093 static inline int get_nr_threads(struct task_struct *tsk)
3094 {
3095 return tsk->signal->nr_threads;
3096 }
3097
3098 static inline bool thread_group_leader(struct task_struct *p)
3099 {
3100 return p->exit_signal >= 0;
3101 }
3102
3103 /* Do to the insanities of de_thread it is possible for a process
3104 * to have the pid of the thread group leader without actually being
3105 * the thread group leader. For iteration through the pids in proc
3106 * all we care about is that we have a task with the appropriate
3107 * pid, we don't actually care if we have the right task.
3108 */
3109 static inline bool has_group_leader_pid(struct task_struct *p)
3110 {
3111 return task_pid(p) == p->signal->leader_pid;
3112 }
3113
3114 static inline
3115 bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
3116 {
3117 return p1->signal == p2->signal;
3118 }
3119
3120 static inline struct task_struct *next_thread(const struct task_struct *p)
3121 {
3122 return list_entry_rcu(p->thread_group.next,
3123 struct task_struct, thread_group);
3124 }
3125
3126 static inline int thread_group_empty(struct task_struct *p)
3127 {
3128 return list_empty(&p->thread_group);
3129 }
3130
3131 #define delay_group_leader(p) \
3132 (thread_group_leader(p) && !thread_group_empty(p))
3133
3134 /*
3135 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
3136 * subscriptions and synchronises with wait4(). Also used in procfs. Also
3137 * pins the final release of task.io_context. Also protects ->cpuset and
3138 * ->cgroup.subsys[]. And ->vfork_done.
3139 *
3140 * Nests both inside and outside of read_lock(&tasklist_lock).
3141 * It must not be nested with write_lock_irq(&tasklist_lock),
3142 * neither inside nor outside.
3143 */
3144 static inline void task_lock(struct task_struct *p)
3145 {
3146 spin_lock(&p->alloc_lock);
3147 }
3148
3149 static inline void task_unlock(struct task_struct *p)
3150 {
3151 spin_unlock(&p->alloc_lock);
3152 }
3153
3154 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
3155 unsigned long *flags);
3156
3157 static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
3158 unsigned long *flags)
3159 {
3160 struct sighand_struct *ret;
3161
3162 ret = __lock_task_sighand(tsk, flags);
3163 (void)__cond_lock(&tsk->sighand->siglock, ret);
3164 return ret;
3165 }
3166
3167 static inline void unlock_task_sighand(struct task_struct *tsk,
3168 unsigned long *flags)
3169 {
3170 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
3171 }
3172
3173 /**
3174 * threadgroup_change_begin - mark the beginning of changes to a threadgroup
3175 * @tsk: task causing the changes
3176 *
3177 * All operations which modify a threadgroup - a new thread joining the
3178 * group, death of a member thread (the assertion of PF_EXITING) and
3179 * exec(2) dethreading the process and replacing the leader - are wrapped
3180 * by threadgroup_change_{begin|end}(). This is to provide a place which
3181 * subsystems needing threadgroup stability can hook into for
3182 * synchronization.
3183 */
3184 static inline void threadgroup_change_begin(struct task_struct *tsk)
3185 {
3186 might_sleep();
3187 cgroup_threadgroup_change_begin(tsk);
3188 }
3189
3190 /**
3191 * threadgroup_change_end - mark the end of changes to a threadgroup
3192 * @tsk: task causing the changes
3193 *
3194 * See threadgroup_change_begin().
3195 */
3196 static inline void threadgroup_change_end(struct task_struct *tsk)
3197 {
3198 cgroup_threadgroup_change_end(tsk);
3199 }
3200
3201 #ifdef CONFIG_THREAD_INFO_IN_TASK
3202
3203 static inline struct thread_info *task_thread_info(struct task_struct *task)
3204 {
3205 return &task->thread_info;
3206 }
3207
3208 /*
3209 * When accessing the stack of a non-current task that might exit, use
3210 * try_get_task_stack() instead. task_stack_page will return a pointer
3211 * that could get freed out from under you.
3212 */
3213 static inline void *task_stack_page(const struct task_struct *task)
3214 {
3215 return task->stack;
3216 }
3217
3218 #define setup_thread_stack(new,old) do { } while(0)
3219
3220 static inline unsigned long *end_of_stack(const struct task_struct *task)
3221 {
3222 return task->stack;
3223 }
3224
3225 #elif !defined(__HAVE_THREAD_FUNCTIONS)
3226
3227 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
3228 #define task_stack_page(task) ((void *)(task)->stack)
3229
3230 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
3231 {
3232 *task_thread_info(p) = *task_thread_info(org);
3233 task_thread_info(p)->task = p;
3234 }
3235
3236 /*
3237 * Return the address of the last usable long on the stack.
3238 *
3239 * When the stack grows down, this is just above the thread
3240 * info struct. Going any lower will corrupt the threadinfo.
3241 *
3242 * When the stack grows up, this is the highest address.
3243 * Beyond that position, we corrupt data on the next page.
3244 */
3245 static inline unsigned long *end_of_stack(struct task_struct *p)
3246 {
3247 #ifdef CONFIG_STACK_GROWSUP
3248 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
3249 #else
3250 return (unsigned long *)(task_thread_info(p) + 1);
3251 #endif
3252 }
3253
3254 #endif
3255
3256 #ifdef CONFIG_THREAD_INFO_IN_TASK
3257 static inline void *try_get_task_stack(struct task_struct *tsk)
3258 {
3259 return atomic_inc_not_zero(&tsk->stack_refcount) ?
3260 task_stack_page(tsk) : NULL;
3261 }
3262
3263 extern void put_task_stack(struct task_struct *tsk);
3264 #else
3265 static inline void *try_get_task_stack(struct task_struct *tsk)
3266 {
3267 return task_stack_page(tsk);
3268 }
3269
3270 static inline void put_task_stack(struct task_struct *tsk) {}
3271 #endif
3272
3273 #define task_stack_end_corrupted(task) \
3274 (*(end_of_stack(task)) != STACK_END_MAGIC)
3275
3276 static inline int object_is_on_stack(void *obj)
3277 {
3278 void *stack = task_stack_page(current);
3279
3280 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3281 }
3282
3283 extern void thread_stack_cache_init(void);
3284
3285 #ifdef CONFIG_DEBUG_STACK_USAGE
3286 static inline unsigned long stack_not_used(struct task_struct *p)
3287 {
3288 unsigned long *n = end_of_stack(p);
3289
3290 do { /* Skip over canary */
3291 # ifdef CONFIG_STACK_GROWSUP
3292 n--;
3293 # else
3294 n++;
3295 # endif
3296 } while (!*n);
3297
3298 # ifdef CONFIG_STACK_GROWSUP
3299 return (unsigned long)end_of_stack(p) - (unsigned long)n;
3300 # else
3301 return (unsigned long)n - (unsigned long)end_of_stack(p);
3302 # endif
3303 }
3304 #endif
3305 extern void set_task_stack_end_magic(struct task_struct *tsk);
3306
3307 /* set thread flags in other task's structures
3308 * - see asm/thread_info.h for TIF_xxxx flags available
3309 */
3310 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3311 {
3312 set_ti_thread_flag(task_thread_info(tsk), flag);
3313 }
3314
3315 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3316 {
3317 clear_ti_thread_flag(task_thread_info(tsk), flag);
3318 }
3319
3320 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3321 {
3322 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3323 }
3324
3325 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3326 {
3327 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3328 }
3329
3330 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3331 {
3332 return test_ti_thread_flag(task_thread_info(tsk), flag);
3333 }
3334
3335 static inline void set_tsk_need_resched(struct task_struct *tsk)
3336 {
3337 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3338 }
3339
3340 static inline void clear_tsk_need_resched(struct task_struct *tsk)
3341 {
3342 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3343 }
3344
3345 static inline int test_tsk_need_resched(struct task_struct *tsk)
3346 {
3347 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3348 }
3349
3350 static inline int restart_syscall(void)
3351 {
3352 set_tsk_thread_flag(current, TIF_SIGPENDING);
3353 return -ERESTARTNOINTR;
3354 }
3355
3356 static inline int signal_pending(struct task_struct *p)
3357 {
3358 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3359 }
3360
3361 static inline int __fatal_signal_pending(struct task_struct *p)
3362 {
3363 return unlikely(sigismember(&p->pending.signal, SIGKILL));
3364 }
3365
3366 static inline int fatal_signal_pending(struct task_struct *p)
3367 {
3368 return signal_pending(p) && __fatal_signal_pending(p);
3369 }
3370
3371 static inline int signal_pending_state(long state, struct task_struct *p)
3372 {
3373 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3374 return 0;
3375 if (!signal_pending(p))
3376 return 0;
3377
3378 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3379 }
3380
3381 /*
3382 * cond_resched() and cond_resched_lock(): latency reduction via
3383 * explicit rescheduling in places that are safe. The return
3384 * value indicates whether a reschedule was done in fact.
3385 * cond_resched_lock() will drop the spinlock before scheduling,
3386 * cond_resched_softirq() will enable bhs before scheduling.
3387 */
3388 #ifndef CONFIG_PREEMPT
3389 extern int _cond_resched(void);
3390 #else
3391 static inline int _cond_resched(void) { return 0; }
3392 #endif
3393
3394 #define cond_resched() ({ \
3395 ___might_sleep(__FILE__, __LINE__, 0); \
3396 _cond_resched(); \
3397 })
3398
3399 extern int __cond_resched_lock(spinlock_t *lock);
3400
3401 #define cond_resched_lock(lock) ({ \
3402 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3403 __cond_resched_lock(lock); \
3404 })
3405
3406 extern int __cond_resched_softirq(void);
3407
3408 #define cond_resched_softirq() ({ \
3409 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3410 __cond_resched_softirq(); \
3411 })
3412
3413 static inline void cond_resched_rcu(void)
3414 {
3415 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3416 rcu_read_unlock();
3417 cond_resched();
3418 rcu_read_lock();
3419 #endif
3420 }
3421
3422 static inline unsigned long get_preempt_disable_ip(struct task_struct *p)
3423 {
3424 #ifdef CONFIG_DEBUG_PREEMPT
3425 return p->preempt_disable_ip;
3426 #else
3427 return 0;
3428 #endif
3429 }
3430
3431 /*
3432 * Does a critical section need to be broken due to another
3433 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3434 * but a general need for low latency)
3435 */
3436 static inline int spin_needbreak(spinlock_t *lock)
3437 {
3438 #ifdef CONFIG_PREEMPT
3439 return spin_is_contended(lock);
3440 #else
3441 return 0;
3442 #endif
3443 }
3444
3445 /*
3446 * Idle thread specific functions to determine the need_resched
3447 * polling state.
3448 */
3449 #ifdef TIF_POLLING_NRFLAG
3450 static inline int tsk_is_polling(struct task_struct *p)
3451 {
3452 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3453 }
3454
3455 static inline void __current_set_polling(void)
3456 {
3457 set_thread_flag(TIF_POLLING_NRFLAG);
3458 }
3459
3460 static inline bool __must_check current_set_polling_and_test(void)
3461 {
3462 __current_set_polling();
3463
3464 /*
3465 * Polling state must be visible before we test NEED_RESCHED,
3466 * paired by resched_curr()
3467 */
3468 smp_mb__after_atomic();
3469
3470 return unlikely(tif_need_resched());
3471 }
3472
3473 static inline void __current_clr_polling(void)
3474 {
3475 clear_thread_flag(TIF_POLLING_NRFLAG);
3476 }
3477
3478 static inline bool __must_check current_clr_polling_and_test(void)
3479 {
3480 __current_clr_polling();
3481
3482 /*
3483 * Polling state must be visible before we test NEED_RESCHED,
3484 * paired by resched_curr()
3485 */
3486 smp_mb__after_atomic();
3487
3488 return unlikely(tif_need_resched());
3489 }
3490
3491 #else
3492 static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3493 static inline void __current_set_polling(void) { }
3494 static inline void __current_clr_polling(void) { }
3495
3496 static inline bool __must_check current_set_polling_and_test(void)
3497 {
3498 return unlikely(tif_need_resched());
3499 }
3500 static inline bool __must_check current_clr_polling_and_test(void)
3501 {
3502 return unlikely(tif_need_resched());
3503 }
3504 #endif
3505
3506 static inline void current_clr_polling(void)
3507 {
3508 __current_clr_polling();
3509
3510 /*
3511 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3512 * Once the bit is cleared, we'll get IPIs with every new
3513 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3514 * fold.
3515 */
3516 smp_mb(); /* paired with resched_curr() */
3517
3518 preempt_fold_need_resched();
3519 }
3520
3521 static __always_inline bool need_resched(void)
3522 {
3523 return unlikely(tif_need_resched());
3524 }
3525
3526 /*
3527 * Thread group CPU time accounting.
3528 */
3529 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3530 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3531
3532 /*
3533 * Reevaluate whether the task has signals pending delivery.
3534 * Wake the task if so.
3535 * This is required every time the blocked sigset_t changes.
3536 * callers must hold sighand->siglock.
3537 */
3538 extern void recalc_sigpending_and_wake(struct task_struct *t);
3539 extern void recalc_sigpending(void);
3540
3541 extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3542
3543 static inline void signal_wake_up(struct task_struct *t, bool resume)
3544 {
3545 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3546 }
3547 static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3548 {
3549 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3550 }
3551
3552 /*
3553 * Wrappers for p->thread_info->cpu access. No-op on UP.
3554 */
3555 #ifdef CONFIG_SMP
3556
3557 static inline unsigned int task_cpu(const struct task_struct *p)
3558 {
3559 #ifdef CONFIG_THREAD_INFO_IN_TASK
3560 return p->cpu;
3561 #else
3562 return task_thread_info(p)->cpu;
3563 #endif
3564 }
3565
3566 static inline int task_node(const struct task_struct *p)
3567 {
3568 return cpu_to_node(task_cpu(p));
3569 }
3570
3571 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3572
3573 #else
3574
3575 static inline unsigned int task_cpu(const struct task_struct *p)
3576 {
3577 return 0;
3578 }
3579
3580 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3581 {
3582 }
3583
3584 #endif /* CONFIG_SMP */
3585
3586 /*
3587 * In order to reduce various lock holder preemption latencies provide an
3588 * interface to see if a vCPU is currently running or not.
3589 *
3590 * This allows us to terminate optimistic spin loops and block, analogous to
3591 * the native optimistic spin heuristic of testing if the lock owner task is
3592 * running or not.
3593 */
3594 #ifndef vcpu_is_preempted
3595 # define vcpu_is_preempted(cpu) false
3596 #endif
3597
3598 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3599 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3600
3601 #ifdef CONFIG_CGROUP_SCHED
3602 extern struct task_group root_task_group;
3603 #endif /* CONFIG_CGROUP_SCHED */
3604
3605 extern int task_can_switch_user(struct user_struct *up,
3606 struct task_struct *tsk);
3607
3608 #ifdef CONFIG_TASK_XACCT
3609 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3610 {
3611 tsk->ioac.rchar += amt;
3612 }
3613
3614 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3615 {
3616 tsk->ioac.wchar += amt;
3617 }
3618
3619 static inline void inc_syscr(struct task_struct *tsk)
3620 {
3621 tsk->ioac.syscr++;
3622 }
3623
3624 static inline void inc_syscw(struct task_struct *tsk)
3625 {
3626 tsk->ioac.syscw++;
3627 }
3628 #else
3629 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3630 {
3631 }
3632
3633 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3634 {
3635 }
3636
3637 static inline void inc_syscr(struct task_struct *tsk)
3638 {
3639 }
3640
3641 static inline void inc_syscw(struct task_struct *tsk)
3642 {
3643 }
3644 #endif
3645
3646 #ifndef TASK_SIZE_OF
3647 #define TASK_SIZE_OF(tsk) TASK_SIZE
3648 #endif
3649
3650 #ifdef CONFIG_MEMCG
3651 extern void mm_update_next_owner(struct mm_struct *mm);
3652 #else
3653 static inline void mm_update_next_owner(struct mm_struct *mm)
3654 {
3655 }
3656 #endif /* CONFIG_MEMCG */
3657
3658 static inline unsigned long task_rlimit(const struct task_struct *tsk,
3659 unsigned int limit)
3660 {
3661 return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3662 }
3663
3664 static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3665 unsigned int limit)
3666 {
3667 return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3668 }
3669
3670 static inline unsigned long rlimit(unsigned int limit)
3671 {
3672 return task_rlimit(current, limit);
3673 }
3674
3675 static inline unsigned long rlimit_max(unsigned int limit)
3676 {
3677 return task_rlimit_max(current, limit);
3678 }
3679
3680 #define SCHED_CPUFREQ_RT (1U << 0)
3681 #define SCHED_CPUFREQ_DL (1U << 1)
3682 #define SCHED_CPUFREQ_IOWAIT (1U << 2)
3683
3684 #define SCHED_CPUFREQ_RT_DL (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL)
3685
3686 #ifdef CONFIG_CPU_FREQ
3687 struct update_util_data {
3688 void (*func)(struct update_util_data *data, u64 time, unsigned int flags);
3689 };
3690
3691 void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3692 void (*func)(struct update_util_data *data, u64 time,
3693 unsigned int flags));
3694 void cpufreq_remove_update_util_hook(int cpu);
3695 #endif /* CONFIG_CPU_FREQ */
3696
3697 #endif