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