1 #ifndef _LINUX_SCHED_H 2 #define _LINUX_SCHED_H 3 4 /* 5 * Define 'struct task_struct' and provide the main scheduler 6 * APIs (schedule(), wakeup variants, etc.) 7 */ 8 9 #include <uapi/linux/sched.h> 10 11 #include <asm/current.h> 12 13 #include <linux/pid.h> 14 #include <linux/sem.h> 15 #include <linux/shm.h> 16 #include <linux/kcov.h> 17 #include <linux/mutex.h> 18 #include <linux/plist.h> 19 #include <linux/hrtimer.h> 20 #include <linux/seccomp.h> 21 #include <linux/nodemask.h> 22 #include <linux/rcupdate.h> 23 #include <linux/resource.h> 24 #include <linux/latencytop.h> 25 #include <linux/sched/prio.h> 26 #include <linux/signal_types.h> 27 #include <linux/mm_types_task.h> 28 #include <linux/task_io_accounting.h> 29 30 /* task_struct member predeclarations (sorted alphabetically): */ 31 struct audit_context; 32 struct backing_dev_info; 33 struct bio_list; 34 struct blk_plug; 35 struct cfs_rq; 36 struct fs_struct; 37 struct futex_pi_state; 38 struct io_context; 39 struct mempolicy; 40 struct nameidata; 41 struct nsproxy; 42 struct perf_event_context; 43 struct pid_namespace; 44 struct pipe_inode_info; 45 struct rcu_node; 46 struct reclaim_state; 47 struct robust_list_head; 48 struct sched_attr; 49 struct sched_param; 50 struct seq_file; 51 struct sighand_struct; 52 struct signal_struct; 53 struct task_delay_info; 54 struct task_group; 55 56 /* 57 * Task state bitmask. NOTE! These bits are also 58 * encoded in fs/proc/array.c: get_task_state(). 59 * 60 * We have two separate sets of flags: task->state 61 * is about runnability, while task->exit_state are 62 * about the task exiting. Confusing, but this way 63 * modifying one set can't modify the other one by 64 * mistake. 65 */ 66 67 /* Used in tsk->state: */ 68 #define TASK_RUNNING 0 69 #define TASK_INTERRUPTIBLE 1 70 #define TASK_UNINTERRUPTIBLE 2 71 #define __TASK_STOPPED 4 72 #define __TASK_TRACED 8 73 /* Used in tsk->exit_state: */ 74 #define EXIT_DEAD 16 75 #define EXIT_ZOMBIE 32 76 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 77 /* Used in tsk->state again: */ 78 #define TASK_DEAD 64 79 #define TASK_WAKEKILL 128 80 #define TASK_WAKING 256 81 #define TASK_PARKED 512 82 #define TASK_NOLOAD 1024 83 #define TASK_NEW 2048 84 #define TASK_STATE_MAX 4096 85 86 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn" 87 88 /* Convenience macros for the sake of set_current_state: */ 89 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 90 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 91 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) 92 93 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 94 95 /* Convenience macros for the sake of wake_up(): */ 96 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 97 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED) 98 99 /* get_task_state(): */ 100 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 101 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 102 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD) 103 104 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0) 105 106 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0) 107 108 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) 109 110 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \ 111 (task->flags & PF_FROZEN) == 0 && \ 112 (task->state & TASK_NOLOAD) == 0) 113 114 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 115 116 #define __set_current_state(state_value) \ 117 do { \ 118 current->task_state_change = _THIS_IP_; \ 119 current->state = (state_value); \ 120 } while (0) 121 #define set_current_state(state_value) \ 122 do { \ 123 current->task_state_change = _THIS_IP_; \ 124 smp_store_mb(current->state, (state_value)); \ 125 } while (0) 126 127 #else 128 /* 129 * set_current_state() includes a barrier so that the write of current->state 130 * is correctly serialised wrt the caller's subsequent test of whether to 131 * actually sleep: 132 * 133 * for (;;) { 134 * set_current_state(TASK_UNINTERRUPTIBLE); 135 * if (!need_sleep) 136 * break; 137 * 138 * schedule(); 139 * } 140 * __set_current_state(TASK_RUNNING); 141 * 142 * If the caller does not need such serialisation (because, for instance, the 143 * condition test and condition change and wakeup are under the same lock) then 144 * use __set_current_state(). 145 * 146 * The above is typically ordered against the wakeup, which does: 147 * 148 * need_sleep = false; 149 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 150 * 151 * Where wake_up_state() (and all other wakeup primitives) imply enough 152 * barriers to order the store of the variable against wakeup. 153 * 154 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 155 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 156 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 157 * 158 * This is obviously fine, since they both store the exact same value. 159 * 160 * Also see the comments of try_to_wake_up(). 161 */ 162 #define __set_current_state(state_value) do { current->state = (state_value); } while (0) 163 #define set_current_state(state_value) smp_store_mb(current->state, (state_value)) 164 #endif 165 166 /* Task command name length: */ 167 #define TASK_COMM_LEN 16 168 169 extern cpumask_var_t cpu_isolated_map; 170 171 extern void scheduler_tick(void); 172 173 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 174 175 extern long schedule_timeout(long timeout); 176 extern long schedule_timeout_interruptible(long timeout); 177 extern long schedule_timeout_killable(long timeout); 178 extern long schedule_timeout_uninterruptible(long timeout); 179 extern long schedule_timeout_idle(long timeout); 180 asmlinkage void schedule(void); 181 extern void schedule_preempt_disabled(void); 182 183 extern int __must_check io_schedule_prepare(void); 184 extern void io_schedule_finish(int token); 185 extern long io_schedule_timeout(long timeout); 186 extern void io_schedule(void); 187 188 /** 189 * struct prev_cputime - snapshot of system and user cputime 190 * @utime: time spent in user mode 191 * @stime: time spent in system mode 192 * @lock: protects the above two fields 193 * 194 * Stores previous user/system time values such that we can guarantee 195 * monotonicity. 196 */ 197 struct prev_cputime { 198 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 199 u64 utime; 200 u64 stime; 201 raw_spinlock_t lock; 202 #endif 203 }; 204 205 /** 206 * struct task_cputime - collected CPU time counts 207 * @utime: time spent in user mode, in nanoseconds 208 * @stime: time spent in kernel mode, in nanoseconds 209 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds 210 * 211 * This structure groups together three kinds of CPU time that are tracked for 212 * threads and thread groups. Most things considering CPU time want to group 213 * these counts together and treat all three of them in parallel. 214 */ 215 struct task_cputime { 216 u64 utime; 217 u64 stime; 218 unsigned long long sum_exec_runtime; 219 }; 220 221 /* Alternate field names when used on cache expirations: */ 222 #define virt_exp utime 223 #define prof_exp stime 224 #define sched_exp sum_exec_runtime 225 226 struct sched_info { 227 #ifdef CONFIG_SCHED_INFO 228 /* Cumulative counters: */ 229 230 /* # of times we have run on this CPU: */ 231 unsigned long pcount; 232 233 /* Time spent waiting on a runqueue: */ 234 unsigned long long run_delay; 235 236 /* Timestamps: */ 237 238 /* When did we last run on a CPU? */ 239 unsigned long long last_arrival; 240 241 /* When were we last queued to run? */ 242 unsigned long long last_queued; 243 244 #endif /* CONFIG_SCHED_INFO */ 245 }; 246 247 /* 248 * Integer metrics need fixed point arithmetic, e.g., sched/fair 249 * has a few: load, load_avg, util_avg, freq, and capacity. 250 * 251 * We define a basic fixed point arithmetic range, and then formalize 252 * all these metrics based on that basic range. 253 */ 254 # define SCHED_FIXEDPOINT_SHIFT 10 255 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 256 257 struct load_weight { 258 unsigned long weight; 259 u32 inv_weight; 260 }; 261 262 /* 263 * The load_avg/util_avg accumulates an infinite geometric series 264 * (see __update_load_avg() in kernel/sched/fair.c). 265 * 266 * [load_avg definition] 267 * 268 * load_avg = runnable% * scale_load_down(load) 269 * 270 * where runnable% is the time ratio that a sched_entity is runnable. 271 * For cfs_rq, it is the aggregated load_avg of all runnable and 272 * blocked sched_entities. 273 * 274 * load_avg may also take frequency scaling into account: 275 * 276 * load_avg = runnable% * scale_load_down(load) * freq% 277 * 278 * where freq% is the CPU frequency normalized to the highest frequency. 279 * 280 * [util_avg definition] 281 * 282 * util_avg = running% * SCHED_CAPACITY_SCALE 283 * 284 * where running% is the time ratio that a sched_entity is running on 285 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable 286 * and blocked sched_entities. 287 * 288 * util_avg may also factor frequency scaling and CPU capacity scaling: 289 * 290 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity% 291 * 292 * where freq% is the same as above, and capacity% is the CPU capacity 293 * normalized to the greatest capacity (due to uarch differences, etc). 294 * 295 * N.B., the above ratios (runnable%, running%, freq%, and capacity%) 296 * themselves are in the range of [0, 1]. To do fixed point arithmetics, 297 * we therefore scale them to as large a range as necessary. This is for 298 * example reflected by util_avg's SCHED_CAPACITY_SCALE. 299 * 300 * [Overflow issue] 301 * 302 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 303 * with the highest load (=88761), always runnable on a single cfs_rq, 304 * and should not overflow as the number already hits PID_MAX_LIMIT. 305 * 306 * For all other cases (including 32-bit kernels), struct load_weight's 307 * weight will overflow first before we do, because: 308 * 309 * Max(load_avg) <= Max(load.weight) 310 * 311 * Then it is the load_weight's responsibility to consider overflow 312 * issues. 313 */ 314 struct sched_avg { 315 u64 last_update_time; 316 u64 load_sum; 317 u32 util_sum; 318 u32 period_contrib; 319 unsigned long load_avg; 320 unsigned long util_avg; 321 }; 322 323 struct sched_statistics { 324 #ifdef CONFIG_SCHEDSTATS 325 u64 wait_start; 326 u64 wait_max; 327 u64 wait_count; 328 u64 wait_sum; 329 u64 iowait_count; 330 u64 iowait_sum; 331 332 u64 sleep_start; 333 u64 sleep_max; 334 s64 sum_sleep_runtime; 335 336 u64 block_start; 337 u64 block_max; 338 u64 exec_max; 339 u64 slice_max; 340 341 u64 nr_migrations_cold; 342 u64 nr_failed_migrations_affine; 343 u64 nr_failed_migrations_running; 344 u64 nr_failed_migrations_hot; 345 u64 nr_forced_migrations; 346 347 u64 nr_wakeups; 348 u64 nr_wakeups_sync; 349 u64 nr_wakeups_migrate; 350 u64 nr_wakeups_local; 351 u64 nr_wakeups_remote; 352 u64 nr_wakeups_affine; 353 u64 nr_wakeups_affine_attempts; 354 u64 nr_wakeups_passive; 355 u64 nr_wakeups_idle; 356 #endif 357 }; 358 359 struct sched_entity { 360 /* For load-balancing: */ 361 struct load_weight load; 362 struct rb_node run_node; 363 struct list_head group_node; 364 unsigned int on_rq; 365 366 u64 exec_start; 367 u64 sum_exec_runtime; 368 u64 vruntime; 369 u64 prev_sum_exec_runtime; 370 371 u64 nr_migrations; 372 373 struct sched_statistics statistics; 374 375 #ifdef CONFIG_FAIR_GROUP_SCHED 376 int depth; 377 struct sched_entity *parent; 378 /* rq on which this entity is (to be) queued: */ 379 struct cfs_rq *cfs_rq; 380 /* rq "owned" by this entity/group: */ 381 struct cfs_rq *my_q; 382 #endif 383 384 #ifdef CONFIG_SMP 385 /* 386 * Per entity load average tracking. 387 * 388 * Put into separate cache line so it does not 389 * collide with read-mostly values above. 390 */ 391 struct sched_avg avg ____cacheline_aligned_in_smp; 392 #endif 393 }; 394 395 struct sched_rt_entity { 396 struct list_head run_list; 397 unsigned long timeout; 398 unsigned long watchdog_stamp; 399 unsigned int time_slice; 400 unsigned short on_rq; 401 unsigned short on_list; 402 403 struct sched_rt_entity *back; 404 #ifdef CONFIG_RT_GROUP_SCHED 405 struct sched_rt_entity *parent; 406 /* rq on which this entity is (to be) queued: */ 407 struct rt_rq *rt_rq; 408 /* rq "owned" by this entity/group: */ 409 struct rt_rq *my_q; 410 #endif 411 }; 412 413 struct sched_dl_entity { 414 struct rb_node rb_node; 415 416 /* 417 * Original scheduling parameters. Copied here from sched_attr 418 * during sched_setattr(), they will remain the same until 419 * the next sched_setattr(). 420 */ 421 u64 dl_runtime; /* Maximum runtime for each instance */ 422 u64 dl_deadline; /* Relative deadline of each instance */ 423 u64 dl_period; /* Separation of two instances (period) */ 424 u64 dl_bw; /* dl_runtime / dl_period */ 425 u64 dl_density; /* dl_runtime / dl_deadline */ 426 427 /* 428 * Actual scheduling parameters. Initialized with the values above, 429 * they are continously updated during task execution. Note that 430 * the remaining runtime could be < 0 in case we are in overrun. 431 */ 432 s64 runtime; /* Remaining runtime for this instance */ 433 u64 deadline; /* Absolute deadline for this instance */ 434 unsigned int flags; /* Specifying the scheduler behaviour */ 435 436 /* 437 * Some bool flags: 438 * 439 * @dl_throttled tells if we exhausted the runtime. If so, the 440 * task has to wait for a replenishment to be performed at the 441 * next firing of dl_timer. 442 * 443 * @dl_boosted tells if we are boosted due to DI. If so we are 444 * outside bandwidth enforcement mechanism (but only until we 445 * exit the critical section); 446 * 447 * @dl_yielded tells if task gave up the CPU before consuming 448 * all its available runtime during the last job. 449 * 450 * @dl_non_contending tells if the task is inactive while still 451 * contributing to the active utilization. In other words, it 452 * indicates if the inactive timer has been armed and its handler 453 * has not been executed yet. This flag is useful to avoid race 454 * conditions between the inactive timer handler and the wakeup 455 * code. 456 */ 457 int dl_throttled; 458 int dl_boosted; 459 int dl_yielded; 460 int dl_non_contending; 461 462 /* 463 * Bandwidth enforcement timer. Each -deadline task has its 464 * own bandwidth to be enforced, thus we need one timer per task. 465 */ 466 struct hrtimer dl_timer; 467 468 /* 469 * Inactive timer, responsible for decreasing the active utilization 470 * at the "0-lag time". When a -deadline task blocks, it contributes 471 * to GRUB's active utilization until the "0-lag time", hence a 472 * timer is needed to decrease the active utilization at the correct 473 * time. 474 */ 475 struct hrtimer inactive_timer; 476 }; 477 478 union rcu_special { 479 struct { 480 u8 blocked; 481 u8 need_qs; 482 u8 exp_need_qs; 483 484 /* Otherwise the compiler can store garbage here: */ 485 u8 pad; 486 } b; /* Bits. */ 487 u32 s; /* Set of bits. */ 488 }; 489 490 enum perf_event_task_context { 491 perf_invalid_context = -1, 492 perf_hw_context = 0, 493 perf_sw_context, 494 perf_nr_task_contexts, 495 }; 496 497 struct wake_q_node { 498 struct wake_q_node *next; 499 }; 500 501 struct task_struct { 502 #ifdef CONFIG_THREAD_INFO_IN_TASK 503 /* 504 * For reasons of header soup (see current_thread_info()), this 505 * must be the first element of task_struct. 506 */ 507 struct thread_info thread_info; 508 #endif 509 /* -1 unrunnable, 0 runnable, >0 stopped: */ 510 volatile long state; 511 void *stack; 512 atomic_t usage; 513 /* Per task flags (PF_*), defined further below: */ 514 unsigned int flags; 515 unsigned int ptrace; 516 517 #ifdef CONFIG_SMP 518 struct llist_node wake_entry; 519 int on_cpu; 520 #ifdef CONFIG_THREAD_INFO_IN_TASK 521 /* Current CPU: */ 522 unsigned int cpu; 523 #endif 524 unsigned int wakee_flips; 525 unsigned long wakee_flip_decay_ts; 526 struct task_struct *last_wakee; 527 528 int wake_cpu; 529 #endif 530 int on_rq; 531 532 int prio; 533 int static_prio; 534 int normal_prio; 535 unsigned int rt_priority; 536 537 const struct sched_class *sched_class; 538 struct sched_entity se; 539 struct sched_rt_entity rt; 540 #ifdef CONFIG_CGROUP_SCHED 541 struct task_group *sched_task_group; 542 #endif 543 struct sched_dl_entity dl; 544 545 #ifdef CONFIG_PREEMPT_NOTIFIERS 546 /* List of struct preempt_notifier: */ 547 struct hlist_head preempt_notifiers; 548 #endif 549 550 #ifdef CONFIG_BLK_DEV_IO_TRACE 551 unsigned int btrace_seq; 552 #endif 553 554 unsigned int policy; 555 int nr_cpus_allowed; 556 cpumask_t cpus_allowed; 557 558 #ifdef CONFIG_PREEMPT_RCU 559 int rcu_read_lock_nesting; 560 union rcu_special rcu_read_unlock_special; 561 struct list_head rcu_node_entry; 562 struct rcu_node *rcu_blocked_node; 563 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 564 565 #ifdef CONFIG_TASKS_RCU 566 unsigned long rcu_tasks_nvcsw; 567 bool rcu_tasks_holdout; 568 struct list_head rcu_tasks_holdout_list; 569 int rcu_tasks_idle_cpu; 570 #endif /* #ifdef CONFIG_TASKS_RCU */ 571 572 struct sched_info sched_info; 573 574 struct list_head tasks; 575 #ifdef CONFIG_SMP 576 struct plist_node pushable_tasks; 577 struct rb_node pushable_dl_tasks; 578 #endif 579 580 struct mm_struct *mm; 581 struct mm_struct *active_mm; 582 583 /* Per-thread vma caching: */ 584 struct vmacache vmacache; 585 586 #ifdef SPLIT_RSS_COUNTING 587 struct task_rss_stat rss_stat; 588 #endif 589 int exit_state; 590 int exit_code; 591 int exit_signal; 592 /* The signal sent when the parent dies: */ 593 int pdeath_signal; 594 /* JOBCTL_*, siglock protected: */ 595 unsigned long jobctl; 596 597 /* Used for emulating ABI behavior of previous Linux versions: */ 598 unsigned int personality; 599 600 /* Scheduler bits, serialized by scheduler locks: */ 601 unsigned sched_reset_on_fork:1; 602 unsigned sched_contributes_to_load:1; 603 unsigned sched_migrated:1; 604 unsigned sched_remote_wakeup:1; 605 /* Force alignment to the next boundary: */ 606 unsigned :0; 607 608 /* Unserialized, strictly 'current' */ 609 610 /* Bit to tell LSMs we're in execve(): */ 611 unsigned in_execve:1; 612 unsigned in_iowait:1; 613 #ifndef TIF_RESTORE_SIGMASK 614 unsigned restore_sigmask:1; 615 #endif 616 #ifdef CONFIG_MEMCG 617 unsigned memcg_may_oom:1; 618 #ifndef CONFIG_SLOB 619 unsigned memcg_kmem_skip_account:1; 620 #endif 621 #endif 622 #ifdef CONFIG_COMPAT_BRK 623 unsigned brk_randomized:1; 624 #endif 625 #ifdef CONFIG_CGROUPS 626 /* disallow userland-initiated cgroup migration */ 627 unsigned no_cgroup_migration:1; 628 #endif 629 630 unsigned long atomic_flags; /* Flags requiring atomic access. */ 631 632 struct restart_block restart_block; 633 634 pid_t pid; 635 pid_t tgid; 636 637 #ifdef CONFIG_CC_STACKPROTECTOR 638 /* Canary value for the -fstack-protector GCC feature: */ 639 unsigned long stack_canary; 640 #endif 641 /* 642 * Pointers to the (original) parent process, youngest child, younger sibling, 643 * older sibling, respectively. (p->father can be replaced with 644 * p->real_parent->pid) 645 */ 646 647 /* Real parent process: */ 648 struct task_struct __rcu *real_parent; 649 650 /* Recipient of SIGCHLD, wait4() reports: */ 651 struct task_struct __rcu *parent; 652 653 /* 654 * Children/sibling form the list of natural children: 655 */ 656 struct list_head children; 657 struct list_head sibling; 658 struct task_struct *group_leader; 659 660 /* 661 * 'ptraced' is the list of tasks this task is using ptrace() on. 662 * 663 * This includes both natural children and PTRACE_ATTACH targets. 664 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 665 */ 666 struct list_head ptraced; 667 struct list_head ptrace_entry; 668 669 /* PID/PID hash table linkage. */ 670 struct pid_link pids[PIDTYPE_MAX]; 671 struct list_head thread_group; 672 struct list_head thread_node; 673 674 struct completion *vfork_done; 675 676 /* CLONE_CHILD_SETTID: */ 677 int __user *set_child_tid; 678 679 /* CLONE_CHILD_CLEARTID: */ 680 int __user *clear_child_tid; 681 682 u64 utime; 683 u64 stime; 684 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 685 u64 utimescaled; 686 u64 stimescaled; 687 #endif 688 u64 gtime; 689 struct prev_cputime prev_cputime; 690 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 691 seqcount_t vtime_seqcount; 692 unsigned long long vtime_snap; 693 enum { 694 /* Task is sleeping or running in a CPU with VTIME inactive: */ 695 VTIME_INACTIVE = 0, 696 /* Task runs in userspace in a CPU with VTIME active: */ 697 VTIME_USER, 698 /* Task runs in kernelspace in a CPU with VTIME active: */ 699 VTIME_SYS, 700 } vtime_snap_whence; 701 #endif 702 703 #ifdef CONFIG_NO_HZ_FULL 704 atomic_t tick_dep_mask; 705 #endif 706 /* Context switch counts: */ 707 unsigned long nvcsw; 708 unsigned long nivcsw; 709 710 /* Monotonic time in nsecs: */ 711 u64 start_time; 712 713 /* Boot based time in nsecs: */ 714 u64 real_start_time; 715 716 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 717 unsigned long min_flt; 718 unsigned long maj_flt; 719 720 #ifdef CONFIG_POSIX_TIMERS 721 struct task_cputime cputime_expires; 722 struct list_head cpu_timers[3]; 723 #endif 724 725 /* Process credentials: */ 726 727 /* Tracer's credentials at attach: */ 728 const struct cred __rcu *ptracer_cred; 729 730 /* Objective and real subjective task credentials (COW): */ 731 const struct cred __rcu *real_cred; 732 733 /* Effective (overridable) subjective task credentials (COW): */ 734 const struct cred __rcu *cred; 735 736 /* 737 * executable name, excluding path. 738 * 739 * - normally initialized setup_new_exec() 740 * - access it with [gs]et_task_comm() 741 * - lock it with task_lock() 742 */ 743 char comm[TASK_COMM_LEN]; 744 745 struct nameidata *nameidata; 746 747 #ifdef CONFIG_SYSVIPC 748 struct sysv_sem sysvsem; 749 struct sysv_shm sysvshm; 750 #endif 751 #ifdef CONFIG_DETECT_HUNG_TASK 752 unsigned long last_switch_count; 753 #endif 754 /* Filesystem information: */ 755 struct fs_struct *fs; 756 757 /* Open file information: */ 758 struct files_struct *files; 759 760 /* Namespaces: */ 761 struct nsproxy *nsproxy; 762 763 /* Signal handlers: */ 764 struct signal_struct *signal; 765 struct sighand_struct *sighand; 766 sigset_t blocked; 767 sigset_t real_blocked; 768 /* Restored if set_restore_sigmask() was used: */ 769 sigset_t saved_sigmask; 770 struct sigpending pending; 771 unsigned long sas_ss_sp; 772 size_t sas_ss_size; 773 unsigned int sas_ss_flags; 774 775 struct callback_head *task_works; 776 777 struct audit_context *audit_context; 778 #ifdef CONFIG_AUDITSYSCALL 779 kuid_t loginuid; 780 unsigned int sessionid; 781 #endif 782 struct seccomp seccomp; 783 784 /* Thread group tracking: */ 785 u32 parent_exec_id; 786 u32 self_exec_id; 787 788 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 789 spinlock_t alloc_lock; 790 791 /* Protection of the PI data structures: */ 792 raw_spinlock_t pi_lock; 793 794 struct wake_q_node wake_q; 795 796 #ifdef CONFIG_RT_MUTEXES 797 /* PI waiters blocked on a rt_mutex held by this task: */ 798 struct rb_root pi_waiters; 799 struct rb_node *pi_waiters_leftmost; 800 /* Updated under owner's pi_lock and rq lock */ 801 struct task_struct *pi_top_task; 802 /* Deadlock detection and priority inheritance handling: */ 803 struct rt_mutex_waiter *pi_blocked_on; 804 #endif 805 806 #ifdef CONFIG_DEBUG_MUTEXES 807 /* Mutex deadlock detection: */ 808 struct mutex_waiter *blocked_on; 809 #endif 810 811 #ifdef CONFIG_TRACE_IRQFLAGS 812 unsigned int irq_events; 813 unsigned long hardirq_enable_ip; 814 unsigned long hardirq_disable_ip; 815 unsigned int hardirq_enable_event; 816 unsigned int hardirq_disable_event; 817 int hardirqs_enabled; 818 int hardirq_context; 819 unsigned long softirq_disable_ip; 820 unsigned long softirq_enable_ip; 821 unsigned int softirq_disable_event; 822 unsigned int softirq_enable_event; 823 int softirqs_enabled; 824 int softirq_context; 825 #endif 826 827 #ifdef CONFIG_LOCKDEP 828 # define MAX_LOCK_DEPTH 48UL 829 u64 curr_chain_key; 830 int lockdep_depth; 831 unsigned int lockdep_recursion; 832 struct held_lock held_locks[MAX_LOCK_DEPTH]; 833 gfp_t lockdep_reclaim_gfp; 834 #endif 835 836 #ifdef CONFIG_UBSAN 837 unsigned int in_ubsan; 838 #endif 839 840 /* Journalling filesystem info: */ 841 void *journal_info; 842 843 /* Stacked block device info: */ 844 struct bio_list *bio_list; 845 846 #ifdef CONFIG_BLOCK 847 /* Stack plugging: */ 848 struct blk_plug *plug; 849 #endif 850 851 /* VM state: */ 852 struct reclaim_state *reclaim_state; 853 854 struct backing_dev_info *backing_dev_info; 855 856 struct io_context *io_context; 857 858 /* Ptrace state: */ 859 unsigned long ptrace_message; 860 siginfo_t *last_siginfo; 861 862 struct task_io_accounting ioac; 863 #ifdef CONFIG_TASK_XACCT 864 /* Accumulated RSS usage: */ 865 u64 acct_rss_mem1; 866 /* Accumulated virtual memory usage: */ 867 u64 acct_vm_mem1; 868 /* stime + utime since last update: */ 869 u64 acct_timexpd; 870 #endif 871 #ifdef CONFIG_CPUSETS 872 /* Protected by ->alloc_lock: */ 873 nodemask_t mems_allowed; 874 /* Seqence number to catch updates: */ 875 seqcount_t mems_allowed_seq; 876 int cpuset_mem_spread_rotor; 877 int cpuset_slab_spread_rotor; 878 #endif 879 #ifdef CONFIG_CGROUPS 880 /* Control Group info protected by css_set_lock: */ 881 struct css_set __rcu *cgroups; 882 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 883 struct list_head cg_list; 884 #endif 885 #ifdef CONFIG_INTEL_RDT_A 886 int closid; 887 #endif 888 #ifdef CONFIG_FUTEX 889 struct robust_list_head __user *robust_list; 890 #ifdef CONFIG_COMPAT 891 struct compat_robust_list_head __user *compat_robust_list; 892 #endif 893 struct list_head pi_state_list; 894 struct futex_pi_state *pi_state_cache; 895 #endif 896 #ifdef CONFIG_PERF_EVENTS 897 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 898 struct mutex perf_event_mutex; 899 struct list_head perf_event_list; 900 #endif 901 #ifdef CONFIG_DEBUG_PREEMPT 902 unsigned long preempt_disable_ip; 903 #endif 904 #ifdef CONFIG_NUMA 905 /* Protected by alloc_lock: */ 906 struct mempolicy *mempolicy; 907 short il_next; 908 short pref_node_fork; 909 #endif 910 #ifdef CONFIG_NUMA_BALANCING 911 int numa_scan_seq; 912 unsigned int numa_scan_period; 913 unsigned int numa_scan_period_max; 914 int numa_preferred_nid; 915 unsigned long numa_migrate_retry; 916 /* Migration stamp: */ 917 u64 node_stamp; 918 u64 last_task_numa_placement; 919 u64 last_sum_exec_runtime; 920 struct callback_head numa_work; 921 922 struct list_head numa_entry; 923 struct numa_group *numa_group; 924 925 /* 926 * numa_faults is an array split into four regions: 927 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 928 * in this precise order. 929 * 930 * faults_memory: Exponential decaying average of faults on a per-node 931 * basis. Scheduling placement decisions are made based on these 932 * counts. The values remain static for the duration of a PTE scan. 933 * faults_cpu: Track the nodes the process was running on when a NUMA 934 * hinting fault was incurred. 935 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 936 * during the current scan window. When the scan completes, the counts 937 * in faults_memory and faults_cpu decay and these values are copied. 938 */ 939 unsigned long *numa_faults; 940 unsigned long total_numa_faults; 941 942 /* 943 * numa_faults_locality tracks if faults recorded during the last 944 * scan window were remote/local or failed to migrate. The task scan 945 * period is adapted based on the locality of the faults with different 946 * weights depending on whether they were shared or private faults 947 */ 948 unsigned long numa_faults_locality[3]; 949 950 unsigned long numa_pages_migrated; 951 #endif /* CONFIG_NUMA_BALANCING */ 952 953 struct tlbflush_unmap_batch tlb_ubc; 954 955 struct rcu_head rcu; 956 957 /* Cache last used pipe for splice(): */ 958 struct pipe_inode_info *splice_pipe; 959 960 struct page_frag task_frag; 961 962 #ifdef CONFIG_TASK_DELAY_ACCT 963 struct task_delay_info *delays; 964 #endif 965 966 #ifdef CONFIG_FAULT_INJECTION 967 int make_it_fail; 968 #endif 969 /* 970 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 971 * balance_dirty_pages() for a dirty throttling pause: 972 */ 973 int nr_dirtied; 974 int nr_dirtied_pause; 975 /* Start of a write-and-pause period: */ 976 unsigned long dirty_paused_when; 977 978 #ifdef CONFIG_LATENCYTOP 979 int latency_record_count; 980 struct latency_record latency_record[LT_SAVECOUNT]; 981 #endif 982 /* 983 * Time slack values; these are used to round up poll() and 984 * select() etc timeout values. These are in nanoseconds. 985 */ 986 u64 timer_slack_ns; 987 u64 default_timer_slack_ns; 988 989 #ifdef CONFIG_KASAN 990 unsigned int kasan_depth; 991 #endif 992 993 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 994 /* Index of current stored address in ret_stack: */ 995 int curr_ret_stack; 996 997 /* Stack of return addresses for return function tracing: */ 998 struct ftrace_ret_stack *ret_stack; 999 1000 /* Timestamp for last schedule: */ 1001 unsigned long long ftrace_timestamp; 1002 1003 /* 1004 * Number of functions that haven't been traced 1005 * because of depth overrun: 1006 */ 1007 atomic_t trace_overrun; 1008 1009 /* Pause tracing: */ 1010 atomic_t tracing_graph_pause; 1011 #endif 1012 1013 #ifdef CONFIG_TRACING 1014 /* State flags for use by tracers: */ 1015 unsigned long trace; 1016 1017 /* Bitmask and counter of trace recursion: */ 1018 unsigned long trace_recursion; 1019 #endif /* CONFIG_TRACING */ 1020 1021 #ifdef CONFIG_KCOV 1022 /* Coverage collection mode enabled for this task (0 if disabled): */ 1023 enum kcov_mode kcov_mode; 1024 1025 /* Size of the kcov_area: */ 1026 unsigned int kcov_size; 1027 1028 /* Buffer for coverage collection: */ 1029 void *kcov_area; 1030 1031 /* KCOV descriptor wired with this task or NULL: */ 1032 struct kcov *kcov; 1033 #endif 1034 1035 #ifdef CONFIG_MEMCG 1036 struct mem_cgroup *memcg_in_oom; 1037 gfp_t memcg_oom_gfp_mask; 1038 int memcg_oom_order; 1039 1040 /* Number of pages to reclaim on returning to userland: */ 1041 unsigned int memcg_nr_pages_over_high; 1042 #endif 1043 1044 #ifdef CONFIG_UPROBES 1045 struct uprobe_task *utask; 1046 #endif 1047 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1048 unsigned int sequential_io; 1049 unsigned int sequential_io_avg; 1050 #endif 1051 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1052 unsigned long task_state_change; 1053 #endif 1054 int pagefault_disabled; 1055 #ifdef CONFIG_MMU 1056 struct task_struct *oom_reaper_list; 1057 #endif 1058 #ifdef CONFIG_VMAP_STACK 1059 struct vm_struct *stack_vm_area; 1060 #endif 1061 #ifdef CONFIG_THREAD_INFO_IN_TASK 1062 /* A live task holds one reference: */ 1063 atomic_t stack_refcount; 1064 #endif 1065 #ifdef CONFIG_LIVEPATCH 1066 int patch_state; 1067 #endif 1068 #ifdef CONFIG_SECURITY 1069 /* Used by LSM modules for access restriction: */ 1070 void *security; 1071 #endif 1072 /* CPU-specific state of this task: */ 1073 struct thread_struct thread; 1074 1075 /* 1076 * WARNING: on x86, 'thread_struct' contains a variable-sized 1077 * structure. It *MUST* be at the end of 'task_struct'. 1078 * 1079 * Do not put anything below here! 1080 */ 1081 }; 1082 1083 static inline struct pid *task_pid(struct task_struct *task) 1084 { 1085 return task->pids[PIDTYPE_PID].pid; 1086 } 1087 1088 static inline struct pid *task_tgid(struct task_struct *task) 1089 { 1090 return task->group_leader->pids[PIDTYPE_PID].pid; 1091 } 1092 1093 /* 1094 * Without tasklist or RCU lock it is not safe to dereference 1095 * the result of task_pgrp/task_session even if task == current, 1096 * we can race with another thread doing sys_setsid/sys_setpgid. 1097 */ 1098 static inline struct pid *task_pgrp(struct task_struct *task) 1099 { 1100 return task->group_leader->pids[PIDTYPE_PGID].pid; 1101 } 1102 1103 static inline struct pid *task_session(struct task_struct *task) 1104 { 1105 return task->group_leader->pids[PIDTYPE_SID].pid; 1106 } 1107 1108 /* 1109 * the helpers to get the task's different pids as they are seen 1110 * from various namespaces 1111 * 1112 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1113 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1114 * current. 1115 * task_xid_nr_ns() : id seen from the ns specified; 1116 * 1117 * see also pid_nr() etc in include/linux/pid.h 1118 */ 1119 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1120 1121 static inline pid_t task_pid_nr(struct task_struct *tsk) 1122 { 1123 return tsk->pid; 1124 } 1125 1126 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1127 { 1128 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1129 } 1130 1131 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1132 { 1133 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1134 } 1135 1136 1137 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1138 { 1139 return tsk->tgid; 1140 } 1141 1142 extern pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns); 1143 1144 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1145 { 1146 return pid_vnr(task_tgid(tsk)); 1147 } 1148 1149 /** 1150 * pid_alive - check that a task structure is not stale 1151 * @p: Task structure to be checked. 1152 * 1153 * Test if a process is not yet dead (at most zombie state) 1154 * If pid_alive fails, then pointers within the task structure 1155 * can be stale and must not be dereferenced. 1156 * 1157 * Return: 1 if the process is alive. 0 otherwise. 1158 */ 1159 static inline int pid_alive(const struct task_struct *p) 1160 { 1161 return p->pids[PIDTYPE_PID].pid != NULL; 1162 } 1163 1164 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1165 { 1166 pid_t pid = 0; 1167 1168 rcu_read_lock(); 1169 if (pid_alive(tsk)) 1170 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1171 rcu_read_unlock(); 1172 1173 return pid; 1174 } 1175 1176 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1177 { 1178 return task_ppid_nr_ns(tsk, &init_pid_ns); 1179 } 1180 1181 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1182 { 1183 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1184 } 1185 1186 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1187 { 1188 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1189 } 1190 1191 1192 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1193 { 1194 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1195 } 1196 1197 static inline pid_t task_session_vnr(struct task_struct *tsk) 1198 { 1199 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1200 } 1201 1202 /* Obsolete, do not use: */ 1203 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1204 { 1205 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1206 } 1207 1208 /** 1209 * is_global_init - check if a task structure is init. Since init 1210 * is free to have sub-threads we need to check tgid. 1211 * @tsk: Task structure to be checked. 1212 * 1213 * Check if a task structure is the first user space task the kernel created. 1214 * 1215 * Return: 1 if the task structure is init. 0 otherwise. 1216 */ 1217 static inline int is_global_init(struct task_struct *tsk) 1218 { 1219 return task_tgid_nr(tsk) == 1; 1220 } 1221 1222 extern struct pid *cad_pid; 1223 1224 /* 1225 * Per process flags 1226 */ 1227 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1228 #define PF_EXITING 0x00000004 /* Getting shut down */ 1229 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */ 1230 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */ 1231 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1232 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1233 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1234 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1235 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1236 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1237 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1238 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1239 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1240 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1241 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1242 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1243 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1244 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1245 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1246 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ 1247 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1248 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1249 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1250 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */ 1251 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1252 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */ 1253 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1254 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1255 1256 /* 1257 * Only the _current_ task can read/write to tsk->flags, but other 1258 * tasks can access tsk->flags in readonly mode for example 1259 * with tsk_used_math (like during threaded core dumping). 1260 * There is however an exception to this rule during ptrace 1261 * or during fork: the ptracer task is allowed to write to the 1262 * child->flags of its traced child (same goes for fork, the parent 1263 * can write to the child->flags), because we're guaranteed the 1264 * child is not running and in turn not changing child->flags 1265 * at the same time the parent does it. 1266 */ 1267 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1268 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1269 #define clear_used_math() clear_stopped_child_used_math(current) 1270 #define set_used_math() set_stopped_child_used_math(current) 1271 1272 #define conditional_stopped_child_used_math(condition, child) \ 1273 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1274 1275 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1276 1277 #define copy_to_stopped_child_used_math(child) \ 1278 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1279 1280 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1281 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1282 #define used_math() tsk_used_math(current) 1283 1284 static inline bool is_percpu_thread(void) 1285 { 1286 #ifdef CONFIG_SMP 1287 return (current->flags & PF_NO_SETAFFINITY) && 1288 (current->nr_cpus_allowed == 1); 1289 #else 1290 return true; 1291 #endif 1292 } 1293 1294 /* Per-process atomic flags. */ 1295 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1296 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1297 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1298 1299 1300 #define TASK_PFA_TEST(name, func) \ 1301 static inline bool task_##func(struct task_struct *p) \ 1302 { return test_bit(PFA_##name, &p->atomic_flags); } 1303 1304 #define TASK_PFA_SET(name, func) \ 1305 static inline void task_set_##func(struct task_struct *p) \ 1306 { set_bit(PFA_##name, &p->atomic_flags); } 1307 1308 #define TASK_PFA_CLEAR(name, func) \ 1309 static inline void task_clear_##func(struct task_struct *p) \ 1310 { clear_bit(PFA_##name, &p->atomic_flags); } 1311 1312 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1313 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1314 1315 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1316 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1317 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1318 1319 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1320 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1321 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1322 1323 static inline void 1324 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1325 { 1326 current->flags &= ~flags; 1327 current->flags |= orig_flags & flags; 1328 } 1329 1330 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1331 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1332 #ifdef CONFIG_SMP 1333 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1334 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1335 #else 1336 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1337 { 1338 } 1339 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1340 { 1341 if (!cpumask_test_cpu(0, new_mask)) 1342 return -EINVAL; 1343 return 0; 1344 } 1345 #endif 1346 1347 #ifndef cpu_relax_yield 1348 #define cpu_relax_yield() cpu_relax() 1349 #endif 1350 1351 extern int yield_to(struct task_struct *p, bool preempt); 1352 extern void set_user_nice(struct task_struct *p, long nice); 1353 extern int task_prio(const struct task_struct *p); 1354 1355 /** 1356 * task_nice - return the nice value of a given task. 1357 * @p: the task in question. 1358 * 1359 * Return: The nice value [ -20 ... 0 ... 19 ]. 1360 */ 1361 static inline int task_nice(const struct task_struct *p) 1362 { 1363 return PRIO_TO_NICE((p)->static_prio); 1364 } 1365 1366 extern int can_nice(const struct task_struct *p, const int nice); 1367 extern int task_curr(const struct task_struct *p); 1368 extern int idle_cpu(int cpu); 1369 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1370 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1371 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1372 extern struct task_struct *idle_task(int cpu); 1373 1374 /** 1375 * is_idle_task - is the specified task an idle task? 1376 * @p: the task in question. 1377 * 1378 * Return: 1 if @p is an idle task. 0 otherwise. 1379 */ 1380 static inline bool is_idle_task(const struct task_struct *p) 1381 { 1382 return !!(p->flags & PF_IDLE); 1383 } 1384 1385 extern struct task_struct *curr_task(int cpu); 1386 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1387 1388 void yield(void); 1389 1390 union thread_union { 1391 #ifndef CONFIG_THREAD_INFO_IN_TASK 1392 struct thread_info thread_info; 1393 #endif 1394 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1395 }; 1396 1397 #ifdef CONFIG_THREAD_INFO_IN_TASK 1398 static inline struct thread_info *task_thread_info(struct task_struct *task) 1399 { 1400 return &task->thread_info; 1401 } 1402 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1403 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1404 #endif 1405 1406 /* 1407 * find a task by one of its numerical ids 1408 * 1409 * find_task_by_pid_ns(): 1410 * finds a task by its pid in the specified namespace 1411 * find_task_by_vpid(): 1412 * finds a task by its virtual pid 1413 * 1414 * see also find_vpid() etc in include/linux/pid.h 1415 */ 1416 1417 extern struct task_struct *find_task_by_vpid(pid_t nr); 1418 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1419 1420 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1421 extern int wake_up_process(struct task_struct *tsk); 1422 extern void wake_up_new_task(struct task_struct *tsk); 1423 1424 #ifdef CONFIG_SMP 1425 extern void kick_process(struct task_struct *tsk); 1426 #else 1427 static inline void kick_process(struct task_struct *tsk) { } 1428 #endif 1429 1430 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1431 1432 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1433 { 1434 __set_task_comm(tsk, from, false); 1435 } 1436 1437 extern char *get_task_comm(char *to, struct task_struct *tsk); 1438 1439 #ifdef CONFIG_SMP 1440 void scheduler_ipi(void); 1441 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 1442 #else 1443 static inline void scheduler_ipi(void) { } 1444 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) 1445 { 1446 return 1; 1447 } 1448 #endif 1449 1450 /* 1451 * Set thread flags in other task's structures. 1452 * See asm/thread_info.h for TIF_xxxx flags available: 1453 */ 1454 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1455 { 1456 set_ti_thread_flag(task_thread_info(tsk), flag); 1457 } 1458 1459 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1460 { 1461 clear_ti_thread_flag(task_thread_info(tsk), flag); 1462 } 1463 1464 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1465 { 1466 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1467 } 1468 1469 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1470 { 1471 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1472 } 1473 1474 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1475 { 1476 return test_ti_thread_flag(task_thread_info(tsk), flag); 1477 } 1478 1479 static inline void set_tsk_need_resched(struct task_struct *tsk) 1480 { 1481 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1482 } 1483 1484 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1485 { 1486 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1487 } 1488 1489 static inline int test_tsk_need_resched(struct task_struct *tsk) 1490 { 1491 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1492 } 1493 1494 /* 1495 * cond_resched() and cond_resched_lock(): latency reduction via 1496 * explicit rescheduling in places that are safe. The return 1497 * value indicates whether a reschedule was done in fact. 1498 * cond_resched_lock() will drop the spinlock before scheduling, 1499 * cond_resched_softirq() will enable bhs before scheduling. 1500 */ 1501 #ifndef CONFIG_PREEMPT 1502 extern int _cond_resched(void); 1503 #else 1504 static inline int _cond_resched(void) { return 0; } 1505 #endif 1506 1507 #define cond_resched() ({ \ 1508 ___might_sleep(__FILE__, __LINE__, 0); \ 1509 _cond_resched(); \ 1510 }) 1511 1512 extern int __cond_resched_lock(spinlock_t *lock); 1513 1514 #define cond_resched_lock(lock) ({ \ 1515 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ 1516 __cond_resched_lock(lock); \ 1517 }) 1518 1519 extern int __cond_resched_softirq(void); 1520 1521 #define cond_resched_softirq() ({ \ 1522 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \ 1523 __cond_resched_softirq(); \ 1524 }) 1525 1526 static inline void cond_resched_rcu(void) 1527 { 1528 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 1529 rcu_read_unlock(); 1530 cond_resched(); 1531 rcu_read_lock(); 1532 #endif 1533 } 1534 1535 /* 1536 * Does a critical section need to be broken due to another 1537 * task waiting?: (technically does not depend on CONFIG_PREEMPT, 1538 * but a general need for low latency) 1539 */ 1540 static inline int spin_needbreak(spinlock_t *lock) 1541 { 1542 #ifdef CONFIG_PREEMPT 1543 return spin_is_contended(lock); 1544 #else 1545 return 0; 1546 #endif 1547 } 1548 1549 static __always_inline bool need_resched(void) 1550 { 1551 return unlikely(tif_need_resched()); 1552 } 1553 1554 /* 1555 * Wrappers for p->thread_info->cpu access. No-op on UP. 1556 */ 1557 #ifdef CONFIG_SMP 1558 1559 static inline unsigned int task_cpu(const struct task_struct *p) 1560 { 1561 #ifdef CONFIG_THREAD_INFO_IN_TASK 1562 return p->cpu; 1563 #else 1564 return task_thread_info(p)->cpu; 1565 #endif 1566 } 1567 1568 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 1569 1570 #else 1571 1572 static inline unsigned int task_cpu(const struct task_struct *p) 1573 { 1574 return 0; 1575 } 1576 1577 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 1578 { 1579 } 1580 1581 #endif /* CONFIG_SMP */ 1582 1583 /* 1584 * In order to reduce various lock holder preemption latencies provide an 1585 * interface to see if a vCPU is currently running or not. 1586 * 1587 * This allows us to terminate optimistic spin loops and block, analogous to 1588 * the native optimistic spin heuristic of testing if the lock owner task is 1589 * running or not. 1590 */ 1591 #ifndef vcpu_is_preempted 1592 # define vcpu_is_preempted(cpu) false 1593 #endif 1594 1595 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 1596 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 1597 1598 #ifndef TASK_SIZE_OF 1599 #define TASK_SIZE_OF(tsk) TASK_SIZE 1600 #endif 1601 1602 #endif 1603