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 0x0000 69 #define TASK_INTERRUPTIBLE 0x0001 70 #define TASK_UNINTERRUPTIBLE 0x0002 71 #define __TASK_STOPPED 0x0004 72 #define __TASK_TRACED 0x0008 73 /* Used in tsk->exit_state: */ 74 #define EXIT_DEAD 0x0010 75 #define EXIT_ZOMBIE 0x0020 76 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 77 /* Used in tsk->state again: */ 78 #define TASK_PARKED 0x0040 79 #define TASK_DEAD 0x0080 80 #define TASK_WAKEKILL 0x0100 81 #define TASK_WAKING 0x0200 82 #define TASK_NOLOAD 0x0400 83 #define TASK_NEW 0x0800 84 #define TASK_STATE_MAX 0x1000 85 86 /* Convenience macros for the sake of set_current_state: */ 87 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 88 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 89 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) 90 91 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 92 93 /* Convenience macros for the sake of wake_up(): */ 94 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 95 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED) 96 97 /* get_task_state(): */ 98 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 99 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 100 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 101 TASK_PARKED) 102 103 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0) 104 105 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0) 106 107 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) 108 109 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \ 110 (task->flags & PF_FROZEN) == 0 && \ 111 (task->state & TASK_NOLOAD) == 0) 112 113 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 114 115 #define __set_current_state(state_value) \ 116 do { \ 117 current->task_state_change = _THIS_IP_; \ 118 current->state = (state_value); \ 119 } while (0) 120 #define set_current_state(state_value) \ 121 do { \ 122 current->task_state_change = _THIS_IP_; \ 123 smp_store_mb(current->state, (state_value)); \ 124 } while (0) 125 126 #else 127 /* 128 * set_current_state() includes a barrier so that the write of current->state 129 * is correctly serialised wrt the caller's subsequent test of whether to 130 * actually sleep: 131 * 132 * for (;;) { 133 * set_current_state(TASK_UNINTERRUPTIBLE); 134 * if (!need_sleep) 135 * break; 136 * 137 * schedule(); 138 * } 139 * __set_current_state(TASK_RUNNING); 140 * 141 * If the caller does not need such serialisation (because, for instance, the 142 * condition test and condition change and wakeup are under the same lock) then 143 * use __set_current_state(). 144 * 145 * The above is typically ordered against the wakeup, which does: 146 * 147 * need_sleep = false; 148 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 149 * 150 * Where wake_up_state() (and all other wakeup primitives) imply enough 151 * barriers to order the store of the variable against wakeup. 152 * 153 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 154 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 155 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 156 * 157 * This is obviously fine, since they both store the exact same value. 158 * 159 * Also see the comments of try_to_wake_up(). 160 */ 161 #define __set_current_state(state_value) do { current->state = (state_value); } while (0) 162 #define set_current_state(state_value) smp_store_mb(current->state, (state_value)) 163 #endif 164 165 /* Task command name length: */ 166 #define TASK_COMM_LEN 16 167 168 extern cpumask_var_t cpu_isolated_map; 169 170 extern void scheduler_tick(void); 171 172 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 173 174 extern long schedule_timeout(long timeout); 175 extern long schedule_timeout_interruptible(long timeout); 176 extern long schedule_timeout_killable(long timeout); 177 extern long schedule_timeout_uninterruptible(long timeout); 178 extern long schedule_timeout_idle(long timeout); 179 asmlinkage void schedule(void); 180 extern void schedule_preempt_disabled(void); 181 182 extern int __must_check io_schedule_prepare(void); 183 extern void io_schedule_finish(int token); 184 extern long io_schedule_timeout(long timeout); 185 extern void io_schedule(void); 186 187 /** 188 * struct prev_cputime - snapshot of system and user cputime 189 * @utime: time spent in user mode 190 * @stime: time spent in system mode 191 * @lock: protects the above two fields 192 * 193 * Stores previous user/system time values such that we can guarantee 194 * monotonicity. 195 */ 196 struct prev_cputime { 197 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 198 u64 utime; 199 u64 stime; 200 raw_spinlock_t lock; 201 #endif 202 }; 203 204 /** 205 * struct task_cputime - collected CPU time counts 206 * @utime: time spent in user mode, in nanoseconds 207 * @stime: time spent in kernel mode, in nanoseconds 208 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds 209 * 210 * This structure groups together three kinds of CPU time that are tracked for 211 * threads and thread groups. Most things considering CPU time want to group 212 * these counts together and treat all three of them in parallel. 213 */ 214 struct task_cputime { 215 u64 utime; 216 u64 stime; 217 unsigned long long sum_exec_runtime; 218 }; 219 220 /* Alternate field names when used on cache expirations: */ 221 #define virt_exp utime 222 #define prof_exp stime 223 #define sched_exp sum_exec_runtime 224 225 enum vtime_state { 226 /* Task is sleeping or running in a CPU with VTIME inactive: */ 227 VTIME_INACTIVE = 0, 228 /* Task runs in userspace in a CPU with VTIME active: */ 229 VTIME_USER, 230 /* Task runs in kernelspace in a CPU with VTIME active: */ 231 VTIME_SYS, 232 }; 233 234 struct vtime { 235 seqcount_t seqcount; 236 unsigned long long starttime; 237 enum vtime_state state; 238 u64 utime; 239 u64 stime; 240 u64 gtime; 241 }; 242 243 struct sched_info { 244 #ifdef CONFIG_SCHED_INFO 245 /* Cumulative counters: */ 246 247 /* # of times we have run on this CPU: */ 248 unsigned long pcount; 249 250 /* Time spent waiting on a runqueue: */ 251 unsigned long long run_delay; 252 253 /* Timestamps: */ 254 255 /* When did we last run on a CPU? */ 256 unsigned long long last_arrival; 257 258 /* When were we last queued to run? */ 259 unsigned long long last_queued; 260 261 #endif /* CONFIG_SCHED_INFO */ 262 }; 263 264 /* 265 * Integer metrics need fixed point arithmetic, e.g., sched/fair 266 * has a few: load, load_avg, util_avg, freq, and capacity. 267 * 268 * We define a basic fixed point arithmetic range, and then formalize 269 * all these metrics based on that basic range. 270 */ 271 # define SCHED_FIXEDPOINT_SHIFT 10 272 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 273 274 struct load_weight { 275 unsigned long weight; 276 u32 inv_weight; 277 }; 278 279 /* 280 * The load_avg/util_avg accumulates an infinite geometric series 281 * (see __update_load_avg() in kernel/sched/fair.c). 282 * 283 * [load_avg definition] 284 * 285 * load_avg = runnable% * scale_load_down(load) 286 * 287 * where runnable% is the time ratio that a sched_entity is runnable. 288 * For cfs_rq, it is the aggregated load_avg of all runnable and 289 * blocked sched_entities. 290 * 291 * load_avg may also take frequency scaling into account: 292 * 293 * load_avg = runnable% * scale_load_down(load) * freq% 294 * 295 * where freq% is the CPU frequency normalized to the highest frequency. 296 * 297 * [util_avg definition] 298 * 299 * util_avg = running% * SCHED_CAPACITY_SCALE 300 * 301 * where running% is the time ratio that a sched_entity is running on 302 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable 303 * and blocked sched_entities. 304 * 305 * util_avg may also factor frequency scaling and CPU capacity scaling: 306 * 307 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity% 308 * 309 * where freq% is the same as above, and capacity% is the CPU capacity 310 * normalized to the greatest capacity (due to uarch differences, etc). 311 * 312 * N.B., the above ratios (runnable%, running%, freq%, and capacity%) 313 * themselves are in the range of [0, 1]. To do fixed point arithmetics, 314 * we therefore scale them to as large a range as necessary. This is for 315 * example reflected by util_avg's SCHED_CAPACITY_SCALE. 316 * 317 * [Overflow issue] 318 * 319 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 320 * with the highest load (=88761), always runnable on a single cfs_rq, 321 * and should not overflow as the number already hits PID_MAX_LIMIT. 322 * 323 * For all other cases (including 32-bit kernels), struct load_weight's 324 * weight will overflow first before we do, because: 325 * 326 * Max(load_avg) <= Max(load.weight) 327 * 328 * Then it is the load_weight's responsibility to consider overflow 329 * issues. 330 */ 331 struct sched_avg { 332 u64 last_update_time; 333 u64 load_sum; 334 u32 util_sum; 335 u32 period_contrib; 336 unsigned long load_avg; 337 unsigned long util_avg; 338 }; 339 340 struct sched_statistics { 341 #ifdef CONFIG_SCHEDSTATS 342 u64 wait_start; 343 u64 wait_max; 344 u64 wait_count; 345 u64 wait_sum; 346 u64 iowait_count; 347 u64 iowait_sum; 348 349 u64 sleep_start; 350 u64 sleep_max; 351 s64 sum_sleep_runtime; 352 353 u64 block_start; 354 u64 block_max; 355 u64 exec_max; 356 u64 slice_max; 357 358 u64 nr_migrations_cold; 359 u64 nr_failed_migrations_affine; 360 u64 nr_failed_migrations_running; 361 u64 nr_failed_migrations_hot; 362 u64 nr_forced_migrations; 363 364 u64 nr_wakeups; 365 u64 nr_wakeups_sync; 366 u64 nr_wakeups_migrate; 367 u64 nr_wakeups_local; 368 u64 nr_wakeups_remote; 369 u64 nr_wakeups_affine; 370 u64 nr_wakeups_affine_attempts; 371 u64 nr_wakeups_passive; 372 u64 nr_wakeups_idle; 373 #endif 374 }; 375 376 struct sched_entity { 377 /* For load-balancing: */ 378 struct load_weight load; 379 struct rb_node run_node; 380 struct list_head group_node; 381 unsigned int on_rq; 382 383 u64 exec_start; 384 u64 sum_exec_runtime; 385 u64 vruntime; 386 u64 prev_sum_exec_runtime; 387 388 u64 nr_migrations; 389 390 struct sched_statistics statistics; 391 392 #ifdef CONFIG_FAIR_GROUP_SCHED 393 int depth; 394 struct sched_entity *parent; 395 /* rq on which this entity is (to be) queued: */ 396 struct cfs_rq *cfs_rq; 397 /* rq "owned" by this entity/group: */ 398 struct cfs_rq *my_q; 399 #endif 400 401 #ifdef CONFIG_SMP 402 /* 403 * Per entity load average tracking. 404 * 405 * Put into separate cache line so it does not 406 * collide with read-mostly values above. 407 */ 408 struct sched_avg avg ____cacheline_aligned_in_smp; 409 #endif 410 }; 411 412 struct sched_rt_entity { 413 struct list_head run_list; 414 unsigned long timeout; 415 unsigned long watchdog_stamp; 416 unsigned int time_slice; 417 unsigned short on_rq; 418 unsigned short on_list; 419 420 struct sched_rt_entity *back; 421 #ifdef CONFIG_RT_GROUP_SCHED 422 struct sched_rt_entity *parent; 423 /* rq on which this entity is (to be) queued: */ 424 struct rt_rq *rt_rq; 425 /* rq "owned" by this entity/group: */ 426 struct rt_rq *my_q; 427 #endif 428 } __randomize_layout; 429 430 struct sched_dl_entity { 431 struct rb_node rb_node; 432 433 /* 434 * Original scheduling parameters. Copied here from sched_attr 435 * during sched_setattr(), they will remain the same until 436 * the next sched_setattr(). 437 */ 438 u64 dl_runtime; /* Maximum runtime for each instance */ 439 u64 dl_deadline; /* Relative deadline of each instance */ 440 u64 dl_period; /* Separation of two instances (period) */ 441 u64 dl_bw; /* dl_runtime / dl_period */ 442 u64 dl_density; /* dl_runtime / dl_deadline */ 443 444 /* 445 * Actual scheduling parameters. Initialized with the values above, 446 * they are continously updated during task execution. Note that 447 * the remaining runtime could be < 0 in case we are in overrun. 448 */ 449 s64 runtime; /* Remaining runtime for this instance */ 450 u64 deadline; /* Absolute deadline for this instance */ 451 unsigned int flags; /* Specifying the scheduler behaviour */ 452 453 /* 454 * Some bool flags: 455 * 456 * @dl_throttled tells if we exhausted the runtime. If so, the 457 * task has to wait for a replenishment to be performed at the 458 * next firing of dl_timer. 459 * 460 * @dl_boosted tells if we are boosted due to DI. If so we are 461 * outside bandwidth enforcement mechanism (but only until we 462 * exit the critical section); 463 * 464 * @dl_yielded tells if task gave up the CPU before consuming 465 * all its available runtime during the last job. 466 * 467 * @dl_non_contending tells if the task is inactive while still 468 * contributing to the active utilization. In other words, it 469 * indicates if the inactive timer has been armed and its handler 470 * has not been executed yet. This flag is useful to avoid race 471 * conditions between the inactive timer handler and the wakeup 472 * code. 473 */ 474 int dl_throttled; 475 int dl_boosted; 476 int dl_yielded; 477 int dl_non_contending; 478 479 /* 480 * Bandwidth enforcement timer. Each -deadline task has its 481 * own bandwidth to be enforced, thus we need one timer per task. 482 */ 483 struct hrtimer dl_timer; 484 485 /* 486 * Inactive timer, responsible for decreasing the active utilization 487 * at the "0-lag time". When a -deadline task blocks, it contributes 488 * to GRUB's active utilization until the "0-lag time", hence a 489 * timer is needed to decrease the active utilization at the correct 490 * time. 491 */ 492 struct hrtimer inactive_timer; 493 }; 494 495 union rcu_special { 496 struct { 497 u8 blocked; 498 u8 need_qs; 499 u8 exp_need_qs; 500 501 /* Otherwise the compiler can store garbage here: */ 502 u8 pad; 503 } b; /* Bits. */ 504 u32 s; /* Set of bits. */ 505 }; 506 507 enum perf_event_task_context { 508 perf_invalid_context = -1, 509 perf_hw_context = 0, 510 perf_sw_context, 511 perf_nr_task_contexts, 512 }; 513 514 struct wake_q_node { 515 struct wake_q_node *next; 516 }; 517 518 struct task_struct { 519 #ifdef CONFIG_THREAD_INFO_IN_TASK 520 /* 521 * For reasons of header soup (see current_thread_info()), this 522 * must be the first element of task_struct. 523 */ 524 struct thread_info thread_info; 525 #endif 526 /* -1 unrunnable, 0 runnable, >0 stopped: */ 527 volatile long state; 528 529 /* 530 * This begins the randomizable portion of task_struct. Only 531 * scheduling-critical items should be added above here. 532 */ 533 randomized_struct_fields_start 534 535 void *stack; 536 atomic_t usage; 537 /* Per task flags (PF_*), defined further below: */ 538 unsigned int flags; 539 unsigned int ptrace; 540 541 #ifdef CONFIG_SMP 542 struct llist_node wake_entry; 543 int on_cpu; 544 #ifdef CONFIG_THREAD_INFO_IN_TASK 545 /* Current CPU: */ 546 unsigned int cpu; 547 #endif 548 unsigned int wakee_flips; 549 unsigned long wakee_flip_decay_ts; 550 struct task_struct *last_wakee; 551 552 int wake_cpu; 553 #endif 554 int on_rq; 555 556 int prio; 557 int static_prio; 558 int normal_prio; 559 unsigned int rt_priority; 560 561 const struct sched_class *sched_class; 562 struct sched_entity se; 563 struct sched_rt_entity rt; 564 #ifdef CONFIG_CGROUP_SCHED 565 struct task_group *sched_task_group; 566 #endif 567 struct sched_dl_entity dl; 568 569 #ifdef CONFIG_PREEMPT_NOTIFIERS 570 /* List of struct preempt_notifier: */ 571 struct hlist_head preempt_notifiers; 572 #endif 573 574 #ifdef CONFIG_BLK_DEV_IO_TRACE 575 unsigned int btrace_seq; 576 #endif 577 578 unsigned int policy; 579 int nr_cpus_allowed; 580 cpumask_t cpus_allowed; 581 582 #ifdef CONFIG_PREEMPT_RCU 583 int rcu_read_lock_nesting; 584 union rcu_special rcu_read_unlock_special; 585 struct list_head rcu_node_entry; 586 struct rcu_node *rcu_blocked_node; 587 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 588 589 #ifdef CONFIG_TASKS_RCU 590 unsigned long rcu_tasks_nvcsw; 591 u8 rcu_tasks_holdout; 592 u8 rcu_tasks_idx; 593 int rcu_tasks_idle_cpu; 594 struct list_head rcu_tasks_holdout_list; 595 #endif /* #ifdef CONFIG_TASKS_RCU */ 596 597 struct sched_info sched_info; 598 599 struct list_head tasks; 600 #ifdef CONFIG_SMP 601 struct plist_node pushable_tasks; 602 struct rb_node pushable_dl_tasks; 603 #endif 604 605 struct mm_struct *mm; 606 struct mm_struct *active_mm; 607 608 /* Per-thread vma caching: */ 609 struct vmacache vmacache; 610 611 #ifdef SPLIT_RSS_COUNTING 612 struct task_rss_stat rss_stat; 613 #endif 614 int exit_state; 615 int exit_code; 616 int exit_signal; 617 /* The signal sent when the parent dies: */ 618 int pdeath_signal; 619 /* JOBCTL_*, siglock protected: */ 620 unsigned long jobctl; 621 622 /* Used for emulating ABI behavior of previous Linux versions: */ 623 unsigned int personality; 624 625 /* Scheduler bits, serialized by scheduler locks: */ 626 unsigned sched_reset_on_fork:1; 627 unsigned sched_contributes_to_load:1; 628 unsigned sched_migrated:1; 629 unsigned sched_remote_wakeup:1; 630 /* Force alignment to the next boundary: */ 631 unsigned :0; 632 633 /* Unserialized, strictly 'current' */ 634 635 /* Bit to tell LSMs we're in execve(): */ 636 unsigned in_execve:1; 637 unsigned in_iowait:1; 638 #ifndef TIF_RESTORE_SIGMASK 639 unsigned restore_sigmask:1; 640 #endif 641 #ifdef CONFIG_MEMCG 642 unsigned memcg_may_oom:1; 643 #ifndef CONFIG_SLOB 644 unsigned memcg_kmem_skip_account:1; 645 #endif 646 #endif 647 #ifdef CONFIG_COMPAT_BRK 648 unsigned brk_randomized:1; 649 #endif 650 #ifdef CONFIG_CGROUPS 651 /* disallow userland-initiated cgroup migration */ 652 unsigned no_cgroup_migration:1; 653 #endif 654 655 unsigned long atomic_flags; /* Flags requiring atomic access. */ 656 657 struct restart_block restart_block; 658 659 pid_t pid; 660 pid_t tgid; 661 662 #ifdef CONFIG_CC_STACKPROTECTOR 663 /* Canary value for the -fstack-protector GCC feature: */ 664 unsigned long stack_canary; 665 #endif 666 /* 667 * Pointers to the (original) parent process, youngest child, younger sibling, 668 * older sibling, respectively. (p->father can be replaced with 669 * p->real_parent->pid) 670 */ 671 672 /* Real parent process: */ 673 struct task_struct __rcu *real_parent; 674 675 /* Recipient of SIGCHLD, wait4() reports: */ 676 struct task_struct __rcu *parent; 677 678 /* 679 * Children/sibling form the list of natural children: 680 */ 681 struct list_head children; 682 struct list_head sibling; 683 struct task_struct *group_leader; 684 685 /* 686 * 'ptraced' is the list of tasks this task is using ptrace() on. 687 * 688 * This includes both natural children and PTRACE_ATTACH targets. 689 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 690 */ 691 struct list_head ptraced; 692 struct list_head ptrace_entry; 693 694 /* PID/PID hash table linkage. */ 695 struct pid_link pids[PIDTYPE_MAX]; 696 struct list_head thread_group; 697 struct list_head thread_node; 698 699 struct completion *vfork_done; 700 701 /* CLONE_CHILD_SETTID: */ 702 int __user *set_child_tid; 703 704 /* CLONE_CHILD_CLEARTID: */ 705 int __user *clear_child_tid; 706 707 u64 utime; 708 u64 stime; 709 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 710 u64 utimescaled; 711 u64 stimescaled; 712 #endif 713 u64 gtime; 714 struct prev_cputime prev_cputime; 715 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 716 struct vtime vtime; 717 #endif 718 719 #ifdef CONFIG_NO_HZ_FULL 720 atomic_t tick_dep_mask; 721 #endif 722 /* Context switch counts: */ 723 unsigned long nvcsw; 724 unsigned long nivcsw; 725 726 /* Monotonic time in nsecs: */ 727 u64 start_time; 728 729 /* Boot based time in nsecs: */ 730 u64 real_start_time; 731 732 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 733 unsigned long min_flt; 734 unsigned long maj_flt; 735 736 #ifdef CONFIG_POSIX_TIMERS 737 struct task_cputime cputime_expires; 738 struct list_head cpu_timers[3]; 739 #endif 740 741 /* Process credentials: */ 742 743 /* Tracer's credentials at attach: */ 744 const struct cred __rcu *ptracer_cred; 745 746 /* Objective and real subjective task credentials (COW): */ 747 const struct cred __rcu *real_cred; 748 749 /* Effective (overridable) subjective task credentials (COW): */ 750 const struct cred __rcu *cred; 751 752 /* 753 * executable name, excluding path. 754 * 755 * - normally initialized setup_new_exec() 756 * - access it with [gs]et_task_comm() 757 * - lock it with task_lock() 758 */ 759 char comm[TASK_COMM_LEN]; 760 761 struct nameidata *nameidata; 762 763 #ifdef CONFIG_SYSVIPC 764 struct sysv_sem sysvsem; 765 struct sysv_shm sysvshm; 766 #endif 767 #ifdef CONFIG_DETECT_HUNG_TASK 768 unsigned long last_switch_count; 769 #endif 770 /* Filesystem information: */ 771 struct fs_struct *fs; 772 773 /* Open file information: */ 774 struct files_struct *files; 775 776 /* Namespaces: */ 777 struct nsproxy *nsproxy; 778 779 /* Signal handlers: */ 780 struct signal_struct *signal; 781 struct sighand_struct *sighand; 782 sigset_t blocked; 783 sigset_t real_blocked; 784 /* Restored if set_restore_sigmask() was used: */ 785 sigset_t saved_sigmask; 786 struct sigpending pending; 787 unsigned long sas_ss_sp; 788 size_t sas_ss_size; 789 unsigned int sas_ss_flags; 790 791 struct callback_head *task_works; 792 793 struct audit_context *audit_context; 794 #ifdef CONFIG_AUDITSYSCALL 795 kuid_t loginuid; 796 unsigned int sessionid; 797 #endif 798 struct seccomp seccomp; 799 800 /* Thread group tracking: */ 801 u32 parent_exec_id; 802 u32 self_exec_id; 803 804 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 805 spinlock_t alloc_lock; 806 807 /* Protection of the PI data structures: */ 808 raw_spinlock_t pi_lock; 809 810 struct wake_q_node wake_q; 811 812 #ifdef CONFIG_RT_MUTEXES 813 /* PI waiters blocked on a rt_mutex held by this task: */ 814 struct rb_root_cached pi_waiters; 815 /* Updated under owner's pi_lock and rq lock */ 816 struct task_struct *pi_top_task; 817 /* Deadlock detection and priority inheritance handling: */ 818 struct rt_mutex_waiter *pi_blocked_on; 819 #endif 820 821 #ifdef CONFIG_DEBUG_MUTEXES 822 /* Mutex deadlock detection: */ 823 struct mutex_waiter *blocked_on; 824 #endif 825 826 #ifdef CONFIG_TRACE_IRQFLAGS 827 unsigned int irq_events; 828 unsigned long hardirq_enable_ip; 829 unsigned long hardirq_disable_ip; 830 unsigned int hardirq_enable_event; 831 unsigned int hardirq_disable_event; 832 int hardirqs_enabled; 833 int hardirq_context; 834 unsigned long softirq_disable_ip; 835 unsigned long softirq_enable_ip; 836 unsigned int softirq_disable_event; 837 unsigned int softirq_enable_event; 838 int softirqs_enabled; 839 int softirq_context; 840 #endif 841 842 #ifdef CONFIG_LOCKDEP 843 # define MAX_LOCK_DEPTH 48UL 844 u64 curr_chain_key; 845 int lockdep_depth; 846 unsigned int lockdep_recursion; 847 struct held_lock held_locks[MAX_LOCK_DEPTH]; 848 #endif 849 850 #ifdef CONFIG_LOCKDEP_CROSSRELEASE 851 #define MAX_XHLOCKS_NR 64UL 852 struct hist_lock *xhlocks; /* Crossrelease history locks */ 853 unsigned int xhlock_idx; 854 /* For restoring at history boundaries */ 855 unsigned int xhlock_idx_hist[XHLOCK_CTX_NR]; 856 unsigned int hist_id; 857 /* For overwrite check at each context exit */ 858 unsigned int hist_id_save[XHLOCK_CTX_NR]; 859 #endif 860 861 #ifdef CONFIG_UBSAN 862 unsigned int in_ubsan; 863 #endif 864 865 /* Journalling filesystem info: */ 866 void *journal_info; 867 868 /* Stacked block device info: */ 869 struct bio_list *bio_list; 870 871 #ifdef CONFIG_BLOCK 872 /* Stack plugging: */ 873 struct blk_plug *plug; 874 #endif 875 876 /* VM state: */ 877 struct reclaim_state *reclaim_state; 878 879 struct backing_dev_info *backing_dev_info; 880 881 struct io_context *io_context; 882 883 /* Ptrace state: */ 884 unsigned long ptrace_message; 885 siginfo_t *last_siginfo; 886 887 struct task_io_accounting ioac; 888 #ifdef CONFIG_TASK_XACCT 889 /* Accumulated RSS usage: */ 890 u64 acct_rss_mem1; 891 /* Accumulated virtual memory usage: */ 892 u64 acct_vm_mem1; 893 /* stime + utime since last update: */ 894 u64 acct_timexpd; 895 #endif 896 #ifdef CONFIG_CPUSETS 897 /* Protected by ->alloc_lock: */ 898 nodemask_t mems_allowed; 899 /* Seqence number to catch updates: */ 900 seqcount_t mems_allowed_seq; 901 int cpuset_mem_spread_rotor; 902 int cpuset_slab_spread_rotor; 903 #endif 904 #ifdef CONFIG_CGROUPS 905 /* Control Group info protected by css_set_lock: */ 906 struct css_set __rcu *cgroups; 907 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 908 struct list_head cg_list; 909 #endif 910 #ifdef CONFIG_INTEL_RDT 911 u32 closid; 912 u32 rmid; 913 #endif 914 #ifdef CONFIG_FUTEX 915 struct robust_list_head __user *robust_list; 916 #ifdef CONFIG_COMPAT 917 struct compat_robust_list_head __user *compat_robust_list; 918 #endif 919 struct list_head pi_state_list; 920 struct futex_pi_state *pi_state_cache; 921 #endif 922 #ifdef CONFIG_PERF_EVENTS 923 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 924 struct mutex perf_event_mutex; 925 struct list_head perf_event_list; 926 #endif 927 #ifdef CONFIG_DEBUG_PREEMPT 928 unsigned long preempt_disable_ip; 929 #endif 930 #ifdef CONFIG_NUMA 931 /* Protected by alloc_lock: */ 932 struct mempolicy *mempolicy; 933 short il_prev; 934 short pref_node_fork; 935 #endif 936 #ifdef CONFIG_NUMA_BALANCING 937 int numa_scan_seq; 938 unsigned int numa_scan_period; 939 unsigned int numa_scan_period_max; 940 int numa_preferred_nid; 941 unsigned long numa_migrate_retry; 942 /* Migration stamp: */ 943 u64 node_stamp; 944 u64 last_task_numa_placement; 945 u64 last_sum_exec_runtime; 946 struct callback_head numa_work; 947 948 struct list_head numa_entry; 949 struct numa_group *numa_group; 950 951 /* 952 * numa_faults is an array split into four regions: 953 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 954 * in this precise order. 955 * 956 * faults_memory: Exponential decaying average of faults on a per-node 957 * basis. Scheduling placement decisions are made based on these 958 * counts. The values remain static for the duration of a PTE scan. 959 * faults_cpu: Track the nodes the process was running on when a NUMA 960 * hinting fault was incurred. 961 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 962 * during the current scan window. When the scan completes, the counts 963 * in faults_memory and faults_cpu decay and these values are copied. 964 */ 965 unsigned long *numa_faults; 966 unsigned long total_numa_faults; 967 968 /* 969 * numa_faults_locality tracks if faults recorded during the last 970 * scan window were remote/local or failed to migrate. The task scan 971 * period is adapted based on the locality of the faults with different 972 * weights depending on whether they were shared or private faults 973 */ 974 unsigned long numa_faults_locality[3]; 975 976 unsigned long numa_pages_migrated; 977 #endif /* CONFIG_NUMA_BALANCING */ 978 979 struct tlbflush_unmap_batch tlb_ubc; 980 981 struct rcu_head rcu; 982 983 /* Cache last used pipe for splice(): */ 984 struct pipe_inode_info *splice_pipe; 985 986 struct page_frag task_frag; 987 988 #ifdef CONFIG_TASK_DELAY_ACCT 989 struct task_delay_info *delays; 990 #endif 991 992 #ifdef CONFIG_FAULT_INJECTION 993 int make_it_fail; 994 unsigned int fail_nth; 995 #endif 996 /* 997 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 998 * balance_dirty_pages() for a dirty throttling pause: 999 */ 1000 int nr_dirtied; 1001 int nr_dirtied_pause; 1002 /* Start of a write-and-pause period: */ 1003 unsigned long dirty_paused_when; 1004 1005 #ifdef CONFIG_LATENCYTOP 1006 int latency_record_count; 1007 struct latency_record latency_record[LT_SAVECOUNT]; 1008 #endif 1009 /* 1010 * Time slack values; these are used to round up poll() and 1011 * select() etc timeout values. These are in nanoseconds. 1012 */ 1013 u64 timer_slack_ns; 1014 u64 default_timer_slack_ns; 1015 1016 #ifdef CONFIG_KASAN 1017 unsigned int kasan_depth; 1018 #endif 1019 1020 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1021 /* Index of current stored address in ret_stack: */ 1022 int curr_ret_stack; 1023 1024 /* Stack of return addresses for return function tracing: */ 1025 struct ftrace_ret_stack *ret_stack; 1026 1027 /* Timestamp for last schedule: */ 1028 unsigned long long ftrace_timestamp; 1029 1030 /* 1031 * Number of functions that haven't been traced 1032 * because of depth overrun: 1033 */ 1034 atomic_t trace_overrun; 1035 1036 /* Pause tracing: */ 1037 atomic_t tracing_graph_pause; 1038 #endif 1039 1040 #ifdef CONFIG_TRACING 1041 /* State flags for use by tracers: */ 1042 unsigned long trace; 1043 1044 /* Bitmask and counter of trace recursion: */ 1045 unsigned long trace_recursion; 1046 #endif /* CONFIG_TRACING */ 1047 1048 #ifdef CONFIG_KCOV 1049 /* Coverage collection mode enabled for this task (0 if disabled): */ 1050 enum kcov_mode kcov_mode; 1051 1052 /* Size of the kcov_area: */ 1053 unsigned int kcov_size; 1054 1055 /* Buffer for coverage collection: */ 1056 void *kcov_area; 1057 1058 /* KCOV descriptor wired with this task or NULL: */ 1059 struct kcov *kcov; 1060 #endif 1061 1062 #ifdef CONFIG_MEMCG 1063 struct mem_cgroup *memcg_in_oom; 1064 gfp_t memcg_oom_gfp_mask; 1065 int memcg_oom_order; 1066 1067 /* Number of pages to reclaim on returning to userland: */ 1068 unsigned int memcg_nr_pages_over_high; 1069 #endif 1070 1071 #ifdef CONFIG_UPROBES 1072 struct uprobe_task *utask; 1073 #endif 1074 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1075 unsigned int sequential_io; 1076 unsigned int sequential_io_avg; 1077 #endif 1078 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1079 unsigned long task_state_change; 1080 #endif 1081 int pagefault_disabled; 1082 #ifdef CONFIG_MMU 1083 struct task_struct *oom_reaper_list; 1084 #endif 1085 #ifdef CONFIG_VMAP_STACK 1086 struct vm_struct *stack_vm_area; 1087 #endif 1088 #ifdef CONFIG_THREAD_INFO_IN_TASK 1089 /* A live task holds one reference: */ 1090 atomic_t stack_refcount; 1091 #endif 1092 #ifdef CONFIG_LIVEPATCH 1093 int patch_state; 1094 #endif 1095 #ifdef CONFIG_SECURITY 1096 /* Used by LSM modules for access restriction: */ 1097 void *security; 1098 #endif 1099 1100 /* 1101 * New fields for task_struct should be added above here, so that 1102 * they are included in the randomized portion of task_struct. 1103 */ 1104 randomized_struct_fields_end 1105 1106 /* CPU-specific state of this task: */ 1107 struct thread_struct thread; 1108 1109 /* 1110 * WARNING: on x86, 'thread_struct' contains a variable-sized 1111 * structure. It *MUST* be at the end of 'task_struct'. 1112 * 1113 * Do not put anything below here! 1114 */ 1115 }; 1116 1117 static inline struct pid *task_pid(struct task_struct *task) 1118 { 1119 return task->pids[PIDTYPE_PID].pid; 1120 } 1121 1122 static inline struct pid *task_tgid(struct task_struct *task) 1123 { 1124 return task->group_leader->pids[PIDTYPE_PID].pid; 1125 } 1126 1127 /* 1128 * Without tasklist or RCU lock it is not safe to dereference 1129 * the result of task_pgrp/task_session even if task == current, 1130 * we can race with another thread doing sys_setsid/sys_setpgid. 1131 */ 1132 static inline struct pid *task_pgrp(struct task_struct *task) 1133 { 1134 return task->group_leader->pids[PIDTYPE_PGID].pid; 1135 } 1136 1137 static inline struct pid *task_session(struct task_struct *task) 1138 { 1139 return task->group_leader->pids[PIDTYPE_SID].pid; 1140 } 1141 1142 /* 1143 * the helpers to get the task's different pids as they are seen 1144 * from various namespaces 1145 * 1146 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1147 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1148 * current. 1149 * task_xid_nr_ns() : id seen from the ns specified; 1150 * 1151 * see also pid_nr() etc in include/linux/pid.h 1152 */ 1153 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1154 1155 static inline pid_t task_pid_nr(struct task_struct *tsk) 1156 { 1157 return tsk->pid; 1158 } 1159 1160 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1161 { 1162 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1163 } 1164 1165 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1166 { 1167 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1168 } 1169 1170 1171 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1172 { 1173 return tsk->tgid; 1174 } 1175 1176 /** 1177 * pid_alive - check that a task structure is not stale 1178 * @p: Task structure to be checked. 1179 * 1180 * Test if a process is not yet dead (at most zombie state) 1181 * If pid_alive fails, then pointers within the task structure 1182 * can be stale and must not be dereferenced. 1183 * 1184 * Return: 1 if the process is alive. 0 otherwise. 1185 */ 1186 static inline int pid_alive(const struct task_struct *p) 1187 { 1188 return p->pids[PIDTYPE_PID].pid != NULL; 1189 } 1190 1191 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1192 { 1193 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1194 } 1195 1196 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1197 { 1198 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1199 } 1200 1201 1202 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1203 { 1204 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1205 } 1206 1207 static inline pid_t task_session_vnr(struct task_struct *tsk) 1208 { 1209 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1210 } 1211 1212 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1213 { 1214 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns); 1215 } 1216 1217 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1218 { 1219 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL); 1220 } 1221 1222 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1223 { 1224 pid_t pid = 0; 1225 1226 rcu_read_lock(); 1227 if (pid_alive(tsk)) 1228 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1229 rcu_read_unlock(); 1230 1231 return pid; 1232 } 1233 1234 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1235 { 1236 return task_ppid_nr_ns(tsk, &init_pid_ns); 1237 } 1238 1239 /* Obsolete, do not use: */ 1240 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1241 { 1242 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1243 } 1244 1245 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1246 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1247 1248 static inline unsigned int __get_task_state(struct task_struct *tsk) 1249 { 1250 unsigned int tsk_state = READ_ONCE(tsk->state); 1251 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT; 1252 1253 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1254 1255 if (tsk_state == TASK_IDLE) 1256 state = TASK_REPORT_IDLE; 1257 1258 return fls(state); 1259 } 1260 1261 static inline char __task_state_to_char(unsigned int state) 1262 { 1263 static const char state_char[] = "RSDTtXZPI"; 1264 1265 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1266 1267 return state_char[state]; 1268 } 1269 1270 static inline char task_state_to_char(struct task_struct *tsk) 1271 { 1272 return __task_state_to_char(__get_task_state(tsk)); 1273 } 1274 1275 /** 1276 * is_global_init - check if a task structure is init. Since init 1277 * is free to have sub-threads we need to check tgid. 1278 * @tsk: Task structure to be checked. 1279 * 1280 * Check if a task structure is the first user space task the kernel created. 1281 * 1282 * Return: 1 if the task structure is init. 0 otherwise. 1283 */ 1284 static inline int is_global_init(struct task_struct *tsk) 1285 { 1286 return task_tgid_nr(tsk) == 1; 1287 } 1288 1289 extern struct pid *cad_pid; 1290 1291 /* 1292 * Per process flags 1293 */ 1294 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1295 #define PF_EXITING 0x00000004 /* Getting shut down */ 1296 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */ 1297 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */ 1298 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1299 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1300 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1301 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1302 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1303 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1304 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1305 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1306 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1307 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1308 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1309 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1310 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1311 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1312 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1313 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ 1314 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1315 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1316 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1317 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */ 1318 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1319 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */ 1320 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1321 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1322 1323 /* 1324 * Only the _current_ task can read/write to tsk->flags, but other 1325 * tasks can access tsk->flags in readonly mode for example 1326 * with tsk_used_math (like during threaded core dumping). 1327 * There is however an exception to this rule during ptrace 1328 * or during fork: the ptracer task is allowed to write to the 1329 * child->flags of its traced child (same goes for fork, the parent 1330 * can write to the child->flags), because we're guaranteed the 1331 * child is not running and in turn not changing child->flags 1332 * at the same time the parent does it. 1333 */ 1334 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1335 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1336 #define clear_used_math() clear_stopped_child_used_math(current) 1337 #define set_used_math() set_stopped_child_used_math(current) 1338 1339 #define conditional_stopped_child_used_math(condition, child) \ 1340 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1341 1342 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1343 1344 #define copy_to_stopped_child_used_math(child) \ 1345 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1346 1347 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1348 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1349 #define used_math() tsk_used_math(current) 1350 1351 static inline bool is_percpu_thread(void) 1352 { 1353 #ifdef CONFIG_SMP 1354 return (current->flags & PF_NO_SETAFFINITY) && 1355 (current->nr_cpus_allowed == 1); 1356 #else 1357 return true; 1358 #endif 1359 } 1360 1361 /* Per-process atomic flags. */ 1362 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1363 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1364 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1365 1366 1367 #define TASK_PFA_TEST(name, func) \ 1368 static inline bool task_##func(struct task_struct *p) \ 1369 { return test_bit(PFA_##name, &p->atomic_flags); } 1370 1371 #define TASK_PFA_SET(name, func) \ 1372 static inline void task_set_##func(struct task_struct *p) \ 1373 { set_bit(PFA_##name, &p->atomic_flags); } 1374 1375 #define TASK_PFA_CLEAR(name, func) \ 1376 static inline void task_clear_##func(struct task_struct *p) \ 1377 { clear_bit(PFA_##name, &p->atomic_flags); } 1378 1379 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1380 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1381 1382 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1383 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1384 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1385 1386 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1387 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1388 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1389 1390 static inline void 1391 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1392 { 1393 current->flags &= ~flags; 1394 current->flags |= orig_flags & flags; 1395 } 1396 1397 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1398 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1399 #ifdef CONFIG_SMP 1400 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1401 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1402 #else 1403 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1404 { 1405 } 1406 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1407 { 1408 if (!cpumask_test_cpu(0, new_mask)) 1409 return -EINVAL; 1410 return 0; 1411 } 1412 #endif 1413 1414 #ifndef cpu_relax_yield 1415 #define cpu_relax_yield() cpu_relax() 1416 #endif 1417 1418 extern int yield_to(struct task_struct *p, bool preempt); 1419 extern void set_user_nice(struct task_struct *p, long nice); 1420 extern int task_prio(const struct task_struct *p); 1421 1422 /** 1423 * task_nice - return the nice value of a given task. 1424 * @p: the task in question. 1425 * 1426 * Return: The nice value [ -20 ... 0 ... 19 ]. 1427 */ 1428 static inline int task_nice(const struct task_struct *p) 1429 { 1430 return PRIO_TO_NICE((p)->static_prio); 1431 } 1432 1433 extern int can_nice(const struct task_struct *p, const int nice); 1434 extern int task_curr(const struct task_struct *p); 1435 extern int idle_cpu(int cpu); 1436 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1437 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1438 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1439 extern struct task_struct *idle_task(int cpu); 1440 1441 /** 1442 * is_idle_task - is the specified task an idle task? 1443 * @p: the task in question. 1444 * 1445 * Return: 1 if @p is an idle task. 0 otherwise. 1446 */ 1447 static inline bool is_idle_task(const struct task_struct *p) 1448 { 1449 return !!(p->flags & PF_IDLE); 1450 } 1451 1452 extern struct task_struct *curr_task(int cpu); 1453 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1454 1455 void yield(void); 1456 1457 union thread_union { 1458 #ifndef CONFIG_THREAD_INFO_IN_TASK 1459 struct thread_info thread_info; 1460 #endif 1461 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1462 }; 1463 1464 #ifdef CONFIG_THREAD_INFO_IN_TASK 1465 static inline struct thread_info *task_thread_info(struct task_struct *task) 1466 { 1467 return &task->thread_info; 1468 } 1469 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1470 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1471 #endif 1472 1473 /* 1474 * find a task by one of its numerical ids 1475 * 1476 * find_task_by_pid_ns(): 1477 * finds a task by its pid in the specified namespace 1478 * find_task_by_vpid(): 1479 * finds a task by its virtual pid 1480 * 1481 * see also find_vpid() etc in include/linux/pid.h 1482 */ 1483 1484 extern struct task_struct *find_task_by_vpid(pid_t nr); 1485 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1486 1487 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1488 extern int wake_up_process(struct task_struct *tsk); 1489 extern void wake_up_new_task(struct task_struct *tsk); 1490 1491 #ifdef CONFIG_SMP 1492 extern void kick_process(struct task_struct *tsk); 1493 #else 1494 static inline void kick_process(struct task_struct *tsk) { } 1495 #endif 1496 1497 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1498 1499 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1500 { 1501 __set_task_comm(tsk, from, false); 1502 } 1503 1504 extern char *get_task_comm(char *to, struct task_struct *tsk); 1505 1506 #ifdef CONFIG_SMP 1507 void scheduler_ipi(void); 1508 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 1509 #else 1510 static inline void scheduler_ipi(void) { } 1511 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) 1512 { 1513 return 1; 1514 } 1515 #endif 1516 1517 /* 1518 * Set thread flags in other task's structures. 1519 * See asm/thread_info.h for TIF_xxxx flags available: 1520 */ 1521 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1522 { 1523 set_ti_thread_flag(task_thread_info(tsk), flag); 1524 } 1525 1526 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1527 { 1528 clear_ti_thread_flag(task_thread_info(tsk), flag); 1529 } 1530 1531 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1532 { 1533 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1534 } 1535 1536 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1537 { 1538 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1539 } 1540 1541 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1542 { 1543 return test_ti_thread_flag(task_thread_info(tsk), flag); 1544 } 1545 1546 static inline void set_tsk_need_resched(struct task_struct *tsk) 1547 { 1548 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1549 } 1550 1551 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1552 { 1553 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1554 } 1555 1556 static inline int test_tsk_need_resched(struct task_struct *tsk) 1557 { 1558 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1559 } 1560 1561 /* 1562 * cond_resched() and cond_resched_lock(): latency reduction via 1563 * explicit rescheduling in places that are safe. The return 1564 * value indicates whether a reschedule was done in fact. 1565 * cond_resched_lock() will drop the spinlock before scheduling, 1566 * cond_resched_softirq() will enable bhs before scheduling. 1567 */ 1568 #ifndef CONFIG_PREEMPT 1569 extern int _cond_resched(void); 1570 #else 1571 static inline int _cond_resched(void) { return 0; } 1572 #endif 1573 1574 #define cond_resched() ({ \ 1575 ___might_sleep(__FILE__, __LINE__, 0); \ 1576 _cond_resched(); \ 1577 }) 1578 1579 extern int __cond_resched_lock(spinlock_t *lock); 1580 1581 #define cond_resched_lock(lock) ({ \ 1582 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ 1583 __cond_resched_lock(lock); \ 1584 }) 1585 1586 extern int __cond_resched_softirq(void); 1587 1588 #define cond_resched_softirq() ({ \ 1589 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \ 1590 __cond_resched_softirq(); \ 1591 }) 1592 1593 static inline void cond_resched_rcu(void) 1594 { 1595 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 1596 rcu_read_unlock(); 1597 cond_resched(); 1598 rcu_read_lock(); 1599 #endif 1600 } 1601 1602 /* 1603 * Does a critical section need to be broken due to another 1604 * task waiting?: (technically does not depend on CONFIG_PREEMPT, 1605 * but a general need for low latency) 1606 */ 1607 static inline int spin_needbreak(spinlock_t *lock) 1608 { 1609 #ifdef CONFIG_PREEMPT 1610 return spin_is_contended(lock); 1611 #else 1612 return 0; 1613 #endif 1614 } 1615 1616 static __always_inline bool need_resched(void) 1617 { 1618 return unlikely(tif_need_resched()); 1619 } 1620 1621 /* 1622 * Wrappers for p->thread_info->cpu access. No-op on UP. 1623 */ 1624 #ifdef CONFIG_SMP 1625 1626 static inline unsigned int task_cpu(const struct task_struct *p) 1627 { 1628 #ifdef CONFIG_THREAD_INFO_IN_TASK 1629 return p->cpu; 1630 #else 1631 return task_thread_info(p)->cpu; 1632 #endif 1633 } 1634 1635 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 1636 1637 #else 1638 1639 static inline unsigned int task_cpu(const struct task_struct *p) 1640 { 1641 return 0; 1642 } 1643 1644 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 1645 { 1646 } 1647 1648 #endif /* CONFIG_SMP */ 1649 1650 /* 1651 * In order to reduce various lock holder preemption latencies provide an 1652 * interface to see if a vCPU is currently running or not. 1653 * 1654 * This allows us to terminate optimistic spin loops and block, analogous to 1655 * the native optimistic spin heuristic of testing if the lock owner task is 1656 * running or not. 1657 */ 1658 #ifndef vcpu_is_preempted 1659 # define vcpu_is_preempted(cpu) false 1660 #endif 1661 1662 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 1663 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 1664 1665 #ifndef TASK_SIZE_OF 1666 #define TASK_SIZE_OF(tsk) TASK_SIZE 1667 #endif 1668 1669 #endif 1670