1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_SCHED_H 3 #define _LINUX_SCHED_H 4 5 /* 6 * Define 'struct task_struct' and provide the main scheduler 7 * APIs (schedule(), wakeup variants, etc.) 8 */ 9 10 #include <uapi/linux/sched.h> 11 12 #include <asm/current.h> 13 14 #include <linux/pid.h> 15 #include <linux/sem.h> 16 #include <linux/shm.h> 17 #include <linux/kcov.h> 18 #include <linux/mutex.h> 19 #include <linux/plist.h> 20 #include <linux/hrtimer.h> 21 #include <linux/seccomp.h> 22 #include <linux/nodemask.h> 23 #include <linux/rcupdate.h> 24 #include <linux/resource.h> 25 #include <linux/latencytop.h> 26 #include <linux/sched/prio.h> 27 #include <linux/signal_types.h> 28 #include <linux/mm_types_task.h> 29 #include <linux/task_io_accounting.h> 30 #include <linux/rseq.h> 31 32 /* task_struct member predeclarations (sorted alphabetically): */ 33 struct audit_context; 34 struct backing_dev_info; 35 struct bio_list; 36 struct blk_plug; 37 struct cfs_rq; 38 struct fs_struct; 39 struct futex_pi_state; 40 struct io_context; 41 struct mempolicy; 42 struct nameidata; 43 struct nsproxy; 44 struct perf_event_context; 45 struct pid_namespace; 46 struct pipe_inode_info; 47 struct rcu_node; 48 struct reclaim_state; 49 struct robust_list_head; 50 struct sched_attr; 51 struct sched_param; 52 struct seq_file; 53 struct sighand_struct; 54 struct signal_struct; 55 struct task_delay_info; 56 struct task_group; 57 58 /* 59 * Task state bitmask. NOTE! These bits are also 60 * encoded in fs/proc/array.c: get_task_state(). 61 * 62 * We have two separate sets of flags: task->state 63 * is about runnability, while task->exit_state are 64 * about the task exiting. Confusing, but this way 65 * modifying one set can't modify the other one by 66 * mistake. 67 */ 68 69 /* Used in tsk->state: */ 70 #define TASK_RUNNING 0x0000 71 #define TASK_INTERRUPTIBLE 0x0001 72 #define TASK_UNINTERRUPTIBLE 0x0002 73 #define __TASK_STOPPED 0x0004 74 #define __TASK_TRACED 0x0008 75 /* Used in tsk->exit_state: */ 76 #define EXIT_DEAD 0x0010 77 #define EXIT_ZOMBIE 0x0020 78 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 79 /* Used in tsk->state again: */ 80 #define TASK_PARKED 0x0040 81 #define TASK_DEAD 0x0080 82 #define TASK_WAKEKILL 0x0100 83 #define TASK_WAKING 0x0200 84 #define TASK_NOLOAD 0x0400 85 #define TASK_NEW 0x0800 86 #define TASK_STATE_MAX 0x1000 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 98 /* get_task_state(): */ 99 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 100 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 101 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 102 TASK_PARKED) 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 /* 117 * Special states are those that do not use the normal wait-loop pattern. See 118 * the comment with set_special_state(). 119 */ 120 #define is_special_task_state(state) \ 121 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD)) 122 123 #define __set_current_state(state_value) \ 124 do { \ 125 WARN_ON_ONCE(is_special_task_state(state_value));\ 126 current->task_state_change = _THIS_IP_; \ 127 current->state = (state_value); \ 128 } while (0) 129 130 #define set_current_state(state_value) \ 131 do { \ 132 WARN_ON_ONCE(is_special_task_state(state_value));\ 133 current->task_state_change = _THIS_IP_; \ 134 smp_store_mb(current->state, (state_value)); \ 135 } while (0) 136 137 #define set_special_state(state_value) \ 138 do { \ 139 unsigned long flags; /* may shadow */ \ 140 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 141 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 142 current->task_state_change = _THIS_IP_; \ 143 current->state = (state_value); \ 144 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 145 } while (0) 146 #else 147 /* 148 * set_current_state() includes a barrier so that the write of current->state 149 * is correctly serialised wrt the caller's subsequent test of whether to 150 * actually sleep: 151 * 152 * for (;;) { 153 * set_current_state(TASK_UNINTERRUPTIBLE); 154 * if (!need_sleep) 155 * break; 156 * 157 * schedule(); 158 * } 159 * __set_current_state(TASK_RUNNING); 160 * 161 * If the caller does not need such serialisation (because, for instance, the 162 * condition test and condition change and wakeup are under the same lock) then 163 * use __set_current_state(). 164 * 165 * The above is typically ordered against the wakeup, which does: 166 * 167 * need_sleep = false; 168 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 169 * 170 * Where wake_up_state() (and all other wakeup primitives) imply enough 171 * barriers to order the store of the variable against wakeup. 172 * 173 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 174 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 175 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 176 * 177 * However, with slightly different timing the wakeup TASK_RUNNING store can 178 * also collide with the TASK_UNINTERRUPTIBLE store. Loosing that store is not 179 * a problem either because that will result in one extra go around the loop 180 * and our @cond test will save the day. 181 * 182 * Also see the comments of try_to_wake_up(). 183 */ 184 #define __set_current_state(state_value) \ 185 current->state = (state_value) 186 187 #define set_current_state(state_value) \ 188 smp_store_mb(current->state, (state_value)) 189 190 /* 191 * set_special_state() should be used for those states when the blocking task 192 * can not use the regular condition based wait-loop. In that case we must 193 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores 194 * will not collide with our state change. 195 */ 196 #define set_special_state(state_value) \ 197 do { \ 198 unsigned long flags; /* may shadow */ \ 199 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 200 current->state = (state_value); \ 201 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 202 } while (0) 203 204 #endif 205 206 /* Task command name length: */ 207 #define TASK_COMM_LEN 16 208 209 extern void scheduler_tick(void); 210 211 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 212 213 extern long schedule_timeout(long timeout); 214 extern long schedule_timeout_interruptible(long timeout); 215 extern long schedule_timeout_killable(long timeout); 216 extern long schedule_timeout_uninterruptible(long timeout); 217 extern long schedule_timeout_idle(long timeout); 218 asmlinkage void schedule(void); 219 extern void schedule_preempt_disabled(void); 220 221 extern int __must_check io_schedule_prepare(void); 222 extern void io_schedule_finish(int token); 223 extern long io_schedule_timeout(long timeout); 224 extern void io_schedule(void); 225 226 /** 227 * struct prev_cputime - snapshot of system and user cputime 228 * @utime: time spent in user mode 229 * @stime: time spent in system mode 230 * @lock: protects the above two fields 231 * 232 * Stores previous user/system time values such that we can guarantee 233 * monotonicity. 234 */ 235 struct prev_cputime { 236 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 237 u64 utime; 238 u64 stime; 239 raw_spinlock_t lock; 240 #endif 241 }; 242 243 /** 244 * struct task_cputime - collected CPU time counts 245 * @utime: time spent in user mode, in nanoseconds 246 * @stime: time spent in kernel mode, in nanoseconds 247 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds 248 * 249 * This structure groups together three kinds of CPU time that are tracked for 250 * threads and thread groups. Most things considering CPU time want to group 251 * these counts together and treat all three of them in parallel. 252 */ 253 struct task_cputime { 254 u64 utime; 255 u64 stime; 256 unsigned long long sum_exec_runtime; 257 }; 258 259 /* Alternate field names when used on cache expirations: */ 260 #define virt_exp utime 261 #define prof_exp stime 262 #define sched_exp sum_exec_runtime 263 264 enum vtime_state { 265 /* Task is sleeping or running in a CPU with VTIME inactive: */ 266 VTIME_INACTIVE = 0, 267 /* Task runs in userspace in a CPU with VTIME active: */ 268 VTIME_USER, 269 /* Task runs in kernelspace in a CPU with VTIME active: */ 270 VTIME_SYS, 271 }; 272 273 struct vtime { 274 seqcount_t seqcount; 275 unsigned long long starttime; 276 enum vtime_state state; 277 u64 utime; 278 u64 stime; 279 u64 gtime; 280 }; 281 282 struct sched_info { 283 #ifdef CONFIG_SCHED_INFO 284 /* Cumulative counters: */ 285 286 /* # of times we have run on this CPU: */ 287 unsigned long pcount; 288 289 /* Time spent waiting on a runqueue: */ 290 unsigned long long run_delay; 291 292 /* Timestamps: */ 293 294 /* When did we last run on a CPU? */ 295 unsigned long long last_arrival; 296 297 /* When were we last queued to run? */ 298 unsigned long long last_queued; 299 300 #endif /* CONFIG_SCHED_INFO */ 301 }; 302 303 /* 304 * Integer metrics need fixed point arithmetic, e.g., sched/fair 305 * has a few: load, load_avg, util_avg, freq, and capacity. 306 * 307 * We define a basic fixed point arithmetic range, and then formalize 308 * all these metrics based on that basic range. 309 */ 310 # define SCHED_FIXEDPOINT_SHIFT 10 311 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 312 313 struct load_weight { 314 unsigned long weight; 315 u32 inv_weight; 316 }; 317 318 /** 319 * struct util_est - Estimation utilization of FAIR tasks 320 * @enqueued: instantaneous estimated utilization of a task/cpu 321 * @ewma: the Exponential Weighted Moving Average (EWMA) 322 * utilization of a task 323 * 324 * Support data structure to track an Exponential Weighted Moving Average 325 * (EWMA) of a FAIR task's utilization. New samples are added to the moving 326 * average each time a task completes an activation. Sample's weight is chosen 327 * so that the EWMA will be relatively insensitive to transient changes to the 328 * task's workload. 329 * 330 * The enqueued attribute has a slightly different meaning for tasks and cpus: 331 * - task: the task's util_avg at last task dequeue time 332 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU 333 * Thus, the util_est.enqueued of a task represents the contribution on the 334 * estimated utilization of the CPU where that task is currently enqueued. 335 * 336 * Only for tasks we track a moving average of the past instantaneous 337 * estimated utilization. This allows to absorb sporadic drops in utilization 338 * of an otherwise almost periodic task. 339 */ 340 struct util_est { 341 unsigned int enqueued; 342 unsigned int ewma; 343 #define UTIL_EST_WEIGHT_SHIFT 2 344 } __attribute__((__aligned__(sizeof(u64)))); 345 346 /* 347 * The load_avg/util_avg accumulates an infinite geometric series 348 * (see __update_load_avg() in kernel/sched/fair.c). 349 * 350 * [load_avg definition] 351 * 352 * load_avg = runnable% * scale_load_down(load) 353 * 354 * where runnable% is the time ratio that a sched_entity is runnable. 355 * For cfs_rq, it is the aggregated load_avg of all runnable and 356 * blocked sched_entities. 357 * 358 * load_avg may also take frequency scaling into account: 359 * 360 * load_avg = runnable% * scale_load_down(load) * freq% 361 * 362 * where freq% is the CPU frequency normalized to the highest frequency. 363 * 364 * [util_avg definition] 365 * 366 * util_avg = running% * SCHED_CAPACITY_SCALE 367 * 368 * where running% is the time ratio that a sched_entity is running on 369 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable 370 * and blocked sched_entities. 371 * 372 * util_avg may also factor frequency scaling and CPU capacity scaling: 373 * 374 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity% 375 * 376 * where freq% is the same as above, and capacity% is the CPU capacity 377 * normalized to the greatest capacity (due to uarch differences, etc). 378 * 379 * N.B., the above ratios (runnable%, running%, freq%, and capacity%) 380 * themselves are in the range of [0, 1]. To do fixed point arithmetics, 381 * we therefore scale them to as large a range as necessary. This is for 382 * example reflected by util_avg's SCHED_CAPACITY_SCALE. 383 * 384 * [Overflow issue] 385 * 386 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 387 * with the highest load (=88761), always runnable on a single cfs_rq, 388 * and should not overflow as the number already hits PID_MAX_LIMIT. 389 * 390 * For all other cases (including 32-bit kernels), struct load_weight's 391 * weight will overflow first before we do, because: 392 * 393 * Max(load_avg) <= Max(load.weight) 394 * 395 * Then it is the load_weight's responsibility to consider overflow 396 * issues. 397 */ 398 struct sched_avg { 399 u64 last_update_time; 400 u64 load_sum; 401 u64 runnable_load_sum; 402 u32 util_sum; 403 u32 period_contrib; 404 unsigned long load_avg; 405 unsigned long runnable_load_avg; 406 unsigned long util_avg; 407 struct util_est util_est; 408 } ____cacheline_aligned; 409 410 struct sched_statistics { 411 #ifdef CONFIG_SCHEDSTATS 412 u64 wait_start; 413 u64 wait_max; 414 u64 wait_count; 415 u64 wait_sum; 416 u64 iowait_count; 417 u64 iowait_sum; 418 419 u64 sleep_start; 420 u64 sleep_max; 421 s64 sum_sleep_runtime; 422 423 u64 block_start; 424 u64 block_max; 425 u64 exec_max; 426 u64 slice_max; 427 428 u64 nr_migrations_cold; 429 u64 nr_failed_migrations_affine; 430 u64 nr_failed_migrations_running; 431 u64 nr_failed_migrations_hot; 432 u64 nr_forced_migrations; 433 434 u64 nr_wakeups; 435 u64 nr_wakeups_sync; 436 u64 nr_wakeups_migrate; 437 u64 nr_wakeups_local; 438 u64 nr_wakeups_remote; 439 u64 nr_wakeups_affine; 440 u64 nr_wakeups_affine_attempts; 441 u64 nr_wakeups_passive; 442 u64 nr_wakeups_idle; 443 #endif 444 }; 445 446 struct sched_entity { 447 /* For load-balancing: */ 448 struct load_weight load; 449 unsigned long runnable_weight; 450 struct rb_node run_node; 451 struct list_head group_node; 452 unsigned int on_rq; 453 454 u64 exec_start; 455 u64 sum_exec_runtime; 456 u64 vruntime; 457 u64 prev_sum_exec_runtime; 458 459 u64 nr_migrations; 460 461 struct sched_statistics statistics; 462 463 #ifdef CONFIG_FAIR_GROUP_SCHED 464 int depth; 465 struct sched_entity *parent; 466 /* rq on which this entity is (to be) queued: */ 467 struct cfs_rq *cfs_rq; 468 /* rq "owned" by this entity/group: */ 469 struct cfs_rq *my_q; 470 #endif 471 472 #ifdef CONFIG_SMP 473 /* 474 * Per entity load average tracking. 475 * 476 * Put into separate cache line so it does not 477 * collide with read-mostly values above. 478 */ 479 struct sched_avg avg; 480 #endif 481 }; 482 483 struct sched_rt_entity { 484 struct list_head run_list; 485 unsigned long timeout; 486 unsigned long watchdog_stamp; 487 unsigned int time_slice; 488 unsigned short on_rq; 489 unsigned short on_list; 490 491 struct sched_rt_entity *back; 492 #ifdef CONFIG_RT_GROUP_SCHED 493 struct sched_rt_entity *parent; 494 /* rq on which this entity is (to be) queued: */ 495 struct rt_rq *rt_rq; 496 /* rq "owned" by this entity/group: */ 497 struct rt_rq *my_q; 498 #endif 499 } __randomize_layout; 500 501 struct sched_dl_entity { 502 struct rb_node rb_node; 503 504 /* 505 * Original scheduling parameters. Copied here from sched_attr 506 * during sched_setattr(), they will remain the same until 507 * the next sched_setattr(). 508 */ 509 u64 dl_runtime; /* Maximum runtime for each instance */ 510 u64 dl_deadline; /* Relative deadline of each instance */ 511 u64 dl_period; /* Separation of two instances (period) */ 512 u64 dl_bw; /* dl_runtime / dl_period */ 513 u64 dl_density; /* dl_runtime / dl_deadline */ 514 515 /* 516 * Actual scheduling parameters. Initialized with the values above, 517 * they are continously updated during task execution. Note that 518 * the remaining runtime could be < 0 in case we are in overrun. 519 */ 520 s64 runtime; /* Remaining runtime for this instance */ 521 u64 deadline; /* Absolute deadline for this instance */ 522 unsigned int flags; /* Specifying the scheduler behaviour */ 523 524 /* 525 * Some bool flags: 526 * 527 * @dl_throttled tells if we exhausted the runtime. If so, the 528 * task has to wait for a replenishment to be performed at the 529 * next firing of dl_timer. 530 * 531 * @dl_boosted tells if we are boosted due to DI. If so we are 532 * outside bandwidth enforcement mechanism (but only until we 533 * exit the critical section); 534 * 535 * @dl_yielded tells if task gave up the CPU before consuming 536 * all its available runtime during the last job. 537 * 538 * @dl_non_contending tells if the task is inactive while still 539 * contributing to the active utilization. In other words, it 540 * indicates if the inactive timer has been armed and its handler 541 * has not been executed yet. This flag is useful to avoid race 542 * conditions between the inactive timer handler and the wakeup 543 * code. 544 * 545 * @dl_overrun tells if the task asked to be informed about runtime 546 * overruns. 547 */ 548 unsigned int dl_throttled : 1; 549 unsigned int dl_boosted : 1; 550 unsigned int dl_yielded : 1; 551 unsigned int dl_non_contending : 1; 552 unsigned int dl_overrun : 1; 553 554 /* 555 * Bandwidth enforcement timer. Each -deadline task has its 556 * own bandwidth to be enforced, thus we need one timer per task. 557 */ 558 struct hrtimer dl_timer; 559 560 /* 561 * Inactive timer, responsible for decreasing the active utilization 562 * at the "0-lag time". When a -deadline task blocks, it contributes 563 * to GRUB's active utilization until the "0-lag time", hence a 564 * timer is needed to decrease the active utilization at the correct 565 * time. 566 */ 567 struct hrtimer inactive_timer; 568 }; 569 570 union rcu_special { 571 struct { 572 u8 blocked; 573 u8 need_qs; 574 u8 exp_need_qs; 575 576 /* Otherwise the compiler can store garbage here: */ 577 u8 pad; 578 } b; /* Bits. */ 579 u32 s; /* Set of bits. */ 580 }; 581 582 enum perf_event_task_context { 583 perf_invalid_context = -1, 584 perf_hw_context = 0, 585 perf_sw_context, 586 perf_nr_task_contexts, 587 }; 588 589 struct wake_q_node { 590 struct wake_q_node *next; 591 }; 592 593 struct task_struct { 594 #ifdef CONFIG_THREAD_INFO_IN_TASK 595 /* 596 * For reasons of header soup (see current_thread_info()), this 597 * must be the first element of task_struct. 598 */ 599 struct thread_info thread_info; 600 #endif 601 /* -1 unrunnable, 0 runnable, >0 stopped: */ 602 volatile long state; 603 604 /* 605 * This begins the randomizable portion of task_struct. Only 606 * scheduling-critical items should be added above here. 607 */ 608 randomized_struct_fields_start 609 610 void *stack; 611 atomic_t usage; 612 /* Per task flags (PF_*), defined further below: */ 613 unsigned int flags; 614 unsigned int ptrace; 615 616 #ifdef CONFIG_SMP 617 struct llist_node wake_entry; 618 int on_cpu; 619 #ifdef CONFIG_THREAD_INFO_IN_TASK 620 /* Current CPU: */ 621 unsigned int cpu; 622 #endif 623 unsigned int wakee_flips; 624 unsigned long wakee_flip_decay_ts; 625 struct task_struct *last_wakee; 626 627 /* 628 * recent_used_cpu is initially set as the last CPU used by a task 629 * that wakes affine another task. Waker/wakee relationships can 630 * push tasks around a CPU where each wakeup moves to the next one. 631 * Tracking a recently used CPU allows a quick search for a recently 632 * used CPU that may be idle. 633 */ 634 int recent_used_cpu; 635 int wake_cpu; 636 #endif 637 int on_rq; 638 639 int prio; 640 int static_prio; 641 int normal_prio; 642 unsigned int rt_priority; 643 644 const struct sched_class *sched_class; 645 struct sched_entity se; 646 struct sched_rt_entity rt; 647 #ifdef CONFIG_CGROUP_SCHED 648 struct task_group *sched_task_group; 649 #endif 650 struct sched_dl_entity dl; 651 652 #ifdef CONFIG_PREEMPT_NOTIFIERS 653 /* List of struct preempt_notifier: */ 654 struct hlist_head preempt_notifiers; 655 #endif 656 657 #ifdef CONFIG_BLK_DEV_IO_TRACE 658 unsigned int btrace_seq; 659 #endif 660 661 unsigned int policy; 662 int nr_cpus_allowed; 663 cpumask_t cpus_allowed; 664 665 #ifdef CONFIG_PREEMPT_RCU 666 int rcu_read_lock_nesting; 667 union rcu_special rcu_read_unlock_special; 668 struct list_head rcu_node_entry; 669 struct rcu_node *rcu_blocked_node; 670 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 671 672 #ifdef CONFIG_TASKS_RCU 673 unsigned long rcu_tasks_nvcsw; 674 u8 rcu_tasks_holdout; 675 u8 rcu_tasks_idx; 676 int rcu_tasks_idle_cpu; 677 struct list_head rcu_tasks_holdout_list; 678 #endif /* #ifdef CONFIG_TASKS_RCU */ 679 680 struct sched_info sched_info; 681 682 struct list_head tasks; 683 #ifdef CONFIG_SMP 684 struct plist_node pushable_tasks; 685 struct rb_node pushable_dl_tasks; 686 #endif 687 688 struct mm_struct *mm; 689 struct mm_struct *active_mm; 690 691 /* Per-thread vma caching: */ 692 struct vmacache vmacache; 693 694 #ifdef SPLIT_RSS_COUNTING 695 struct task_rss_stat rss_stat; 696 #endif 697 int exit_state; 698 int exit_code; 699 int exit_signal; 700 /* The signal sent when the parent dies: */ 701 int pdeath_signal; 702 /* JOBCTL_*, siglock protected: */ 703 unsigned long jobctl; 704 705 /* Used for emulating ABI behavior of previous Linux versions: */ 706 unsigned int personality; 707 708 /* Scheduler bits, serialized by scheduler locks: */ 709 unsigned sched_reset_on_fork:1; 710 unsigned sched_contributes_to_load:1; 711 unsigned sched_migrated:1; 712 unsigned sched_remote_wakeup:1; 713 /* Force alignment to the next boundary: */ 714 unsigned :0; 715 716 /* Unserialized, strictly 'current' */ 717 718 /* Bit to tell LSMs we're in execve(): */ 719 unsigned in_execve:1; 720 unsigned in_iowait:1; 721 #ifndef TIF_RESTORE_SIGMASK 722 unsigned restore_sigmask:1; 723 #endif 724 #ifdef CONFIG_MEMCG 725 unsigned memcg_may_oom:1; 726 #ifndef CONFIG_SLOB 727 unsigned memcg_kmem_skip_account:1; 728 #endif 729 #endif 730 #ifdef CONFIG_COMPAT_BRK 731 unsigned brk_randomized:1; 732 #endif 733 #ifdef CONFIG_CGROUPS 734 /* disallow userland-initiated cgroup migration */ 735 unsigned no_cgroup_migration:1; 736 #endif 737 738 unsigned long atomic_flags; /* Flags requiring atomic access. */ 739 740 struct restart_block restart_block; 741 742 pid_t pid; 743 pid_t tgid; 744 745 #ifdef CONFIG_STACKPROTECTOR 746 /* Canary value for the -fstack-protector GCC feature: */ 747 unsigned long stack_canary; 748 #endif 749 /* 750 * Pointers to the (original) parent process, youngest child, younger sibling, 751 * older sibling, respectively. (p->father can be replaced with 752 * p->real_parent->pid) 753 */ 754 755 /* Real parent process: */ 756 struct task_struct __rcu *real_parent; 757 758 /* Recipient of SIGCHLD, wait4() reports: */ 759 struct task_struct __rcu *parent; 760 761 /* 762 * Children/sibling form the list of natural children: 763 */ 764 struct list_head children; 765 struct list_head sibling; 766 struct task_struct *group_leader; 767 768 /* 769 * 'ptraced' is the list of tasks this task is using ptrace() on. 770 * 771 * This includes both natural children and PTRACE_ATTACH targets. 772 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 773 */ 774 struct list_head ptraced; 775 struct list_head ptrace_entry; 776 777 /* PID/PID hash table linkage. */ 778 struct pid_link pids[PIDTYPE_MAX]; 779 struct list_head thread_group; 780 struct list_head thread_node; 781 782 struct completion *vfork_done; 783 784 /* CLONE_CHILD_SETTID: */ 785 int __user *set_child_tid; 786 787 /* CLONE_CHILD_CLEARTID: */ 788 int __user *clear_child_tid; 789 790 u64 utime; 791 u64 stime; 792 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 793 u64 utimescaled; 794 u64 stimescaled; 795 #endif 796 u64 gtime; 797 struct prev_cputime prev_cputime; 798 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 799 struct vtime vtime; 800 #endif 801 802 #ifdef CONFIG_NO_HZ_FULL 803 atomic_t tick_dep_mask; 804 #endif 805 /* Context switch counts: */ 806 unsigned long nvcsw; 807 unsigned long nivcsw; 808 809 /* Monotonic time in nsecs: */ 810 u64 start_time; 811 812 /* Boot based time in nsecs: */ 813 u64 real_start_time; 814 815 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 816 unsigned long min_flt; 817 unsigned long maj_flt; 818 819 #ifdef CONFIG_POSIX_TIMERS 820 struct task_cputime cputime_expires; 821 struct list_head cpu_timers[3]; 822 #endif 823 824 /* Process credentials: */ 825 826 /* Tracer's credentials at attach: */ 827 const struct cred __rcu *ptracer_cred; 828 829 /* Objective and real subjective task credentials (COW): */ 830 const struct cred __rcu *real_cred; 831 832 /* Effective (overridable) subjective task credentials (COW): */ 833 const struct cred __rcu *cred; 834 835 /* 836 * executable name, excluding path. 837 * 838 * - normally initialized setup_new_exec() 839 * - access it with [gs]et_task_comm() 840 * - lock it with task_lock() 841 */ 842 char comm[TASK_COMM_LEN]; 843 844 struct nameidata *nameidata; 845 846 #ifdef CONFIG_SYSVIPC 847 struct sysv_sem sysvsem; 848 struct sysv_shm sysvshm; 849 #endif 850 #ifdef CONFIG_DETECT_HUNG_TASK 851 unsigned long last_switch_count; 852 #endif 853 /* Filesystem information: */ 854 struct fs_struct *fs; 855 856 /* Open file information: */ 857 struct files_struct *files; 858 859 /* Namespaces: */ 860 struct nsproxy *nsproxy; 861 862 /* Signal handlers: */ 863 struct signal_struct *signal; 864 struct sighand_struct *sighand; 865 sigset_t blocked; 866 sigset_t real_blocked; 867 /* Restored if set_restore_sigmask() was used: */ 868 sigset_t saved_sigmask; 869 struct sigpending pending; 870 unsigned long sas_ss_sp; 871 size_t sas_ss_size; 872 unsigned int sas_ss_flags; 873 874 struct callback_head *task_works; 875 876 struct audit_context *audit_context; 877 #ifdef CONFIG_AUDITSYSCALL 878 kuid_t loginuid; 879 unsigned int sessionid; 880 #endif 881 struct seccomp seccomp; 882 883 /* Thread group tracking: */ 884 u32 parent_exec_id; 885 u32 self_exec_id; 886 887 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 888 spinlock_t alloc_lock; 889 890 /* Protection of the PI data structures: */ 891 raw_spinlock_t pi_lock; 892 893 struct wake_q_node wake_q; 894 895 #ifdef CONFIG_RT_MUTEXES 896 /* PI waiters blocked on a rt_mutex held by this task: */ 897 struct rb_root_cached pi_waiters; 898 /* Updated under owner's pi_lock and rq lock */ 899 struct task_struct *pi_top_task; 900 /* Deadlock detection and priority inheritance handling: */ 901 struct rt_mutex_waiter *pi_blocked_on; 902 #endif 903 904 #ifdef CONFIG_DEBUG_MUTEXES 905 /* Mutex deadlock detection: */ 906 struct mutex_waiter *blocked_on; 907 #endif 908 909 #ifdef CONFIG_TRACE_IRQFLAGS 910 unsigned int irq_events; 911 unsigned long hardirq_enable_ip; 912 unsigned long hardirq_disable_ip; 913 unsigned int hardirq_enable_event; 914 unsigned int hardirq_disable_event; 915 int hardirqs_enabled; 916 int hardirq_context; 917 unsigned long softirq_disable_ip; 918 unsigned long softirq_enable_ip; 919 unsigned int softirq_disable_event; 920 unsigned int softirq_enable_event; 921 int softirqs_enabled; 922 int softirq_context; 923 #endif 924 925 #ifdef CONFIG_LOCKDEP 926 # define MAX_LOCK_DEPTH 48UL 927 u64 curr_chain_key; 928 int lockdep_depth; 929 unsigned int lockdep_recursion; 930 struct held_lock held_locks[MAX_LOCK_DEPTH]; 931 #endif 932 933 #ifdef CONFIG_UBSAN 934 unsigned int in_ubsan; 935 #endif 936 937 /* Journalling filesystem info: */ 938 void *journal_info; 939 940 /* Stacked block device info: */ 941 struct bio_list *bio_list; 942 943 #ifdef CONFIG_BLOCK 944 /* Stack plugging: */ 945 struct blk_plug *plug; 946 #endif 947 948 /* VM state: */ 949 struct reclaim_state *reclaim_state; 950 951 struct backing_dev_info *backing_dev_info; 952 953 struct io_context *io_context; 954 955 /* Ptrace state: */ 956 unsigned long ptrace_message; 957 siginfo_t *last_siginfo; 958 959 struct task_io_accounting ioac; 960 #ifdef CONFIG_TASK_XACCT 961 /* Accumulated RSS usage: */ 962 u64 acct_rss_mem1; 963 /* Accumulated virtual memory usage: */ 964 u64 acct_vm_mem1; 965 /* stime + utime since last update: */ 966 u64 acct_timexpd; 967 #endif 968 #ifdef CONFIG_CPUSETS 969 /* Protected by ->alloc_lock: */ 970 nodemask_t mems_allowed; 971 /* Seqence number to catch updates: */ 972 seqcount_t mems_allowed_seq; 973 int cpuset_mem_spread_rotor; 974 int cpuset_slab_spread_rotor; 975 #endif 976 #ifdef CONFIG_CGROUPS 977 /* Control Group info protected by css_set_lock: */ 978 struct css_set __rcu *cgroups; 979 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 980 struct list_head cg_list; 981 #endif 982 #ifdef CONFIG_INTEL_RDT 983 u32 closid; 984 u32 rmid; 985 #endif 986 #ifdef CONFIG_FUTEX 987 struct robust_list_head __user *robust_list; 988 #ifdef CONFIG_COMPAT 989 struct compat_robust_list_head __user *compat_robust_list; 990 #endif 991 struct list_head pi_state_list; 992 struct futex_pi_state *pi_state_cache; 993 #endif 994 #ifdef CONFIG_PERF_EVENTS 995 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 996 struct mutex perf_event_mutex; 997 struct list_head perf_event_list; 998 #endif 999 #ifdef CONFIG_DEBUG_PREEMPT 1000 unsigned long preempt_disable_ip; 1001 #endif 1002 #ifdef CONFIG_NUMA 1003 /* Protected by alloc_lock: */ 1004 struct mempolicy *mempolicy; 1005 short il_prev; 1006 short pref_node_fork; 1007 #endif 1008 #ifdef CONFIG_NUMA_BALANCING 1009 int numa_scan_seq; 1010 unsigned int numa_scan_period; 1011 unsigned int numa_scan_period_max; 1012 int numa_preferred_nid; 1013 unsigned long numa_migrate_retry; 1014 /* Migration stamp: */ 1015 u64 node_stamp; 1016 u64 last_task_numa_placement; 1017 u64 last_sum_exec_runtime; 1018 struct callback_head numa_work; 1019 1020 struct list_head numa_entry; 1021 struct numa_group *numa_group; 1022 1023 /* 1024 * numa_faults is an array split into four regions: 1025 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1026 * in this precise order. 1027 * 1028 * faults_memory: Exponential decaying average of faults on a per-node 1029 * basis. Scheduling placement decisions are made based on these 1030 * counts. The values remain static for the duration of a PTE scan. 1031 * faults_cpu: Track the nodes the process was running on when a NUMA 1032 * hinting fault was incurred. 1033 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1034 * during the current scan window. When the scan completes, the counts 1035 * in faults_memory and faults_cpu decay and these values are copied. 1036 */ 1037 unsigned long *numa_faults; 1038 unsigned long total_numa_faults; 1039 1040 /* 1041 * numa_faults_locality tracks if faults recorded during the last 1042 * scan window were remote/local or failed to migrate. The task scan 1043 * period is adapted based on the locality of the faults with different 1044 * weights depending on whether they were shared or private faults 1045 */ 1046 unsigned long numa_faults_locality[3]; 1047 1048 unsigned long numa_pages_migrated; 1049 #endif /* CONFIG_NUMA_BALANCING */ 1050 1051 #ifdef CONFIG_RSEQ 1052 struct rseq __user *rseq; 1053 u32 rseq_len; 1054 u32 rseq_sig; 1055 /* 1056 * RmW on rseq_event_mask must be performed atomically 1057 * with respect to preemption. 1058 */ 1059 unsigned long rseq_event_mask; 1060 #endif 1061 1062 struct tlbflush_unmap_batch tlb_ubc; 1063 1064 struct rcu_head rcu; 1065 1066 /* Cache last used pipe for splice(): */ 1067 struct pipe_inode_info *splice_pipe; 1068 1069 struct page_frag task_frag; 1070 1071 #ifdef CONFIG_TASK_DELAY_ACCT 1072 struct task_delay_info *delays; 1073 #endif 1074 1075 #ifdef CONFIG_FAULT_INJECTION 1076 int make_it_fail; 1077 unsigned int fail_nth; 1078 #endif 1079 /* 1080 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1081 * balance_dirty_pages() for a dirty throttling pause: 1082 */ 1083 int nr_dirtied; 1084 int nr_dirtied_pause; 1085 /* Start of a write-and-pause period: */ 1086 unsigned long dirty_paused_when; 1087 1088 #ifdef CONFIG_LATENCYTOP 1089 int latency_record_count; 1090 struct latency_record latency_record[LT_SAVECOUNT]; 1091 #endif 1092 /* 1093 * Time slack values; these are used to round up poll() and 1094 * select() etc timeout values. These are in nanoseconds. 1095 */ 1096 u64 timer_slack_ns; 1097 u64 default_timer_slack_ns; 1098 1099 #ifdef CONFIG_KASAN 1100 unsigned int kasan_depth; 1101 #endif 1102 1103 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1104 /* Index of current stored address in ret_stack: */ 1105 int curr_ret_stack; 1106 1107 /* Stack of return addresses for return function tracing: */ 1108 struct ftrace_ret_stack *ret_stack; 1109 1110 /* Timestamp for last schedule: */ 1111 unsigned long long ftrace_timestamp; 1112 1113 /* 1114 * Number of functions that haven't been traced 1115 * because of depth overrun: 1116 */ 1117 atomic_t trace_overrun; 1118 1119 /* Pause tracing: */ 1120 atomic_t tracing_graph_pause; 1121 #endif 1122 1123 #ifdef CONFIG_TRACING 1124 /* State flags for use by tracers: */ 1125 unsigned long trace; 1126 1127 /* Bitmask and counter of trace recursion: */ 1128 unsigned long trace_recursion; 1129 #endif /* CONFIG_TRACING */ 1130 1131 #ifdef CONFIG_KCOV 1132 /* Coverage collection mode enabled for this task (0 if disabled): */ 1133 unsigned int kcov_mode; 1134 1135 /* Size of the kcov_area: */ 1136 unsigned int kcov_size; 1137 1138 /* Buffer for coverage collection: */ 1139 void *kcov_area; 1140 1141 /* KCOV descriptor wired with this task or NULL: */ 1142 struct kcov *kcov; 1143 #endif 1144 1145 #ifdef CONFIG_MEMCG 1146 struct mem_cgroup *memcg_in_oom; 1147 gfp_t memcg_oom_gfp_mask; 1148 int memcg_oom_order; 1149 1150 /* Number of pages to reclaim on returning to userland: */ 1151 unsigned int memcg_nr_pages_over_high; 1152 #endif 1153 1154 #ifdef CONFIG_UPROBES 1155 struct uprobe_task *utask; 1156 #endif 1157 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1158 unsigned int sequential_io; 1159 unsigned int sequential_io_avg; 1160 #endif 1161 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1162 unsigned long task_state_change; 1163 #endif 1164 int pagefault_disabled; 1165 #ifdef CONFIG_MMU 1166 struct task_struct *oom_reaper_list; 1167 #endif 1168 #ifdef CONFIG_VMAP_STACK 1169 struct vm_struct *stack_vm_area; 1170 #endif 1171 #ifdef CONFIG_THREAD_INFO_IN_TASK 1172 /* A live task holds one reference: */ 1173 atomic_t stack_refcount; 1174 #endif 1175 #ifdef CONFIG_LIVEPATCH 1176 int patch_state; 1177 #endif 1178 #ifdef CONFIG_SECURITY 1179 /* Used by LSM modules for access restriction: */ 1180 void *security; 1181 #endif 1182 1183 /* 1184 * New fields for task_struct should be added above here, so that 1185 * they are included in the randomized portion of task_struct. 1186 */ 1187 randomized_struct_fields_end 1188 1189 /* CPU-specific state of this task: */ 1190 struct thread_struct thread; 1191 1192 /* 1193 * WARNING: on x86, 'thread_struct' contains a variable-sized 1194 * structure. It *MUST* be at the end of 'task_struct'. 1195 * 1196 * Do not put anything below here! 1197 */ 1198 }; 1199 1200 static inline struct pid *task_pid(struct task_struct *task) 1201 { 1202 return task->pids[PIDTYPE_PID].pid; 1203 } 1204 1205 static inline struct pid *task_tgid(struct task_struct *task) 1206 { 1207 return task->group_leader->pids[PIDTYPE_PID].pid; 1208 } 1209 1210 /* 1211 * Without tasklist or RCU lock it is not safe to dereference 1212 * the result of task_pgrp/task_session even if task == current, 1213 * we can race with another thread doing sys_setsid/sys_setpgid. 1214 */ 1215 static inline struct pid *task_pgrp(struct task_struct *task) 1216 { 1217 return task->group_leader->pids[PIDTYPE_PGID].pid; 1218 } 1219 1220 static inline struct pid *task_session(struct task_struct *task) 1221 { 1222 return task->group_leader->pids[PIDTYPE_SID].pid; 1223 } 1224 1225 /* 1226 * the helpers to get the task's different pids as they are seen 1227 * from various namespaces 1228 * 1229 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1230 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1231 * current. 1232 * task_xid_nr_ns() : id seen from the ns specified; 1233 * 1234 * see also pid_nr() etc in include/linux/pid.h 1235 */ 1236 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1237 1238 static inline pid_t task_pid_nr(struct task_struct *tsk) 1239 { 1240 return tsk->pid; 1241 } 1242 1243 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1244 { 1245 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1246 } 1247 1248 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1249 { 1250 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1251 } 1252 1253 1254 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1255 { 1256 return tsk->tgid; 1257 } 1258 1259 /** 1260 * pid_alive - check that a task structure is not stale 1261 * @p: Task structure to be checked. 1262 * 1263 * Test if a process is not yet dead (at most zombie state) 1264 * If pid_alive fails, then pointers within the task structure 1265 * can be stale and must not be dereferenced. 1266 * 1267 * Return: 1 if the process is alive. 0 otherwise. 1268 */ 1269 static inline int pid_alive(const struct task_struct *p) 1270 { 1271 return p->pids[PIDTYPE_PID].pid != NULL; 1272 } 1273 1274 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1275 { 1276 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1277 } 1278 1279 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1280 { 1281 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1282 } 1283 1284 1285 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1286 { 1287 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1288 } 1289 1290 static inline pid_t task_session_vnr(struct task_struct *tsk) 1291 { 1292 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1293 } 1294 1295 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1296 { 1297 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns); 1298 } 1299 1300 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1301 { 1302 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL); 1303 } 1304 1305 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1306 { 1307 pid_t pid = 0; 1308 1309 rcu_read_lock(); 1310 if (pid_alive(tsk)) 1311 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1312 rcu_read_unlock(); 1313 1314 return pid; 1315 } 1316 1317 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1318 { 1319 return task_ppid_nr_ns(tsk, &init_pid_ns); 1320 } 1321 1322 /* Obsolete, do not use: */ 1323 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1324 { 1325 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1326 } 1327 1328 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1329 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1330 1331 static inline unsigned int task_state_index(struct task_struct *tsk) 1332 { 1333 unsigned int tsk_state = READ_ONCE(tsk->state); 1334 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT; 1335 1336 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1337 1338 if (tsk_state == TASK_IDLE) 1339 state = TASK_REPORT_IDLE; 1340 1341 return fls(state); 1342 } 1343 1344 static inline char task_index_to_char(unsigned int state) 1345 { 1346 static const char state_char[] = "RSDTtXZPI"; 1347 1348 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1349 1350 return state_char[state]; 1351 } 1352 1353 static inline char task_state_to_char(struct task_struct *tsk) 1354 { 1355 return task_index_to_char(task_state_index(tsk)); 1356 } 1357 1358 /** 1359 * is_global_init - check if a task structure is init. Since init 1360 * is free to have sub-threads we need to check tgid. 1361 * @tsk: Task structure to be checked. 1362 * 1363 * Check if a task structure is the first user space task the kernel created. 1364 * 1365 * Return: 1 if the task structure is init. 0 otherwise. 1366 */ 1367 static inline int is_global_init(struct task_struct *tsk) 1368 { 1369 return task_tgid_nr(tsk) == 1; 1370 } 1371 1372 extern struct pid *cad_pid; 1373 1374 /* 1375 * Per process flags 1376 */ 1377 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1378 #define PF_EXITING 0x00000004 /* Getting shut down */ 1379 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */ 1380 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */ 1381 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1382 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1383 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1384 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1385 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1386 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1387 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1388 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1389 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1390 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1391 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1392 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1393 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1394 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1395 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1396 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ 1397 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1398 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1399 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1400 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */ 1401 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1402 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */ 1403 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1404 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1405 1406 /* 1407 * Only the _current_ task can read/write to tsk->flags, but other 1408 * tasks can access tsk->flags in readonly mode for example 1409 * with tsk_used_math (like during threaded core dumping). 1410 * There is however an exception to this rule during ptrace 1411 * or during fork: the ptracer task is allowed to write to the 1412 * child->flags of its traced child (same goes for fork, the parent 1413 * can write to the child->flags), because we're guaranteed the 1414 * child is not running and in turn not changing child->flags 1415 * at the same time the parent does it. 1416 */ 1417 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1418 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1419 #define clear_used_math() clear_stopped_child_used_math(current) 1420 #define set_used_math() set_stopped_child_used_math(current) 1421 1422 #define conditional_stopped_child_used_math(condition, child) \ 1423 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1424 1425 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1426 1427 #define copy_to_stopped_child_used_math(child) \ 1428 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1429 1430 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1431 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1432 #define used_math() tsk_used_math(current) 1433 1434 static inline bool is_percpu_thread(void) 1435 { 1436 #ifdef CONFIG_SMP 1437 return (current->flags & PF_NO_SETAFFINITY) && 1438 (current->nr_cpus_allowed == 1); 1439 #else 1440 return true; 1441 #endif 1442 } 1443 1444 /* Per-process atomic flags. */ 1445 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1446 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1447 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1448 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1449 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1450 1451 #define TASK_PFA_TEST(name, func) \ 1452 static inline bool task_##func(struct task_struct *p) \ 1453 { return test_bit(PFA_##name, &p->atomic_flags); } 1454 1455 #define TASK_PFA_SET(name, func) \ 1456 static inline void task_set_##func(struct task_struct *p) \ 1457 { set_bit(PFA_##name, &p->atomic_flags); } 1458 1459 #define TASK_PFA_CLEAR(name, func) \ 1460 static inline void task_clear_##func(struct task_struct *p) \ 1461 { clear_bit(PFA_##name, &p->atomic_flags); } 1462 1463 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1464 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1465 1466 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1467 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1468 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1469 1470 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1471 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1472 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1473 1474 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1475 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1476 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1477 1478 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1479 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1480 1481 static inline void 1482 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1483 { 1484 current->flags &= ~flags; 1485 current->flags |= orig_flags & flags; 1486 } 1487 1488 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1489 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1490 #ifdef CONFIG_SMP 1491 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1492 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1493 #else 1494 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1495 { 1496 } 1497 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1498 { 1499 if (!cpumask_test_cpu(0, new_mask)) 1500 return -EINVAL; 1501 return 0; 1502 } 1503 #endif 1504 1505 #ifndef cpu_relax_yield 1506 #define cpu_relax_yield() cpu_relax() 1507 #endif 1508 1509 extern int yield_to(struct task_struct *p, bool preempt); 1510 extern void set_user_nice(struct task_struct *p, long nice); 1511 extern int task_prio(const struct task_struct *p); 1512 1513 /** 1514 * task_nice - return the nice value of a given task. 1515 * @p: the task in question. 1516 * 1517 * Return: The nice value [ -20 ... 0 ... 19 ]. 1518 */ 1519 static inline int task_nice(const struct task_struct *p) 1520 { 1521 return PRIO_TO_NICE((p)->static_prio); 1522 } 1523 1524 extern int can_nice(const struct task_struct *p, const int nice); 1525 extern int task_curr(const struct task_struct *p); 1526 extern int idle_cpu(int cpu); 1527 extern int available_idle_cpu(int cpu); 1528 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1529 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1530 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1531 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1532 extern struct task_struct *idle_task(int cpu); 1533 1534 /** 1535 * is_idle_task - is the specified task an idle task? 1536 * @p: the task in question. 1537 * 1538 * Return: 1 if @p is an idle task. 0 otherwise. 1539 */ 1540 static inline bool is_idle_task(const struct task_struct *p) 1541 { 1542 return !!(p->flags & PF_IDLE); 1543 } 1544 1545 extern struct task_struct *curr_task(int cpu); 1546 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1547 1548 void yield(void); 1549 1550 union thread_union { 1551 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1552 struct task_struct task; 1553 #endif 1554 #ifndef CONFIG_THREAD_INFO_IN_TASK 1555 struct thread_info thread_info; 1556 #endif 1557 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1558 }; 1559 1560 #ifndef CONFIG_THREAD_INFO_IN_TASK 1561 extern struct thread_info init_thread_info; 1562 #endif 1563 1564 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1565 1566 #ifdef CONFIG_THREAD_INFO_IN_TASK 1567 static inline struct thread_info *task_thread_info(struct task_struct *task) 1568 { 1569 return &task->thread_info; 1570 } 1571 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1572 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1573 #endif 1574 1575 /* 1576 * find a task by one of its numerical ids 1577 * 1578 * find_task_by_pid_ns(): 1579 * finds a task by its pid in the specified namespace 1580 * find_task_by_vpid(): 1581 * finds a task by its virtual pid 1582 * 1583 * see also find_vpid() etc in include/linux/pid.h 1584 */ 1585 1586 extern struct task_struct *find_task_by_vpid(pid_t nr); 1587 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1588 1589 /* 1590 * find a task by its virtual pid and get the task struct 1591 */ 1592 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1593 1594 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1595 extern int wake_up_process(struct task_struct *tsk); 1596 extern void wake_up_new_task(struct task_struct *tsk); 1597 1598 #ifdef CONFIG_SMP 1599 extern void kick_process(struct task_struct *tsk); 1600 #else 1601 static inline void kick_process(struct task_struct *tsk) { } 1602 #endif 1603 1604 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1605 1606 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1607 { 1608 __set_task_comm(tsk, from, false); 1609 } 1610 1611 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1612 #define get_task_comm(buf, tsk) ({ \ 1613 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1614 __get_task_comm(buf, sizeof(buf), tsk); \ 1615 }) 1616 1617 #ifdef CONFIG_SMP 1618 void scheduler_ipi(void); 1619 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 1620 #else 1621 static inline void scheduler_ipi(void) { } 1622 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) 1623 { 1624 return 1; 1625 } 1626 #endif 1627 1628 /* 1629 * Set thread flags in other task's structures. 1630 * See asm/thread_info.h for TIF_xxxx flags available: 1631 */ 1632 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1633 { 1634 set_ti_thread_flag(task_thread_info(tsk), flag); 1635 } 1636 1637 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1638 { 1639 clear_ti_thread_flag(task_thread_info(tsk), flag); 1640 } 1641 1642 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1643 bool value) 1644 { 1645 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1646 } 1647 1648 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1649 { 1650 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1651 } 1652 1653 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1654 { 1655 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1656 } 1657 1658 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1659 { 1660 return test_ti_thread_flag(task_thread_info(tsk), flag); 1661 } 1662 1663 static inline void set_tsk_need_resched(struct task_struct *tsk) 1664 { 1665 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1666 } 1667 1668 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1669 { 1670 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1671 } 1672 1673 static inline int test_tsk_need_resched(struct task_struct *tsk) 1674 { 1675 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1676 } 1677 1678 /* 1679 * cond_resched() and cond_resched_lock(): latency reduction via 1680 * explicit rescheduling in places that are safe. The return 1681 * value indicates whether a reschedule was done in fact. 1682 * cond_resched_lock() will drop the spinlock before scheduling, 1683 */ 1684 #ifndef CONFIG_PREEMPT 1685 extern int _cond_resched(void); 1686 #else 1687 static inline int _cond_resched(void) { return 0; } 1688 #endif 1689 1690 #define cond_resched() ({ \ 1691 ___might_sleep(__FILE__, __LINE__, 0); \ 1692 _cond_resched(); \ 1693 }) 1694 1695 extern int __cond_resched_lock(spinlock_t *lock); 1696 1697 #define cond_resched_lock(lock) ({ \ 1698 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ 1699 __cond_resched_lock(lock); \ 1700 }) 1701 1702 static inline void cond_resched_rcu(void) 1703 { 1704 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 1705 rcu_read_unlock(); 1706 cond_resched(); 1707 rcu_read_lock(); 1708 #endif 1709 } 1710 1711 /* 1712 * Does a critical section need to be broken due to another 1713 * task waiting?: (technically does not depend on CONFIG_PREEMPT, 1714 * but a general need for low latency) 1715 */ 1716 static inline int spin_needbreak(spinlock_t *lock) 1717 { 1718 #ifdef CONFIG_PREEMPT 1719 return spin_is_contended(lock); 1720 #else 1721 return 0; 1722 #endif 1723 } 1724 1725 static __always_inline bool need_resched(void) 1726 { 1727 return unlikely(tif_need_resched()); 1728 } 1729 1730 /* 1731 * Wrappers for p->thread_info->cpu access. No-op on UP. 1732 */ 1733 #ifdef CONFIG_SMP 1734 1735 static inline unsigned int task_cpu(const struct task_struct *p) 1736 { 1737 #ifdef CONFIG_THREAD_INFO_IN_TASK 1738 return p->cpu; 1739 #else 1740 return task_thread_info(p)->cpu; 1741 #endif 1742 } 1743 1744 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 1745 1746 #else 1747 1748 static inline unsigned int task_cpu(const struct task_struct *p) 1749 { 1750 return 0; 1751 } 1752 1753 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 1754 { 1755 } 1756 1757 #endif /* CONFIG_SMP */ 1758 1759 /* 1760 * In order to reduce various lock holder preemption latencies provide an 1761 * interface to see if a vCPU is currently running or not. 1762 * 1763 * This allows us to terminate optimistic spin loops and block, analogous to 1764 * the native optimistic spin heuristic of testing if the lock owner task is 1765 * running or not. 1766 */ 1767 #ifndef vcpu_is_preempted 1768 # define vcpu_is_preempted(cpu) false 1769 #endif 1770 1771 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 1772 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 1773 1774 #ifndef TASK_SIZE_OF 1775 #define TASK_SIZE_OF(tsk) TASK_SIZE 1776 #endif 1777 1778 #ifdef CONFIG_RSEQ 1779 1780 /* 1781 * Map the event mask on the user-space ABI enum rseq_cs_flags 1782 * for direct mask checks. 1783 */ 1784 enum rseq_event_mask_bits { 1785 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT, 1786 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT, 1787 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT, 1788 }; 1789 1790 enum rseq_event_mask { 1791 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT), 1792 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT), 1793 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT), 1794 }; 1795 1796 static inline void rseq_set_notify_resume(struct task_struct *t) 1797 { 1798 if (t->rseq) 1799 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME); 1800 } 1801 1802 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs); 1803 1804 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 1805 struct pt_regs *regs) 1806 { 1807 if (current->rseq) 1808 __rseq_handle_notify_resume(ksig, regs); 1809 } 1810 1811 static inline void rseq_signal_deliver(struct ksignal *ksig, 1812 struct pt_regs *regs) 1813 { 1814 preempt_disable(); 1815 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask); 1816 preempt_enable(); 1817 rseq_handle_notify_resume(ksig, regs); 1818 } 1819 1820 /* rseq_preempt() requires preemption to be disabled. */ 1821 static inline void rseq_preempt(struct task_struct *t) 1822 { 1823 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask); 1824 rseq_set_notify_resume(t); 1825 } 1826 1827 /* rseq_migrate() requires preemption to be disabled. */ 1828 static inline void rseq_migrate(struct task_struct *t) 1829 { 1830 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask); 1831 rseq_set_notify_resume(t); 1832 } 1833 1834 /* 1835 * If parent process has a registered restartable sequences area, the 1836 * child inherits. Only applies when forking a process, not a thread. 1837 */ 1838 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 1839 { 1840 if (clone_flags & CLONE_THREAD) { 1841 t->rseq = NULL; 1842 t->rseq_len = 0; 1843 t->rseq_sig = 0; 1844 t->rseq_event_mask = 0; 1845 } else { 1846 t->rseq = current->rseq; 1847 t->rseq_len = current->rseq_len; 1848 t->rseq_sig = current->rseq_sig; 1849 t->rseq_event_mask = current->rseq_event_mask; 1850 } 1851 } 1852 1853 static inline void rseq_execve(struct task_struct *t) 1854 { 1855 t->rseq = NULL; 1856 t->rseq_len = 0; 1857 t->rseq_sig = 0; 1858 t->rseq_event_mask = 0; 1859 } 1860 1861 #else 1862 1863 static inline void rseq_set_notify_resume(struct task_struct *t) 1864 { 1865 } 1866 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 1867 struct pt_regs *regs) 1868 { 1869 } 1870 static inline void rseq_signal_deliver(struct ksignal *ksig, 1871 struct pt_regs *regs) 1872 { 1873 } 1874 static inline void rseq_preempt(struct task_struct *t) 1875 { 1876 } 1877 static inline void rseq_migrate(struct task_struct *t) 1878 { 1879 } 1880 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 1881 { 1882 } 1883 static inline void rseq_execve(struct task_struct *t) 1884 { 1885 } 1886 1887 #endif 1888 1889 #ifdef CONFIG_DEBUG_RSEQ 1890 1891 void rseq_syscall(struct pt_regs *regs); 1892 1893 #else 1894 1895 static inline void rseq_syscall(struct pt_regs *regs) 1896 { 1897 } 1898 1899 #endif 1900 1901 #endif 1902