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/psi_types.h> 29 #include <linux/mm_types_task.h> 30 #include <linux/task_io_accounting.h> 31 #include <linux/rseq.h> 32 33 /* task_struct member predeclarations (sorted alphabetically): */ 34 struct audit_context; 35 struct backing_dev_info; 36 struct bio_list; 37 struct blk_plug; 38 struct cfs_rq; 39 struct fs_struct; 40 struct futex_pi_state; 41 struct io_context; 42 struct mempolicy; 43 struct nameidata; 44 struct nsproxy; 45 struct perf_event_context; 46 struct pid_namespace; 47 struct pipe_inode_info; 48 struct rcu_node; 49 struct reclaim_state; 50 struct robust_list_head; 51 struct sched_attr; 52 struct sched_param; 53 struct seq_file; 54 struct sighand_struct; 55 struct signal_struct; 56 struct task_delay_info; 57 struct task_group; 58 59 /* 60 * Task state bitmask. NOTE! These bits are also 61 * encoded in fs/proc/array.c: get_task_state(). 62 * 63 * We have two separate sets of flags: task->state 64 * is about runnability, while task->exit_state are 65 * about the task exiting. Confusing, but this way 66 * modifying one set can't modify the other one by 67 * mistake. 68 */ 69 70 /* Used in tsk->state: */ 71 #define TASK_RUNNING 0x0000 72 #define TASK_INTERRUPTIBLE 0x0001 73 #define TASK_UNINTERRUPTIBLE 0x0002 74 #define __TASK_STOPPED 0x0004 75 #define __TASK_TRACED 0x0008 76 /* Used in tsk->exit_state: */ 77 #define EXIT_DEAD 0x0010 78 #define EXIT_ZOMBIE 0x0020 79 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 80 /* Used in tsk->state again: */ 81 #define TASK_PARKED 0x0040 82 #define TASK_DEAD 0x0080 83 #define TASK_WAKEKILL 0x0100 84 #define TASK_WAKING 0x0200 85 #define TASK_NOLOAD 0x0400 86 #define TASK_NEW 0x0800 87 #define TASK_STATE_MAX 0x1000 88 89 /* Convenience macros for the sake of set_current_state: */ 90 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 91 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 92 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) 93 94 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 95 96 /* Convenience macros for the sake of wake_up(): */ 97 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 98 99 /* get_task_state(): */ 100 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 101 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 102 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 103 TASK_PARKED) 104 105 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0) 106 107 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0) 108 109 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) 110 111 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \ 112 (task->flags & PF_FROZEN) == 0 && \ 113 (task->state & TASK_NOLOAD) == 0) 114 115 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 116 117 /* 118 * Special states are those that do not use the normal wait-loop pattern. See 119 * the comment with set_special_state(). 120 */ 121 #define is_special_task_state(state) \ 122 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD)) 123 124 #define __set_current_state(state_value) \ 125 do { \ 126 WARN_ON_ONCE(is_special_task_state(state_value));\ 127 current->task_state_change = _THIS_IP_; \ 128 current->state = (state_value); \ 129 } while (0) 130 131 #define set_current_state(state_value) \ 132 do { \ 133 WARN_ON_ONCE(is_special_task_state(state_value));\ 134 current->task_state_change = _THIS_IP_; \ 135 smp_store_mb(current->state, (state_value)); \ 136 } while (0) 137 138 #define set_special_state(state_value) \ 139 do { \ 140 unsigned long flags; /* may shadow */ \ 141 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 142 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 143 current->task_state_change = _THIS_IP_; \ 144 current->state = (state_value); \ 145 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 146 } while (0) 147 #else 148 /* 149 * set_current_state() includes a barrier so that the write of current->state 150 * is correctly serialised wrt the caller's subsequent test of whether to 151 * actually sleep: 152 * 153 * for (;;) { 154 * set_current_state(TASK_UNINTERRUPTIBLE); 155 * if (!need_sleep) 156 * break; 157 * 158 * schedule(); 159 * } 160 * __set_current_state(TASK_RUNNING); 161 * 162 * If the caller does not need such serialisation (because, for instance, the 163 * condition test and condition change and wakeup are under the same lock) then 164 * use __set_current_state(). 165 * 166 * The above is typically ordered against the wakeup, which does: 167 * 168 * need_sleep = false; 169 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 170 * 171 * where wake_up_state() executes a full memory barrier before accessing the 172 * task state. 173 * 174 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 175 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 176 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 177 * 178 * However, with slightly different timing the wakeup TASK_RUNNING store can 179 * also collide with the TASK_UNINTERRUPTIBLE store. Loosing that store is not 180 * a problem either because that will result in one extra go around the loop 181 * and our @cond test will save the day. 182 * 183 * Also see the comments of try_to_wake_up(). 184 */ 185 #define __set_current_state(state_value) \ 186 current->state = (state_value) 187 188 #define set_current_state(state_value) \ 189 smp_store_mb(current->state, (state_value)) 190 191 /* 192 * set_special_state() should be used for those states when the blocking task 193 * can not use the regular condition based wait-loop. In that case we must 194 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores 195 * will not collide with our state change. 196 */ 197 #define set_special_state(state_value) \ 198 do { \ 199 unsigned long flags; /* may shadow */ \ 200 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 201 current->state = (state_value); \ 202 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 203 } while (0) 204 205 #endif 206 207 /* Task command name length: */ 208 #define TASK_COMM_LEN 16 209 210 extern void scheduler_tick(void); 211 212 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 213 214 extern long schedule_timeout(long timeout); 215 extern long schedule_timeout_interruptible(long timeout); 216 extern long schedule_timeout_killable(long timeout); 217 extern long schedule_timeout_uninterruptible(long timeout); 218 extern long schedule_timeout_idle(long timeout); 219 asmlinkage void schedule(void); 220 extern void schedule_preempt_disabled(void); 221 222 extern int __must_check io_schedule_prepare(void); 223 extern void io_schedule_finish(int token); 224 extern long io_schedule_timeout(long timeout); 225 extern void io_schedule(void); 226 227 /** 228 * struct prev_cputime - snapshot of system and user cputime 229 * @utime: time spent in user mode 230 * @stime: time spent in system mode 231 * @lock: protects the above two fields 232 * 233 * Stores previous user/system time values such that we can guarantee 234 * monotonicity. 235 */ 236 struct prev_cputime { 237 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 238 u64 utime; 239 u64 stime; 240 raw_spinlock_t lock; 241 #endif 242 }; 243 244 /** 245 * struct task_cputime - collected CPU time counts 246 * @utime: time spent in user mode, in nanoseconds 247 * @stime: time spent in kernel mode, in nanoseconds 248 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds 249 * 250 * This structure groups together three kinds of CPU time that are tracked for 251 * threads and thread groups. Most things considering CPU time want to group 252 * these counts together and treat all three of them in parallel. 253 */ 254 struct task_cputime { 255 u64 utime; 256 u64 stime; 257 unsigned long long sum_exec_runtime; 258 }; 259 260 /* Alternate field names when used on cache expirations: */ 261 #define virt_exp utime 262 #define prof_exp stime 263 #define sched_exp sum_exec_runtime 264 265 enum vtime_state { 266 /* Task is sleeping or running in a CPU with VTIME inactive: */ 267 VTIME_INACTIVE = 0, 268 /* Task runs in userspace in a CPU with VTIME active: */ 269 VTIME_USER, 270 /* Task runs in kernelspace in a CPU with VTIME active: */ 271 VTIME_SYS, 272 }; 273 274 struct vtime { 275 seqcount_t seqcount; 276 unsigned long long starttime; 277 enum vtime_state state; 278 u64 utime; 279 u64 stime; 280 u64 gtime; 281 }; 282 283 struct sched_info { 284 #ifdef CONFIG_SCHED_INFO 285 /* Cumulative counters: */ 286 287 /* # of times we have run on this CPU: */ 288 unsigned long pcount; 289 290 /* Time spent waiting on a runqueue: */ 291 unsigned long long run_delay; 292 293 /* Timestamps: */ 294 295 /* When did we last run on a CPU? */ 296 unsigned long long last_arrival; 297 298 /* When were we last queued to run? */ 299 unsigned long long last_queued; 300 301 #endif /* CONFIG_SCHED_INFO */ 302 }; 303 304 /* 305 * Integer metrics need fixed point arithmetic, e.g., sched/fair 306 * has a few: load, load_avg, util_avg, freq, and capacity. 307 * 308 * We define a basic fixed point arithmetic range, and then formalize 309 * all these metrics based on that basic range. 310 */ 311 # define SCHED_FIXEDPOINT_SHIFT 10 312 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 313 314 struct load_weight { 315 unsigned long weight; 316 u32 inv_weight; 317 }; 318 319 /** 320 * struct util_est - Estimation utilization of FAIR tasks 321 * @enqueued: instantaneous estimated utilization of a task/cpu 322 * @ewma: the Exponential Weighted Moving Average (EWMA) 323 * utilization of a task 324 * 325 * Support data structure to track an Exponential Weighted Moving Average 326 * (EWMA) of a FAIR task's utilization. New samples are added to the moving 327 * average each time a task completes an activation. Sample's weight is chosen 328 * so that the EWMA will be relatively insensitive to transient changes to the 329 * task's workload. 330 * 331 * The enqueued attribute has a slightly different meaning for tasks and cpus: 332 * - task: the task's util_avg at last task dequeue time 333 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU 334 * Thus, the util_est.enqueued of a task represents the contribution on the 335 * estimated utilization of the CPU where that task is currently enqueued. 336 * 337 * Only for tasks we track a moving average of the past instantaneous 338 * estimated utilization. This allows to absorb sporadic drops in utilization 339 * of an otherwise almost periodic task. 340 */ 341 struct util_est { 342 unsigned int enqueued; 343 unsigned int ewma; 344 #define UTIL_EST_WEIGHT_SHIFT 2 345 } __attribute__((__aligned__(sizeof(u64)))); 346 347 /* 348 * The load_avg/util_avg accumulates an infinite geometric series 349 * (see __update_load_avg() in kernel/sched/fair.c). 350 * 351 * [load_avg definition] 352 * 353 * load_avg = runnable% * scale_load_down(load) 354 * 355 * where runnable% is the time ratio that a sched_entity is runnable. 356 * For cfs_rq, it is the aggregated load_avg of all runnable and 357 * blocked sched_entities. 358 * 359 * load_avg may also take frequency scaling into account: 360 * 361 * load_avg = runnable% * scale_load_down(load) * freq% 362 * 363 * where freq% is the CPU frequency normalized to the highest frequency. 364 * 365 * [util_avg definition] 366 * 367 * util_avg = running% * SCHED_CAPACITY_SCALE 368 * 369 * where running% is the time ratio that a sched_entity is running on 370 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable 371 * and blocked sched_entities. 372 * 373 * util_avg may also factor frequency scaling and CPU capacity scaling: 374 * 375 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity% 376 * 377 * where freq% is the same as above, and capacity% is the CPU capacity 378 * normalized to the greatest capacity (due to uarch differences, etc). 379 * 380 * N.B., the above ratios (runnable%, running%, freq%, and capacity%) 381 * themselves are in the range of [0, 1]. To do fixed point arithmetics, 382 * we therefore scale them to as large a range as necessary. This is for 383 * example reflected by util_avg's SCHED_CAPACITY_SCALE. 384 * 385 * [Overflow issue] 386 * 387 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 388 * with the highest load (=88761), always runnable on a single cfs_rq, 389 * and should not overflow as the number already hits PID_MAX_LIMIT. 390 * 391 * For all other cases (including 32-bit kernels), struct load_weight's 392 * weight will overflow first before we do, because: 393 * 394 * Max(load_avg) <= Max(load.weight) 395 * 396 * Then it is the load_weight's responsibility to consider overflow 397 * issues. 398 */ 399 struct sched_avg { 400 u64 last_update_time; 401 u64 load_sum; 402 u64 runnable_load_sum; 403 u32 util_sum; 404 u32 period_contrib; 405 unsigned long load_avg; 406 unsigned long runnable_load_avg; 407 unsigned long util_avg; 408 struct util_est util_est; 409 } ____cacheline_aligned; 410 411 struct sched_statistics { 412 #ifdef CONFIG_SCHEDSTATS 413 u64 wait_start; 414 u64 wait_max; 415 u64 wait_count; 416 u64 wait_sum; 417 u64 iowait_count; 418 u64 iowait_sum; 419 420 u64 sleep_start; 421 u64 sleep_max; 422 s64 sum_sleep_runtime; 423 424 u64 block_start; 425 u64 block_max; 426 u64 exec_max; 427 u64 slice_max; 428 429 u64 nr_migrations_cold; 430 u64 nr_failed_migrations_affine; 431 u64 nr_failed_migrations_running; 432 u64 nr_failed_migrations_hot; 433 u64 nr_forced_migrations; 434 435 u64 nr_wakeups; 436 u64 nr_wakeups_sync; 437 u64 nr_wakeups_migrate; 438 u64 nr_wakeups_local; 439 u64 nr_wakeups_remote; 440 u64 nr_wakeups_affine; 441 u64 nr_wakeups_affine_attempts; 442 u64 nr_wakeups_passive; 443 u64 nr_wakeups_idle; 444 #endif 445 }; 446 447 struct sched_entity { 448 /* For load-balancing: */ 449 struct load_weight load; 450 unsigned long runnable_weight; 451 struct rb_node run_node; 452 struct list_head group_node; 453 unsigned int on_rq; 454 455 u64 exec_start; 456 u64 sum_exec_runtime; 457 u64 vruntime; 458 u64 prev_sum_exec_runtime; 459 460 u64 nr_migrations; 461 462 struct sched_statistics statistics; 463 464 #ifdef CONFIG_FAIR_GROUP_SCHED 465 int depth; 466 struct sched_entity *parent; 467 /* rq on which this entity is (to be) queued: */ 468 struct cfs_rq *cfs_rq; 469 /* rq "owned" by this entity/group: */ 470 struct cfs_rq *my_q; 471 #endif 472 473 #ifdef CONFIG_SMP 474 /* 475 * Per entity load average tracking. 476 * 477 * Put into separate cache line so it does not 478 * collide with read-mostly values above. 479 */ 480 struct sched_avg avg; 481 #endif 482 }; 483 484 struct sched_rt_entity { 485 struct list_head run_list; 486 unsigned long timeout; 487 unsigned long watchdog_stamp; 488 unsigned int time_slice; 489 unsigned short on_rq; 490 unsigned short on_list; 491 492 struct sched_rt_entity *back; 493 #ifdef CONFIG_RT_GROUP_SCHED 494 struct sched_rt_entity *parent; 495 /* rq on which this entity is (to be) queued: */ 496 struct rt_rq *rt_rq; 497 /* rq "owned" by this entity/group: */ 498 struct rt_rq *my_q; 499 #endif 500 } __randomize_layout; 501 502 struct sched_dl_entity { 503 struct rb_node rb_node; 504 505 /* 506 * Original scheduling parameters. Copied here from sched_attr 507 * during sched_setattr(), they will remain the same until 508 * the next sched_setattr(). 509 */ 510 u64 dl_runtime; /* Maximum runtime for each instance */ 511 u64 dl_deadline; /* Relative deadline of each instance */ 512 u64 dl_period; /* Separation of two instances (period) */ 513 u64 dl_bw; /* dl_runtime / dl_period */ 514 u64 dl_density; /* dl_runtime / dl_deadline */ 515 516 /* 517 * Actual scheduling parameters. Initialized with the values above, 518 * they are continously updated during task execution. Note that 519 * the remaining runtime could be < 0 in case we are in overrun. 520 */ 521 s64 runtime; /* Remaining runtime for this instance */ 522 u64 deadline; /* Absolute deadline for this instance */ 523 unsigned int flags; /* Specifying the scheduler behaviour */ 524 525 /* 526 * Some bool flags: 527 * 528 * @dl_throttled tells if we exhausted the runtime. If so, the 529 * task has to wait for a replenishment to be performed at the 530 * next firing of dl_timer. 531 * 532 * @dl_boosted tells if we are boosted due to DI. If so we are 533 * outside bandwidth enforcement mechanism (but only until we 534 * exit the critical section); 535 * 536 * @dl_yielded tells if task gave up the CPU before consuming 537 * all its available runtime during the last job. 538 * 539 * @dl_non_contending tells if the task is inactive while still 540 * contributing to the active utilization. In other words, it 541 * indicates if the inactive timer has been armed and its handler 542 * has not been executed yet. This flag is useful to avoid race 543 * conditions between the inactive timer handler and the wakeup 544 * code. 545 * 546 * @dl_overrun tells if the task asked to be informed about runtime 547 * overruns. 548 */ 549 unsigned int dl_throttled : 1; 550 unsigned int dl_boosted : 1; 551 unsigned int dl_yielded : 1; 552 unsigned int dl_non_contending : 1; 553 unsigned int dl_overrun : 1; 554 555 /* 556 * Bandwidth enforcement timer. Each -deadline task has its 557 * own bandwidth to be enforced, thus we need one timer per task. 558 */ 559 struct hrtimer dl_timer; 560 561 /* 562 * Inactive timer, responsible for decreasing the active utilization 563 * at the "0-lag time". When a -deadline task blocks, it contributes 564 * to GRUB's active utilization until the "0-lag time", hence a 565 * timer is needed to decrease the active utilization at the correct 566 * time. 567 */ 568 struct hrtimer inactive_timer; 569 }; 570 571 union rcu_special { 572 struct { 573 u8 blocked; 574 u8 need_qs; 575 } b; /* Bits. */ 576 u16 s; /* Set of bits. */ 577 }; 578 579 enum perf_event_task_context { 580 perf_invalid_context = -1, 581 perf_hw_context = 0, 582 perf_sw_context, 583 perf_nr_task_contexts, 584 }; 585 586 struct wake_q_node { 587 struct wake_q_node *next; 588 }; 589 590 struct task_struct { 591 #ifdef CONFIG_THREAD_INFO_IN_TASK 592 /* 593 * For reasons of header soup (see current_thread_info()), this 594 * must be the first element of task_struct. 595 */ 596 struct thread_info thread_info; 597 #endif 598 /* -1 unrunnable, 0 runnable, >0 stopped: */ 599 volatile long state; 600 601 /* 602 * This begins the randomizable portion of task_struct. Only 603 * scheduling-critical items should be added above here. 604 */ 605 randomized_struct_fields_start 606 607 void *stack; 608 atomic_t usage; 609 /* Per task flags (PF_*), defined further below: */ 610 unsigned int flags; 611 unsigned int ptrace; 612 613 #ifdef CONFIG_SMP 614 struct llist_node wake_entry; 615 int on_cpu; 616 #ifdef CONFIG_THREAD_INFO_IN_TASK 617 /* Current CPU: */ 618 unsigned int cpu; 619 #endif 620 unsigned int wakee_flips; 621 unsigned long wakee_flip_decay_ts; 622 struct task_struct *last_wakee; 623 624 /* 625 * recent_used_cpu is initially set as the last CPU used by a task 626 * that wakes affine another task. Waker/wakee relationships can 627 * push tasks around a CPU where each wakeup moves to the next one. 628 * Tracking a recently used CPU allows a quick search for a recently 629 * used CPU that may be idle. 630 */ 631 int recent_used_cpu; 632 int wake_cpu; 633 #endif 634 int on_rq; 635 636 int prio; 637 int static_prio; 638 int normal_prio; 639 unsigned int rt_priority; 640 641 const struct sched_class *sched_class; 642 struct sched_entity se; 643 struct sched_rt_entity rt; 644 #ifdef CONFIG_CGROUP_SCHED 645 struct task_group *sched_task_group; 646 #endif 647 struct sched_dl_entity dl; 648 649 #ifdef CONFIG_PREEMPT_NOTIFIERS 650 /* List of struct preempt_notifier: */ 651 struct hlist_head preempt_notifiers; 652 #endif 653 654 #ifdef CONFIG_BLK_DEV_IO_TRACE 655 unsigned int btrace_seq; 656 #endif 657 658 unsigned int policy; 659 int nr_cpus_allowed; 660 cpumask_t cpus_allowed; 661 662 #ifdef CONFIG_PREEMPT_RCU 663 int rcu_read_lock_nesting; 664 union rcu_special rcu_read_unlock_special; 665 struct list_head rcu_node_entry; 666 struct rcu_node *rcu_blocked_node; 667 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 668 669 #ifdef CONFIG_TASKS_RCU 670 unsigned long rcu_tasks_nvcsw; 671 u8 rcu_tasks_holdout; 672 u8 rcu_tasks_idx; 673 int rcu_tasks_idle_cpu; 674 struct list_head rcu_tasks_holdout_list; 675 #endif /* #ifdef CONFIG_TASKS_RCU */ 676 677 struct sched_info sched_info; 678 679 struct list_head tasks; 680 #ifdef CONFIG_SMP 681 struct plist_node pushable_tasks; 682 struct rb_node pushable_dl_tasks; 683 #endif 684 685 struct mm_struct *mm; 686 struct mm_struct *active_mm; 687 688 /* Per-thread vma caching: */ 689 struct vmacache vmacache; 690 691 #ifdef SPLIT_RSS_COUNTING 692 struct task_rss_stat rss_stat; 693 #endif 694 int exit_state; 695 int exit_code; 696 int exit_signal; 697 /* The signal sent when the parent dies: */ 698 int pdeath_signal; 699 /* JOBCTL_*, siglock protected: */ 700 unsigned long jobctl; 701 702 /* Used for emulating ABI behavior of previous Linux versions: */ 703 unsigned int personality; 704 705 /* Scheduler bits, serialized by scheduler locks: */ 706 unsigned sched_reset_on_fork:1; 707 unsigned sched_contributes_to_load:1; 708 unsigned sched_migrated:1; 709 unsigned sched_remote_wakeup:1; 710 #ifdef CONFIG_PSI 711 unsigned sched_psi_wake_requeue:1; 712 #endif 713 714 /* Force alignment to the next boundary: */ 715 unsigned :0; 716 717 /* Unserialized, strictly 'current' */ 718 719 /* Bit to tell LSMs we're in execve(): */ 720 unsigned in_execve:1; 721 unsigned in_iowait:1; 722 #ifndef TIF_RESTORE_SIGMASK 723 unsigned restore_sigmask:1; 724 #endif 725 #ifdef CONFIG_MEMCG 726 unsigned in_user_fault:1; 727 #endif 728 #ifdef CONFIG_COMPAT_BRK 729 unsigned brk_randomized:1; 730 #endif 731 #ifdef CONFIG_CGROUPS 732 /* disallow userland-initiated cgroup migration */ 733 unsigned no_cgroup_migration:1; 734 #endif 735 #ifdef CONFIG_BLK_CGROUP 736 /* to be used once the psi infrastructure lands upstream. */ 737 unsigned use_memdelay:1; 738 #endif 739 740 /* 741 * May usercopy functions fault on kernel addresses? 742 * This is not just a single bit because this can potentially nest. 743 */ 744 unsigned int kernel_uaccess_faults_ok; 745 746 unsigned long atomic_flags; /* Flags requiring atomic access. */ 747 748 struct restart_block restart_block; 749 750 pid_t pid; 751 pid_t tgid; 752 753 #ifdef CONFIG_STACKPROTECTOR 754 /* Canary value for the -fstack-protector GCC feature: */ 755 unsigned long stack_canary; 756 #endif 757 /* 758 * Pointers to the (original) parent process, youngest child, younger sibling, 759 * older sibling, respectively. (p->father can be replaced with 760 * p->real_parent->pid) 761 */ 762 763 /* Real parent process: */ 764 struct task_struct __rcu *real_parent; 765 766 /* Recipient of SIGCHLD, wait4() reports: */ 767 struct task_struct __rcu *parent; 768 769 /* 770 * Children/sibling form the list of natural children: 771 */ 772 struct list_head children; 773 struct list_head sibling; 774 struct task_struct *group_leader; 775 776 /* 777 * 'ptraced' is the list of tasks this task is using ptrace() on. 778 * 779 * This includes both natural children and PTRACE_ATTACH targets. 780 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 781 */ 782 struct list_head ptraced; 783 struct list_head ptrace_entry; 784 785 /* PID/PID hash table linkage. */ 786 struct pid *thread_pid; 787 struct hlist_node pid_links[PIDTYPE_MAX]; 788 struct list_head thread_group; 789 struct list_head thread_node; 790 791 struct completion *vfork_done; 792 793 /* CLONE_CHILD_SETTID: */ 794 int __user *set_child_tid; 795 796 /* CLONE_CHILD_CLEARTID: */ 797 int __user *clear_child_tid; 798 799 u64 utime; 800 u64 stime; 801 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 802 u64 utimescaled; 803 u64 stimescaled; 804 #endif 805 u64 gtime; 806 struct prev_cputime prev_cputime; 807 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 808 struct vtime vtime; 809 #endif 810 811 #ifdef CONFIG_NO_HZ_FULL 812 atomic_t tick_dep_mask; 813 #endif 814 /* Context switch counts: */ 815 unsigned long nvcsw; 816 unsigned long nivcsw; 817 818 /* Monotonic time in nsecs: */ 819 u64 start_time; 820 821 /* Boot based time in nsecs: */ 822 u64 real_start_time; 823 824 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 825 unsigned long min_flt; 826 unsigned long maj_flt; 827 828 #ifdef CONFIG_POSIX_TIMERS 829 struct task_cputime cputime_expires; 830 struct list_head cpu_timers[3]; 831 #endif 832 833 /* Process credentials: */ 834 835 /* Tracer's credentials at attach: */ 836 const struct cred __rcu *ptracer_cred; 837 838 /* Objective and real subjective task credentials (COW): */ 839 const struct cred __rcu *real_cred; 840 841 /* Effective (overridable) subjective task credentials (COW): */ 842 const struct cred __rcu *cred; 843 844 /* 845 * executable name, excluding path. 846 * 847 * - normally initialized setup_new_exec() 848 * - access it with [gs]et_task_comm() 849 * - lock it with task_lock() 850 */ 851 char comm[TASK_COMM_LEN]; 852 853 struct nameidata *nameidata; 854 855 #ifdef CONFIG_SYSVIPC 856 struct sysv_sem sysvsem; 857 struct sysv_shm sysvshm; 858 #endif 859 #ifdef CONFIG_DETECT_HUNG_TASK 860 unsigned long last_switch_count; 861 unsigned long last_switch_time; 862 #endif 863 /* Filesystem information: */ 864 struct fs_struct *fs; 865 866 /* Open file information: */ 867 struct files_struct *files; 868 869 /* Namespaces: */ 870 struct nsproxy *nsproxy; 871 872 /* Signal handlers: */ 873 struct signal_struct *signal; 874 struct sighand_struct *sighand; 875 sigset_t blocked; 876 sigset_t real_blocked; 877 /* Restored if set_restore_sigmask() was used: */ 878 sigset_t saved_sigmask; 879 struct sigpending pending; 880 unsigned long sas_ss_sp; 881 size_t sas_ss_size; 882 unsigned int sas_ss_flags; 883 884 struct callback_head *task_works; 885 886 struct audit_context *audit_context; 887 #ifdef CONFIG_AUDITSYSCALL 888 kuid_t loginuid; 889 unsigned int sessionid; 890 #endif 891 struct seccomp seccomp; 892 893 /* Thread group tracking: */ 894 u32 parent_exec_id; 895 u32 self_exec_id; 896 897 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 898 spinlock_t alloc_lock; 899 900 /* Protection of the PI data structures: */ 901 raw_spinlock_t pi_lock; 902 903 struct wake_q_node wake_q; 904 905 #ifdef CONFIG_RT_MUTEXES 906 /* PI waiters blocked on a rt_mutex held by this task: */ 907 struct rb_root_cached pi_waiters; 908 /* Updated under owner's pi_lock and rq lock */ 909 struct task_struct *pi_top_task; 910 /* Deadlock detection and priority inheritance handling: */ 911 struct rt_mutex_waiter *pi_blocked_on; 912 #endif 913 914 #ifdef CONFIG_DEBUG_MUTEXES 915 /* Mutex deadlock detection: */ 916 struct mutex_waiter *blocked_on; 917 #endif 918 919 #ifdef CONFIG_TRACE_IRQFLAGS 920 unsigned int irq_events; 921 unsigned long hardirq_enable_ip; 922 unsigned long hardirq_disable_ip; 923 unsigned int hardirq_enable_event; 924 unsigned int hardirq_disable_event; 925 int hardirqs_enabled; 926 int hardirq_context; 927 unsigned long softirq_disable_ip; 928 unsigned long softirq_enable_ip; 929 unsigned int softirq_disable_event; 930 unsigned int softirq_enable_event; 931 int softirqs_enabled; 932 int softirq_context; 933 #endif 934 935 #ifdef CONFIG_LOCKDEP 936 # define MAX_LOCK_DEPTH 48UL 937 u64 curr_chain_key; 938 int lockdep_depth; 939 unsigned int lockdep_recursion; 940 struct held_lock held_locks[MAX_LOCK_DEPTH]; 941 #endif 942 943 #ifdef CONFIG_UBSAN 944 unsigned int in_ubsan; 945 #endif 946 947 /* Journalling filesystem info: */ 948 void *journal_info; 949 950 /* Stacked block device info: */ 951 struct bio_list *bio_list; 952 953 #ifdef CONFIG_BLOCK 954 /* Stack plugging: */ 955 struct blk_plug *plug; 956 #endif 957 958 /* VM state: */ 959 struct reclaim_state *reclaim_state; 960 961 struct backing_dev_info *backing_dev_info; 962 963 struct io_context *io_context; 964 965 /* Ptrace state: */ 966 unsigned long ptrace_message; 967 kernel_siginfo_t *last_siginfo; 968 969 struct task_io_accounting ioac; 970 #ifdef CONFIG_PSI 971 /* Pressure stall state */ 972 unsigned int psi_flags; 973 #endif 974 #ifdef CONFIG_TASK_XACCT 975 /* Accumulated RSS usage: */ 976 u64 acct_rss_mem1; 977 /* Accumulated virtual memory usage: */ 978 u64 acct_vm_mem1; 979 /* stime + utime since last update: */ 980 u64 acct_timexpd; 981 #endif 982 #ifdef CONFIG_CPUSETS 983 /* Protected by ->alloc_lock: */ 984 nodemask_t mems_allowed; 985 /* Seqence number to catch updates: */ 986 seqcount_t mems_allowed_seq; 987 int cpuset_mem_spread_rotor; 988 int cpuset_slab_spread_rotor; 989 #endif 990 #ifdef CONFIG_CGROUPS 991 /* Control Group info protected by css_set_lock: */ 992 struct css_set __rcu *cgroups; 993 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 994 struct list_head cg_list; 995 #endif 996 #ifdef CONFIG_INTEL_RDT 997 u32 closid; 998 u32 rmid; 999 #endif 1000 #ifdef CONFIG_FUTEX 1001 struct robust_list_head __user *robust_list; 1002 #ifdef CONFIG_COMPAT 1003 struct compat_robust_list_head __user *compat_robust_list; 1004 #endif 1005 struct list_head pi_state_list; 1006 struct futex_pi_state *pi_state_cache; 1007 #endif 1008 #ifdef CONFIG_PERF_EVENTS 1009 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 1010 struct mutex perf_event_mutex; 1011 struct list_head perf_event_list; 1012 #endif 1013 #ifdef CONFIG_DEBUG_PREEMPT 1014 unsigned long preempt_disable_ip; 1015 #endif 1016 #ifdef CONFIG_NUMA 1017 /* Protected by alloc_lock: */ 1018 struct mempolicy *mempolicy; 1019 short il_prev; 1020 short pref_node_fork; 1021 #endif 1022 #ifdef CONFIG_NUMA_BALANCING 1023 int numa_scan_seq; 1024 unsigned int numa_scan_period; 1025 unsigned int numa_scan_period_max; 1026 int numa_preferred_nid; 1027 unsigned long numa_migrate_retry; 1028 /* Migration stamp: */ 1029 u64 node_stamp; 1030 u64 last_task_numa_placement; 1031 u64 last_sum_exec_runtime; 1032 struct callback_head numa_work; 1033 1034 struct numa_group *numa_group; 1035 1036 /* 1037 * numa_faults is an array split into four regions: 1038 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1039 * in this precise order. 1040 * 1041 * faults_memory: Exponential decaying average of faults on a per-node 1042 * basis. Scheduling placement decisions are made based on these 1043 * counts. The values remain static for the duration of a PTE scan. 1044 * faults_cpu: Track the nodes the process was running on when a NUMA 1045 * hinting fault was incurred. 1046 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1047 * during the current scan window. When the scan completes, the counts 1048 * in faults_memory and faults_cpu decay and these values are copied. 1049 */ 1050 unsigned long *numa_faults; 1051 unsigned long total_numa_faults; 1052 1053 /* 1054 * numa_faults_locality tracks if faults recorded during the last 1055 * scan window were remote/local or failed to migrate. The task scan 1056 * period is adapted based on the locality of the faults with different 1057 * weights depending on whether they were shared or private faults 1058 */ 1059 unsigned long numa_faults_locality[3]; 1060 1061 unsigned long numa_pages_migrated; 1062 #endif /* CONFIG_NUMA_BALANCING */ 1063 1064 #ifdef CONFIG_RSEQ 1065 struct rseq __user *rseq; 1066 u32 rseq_len; 1067 u32 rseq_sig; 1068 /* 1069 * RmW on rseq_event_mask must be performed atomically 1070 * with respect to preemption. 1071 */ 1072 unsigned long rseq_event_mask; 1073 #endif 1074 1075 struct tlbflush_unmap_batch tlb_ubc; 1076 1077 struct rcu_head rcu; 1078 1079 /* Cache last used pipe for splice(): */ 1080 struct pipe_inode_info *splice_pipe; 1081 1082 struct page_frag task_frag; 1083 1084 #ifdef CONFIG_TASK_DELAY_ACCT 1085 struct task_delay_info *delays; 1086 #endif 1087 1088 #ifdef CONFIG_FAULT_INJECTION 1089 int make_it_fail; 1090 unsigned int fail_nth; 1091 #endif 1092 /* 1093 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1094 * balance_dirty_pages() for a dirty throttling pause: 1095 */ 1096 int nr_dirtied; 1097 int nr_dirtied_pause; 1098 /* Start of a write-and-pause period: */ 1099 unsigned long dirty_paused_when; 1100 1101 #ifdef CONFIG_LATENCYTOP 1102 int latency_record_count; 1103 struct latency_record latency_record[LT_SAVECOUNT]; 1104 #endif 1105 /* 1106 * Time slack values; these are used to round up poll() and 1107 * select() etc timeout values. These are in nanoseconds. 1108 */ 1109 u64 timer_slack_ns; 1110 u64 default_timer_slack_ns; 1111 1112 #ifdef CONFIG_KASAN 1113 unsigned int kasan_depth; 1114 #endif 1115 1116 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1117 /* Index of current stored address in ret_stack: */ 1118 int curr_ret_stack; 1119 1120 /* Stack of return addresses for return function tracing: */ 1121 struct ftrace_ret_stack *ret_stack; 1122 1123 /* Timestamp for last schedule: */ 1124 unsigned long long ftrace_timestamp; 1125 1126 /* 1127 * Number of functions that haven't been traced 1128 * because of depth overrun: 1129 */ 1130 atomic_t trace_overrun; 1131 1132 /* Pause tracing: */ 1133 atomic_t tracing_graph_pause; 1134 #endif 1135 1136 #ifdef CONFIG_TRACING 1137 /* State flags for use by tracers: */ 1138 unsigned long trace; 1139 1140 /* Bitmask and counter of trace recursion: */ 1141 unsigned long trace_recursion; 1142 #endif /* CONFIG_TRACING */ 1143 1144 #ifdef CONFIG_KCOV 1145 /* Coverage collection mode enabled for this task (0 if disabled): */ 1146 unsigned int kcov_mode; 1147 1148 /* Size of the kcov_area: */ 1149 unsigned int kcov_size; 1150 1151 /* Buffer for coverage collection: */ 1152 void *kcov_area; 1153 1154 /* KCOV descriptor wired with this task or NULL: */ 1155 struct kcov *kcov; 1156 #endif 1157 1158 #ifdef CONFIG_MEMCG 1159 struct mem_cgroup *memcg_in_oom; 1160 gfp_t memcg_oom_gfp_mask; 1161 int memcg_oom_order; 1162 1163 /* Number of pages to reclaim on returning to userland: */ 1164 unsigned int memcg_nr_pages_over_high; 1165 1166 /* Used by memcontrol for targeted memcg charge: */ 1167 struct mem_cgroup *active_memcg; 1168 #endif 1169 1170 #ifdef CONFIG_BLK_CGROUP 1171 struct request_queue *throttle_queue; 1172 #endif 1173 1174 #ifdef CONFIG_UPROBES 1175 struct uprobe_task *utask; 1176 #endif 1177 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1178 unsigned int sequential_io; 1179 unsigned int sequential_io_avg; 1180 #endif 1181 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1182 unsigned long task_state_change; 1183 #endif 1184 int pagefault_disabled; 1185 #ifdef CONFIG_MMU 1186 struct task_struct *oom_reaper_list; 1187 #endif 1188 #ifdef CONFIG_VMAP_STACK 1189 struct vm_struct *stack_vm_area; 1190 #endif 1191 #ifdef CONFIG_THREAD_INFO_IN_TASK 1192 /* A live task holds one reference: */ 1193 atomic_t stack_refcount; 1194 #endif 1195 #ifdef CONFIG_LIVEPATCH 1196 int patch_state; 1197 #endif 1198 #ifdef CONFIG_SECURITY 1199 /* Used by LSM modules for access restriction: */ 1200 void *security; 1201 #endif 1202 1203 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1204 unsigned long lowest_stack; 1205 unsigned long prev_lowest_stack; 1206 #endif 1207 1208 /* 1209 * New fields for task_struct should be added above here, so that 1210 * they are included in the randomized portion of task_struct. 1211 */ 1212 randomized_struct_fields_end 1213 1214 /* CPU-specific state of this task: */ 1215 struct thread_struct thread; 1216 1217 /* 1218 * WARNING: on x86, 'thread_struct' contains a variable-sized 1219 * structure. It *MUST* be at the end of 'task_struct'. 1220 * 1221 * Do not put anything below here! 1222 */ 1223 }; 1224 1225 static inline struct pid *task_pid(struct task_struct *task) 1226 { 1227 return task->thread_pid; 1228 } 1229 1230 /* 1231 * the helpers to get the task's different pids as they are seen 1232 * from various namespaces 1233 * 1234 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1235 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1236 * current. 1237 * task_xid_nr_ns() : id seen from the ns specified; 1238 * 1239 * see also pid_nr() etc in include/linux/pid.h 1240 */ 1241 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1242 1243 static inline pid_t task_pid_nr(struct task_struct *tsk) 1244 { 1245 return tsk->pid; 1246 } 1247 1248 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1249 { 1250 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1251 } 1252 1253 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1254 { 1255 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1256 } 1257 1258 1259 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1260 { 1261 return tsk->tgid; 1262 } 1263 1264 /** 1265 * pid_alive - check that a task structure is not stale 1266 * @p: Task structure to be checked. 1267 * 1268 * Test if a process is not yet dead (at most zombie state) 1269 * If pid_alive fails, then pointers within the task structure 1270 * can be stale and must not be dereferenced. 1271 * 1272 * Return: 1 if the process is alive. 0 otherwise. 1273 */ 1274 static inline int pid_alive(const struct task_struct *p) 1275 { 1276 return p->thread_pid != NULL; 1277 } 1278 1279 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1280 { 1281 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1282 } 1283 1284 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1285 { 1286 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1287 } 1288 1289 1290 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1291 { 1292 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1293 } 1294 1295 static inline pid_t task_session_vnr(struct task_struct *tsk) 1296 { 1297 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1298 } 1299 1300 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1301 { 1302 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns); 1303 } 1304 1305 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1306 { 1307 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL); 1308 } 1309 1310 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1311 { 1312 pid_t pid = 0; 1313 1314 rcu_read_lock(); 1315 if (pid_alive(tsk)) 1316 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1317 rcu_read_unlock(); 1318 1319 return pid; 1320 } 1321 1322 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1323 { 1324 return task_ppid_nr_ns(tsk, &init_pid_ns); 1325 } 1326 1327 /* Obsolete, do not use: */ 1328 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1329 { 1330 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1331 } 1332 1333 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1334 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1335 1336 static inline unsigned int task_state_index(struct task_struct *tsk) 1337 { 1338 unsigned int tsk_state = READ_ONCE(tsk->state); 1339 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT; 1340 1341 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1342 1343 if (tsk_state == TASK_IDLE) 1344 state = TASK_REPORT_IDLE; 1345 1346 return fls(state); 1347 } 1348 1349 static inline char task_index_to_char(unsigned int state) 1350 { 1351 static const char state_char[] = "RSDTtXZPI"; 1352 1353 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1354 1355 return state_char[state]; 1356 } 1357 1358 static inline char task_state_to_char(struct task_struct *tsk) 1359 { 1360 return task_index_to_char(task_state_index(tsk)); 1361 } 1362 1363 /** 1364 * is_global_init - check if a task structure is init. Since init 1365 * is free to have sub-threads we need to check tgid. 1366 * @tsk: Task structure to be checked. 1367 * 1368 * Check if a task structure is the first user space task the kernel created. 1369 * 1370 * Return: 1 if the task structure is init. 0 otherwise. 1371 */ 1372 static inline int is_global_init(struct task_struct *tsk) 1373 { 1374 return task_tgid_nr(tsk) == 1; 1375 } 1376 1377 extern struct pid *cad_pid; 1378 1379 /* 1380 * Per process flags 1381 */ 1382 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1383 #define PF_EXITING 0x00000004 /* Getting shut down */ 1384 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */ 1385 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */ 1386 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1387 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1388 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1389 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1390 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1391 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1392 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1393 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1394 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1395 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1396 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1397 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1398 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1399 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1400 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1401 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ 1402 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1403 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1404 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1405 #define PF_MEMSTALL 0x01000000 /* Stalled due to lack of memory */ 1406 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */ 1407 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1408 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */ 1409 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1410 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1411 1412 /* 1413 * Only the _current_ task can read/write to tsk->flags, but other 1414 * tasks can access tsk->flags in readonly mode for example 1415 * with tsk_used_math (like during threaded core dumping). 1416 * There is however an exception to this rule during ptrace 1417 * or during fork: the ptracer task is allowed to write to the 1418 * child->flags of its traced child (same goes for fork, the parent 1419 * can write to the child->flags), because we're guaranteed the 1420 * child is not running and in turn not changing child->flags 1421 * at the same time the parent does it. 1422 */ 1423 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1424 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1425 #define clear_used_math() clear_stopped_child_used_math(current) 1426 #define set_used_math() set_stopped_child_used_math(current) 1427 1428 #define conditional_stopped_child_used_math(condition, child) \ 1429 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1430 1431 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1432 1433 #define copy_to_stopped_child_used_math(child) \ 1434 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1435 1436 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1437 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1438 #define used_math() tsk_used_math(current) 1439 1440 static inline bool is_percpu_thread(void) 1441 { 1442 #ifdef CONFIG_SMP 1443 return (current->flags & PF_NO_SETAFFINITY) && 1444 (current->nr_cpus_allowed == 1); 1445 #else 1446 return true; 1447 #endif 1448 } 1449 1450 /* Per-process atomic flags. */ 1451 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1452 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1453 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1454 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1455 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1456 1457 #define TASK_PFA_TEST(name, func) \ 1458 static inline bool task_##func(struct task_struct *p) \ 1459 { return test_bit(PFA_##name, &p->atomic_flags); } 1460 1461 #define TASK_PFA_SET(name, func) \ 1462 static inline void task_set_##func(struct task_struct *p) \ 1463 { set_bit(PFA_##name, &p->atomic_flags); } 1464 1465 #define TASK_PFA_CLEAR(name, func) \ 1466 static inline void task_clear_##func(struct task_struct *p) \ 1467 { clear_bit(PFA_##name, &p->atomic_flags); } 1468 1469 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1470 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1471 1472 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1473 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1474 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1475 1476 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1477 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1478 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1479 1480 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1481 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1482 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1483 1484 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1485 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1486 1487 static inline void 1488 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1489 { 1490 current->flags &= ~flags; 1491 current->flags |= orig_flags & flags; 1492 } 1493 1494 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1495 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1496 #ifdef CONFIG_SMP 1497 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1498 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1499 #else 1500 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1501 { 1502 } 1503 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1504 { 1505 if (!cpumask_test_cpu(0, new_mask)) 1506 return -EINVAL; 1507 return 0; 1508 } 1509 #endif 1510 1511 #ifndef cpu_relax_yield 1512 #define cpu_relax_yield() cpu_relax() 1513 #endif 1514 1515 extern int yield_to(struct task_struct *p, bool preempt); 1516 extern void set_user_nice(struct task_struct *p, long nice); 1517 extern int task_prio(const struct task_struct *p); 1518 1519 /** 1520 * task_nice - return the nice value of a given task. 1521 * @p: the task in question. 1522 * 1523 * Return: The nice value [ -20 ... 0 ... 19 ]. 1524 */ 1525 static inline int task_nice(const struct task_struct *p) 1526 { 1527 return PRIO_TO_NICE((p)->static_prio); 1528 } 1529 1530 extern int can_nice(const struct task_struct *p, const int nice); 1531 extern int task_curr(const struct task_struct *p); 1532 extern int idle_cpu(int cpu); 1533 extern int available_idle_cpu(int cpu); 1534 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1535 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1536 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1537 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1538 extern struct task_struct *idle_task(int cpu); 1539 1540 /** 1541 * is_idle_task - is the specified task an idle task? 1542 * @p: the task in question. 1543 * 1544 * Return: 1 if @p is an idle task. 0 otherwise. 1545 */ 1546 static inline bool is_idle_task(const struct task_struct *p) 1547 { 1548 return !!(p->flags & PF_IDLE); 1549 } 1550 1551 extern struct task_struct *curr_task(int cpu); 1552 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1553 1554 void yield(void); 1555 1556 union thread_union { 1557 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1558 struct task_struct task; 1559 #endif 1560 #ifndef CONFIG_THREAD_INFO_IN_TASK 1561 struct thread_info thread_info; 1562 #endif 1563 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1564 }; 1565 1566 #ifndef CONFIG_THREAD_INFO_IN_TASK 1567 extern struct thread_info init_thread_info; 1568 #endif 1569 1570 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1571 1572 #ifdef CONFIG_THREAD_INFO_IN_TASK 1573 static inline struct thread_info *task_thread_info(struct task_struct *task) 1574 { 1575 return &task->thread_info; 1576 } 1577 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1578 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1579 #endif 1580 1581 /* 1582 * find a task by one of its numerical ids 1583 * 1584 * find_task_by_pid_ns(): 1585 * finds a task by its pid in the specified namespace 1586 * find_task_by_vpid(): 1587 * finds a task by its virtual pid 1588 * 1589 * see also find_vpid() etc in include/linux/pid.h 1590 */ 1591 1592 extern struct task_struct *find_task_by_vpid(pid_t nr); 1593 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1594 1595 /* 1596 * find a task by its virtual pid and get the task struct 1597 */ 1598 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1599 1600 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1601 extern int wake_up_process(struct task_struct *tsk); 1602 extern void wake_up_new_task(struct task_struct *tsk); 1603 1604 #ifdef CONFIG_SMP 1605 extern void kick_process(struct task_struct *tsk); 1606 #else 1607 static inline void kick_process(struct task_struct *tsk) { } 1608 #endif 1609 1610 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1611 1612 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1613 { 1614 __set_task_comm(tsk, from, false); 1615 } 1616 1617 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1618 #define get_task_comm(buf, tsk) ({ \ 1619 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1620 __get_task_comm(buf, sizeof(buf), tsk); \ 1621 }) 1622 1623 #ifdef CONFIG_SMP 1624 void scheduler_ipi(void); 1625 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 1626 #else 1627 static inline void scheduler_ipi(void) { } 1628 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) 1629 { 1630 return 1; 1631 } 1632 #endif 1633 1634 /* 1635 * Set thread flags in other task's structures. 1636 * See asm/thread_info.h for TIF_xxxx flags available: 1637 */ 1638 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1639 { 1640 set_ti_thread_flag(task_thread_info(tsk), flag); 1641 } 1642 1643 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1644 { 1645 clear_ti_thread_flag(task_thread_info(tsk), flag); 1646 } 1647 1648 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1649 bool value) 1650 { 1651 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1652 } 1653 1654 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1655 { 1656 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1657 } 1658 1659 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1660 { 1661 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1662 } 1663 1664 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1665 { 1666 return test_ti_thread_flag(task_thread_info(tsk), flag); 1667 } 1668 1669 static inline void set_tsk_need_resched(struct task_struct *tsk) 1670 { 1671 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1672 } 1673 1674 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1675 { 1676 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1677 } 1678 1679 static inline int test_tsk_need_resched(struct task_struct *tsk) 1680 { 1681 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1682 } 1683 1684 /* 1685 * cond_resched() and cond_resched_lock(): latency reduction via 1686 * explicit rescheduling in places that are safe. The return 1687 * value indicates whether a reschedule was done in fact. 1688 * cond_resched_lock() will drop the spinlock before scheduling, 1689 */ 1690 #ifndef CONFIG_PREEMPT 1691 extern int _cond_resched(void); 1692 #else 1693 static inline int _cond_resched(void) { return 0; } 1694 #endif 1695 1696 #define cond_resched() ({ \ 1697 ___might_sleep(__FILE__, __LINE__, 0); \ 1698 _cond_resched(); \ 1699 }) 1700 1701 extern int __cond_resched_lock(spinlock_t *lock); 1702 1703 #define cond_resched_lock(lock) ({ \ 1704 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ 1705 __cond_resched_lock(lock); \ 1706 }) 1707 1708 static inline void cond_resched_rcu(void) 1709 { 1710 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 1711 rcu_read_unlock(); 1712 cond_resched(); 1713 rcu_read_lock(); 1714 #endif 1715 } 1716 1717 /* 1718 * Does a critical section need to be broken due to another 1719 * task waiting?: (technically does not depend on CONFIG_PREEMPT, 1720 * but a general need for low latency) 1721 */ 1722 static inline int spin_needbreak(spinlock_t *lock) 1723 { 1724 #ifdef CONFIG_PREEMPT 1725 return spin_is_contended(lock); 1726 #else 1727 return 0; 1728 #endif 1729 } 1730 1731 static __always_inline bool need_resched(void) 1732 { 1733 return unlikely(tif_need_resched()); 1734 } 1735 1736 /* 1737 * Wrappers for p->thread_info->cpu access. No-op on UP. 1738 */ 1739 #ifdef CONFIG_SMP 1740 1741 static inline unsigned int task_cpu(const struct task_struct *p) 1742 { 1743 #ifdef CONFIG_THREAD_INFO_IN_TASK 1744 return p->cpu; 1745 #else 1746 return task_thread_info(p)->cpu; 1747 #endif 1748 } 1749 1750 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 1751 1752 #else 1753 1754 static inline unsigned int task_cpu(const struct task_struct *p) 1755 { 1756 return 0; 1757 } 1758 1759 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 1760 { 1761 } 1762 1763 #endif /* CONFIG_SMP */ 1764 1765 /* 1766 * In order to reduce various lock holder preemption latencies provide an 1767 * interface to see if a vCPU is currently running or not. 1768 * 1769 * This allows us to terminate optimistic spin loops and block, analogous to 1770 * the native optimistic spin heuristic of testing if the lock owner task is 1771 * running or not. 1772 */ 1773 #ifndef vcpu_is_preempted 1774 # define vcpu_is_preempted(cpu) false 1775 #endif 1776 1777 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 1778 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 1779 1780 #ifndef TASK_SIZE_OF 1781 #define TASK_SIZE_OF(tsk) TASK_SIZE 1782 #endif 1783 1784 #ifdef CONFIG_RSEQ 1785 1786 /* 1787 * Map the event mask on the user-space ABI enum rseq_cs_flags 1788 * for direct mask checks. 1789 */ 1790 enum rseq_event_mask_bits { 1791 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT, 1792 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT, 1793 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT, 1794 }; 1795 1796 enum rseq_event_mask { 1797 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT), 1798 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT), 1799 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT), 1800 }; 1801 1802 static inline void rseq_set_notify_resume(struct task_struct *t) 1803 { 1804 if (t->rseq) 1805 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME); 1806 } 1807 1808 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs); 1809 1810 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 1811 struct pt_regs *regs) 1812 { 1813 if (current->rseq) 1814 __rseq_handle_notify_resume(ksig, regs); 1815 } 1816 1817 static inline void rseq_signal_deliver(struct ksignal *ksig, 1818 struct pt_regs *regs) 1819 { 1820 preempt_disable(); 1821 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask); 1822 preempt_enable(); 1823 rseq_handle_notify_resume(ksig, regs); 1824 } 1825 1826 /* rseq_preempt() requires preemption to be disabled. */ 1827 static inline void rseq_preempt(struct task_struct *t) 1828 { 1829 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask); 1830 rseq_set_notify_resume(t); 1831 } 1832 1833 /* rseq_migrate() requires preemption to be disabled. */ 1834 static inline void rseq_migrate(struct task_struct *t) 1835 { 1836 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask); 1837 rseq_set_notify_resume(t); 1838 } 1839 1840 /* 1841 * If parent process has a registered restartable sequences area, the 1842 * child inherits. Only applies when forking a process, not a thread. 1843 */ 1844 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 1845 { 1846 if (clone_flags & CLONE_THREAD) { 1847 t->rseq = NULL; 1848 t->rseq_len = 0; 1849 t->rseq_sig = 0; 1850 t->rseq_event_mask = 0; 1851 } else { 1852 t->rseq = current->rseq; 1853 t->rseq_len = current->rseq_len; 1854 t->rseq_sig = current->rseq_sig; 1855 t->rseq_event_mask = current->rseq_event_mask; 1856 } 1857 } 1858 1859 static inline void rseq_execve(struct task_struct *t) 1860 { 1861 t->rseq = NULL; 1862 t->rseq_len = 0; 1863 t->rseq_sig = 0; 1864 t->rseq_event_mask = 0; 1865 } 1866 1867 #else 1868 1869 static inline void rseq_set_notify_resume(struct task_struct *t) 1870 { 1871 } 1872 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 1873 struct pt_regs *regs) 1874 { 1875 } 1876 static inline void rseq_signal_deliver(struct ksignal *ksig, 1877 struct pt_regs *regs) 1878 { 1879 } 1880 static inline void rseq_preempt(struct task_struct *t) 1881 { 1882 } 1883 static inline void rseq_migrate(struct task_struct *t) 1884 { 1885 } 1886 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 1887 { 1888 } 1889 static inline void rseq_execve(struct task_struct *t) 1890 { 1891 } 1892 1893 #endif 1894 1895 #ifdef CONFIG_DEBUG_RSEQ 1896 1897 void rseq_syscall(struct pt_regs *regs); 1898 1899 #else 1900 1901 static inline void rseq_syscall(struct pt_regs *regs) 1902 { 1903 } 1904 1905 #endif 1906 1907 #endif 1908