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/refcount.h> 25 #include <linux/resource.h> 26 #include <linux/latencytop.h> 27 #include <linux/sched/prio.h> 28 #include <linux/signal_types.h> 29 #include <linux/psi_types.h> 30 #include <linux/mm_types_task.h> 31 #include <linux/task_io_accounting.h> 32 #include <linux/rseq.h> 33 34 /* task_struct member predeclarations (sorted alphabetically): */ 35 struct audit_context; 36 struct backing_dev_info; 37 struct bio_list; 38 struct blk_plug; 39 struct cfs_rq; 40 struct fs_struct; 41 struct futex_pi_state; 42 struct io_context; 43 struct mempolicy; 44 struct nameidata; 45 struct nsproxy; 46 struct perf_event_context; 47 struct pid_namespace; 48 struct pipe_inode_info; 49 struct rcu_node; 50 struct reclaim_state; 51 struct robust_list_head; 52 struct sched_attr; 53 struct sched_param; 54 struct seq_file; 55 struct sighand_struct; 56 struct signal_struct; 57 struct task_delay_info; 58 struct task_group; 59 60 /* 61 * Task state bitmask. NOTE! These bits are also 62 * encoded in fs/proc/array.c: get_task_state(). 63 * 64 * We have two separate sets of flags: task->state 65 * is about runnability, while task->exit_state are 66 * about the task exiting. Confusing, but this way 67 * modifying one set can't modify the other one by 68 * mistake. 69 */ 70 71 /* Used in tsk->state: */ 72 #define TASK_RUNNING 0x0000 73 #define TASK_INTERRUPTIBLE 0x0001 74 #define TASK_UNINTERRUPTIBLE 0x0002 75 #define __TASK_STOPPED 0x0004 76 #define __TASK_TRACED 0x0008 77 /* Used in tsk->exit_state: */ 78 #define EXIT_DEAD 0x0010 79 #define EXIT_ZOMBIE 0x0020 80 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 81 /* Used in tsk->state again: */ 82 #define TASK_PARKED 0x0040 83 #define TASK_DEAD 0x0080 84 #define TASK_WAKEKILL 0x0100 85 #define TASK_WAKING 0x0200 86 #define TASK_NOLOAD 0x0400 87 #define TASK_NEW 0x0800 88 #define TASK_STATE_MAX 0x1000 89 90 /* Convenience macros for the sake of set_current_state: */ 91 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 92 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 93 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) 94 95 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 96 97 /* Convenience macros for the sake of wake_up(): */ 98 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 99 100 /* get_task_state(): */ 101 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 102 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 103 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 104 TASK_PARKED) 105 106 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0) 107 108 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0) 109 110 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) 111 112 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \ 113 (task->flags & PF_FROZEN) == 0 && \ 114 (task->state & TASK_NOLOAD) == 0) 115 116 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 117 118 /* 119 * Special states are those that do not use the normal wait-loop pattern. See 120 * the comment with set_special_state(). 121 */ 122 #define is_special_task_state(state) \ 123 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD)) 124 125 #define __set_current_state(state_value) \ 126 do { \ 127 WARN_ON_ONCE(is_special_task_state(state_value));\ 128 current->task_state_change = _THIS_IP_; \ 129 current->state = (state_value); \ 130 } while (0) 131 132 #define set_current_state(state_value) \ 133 do { \ 134 WARN_ON_ONCE(is_special_task_state(state_value));\ 135 current->task_state_change = _THIS_IP_; \ 136 smp_store_mb(current->state, (state_value)); \ 137 } while (0) 138 139 #define set_special_state(state_value) \ 140 do { \ 141 unsigned long flags; /* may shadow */ \ 142 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 143 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 144 current->task_state_change = _THIS_IP_; \ 145 current->state = (state_value); \ 146 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 147 } while (0) 148 #else 149 /* 150 * set_current_state() includes a barrier so that the write of current->state 151 * is correctly serialised wrt the caller's subsequent test of whether to 152 * actually sleep: 153 * 154 * for (;;) { 155 * set_current_state(TASK_UNINTERRUPTIBLE); 156 * if (!need_sleep) 157 * break; 158 * 159 * schedule(); 160 * } 161 * __set_current_state(TASK_RUNNING); 162 * 163 * If the caller does not need such serialisation (because, for instance, the 164 * condition test and condition change and wakeup are under the same lock) then 165 * use __set_current_state(). 166 * 167 * The above is typically ordered against the wakeup, which does: 168 * 169 * need_sleep = false; 170 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 171 * 172 * where wake_up_state() executes a full memory barrier before accessing the 173 * task state. 174 * 175 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 176 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 177 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 178 * 179 * However, with slightly different timing the wakeup TASK_RUNNING store can 180 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 181 * a problem either because that will result in one extra go around the loop 182 * and our @cond test will save the day. 183 * 184 * Also see the comments of try_to_wake_up(). 185 */ 186 #define __set_current_state(state_value) \ 187 current->state = (state_value) 188 189 #define set_current_state(state_value) \ 190 smp_store_mb(current->state, (state_value)) 191 192 /* 193 * set_special_state() should be used for those states when the blocking task 194 * can not use the regular condition based wait-loop. In that case we must 195 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores 196 * will not collide with our state change. 197 */ 198 #define set_special_state(state_value) \ 199 do { \ 200 unsigned long flags; /* may shadow */ \ 201 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 202 current->state = (state_value); \ 203 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 204 } while (0) 205 206 #endif 207 208 /* Task command name length: */ 209 #define TASK_COMM_LEN 16 210 211 extern void scheduler_tick(void); 212 213 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 214 215 extern long schedule_timeout(long timeout); 216 extern long schedule_timeout_interruptible(long timeout); 217 extern long schedule_timeout_killable(long timeout); 218 extern long schedule_timeout_uninterruptible(long timeout); 219 extern long schedule_timeout_idle(long timeout); 220 asmlinkage void schedule(void); 221 extern void schedule_preempt_disabled(void); 222 223 extern int __must_check io_schedule_prepare(void); 224 extern void io_schedule_finish(int token); 225 extern long io_schedule_timeout(long timeout); 226 extern void io_schedule(void); 227 228 /** 229 * struct prev_cputime - snapshot of system and user cputime 230 * @utime: time spent in user mode 231 * @stime: time spent in system mode 232 * @lock: protects the above two fields 233 * 234 * Stores previous user/system time values such that we can guarantee 235 * monotonicity. 236 */ 237 struct prev_cputime { 238 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 239 u64 utime; 240 u64 stime; 241 raw_spinlock_t lock; 242 #endif 243 }; 244 245 /** 246 * struct task_cputime - collected CPU time counts 247 * @utime: time spent in user mode, in nanoseconds 248 * @stime: time spent in kernel mode, in nanoseconds 249 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds 250 * 251 * This structure groups together three kinds of CPU time that are tracked for 252 * threads and thread groups. Most things considering CPU time want to group 253 * these counts together and treat all three of them in parallel. 254 */ 255 struct task_cputime { 256 u64 utime; 257 u64 stime; 258 unsigned long long sum_exec_runtime; 259 }; 260 261 /* Alternate field names when used on cache expirations: */ 262 #define virt_exp utime 263 #define prof_exp stime 264 #define sched_exp sum_exec_runtime 265 266 enum vtime_state { 267 /* Task is sleeping or running in a CPU with VTIME inactive: */ 268 VTIME_INACTIVE = 0, 269 /* Task runs in userspace in a CPU with VTIME active: */ 270 VTIME_USER, 271 /* Task runs in kernelspace in a CPU with VTIME active: */ 272 VTIME_SYS, 273 }; 274 275 struct vtime { 276 seqcount_t seqcount; 277 unsigned long long starttime; 278 enum vtime_state state; 279 u64 utime; 280 u64 stime; 281 u64 gtime; 282 }; 283 284 struct sched_info { 285 #ifdef CONFIG_SCHED_INFO 286 /* Cumulative counters: */ 287 288 /* # of times we have run on this CPU: */ 289 unsigned long pcount; 290 291 /* Time spent waiting on a runqueue: */ 292 unsigned long long run_delay; 293 294 /* Timestamps: */ 295 296 /* When did we last run on a CPU? */ 297 unsigned long long last_arrival; 298 299 /* When were we last queued to run? */ 300 unsigned long long last_queued; 301 302 #endif /* CONFIG_SCHED_INFO */ 303 }; 304 305 /* 306 * Integer metrics need fixed point arithmetic, e.g., sched/fair 307 * has a few: load, load_avg, util_avg, freq, and capacity. 308 * 309 * We define a basic fixed point arithmetic range, and then formalize 310 * all these metrics based on that basic range. 311 */ 312 # define SCHED_FIXEDPOINT_SHIFT 10 313 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 314 315 struct load_weight { 316 unsigned long weight; 317 u32 inv_weight; 318 }; 319 320 /** 321 * struct util_est - Estimation utilization of FAIR tasks 322 * @enqueued: instantaneous estimated utilization of a task/cpu 323 * @ewma: the Exponential Weighted Moving Average (EWMA) 324 * utilization of a task 325 * 326 * Support data structure to track an Exponential Weighted Moving Average 327 * (EWMA) of a FAIR task's utilization. New samples are added to the moving 328 * average each time a task completes an activation. Sample's weight is chosen 329 * so that the EWMA will be relatively insensitive to transient changes to the 330 * task's workload. 331 * 332 * The enqueued attribute has a slightly different meaning for tasks and cpus: 333 * - task: the task's util_avg at last task dequeue time 334 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU 335 * Thus, the util_est.enqueued of a task represents the contribution on the 336 * estimated utilization of the CPU where that task is currently enqueued. 337 * 338 * Only for tasks we track a moving average of the past instantaneous 339 * estimated utilization. This allows to absorb sporadic drops in utilization 340 * of an otherwise almost periodic task. 341 */ 342 struct util_est { 343 unsigned int enqueued; 344 unsigned int ewma; 345 #define UTIL_EST_WEIGHT_SHIFT 2 346 } __attribute__((__aligned__(sizeof(u64)))); 347 348 /* 349 * The load_avg/util_avg accumulates an infinite geometric series 350 * (see __update_load_avg() in kernel/sched/fair.c). 351 * 352 * [load_avg definition] 353 * 354 * load_avg = runnable% * scale_load_down(load) 355 * 356 * where runnable% is the time ratio that a sched_entity is runnable. 357 * For cfs_rq, it is the aggregated load_avg of all runnable and 358 * blocked sched_entities. 359 * 360 * [util_avg definition] 361 * 362 * util_avg = running% * SCHED_CAPACITY_SCALE 363 * 364 * where running% is the time ratio that a sched_entity is running on 365 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable 366 * and blocked sched_entities. 367 * 368 * load_avg and util_avg don't direcly factor frequency scaling and CPU 369 * capacity scaling. The scaling is done through the rq_clock_pelt that 370 * is used for computing those signals (see update_rq_clock_pelt()) 371 * 372 * N.B., the above ratios (runnable% and running%) themselves are in the 373 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 374 * to as large a range as necessary. This is for example reflected by 375 * util_avg's SCHED_CAPACITY_SCALE. 376 * 377 * [Overflow issue] 378 * 379 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 380 * with the highest load (=88761), always runnable on a single cfs_rq, 381 * and should not overflow as the number already hits PID_MAX_LIMIT. 382 * 383 * For all other cases (including 32-bit kernels), struct load_weight's 384 * weight will overflow first before we do, because: 385 * 386 * Max(load_avg) <= Max(load.weight) 387 * 388 * Then it is the load_weight's responsibility to consider overflow 389 * issues. 390 */ 391 struct sched_avg { 392 u64 last_update_time; 393 u64 load_sum; 394 u64 runnable_load_sum; 395 u32 util_sum; 396 u32 period_contrib; 397 unsigned long load_avg; 398 unsigned long runnable_load_avg; 399 unsigned long util_avg; 400 struct util_est util_est; 401 } ____cacheline_aligned; 402 403 struct sched_statistics { 404 #ifdef CONFIG_SCHEDSTATS 405 u64 wait_start; 406 u64 wait_max; 407 u64 wait_count; 408 u64 wait_sum; 409 u64 iowait_count; 410 u64 iowait_sum; 411 412 u64 sleep_start; 413 u64 sleep_max; 414 s64 sum_sleep_runtime; 415 416 u64 block_start; 417 u64 block_max; 418 u64 exec_max; 419 u64 slice_max; 420 421 u64 nr_migrations_cold; 422 u64 nr_failed_migrations_affine; 423 u64 nr_failed_migrations_running; 424 u64 nr_failed_migrations_hot; 425 u64 nr_forced_migrations; 426 427 u64 nr_wakeups; 428 u64 nr_wakeups_sync; 429 u64 nr_wakeups_migrate; 430 u64 nr_wakeups_local; 431 u64 nr_wakeups_remote; 432 u64 nr_wakeups_affine; 433 u64 nr_wakeups_affine_attempts; 434 u64 nr_wakeups_passive; 435 u64 nr_wakeups_idle; 436 #endif 437 }; 438 439 struct sched_entity { 440 /* For load-balancing: */ 441 struct load_weight load; 442 unsigned long runnable_weight; 443 struct rb_node run_node; 444 struct list_head group_node; 445 unsigned int on_rq; 446 447 u64 exec_start; 448 u64 sum_exec_runtime; 449 u64 vruntime; 450 u64 prev_sum_exec_runtime; 451 452 u64 nr_migrations; 453 454 struct sched_statistics statistics; 455 456 #ifdef CONFIG_FAIR_GROUP_SCHED 457 int depth; 458 struct sched_entity *parent; 459 /* rq on which this entity is (to be) queued: */ 460 struct cfs_rq *cfs_rq; 461 /* rq "owned" by this entity/group: */ 462 struct cfs_rq *my_q; 463 #endif 464 465 #ifdef CONFIG_SMP 466 /* 467 * Per entity load average tracking. 468 * 469 * Put into separate cache line so it does not 470 * collide with read-mostly values above. 471 */ 472 struct sched_avg avg; 473 #endif 474 }; 475 476 struct sched_rt_entity { 477 struct list_head run_list; 478 unsigned long timeout; 479 unsigned long watchdog_stamp; 480 unsigned int time_slice; 481 unsigned short on_rq; 482 unsigned short on_list; 483 484 struct sched_rt_entity *back; 485 #ifdef CONFIG_RT_GROUP_SCHED 486 struct sched_rt_entity *parent; 487 /* rq on which this entity is (to be) queued: */ 488 struct rt_rq *rt_rq; 489 /* rq "owned" by this entity/group: */ 490 struct rt_rq *my_q; 491 #endif 492 } __randomize_layout; 493 494 struct sched_dl_entity { 495 struct rb_node rb_node; 496 497 /* 498 * Original scheduling parameters. Copied here from sched_attr 499 * during sched_setattr(), they will remain the same until 500 * the next sched_setattr(). 501 */ 502 u64 dl_runtime; /* Maximum runtime for each instance */ 503 u64 dl_deadline; /* Relative deadline of each instance */ 504 u64 dl_period; /* Separation of two instances (period) */ 505 u64 dl_bw; /* dl_runtime / dl_period */ 506 u64 dl_density; /* dl_runtime / dl_deadline */ 507 508 /* 509 * Actual scheduling parameters. Initialized with the values above, 510 * they are continuously updated during task execution. Note that 511 * the remaining runtime could be < 0 in case we are in overrun. 512 */ 513 s64 runtime; /* Remaining runtime for this instance */ 514 u64 deadline; /* Absolute deadline for this instance */ 515 unsigned int flags; /* Specifying the scheduler behaviour */ 516 517 /* 518 * Some bool flags: 519 * 520 * @dl_throttled tells if we exhausted the runtime. If so, the 521 * task has to wait for a replenishment to be performed at the 522 * next firing of dl_timer. 523 * 524 * @dl_boosted tells if we are boosted due to DI. If so we are 525 * outside bandwidth enforcement mechanism (but only until we 526 * exit the critical section); 527 * 528 * @dl_yielded tells if task gave up the CPU before consuming 529 * all its available runtime during the last job. 530 * 531 * @dl_non_contending tells if the task is inactive while still 532 * contributing to the active utilization. In other words, it 533 * indicates if the inactive timer has been armed and its handler 534 * has not been executed yet. This flag is useful to avoid race 535 * conditions between the inactive timer handler and the wakeup 536 * code. 537 * 538 * @dl_overrun tells if the task asked to be informed about runtime 539 * overruns. 540 */ 541 unsigned int dl_throttled : 1; 542 unsigned int dl_boosted : 1; 543 unsigned int dl_yielded : 1; 544 unsigned int dl_non_contending : 1; 545 unsigned int dl_overrun : 1; 546 547 /* 548 * Bandwidth enforcement timer. Each -deadline task has its 549 * own bandwidth to be enforced, thus we need one timer per task. 550 */ 551 struct hrtimer dl_timer; 552 553 /* 554 * Inactive timer, responsible for decreasing the active utilization 555 * at the "0-lag time". When a -deadline task blocks, it contributes 556 * to GRUB's active utilization until the "0-lag time", hence a 557 * timer is needed to decrease the active utilization at the correct 558 * time. 559 */ 560 struct hrtimer inactive_timer; 561 }; 562 563 union rcu_special { 564 struct { 565 u8 blocked; 566 u8 need_qs; 567 u8 exp_hint; /* Hint for performance. */ 568 u8 pad; /* No garbage from compiler! */ 569 } b; /* Bits. */ 570 u32 s; /* Set of bits. */ 571 }; 572 573 enum perf_event_task_context { 574 perf_invalid_context = -1, 575 perf_hw_context = 0, 576 perf_sw_context, 577 perf_nr_task_contexts, 578 }; 579 580 struct wake_q_node { 581 struct wake_q_node *next; 582 }; 583 584 struct task_struct { 585 #ifdef CONFIG_THREAD_INFO_IN_TASK 586 /* 587 * For reasons of header soup (see current_thread_info()), this 588 * must be the first element of task_struct. 589 */ 590 struct thread_info thread_info; 591 #endif 592 /* -1 unrunnable, 0 runnable, >0 stopped: */ 593 volatile long state; 594 595 /* 596 * This begins the randomizable portion of task_struct. Only 597 * scheduling-critical items should be added above here. 598 */ 599 randomized_struct_fields_start 600 601 void *stack; 602 refcount_t usage; 603 /* Per task flags (PF_*), defined further below: */ 604 unsigned int flags; 605 unsigned int ptrace; 606 607 #ifdef CONFIG_SMP 608 struct llist_node wake_entry; 609 int on_cpu; 610 #ifdef CONFIG_THREAD_INFO_IN_TASK 611 /* Current CPU: */ 612 unsigned int cpu; 613 #endif 614 unsigned int wakee_flips; 615 unsigned long wakee_flip_decay_ts; 616 struct task_struct *last_wakee; 617 618 /* 619 * recent_used_cpu is initially set as the last CPU used by a task 620 * that wakes affine another task. Waker/wakee relationships can 621 * push tasks around a CPU where each wakeup moves to the next one. 622 * Tracking a recently used CPU allows a quick search for a recently 623 * used CPU that may be idle. 624 */ 625 int recent_used_cpu; 626 int wake_cpu; 627 #endif 628 int on_rq; 629 630 int prio; 631 int static_prio; 632 int normal_prio; 633 unsigned int rt_priority; 634 635 const struct sched_class *sched_class; 636 struct sched_entity se; 637 struct sched_rt_entity rt; 638 #ifdef CONFIG_CGROUP_SCHED 639 struct task_group *sched_task_group; 640 #endif 641 struct sched_dl_entity dl; 642 643 #ifdef CONFIG_PREEMPT_NOTIFIERS 644 /* List of struct preempt_notifier: */ 645 struct hlist_head preempt_notifiers; 646 #endif 647 648 #ifdef CONFIG_BLK_DEV_IO_TRACE 649 unsigned int btrace_seq; 650 #endif 651 652 unsigned int policy; 653 int nr_cpus_allowed; 654 cpumask_t cpus_allowed; 655 656 #ifdef CONFIG_PREEMPT_RCU 657 int rcu_read_lock_nesting; 658 union rcu_special rcu_read_unlock_special; 659 struct list_head rcu_node_entry; 660 struct rcu_node *rcu_blocked_node; 661 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 662 663 #ifdef CONFIG_TASKS_RCU 664 unsigned long rcu_tasks_nvcsw; 665 u8 rcu_tasks_holdout; 666 u8 rcu_tasks_idx; 667 int rcu_tasks_idle_cpu; 668 struct list_head rcu_tasks_holdout_list; 669 #endif /* #ifdef CONFIG_TASKS_RCU */ 670 671 struct sched_info sched_info; 672 673 struct list_head tasks; 674 #ifdef CONFIG_SMP 675 struct plist_node pushable_tasks; 676 struct rb_node pushable_dl_tasks; 677 #endif 678 679 struct mm_struct *mm; 680 struct mm_struct *active_mm; 681 682 /* Per-thread vma caching: */ 683 struct vmacache vmacache; 684 685 #ifdef SPLIT_RSS_COUNTING 686 struct task_rss_stat rss_stat; 687 #endif 688 int exit_state; 689 int exit_code; 690 int exit_signal; 691 /* The signal sent when the parent dies: */ 692 int pdeath_signal; 693 /* JOBCTL_*, siglock protected: */ 694 unsigned long jobctl; 695 696 /* Used for emulating ABI behavior of previous Linux versions: */ 697 unsigned int personality; 698 699 /* Scheduler bits, serialized by scheduler locks: */ 700 unsigned sched_reset_on_fork:1; 701 unsigned sched_contributes_to_load:1; 702 unsigned sched_migrated:1; 703 unsigned sched_remote_wakeup:1; 704 #ifdef CONFIG_PSI 705 unsigned sched_psi_wake_requeue:1; 706 #endif 707 708 /* Force alignment to the next boundary: */ 709 unsigned :0; 710 711 /* Unserialized, strictly 'current' */ 712 713 /* Bit to tell LSMs we're in execve(): */ 714 unsigned in_execve:1; 715 unsigned in_iowait:1; 716 #ifndef TIF_RESTORE_SIGMASK 717 unsigned restore_sigmask:1; 718 #endif 719 #ifdef CONFIG_MEMCG 720 unsigned in_user_fault:1; 721 #endif 722 #ifdef CONFIG_COMPAT_BRK 723 unsigned brk_randomized:1; 724 #endif 725 #ifdef CONFIG_CGROUPS 726 /* disallow userland-initiated cgroup migration */ 727 unsigned no_cgroup_migration:1; 728 #endif 729 #ifdef CONFIG_BLK_CGROUP 730 /* to be used once the psi infrastructure lands upstream. */ 731 unsigned use_memdelay:1; 732 #endif 733 734 unsigned long atomic_flags; /* Flags requiring atomic access. */ 735 736 struct restart_block restart_block; 737 738 pid_t pid; 739 pid_t tgid; 740 741 #ifdef CONFIG_STACKPROTECTOR 742 /* Canary value for the -fstack-protector GCC feature: */ 743 unsigned long stack_canary; 744 #endif 745 /* 746 * Pointers to the (original) parent process, youngest child, younger sibling, 747 * older sibling, respectively. (p->father can be replaced with 748 * p->real_parent->pid) 749 */ 750 751 /* Real parent process: */ 752 struct task_struct __rcu *real_parent; 753 754 /* Recipient of SIGCHLD, wait4() reports: */ 755 struct task_struct __rcu *parent; 756 757 /* 758 * Children/sibling form the list of natural children: 759 */ 760 struct list_head children; 761 struct list_head sibling; 762 struct task_struct *group_leader; 763 764 /* 765 * 'ptraced' is the list of tasks this task is using ptrace() on. 766 * 767 * This includes both natural children and PTRACE_ATTACH targets. 768 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 769 */ 770 struct list_head ptraced; 771 struct list_head ptrace_entry; 772 773 /* PID/PID hash table linkage. */ 774 struct pid *thread_pid; 775 struct hlist_node pid_links[PIDTYPE_MAX]; 776 struct list_head thread_group; 777 struct list_head thread_node; 778 779 struct completion *vfork_done; 780 781 /* CLONE_CHILD_SETTID: */ 782 int __user *set_child_tid; 783 784 /* CLONE_CHILD_CLEARTID: */ 785 int __user *clear_child_tid; 786 787 u64 utime; 788 u64 stime; 789 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 790 u64 utimescaled; 791 u64 stimescaled; 792 #endif 793 u64 gtime; 794 struct prev_cputime prev_cputime; 795 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 796 struct vtime vtime; 797 #endif 798 799 #ifdef CONFIG_NO_HZ_FULL 800 atomic_t tick_dep_mask; 801 #endif 802 /* Context switch counts: */ 803 unsigned long nvcsw; 804 unsigned long nivcsw; 805 806 /* Monotonic time in nsecs: */ 807 u64 start_time; 808 809 /* Boot based time in nsecs: */ 810 u64 real_start_time; 811 812 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 813 unsigned long min_flt; 814 unsigned long maj_flt; 815 816 #ifdef CONFIG_POSIX_TIMERS 817 struct task_cputime cputime_expires; 818 struct list_head cpu_timers[3]; 819 #endif 820 821 /* Process credentials: */ 822 823 /* Tracer's credentials at attach: */ 824 const struct cred __rcu *ptracer_cred; 825 826 /* Objective and real subjective task credentials (COW): */ 827 const struct cred __rcu *real_cred; 828 829 /* Effective (overridable) subjective task credentials (COW): */ 830 const struct cred __rcu *cred; 831 832 /* 833 * executable name, excluding path. 834 * 835 * - normally initialized setup_new_exec() 836 * - access it with [gs]et_task_comm() 837 * - lock it with task_lock() 838 */ 839 char comm[TASK_COMM_LEN]; 840 841 struct nameidata *nameidata; 842 843 #ifdef CONFIG_SYSVIPC 844 struct sysv_sem sysvsem; 845 struct sysv_shm sysvshm; 846 #endif 847 #ifdef CONFIG_DETECT_HUNG_TASK 848 unsigned long last_switch_count; 849 unsigned long last_switch_time; 850 #endif 851 /* Filesystem information: */ 852 struct fs_struct *fs; 853 854 /* Open file information: */ 855 struct files_struct *files; 856 857 /* Namespaces: */ 858 struct nsproxy *nsproxy; 859 860 /* Signal handlers: */ 861 struct signal_struct *signal; 862 struct sighand_struct *sighand; 863 sigset_t blocked; 864 sigset_t real_blocked; 865 /* Restored if set_restore_sigmask() was used: */ 866 sigset_t saved_sigmask; 867 struct sigpending pending; 868 unsigned long sas_ss_sp; 869 size_t sas_ss_size; 870 unsigned int sas_ss_flags; 871 872 struct callback_head *task_works; 873 874 struct audit_context *audit_context; 875 #ifdef CONFIG_AUDITSYSCALL 876 kuid_t loginuid; 877 unsigned int sessionid; 878 #endif 879 struct seccomp seccomp; 880 881 /* Thread group tracking: */ 882 u32 parent_exec_id; 883 u32 self_exec_id; 884 885 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 886 spinlock_t alloc_lock; 887 888 /* Protection of the PI data structures: */ 889 raw_spinlock_t pi_lock; 890 891 struct wake_q_node wake_q; 892 893 #ifdef CONFIG_RT_MUTEXES 894 /* PI waiters blocked on a rt_mutex held by this task: */ 895 struct rb_root_cached pi_waiters; 896 /* Updated under owner's pi_lock and rq lock */ 897 struct task_struct *pi_top_task; 898 /* Deadlock detection and priority inheritance handling: */ 899 struct rt_mutex_waiter *pi_blocked_on; 900 #endif 901 902 #ifdef CONFIG_DEBUG_MUTEXES 903 /* Mutex deadlock detection: */ 904 struct mutex_waiter *blocked_on; 905 #endif 906 907 #ifdef CONFIG_TRACE_IRQFLAGS 908 unsigned int irq_events; 909 unsigned long hardirq_enable_ip; 910 unsigned long hardirq_disable_ip; 911 unsigned int hardirq_enable_event; 912 unsigned int hardirq_disable_event; 913 int hardirqs_enabled; 914 int hardirq_context; 915 unsigned long softirq_disable_ip; 916 unsigned long softirq_enable_ip; 917 unsigned int softirq_disable_event; 918 unsigned int softirq_enable_event; 919 int softirqs_enabled; 920 int softirq_context; 921 #endif 922 923 #ifdef CONFIG_LOCKDEP 924 # define MAX_LOCK_DEPTH 48UL 925 u64 curr_chain_key; 926 int lockdep_depth; 927 unsigned int lockdep_recursion; 928 struct held_lock held_locks[MAX_LOCK_DEPTH]; 929 #endif 930 931 #ifdef CONFIG_UBSAN 932 unsigned int in_ubsan; 933 #endif 934 935 /* Journalling filesystem info: */ 936 void *journal_info; 937 938 /* Stacked block device info: */ 939 struct bio_list *bio_list; 940 941 #ifdef CONFIG_BLOCK 942 /* Stack plugging: */ 943 struct blk_plug *plug; 944 #endif 945 946 /* VM state: */ 947 struct reclaim_state *reclaim_state; 948 949 struct backing_dev_info *backing_dev_info; 950 951 struct io_context *io_context; 952 953 /* Ptrace state: */ 954 unsigned long ptrace_message; 955 kernel_siginfo_t *last_siginfo; 956 957 struct task_io_accounting ioac; 958 #ifdef CONFIG_PSI 959 /* Pressure stall state */ 960 unsigned int psi_flags; 961 #endif 962 #ifdef CONFIG_TASK_XACCT 963 /* Accumulated RSS usage: */ 964 u64 acct_rss_mem1; 965 /* Accumulated virtual memory usage: */ 966 u64 acct_vm_mem1; 967 /* stime + utime since last update: */ 968 u64 acct_timexpd; 969 #endif 970 #ifdef CONFIG_CPUSETS 971 /* Protected by ->alloc_lock: */ 972 nodemask_t mems_allowed; 973 /* Seqence number to catch updates: */ 974 seqcount_t mems_allowed_seq; 975 int cpuset_mem_spread_rotor; 976 int cpuset_slab_spread_rotor; 977 #endif 978 #ifdef CONFIG_CGROUPS 979 /* Control Group info protected by css_set_lock: */ 980 struct css_set __rcu *cgroups; 981 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 982 struct list_head cg_list; 983 #endif 984 #ifdef CONFIG_X86_CPU_RESCTRL 985 u32 closid; 986 u32 rmid; 987 #endif 988 #ifdef CONFIG_FUTEX 989 struct robust_list_head __user *robust_list; 990 #ifdef CONFIG_COMPAT 991 struct compat_robust_list_head __user *compat_robust_list; 992 #endif 993 struct list_head pi_state_list; 994 struct futex_pi_state *pi_state_cache; 995 #endif 996 #ifdef CONFIG_PERF_EVENTS 997 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 998 struct mutex perf_event_mutex; 999 struct list_head perf_event_list; 1000 #endif 1001 #ifdef CONFIG_DEBUG_PREEMPT 1002 unsigned long preempt_disable_ip; 1003 #endif 1004 #ifdef CONFIG_NUMA 1005 /* Protected by alloc_lock: */ 1006 struct mempolicy *mempolicy; 1007 short il_prev; 1008 short pref_node_fork; 1009 #endif 1010 #ifdef CONFIG_NUMA_BALANCING 1011 int numa_scan_seq; 1012 unsigned int numa_scan_period; 1013 unsigned int numa_scan_period_max; 1014 int numa_preferred_nid; 1015 unsigned long numa_migrate_retry; 1016 /* Migration stamp: */ 1017 u64 node_stamp; 1018 u64 last_task_numa_placement; 1019 u64 last_sum_exec_runtime; 1020 struct callback_head numa_work; 1021 1022 struct numa_group *numa_group; 1023 1024 /* 1025 * numa_faults is an array split into four regions: 1026 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1027 * in this precise order. 1028 * 1029 * faults_memory: Exponential decaying average of faults on a per-node 1030 * basis. Scheduling placement decisions are made based on these 1031 * counts. The values remain static for the duration of a PTE scan. 1032 * faults_cpu: Track the nodes the process was running on when a NUMA 1033 * hinting fault was incurred. 1034 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1035 * during the current scan window. When the scan completes, the counts 1036 * in faults_memory and faults_cpu decay and these values are copied. 1037 */ 1038 unsigned long *numa_faults; 1039 unsigned long total_numa_faults; 1040 1041 /* 1042 * numa_faults_locality tracks if faults recorded during the last 1043 * scan window were remote/local or failed to migrate. The task scan 1044 * period is adapted based on the locality of the faults with different 1045 * weights depending on whether they were shared or private faults 1046 */ 1047 unsigned long numa_faults_locality[3]; 1048 1049 unsigned long numa_pages_migrated; 1050 #endif /* CONFIG_NUMA_BALANCING */ 1051 1052 #ifdef CONFIG_RSEQ 1053 struct rseq __user *rseq; 1054 u32 rseq_len; 1055 u32 rseq_sig; 1056 /* 1057 * RmW on rseq_event_mask must be performed atomically 1058 * with respect to preemption. 1059 */ 1060 unsigned long rseq_event_mask; 1061 #endif 1062 1063 struct tlbflush_unmap_batch tlb_ubc; 1064 1065 struct rcu_head rcu; 1066 1067 /* Cache last used pipe for splice(): */ 1068 struct pipe_inode_info *splice_pipe; 1069 1070 struct page_frag task_frag; 1071 1072 #ifdef CONFIG_TASK_DELAY_ACCT 1073 struct task_delay_info *delays; 1074 #endif 1075 1076 #ifdef CONFIG_FAULT_INJECTION 1077 int make_it_fail; 1078 unsigned int fail_nth; 1079 #endif 1080 /* 1081 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1082 * balance_dirty_pages() for a dirty throttling pause: 1083 */ 1084 int nr_dirtied; 1085 int nr_dirtied_pause; 1086 /* Start of a write-and-pause period: */ 1087 unsigned long dirty_paused_when; 1088 1089 #ifdef CONFIG_LATENCYTOP 1090 int latency_record_count; 1091 struct latency_record latency_record[LT_SAVECOUNT]; 1092 #endif 1093 /* 1094 * Time slack values; these are used to round up poll() and 1095 * select() etc timeout values. These are in nanoseconds. 1096 */ 1097 u64 timer_slack_ns; 1098 u64 default_timer_slack_ns; 1099 1100 #ifdef CONFIG_KASAN 1101 unsigned int kasan_depth; 1102 #endif 1103 1104 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1105 /* Index of current stored address in ret_stack: */ 1106 int curr_ret_stack; 1107 int curr_ret_depth; 1108 1109 /* Stack of return addresses for return function tracing: */ 1110 struct ftrace_ret_stack *ret_stack; 1111 1112 /* Timestamp for last schedule: */ 1113 unsigned long long ftrace_timestamp; 1114 1115 /* 1116 * Number of functions that haven't been traced 1117 * because of depth overrun: 1118 */ 1119 atomic_t trace_overrun; 1120 1121 /* Pause tracing: */ 1122 atomic_t tracing_graph_pause; 1123 #endif 1124 1125 #ifdef CONFIG_TRACING 1126 /* State flags for use by tracers: */ 1127 unsigned long trace; 1128 1129 /* Bitmask and counter of trace recursion: */ 1130 unsigned long trace_recursion; 1131 #endif /* CONFIG_TRACING */ 1132 1133 #ifdef CONFIG_KCOV 1134 /* Coverage collection mode enabled for this task (0 if disabled): */ 1135 unsigned int kcov_mode; 1136 1137 /* Size of the kcov_area: */ 1138 unsigned int kcov_size; 1139 1140 /* Buffer for coverage collection: */ 1141 void *kcov_area; 1142 1143 /* KCOV descriptor wired with this task or NULL: */ 1144 struct kcov *kcov; 1145 #endif 1146 1147 #ifdef CONFIG_MEMCG 1148 struct mem_cgroup *memcg_in_oom; 1149 gfp_t memcg_oom_gfp_mask; 1150 int memcg_oom_order; 1151 1152 /* Number of pages to reclaim on returning to userland: */ 1153 unsigned int memcg_nr_pages_over_high; 1154 1155 /* Used by memcontrol for targeted memcg charge: */ 1156 struct mem_cgroup *active_memcg; 1157 #endif 1158 1159 #ifdef CONFIG_BLK_CGROUP 1160 struct request_queue *throttle_queue; 1161 #endif 1162 1163 #ifdef CONFIG_UPROBES 1164 struct uprobe_task *utask; 1165 #endif 1166 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1167 unsigned int sequential_io; 1168 unsigned int sequential_io_avg; 1169 #endif 1170 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1171 unsigned long task_state_change; 1172 #endif 1173 int pagefault_disabled; 1174 #ifdef CONFIG_MMU 1175 struct task_struct *oom_reaper_list; 1176 #endif 1177 #ifdef CONFIG_VMAP_STACK 1178 struct vm_struct *stack_vm_area; 1179 #endif 1180 #ifdef CONFIG_THREAD_INFO_IN_TASK 1181 /* A live task holds one reference: */ 1182 refcount_t stack_refcount; 1183 #endif 1184 #ifdef CONFIG_LIVEPATCH 1185 int patch_state; 1186 #endif 1187 #ifdef CONFIG_SECURITY 1188 /* Used by LSM modules for access restriction: */ 1189 void *security; 1190 #endif 1191 1192 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1193 unsigned long lowest_stack; 1194 unsigned long prev_lowest_stack; 1195 #endif 1196 1197 /* 1198 * New fields for task_struct should be added above here, so that 1199 * they are included in the randomized portion of task_struct. 1200 */ 1201 randomized_struct_fields_end 1202 1203 /* CPU-specific state of this task: */ 1204 struct thread_struct thread; 1205 1206 /* 1207 * WARNING: on x86, 'thread_struct' contains a variable-sized 1208 * structure. It *MUST* be at the end of 'task_struct'. 1209 * 1210 * Do not put anything below here! 1211 */ 1212 }; 1213 1214 static inline struct pid *task_pid(struct task_struct *task) 1215 { 1216 return task->thread_pid; 1217 } 1218 1219 /* 1220 * the helpers to get the task's different pids as they are seen 1221 * from various namespaces 1222 * 1223 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1224 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1225 * current. 1226 * task_xid_nr_ns() : id seen from the ns specified; 1227 * 1228 * see also pid_nr() etc in include/linux/pid.h 1229 */ 1230 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1231 1232 static inline pid_t task_pid_nr(struct task_struct *tsk) 1233 { 1234 return tsk->pid; 1235 } 1236 1237 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1238 { 1239 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1240 } 1241 1242 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1243 { 1244 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1245 } 1246 1247 1248 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1249 { 1250 return tsk->tgid; 1251 } 1252 1253 /** 1254 * pid_alive - check that a task structure is not stale 1255 * @p: Task structure to be checked. 1256 * 1257 * Test if a process is not yet dead (at most zombie state) 1258 * If pid_alive fails, then pointers within the task structure 1259 * can be stale and must not be dereferenced. 1260 * 1261 * Return: 1 if the process is alive. 0 otherwise. 1262 */ 1263 static inline int pid_alive(const struct task_struct *p) 1264 { 1265 return p->thread_pid != NULL; 1266 } 1267 1268 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1269 { 1270 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1271 } 1272 1273 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1274 { 1275 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1276 } 1277 1278 1279 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1280 { 1281 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1282 } 1283 1284 static inline pid_t task_session_vnr(struct task_struct *tsk) 1285 { 1286 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1287 } 1288 1289 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1290 { 1291 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns); 1292 } 1293 1294 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1295 { 1296 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL); 1297 } 1298 1299 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1300 { 1301 pid_t pid = 0; 1302 1303 rcu_read_lock(); 1304 if (pid_alive(tsk)) 1305 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1306 rcu_read_unlock(); 1307 1308 return pid; 1309 } 1310 1311 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1312 { 1313 return task_ppid_nr_ns(tsk, &init_pid_ns); 1314 } 1315 1316 /* Obsolete, do not use: */ 1317 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1318 { 1319 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1320 } 1321 1322 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1323 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1324 1325 static inline unsigned int task_state_index(struct task_struct *tsk) 1326 { 1327 unsigned int tsk_state = READ_ONCE(tsk->state); 1328 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT; 1329 1330 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1331 1332 if (tsk_state == TASK_IDLE) 1333 state = TASK_REPORT_IDLE; 1334 1335 return fls(state); 1336 } 1337 1338 static inline char task_index_to_char(unsigned int state) 1339 { 1340 static const char state_char[] = "RSDTtXZPI"; 1341 1342 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1343 1344 return state_char[state]; 1345 } 1346 1347 static inline char task_state_to_char(struct task_struct *tsk) 1348 { 1349 return task_index_to_char(task_state_index(tsk)); 1350 } 1351 1352 /** 1353 * is_global_init - check if a task structure is init. Since init 1354 * is free to have sub-threads we need to check tgid. 1355 * @tsk: Task structure to be checked. 1356 * 1357 * Check if a task structure is the first user space task the kernel created. 1358 * 1359 * Return: 1 if the task structure is init. 0 otherwise. 1360 */ 1361 static inline int is_global_init(struct task_struct *tsk) 1362 { 1363 return task_tgid_nr(tsk) == 1; 1364 } 1365 1366 extern struct pid *cad_pid; 1367 1368 /* 1369 * Per process flags 1370 */ 1371 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1372 #define PF_EXITING 0x00000004 /* Getting shut down */ 1373 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */ 1374 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */ 1375 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1376 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1377 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1378 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1379 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1380 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1381 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1382 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1383 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1384 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1385 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1386 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1387 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1388 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1389 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1390 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ 1391 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1392 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1393 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1394 #define PF_MEMSTALL 0x01000000 /* Stalled due to lack of memory */ 1395 #define PF_UMH 0x02000000 /* I'm an Usermodehelper process */ 1396 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */ 1397 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1398 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1399 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1400 1401 /* 1402 * Only the _current_ task can read/write to tsk->flags, but other 1403 * tasks can access tsk->flags in readonly mode for example 1404 * with tsk_used_math (like during threaded core dumping). 1405 * There is however an exception to this rule during ptrace 1406 * or during fork: the ptracer task is allowed to write to the 1407 * child->flags of its traced child (same goes for fork, the parent 1408 * can write to the child->flags), because we're guaranteed the 1409 * child is not running and in turn not changing child->flags 1410 * at the same time the parent does it. 1411 */ 1412 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1413 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1414 #define clear_used_math() clear_stopped_child_used_math(current) 1415 #define set_used_math() set_stopped_child_used_math(current) 1416 1417 #define conditional_stopped_child_used_math(condition, child) \ 1418 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1419 1420 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1421 1422 #define copy_to_stopped_child_used_math(child) \ 1423 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1424 1425 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1426 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1427 #define used_math() tsk_used_math(current) 1428 1429 static inline bool is_percpu_thread(void) 1430 { 1431 #ifdef CONFIG_SMP 1432 return (current->flags & PF_NO_SETAFFINITY) && 1433 (current->nr_cpus_allowed == 1); 1434 #else 1435 return true; 1436 #endif 1437 } 1438 1439 /* Per-process atomic flags. */ 1440 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1441 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1442 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1443 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1444 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1445 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1446 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1447 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1448 1449 #define TASK_PFA_TEST(name, func) \ 1450 static inline bool task_##func(struct task_struct *p) \ 1451 { return test_bit(PFA_##name, &p->atomic_flags); } 1452 1453 #define TASK_PFA_SET(name, func) \ 1454 static inline void task_set_##func(struct task_struct *p) \ 1455 { set_bit(PFA_##name, &p->atomic_flags); } 1456 1457 #define TASK_PFA_CLEAR(name, func) \ 1458 static inline void task_clear_##func(struct task_struct *p) \ 1459 { clear_bit(PFA_##name, &p->atomic_flags); } 1460 1461 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1462 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1463 1464 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1465 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1466 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1467 1468 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1469 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1470 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1471 1472 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1473 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1474 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1475 1476 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1477 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1478 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1479 1480 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1481 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1482 1483 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1484 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1485 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1486 1487 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1488 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1489 1490 static inline void 1491 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1492 { 1493 current->flags &= ~flags; 1494 current->flags |= orig_flags & flags; 1495 } 1496 1497 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1498 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1499 #ifdef CONFIG_SMP 1500 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1501 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1502 #else 1503 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1504 { 1505 } 1506 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1507 { 1508 if (!cpumask_test_cpu(0, new_mask)) 1509 return -EINVAL; 1510 return 0; 1511 } 1512 #endif 1513 1514 #ifndef cpu_relax_yield 1515 #define cpu_relax_yield() cpu_relax() 1516 #endif 1517 1518 extern int yield_to(struct task_struct *p, bool preempt); 1519 extern void set_user_nice(struct task_struct *p, long nice); 1520 extern int task_prio(const struct task_struct *p); 1521 1522 /** 1523 * task_nice - return the nice value of a given task. 1524 * @p: the task in question. 1525 * 1526 * Return: The nice value [ -20 ... 0 ... 19 ]. 1527 */ 1528 static inline int task_nice(const struct task_struct *p) 1529 { 1530 return PRIO_TO_NICE((p)->static_prio); 1531 } 1532 1533 extern int can_nice(const struct task_struct *p, const int nice); 1534 extern int task_curr(const struct task_struct *p); 1535 extern int idle_cpu(int cpu); 1536 extern int available_idle_cpu(int cpu); 1537 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1538 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1539 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1540 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1541 extern struct task_struct *idle_task(int cpu); 1542 1543 /** 1544 * is_idle_task - is the specified task an idle task? 1545 * @p: the task in question. 1546 * 1547 * Return: 1 if @p is an idle task. 0 otherwise. 1548 */ 1549 static inline bool is_idle_task(const struct task_struct *p) 1550 { 1551 return !!(p->flags & PF_IDLE); 1552 } 1553 1554 extern struct task_struct *curr_task(int cpu); 1555 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1556 1557 void yield(void); 1558 1559 union thread_union { 1560 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1561 struct task_struct task; 1562 #endif 1563 #ifndef CONFIG_THREAD_INFO_IN_TASK 1564 struct thread_info thread_info; 1565 #endif 1566 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1567 }; 1568 1569 #ifndef CONFIG_THREAD_INFO_IN_TASK 1570 extern struct thread_info init_thread_info; 1571 #endif 1572 1573 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1574 1575 #ifdef CONFIG_THREAD_INFO_IN_TASK 1576 static inline struct thread_info *task_thread_info(struct task_struct *task) 1577 { 1578 return &task->thread_info; 1579 } 1580 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1581 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1582 #endif 1583 1584 /* 1585 * find a task by one of its numerical ids 1586 * 1587 * find_task_by_pid_ns(): 1588 * finds a task by its pid in the specified namespace 1589 * find_task_by_vpid(): 1590 * finds a task by its virtual pid 1591 * 1592 * see also find_vpid() etc in include/linux/pid.h 1593 */ 1594 1595 extern struct task_struct *find_task_by_vpid(pid_t nr); 1596 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1597 1598 /* 1599 * find a task by its virtual pid and get the task struct 1600 */ 1601 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1602 1603 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1604 extern int wake_up_process(struct task_struct *tsk); 1605 extern void wake_up_new_task(struct task_struct *tsk); 1606 1607 #ifdef CONFIG_SMP 1608 extern void kick_process(struct task_struct *tsk); 1609 #else 1610 static inline void kick_process(struct task_struct *tsk) { } 1611 #endif 1612 1613 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1614 1615 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1616 { 1617 __set_task_comm(tsk, from, false); 1618 } 1619 1620 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1621 #define get_task_comm(buf, tsk) ({ \ 1622 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1623 __get_task_comm(buf, sizeof(buf), tsk); \ 1624 }) 1625 1626 #ifdef CONFIG_SMP 1627 void scheduler_ipi(void); 1628 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 1629 #else 1630 static inline void scheduler_ipi(void) { } 1631 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) 1632 { 1633 return 1; 1634 } 1635 #endif 1636 1637 /* 1638 * Set thread flags in other task's structures. 1639 * See asm/thread_info.h for TIF_xxxx flags available: 1640 */ 1641 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1642 { 1643 set_ti_thread_flag(task_thread_info(tsk), flag); 1644 } 1645 1646 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1647 { 1648 clear_ti_thread_flag(task_thread_info(tsk), flag); 1649 } 1650 1651 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1652 bool value) 1653 { 1654 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1655 } 1656 1657 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1658 { 1659 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1660 } 1661 1662 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1663 { 1664 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1665 } 1666 1667 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1668 { 1669 return test_ti_thread_flag(task_thread_info(tsk), flag); 1670 } 1671 1672 static inline void set_tsk_need_resched(struct task_struct *tsk) 1673 { 1674 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1675 } 1676 1677 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1678 { 1679 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1680 } 1681 1682 static inline int test_tsk_need_resched(struct task_struct *tsk) 1683 { 1684 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1685 } 1686 1687 /* 1688 * cond_resched() and cond_resched_lock(): latency reduction via 1689 * explicit rescheduling in places that are safe. The return 1690 * value indicates whether a reschedule was done in fact. 1691 * cond_resched_lock() will drop the spinlock before scheduling, 1692 */ 1693 #ifndef CONFIG_PREEMPT 1694 extern int _cond_resched(void); 1695 #else 1696 static inline int _cond_resched(void) { return 0; } 1697 #endif 1698 1699 #define cond_resched() ({ \ 1700 ___might_sleep(__FILE__, __LINE__, 0); \ 1701 _cond_resched(); \ 1702 }) 1703 1704 extern int __cond_resched_lock(spinlock_t *lock); 1705 1706 #define cond_resched_lock(lock) ({ \ 1707 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ 1708 __cond_resched_lock(lock); \ 1709 }) 1710 1711 static inline void cond_resched_rcu(void) 1712 { 1713 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 1714 rcu_read_unlock(); 1715 cond_resched(); 1716 rcu_read_lock(); 1717 #endif 1718 } 1719 1720 /* 1721 * Does a critical section need to be broken due to another 1722 * task waiting?: (technically does not depend on CONFIG_PREEMPT, 1723 * but a general need for low latency) 1724 */ 1725 static inline int spin_needbreak(spinlock_t *lock) 1726 { 1727 #ifdef CONFIG_PREEMPT 1728 return spin_is_contended(lock); 1729 #else 1730 return 0; 1731 #endif 1732 } 1733 1734 static __always_inline bool need_resched(void) 1735 { 1736 return unlikely(tif_need_resched()); 1737 } 1738 1739 /* 1740 * Wrappers for p->thread_info->cpu access. No-op on UP. 1741 */ 1742 #ifdef CONFIG_SMP 1743 1744 static inline unsigned int task_cpu(const struct task_struct *p) 1745 { 1746 #ifdef CONFIG_THREAD_INFO_IN_TASK 1747 return READ_ONCE(p->cpu); 1748 #else 1749 return READ_ONCE(task_thread_info(p)->cpu); 1750 #endif 1751 } 1752 1753 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 1754 1755 #else 1756 1757 static inline unsigned int task_cpu(const struct task_struct *p) 1758 { 1759 return 0; 1760 } 1761 1762 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 1763 { 1764 } 1765 1766 #endif /* CONFIG_SMP */ 1767 1768 /* 1769 * In order to reduce various lock holder preemption latencies provide an 1770 * interface to see if a vCPU is currently running or not. 1771 * 1772 * This allows us to terminate optimistic spin loops and block, analogous to 1773 * the native optimistic spin heuristic of testing if the lock owner task is 1774 * running or not. 1775 */ 1776 #ifndef vcpu_is_preempted 1777 # define vcpu_is_preempted(cpu) false 1778 #endif 1779 1780 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 1781 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 1782 1783 #ifndef TASK_SIZE_OF 1784 #define TASK_SIZE_OF(tsk) TASK_SIZE 1785 #endif 1786 1787 #ifdef CONFIG_RSEQ 1788 1789 /* 1790 * Map the event mask on the user-space ABI enum rseq_cs_flags 1791 * for direct mask checks. 1792 */ 1793 enum rseq_event_mask_bits { 1794 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT, 1795 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT, 1796 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT, 1797 }; 1798 1799 enum rseq_event_mask { 1800 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT), 1801 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT), 1802 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT), 1803 }; 1804 1805 static inline void rseq_set_notify_resume(struct task_struct *t) 1806 { 1807 if (t->rseq) 1808 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME); 1809 } 1810 1811 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs); 1812 1813 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 1814 struct pt_regs *regs) 1815 { 1816 if (current->rseq) 1817 __rseq_handle_notify_resume(ksig, regs); 1818 } 1819 1820 static inline void rseq_signal_deliver(struct ksignal *ksig, 1821 struct pt_regs *regs) 1822 { 1823 preempt_disable(); 1824 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask); 1825 preempt_enable(); 1826 rseq_handle_notify_resume(ksig, regs); 1827 } 1828 1829 /* rseq_preempt() requires preemption to be disabled. */ 1830 static inline void rseq_preempt(struct task_struct *t) 1831 { 1832 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask); 1833 rseq_set_notify_resume(t); 1834 } 1835 1836 /* rseq_migrate() requires preemption to be disabled. */ 1837 static inline void rseq_migrate(struct task_struct *t) 1838 { 1839 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask); 1840 rseq_set_notify_resume(t); 1841 } 1842 1843 /* 1844 * If parent process has a registered restartable sequences area, the 1845 * child inherits. Only applies when forking a process, not a thread. 1846 */ 1847 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 1848 { 1849 if (clone_flags & CLONE_THREAD) { 1850 t->rseq = NULL; 1851 t->rseq_len = 0; 1852 t->rseq_sig = 0; 1853 t->rseq_event_mask = 0; 1854 } else { 1855 t->rseq = current->rseq; 1856 t->rseq_len = current->rseq_len; 1857 t->rseq_sig = current->rseq_sig; 1858 t->rseq_event_mask = current->rseq_event_mask; 1859 } 1860 } 1861 1862 static inline void rseq_execve(struct task_struct *t) 1863 { 1864 t->rseq = NULL; 1865 t->rseq_len = 0; 1866 t->rseq_sig = 0; 1867 t->rseq_event_mask = 0; 1868 } 1869 1870 #else 1871 1872 static inline void rseq_set_notify_resume(struct task_struct *t) 1873 { 1874 } 1875 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 1876 struct pt_regs *regs) 1877 { 1878 } 1879 static inline void rseq_signal_deliver(struct ksignal *ksig, 1880 struct pt_regs *regs) 1881 { 1882 } 1883 static inline void rseq_preempt(struct task_struct *t) 1884 { 1885 } 1886 static inline void rseq_migrate(struct task_struct *t) 1887 { 1888 } 1889 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 1890 { 1891 } 1892 static inline void rseq_execve(struct task_struct *t) 1893 { 1894 } 1895 1896 #endif 1897 1898 void __exit_umh(struct task_struct *tsk); 1899 1900 static inline void exit_umh(struct task_struct *tsk) 1901 { 1902 if (unlikely(tsk->flags & PF_UMH)) 1903 __exit_umh(tsk); 1904 } 1905 1906 #ifdef CONFIG_DEBUG_RSEQ 1907 1908 void rseq_syscall(struct pt_regs *regs); 1909 1910 #else 1911 1912 static inline void rseq_syscall(struct pt_regs *regs) 1913 { 1914 } 1915 1916 #endif 1917 1918 #endif 1919