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/mutex.h> 18 #include <linux/plist.h> 19 #include <linux/hrtimer.h> 20 #include <linux/irqflags.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/sched/types.h> 29 #include <linux/signal_types.h> 30 #include <linux/syscall_user_dispatch.h> 31 #include <linux/mm_types_task.h> 32 #include <linux/task_io_accounting.h> 33 #include <linux/posix-timers.h> 34 #include <linux/rseq.h> 35 #include <linux/seqlock.h> 36 #include <linux/kcsan.h> 37 #include <asm/kmap_size.h> 38 39 /* task_struct member predeclarations (sorted alphabetically): */ 40 struct audit_context; 41 struct backing_dev_info; 42 struct bio_list; 43 struct blk_plug; 44 struct capture_control; 45 struct cfs_rq; 46 struct fs_struct; 47 struct futex_pi_state; 48 struct io_context; 49 struct io_uring_task; 50 struct mempolicy; 51 struct nameidata; 52 struct nsproxy; 53 struct perf_event_context; 54 struct pid_namespace; 55 struct pipe_inode_info; 56 struct rcu_node; 57 struct reclaim_state; 58 struct robust_list_head; 59 struct root_domain; 60 struct rq; 61 struct sched_attr; 62 struct sched_param; 63 struct seq_file; 64 struct sighand_struct; 65 struct signal_struct; 66 struct task_delay_info; 67 struct task_group; 68 69 /* 70 * Task state bitmask. NOTE! These bits are also 71 * encoded in fs/proc/array.c: get_task_state(). 72 * 73 * We have two separate sets of flags: task->state 74 * is about runnability, while task->exit_state are 75 * about the task exiting. Confusing, but this way 76 * modifying one set can't modify the other one by 77 * mistake. 78 */ 79 80 /* Used in tsk->state: */ 81 #define TASK_RUNNING 0x0000 82 #define TASK_INTERRUPTIBLE 0x0001 83 #define TASK_UNINTERRUPTIBLE 0x0002 84 #define __TASK_STOPPED 0x0004 85 #define __TASK_TRACED 0x0008 86 /* Used in tsk->exit_state: */ 87 #define EXIT_DEAD 0x0010 88 #define EXIT_ZOMBIE 0x0020 89 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 90 /* Used in tsk->state again: */ 91 #define TASK_PARKED 0x0040 92 #define TASK_DEAD 0x0080 93 #define TASK_WAKEKILL 0x0100 94 #define TASK_WAKING 0x0200 95 #define TASK_NOLOAD 0x0400 96 #define TASK_NEW 0x0800 97 #define TASK_STATE_MAX 0x1000 98 99 /* Convenience macros for the sake of set_current_state: */ 100 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 101 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 102 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) 103 104 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 105 106 /* Convenience macros for the sake of wake_up(): */ 107 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 108 109 /* get_task_state(): */ 110 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 111 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 112 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 113 TASK_PARKED) 114 115 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0) 116 117 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0) 118 119 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) 120 121 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 122 123 /* 124 * Special states are those that do not use the normal wait-loop pattern. See 125 * the comment with set_special_state(). 126 */ 127 #define is_special_task_state(state) \ 128 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD)) 129 130 #define __set_current_state(state_value) \ 131 do { \ 132 WARN_ON_ONCE(is_special_task_state(state_value));\ 133 current->task_state_change = _THIS_IP_; \ 134 current->state = (state_value); \ 135 } while (0) 136 137 #define set_current_state(state_value) \ 138 do { \ 139 WARN_ON_ONCE(is_special_task_state(state_value));\ 140 current->task_state_change = _THIS_IP_; \ 141 smp_store_mb(current->state, (state_value)); \ 142 } while (0) 143 144 #define set_special_state(state_value) \ 145 do { \ 146 unsigned long flags; /* may shadow */ \ 147 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 148 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 149 current->task_state_change = _THIS_IP_; \ 150 current->state = (state_value); \ 151 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 152 } while (0) 153 #else 154 /* 155 * set_current_state() includes a barrier so that the write of current->state 156 * is correctly serialised wrt the caller's subsequent test of whether to 157 * actually sleep: 158 * 159 * for (;;) { 160 * set_current_state(TASK_UNINTERRUPTIBLE); 161 * if (CONDITION) 162 * break; 163 * 164 * schedule(); 165 * } 166 * __set_current_state(TASK_RUNNING); 167 * 168 * If the caller does not need such serialisation (because, for instance, the 169 * CONDITION test and condition change and wakeup are under the same lock) then 170 * use __set_current_state(). 171 * 172 * The above is typically ordered against the wakeup, which does: 173 * 174 * CONDITION = 1; 175 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 176 * 177 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 178 * accessing p->state. 179 * 180 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 181 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 182 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 183 * 184 * However, with slightly different timing the wakeup TASK_RUNNING store can 185 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 186 * a problem either because that will result in one extra go around the loop 187 * and our @cond test will save the day. 188 * 189 * Also see the comments of try_to_wake_up(). 190 */ 191 #define __set_current_state(state_value) \ 192 current->state = (state_value) 193 194 #define set_current_state(state_value) \ 195 smp_store_mb(current->state, (state_value)) 196 197 /* 198 * set_special_state() should be used for those states when the blocking task 199 * can not use the regular condition based wait-loop. In that case we must 200 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores 201 * will not collide with our state change. 202 */ 203 #define set_special_state(state_value) \ 204 do { \ 205 unsigned long flags; /* may shadow */ \ 206 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 207 current->state = (state_value); \ 208 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 209 } while (0) 210 211 #endif 212 213 /* Task command name length: */ 214 #define TASK_COMM_LEN 16 215 216 extern void scheduler_tick(void); 217 218 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 219 220 extern long schedule_timeout(long timeout); 221 extern long schedule_timeout_interruptible(long timeout); 222 extern long schedule_timeout_killable(long timeout); 223 extern long schedule_timeout_uninterruptible(long timeout); 224 extern long schedule_timeout_idle(long timeout); 225 asmlinkage void schedule(void); 226 extern void schedule_preempt_disabled(void); 227 asmlinkage void preempt_schedule_irq(void); 228 229 extern int __must_check io_schedule_prepare(void); 230 extern void io_schedule_finish(int token); 231 extern long io_schedule_timeout(long timeout); 232 extern void io_schedule(void); 233 234 /** 235 * struct prev_cputime - snapshot of system and user cputime 236 * @utime: time spent in user mode 237 * @stime: time spent in system mode 238 * @lock: protects the above two fields 239 * 240 * Stores previous user/system time values such that we can guarantee 241 * monotonicity. 242 */ 243 struct prev_cputime { 244 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 245 u64 utime; 246 u64 stime; 247 raw_spinlock_t lock; 248 #endif 249 }; 250 251 enum vtime_state { 252 /* Task is sleeping or running in a CPU with VTIME inactive: */ 253 VTIME_INACTIVE = 0, 254 /* Task is idle */ 255 VTIME_IDLE, 256 /* Task runs in kernelspace in a CPU with VTIME active: */ 257 VTIME_SYS, 258 /* Task runs in userspace in a CPU with VTIME active: */ 259 VTIME_USER, 260 /* Task runs as guests in a CPU with VTIME active: */ 261 VTIME_GUEST, 262 }; 263 264 struct vtime { 265 seqcount_t seqcount; 266 unsigned long long starttime; 267 enum vtime_state state; 268 unsigned int cpu; 269 u64 utime; 270 u64 stime; 271 u64 gtime; 272 }; 273 274 /* 275 * Utilization clamp constraints. 276 * @UCLAMP_MIN: Minimum utilization 277 * @UCLAMP_MAX: Maximum utilization 278 * @UCLAMP_CNT: Utilization clamp constraints count 279 */ 280 enum uclamp_id { 281 UCLAMP_MIN = 0, 282 UCLAMP_MAX, 283 UCLAMP_CNT 284 }; 285 286 #ifdef CONFIG_SMP 287 extern struct root_domain def_root_domain; 288 extern struct mutex sched_domains_mutex; 289 #endif 290 291 struct sched_info { 292 #ifdef CONFIG_SCHED_INFO 293 /* Cumulative counters: */ 294 295 /* # of times we have run on this CPU: */ 296 unsigned long pcount; 297 298 /* Time spent waiting on a runqueue: */ 299 unsigned long long run_delay; 300 301 /* Timestamps: */ 302 303 /* When did we last run on a CPU? */ 304 unsigned long long last_arrival; 305 306 /* When were we last queued to run? */ 307 unsigned long long last_queued; 308 309 #endif /* CONFIG_SCHED_INFO */ 310 }; 311 312 /* 313 * Integer metrics need fixed point arithmetic, e.g., sched/fair 314 * has a few: load, load_avg, util_avg, freq, and capacity. 315 * 316 * We define a basic fixed point arithmetic range, and then formalize 317 * all these metrics based on that basic range. 318 */ 319 # define SCHED_FIXEDPOINT_SHIFT 10 320 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 321 322 /* Increase resolution of cpu_capacity calculations */ 323 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 324 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 325 326 struct load_weight { 327 unsigned long weight; 328 u32 inv_weight; 329 }; 330 331 /** 332 * struct util_est - Estimation utilization of FAIR tasks 333 * @enqueued: instantaneous estimated utilization of a task/cpu 334 * @ewma: the Exponential Weighted Moving Average (EWMA) 335 * utilization of a task 336 * 337 * Support data structure to track an Exponential Weighted Moving Average 338 * (EWMA) of a FAIR task's utilization. New samples are added to the moving 339 * average each time a task completes an activation. Sample's weight is chosen 340 * so that the EWMA will be relatively insensitive to transient changes to the 341 * task's workload. 342 * 343 * The enqueued attribute has a slightly different meaning for tasks and cpus: 344 * - task: the task's util_avg at last task dequeue time 345 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU 346 * Thus, the util_est.enqueued of a task represents the contribution on the 347 * estimated utilization of the CPU where that task is currently enqueued. 348 * 349 * Only for tasks we track a moving average of the past instantaneous 350 * estimated utilization. This allows to absorb sporadic drops in utilization 351 * of an otherwise almost periodic task. 352 */ 353 struct util_est { 354 unsigned int enqueued; 355 unsigned int ewma; 356 #define UTIL_EST_WEIGHT_SHIFT 2 357 } __attribute__((__aligned__(sizeof(u64)))); 358 359 /* 360 * The load/runnable/util_avg accumulates an infinite geometric series 361 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 362 * 363 * [load_avg definition] 364 * 365 * load_avg = runnable% * scale_load_down(load) 366 * 367 * [runnable_avg definition] 368 * 369 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 370 * 371 * [util_avg definition] 372 * 373 * util_avg = running% * SCHED_CAPACITY_SCALE 374 * 375 * where runnable% is the time ratio that a sched_entity is runnable and 376 * running% the time ratio that a sched_entity is running. 377 * 378 * For cfs_rq, they are the aggregated values of all runnable and blocked 379 * sched_entities. 380 * 381 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 382 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 383 * for computing those signals (see update_rq_clock_pelt()) 384 * 385 * N.B., the above ratios (runnable% and running%) themselves are in the 386 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 387 * to as large a range as necessary. This is for example reflected by 388 * util_avg's SCHED_CAPACITY_SCALE. 389 * 390 * [Overflow issue] 391 * 392 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 393 * with the highest load (=88761), always runnable on a single cfs_rq, 394 * and should not overflow as the number already hits PID_MAX_LIMIT. 395 * 396 * For all other cases (including 32-bit kernels), struct load_weight's 397 * weight will overflow first before we do, because: 398 * 399 * Max(load_avg) <= Max(load.weight) 400 * 401 * Then it is the load_weight's responsibility to consider overflow 402 * issues. 403 */ 404 struct sched_avg { 405 u64 last_update_time; 406 u64 load_sum; 407 u64 runnable_sum; 408 u32 util_sum; 409 u32 period_contrib; 410 unsigned long load_avg; 411 unsigned long runnable_avg; 412 unsigned long util_avg; 413 struct util_est util_est; 414 } ____cacheline_aligned; 415 416 struct sched_statistics { 417 #ifdef CONFIG_SCHEDSTATS 418 u64 wait_start; 419 u64 wait_max; 420 u64 wait_count; 421 u64 wait_sum; 422 u64 iowait_count; 423 u64 iowait_sum; 424 425 u64 sleep_start; 426 u64 sleep_max; 427 s64 sum_sleep_runtime; 428 429 u64 block_start; 430 u64 block_max; 431 u64 exec_max; 432 u64 slice_max; 433 434 u64 nr_migrations_cold; 435 u64 nr_failed_migrations_affine; 436 u64 nr_failed_migrations_running; 437 u64 nr_failed_migrations_hot; 438 u64 nr_forced_migrations; 439 440 u64 nr_wakeups; 441 u64 nr_wakeups_sync; 442 u64 nr_wakeups_migrate; 443 u64 nr_wakeups_local; 444 u64 nr_wakeups_remote; 445 u64 nr_wakeups_affine; 446 u64 nr_wakeups_affine_attempts; 447 u64 nr_wakeups_passive; 448 u64 nr_wakeups_idle; 449 #endif 450 }; 451 452 struct sched_entity { 453 /* For load-balancing: */ 454 struct load_weight load; 455 struct rb_node run_node; 456 struct list_head group_node; 457 unsigned int on_rq; 458 459 u64 exec_start; 460 u64 sum_exec_runtime; 461 u64 vruntime; 462 u64 prev_sum_exec_runtime; 463 464 u64 nr_migrations; 465 466 struct sched_statistics statistics; 467 468 #ifdef CONFIG_FAIR_GROUP_SCHED 469 int depth; 470 struct sched_entity *parent; 471 /* rq on which this entity is (to be) queued: */ 472 struct cfs_rq *cfs_rq; 473 /* rq "owned" by this entity/group: */ 474 struct cfs_rq *my_q; 475 /* cached value of my_q->h_nr_running */ 476 unsigned long runnable_weight; 477 #endif 478 479 #ifdef CONFIG_SMP 480 /* 481 * Per entity load average tracking. 482 * 483 * Put into separate cache line so it does not 484 * collide with read-mostly values above. 485 */ 486 struct sched_avg avg; 487 #endif 488 }; 489 490 struct sched_rt_entity { 491 struct list_head run_list; 492 unsigned long timeout; 493 unsigned long watchdog_stamp; 494 unsigned int time_slice; 495 unsigned short on_rq; 496 unsigned short on_list; 497 498 struct sched_rt_entity *back; 499 #ifdef CONFIG_RT_GROUP_SCHED 500 struct sched_rt_entity *parent; 501 /* rq on which this entity is (to be) queued: */ 502 struct rt_rq *rt_rq; 503 /* rq "owned" by this entity/group: */ 504 struct rt_rq *my_q; 505 #endif 506 } __randomize_layout; 507 508 struct sched_dl_entity { 509 struct rb_node rb_node; 510 511 /* 512 * Original scheduling parameters. Copied here from sched_attr 513 * during sched_setattr(), they will remain the same until 514 * the next sched_setattr(). 515 */ 516 u64 dl_runtime; /* Maximum runtime for each instance */ 517 u64 dl_deadline; /* Relative deadline of each instance */ 518 u64 dl_period; /* Separation of two instances (period) */ 519 u64 dl_bw; /* dl_runtime / dl_period */ 520 u64 dl_density; /* dl_runtime / dl_deadline */ 521 522 /* 523 * Actual scheduling parameters. Initialized with the values above, 524 * they are continuously updated during task execution. Note that 525 * the remaining runtime could be < 0 in case we are in overrun. 526 */ 527 s64 runtime; /* Remaining runtime for this instance */ 528 u64 deadline; /* Absolute deadline for this instance */ 529 unsigned int flags; /* Specifying the scheduler behaviour */ 530 531 /* 532 * Some bool flags: 533 * 534 * @dl_throttled tells if we exhausted the runtime. If so, the 535 * task has to wait for a replenishment to be performed at the 536 * next firing of dl_timer. 537 * 538 * @dl_boosted tells if we are boosted due to DI. If so we are 539 * outside bandwidth enforcement mechanism (but only until we 540 * exit the critical section); 541 * 542 * @dl_yielded tells if task gave up the CPU before consuming 543 * all its available runtime during the last job. 544 * 545 * @dl_non_contending tells if the task is inactive while still 546 * contributing to the active utilization. In other words, it 547 * indicates if the inactive timer has been armed and its handler 548 * has not been executed yet. This flag is useful to avoid race 549 * conditions between the inactive timer handler and the wakeup 550 * code. 551 * 552 * @dl_overrun tells if the task asked to be informed about runtime 553 * overruns. 554 */ 555 unsigned int dl_throttled : 1; 556 unsigned int dl_yielded : 1; 557 unsigned int dl_non_contending : 1; 558 unsigned int dl_overrun : 1; 559 560 /* 561 * Bandwidth enforcement timer. Each -deadline task has its 562 * own bandwidth to be enforced, thus we need one timer per task. 563 */ 564 struct hrtimer dl_timer; 565 566 /* 567 * Inactive timer, responsible for decreasing the active utilization 568 * at the "0-lag time". When a -deadline task blocks, it contributes 569 * to GRUB's active utilization until the "0-lag time", hence a 570 * timer is needed to decrease the active utilization at the correct 571 * time. 572 */ 573 struct hrtimer inactive_timer; 574 575 #ifdef CONFIG_RT_MUTEXES 576 /* 577 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 578 * pi_se points to the donor, otherwise points to the dl_se it belongs 579 * to (the original one/itself). 580 */ 581 struct sched_dl_entity *pi_se; 582 #endif 583 }; 584 585 #ifdef CONFIG_UCLAMP_TASK 586 /* Number of utilization clamp buckets (shorter alias) */ 587 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 588 589 /* 590 * Utilization clamp for a scheduling entity 591 * @value: clamp value "assigned" to a se 592 * @bucket_id: bucket index corresponding to the "assigned" value 593 * @active: the se is currently refcounted in a rq's bucket 594 * @user_defined: the requested clamp value comes from user-space 595 * 596 * The bucket_id is the index of the clamp bucket matching the clamp value 597 * which is pre-computed and stored to avoid expensive integer divisions from 598 * the fast path. 599 * 600 * The active bit is set whenever a task has got an "effective" value assigned, 601 * which can be different from the clamp value "requested" from user-space. 602 * This allows to know a task is refcounted in the rq's bucket corresponding 603 * to the "effective" bucket_id. 604 * 605 * The user_defined bit is set whenever a task has got a task-specific clamp 606 * value requested from userspace, i.e. the system defaults apply to this task 607 * just as a restriction. This allows to relax default clamps when a less 608 * restrictive task-specific value has been requested, thus allowing to 609 * implement a "nice" semantic. For example, a task running with a 20% 610 * default boost can still drop its own boosting to 0%. 611 */ 612 struct uclamp_se { 613 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 614 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 615 unsigned int active : 1; 616 unsigned int user_defined : 1; 617 }; 618 #endif /* CONFIG_UCLAMP_TASK */ 619 620 union rcu_special { 621 struct { 622 u8 blocked; 623 u8 need_qs; 624 u8 exp_hint; /* Hint for performance. */ 625 u8 need_mb; /* Readers need smp_mb(). */ 626 } b; /* Bits. */ 627 u32 s; /* Set of bits. */ 628 }; 629 630 enum perf_event_task_context { 631 perf_invalid_context = -1, 632 perf_hw_context = 0, 633 perf_sw_context, 634 perf_nr_task_contexts, 635 }; 636 637 struct wake_q_node { 638 struct wake_q_node *next; 639 }; 640 641 struct kmap_ctrl { 642 #ifdef CONFIG_KMAP_LOCAL 643 int idx; 644 pte_t pteval[KM_MAX_IDX]; 645 #endif 646 }; 647 648 struct task_struct { 649 #ifdef CONFIG_THREAD_INFO_IN_TASK 650 /* 651 * For reasons of header soup (see current_thread_info()), this 652 * must be the first element of task_struct. 653 */ 654 struct thread_info thread_info; 655 #endif 656 /* -1 unrunnable, 0 runnable, >0 stopped: */ 657 volatile long state; 658 659 /* 660 * This begins the randomizable portion of task_struct. Only 661 * scheduling-critical items should be added above here. 662 */ 663 randomized_struct_fields_start 664 665 void *stack; 666 refcount_t usage; 667 /* Per task flags (PF_*), defined further below: */ 668 unsigned int flags; 669 unsigned int ptrace; 670 671 #ifdef CONFIG_SMP 672 int on_cpu; 673 struct __call_single_node wake_entry; 674 #ifdef CONFIG_THREAD_INFO_IN_TASK 675 /* Current CPU: */ 676 unsigned int cpu; 677 #endif 678 unsigned int wakee_flips; 679 unsigned long wakee_flip_decay_ts; 680 struct task_struct *last_wakee; 681 682 /* 683 * recent_used_cpu is initially set as the last CPU used by a task 684 * that wakes affine another task. Waker/wakee relationships can 685 * push tasks around a CPU where each wakeup moves to the next one. 686 * Tracking a recently used CPU allows a quick search for a recently 687 * used CPU that may be idle. 688 */ 689 int recent_used_cpu; 690 int wake_cpu; 691 #endif 692 int on_rq; 693 694 int prio; 695 int static_prio; 696 int normal_prio; 697 unsigned int rt_priority; 698 699 const struct sched_class *sched_class; 700 struct sched_entity se; 701 struct sched_rt_entity rt; 702 #ifdef CONFIG_CGROUP_SCHED 703 struct task_group *sched_task_group; 704 #endif 705 struct sched_dl_entity dl; 706 707 #ifdef CONFIG_UCLAMP_TASK 708 /* 709 * Clamp values requested for a scheduling entity. 710 * Must be updated with task_rq_lock() held. 711 */ 712 struct uclamp_se uclamp_req[UCLAMP_CNT]; 713 /* 714 * Effective clamp values used for a scheduling entity. 715 * Must be updated with task_rq_lock() held. 716 */ 717 struct uclamp_se uclamp[UCLAMP_CNT]; 718 #endif 719 720 #ifdef CONFIG_PREEMPT_NOTIFIERS 721 /* List of struct preempt_notifier: */ 722 struct hlist_head preempt_notifiers; 723 #endif 724 725 #ifdef CONFIG_BLK_DEV_IO_TRACE 726 unsigned int btrace_seq; 727 #endif 728 729 unsigned int policy; 730 int nr_cpus_allowed; 731 const cpumask_t *cpus_ptr; 732 cpumask_t cpus_mask; 733 void *migration_pending; 734 #ifdef CONFIG_SMP 735 unsigned short migration_disabled; 736 #endif 737 unsigned short migration_flags; 738 739 #ifdef CONFIG_PREEMPT_RCU 740 int rcu_read_lock_nesting; 741 union rcu_special rcu_read_unlock_special; 742 struct list_head rcu_node_entry; 743 struct rcu_node *rcu_blocked_node; 744 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 745 746 #ifdef CONFIG_TASKS_RCU 747 unsigned long rcu_tasks_nvcsw; 748 u8 rcu_tasks_holdout; 749 u8 rcu_tasks_idx; 750 int rcu_tasks_idle_cpu; 751 struct list_head rcu_tasks_holdout_list; 752 #endif /* #ifdef CONFIG_TASKS_RCU */ 753 754 #ifdef CONFIG_TASKS_TRACE_RCU 755 int trc_reader_nesting; 756 int trc_ipi_to_cpu; 757 union rcu_special trc_reader_special; 758 bool trc_reader_checked; 759 struct list_head trc_holdout_list; 760 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 761 762 struct sched_info sched_info; 763 764 struct list_head tasks; 765 #ifdef CONFIG_SMP 766 struct plist_node pushable_tasks; 767 struct rb_node pushable_dl_tasks; 768 #endif 769 770 struct mm_struct *mm; 771 struct mm_struct *active_mm; 772 773 /* Per-thread vma caching: */ 774 struct vmacache vmacache; 775 776 #ifdef SPLIT_RSS_COUNTING 777 struct task_rss_stat rss_stat; 778 #endif 779 int exit_state; 780 int exit_code; 781 int exit_signal; 782 /* The signal sent when the parent dies: */ 783 int pdeath_signal; 784 /* JOBCTL_*, siglock protected: */ 785 unsigned long jobctl; 786 787 /* Used for emulating ABI behavior of previous Linux versions: */ 788 unsigned int personality; 789 790 /* Scheduler bits, serialized by scheduler locks: */ 791 unsigned sched_reset_on_fork:1; 792 unsigned sched_contributes_to_load:1; 793 unsigned sched_migrated:1; 794 #ifdef CONFIG_PSI 795 unsigned sched_psi_wake_requeue:1; 796 #endif 797 798 /* Force alignment to the next boundary: */ 799 unsigned :0; 800 801 /* Unserialized, strictly 'current' */ 802 803 /* 804 * This field must not be in the scheduler word above due to wakelist 805 * queueing no longer being serialized by p->on_cpu. However: 806 * 807 * p->XXX = X; ttwu() 808 * schedule() if (p->on_rq && ..) // false 809 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 810 * deactivate_task() ttwu_queue_wakelist()) 811 * p->on_rq = 0; p->sched_remote_wakeup = Y; 812 * 813 * guarantees all stores of 'current' are visible before 814 * ->sched_remote_wakeup gets used, so it can be in this word. 815 */ 816 unsigned sched_remote_wakeup:1; 817 818 /* Bit to tell LSMs we're in execve(): */ 819 unsigned in_execve:1; 820 unsigned in_iowait:1; 821 #ifndef TIF_RESTORE_SIGMASK 822 unsigned restore_sigmask:1; 823 #endif 824 #ifdef CONFIG_MEMCG 825 unsigned in_user_fault:1; 826 #endif 827 #ifdef CONFIG_COMPAT_BRK 828 unsigned brk_randomized:1; 829 #endif 830 #ifdef CONFIG_CGROUPS 831 /* disallow userland-initiated cgroup migration */ 832 unsigned no_cgroup_migration:1; 833 /* task is frozen/stopped (used by the cgroup freezer) */ 834 unsigned frozen:1; 835 #endif 836 #ifdef CONFIG_BLK_CGROUP 837 unsigned use_memdelay:1; 838 #endif 839 #ifdef CONFIG_PSI 840 /* Stalled due to lack of memory */ 841 unsigned in_memstall:1; 842 #endif 843 844 unsigned long atomic_flags; /* Flags requiring atomic access. */ 845 846 struct restart_block restart_block; 847 848 pid_t pid; 849 pid_t tgid; 850 851 #ifdef CONFIG_STACKPROTECTOR 852 /* Canary value for the -fstack-protector GCC feature: */ 853 unsigned long stack_canary; 854 #endif 855 /* 856 * Pointers to the (original) parent process, youngest child, younger sibling, 857 * older sibling, respectively. (p->father can be replaced with 858 * p->real_parent->pid) 859 */ 860 861 /* Real parent process: */ 862 struct task_struct __rcu *real_parent; 863 864 /* Recipient of SIGCHLD, wait4() reports: */ 865 struct task_struct __rcu *parent; 866 867 /* 868 * Children/sibling form the list of natural children: 869 */ 870 struct list_head children; 871 struct list_head sibling; 872 struct task_struct *group_leader; 873 874 /* 875 * 'ptraced' is the list of tasks this task is using ptrace() on. 876 * 877 * This includes both natural children and PTRACE_ATTACH targets. 878 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 879 */ 880 struct list_head ptraced; 881 struct list_head ptrace_entry; 882 883 /* PID/PID hash table linkage. */ 884 struct pid *thread_pid; 885 struct hlist_node pid_links[PIDTYPE_MAX]; 886 struct list_head thread_group; 887 struct list_head thread_node; 888 889 struct completion *vfork_done; 890 891 /* CLONE_CHILD_SETTID: */ 892 int __user *set_child_tid; 893 894 /* CLONE_CHILD_CLEARTID: */ 895 int __user *clear_child_tid; 896 897 /* PF_IO_WORKER */ 898 void *pf_io_worker; 899 900 u64 utime; 901 u64 stime; 902 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 903 u64 utimescaled; 904 u64 stimescaled; 905 #endif 906 u64 gtime; 907 struct prev_cputime prev_cputime; 908 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 909 struct vtime vtime; 910 #endif 911 912 #ifdef CONFIG_NO_HZ_FULL 913 atomic_t tick_dep_mask; 914 #endif 915 /* Context switch counts: */ 916 unsigned long nvcsw; 917 unsigned long nivcsw; 918 919 /* Monotonic time in nsecs: */ 920 u64 start_time; 921 922 /* Boot based time in nsecs: */ 923 u64 start_boottime; 924 925 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 926 unsigned long min_flt; 927 unsigned long maj_flt; 928 929 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 930 struct posix_cputimers posix_cputimers; 931 932 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 933 struct posix_cputimers_work posix_cputimers_work; 934 #endif 935 936 /* Process credentials: */ 937 938 /* Tracer's credentials at attach: */ 939 const struct cred __rcu *ptracer_cred; 940 941 /* Objective and real subjective task credentials (COW): */ 942 const struct cred __rcu *real_cred; 943 944 /* Effective (overridable) subjective task credentials (COW): */ 945 const struct cred __rcu *cred; 946 947 #ifdef CONFIG_KEYS 948 /* Cached requested key. */ 949 struct key *cached_requested_key; 950 #endif 951 952 /* 953 * executable name, excluding path. 954 * 955 * - normally initialized setup_new_exec() 956 * - access it with [gs]et_task_comm() 957 * - lock it with task_lock() 958 */ 959 char comm[TASK_COMM_LEN]; 960 961 struct nameidata *nameidata; 962 963 #ifdef CONFIG_SYSVIPC 964 struct sysv_sem sysvsem; 965 struct sysv_shm sysvshm; 966 #endif 967 #ifdef CONFIG_DETECT_HUNG_TASK 968 unsigned long last_switch_count; 969 unsigned long last_switch_time; 970 #endif 971 /* Filesystem information: */ 972 struct fs_struct *fs; 973 974 /* Open file information: */ 975 struct files_struct *files; 976 977 #ifdef CONFIG_IO_URING 978 struct io_uring_task *io_uring; 979 #endif 980 981 /* Namespaces: */ 982 struct nsproxy *nsproxy; 983 984 /* Signal handlers: */ 985 struct signal_struct *signal; 986 struct sighand_struct __rcu *sighand; 987 struct sigqueue *sigqueue_cache; 988 sigset_t blocked; 989 sigset_t real_blocked; 990 /* Restored if set_restore_sigmask() was used: */ 991 sigset_t saved_sigmask; 992 struct sigpending pending; 993 unsigned long sas_ss_sp; 994 size_t sas_ss_size; 995 unsigned int sas_ss_flags; 996 997 struct callback_head *task_works; 998 999 #ifdef CONFIG_AUDIT 1000 #ifdef CONFIG_AUDITSYSCALL 1001 struct audit_context *audit_context; 1002 #endif 1003 kuid_t loginuid; 1004 unsigned int sessionid; 1005 #endif 1006 struct seccomp seccomp; 1007 struct syscall_user_dispatch syscall_dispatch; 1008 1009 /* Thread group tracking: */ 1010 u64 parent_exec_id; 1011 u64 self_exec_id; 1012 1013 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1014 spinlock_t alloc_lock; 1015 1016 /* Protection of the PI data structures: */ 1017 raw_spinlock_t pi_lock; 1018 1019 struct wake_q_node wake_q; 1020 1021 #ifdef CONFIG_RT_MUTEXES 1022 /* PI waiters blocked on a rt_mutex held by this task: */ 1023 struct rb_root_cached pi_waiters; 1024 /* Updated under owner's pi_lock and rq lock */ 1025 struct task_struct *pi_top_task; 1026 /* Deadlock detection and priority inheritance handling: */ 1027 struct rt_mutex_waiter *pi_blocked_on; 1028 #endif 1029 1030 #ifdef CONFIG_DEBUG_MUTEXES 1031 /* Mutex deadlock detection: */ 1032 struct mutex_waiter *blocked_on; 1033 #endif 1034 1035 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1036 int non_block_count; 1037 #endif 1038 1039 #ifdef CONFIG_TRACE_IRQFLAGS 1040 struct irqtrace_events irqtrace; 1041 unsigned int hardirq_threaded; 1042 u64 hardirq_chain_key; 1043 int softirqs_enabled; 1044 int softirq_context; 1045 int irq_config; 1046 #endif 1047 #ifdef CONFIG_PREEMPT_RT 1048 int softirq_disable_cnt; 1049 #endif 1050 1051 #ifdef CONFIG_LOCKDEP 1052 # define MAX_LOCK_DEPTH 48UL 1053 u64 curr_chain_key; 1054 int lockdep_depth; 1055 unsigned int lockdep_recursion; 1056 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1057 #endif 1058 1059 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1060 unsigned int in_ubsan; 1061 #endif 1062 1063 /* Journalling filesystem info: */ 1064 void *journal_info; 1065 1066 /* Stacked block device info: */ 1067 struct bio_list *bio_list; 1068 1069 #ifdef CONFIG_BLOCK 1070 /* Stack plugging: */ 1071 struct blk_plug *plug; 1072 #endif 1073 1074 /* VM state: */ 1075 struct reclaim_state *reclaim_state; 1076 1077 struct backing_dev_info *backing_dev_info; 1078 1079 struct io_context *io_context; 1080 1081 #ifdef CONFIG_COMPACTION 1082 struct capture_control *capture_control; 1083 #endif 1084 /* Ptrace state: */ 1085 unsigned long ptrace_message; 1086 kernel_siginfo_t *last_siginfo; 1087 1088 struct task_io_accounting ioac; 1089 #ifdef CONFIG_PSI 1090 /* Pressure stall state */ 1091 unsigned int psi_flags; 1092 #endif 1093 #ifdef CONFIG_TASK_XACCT 1094 /* Accumulated RSS usage: */ 1095 u64 acct_rss_mem1; 1096 /* Accumulated virtual memory usage: */ 1097 u64 acct_vm_mem1; 1098 /* stime + utime since last update: */ 1099 u64 acct_timexpd; 1100 #endif 1101 #ifdef CONFIG_CPUSETS 1102 /* Protected by ->alloc_lock: */ 1103 nodemask_t mems_allowed; 1104 /* Sequence number to catch updates: */ 1105 seqcount_spinlock_t mems_allowed_seq; 1106 int cpuset_mem_spread_rotor; 1107 int cpuset_slab_spread_rotor; 1108 #endif 1109 #ifdef CONFIG_CGROUPS 1110 /* Control Group info protected by css_set_lock: */ 1111 struct css_set __rcu *cgroups; 1112 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1113 struct list_head cg_list; 1114 #endif 1115 #ifdef CONFIG_X86_CPU_RESCTRL 1116 u32 closid; 1117 u32 rmid; 1118 #endif 1119 #ifdef CONFIG_FUTEX 1120 struct robust_list_head __user *robust_list; 1121 #ifdef CONFIG_COMPAT 1122 struct compat_robust_list_head __user *compat_robust_list; 1123 #endif 1124 struct list_head pi_state_list; 1125 struct futex_pi_state *pi_state_cache; 1126 struct mutex futex_exit_mutex; 1127 unsigned int futex_state; 1128 #endif 1129 #ifdef CONFIG_PERF_EVENTS 1130 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 1131 struct mutex perf_event_mutex; 1132 struct list_head perf_event_list; 1133 #endif 1134 #ifdef CONFIG_DEBUG_PREEMPT 1135 unsigned long preempt_disable_ip; 1136 #endif 1137 #ifdef CONFIG_NUMA 1138 /* Protected by alloc_lock: */ 1139 struct mempolicy *mempolicy; 1140 short il_prev; 1141 short pref_node_fork; 1142 #endif 1143 #ifdef CONFIG_NUMA_BALANCING 1144 int numa_scan_seq; 1145 unsigned int numa_scan_period; 1146 unsigned int numa_scan_period_max; 1147 int numa_preferred_nid; 1148 unsigned long numa_migrate_retry; 1149 /* Migration stamp: */ 1150 u64 node_stamp; 1151 u64 last_task_numa_placement; 1152 u64 last_sum_exec_runtime; 1153 struct callback_head numa_work; 1154 1155 /* 1156 * This pointer is only modified for current in syscall and 1157 * pagefault context (and for tasks being destroyed), so it can be read 1158 * from any of the following contexts: 1159 * - RCU read-side critical section 1160 * - current->numa_group from everywhere 1161 * - task's runqueue locked, task not running 1162 */ 1163 struct numa_group __rcu *numa_group; 1164 1165 /* 1166 * numa_faults is an array split into four regions: 1167 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1168 * in this precise order. 1169 * 1170 * faults_memory: Exponential decaying average of faults on a per-node 1171 * basis. Scheduling placement decisions are made based on these 1172 * counts. The values remain static for the duration of a PTE scan. 1173 * faults_cpu: Track the nodes the process was running on when a NUMA 1174 * hinting fault was incurred. 1175 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1176 * during the current scan window. When the scan completes, the counts 1177 * in faults_memory and faults_cpu decay and these values are copied. 1178 */ 1179 unsigned long *numa_faults; 1180 unsigned long total_numa_faults; 1181 1182 /* 1183 * numa_faults_locality tracks if faults recorded during the last 1184 * scan window were remote/local or failed to migrate. The task scan 1185 * period is adapted based on the locality of the faults with different 1186 * weights depending on whether they were shared or private faults 1187 */ 1188 unsigned long numa_faults_locality[3]; 1189 1190 unsigned long numa_pages_migrated; 1191 #endif /* CONFIG_NUMA_BALANCING */ 1192 1193 #ifdef CONFIG_RSEQ 1194 struct rseq __user *rseq; 1195 u32 rseq_sig; 1196 /* 1197 * RmW on rseq_event_mask must be performed atomically 1198 * with respect to preemption. 1199 */ 1200 unsigned long rseq_event_mask; 1201 #endif 1202 1203 struct tlbflush_unmap_batch tlb_ubc; 1204 1205 union { 1206 refcount_t rcu_users; 1207 struct rcu_head rcu; 1208 }; 1209 1210 /* Cache last used pipe for splice(): */ 1211 struct pipe_inode_info *splice_pipe; 1212 1213 struct page_frag task_frag; 1214 1215 #ifdef CONFIG_TASK_DELAY_ACCT 1216 struct task_delay_info *delays; 1217 #endif 1218 1219 #ifdef CONFIG_FAULT_INJECTION 1220 int make_it_fail; 1221 unsigned int fail_nth; 1222 #endif 1223 /* 1224 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1225 * balance_dirty_pages() for a dirty throttling pause: 1226 */ 1227 int nr_dirtied; 1228 int nr_dirtied_pause; 1229 /* Start of a write-and-pause period: */ 1230 unsigned long dirty_paused_when; 1231 1232 #ifdef CONFIG_LATENCYTOP 1233 int latency_record_count; 1234 struct latency_record latency_record[LT_SAVECOUNT]; 1235 #endif 1236 /* 1237 * Time slack values; these are used to round up poll() and 1238 * select() etc timeout values. These are in nanoseconds. 1239 */ 1240 u64 timer_slack_ns; 1241 u64 default_timer_slack_ns; 1242 1243 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1244 unsigned int kasan_depth; 1245 #endif 1246 1247 #ifdef CONFIG_KCSAN 1248 struct kcsan_ctx kcsan_ctx; 1249 #ifdef CONFIG_TRACE_IRQFLAGS 1250 struct irqtrace_events kcsan_save_irqtrace; 1251 #endif 1252 #endif 1253 1254 #if IS_ENABLED(CONFIG_KUNIT) 1255 struct kunit *kunit_test; 1256 #endif 1257 1258 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1259 /* Index of current stored address in ret_stack: */ 1260 int curr_ret_stack; 1261 int curr_ret_depth; 1262 1263 /* Stack of return addresses for return function tracing: */ 1264 struct ftrace_ret_stack *ret_stack; 1265 1266 /* Timestamp for last schedule: */ 1267 unsigned long long ftrace_timestamp; 1268 1269 /* 1270 * Number of functions that haven't been traced 1271 * because of depth overrun: 1272 */ 1273 atomic_t trace_overrun; 1274 1275 /* Pause tracing: */ 1276 atomic_t tracing_graph_pause; 1277 #endif 1278 1279 #ifdef CONFIG_TRACING 1280 /* State flags for use by tracers: */ 1281 unsigned long trace; 1282 1283 /* Bitmask and counter of trace recursion: */ 1284 unsigned long trace_recursion; 1285 #endif /* CONFIG_TRACING */ 1286 1287 #ifdef CONFIG_KCOV 1288 /* See kernel/kcov.c for more details. */ 1289 1290 /* Coverage collection mode enabled for this task (0 if disabled): */ 1291 unsigned int kcov_mode; 1292 1293 /* Size of the kcov_area: */ 1294 unsigned int kcov_size; 1295 1296 /* Buffer for coverage collection: */ 1297 void *kcov_area; 1298 1299 /* KCOV descriptor wired with this task or NULL: */ 1300 struct kcov *kcov; 1301 1302 /* KCOV common handle for remote coverage collection: */ 1303 u64 kcov_handle; 1304 1305 /* KCOV sequence number: */ 1306 int kcov_sequence; 1307 1308 /* Collect coverage from softirq context: */ 1309 unsigned int kcov_softirq; 1310 #endif 1311 1312 #ifdef CONFIG_MEMCG 1313 struct mem_cgroup *memcg_in_oom; 1314 gfp_t memcg_oom_gfp_mask; 1315 int memcg_oom_order; 1316 1317 /* Number of pages to reclaim on returning to userland: */ 1318 unsigned int memcg_nr_pages_over_high; 1319 1320 /* Used by memcontrol for targeted memcg charge: */ 1321 struct mem_cgroup *active_memcg; 1322 #endif 1323 1324 #ifdef CONFIG_BLK_CGROUP 1325 struct request_queue *throttle_queue; 1326 #endif 1327 1328 #ifdef CONFIG_UPROBES 1329 struct uprobe_task *utask; 1330 #endif 1331 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1332 unsigned int sequential_io; 1333 unsigned int sequential_io_avg; 1334 #endif 1335 struct kmap_ctrl kmap_ctrl; 1336 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1337 unsigned long task_state_change; 1338 #endif 1339 int pagefault_disabled; 1340 #ifdef CONFIG_MMU 1341 struct task_struct *oom_reaper_list; 1342 #endif 1343 #ifdef CONFIG_VMAP_STACK 1344 struct vm_struct *stack_vm_area; 1345 #endif 1346 #ifdef CONFIG_THREAD_INFO_IN_TASK 1347 /* A live task holds one reference: */ 1348 refcount_t stack_refcount; 1349 #endif 1350 #ifdef CONFIG_LIVEPATCH 1351 int patch_state; 1352 #endif 1353 #ifdef CONFIG_SECURITY 1354 /* Used by LSM modules for access restriction: */ 1355 void *security; 1356 #endif 1357 1358 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1359 unsigned long lowest_stack; 1360 unsigned long prev_lowest_stack; 1361 #endif 1362 1363 #ifdef CONFIG_X86_MCE 1364 void __user *mce_vaddr; 1365 __u64 mce_kflags; 1366 u64 mce_addr; 1367 __u64 mce_ripv : 1, 1368 mce_whole_page : 1, 1369 __mce_reserved : 62; 1370 struct callback_head mce_kill_me; 1371 #endif 1372 1373 #ifdef CONFIG_KRETPROBES 1374 struct llist_head kretprobe_instances; 1375 #endif 1376 1377 /* 1378 * New fields for task_struct should be added above here, so that 1379 * they are included in the randomized portion of task_struct. 1380 */ 1381 randomized_struct_fields_end 1382 1383 /* CPU-specific state of this task: */ 1384 struct thread_struct thread; 1385 1386 /* 1387 * WARNING: on x86, 'thread_struct' contains a variable-sized 1388 * structure. It *MUST* be at the end of 'task_struct'. 1389 * 1390 * Do not put anything below here! 1391 */ 1392 }; 1393 1394 static inline struct pid *task_pid(struct task_struct *task) 1395 { 1396 return task->thread_pid; 1397 } 1398 1399 /* 1400 * the helpers to get the task's different pids as they are seen 1401 * from various namespaces 1402 * 1403 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1404 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1405 * current. 1406 * task_xid_nr_ns() : id seen from the ns specified; 1407 * 1408 * see also pid_nr() etc in include/linux/pid.h 1409 */ 1410 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1411 1412 static inline pid_t task_pid_nr(struct task_struct *tsk) 1413 { 1414 return tsk->pid; 1415 } 1416 1417 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1418 { 1419 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1420 } 1421 1422 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1423 { 1424 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1425 } 1426 1427 1428 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1429 { 1430 return tsk->tgid; 1431 } 1432 1433 /** 1434 * pid_alive - check that a task structure is not stale 1435 * @p: Task structure to be checked. 1436 * 1437 * Test if a process is not yet dead (at most zombie state) 1438 * If pid_alive fails, then pointers within the task structure 1439 * can be stale and must not be dereferenced. 1440 * 1441 * Return: 1 if the process is alive. 0 otherwise. 1442 */ 1443 static inline int pid_alive(const struct task_struct *p) 1444 { 1445 return p->thread_pid != NULL; 1446 } 1447 1448 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1449 { 1450 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1451 } 1452 1453 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1454 { 1455 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1456 } 1457 1458 1459 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1460 { 1461 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1462 } 1463 1464 static inline pid_t task_session_vnr(struct task_struct *tsk) 1465 { 1466 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1467 } 1468 1469 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1470 { 1471 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns); 1472 } 1473 1474 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1475 { 1476 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL); 1477 } 1478 1479 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1480 { 1481 pid_t pid = 0; 1482 1483 rcu_read_lock(); 1484 if (pid_alive(tsk)) 1485 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1486 rcu_read_unlock(); 1487 1488 return pid; 1489 } 1490 1491 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1492 { 1493 return task_ppid_nr_ns(tsk, &init_pid_ns); 1494 } 1495 1496 /* Obsolete, do not use: */ 1497 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1498 { 1499 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1500 } 1501 1502 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1503 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1504 1505 static inline unsigned int task_state_index(struct task_struct *tsk) 1506 { 1507 unsigned int tsk_state = READ_ONCE(tsk->state); 1508 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT; 1509 1510 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1511 1512 if (tsk_state == TASK_IDLE) 1513 state = TASK_REPORT_IDLE; 1514 1515 return fls(state); 1516 } 1517 1518 static inline char task_index_to_char(unsigned int state) 1519 { 1520 static const char state_char[] = "RSDTtXZPI"; 1521 1522 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1523 1524 return state_char[state]; 1525 } 1526 1527 static inline char task_state_to_char(struct task_struct *tsk) 1528 { 1529 return task_index_to_char(task_state_index(tsk)); 1530 } 1531 1532 /** 1533 * is_global_init - check if a task structure is init. Since init 1534 * is free to have sub-threads we need to check tgid. 1535 * @tsk: Task structure to be checked. 1536 * 1537 * Check if a task structure is the first user space task the kernel created. 1538 * 1539 * Return: 1 if the task structure is init. 0 otherwise. 1540 */ 1541 static inline int is_global_init(struct task_struct *tsk) 1542 { 1543 return task_tgid_nr(tsk) == 1; 1544 } 1545 1546 extern struct pid *cad_pid; 1547 1548 /* 1549 * Per process flags 1550 */ 1551 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1552 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1553 #define PF_EXITING 0x00000004 /* Getting shut down */ 1554 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1555 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1556 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1557 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1558 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1559 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1560 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1561 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1562 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1563 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1564 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1565 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1566 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1567 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1568 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1569 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1570 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1571 * I am cleaning dirty pages from some other bdi. */ 1572 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1573 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1574 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1575 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1576 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1577 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */ 1578 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1579 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1580 1581 /* 1582 * Only the _current_ task can read/write to tsk->flags, but other 1583 * tasks can access tsk->flags in readonly mode for example 1584 * with tsk_used_math (like during threaded core dumping). 1585 * There is however an exception to this rule during ptrace 1586 * or during fork: the ptracer task is allowed to write to the 1587 * child->flags of its traced child (same goes for fork, the parent 1588 * can write to the child->flags), because we're guaranteed the 1589 * child is not running and in turn not changing child->flags 1590 * at the same time the parent does it. 1591 */ 1592 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1593 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1594 #define clear_used_math() clear_stopped_child_used_math(current) 1595 #define set_used_math() set_stopped_child_used_math(current) 1596 1597 #define conditional_stopped_child_used_math(condition, child) \ 1598 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1599 1600 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1601 1602 #define copy_to_stopped_child_used_math(child) \ 1603 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1604 1605 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1606 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1607 #define used_math() tsk_used_math(current) 1608 1609 static inline bool is_percpu_thread(void) 1610 { 1611 #ifdef CONFIG_SMP 1612 return (current->flags & PF_NO_SETAFFINITY) && 1613 (current->nr_cpus_allowed == 1); 1614 #else 1615 return true; 1616 #endif 1617 } 1618 1619 /* Per-process atomic flags. */ 1620 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1621 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1622 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1623 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1624 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1625 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1626 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1627 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1628 1629 #define TASK_PFA_TEST(name, func) \ 1630 static inline bool task_##func(struct task_struct *p) \ 1631 { return test_bit(PFA_##name, &p->atomic_flags); } 1632 1633 #define TASK_PFA_SET(name, func) \ 1634 static inline void task_set_##func(struct task_struct *p) \ 1635 { set_bit(PFA_##name, &p->atomic_flags); } 1636 1637 #define TASK_PFA_CLEAR(name, func) \ 1638 static inline void task_clear_##func(struct task_struct *p) \ 1639 { clear_bit(PFA_##name, &p->atomic_flags); } 1640 1641 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1642 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1643 1644 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1645 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1646 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1647 1648 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1649 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1650 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1651 1652 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1653 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1654 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1655 1656 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1657 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1658 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1659 1660 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1661 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1662 1663 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1664 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1665 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1666 1667 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1668 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1669 1670 static inline void 1671 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1672 { 1673 current->flags &= ~flags; 1674 current->flags |= orig_flags & flags; 1675 } 1676 1677 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1678 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1679 #ifdef CONFIG_SMP 1680 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1681 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1682 #else 1683 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1684 { 1685 } 1686 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1687 { 1688 if (!cpumask_test_cpu(0, new_mask)) 1689 return -EINVAL; 1690 return 0; 1691 } 1692 #endif 1693 1694 extern int yield_to(struct task_struct *p, bool preempt); 1695 extern void set_user_nice(struct task_struct *p, long nice); 1696 extern int task_prio(const struct task_struct *p); 1697 1698 /** 1699 * task_nice - return the nice value of a given task. 1700 * @p: the task in question. 1701 * 1702 * Return: The nice value [ -20 ... 0 ... 19 ]. 1703 */ 1704 static inline int task_nice(const struct task_struct *p) 1705 { 1706 return PRIO_TO_NICE((p)->static_prio); 1707 } 1708 1709 extern int can_nice(const struct task_struct *p, const int nice); 1710 extern int task_curr(const struct task_struct *p); 1711 extern int idle_cpu(int cpu); 1712 extern int available_idle_cpu(int cpu); 1713 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1714 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1715 extern void sched_set_fifo(struct task_struct *p); 1716 extern void sched_set_fifo_low(struct task_struct *p); 1717 extern void sched_set_normal(struct task_struct *p, int nice); 1718 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1719 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1720 extern struct task_struct *idle_task(int cpu); 1721 1722 /** 1723 * is_idle_task - is the specified task an idle task? 1724 * @p: the task in question. 1725 * 1726 * Return: 1 if @p is an idle task. 0 otherwise. 1727 */ 1728 static __always_inline bool is_idle_task(const struct task_struct *p) 1729 { 1730 return !!(p->flags & PF_IDLE); 1731 } 1732 1733 extern struct task_struct *curr_task(int cpu); 1734 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1735 1736 void yield(void); 1737 1738 union thread_union { 1739 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1740 struct task_struct task; 1741 #endif 1742 #ifndef CONFIG_THREAD_INFO_IN_TASK 1743 struct thread_info thread_info; 1744 #endif 1745 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1746 }; 1747 1748 #ifndef CONFIG_THREAD_INFO_IN_TASK 1749 extern struct thread_info init_thread_info; 1750 #endif 1751 1752 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1753 1754 #ifdef CONFIG_THREAD_INFO_IN_TASK 1755 static inline struct thread_info *task_thread_info(struct task_struct *task) 1756 { 1757 return &task->thread_info; 1758 } 1759 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1760 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1761 #endif 1762 1763 /* 1764 * find a task by one of its numerical ids 1765 * 1766 * find_task_by_pid_ns(): 1767 * finds a task by its pid in the specified namespace 1768 * find_task_by_vpid(): 1769 * finds a task by its virtual pid 1770 * 1771 * see also find_vpid() etc in include/linux/pid.h 1772 */ 1773 1774 extern struct task_struct *find_task_by_vpid(pid_t nr); 1775 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1776 1777 /* 1778 * find a task by its virtual pid and get the task struct 1779 */ 1780 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1781 1782 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1783 extern int wake_up_process(struct task_struct *tsk); 1784 extern void wake_up_new_task(struct task_struct *tsk); 1785 1786 #ifdef CONFIG_SMP 1787 extern void kick_process(struct task_struct *tsk); 1788 #else 1789 static inline void kick_process(struct task_struct *tsk) { } 1790 #endif 1791 1792 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1793 1794 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1795 { 1796 __set_task_comm(tsk, from, false); 1797 } 1798 1799 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1800 #define get_task_comm(buf, tsk) ({ \ 1801 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1802 __get_task_comm(buf, sizeof(buf), tsk); \ 1803 }) 1804 1805 #ifdef CONFIG_SMP 1806 static __always_inline void scheduler_ipi(void) 1807 { 1808 /* 1809 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1810 * TIF_NEED_RESCHED remotely (for the first time) will also send 1811 * this IPI. 1812 */ 1813 preempt_fold_need_resched(); 1814 } 1815 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 1816 #else 1817 static inline void scheduler_ipi(void) { } 1818 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) 1819 { 1820 return 1; 1821 } 1822 #endif 1823 1824 /* 1825 * Set thread flags in other task's structures. 1826 * See asm/thread_info.h for TIF_xxxx flags available: 1827 */ 1828 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1829 { 1830 set_ti_thread_flag(task_thread_info(tsk), flag); 1831 } 1832 1833 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1834 { 1835 clear_ti_thread_flag(task_thread_info(tsk), flag); 1836 } 1837 1838 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1839 bool value) 1840 { 1841 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1842 } 1843 1844 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1845 { 1846 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1847 } 1848 1849 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1850 { 1851 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1852 } 1853 1854 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1855 { 1856 return test_ti_thread_flag(task_thread_info(tsk), flag); 1857 } 1858 1859 static inline void set_tsk_need_resched(struct task_struct *tsk) 1860 { 1861 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1862 } 1863 1864 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1865 { 1866 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1867 } 1868 1869 static inline int test_tsk_need_resched(struct task_struct *tsk) 1870 { 1871 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1872 } 1873 1874 /* 1875 * cond_resched() and cond_resched_lock(): latency reduction via 1876 * explicit rescheduling in places that are safe. The return 1877 * value indicates whether a reschedule was done in fact. 1878 * cond_resched_lock() will drop the spinlock before scheduling, 1879 */ 1880 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 1881 extern int __cond_resched(void); 1882 1883 #ifdef CONFIG_PREEMPT_DYNAMIC 1884 1885 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 1886 1887 static __always_inline int _cond_resched(void) 1888 { 1889 return static_call_mod(cond_resched)(); 1890 } 1891 1892 #else 1893 1894 static inline int _cond_resched(void) 1895 { 1896 return __cond_resched(); 1897 } 1898 1899 #endif /* CONFIG_PREEMPT_DYNAMIC */ 1900 1901 #else 1902 1903 static inline int _cond_resched(void) { return 0; } 1904 1905 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */ 1906 1907 #define cond_resched() ({ \ 1908 ___might_sleep(__FILE__, __LINE__, 0); \ 1909 _cond_resched(); \ 1910 }) 1911 1912 extern int __cond_resched_lock(spinlock_t *lock); 1913 extern int __cond_resched_rwlock_read(rwlock_t *lock); 1914 extern int __cond_resched_rwlock_write(rwlock_t *lock); 1915 1916 #define cond_resched_lock(lock) ({ \ 1917 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ 1918 __cond_resched_lock(lock); \ 1919 }) 1920 1921 #define cond_resched_rwlock_read(lock) ({ \ 1922 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \ 1923 __cond_resched_rwlock_read(lock); \ 1924 }) 1925 1926 #define cond_resched_rwlock_write(lock) ({ \ 1927 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \ 1928 __cond_resched_rwlock_write(lock); \ 1929 }) 1930 1931 static inline void cond_resched_rcu(void) 1932 { 1933 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 1934 rcu_read_unlock(); 1935 cond_resched(); 1936 rcu_read_lock(); 1937 #endif 1938 } 1939 1940 /* 1941 * Does a critical section need to be broken due to another 1942 * task waiting?: (technically does not depend on CONFIG_PREEMPTION, 1943 * but a general need for low latency) 1944 */ 1945 static inline int spin_needbreak(spinlock_t *lock) 1946 { 1947 #ifdef CONFIG_PREEMPTION 1948 return spin_is_contended(lock); 1949 #else 1950 return 0; 1951 #endif 1952 } 1953 1954 /* 1955 * Check if a rwlock is contended. 1956 * Returns non-zero if there is another task waiting on the rwlock. 1957 * Returns zero if the lock is not contended or the system / underlying 1958 * rwlock implementation does not support contention detection. 1959 * Technically does not depend on CONFIG_PREEMPTION, but a general need 1960 * for low latency. 1961 */ 1962 static inline int rwlock_needbreak(rwlock_t *lock) 1963 { 1964 #ifdef CONFIG_PREEMPTION 1965 return rwlock_is_contended(lock); 1966 #else 1967 return 0; 1968 #endif 1969 } 1970 1971 static __always_inline bool need_resched(void) 1972 { 1973 return unlikely(tif_need_resched()); 1974 } 1975 1976 /* 1977 * Wrappers for p->thread_info->cpu access. No-op on UP. 1978 */ 1979 #ifdef CONFIG_SMP 1980 1981 static inline unsigned int task_cpu(const struct task_struct *p) 1982 { 1983 #ifdef CONFIG_THREAD_INFO_IN_TASK 1984 return READ_ONCE(p->cpu); 1985 #else 1986 return READ_ONCE(task_thread_info(p)->cpu); 1987 #endif 1988 } 1989 1990 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 1991 1992 #else 1993 1994 static inline unsigned int task_cpu(const struct task_struct *p) 1995 { 1996 return 0; 1997 } 1998 1999 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2000 { 2001 } 2002 2003 #endif /* CONFIG_SMP */ 2004 2005 /* 2006 * In order to reduce various lock holder preemption latencies provide an 2007 * interface to see if a vCPU is currently running or not. 2008 * 2009 * This allows us to terminate optimistic spin loops and block, analogous to 2010 * the native optimistic spin heuristic of testing if the lock owner task is 2011 * running or not. 2012 */ 2013 #ifndef vcpu_is_preempted 2014 static inline bool vcpu_is_preempted(int cpu) 2015 { 2016 return false; 2017 } 2018 #endif 2019 2020 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2021 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2022 2023 #ifndef TASK_SIZE_OF 2024 #define TASK_SIZE_OF(tsk) TASK_SIZE 2025 #endif 2026 2027 #ifdef CONFIG_SMP 2028 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2029 unsigned long sched_cpu_util(int cpu, unsigned long max); 2030 #endif /* CONFIG_SMP */ 2031 2032 #ifdef CONFIG_RSEQ 2033 2034 /* 2035 * Map the event mask on the user-space ABI enum rseq_cs_flags 2036 * for direct mask checks. 2037 */ 2038 enum rseq_event_mask_bits { 2039 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT, 2040 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT, 2041 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT, 2042 }; 2043 2044 enum rseq_event_mask { 2045 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT), 2046 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT), 2047 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT), 2048 }; 2049 2050 static inline void rseq_set_notify_resume(struct task_struct *t) 2051 { 2052 if (t->rseq) 2053 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME); 2054 } 2055 2056 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs); 2057 2058 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2059 struct pt_regs *regs) 2060 { 2061 if (current->rseq) 2062 __rseq_handle_notify_resume(ksig, regs); 2063 } 2064 2065 static inline void rseq_signal_deliver(struct ksignal *ksig, 2066 struct pt_regs *regs) 2067 { 2068 preempt_disable(); 2069 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask); 2070 preempt_enable(); 2071 rseq_handle_notify_resume(ksig, regs); 2072 } 2073 2074 /* rseq_preempt() requires preemption to be disabled. */ 2075 static inline void rseq_preempt(struct task_struct *t) 2076 { 2077 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask); 2078 rseq_set_notify_resume(t); 2079 } 2080 2081 /* rseq_migrate() requires preemption to be disabled. */ 2082 static inline void rseq_migrate(struct task_struct *t) 2083 { 2084 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask); 2085 rseq_set_notify_resume(t); 2086 } 2087 2088 /* 2089 * If parent process has a registered restartable sequences area, the 2090 * child inherits. Unregister rseq for a clone with CLONE_VM set. 2091 */ 2092 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2093 { 2094 if (clone_flags & CLONE_VM) { 2095 t->rseq = NULL; 2096 t->rseq_sig = 0; 2097 t->rseq_event_mask = 0; 2098 } else { 2099 t->rseq = current->rseq; 2100 t->rseq_sig = current->rseq_sig; 2101 t->rseq_event_mask = current->rseq_event_mask; 2102 } 2103 } 2104 2105 static inline void rseq_execve(struct task_struct *t) 2106 { 2107 t->rseq = NULL; 2108 t->rseq_sig = 0; 2109 t->rseq_event_mask = 0; 2110 } 2111 2112 #else 2113 2114 static inline void rseq_set_notify_resume(struct task_struct *t) 2115 { 2116 } 2117 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2118 struct pt_regs *regs) 2119 { 2120 } 2121 static inline void rseq_signal_deliver(struct ksignal *ksig, 2122 struct pt_regs *regs) 2123 { 2124 } 2125 static inline void rseq_preempt(struct task_struct *t) 2126 { 2127 } 2128 static inline void rseq_migrate(struct task_struct *t) 2129 { 2130 } 2131 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2132 { 2133 } 2134 static inline void rseq_execve(struct task_struct *t) 2135 { 2136 } 2137 2138 #endif 2139 2140 #ifdef CONFIG_DEBUG_RSEQ 2141 2142 void rseq_syscall(struct pt_regs *regs); 2143 2144 #else 2145 2146 static inline void rseq_syscall(struct pt_regs *regs) 2147 { 2148 } 2149 2150 #endif 2151 2152 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq); 2153 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len); 2154 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq); 2155 2156 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq); 2157 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq); 2158 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq); 2159 2160 int sched_trace_rq_cpu(struct rq *rq); 2161 int sched_trace_rq_cpu_capacity(struct rq *rq); 2162 int sched_trace_rq_nr_running(struct rq *rq); 2163 2164 const struct cpumask *sched_trace_rd_span(struct root_domain *rd); 2165 2166 #endif 2167