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 bpf_local_storage; 45 struct capture_control; 46 struct cfs_rq; 47 struct fs_struct; 48 struct futex_pi_state; 49 struct io_context; 50 struct io_uring_task; 51 struct mempolicy; 52 struct nameidata; 53 struct nsproxy; 54 struct perf_event_context; 55 struct pid_namespace; 56 struct pipe_inode_info; 57 struct rcu_node; 58 struct reclaim_state; 59 struct robust_list_head; 60 struct root_domain; 61 struct rq; 62 struct sched_attr; 63 struct sched_param; 64 struct seq_file; 65 struct sighand_struct; 66 struct signal_struct; 67 struct task_delay_info; 68 struct task_group; 69 70 /* 71 * Task state bitmask. NOTE! These bits are also 72 * encoded in fs/proc/array.c: get_task_state(). 73 * 74 * We have two separate sets of flags: task->state 75 * is about runnability, while task->exit_state are 76 * about the task exiting. Confusing, but this way 77 * modifying one set can't modify the other one by 78 * mistake. 79 */ 80 81 /* Used in tsk->state: */ 82 #define TASK_RUNNING 0x0000 83 #define TASK_INTERRUPTIBLE 0x0001 84 #define TASK_UNINTERRUPTIBLE 0x0002 85 #define __TASK_STOPPED 0x0004 86 #define __TASK_TRACED 0x0008 87 /* Used in tsk->exit_state: */ 88 #define EXIT_DEAD 0x0010 89 #define EXIT_ZOMBIE 0x0020 90 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 91 /* Used in tsk->state again: */ 92 #define TASK_PARKED 0x0040 93 #define TASK_DEAD 0x0080 94 #define TASK_WAKEKILL 0x0100 95 #define TASK_WAKING 0x0200 96 #define TASK_NOLOAD 0x0400 97 #define TASK_NEW 0x0800 98 #define TASK_STATE_MAX 0x1000 99 100 /* Convenience macros for the sake of set_current_state: */ 101 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 102 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 103 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) 104 105 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 106 107 /* Convenience macros for the sake of wake_up(): */ 108 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 109 110 /* get_task_state(): */ 111 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 112 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 113 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 114 TASK_PARKED) 115 116 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0) 117 118 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0) 119 120 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) 121 122 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 123 124 /* 125 * Special states are those that do not use the normal wait-loop pattern. See 126 * the comment with set_special_state(). 127 */ 128 #define is_special_task_state(state) \ 129 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD)) 130 131 #define __set_current_state(state_value) \ 132 do { \ 133 WARN_ON_ONCE(is_special_task_state(state_value));\ 134 current->task_state_change = _THIS_IP_; \ 135 current->state = (state_value); \ 136 } while (0) 137 138 #define set_current_state(state_value) \ 139 do { \ 140 WARN_ON_ONCE(is_special_task_state(state_value));\ 141 current->task_state_change = _THIS_IP_; \ 142 smp_store_mb(current->state, (state_value)); \ 143 } while (0) 144 145 #define set_special_state(state_value) \ 146 do { \ 147 unsigned long flags; /* may shadow */ \ 148 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 149 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 150 current->task_state_change = _THIS_IP_; \ 151 current->state = (state_value); \ 152 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 153 } while (0) 154 #else 155 /* 156 * set_current_state() includes a barrier so that the write of current->state 157 * is correctly serialised wrt the caller's subsequent test of whether to 158 * actually sleep: 159 * 160 * for (;;) { 161 * set_current_state(TASK_UNINTERRUPTIBLE); 162 * if (CONDITION) 163 * break; 164 * 165 * schedule(); 166 * } 167 * __set_current_state(TASK_RUNNING); 168 * 169 * If the caller does not need such serialisation (because, for instance, the 170 * CONDITION test and condition change and wakeup are under the same lock) then 171 * use __set_current_state(). 172 * 173 * The above is typically ordered against the wakeup, which does: 174 * 175 * CONDITION = 1; 176 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 177 * 178 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 179 * accessing p->state. 180 * 181 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 182 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 183 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 184 * 185 * However, with slightly different timing the wakeup TASK_RUNNING store can 186 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 187 * a problem either because that will result in one extra go around the loop 188 * and our @cond test will save the day. 189 * 190 * Also see the comments of try_to_wake_up(). 191 */ 192 #define __set_current_state(state_value) \ 193 current->state = (state_value) 194 195 #define set_current_state(state_value) \ 196 smp_store_mb(current->state, (state_value)) 197 198 /* 199 * set_special_state() should be used for those states when the blocking task 200 * can not use the regular condition based wait-loop. In that case we must 201 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores 202 * will not collide with our state change. 203 */ 204 #define set_special_state(state_value) \ 205 do { \ 206 unsigned long flags; /* may shadow */ \ 207 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 208 current->state = (state_value); \ 209 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 210 } while (0) 211 212 #endif 213 214 /* Task command name length: */ 215 #define TASK_COMM_LEN 16 216 217 extern void scheduler_tick(void); 218 219 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 220 221 extern long schedule_timeout(long timeout); 222 extern long schedule_timeout_interruptible(long timeout); 223 extern long schedule_timeout_killable(long timeout); 224 extern long schedule_timeout_uninterruptible(long timeout); 225 extern long schedule_timeout_idle(long timeout); 226 asmlinkage void schedule(void); 227 extern void schedule_preempt_disabled(void); 228 asmlinkage void preempt_schedule_irq(void); 229 230 extern int __must_check io_schedule_prepare(void); 231 extern void io_schedule_finish(int token); 232 extern long io_schedule_timeout(long timeout); 233 extern void io_schedule(void); 234 235 /** 236 * struct prev_cputime - snapshot of system and user cputime 237 * @utime: time spent in user mode 238 * @stime: time spent in system mode 239 * @lock: protects the above two fields 240 * 241 * Stores previous user/system time values such that we can guarantee 242 * monotonicity. 243 */ 244 struct prev_cputime { 245 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 246 u64 utime; 247 u64 stime; 248 raw_spinlock_t lock; 249 #endif 250 }; 251 252 enum vtime_state { 253 /* Task is sleeping or running in a CPU with VTIME inactive: */ 254 VTIME_INACTIVE = 0, 255 /* Task is idle */ 256 VTIME_IDLE, 257 /* Task runs in kernelspace in a CPU with VTIME active: */ 258 VTIME_SYS, 259 /* Task runs in userspace in a CPU with VTIME active: */ 260 VTIME_USER, 261 /* Task runs as guests in a CPU with VTIME active: */ 262 VTIME_GUEST, 263 }; 264 265 struct vtime { 266 seqcount_t seqcount; 267 unsigned long long starttime; 268 enum vtime_state state; 269 unsigned int cpu; 270 u64 utime; 271 u64 stime; 272 u64 gtime; 273 }; 274 275 /* 276 * Utilization clamp constraints. 277 * @UCLAMP_MIN: Minimum utilization 278 * @UCLAMP_MAX: Maximum utilization 279 * @UCLAMP_CNT: Utilization clamp constraints count 280 */ 281 enum uclamp_id { 282 UCLAMP_MIN = 0, 283 UCLAMP_MAX, 284 UCLAMP_CNT 285 }; 286 287 #ifdef CONFIG_SMP 288 extern struct root_domain def_root_domain; 289 extern struct mutex sched_domains_mutex; 290 #endif 291 292 struct sched_info { 293 #ifdef CONFIG_SCHED_INFO 294 /* Cumulative counters: */ 295 296 /* # of times we have run on this CPU: */ 297 unsigned long pcount; 298 299 /* Time spent waiting on a runqueue: */ 300 unsigned long long run_delay; 301 302 /* Timestamps: */ 303 304 /* When did we last run on a CPU? */ 305 unsigned long long last_arrival; 306 307 /* When were we last queued to run? */ 308 unsigned long long last_queued; 309 310 #endif /* CONFIG_SCHED_INFO */ 311 }; 312 313 /* 314 * Integer metrics need fixed point arithmetic, e.g., sched/fair 315 * has a few: load, load_avg, util_avg, freq, and capacity. 316 * 317 * We define a basic fixed point arithmetic range, and then formalize 318 * all these metrics based on that basic range. 319 */ 320 # define SCHED_FIXEDPOINT_SHIFT 10 321 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 322 323 /* Increase resolution of cpu_capacity calculations */ 324 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 325 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 326 327 struct load_weight { 328 unsigned long weight; 329 u32 inv_weight; 330 }; 331 332 /** 333 * struct util_est - Estimation utilization of FAIR tasks 334 * @enqueued: instantaneous estimated utilization of a task/cpu 335 * @ewma: the Exponential Weighted Moving Average (EWMA) 336 * utilization of a task 337 * 338 * Support data structure to track an Exponential Weighted Moving Average 339 * (EWMA) of a FAIR task's utilization. New samples are added to the moving 340 * average each time a task completes an activation. Sample's weight is chosen 341 * so that the EWMA will be relatively insensitive to transient changes to the 342 * task's workload. 343 * 344 * The enqueued attribute has a slightly different meaning for tasks and cpus: 345 * - task: the task's util_avg at last task dequeue time 346 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU 347 * Thus, the util_est.enqueued of a task represents the contribution on the 348 * estimated utilization of the CPU where that task is currently enqueued. 349 * 350 * Only for tasks we track a moving average of the past instantaneous 351 * estimated utilization. This allows to absorb sporadic drops in utilization 352 * of an otherwise almost periodic task. 353 */ 354 struct util_est { 355 unsigned int enqueued; 356 unsigned int ewma; 357 #define UTIL_EST_WEIGHT_SHIFT 2 358 } __attribute__((__aligned__(sizeof(u64)))); 359 360 /* 361 * The load/runnable/util_avg accumulates an infinite geometric series 362 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 363 * 364 * [load_avg definition] 365 * 366 * load_avg = runnable% * scale_load_down(load) 367 * 368 * [runnable_avg definition] 369 * 370 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 371 * 372 * [util_avg definition] 373 * 374 * util_avg = running% * SCHED_CAPACITY_SCALE 375 * 376 * where runnable% is the time ratio that a sched_entity is runnable and 377 * running% the time ratio that a sched_entity is running. 378 * 379 * For cfs_rq, they are the aggregated values of all runnable and blocked 380 * sched_entities. 381 * 382 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 383 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 384 * for computing those signals (see update_rq_clock_pelt()) 385 * 386 * N.B., the above ratios (runnable% and running%) themselves are in the 387 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 388 * to as large a range as necessary. This is for example reflected by 389 * util_avg's SCHED_CAPACITY_SCALE. 390 * 391 * [Overflow issue] 392 * 393 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 394 * with the highest load (=88761), always runnable on a single cfs_rq, 395 * and should not overflow as the number already hits PID_MAX_LIMIT. 396 * 397 * For all other cases (including 32-bit kernels), struct load_weight's 398 * weight will overflow first before we do, because: 399 * 400 * Max(load_avg) <= Max(load.weight) 401 * 402 * Then it is the load_weight's responsibility to consider overflow 403 * issues. 404 */ 405 struct sched_avg { 406 u64 last_update_time; 407 u64 load_sum; 408 u64 runnable_sum; 409 u32 util_sum; 410 u32 period_contrib; 411 unsigned long load_avg; 412 unsigned long runnable_avg; 413 unsigned long util_avg; 414 struct util_est util_est; 415 } ____cacheline_aligned; 416 417 struct sched_statistics { 418 #ifdef CONFIG_SCHEDSTATS 419 u64 wait_start; 420 u64 wait_max; 421 u64 wait_count; 422 u64 wait_sum; 423 u64 iowait_count; 424 u64 iowait_sum; 425 426 u64 sleep_start; 427 u64 sleep_max; 428 s64 sum_sleep_runtime; 429 430 u64 block_start; 431 u64 block_max; 432 u64 exec_max; 433 u64 slice_max; 434 435 u64 nr_migrations_cold; 436 u64 nr_failed_migrations_affine; 437 u64 nr_failed_migrations_running; 438 u64 nr_failed_migrations_hot; 439 u64 nr_forced_migrations; 440 441 u64 nr_wakeups; 442 u64 nr_wakeups_sync; 443 u64 nr_wakeups_migrate; 444 u64 nr_wakeups_local; 445 u64 nr_wakeups_remote; 446 u64 nr_wakeups_affine; 447 u64 nr_wakeups_affine_attempts; 448 u64 nr_wakeups_passive; 449 u64 nr_wakeups_idle; 450 #endif 451 }; 452 453 struct sched_entity { 454 /* For load-balancing: */ 455 struct load_weight load; 456 struct rb_node run_node; 457 struct list_head group_node; 458 unsigned int on_rq; 459 460 u64 exec_start; 461 u64 sum_exec_runtime; 462 u64 vruntime; 463 u64 prev_sum_exec_runtime; 464 465 u64 nr_migrations; 466 467 struct sched_statistics statistics; 468 469 #ifdef CONFIG_FAIR_GROUP_SCHED 470 int depth; 471 struct sched_entity *parent; 472 /* rq on which this entity is (to be) queued: */ 473 struct cfs_rq *cfs_rq; 474 /* rq "owned" by this entity/group: */ 475 struct cfs_rq *my_q; 476 /* cached value of my_q->h_nr_running */ 477 unsigned long runnable_weight; 478 #endif 479 480 #ifdef CONFIG_SMP 481 /* 482 * Per entity load average tracking. 483 * 484 * Put into separate cache line so it does not 485 * collide with read-mostly values above. 486 */ 487 struct sched_avg avg; 488 #endif 489 }; 490 491 struct sched_rt_entity { 492 struct list_head run_list; 493 unsigned long timeout; 494 unsigned long watchdog_stamp; 495 unsigned int time_slice; 496 unsigned short on_rq; 497 unsigned short on_list; 498 499 struct sched_rt_entity *back; 500 #ifdef CONFIG_RT_GROUP_SCHED 501 struct sched_rt_entity *parent; 502 /* rq on which this entity is (to be) queued: */ 503 struct rt_rq *rt_rq; 504 /* rq "owned" by this entity/group: */ 505 struct rt_rq *my_q; 506 #endif 507 } __randomize_layout; 508 509 struct sched_dl_entity { 510 struct rb_node rb_node; 511 512 /* 513 * Original scheduling parameters. Copied here from sched_attr 514 * during sched_setattr(), they will remain the same until 515 * the next sched_setattr(). 516 */ 517 u64 dl_runtime; /* Maximum runtime for each instance */ 518 u64 dl_deadline; /* Relative deadline of each instance */ 519 u64 dl_period; /* Separation of two instances (period) */ 520 u64 dl_bw; /* dl_runtime / dl_period */ 521 u64 dl_density; /* dl_runtime / dl_deadline */ 522 523 /* 524 * Actual scheduling parameters. Initialized with the values above, 525 * they are continuously updated during task execution. Note that 526 * the remaining runtime could be < 0 in case we are in overrun. 527 */ 528 s64 runtime; /* Remaining runtime for this instance */ 529 u64 deadline; /* Absolute deadline for this instance */ 530 unsigned int flags; /* Specifying the scheduler behaviour */ 531 532 /* 533 * Some bool flags: 534 * 535 * @dl_throttled tells if we exhausted the runtime. If so, the 536 * task has to wait for a replenishment to be performed at the 537 * next firing of dl_timer. 538 * 539 * @dl_boosted tells if we are boosted due to DI. If so we are 540 * outside bandwidth enforcement mechanism (but only until we 541 * exit the critical section); 542 * 543 * @dl_yielded tells if task gave up the CPU before consuming 544 * all its available runtime during the last job. 545 * 546 * @dl_non_contending tells if the task is inactive while still 547 * contributing to the active utilization. In other words, it 548 * indicates if the inactive timer has been armed and its handler 549 * has not been executed yet. This flag is useful to avoid race 550 * conditions between the inactive timer handler and the wakeup 551 * code. 552 * 553 * @dl_overrun tells if the task asked to be informed about runtime 554 * overruns. 555 */ 556 unsigned int dl_throttled : 1; 557 unsigned int dl_yielded : 1; 558 unsigned int dl_non_contending : 1; 559 unsigned int dl_overrun : 1; 560 561 /* 562 * Bandwidth enforcement timer. Each -deadline task has its 563 * own bandwidth to be enforced, thus we need one timer per task. 564 */ 565 struct hrtimer dl_timer; 566 567 /* 568 * Inactive timer, responsible for decreasing the active utilization 569 * at the "0-lag time". When a -deadline task blocks, it contributes 570 * to GRUB's active utilization until the "0-lag time", hence a 571 * timer is needed to decrease the active utilization at the correct 572 * time. 573 */ 574 struct hrtimer inactive_timer; 575 576 #ifdef CONFIG_RT_MUTEXES 577 /* 578 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 579 * pi_se points to the donor, otherwise points to the dl_se it belongs 580 * to (the original one/itself). 581 */ 582 struct sched_dl_entity *pi_se; 583 #endif 584 }; 585 586 #ifdef CONFIG_UCLAMP_TASK 587 /* Number of utilization clamp buckets (shorter alias) */ 588 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 589 590 /* 591 * Utilization clamp for a scheduling entity 592 * @value: clamp value "assigned" to a se 593 * @bucket_id: bucket index corresponding to the "assigned" value 594 * @active: the se is currently refcounted in a rq's bucket 595 * @user_defined: the requested clamp value comes from user-space 596 * 597 * The bucket_id is the index of the clamp bucket matching the clamp value 598 * which is pre-computed and stored to avoid expensive integer divisions from 599 * the fast path. 600 * 601 * The active bit is set whenever a task has got an "effective" value assigned, 602 * which can be different from the clamp value "requested" from user-space. 603 * This allows to know a task is refcounted in the rq's bucket corresponding 604 * to the "effective" bucket_id. 605 * 606 * The user_defined bit is set whenever a task has got a task-specific clamp 607 * value requested from userspace, i.e. the system defaults apply to this task 608 * just as a restriction. This allows to relax default clamps when a less 609 * restrictive task-specific value has been requested, thus allowing to 610 * implement a "nice" semantic. For example, a task running with a 20% 611 * default boost can still drop its own boosting to 0%. 612 */ 613 struct uclamp_se { 614 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 615 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 616 unsigned int active : 1; 617 unsigned int user_defined : 1; 618 }; 619 #endif /* CONFIG_UCLAMP_TASK */ 620 621 union rcu_special { 622 struct { 623 u8 blocked; 624 u8 need_qs; 625 u8 exp_hint; /* Hint for performance. */ 626 u8 need_mb; /* Readers need smp_mb(). */ 627 } b; /* Bits. */ 628 u32 s; /* Set of bits. */ 629 }; 630 631 enum perf_event_task_context { 632 perf_invalid_context = -1, 633 perf_hw_context = 0, 634 perf_sw_context, 635 perf_nr_task_contexts, 636 }; 637 638 struct wake_q_node { 639 struct wake_q_node *next; 640 }; 641 642 struct kmap_ctrl { 643 #ifdef CONFIG_KMAP_LOCAL 644 int idx; 645 pte_t pteval[KM_MAX_IDX]; 646 #endif 647 }; 648 649 struct task_struct { 650 #ifdef CONFIG_THREAD_INFO_IN_TASK 651 /* 652 * For reasons of header soup (see current_thread_info()), this 653 * must be the first element of task_struct. 654 */ 655 struct thread_info thread_info; 656 #endif 657 /* -1 unrunnable, 0 runnable, >0 stopped: */ 658 volatile long state; 659 660 /* 661 * This begins the randomizable portion of task_struct. Only 662 * scheduling-critical items should be added above here. 663 */ 664 randomized_struct_fields_start 665 666 void *stack; 667 refcount_t usage; 668 /* Per task flags (PF_*), defined further below: */ 669 unsigned int flags; 670 unsigned int ptrace; 671 672 #ifdef CONFIG_SMP 673 int on_cpu; 674 struct __call_single_node wake_entry; 675 #ifdef CONFIG_THREAD_INFO_IN_TASK 676 /* Current CPU: */ 677 unsigned int cpu; 678 #endif 679 unsigned int wakee_flips; 680 unsigned long wakee_flip_decay_ts; 681 struct task_struct *last_wakee; 682 683 /* 684 * recent_used_cpu is initially set as the last CPU used by a task 685 * that wakes affine another task. Waker/wakee relationships can 686 * push tasks around a CPU where each wakeup moves to the next one. 687 * Tracking a recently used CPU allows a quick search for a recently 688 * used CPU that may be idle. 689 */ 690 int recent_used_cpu; 691 int wake_cpu; 692 #endif 693 int on_rq; 694 695 int prio; 696 int static_prio; 697 int normal_prio; 698 unsigned int rt_priority; 699 700 const struct sched_class *sched_class; 701 struct sched_entity se; 702 struct sched_rt_entity rt; 703 #ifdef CONFIG_CGROUP_SCHED 704 struct task_group *sched_task_group; 705 #endif 706 struct sched_dl_entity dl; 707 708 #ifdef CONFIG_UCLAMP_TASK 709 /* 710 * Clamp values requested for a scheduling entity. 711 * Must be updated with task_rq_lock() held. 712 */ 713 struct uclamp_se uclamp_req[UCLAMP_CNT]; 714 /* 715 * Effective clamp values used for a scheduling entity. 716 * Must be updated with task_rq_lock() held. 717 */ 718 struct uclamp_se uclamp[UCLAMP_CNT]; 719 #endif 720 721 #ifdef CONFIG_PREEMPT_NOTIFIERS 722 /* List of struct preempt_notifier: */ 723 struct hlist_head preempt_notifiers; 724 #endif 725 726 #ifdef CONFIG_BLK_DEV_IO_TRACE 727 unsigned int btrace_seq; 728 #endif 729 730 unsigned int policy; 731 int nr_cpus_allowed; 732 const cpumask_t *cpus_ptr; 733 cpumask_t cpus_mask; 734 void *migration_pending; 735 #ifdef CONFIG_SMP 736 unsigned short migration_disabled; 737 #endif 738 unsigned short migration_flags; 739 740 #ifdef CONFIG_PREEMPT_RCU 741 int rcu_read_lock_nesting; 742 union rcu_special rcu_read_unlock_special; 743 struct list_head rcu_node_entry; 744 struct rcu_node *rcu_blocked_node; 745 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 746 747 #ifdef CONFIG_TASKS_RCU 748 unsigned long rcu_tasks_nvcsw; 749 u8 rcu_tasks_holdout; 750 u8 rcu_tasks_idx; 751 int rcu_tasks_idle_cpu; 752 struct list_head rcu_tasks_holdout_list; 753 #endif /* #ifdef CONFIG_TASKS_RCU */ 754 755 #ifdef CONFIG_TASKS_TRACE_RCU 756 int trc_reader_nesting; 757 int trc_ipi_to_cpu; 758 union rcu_special trc_reader_special; 759 bool trc_reader_checked; 760 struct list_head trc_holdout_list; 761 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 762 763 struct sched_info sched_info; 764 765 struct list_head tasks; 766 #ifdef CONFIG_SMP 767 struct plist_node pushable_tasks; 768 struct rb_node pushable_dl_tasks; 769 #endif 770 771 struct mm_struct *mm; 772 struct mm_struct *active_mm; 773 774 /* Per-thread vma caching: */ 775 struct vmacache vmacache; 776 777 #ifdef SPLIT_RSS_COUNTING 778 struct task_rss_stat rss_stat; 779 #endif 780 int exit_state; 781 int exit_code; 782 int exit_signal; 783 /* The signal sent when the parent dies: */ 784 int pdeath_signal; 785 /* JOBCTL_*, siglock protected: */ 786 unsigned long jobctl; 787 788 /* Used for emulating ABI behavior of previous Linux versions: */ 789 unsigned int personality; 790 791 /* Scheduler bits, serialized by scheduler locks: */ 792 unsigned sched_reset_on_fork:1; 793 unsigned sched_contributes_to_load:1; 794 unsigned sched_migrated:1; 795 #ifdef CONFIG_PSI 796 unsigned sched_psi_wake_requeue:1; 797 #endif 798 799 /* Force alignment to the next boundary: */ 800 unsigned :0; 801 802 /* Unserialized, strictly 'current' */ 803 804 /* 805 * This field must not be in the scheduler word above due to wakelist 806 * queueing no longer being serialized by p->on_cpu. However: 807 * 808 * p->XXX = X; ttwu() 809 * schedule() if (p->on_rq && ..) // false 810 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 811 * deactivate_task() ttwu_queue_wakelist()) 812 * p->on_rq = 0; p->sched_remote_wakeup = Y; 813 * 814 * guarantees all stores of 'current' are visible before 815 * ->sched_remote_wakeup gets used, so it can be in this word. 816 */ 817 unsigned sched_remote_wakeup:1; 818 819 /* Bit to tell LSMs we're in execve(): */ 820 unsigned in_execve:1; 821 unsigned in_iowait:1; 822 #ifndef TIF_RESTORE_SIGMASK 823 unsigned restore_sigmask:1; 824 #endif 825 #ifdef CONFIG_MEMCG 826 unsigned in_user_fault:1; 827 #endif 828 #ifdef CONFIG_COMPAT_BRK 829 unsigned brk_randomized:1; 830 #endif 831 #ifdef CONFIG_CGROUPS 832 /* disallow userland-initiated cgroup migration */ 833 unsigned no_cgroup_migration:1; 834 /* task is frozen/stopped (used by the cgroup freezer) */ 835 unsigned frozen:1; 836 #endif 837 #ifdef CONFIG_BLK_CGROUP 838 unsigned use_memdelay:1; 839 #endif 840 #ifdef CONFIG_PSI 841 /* Stalled due to lack of memory */ 842 unsigned in_memstall:1; 843 #endif 844 #ifdef CONFIG_PAGE_OWNER 845 /* Used by page_owner=on to detect recursion in page tracking. */ 846 unsigned in_page_owner:1; 847 #endif 848 849 unsigned long atomic_flags; /* Flags requiring atomic access. */ 850 851 struct restart_block restart_block; 852 853 pid_t pid; 854 pid_t tgid; 855 856 #ifdef CONFIG_STACKPROTECTOR 857 /* Canary value for the -fstack-protector GCC feature: */ 858 unsigned long stack_canary; 859 #endif 860 /* 861 * Pointers to the (original) parent process, youngest child, younger sibling, 862 * older sibling, respectively. (p->father can be replaced with 863 * p->real_parent->pid) 864 */ 865 866 /* Real parent process: */ 867 struct task_struct __rcu *real_parent; 868 869 /* Recipient of SIGCHLD, wait4() reports: */ 870 struct task_struct __rcu *parent; 871 872 /* 873 * Children/sibling form the list of natural children: 874 */ 875 struct list_head children; 876 struct list_head sibling; 877 struct task_struct *group_leader; 878 879 /* 880 * 'ptraced' is the list of tasks this task is using ptrace() on. 881 * 882 * This includes both natural children and PTRACE_ATTACH targets. 883 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 884 */ 885 struct list_head ptraced; 886 struct list_head ptrace_entry; 887 888 /* PID/PID hash table linkage. */ 889 struct pid *thread_pid; 890 struct hlist_node pid_links[PIDTYPE_MAX]; 891 struct list_head thread_group; 892 struct list_head thread_node; 893 894 struct completion *vfork_done; 895 896 /* CLONE_CHILD_SETTID: */ 897 int __user *set_child_tid; 898 899 /* CLONE_CHILD_CLEARTID: */ 900 int __user *clear_child_tid; 901 902 /* PF_IO_WORKER */ 903 void *pf_io_worker; 904 905 u64 utime; 906 u64 stime; 907 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 908 u64 utimescaled; 909 u64 stimescaled; 910 #endif 911 u64 gtime; 912 struct prev_cputime prev_cputime; 913 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 914 struct vtime vtime; 915 #endif 916 917 #ifdef CONFIG_NO_HZ_FULL 918 atomic_t tick_dep_mask; 919 #endif 920 /* Context switch counts: */ 921 unsigned long nvcsw; 922 unsigned long nivcsw; 923 924 /* Monotonic time in nsecs: */ 925 u64 start_time; 926 927 /* Boot based time in nsecs: */ 928 u64 start_boottime; 929 930 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 931 unsigned long min_flt; 932 unsigned long maj_flt; 933 934 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 935 struct posix_cputimers posix_cputimers; 936 937 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 938 struct posix_cputimers_work posix_cputimers_work; 939 #endif 940 941 /* Process credentials: */ 942 943 /* Tracer's credentials at attach: */ 944 const struct cred __rcu *ptracer_cred; 945 946 /* Objective and real subjective task credentials (COW): */ 947 const struct cred __rcu *real_cred; 948 949 /* Effective (overridable) subjective task credentials (COW): */ 950 const struct cred __rcu *cred; 951 952 #ifdef CONFIG_KEYS 953 /* Cached requested key. */ 954 struct key *cached_requested_key; 955 #endif 956 957 /* 958 * executable name, excluding path. 959 * 960 * - normally initialized setup_new_exec() 961 * - access it with [gs]et_task_comm() 962 * - lock it with task_lock() 963 */ 964 char comm[TASK_COMM_LEN]; 965 966 struct nameidata *nameidata; 967 968 #ifdef CONFIG_SYSVIPC 969 struct sysv_sem sysvsem; 970 struct sysv_shm sysvshm; 971 #endif 972 #ifdef CONFIG_DETECT_HUNG_TASK 973 unsigned long last_switch_count; 974 unsigned long last_switch_time; 975 #endif 976 /* Filesystem information: */ 977 struct fs_struct *fs; 978 979 /* Open file information: */ 980 struct files_struct *files; 981 982 #ifdef CONFIG_IO_URING 983 struct io_uring_task *io_uring; 984 #endif 985 986 /* Namespaces: */ 987 struct nsproxy *nsproxy; 988 989 /* Signal handlers: */ 990 struct signal_struct *signal; 991 struct sighand_struct __rcu *sighand; 992 struct sigqueue *sigqueue_cache; 993 sigset_t blocked; 994 sigset_t real_blocked; 995 /* Restored if set_restore_sigmask() was used: */ 996 sigset_t saved_sigmask; 997 struct sigpending pending; 998 unsigned long sas_ss_sp; 999 size_t sas_ss_size; 1000 unsigned int sas_ss_flags; 1001 1002 struct callback_head *task_works; 1003 1004 #ifdef CONFIG_AUDIT 1005 #ifdef CONFIG_AUDITSYSCALL 1006 struct audit_context *audit_context; 1007 #endif 1008 kuid_t loginuid; 1009 unsigned int sessionid; 1010 #endif 1011 struct seccomp seccomp; 1012 struct syscall_user_dispatch syscall_dispatch; 1013 1014 /* Thread group tracking: */ 1015 u64 parent_exec_id; 1016 u64 self_exec_id; 1017 1018 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1019 spinlock_t alloc_lock; 1020 1021 /* Protection of the PI data structures: */ 1022 raw_spinlock_t pi_lock; 1023 1024 struct wake_q_node wake_q; 1025 1026 #ifdef CONFIG_RT_MUTEXES 1027 /* PI waiters blocked on a rt_mutex held by this task: */ 1028 struct rb_root_cached pi_waiters; 1029 /* Updated under owner's pi_lock and rq lock */ 1030 struct task_struct *pi_top_task; 1031 /* Deadlock detection and priority inheritance handling: */ 1032 struct rt_mutex_waiter *pi_blocked_on; 1033 #endif 1034 1035 #ifdef CONFIG_DEBUG_MUTEXES 1036 /* Mutex deadlock detection: */ 1037 struct mutex_waiter *blocked_on; 1038 #endif 1039 1040 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1041 int non_block_count; 1042 #endif 1043 1044 #ifdef CONFIG_TRACE_IRQFLAGS 1045 struct irqtrace_events irqtrace; 1046 unsigned int hardirq_threaded; 1047 u64 hardirq_chain_key; 1048 int softirqs_enabled; 1049 int softirq_context; 1050 int irq_config; 1051 #endif 1052 #ifdef CONFIG_PREEMPT_RT 1053 int softirq_disable_cnt; 1054 #endif 1055 1056 #ifdef CONFIG_LOCKDEP 1057 # define MAX_LOCK_DEPTH 48UL 1058 u64 curr_chain_key; 1059 int lockdep_depth; 1060 unsigned int lockdep_recursion; 1061 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1062 #endif 1063 1064 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1065 unsigned int in_ubsan; 1066 #endif 1067 1068 /* Journalling filesystem info: */ 1069 void *journal_info; 1070 1071 /* Stacked block device info: */ 1072 struct bio_list *bio_list; 1073 1074 #ifdef CONFIG_BLOCK 1075 /* Stack plugging: */ 1076 struct blk_plug *plug; 1077 #endif 1078 1079 /* VM state: */ 1080 struct reclaim_state *reclaim_state; 1081 1082 struct backing_dev_info *backing_dev_info; 1083 1084 struct io_context *io_context; 1085 1086 #ifdef CONFIG_COMPACTION 1087 struct capture_control *capture_control; 1088 #endif 1089 /* Ptrace state: */ 1090 unsigned long ptrace_message; 1091 kernel_siginfo_t *last_siginfo; 1092 1093 struct task_io_accounting ioac; 1094 #ifdef CONFIG_PSI 1095 /* Pressure stall state */ 1096 unsigned int psi_flags; 1097 #endif 1098 #ifdef CONFIG_TASK_XACCT 1099 /* Accumulated RSS usage: */ 1100 u64 acct_rss_mem1; 1101 /* Accumulated virtual memory usage: */ 1102 u64 acct_vm_mem1; 1103 /* stime + utime since last update: */ 1104 u64 acct_timexpd; 1105 #endif 1106 #ifdef CONFIG_CPUSETS 1107 /* Protected by ->alloc_lock: */ 1108 nodemask_t mems_allowed; 1109 /* Sequence number to catch updates: */ 1110 seqcount_spinlock_t mems_allowed_seq; 1111 int cpuset_mem_spread_rotor; 1112 int cpuset_slab_spread_rotor; 1113 #endif 1114 #ifdef CONFIG_CGROUPS 1115 /* Control Group info protected by css_set_lock: */ 1116 struct css_set __rcu *cgroups; 1117 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1118 struct list_head cg_list; 1119 #endif 1120 #ifdef CONFIG_X86_CPU_RESCTRL 1121 u32 closid; 1122 u32 rmid; 1123 #endif 1124 #ifdef CONFIG_FUTEX 1125 struct robust_list_head __user *robust_list; 1126 #ifdef CONFIG_COMPAT 1127 struct compat_robust_list_head __user *compat_robust_list; 1128 #endif 1129 struct list_head pi_state_list; 1130 struct futex_pi_state *pi_state_cache; 1131 struct mutex futex_exit_mutex; 1132 unsigned int futex_state; 1133 #endif 1134 #ifdef CONFIG_PERF_EVENTS 1135 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 1136 struct mutex perf_event_mutex; 1137 struct list_head perf_event_list; 1138 #endif 1139 #ifdef CONFIG_DEBUG_PREEMPT 1140 unsigned long preempt_disable_ip; 1141 #endif 1142 #ifdef CONFIG_NUMA 1143 /* Protected by alloc_lock: */ 1144 struct mempolicy *mempolicy; 1145 short il_prev; 1146 short pref_node_fork; 1147 #endif 1148 #ifdef CONFIG_NUMA_BALANCING 1149 int numa_scan_seq; 1150 unsigned int numa_scan_period; 1151 unsigned int numa_scan_period_max; 1152 int numa_preferred_nid; 1153 unsigned long numa_migrate_retry; 1154 /* Migration stamp: */ 1155 u64 node_stamp; 1156 u64 last_task_numa_placement; 1157 u64 last_sum_exec_runtime; 1158 struct callback_head numa_work; 1159 1160 /* 1161 * This pointer is only modified for current in syscall and 1162 * pagefault context (and for tasks being destroyed), so it can be read 1163 * from any of the following contexts: 1164 * - RCU read-side critical section 1165 * - current->numa_group from everywhere 1166 * - task's runqueue locked, task not running 1167 */ 1168 struct numa_group __rcu *numa_group; 1169 1170 /* 1171 * numa_faults is an array split into four regions: 1172 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1173 * in this precise order. 1174 * 1175 * faults_memory: Exponential decaying average of faults on a per-node 1176 * basis. Scheduling placement decisions are made based on these 1177 * counts. The values remain static for the duration of a PTE scan. 1178 * faults_cpu: Track the nodes the process was running on when a NUMA 1179 * hinting fault was incurred. 1180 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1181 * during the current scan window. When the scan completes, the counts 1182 * in faults_memory and faults_cpu decay and these values are copied. 1183 */ 1184 unsigned long *numa_faults; 1185 unsigned long total_numa_faults; 1186 1187 /* 1188 * numa_faults_locality tracks if faults recorded during the last 1189 * scan window were remote/local or failed to migrate. The task scan 1190 * period is adapted based on the locality of the faults with different 1191 * weights depending on whether they were shared or private faults 1192 */ 1193 unsigned long numa_faults_locality[3]; 1194 1195 unsigned long numa_pages_migrated; 1196 #endif /* CONFIG_NUMA_BALANCING */ 1197 1198 #ifdef CONFIG_RSEQ 1199 struct rseq __user *rseq; 1200 u32 rseq_sig; 1201 /* 1202 * RmW on rseq_event_mask must be performed atomically 1203 * with respect to preemption. 1204 */ 1205 unsigned long rseq_event_mask; 1206 #endif 1207 1208 struct tlbflush_unmap_batch tlb_ubc; 1209 1210 union { 1211 refcount_t rcu_users; 1212 struct rcu_head rcu; 1213 }; 1214 1215 /* Cache last used pipe for splice(): */ 1216 struct pipe_inode_info *splice_pipe; 1217 1218 struct page_frag task_frag; 1219 1220 #ifdef CONFIG_TASK_DELAY_ACCT 1221 struct task_delay_info *delays; 1222 #endif 1223 1224 #ifdef CONFIG_FAULT_INJECTION 1225 int make_it_fail; 1226 unsigned int fail_nth; 1227 #endif 1228 /* 1229 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1230 * balance_dirty_pages() for a dirty throttling pause: 1231 */ 1232 int nr_dirtied; 1233 int nr_dirtied_pause; 1234 /* Start of a write-and-pause period: */ 1235 unsigned long dirty_paused_when; 1236 1237 #ifdef CONFIG_LATENCYTOP 1238 int latency_record_count; 1239 struct latency_record latency_record[LT_SAVECOUNT]; 1240 #endif 1241 /* 1242 * Time slack values; these are used to round up poll() and 1243 * select() etc timeout values. These are in nanoseconds. 1244 */ 1245 u64 timer_slack_ns; 1246 u64 default_timer_slack_ns; 1247 1248 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1249 unsigned int kasan_depth; 1250 #endif 1251 1252 #ifdef CONFIG_KCSAN 1253 struct kcsan_ctx kcsan_ctx; 1254 #ifdef CONFIG_TRACE_IRQFLAGS 1255 struct irqtrace_events kcsan_save_irqtrace; 1256 #endif 1257 #endif 1258 1259 #if IS_ENABLED(CONFIG_KUNIT) 1260 struct kunit *kunit_test; 1261 #endif 1262 1263 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1264 /* Index of current stored address in ret_stack: */ 1265 int curr_ret_stack; 1266 int curr_ret_depth; 1267 1268 /* Stack of return addresses for return function tracing: */ 1269 struct ftrace_ret_stack *ret_stack; 1270 1271 /* Timestamp for last schedule: */ 1272 unsigned long long ftrace_timestamp; 1273 1274 /* 1275 * Number of functions that haven't been traced 1276 * because of depth overrun: 1277 */ 1278 atomic_t trace_overrun; 1279 1280 /* Pause tracing: */ 1281 atomic_t tracing_graph_pause; 1282 #endif 1283 1284 #ifdef CONFIG_TRACING 1285 /* State flags for use by tracers: */ 1286 unsigned long trace; 1287 1288 /* Bitmask and counter of trace recursion: */ 1289 unsigned long trace_recursion; 1290 #endif /* CONFIG_TRACING */ 1291 1292 #ifdef CONFIG_KCOV 1293 /* See kernel/kcov.c for more details. */ 1294 1295 /* Coverage collection mode enabled for this task (0 if disabled): */ 1296 unsigned int kcov_mode; 1297 1298 /* Size of the kcov_area: */ 1299 unsigned int kcov_size; 1300 1301 /* Buffer for coverage collection: */ 1302 void *kcov_area; 1303 1304 /* KCOV descriptor wired with this task or NULL: */ 1305 struct kcov *kcov; 1306 1307 /* KCOV common handle for remote coverage collection: */ 1308 u64 kcov_handle; 1309 1310 /* KCOV sequence number: */ 1311 int kcov_sequence; 1312 1313 /* Collect coverage from softirq context: */ 1314 unsigned int kcov_softirq; 1315 #endif 1316 1317 #ifdef CONFIG_MEMCG 1318 struct mem_cgroup *memcg_in_oom; 1319 gfp_t memcg_oom_gfp_mask; 1320 int memcg_oom_order; 1321 1322 /* Number of pages to reclaim on returning to userland: */ 1323 unsigned int memcg_nr_pages_over_high; 1324 1325 /* Used by memcontrol for targeted memcg charge: */ 1326 struct mem_cgroup *active_memcg; 1327 #endif 1328 1329 #ifdef CONFIG_BLK_CGROUP 1330 struct request_queue *throttle_queue; 1331 #endif 1332 1333 #ifdef CONFIG_UPROBES 1334 struct uprobe_task *utask; 1335 #endif 1336 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1337 unsigned int sequential_io; 1338 unsigned int sequential_io_avg; 1339 #endif 1340 struct kmap_ctrl kmap_ctrl; 1341 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1342 unsigned long task_state_change; 1343 #endif 1344 int pagefault_disabled; 1345 #ifdef CONFIG_MMU 1346 struct task_struct *oom_reaper_list; 1347 #endif 1348 #ifdef CONFIG_VMAP_STACK 1349 struct vm_struct *stack_vm_area; 1350 #endif 1351 #ifdef CONFIG_THREAD_INFO_IN_TASK 1352 /* A live task holds one reference: */ 1353 refcount_t stack_refcount; 1354 #endif 1355 #ifdef CONFIG_LIVEPATCH 1356 int patch_state; 1357 #endif 1358 #ifdef CONFIG_SECURITY 1359 /* Used by LSM modules for access restriction: */ 1360 void *security; 1361 #endif 1362 #ifdef CONFIG_BPF_SYSCALL 1363 /* Used by BPF task local storage */ 1364 struct bpf_local_storage __rcu *bpf_storage; 1365 #endif 1366 1367 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1368 unsigned long lowest_stack; 1369 unsigned long prev_lowest_stack; 1370 #endif 1371 1372 #ifdef CONFIG_X86_MCE 1373 void __user *mce_vaddr; 1374 __u64 mce_kflags; 1375 u64 mce_addr; 1376 __u64 mce_ripv : 1, 1377 mce_whole_page : 1, 1378 __mce_reserved : 62; 1379 struct callback_head mce_kill_me; 1380 #endif 1381 1382 #ifdef CONFIG_KRETPROBES 1383 struct llist_head kretprobe_instances; 1384 #endif 1385 1386 /* 1387 * New fields for task_struct should be added above here, so that 1388 * they are included in the randomized portion of task_struct. 1389 */ 1390 randomized_struct_fields_end 1391 1392 /* CPU-specific state of this task: */ 1393 struct thread_struct thread; 1394 1395 /* 1396 * WARNING: on x86, 'thread_struct' contains a variable-sized 1397 * structure. It *MUST* be at the end of 'task_struct'. 1398 * 1399 * Do not put anything below here! 1400 */ 1401 }; 1402 1403 static inline struct pid *task_pid(struct task_struct *task) 1404 { 1405 return task->thread_pid; 1406 } 1407 1408 /* 1409 * the helpers to get the task's different pids as they are seen 1410 * from various namespaces 1411 * 1412 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1413 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1414 * current. 1415 * task_xid_nr_ns() : id seen from the ns specified; 1416 * 1417 * see also pid_nr() etc in include/linux/pid.h 1418 */ 1419 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1420 1421 static inline pid_t task_pid_nr(struct task_struct *tsk) 1422 { 1423 return tsk->pid; 1424 } 1425 1426 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1427 { 1428 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1429 } 1430 1431 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1432 { 1433 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1434 } 1435 1436 1437 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1438 { 1439 return tsk->tgid; 1440 } 1441 1442 /** 1443 * pid_alive - check that a task structure is not stale 1444 * @p: Task structure to be checked. 1445 * 1446 * Test if a process is not yet dead (at most zombie state) 1447 * If pid_alive fails, then pointers within the task structure 1448 * can be stale and must not be dereferenced. 1449 * 1450 * Return: 1 if the process is alive. 0 otherwise. 1451 */ 1452 static inline int pid_alive(const struct task_struct *p) 1453 { 1454 return p->thread_pid != NULL; 1455 } 1456 1457 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1458 { 1459 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1460 } 1461 1462 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1463 { 1464 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1465 } 1466 1467 1468 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1469 { 1470 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1471 } 1472 1473 static inline pid_t task_session_vnr(struct task_struct *tsk) 1474 { 1475 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1476 } 1477 1478 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1479 { 1480 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns); 1481 } 1482 1483 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1484 { 1485 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL); 1486 } 1487 1488 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1489 { 1490 pid_t pid = 0; 1491 1492 rcu_read_lock(); 1493 if (pid_alive(tsk)) 1494 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1495 rcu_read_unlock(); 1496 1497 return pid; 1498 } 1499 1500 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1501 { 1502 return task_ppid_nr_ns(tsk, &init_pid_ns); 1503 } 1504 1505 /* Obsolete, do not use: */ 1506 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1507 { 1508 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1509 } 1510 1511 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1512 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1513 1514 static inline unsigned int task_state_index(struct task_struct *tsk) 1515 { 1516 unsigned int tsk_state = READ_ONCE(tsk->state); 1517 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT; 1518 1519 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1520 1521 if (tsk_state == TASK_IDLE) 1522 state = TASK_REPORT_IDLE; 1523 1524 return fls(state); 1525 } 1526 1527 static inline char task_index_to_char(unsigned int state) 1528 { 1529 static const char state_char[] = "RSDTtXZPI"; 1530 1531 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1532 1533 return state_char[state]; 1534 } 1535 1536 static inline char task_state_to_char(struct task_struct *tsk) 1537 { 1538 return task_index_to_char(task_state_index(tsk)); 1539 } 1540 1541 /** 1542 * is_global_init - check if a task structure is init. Since init 1543 * is free to have sub-threads we need to check tgid. 1544 * @tsk: Task structure to be checked. 1545 * 1546 * Check if a task structure is the first user space task the kernel created. 1547 * 1548 * Return: 1 if the task structure is init. 0 otherwise. 1549 */ 1550 static inline int is_global_init(struct task_struct *tsk) 1551 { 1552 return task_tgid_nr(tsk) == 1; 1553 } 1554 1555 extern struct pid *cad_pid; 1556 1557 /* 1558 * Per process flags 1559 */ 1560 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1561 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1562 #define PF_EXITING 0x00000004 /* Getting shut down */ 1563 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1564 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1565 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1566 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1567 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1568 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1569 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1570 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1571 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1572 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1573 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1574 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1575 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1576 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1577 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1578 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1579 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1580 * I am cleaning dirty pages from some other bdi. */ 1581 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1582 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1583 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1584 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1585 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1586 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */ 1587 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1588 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1589 1590 /* 1591 * Only the _current_ task can read/write to tsk->flags, but other 1592 * tasks can access tsk->flags in readonly mode for example 1593 * with tsk_used_math (like during threaded core dumping). 1594 * There is however an exception to this rule during ptrace 1595 * or during fork: the ptracer task is allowed to write to the 1596 * child->flags of its traced child (same goes for fork, the parent 1597 * can write to the child->flags), because we're guaranteed the 1598 * child is not running and in turn not changing child->flags 1599 * at the same time the parent does it. 1600 */ 1601 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1602 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1603 #define clear_used_math() clear_stopped_child_used_math(current) 1604 #define set_used_math() set_stopped_child_used_math(current) 1605 1606 #define conditional_stopped_child_used_math(condition, child) \ 1607 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1608 1609 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1610 1611 #define copy_to_stopped_child_used_math(child) \ 1612 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1613 1614 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1615 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1616 #define used_math() tsk_used_math(current) 1617 1618 static inline bool is_percpu_thread(void) 1619 { 1620 #ifdef CONFIG_SMP 1621 return (current->flags & PF_NO_SETAFFINITY) && 1622 (current->nr_cpus_allowed == 1); 1623 #else 1624 return true; 1625 #endif 1626 } 1627 1628 /* Per-process atomic flags. */ 1629 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1630 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1631 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1632 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1633 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1634 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1635 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1636 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1637 1638 #define TASK_PFA_TEST(name, func) \ 1639 static inline bool task_##func(struct task_struct *p) \ 1640 { return test_bit(PFA_##name, &p->atomic_flags); } 1641 1642 #define TASK_PFA_SET(name, func) \ 1643 static inline void task_set_##func(struct task_struct *p) \ 1644 { set_bit(PFA_##name, &p->atomic_flags); } 1645 1646 #define TASK_PFA_CLEAR(name, func) \ 1647 static inline void task_clear_##func(struct task_struct *p) \ 1648 { clear_bit(PFA_##name, &p->atomic_flags); } 1649 1650 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1651 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1652 1653 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1654 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1655 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1656 1657 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1658 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1659 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1660 1661 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1662 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1663 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1664 1665 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1666 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1667 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1668 1669 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1670 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1671 1672 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1673 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1674 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1675 1676 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1677 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1678 1679 static inline void 1680 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1681 { 1682 current->flags &= ~flags; 1683 current->flags |= orig_flags & flags; 1684 } 1685 1686 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1687 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1688 #ifdef CONFIG_SMP 1689 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1690 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1691 #else 1692 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1693 { 1694 } 1695 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1696 { 1697 if (!cpumask_test_cpu(0, new_mask)) 1698 return -EINVAL; 1699 return 0; 1700 } 1701 #endif 1702 1703 extern int yield_to(struct task_struct *p, bool preempt); 1704 extern void set_user_nice(struct task_struct *p, long nice); 1705 extern int task_prio(const struct task_struct *p); 1706 1707 /** 1708 * task_nice - return the nice value of a given task. 1709 * @p: the task in question. 1710 * 1711 * Return: The nice value [ -20 ... 0 ... 19 ]. 1712 */ 1713 static inline int task_nice(const struct task_struct *p) 1714 { 1715 return PRIO_TO_NICE((p)->static_prio); 1716 } 1717 1718 extern int can_nice(const struct task_struct *p, const int nice); 1719 extern int task_curr(const struct task_struct *p); 1720 extern int idle_cpu(int cpu); 1721 extern int available_idle_cpu(int cpu); 1722 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1723 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1724 extern void sched_set_fifo(struct task_struct *p); 1725 extern void sched_set_fifo_low(struct task_struct *p); 1726 extern void sched_set_normal(struct task_struct *p, int nice); 1727 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1728 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1729 extern struct task_struct *idle_task(int cpu); 1730 1731 /** 1732 * is_idle_task - is the specified task an idle task? 1733 * @p: the task in question. 1734 * 1735 * Return: 1 if @p is an idle task. 0 otherwise. 1736 */ 1737 static __always_inline bool is_idle_task(const struct task_struct *p) 1738 { 1739 return !!(p->flags & PF_IDLE); 1740 } 1741 1742 extern struct task_struct *curr_task(int cpu); 1743 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1744 1745 void yield(void); 1746 1747 union thread_union { 1748 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1749 struct task_struct task; 1750 #endif 1751 #ifndef CONFIG_THREAD_INFO_IN_TASK 1752 struct thread_info thread_info; 1753 #endif 1754 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1755 }; 1756 1757 #ifndef CONFIG_THREAD_INFO_IN_TASK 1758 extern struct thread_info init_thread_info; 1759 #endif 1760 1761 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1762 1763 #ifdef CONFIG_THREAD_INFO_IN_TASK 1764 static inline struct thread_info *task_thread_info(struct task_struct *task) 1765 { 1766 return &task->thread_info; 1767 } 1768 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1769 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1770 #endif 1771 1772 /* 1773 * find a task by one of its numerical ids 1774 * 1775 * find_task_by_pid_ns(): 1776 * finds a task by its pid in the specified namespace 1777 * find_task_by_vpid(): 1778 * finds a task by its virtual pid 1779 * 1780 * see also find_vpid() etc in include/linux/pid.h 1781 */ 1782 1783 extern struct task_struct *find_task_by_vpid(pid_t nr); 1784 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1785 1786 /* 1787 * find a task by its virtual pid and get the task struct 1788 */ 1789 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1790 1791 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1792 extern int wake_up_process(struct task_struct *tsk); 1793 extern void wake_up_new_task(struct task_struct *tsk); 1794 1795 #ifdef CONFIG_SMP 1796 extern void kick_process(struct task_struct *tsk); 1797 #else 1798 static inline void kick_process(struct task_struct *tsk) { } 1799 #endif 1800 1801 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1802 1803 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1804 { 1805 __set_task_comm(tsk, from, false); 1806 } 1807 1808 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1809 #define get_task_comm(buf, tsk) ({ \ 1810 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1811 __get_task_comm(buf, sizeof(buf), tsk); \ 1812 }) 1813 1814 #ifdef CONFIG_SMP 1815 static __always_inline void scheduler_ipi(void) 1816 { 1817 /* 1818 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1819 * TIF_NEED_RESCHED remotely (for the first time) will also send 1820 * this IPI. 1821 */ 1822 preempt_fold_need_resched(); 1823 } 1824 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 1825 #else 1826 static inline void scheduler_ipi(void) { } 1827 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) 1828 { 1829 return 1; 1830 } 1831 #endif 1832 1833 /* 1834 * Set thread flags in other task's structures. 1835 * See asm/thread_info.h for TIF_xxxx flags available: 1836 */ 1837 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1838 { 1839 set_ti_thread_flag(task_thread_info(tsk), flag); 1840 } 1841 1842 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1843 { 1844 clear_ti_thread_flag(task_thread_info(tsk), flag); 1845 } 1846 1847 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1848 bool value) 1849 { 1850 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1851 } 1852 1853 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1854 { 1855 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1856 } 1857 1858 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1859 { 1860 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1861 } 1862 1863 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1864 { 1865 return test_ti_thread_flag(task_thread_info(tsk), flag); 1866 } 1867 1868 static inline void set_tsk_need_resched(struct task_struct *tsk) 1869 { 1870 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1871 } 1872 1873 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1874 { 1875 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1876 } 1877 1878 static inline int test_tsk_need_resched(struct task_struct *tsk) 1879 { 1880 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1881 } 1882 1883 /* 1884 * cond_resched() and cond_resched_lock(): latency reduction via 1885 * explicit rescheduling in places that are safe. The return 1886 * value indicates whether a reschedule was done in fact. 1887 * cond_resched_lock() will drop the spinlock before scheduling, 1888 */ 1889 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 1890 extern int __cond_resched(void); 1891 1892 #ifdef CONFIG_PREEMPT_DYNAMIC 1893 1894 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 1895 1896 static __always_inline int _cond_resched(void) 1897 { 1898 return static_call_mod(cond_resched)(); 1899 } 1900 1901 #else 1902 1903 static inline int _cond_resched(void) 1904 { 1905 return __cond_resched(); 1906 } 1907 1908 #endif /* CONFIG_PREEMPT_DYNAMIC */ 1909 1910 #else 1911 1912 static inline int _cond_resched(void) { return 0; } 1913 1914 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */ 1915 1916 #define cond_resched() ({ \ 1917 ___might_sleep(__FILE__, __LINE__, 0); \ 1918 _cond_resched(); \ 1919 }) 1920 1921 extern int __cond_resched_lock(spinlock_t *lock); 1922 extern int __cond_resched_rwlock_read(rwlock_t *lock); 1923 extern int __cond_resched_rwlock_write(rwlock_t *lock); 1924 1925 #define cond_resched_lock(lock) ({ \ 1926 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ 1927 __cond_resched_lock(lock); \ 1928 }) 1929 1930 #define cond_resched_rwlock_read(lock) ({ \ 1931 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \ 1932 __cond_resched_rwlock_read(lock); \ 1933 }) 1934 1935 #define cond_resched_rwlock_write(lock) ({ \ 1936 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \ 1937 __cond_resched_rwlock_write(lock); \ 1938 }) 1939 1940 static inline void cond_resched_rcu(void) 1941 { 1942 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 1943 rcu_read_unlock(); 1944 cond_resched(); 1945 rcu_read_lock(); 1946 #endif 1947 } 1948 1949 /* 1950 * Does a critical section need to be broken due to another 1951 * task waiting?: (technically does not depend on CONFIG_PREEMPTION, 1952 * but a general need for low latency) 1953 */ 1954 static inline int spin_needbreak(spinlock_t *lock) 1955 { 1956 #ifdef CONFIG_PREEMPTION 1957 return spin_is_contended(lock); 1958 #else 1959 return 0; 1960 #endif 1961 } 1962 1963 /* 1964 * Check if a rwlock is contended. 1965 * Returns non-zero if there is another task waiting on the rwlock. 1966 * Returns zero if the lock is not contended or the system / underlying 1967 * rwlock implementation does not support contention detection. 1968 * Technically does not depend on CONFIG_PREEMPTION, but a general need 1969 * for low latency. 1970 */ 1971 static inline int rwlock_needbreak(rwlock_t *lock) 1972 { 1973 #ifdef CONFIG_PREEMPTION 1974 return rwlock_is_contended(lock); 1975 #else 1976 return 0; 1977 #endif 1978 } 1979 1980 static __always_inline bool need_resched(void) 1981 { 1982 return unlikely(tif_need_resched()); 1983 } 1984 1985 /* 1986 * Wrappers for p->thread_info->cpu access. No-op on UP. 1987 */ 1988 #ifdef CONFIG_SMP 1989 1990 static inline unsigned int task_cpu(const struct task_struct *p) 1991 { 1992 #ifdef CONFIG_THREAD_INFO_IN_TASK 1993 return READ_ONCE(p->cpu); 1994 #else 1995 return READ_ONCE(task_thread_info(p)->cpu); 1996 #endif 1997 } 1998 1999 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2000 2001 #else 2002 2003 static inline unsigned int task_cpu(const struct task_struct *p) 2004 { 2005 return 0; 2006 } 2007 2008 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2009 { 2010 } 2011 2012 #endif /* CONFIG_SMP */ 2013 2014 /* 2015 * In order to reduce various lock holder preemption latencies provide an 2016 * interface to see if a vCPU is currently running or not. 2017 * 2018 * This allows us to terminate optimistic spin loops and block, analogous to 2019 * the native optimistic spin heuristic of testing if the lock owner task is 2020 * running or not. 2021 */ 2022 #ifndef vcpu_is_preempted 2023 static inline bool vcpu_is_preempted(int cpu) 2024 { 2025 return false; 2026 } 2027 #endif 2028 2029 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2030 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2031 2032 #ifndef TASK_SIZE_OF 2033 #define TASK_SIZE_OF(tsk) TASK_SIZE 2034 #endif 2035 2036 #ifdef CONFIG_SMP 2037 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2038 unsigned long sched_cpu_util(int cpu, unsigned long max); 2039 #endif /* CONFIG_SMP */ 2040 2041 #ifdef CONFIG_RSEQ 2042 2043 /* 2044 * Map the event mask on the user-space ABI enum rseq_cs_flags 2045 * for direct mask checks. 2046 */ 2047 enum rseq_event_mask_bits { 2048 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT, 2049 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT, 2050 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT, 2051 }; 2052 2053 enum rseq_event_mask { 2054 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT), 2055 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT), 2056 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT), 2057 }; 2058 2059 static inline void rseq_set_notify_resume(struct task_struct *t) 2060 { 2061 if (t->rseq) 2062 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME); 2063 } 2064 2065 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs); 2066 2067 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2068 struct pt_regs *regs) 2069 { 2070 if (current->rseq) 2071 __rseq_handle_notify_resume(ksig, regs); 2072 } 2073 2074 static inline void rseq_signal_deliver(struct ksignal *ksig, 2075 struct pt_regs *regs) 2076 { 2077 preempt_disable(); 2078 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask); 2079 preempt_enable(); 2080 rseq_handle_notify_resume(ksig, regs); 2081 } 2082 2083 /* rseq_preempt() requires preemption to be disabled. */ 2084 static inline void rseq_preempt(struct task_struct *t) 2085 { 2086 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask); 2087 rseq_set_notify_resume(t); 2088 } 2089 2090 /* rseq_migrate() requires preemption to be disabled. */ 2091 static inline void rseq_migrate(struct task_struct *t) 2092 { 2093 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask); 2094 rseq_set_notify_resume(t); 2095 } 2096 2097 /* 2098 * If parent process has a registered restartable sequences area, the 2099 * child inherits. Unregister rseq for a clone with CLONE_VM set. 2100 */ 2101 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2102 { 2103 if (clone_flags & CLONE_VM) { 2104 t->rseq = NULL; 2105 t->rseq_sig = 0; 2106 t->rseq_event_mask = 0; 2107 } else { 2108 t->rseq = current->rseq; 2109 t->rseq_sig = current->rseq_sig; 2110 t->rseq_event_mask = current->rseq_event_mask; 2111 } 2112 } 2113 2114 static inline void rseq_execve(struct task_struct *t) 2115 { 2116 t->rseq = NULL; 2117 t->rseq_sig = 0; 2118 t->rseq_event_mask = 0; 2119 } 2120 2121 #else 2122 2123 static inline void rseq_set_notify_resume(struct task_struct *t) 2124 { 2125 } 2126 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2127 struct pt_regs *regs) 2128 { 2129 } 2130 static inline void rseq_signal_deliver(struct ksignal *ksig, 2131 struct pt_regs *regs) 2132 { 2133 } 2134 static inline void rseq_preempt(struct task_struct *t) 2135 { 2136 } 2137 static inline void rseq_migrate(struct task_struct *t) 2138 { 2139 } 2140 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2141 { 2142 } 2143 static inline void rseq_execve(struct task_struct *t) 2144 { 2145 } 2146 2147 #endif 2148 2149 #ifdef CONFIG_DEBUG_RSEQ 2150 2151 void rseq_syscall(struct pt_regs *regs); 2152 2153 #else 2154 2155 static inline void rseq_syscall(struct pt_regs *regs) 2156 { 2157 } 2158 2159 #endif 2160 2161 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq); 2162 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len); 2163 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq); 2164 2165 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq); 2166 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq); 2167 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq); 2168 2169 int sched_trace_rq_cpu(struct rq *rq); 2170 int sched_trace_rq_cpu_capacity(struct rq *rq); 2171 int sched_trace_rq_nr_running(struct rq *rq); 2172 2173 const struct cpumask *sched_trace_rd_span(struct root_domain *rd); 2174 2175 #endif 2176