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 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 355 * updates. When a task is dequeued, its util_est should not be updated if its 356 * util_avg has not been updated in the meantime. 357 * This information is mapped into the MSB bit of util_est.enqueued at dequeue 358 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg 359 * for a task) it is safe to use MSB. 360 */ 361 struct util_est { 362 unsigned int enqueued; 363 unsigned int ewma; 364 #define UTIL_EST_WEIGHT_SHIFT 2 365 #define UTIL_AVG_UNCHANGED 0x80000000 366 } __attribute__((__aligned__(sizeof(u64)))); 367 368 /* 369 * The load/runnable/util_avg accumulates an infinite geometric series 370 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 371 * 372 * [load_avg definition] 373 * 374 * load_avg = runnable% * scale_load_down(load) 375 * 376 * [runnable_avg definition] 377 * 378 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 379 * 380 * [util_avg definition] 381 * 382 * util_avg = running% * SCHED_CAPACITY_SCALE 383 * 384 * where runnable% is the time ratio that a sched_entity is runnable and 385 * running% the time ratio that a sched_entity is running. 386 * 387 * For cfs_rq, they are the aggregated values of all runnable and blocked 388 * sched_entities. 389 * 390 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 391 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 392 * for computing those signals (see update_rq_clock_pelt()) 393 * 394 * N.B., the above ratios (runnable% and running%) themselves are in the 395 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 396 * to as large a range as necessary. This is for example reflected by 397 * util_avg's SCHED_CAPACITY_SCALE. 398 * 399 * [Overflow issue] 400 * 401 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 402 * with the highest load (=88761), always runnable on a single cfs_rq, 403 * and should not overflow as the number already hits PID_MAX_LIMIT. 404 * 405 * For all other cases (including 32-bit kernels), struct load_weight's 406 * weight will overflow first before we do, because: 407 * 408 * Max(load_avg) <= Max(load.weight) 409 * 410 * Then it is the load_weight's responsibility to consider overflow 411 * issues. 412 */ 413 struct sched_avg { 414 u64 last_update_time; 415 u64 load_sum; 416 u64 runnable_sum; 417 u32 util_sum; 418 u32 period_contrib; 419 unsigned long load_avg; 420 unsigned long runnable_avg; 421 unsigned long util_avg; 422 struct util_est util_est; 423 } ____cacheline_aligned; 424 425 struct sched_statistics { 426 #ifdef CONFIG_SCHEDSTATS 427 u64 wait_start; 428 u64 wait_max; 429 u64 wait_count; 430 u64 wait_sum; 431 u64 iowait_count; 432 u64 iowait_sum; 433 434 u64 sleep_start; 435 u64 sleep_max; 436 s64 sum_sleep_runtime; 437 438 u64 block_start; 439 u64 block_max; 440 u64 exec_max; 441 u64 slice_max; 442 443 u64 nr_migrations_cold; 444 u64 nr_failed_migrations_affine; 445 u64 nr_failed_migrations_running; 446 u64 nr_failed_migrations_hot; 447 u64 nr_forced_migrations; 448 449 u64 nr_wakeups; 450 u64 nr_wakeups_sync; 451 u64 nr_wakeups_migrate; 452 u64 nr_wakeups_local; 453 u64 nr_wakeups_remote; 454 u64 nr_wakeups_affine; 455 u64 nr_wakeups_affine_attempts; 456 u64 nr_wakeups_passive; 457 u64 nr_wakeups_idle; 458 #endif 459 }; 460 461 struct sched_entity { 462 /* For load-balancing: */ 463 struct load_weight load; 464 struct rb_node run_node; 465 struct list_head group_node; 466 unsigned int on_rq; 467 468 u64 exec_start; 469 u64 sum_exec_runtime; 470 u64 vruntime; 471 u64 prev_sum_exec_runtime; 472 473 u64 nr_migrations; 474 475 struct sched_statistics statistics; 476 477 #ifdef CONFIG_FAIR_GROUP_SCHED 478 int depth; 479 struct sched_entity *parent; 480 /* rq on which this entity is (to be) queued: */ 481 struct cfs_rq *cfs_rq; 482 /* rq "owned" by this entity/group: */ 483 struct cfs_rq *my_q; 484 /* cached value of my_q->h_nr_running */ 485 unsigned long runnable_weight; 486 #endif 487 488 #ifdef CONFIG_SMP 489 /* 490 * Per entity load average tracking. 491 * 492 * Put into separate cache line so it does not 493 * collide with read-mostly values above. 494 */ 495 struct sched_avg avg; 496 #endif 497 }; 498 499 struct sched_rt_entity { 500 struct list_head run_list; 501 unsigned long timeout; 502 unsigned long watchdog_stamp; 503 unsigned int time_slice; 504 unsigned short on_rq; 505 unsigned short on_list; 506 507 struct sched_rt_entity *back; 508 #ifdef CONFIG_RT_GROUP_SCHED 509 struct sched_rt_entity *parent; 510 /* rq on which this entity is (to be) queued: */ 511 struct rt_rq *rt_rq; 512 /* rq "owned" by this entity/group: */ 513 struct rt_rq *my_q; 514 #endif 515 } __randomize_layout; 516 517 struct sched_dl_entity { 518 struct rb_node rb_node; 519 520 /* 521 * Original scheduling parameters. Copied here from sched_attr 522 * during sched_setattr(), they will remain the same until 523 * the next sched_setattr(). 524 */ 525 u64 dl_runtime; /* Maximum runtime for each instance */ 526 u64 dl_deadline; /* Relative deadline of each instance */ 527 u64 dl_period; /* Separation of two instances (period) */ 528 u64 dl_bw; /* dl_runtime / dl_period */ 529 u64 dl_density; /* dl_runtime / dl_deadline */ 530 531 /* 532 * Actual scheduling parameters. Initialized with the values above, 533 * they are continuously updated during task execution. Note that 534 * the remaining runtime could be < 0 in case we are in overrun. 535 */ 536 s64 runtime; /* Remaining runtime for this instance */ 537 u64 deadline; /* Absolute deadline for this instance */ 538 unsigned int flags; /* Specifying the scheduler behaviour */ 539 540 /* 541 * Some bool flags: 542 * 543 * @dl_throttled tells if we exhausted the runtime. If so, the 544 * task has to wait for a replenishment to be performed at the 545 * next firing of dl_timer. 546 * 547 * @dl_boosted tells if we are boosted due to DI. If so we are 548 * outside bandwidth enforcement mechanism (but only until we 549 * exit the critical section); 550 * 551 * @dl_yielded tells if task gave up the CPU before consuming 552 * all its available runtime during the last job. 553 * 554 * @dl_non_contending tells if the task is inactive while still 555 * contributing to the active utilization. In other words, it 556 * indicates if the inactive timer has been armed and its handler 557 * has not been executed yet. This flag is useful to avoid race 558 * conditions between the inactive timer handler and the wakeup 559 * code. 560 * 561 * @dl_overrun tells if the task asked to be informed about runtime 562 * overruns. 563 */ 564 unsigned int dl_throttled : 1; 565 unsigned int dl_yielded : 1; 566 unsigned int dl_non_contending : 1; 567 unsigned int dl_overrun : 1; 568 569 /* 570 * Bandwidth enforcement timer. Each -deadline task has its 571 * own bandwidth to be enforced, thus we need one timer per task. 572 */ 573 struct hrtimer dl_timer; 574 575 /* 576 * Inactive timer, responsible for decreasing the active utilization 577 * at the "0-lag time". When a -deadline task blocks, it contributes 578 * to GRUB's active utilization until the "0-lag time", hence a 579 * timer is needed to decrease the active utilization at the correct 580 * time. 581 */ 582 struct hrtimer inactive_timer; 583 584 #ifdef CONFIG_RT_MUTEXES 585 /* 586 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 587 * pi_se points to the donor, otherwise points to the dl_se it belongs 588 * to (the original one/itself). 589 */ 590 struct sched_dl_entity *pi_se; 591 #endif 592 }; 593 594 #ifdef CONFIG_UCLAMP_TASK 595 /* Number of utilization clamp buckets (shorter alias) */ 596 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 597 598 /* 599 * Utilization clamp for a scheduling entity 600 * @value: clamp value "assigned" to a se 601 * @bucket_id: bucket index corresponding to the "assigned" value 602 * @active: the se is currently refcounted in a rq's bucket 603 * @user_defined: the requested clamp value comes from user-space 604 * 605 * The bucket_id is the index of the clamp bucket matching the clamp value 606 * which is pre-computed and stored to avoid expensive integer divisions from 607 * the fast path. 608 * 609 * The active bit is set whenever a task has got an "effective" value assigned, 610 * which can be different from the clamp value "requested" from user-space. 611 * This allows to know a task is refcounted in the rq's bucket corresponding 612 * to the "effective" bucket_id. 613 * 614 * The user_defined bit is set whenever a task has got a task-specific clamp 615 * value requested from userspace, i.e. the system defaults apply to this task 616 * just as a restriction. This allows to relax default clamps when a less 617 * restrictive task-specific value has been requested, thus allowing to 618 * implement a "nice" semantic. For example, a task running with a 20% 619 * default boost can still drop its own boosting to 0%. 620 */ 621 struct uclamp_se { 622 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 623 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 624 unsigned int active : 1; 625 unsigned int user_defined : 1; 626 }; 627 #endif /* CONFIG_UCLAMP_TASK */ 628 629 union rcu_special { 630 struct { 631 u8 blocked; 632 u8 need_qs; 633 u8 exp_hint; /* Hint for performance. */ 634 u8 need_mb; /* Readers need smp_mb(). */ 635 } b; /* Bits. */ 636 u32 s; /* Set of bits. */ 637 }; 638 639 enum perf_event_task_context { 640 perf_invalid_context = -1, 641 perf_hw_context = 0, 642 perf_sw_context, 643 perf_nr_task_contexts, 644 }; 645 646 struct wake_q_node { 647 struct wake_q_node *next; 648 }; 649 650 struct kmap_ctrl { 651 #ifdef CONFIG_KMAP_LOCAL 652 int idx; 653 pte_t pteval[KM_MAX_IDX]; 654 #endif 655 }; 656 657 struct task_struct { 658 #ifdef CONFIG_THREAD_INFO_IN_TASK 659 /* 660 * For reasons of header soup (see current_thread_info()), this 661 * must be the first element of task_struct. 662 */ 663 struct thread_info thread_info; 664 #endif 665 /* -1 unrunnable, 0 runnable, >0 stopped: */ 666 volatile long state; 667 668 /* 669 * This begins the randomizable portion of task_struct. Only 670 * scheduling-critical items should be added above here. 671 */ 672 randomized_struct_fields_start 673 674 void *stack; 675 refcount_t usage; 676 /* Per task flags (PF_*), defined further below: */ 677 unsigned int flags; 678 unsigned int ptrace; 679 680 #ifdef CONFIG_SMP 681 int on_cpu; 682 struct __call_single_node wake_entry; 683 #ifdef CONFIG_THREAD_INFO_IN_TASK 684 /* Current CPU: */ 685 unsigned int cpu; 686 #endif 687 unsigned int wakee_flips; 688 unsigned long wakee_flip_decay_ts; 689 struct task_struct *last_wakee; 690 691 /* 692 * recent_used_cpu is initially set as the last CPU used by a task 693 * that wakes affine another task. Waker/wakee relationships can 694 * push tasks around a CPU where each wakeup moves to the next one. 695 * Tracking a recently used CPU allows a quick search for a recently 696 * used CPU that may be idle. 697 */ 698 int recent_used_cpu; 699 int wake_cpu; 700 #endif 701 int on_rq; 702 703 int prio; 704 int static_prio; 705 int normal_prio; 706 unsigned int rt_priority; 707 708 const struct sched_class *sched_class; 709 struct sched_entity se; 710 struct sched_rt_entity rt; 711 #ifdef CONFIG_CGROUP_SCHED 712 struct task_group *sched_task_group; 713 #endif 714 struct sched_dl_entity dl; 715 716 #ifdef CONFIG_UCLAMP_TASK 717 /* 718 * Clamp values requested for a scheduling entity. 719 * Must be updated with task_rq_lock() held. 720 */ 721 struct uclamp_se uclamp_req[UCLAMP_CNT]; 722 /* 723 * Effective clamp values used for a scheduling entity. 724 * Must be updated with task_rq_lock() held. 725 */ 726 struct uclamp_se uclamp[UCLAMP_CNT]; 727 #endif 728 729 #ifdef CONFIG_PREEMPT_NOTIFIERS 730 /* List of struct preempt_notifier: */ 731 struct hlist_head preempt_notifiers; 732 #endif 733 734 #ifdef CONFIG_BLK_DEV_IO_TRACE 735 unsigned int btrace_seq; 736 #endif 737 738 unsigned int policy; 739 int nr_cpus_allowed; 740 const cpumask_t *cpus_ptr; 741 cpumask_t cpus_mask; 742 void *migration_pending; 743 #ifdef CONFIG_SMP 744 unsigned short migration_disabled; 745 #endif 746 unsigned short migration_flags; 747 748 #ifdef CONFIG_PREEMPT_RCU 749 int rcu_read_lock_nesting; 750 union rcu_special rcu_read_unlock_special; 751 struct list_head rcu_node_entry; 752 struct rcu_node *rcu_blocked_node; 753 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 754 755 #ifdef CONFIG_TASKS_RCU 756 unsigned long rcu_tasks_nvcsw; 757 u8 rcu_tasks_holdout; 758 u8 rcu_tasks_idx; 759 int rcu_tasks_idle_cpu; 760 struct list_head rcu_tasks_holdout_list; 761 #endif /* #ifdef CONFIG_TASKS_RCU */ 762 763 #ifdef CONFIG_TASKS_TRACE_RCU 764 int trc_reader_nesting; 765 int trc_ipi_to_cpu; 766 union rcu_special trc_reader_special; 767 bool trc_reader_checked; 768 struct list_head trc_holdout_list; 769 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 770 771 struct sched_info sched_info; 772 773 struct list_head tasks; 774 #ifdef CONFIG_SMP 775 struct plist_node pushable_tasks; 776 struct rb_node pushable_dl_tasks; 777 #endif 778 779 struct mm_struct *mm; 780 struct mm_struct *active_mm; 781 782 /* Per-thread vma caching: */ 783 struct vmacache vmacache; 784 785 #ifdef SPLIT_RSS_COUNTING 786 struct task_rss_stat rss_stat; 787 #endif 788 int exit_state; 789 int exit_code; 790 int exit_signal; 791 /* The signal sent when the parent dies: */ 792 int pdeath_signal; 793 /* JOBCTL_*, siglock protected: */ 794 unsigned long jobctl; 795 796 /* Used for emulating ABI behavior of previous Linux versions: */ 797 unsigned int personality; 798 799 /* Scheduler bits, serialized by scheduler locks: */ 800 unsigned sched_reset_on_fork:1; 801 unsigned sched_contributes_to_load:1; 802 unsigned sched_migrated:1; 803 #ifdef CONFIG_PSI 804 unsigned sched_psi_wake_requeue:1; 805 #endif 806 807 /* Force alignment to the next boundary: */ 808 unsigned :0; 809 810 /* Unserialized, strictly 'current' */ 811 812 /* 813 * This field must not be in the scheduler word above due to wakelist 814 * queueing no longer being serialized by p->on_cpu. However: 815 * 816 * p->XXX = X; ttwu() 817 * schedule() if (p->on_rq && ..) // false 818 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 819 * deactivate_task() ttwu_queue_wakelist()) 820 * p->on_rq = 0; p->sched_remote_wakeup = Y; 821 * 822 * guarantees all stores of 'current' are visible before 823 * ->sched_remote_wakeup gets used, so it can be in this word. 824 */ 825 unsigned sched_remote_wakeup:1; 826 827 /* Bit to tell LSMs we're in execve(): */ 828 unsigned in_execve:1; 829 unsigned in_iowait:1; 830 #ifndef TIF_RESTORE_SIGMASK 831 unsigned restore_sigmask:1; 832 #endif 833 #ifdef CONFIG_MEMCG 834 unsigned in_user_fault:1; 835 #endif 836 #ifdef CONFIG_COMPAT_BRK 837 unsigned brk_randomized:1; 838 #endif 839 #ifdef CONFIG_CGROUPS 840 /* disallow userland-initiated cgroup migration */ 841 unsigned no_cgroup_migration:1; 842 /* task is frozen/stopped (used by the cgroup freezer) */ 843 unsigned frozen:1; 844 #endif 845 #ifdef CONFIG_BLK_CGROUP 846 unsigned use_memdelay:1; 847 #endif 848 #ifdef CONFIG_PSI 849 /* Stalled due to lack of memory */ 850 unsigned in_memstall:1; 851 #endif 852 #ifdef CONFIG_PAGE_OWNER 853 /* Used by page_owner=on to detect recursion in page tracking. */ 854 unsigned in_page_owner:1; 855 #endif 856 857 unsigned long atomic_flags; /* Flags requiring atomic access. */ 858 859 struct restart_block restart_block; 860 861 pid_t pid; 862 pid_t tgid; 863 864 #ifdef CONFIG_STACKPROTECTOR 865 /* Canary value for the -fstack-protector GCC feature: */ 866 unsigned long stack_canary; 867 #endif 868 /* 869 * Pointers to the (original) parent process, youngest child, younger sibling, 870 * older sibling, respectively. (p->father can be replaced with 871 * p->real_parent->pid) 872 */ 873 874 /* Real parent process: */ 875 struct task_struct __rcu *real_parent; 876 877 /* Recipient of SIGCHLD, wait4() reports: */ 878 struct task_struct __rcu *parent; 879 880 /* 881 * Children/sibling form the list of natural children: 882 */ 883 struct list_head children; 884 struct list_head sibling; 885 struct task_struct *group_leader; 886 887 /* 888 * 'ptraced' is the list of tasks this task is using ptrace() on. 889 * 890 * This includes both natural children and PTRACE_ATTACH targets. 891 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 892 */ 893 struct list_head ptraced; 894 struct list_head ptrace_entry; 895 896 /* PID/PID hash table linkage. */ 897 struct pid *thread_pid; 898 struct hlist_node pid_links[PIDTYPE_MAX]; 899 struct list_head thread_group; 900 struct list_head thread_node; 901 902 struct completion *vfork_done; 903 904 /* CLONE_CHILD_SETTID: */ 905 int __user *set_child_tid; 906 907 /* CLONE_CHILD_CLEARTID: */ 908 int __user *clear_child_tid; 909 910 /* PF_IO_WORKER */ 911 void *pf_io_worker; 912 913 u64 utime; 914 u64 stime; 915 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 916 u64 utimescaled; 917 u64 stimescaled; 918 #endif 919 u64 gtime; 920 struct prev_cputime prev_cputime; 921 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 922 struct vtime vtime; 923 #endif 924 925 #ifdef CONFIG_NO_HZ_FULL 926 atomic_t tick_dep_mask; 927 #endif 928 /* Context switch counts: */ 929 unsigned long nvcsw; 930 unsigned long nivcsw; 931 932 /* Monotonic time in nsecs: */ 933 u64 start_time; 934 935 /* Boot based time in nsecs: */ 936 u64 start_boottime; 937 938 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 939 unsigned long min_flt; 940 unsigned long maj_flt; 941 942 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 943 struct posix_cputimers posix_cputimers; 944 945 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 946 struct posix_cputimers_work posix_cputimers_work; 947 #endif 948 949 /* Process credentials: */ 950 951 /* Tracer's credentials at attach: */ 952 const struct cred __rcu *ptracer_cred; 953 954 /* Objective and real subjective task credentials (COW): */ 955 const struct cred __rcu *real_cred; 956 957 /* Effective (overridable) subjective task credentials (COW): */ 958 const struct cred __rcu *cred; 959 960 #ifdef CONFIG_KEYS 961 /* Cached requested key. */ 962 struct key *cached_requested_key; 963 #endif 964 965 /* 966 * executable name, excluding path. 967 * 968 * - normally initialized setup_new_exec() 969 * - access it with [gs]et_task_comm() 970 * - lock it with task_lock() 971 */ 972 char comm[TASK_COMM_LEN]; 973 974 struct nameidata *nameidata; 975 976 #ifdef CONFIG_SYSVIPC 977 struct sysv_sem sysvsem; 978 struct sysv_shm sysvshm; 979 #endif 980 #ifdef CONFIG_DETECT_HUNG_TASK 981 unsigned long last_switch_count; 982 unsigned long last_switch_time; 983 #endif 984 /* Filesystem information: */ 985 struct fs_struct *fs; 986 987 /* Open file information: */ 988 struct files_struct *files; 989 990 #ifdef CONFIG_IO_URING 991 struct io_uring_task *io_uring; 992 #endif 993 994 /* Namespaces: */ 995 struct nsproxy *nsproxy; 996 997 /* Signal handlers: */ 998 struct signal_struct *signal; 999 struct sighand_struct __rcu *sighand; 1000 struct sigqueue *sigqueue_cache; 1001 sigset_t blocked; 1002 sigset_t real_blocked; 1003 /* Restored if set_restore_sigmask() was used: */ 1004 sigset_t saved_sigmask; 1005 struct sigpending pending; 1006 unsigned long sas_ss_sp; 1007 size_t sas_ss_size; 1008 unsigned int sas_ss_flags; 1009 1010 struct callback_head *task_works; 1011 1012 #ifdef CONFIG_AUDIT 1013 #ifdef CONFIG_AUDITSYSCALL 1014 struct audit_context *audit_context; 1015 #endif 1016 kuid_t loginuid; 1017 unsigned int sessionid; 1018 #endif 1019 struct seccomp seccomp; 1020 struct syscall_user_dispatch syscall_dispatch; 1021 1022 /* Thread group tracking: */ 1023 u64 parent_exec_id; 1024 u64 self_exec_id; 1025 1026 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1027 spinlock_t alloc_lock; 1028 1029 /* Protection of the PI data structures: */ 1030 raw_spinlock_t pi_lock; 1031 1032 struct wake_q_node wake_q; 1033 1034 #ifdef CONFIG_RT_MUTEXES 1035 /* PI waiters blocked on a rt_mutex held by this task: */ 1036 struct rb_root_cached pi_waiters; 1037 /* Updated under owner's pi_lock and rq lock */ 1038 struct task_struct *pi_top_task; 1039 /* Deadlock detection and priority inheritance handling: */ 1040 struct rt_mutex_waiter *pi_blocked_on; 1041 #endif 1042 1043 #ifdef CONFIG_DEBUG_MUTEXES 1044 /* Mutex deadlock detection: */ 1045 struct mutex_waiter *blocked_on; 1046 #endif 1047 1048 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1049 int non_block_count; 1050 #endif 1051 1052 #ifdef CONFIG_TRACE_IRQFLAGS 1053 struct irqtrace_events irqtrace; 1054 unsigned int hardirq_threaded; 1055 u64 hardirq_chain_key; 1056 int softirqs_enabled; 1057 int softirq_context; 1058 int irq_config; 1059 #endif 1060 #ifdef CONFIG_PREEMPT_RT 1061 int softirq_disable_cnt; 1062 #endif 1063 1064 #ifdef CONFIG_LOCKDEP 1065 # define MAX_LOCK_DEPTH 48UL 1066 u64 curr_chain_key; 1067 int lockdep_depth; 1068 unsigned int lockdep_recursion; 1069 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1070 #endif 1071 1072 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1073 unsigned int in_ubsan; 1074 #endif 1075 1076 /* Journalling filesystem info: */ 1077 void *journal_info; 1078 1079 /* Stacked block device info: */ 1080 struct bio_list *bio_list; 1081 1082 #ifdef CONFIG_BLOCK 1083 /* Stack plugging: */ 1084 struct blk_plug *plug; 1085 #endif 1086 1087 /* VM state: */ 1088 struct reclaim_state *reclaim_state; 1089 1090 struct backing_dev_info *backing_dev_info; 1091 1092 struct io_context *io_context; 1093 1094 #ifdef CONFIG_COMPACTION 1095 struct capture_control *capture_control; 1096 #endif 1097 /* Ptrace state: */ 1098 unsigned long ptrace_message; 1099 kernel_siginfo_t *last_siginfo; 1100 1101 struct task_io_accounting ioac; 1102 #ifdef CONFIG_PSI 1103 /* Pressure stall state */ 1104 unsigned int psi_flags; 1105 #endif 1106 #ifdef CONFIG_TASK_XACCT 1107 /* Accumulated RSS usage: */ 1108 u64 acct_rss_mem1; 1109 /* Accumulated virtual memory usage: */ 1110 u64 acct_vm_mem1; 1111 /* stime + utime since last update: */ 1112 u64 acct_timexpd; 1113 #endif 1114 #ifdef CONFIG_CPUSETS 1115 /* Protected by ->alloc_lock: */ 1116 nodemask_t mems_allowed; 1117 /* Sequence number to catch updates: */ 1118 seqcount_spinlock_t mems_allowed_seq; 1119 int cpuset_mem_spread_rotor; 1120 int cpuset_slab_spread_rotor; 1121 #endif 1122 #ifdef CONFIG_CGROUPS 1123 /* Control Group info protected by css_set_lock: */ 1124 struct css_set __rcu *cgroups; 1125 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1126 struct list_head cg_list; 1127 #endif 1128 #ifdef CONFIG_X86_CPU_RESCTRL 1129 u32 closid; 1130 u32 rmid; 1131 #endif 1132 #ifdef CONFIG_FUTEX 1133 struct robust_list_head __user *robust_list; 1134 #ifdef CONFIG_COMPAT 1135 struct compat_robust_list_head __user *compat_robust_list; 1136 #endif 1137 struct list_head pi_state_list; 1138 struct futex_pi_state *pi_state_cache; 1139 struct mutex futex_exit_mutex; 1140 unsigned int futex_state; 1141 #endif 1142 #ifdef CONFIG_PERF_EVENTS 1143 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 1144 struct mutex perf_event_mutex; 1145 struct list_head perf_event_list; 1146 #endif 1147 #ifdef CONFIG_DEBUG_PREEMPT 1148 unsigned long preempt_disable_ip; 1149 #endif 1150 #ifdef CONFIG_NUMA 1151 /* Protected by alloc_lock: */ 1152 struct mempolicy *mempolicy; 1153 short il_prev; 1154 short pref_node_fork; 1155 #endif 1156 #ifdef CONFIG_NUMA_BALANCING 1157 int numa_scan_seq; 1158 unsigned int numa_scan_period; 1159 unsigned int numa_scan_period_max; 1160 int numa_preferred_nid; 1161 unsigned long numa_migrate_retry; 1162 /* Migration stamp: */ 1163 u64 node_stamp; 1164 u64 last_task_numa_placement; 1165 u64 last_sum_exec_runtime; 1166 struct callback_head numa_work; 1167 1168 /* 1169 * This pointer is only modified for current in syscall and 1170 * pagefault context (and for tasks being destroyed), so it can be read 1171 * from any of the following contexts: 1172 * - RCU read-side critical section 1173 * - current->numa_group from everywhere 1174 * - task's runqueue locked, task not running 1175 */ 1176 struct numa_group __rcu *numa_group; 1177 1178 /* 1179 * numa_faults is an array split into four regions: 1180 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1181 * in this precise order. 1182 * 1183 * faults_memory: Exponential decaying average of faults on a per-node 1184 * basis. Scheduling placement decisions are made based on these 1185 * counts. The values remain static for the duration of a PTE scan. 1186 * faults_cpu: Track the nodes the process was running on when a NUMA 1187 * hinting fault was incurred. 1188 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1189 * during the current scan window. When the scan completes, the counts 1190 * in faults_memory and faults_cpu decay and these values are copied. 1191 */ 1192 unsigned long *numa_faults; 1193 unsigned long total_numa_faults; 1194 1195 /* 1196 * numa_faults_locality tracks if faults recorded during the last 1197 * scan window were remote/local or failed to migrate. The task scan 1198 * period is adapted based on the locality of the faults with different 1199 * weights depending on whether they were shared or private faults 1200 */ 1201 unsigned long numa_faults_locality[3]; 1202 1203 unsigned long numa_pages_migrated; 1204 #endif /* CONFIG_NUMA_BALANCING */ 1205 1206 #ifdef CONFIG_RSEQ 1207 struct rseq __user *rseq; 1208 u32 rseq_sig; 1209 /* 1210 * RmW on rseq_event_mask must be performed atomically 1211 * with respect to preemption. 1212 */ 1213 unsigned long rseq_event_mask; 1214 #endif 1215 1216 struct tlbflush_unmap_batch tlb_ubc; 1217 1218 union { 1219 refcount_t rcu_users; 1220 struct rcu_head rcu; 1221 }; 1222 1223 /* Cache last used pipe for splice(): */ 1224 struct pipe_inode_info *splice_pipe; 1225 1226 struct page_frag task_frag; 1227 1228 #ifdef CONFIG_TASK_DELAY_ACCT 1229 struct task_delay_info *delays; 1230 #endif 1231 1232 #ifdef CONFIG_FAULT_INJECTION 1233 int make_it_fail; 1234 unsigned int fail_nth; 1235 #endif 1236 /* 1237 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1238 * balance_dirty_pages() for a dirty throttling pause: 1239 */ 1240 int nr_dirtied; 1241 int nr_dirtied_pause; 1242 /* Start of a write-and-pause period: */ 1243 unsigned long dirty_paused_when; 1244 1245 #ifdef CONFIG_LATENCYTOP 1246 int latency_record_count; 1247 struct latency_record latency_record[LT_SAVECOUNT]; 1248 #endif 1249 /* 1250 * Time slack values; these are used to round up poll() and 1251 * select() etc timeout values. These are in nanoseconds. 1252 */ 1253 u64 timer_slack_ns; 1254 u64 default_timer_slack_ns; 1255 1256 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1257 unsigned int kasan_depth; 1258 #endif 1259 1260 #ifdef CONFIG_KCSAN 1261 struct kcsan_ctx kcsan_ctx; 1262 #ifdef CONFIG_TRACE_IRQFLAGS 1263 struct irqtrace_events kcsan_save_irqtrace; 1264 #endif 1265 #endif 1266 1267 #if IS_ENABLED(CONFIG_KUNIT) 1268 struct kunit *kunit_test; 1269 #endif 1270 1271 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1272 /* Index of current stored address in ret_stack: */ 1273 int curr_ret_stack; 1274 int curr_ret_depth; 1275 1276 /* Stack of return addresses for return function tracing: */ 1277 struct ftrace_ret_stack *ret_stack; 1278 1279 /* Timestamp for last schedule: */ 1280 unsigned long long ftrace_timestamp; 1281 1282 /* 1283 * Number of functions that haven't been traced 1284 * because of depth overrun: 1285 */ 1286 atomic_t trace_overrun; 1287 1288 /* Pause tracing: */ 1289 atomic_t tracing_graph_pause; 1290 #endif 1291 1292 #ifdef CONFIG_TRACING 1293 /* State flags for use by tracers: */ 1294 unsigned long trace; 1295 1296 /* Bitmask and counter of trace recursion: */ 1297 unsigned long trace_recursion; 1298 #endif /* CONFIG_TRACING */ 1299 1300 #ifdef CONFIG_KCOV 1301 /* See kernel/kcov.c for more details. */ 1302 1303 /* Coverage collection mode enabled for this task (0 if disabled): */ 1304 unsigned int kcov_mode; 1305 1306 /* Size of the kcov_area: */ 1307 unsigned int kcov_size; 1308 1309 /* Buffer for coverage collection: */ 1310 void *kcov_area; 1311 1312 /* KCOV descriptor wired with this task or NULL: */ 1313 struct kcov *kcov; 1314 1315 /* KCOV common handle for remote coverage collection: */ 1316 u64 kcov_handle; 1317 1318 /* KCOV sequence number: */ 1319 int kcov_sequence; 1320 1321 /* Collect coverage from softirq context: */ 1322 unsigned int kcov_softirq; 1323 #endif 1324 1325 #ifdef CONFIG_MEMCG 1326 struct mem_cgroup *memcg_in_oom; 1327 gfp_t memcg_oom_gfp_mask; 1328 int memcg_oom_order; 1329 1330 /* Number of pages to reclaim on returning to userland: */ 1331 unsigned int memcg_nr_pages_over_high; 1332 1333 /* Used by memcontrol for targeted memcg charge: */ 1334 struct mem_cgroup *active_memcg; 1335 #endif 1336 1337 #ifdef CONFIG_BLK_CGROUP 1338 struct request_queue *throttle_queue; 1339 #endif 1340 1341 #ifdef CONFIG_UPROBES 1342 struct uprobe_task *utask; 1343 #endif 1344 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1345 unsigned int sequential_io; 1346 unsigned int sequential_io_avg; 1347 #endif 1348 struct kmap_ctrl kmap_ctrl; 1349 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1350 unsigned long task_state_change; 1351 #endif 1352 int pagefault_disabled; 1353 #ifdef CONFIG_MMU 1354 struct task_struct *oom_reaper_list; 1355 #endif 1356 #ifdef CONFIG_VMAP_STACK 1357 struct vm_struct *stack_vm_area; 1358 #endif 1359 #ifdef CONFIG_THREAD_INFO_IN_TASK 1360 /* A live task holds one reference: */ 1361 refcount_t stack_refcount; 1362 #endif 1363 #ifdef CONFIG_LIVEPATCH 1364 int patch_state; 1365 #endif 1366 #ifdef CONFIG_SECURITY 1367 /* Used by LSM modules for access restriction: */ 1368 void *security; 1369 #endif 1370 #ifdef CONFIG_BPF_SYSCALL 1371 /* Used by BPF task local storage */ 1372 struct bpf_local_storage __rcu *bpf_storage; 1373 #endif 1374 1375 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1376 unsigned long lowest_stack; 1377 unsigned long prev_lowest_stack; 1378 #endif 1379 1380 #ifdef CONFIG_X86_MCE 1381 void __user *mce_vaddr; 1382 __u64 mce_kflags; 1383 u64 mce_addr; 1384 __u64 mce_ripv : 1, 1385 mce_whole_page : 1, 1386 __mce_reserved : 62; 1387 struct callback_head mce_kill_me; 1388 #endif 1389 1390 #ifdef CONFIG_KRETPROBES 1391 struct llist_head kretprobe_instances; 1392 #endif 1393 1394 /* 1395 * New fields for task_struct should be added above here, so that 1396 * they are included in the randomized portion of task_struct. 1397 */ 1398 randomized_struct_fields_end 1399 1400 /* CPU-specific state of this task: */ 1401 struct thread_struct thread; 1402 1403 /* 1404 * WARNING: on x86, 'thread_struct' contains a variable-sized 1405 * structure. It *MUST* be at the end of 'task_struct'. 1406 * 1407 * Do not put anything below here! 1408 */ 1409 }; 1410 1411 static inline struct pid *task_pid(struct task_struct *task) 1412 { 1413 return task->thread_pid; 1414 } 1415 1416 /* 1417 * the helpers to get the task's different pids as they are seen 1418 * from various namespaces 1419 * 1420 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1421 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1422 * current. 1423 * task_xid_nr_ns() : id seen from the ns specified; 1424 * 1425 * see also pid_nr() etc in include/linux/pid.h 1426 */ 1427 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1428 1429 static inline pid_t task_pid_nr(struct task_struct *tsk) 1430 { 1431 return tsk->pid; 1432 } 1433 1434 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1435 { 1436 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1437 } 1438 1439 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1440 { 1441 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1442 } 1443 1444 1445 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1446 { 1447 return tsk->tgid; 1448 } 1449 1450 /** 1451 * pid_alive - check that a task structure is not stale 1452 * @p: Task structure to be checked. 1453 * 1454 * Test if a process is not yet dead (at most zombie state) 1455 * If pid_alive fails, then pointers within the task structure 1456 * can be stale and must not be dereferenced. 1457 * 1458 * Return: 1 if the process is alive. 0 otherwise. 1459 */ 1460 static inline int pid_alive(const struct task_struct *p) 1461 { 1462 return p->thread_pid != NULL; 1463 } 1464 1465 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1466 { 1467 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1468 } 1469 1470 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1471 { 1472 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1473 } 1474 1475 1476 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1477 { 1478 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1479 } 1480 1481 static inline pid_t task_session_vnr(struct task_struct *tsk) 1482 { 1483 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1484 } 1485 1486 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1487 { 1488 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns); 1489 } 1490 1491 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1492 { 1493 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL); 1494 } 1495 1496 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1497 { 1498 pid_t pid = 0; 1499 1500 rcu_read_lock(); 1501 if (pid_alive(tsk)) 1502 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1503 rcu_read_unlock(); 1504 1505 return pid; 1506 } 1507 1508 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1509 { 1510 return task_ppid_nr_ns(tsk, &init_pid_ns); 1511 } 1512 1513 /* Obsolete, do not use: */ 1514 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1515 { 1516 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1517 } 1518 1519 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1520 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1521 1522 static inline unsigned int task_state_index(struct task_struct *tsk) 1523 { 1524 unsigned int tsk_state = READ_ONCE(tsk->state); 1525 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT; 1526 1527 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1528 1529 if (tsk_state == TASK_IDLE) 1530 state = TASK_REPORT_IDLE; 1531 1532 return fls(state); 1533 } 1534 1535 static inline char task_index_to_char(unsigned int state) 1536 { 1537 static const char state_char[] = "RSDTtXZPI"; 1538 1539 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1540 1541 return state_char[state]; 1542 } 1543 1544 static inline char task_state_to_char(struct task_struct *tsk) 1545 { 1546 return task_index_to_char(task_state_index(tsk)); 1547 } 1548 1549 /** 1550 * is_global_init - check if a task structure is init. Since init 1551 * is free to have sub-threads we need to check tgid. 1552 * @tsk: Task structure to be checked. 1553 * 1554 * Check if a task structure is the first user space task the kernel created. 1555 * 1556 * Return: 1 if the task structure is init. 0 otherwise. 1557 */ 1558 static inline int is_global_init(struct task_struct *tsk) 1559 { 1560 return task_tgid_nr(tsk) == 1; 1561 } 1562 1563 extern struct pid *cad_pid; 1564 1565 /* 1566 * Per process flags 1567 */ 1568 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1569 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1570 #define PF_EXITING 0x00000004 /* Getting shut down */ 1571 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1572 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1573 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1574 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1575 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1576 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1577 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1578 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1579 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1580 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1581 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1582 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1583 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1584 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1585 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1586 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1587 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1588 * I am cleaning dirty pages from some other bdi. */ 1589 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1590 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1591 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1592 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1593 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1594 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */ 1595 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1596 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1597 1598 /* 1599 * Only the _current_ task can read/write to tsk->flags, but other 1600 * tasks can access tsk->flags in readonly mode for example 1601 * with tsk_used_math (like during threaded core dumping). 1602 * There is however an exception to this rule during ptrace 1603 * or during fork: the ptracer task is allowed to write to the 1604 * child->flags of its traced child (same goes for fork, the parent 1605 * can write to the child->flags), because we're guaranteed the 1606 * child is not running and in turn not changing child->flags 1607 * at the same time the parent does it. 1608 */ 1609 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1610 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1611 #define clear_used_math() clear_stopped_child_used_math(current) 1612 #define set_used_math() set_stopped_child_used_math(current) 1613 1614 #define conditional_stopped_child_used_math(condition, child) \ 1615 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1616 1617 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1618 1619 #define copy_to_stopped_child_used_math(child) \ 1620 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1621 1622 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1623 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1624 #define used_math() tsk_used_math(current) 1625 1626 static inline bool is_percpu_thread(void) 1627 { 1628 #ifdef CONFIG_SMP 1629 return (current->flags & PF_NO_SETAFFINITY) && 1630 (current->nr_cpus_allowed == 1); 1631 #else 1632 return true; 1633 #endif 1634 } 1635 1636 /* Per-process atomic flags. */ 1637 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1638 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1639 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1640 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1641 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1642 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1643 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1644 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1645 1646 #define TASK_PFA_TEST(name, func) \ 1647 static inline bool task_##func(struct task_struct *p) \ 1648 { return test_bit(PFA_##name, &p->atomic_flags); } 1649 1650 #define TASK_PFA_SET(name, func) \ 1651 static inline void task_set_##func(struct task_struct *p) \ 1652 { set_bit(PFA_##name, &p->atomic_flags); } 1653 1654 #define TASK_PFA_CLEAR(name, func) \ 1655 static inline void task_clear_##func(struct task_struct *p) \ 1656 { clear_bit(PFA_##name, &p->atomic_flags); } 1657 1658 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1659 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1660 1661 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1662 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1663 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1664 1665 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1666 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1667 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1668 1669 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1670 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1671 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1672 1673 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1674 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1675 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1676 1677 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1678 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1679 1680 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1681 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1682 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1683 1684 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1685 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1686 1687 static inline void 1688 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1689 { 1690 current->flags &= ~flags; 1691 current->flags |= orig_flags & flags; 1692 } 1693 1694 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1695 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1696 #ifdef CONFIG_SMP 1697 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1698 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1699 #else 1700 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1701 { 1702 } 1703 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1704 { 1705 if (!cpumask_test_cpu(0, new_mask)) 1706 return -EINVAL; 1707 return 0; 1708 } 1709 #endif 1710 1711 extern int yield_to(struct task_struct *p, bool preempt); 1712 extern void set_user_nice(struct task_struct *p, long nice); 1713 extern int task_prio(const struct task_struct *p); 1714 1715 /** 1716 * task_nice - return the nice value of a given task. 1717 * @p: the task in question. 1718 * 1719 * Return: The nice value [ -20 ... 0 ... 19 ]. 1720 */ 1721 static inline int task_nice(const struct task_struct *p) 1722 { 1723 return PRIO_TO_NICE((p)->static_prio); 1724 } 1725 1726 extern int can_nice(const struct task_struct *p, const int nice); 1727 extern int task_curr(const struct task_struct *p); 1728 extern int idle_cpu(int cpu); 1729 extern int available_idle_cpu(int cpu); 1730 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1731 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1732 extern void sched_set_fifo(struct task_struct *p); 1733 extern void sched_set_fifo_low(struct task_struct *p); 1734 extern void sched_set_normal(struct task_struct *p, int nice); 1735 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1736 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1737 extern struct task_struct *idle_task(int cpu); 1738 1739 /** 1740 * is_idle_task - is the specified task an idle task? 1741 * @p: the task in question. 1742 * 1743 * Return: 1 if @p is an idle task. 0 otherwise. 1744 */ 1745 static __always_inline bool is_idle_task(const struct task_struct *p) 1746 { 1747 return !!(p->flags & PF_IDLE); 1748 } 1749 1750 extern struct task_struct *curr_task(int cpu); 1751 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1752 1753 void yield(void); 1754 1755 union thread_union { 1756 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1757 struct task_struct task; 1758 #endif 1759 #ifndef CONFIG_THREAD_INFO_IN_TASK 1760 struct thread_info thread_info; 1761 #endif 1762 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1763 }; 1764 1765 #ifndef CONFIG_THREAD_INFO_IN_TASK 1766 extern struct thread_info init_thread_info; 1767 #endif 1768 1769 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1770 1771 #ifdef CONFIG_THREAD_INFO_IN_TASK 1772 static inline struct thread_info *task_thread_info(struct task_struct *task) 1773 { 1774 return &task->thread_info; 1775 } 1776 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1777 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1778 #endif 1779 1780 /* 1781 * find a task by one of its numerical ids 1782 * 1783 * find_task_by_pid_ns(): 1784 * finds a task by its pid in the specified namespace 1785 * find_task_by_vpid(): 1786 * finds a task by its virtual pid 1787 * 1788 * see also find_vpid() etc in include/linux/pid.h 1789 */ 1790 1791 extern struct task_struct *find_task_by_vpid(pid_t nr); 1792 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1793 1794 /* 1795 * find a task by its virtual pid and get the task struct 1796 */ 1797 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1798 1799 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1800 extern int wake_up_process(struct task_struct *tsk); 1801 extern void wake_up_new_task(struct task_struct *tsk); 1802 1803 #ifdef CONFIG_SMP 1804 extern void kick_process(struct task_struct *tsk); 1805 #else 1806 static inline void kick_process(struct task_struct *tsk) { } 1807 #endif 1808 1809 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1810 1811 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1812 { 1813 __set_task_comm(tsk, from, false); 1814 } 1815 1816 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1817 #define get_task_comm(buf, tsk) ({ \ 1818 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1819 __get_task_comm(buf, sizeof(buf), tsk); \ 1820 }) 1821 1822 #ifdef CONFIG_SMP 1823 static __always_inline void scheduler_ipi(void) 1824 { 1825 /* 1826 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1827 * TIF_NEED_RESCHED remotely (for the first time) will also send 1828 * this IPI. 1829 */ 1830 preempt_fold_need_resched(); 1831 } 1832 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 1833 #else 1834 static inline void scheduler_ipi(void) { } 1835 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) 1836 { 1837 return 1; 1838 } 1839 #endif 1840 1841 /* 1842 * Set thread flags in other task's structures. 1843 * See asm/thread_info.h for TIF_xxxx flags available: 1844 */ 1845 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1846 { 1847 set_ti_thread_flag(task_thread_info(tsk), flag); 1848 } 1849 1850 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1851 { 1852 clear_ti_thread_flag(task_thread_info(tsk), flag); 1853 } 1854 1855 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1856 bool value) 1857 { 1858 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1859 } 1860 1861 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1862 { 1863 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1864 } 1865 1866 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1867 { 1868 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1869 } 1870 1871 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1872 { 1873 return test_ti_thread_flag(task_thread_info(tsk), flag); 1874 } 1875 1876 static inline void set_tsk_need_resched(struct task_struct *tsk) 1877 { 1878 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1879 } 1880 1881 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1882 { 1883 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1884 } 1885 1886 static inline int test_tsk_need_resched(struct task_struct *tsk) 1887 { 1888 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1889 } 1890 1891 /* 1892 * cond_resched() and cond_resched_lock(): latency reduction via 1893 * explicit rescheduling in places that are safe. The return 1894 * value indicates whether a reschedule was done in fact. 1895 * cond_resched_lock() will drop the spinlock before scheduling, 1896 */ 1897 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 1898 extern int __cond_resched(void); 1899 1900 #ifdef CONFIG_PREEMPT_DYNAMIC 1901 1902 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 1903 1904 static __always_inline int _cond_resched(void) 1905 { 1906 return static_call_mod(cond_resched)(); 1907 } 1908 1909 #else 1910 1911 static inline int _cond_resched(void) 1912 { 1913 return __cond_resched(); 1914 } 1915 1916 #endif /* CONFIG_PREEMPT_DYNAMIC */ 1917 1918 #else 1919 1920 static inline int _cond_resched(void) { return 0; } 1921 1922 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */ 1923 1924 #define cond_resched() ({ \ 1925 ___might_sleep(__FILE__, __LINE__, 0); \ 1926 _cond_resched(); \ 1927 }) 1928 1929 extern int __cond_resched_lock(spinlock_t *lock); 1930 extern int __cond_resched_rwlock_read(rwlock_t *lock); 1931 extern int __cond_resched_rwlock_write(rwlock_t *lock); 1932 1933 #define cond_resched_lock(lock) ({ \ 1934 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ 1935 __cond_resched_lock(lock); \ 1936 }) 1937 1938 #define cond_resched_rwlock_read(lock) ({ \ 1939 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \ 1940 __cond_resched_rwlock_read(lock); \ 1941 }) 1942 1943 #define cond_resched_rwlock_write(lock) ({ \ 1944 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \ 1945 __cond_resched_rwlock_write(lock); \ 1946 }) 1947 1948 static inline void cond_resched_rcu(void) 1949 { 1950 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 1951 rcu_read_unlock(); 1952 cond_resched(); 1953 rcu_read_lock(); 1954 #endif 1955 } 1956 1957 /* 1958 * Does a critical section need to be broken due to another 1959 * task waiting?: (technically does not depend on CONFIG_PREEMPTION, 1960 * but a general need for low latency) 1961 */ 1962 static inline int spin_needbreak(spinlock_t *lock) 1963 { 1964 #ifdef CONFIG_PREEMPTION 1965 return spin_is_contended(lock); 1966 #else 1967 return 0; 1968 #endif 1969 } 1970 1971 /* 1972 * Check if a rwlock is contended. 1973 * Returns non-zero if there is another task waiting on the rwlock. 1974 * Returns zero if the lock is not contended or the system / underlying 1975 * rwlock implementation does not support contention detection. 1976 * Technically does not depend on CONFIG_PREEMPTION, but a general need 1977 * for low latency. 1978 */ 1979 static inline int rwlock_needbreak(rwlock_t *lock) 1980 { 1981 #ifdef CONFIG_PREEMPTION 1982 return rwlock_is_contended(lock); 1983 #else 1984 return 0; 1985 #endif 1986 } 1987 1988 static __always_inline bool need_resched(void) 1989 { 1990 return unlikely(tif_need_resched()); 1991 } 1992 1993 /* 1994 * Wrappers for p->thread_info->cpu access. No-op on UP. 1995 */ 1996 #ifdef CONFIG_SMP 1997 1998 static inline unsigned int task_cpu(const struct task_struct *p) 1999 { 2000 #ifdef CONFIG_THREAD_INFO_IN_TASK 2001 return READ_ONCE(p->cpu); 2002 #else 2003 return READ_ONCE(task_thread_info(p)->cpu); 2004 #endif 2005 } 2006 2007 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2008 2009 #else 2010 2011 static inline unsigned int task_cpu(const struct task_struct *p) 2012 { 2013 return 0; 2014 } 2015 2016 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2017 { 2018 } 2019 2020 #endif /* CONFIG_SMP */ 2021 2022 /* 2023 * In order to reduce various lock holder preemption latencies provide an 2024 * interface to see if a vCPU is currently running or not. 2025 * 2026 * This allows us to terminate optimistic spin loops and block, analogous to 2027 * the native optimistic spin heuristic of testing if the lock owner task is 2028 * running or not. 2029 */ 2030 #ifndef vcpu_is_preempted 2031 static inline bool vcpu_is_preempted(int cpu) 2032 { 2033 return false; 2034 } 2035 #endif 2036 2037 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2038 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2039 2040 #ifndef TASK_SIZE_OF 2041 #define TASK_SIZE_OF(tsk) TASK_SIZE 2042 #endif 2043 2044 #ifdef CONFIG_SMP 2045 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2046 unsigned long sched_cpu_util(int cpu, unsigned long max); 2047 #endif /* CONFIG_SMP */ 2048 2049 #ifdef CONFIG_RSEQ 2050 2051 /* 2052 * Map the event mask on the user-space ABI enum rseq_cs_flags 2053 * for direct mask checks. 2054 */ 2055 enum rseq_event_mask_bits { 2056 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT, 2057 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT, 2058 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT, 2059 }; 2060 2061 enum rseq_event_mask { 2062 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT), 2063 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT), 2064 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT), 2065 }; 2066 2067 static inline void rseq_set_notify_resume(struct task_struct *t) 2068 { 2069 if (t->rseq) 2070 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME); 2071 } 2072 2073 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs); 2074 2075 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2076 struct pt_regs *regs) 2077 { 2078 if (current->rseq) 2079 __rseq_handle_notify_resume(ksig, regs); 2080 } 2081 2082 static inline void rseq_signal_deliver(struct ksignal *ksig, 2083 struct pt_regs *regs) 2084 { 2085 preempt_disable(); 2086 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask); 2087 preempt_enable(); 2088 rseq_handle_notify_resume(ksig, regs); 2089 } 2090 2091 /* rseq_preempt() requires preemption to be disabled. */ 2092 static inline void rseq_preempt(struct task_struct *t) 2093 { 2094 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask); 2095 rseq_set_notify_resume(t); 2096 } 2097 2098 /* rseq_migrate() requires preemption to be disabled. */ 2099 static inline void rseq_migrate(struct task_struct *t) 2100 { 2101 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask); 2102 rseq_set_notify_resume(t); 2103 } 2104 2105 /* 2106 * If parent process has a registered restartable sequences area, the 2107 * child inherits. Unregister rseq for a clone with CLONE_VM set. 2108 */ 2109 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2110 { 2111 if (clone_flags & CLONE_VM) { 2112 t->rseq = NULL; 2113 t->rseq_sig = 0; 2114 t->rseq_event_mask = 0; 2115 } else { 2116 t->rseq = current->rseq; 2117 t->rseq_sig = current->rseq_sig; 2118 t->rseq_event_mask = current->rseq_event_mask; 2119 } 2120 } 2121 2122 static inline void rseq_execve(struct task_struct *t) 2123 { 2124 t->rseq = NULL; 2125 t->rseq_sig = 0; 2126 t->rseq_event_mask = 0; 2127 } 2128 2129 #else 2130 2131 static inline void rseq_set_notify_resume(struct task_struct *t) 2132 { 2133 } 2134 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2135 struct pt_regs *regs) 2136 { 2137 } 2138 static inline void rseq_signal_deliver(struct ksignal *ksig, 2139 struct pt_regs *regs) 2140 { 2141 } 2142 static inline void rseq_preempt(struct task_struct *t) 2143 { 2144 } 2145 static inline void rseq_migrate(struct task_struct *t) 2146 { 2147 } 2148 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2149 { 2150 } 2151 static inline void rseq_execve(struct task_struct *t) 2152 { 2153 } 2154 2155 #endif 2156 2157 #ifdef CONFIG_DEBUG_RSEQ 2158 2159 void rseq_syscall(struct pt_regs *regs); 2160 2161 #else 2162 2163 static inline void rseq_syscall(struct pt_regs *regs) 2164 { 2165 } 2166 2167 #endif 2168 2169 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq); 2170 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len); 2171 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq); 2172 2173 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq); 2174 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq); 2175 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq); 2176 2177 int sched_trace_rq_cpu(struct rq *rq); 2178 int sched_trace_rq_cpu_capacity(struct rq *rq); 2179 int sched_trace_rq_nr_running(struct rq *rq); 2180 2181 const struct cpumask *sched_trace_rd_span(struct root_domain *rd); 2182 2183 #endif 2184