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/kmsan_types.h> 18 #include <linux/mutex.h> 19 #include <linux/plist.h> 20 #include <linux/hrtimer.h> 21 #include <linux/irqflags.h> 22 #include <linux/seccomp.h> 23 #include <linux/nodemask.h> 24 #include <linux/rcupdate.h> 25 #include <linux/refcount.h> 26 #include <linux/resource.h> 27 #include <linux/latencytop.h> 28 #include <linux/sched/prio.h> 29 #include <linux/sched/types.h> 30 #include <linux/signal_types.h> 31 #include <linux/syscall_user_dispatch.h> 32 #include <linux/mm_types_task.h> 33 #include <linux/task_io_accounting.h> 34 #include <linux/posix-timers.h> 35 #include <linux/rseq.h> 36 #include <linux/seqlock.h> 37 #include <linux/kcsan.h> 38 #include <linux/rv.h> 39 #include <linux/livepatch_sched.h> 40 #include <asm/kmap_size.h> 41 42 /* task_struct member predeclarations (sorted alphabetically): */ 43 struct audit_context; 44 struct bio_list; 45 struct blk_plug; 46 struct bpf_local_storage; 47 struct bpf_run_ctx; 48 struct capture_control; 49 struct cfs_rq; 50 struct fs_struct; 51 struct futex_pi_state; 52 struct io_context; 53 struct io_uring_task; 54 struct mempolicy; 55 struct nameidata; 56 struct nsproxy; 57 struct perf_event_context; 58 struct pid_namespace; 59 struct pipe_inode_info; 60 struct rcu_node; 61 struct reclaim_state; 62 struct robust_list_head; 63 struct root_domain; 64 struct rq; 65 struct sched_attr; 66 struct seq_file; 67 struct sighand_struct; 68 struct signal_struct; 69 struct task_delay_info; 70 struct task_group; 71 struct user_event_mm; 72 73 /* 74 * Task state bitmask. NOTE! These bits are also 75 * encoded in fs/proc/array.c: get_task_state(). 76 * 77 * We have two separate sets of flags: task->__state 78 * is about runnability, while task->exit_state are 79 * about the task exiting. Confusing, but this way 80 * modifying one set can't modify the other one by 81 * mistake. 82 */ 83 84 /* Used in tsk->__state: */ 85 #define TASK_RUNNING 0x00000000 86 #define TASK_INTERRUPTIBLE 0x00000001 87 #define TASK_UNINTERRUPTIBLE 0x00000002 88 #define __TASK_STOPPED 0x00000004 89 #define __TASK_TRACED 0x00000008 90 /* Used in tsk->exit_state: */ 91 #define EXIT_DEAD 0x00000010 92 #define EXIT_ZOMBIE 0x00000020 93 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 94 /* Used in tsk->__state again: */ 95 #define TASK_PARKED 0x00000040 96 #define TASK_DEAD 0x00000080 97 #define TASK_WAKEKILL 0x00000100 98 #define TASK_WAKING 0x00000200 99 #define TASK_NOLOAD 0x00000400 100 #define TASK_NEW 0x00000800 101 #define TASK_RTLOCK_WAIT 0x00001000 102 #define TASK_FREEZABLE 0x00002000 103 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP)) 104 #define TASK_FROZEN 0x00008000 105 #define TASK_STATE_MAX 0x00010000 106 107 #define TASK_ANY (TASK_STATE_MAX-1) 108 109 /* 110 * DO NOT ADD ANY NEW USERS ! 111 */ 112 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE) 113 114 /* Convenience macros for the sake of set_current_state: */ 115 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 116 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 117 #define TASK_TRACED __TASK_TRACED 118 119 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 120 121 /* Convenience macros for the sake of wake_up(): */ 122 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 123 124 /* get_task_state(): */ 125 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 126 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 127 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 128 TASK_PARKED) 129 130 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 131 132 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 133 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 134 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 135 136 /* 137 * Special states are those that do not use the normal wait-loop pattern. See 138 * the comment with set_special_state(). 139 */ 140 #define is_special_task_state(state) \ 141 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD)) 142 143 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 144 # define debug_normal_state_change(state_value) \ 145 do { \ 146 WARN_ON_ONCE(is_special_task_state(state_value)); \ 147 current->task_state_change = _THIS_IP_; \ 148 } while (0) 149 150 # define debug_special_state_change(state_value) \ 151 do { \ 152 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 153 current->task_state_change = _THIS_IP_; \ 154 } while (0) 155 156 # define debug_rtlock_wait_set_state() \ 157 do { \ 158 current->saved_state_change = current->task_state_change;\ 159 current->task_state_change = _THIS_IP_; \ 160 } while (0) 161 162 # define debug_rtlock_wait_restore_state() \ 163 do { \ 164 current->task_state_change = current->saved_state_change;\ 165 } while (0) 166 167 #else 168 # define debug_normal_state_change(cond) do { } while (0) 169 # define debug_special_state_change(cond) do { } while (0) 170 # define debug_rtlock_wait_set_state() do { } while (0) 171 # define debug_rtlock_wait_restore_state() do { } while (0) 172 #endif 173 174 /* 175 * set_current_state() includes a barrier so that the write of current->__state 176 * is correctly serialised wrt the caller's subsequent test of whether to 177 * actually sleep: 178 * 179 * for (;;) { 180 * set_current_state(TASK_UNINTERRUPTIBLE); 181 * if (CONDITION) 182 * break; 183 * 184 * schedule(); 185 * } 186 * __set_current_state(TASK_RUNNING); 187 * 188 * If the caller does not need such serialisation (because, for instance, the 189 * CONDITION test and condition change and wakeup are under the same lock) then 190 * use __set_current_state(). 191 * 192 * The above is typically ordered against the wakeup, which does: 193 * 194 * CONDITION = 1; 195 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 196 * 197 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 198 * accessing p->__state. 199 * 200 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is, 201 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 202 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 203 * 204 * However, with slightly different timing the wakeup TASK_RUNNING store can 205 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 206 * a problem either because that will result in one extra go around the loop 207 * and our @cond test will save the day. 208 * 209 * Also see the comments of try_to_wake_up(). 210 */ 211 #define __set_current_state(state_value) \ 212 do { \ 213 debug_normal_state_change((state_value)); \ 214 WRITE_ONCE(current->__state, (state_value)); \ 215 } while (0) 216 217 #define set_current_state(state_value) \ 218 do { \ 219 debug_normal_state_change((state_value)); \ 220 smp_store_mb(current->__state, (state_value)); \ 221 } while (0) 222 223 /* 224 * set_special_state() should be used for those states when the blocking task 225 * can not use the regular condition based wait-loop. In that case we must 226 * serialize against wakeups such that any possible in-flight TASK_RUNNING 227 * stores will not collide with our state change. 228 */ 229 #define set_special_state(state_value) \ 230 do { \ 231 unsigned long flags; /* may shadow */ \ 232 \ 233 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 234 debug_special_state_change((state_value)); \ 235 WRITE_ONCE(current->__state, (state_value)); \ 236 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 237 } while (0) 238 239 /* 240 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 241 * 242 * RT's spin/rwlock substitutions are state preserving. The state of the 243 * task when blocking on the lock is saved in task_struct::saved_state and 244 * restored after the lock has been acquired. These operations are 245 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 246 * lock related wakeups while the task is blocked on the lock are 247 * redirected to operate on task_struct::saved_state to ensure that these 248 * are not dropped. On restore task_struct::saved_state is set to 249 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 250 * 251 * The lock operation looks like this: 252 * 253 * current_save_and_set_rtlock_wait_state(); 254 * for (;;) { 255 * if (try_lock()) 256 * break; 257 * raw_spin_unlock_irq(&lock->wait_lock); 258 * schedule_rtlock(); 259 * raw_spin_lock_irq(&lock->wait_lock); 260 * set_current_state(TASK_RTLOCK_WAIT); 261 * } 262 * current_restore_rtlock_saved_state(); 263 */ 264 #define current_save_and_set_rtlock_wait_state() \ 265 do { \ 266 lockdep_assert_irqs_disabled(); \ 267 raw_spin_lock(¤t->pi_lock); \ 268 current->saved_state = current->__state; \ 269 debug_rtlock_wait_set_state(); \ 270 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 271 raw_spin_unlock(¤t->pi_lock); \ 272 } while (0); 273 274 #define current_restore_rtlock_saved_state() \ 275 do { \ 276 lockdep_assert_irqs_disabled(); \ 277 raw_spin_lock(¤t->pi_lock); \ 278 debug_rtlock_wait_restore_state(); \ 279 WRITE_ONCE(current->__state, current->saved_state); \ 280 current->saved_state = TASK_RUNNING; \ 281 raw_spin_unlock(¤t->pi_lock); \ 282 } while (0); 283 284 #define get_current_state() READ_ONCE(current->__state) 285 286 /* 287 * Define the task command name length as enum, then it can be visible to 288 * BPF programs. 289 */ 290 enum { 291 TASK_COMM_LEN = 16, 292 }; 293 294 extern void scheduler_tick(void); 295 296 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 297 298 extern long schedule_timeout(long timeout); 299 extern long schedule_timeout_interruptible(long timeout); 300 extern long schedule_timeout_killable(long timeout); 301 extern long schedule_timeout_uninterruptible(long timeout); 302 extern long schedule_timeout_idle(long timeout); 303 asmlinkage void schedule(void); 304 extern void schedule_preempt_disabled(void); 305 asmlinkage void preempt_schedule_irq(void); 306 #ifdef CONFIG_PREEMPT_RT 307 extern void schedule_rtlock(void); 308 #endif 309 310 extern int __must_check io_schedule_prepare(void); 311 extern void io_schedule_finish(int token); 312 extern long io_schedule_timeout(long timeout); 313 extern void io_schedule(void); 314 315 /** 316 * struct prev_cputime - snapshot of system and user cputime 317 * @utime: time spent in user mode 318 * @stime: time spent in system mode 319 * @lock: protects the above two fields 320 * 321 * Stores previous user/system time values such that we can guarantee 322 * monotonicity. 323 */ 324 struct prev_cputime { 325 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 326 u64 utime; 327 u64 stime; 328 raw_spinlock_t lock; 329 #endif 330 }; 331 332 enum vtime_state { 333 /* Task is sleeping or running in a CPU with VTIME inactive: */ 334 VTIME_INACTIVE = 0, 335 /* Task is idle */ 336 VTIME_IDLE, 337 /* Task runs in kernelspace in a CPU with VTIME active: */ 338 VTIME_SYS, 339 /* Task runs in userspace in a CPU with VTIME active: */ 340 VTIME_USER, 341 /* Task runs as guests in a CPU with VTIME active: */ 342 VTIME_GUEST, 343 }; 344 345 struct vtime { 346 seqcount_t seqcount; 347 unsigned long long starttime; 348 enum vtime_state state; 349 unsigned int cpu; 350 u64 utime; 351 u64 stime; 352 u64 gtime; 353 }; 354 355 /* 356 * Utilization clamp constraints. 357 * @UCLAMP_MIN: Minimum utilization 358 * @UCLAMP_MAX: Maximum utilization 359 * @UCLAMP_CNT: Utilization clamp constraints count 360 */ 361 enum uclamp_id { 362 UCLAMP_MIN = 0, 363 UCLAMP_MAX, 364 UCLAMP_CNT 365 }; 366 367 #ifdef CONFIG_SMP 368 extern struct root_domain def_root_domain; 369 extern struct mutex sched_domains_mutex; 370 #endif 371 372 struct sched_param { 373 int sched_priority; 374 }; 375 376 struct sched_info { 377 #ifdef CONFIG_SCHED_INFO 378 /* Cumulative counters: */ 379 380 /* # of times we have run on this CPU: */ 381 unsigned long pcount; 382 383 /* Time spent waiting on a runqueue: */ 384 unsigned long long run_delay; 385 386 /* Timestamps: */ 387 388 /* When did we last run on a CPU? */ 389 unsigned long long last_arrival; 390 391 /* When were we last queued to run? */ 392 unsigned long long last_queued; 393 394 #endif /* CONFIG_SCHED_INFO */ 395 }; 396 397 /* 398 * Integer metrics need fixed point arithmetic, e.g., sched/fair 399 * has a few: load, load_avg, util_avg, freq, and capacity. 400 * 401 * We define a basic fixed point arithmetic range, and then formalize 402 * all these metrics based on that basic range. 403 */ 404 # define SCHED_FIXEDPOINT_SHIFT 10 405 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 406 407 /* Increase resolution of cpu_capacity calculations */ 408 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 409 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 410 411 struct load_weight { 412 unsigned long weight; 413 u32 inv_weight; 414 }; 415 416 /** 417 * struct util_est - Estimation utilization of FAIR tasks 418 * @enqueued: instantaneous estimated utilization of a task/cpu 419 * @ewma: the Exponential Weighted Moving Average (EWMA) 420 * utilization of a task 421 * 422 * Support data structure to track an Exponential Weighted Moving Average 423 * (EWMA) of a FAIR task's utilization. New samples are added to the moving 424 * average each time a task completes an activation. Sample's weight is chosen 425 * so that the EWMA will be relatively insensitive to transient changes to the 426 * task's workload. 427 * 428 * The enqueued attribute has a slightly different meaning for tasks and cpus: 429 * - task: the task's util_avg at last task dequeue time 430 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU 431 * Thus, the util_est.enqueued of a task represents the contribution on the 432 * estimated utilization of the CPU where that task is currently enqueued. 433 * 434 * Only for tasks we track a moving average of the past instantaneous 435 * estimated utilization. This allows to absorb sporadic drops in utilization 436 * of an otherwise almost periodic task. 437 * 438 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 439 * updates. When a task is dequeued, its util_est should not be updated if its 440 * util_avg has not been updated in the meantime. 441 * This information is mapped into the MSB bit of util_est.enqueued at dequeue 442 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg 443 * for a task) it is safe to use MSB. 444 */ 445 struct util_est { 446 unsigned int enqueued; 447 unsigned int ewma; 448 #define UTIL_EST_WEIGHT_SHIFT 2 449 #define UTIL_AVG_UNCHANGED 0x80000000 450 } __attribute__((__aligned__(sizeof(u64)))); 451 452 /* 453 * The load/runnable/util_avg accumulates an infinite geometric series 454 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 455 * 456 * [load_avg definition] 457 * 458 * load_avg = runnable% * scale_load_down(load) 459 * 460 * [runnable_avg definition] 461 * 462 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 463 * 464 * [util_avg definition] 465 * 466 * util_avg = running% * SCHED_CAPACITY_SCALE 467 * 468 * where runnable% is the time ratio that a sched_entity is runnable and 469 * running% the time ratio that a sched_entity is running. 470 * 471 * For cfs_rq, they are the aggregated values of all runnable and blocked 472 * sched_entities. 473 * 474 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 475 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 476 * for computing those signals (see update_rq_clock_pelt()) 477 * 478 * N.B., the above ratios (runnable% and running%) themselves are in the 479 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 480 * to as large a range as necessary. This is for example reflected by 481 * util_avg's SCHED_CAPACITY_SCALE. 482 * 483 * [Overflow issue] 484 * 485 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 486 * with the highest load (=88761), always runnable on a single cfs_rq, 487 * and should not overflow as the number already hits PID_MAX_LIMIT. 488 * 489 * For all other cases (including 32-bit kernels), struct load_weight's 490 * weight will overflow first before we do, because: 491 * 492 * Max(load_avg) <= Max(load.weight) 493 * 494 * Then it is the load_weight's responsibility to consider overflow 495 * issues. 496 */ 497 struct sched_avg { 498 u64 last_update_time; 499 u64 load_sum; 500 u64 runnable_sum; 501 u32 util_sum; 502 u32 period_contrib; 503 unsigned long load_avg; 504 unsigned long runnable_avg; 505 unsigned long util_avg; 506 struct util_est util_est; 507 } ____cacheline_aligned; 508 509 struct sched_statistics { 510 #ifdef CONFIG_SCHEDSTATS 511 u64 wait_start; 512 u64 wait_max; 513 u64 wait_count; 514 u64 wait_sum; 515 u64 iowait_count; 516 u64 iowait_sum; 517 518 u64 sleep_start; 519 u64 sleep_max; 520 s64 sum_sleep_runtime; 521 522 u64 block_start; 523 u64 block_max; 524 s64 sum_block_runtime; 525 526 u64 exec_max; 527 u64 slice_max; 528 529 u64 nr_migrations_cold; 530 u64 nr_failed_migrations_affine; 531 u64 nr_failed_migrations_running; 532 u64 nr_failed_migrations_hot; 533 u64 nr_forced_migrations; 534 535 u64 nr_wakeups; 536 u64 nr_wakeups_sync; 537 u64 nr_wakeups_migrate; 538 u64 nr_wakeups_local; 539 u64 nr_wakeups_remote; 540 u64 nr_wakeups_affine; 541 u64 nr_wakeups_affine_attempts; 542 u64 nr_wakeups_passive; 543 u64 nr_wakeups_idle; 544 545 #ifdef CONFIG_SCHED_CORE 546 u64 core_forceidle_sum; 547 #endif 548 #endif /* CONFIG_SCHEDSTATS */ 549 } ____cacheline_aligned; 550 551 struct sched_entity { 552 /* For load-balancing: */ 553 struct load_weight load; 554 struct rb_node run_node; 555 u64 deadline; 556 u64 min_deadline; 557 558 struct list_head group_node; 559 unsigned int on_rq; 560 561 u64 exec_start; 562 u64 sum_exec_runtime; 563 u64 prev_sum_exec_runtime; 564 u64 vruntime; 565 s64 vlag; 566 u64 slice; 567 568 u64 nr_migrations; 569 570 #ifdef CONFIG_FAIR_GROUP_SCHED 571 int depth; 572 struct sched_entity *parent; 573 /* rq on which this entity is (to be) queued: */ 574 struct cfs_rq *cfs_rq; 575 /* rq "owned" by this entity/group: */ 576 struct cfs_rq *my_q; 577 /* cached value of my_q->h_nr_running */ 578 unsigned long runnable_weight; 579 #endif 580 581 #ifdef CONFIG_SMP 582 /* 583 * Per entity load average tracking. 584 * 585 * Put into separate cache line so it does not 586 * collide with read-mostly values above. 587 */ 588 struct sched_avg avg; 589 #endif 590 }; 591 592 struct sched_rt_entity { 593 struct list_head run_list; 594 unsigned long timeout; 595 unsigned long watchdog_stamp; 596 unsigned int time_slice; 597 unsigned short on_rq; 598 unsigned short on_list; 599 600 struct sched_rt_entity *back; 601 #ifdef CONFIG_RT_GROUP_SCHED 602 struct sched_rt_entity *parent; 603 /* rq on which this entity is (to be) queued: */ 604 struct rt_rq *rt_rq; 605 /* rq "owned" by this entity/group: */ 606 struct rt_rq *my_q; 607 #endif 608 } __randomize_layout; 609 610 struct sched_dl_entity { 611 struct rb_node rb_node; 612 613 /* 614 * Original scheduling parameters. Copied here from sched_attr 615 * during sched_setattr(), they will remain the same until 616 * the next sched_setattr(). 617 */ 618 u64 dl_runtime; /* Maximum runtime for each instance */ 619 u64 dl_deadline; /* Relative deadline of each instance */ 620 u64 dl_period; /* Separation of two instances (period) */ 621 u64 dl_bw; /* dl_runtime / dl_period */ 622 u64 dl_density; /* dl_runtime / dl_deadline */ 623 624 /* 625 * Actual scheduling parameters. Initialized with the values above, 626 * they are continuously updated during task execution. Note that 627 * the remaining runtime could be < 0 in case we are in overrun. 628 */ 629 s64 runtime; /* Remaining runtime for this instance */ 630 u64 deadline; /* Absolute deadline for this instance */ 631 unsigned int flags; /* Specifying the scheduler behaviour */ 632 633 /* 634 * Some bool flags: 635 * 636 * @dl_throttled tells if we exhausted the runtime. If so, the 637 * task has to wait for a replenishment to be performed at the 638 * next firing of dl_timer. 639 * 640 * @dl_yielded tells if task gave up the CPU before consuming 641 * all its available runtime during the last job. 642 * 643 * @dl_non_contending tells if the task is inactive while still 644 * contributing to the active utilization. In other words, it 645 * indicates if the inactive timer has been armed and its handler 646 * has not been executed yet. This flag is useful to avoid race 647 * conditions between the inactive timer handler and the wakeup 648 * code. 649 * 650 * @dl_overrun tells if the task asked to be informed about runtime 651 * overruns. 652 */ 653 unsigned int dl_throttled : 1; 654 unsigned int dl_yielded : 1; 655 unsigned int dl_non_contending : 1; 656 unsigned int dl_overrun : 1; 657 658 /* 659 * Bandwidth enforcement timer. Each -deadline task has its 660 * own bandwidth to be enforced, thus we need one timer per task. 661 */ 662 struct hrtimer dl_timer; 663 664 /* 665 * Inactive timer, responsible for decreasing the active utilization 666 * at the "0-lag time". When a -deadline task blocks, it contributes 667 * to GRUB's active utilization until the "0-lag time", hence a 668 * timer is needed to decrease the active utilization at the correct 669 * time. 670 */ 671 struct hrtimer inactive_timer; 672 673 #ifdef CONFIG_RT_MUTEXES 674 /* 675 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 676 * pi_se points to the donor, otherwise points to the dl_se it belongs 677 * to (the original one/itself). 678 */ 679 struct sched_dl_entity *pi_se; 680 #endif 681 }; 682 683 #ifdef CONFIG_UCLAMP_TASK 684 /* Number of utilization clamp buckets (shorter alias) */ 685 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 686 687 /* 688 * Utilization clamp for a scheduling entity 689 * @value: clamp value "assigned" to a se 690 * @bucket_id: bucket index corresponding to the "assigned" value 691 * @active: the se is currently refcounted in a rq's bucket 692 * @user_defined: the requested clamp value comes from user-space 693 * 694 * The bucket_id is the index of the clamp bucket matching the clamp value 695 * which is pre-computed and stored to avoid expensive integer divisions from 696 * the fast path. 697 * 698 * The active bit is set whenever a task has got an "effective" value assigned, 699 * which can be different from the clamp value "requested" from user-space. 700 * This allows to know a task is refcounted in the rq's bucket corresponding 701 * to the "effective" bucket_id. 702 * 703 * The user_defined bit is set whenever a task has got a task-specific clamp 704 * value requested from userspace, i.e. the system defaults apply to this task 705 * just as a restriction. This allows to relax default clamps when a less 706 * restrictive task-specific value has been requested, thus allowing to 707 * implement a "nice" semantic. For example, a task running with a 20% 708 * default boost can still drop its own boosting to 0%. 709 */ 710 struct uclamp_se { 711 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 712 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 713 unsigned int active : 1; 714 unsigned int user_defined : 1; 715 }; 716 #endif /* CONFIG_UCLAMP_TASK */ 717 718 union rcu_special { 719 struct { 720 u8 blocked; 721 u8 need_qs; 722 u8 exp_hint; /* Hint for performance. */ 723 u8 need_mb; /* Readers need smp_mb(). */ 724 } b; /* Bits. */ 725 u32 s; /* Set of bits. */ 726 }; 727 728 enum perf_event_task_context { 729 perf_invalid_context = -1, 730 perf_hw_context = 0, 731 perf_sw_context, 732 perf_nr_task_contexts, 733 }; 734 735 struct wake_q_node { 736 struct wake_q_node *next; 737 }; 738 739 struct kmap_ctrl { 740 #ifdef CONFIG_KMAP_LOCAL 741 int idx; 742 pte_t pteval[KM_MAX_IDX]; 743 #endif 744 }; 745 746 struct task_struct { 747 #ifdef CONFIG_THREAD_INFO_IN_TASK 748 /* 749 * For reasons of header soup (see current_thread_info()), this 750 * must be the first element of task_struct. 751 */ 752 struct thread_info thread_info; 753 #endif 754 unsigned int __state; 755 756 /* saved state for "spinlock sleepers" */ 757 unsigned int saved_state; 758 759 /* 760 * This begins the randomizable portion of task_struct. Only 761 * scheduling-critical items should be added above here. 762 */ 763 randomized_struct_fields_start 764 765 void *stack; 766 refcount_t usage; 767 /* Per task flags (PF_*), defined further below: */ 768 unsigned int flags; 769 unsigned int ptrace; 770 771 #ifdef CONFIG_SMP 772 int on_cpu; 773 struct __call_single_node wake_entry; 774 unsigned int wakee_flips; 775 unsigned long wakee_flip_decay_ts; 776 struct task_struct *last_wakee; 777 778 /* 779 * recent_used_cpu is initially set as the last CPU used by a task 780 * that wakes affine another task. Waker/wakee relationships can 781 * push tasks around a CPU where each wakeup moves to the next one. 782 * Tracking a recently used CPU allows a quick search for a recently 783 * used CPU that may be idle. 784 */ 785 int recent_used_cpu; 786 int wake_cpu; 787 #endif 788 int on_rq; 789 790 int prio; 791 int static_prio; 792 int normal_prio; 793 unsigned int rt_priority; 794 795 struct sched_entity se; 796 struct sched_rt_entity rt; 797 struct sched_dl_entity dl; 798 const struct sched_class *sched_class; 799 800 #ifdef CONFIG_SCHED_CORE 801 struct rb_node core_node; 802 unsigned long core_cookie; 803 unsigned int core_occupation; 804 #endif 805 806 #ifdef CONFIG_CGROUP_SCHED 807 struct task_group *sched_task_group; 808 #endif 809 810 #ifdef CONFIG_UCLAMP_TASK 811 /* 812 * Clamp values requested for a scheduling entity. 813 * Must be updated with task_rq_lock() held. 814 */ 815 struct uclamp_se uclamp_req[UCLAMP_CNT]; 816 /* 817 * Effective clamp values used for a scheduling entity. 818 * Must be updated with task_rq_lock() held. 819 */ 820 struct uclamp_se uclamp[UCLAMP_CNT]; 821 #endif 822 823 struct sched_statistics stats; 824 825 #ifdef CONFIG_PREEMPT_NOTIFIERS 826 /* List of struct preempt_notifier: */ 827 struct hlist_head preempt_notifiers; 828 #endif 829 830 #ifdef CONFIG_BLK_DEV_IO_TRACE 831 unsigned int btrace_seq; 832 #endif 833 834 unsigned int policy; 835 int nr_cpus_allowed; 836 const cpumask_t *cpus_ptr; 837 cpumask_t *user_cpus_ptr; 838 cpumask_t cpus_mask; 839 void *migration_pending; 840 #ifdef CONFIG_SMP 841 unsigned short migration_disabled; 842 #endif 843 unsigned short migration_flags; 844 845 #ifdef CONFIG_PREEMPT_RCU 846 int rcu_read_lock_nesting; 847 union rcu_special rcu_read_unlock_special; 848 struct list_head rcu_node_entry; 849 struct rcu_node *rcu_blocked_node; 850 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 851 852 #ifdef CONFIG_TASKS_RCU 853 unsigned long rcu_tasks_nvcsw; 854 u8 rcu_tasks_holdout; 855 u8 rcu_tasks_idx; 856 int rcu_tasks_idle_cpu; 857 struct list_head rcu_tasks_holdout_list; 858 #endif /* #ifdef CONFIG_TASKS_RCU */ 859 860 #ifdef CONFIG_TASKS_TRACE_RCU 861 int trc_reader_nesting; 862 int trc_ipi_to_cpu; 863 union rcu_special trc_reader_special; 864 struct list_head trc_holdout_list; 865 struct list_head trc_blkd_node; 866 int trc_blkd_cpu; 867 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 868 869 struct sched_info sched_info; 870 871 struct list_head tasks; 872 #ifdef CONFIG_SMP 873 struct plist_node pushable_tasks; 874 struct rb_node pushable_dl_tasks; 875 #endif 876 877 struct mm_struct *mm; 878 struct mm_struct *active_mm; 879 struct address_space *faults_disabled_mapping; 880 881 int exit_state; 882 int exit_code; 883 int exit_signal; 884 /* The signal sent when the parent dies: */ 885 int pdeath_signal; 886 /* JOBCTL_*, siglock protected: */ 887 unsigned long jobctl; 888 889 /* Used for emulating ABI behavior of previous Linux versions: */ 890 unsigned int personality; 891 892 /* Scheduler bits, serialized by scheduler locks: */ 893 unsigned sched_reset_on_fork:1; 894 unsigned sched_contributes_to_load:1; 895 unsigned sched_migrated:1; 896 897 /* Force alignment to the next boundary: */ 898 unsigned :0; 899 900 /* Unserialized, strictly 'current' */ 901 902 /* 903 * This field must not be in the scheduler word above due to wakelist 904 * queueing no longer being serialized by p->on_cpu. However: 905 * 906 * p->XXX = X; ttwu() 907 * schedule() if (p->on_rq && ..) // false 908 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 909 * deactivate_task() ttwu_queue_wakelist()) 910 * p->on_rq = 0; p->sched_remote_wakeup = Y; 911 * 912 * guarantees all stores of 'current' are visible before 913 * ->sched_remote_wakeup gets used, so it can be in this word. 914 */ 915 unsigned sched_remote_wakeup:1; 916 #ifdef CONFIG_RT_MUTEXES 917 unsigned sched_rt_mutex:1; 918 #endif 919 920 /* Bit to tell LSMs we're in execve(): */ 921 unsigned in_execve:1; 922 unsigned in_iowait:1; 923 #ifndef TIF_RESTORE_SIGMASK 924 unsigned restore_sigmask:1; 925 #endif 926 #ifdef CONFIG_MEMCG 927 unsigned in_user_fault:1; 928 #endif 929 #ifdef CONFIG_LRU_GEN 930 /* whether the LRU algorithm may apply to this access */ 931 unsigned in_lru_fault:1; 932 #endif 933 #ifdef CONFIG_COMPAT_BRK 934 unsigned brk_randomized:1; 935 #endif 936 #ifdef CONFIG_CGROUPS 937 /* disallow userland-initiated cgroup migration */ 938 unsigned no_cgroup_migration:1; 939 /* task is frozen/stopped (used by the cgroup freezer) */ 940 unsigned frozen:1; 941 #endif 942 #ifdef CONFIG_BLK_CGROUP 943 unsigned use_memdelay:1; 944 #endif 945 #ifdef CONFIG_PSI 946 /* Stalled due to lack of memory */ 947 unsigned in_memstall:1; 948 #endif 949 #ifdef CONFIG_PAGE_OWNER 950 /* Used by page_owner=on to detect recursion in page tracking. */ 951 unsigned in_page_owner:1; 952 #endif 953 #ifdef CONFIG_EVENTFD 954 /* Recursion prevention for eventfd_signal() */ 955 unsigned in_eventfd:1; 956 #endif 957 #ifdef CONFIG_IOMMU_SVA 958 unsigned pasid_activated:1; 959 #endif 960 #ifdef CONFIG_CPU_SUP_INTEL 961 unsigned reported_split_lock:1; 962 #endif 963 #ifdef CONFIG_TASK_DELAY_ACCT 964 /* delay due to memory thrashing */ 965 unsigned in_thrashing:1; 966 #endif 967 968 unsigned long atomic_flags; /* Flags requiring atomic access. */ 969 970 struct restart_block restart_block; 971 972 pid_t pid; 973 pid_t tgid; 974 975 #ifdef CONFIG_STACKPROTECTOR 976 /* Canary value for the -fstack-protector GCC feature: */ 977 unsigned long stack_canary; 978 #endif 979 /* 980 * Pointers to the (original) parent process, youngest child, younger sibling, 981 * older sibling, respectively. (p->father can be replaced with 982 * p->real_parent->pid) 983 */ 984 985 /* Real parent process: */ 986 struct task_struct __rcu *real_parent; 987 988 /* Recipient of SIGCHLD, wait4() reports: */ 989 struct task_struct __rcu *parent; 990 991 /* 992 * Children/sibling form the list of natural children: 993 */ 994 struct list_head children; 995 struct list_head sibling; 996 struct task_struct *group_leader; 997 998 /* 999 * 'ptraced' is the list of tasks this task is using ptrace() on. 1000 * 1001 * This includes both natural children and PTRACE_ATTACH targets. 1002 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1003 */ 1004 struct list_head ptraced; 1005 struct list_head ptrace_entry; 1006 1007 /* PID/PID hash table linkage. */ 1008 struct pid *thread_pid; 1009 struct hlist_node pid_links[PIDTYPE_MAX]; 1010 struct list_head thread_group; 1011 struct list_head thread_node; 1012 1013 struct completion *vfork_done; 1014 1015 /* CLONE_CHILD_SETTID: */ 1016 int __user *set_child_tid; 1017 1018 /* CLONE_CHILD_CLEARTID: */ 1019 int __user *clear_child_tid; 1020 1021 /* PF_KTHREAD | PF_IO_WORKER */ 1022 void *worker_private; 1023 1024 u64 utime; 1025 u64 stime; 1026 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1027 u64 utimescaled; 1028 u64 stimescaled; 1029 #endif 1030 u64 gtime; 1031 struct prev_cputime prev_cputime; 1032 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1033 struct vtime vtime; 1034 #endif 1035 1036 #ifdef CONFIG_NO_HZ_FULL 1037 atomic_t tick_dep_mask; 1038 #endif 1039 /* Context switch counts: */ 1040 unsigned long nvcsw; 1041 unsigned long nivcsw; 1042 1043 /* Monotonic time in nsecs: */ 1044 u64 start_time; 1045 1046 /* Boot based time in nsecs: */ 1047 u64 start_boottime; 1048 1049 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1050 unsigned long min_flt; 1051 unsigned long maj_flt; 1052 1053 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1054 struct posix_cputimers posix_cputimers; 1055 1056 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1057 struct posix_cputimers_work posix_cputimers_work; 1058 #endif 1059 1060 /* Process credentials: */ 1061 1062 /* Tracer's credentials at attach: */ 1063 const struct cred __rcu *ptracer_cred; 1064 1065 /* Objective and real subjective task credentials (COW): */ 1066 const struct cred __rcu *real_cred; 1067 1068 /* Effective (overridable) subjective task credentials (COW): */ 1069 const struct cred __rcu *cred; 1070 1071 #ifdef CONFIG_KEYS 1072 /* Cached requested key. */ 1073 struct key *cached_requested_key; 1074 #endif 1075 1076 /* 1077 * executable name, excluding path. 1078 * 1079 * - normally initialized setup_new_exec() 1080 * - access it with [gs]et_task_comm() 1081 * - lock it with task_lock() 1082 */ 1083 char comm[TASK_COMM_LEN]; 1084 1085 struct nameidata *nameidata; 1086 1087 #ifdef CONFIG_SYSVIPC 1088 struct sysv_sem sysvsem; 1089 struct sysv_shm sysvshm; 1090 #endif 1091 #ifdef CONFIG_DETECT_HUNG_TASK 1092 unsigned long last_switch_count; 1093 unsigned long last_switch_time; 1094 #endif 1095 /* Filesystem information: */ 1096 struct fs_struct *fs; 1097 1098 /* Open file information: */ 1099 struct files_struct *files; 1100 1101 #ifdef CONFIG_IO_URING 1102 struct io_uring_task *io_uring; 1103 #endif 1104 1105 /* Namespaces: */ 1106 struct nsproxy *nsproxy; 1107 1108 /* Signal handlers: */ 1109 struct signal_struct *signal; 1110 struct sighand_struct __rcu *sighand; 1111 sigset_t blocked; 1112 sigset_t real_blocked; 1113 /* Restored if set_restore_sigmask() was used: */ 1114 sigset_t saved_sigmask; 1115 struct sigpending pending; 1116 unsigned long sas_ss_sp; 1117 size_t sas_ss_size; 1118 unsigned int sas_ss_flags; 1119 1120 struct callback_head *task_works; 1121 1122 #ifdef CONFIG_AUDIT 1123 #ifdef CONFIG_AUDITSYSCALL 1124 struct audit_context *audit_context; 1125 #endif 1126 kuid_t loginuid; 1127 unsigned int sessionid; 1128 #endif 1129 struct seccomp seccomp; 1130 struct syscall_user_dispatch syscall_dispatch; 1131 1132 /* Thread group tracking: */ 1133 u64 parent_exec_id; 1134 u64 self_exec_id; 1135 1136 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1137 spinlock_t alloc_lock; 1138 1139 /* Protection of the PI data structures: */ 1140 raw_spinlock_t pi_lock; 1141 1142 struct wake_q_node wake_q; 1143 1144 #ifdef CONFIG_RT_MUTEXES 1145 /* PI waiters blocked on a rt_mutex held by this task: */ 1146 struct rb_root_cached pi_waiters; 1147 /* Updated under owner's pi_lock and rq lock */ 1148 struct task_struct *pi_top_task; 1149 /* Deadlock detection and priority inheritance handling: */ 1150 struct rt_mutex_waiter *pi_blocked_on; 1151 #endif 1152 1153 #ifdef CONFIG_DEBUG_MUTEXES 1154 /* Mutex deadlock detection: */ 1155 struct mutex_waiter *blocked_on; 1156 #endif 1157 1158 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1159 int non_block_count; 1160 #endif 1161 1162 #ifdef CONFIG_TRACE_IRQFLAGS 1163 struct irqtrace_events irqtrace; 1164 unsigned int hardirq_threaded; 1165 u64 hardirq_chain_key; 1166 int softirqs_enabled; 1167 int softirq_context; 1168 int irq_config; 1169 #endif 1170 #ifdef CONFIG_PREEMPT_RT 1171 int softirq_disable_cnt; 1172 #endif 1173 1174 #ifdef CONFIG_LOCKDEP 1175 # define MAX_LOCK_DEPTH 48UL 1176 u64 curr_chain_key; 1177 int lockdep_depth; 1178 unsigned int lockdep_recursion; 1179 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1180 #endif 1181 1182 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1183 unsigned int in_ubsan; 1184 #endif 1185 1186 /* Journalling filesystem info: */ 1187 void *journal_info; 1188 1189 /* Stacked block device info: */ 1190 struct bio_list *bio_list; 1191 1192 /* Stack plugging: */ 1193 struct blk_plug *plug; 1194 1195 /* VM state: */ 1196 struct reclaim_state *reclaim_state; 1197 1198 struct io_context *io_context; 1199 1200 #ifdef CONFIG_COMPACTION 1201 struct capture_control *capture_control; 1202 #endif 1203 /* Ptrace state: */ 1204 unsigned long ptrace_message; 1205 kernel_siginfo_t *last_siginfo; 1206 1207 struct task_io_accounting ioac; 1208 #ifdef CONFIG_PSI 1209 /* Pressure stall state */ 1210 unsigned int psi_flags; 1211 #endif 1212 #ifdef CONFIG_TASK_XACCT 1213 /* Accumulated RSS usage: */ 1214 u64 acct_rss_mem1; 1215 /* Accumulated virtual memory usage: */ 1216 u64 acct_vm_mem1; 1217 /* stime + utime since last update: */ 1218 u64 acct_timexpd; 1219 #endif 1220 #ifdef CONFIG_CPUSETS 1221 /* Protected by ->alloc_lock: */ 1222 nodemask_t mems_allowed; 1223 /* Sequence number to catch updates: */ 1224 seqcount_spinlock_t mems_allowed_seq; 1225 int cpuset_mem_spread_rotor; 1226 int cpuset_slab_spread_rotor; 1227 #endif 1228 #ifdef CONFIG_CGROUPS 1229 /* Control Group info protected by css_set_lock: */ 1230 struct css_set __rcu *cgroups; 1231 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1232 struct list_head cg_list; 1233 #endif 1234 #ifdef CONFIG_X86_CPU_RESCTRL 1235 u32 closid; 1236 u32 rmid; 1237 #endif 1238 #ifdef CONFIG_FUTEX 1239 struct robust_list_head __user *robust_list; 1240 #ifdef CONFIG_COMPAT 1241 struct compat_robust_list_head __user *compat_robust_list; 1242 #endif 1243 struct list_head pi_state_list; 1244 struct futex_pi_state *pi_state_cache; 1245 struct mutex futex_exit_mutex; 1246 unsigned int futex_state; 1247 #endif 1248 #ifdef CONFIG_PERF_EVENTS 1249 struct perf_event_context *perf_event_ctxp; 1250 struct mutex perf_event_mutex; 1251 struct list_head perf_event_list; 1252 #endif 1253 #ifdef CONFIG_DEBUG_PREEMPT 1254 unsigned long preempt_disable_ip; 1255 #endif 1256 #ifdef CONFIG_NUMA 1257 /* Protected by alloc_lock: */ 1258 struct mempolicy *mempolicy; 1259 short il_prev; 1260 short pref_node_fork; 1261 #endif 1262 #ifdef CONFIG_NUMA_BALANCING 1263 int numa_scan_seq; 1264 unsigned int numa_scan_period; 1265 unsigned int numa_scan_period_max; 1266 int numa_preferred_nid; 1267 unsigned long numa_migrate_retry; 1268 /* Migration stamp: */ 1269 u64 node_stamp; 1270 u64 last_task_numa_placement; 1271 u64 last_sum_exec_runtime; 1272 struct callback_head numa_work; 1273 1274 /* 1275 * This pointer is only modified for current in syscall and 1276 * pagefault context (and for tasks being destroyed), so it can be read 1277 * from any of the following contexts: 1278 * - RCU read-side critical section 1279 * - current->numa_group from everywhere 1280 * - task's runqueue locked, task not running 1281 */ 1282 struct numa_group __rcu *numa_group; 1283 1284 /* 1285 * numa_faults is an array split into four regions: 1286 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1287 * in this precise order. 1288 * 1289 * faults_memory: Exponential decaying average of faults on a per-node 1290 * basis. Scheduling placement decisions are made based on these 1291 * counts. The values remain static for the duration of a PTE scan. 1292 * faults_cpu: Track the nodes the process was running on when a NUMA 1293 * hinting fault was incurred. 1294 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1295 * during the current scan window. When the scan completes, the counts 1296 * in faults_memory and faults_cpu decay and these values are copied. 1297 */ 1298 unsigned long *numa_faults; 1299 unsigned long total_numa_faults; 1300 1301 /* 1302 * numa_faults_locality tracks if faults recorded during the last 1303 * scan window were remote/local or failed to migrate. The task scan 1304 * period is adapted based on the locality of the faults with different 1305 * weights depending on whether they were shared or private faults 1306 */ 1307 unsigned long numa_faults_locality[3]; 1308 1309 unsigned long numa_pages_migrated; 1310 #endif /* CONFIG_NUMA_BALANCING */ 1311 1312 #ifdef CONFIG_RSEQ 1313 struct rseq __user *rseq; 1314 u32 rseq_len; 1315 u32 rseq_sig; 1316 /* 1317 * RmW on rseq_event_mask must be performed atomically 1318 * with respect to preemption. 1319 */ 1320 unsigned long rseq_event_mask; 1321 #endif 1322 1323 #ifdef CONFIG_SCHED_MM_CID 1324 int mm_cid; /* Current cid in mm */ 1325 int last_mm_cid; /* Most recent cid in mm */ 1326 int migrate_from_cpu; 1327 int mm_cid_active; /* Whether cid bitmap is active */ 1328 struct callback_head cid_work; 1329 #endif 1330 1331 struct tlbflush_unmap_batch tlb_ubc; 1332 1333 /* Cache last used pipe for splice(): */ 1334 struct pipe_inode_info *splice_pipe; 1335 1336 struct page_frag task_frag; 1337 1338 #ifdef CONFIG_TASK_DELAY_ACCT 1339 struct task_delay_info *delays; 1340 #endif 1341 1342 #ifdef CONFIG_FAULT_INJECTION 1343 int make_it_fail; 1344 unsigned int fail_nth; 1345 #endif 1346 /* 1347 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1348 * balance_dirty_pages() for a dirty throttling pause: 1349 */ 1350 int nr_dirtied; 1351 int nr_dirtied_pause; 1352 /* Start of a write-and-pause period: */ 1353 unsigned long dirty_paused_when; 1354 1355 #ifdef CONFIG_LATENCYTOP 1356 int latency_record_count; 1357 struct latency_record latency_record[LT_SAVECOUNT]; 1358 #endif 1359 /* 1360 * Time slack values; these are used to round up poll() and 1361 * select() etc timeout values. These are in nanoseconds. 1362 */ 1363 u64 timer_slack_ns; 1364 u64 default_timer_slack_ns; 1365 1366 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1367 unsigned int kasan_depth; 1368 #endif 1369 1370 #ifdef CONFIG_KCSAN 1371 struct kcsan_ctx kcsan_ctx; 1372 #ifdef CONFIG_TRACE_IRQFLAGS 1373 struct irqtrace_events kcsan_save_irqtrace; 1374 #endif 1375 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1376 int kcsan_stack_depth; 1377 #endif 1378 #endif 1379 1380 #ifdef CONFIG_KMSAN 1381 struct kmsan_ctx kmsan_ctx; 1382 #endif 1383 1384 #if IS_ENABLED(CONFIG_KUNIT) 1385 struct kunit *kunit_test; 1386 #endif 1387 1388 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1389 /* Index of current stored address in ret_stack: */ 1390 int curr_ret_stack; 1391 int curr_ret_depth; 1392 1393 /* Stack of return addresses for return function tracing: */ 1394 struct ftrace_ret_stack *ret_stack; 1395 1396 /* Timestamp for last schedule: */ 1397 unsigned long long ftrace_timestamp; 1398 1399 /* 1400 * Number of functions that haven't been traced 1401 * because of depth overrun: 1402 */ 1403 atomic_t trace_overrun; 1404 1405 /* Pause tracing: */ 1406 atomic_t tracing_graph_pause; 1407 #endif 1408 1409 #ifdef CONFIG_TRACING 1410 /* Bitmask and counter of trace recursion: */ 1411 unsigned long trace_recursion; 1412 #endif /* CONFIG_TRACING */ 1413 1414 #ifdef CONFIG_KCOV 1415 /* See kernel/kcov.c for more details. */ 1416 1417 /* Coverage collection mode enabled for this task (0 if disabled): */ 1418 unsigned int kcov_mode; 1419 1420 /* Size of the kcov_area: */ 1421 unsigned int kcov_size; 1422 1423 /* Buffer for coverage collection: */ 1424 void *kcov_area; 1425 1426 /* KCOV descriptor wired with this task or NULL: */ 1427 struct kcov *kcov; 1428 1429 /* KCOV common handle for remote coverage collection: */ 1430 u64 kcov_handle; 1431 1432 /* KCOV sequence number: */ 1433 int kcov_sequence; 1434 1435 /* Collect coverage from softirq context: */ 1436 unsigned int kcov_softirq; 1437 #endif 1438 1439 #ifdef CONFIG_MEMCG 1440 struct mem_cgroup *memcg_in_oom; 1441 gfp_t memcg_oom_gfp_mask; 1442 int memcg_oom_order; 1443 1444 /* Number of pages to reclaim on returning to userland: */ 1445 unsigned int memcg_nr_pages_over_high; 1446 1447 /* Used by memcontrol for targeted memcg charge: */ 1448 struct mem_cgroup *active_memcg; 1449 #endif 1450 1451 #ifdef CONFIG_BLK_CGROUP 1452 struct gendisk *throttle_disk; 1453 #endif 1454 1455 #ifdef CONFIG_UPROBES 1456 struct uprobe_task *utask; 1457 #endif 1458 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1459 unsigned int sequential_io; 1460 unsigned int sequential_io_avg; 1461 #endif 1462 struct kmap_ctrl kmap_ctrl; 1463 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1464 unsigned long task_state_change; 1465 # ifdef CONFIG_PREEMPT_RT 1466 unsigned long saved_state_change; 1467 # endif 1468 #endif 1469 struct rcu_head rcu; 1470 refcount_t rcu_users; 1471 int pagefault_disabled; 1472 #ifdef CONFIG_MMU 1473 struct task_struct *oom_reaper_list; 1474 struct timer_list oom_reaper_timer; 1475 #endif 1476 #ifdef CONFIG_VMAP_STACK 1477 struct vm_struct *stack_vm_area; 1478 #endif 1479 #ifdef CONFIG_THREAD_INFO_IN_TASK 1480 /* A live task holds one reference: */ 1481 refcount_t stack_refcount; 1482 #endif 1483 #ifdef CONFIG_LIVEPATCH 1484 int patch_state; 1485 #endif 1486 #ifdef CONFIG_SECURITY 1487 /* Used by LSM modules for access restriction: */ 1488 void *security; 1489 #endif 1490 #ifdef CONFIG_BPF_SYSCALL 1491 /* Used by BPF task local storage */ 1492 struct bpf_local_storage __rcu *bpf_storage; 1493 /* Used for BPF run context */ 1494 struct bpf_run_ctx *bpf_ctx; 1495 #endif 1496 1497 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1498 unsigned long lowest_stack; 1499 unsigned long prev_lowest_stack; 1500 #endif 1501 1502 #ifdef CONFIG_X86_MCE 1503 void __user *mce_vaddr; 1504 __u64 mce_kflags; 1505 u64 mce_addr; 1506 __u64 mce_ripv : 1, 1507 mce_whole_page : 1, 1508 __mce_reserved : 62; 1509 struct callback_head mce_kill_me; 1510 int mce_count; 1511 #endif 1512 1513 #ifdef CONFIG_KRETPROBES 1514 struct llist_head kretprobe_instances; 1515 #endif 1516 #ifdef CONFIG_RETHOOK 1517 struct llist_head rethooks; 1518 #endif 1519 1520 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1521 /* 1522 * If L1D flush is supported on mm context switch 1523 * then we use this callback head to queue kill work 1524 * to kill tasks that are not running on SMT disabled 1525 * cores 1526 */ 1527 struct callback_head l1d_flush_kill; 1528 #endif 1529 1530 #ifdef CONFIG_RV 1531 /* 1532 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS. 1533 * If we find justification for more monitors, we can think 1534 * about adding more or developing a dynamic method. So far, 1535 * none of these are justified. 1536 */ 1537 union rv_task_monitor rv[RV_PER_TASK_MONITORS]; 1538 #endif 1539 1540 #ifdef CONFIG_USER_EVENTS 1541 struct user_event_mm *user_event_mm; 1542 #endif 1543 1544 /* 1545 * New fields for task_struct should be added above here, so that 1546 * they are included in the randomized portion of task_struct. 1547 */ 1548 randomized_struct_fields_end 1549 1550 /* CPU-specific state of this task: */ 1551 struct thread_struct thread; 1552 1553 /* 1554 * WARNING: on x86, 'thread_struct' contains a variable-sized 1555 * structure. It *MUST* be at the end of 'task_struct'. 1556 * 1557 * Do not put anything below here! 1558 */ 1559 }; 1560 1561 static inline struct pid *task_pid(struct task_struct *task) 1562 { 1563 return task->thread_pid; 1564 } 1565 1566 /* 1567 * the helpers to get the task's different pids as they are seen 1568 * from various namespaces 1569 * 1570 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1571 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1572 * current. 1573 * task_xid_nr_ns() : id seen from the ns specified; 1574 * 1575 * see also pid_nr() etc in include/linux/pid.h 1576 */ 1577 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1578 1579 static inline pid_t task_pid_nr(struct task_struct *tsk) 1580 { 1581 return tsk->pid; 1582 } 1583 1584 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1585 { 1586 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1587 } 1588 1589 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1590 { 1591 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1592 } 1593 1594 1595 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1596 { 1597 return tsk->tgid; 1598 } 1599 1600 /** 1601 * pid_alive - check that a task structure is not stale 1602 * @p: Task structure to be checked. 1603 * 1604 * Test if a process is not yet dead (at most zombie state) 1605 * If pid_alive fails, then pointers within the task structure 1606 * can be stale and must not be dereferenced. 1607 * 1608 * Return: 1 if the process is alive. 0 otherwise. 1609 */ 1610 static inline int pid_alive(const struct task_struct *p) 1611 { 1612 return p->thread_pid != NULL; 1613 } 1614 1615 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1616 { 1617 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1618 } 1619 1620 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1621 { 1622 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1623 } 1624 1625 1626 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1627 { 1628 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1629 } 1630 1631 static inline pid_t task_session_vnr(struct task_struct *tsk) 1632 { 1633 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1634 } 1635 1636 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1637 { 1638 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns); 1639 } 1640 1641 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1642 { 1643 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL); 1644 } 1645 1646 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1647 { 1648 pid_t pid = 0; 1649 1650 rcu_read_lock(); 1651 if (pid_alive(tsk)) 1652 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1653 rcu_read_unlock(); 1654 1655 return pid; 1656 } 1657 1658 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1659 { 1660 return task_ppid_nr_ns(tsk, &init_pid_ns); 1661 } 1662 1663 /* Obsolete, do not use: */ 1664 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1665 { 1666 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1667 } 1668 1669 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1670 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1671 1672 static inline unsigned int __task_state_index(unsigned int tsk_state, 1673 unsigned int tsk_exit_state) 1674 { 1675 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1676 1677 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1678 1679 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1680 state = TASK_REPORT_IDLE; 1681 1682 /* 1683 * We're lying here, but rather than expose a completely new task state 1684 * to userspace, we can make this appear as if the task has gone through 1685 * a regular rt_mutex_lock() call. 1686 */ 1687 if (tsk_state & TASK_RTLOCK_WAIT) 1688 state = TASK_UNINTERRUPTIBLE; 1689 1690 return fls(state); 1691 } 1692 1693 static inline unsigned int task_state_index(struct task_struct *tsk) 1694 { 1695 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1696 } 1697 1698 static inline char task_index_to_char(unsigned int state) 1699 { 1700 static const char state_char[] = "RSDTtXZPI"; 1701 1702 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1703 1704 return state_char[state]; 1705 } 1706 1707 static inline char task_state_to_char(struct task_struct *tsk) 1708 { 1709 return task_index_to_char(task_state_index(tsk)); 1710 } 1711 1712 /** 1713 * is_global_init - check if a task structure is init. Since init 1714 * is free to have sub-threads we need to check tgid. 1715 * @tsk: Task structure to be checked. 1716 * 1717 * Check if a task structure is the first user space task the kernel created. 1718 * 1719 * Return: 1 if the task structure is init. 0 otherwise. 1720 */ 1721 static inline int is_global_init(struct task_struct *tsk) 1722 { 1723 return task_tgid_nr(tsk) == 1; 1724 } 1725 1726 extern struct pid *cad_pid; 1727 1728 /* 1729 * Per process flags 1730 */ 1731 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1732 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1733 #define PF_EXITING 0x00000004 /* Getting shut down */ 1734 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1735 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1736 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1737 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1738 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1739 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1740 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1741 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1742 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1743 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1744 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1745 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1746 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1747 #define PF__HOLE__00010000 0x00010000 1748 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1749 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1750 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1751 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1752 * I am cleaning dirty pages from some other bdi. */ 1753 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1754 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1755 #define PF__HOLE__00800000 0x00800000 1756 #define PF__HOLE__01000000 0x01000000 1757 #define PF__HOLE__02000000 0x02000000 1758 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1759 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1760 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */ 1761 #define PF__HOLE__20000000 0x20000000 1762 #define PF__HOLE__40000000 0x40000000 1763 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1764 1765 /* 1766 * Only the _current_ task can read/write to tsk->flags, but other 1767 * tasks can access tsk->flags in readonly mode for example 1768 * with tsk_used_math (like during threaded core dumping). 1769 * There is however an exception to this rule during ptrace 1770 * or during fork: the ptracer task is allowed to write to the 1771 * child->flags of its traced child (same goes for fork, the parent 1772 * can write to the child->flags), because we're guaranteed the 1773 * child is not running and in turn not changing child->flags 1774 * at the same time the parent does it. 1775 */ 1776 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1777 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1778 #define clear_used_math() clear_stopped_child_used_math(current) 1779 #define set_used_math() set_stopped_child_used_math(current) 1780 1781 #define conditional_stopped_child_used_math(condition, child) \ 1782 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1783 1784 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1785 1786 #define copy_to_stopped_child_used_math(child) \ 1787 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1788 1789 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1790 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1791 #define used_math() tsk_used_math(current) 1792 1793 static __always_inline bool is_percpu_thread(void) 1794 { 1795 #ifdef CONFIG_SMP 1796 return (current->flags & PF_NO_SETAFFINITY) && 1797 (current->nr_cpus_allowed == 1); 1798 #else 1799 return true; 1800 #endif 1801 } 1802 1803 /* Per-process atomic flags. */ 1804 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1805 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1806 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1807 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1808 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1809 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1810 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1811 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1812 1813 #define TASK_PFA_TEST(name, func) \ 1814 static inline bool task_##func(struct task_struct *p) \ 1815 { return test_bit(PFA_##name, &p->atomic_flags); } 1816 1817 #define TASK_PFA_SET(name, func) \ 1818 static inline void task_set_##func(struct task_struct *p) \ 1819 { set_bit(PFA_##name, &p->atomic_flags); } 1820 1821 #define TASK_PFA_CLEAR(name, func) \ 1822 static inline void task_clear_##func(struct task_struct *p) \ 1823 { clear_bit(PFA_##name, &p->atomic_flags); } 1824 1825 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1826 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1827 1828 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1829 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1830 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1831 1832 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1833 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1834 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1835 1836 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1837 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1838 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1839 1840 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1841 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1842 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1843 1844 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1845 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1846 1847 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1848 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1849 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1850 1851 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1852 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1853 1854 static inline void 1855 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1856 { 1857 current->flags &= ~flags; 1858 current->flags |= orig_flags & flags; 1859 } 1860 1861 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1862 extern int task_can_attach(struct task_struct *p); 1863 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1864 extern void dl_bw_free(int cpu, u64 dl_bw); 1865 #ifdef CONFIG_SMP 1866 1867 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1868 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1869 1870 /** 1871 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1872 * @p: the task 1873 * @new_mask: CPU affinity mask 1874 * 1875 * Return: zero if successful, or a negative error code 1876 */ 1877 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1878 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1879 extern void release_user_cpus_ptr(struct task_struct *p); 1880 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1881 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1882 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1883 #else 1884 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1885 { 1886 } 1887 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1888 { 1889 if (!cpumask_test_cpu(0, new_mask)) 1890 return -EINVAL; 1891 return 0; 1892 } 1893 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node) 1894 { 1895 if (src->user_cpus_ptr) 1896 return -EINVAL; 1897 return 0; 1898 } 1899 static inline void release_user_cpus_ptr(struct task_struct *p) 1900 { 1901 WARN_ON(p->user_cpus_ptr); 1902 } 1903 1904 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) 1905 { 1906 return 0; 1907 } 1908 #endif 1909 1910 extern int yield_to(struct task_struct *p, bool preempt); 1911 extern void set_user_nice(struct task_struct *p, long nice); 1912 extern int task_prio(const struct task_struct *p); 1913 1914 /** 1915 * task_nice - return the nice value of a given task. 1916 * @p: the task in question. 1917 * 1918 * Return: The nice value [ -20 ... 0 ... 19 ]. 1919 */ 1920 static inline int task_nice(const struct task_struct *p) 1921 { 1922 return PRIO_TO_NICE((p)->static_prio); 1923 } 1924 1925 extern int can_nice(const struct task_struct *p, const int nice); 1926 extern int task_curr(const struct task_struct *p); 1927 extern int idle_cpu(int cpu); 1928 extern int available_idle_cpu(int cpu); 1929 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1930 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1931 extern void sched_set_fifo(struct task_struct *p); 1932 extern void sched_set_fifo_low(struct task_struct *p); 1933 extern void sched_set_normal(struct task_struct *p, int nice); 1934 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1935 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1936 extern struct task_struct *idle_task(int cpu); 1937 1938 /** 1939 * is_idle_task - is the specified task an idle task? 1940 * @p: the task in question. 1941 * 1942 * Return: 1 if @p is an idle task. 0 otherwise. 1943 */ 1944 static __always_inline bool is_idle_task(const struct task_struct *p) 1945 { 1946 return !!(p->flags & PF_IDLE); 1947 } 1948 1949 extern struct task_struct *curr_task(int cpu); 1950 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1951 1952 void yield(void); 1953 1954 union thread_union { 1955 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1956 struct task_struct task; 1957 #endif 1958 #ifndef CONFIG_THREAD_INFO_IN_TASK 1959 struct thread_info thread_info; 1960 #endif 1961 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1962 }; 1963 1964 #ifndef CONFIG_THREAD_INFO_IN_TASK 1965 extern struct thread_info init_thread_info; 1966 #endif 1967 1968 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1969 1970 #ifdef CONFIG_THREAD_INFO_IN_TASK 1971 # define task_thread_info(task) (&(task)->thread_info) 1972 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1973 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1974 #endif 1975 1976 /* 1977 * find a task by one of its numerical ids 1978 * 1979 * find_task_by_pid_ns(): 1980 * finds a task by its pid in the specified namespace 1981 * find_task_by_vpid(): 1982 * finds a task by its virtual pid 1983 * 1984 * see also find_vpid() etc in include/linux/pid.h 1985 */ 1986 1987 extern struct task_struct *find_task_by_vpid(pid_t nr); 1988 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1989 1990 /* 1991 * find a task by its virtual pid and get the task struct 1992 */ 1993 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1994 1995 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1996 extern int wake_up_process(struct task_struct *tsk); 1997 extern void wake_up_new_task(struct task_struct *tsk); 1998 1999 #ifdef CONFIG_SMP 2000 extern void kick_process(struct task_struct *tsk); 2001 #else 2002 static inline void kick_process(struct task_struct *tsk) { } 2003 #endif 2004 2005 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 2006 2007 static inline void set_task_comm(struct task_struct *tsk, const char *from) 2008 { 2009 __set_task_comm(tsk, from, false); 2010 } 2011 2012 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 2013 #define get_task_comm(buf, tsk) ({ \ 2014 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 2015 __get_task_comm(buf, sizeof(buf), tsk); \ 2016 }) 2017 2018 #ifdef CONFIG_SMP 2019 static __always_inline void scheduler_ipi(void) 2020 { 2021 /* 2022 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 2023 * TIF_NEED_RESCHED remotely (for the first time) will also send 2024 * this IPI. 2025 */ 2026 preempt_fold_need_resched(); 2027 } 2028 #else 2029 static inline void scheduler_ipi(void) { } 2030 #endif 2031 2032 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 2033 2034 /* 2035 * Set thread flags in other task's structures. 2036 * See asm/thread_info.h for TIF_xxxx flags available: 2037 */ 2038 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 2039 { 2040 set_ti_thread_flag(task_thread_info(tsk), flag); 2041 } 2042 2043 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2044 { 2045 clear_ti_thread_flag(task_thread_info(tsk), flag); 2046 } 2047 2048 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 2049 bool value) 2050 { 2051 update_ti_thread_flag(task_thread_info(tsk), flag, value); 2052 } 2053 2054 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2055 { 2056 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2057 } 2058 2059 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2060 { 2061 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2062 } 2063 2064 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2065 { 2066 return test_ti_thread_flag(task_thread_info(tsk), flag); 2067 } 2068 2069 static inline void set_tsk_need_resched(struct task_struct *tsk) 2070 { 2071 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2072 } 2073 2074 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2075 { 2076 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2077 } 2078 2079 static inline int test_tsk_need_resched(struct task_struct *tsk) 2080 { 2081 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2082 } 2083 2084 /* 2085 * cond_resched() and cond_resched_lock(): latency reduction via 2086 * explicit rescheduling in places that are safe. The return 2087 * value indicates whether a reschedule was done in fact. 2088 * cond_resched_lock() will drop the spinlock before scheduling, 2089 */ 2090 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2091 extern int __cond_resched(void); 2092 2093 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2094 2095 void sched_dynamic_klp_enable(void); 2096 void sched_dynamic_klp_disable(void); 2097 2098 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2099 2100 static __always_inline int _cond_resched(void) 2101 { 2102 return static_call_mod(cond_resched)(); 2103 } 2104 2105 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2106 2107 extern int dynamic_cond_resched(void); 2108 2109 static __always_inline int _cond_resched(void) 2110 { 2111 return dynamic_cond_resched(); 2112 } 2113 2114 #else /* !CONFIG_PREEMPTION */ 2115 2116 static inline int _cond_resched(void) 2117 { 2118 klp_sched_try_switch(); 2119 return __cond_resched(); 2120 } 2121 2122 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2123 2124 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2125 2126 static inline int _cond_resched(void) 2127 { 2128 klp_sched_try_switch(); 2129 return 0; 2130 } 2131 2132 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2133 2134 #define cond_resched() ({ \ 2135 __might_resched(__FILE__, __LINE__, 0); \ 2136 _cond_resched(); \ 2137 }) 2138 2139 extern int __cond_resched_lock(spinlock_t *lock); 2140 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2141 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2142 2143 #define MIGHT_RESCHED_RCU_SHIFT 8 2144 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2145 2146 #ifndef CONFIG_PREEMPT_RT 2147 /* 2148 * Non RT kernels have an elevated preempt count due to the held lock, 2149 * but are not allowed to be inside a RCU read side critical section 2150 */ 2151 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2152 #else 2153 /* 2154 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2155 * cond_resched*lock() has to take that into account because it checks for 2156 * preempt_count() and rcu_preempt_depth(). 2157 */ 2158 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2159 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2160 #endif 2161 2162 #define cond_resched_lock(lock) ({ \ 2163 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2164 __cond_resched_lock(lock); \ 2165 }) 2166 2167 #define cond_resched_rwlock_read(lock) ({ \ 2168 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2169 __cond_resched_rwlock_read(lock); \ 2170 }) 2171 2172 #define cond_resched_rwlock_write(lock) ({ \ 2173 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2174 __cond_resched_rwlock_write(lock); \ 2175 }) 2176 2177 static inline void cond_resched_rcu(void) 2178 { 2179 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 2180 rcu_read_unlock(); 2181 cond_resched(); 2182 rcu_read_lock(); 2183 #endif 2184 } 2185 2186 #ifdef CONFIG_PREEMPT_DYNAMIC 2187 2188 extern bool preempt_model_none(void); 2189 extern bool preempt_model_voluntary(void); 2190 extern bool preempt_model_full(void); 2191 2192 #else 2193 2194 static inline bool preempt_model_none(void) 2195 { 2196 return IS_ENABLED(CONFIG_PREEMPT_NONE); 2197 } 2198 static inline bool preempt_model_voluntary(void) 2199 { 2200 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY); 2201 } 2202 static inline bool preempt_model_full(void) 2203 { 2204 return IS_ENABLED(CONFIG_PREEMPT); 2205 } 2206 2207 #endif 2208 2209 static inline bool preempt_model_rt(void) 2210 { 2211 return IS_ENABLED(CONFIG_PREEMPT_RT); 2212 } 2213 2214 /* 2215 * Does the preemption model allow non-cooperative preemption? 2216 * 2217 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with 2218 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the 2219 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the 2220 * PREEMPT_NONE model. 2221 */ 2222 static inline bool preempt_model_preemptible(void) 2223 { 2224 return preempt_model_full() || preempt_model_rt(); 2225 } 2226 2227 /* 2228 * Does a critical section need to be broken due to another 2229 * task waiting?: (technically does not depend on CONFIG_PREEMPTION, 2230 * but a general need for low latency) 2231 */ 2232 static inline int spin_needbreak(spinlock_t *lock) 2233 { 2234 #ifdef CONFIG_PREEMPTION 2235 return spin_is_contended(lock); 2236 #else 2237 return 0; 2238 #endif 2239 } 2240 2241 /* 2242 * Check if a rwlock is contended. 2243 * Returns non-zero if there is another task waiting on the rwlock. 2244 * Returns zero if the lock is not contended or the system / underlying 2245 * rwlock implementation does not support contention detection. 2246 * Technically does not depend on CONFIG_PREEMPTION, but a general need 2247 * for low latency. 2248 */ 2249 static inline int rwlock_needbreak(rwlock_t *lock) 2250 { 2251 #ifdef CONFIG_PREEMPTION 2252 return rwlock_is_contended(lock); 2253 #else 2254 return 0; 2255 #endif 2256 } 2257 2258 static __always_inline bool need_resched(void) 2259 { 2260 return unlikely(tif_need_resched()); 2261 } 2262 2263 /* 2264 * Wrappers for p->thread_info->cpu access. No-op on UP. 2265 */ 2266 #ifdef CONFIG_SMP 2267 2268 static inline unsigned int task_cpu(const struct task_struct *p) 2269 { 2270 return READ_ONCE(task_thread_info(p)->cpu); 2271 } 2272 2273 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2274 2275 #else 2276 2277 static inline unsigned int task_cpu(const struct task_struct *p) 2278 { 2279 return 0; 2280 } 2281 2282 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2283 { 2284 } 2285 2286 #endif /* CONFIG_SMP */ 2287 2288 extern bool sched_task_on_rq(struct task_struct *p); 2289 extern unsigned long get_wchan(struct task_struct *p); 2290 extern struct task_struct *cpu_curr_snapshot(int cpu); 2291 2292 /* 2293 * In order to reduce various lock holder preemption latencies provide an 2294 * interface to see if a vCPU is currently running or not. 2295 * 2296 * This allows us to terminate optimistic spin loops and block, analogous to 2297 * the native optimistic spin heuristic of testing if the lock owner task is 2298 * running or not. 2299 */ 2300 #ifndef vcpu_is_preempted 2301 static inline bool vcpu_is_preempted(int cpu) 2302 { 2303 return false; 2304 } 2305 #endif 2306 2307 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2308 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2309 2310 #ifndef TASK_SIZE_OF 2311 #define TASK_SIZE_OF(tsk) TASK_SIZE 2312 #endif 2313 2314 #ifdef CONFIG_SMP 2315 static inline bool owner_on_cpu(struct task_struct *owner) 2316 { 2317 /* 2318 * As lock holder preemption issue, we both skip spinning if 2319 * task is not on cpu or its cpu is preempted 2320 */ 2321 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2322 } 2323 2324 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2325 unsigned long sched_cpu_util(int cpu); 2326 #endif /* CONFIG_SMP */ 2327 2328 #ifdef CONFIG_RSEQ 2329 2330 /* 2331 * Map the event mask on the user-space ABI enum rseq_cs_flags 2332 * for direct mask checks. 2333 */ 2334 enum rseq_event_mask_bits { 2335 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT, 2336 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT, 2337 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT, 2338 }; 2339 2340 enum rseq_event_mask { 2341 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT), 2342 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT), 2343 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT), 2344 }; 2345 2346 static inline void rseq_set_notify_resume(struct task_struct *t) 2347 { 2348 if (t->rseq) 2349 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME); 2350 } 2351 2352 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs); 2353 2354 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2355 struct pt_regs *regs) 2356 { 2357 if (current->rseq) 2358 __rseq_handle_notify_resume(ksig, regs); 2359 } 2360 2361 static inline void rseq_signal_deliver(struct ksignal *ksig, 2362 struct pt_regs *regs) 2363 { 2364 preempt_disable(); 2365 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask); 2366 preempt_enable(); 2367 rseq_handle_notify_resume(ksig, regs); 2368 } 2369 2370 /* rseq_preempt() requires preemption to be disabled. */ 2371 static inline void rseq_preempt(struct task_struct *t) 2372 { 2373 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask); 2374 rseq_set_notify_resume(t); 2375 } 2376 2377 /* rseq_migrate() requires preemption to be disabled. */ 2378 static inline void rseq_migrate(struct task_struct *t) 2379 { 2380 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask); 2381 rseq_set_notify_resume(t); 2382 } 2383 2384 /* 2385 * If parent process has a registered restartable sequences area, the 2386 * child inherits. Unregister rseq for a clone with CLONE_VM set. 2387 */ 2388 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2389 { 2390 if (clone_flags & CLONE_VM) { 2391 t->rseq = NULL; 2392 t->rseq_len = 0; 2393 t->rseq_sig = 0; 2394 t->rseq_event_mask = 0; 2395 } else { 2396 t->rseq = current->rseq; 2397 t->rseq_len = current->rseq_len; 2398 t->rseq_sig = current->rseq_sig; 2399 t->rseq_event_mask = current->rseq_event_mask; 2400 } 2401 } 2402 2403 static inline void rseq_execve(struct task_struct *t) 2404 { 2405 t->rseq = NULL; 2406 t->rseq_len = 0; 2407 t->rseq_sig = 0; 2408 t->rseq_event_mask = 0; 2409 } 2410 2411 #else 2412 2413 static inline void rseq_set_notify_resume(struct task_struct *t) 2414 { 2415 } 2416 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2417 struct pt_regs *regs) 2418 { 2419 } 2420 static inline void rseq_signal_deliver(struct ksignal *ksig, 2421 struct pt_regs *regs) 2422 { 2423 } 2424 static inline void rseq_preempt(struct task_struct *t) 2425 { 2426 } 2427 static inline void rseq_migrate(struct task_struct *t) 2428 { 2429 } 2430 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2431 { 2432 } 2433 static inline void rseq_execve(struct task_struct *t) 2434 { 2435 } 2436 2437 #endif 2438 2439 #ifdef CONFIG_DEBUG_RSEQ 2440 2441 void rseq_syscall(struct pt_regs *regs); 2442 2443 #else 2444 2445 static inline void rseq_syscall(struct pt_regs *regs) 2446 { 2447 } 2448 2449 #endif 2450 2451 #ifdef CONFIG_SCHED_CORE 2452 extern void sched_core_free(struct task_struct *tsk); 2453 extern void sched_core_fork(struct task_struct *p); 2454 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2455 unsigned long uaddr); 2456 extern int sched_core_idle_cpu(int cpu); 2457 #else 2458 static inline void sched_core_free(struct task_struct *tsk) { } 2459 static inline void sched_core_fork(struct task_struct *p) { } 2460 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2461 #endif 2462 2463 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2464 2465 #endif 2466