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