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 #include <asm/processor.h> 14 #include <linux/thread_info.h> 15 #include <linux/preempt.h> 16 #include <linux/cpumask_types.h> 17 18 #include <linux/cache.h> 19 #include <linux/irqflags_types.h> 20 #include <linux/smp_types.h> 21 #include <linux/pid_types.h> 22 #include <linux/sem_types.h> 23 #include <linux/shm.h> 24 #include <linux/kmsan_types.h> 25 #include <linux/mutex_types.h> 26 #include <linux/plist_types.h> 27 #include <linux/hrtimer_types.h> 28 #include <linux/timer_types.h> 29 #include <linux/seccomp_types.h> 30 #include <linux/nodemask_types.h> 31 #include <linux/refcount_types.h> 32 #include <linux/resource.h> 33 #include <linux/latencytop.h> 34 #include <linux/sched/prio.h> 35 #include <linux/sched/types.h> 36 #include <linux/signal_types.h> 37 #include <linux/spinlock.h> 38 #include <linux/syscall_user_dispatch_types.h> 39 #include <linux/mm_types_task.h> 40 #include <linux/netdevice_xmit.h> 41 #include <linux/task_io_accounting.h> 42 #include <linux/posix-timers_types.h> 43 #include <linux/restart_block.h> 44 #include <uapi/linux/rseq.h> 45 #include <linux/seqlock_types.h> 46 #include <linux/kcsan.h> 47 #include <linux/rv.h> 48 #include <linux/uidgid_types.h> 49 #include <linux/tracepoint-defs.h> 50 #include <linux/unwind_deferred_types.h> 51 #include <asm/kmap_size.h> 52 53 /* task_struct member predeclarations (sorted alphabetically): */ 54 struct audit_context; 55 struct bio_list; 56 struct blk_plug; 57 struct bpf_local_storage; 58 struct bpf_run_ctx; 59 struct bpf_net_context; 60 struct capture_control; 61 struct cfs_rq; 62 struct fs_struct; 63 struct futex_pi_state; 64 struct io_context; 65 struct io_uring_task; 66 struct mempolicy; 67 struct nameidata; 68 struct nsproxy; 69 struct perf_event_context; 70 struct perf_ctx_data; 71 struct pid_namespace; 72 struct pipe_inode_info; 73 struct rcu_node; 74 struct reclaim_state; 75 struct robust_list_head; 76 struct root_domain; 77 struct rq; 78 struct sched_attr; 79 struct sched_dl_entity; 80 struct seq_file; 81 struct sighand_struct; 82 struct signal_struct; 83 struct task_delay_info; 84 struct task_group; 85 struct task_struct; 86 struct user_event_mm; 87 88 #include <linux/sched/ext.h> 89 90 /* 91 * Task state bitmask. NOTE! These bits are also 92 * encoded in fs/proc/array.c: get_task_state(). 93 * 94 * We have two separate sets of flags: task->__state 95 * is about runnability, while task->exit_state are 96 * about the task exiting. Confusing, but this way 97 * modifying one set can't modify the other one by 98 * mistake. 99 */ 100 101 /* Used in tsk->__state: */ 102 #define TASK_RUNNING 0x00000000 103 #define TASK_INTERRUPTIBLE 0x00000001 104 #define TASK_UNINTERRUPTIBLE 0x00000002 105 #define __TASK_STOPPED 0x00000004 106 #define __TASK_TRACED 0x00000008 107 /* Used in tsk->exit_state: */ 108 #define EXIT_DEAD 0x00000010 109 #define EXIT_ZOMBIE 0x00000020 110 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 111 /* Used in tsk->__state again: */ 112 #define TASK_PARKED 0x00000040 113 #define TASK_DEAD 0x00000080 114 #define TASK_WAKEKILL 0x00000100 115 #define TASK_WAKING 0x00000200 116 #define TASK_NOLOAD 0x00000400 117 #define TASK_NEW 0x00000800 118 #define TASK_RTLOCK_WAIT 0x00001000 119 #define TASK_FREEZABLE 0x00002000 120 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP)) 121 #define TASK_FROZEN 0x00008000 122 #define TASK_STATE_MAX 0x00010000 123 124 #define TASK_ANY (TASK_STATE_MAX-1) 125 126 /* 127 * DO NOT ADD ANY NEW USERS ! 128 */ 129 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE) 130 131 /* Convenience macros for the sake of set_current_state: */ 132 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 133 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 134 #define TASK_TRACED __TASK_TRACED 135 136 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 137 138 /* Convenience macros for the sake of wake_up(): */ 139 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 140 141 /* get_task_state(): */ 142 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 143 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 144 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 145 TASK_PARKED) 146 147 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 148 149 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 150 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 151 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 152 153 /* 154 * Special states are those that do not use the normal wait-loop pattern. See 155 * the comment with set_special_state(). 156 */ 157 #define is_special_task_state(state) \ 158 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \ 159 TASK_DEAD | TASK_FROZEN)) 160 161 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 162 # define debug_normal_state_change(state_value) \ 163 do { \ 164 WARN_ON_ONCE(is_special_task_state(state_value)); \ 165 current->task_state_change = _THIS_IP_; \ 166 } while (0) 167 168 # define debug_special_state_change(state_value) \ 169 do { \ 170 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 171 current->task_state_change = _THIS_IP_; \ 172 } while (0) 173 174 # define debug_rtlock_wait_set_state() \ 175 do { \ 176 current->saved_state_change = current->task_state_change;\ 177 current->task_state_change = _THIS_IP_; \ 178 } while (0) 179 180 # define debug_rtlock_wait_restore_state() \ 181 do { \ 182 current->task_state_change = current->saved_state_change;\ 183 } while (0) 184 185 #else 186 # define debug_normal_state_change(cond) do { } while (0) 187 # define debug_special_state_change(cond) do { } while (0) 188 # define debug_rtlock_wait_set_state() do { } while (0) 189 # define debug_rtlock_wait_restore_state() do { } while (0) 190 #endif 191 192 #define trace_set_current_state(state_value) \ 193 do { \ 194 if (tracepoint_enabled(sched_set_state_tp)) \ 195 __trace_set_current_state(state_value); \ 196 } while (0) 197 198 /* 199 * set_current_state() includes a barrier so that the write of current->__state 200 * is correctly serialised wrt the caller's subsequent test of whether to 201 * actually sleep: 202 * 203 * for (;;) { 204 * set_current_state(TASK_UNINTERRUPTIBLE); 205 * if (CONDITION) 206 * break; 207 * 208 * schedule(); 209 * } 210 * __set_current_state(TASK_RUNNING); 211 * 212 * If the caller does not need such serialisation (because, for instance, the 213 * CONDITION test and condition change and wakeup are under the same lock) then 214 * use __set_current_state(). 215 * 216 * The above is typically ordered against the wakeup, which does: 217 * 218 * CONDITION = 1; 219 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 220 * 221 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 222 * accessing p->__state. 223 * 224 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is, 225 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 226 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 227 * 228 * However, with slightly different timing the wakeup TASK_RUNNING store can 229 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 230 * a problem either because that will result in one extra go around the loop 231 * and our @cond test will save the day. 232 * 233 * Also see the comments of try_to_wake_up(). 234 */ 235 #define __set_current_state(state_value) \ 236 do { \ 237 debug_normal_state_change((state_value)); \ 238 trace_set_current_state(state_value); \ 239 WRITE_ONCE(current->__state, (state_value)); \ 240 } while (0) 241 242 #define set_current_state(state_value) \ 243 do { \ 244 debug_normal_state_change((state_value)); \ 245 trace_set_current_state(state_value); \ 246 smp_store_mb(current->__state, (state_value)); \ 247 } while (0) 248 249 /* 250 * set_special_state() should be used for those states when the blocking task 251 * can not use the regular condition based wait-loop. In that case we must 252 * serialize against wakeups such that any possible in-flight TASK_RUNNING 253 * stores will not collide with our state change. 254 */ 255 #define set_special_state(state_value) \ 256 do { \ 257 unsigned long flags; /* may shadow */ \ 258 \ 259 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 260 debug_special_state_change((state_value)); \ 261 trace_set_current_state(state_value); \ 262 WRITE_ONCE(current->__state, (state_value)); \ 263 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 264 } while (0) 265 266 /* 267 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 268 * 269 * RT's spin/rwlock substitutions are state preserving. The state of the 270 * task when blocking on the lock is saved in task_struct::saved_state and 271 * restored after the lock has been acquired. These operations are 272 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 273 * lock related wakeups while the task is blocked on the lock are 274 * redirected to operate on task_struct::saved_state to ensure that these 275 * are not dropped. On restore task_struct::saved_state is set to 276 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 277 * 278 * The lock operation looks like this: 279 * 280 * current_save_and_set_rtlock_wait_state(); 281 * for (;;) { 282 * if (try_lock()) 283 * break; 284 * raw_spin_unlock_irq(&lock->wait_lock); 285 * schedule_rtlock(); 286 * raw_spin_lock_irq(&lock->wait_lock); 287 * set_current_state(TASK_RTLOCK_WAIT); 288 * } 289 * current_restore_rtlock_saved_state(); 290 */ 291 #define current_save_and_set_rtlock_wait_state() \ 292 do { \ 293 lockdep_assert_irqs_disabled(); \ 294 raw_spin_lock(¤t->pi_lock); \ 295 current->saved_state = current->__state; \ 296 debug_rtlock_wait_set_state(); \ 297 trace_set_current_state(TASK_RTLOCK_WAIT); \ 298 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 299 raw_spin_unlock(¤t->pi_lock); \ 300 } while (0); 301 302 #define current_restore_rtlock_saved_state() \ 303 do { \ 304 lockdep_assert_irqs_disabled(); \ 305 raw_spin_lock(¤t->pi_lock); \ 306 debug_rtlock_wait_restore_state(); \ 307 trace_set_current_state(current->saved_state); \ 308 WRITE_ONCE(current->__state, current->saved_state); \ 309 current->saved_state = TASK_RUNNING; \ 310 raw_spin_unlock(¤t->pi_lock); \ 311 } while (0); 312 313 #define get_current_state() READ_ONCE(current->__state) 314 315 /* 316 * Define the task command name length as enum, then it can be visible to 317 * BPF programs. 318 */ 319 enum { 320 TASK_COMM_LEN = 16, 321 }; 322 323 extern void sched_tick(void); 324 325 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 326 327 extern long schedule_timeout(long timeout); 328 extern long schedule_timeout_interruptible(long timeout); 329 extern long schedule_timeout_killable(long timeout); 330 extern long schedule_timeout_uninterruptible(long timeout); 331 extern long schedule_timeout_idle(long timeout); 332 asmlinkage void schedule(void); 333 extern void schedule_preempt_disabled(void); 334 asmlinkage void preempt_schedule_irq(void); 335 #ifdef CONFIG_PREEMPT_RT 336 extern void schedule_rtlock(void); 337 #endif 338 339 extern int __must_check io_schedule_prepare(void); 340 extern void io_schedule_finish(int token); 341 extern long io_schedule_timeout(long timeout); 342 extern void io_schedule(void); 343 344 /* wrapper functions to trace from this header file */ 345 DECLARE_TRACEPOINT(sched_set_state_tp); 346 extern void __trace_set_current_state(int state_value); 347 DECLARE_TRACEPOINT(sched_set_need_resched_tp); 348 extern void __trace_set_need_resched(struct task_struct *curr, int tif); 349 350 /** 351 * struct prev_cputime - snapshot of system and user cputime 352 * @utime: time spent in user mode 353 * @stime: time spent in system mode 354 * @lock: protects the above two fields 355 * 356 * Stores previous user/system time values such that we can guarantee 357 * monotonicity. 358 */ 359 struct prev_cputime { 360 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 361 u64 utime; 362 u64 stime; 363 raw_spinlock_t lock; 364 #endif 365 }; 366 367 enum vtime_state { 368 /* Task is sleeping or running in a CPU with VTIME inactive: */ 369 VTIME_INACTIVE = 0, 370 /* Task is idle */ 371 VTIME_IDLE, 372 /* Task runs in kernelspace in a CPU with VTIME active: */ 373 VTIME_SYS, 374 /* Task runs in userspace in a CPU with VTIME active: */ 375 VTIME_USER, 376 /* Task runs as guests in a CPU with VTIME active: */ 377 VTIME_GUEST, 378 }; 379 380 struct vtime { 381 seqcount_t seqcount; 382 unsigned long long starttime; 383 enum vtime_state state; 384 unsigned int cpu; 385 u64 utime; 386 u64 stime; 387 u64 gtime; 388 }; 389 390 /* 391 * Utilization clamp constraints. 392 * @UCLAMP_MIN: Minimum utilization 393 * @UCLAMP_MAX: Maximum utilization 394 * @UCLAMP_CNT: Utilization clamp constraints count 395 */ 396 enum uclamp_id { 397 UCLAMP_MIN = 0, 398 UCLAMP_MAX, 399 UCLAMP_CNT 400 }; 401 402 extern struct root_domain def_root_domain; 403 extern struct mutex sched_domains_mutex; 404 extern void sched_domains_mutex_lock(void); 405 extern void sched_domains_mutex_unlock(void); 406 407 struct sched_param { 408 int sched_priority; 409 }; 410 411 struct sched_info { 412 #ifdef CONFIG_SCHED_INFO 413 /* Cumulative counters: */ 414 415 /* # of times we have run on this CPU: */ 416 unsigned long pcount; 417 418 /* Time spent waiting on a runqueue: */ 419 unsigned long long run_delay; 420 421 /* Max time spent waiting on a runqueue: */ 422 unsigned long long max_run_delay; 423 424 /* Min time spent waiting on a runqueue: */ 425 unsigned long long min_run_delay; 426 427 /* Timestamps: */ 428 429 /* When did we last run on a CPU? */ 430 unsigned long long last_arrival; 431 432 /* When were we last queued to run? */ 433 unsigned long long last_queued; 434 435 #endif /* CONFIG_SCHED_INFO */ 436 }; 437 438 /* 439 * Integer metrics need fixed point arithmetic, e.g., sched/fair 440 * has a few: load, load_avg, util_avg, freq, and capacity. 441 * 442 * We define a basic fixed point arithmetic range, and then formalize 443 * all these metrics based on that basic range. 444 */ 445 # define SCHED_FIXEDPOINT_SHIFT 10 446 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 447 448 /* Increase resolution of cpu_capacity calculations */ 449 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 450 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 451 452 struct load_weight { 453 unsigned long weight; 454 u32 inv_weight; 455 }; 456 457 /* 458 * The load/runnable/util_avg accumulates an infinite geometric series 459 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 460 * 461 * [load_avg definition] 462 * 463 * load_avg = runnable% * scale_load_down(load) 464 * 465 * [runnable_avg definition] 466 * 467 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 468 * 469 * [util_avg definition] 470 * 471 * util_avg = running% * SCHED_CAPACITY_SCALE 472 * 473 * where runnable% is the time ratio that a sched_entity is runnable and 474 * running% the time ratio that a sched_entity is running. 475 * 476 * For cfs_rq, they are the aggregated values of all runnable and blocked 477 * sched_entities. 478 * 479 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 480 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 481 * for computing those signals (see update_rq_clock_pelt()) 482 * 483 * N.B., the above ratios (runnable% and running%) themselves are in the 484 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 485 * to as large a range as necessary. This is for example reflected by 486 * util_avg's SCHED_CAPACITY_SCALE. 487 * 488 * [Overflow issue] 489 * 490 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 491 * with the highest load (=88761), always runnable on a single cfs_rq, 492 * and should not overflow as the number already hits PID_MAX_LIMIT. 493 * 494 * For all other cases (including 32-bit kernels), struct load_weight's 495 * weight will overflow first before we do, because: 496 * 497 * Max(load_avg) <= Max(load.weight) 498 * 499 * Then it is the load_weight's responsibility to consider overflow 500 * issues. 501 */ 502 struct sched_avg { 503 u64 last_update_time; 504 u64 load_sum; 505 u64 runnable_sum; 506 u32 util_sum; 507 u32 period_contrib; 508 unsigned long load_avg; 509 unsigned long runnable_avg; 510 unsigned long util_avg; 511 unsigned int util_est; 512 } ____cacheline_aligned; 513 514 /* 515 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 516 * updates. When a task is dequeued, its util_est should not be updated if its 517 * util_avg has not been updated in the meantime. 518 * This information is mapped into the MSB bit of util_est at dequeue time. 519 * Since max value of util_est for a task is 1024 (PELT util_avg for a task) 520 * it is safe to use MSB. 521 */ 522 #define UTIL_EST_WEIGHT_SHIFT 2 523 #define UTIL_AVG_UNCHANGED 0x80000000 524 525 struct sched_statistics { 526 #ifdef CONFIG_SCHEDSTATS 527 u64 wait_start; 528 u64 wait_max; 529 u64 wait_count; 530 u64 wait_sum; 531 u64 iowait_count; 532 u64 iowait_sum; 533 534 u64 sleep_start; 535 u64 sleep_max; 536 s64 sum_sleep_runtime; 537 538 u64 block_start; 539 u64 block_max; 540 s64 sum_block_runtime; 541 542 s64 exec_max; 543 u64 slice_max; 544 545 u64 nr_migrations_cold; 546 u64 nr_failed_migrations_affine; 547 u64 nr_failed_migrations_running; 548 u64 nr_failed_migrations_hot; 549 u64 nr_forced_migrations; 550 551 u64 nr_wakeups; 552 u64 nr_wakeups_sync; 553 u64 nr_wakeups_migrate; 554 u64 nr_wakeups_local; 555 u64 nr_wakeups_remote; 556 u64 nr_wakeups_affine; 557 u64 nr_wakeups_affine_attempts; 558 u64 nr_wakeups_passive; 559 u64 nr_wakeups_idle; 560 561 #ifdef CONFIG_SCHED_CORE 562 u64 core_forceidle_sum; 563 #endif 564 #endif /* CONFIG_SCHEDSTATS */ 565 } ____cacheline_aligned; 566 567 struct sched_entity { 568 /* For load-balancing: */ 569 struct load_weight load; 570 struct rb_node run_node; 571 u64 deadline; 572 u64 min_vruntime; 573 u64 min_slice; 574 575 struct list_head group_node; 576 unsigned char on_rq; 577 unsigned char sched_delayed; 578 unsigned char rel_deadline; 579 unsigned char custom_slice; 580 /* hole */ 581 582 u64 exec_start; 583 u64 sum_exec_runtime; 584 u64 prev_sum_exec_runtime; 585 u64 vruntime; 586 union { 587 /* 588 * When !@on_rq this field is vlag. 589 * When cfs_rq->curr == se (which implies @on_rq) 590 * this field is vprot. See protect_slice(). 591 */ 592 s64 vlag; 593 u64 vprot; 594 }; 595 u64 slice; 596 597 u64 nr_migrations; 598 599 #ifdef CONFIG_FAIR_GROUP_SCHED 600 int depth; 601 struct sched_entity *parent; 602 /* rq on which this entity is (to be) queued: */ 603 struct cfs_rq *cfs_rq; 604 /* rq "owned" by this entity/group: */ 605 struct cfs_rq *my_q; 606 /* cached value of my_q->h_nr_running */ 607 unsigned long runnable_weight; 608 #endif 609 610 /* 611 * Per entity load average tracking. 612 * 613 * Put into separate cache line so it does not 614 * collide with read-mostly values above. 615 */ 616 struct sched_avg avg; 617 }; 618 619 struct sched_rt_entity { 620 struct list_head run_list; 621 unsigned long timeout; 622 unsigned long watchdog_stamp; 623 unsigned int time_slice; 624 unsigned short on_rq; 625 unsigned short on_list; 626 627 struct sched_rt_entity *back; 628 #ifdef CONFIG_RT_GROUP_SCHED 629 struct sched_rt_entity *parent; 630 /* rq on which this entity is (to be) queued: */ 631 struct rt_rq *rt_rq; 632 /* rq "owned" by this entity/group: */ 633 struct rt_rq *my_q; 634 #endif 635 } __randomize_layout; 636 637 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *); 638 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *); 639 640 struct sched_dl_entity { 641 struct rb_node rb_node; 642 643 /* 644 * Original scheduling parameters. Copied here from sched_attr 645 * during sched_setattr(), they will remain the same until 646 * the next sched_setattr(). 647 */ 648 u64 dl_runtime; /* Maximum runtime for each instance */ 649 u64 dl_deadline; /* Relative deadline of each instance */ 650 u64 dl_period; /* Separation of two instances (period) */ 651 u64 dl_bw; /* dl_runtime / dl_period */ 652 u64 dl_density; /* dl_runtime / dl_deadline */ 653 654 /* 655 * Actual scheduling parameters. Initialized with the values above, 656 * they are continuously updated during task execution. Note that 657 * the remaining runtime could be < 0 in case we are in overrun. 658 */ 659 s64 runtime; /* Remaining runtime for this instance */ 660 u64 deadline; /* Absolute deadline for this instance */ 661 unsigned int flags; /* Specifying the scheduler behaviour */ 662 663 /* 664 * Some bool flags: 665 * 666 * @dl_throttled tells if we exhausted the runtime. If so, the 667 * task has to wait for a replenishment to be performed at the 668 * next firing of dl_timer. 669 * 670 * @dl_yielded tells if task gave up the CPU before consuming 671 * all its available runtime during the last job. 672 * 673 * @dl_non_contending tells if the task is inactive while still 674 * contributing to the active utilization. In other words, it 675 * indicates if the inactive timer has been armed and its handler 676 * has not been executed yet. This flag is useful to avoid race 677 * conditions between the inactive timer handler and the wakeup 678 * code. 679 * 680 * @dl_overrun tells if the task asked to be informed about runtime 681 * overruns. 682 * 683 * @dl_server tells if this is a server entity. 684 * 685 * @dl_defer tells if this is a deferred or regular server. For 686 * now only defer server exists. 687 * 688 * @dl_defer_armed tells if the deferrable server is waiting 689 * for the replenishment timer to activate it. 690 * 691 * @dl_server_active tells if the dlserver is active(started). 692 * dlserver is started on first cfs enqueue on an idle runqueue 693 * and is stopped when a dequeue results in 0 cfs tasks on the 694 * runqueue. In other words, dlserver is active only when cpu's 695 * runqueue has atleast one cfs task. 696 * 697 * @dl_defer_running tells if the deferrable server is actually 698 * running, skipping the defer phase. 699 */ 700 unsigned int dl_throttled : 1; 701 unsigned int dl_yielded : 1; 702 unsigned int dl_non_contending : 1; 703 unsigned int dl_overrun : 1; 704 unsigned int dl_server : 1; 705 unsigned int dl_server_active : 1; 706 unsigned int dl_defer : 1; 707 unsigned int dl_defer_armed : 1; 708 unsigned int dl_defer_running : 1; 709 unsigned int dl_server_idle : 1; 710 711 /* 712 * Bandwidth enforcement timer. Each -deadline task has its 713 * own bandwidth to be enforced, thus we need one timer per task. 714 */ 715 struct hrtimer dl_timer; 716 717 /* 718 * Inactive timer, responsible for decreasing the active utilization 719 * at the "0-lag time". When a -deadline task blocks, it contributes 720 * to GRUB's active utilization until the "0-lag time", hence a 721 * timer is needed to decrease the active utilization at the correct 722 * time. 723 */ 724 struct hrtimer inactive_timer; 725 726 /* 727 * Bits for DL-server functionality. Also see the comment near 728 * dl_server_update(). 729 * 730 * @rq the runqueue this server is for 731 * 732 * @server_has_tasks() returns true if @server_pick return a 733 * runnable task. 734 */ 735 struct rq *rq; 736 dl_server_has_tasks_f server_has_tasks; 737 dl_server_pick_f server_pick_task; 738 739 #ifdef CONFIG_RT_MUTEXES 740 /* 741 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 742 * pi_se points to the donor, otherwise points to the dl_se it belongs 743 * to (the original one/itself). 744 */ 745 struct sched_dl_entity *pi_se; 746 #endif 747 }; 748 749 #ifdef CONFIG_UCLAMP_TASK 750 /* Number of utilization clamp buckets (shorter alias) */ 751 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 752 753 /* 754 * Utilization clamp for a scheduling entity 755 * @value: clamp value "assigned" to a se 756 * @bucket_id: bucket index corresponding to the "assigned" value 757 * @active: the se is currently refcounted in a rq's bucket 758 * @user_defined: the requested clamp value comes from user-space 759 * 760 * The bucket_id is the index of the clamp bucket matching the clamp value 761 * which is pre-computed and stored to avoid expensive integer divisions from 762 * the fast path. 763 * 764 * The active bit is set whenever a task has got an "effective" value assigned, 765 * which can be different from the clamp value "requested" from user-space. 766 * This allows to know a task is refcounted in the rq's bucket corresponding 767 * to the "effective" bucket_id. 768 * 769 * The user_defined bit is set whenever a task has got a task-specific clamp 770 * value requested from userspace, i.e. the system defaults apply to this task 771 * just as a restriction. This allows to relax default clamps when a less 772 * restrictive task-specific value has been requested, thus allowing to 773 * implement a "nice" semantic. For example, a task running with a 20% 774 * default boost can still drop its own boosting to 0%. 775 */ 776 struct uclamp_se { 777 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 778 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 779 unsigned int active : 1; 780 unsigned int user_defined : 1; 781 }; 782 #endif /* CONFIG_UCLAMP_TASK */ 783 784 union rcu_special { 785 struct { 786 u8 blocked; 787 u8 need_qs; 788 u8 exp_hint; /* Hint for performance. */ 789 u8 need_mb; /* Readers need smp_mb(). */ 790 } b; /* Bits. */ 791 u32 s; /* Set of bits. */ 792 }; 793 794 enum perf_event_task_context { 795 perf_invalid_context = -1, 796 perf_hw_context = 0, 797 perf_sw_context, 798 perf_nr_task_contexts, 799 }; 800 801 /* 802 * Number of contexts where an event can trigger: 803 * task, softirq, hardirq, nmi. 804 */ 805 #define PERF_NR_CONTEXTS 4 806 807 struct wake_q_node { 808 struct wake_q_node *next; 809 }; 810 811 struct kmap_ctrl { 812 #ifdef CONFIG_KMAP_LOCAL 813 int idx; 814 pte_t pteval[KM_MAX_IDX]; 815 #endif 816 }; 817 818 struct task_struct { 819 #ifdef CONFIG_THREAD_INFO_IN_TASK 820 /* 821 * For reasons of header soup (see current_thread_info()), this 822 * must be the first element of task_struct. 823 */ 824 struct thread_info thread_info; 825 #endif 826 unsigned int __state; 827 828 /* saved state for "spinlock sleepers" */ 829 unsigned int saved_state; 830 831 /* 832 * This begins the randomizable portion of task_struct. Only 833 * scheduling-critical items should be added above here. 834 */ 835 randomized_struct_fields_start 836 837 void *stack; 838 refcount_t usage; 839 /* Per task flags (PF_*), defined further below: */ 840 unsigned int flags; 841 unsigned int ptrace; 842 843 #ifdef CONFIG_MEM_ALLOC_PROFILING 844 struct alloc_tag *alloc_tag; 845 #endif 846 847 int on_cpu; 848 struct __call_single_node wake_entry; 849 unsigned int wakee_flips; 850 unsigned long wakee_flip_decay_ts; 851 struct task_struct *last_wakee; 852 853 /* 854 * recent_used_cpu is initially set as the last CPU used by a task 855 * that wakes affine another task. Waker/wakee relationships can 856 * push tasks around a CPU where each wakeup moves to the next one. 857 * Tracking a recently used CPU allows a quick search for a recently 858 * used CPU that may be idle. 859 */ 860 int recent_used_cpu; 861 int wake_cpu; 862 int on_rq; 863 864 int prio; 865 int static_prio; 866 int normal_prio; 867 unsigned int rt_priority; 868 869 struct sched_entity se; 870 struct sched_rt_entity rt; 871 struct sched_dl_entity dl; 872 struct sched_dl_entity *dl_server; 873 #ifdef CONFIG_SCHED_CLASS_EXT 874 struct sched_ext_entity scx; 875 #endif 876 const struct sched_class *sched_class; 877 878 #ifdef CONFIG_SCHED_CORE 879 struct rb_node core_node; 880 unsigned long core_cookie; 881 unsigned int core_occupation; 882 #endif 883 884 #ifdef CONFIG_CGROUP_SCHED 885 struct task_group *sched_task_group; 886 #endif 887 888 889 #ifdef CONFIG_UCLAMP_TASK 890 /* 891 * Clamp values requested for a scheduling entity. 892 * Must be updated with task_rq_lock() held. 893 */ 894 struct uclamp_se uclamp_req[UCLAMP_CNT]; 895 /* 896 * Effective clamp values used for a scheduling entity. 897 * Must be updated with task_rq_lock() held. 898 */ 899 struct uclamp_se uclamp[UCLAMP_CNT]; 900 #endif 901 902 struct sched_statistics stats; 903 904 #ifdef CONFIG_PREEMPT_NOTIFIERS 905 /* List of struct preempt_notifier: */ 906 struct hlist_head preempt_notifiers; 907 #endif 908 909 #ifdef CONFIG_BLK_DEV_IO_TRACE 910 unsigned int btrace_seq; 911 #endif 912 913 unsigned int policy; 914 unsigned long max_allowed_capacity; 915 int nr_cpus_allowed; 916 const cpumask_t *cpus_ptr; 917 cpumask_t *user_cpus_ptr; 918 cpumask_t cpus_mask; 919 void *migration_pending; 920 unsigned short migration_disabled; 921 unsigned short migration_flags; 922 923 #ifdef CONFIG_PREEMPT_RCU 924 int rcu_read_lock_nesting; 925 union rcu_special rcu_read_unlock_special; 926 struct list_head rcu_node_entry; 927 struct rcu_node *rcu_blocked_node; 928 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 929 930 #ifdef CONFIG_TASKS_RCU 931 unsigned long rcu_tasks_nvcsw; 932 u8 rcu_tasks_holdout; 933 u8 rcu_tasks_idx; 934 int rcu_tasks_idle_cpu; 935 struct list_head rcu_tasks_holdout_list; 936 int rcu_tasks_exit_cpu; 937 struct list_head rcu_tasks_exit_list; 938 #endif /* #ifdef CONFIG_TASKS_RCU */ 939 940 #ifdef CONFIG_TASKS_TRACE_RCU 941 int trc_reader_nesting; 942 int trc_ipi_to_cpu; 943 union rcu_special trc_reader_special; 944 struct list_head trc_holdout_list; 945 struct list_head trc_blkd_node; 946 int trc_blkd_cpu; 947 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 948 949 struct sched_info sched_info; 950 951 struct list_head tasks; 952 struct plist_node pushable_tasks; 953 struct rb_node pushable_dl_tasks; 954 955 struct mm_struct *mm; 956 struct mm_struct *active_mm; 957 struct address_space *faults_disabled_mapping; 958 959 int exit_state; 960 int exit_code; 961 int exit_signal; 962 /* The signal sent when the parent dies: */ 963 int pdeath_signal; 964 /* JOBCTL_*, siglock protected: */ 965 unsigned long jobctl; 966 967 /* Used for emulating ABI behavior of previous Linux versions: */ 968 unsigned int personality; 969 970 /* Scheduler bits, serialized by scheduler locks: */ 971 unsigned sched_reset_on_fork:1; 972 unsigned sched_contributes_to_load:1; 973 unsigned sched_migrated:1; 974 unsigned sched_task_hot:1; 975 976 /* Force alignment to the next boundary: */ 977 unsigned :0; 978 979 /* Unserialized, strictly 'current' */ 980 981 /* 982 * This field must not be in the scheduler word above due to wakelist 983 * queueing no longer being serialized by p->on_cpu. However: 984 * 985 * p->XXX = X; ttwu() 986 * schedule() if (p->on_rq && ..) // false 987 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 988 * deactivate_task() ttwu_queue_wakelist()) 989 * p->on_rq = 0; p->sched_remote_wakeup = Y; 990 * 991 * guarantees all stores of 'current' are visible before 992 * ->sched_remote_wakeup gets used, so it can be in this word. 993 */ 994 unsigned sched_remote_wakeup:1; 995 #ifdef CONFIG_RT_MUTEXES 996 unsigned sched_rt_mutex:1; 997 #endif 998 999 /* Bit to tell TOMOYO we're in execve(): */ 1000 unsigned in_execve:1; 1001 unsigned in_iowait:1; 1002 #ifndef TIF_RESTORE_SIGMASK 1003 unsigned restore_sigmask:1; 1004 #endif 1005 #ifdef CONFIG_MEMCG_V1 1006 unsigned in_user_fault:1; 1007 #endif 1008 #ifdef CONFIG_LRU_GEN 1009 /* whether the LRU algorithm may apply to this access */ 1010 unsigned in_lru_fault:1; 1011 #endif 1012 #ifdef CONFIG_COMPAT_BRK 1013 unsigned brk_randomized:1; 1014 #endif 1015 #ifdef CONFIG_CGROUPS 1016 /* disallow userland-initiated cgroup migration */ 1017 unsigned no_cgroup_migration:1; 1018 /* task is frozen/stopped (used by the cgroup freezer) */ 1019 unsigned frozen:1; 1020 #endif 1021 #ifdef CONFIG_BLK_CGROUP 1022 unsigned use_memdelay:1; 1023 #endif 1024 #ifdef CONFIG_PSI 1025 /* Stalled due to lack of memory */ 1026 unsigned in_memstall:1; 1027 #endif 1028 #ifdef CONFIG_PAGE_OWNER 1029 /* Used by page_owner=on to detect recursion in page tracking. */ 1030 unsigned in_page_owner:1; 1031 #endif 1032 #ifdef CONFIG_EVENTFD 1033 /* Recursion prevention for eventfd_signal() */ 1034 unsigned in_eventfd:1; 1035 #endif 1036 #ifdef CONFIG_ARCH_HAS_CPU_PASID 1037 unsigned pasid_activated:1; 1038 #endif 1039 #ifdef CONFIG_X86_BUS_LOCK_DETECT 1040 unsigned reported_split_lock:1; 1041 #endif 1042 #ifdef CONFIG_TASK_DELAY_ACCT 1043 /* delay due to memory thrashing */ 1044 unsigned in_thrashing:1; 1045 #endif 1046 unsigned in_nf_duplicate:1; 1047 #ifdef CONFIG_PREEMPT_RT 1048 struct netdev_xmit net_xmit; 1049 #endif 1050 unsigned long atomic_flags; /* Flags requiring atomic access. */ 1051 1052 struct restart_block restart_block; 1053 1054 pid_t pid; 1055 pid_t tgid; 1056 1057 #ifdef CONFIG_STACKPROTECTOR 1058 /* Canary value for the -fstack-protector GCC feature: */ 1059 unsigned long stack_canary; 1060 #endif 1061 /* 1062 * Pointers to the (original) parent process, youngest child, younger sibling, 1063 * older sibling, respectively. (p->father can be replaced with 1064 * p->real_parent->pid) 1065 */ 1066 1067 /* Real parent process: */ 1068 struct task_struct __rcu *real_parent; 1069 1070 /* Recipient of SIGCHLD, wait4() reports: */ 1071 struct task_struct __rcu *parent; 1072 1073 /* 1074 * Children/sibling form the list of natural children: 1075 */ 1076 struct list_head children; 1077 struct list_head sibling; 1078 struct task_struct *group_leader; 1079 1080 /* 1081 * 'ptraced' is the list of tasks this task is using ptrace() on. 1082 * 1083 * This includes both natural children and PTRACE_ATTACH targets. 1084 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1085 */ 1086 struct list_head ptraced; 1087 struct list_head ptrace_entry; 1088 1089 /* PID/PID hash table linkage. */ 1090 struct pid *thread_pid; 1091 struct hlist_node pid_links[PIDTYPE_MAX]; 1092 struct list_head thread_node; 1093 1094 struct completion *vfork_done; 1095 1096 /* CLONE_CHILD_SETTID: */ 1097 int __user *set_child_tid; 1098 1099 /* CLONE_CHILD_CLEARTID: */ 1100 int __user *clear_child_tid; 1101 1102 /* PF_KTHREAD | PF_IO_WORKER */ 1103 void *worker_private; 1104 1105 u64 utime; 1106 u64 stime; 1107 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1108 u64 utimescaled; 1109 u64 stimescaled; 1110 #endif 1111 u64 gtime; 1112 struct prev_cputime prev_cputime; 1113 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1114 struct vtime vtime; 1115 #endif 1116 1117 #ifdef CONFIG_NO_HZ_FULL 1118 atomic_t tick_dep_mask; 1119 #endif 1120 /* Context switch counts: */ 1121 unsigned long nvcsw; 1122 unsigned long nivcsw; 1123 1124 /* Monotonic time in nsecs: */ 1125 u64 start_time; 1126 1127 /* Boot based time in nsecs: */ 1128 u64 start_boottime; 1129 1130 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1131 unsigned long min_flt; 1132 unsigned long maj_flt; 1133 1134 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1135 struct posix_cputimers posix_cputimers; 1136 1137 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1138 struct posix_cputimers_work posix_cputimers_work; 1139 #endif 1140 1141 /* Process credentials: */ 1142 1143 /* Tracer's credentials at attach: */ 1144 const struct cred __rcu *ptracer_cred; 1145 1146 /* Objective and real subjective task credentials (COW): */ 1147 const struct cred __rcu *real_cred; 1148 1149 /* Effective (overridable) subjective task credentials (COW): */ 1150 const struct cred __rcu *cred; 1151 1152 #ifdef CONFIG_KEYS 1153 /* Cached requested key. */ 1154 struct key *cached_requested_key; 1155 #endif 1156 1157 /* 1158 * executable name, excluding path. 1159 * 1160 * - normally initialized begin_new_exec() 1161 * - set it with set_task_comm() 1162 * - strscpy_pad() to ensure it is always NUL-terminated and 1163 * zero-padded 1164 * - task_lock() to ensure the operation is atomic and the name is 1165 * fully updated. 1166 */ 1167 char comm[TASK_COMM_LEN]; 1168 1169 struct nameidata *nameidata; 1170 1171 #ifdef CONFIG_SYSVIPC 1172 struct sysv_sem sysvsem; 1173 struct sysv_shm sysvshm; 1174 #endif 1175 #ifdef CONFIG_DETECT_HUNG_TASK 1176 unsigned long last_switch_count; 1177 unsigned long last_switch_time; 1178 #endif 1179 /* Filesystem information: */ 1180 struct fs_struct *fs; 1181 1182 /* Open file information: */ 1183 struct files_struct *files; 1184 1185 #ifdef CONFIG_IO_URING 1186 struct io_uring_task *io_uring; 1187 #endif 1188 1189 /* Namespaces: */ 1190 struct nsproxy *nsproxy; 1191 1192 /* Signal handlers: */ 1193 struct signal_struct *signal; 1194 struct sighand_struct __rcu *sighand; 1195 sigset_t blocked; 1196 sigset_t real_blocked; 1197 /* Restored if set_restore_sigmask() was used: */ 1198 sigset_t saved_sigmask; 1199 struct sigpending pending; 1200 unsigned long sas_ss_sp; 1201 size_t sas_ss_size; 1202 unsigned int sas_ss_flags; 1203 1204 struct callback_head *task_works; 1205 1206 #ifdef CONFIG_AUDIT 1207 #ifdef CONFIG_AUDITSYSCALL 1208 struct audit_context *audit_context; 1209 #endif 1210 kuid_t loginuid; 1211 unsigned int sessionid; 1212 #endif 1213 struct seccomp seccomp; 1214 struct syscall_user_dispatch syscall_dispatch; 1215 1216 /* Thread group tracking: */ 1217 u64 parent_exec_id; 1218 u64 self_exec_id; 1219 1220 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1221 spinlock_t alloc_lock; 1222 1223 /* Protection of the PI data structures: */ 1224 raw_spinlock_t pi_lock; 1225 1226 struct wake_q_node wake_q; 1227 1228 #ifdef CONFIG_RT_MUTEXES 1229 /* PI waiters blocked on a rt_mutex held by this task: */ 1230 struct rb_root_cached pi_waiters; 1231 /* Updated under owner's pi_lock and rq lock */ 1232 struct task_struct *pi_top_task; 1233 /* Deadlock detection and priority inheritance handling: */ 1234 struct rt_mutex_waiter *pi_blocked_on; 1235 #endif 1236 1237 struct mutex *blocked_on; /* lock we're blocked on */ 1238 1239 #ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER 1240 /* 1241 * Encoded lock address causing task block (lower 2 bits = type from 1242 * <linux/hung_task.h>). Accessed via hung_task_*() helpers. 1243 */ 1244 unsigned long blocker; 1245 #endif 1246 1247 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1248 int non_block_count; 1249 #endif 1250 1251 #ifdef CONFIG_TRACE_IRQFLAGS 1252 struct irqtrace_events irqtrace; 1253 unsigned int hardirq_threaded; 1254 u64 hardirq_chain_key; 1255 int softirqs_enabled; 1256 int softirq_context; 1257 int irq_config; 1258 #endif 1259 #ifdef CONFIG_PREEMPT_RT 1260 int softirq_disable_cnt; 1261 #endif 1262 1263 #ifdef CONFIG_LOCKDEP 1264 # define MAX_LOCK_DEPTH 48UL 1265 u64 curr_chain_key; 1266 int lockdep_depth; 1267 unsigned int lockdep_recursion; 1268 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1269 #endif 1270 1271 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1272 unsigned int in_ubsan; 1273 #endif 1274 1275 /* Journalling filesystem info: */ 1276 void *journal_info; 1277 1278 /* Stacked block device info: */ 1279 struct bio_list *bio_list; 1280 1281 /* Stack plugging: */ 1282 struct blk_plug *plug; 1283 1284 /* VM state: */ 1285 struct reclaim_state *reclaim_state; 1286 1287 struct io_context *io_context; 1288 1289 #ifdef CONFIG_COMPACTION 1290 struct capture_control *capture_control; 1291 #endif 1292 /* Ptrace state: */ 1293 unsigned long ptrace_message; 1294 kernel_siginfo_t *last_siginfo; 1295 1296 struct task_io_accounting ioac; 1297 #ifdef CONFIG_PSI 1298 /* Pressure stall state */ 1299 unsigned int psi_flags; 1300 #endif 1301 #ifdef CONFIG_TASK_XACCT 1302 /* Accumulated RSS usage: */ 1303 u64 acct_rss_mem1; 1304 /* Accumulated virtual memory usage: */ 1305 u64 acct_vm_mem1; 1306 /* stime + utime since last update: */ 1307 u64 acct_timexpd; 1308 #endif 1309 #ifdef CONFIG_CPUSETS 1310 /* Protected by ->alloc_lock: */ 1311 nodemask_t mems_allowed; 1312 /* Sequence number to catch updates: */ 1313 seqcount_spinlock_t mems_allowed_seq; 1314 int cpuset_mem_spread_rotor; 1315 #endif 1316 #ifdef CONFIG_CGROUPS 1317 /* Control Group info protected by css_set_lock: */ 1318 struct css_set __rcu *cgroups; 1319 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1320 struct list_head cg_list; 1321 #endif 1322 #ifdef CONFIG_X86_CPU_RESCTRL 1323 u32 closid; 1324 u32 rmid; 1325 #endif 1326 #ifdef CONFIG_FUTEX 1327 struct robust_list_head __user *robust_list; 1328 #ifdef CONFIG_COMPAT 1329 struct compat_robust_list_head __user *compat_robust_list; 1330 #endif 1331 struct list_head pi_state_list; 1332 struct futex_pi_state *pi_state_cache; 1333 struct mutex futex_exit_mutex; 1334 unsigned int futex_state; 1335 #endif 1336 #ifdef CONFIG_PERF_EVENTS 1337 u8 perf_recursion[PERF_NR_CONTEXTS]; 1338 struct perf_event_context *perf_event_ctxp; 1339 struct mutex perf_event_mutex; 1340 struct list_head perf_event_list; 1341 struct perf_ctx_data __rcu *perf_ctx_data; 1342 #endif 1343 #ifdef CONFIG_DEBUG_PREEMPT 1344 unsigned long preempt_disable_ip; 1345 #endif 1346 #ifdef CONFIG_NUMA 1347 /* Protected by alloc_lock: */ 1348 struct mempolicy *mempolicy; 1349 short il_prev; 1350 u8 il_weight; 1351 short pref_node_fork; 1352 #endif 1353 #ifdef CONFIG_NUMA_BALANCING 1354 int numa_scan_seq; 1355 unsigned int numa_scan_period; 1356 unsigned int numa_scan_period_max; 1357 int numa_preferred_nid; 1358 unsigned long numa_migrate_retry; 1359 /* Migration stamp: */ 1360 u64 node_stamp; 1361 u64 last_task_numa_placement; 1362 u64 last_sum_exec_runtime; 1363 struct callback_head numa_work; 1364 1365 /* 1366 * This pointer is only modified for current in syscall and 1367 * pagefault context (and for tasks being destroyed), so it can be read 1368 * from any of the following contexts: 1369 * - RCU read-side critical section 1370 * - current->numa_group from everywhere 1371 * - task's runqueue locked, task not running 1372 */ 1373 struct numa_group __rcu *numa_group; 1374 1375 /* 1376 * numa_faults is an array split into four regions: 1377 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1378 * in this precise order. 1379 * 1380 * faults_memory: Exponential decaying average of faults on a per-node 1381 * basis. Scheduling placement decisions are made based on these 1382 * counts. The values remain static for the duration of a PTE scan. 1383 * faults_cpu: Track the nodes the process was running on when a NUMA 1384 * hinting fault was incurred. 1385 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1386 * during the current scan window. When the scan completes, the counts 1387 * in faults_memory and faults_cpu decay and these values are copied. 1388 */ 1389 unsigned long *numa_faults; 1390 unsigned long total_numa_faults; 1391 1392 /* 1393 * numa_faults_locality tracks if faults recorded during the last 1394 * scan window were remote/local or failed to migrate. The task scan 1395 * period is adapted based on the locality of the faults with different 1396 * weights depending on whether they were shared or private faults 1397 */ 1398 unsigned long numa_faults_locality[3]; 1399 1400 unsigned long numa_pages_migrated; 1401 #endif /* CONFIG_NUMA_BALANCING */ 1402 1403 #ifdef CONFIG_RSEQ 1404 struct rseq __user *rseq; 1405 u32 rseq_len; 1406 u32 rseq_sig; 1407 /* 1408 * RmW on rseq_event_mask must be performed atomically 1409 * with respect to preemption. 1410 */ 1411 unsigned long rseq_event_mask; 1412 # ifdef CONFIG_DEBUG_RSEQ 1413 /* 1414 * This is a place holder to save a copy of the rseq fields for 1415 * validation of read-only fields. The struct rseq has a 1416 * variable-length array at the end, so it cannot be used 1417 * directly. Reserve a size large enough for the known fields. 1418 */ 1419 char rseq_fields[sizeof(struct rseq)]; 1420 # endif 1421 #endif 1422 1423 #ifdef CONFIG_SCHED_MM_CID 1424 int mm_cid; /* Current cid in mm */ 1425 int last_mm_cid; /* Most recent cid in mm */ 1426 int migrate_from_cpu; 1427 int mm_cid_active; /* Whether cid bitmap is active */ 1428 struct callback_head cid_work; 1429 #endif 1430 1431 struct tlbflush_unmap_batch tlb_ubc; 1432 1433 /* Cache last used pipe for splice(): */ 1434 struct pipe_inode_info *splice_pipe; 1435 1436 struct page_frag task_frag; 1437 1438 #ifdef CONFIG_TASK_DELAY_ACCT 1439 struct task_delay_info *delays; 1440 #endif 1441 1442 #ifdef CONFIG_FAULT_INJECTION 1443 int make_it_fail; 1444 unsigned int fail_nth; 1445 #endif 1446 /* 1447 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1448 * balance_dirty_pages() for a dirty throttling pause: 1449 */ 1450 int nr_dirtied; 1451 int nr_dirtied_pause; 1452 /* Start of a write-and-pause period: */ 1453 unsigned long dirty_paused_when; 1454 1455 #ifdef CONFIG_LATENCYTOP 1456 int latency_record_count; 1457 struct latency_record latency_record[LT_SAVECOUNT]; 1458 #endif 1459 /* 1460 * Time slack values; these are used to round up poll() and 1461 * select() etc timeout values. These are in nanoseconds. 1462 */ 1463 u64 timer_slack_ns; 1464 u64 default_timer_slack_ns; 1465 1466 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1467 unsigned int kasan_depth; 1468 #endif 1469 1470 #ifdef CONFIG_KCSAN 1471 struct kcsan_ctx kcsan_ctx; 1472 #ifdef CONFIG_TRACE_IRQFLAGS 1473 struct irqtrace_events kcsan_save_irqtrace; 1474 #endif 1475 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1476 int kcsan_stack_depth; 1477 #endif 1478 #endif 1479 1480 #ifdef CONFIG_KMSAN 1481 struct kmsan_ctx kmsan_ctx; 1482 #endif 1483 1484 #if IS_ENABLED(CONFIG_KUNIT) 1485 struct kunit *kunit_test; 1486 #endif 1487 1488 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1489 /* Index of current stored address in ret_stack: */ 1490 int curr_ret_stack; 1491 int curr_ret_depth; 1492 1493 /* Stack of return addresses for return function tracing: */ 1494 unsigned long *ret_stack; 1495 1496 /* Timestamp for last schedule: */ 1497 unsigned long long ftrace_timestamp; 1498 unsigned long long ftrace_sleeptime; 1499 1500 /* 1501 * Number of functions that haven't been traced 1502 * because of depth overrun: 1503 */ 1504 atomic_t trace_overrun; 1505 1506 /* Pause tracing: */ 1507 atomic_t tracing_graph_pause; 1508 #endif 1509 1510 #ifdef CONFIG_TRACING 1511 /* Bitmask and counter of trace recursion: */ 1512 unsigned long trace_recursion; 1513 #endif /* CONFIG_TRACING */ 1514 1515 #ifdef CONFIG_KCOV 1516 /* See kernel/kcov.c for more details. */ 1517 1518 /* Coverage collection mode enabled for this task (0 if disabled): */ 1519 unsigned int kcov_mode; 1520 1521 /* Size of the kcov_area: */ 1522 unsigned int kcov_size; 1523 1524 /* Buffer for coverage collection: */ 1525 void *kcov_area; 1526 1527 /* KCOV descriptor wired with this task or NULL: */ 1528 struct kcov *kcov; 1529 1530 /* KCOV common handle for remote coverage collection: */ 1531 u64 kcov_handle; 1532 1533 /* KCOV sequence number: */ 1534 int kcov_sequence; 1535 1536 /* Collect coverage from softirq context: */ 1537 unsigned int kcov_softirq; 1538 #endif 1539 1540 #ifdef CONFIG_MEMCG_V1 1541 struct mem_cgroup *memcg_in_oom; 1542 #endif 1543 1544 #ifdef CONFIG_MEMCG 1545 /* Number of pages to reclaim on returning to userland: */ 1546 unsigned int memcg_nr_pages_over_high; 1547 1548 /* Used by memcontrol for targeted memcg charge: */ 1549 struct mem_cgroup *active_memcg; 1550 1551 /* Cache for current->cgroups->memcg->objcg lookups: */ 1552 struct obj_cgroup *objcg; 1553 #endif 1554 1555 #ifdef CONFIG_BLK_CGROUP 1556 struct gendisk *throttle_disk; 1557 #endif 1558 1559 #ifdef CONFIG_UPROBES 1560 struct uprobe_task *utask; 1561 #endif 1562 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1563 unsigned int sequential_io; 1564 unsigned int sequential_io_avg; 1565 #endif 1566 struct kmap_ctrl kmap_ctrl; 1567 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1568 unsigned long task_state_change; 1569 # ifdef CONFIG_PREEMPT_RT 1570 unsigned long saved_state_change; 1571 # endif 1572 #endif 1573 struct rcu_head rcu; 1574 refcount_t rcu_users; 1575 int pagefault_disabled; 1576 #ifdef CONFIG_MMU 1577 struct task_struct *oom_reaper_list; 1578 struct timer_list oom_reaper_timer; 1579 #endif 1580 #ifdef CONFIG_VMAP_STACK 1581 struct vm_struct *stack_vm_area; 1582 #endif 1583 #ifdef CONFIG_THREAD_INFO_IN_TASK 1584 /* A live task holds one reference: */ 1585 refcount_t stack_refcount; 1586 #endif 1587 #ifdef CONFIG_LIVEPATCH 1588 int patch_state; 1589 #endif 1590 #ifdef CONFIG_SECURITY 1591 /* Used by LSM modules for access restriction: */ 1592 void *security; 1593 #endif 1594 #ifdef CONFIG_BPF_SYSCALL 1595 /* Used by BPF task local storage */ 1596 struct bpf_local_storage __rcu *bpf_storage; 1597 /* Used for BPF run context */ 1598 struct bpf_run_ctx *bpf_ctx; 1599 #endif 1600 /* Used by BPF for per-TASK xdp storage */ 1601 struct bpf_net_context *bpf_net_context; 1602 1603 #ifdef CONFIG_KSTACK_ERASE 1604 unsigned long lowest_stack; 1605 #endif 1606 #ifdef CONFIG_KSTACK_ERASE_METRICS 1607 unsigned long prev_lowest_stack; 1608 #endif 1609 1610 #ifdef CONFIG_X86_MCE 1611 void __user *mce_vaddr; 1612 __u64 mce_kflags; 1613 u64 mce_addr; 1614 __u64 mce_ripv : 1, 1615 mce_whole_page : 1, 1616 __mce_reserved : 62; 1617 struct callback_head mce_kill_me; 1618 int mce_count; 1619 #endif 1620 1621 #ifdef CONFIG_KRETPROBES 1622 struct llist_head kretprobe_instances; 1623 #endif 1624 #ifdef CONFIG_RETHOOK 1625 struct llist_head rethooks; 1626 #endif 1627 1628 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1629 /* 1630 * If L1D flush is supported on mm context switch 1631 * then we use this callback head to queue kill work 1632 * to kill tasks that are not running on SMT disabled 1633 * cores 1634 */ 1635 struct callback_head l1d_flush_kill; 1636 #endif 1637 1638 #ifdef CONFIG_RV 1639 /* 1640 * Per-task RV monitor, fixed in CONFIG_RV_PER_TASK_MONITORS. 1641 * If memory becomes a concern, we can think about a dynamic method. 1642 */ 1643 union rv_task_monitor rv[CONFIG_RV_PER_TASK_MONITORS]; 1644 #endif 1645 1646 #ifdef CONFIG_USER_EVENTS 1647 struct user_event_mm *user_event_mm; 1648 #endif 1649 1650 #ifdef CONFIG_UNWIND_USER 1651 struct unwind_task_info unwind_info; 1652 #endif 1653 1654 /* CPU-specific state of this task: */ 1655 struct thread_struct thread; 1656 1657 /* 1658 * New fields for task_struct should be added above here, so that 1659 * they are included in the randomized portion of task_struct. 1660 */ 1661 randomized_struct_fields_end 1662 } __attribute__ ((aligned (64))); 1663 1664 #ifdef CONFIG_SCHED_PROXY_EXEC 1665 DECLARE_STATIC_KEY_TRUE(__sched_proxy_exec); 1666 static inline bool sched_proxy_exec(void) 1667 { 1668 return static_branch_likely(&__sched_proxy_exec); 1669 } 1670 #else 1671 static inline bool sched_proxy_exec(void) 1672 { 1673 return false; 1674 } 1675 #endif 1676 1677 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1678 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1679 1680 static inline unsigned int __task_state_index(unsigned int tsk_state, 1681 unsigned int tsk_exit_state) 1682 { 1683 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1684 1685 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1686 1687 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1688 state = TASK_REPORT_IDLE; 1689 1690 /* 1691 * We're lying here, but rather than expose a completely new task state 1692 * to userspace, we can make this appear as if the task has gone through 1693 * a regular rt_mutex_lock() call. 1694 * Report frozen tasks as uninterruptible. 1695 */ 1696 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN)) 1697 state = TASK_UNINTERRUPTIBLE; 1698 1699 return fls(state); 1700 } 1701 1702 static inline unsigned int task_state_index(struct task_struct *tsk) 1703 { 1704 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1705 } 1706 1707 static inline char task_index_to_char(unsigned int state) 1708 { 1709 static const char state_char[] = "RSDTtXZPI"; 1710 1711 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1)); 1712 1713 return state_char[state]; 1714 } 1715 1716 static inline char task_state_to_char(struct task_struct *tsk) 1717 { 1718 return task_index_to_char(task_state_index(tsk)); 1719 } 1720 1721 extern struct pid *cad_pid; 1722 1723 /* 1724 * Per process flags 1725 */ 1726 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1727 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1728 #define PF_EXITING 0x00000004 /* Getting shut down */ 1729 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1730 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1731 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1732 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1733 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1734 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1735 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1736 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1737 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1738 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1739 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1740 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1741 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1742 #define PF_KCOMPACTD 0x00010000 /* I am kcompactd */ 1743 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1744 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1745 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1746 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1747 * I am cleaning dirty pages from some other bdi. */ 1748 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1749 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1750 #define PF__HOLE__00800000 0x00800000 1751 #define PF__HOLE__01000000 0x01000000 1752 #define PF__HOLE__02000000 0x02000000 1753 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1754 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1755 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1756 * See memalloc_pin_save() */ 1757 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1758 #define PF__HOLE__40000000 0x40000000 1759 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1760 1761 /* 1762 * Only the _current_ task can read/write to tsk->flags, but other 1763 * tasks can access tsk->flags in readonly mode for example 1764 * with tsk_used_math (like during threaded core dumping). 1765 * There is however an exception to this rule during ptrace 1766 * or during fork: the ptracer task is allowed to write to the 1767 * child->flags of its traced child (same goes for fork, the parent 1768 * can write to the child->flags), because we're guaranteed the 1769 * child is not running and in turn not changing child->flags 1770 * at the same time the parent does it. 1771 */ 1772 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1773 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1774 #define clear_used_math() clear_stopped_child_used_math(current) 1775 #define set_used_math() set_stopped_child_used_math(current) 1776 1777 #define conditional_stopped_child_used_math(condition, child) \ 1778 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1779 1780 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1781 1782 #define copy_to_stopped_child_used_math(child) \ 1783 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1784 1785 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1786 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1787 #define used_math() tsk_used_math(current) 1788 1789 static __always_inline bool is_percpu_thread(void) 1790 { 1791 return (current->flags & PF_NO_SETAFFINITY) && 1792 (current->nr_cpus_allowed == 1); 1793 } 1794 1795 /* Per-process atomic flags. */ 1796 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1797 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1798 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1799 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1800 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1801 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1802 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1803 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1804 1805 #define TASK_PFA_TEST(name, func) \ 1806 static inline bool task_##func(struct task_struct *p) \ 1807 { return test_bit(PFA_##name, &p->atomic_flags); } 1808 1809 #define TASK_PFA_SET(name, func) \ 1810 static inline void task_set_##func(struct task_struct *p) \ 1811 { set_bit(PFA_##name, &p->atomic_flags); } 1812 1813 #define TASK_PFA_CLEAR(name, func) \ 1814 static inline void task_clear_##func(struct task_struct *p) \ 1815 { clear_bit(PFA_##name, &p->atomic_flags); } 1816 1817 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1818 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1819 1820 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1821 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1822 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1823 1824 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1825 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1826 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1827 1828 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1829 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1830 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1831 1832 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1833 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1834 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1835 1836 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1837 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1838 1839 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1840 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1841 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1842 1843 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1844 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1845 1846 static inline void 1847 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1848 { 1849 current->flags &= ~flags; 1850 current->flags |= orig_flags & flags; 1851 } 1852 1853 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1854 extern int task_can_attach(struct task_struct *p); 1855 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1856 extern void dl_bw_free(int cpu, u64 dl_bw); 1857 1858 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1859 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1860 1861 /** 1862 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1863 * @p: the task 1864 * @new_mask: CPU affinity mask 1865 * 1866 * Return: zero if successful, or a negative error code 1867 */ 1868 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1869 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1870 extern void release_user_cpus_ptr(struct task_struct *p); 1871 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1872 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1873 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1874 1875 extern int yield_to(struct task_struct *p, bool preempt); 1876 extern void set_user_nice(struct task_struct *p, long nice); 1877 extern int task_prio(const struct task_struct *p); 1878 1879 /** 1880 * task_nice - return the nice value of a given task. 1881 * @p: the task in question. 1882 * 1883 * Return: The nice value [ -20 ... 0 ... 19 ]. 1884 */ 1885 static inline int task_nice(const struct task_struct *p) 1886 { 1887 return PRIO_TO_NICE((p)->static_prio); 1888 } 1889 1890 extern int can_nice(const struct task_struct *p, const int nice); 1891 extern int task_curr(const struct task_struct *p); 1892 extern int idle_cpu(int cpu); 1893 extern int available_idle_cpu(int cpu); 1894 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1895 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1896 extern void sched_set_fifo(struct task_struct *p); 1897 extern void sched_set_fifo_low(struct task_struct *p); 1898 extern void sched_set_normal(struct task_struct *p, int nice); 1899 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1900 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1901 extern struct task_struct *idle_task(int cpu); 1902 1903 /** 1904 * is_idle_task - is the specified task an idle task? 1905 * @p: the task in question. 1906 * 1907 * Return: 1 if @p is an idle task. 0 otherwise. 1908 */ 1909 static __always_inline bool is_idle_task(const struct task_struct *p) 1910 { 1911 return !!(p->flags & PF_IDLE); 1912 } 1913 1914 extern struct task_struct *curr_task(int cpu); 1915 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1916 1917 void yield(void); 1918 1919 union thread_union { 1920 struct task_struct task; 1921 #ifndef CONFIG_THREAD_INFO_IN_TASK 1922 struct thread_info thread_info; 1923 #endif 1924 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1925 }; 1926 1927 #ifndef CONFIG_THREAD_INFO_IN_TASK 1928 extern struct thread_info init_thread_info; 1929 #endif 1930 1931 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1932 1933 #ifdef CONFIG_THREAD_INFO_IN_TASK 1934 # define task_thread_info(task) (&(task)->thread_info) 1935 #else 1936 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1937 #endif 1938 1939 /* 1940 * find a task by one of its numerical ids 1941 * 1942 * find_task_by_pid_ns(): 1943 * finds a task by its pid in the specified namespace 1944 * find_task_by_vpid(): 1945 * finds a task by its virtual pid 1946 * 1947 * see also find_vpid() etc in include/linux/pid.h 1948 */ 1949 1950 extern struct task_struct *find_task_by_vpid(pid_t nr); 1951 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1952 1953 /* 1954 * find a task by its virtual pid and get the task struct 1955 */ 1956 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1957 1958 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1959 extern int wake_up_process(struct task_struct *tsk); 1960 extern void wake_up_new_task(struct task_struct *tsk); 1961 1962 extern void kick_process(struct task_struct *tsk); 1963 1964 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1965 #define set_task_comm(tsk, from) ({ \ 1966 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \ 1967 __set_task_comm(tsk, from, false); \ 1968 }) 1969 1970 /* 1971 * - Why not use task_lock()? 1972 * User space can randomly change their names anyway, so locking for readers 1973 * doesn't make sense. For writers, locking is probably necessary, as a race 1974 * condition could lead to long-term mixed results. 1975 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is 1976 * always NUL-terminated and zero-padded. Therefore the race condition between 1977 * reader and writer is not an issue. 1978 * 1979 * - BUILD_BUG_ON() can help prevent the buf from being truncated. 1980 * Since the callers don't perform any return value checks, this safeguard is 1981 * necessary. 1982 */ 1983 #define get_task_comm(buf, tsk) ({ \ 1984 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \ 1985 strscpy_pad(buf, (tsk)->comm); \ 1986 buf; \ 1987 }) 1988 1989 static __always_inline void scheduler_ipi(void) 1990 { 1991 /* 1992 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1993 * TIF_NEED_RESCHED remotely (for the first time) will also send 1994 * this IPI. 1995 */ 1996 preempt_fold_need_resched(); 1997 } 1998 1999 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 2000 2001 /* 2002 * Set thread flags in other task's structures. 2003 * See asm/thread_info.h for TIF_xxxx flags available: 2004 */ 2005 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 2006 { 2007 set_ti_thread_flag(task_thread_info(tsk), flag); 2008 } 2009 2010 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2011 { 2012 clear_ti_thread_flag(task_thread_info(tsk), flag); 2013 } 2014 2015 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 2016 bool value) 2017 { 2018 update_ti_thread_flag(task_thread_info(tsk), flag, value); 2019 } 2020 2021 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2022 { 2023 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2024 } 2025 2026 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2027 { 2028 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2029 } 2030 2031 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2032 { 2033 return test_ti_thread_flag(task_thread_info(tsk), flag); 2034 } 2035 2036 static inline void set_tsk_need_resched(struct task_struct *tsk) 2037 { 2038 if (tracepoint_enabled(sched_set_need_resched_tp) && 2039 !test_tsk_thread_flag(tsk, TIF_NEED_RESCHED)) 2040 __trace_set_need_resched(tsk, TIF_NEED_RESCHED); 2041 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2042 } 2043 2044 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2045 { 2046 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY, 2047 (atomic_long_t *)&task_thread_info(tsk)->flags); 2048 } 2049 2050 static inline int test_tsk_need_resched(struct task_struct *tsk) 2051 { 2052 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2053 } 2054 2055 /* 2056 * cond_resched() and cond_resched_lock(): latency reduction via 2057 * explicit rescheduling in places that are safe. The return 2058 * value indicates whether a reschedule was done in fact. 2059 * cond_resched_lock() will drop the spinlock before scheduling, 2060 */ 2061 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2062 extern int __cond_resched(void); 2063 2064 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2065 2066 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2067 2068 static __always_inline int _cond_resched(void) 2069 { 2070 return static_call_mod(cond_resched)(); 2071 } 2072 2073 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2074 2075 extern int dynamic_cond_resched(void); 2076 2077 static __always_inline int _cond_resched(void) 2078 { 2079 return dynamic_cond_resched(); 2080 } 2081 2082 #else /* !CONFIG_PREEMPTION */ 2083 2084 static inline int _cond_resched(void) 2085 { 2086 return __cond_resched(); 2087 } 2088 2089 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2090 2091 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2092 2093 static inline int _cond_resched(void) 2094 { 2095 return 0; 2096 } 2097 2098 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2099 2100 #define cond_resched() ({ \ 2101 __might_resched(__FILE__, __LINE__, 0); \ 2102 _cond_resched(); \ 2103 }) 2104 2105 extern int __cond_resched_lock(spinlock_t *lock); 2106 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2107 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2108 2109 #define MIGHT_RESCHED_RCU_SHIFT 8 2110 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2111 2112 #ifndef CONFIG_PREEMPT_RT 2113 /* 2114 * Non RT kernels have an elevated preempt count due to the held lock, 2115 * but are not allowed to be inside a RCU read side critical section 2116 */ 2117 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2118 #else 2119 /* 2120 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2121 * cond_resched*lock() has to take that into account because it checks for 2122 * preempt_count() and rcu_preempt_depth(). 2123 */ 2124 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2125 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2126 #endif 2127 2128 #define cond_resched_lock(lock) ({ \ 2129 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2130 __cond_resched_lock(lock); \ 2131 }) 2132 2133 #define cond_resched_rwlock_read(lock) ({ \ 2134 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2135 __cond_resched_rwlock_read(lock); \ 2136 }) 2137 2138 #define cond_resched_rwlock_write(lock) ({ \ 2139 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2140 __cond_resched_rwlock_write(lock); \ 2141 }) 2142 2143 #ifndef CONFIG_PREEMPT_RT 2144 static inline struct mutex *__get_task_blocked_on(struct task_struct *p) 2145 { 2146 struct mutex *m = p->blocked_on; 2147 2148 if (m) 2149 lockdep_assert_held_once(&m->wait_lock); 2150 return m; 2151 } 2152 2153 static inline void __set_task_blocked_on(struct task_struct *p, struct mutex *m) 2154 { 2155 struct mutex *blocked_on = READ_ONCE(p->blocked_on); 2156 2157 WARN_ON_ONCE(!m); 2158 /* The task should only be setting itself as blocked */ 2159 WARN_ON_ONCE(p != current); 2160 /* Currently we serialize blocked_on under the mutex::wait_lock */ 2161 lockdep_assert_held_once(&m->wait_lock); 2162 /* 2163 * Check ensure we don't overwrite existing mutex value 2164 * with a different mutex. Note, setting it to the same 2165 * lock repeatedly is ok. 2166 */ 2167 WARN_ON_ONCE(blocked_on && blocked_on != m); 2168 WRITE_ONCE(p->blocked_on, m); 2169 } 2170 2171 static inline void set_task_blocked_on(struct task_struct *p, struct mutex *m) 2172 { 2173 guard(raw_spinlock_irqsave)(&m->wait_lock); 2174 __set_task_blocked_on(p, m); 2175 } 2176 2177 static inline void __clear_task_blocked_on(struct task_struct *p, struct mutex *m) 2178 { 2179 if (m) { 2180 struct mutex *blocked_on = READ_ONCE(p->blocked_on); 2181 2182 /* Currently we serialize blocked_on under the mutex::wait_lock */ 2183 lockdep_assert_held_once(&m->wait_lock); 2184 /* 2185 * There may be cases where we re-clear already cleared 2186 * blocked_on relationships, but make sure we are not 2187 * clearing the relationship with a different lock. 2188 */ 2189 WARN_ON_ONCE(blocked_on && blocked_on != m); 2190 } 2191 WRITE_ONCE(p->blocked_on, NULL); 2192 } 2193 2194 static inline void clear_task_blocked_on(struct task_struct *p, struct mutex *m) 2195 { 2196 guard(raw_spinlock_irqsave)(&m->wait_lock); 2197 __clear_task_blocked_on(p, m); 2198 } 2199 #else 2200 static inline void __clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m) 2201 { 2202 } 2203 2204 static inline void clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m) 2205 { 2206 } 2207 #endif /* !CONFIG_PREEMPT_RT */ 2208 2209 static __always_inline bool need_resched(void) 2210 { 2211 return unlikely(tif_need_resched()); 2212 } 2213 2214 /* 2215 * Wrappers for p->thread_info->cpu access. No-op on UP. 2216 */ 2217 #ifdef CONFIG_SMP 2218 2219 static inline unsigned int task_cpu(const struct task_struct *p) 2220 { 2221 return READ_ONCE(task_thread_info(p)->cpu); 2222 } 2223 2224 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2225 2226 #else 2227 2228 static inline unsigned int task_cpu(const struct task_struct *p) 2229 { 2230 return 0; 2231 } 2232 2233 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2234 { 2235 } 2236 2237 #endif /* CONFIG_SMP */ 2238 2239 static inline bool task_is_runnable(struct task_struct *p) 2240 { 2241 return p->on_rq && !p->se.sched_delayed; 2242 } 2243 2244 extern bool sched_task_on_rq(struct task_struct *p); 2245 extern unsigned long get_wchan(struct task_struct *p); 2246 extern struct task_struct *cpu_curr_snapshot(int cpu); 2247 2248 /* 2249 * In order to reduce various lock holder preemption latencies provide an 2250 * interface to see if a vCPU is currently running or not. 2251 * 2252 * This allows us to terminate optimistic spin loops and block, analogous to 2253 * the native optimistic spin heuristic of testing if the lock owner task is 2254 * running or not. 2255 */ 2256 #ifndef vcpu_is_preempted 2257 static inline bool vcpu_is_preempted(int cpu) 2258 { 2259 return false; 2260 } 2261 #endif 2262 2263 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2264 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2265 2266 #ifndef TASK_SIZE_OF 2267 #define TASK_SIZE_OF(tsk) TASK_SIZE 2268 #endif 2269 2270 static inline bool owner_on_cpu(struct task_struct *owner) 2271 { 2272 /* 2273 * As lock holder preemption issue, we both skip spinning if 2274 * task is not on cpu or its cpu is preempted 2275 */ 2276 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2277 } 2278 2279 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2280 unsigned long sched_cpu_util(int cpu); 2281 2282 #ifdef CONFIG_SCHED_CORE 2283 extern void sched_core_free(struct task_struct *tsk); 2284 extern void sched_core_fork(struct task_struct *p); 2285 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2286 unsigned long uaddr); 2287 extern int sched_core_idle_cpu(int cpu); 2288 #else 2289 static inline void sched_core_free(struct task_struct *tsk) { } 2290 static inline void sched_core_fork(struct task_struct *p) { } 2291 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2292 #endif 2293 2294 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2295 2296 #ifdef CONFIG_MEM_ALLOC_PROFILING 2297 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2298 { 2299 swap(current->alloc_tag, tag); 2300 return tag; 2301 } 2302 2303 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2304 { 2305 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2306 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2307 #endif 2308 current->alloc_tag = old; 2309 } 2310 #else 2311 #define alloc_tag_save(_tag) NULL 2312 #define alloc_tag_restore(_tag, _old) do {} while (0) 2313 #endif 2314 2315 #endif 2316