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; 709 710 /* 711 * Bandwidth enforcement timer. Each -deadline task has its 712 * own bandwidth to be enforced, thus we need one timer per task. 713 */ 714 struct hrtimer dl_timer; 715 716 /* 717 * Inactive timer, responsible for decreasing the active utilization 718 * at the "0-lag time". When a -deadline task blocks, it contributes 719 * to GRUB's active utilization until the "0-lag time", hence a 720 * timer is needed to decrease the active utilization at the correct 721 * time. 722 */ 723 struct hrtimer inactive_timer; 724 725 /* 726 * Bits for DL-server functionality. Also see the comment near 727 * dl_server_update(). 728 * 729 * @rq the runqueue this server is for 730 * 731 * @server_has_tasks() returns true if @server_pick return a 732 * runnable task. 733 */ 734 struct rq *rq; 735 dl_server_pick_f server_pick_task; 736 737 #ifdef CONFIG_RT_MUTEXES 738 /* 739 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 740 * pi_se points to the donor, otherwise points to the dl_se it belongs 741 * to (the original one/itself). 742 */ 743 struct sched_dl_entity *pi_se; 744 #endif 745 }; 746 747 #ifdef CONFIG_UCLAMP_TASK 748 /* Number of utilization clamp buckets (shorter alias) */ 749 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 750 751 /* 752 * Utilization clamp for a scheduling entity 753 * @value: clamp value "assigned" to a se 754 * @bucket_id: bucket index corresponding to the "assigned" value 755 * @active: the se is currently refcounted in a rq's bucket 756 * @user_defined: the requested clamp value comes from user-space 757 * 758 * The bucket_id is the index of the clamp bucket matching the clamp value 759 * which is pre-computed and stored to avoid expensive integer divisions from 760 * the fast path. 761 * 762 * The active bit is set whenever a task has got an "effective" value assigned, 763 * which can be different from the clamp value "requested" from user-space. 764 * This allows to know a task is refcounted in the rq's bucket corresponding 765 * to the "effective" bucket_id. 766 * 767 * The user_defined bit is set whenever a task has got a task-specific clamp 768 * value requested from userspace, i.e. the system defaults apply to this task 769 * just as a restriction. This allows to relax default clamps when a less 770 * restrictive task-specific value has been requested, thus allowing to 771 * implement a "nice" semantic. For example, a task running with a 20% 772 * default boost can still drop its own boosting to 0%. 773 */ 774 struct uclamp_se { 775 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 776 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 777 unsigned int active : 1; 778 unsigned int user_defined : 1; 779 }; 780 #endif /* CONFIG_UCLAMP_TASK */ 781 782 union rcu_special { 783 struct { 784 u8 blocked; 785 u8 need_qs; 786 u8 exp_hint; /* Hint for performance. */ 787 u8 need_mb; /* Readers need smp_mb(). */ 788 } b; /* Bits. */ 789 u32 s; /* Set of bits. */ 790 }; 791 792 enum perf_event_task_context { 793 perf_invalid_context = -1, 794 perf_hw_context = 0, 795 perf_sw_context, 796 perf_nr_task_contexts, 797 }; 798 799 /* 800 * Number of contexts where an event can trigger: 801 * task, softirq, hardirq, nmi. 802 */ 803 #define PERF_NR_CONTEXTS 4 804 805 struct wake_q_node { 806 struct wake_q_node *next; 807 }; 808 809 struct kmap_ctrl { 810 #ifdef CONFIG_KMAP_LOCAL 811 int idx; 812 pte_t pteval[KM_MAX_IDX]; 813 #endif 814 }; 815 816 struct task_struct { 817 #ifdef CONFIG_THREAD_INFO_IN_TASK 818 /* 819 * For reasons of header soup (see current_thread_info()), this 820 * must be the first element of task_struct. 821 */ 822 struct thread_info thread_info; 823 #endif 824 unsigned int __state; 825 826 /* saved state for "spinlock sleepers" */ 827 unsigned int saved_state; 828 829 /* 830 * This begins the randomizable portion of task_struct. Only 831 * scheduling-critical items should be added above here. 832 */ 833 randomized_struct_fields_start 834 835 void *stack; 836 refcount_t usage; 837 /* Per task flags (PF_*), defined further below: */ 838 unsigned int flags; 839 unsigned int ptrace; 840 841 #ifdef CONFIG_MEM_ALLOC_PROFILING 842 struct alloc_tag *alloc_tag; 843 #endif 844 845 int on_cpu; 846 struct __call_single_node wake_entry; 847 unsigned int wakee_flips; 848 unsigned long wakee_flip_decay_ts; 849 struct task_struct *last_wakee; 850 851 /* 852 * recent_used_cpu is initially set as the last CPU used by a task 853 * that wakes affine another task. Waker/wakee relationships can 854 * push tasks around a CPU where each wakeup moves to the next one. 855 * Tracking a recently used CPU allows a quick search for a recently 856 * used CPU that may be idle. 857 */ 858 int recent_used_cpu; 859 int wake_cpu; 860 int on_rq; 861 862 int prio; 863 int static_prio; 864 int normal_prio; 865 unsigned int rt_priority; 866 867 struct sched_entity se; 868 struct sched_rt_entity rt; 869 struct sched_dl_entity dl; 870 struct sched_dl_entity *dl_server; 871 #ifdef CONFIG_SCHED_CLASS_EXT 872 struct sched_ext_entity scx; 873 #endif 874 const struct sched_class *sched_class; 875 876 #ifdef CONFIG_SCHED_CORE 877 struct rb_node core_node; 878 unsigned long core_cookie; 879 unsigned int core_occupation; 880 #endif 881 882 #ifdef CONFIG_CGROUP_SCHED 883 struct task_group *sched_task_group; 884 #endif 885 886 887 #ifdef CONFIG_UCLAMP_TASK 888 /* 889 * Clamp values requested for a scheduling entity. 890 * Must be updated with task_rq_lock() held. 891 */ 892 struct uclamp_se uclamp_req[UCLAMP_CNT]; 893 /* 894 * Effective clamp values used for a scheduling entity. 895 * Must be updated with task_rq_lock() held. 896 */ 897 struct uclamp_se uclamp[UCLAMP_CNT]; 898 #endif 899 900 struct sched_statistics stats; 901 902 #ifdef CONFIG_PREEMPT_NOTIFIERS 903 /* List of struct preempt_notifier: */ 904 struct hlist_head preempt_notifiers; 905 #endif 906 907 #ifdef CONFIG_BLK_DEV_IO_TRACE 908 unsigned int btrace_seq; 909 #endif 910 911 unsigned int policy; 912 unsigned long max_allowed_capacity; 913 int nr_cpus_allowed; 914 const cpumask_t *cpus_ptr; 915 cpumask_t *user_cpus_ptr; 916 cpumask_t cpus_mask; 917 void *migration_pending; 918 unsigned short migration_disabled; 919 unsigned short migration_flags; 920 921 #ifdef CONFIG_PREEMPT_RCU 922 int rcu_read_lock_nesting; 923 union rcu_special rcu_read_unlock_special; 924 struct list_head rcu_node_entry; 925 struct rcu_node *rcu_blocked_node; 926 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 927 928 #ifdef CONFIG_TASKS_RCU 929 unsigned long rcu_tasks_nvcsw; 930 u8 rcu_tasks_holdout; 931 u8 rcu_tasks_idx; 932 int rcu_tasks_idle_cpu; 933 struct list_head rcu_tasks_holdout_list; 934 int rcu_tasks_exit_cpu; 935 struct list_head rcu_tasks_exit_list; 936 #endif /* #ifdef CONFIG_TASKS_RCU */ 937 938 #ifdef CONFIG_TASKS_TRACE_RCU 939 int trc_reader_nesting; 940 int trc_ipi_to_cpu; 941 union rcu_special trc_reader_special; 942 struct list_head trc_holdout_list; 943 struct list_head trc_blkd_node; 944 int trc_blkd_cpu; 945 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 946 947 struct sched_info sched_info; 948 949 struct list_head tasks; 950 struct plist_node pushable_tasks; 951 struct rb_node pushable_dl_tasks; 952 953 struct mm_struct *mm; 954 struct mm_struct *active_mm; 955 struct address_space *faults_disabled_mapping; 956 957 int exit_state; 958 int exit_code; 959 int exit_signal; 960 /* The signal sent when the parent dies: */ 961 int pdeath_signal; 962 /* JOBCTL_*, siglock protected: */ 963 unsigned long jobctl; 964 965 /* Used for emulating ABI behavior of previous Linux versions: */ 966 unsigned int personality; 967 968 /* Scheduler bits, serialized by scheduler locks: */ 969 unsigned sched_reset_on_fork:1; 970 unsigned sched_contributes_to_load:1; 971 unsigned sched_migrated:1; 972 unsigned sched_task_hot:1; 973 974 /* Force alignment to the next boundary: */ 975 unsigned :0; 976 977 /* Unserialized, strictly 'current' */ 978 979 /* 980 * This field must not be in the scheduler word above due to wakelist 981 * queueing no longer being serialized by p->on_cpu. However: 982 * 983 * p->XXX = X; ttwu() 984 * schedule() if (p->on_rq && ..) // false 985 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 986 * deactivate_task() ttwu_queue_wakelist()) 987 * p->on_rq = 0; p->sched_remote_wakeup = Y; 988 * 989 * guarantees all stores of 'current' are visible before 990 * ->sched_remote_wakeup gets used, so it can be in this word. 991 */ 992 unsigned sched_remote_wakeup:1; 993 #ifdef CONFIG_RT_MUTEXES 994 unsigned sched_rt_mutex:1; 995 #endif 996 997 /* Bit to tell TOMOYO we're in execve(): */ 998 unsigned in_execve:1; 999 unsigned in_iowait:1; 1000 #ifndef TIF_RESTORE_SIGMASK 1001 unsigned restore_sigmask:1; 1002 #endif 1003 #ifdef CONFIG_MEMCG_V1 1004 unsigned in_user_fault:1; 1005 #endif 1006 #ifdef CONFIG_LRU_GEN 1007 /* whether the LRU algorithm may apply to this access */ 1008 unsigned in_lru_fault:1; 1009 #endif 1010 #ifdef CONFIG_COMPAT_BRK 1011 unsigned brk_randomized:1; 1012 #endif 1013 #ifdef CONFIG_CGROUPS 1014 /* disallow userland-initiated cgroup migration */ 1015 unsigned no_cgroup_migration:1; 1016 /* task is frozen/stopped (used by the cgroup freezer) */ 1017 unsigned frozen:1; 1018 #endif 1019 #ifdef CONFIG_BLK_CGROUP 1020 unsigned use_memdelay:1; 1021 #endif 1022 #ifdef CONFIG_PSI 1023 /* Stalled due to lack of memory */ 1024 unsigned in_memstall:1; 1025 #endif 1026 #ifdef CONFIG_PAGE_OWNER 1027 /* Used by page_owner=on to detect recursion in page tracking. */ 1028 unsigned in_page_owner:1; 1029 #endif 1030 #ifdef CONFIG_EVENTFD 1031 /* Recursion prevention for eventfd_signal() */ 1032 unsigned in_eventfd:1; 1033 #endif 1034 #ifdef CONFIG_ARCH_HAS_CPU_PASID 1035 unsigned pasid_activated:1; 1036 #endif 1037 #ifdef CONFIG_X86_BUS_LOCK_DETECT 1038 unsigned reported_split_lock:1; 1039 #endif 1040 #ifdef CONFIG_TASK_DELAY_ACCT 1041 /* delay due to memory thrashing */ 1042 unsigned in_thrashing:1; 1043 #endif 1044 unsigned in_nf_duplicate:1; 1045 #ifdef CONFIG_PREEMPT_RT 1046 struct netdev_xmit net_xmit; 1047 #endif 1048 unsigned long atomic_flags; /* Flags requiring atomic access. */ 1049 1050 struct restart_block restart_block; 1051 1052 pid_t pid; 1053 pid_t tgid; 1054 1055 #ifdef CONFIG_STACKPROTECTOR 1056 /* Canary value for the -fstack-protector GCC feature: */ 1057 unsigned long stack_canary; 1058 #endif 1059 /* 1060 * Pointers to the (original) parent process, youngest child, younger sibling, 1061 * older sibling, respectively. (p->father can be replaced with 1062 * p->real_parent->pid) 1063 */ 1064 1065 /* Real parent process: */ 1066 struct task_struct __rcu *real_parent; 1067 1068 /* Recipient of SIGCHLD, wait4() reports: */ 1069 struct task_struct __rcu *parent; 1070 1071 /* 1072 * Children/sibling form the list of natural children: 1073 */ 1074 struct list_head children; 1075 struct list_head sibling; 1076 struct task_struct *group_leader; 1077 1078 /* 1079 * 'ptraced' is the list of tasks this task is using ptrace() on. 1080 * 1081 * This includes both natural children and PTRACE_ATTACH targets. 1082 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1083 */ 1084 struct list_head ptraced; 1085 struct list_head ptrace_entry; 1086 1087 /* PID/PID hash table linkage. */ 1088 struct pid *thread_pid; 1089 struct hlist_node pid_links[PIDTYPE_MAX]; 1090 struct list_head thread_node; 1091 1092 struct completion *vfork_done; 1093 1094 /* CLONE_CHILD_SETTID: */ 1095 int __user *set_child_tid; 1096 1097 /* CLONE_CHILD_CLEARTID: */ 1098 int __user *clear_child_tid; 1099 1100 /* PF_KTHREAD | PF_IO_WORKER */ 1101 void *worker_private; 1102 1103 u64 utime; 1104 u64 stime; 1105 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1106 u64 utimescaled; 1107 u64 stimescaled; 1108 #endif 1109 u64 gtime; 1110 struct prev_cputime prev_cputime; 1111 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1112 struct vtime vtime; 1113 #endif 1114 1115 #ifdef CONFIG_NO_HZ_FULL 1116 atomic_t tick_dep_mask; 1117 #endif 1118 /* Context switch counts: */ 1119 unsigned long nvcsw; 1120 unsigned long nivcsw; 1121 1122 /* Monotonic time in nsecs: */ 1123 u64 start_time; 1124 1125 /* Boot based time in nsecs: */ 1126 u64 start_boottime; 1127 1128 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1129 unsigned long min_flt; 1130 unsigned long maj_flt; 1131 1132 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1133 struct posix_cputimers posix_cputimers; 1134 1135 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1136 struct posix_cputimers_work posix_cputimers_work; 1137 #endif 1138 1139 /* Process credentials: */ 1140 1141 /* Tracer's credentials at attach: */ 1142 const struct cred __rcu *ptracer_cred; 1143 1144 /* Objective and real subjective task credentials (COW): */ 1145 const struct cred __rcu *real_cred; 1146 1147 /* Effective (overridable) subjective task credentials (COW): */ 1148 const struct cred __rcu *cred; 1149 1150 #ifdef CONFIG_KEYS 1151 /* Cached requested key. */ 1152 struct key *cached_requested_key; 1153 #endif 1154 1155 /* 1156 * executable name, excluding path. 1157 * 1158 * - normally initialized begin_new_exec() 1159 * - set it with set_task_comm() 1160 * - strscpy_pad() to ensure it is always NUL-terminated and 1161 * zero-padded 1162 * - task_lock() to ensure the operation is atomic and the name is 1163 * fully updated. 1164 */ 1165 char comm[TASK_COMM_LEN]; 1166 1167 struct nameidata *nameidata; 1168 1169 #ifdef CONFIG_SYSVIPC 1170 struct sysv_sem sysvsem; 1171 struct sysv_shm sysvshm; 1172 #endif 1173 #ifdef CONFIG_DETECT_HUNG_TASK 1174 unsigned long last_switch_count; 1175 unsigned long last_switch_time; 1176 #endif 1177 /* Filesystem information: */ 1178 struct fs_struct *fs; 1179 1180 /* Open file information: */ 1181 struct files_struct *files; 1182 1183 #ifdef CONFIG_IO_URING 1184 struct io_uring_task *io_uring; 1185 #endif 1186 1187 /* Namespaces: */ 1188 struct nsproxy *nsproxy; 1189 1190 /* Signal handlers: */ 1191 struct signal_struct *signal; 1192 struct sighand_struct __rcu *sighand; 1193 sigset_t blocked; 1194 sigset_t real_blocked; 1195 /* Restored if set_restore_sigmask() was used: */ 1196 sigset_t saved_sigmask; 1197 struct sigpending pending; 1198 unsigned long sas_ss_sp; 1199 size_t sas_ss_size; 1200 unsigned int sas_ss_flags; 1201 1202 struct callback_head *task_works; 1203 1204 #ifdef CONFIG_AUDIT 1205 #ifdef CONFIG_AUDITSYSCALL 1206 struct audit_context *audit_context; 1207 #endif 1208 kuid_t loginuid; 1209 unsigned int sessionid; 1210 #endif 1211 struct seccomp seccomp; 1212 struct syscall_user_dispatch syscall_dispatch; 1213 1214 /* Thread group tracking: */ 1215 u64 parent_exec_id; 1216 u64 self_exec_id; 1217 1218 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1219 spinlock_t alloc_lock; 1220 1221 /* Protection of the PI data structures: */ 1222 raw_spinlock_t pi_lock; 1223 1224 struct wake_q_node wake_q; 1225 1226 #ifdef CONFIG_RT_MUTEXES 1227 /* PI waiters blocked on a rt_mutex held by this task: */ 1228 struct rb_root_cached pi_waiters; 1229 /* Updated under owner's pi_lock and rq lock */ 1230 struct task_struct *pi_top_task; 1231 /* Deadlock detection and priority inheritance handling: */ 1232 struct rt_mutex_waiter *pi_blocked_on; 1233 #endif 1234 1235 struct mutex *blocked_on; /* lock we're blocked on */ 1236 1237 #ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER 1238 /* 1239 * Encoded lock address causing task block (lower 2 bits = type from 1240 * <linux/hung_task.h>). Accessed via hung_task_*() helpers. 1241 */ 1242 unsigned long blocker; 1243 #endif 1244 1245 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1246 int non_block_count; 1247 #endif 1248 1249 #ifdef CONFIG_TRACE_IRQFLAGS 1250 struct irqtrace_events irqtrace; 1251 unsigned int hardirq_threaded; 1252 u64 hardirq_chain_key; 1253 int softirqs_enabled; 1254 int softirq_context; 1255 int irq_config; 1256 #endif 1257 #ifdef CONFIG_PREEMPT_RT 1258 int softirq_disable_cnt; 1259 #endif 1260 1261 #ifdef CONFIG_LOCKDEP 1262 # define MAX_LOCK_DEPTH 48UL 1263 u64 curr_chain_key; 1264 int lockdep_depth; 1265 unsigned int lockdep_recursion; 1266 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1267 #endif 1268 1269 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1270 unsigned int in_ubsan; 1271 #endif 1272 1273 /* Journalling filesystem info: */ 1274 void *journal_info; 1275 1276 /* Stacked block device info: */ 1277 struct bio_list *bio_list; 1278 1279 /* Stack plugging: */ 1280 struct blk_plug *plug; 1281 1282 /* VM state: */ 1283 struct reclaim_state *reclaim_state; 1284 1285 struct io_context *io_context; 1286 1287 #ifdef CONFIG_COMPACTION 1288 struct capture_control *capture_control; 1289 #endif 1290 /* Ptrace state: */ 1291 unsigned long ptrace_message; 1292 kernel_siginfo_t *last_siginfo; 1293 1294 struct task_io_accounting ioac; 1295 #ifdef CONFIG_PSI 1296 /* Pressure stall state */ 1297 unsigned int psi_flags; 1298 #endif 1299 #ifdef CONFIG_TASK_XACCT 1300 /* Accumulated RSS usage: */ 1301 u64 acct_rss_mem1; 1302 /* Accumulated virtual memory usage: */ 1303 u64 acct_vm_mem1; 1304 /* stime + utime since last update: */ 1305 u64 acct_timexpd; 1306 #endif 1307 #ifdef CONFIG_CPUSETS 1308 /* Protected by ->alloc_lock: */ 1309 nodemask_t mems_allowed; 1310 /* Sequence number to catch updates: */ 1311 seqcount_spinlock_t mems_allowed_seq; 1312 int cpuset_mem_spread_rotor; 1313 #endif 1314 #ifdef CONFIG_CGROUPS 1315 /* Control Group info protected by css_set_lock: */ 1316 struct css_set __rcu *cgroups; 1317 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1318 struct list_head cg_list; 1319 #endif 1320 #ifdef CONFIG_X86_CPU_RESCTRL 1321 u32 closid; 1322 u32 rmid; 1323 #endif 1324 #ifdef CONFIG_FUTEX 1325 struct robust_list_head __user *robust_list; 1326 #ifdef CONFIG_COMPAT 1327 struct compat_robust_list_head __user *compat_robust_list; 1328 #endif 1329 struct list_head pi_state_list; 1330 struct futex_pi_state *pi_state_cache; 1331 struct mutex futex_exit_mutex; 1332 unsigned int futex_state; 1333 #endif 1334 #ifdef CONFIG_PERF_EVENTS 1335 u8 perf_recursion[PERF_NR_CONTEXTS]; 1336 struct perf_event_context *perf_event_ctxp; 1337 struct mutex perf_event_mutex; 1338 struct list_head perf_event_list; 1339 struct perf_ctx_data __rcu *perf_ctx_data; 1340 #endif 1341 #ifdef CONFIG_DEBUG_PREEMPT 1342 unsigned long preempt_disable_ip; 1343 #endif 1344 #ifdef CONFIG_NUMA 1345 /* Protected by alloc_lock: */ 1346 struct mempolicy *mempolicy; 1347 short il_prev; 1348 u8 il_weight; 1349 short pref_node_fork; 1350 #endif 1351 #ifdef CONFIG_NUMA_BALANCING 1352 int numa_scan_seq; 1353 unsigned int numa_scan_period; 1354 unsigned int numa_scan_period_max; 1355 int numa_preferred_nid; 1356 unsigned long numa_migrate_retry; 1357 /* Migration stamp: */ 1358 u64 node_stamp; 1359 u64 last_task_numa_placement; 1360 u64 last_sum_exec_runtime; 1361 struct callback_head numa_work; 1362 1363 /* 1364 * This pointer is only modified for current in syscall and 1365 * pagefault context (and for tasks being destroyed), so it can be read 1366 * from any of the following contexts: 1367 * - RCU read-side critical section 1368 * - current->numa_group from everywhere 1369 * - task's runqueue locked, task not running 1370 */ 1371 struct numa_group __rcu *numa_group; 1372 1373 /* 1374 * numa_faults is an array split into four regions: 1375 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1376 * in this precise order. 1377 * 1378 * faults_memory: Exponential decaying average of faults on a per-node 1379 * basis. Scheduling placement decisions are made based on these 1380 * counts. The values remain static for the duration of a PTE scan. 1381 * faults_cpu: Track the nodes the process was running on when a NUMA 1382 * hinting fault was incurred. 1383 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1384 * during the current scan window. When the scan completes, the counts 1385 * in faults_memory and faults_cpu decay and these values are copied. 1386 */ 1387 unsigned long *numa_faults; 1388 unsigned long total_numa_faults; 1389 1390 /* 1391 * numa_faults_locality tracks if faults recorded during the last 1392 * scan window were remote/local or failed to migrate. The task scan 1393 * period is adapted based on the locality of the faults with different 1394 * weights depending on whether they were shared or private faults 1395 */ 1396 unsigned long numa_faults_locality[3]; 1397 1398 unsigned long numa_pages_migrated; 1399 #endif /* CONFIG_NUMA_BALANCING */ 1400 1401 #ifdef CONFIG_RSEQ 1402 struct rseq __user *rseq; 1403 u32 rseq_len; 1404 u32 rseq_sig; 1405 /* 1406 * RmW on rseq_event_mask must be performed atomically 1407 * with respect to preemption. 1408 */ 1409 unsigned long rseq_event_mask; 1410 # ifdef CONFIG_DEBUG_RSEQ 1411 /* 1412 * This is a place holder to save a copy of the rseq fields for 1413 * validation of read-only fields. The struct rseq has a 1414 * variable-length array at the end, so it cannot be used 1415 * directly. Reserve a size large enough for the known fields. 1416 */ 1417 char rseq_fields[sizeof(struct rseq)]; 1418 # endif 1419 #endif 1420 1421 #ifdef CONFIG_SCHED_MM_CID 1422 int mm_cid; /* Current cid in mm */ 1423 int last_mm_cid; /* Most recent cid in mm */ 1424 int migrate_from_cpu; 1425 int mm_cid_active; /* Whether cid bitmap is active */ 1426 struct callback_head cid_work; 1427 #endif 1428 1429 struct tlbflush_unmap_batch tlb_ubc; 1430 1431 /* Cache last used pipe for splice(): */ 1432 struct pipe_inode_info *splice_pipe; 1433 1434 struct page_frag task_frag; 1435 1436 #ifdef CONFIG_TASK_DELAY_ACCT 1437 struct task_delay_info *delays; 1438 #endif 1439 1440 #ifdef CONFIG_FAULT_INJECTION 1441 int make_it_fail; 1442 unsigned int fail_nth; 1443 #endif 1444 /* 1445 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1446 * balance_dirty_pages() for a dirty throttling pause: 1447 */ 1448 int nr_dirtied; 1449 int nr_dirtied_pause; 1450 /* Start of a write-and-pause period: */ 1451 unsigned long dirty_paused_when; 1452 1453 #ifdef CONFIG_LATENCYTOP 1454 int latency_record_count; 1455 struct latency_record latency_record[LT_SAVECOUNT]; 1456 #endif 1457 /* 1458 * Time slack values; these are used to round up poll() and 1459 * select() etc timeout values. These are in nanoseconds. 1460 */ 1461 u64 timer_slack_ns; 1462 u64 default_timer_slack_ns; 1463 1464 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1465 unsigned int kasan_depth; 1466 #endif 1467 1468 #ifdef CONFIG_KCSAN 1469 struct kcsan_ctx kcsan_ctx; 1470 #ifdef CONFIG_TRACE_IRQFLAGS 1471 struct irqtrace_events kcsan_save_irqtrace; 1472 #endif 1473 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1474 int kcsan_stack_depth; 1475 #endif 1476 #endif 1477 1478 #ifdef CONFIG_KMSAN 1479 struct kmsan_ctx kmsan_ctx; 1480 #endif 1481 1482 #if IS_ENABLED(CONFIG_KUNIT) 1483 struct kunit *kunit_test; 1484 #endif 1485 1486 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1487 /* Index of current stored address in ret_stack: */ 1488 int curr_ret_stack; 1489 int curr_ret_depth; 1490 1491 /* Stack of return addresses for return function tracing: */ 1492 unsigned long *ret_stack; 1493 1494 /* Timestamp for last schedule: */ 1495 unsigned long long ftrace_timestamp; 1496 unsigned long long ftrace_sleeptime; 1497 1498 /* 1499 * Number of functions that haven't been traced 1500 * because of depth overrun: 1501 */ 1502 atomic_t trace_overrun; 1503 1504 /* Pause tracing: */ 1505 atomic_t tracing_graph_pause; 1506 #endif 1507 1508 #ifdef CONFIG_TRACING 1509 /* Bitmask and counter of trace recursion: */ 1510 unsigned long trace_recursion; 1511 #endif /* CONFIG_TRACING */ 1512 1513 #ifdef CONFIG_KCOV 1514 /* See kernel/kcov.c for more details. */ 1515 1516 /* Coverage collection mode enabled for this task (0 if disabled): */ 1517 unsigned int kcov_mode; 1518 1519 /* Size of the kcov_area: */ 1520 unsigned int kcov_size; 1521 1522 /* Buffer for coverage collection: */ 1523 void *kcov_area; 1524 1525 /* KCOV descriptor wired with this task or NULL: */ 1526 struct kcov *kcov; 1527 1528 /* KCOV common handle for remote coverage collection: */ 1529 u64 kcov_handle; 1530 1531 /* KCOV sequence number: */ 1532 int kcov_sequence; 1533 1534 /* Collect coverage from softirq context: */ 1535 unsigned int kcov_softirq; 1536 #endif 1537 1538 #ifdef CONFIG_MEMCG_V1 1539 struct mem_cgroup *memcg_in_oom; 1540 #endif 1541 1542 #ifdef CONFIG_MEMCG 1543 /* Number of pages to reclaim on returning to userland: */ 1544 unsigned int memcg_nr_pages_over_high; 1545 1546 /* Used by memcontrol for targeted memcg charge: */ 1547 struct mem_cgroup *active_memcg; 1548 1549 /* Cache for current->cgroups->memcg->objcg lookups: */ 1550 struct obj_cgroup *objcg; 1551 #endif 1552 1553 #ifdef CONFIG_BLK_CGROUP 1554 struct gendisk *throttle_disk; 1555 #endif 1556 1557 #ifdef CONFIG_UPROBES 1558 struct uprobe_task *utask; 1559 #endif 1560 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1561 unsigned int sequential_io; 1562 unsigned int sequential_io_avg; 1563 #endif 1564 struct kmap_ctrl kmap_ctrl; 1565 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1566 unsigned long task_state_change; 1567 # ifdef CONFIG_PREEMPT_RT 1568 unsigned long saved_state_change; 1569 # endif 1570 #endif 1571 struct rcu_head rcu; 1572 refcount_t rcu_users; 1573 int pagefault_disabled; 1574 #ifdef CONFIG_MMU 1575 struct task_struct *oom_reaper_list; 1576 struct timer_list oom_reaper_timer; 1577 #endif 1578 #ifdef CONFIG_VMAP_STACK 1579 struct vm_struct *stack_vm_area; 1580 #endif 1581 #ifdef CONFIG_THREAD_INFO_IN_TASK 1582 /* A live task holds one reference: */ 1583 refcount_t stack_refcount; 1584 #endif 1585 #ifdef CONFIG_LIVEPATCH 1586 int patch_state; 1587 #endif 1588 #ifdef CONFIG_SECURITY 1589 /* Used by LSM modules for access restriction: */ 1590 void *security; 1591 #endif 1592 #ifdef CONFIG_BPF_SYSCALL 1593 /* Used by BPF task local storage */ 1594 struct bpf_local_storage __rcu *bpf_storage; 1595 /* Used for BPF run context */ 1596 struct bpf_run_ctx *bpf_ctx; 1597 #endif 1598 /* Used by BPF for per-TASK xdp storage */ 1599 struct bpf_net_context *bpf_net_context; 1600 1601 #ifdef CONFIG_KSTACK_ERASE 1602 unsigned long lowest_stack; 1603 #endif 1604 #ifdef CONFIG_KSTACK_ERASE_METRICS 1605 unsigned long prev_lowest_stack; 1606 #endif 1607 1608 #ifdef CONFIG_X86_MCE 1609 void __user *mce_vaddr; 1610 __u64 mce_kflags; 1611 u64 mce_addr; 1612 __u64 mce_ripv : 1, 1613 mce_whole_page : 1, 1614 __mce_reserved : 62; 1615 struct callback_head mce_kill_me; 1616 int mce_count; 1617 #endif 1618 1619 #ifdef CONFIG_KRETPROBES 1620 struct llist_head kretprobe_instances; 1621 #endif 1622 #ifdef CONFIG_RETHOOK 1623 struct llist_head rethooks; 1624 #endif 1625 1626 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1627 /* 1628 * If L1D flush is supported on mm context switch 1629 * then we use this callback head to queue kill work 1630 * to kill tasks that are not running on SMT disabled 1631 * cores 1632 */ 1633 struct callback_head l1d_flush_kill; 1634 #endif 1635 1636 #ifdef CONFIG_RV 1637 /* 1638 * Per-task RV monitor, fixed in CONFIG_RV_PER_TASK_MONITORS. 1639 * If memory becomes a concern, we can think about a dynamic method. 1640 */ 1641 union rv_task_monitor rv[CONFIG_RV_PER_TASK_MONITORS]; 1642 #endif 1643 1644 #ifdef CONFIG_USER_EVENTS 1645 struct user_event_mm *user_event_mm; 1646 #endif 1647 1648 #ifdef CONFIG_UNWIND_USER 1649 struct unwind_task_info unwind_info; 1650 #endif 1651 1652 /* CPU-specific state of this task: */ 1653 struct thread_struct thread; 1654 1655 /* 1656 * New fields for task_struct should be added above here, so that 1657 * they are included in the randomized portion of task_struct. 1658 */ 1659 randomized_struct_fields_end 1660 } __attribute__ ((aligned (64))); 1661 1662 #ifdef CONFIG_SCHED_PROXY_EXEC 1663 DECLARE_STATIC_KEY_TRUE(__sched_proxy_exec); 1664 static inline bool sched_proxy_exec(void) 1665 { 1666 return static_branch_likely(&__sched_proxy_exec); 1667 } 1668 #else 1669 static inline bool sched_proxy_exec(void) 1670 { 1671 return false; 1672 } 1673 #endif 1674 1675 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1676 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1677 1678 static inline unsigned int __task_state_index(unsigned int tsk_state, 1679 unsigned int tsk_exit_state) 1680 { 1681 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1682 1683 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1684 1685 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1686 state = TASK_REPORT_IDLE; 1687 1688 /* 1689 * We're lying here, but rather than expose a completely new task state 1690 * to userspace, we can make this appear as if the task has gone through 1691 * a regular rt_mutex_lock() call. 1692 * Report frozen tasks as uninterruptible. 1693 */ 1694 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN)) 1695 state = TASK_UNINTERRUPTIBLE; 1696 1697 return fls(state); 1698 } 1699 1700 static inline unsigned int task_state_index(struct task_struct *tsk) 1701 { 1702 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1703 } 1704 1705 static inline char task_index_to_char(unsigned int state) 1706 { 1707 static const char state_char[] = "RSDTtXZPI"; 1708 1709 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1)); 1710 1711 return state_char[state]; 1712 } 1713 1714 static inline char task_state_to_char(struct task_struct *tsk) 1715 { 1716 return task_index_to_char(task_state_index(tsk)); 1717 } 1718 1719 extern struct pid *cad_pid; 1720 1721 /* 1722 * Per process flags 1723 */ 1724 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1725 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1726 #define PF_EXITING 0x00000004 /* Getting shut down */ 1727 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1728 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1729 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1730 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1731 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1732 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1733 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1734 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1735 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1736 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1737 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1738 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1739 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1740 #define PF_KCOMPACTD 0x00010000 /* I am kcompactd */ 1741 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1742 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1743 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1744 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1745 * I am cleaning dirty pages from some other bdi. */ 1746 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1747 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1748 #define PF__HOLE__00800000 0x00800000 1749 #define PF__HOLE__01000000 0x01000000 1750 #define PF__HOLE__02000000 0x02000000 1751 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1752 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1753 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1754 * See memalloc_pin_save() */ 1755 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1756 #define PF__HOLE__40000000 0x40000000 1757 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1758 1759 /* 1760 * Only the _current_ task can read/write to tsk->flags, but other 1761 * tasks can access tsk->flags in readonly mode for example 1762 * with tsk_used_math (like during threaded core dumping). 1763 * There is however an exception to this rule during ptrace 1764 * or during fork: the ptracer task is allowed to write to the 1765 * child->flags of its traced child (same goes for fork, the parent 1766 * can write to the child->flags), because we're guaranteed the 1767 * child is not running and in turn not changing child->flags 1768 * at the same time the parent does it. 1769 */ 1770 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1771 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1772 #define clear_used_math() clear_stopped_child_used_math(current) 1773 #define set_used_math() set_stopped_child_used_math(current) 1774 1775 #define conditional_stopped_child_used_math(condition, child) \ 1776 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1777 1778 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1779 1780 #define copy_to_stopped_child_used_math(child) \ 1781 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1782 1783 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1784 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1785 #define used_math() tsk_used_math(current) 1786 1787 static __always_inline bool is_percpu_thread(void) 1788 { 1789 return (current->flags & PF_NO_SETAFFINITY) && 1790 (current->nr_cpus_allowed == 1); 1791 } 1792 1793 /* Per-process atomic flags. */ 1794 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1795 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1796 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1797 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1798 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1799 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1800 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1801 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1802 1803 #define TASK_PFA_TEST(name, func) \ 1804 static inline bool task_##func(struct task_struct *p) \ 1805 { return test_bit(PFA_##name, &p->atomic_flags); } 1806 1807 #define TASK_PFA_SET(name, func) \ 1808 static inline void task_set_##func(struct task_struct *p) \ 1809 { set_bit(PFA_##name, &p->atomic_flags); } 1810 1811 #define TASK_PFA_CLEAR(name, func) \ 1812 static inline void task_clear_##func(struct task_struct *p) \ 1813 { clear_bit(PFA_##name, &p->atomic_flags); } 1814 1815 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1816 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1817 1818 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1819 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1820 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1821 1822 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1823 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1824 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1825 1826 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1827 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1828 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1829 1830 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1831 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1832 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1833 1834 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1835 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1836 1837 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1838 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1839 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1840 1841 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1842 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1843 1844 static inline void 1845 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1846 { 1847 current->flags &= ~flags; 1848 current->flags |= orig_flags & flags; 1849 } 1850 1851 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1852 extern int task_can_attach(struct task_struct *p); 1853 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1854 extern void dl_bw_free(int cpu, u64 dl_bw); 1855 1856 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1857 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1858 1859 /** 1860 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1861 * @p: the task 1862 * @new_mask: CPU affinity mask 1863 * 1864 * Return: zero if successful, or a negative error code 1865 */ 1866 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1867 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1868 extern void release_user_cpus_ptr(struct task_struct *p); 1869 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1870 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1871 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1872 1873 extern int yield_to(struct task_struct *p, bool preempt); 1874 extern void set_user_nice(struct task_struct *p, long nice); 1875 extern int task_prio(const struct task_struct *p); 1876 1877 /** 1878 * task_nice - return the nice value of a given task. 1879 * @p: the task in question. 1880 * 1881 * Return: The nice value [ -20 ... 0 ... 19 ]. 1882 */ 1883 static inline int task_nice(const struct task_struct *p) 1884 { 1885 return PRIO_TO_NICE((p)->static_prio); 1886 } 1887 1888 extern int can_nice(const struct task_struct *p, const int nice); 1889 extern int task_curr(const struct task_struct *p); 1890 extern int idle_cpu(int cpu); 1891 extern int available_idle_cpu(int cpu); 1892 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1893 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1894 extern void sched_set_fifo(struct task_struct *p); 1895 extern void sched_set_fifo_low(struct task_struct *p); 1896 extern void sched_set_normal(struct task_struct *p, int nice); 1897 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1898 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1899 extern struct task_struct *idle_task(int cpu); 1900 1901 /** 1902 * is_idle_task - is the specified task an idle task? 1903 * @p: the task in question. 1904 * 1905 * Return: 1 if @p is an idle task. 0 otherwise. 1906 */ 1907 static __always_inline bool is_idle_task(const struct task_struct *p) 1908 { 1909 return !!(p->flags & PF_IDLE); 1910 } 1911 1912 extern struct task_struct *curr_task(int cpu); 1913 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1914 1915 void yield(void); 1916 1917 union thread_union { 1918 struct task_struct task; 1919 #ifndef CONFIG_THREAD_INFO_IN_TASK 1920 struct thread_info thread_info; 1921 #endif 1922 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1923 }; 1924 1925 #ifndef CONFIG_THREAD_INFO_IN_TASK 1926 extern struct thread_info init_thread_info; 1927 #endif 1928 1929 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1930 1931 #ifdef CONFIG_THREAD_INFO_IN_TASK 1932 # define task_thread_info(task) (&(task)->thread_info) 1933 #else 1934 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1935 #endif 1936 1937 /* 1938 * find a task by one of its numerical ids 1939 * 1940 * find_task_by_pid_ns(): 1941 * finds a task by its pid in the specified namespace 1942 * find_task_by_vpid(): 1943 * finds a task by its virtual pid 1944 * 1945 * see also find_vpid() etc in include/linux/pid.h 1946 */ 1947 1948 extern struct task_struct *find_task_by_vpid(pid_t nr); 1949 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1950 1951 /* 1952 * find a task by its virtual pid and get the task struct 1953 */ 1954 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1955 1956 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1957 extern int wake_up_process(struct task_struct *tsk); 1958 extern void wake_up_new_task(struct task_struct *tsk); 1959 1960 extern void kick_process(struct task_struct *tsk); 1961 1962 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1963 #define set_task_comm(tsk, from) ({ \ 1964 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \ 1965 __set_task_comm(tsk, from, false); \ 1966 }) 1967 1968 /* 1969 * - Why not use task_lock()? 1970 * User space can randomly change their names anyway, so locking for readers 1971 * doesn't make sense. For writers, locking is probably necessary, as a race 1972 * condition could lead to long-term mixed results. 1973 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is 1974 * always NUL-terminated and zero-padded. Therefore the race condition between 1975 * reader and writer is not an issue. 1976 * 1977 * - BUILD_BUG_ON() can help prevent the buf from being truncated. 1978 * Since the callers don't perform any return value checks, this safeguard is 1979 * necessary. 1980 */ 1981 #define get_task_comm(buf, tsk) ({ \ 1982 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \ 1983 strscpy_pad(buf, (tsk)->comm); \ 1984 buf; \ 1985 }) 1986 1987 static __always_inline void scheduler_ipi(void) 1988 { 1989 /* 1990 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1991 * TIF_NEED_RESCHED remotely (for the first time) will also send 1992 * this IPI. 1993 */ 1994 preempt_fold_need_resched(); 1995 } 1996 1997 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 1998 1999 /* 2000 * Set thread flags in other task's structures. 2001 * See asm/thread_info.h for TIF_xxxx flags available: 2002 */ 2003 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 2004 { 2005 set_ti_thread_flag(task_thread_info(tsk), flag); 2006 } 2007 2008 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2009 { 2010 clear_ti_thread_flag(task_thread_info(tsk), flag); 2011 } 2012 2013 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 2014 bool value) 2015 { 2016 update_ti_thread_flag(task_thread_info(tsk), flag, value); 2017 } 2018 2019 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2020 { 2021 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2022 } 2023 2024 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2025 { 2026 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2027 } 2028 2029 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2030 { 2031 return test_ti_thread_flag(task_thread_info(tsk), flag); 2032 } 2033 2034 static inline void set_tsk_need_resched(struct task_struct *tsk) 2035 { 2036 if (tracepoint_enabled(sched_set_need_resched_tp) && 2037 !test_tsk_thread_flag(tsk, TIF_NEED_RESCHED)) 2038 __trace_set_need_resched(tsk, TIF_NEED_RESCHED); 2039 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2040 } 2041 2042 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2043 { 2044 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY, 2045 (atomic_long_t *)&task_thread_info(tsk)->flags); 2046 } 2047 2048 static inline int test_tsk_need_resched(struct task_struct *tsk) 2049 { 2050 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2051 } 2052 2053 /* 2054 * cond_resched() and cond_resched_lock(): latency reduction via 2055 * explicit rescheduling in places that are safe. The return 2056 * value indicates whether a reschedule was done in fact. 2057 * cond_resched_lock() will drop the spinlock before scheduling, 2058 */ 2059 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2060 extern int __cond_resched(void); 2061 2062 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2063 2064 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2065 2066 static __always_inline int _cond_resched(void) 2067 { 2068 return static_call_mod(cond_resched)(); 2069 } 2070 2071 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2072 2073 extern int dynamic_cond_resched(void); 2074 2075 static __always_inline int _cond_resched(void) 2076 { 2077 return dynamic_cond_resched(); 2078 } 2079 2080 #else /* !CONFIG_PREEMPTION */ 2081 2082 static inline int _cond_resched(void) 2083 { 2084 return __cond_resched(); 2085 } 2086 2087 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2088 2089 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2090 2091 static inline int _cond_resched(void) 2092 { 2093 return 0; 2094 } 2095 2096 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2097 2098 #define cond_resched() ({ \ 2099 __might_resched(__FILE__, __LINE__, 0); \ 2100 _cond_resched(); \ 2101 }) 2102 2103 extern int __cond_resched_lock(spinlock_t *lock); 2104 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2105 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2106 2107 #define MIGHT_RESCHED_RCU_SHIFT 8 2108 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2109 2110 #ifndef CONFIG_PREEMPT_RT 2111 /* 2112 * Non RT kernels have an elevated preempt count due to the held lock, 2113 * but are not allowed to be inside a RCU read side critical section 2114 */ 2115 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2116 #else 2117 /* 2118 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2119 * cond_resched*lock() has to take that into account because it checks for 2120 * preempt_count() and rcu_preempt_depth(). 2121 */ 2122 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2123 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2124 #endif 2125 2126 #define cond_resched_lock(lock) ({ \ 2127 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2128 __cond_resched_lock(lock); \ 2129 }) 2130 2131 #define cond_resched_rwlock_read(lock) ({ \ 2132 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2133 __cond_resched_rwlock_read(lock); \ 2134 }) 2135 2136 #define cond_resched_rwlock_write(lock) ({ \ 2137 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2138 __cond_resched_rwlock_write(lock); \ 2139 }) 2140 2141 #ifndef CONFIG_PREEMPT_RT 2142 static inline struct mutex *__get_task_blocked_on(struct task_struct *p) 2143 { 2144 struct mutex *m = p->blocked_on; 2145 2146 if (m) 2147 lockdep_assert_held_once(&m->wait_lock); 2148 return m; 2149 } 2150 2151 static inline void __set_task_blocked_on(struct task_struct *p, struct mutex *m) 2152 { 2153 struct mutex *blocked_on = READ_ONCE(p->blocked_on); 2154 2155 WARN_ON_ONCE(!m); 2156 /* The task should only be setting itself as blocked */ 2157 WARN_ON_ONCE(p != current); 2158 /* Currently we serialize blocked_on under the mutex::wait_lock */ 2159 lockdep_assert_held_once(&m->wait_lock); 2160 /* 2161 * Check ensure we don't overwrite existing mutex value 2162 * with a different mutex. Note, setting it to the same 2163 * lock repeatedly is ok. 2164 */ 2165 WARN_ON_ONCE(blocked_on && blocked_on != m); 2166 WRITE_ONCE(p->blocked_on, m); 2167 } 2168 2169 static inline void set_task_blocked_on(struct task_struct *p, struct mutex *m) 2170 { 2171 guard(raw_spinlock_irqsave)(&m->wait_lock); 2172 __set_task_blocked_on(p, m); 2173 } 2174 2175 static inline void __clear_task_blocked_on(struct task_struct *p, struct mutex *m) 2176 { 2177 if (m) { 2178 struct mutex *blocked_on = READ_ONCE(p->blocked_on); 2179 2180 /* Currently we serialize blocked_on under the mutex::wait_lock */ 2181 lockdep_assert_held_once(&m->wait_lock); 2182 /* 2183 * There may be cases where we re-clear already cleared 2184 * blocked_on relationships, but make sure we are not 2185 * clearing the relationship with a different lock. 2186 */ 2187 WARN_ON_ONCE(blocked_on && blocked_on != m); 2188 } 2189 WRITE_ONCE(p->blocked_on, NULL); 2190 } 2191 2192 static inline void clear_task_blocked_on(struct task_struct *p, struct mutex *m) 2193 { 2194 guard(raw_spinlock_irqsave)(&m->wait_lock); 2195 __clear_task_blocked_on(p, m); 2196 } 2197 #else 2198 static inline void __clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m) 2199 { 2200 } 2201 2202 static inline void clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m) 2203 { 2204 } 2205 #endif /* !CONFIG_PREEMPT_RT */ 2206 2207 static __always_inline bool need_resched(void) 2208 { 2209 return unlikely(tif_need_resched()); 2210 } 2211 2212 /* 2213 * Wrappers for p->thread_info->cpu access. No-op on UP. 2214 */ 2215 #ifdef CONFIG_SMP 2216 2217 static inline unsigned int task_cpu(const struct task_struct *p) 2218 { 2219 return READ_ONCE(task_thread_info(p)->cpu); 2220 } 2221 2222 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2223 2224 #else 2225 2226 static inline unsigned int task_cpu(const struct task_struct *p) 2227 { 2228 return 0; 2229 } 2230 2231 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2232 { 2233 } 2234 2235 #endif /* CONFIG_SMP */ 2236 2237 static inline bool task_is_runnable(struct task_struct *p) 2238 { 2239 return p->on_rq && !p->se.sched_delayed; 2240 } 2241 2242 extern bool sched_task_on_rq(struct task_struct *p); 2243 extern unsigned long get_wchan(struct task_struct *p); 2244 extern struct task_struct *cpu_curr_snapshot(int cpu); 2245 2246 /* 2247 * In order to reduce various lock holder preemption latencies provide an 2248 * interface to see if a vCPU is currently running or not. 2249 * 2250 * This allows us to terminate optimistic spin loops and block, analogous to 2251 * the native optimistic spin heuristic of testing if the lock owner task is 2252 * running or not. 2253 */ 2254 #ifndef vcpu_is_preempted 2255 static inline bool vcpu_is_preempted(int cpu) 2256 { 2257 return false; 2258 } 2259 #endif 2260 2261 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2262 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2263 2264 #ifndef TASK_SIZE_OF 2265 #define TASK_SIZE_OF(tsk) TASK_SIZE 2266 #endif 2267 2268 static inline bool owner_on_cpu(struct task_struct *owner) 2269 { 2270 /* 2271 * As lock holder preemption issue, we both skip spinning if 2272 * task is not on cpu or its cpu is preempted 2273 */ 2274 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2275 } 2276 2277 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2278 unsigned long sched_cpu_util(int cpu); 2279 2280 #ifdef CONFIG_SCHED_CORE 2281 extern void sched_core_free(struct task_struct *tsk); 2282 extern void sched_core_fork(struct task_struct *p); 2283 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2284 unsigned long uaddr); 2285 extern int sched_core_idle_cpu(int cpu); 2286 #else 2287 static inline void sched_core_free(struct task_struct *tsk) { } 2288 static inline void sched_core_fork(struct task_struct *p) { } 2289 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2290 #endif 2291 2292 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2293 2294 #ifdef CONFIG_MEM_ALLOC_PROFILING 2295 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2296 { 2297 swap(current->alloc_tag, tag); 2298 return tag; 2299 } 2300 2301 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2302 { 2303 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2304 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2305 #endif 2306 current->alloc_tag = old; 2307 } 2308 #else 2309 #define alloc_tag_save(_tag) NULL 2310 #define alloc_tag_restore(_tag, _old) do {} while (0) 2311 #endif 2312 2313 #endif 2314