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/syscall_user_dispatch_types.h> 38 #include <linux/mm_types_task.h> 39 #include <linux/netdevice_xmit.h> 40 #include <linux/task_io_accounting.h> 41 #include <linux/posix-timers_types.h> 42 #include <linux/restart_block.h> 43 #include <uapi/linux/rseq.h> 44 #include <linux/seqlock_types.h> 45 #include <linux/kcsan.h> 46 #include <linux/rv.h> 47 #include <linux/livepatch_sched.h> 48 #include <linux/uidgid_types.h> 49 #include <asm/kmap_size.h> 50 51 /* task_struct member predeclarations (sorted alphabetically): */ 52 struct audit_context; 53 struct bio_list; 54 struct blk_plug; 55 struct bpf_local_storage; 56 struct bpf_run_ctx; 57 struct bpf_net_context; 58 struct capture_control; 59 struct cfs_rq; 60 struct fs_struct; 61 struct futex_pi_state; 62 struct io_context; 63 struct io_uring_task; 64 struct mempolicy; 65 struct nameidata; 66 struct nsproxy; 67 struct perf_event_context; 68 struct perf_ctx_data; 69 struct pid_namespace; 70 struct pipe_inode_info; 71 struct rcu_node; 72 struct reclaim_state; 73 struct robust_list_head; 74 struct root_domain; 75 struct rq; 76 struct sched_attr; 77 struct sched_dl_entity; 78 struct seq_file; 79 struct sighand_struct; 80 struct signal_struct; 81 struct task_delay_info; 82 struct task_group; 83 struct task_struct; 84 struct user_event_mm; 85 86 #include <linux/sched/ext.h> 87 88 /* 89 * Task state bitmask. NOTE! These bits are also 90 * encoded in fs/proc/array.c: get_task_state(). 91 * 92 * We have two separate sets of flags: task->__state 93 * is about runnability, while task->exit_state are 94 * about the task exiting. Confusing, but this way 95 * modifying one set can't modify the other one by 96 * mistake. 97 */ 98 99 /* Used in tsk->__state: */ 100 #define TASK_RUNNING 0x00000000 101 #define TASK_INTERRUPTIBLE 0x00000001 102 #define TASK_UNINTERRUPTIBLE 0x00000002 103 #define __TASK_STOPPED 0x00000004 104 #define __TASK_TRACED 0x00000008 105 /* Used in tsk->exit_state: */ 106 #define EXIT_DEAD 0x00000010 107 #define EXIT_ZOMBIE 0x00000020 108 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 109 /* Used in tsk->__state again: */ 110 #define TASK_PARKED 0x00000040 111 #define TASK_DEAD 0x00000080 112 #define TASK_WAKEKILL 0x00000100 113 #define TASK_WAKING 0x00000200 114 #define TASK_NOLOAD 0x00000400 115 #define TASK_NEW 0x00000800 116 #define TASK_RTLOCK_WAIT 0x00001000 117 #define TASK_FREEZABLE 0x00002000 118 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP)) 119 #define TASK_FROZEN 0x00008000 120 #define TASK_STATE_MAX 0x00010000 121 122 #define TASK_ANY (TASK_STATE_MAX-1) 123 124 /* 125 * DO NOT ADD ANY NEW USERS ! 126 */ 127 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE) 128 129 /* Convenience macros for the sake of set_current_state: */ 130 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 131 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 132 #define TASK_TRACED __TASK_TRACED 133 134 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 135 136 /* Convenience macros for the sake of wake_up(): */ 137 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 138 139 /* get_task_state(): */ 140 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 141 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 142 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 143 TASK_PARKED) 144 145 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 146 147 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 148 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 149 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 150 151 /* 152 * Special states are those that do not use the normal wait-loop pattern. See 153 * the comment with set_special_state(). 154 */ 155 #define is_special_task_state(state) \ 156 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \ 157 TASK_DEAD | TASK_FROZEN)) 158 159 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 160 # define debug_normal_state_change(state_value) \ 161 do { \ 162 WARN_ON_ONCE(is_special_task_state(state_value)); \ 163 current->task_state_change = _THIS_IP_; \ 164 } while (0) 165 166 # define debug_special_state_change(state_value) \ 167 do { \ 168 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 169 current->task_state_change = _THIS_IP_; \ 170 } while (0) 171 172 # define debug_rtlock_wait_set_state() \ 173 do { \ 174 current->saved_state_change = current->task_state_change;\ 175 current->task_state_change = _THIS_IP_; \ 176 } while (0) 177 178 # define debug_rtlock_wait_restore_state() \ 179 do { \ 180 current->task_state_change = current->saved_state_change;\ 181 } while (0) 182 183 #else 184 # define debug_normal_state_change(cond) do { } while (0) 185 # define debug_special_state_change(cond) do { } while (0) 186 # define debug_rtlock_wait_set_state() do { } while (0) 187 # define debug_rtlock_wait_restore_state() do { } while (0) 188 #endif 189 190 /* 191 * set_current_state() includes a barrier so that the write of current->__state 192 * is correctly serialised wrt the caller's subsequent test of whether to 193 * actually sleep: 194 * 195 * for (;;) { 196 * set_current_state(TASK_UNINTERRUPTIBLE); 197 * if (CONDITION) 198 * break; 199 * 200 * schedule(); 201 * } 202 * __set_current_state(TASK_RUNNING); 203 * 204 * If the caller does not need such serialisation (because, for instance, the 205 * CONDITION test and condition change and wakeup are under the same lock) then 206 * use __set_current_state(). 207 * 208 * The above is typically ordered against the wakeup, which does: 209 * 210 * CONDITION = 1; 211 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 212 * 213 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 214 * accessing p->__state. 215 * 216 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is, 217 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 218 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 219 * 220 * However, with slightly different timing the wakeup TASK_RUNNING store can 221 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 222 * a problem either because that will result in one extra go around the loop 223 * and our @cond test will save the day. 224 * 225 * Also see the comments of try_to_wake_up(). 226 */ 227 #define __set_current_state(state_value) \ 228 do { \ 229 debug_normal_state_change((state_value)); \ 230 WRITE_ONCE(current->__state, (state_value)); \ 231 } while (0) 232 233 #define set_current_state(state_value) \ 234 do { \ 235 debug_normal_state_change((state_value)); \ 236 smp_store_mb(current->__state, (state_value)); \ 237 } while (0) 238 239 /* 240 * set_special_state() should be used for those states when the blocking task 241 * can not use the regular condition based wait-loop. In that case we must 242 * serialize against wakeups such that any possible in-flight TASK_RUNNING 243 * stores will not collide with our state change. 244 */ 245 #define set_special_state(state_value) \ 246 do { \ 247 unsigned long flags; /* may shadow */ \ 248 \ 249 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 250 debug_special_state_change((state_value)); \ 251 WRITE_ONCE(current->__state, (state_value)); \ 252 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 253 } while (0) 254 255 /* 256 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 257 * 258 * RT's spin/rwlock substitutions are state preserving. The state of the 259 * task when blocking on the lock is saved in task_struct::saved_state and 260 * restored after the lock has been acquired. These operations are 261 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 262 * lock related wakeups while the task is blocked on the lock are 263 * redirected to operate on task_struct::saved_state to ensure that these 264 * are not dropped. On restore task_struct::saved_state is set to 265 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 266 * 267 * The lock operation looks like this: 268 * 269 * current_save_and_set_rtlock_wait_state(); 270 * for (;;) { 271 * if (try_lock()) 272 * break; 273 * raw_spin_unlock_irq(&lock->wait_lock); 274 * schedule_rtlock(); 275 * raw_spin_lock_irq(&lock->wait_lock); 276 * set_current_state(TASK_RTLOCK_WAIT); 277 * } 278 * current_restore_rtlock_saved_state(); 279 */ 280 #define current_save_and_set_rtlock_wait_state() \ 281 do { \ 282 lockdep_assert_irqs_disabled(); \ 283 raw_spin_lock(¤t->pi_lock); \ 284 current->saved_state = current->__state; \ 285 debug_rtlock_wait_set_state(); \ 286 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 287 raw_spin_unlock(¤t->pi_lock); \ 288 } while (0); 289 290 #define current_restore_rtlock_saved_state() \ 291 do { \ 292 lockdep_assert_irqs_disabled(); \ 293 raw_spin_lock(¤t->pi_lock); \ 294 debug_rtlock_wait_restore_state(); \ 295 WRITE_ONCE(current->__state, current->saved_state); \ 296 current->saved_state = TASK_RUNNING; \ 297 raw_spin_unlock(¤t->pi_lock); \ 298 } while (0); 299 300 #define get_current_state() READ_ONCE(current->__state) 301 302 /* 303 * Define the task command name length as enum, then it can be visible to 304 * BPF programs. 305 */ 306 enum { 307 TASK_COMM_LEN = 16, 308 }; 309 310 extern void sched_tick(void); 311 312 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 313 314 extern long schedule_timeout(long timeout); 315 extern long schedule_timeout_interruptible(long timeout); 316 extern long schedule_timeout_killable(long timeout); 317 extern long schedule_timeout_uninterruptible(long timeout); 318 extern long schedule_timeout_idle(long timeout); 319 asmlinkage void schedule(void); 320 extern void schedule_preempt_disabled(void); 321 asmlinkage void preempt_schedule_irq(void); 322 #ifdef CONFIG_PREEMPT_RT 323 extern void schedule_rtlock(void); 324 #endif 325 326 extern int __must_check io_schedule_prepare(void); 327 extern void io_schedule_finish(int token); 328 extern long io_schedule_timeout(long timeout); 329 extern void io_schedule(void); 330 331 /** 332 * struct prev_cputime - snapshot of system and user cputime 333 * @utime: time spent in user mode 334 * @stime: time spent in system mode 335 * @lock: protects the above two fields 336 * 337 * Stores previous user/system time values such that we can guarantee 338 * monotonicity. 339 */ 340 struct prev_cputime { 341 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 342 u64 utime; 343 u64 stime; 344 raw_spinlock_t lock; 345 #endif 346 }; 347 348 enum vtime_state { 349 /* Task is sleeping or running in a CPU with VTIME inactive: */ 350 VTIME_INACTIVE = 0, 351 /* Task is idle */ 352 VTIME_IDLE, 353 /* Task runs in kernelspace in a CPU with VTIME active: */ 354 VTIME_SYS, 355 /* Task runs in userspace in a CPU with VTIME active: */ 356 VTIME_USER, 357 /* Task runs as guests in a CPU with VTIME active: */ 358 VTIME_GUEST, 359 }; 360 361 struct vtime { 362 seqcount_t seqcount; 363 unsigned long long starttime; 364 enum vtime_state state; 365 unsigned int cpu; 366 u64 utime; 367 u64 stime; 368 u64 gtime; 369 }; 370 371 /* 372 * Utilization clamp constraints. 373 * @UCLAMP_MIN: Minimum utilization 374 * @UCLAMP_MAX: Maximum utilization 375 * @UCLAMP_CNT: Utilization clamp constraints count 376 */ 377 enum uclamp_id { 378 UCLAMP_MIN = 0, 379 UCLAMP_MAX, 380 UCLAMP_CNT 381 }; 382 383 #ifdef CONFIG_SMP 384 extern struct root_domain def_root_domain; 385 extern struct mutex sched_domains_mutex; 386 extern void sched_domains_mutex_lock(void); 387 extern void sched_domains_mutex_unlock(void); 388 #else 389 static inline void sched_domains_mutex_lock(void) { } 390 static inline void sched_domains_mutex_unlock(void) { } 391 #endif 392 393 struct sched_param { 394 int sched_priority; 395 }; 396 397 struct sched_info { 398 #ifdef CONFIG_SCHED_INFO 399 /* Cumulative counters: */ 400 401 /* # of times we have run on this CPU: */ 402 unsigned long pcount; 403 404 /* Time spent waiting on a runqueue: */ 405 unsigned long long run_delay; 406 407 /* Max time spent waiting on a runqueue: */ 408 unsigned long long max_run_delay; 409 410 /* Min time spent waiting on a runqueue: */ 411 unsigned long long min_run_delay; 412 413 /* Timestamps: */ 414 415 /* When did we last run on a CPU? */ 416 unsigned long long last_arrival; 417 418 /* When were we last queued to run? */ 419 unsigned long long last_queued; 420 421 #endif /* CONFIG_SCHED_INFO */ 422 }; 423 424 /* 425 * Integer metrics need fixed point arithmetic, e.g., sched/fair 426 * has a few: load, load_avg, util_avg, freq, and capacity. 427 * 428 * We define a basic fixed point arithmetic range, and then formalize 429 * all these metrics based on that basic range. 430 */ 431 # define SCHED_FIXEDPOINT_SHIFT 10 432 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 433 434 /* Increase resolution of cpu_capacity calculations */ 435 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 436 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 437 438 struct load_weight { 439 unsigned long weight; 440 u32 inv_weight; 441 }; 442 443 /* 444 * The load/runnable/util_avg accumulates an infinite geometric series 445 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 446 * 447 * [load_avg definition] 448 * 449 * load_avg = runnable% * scale_load_down(load) 450 * 451 * [runnable_avg definition] 452 * 453 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 454 * 455 * [util_avg definition] 456 * 457 * util_avg = running% * SCHED_CAPACITY_SCALE 458 * 459 * where runnable% is the time ratio that a sched_entity is runnable and 460 * running% the time ratio that a sched_entity is running. 461 * 462 * For cfs_rq, they are the aggregated values of all runnable and blocked 463 * sched_entities. 464 * 465 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 466 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 467 * for computing those signals (see update_rq_clock_pelt()) 468 * 469 * N.B., the above ratios (runnable% and running%) themselves are in the 470 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 471 * to as large a range as necessary. This is for example reflected by 472 * util_avg's SCHED_CAPACITY_SCALE. 473 * 474 * [Overflow issue] 475 * 476 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 477 * with the highest load (=88761), always runnable on a single cfs_rq, 478 * and should not overflow as the number already hits PID_MAX_LIMIT. 479 * 480 * For all other cases (including 32-bit kernels), struct load_weight's 481 * weight will overflow first before we do, because: 482 * 483 * Max(load_avg) <= Max(load.weight) 484 * 485 * Then it is the load_weight's responsibility to consider overflow 486 * issues. 487 */ 488 struct sched_avg { 489 u64 last_update_time; 490 u64 load_sum; 491 u64 runnable_sum; 492 u32 util_sum; 493 u32 period_contrib; 494 unsigned long load_avg; 495 unsigned long runnable_avg; 496 unsigned long util_avg; 497 unsigned int util_est; 498 } ____cacheline_aligned; 499 500 /* 501 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 502 * updates. When a task is dequeued, its util_est should not be updated if its 503 * util_avg has not been updated in the meantime. 504 * This information is mapped into the MSB bit of util_est at dequeue time. 505 * Since max value of util_est for a task is 1024 (PELT util_avg for a task) 506 * it is safe to use MSB. 507 */ 508 #define UTIL_EST_WEIGHT_SHIFT 2 509 #define UTIL_AVG_UNCHANGED 0x80000000 510 511 struct sched_statistics { 512 #ifdef CONFIG_SCHEDSTATS 513 u64 wait_start; 514 u64 wait_max; 515 u64 wait_count; 516 u64 wait_sum; 517 u64 iowait_count; 518 u64 iowait_sum; 519 520 u64 sleep_start; 521 u64 sleep_max; 522 s64 sum_sleep_runtime; 523 524 u64 block_start; 525 u64 block_max; 526 s64 sum_block_runtime; 527 528 s64 exec_max; 529 u64 slice_max; 530 531 u64 nr_migrations_cold; 532 u64 nr_failed_migrations_affine; 533 u64 nr_failed_migrations_running; 534 u64 nr_failed_migrations_hot; 535 u64 nr_forced_migrations; 536 537 u64 nr_wakeups; 538 u64 nr_wakeups_sync; 539 u64 nr_wakeups_migrate; 540 u64 nr_wakeups_local; 541 u64 nr_wakeups_remote; 542 u64 nr_wakeups_affine; 543 u64 nr_wakeups_affine_attempts; 544 u64 nr_wakeups_passive; 545 u64 nr_wakeups_idle; 546 547 #ifdef CONFIG_SCHED_CORE 548 u64 core_forceidle_sum; 549 #endif 550 #endif /* CONFIG_SCHEDSTATS */ 551 } ____cacheline_aligned; 552 553 struct sched_entity { 554 /* For load-balancing: */ 555 struct load_weight load; 556 struct rb_node run_node; 557 u64 deadline; 558 u64 min_vruntime; 559 u64 min_slice; 560 561 struct list_head group_node; 562 unsigned char on_rq; 563 unsigned char sched_delayed; 564 unsigned char rel_deadline; 565 unsigned char custom_slice; 566 /* hole */ 567 568 u64 exec_start; 569 u64 sum_exec_runtime; 570 u64 prev_sum_exec_runtime; 571 u64 vruntime; 572 s64 vlag; 573 u64 slice; 574 575 u64 nr_migrations; 576 577 #ifdef CONFIG_FAIR_GROUP_SCHED 578 int depth; 579 struct sched_entity *parent; 580 /* rq on which this entity is (to be) queued: */ 581 struct cfs_rq *cfs_rq; 582 /* rq "owned" by this entity/group: */ 583 struct cfs_rq *my_q; 584 /* cached value of my_q->h_nr_running */ 585 unsigned long runnable_weight; 586 #endif 587 588 #ifdef CONFIG_SMP 589 /* 590 * Per entity load average tracking. 591 * 592 * Put into separate cache line so it does not 593 * collide with read-mostly values above. 594 */ 595 struct sched_avg avg; 596 #endif 597 }; 598 599 struct sched_rt_entity { 600 struct list_head run_list; 601 unsigned long timeout; 602 unsigned long watchdog_stamp; 603 unsigned int time_slice; 604 unsigned short on_rq; 605 unsigned short on_list; 606 607 struct sched_rt_entity *back; 608 #ifdef CONFIG_RT_GROUP_SCHED 609 struct sched_rt_entity *parent; 610 /* rq on which this entity is (to be) queued: */ 611 struct rt_rq *rt_rq; 612 /* rq "owned" by this entity/group: */ 613 struct rt_rq *my_q; 614 #endif 615 } __randomize_layout; 616 617 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *); 618 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *); 619 620 struct sched_dl_entity { 621 struct rb_node rb_node; 622 623 /* 624 * Original scheduling parameters. Copied here from sched_attr 625 * during sched_setattr(), they will remain the same until 626 * the next sched_setattr(). 627 */ 628 u64 dl_runtime; /* Maximum runtime for each instance */ 629 u64 dl_deadline; /* Relative deadline of each instance */ 630 u64 dl_period; /* Separation of two instances (period) */ 631 u64 dl_bw; /* dl_runtime / dl_period */ 632 u64 dl_density; /* dl_runtime / dl_deadline */ 633 634 /* 635 * Actual scheduling parameters. Initialized with the values above, 636 * they are continuously updated during task execution. Note that 637 * the remaining runtime could be < 0 in case we are in overrun. 638 */ 639 s64 runtime; /* Remaining runtime for this instance */ 640 u64 deadline; /* Absolute deadline for this instance */ 641 unsigned int flags; /* Specifying the scheduler behaviour */ 642 643 /* 644 * Some bool flags: 645 * 646 * @dl_throttled tells if we exhausted the runtime. If so, the 647 * task has to wait for a replenishment to be performed at the 648 * next firing of dl_timer. 649 * 650 * @dl_yielded tells if task gave up the CPU before consuming 651 * all its available runtime during the last job. 652 * 653 * @dl_non_contending tells if the task is inactive while still 654 * contributing to the active utilization. In other words, it 655 * indicates if the inactive timer has been armed and its handler 656 * has not been executed yet. This flag is useful to avoid race 657 * conditions between the inactive timer handler and the wakeup 658 * code. 659 * 660 * @dl_overrun tells if the task asked to be informed about runtime 661 * overruns. 662 * 663 * @dl_server tells if this is a server entity. 664 * 665 * @dl_defer tells if this is a deferred or regular server. For 666 * now only defer server exists. 667 * 668 * @dl_defer_armed tells if the deferrable server is waiting 669 * for the replenishment timer to activate it. 670 * 671 * @dl_server_active tells if the dlserver is active(started). 672 * dlserver is started on first cfs enqueue on an idle runqueue 673 * and is stopped when a dequeue results in 0 cfs tasks on the 674 * runqueue. In other words, dlserver is active only when cpu's 675 * runqueue has atleast one cfs task. 676 * 677 * @dl_defer_running tells if the deferrable server is actually 678 * running, skipping the defer phase. 679 */ 680 unsigned int dl_throttled : 1; 681 unsigned int dl_yielded : 1; 682 unsigned int dl_non_contending : 1; 683 unsigned int dl_overrun : 1; 684 unsigned int dl_server : 1; 685 unsigned int dl_server_active : 1; 686 unsigned int dl_defer : 1; 687 unsigned int dl_defer_armed : 1; 688 unsigned int dl_defer_running : 1; 689 690 /* 691 * Bandwidth enforcement timer. Each -deadline task has its 692 * own bandwidth to be enforced, thus we need one timer per task. 693 */ 694 struct hrtimer dl_timer; 695 696 /* 697 * Inactive timer, responsible for decreasing the active utilization 698 * at the "0-lag time". When a -deadline task blocks, it contributes 699 * to GRUB's active utilization until the "0-lag time", hence a 700 * timer is needed to decrease the active utilization at the correct 701 * time. 702 */ 703 struct hrtimer inactive_timer; 704 705 /* 706 * Bits for DL-server functionality. Also see the comment near 707 * dl_server_update(). 708 * 709 * @rq the runqueue this server is for 710 * 711 * @server_has_tasks() returns true if @server_pick return a 712 * runnable task. 713 */ 714 struct rq *rq; 715 dl_server_has_tasks_f server_has_tasks; 716 dl_server_pick_f server_pick_task; 717 718 #ifdef CONFIG_RT_MUTEXES 719 /* 720 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 721 * pi_se points to the donor, otherwise points to the dl_se it belongs 722 * to (the original one/itself). 723 */ 724 struct sched_dl_entity *pi_se; 725 #endif 726 }; 727 728 #ifdef CONFIG_UCLAMP_TASK 729 /* Number of utilization clamp buckets (shorter alias) */ 730 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 731 732 /* 733 * Utilization clamp for a scheduling entity 734 * @value: clamp value "assigned" to a se 735 * @bucket_id: bucket index corresponding to the "assigned" value 736 * @active: the se is currently refcounted in a rq's bucket 737 * @user_defined: the requested clamp value comes from user-space 738 * 739 * The bucket_id is the index of the clamp bucket matching the clamp value 740 * which is pre-computed and stored to avoid expensive integer divisions from 741 * the fast path. 742 * 743 * The active bit is set whenever a task has got an "effective" value assigned, 744 * which can be different from the clamp value "requested" from user-space. 745 * This allows to know a task is refcounted in the rq's bucket corresponding 746 * to the "effective" bucket_id. 747 * 748 * The user_defined bit is set whenever a task has got a task-specific clamp 749 * value requested from userspace, i.e. the system defaults apply to this task 750 * just as a restriction. This allows to relax default clamps when a less 751 * restrictive task-specific value has been requested, thus allowing to 752 * implement a "nice" semantic. For example, a task running with a 20% 753 * default boost can still drop its own boosting to 0%. 754 */ 755 struct uclamp_se { 756 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 757 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 758 unsigned int active : 1; 759 unsigned int user_defined : 1; 760 }; 761 #endif /* CONFIG_UCLAMP_TASK */ 762 763 union rcu_special { 764 struct { 765 u8 blocked; 766 u8 need_qs; 767 u8 exp_hint; /* Hint for performance. */ 768 u8 need_mb; /* Readers need smp_mb(). */ 769 } b; /* Bits. */ 770 u32 s; /* Set of bits. */ 771 }; 772 773 enum perf_event_task_context { 774 perf_invalid_context = -1, 775 perf_hw_context = 0, 776 perf_sw_context, 777 perf_nr_task_contexts, 778 }; 779 780 /* 781 * Number of contexts where an event can trigger: 782 * task, softirq, hardirq, nmi. 783 */ 784 #define PERF_NR_CONTEXTS 4 785 786 struct wake_q_node { 787 struct wake_q_node *next; 788 }; 789 790 struct kmap_ctrl { 791 #ifdef CONFIG_KMAP_LOCAL 792 int idx; 793 pte_t pteval[KM_MAX_IDX]; 794 #endif 795 }; 796 797 struct task_struct { 798 #ifdef CONFIG_THREAD_INFO_IN_TASK 799 /* 800 * For reasons of header soup (see current_thread_info()), this 801 * must be the first element of task_struct. 802 */ 803 struct thread_info thread_info; 804 #endif 805 unsigned int __state; 806 807 /* saved state for "spinlock sleepers" */ 808 unsigned int saved_state; 809 810 /* 811 * This begins the randomizable portion of task_struct. Only 812 * scheduling-critical items should be added above here. 813 */ 814 randomized_struct_fields_start 815 816 void *stack; 817 refcount_t usage; 818 /* Per task flags (PF_*), defined further below: */ 819 unsigned int flags; 820 unsigned int ptrace; 821 822 #ifdef CONFIG_MEM_ALLOC_PROFILING 823 struct alloc_tag *alloc_tag; 824 #endif 825 826 #ifdef CONFIG_SMP 827 int on_cpu; 828 struct __call_single_node wake_entry; 829 unsigned int wakee_flips; 830 unsigned long wakee_flip_decay_ts; 831 struct task_struct *last_wakee; 832 833 /* 834 * recent_used_cpu is initially set as the last CPU used by a task 835 * that wakes affine another task. Waker/wakee relationships can 836 * push tasks around a CPU where each wakeup moves to the next one. 837 * Tracking a recently used CPU allows a quick search for a recently 838 * used CPU that may be idle. 839 */ 840 int recent_used_cpu; 841 int wake_cpu; 842 #endif 843 int on_rq; 844 845 int prio; 846 int static_prio; 847 int normal_prio; 848 unsigned int rt_priority; 849 850 struct sched_entity se; 851 struct sched_rt_entity rt; 852 struct sched_dl_entity dl; 853 struct sched_dl_entity *dl_server; 854 #ifdef CONFIG_SCHED_CLASS_EXT 855 struct sched_ext_entity scx; 856 #endif 857 const struct sched_class *sched_class; 858 859 #ifdef CONFIG_SCHED_CORE 860 struct rb_node core_node; 861 unsigned long core_cookie; 862 unsigned int core_occupation; 863 #endif 864 865 #ifdef CONFIG_CGROUP_SCHED 866 struct task_group *sched_task_group; 867 #endif 868 869 870 #ifdef CONFIG_UCLAMP_TASK 871 /* 872 * Clamp values requested for a scheduling entity. 873 * Must be updated with task_rq_lock() held. 874 */ 875 struct uclamp_se uclamp_req[UCLAMP_CNT]; 876 /* 877 * Effective clamp values used for a scheduling entity. 878 * Must be updated with task_rq_lock() held. 879 */ 880 struct uclamp_se uclamp[UCLAMP_CNT]; 881 #endif 882 883 struct sched_statistics stats; 884 885 #ifdef CONFIG_PREEMPT_NOTIFIERS 886 /* List of struct preempt_notifier: */ 887 struct hlist_head preempt_notifiers; 888 #endif 889 890 #ifdef CONFIG_BLK_DEV_IO_TRACE 891 unsigned int btrace_seq; 892 #endif 893 894 unsigned int policy; 895 unsigned long max_allowed_capacity; 896 int nr_cpus_allowed; 897 const cpumask_t *cpus_ptr; 898 cpumask_t *user_cpus_ptr; 899 cpumask_t cpus_mask; 900 void *migration_pending; 901 #ifdef CONFIG_SMP 902 unsigned short migration_disabled; 903 #endif 904 unsigned short migration_flags; 905 906 #ifdef CONFIG_PREEMPT_RCU 907 int rcu_read_lock_nesting; 908 union rcu_special rcu_read_unlock_special; 909 struct list_head rcu_node_entry; 910 struct rcu_node *rcu_blocked_node; 911 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 912 913 #ifdef CONFIG_TASKS_RCU 914 unsigned long rcu_tasks_nvcsw; 915 u8 rcu_tasks_holdout; 916 u8 rcu_tasks_idx; 917 int rcu_tasks_idle_cpu; 918 struct list_head rcu_tasks_holdout_list; 919 int rcu_tasks_exit_cpu; 920 struct list_head rcu_tasks_exit_list; 921 #endif /* #ifdef CONFIG_TASKS_RCU */ 922 923 #ifdef CONFIG_TASKS_TRACE_RCU 924 int trc_reader_nesting; 925 int trc_ipi_to_cpu; 926 union rcu_special trc_reader_special; 927 struct list_head trc_holdout_list; 928 struct list_head trc_blkd_node; 929 int trc_blkd_cpu; 930 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 931 932 struct sched_info sched_info; 933 934 struct list_head tasks; 935 #ifdef CONFIG_SMP 936 struct plist_node pushable_tasks; 937 struct rb_node pushable_dl_tasks; 938 #endif 939 940 struct mm_struct *mm; 941 struct mm_struct *active_mm; 942 struct address_space *faults_disabled_mapping; 943 944 int exit_state; 945 int exit_code; 946 int exit_signal; 947 /* The signal sent when the parent dies: */ 948 int pdeath_signal; 949 /* JOBCTL_*, siglock protected: */ 950 unsigned long jobctl; 951 952 /* Used for emulating ABI behavior of previous Linux versions: */ 953 unsigned int personality; 954 955 /* Scheduler bits, serialized by scheduler locks: */ 956 unsigned sched_reset_on_fork:1; 957 unsigned sched_contributes_to_load:1; 958 unsigned sched_migrated:1; 959 unsigned sched_task_hot:1; 960 961 /* Force alignment to the next boundary: */ 962 unsigned :0; 963 964 /* Unserialized, strictly 'current' */ 965 966 /* 967 * This field must not be in the scheduler word above due to wakelist 968 * queueing no longer being serialized by p->on_cpu. However: 969 * 970 * p->XXX = X; ttwu() 971 * schedule() if (p->on_rq && ..) // false 972 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 973 * deactivate_task() ttwu_queue_wakelist()) 974 * p->on_rq = 0; p->sched_remote_wakeup = Y; 975 * 976 * guarantees all stores of 'current' are visible before 977 * ->sched_remote_wakeup gets used, so it can be in this word. 978 */ 979 unsigned sched_remote_wakeup:1; 980 #ifdef CONFIG_RT_MUTEXES 981 unsigned sched_rt_mutex:1; 982 #endif 983 984 /* Bit to tell TOMOYO we're in execve(): */ 985 unsigned in_execve:1; 986 unsigned in_iowait:1; 987 #ifndef TIF_RESTORE_SIGMASK 988 unsigned restore_sigmask:1; 989 #endif 990 #ifdef CONFIG_MEMCG_V1 991 unsigned in_user_fault:1; 992 #endif 993 #ifdef CONFIG_LRU_GEN 994 /* whether the LRU algorithm may apply to this access */ 995 unsigned in_lru_fault:1; 996 #endif 997 #ifdef CONFIG_COMPAT_BRK 998 unsigned brk_randomized:1; 999 #endif 1000 #ifdef CONFIG_CGROUPS 1001 /* disallow userland-initiated cgroup migration */ 1002 unsigned no_cgroup_migration:1; 1003 /* task is frozen/stopped (used by the cgroup freezer) */ 1004 unsigned frozen:1; 1005 #endif 1006 #ifdef CONFIG_BLK_CGROUP 1007 unsigned use_memdelay:1; 1008 #endif 1009 #ifdef CONFIG_PSI 1010 /* Stalled due to lack of memory */ 1011 unsigned in_memstall:1; 1012 #endif 1013 #ifdef CONFIG_PAGE_OWNER 1014 /* Used by page_owner=on to detect recursion in page tracking. */ 1015 unsigned in_page_owner:1; 1016 #endif 1017 #ifdef CONFIG_EVENTFD 1018 /* Recursion prevention for eventfd_signal() */ 1019 unsigned in_eventfd:1; 1020 #endif 1021 #ifdef CONFIG_ARCH_HAS_CPU_PASID 1022 unsigned pasid_activated:1; 1023 #endif 1024 #ifdef CONFIG_X86_BUS_LOCK_DETECT 1025 unsigned reported_split_lock:1; 1026 #endif 1027 #ifdef CONFIG_TASK_DELAY_ACCT 1028 /* delay due to memory thrashing */ 1029 unsigned in_thrashing:1; 1030 #endif 1031 #ifdef CONFIG_PREEMPT_RT 1032 struct netdev_xmit net_xmit; 1033 #endif 1034 unsigned long atomic_flags; /* Flags requiring atomic access. */ 1035 1036 struct restart_block restart_block; 1037 1038 pid_t pid; 1039 pid_t tgid; 1040 1041 #ifdef CONFIG_STACKPROTECTOR 1042 /* Canary value for the -fstack-protector GCC feature: */ 1043 unsigned long stack_canary; 1044 #endif 1045 /* 1046 * Pointers to the (original) parent process, youngest child, younger sibling, 1047 * older sibling, respectively. (p->father can be replaced with 1048 * p->real_parent->pid) 1049 */ 1050 1051 /* Real parent process: */ 1052 struct task_struct __rcu *real_parent; 1053 1054 /* Recipient of SIGCHLD, wait4() reports: */ 1055 struct task_struct __rcu *parent; 1056 1057 /* 1058 * Children/sibling form the list of natural children: 1059 */ 1060 struct list_head children; 1061 struct list_head sibling; 1062 struct task_struct *group_leader; 1063 1064 /* 1065 * 'ptraced' is the list of tasks this task is using ptrace() on. 1066 * 1067 * This includes both natural children and PTRACE_ATTACH targets. 1068 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1069 */ 1070 struct list_head ptraced; 1071 struct list_head ptrace_entry; 1072 1073 /* PID/PID hash table linkage. */ 1074 struct pid *thread_pid; 1075 struct hlist_node pid_links[PIDTYPE_MAX]; 1076 struct list_head thread_node; 1077 1078 struct completion *vfork_done; 1079 1080 /* CLONE_CHILD_SETTID: */ 1081 int __user *set_child_tid; 1082 1083 /* CLONE_CHILD_CLEARTID: */ 1084 int __user *clear_child_tid; 1085 1086 /* PF_KTHREAD | PF_IO_WORKER */ 1087 void *worker_private; 1088 1089 u64 utime; 1090 u64 stime; 1091 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1092 u64 utimescaled; 1093 u64 stimescaled; 1094 #endif 1095 u64 gtime; 1096 struct prev_cputime prev_cputime; 1097 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1098 struct vtime vtime; 1099 #endif 1100 1101 #ifdef CONFIG_NO_HZ_FULL 1102 atomic_t tick_dep_mask; 1103 #endif 1104 /* Context switch counts: */ 1105 unsigned long nvcsw; 1106 unsigned long nivcsw; 1107 1108 /* Monotonic time in nsecs: */ 1109 u64 start_time; 1110 1111 /* Boot based time in nsecs: */ 1112 u64 start_boottime; 1113 1114 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1115 unsigned long min_flt; 1116 unsigned long maj_flt; 1117 1118 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1119 struct posix_cputimers posix_cputimers; 1120 1121 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1122 struct posix_cputimers_work posix_cputimers_work; 1123 #endif 1124 1125 /* Process credentials: */ 1126 1127 /* Tracer's credentials at attach: */ 1128 const struct cred __rcu *ptracer_cred; 1129 1130 /* Objective and real subjective task credentials (COW): */ 1131 const struct cred __rcu *real_cred; 1132 1133 /* Effective (overridable) subjective task credentials (COW): */ 1134 const struct cred __rcu *cred; 1135 1136 #ifdef CONFIG_KEYS 1137 /* Cached requested key. */ 1138 struct key *cached_requested_key; 1139 #endif 1140 1141 /* 1142 * executable name, excluding path. 1143 * 1144 * - normally initialized begin_new_exec() 1145 * - set it with set_task_comm() 1146 * - strscpy_pad() to ensure it is always NUL-terminated and 1147 * zero-padded 1148 * - task_lock() to ensure the operation is atomic and the name is 1149 * fully updated. 1150 */ 1151 char comm[TASK_COMM_LEN]; 1152 1153 struct nameidata *nameidata; 1154 1155 #ifdef CONFIG_SYSVIPC 1156 struct sysv_sem sysvsem; 1157 struct sysv_shm sysvshm; 1158 #endif 1159 #ifdef CONFIG_DETECT_HUNG_TASK 1160 unsigned long last_switch_count; 1161 unsigned long last_switch_time; 1162 #endif 1163 /* Filesystem information: */ 1164 struct fs_struct *fs; 1165 1166 /* Open file information: */ 1167 struct files_struct *files; 1168 1169 #ifdef CONFIG_IO_URING 1170 struct io_uring_task *io_uring; 1171 #endif 1172 1173 /* Namespaces: */ 1174 struct nsproxy *nsproxy; 1175 1176 /* Signal handlers: */ 1177 struct signal_struct *signal; 1178 struct sighand_struct __rcu *sighand; 1179 sigset_t blocked; 1180 sigset_t real_blocked; 1181 /* Restored if set_restore_sigmask() was used: */ 1182 sigset_t saved_sigmask; 1183 struct sigpending pending; 1184 unsigned long sas_ss_sp; 1185 size_t sas_ss_size; 1186 unsigned int sas_ss_flags; 1187 1188 struct callback_head *task_works; 1189 1190 #ifdef CONFIG_AUDIT 1191 #ifdef CONFIG_AUDITSYSCALL 1192 struct audit_context *audit_context; 1193 #endif 1194 kuid_t loginuid; 1195 unsigned int sessionid; 1196 #endif 1197 struct seccomp seccomp; 1198 struct syscall_user_dispatch syscall_dispatch; 1199 1200 /* Thread group tracking: */ 1201 u64 parent_exec_id; 1202 u64 self_exec_id; 1203 1204 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1205 spinlock_t alloc_lock; 1206 1207 /* Protection of the PI data structures: */ 1208 raw_spinlock_t pi_lock; 1209 1210 struct wake_q_node wake_q; 1211 1212 #ifdef CONFIG_RT_MUTEXES 1213 /* PI waiters blocked on a rt_mutex held by this task: */ 1214 struct rb_root_cached pi_waiters; 1215 /* Updated under owner's pi_lock and rq lock */ 1216 struct task_struct *pi_top_task; 1217 /* Deadlock detection and priority inheritance handling: */ 1218 struct rt_mutex_waiter *pi_blocked_on; 1219 #endif 1220 1221 #ifdef CONFIG_DEBUG_MUTEXES 1222 /* Mutex deadlock detection: */ 1223 struct mutex_waiter *blocked_on; 1224 #endif 1225 1226 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1227 int non_block_count; 1228 #endif 1229 1230 #ifdef CONFIG_TRACE_IRQFLAGS 1231 struct irqtrace_events irqtrace; 1232 unsigned int hardirq_threaded; 1233 u64 hardirq_chain_key; 1234 int softirqs_enabled; 1235 int softirq_context; 1236 int irq_config; 1237 #endif 1238 #ifdef CONFIG_PREEMPT_RT 1239 int softirq_disable_cnt; 1240 #endif 1241 1242 #ifdef CONFIG_LOCKDEP 1243 # define MAX_LOCK_DEPTH 48UL 1244 u64 curr_chain_key; 1245 int lockdep_depth; 1246 unsigned int lockdep_recursion; 1247 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1248 #endif 1249 1250 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1251 unsigned int in_ubsan; 1252 #endif 1253 1254 /* Journalling filesystem info: */ 1255 void *journal_info; 1256 1257 /* Stacked block device info: */ 1258 struct bio_list *bio_list; 1259 1260 /* Stack plugging: */ 1261 struct blk_plug *plug; 1262 1263 /* VM state: */ 1264 struct reclaim_state *reclaim_state; 1265 1266 struct io_context *io_context; 1267 1268 #ifdef CONFIG_COMPACTION 1269 struct capture_control *capture_control; 1270 #endif 1271 /* Ptrace state: */ 1272 unsigned long ptrace_message; 1273 kernel_siginfo_t *last_siginfo; 1274 1275 struct task_io_accounting ioac; 1276 #ifdef CONFIG_PSI 1277 /* Pressure stall state */ 1278 unsigned int psi_flags; 1279 #endif 1280 #ifdef CONFIG_TASK_XACCT 1281 /* Accumulated RSS usage: */ 1282 u64 acct_rss_mem1; 1283 /* Accumulated virtual memory usage: */ 1284 u64 acct_vm_mem1; 1285 /* stime + utime since last update: */ 1286 u64 acct_timexpd; 1287 #endif 1288 #ifdef CONFIG_CPUSETS 1289 /* Protected by ->alloc_lock: */ 1290 nodemask_t mems_allowed; 1291 /* Sequence number to catch updates: */ 1292 seqcount_spinlock_t mems_allowed_seq; 1293 int cpuset_mem_spread_rotor; 1294 #endif 1295 #ifdef CONFIG_CGROUPS 1296 /* Control Group info protected by css_set_lock: */ 1297 struct css_set __rcu *cgroups; 1298 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1299 struct list_head cg_list; 1300 #endif 1301 #ifdef CONFIG_X86_CPU_RESCTRL 1302 u32 closid; 1303 u32 rmid; 1304 #endif 1305 #ifdef CONFIG_FUTEX 1306 struct robust_list_head __user *robust_list; 1307 #ifdef CONFIG_COMPAT 1308 struct compat_robust_list_head __user *compat_robust_list; 1309 #endif 1310 struct list_head pi_state_list; 1311 struct futex_pi_state *pi_state_cache; 1312 struct mutex futex_exit_mutex; 1313 unsigned int futex_state; 1314 #endif 1315 #ifdef CONFIG_PERF_EVENTS 1316 u8 perf_recursion[PERF_NR_CONTEXTS]; 1317 struct perf_event_context *perf_event_ctxp; 1318 struct mutex perf_event_mutex; 1319 struct list_head perf_event_list; 1320 struct perf_ctx_data __rcu *perf_ctx_data; 1321 #endif 1322 #ifdef CONFIG_DEBUG_PREEMPT 1323 unsigned long preempt_disable_ip; 1324 #endif 1325 #ifdef CONFIG_NUMA 1326 /* Protected by alloc_lock: */ 1327 struct mempolicy *mempolicy; 1328 short il_prev; 1329 u8 il_weight; 1330 short pref_node_fork; 1331 #endif 1332 #ifdef CONFIG_NUMA_BALANCING 1333 int numa_scan_seq; 1334 unsigned int numa_scan_period; 1335 unsigned int numa_scan_period_max; 1336 int numa_preferred_nid; 1337 unsigned long numa_migrate_retry; 1338 /* Migration stamp: */ 1339 u64 node_stamp; 1340 u64 last_task_numa_placement; 1341 u64 last_sum_exec_runtime; 1342 struct callback_head numa_work; 1343 1344 /* 1345 * This pointer is only modified for current in syscall and 1346 * pagefault context (and for tasks being destroyed), so it can be read 1347 * from any of the following contexts: 1348 * - RCU read-side critical section 1349 * - current->numa_group from everywhere 1350 * - task's runqueue locked, task not running 1351 */ 1352 struct numa_group __rcu *numa_group; 1353 1354 /* 1355 * numa_faults is an array split into four regions: 1356 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1357 * in this precise order. 1358 * 1359 * faults_memory: Exponential decaying average of faults on a per-node 1360 * basis. Scheduling placement decisions are made based on these 1361 * counts. The values remain static for the duration of a PTE scan. 1362 * faults_cpu: Track the nodes the process was running on when a NUMA 1363 * hinting fault was incurred. 1364 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1365 * during the current scan window. When the scan completes, the counts 1366 * in faults_memory and faults_cpu decay and these values are copied. 1367 */ 1368 unsigned long *numa_faults; 1369 unsigned long total_numa_faults; 1370 1371 /* 1372 * numa_faults_locality tracks if faults recorded during the last 1373 * scan window were remote/local or failed to migrate. The task scan 1374 * period is adapted based on the locality of the faults with different 1375 * weights depending on whether they were shared or private faults 1376 */ 1377 unsigned long numa_faults_locality[3]; 1378 1379 unsigned long numa_pages_migrated; 1380 #endif /* CONFIG_NUMA_BALANCING */ 1381 1382 #ifdef CONFIG_RSEQ 1383 struct rseq __user *rseq; 1384 u32 rseq_len; 1385 u32 rseq_sig; 1386 /* 1387 * RmW on rseq_event_mask must be performed atomically 1388 * with respect to preemption. 1389 */ 1390 unsigned long rseq_event_mask; 1391 # ifdef CONFIG_DEBUG_RSEQ 1392 /* 1393 * This is a place holder to save a copy of the rseq fields for 1394 * validation of read-only fields. The struct rseq has a 1395 * variable-length array at the end, so it cannot be used 1396 * directly. Reserve a size large enough for the known fields. 1397 */ 1398 char rseq_fields[sizeof(struct rseq)]; 1399 # endif 1400 #endif 1401 1402 #ifdef CONFIG_SCHED_MM_CID 1403 int mm_cid; /* Current cid in mm */ 1404 int last_mm_cid; /* Most recent cid in mm */ 1405 int migrate_from_cpu; 1406 int mm_cid_active; /* Whether cid bitmap is active */ 1407 struct callback_head cid_work; 1408 #endif 1409 1410 struct tlbflush_unmap_batch tlb_ubc; 1411 1412 /* Cache last used pipe for splice(): */ 1413 struct pipe_inode_info *splice_pipe; 1414 1415 struct page_frag task_frag; 1416 1417 #ifdef CONFIG_TASK_DELAY_ACCT 1418 struct task_delay_info *delays; 1419 #endif 1420 1421 #ifdef CONFIG_FAULT_INJECTION 1422 int make_it_fail; 1423 unsigned int fail_nth; 1424 #endif 1425 /* 1426 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1427 * balance_dirty_pages() for a dirty throttling pause: 1428 */ 1429 int nr_dirtied; 1430 int nr_dirtied_pause; 1431 /* Start of a write-and-pause period: */ 1432 unsigned long dirty_paused_when; 1433 1434 #ifdef CONFIG_LATENCYTOP 1435 int latency_record_count; 1436 struct latency_record latency_record[LT_SAVECOUNT]; 1437 #endif 1438 /* 1439 * Time slack values; these are used to round up poll() and 1440 * select() etc timeout values. These are in nanoseconds. 1441 */ 1442 u64 timer_slack_ns; 1443 u64 default_timer_slack_ns; 1444 1445 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1446 unsigned int kasan_depth; 1447 #endif 1448 1449 #ifdef CONFIG_KCSAN 1450 struct kcsan_ctx kcsan_ctx; 1451 #ifdef CONFIG_TRACE_IRQFLAGS 1452 struct irqtrace_events kcsan_save_irqtrace; 1453 #endif 1454 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1455 int kcsan_stack_depth; 1456 #endif 1457 #endif 1458 1459 #ifdef CONFIG_KMSAN 1460 struct kmsan_ctx kmsan_ctx; 1461 #endif 1462 1463 #if IS_ENABLED(CONFIG_KUNIT) 1464 struct kunit *kunit_test; 1465 #endif 1466 1467 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1468 /* Index of current stored address in ret_stack: */ 1469 int curr_ret_stack; 1470 int curr_ret_depth; 1471 1472 /* Stack of return addresses for return function tracing: */ 1473 unsigned long *ret_stack; 1474 1475 /* Timestamp for last schedule: */ 1476 unsigned long long ftrace_timestamp; 1477 unsigned long long ftrace_sleeptime; 1478 1479 /* 1480 * Number of functions that haven't been traced 1481 * because of depth overrun: 1482 */ 1483 atomic_t trace_overrun; 1484 1485 /* Pause tracing: */ 1486 atomic_t tracing_graph_pause; 1487 #endif 1488 1489 #ifdef CONFIG_TRACING 1490 /* Bitmask and counter of trace recursion: */ 1491 unsigned long trace_recursion; 1492 #endif /* CONFIG_TRACING */ 1493 1494 #ifdef CONFIG_KCOV 1495 /* See kernel/kcov.c for more details. */ 1496 1497 /* Coverage collection mode enabled for this task (0 if disabled): */ 1498 unsigned int kcov_mode; 1499 1500 /* Size of the kcov_area: */ 1501 unsigned int kcov_size; 1502 1503 /* Buffer for coverage collection: */ 1504 void *kcov_area; 1505 1506 /* KCOV descriptor wired with this task or NULL: */ 1507 struct kcov *kcov; 1508 1509 /* KCOV common handle for remote coverage collection: */ 1510 u64 kcov_handle; 1511 1512 /* KCOV sequence number: */ 1513 int kcov_sequence; 1514 1515 /* Collect coverage from softirq context: */ 1516 unsigned int kcov_softirq; 1517 #endif 1518 1519 #ifdef CONFIG_MEMCG_V1 1520 struct mem_cgroup *memcg_in_oom; 1521 #endif 1522 1523 #ifdef CONFIG_MEMCG 1524 /* Number of pages to reclaim on returning to userland: */ 1525 unsigned int memcg_nr_pages_over_high; 1526 1527 /* Used by memcontrol for targeted memcg charge: */ 1528 struct mem_cgroup *active_memcg; 1529 1530 /* Cache for current->cgroups->memcg->objcg lookups: */ 1531 struct obj_cgroup *objcg; 1532 #endif 1533 1534 #ifdef CONFIG_BLK_CGROUP 1535 struct gendisk *throttle_disk; 1536 #endif 1537 1538 #ifdef CONFIG_UPROBES 1539 struct uprobe_task *utask; 1540 #endif 1541 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1542 unsigned int sequential_io; 1543 unsigned int sequential_io_avg; 1544 #endif 1545 struct kmap_ctrl kmap_ctrl; 1546 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1547 unsigned long task_state_change; 1548 # ifdef CONFIG_PREEMPT_RT 1549 unsigned long saved_state_change; 1550 # endif 1551 #endif 1552 struct rcu_head rcu; 1553 refcount_t rcu_users; 1554 int pagefault_disabled; 1555 #ifdef CONFIG_MMU 1556 struct task_struct *oom_reaper_list; 1557 struct timer_list oom_reaper_timer; 1558 #endif 1559 #ifdef CONFIG_VMAP_STACK 1560 struct vm_struct *stack_vm_area; 1561 #endif 1562 #ifdef CONFIG_THREAD_INFO_IN_TASK 1563 /* A live task holds one reference: */ 1564 refcount_t stack_refcount; 1565 #endif 1566 #ifdef CONFIG_LIVEPATCH 1567 int patch_state; 1568 #endif 1569 #ifdef CONFIG_SECURITY 1570 /* Used by LSM modules for access restriction: */ 1571 void *security; 1572 #endif 1573 #ifdef CONFIG_BPF_SYSCALL 1574 /* Used by BPF task local storage */ 1575 struct bpf_local_storage __rcu *bpf_storage; 1576 /* Used for BPF run context */ 1577 struct bpf_run_ctx *bpf_ctx; 1578 #endif 1579 /* Used by BPF for per-TASK xdp storage */ 1580 struct bpf_net_context *bpf_net_context; 1581 1582 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1583 unsigned long lowest_stack; 1584 unsigned long prev_lowest_stack; 1585 #endif 1586 1587 #ifdef CONFIG_X86_MCE 1588 void __user *mce_vaddr; 1589 __u64 mce_kflags; 1590 u64 mce_addr; 1591 __u64 mce_ripv : 1, 1592 mce_whole_page : 1, 1593 __mce_reserved : 62; 1594 struct callback_head mce_kill_me; 1595 int mce_count; 1596 #endif 1597 1598 #ifdef CONFIG_KRETPROBES 1599 struct llist_head kretprobe_instances; 1600 #endif 1601 #ifdef CONFIG_RETHOOK 1602 struct llist_head rethooks; 1603 #endif 1604 1605 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1606 /* 1607 * If L1D flush is supported on mm context switch 1608 * then we use this callback head to queue kill work 1609 * to kill tasks that are not running on SMT disabled 1610 * cores 1611 */ 1612 struct callback_head l1d_flush_kill; 1613 #endif 1614 1615 #ifdef CONFIG_RV 1616 /* 1617 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS. 1618 * If we find justification for more monitors, we can think 1619 * about adding more or developing a dynamic method. So far, 1620 * none of these are justified. 1621 */ 1622 union rv_task_monitor rv[RV_PER_TASK_MONITORS]; 1623 #endif 1624 1625 #ifdef CONFIG_USER_EVENTS 1626 struct user_event_mm *user_event_mm; 1627 #endif 1628 1629 /* 1630 * New fields for task_struct should be added above here, so that 1631 * they are included in the randomized portion of task_struct. 1632 */ 1633 randomized_struct_fields_end 1634 1635 /* CPU-specific state of this task: */ 1636 struct thread_struct thread; 1637 1638 /* 1639 * WARNING: on x86, 'thread_struct' contains a variable-sized 1640 * structure. It *MUST* be at the end of 'task_struct'. 1641 * 1642 * Do not put anything below here! 1643 */ 1644 }; 1645 1646 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1647 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1648 1649 static inline unsigned int __task_state_index(unsigned int tsk_state, 1650 unsigned int tsk_exit_state) 1651 { 1652 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1653 1654 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1655 1656 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1657 state = TASK_REPORT_IDLE; 1658 1659 /* 1660 * We're lying here, but rather than expose a completely new task state 1661 * to userspace, we can make this appear as if the task has gone through 1662 * a regular rt_mutex_lock() call. 1663 * Report frozen tasks as uninterruptible. 1664 */ 1665 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN)) 1666 state = TASK_UNINTERRUPTIBLE; 1667 1668 return fls(state); 1669 } 1670 1671 static inline unsigned int task_state_index(struct task_struct *tsk) 1672 { 1673 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1674 } 1675 1676 static inline char task_index_to_char(unsigned int state) 1677 { 1678 static const char state_char[] = "RSDTtXZPI"; 1679 1680 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1)); 1681 1682 return state_char[state]; 1683 } 1684 1685 static inline char task_state_to_char(struct task_struct *tsk) 1686 { 1687 return task_index_to_char(task_state_index(tsk)); 1688 } 1689 1690 extern struct pid *cad_pid; 1691 1692 /* 1693 * Per process flags 1694 */ 1695 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1696 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1697 #define PF_EXITING 0x00000004 /* Getting shut down */ 1698 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1699 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1700 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1701 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1702 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1703 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1704 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1705 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1706 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1707 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1708 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1709 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1710 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1711 #define PF_KCOMPACTD 0x00010000 /* I am kcompactd */ 1712 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1713 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1714 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1715 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1716 * I am cleaning dirty pages from some other bdi. */ 1717 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1718 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1719 #define PF__HOLE__00800000 0x00800000 1720 #define PF__HOLE__01000000 0x01000000 1721 #define PF__HOLE__02000000 0x02000000 1722 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1723 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1724 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1725 * See memalloc_pin_save() */ 1726 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1727 #define PF__HOLE__40000000 0x40000000 1728 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1729 1730 /* 1731 * Only the _current_ task can read/write to tsk->flags, but other 1732 * tasks can access tsk->flags in readonly mode for example 1733 * with tsk_used_math (like during threaded core dumping). 1734 * There is however an exception to this rule during ptrace 1735 * or during fork: the ptracer task is allowed to write to the 1736 * child->flags of its traced child (same goes for fork, the parent 1737 * can write to the child->flags), because we're guaranteed the 1738 * child is not running and in turn not changing child->flags 1739 * at the same time the parent does it. 1740 */ 1741 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1742 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1743 #define clear_used_math() clear_stopped_child_used_math(current) 1744 #define set_used_math() set_stopped_child_used_math(current) 1745 1746 #define conditional_stopped_child_used_math(condition, child) \ 1747 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1748 1749 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1750 1751 #define copy_to_stopped_child_used_math(child) \ 1752 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1753 1754 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1755 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1756 #define used_math() tsk_used_math(current) 1757 1758 static __always_inline bool is_percpu_thread(void) 1759 { 1760 #ifdef CONFIG_SMP 1761 return (current->flags & PF_NO_SETAFFINITY) && 1762 (current->nr_cpus_allowed == 1); 1763 #else 1764 return true; 1765 #endif 1766 } 1767 1768 /* Per-process atomic flags. */ 1769 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1770 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1771 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1772 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1773 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1774 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1775 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1776 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1777 1778 #define TASK_PFA_TEST(name, func) \ 1779 static inline bool task_##func(struct task_struct *p) \ 1780 { return test_bit(PFA_##name, &p->atomic_flags); } 1781 1782 #define TASK_PFA_SET(name, func) \ 1783 static inline void task_set_##func(struct task_struct *p) \ 1784 { set_bit(PFA_##name, &p->atomic_flags); } 1785 1786 #define TASK_PFA_CLEAR(name, func) \ 1787 static inline void task_clear_##func(struct task_struct *p) \ 1788 { clear_bit(PFA_##name, &p->atomic_flags); } 1789 1790 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1791 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1792 1793 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1794 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1795 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1796 1797 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1798 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1799 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1800 1801 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1802 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1803 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1804 1805 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1806 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1807 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1808 1809 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1810 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1811 1812 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1813 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1814 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1815 1816 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1817 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1818 1819 static inline void 1820 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1821 { 1822 current->flags &= ~flags; 1823 current->flags |= orig_flags & flags; 1824 } 1825 1826 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1827 extern int task_can_attach(struct task_struct *p); 1828 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1829 extern void dl_bw_free(int cpu, u64 dl_bw); 1830 #ifdef CONFIG_SMP 1831 1832 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1833 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1834 1835 /** 1836 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1837 * @p: the task 1838 * @new_mask: CPU affinity mask 1839 * 1840 * Return: zero if successful, or a negative error code 1841 */ 1842 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1843 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1844 extern void release_user_cpus_ptr(struct task_struct *p); 1845 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1846 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1847 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1848 #else 1849 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1850 { 1851 } 1852 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1853 { 1854 /* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */ 1855 if ((*cpumask_bits(new_mask) & 1) == 0) 1856 return -EINVAL; 1857 return 0; 1858 } 1859 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node) 1860 { 1861 if (src->user_cpus_ptr) 1862 return -EINVAL; 1863 return 0; 1864 } 1865 static inline void release_user_cpus_ptr(struct task_struct *p) 1866 { 1867 WARN_ON(p->user_cpus_ptr); 1868 } 1869 1870 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) 1871 { 1872 return 0; 1873 } 1874 #endif 1875 1876 extern int yield_to(struct task_struct *p, bool preempt); 1877 extern void set_user_nice(struct task_struct *p, long nice); 1878 extern int task_prio(const struct task_struct *p); 1879 1880 /** 1881 * task_nice - return the nice value of a given task. 1882 * @p: the task in question. 1883 * 1884 * Return: The nice value [ -20 ... 0 ... 19 ]. 1885 */ 1886 static inline int task_nice(const struct task_struct *p) 1887 { 1888 return PRIO_TO_NICE((p)->static_prio); 1889 } 1890 1891 extern int can_nice(const struct task_struct *p, const int nice); 1892 extern int task_curr(const struct task_struct *p); 1893 extern int idle_cpu(int cpu); 1894 extern int available_idle_cpu(int cpu); 1895 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1896 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1897 extern void sched_set_fifo(struct task_struct *p); 1898 extern void sched_set_fifo_low(struct task_struct *p); 1899 extern void sched_set_normal(struct task_struct *p, int nice); 1900 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1901 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1902 extern struct task_struct *idle_task(int cpu); 1903 1904 /** 1905 * is_idle_task - is the specified task an idle task? 1906 * @p: the task in question. 1907 * 1908 * Return: 1 if @p is an idle task. 0 otherwise. 1909 */ 1910 static __always_inline bool is_idle_task(const struct task_struct *p) 1911 { 1912 return !!(p->flags & PF_IDLE); 1913 } 1914 1915 extern struct task_struct *curr_task(int cpu); 1916 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1917 1918 void yield(void); 1919 1920 union thread_union { 1921 struct task_struct task; 1922 #ifndef CONFIG_THREAD_INFO_IN_TASK 1923 struct thread_info thread_info; 1924 #endif 1925 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1926 }; 1927 1928 #ifndef CONFIG_THREAD_INFO_IN_TASK 1929 extern struct thread_info init_thread_info; 1930 #endif 1931 1932 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1933 1934 #ifdef CONFIG_THREAD_INFO_IN_TASK 1935 # define task_thread_info(task) (&(task)->thread_info) 1936 #else 1937 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1938 #endif 1939 1940 /* 1941 * find a task by one of its numerical ids 1942 * 1943 * find_task_by_pid_ns(): 1944 * finds a task by its pid in the specified namespace 1945 * find_task_by_vpid(): 1946 * finds a task by its virtual pid 1947 * 1948 * see also find_vpid() etc in include/linux/pid.h 1949 */ 1950 1951 extern struct task_struct *find_task_by_vpid(pid_t nr); 1952 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1953 1954 /* 1955 * find a task by its virtual pid and get the task struct 1956 */ 1957 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1958 1959 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1960 extern int wake_up_process(struct task_struct *tsk); 1961 extern void wake_up_new_task(struct task_struct *tsk); 1962 1963 #ifdef CONFIG_SMP 1964 extern void kick_process(struct task_struct *tsk); 1965 #else 1966 static inline void kick_process(struct task_struct *tsk) { } 1967 #endif 1968 1969 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1970 #define set_task_comm(tsk, from) ({ \ 1971 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \ 1972 __set_task_comm(tsk, from, false); \ 1973 }) 1974 1975 /* 1976 * - Why not use task_lock()? 1977 * User space can randomly change their names anyway, so locking for readers 1978 * doesn't make sense. For writers, locking is probably necessary, as a race 1979 * condition could lead to long-term mixed results. 1980 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is 1981 * always NUL-terminated and zero-padded. Therefore the race condition between 1982 * reader and writer is not an issue. 1983 * 1984 * - BUILD_BUG_ON() can help prevent the buf from being truncated. 1985 * Since the callers don't perform any return value checks, this safeguard is 1986 * necessary. 1987 */ 1988 #define get_task_comm(buf, tsk) ({ \ 1989 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \ 1990 strscpy_pad(buf, (tsk)->comm); \ 1991 buf; \ 1992 }) 1993 1994 #ifdef CONFIG_SMP 1995 static __always_inline void scheduler_ipi(void) 1996 { 1997 /* 1998 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1999 * TIF_NEED_RESCHED remotely (for the first time) will also send 2000 * this IPI. 2001 */ 2002 preempt_fold_need_resched(); 2003 } 2004 #else 2005 static inline void scheduler_ipi(void) { } 2006 #endif 2007 2008 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 2009 2010 /* 2011 * Set thread flags in other task's structures. 2012 * See asm/thread_info.h for TIF_xxxx flags available: 2013 */ 2014 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 2015 { 2016 set_ti_thread_flag(task_thread_info(tsk), flag); 2017 } 2018 2019 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2020 { 2021 clear_ti_thread_flag(task_thread_info(tsk), flag); 2022 } 2023 2024 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 2025 bool value) 2026 { 2027 update_ti_thread_flag(task_thread_info(tsk), flag, value); 2028 } 2029 2030 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2031 { 2032 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2033 } 2034 2035 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2036 { 2037 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2038 } 2039 2040 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2041 { 2042 return test_ti_thread_flag(task_thread_info(tsk), flag); 2043 } 2044 2045 static inline void set_tsk_need_resched(struct task_struct *tsk) 2046 { 2047 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2048 } 2049 2050 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2051 { 2052 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY, 2053 (atomic_long_t *)&task_thread_info(tsk)->flags); 2054 } 2055 2056 static inline int test_tsk_need_resched(struct task_struct *tsk) 2057 { 2058 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2059 } 2060 2061 /* 2062 * cond_resched() and cond_resched_lock(): latency reduction via 2063 * explicit rescheduling in places that are safe. The return 2064 * value indicates whether a reschedule was done in fact. 2065 * cond_resched_lock() will drop the spinlock before scheduling, 2066 */ 2067 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2068 extern int __cond_resched(void); 2069 2070 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2071 2072 void sched_dynamic_klp_enable(void); 2073 void sched_dynamic_klp_disable(void); 2074 2075 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2076 2077 static __always_inline int _cond_resched(void) 2078 { 2079 return static_call_mod(cond_resched)(); 2080 } 2081 2082 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2083 2084 extern int dynamic_cond_resched(void); 2085 2086 static __always_inline int _cond_resched(void) 2087 { 2088 return dynamic_cond_resched(); 2089 } 2090 2091 #else /* !CONFIG_PREEMPTION */ 2092 2093 static inline int _cond_resched(void) 2094 { 2095 klp_sched_try_switch(); 2096 return __cond_resched(); 2097 } 2098 2099 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2100 2101 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2102 2103 static inline int _cond_resched(void) 2104 { 2105 klp_sched_try_switch(); 2106 return 0; 2107 } 2108 2109 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2110 2111 #define cond_resched() ({ \ 2112 __might_resched(__FILE__, __LINE__, 0); \ 2113 _cond_resched(); \ 2114 }) 2115 2116 extern int __cond_resched_lock(spinlock_t *lock); 2117 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2118 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2119 2120 #define MIGHT_RESCHED_RCU_SHIFT 8 2121 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2122 2123 #ifndef CONFIG_PREEMPT_RT 2124 /* 2125 * Non RT kernels have an elevated preempt count due to the held lock, 2126 * but are not allowed to be inside a RCU read side critical section 2127 */ 2128 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2129 #else 2130 /* 2131 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2132 * cond_resched*lock() has to take that into account because it checks for 2133 * preempt_count() and rcu_preempt_depth(). 2134 */ 2135 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2136 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2137 #endif 2138 2139 #define cond_resched_lock(lock) ({ \ 2140 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2141 __cond_resched_lock(lock); \ 2142 }) 2143 2144 #define cond_resched_rwlock_read(lock) ({ \ 2145 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2146 __cond_resched_rwlock_read(lock); \ 2147 }) 2148 2149 #define cond_resched_rwlock_write(lock) ({ \ 2150 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2151 __cond_resched_rwlock_write(lock); \ 2152 }) 2153 2154 static __always_inline bool need_resched(void) 2155 { 2156 return unlikely(tif_need_resched()); 2157 } 2158 2159 /* 2160 * Wrappers for p->thread_info->cpu access. No-op on UP. 2161 */ 2162 #ifdef CONFIG_SMP 2163 2164 static inline unsigned int task_cpu(const struct task_struct *p) 2165 { 2166 return READ_ONCE(task_thread_info(p)->cpu); 2167 } 2168 2169 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2170 2171 #else 2172 2173 static inline unsigned int task_cpu(const struct task_struct *p) 2174 { 2175 return 0; 2176 } 2177 2178 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2179 { 2180 } 2181 2182 #endif /* CONFIG_SMP */ 2183 2184 static inline bool task_is_runnable(struct task_struct *p) 2185 { 2186 return p->on_rq && !p->se.sched_delayed; 2187 } 2188 2189 extern bool sched_task_on_rq(struct task_struct *p); 2190 extern unsigned long get_wchan(struct task_struct *p); 2191 extern struct task_struct *cpu_curr_snapshot(int cpu); 2192 2193 #include <linux/spinlock.h> 2194 2195 /* 2196 * In order to reduce various lock holder preemption latencies provide an 2197 * interface to see if a vCPU is currently running or not. 2198 * 2199 * This allows us to terminate optimistic spin loops and block, analogous to 2200 * the native optimistic spin heuristic of testing if the lock owner task is 2201 * running or not. 2202 */ 2203 #ifndef vcpu_is_preempted 2204 static inline bool vcpu_is_preempted(int cpu) 2205 { 2206 return false; 2207 } 2208 #endif 2209 2210 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2211 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2212 2213 #ifndef TASK_SIZE_OF 2214 #define TASK_SIZE_OF(tsk) TASK_SIZE 2215 #endif 2216 2217 #ifdef CONFIG_SMP 2218 static inline bool owner_on_cpu(struct task_struct *owner) 2219 { 2220 /* 2221 * As lock holder preemption issue, we both skip spinning if 2222 * task is not on cpu or its cpu is preempted 2223 */ 2224 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2225 } 2226 2227 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2228 unsigned long sched_cpu_util(int cpu); 2229 #endif /* CONFIG_SMP */ 2230 2231 #ifdef CONFIG_SCHED_CORE 2232 extern void sched_core_free(struct task_struct *tsk); 2233 extern void sched_core_fork(struct task_struct *p); 2234 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2235 unsigned long uaddr); 2236 extern int sched_core_idle_cpu(int cpu); 2237 #else 2238 static inline void sched_core_free(struct task_struct *tsk) { } 2239 static inline void sched_core_fork(struct task_struct *p) { } 2240 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2241 #endif 2242 2243 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2244 2245 #ifdef CONFIG_MEM_ALLOC_PROFILING 2246 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2247 { 2248 swap(current->alloc_tag, tag); 2249 return tag; 2250 } 2251 2252 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2253 { 2254 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2255 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2256 #endif 2257 current->alloc_tag = old; 2258 } 2259 #else 2260 #define alloc_tag_save(_tag) NULL 2261 #define alloc_tag_restore(_tag, _old) do {} while (0) 2262 #endif 2263 2264 #endif 2265