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 pid_namespace; 69 struct pipe_inode_info; 70 struct rcu_node; 71 struct reclaim_state; 72 struct robust_list_head; 73 struct root_domain; 74 struct rq; 75 struct sched_attr; 76 struct sched_dl_entity; 77 struct seq_file; 78 struct sighand_struct; 79 struct signal_struct; 80 struct task_delay_info; 81 struct task_group; 82 struct task_struct; 83 struct user_event_mm; 84 85 #include <linux/sched/ext.h> 86 87 /* 88 * Task state bitmask. NOTE! These bits are also 89 * encoded in fs/proc/array.c: get_task_state(). 90 * 91 * We have two separate sets of flags: task->__state 92 * is about runnability, while task->exit_state are 93 * about the task exiting. Confusing, but this way 94 * modifying one set can't modify the other one by 95 * mistake. 96 */ 97 98 /* Used in tsk->__state: */ 99 #define TASK_RUNNING 0x00000000 100 #define TASK_INTERRUPTIBLE 0x00000001 101 #define TASK_UNINTERRUPTIBLE 0x00000002 102 #define __TASK_STOPPED 0x00000004 103 #define __TASK_TRACED 0x00000008 104 /* Used in tsk->exit_state: */ 105 #define EXIT_DEAD 0x00000010 106 #define EXIT_ZOMBIE 0x00000020 107 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 108 /* Used in tsk->__state again: */ 109 #define TASK_PARKED 0x00000040 110 #define TASK_DEAD 0x00000080 111 #define TASK_WAKEKILL 0x00000100 112 #define TASK_WAKING 0x00000200 113 #define TASK_NOLOAD 0x00000400 114 #define TASK_NEW 0x00000800 115 #define TASK_RTLOCK_WAIT 0x00001000 116 #define TASK_FREEZABLE 0x00002000 117 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP)) 118 #define TASK_FROZEN 0x00008000 119 #define TASK_STATE_MAX 0x00010000 120 121 #define TASK_ANY (TASK_STATE_MAX-1) 122 123 /* 124 * DO NOT ADD ANY NEW USERS ! 125 */ 126 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE) 127 128 /* Convenience macros for the sake of set_current_state: */ 129 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 130 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 131 #define TASK_TRACED __TASK_TRACED 132 133 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 134 135 /* Convenience macros for the sake of wake_up(): */ 136 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 137 138 /* get_task_state(): */ 139 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 140 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 141 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 142 TASK_PARKED) 143 144 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 145 146 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 147 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 148 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 149 150 /* 151 * Special states are those that do not use the normal wait-loop pattern. See 152 * the comment with set_special_state(). 153 */ 154 #define is_special_task_state(state) \ 155 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \ 156 TASK_DEAD | TASK_FROZEN)) 157 158 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 159 # define debug_normal_state_change(state_value) \ 160 do { \ 161 WARN_ON_ONCE(is_special_task_state(state_value)); \ 162 current->task_state_change = _THIS_IP_; \ 163 } while (0) 164 165 # define debug_special_state_change(state_value) \ 166 do { \ 167 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 168 current->task_state_change = _THIS_IP_; \ 169 } while (0) 170 171 # define debug_rtlock_wait_set_state() \ 172 do { \ 173 current->saved_state_change = current->task_state_change;\ 174 current->task_state_change = _THIS_IP_; \ 175 } while (0) 176 177 # define debug_rtlock_wait_restore_state() \ 178 do { \ 179 current->task_state_change = current->saved_state_change;\ 180 } while (0) 181 182 #else 183 # define debug_normal_state_change(cond) do { } while (0) 184 # define debug_special_state_change(cond) do { } while (0) 185 # define debug_rtlock_wait_set_state() do { } while (0) 186 # define debug_rtlock_wait_restore_state() do { } while (0) 187 #endif 188 189 /* 190 * set_current_state() includes a barrier so that the write of current->__state 191 * is correctly serialised wrt the caller's subsequent test of whether to 192 * actually sleep: 193 * 194 * for (;;) { 195 * set_current_state(TASK_UNINTERRUPTIBLE); 196 * if (CONDITION) 197 * break; 198 * 199 * schedule(); 200 * } 201 * __set_current_state(TASK_RUNNING); 202 * 203 * If the caller does not need such serialisation (because, for instance, the 204 * CONDITION test and condition change and wakeup are under the same lock) then 205 * use __set_current_state(). 206 * 207 * The above is typically ordered against the wakeup, which does: 208 * 209 * CONDITION = 1; 210 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 211 * 212 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 213 * accessing p->__state. 214 * 215 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is, 216 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 217 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 218 * 219 * However, with slightly different timing the wakeup TASK_RUNNING store can 220 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 221 * a problem either because that will result in one extra go around the loop 222 * and our @cond test will save the day. 223 * 224 * Also see the comments of try_to_wake_up(). 225 */ 226 #define __set_current_state(state_value) \ 227 do { \ 228 debug_normal_state_change((state_value)); \ 229 WRITE_ONCE(current->__state, (state_value)); \ 230 } while (0) 231 232 #define set_current_state(state_value) \ 233 do { \ 234 debug_normal_state_change((state_value)); \ 235 smp_store_mb(current->__state, (state_value)); \ 236 } while (0) 237 238 /* 239 * set_special_state() should be used for those states when the blocking task 240 * can not use the regular condition based wait-loop. In that case we must 241 * serialize against wakeups such that any possible in-flight TASK_RUNNING 242 * stores will not collide with our state change. 243 */ 244 #define set_special_state(state_value) \ 245 do { \ 246 unsigned long flags; /* may shadow */ \ 247 \ 248 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 249 debug_special_state_change((state_value)); \ 250 WRITE_ONCE(current->__state, (state_value)); \ 251 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 252 } while (0) 253 254 /* 255 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 256 * 257 * RT's spin/rwlock substitutions are state preserving. The state of the 258 * task when blocking on the lock is saved in task_struct::saved_state and 259 * restored after the lock has been acquired. These operations are 260 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 261 * lock related wakeups while the task is blocked on the lock are 262 * redirected to operate on task_struct::saved_state to ensure that these 263 * are not dropped. On restore task_struct::saved_state is set to 264 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 265 * 266 * The lock operation looks like this: 267 * 268 * current_save_and_set_rtlock_wait_state(); 269 * for (;;) { 270 * if (try_lock()) 271 * break; 272 * raw_spin_unlock_irq(&lock->wait_lock); 273 * schedule_rtlock(); 274 * raw_spin_lock_irq(&lock->wait_lock); 275 * set_current_state(TASK_RTLOCK_WAIT); 276 * } 277 * current_restore_rtlock_saved_state(); 278 */ 279 #define current_save_and_set_rtlock_wait_state() \ 280 do { \ 281 lockdep_assert_irqs_disabled(); \ 282 raw_spin_lock(¤t->pi_lock); \ 283 current->saved_state = current->__state; \ 284 debug_rtlock_wait_set_state(); \ 285 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 286 raw_spin_unlock(¤t->pi_lock); \ 287 } while (0); 288 289 #define current_restore_rtlock_saved_state() \ 290 do { \ 291 lockdep_assert_irqs_disabled(); \ 292 raw_spin_lock(¤t->pi_lock); \ 293 debug_rtlock_wait_restore_state(); \ 294 WRITE_ONCE(current->__state, current->saved_state); \ 295 current->saved_state = TASK_RUNNING; \ 296 raw_spin_unlock(¤t->pi_lock); \ 297 } while (0); 298 299 #define get_current_state() READ_ONCE(current->__state) 300 301 /* 302 * Define the task command name length as enum, then it can be visible to 303 * BPF programs. 304 */ 305 enum { 306 TASK_COMM_LEN = 16, 307 }; 308 309 extern void sched_tick(void); 310 311 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 312 313 extern long schedule_timeout(long timeout); 314 extern long schedule_timeout_interruptible(long timeout); 315 extern long schedule_timeout_killable(long timeout); 316 extern long schedule_timeout_uninterruptible(long timeout); 317 extern long schedule_timeout_idle(long timeout); 318 asmlinkage void schedule(void); 319 extern void schedule_preempt_disabled(void); 320 asmlinkage void preempt_schedule_irq(void); 321 #ifdef CONFIG_PREEMPT_RT 322 extern void schedule_rtlock(void); 323 #endif 324 325 extern int __must_check io_schedule_prepare(void); 326 extern void io_schedule_finish(int token); 327 extern long io_schedule_timeout(long timeout); 328 extern void io_schedule(void); 329 330 /** 331 * struct prev_cputime - snapshot of system and user cputime 332 * @utime: time spent in user mode 333 * @stime: time spent in system mode 334 * @lock: protects the above two fields 335 * 336 * Stores previous user/system time values such that we can guarantee 337 * monotonicity. 338 */ 339 struct prev_cputime { 340 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 341 u64 utime; 342 u64 stime; 343 raw_spinlock_t lock; 344 #endif 345 }; 346 347 enum vtime_state { 348 /* Task is sleeping or running in a CPU with VTIME inactive: */ 349 VTIME_INACTIVE = 0, 350 /* Task is idle */ 351 VTIME_IDLE, 352 /* Task runs in kernelspace in a CPU with VTIME active: */ 353 VTIME_SYS, 354 /* Task runs in userspace in a CPU with VTIME active: */ 355 VTIME_USER, 356 /* Task runs as guests in a CPU with VTIME active: */ 357 VTIME_GUEST, 358 }; 359 360 struct vtime { 361 seqcount_t seqcount; 362 unsigned long long starttime; 363 enum vtime_state state; 364 unsigned int cpu; 365 u64 utime; 366 u64 stime; 367 u64 gtime; 368 }; 369 370 /* 371 * Utilization clamp constraints. 372 * @UCLAMP_MIN: Minimum utilization 373 * @UCLAMP_MAX: Maximum utilization 374 * @UCLAMP_CNT: Utilization clamp constraints count 375 */ 376 enum uclamp_id { 377 UCLAMP_MIN = 0, 378 UCLAMP_MAX, 379 UCLAMP_CNT 380 }; 381 382 #ifdef CONFIG_SMP 383 extern struct root_domain def_root_domain; 384 extern struct mutex sched_domains_mutex; 385 #endif 386 387 struct sched_param { 388 int sched_priority; 389 }; 390 391 struct sched_info { 392 #ifdef CONFIG_SCHED_INFO 393 /* Cumulative counters: */ 394 395 /* # of times we have run on this CPU: */ 396 unsigned long pcount; 397 398 /* Time spent waiting on a runqueue: */ 399 unsigned long long run_delay; 400 401 /* Max time spent waiting on a runqueue: */ 402 unsigned long long max_run_delay; 403 404 /* Min time spent waiting on a runqueue: */ 405 unsigned long long min_run_delay; 406 407 /* Timestamps: */ 408 409 /* When did we last run on a CPU? */ 410 unsigned long long last_arrival; 411 412 /* When were we last queued to run? */ 413 unsigned long long last_queued; 414 415 #endif /* CONFIG_SCHED_INFO */ 416 }; 417 418 /* 419 * Integer metrics need fixed point arithmetic, e.g., sched/fair 420 * has a few: load, load_avg, util_avg, freq, and capacity. 421 * 422 * We define a basic fixed point arithmetic range, and then formalize 423 * all these metrics based on that basic range. 424 */ 425 # define SCHED_FIXEDPOINT_SHIFT 10 426 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 427 428 /* Increase resolution of cpu_capacity calculations */ 429 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 430 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 431 432 struct load_weight { 433 unsigned long weight; 434 u32 inv_weight; 435 }; 436 437 /* 438 * The load/runnable/util_avg accumulates an infinite geometric series 439 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 440 * 441 * [load_avg definition] 442 * 443 * load_avg = runnable% * scale_load_down(load) 444 * 445 * [runnable_avg definition] 446 * 447 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 448 * 449 * [util_avg definition] 450 * 451 * util_avg = running% * SCHED_CAPACITY_SCALE 452 * 453 * where runnable% is the time ratio that a sched_entity is runnable and 454 * running% the time ratio that a sched_entity is running. 455 * 456 * For cfs_rq, they are the aggregated values of all runnable and blocked 457 * sched_entities. 458 * 459 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 460 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 461 * for computing those signals (see update_rq_clock_pelt()) 462 * 463 * N.B., the above ratios (runnable% and running%) themselves are in the 464 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 465 * to as large a range as necessary. This is for example reflected by 466 * util_avg's SCHED_CAPACITY_SCALE. 467 * 468 * [Overflow issue] 469 * 470 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 471 * with the highest load (=88761), always runnable on a single cfs_rq, 472 * and should not overflow as the number already hits PID_MAX_LIMIT. 473 * 474 * For all other cases (including 32-bit kernels), struct load_weight's 475 * weight will overflow first before we do, because: 476 * 477 * Max(load_avg) <= Max(load.weight) 478 * 479 * Then it is the load_weight's responsibility to consider overflow 480 * issues. 481 */ 482 struct sched_avg { 483 u64 last_update_time; 484 u64 load_sum; 485 u64 runnable_sum; 486 u32 util_sum; 487 u32 period_contrib; 488 unsigned long load_avg; 489 unsigned long runnable_avg; 490 unsigned long util_avg; 491 unsigned int util_est; 492 } ____cacheline_aligned; 493 494 /* 495 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 496 * updates. When a task is dequeued, its util_est should not be updated if its 497 * util_avg has not been updated in the meantime. 498 * This information is mapped into the MSB bit of util_est at dequeue time. 499 * Since max value of util_est for a task is 1024 (PELT util_avg for a task) 500 * it is safe to use MSB. 501 */ 502 #define UTIL_EST_WEIGHT_SHIFT 2 503 #define UTIL_AVG_UNCHANGED 0x80000000 504 505 struct sched_statistics { 506 #ifdef CONFIG_SCHEDSTATS 507 u64 wait_start; 508 u64 wait_max; 509 u64 wait_count; 510 u64 wait_sum; 511 u64 iowait_count; 512 u64 iowait_sum; 513 514 u64 sleep_start; 515 u64 sleep_max; 516 s64 sum_sleep_runtime; 517 518 u64 block_start; 519 u64 block_max; 520 s64 sum_block_runtime; 521 522 s64 exec_max; 523 u64 slice_max; 524 525 u64 nr_migrations_cold; 526 u64 nr_failed_migrations_affine; 527 u64 nr_failed_migrations_running; 528 u64 nr_failed_migrations_hot; 529 u64 nr_forced_migrations; 530 531 u64 nr_wakeups; 532 u64 nr_wakeups_sync; 533 u64 nr_wakeups_migrate; 534 u64 nr_wakeups_local; 535 u64 nr_wakeups_remote; 536 u64 nr_wakeups_affine; 537 u64 nr_wakeups_affine_attempts; 538 u64 nr_wakeups_passive; 539 u64 nr_wakeups_idle; 540 541 #ifdef CONFIG_SCHED_CORE 542 u64 core_forceidle_sum; 543 #endif 544 #endif /* CONFIG_SCHEDSTATS */ 545 } ____cacheline_aligned; 546 547 struct sched_entity { 548 /* For load-balancing: */ 549 struct load_weight load; 550 struct rb_node run_node; 551 u64 deadline; 552 u64 min_vruntime; 553 u64 min_slice; 554 555 struct list_head group_node; 556 unsigned char on_rq; 557 unsigned char sched_delayed; 558 unsigned char rel_deadline; 559 unsigned char custom_slice; 560 /* hole */ 561 562 u64 exec_start; 563 u64 sum_exec_runtime; 564 u64 prev_sum_exec_runtime; 565 u64 vruntime; 566 s64 vlag; 567 u64 slice; 568 569 u64 nr_migrations; 570 571 #ifdef CONFIG_FAIR_GROUP_SCHED 572 int depth; 573 struct sched_entity *parent; 574 /* rq on which this entity is (to be) queued: */ 575 struct cfs_rq *cfs_rq; 576 /* rq "owned" by this entity/group: */ 577 struct cfs_rq *my_q; 578 /* cached value of my_q->h_nr_running */ 579 unsigned long runnable_weight; 580 #endif 581 582 #ifdef CONFIG_SMP 583 /* 584 * Per entity load average tracking. 585 * 586 * Put into separate cache line so it does not 587 * collide with read-mostly values above. 588 */ 589 struct sched_avg avg; 590 #endif 591 }; 592 593 struct sched_rt_entity { 594 struct list_head run_list; 595 unsigned long timeout; 596 unsigned long watchdog_stamp; 597 unsigned int time_slice; 598 unsigned short on_rq; 599 unsigned short on_list; 600 601 struct sched_rt_entity *back; 602 #ifdef CONFIG_RT_GROUP_SCHED 603 struct sched_rt_entity *parent; 604 /* rq on which this entity is (to be) queued: */ 605 struct rt_rq *rt_rq; 606 /* rq "owned" by this entity/group: */ 607 struct rt_rq *my_q; 608 #endif 609 } __randomize_layout; 610 611 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *); 612 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *); 613 614 struct sched_dl_entity { 615 struct rb_node rb_node; 616 617 /* 618 * Original scheduling parameters. Copied here from sched_attr 619 * during sched_setattr(), they will remain the same until 620 * the next sched_setattr(). 621 */ 622 u64 dl_runtime; /* Maximum runtime for each instance */ 623 u64 dl_deadline; /* Relative deadline of each instance */ 624 u64 dl_period; /* Separation of two instances (period) */ 625 u64 dl_bw; /* dl_runtime / dl_period */ 626 u64 dl_density; /* dl_runtime / dl_deadline */ 627 628 /* 629 * Actual scheduling parameters. Initialized with the values above, 630 * they are continuously updated during task execution. Note that 631 * the remaining runtime could be < 0 in case we are in overrun. 632 */ 633 s64 runtime; /* Remaining runtime for this instance */ 634 u64 deadline; /* Absolute deadline for this instance */ 635 unsigned int flags; /* Specifying the scheduler behaviour */ 636 637 /* 638 * Some bool flags: 639 * 640 * @dl_throttled tells if we exhausted the runtime. If so, the 641 * task has to wait for a replenishment to be performed at the 642 * next firing of dl_timer. 643 * 644 * @dl_yielded tells if task gave up the CPU before consuming 645 * all its available runtime during the last job. 646 * 647 * @dl_non_contending tells if the task is inactive while still 648 * contributing to the active utilization. In other words, it 649 * indicates if the inactive timer has been armed and its handler 650 * has not been executed yet. This flag is useful to avoid race 651 * conditions between the inactive timer handler and the wakeup 652 * code. 653 * 654 * @dl_overrun tells if the task asked to be informed about runtime 655 * overruns. 656 * 657 * @dl_server tells if this is a server entity. 658 * 659 * @dl_defer tells if this is a deferred or regular server. For 660 * now only defer server exists. 661 * 662 * @dl_defer_armed tells if the deferrable server is waiting 663 * for the replenishment timer to activate it. 664 * 665 * @dl_server_active tells if the dlserver is active(started). 666 * dlserver is started on first cfs enqueue on an idle runqueue 667 * and is stopped when a dequeue results in 0 cfs tasks on the 668 * runqueue. In other words, dlserver is active only when cpu's 669 * runqueue has atleast one cfs task. 670 * 671 * @dl_defer_running tells if the deferrable server is actually 672 * running, skipping the defer phase. 673 */ 674 unsigned int dl_throttled : 1; 675 unsigned int dl_yielded : 1; 676 unsigned int dl_non_contending : 1; 677 unsigned int dl_overrun : 1; 678 unsigned int dl_server : 1; 679 unsigned int dl_server_active : 1; 680 unsigned int dl_defer : 1; 681 unsigned int dl_defer_armed : 1; 682 unsigned int dl_defer_running : 1; 683 684 /* 685 * Bandwidth enforcement timer. Each -deadline task has its 686 * own bandwidth to be enforced, thus we need one timer per task. 687 */ 688 struct hrtimer dl_timer; 689 690 /* 691 * Inactive timer, responsible for decreasing the active utilization 692 * at the "0-lag time". When a -deadline task blocks, it contributes 693 * to GRUB's active utilization until the "0-lag time", hence a 694 * timer is needed to decrease the active utilization at the correct 695 * time. 696 */ 697 struct hrtimer inactive_timer; 698 699 /* 700 * Bits for DL-server functionality. Also see the comment near 701 * dl_server_update(). 702 * 703 * @rq the runqueue this server is for 704 * 705 * @server_has_tasks() returns true if @server_pick return a 706 * runnable task. 707 */ 708 struct rq *rq; 709 dl_server_has_tasks_f server_has_tasks; 710 dl_server_pick_f server_pick_task; 711 712 #ifdef CONFIG_RT_MUTEXES 713 /* 714 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 715 * pi_se points to the donor, otherwise points to the dl_se it belongs 716 * to (the original one/itself). 717 */ 718 struct sched_dl_entity *pi_se; 719 #endif 720 }; 721 722 #ifdef CONFIG_UCLAMP_TASK 723 /* Number of utilization clamp buckets (shorter alias) */ 724 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 725 726 /* 727 * Utilization clamp for a scheduling entity 728 * @value: clamp value "assigned" to a se 729 * @bucket_id: bucket index corresponding to the "assigned" value 730 * @active: the se is currently refcounted in a rq's bucket 731 * @user_defined: the requested clamp value comes from user-space 732 * 733 * The bucket_id is the index of the clamp bucket matching the clamp value 734 * which is pre-computed and stored to avoid expensive integer divisions from 735 * the fast path. 736 * 737 * The active bit is set whenever a task has got an "effective" value assigned, 738 * which can be different from the clamp value "requested" from user-space. 739 * This allows to know a task is refcounted in the rq's bucket corresponding 740 * to the "effective" bucket_id. 741 * 742 * The user_defined bit is set whenever a task has got a task-specific clamp 743 * value requested from userspace, i.e. the system defaults apply to this task 744 * just as a restriction. This allows to relax default clamps when a less 745 * restrictive task-specific value has been requested, thus allowing to 746 * implement a "nice" semantic. For example, a task running with a 20% 747 * default boost can still drop its own boosting to 0%. 748 */ 749 struct uclamp_se { 750 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 751 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 752 unsigned int active : 1; 753 unsigned int user_defined : 1; 754 }; 755 #endif /* CONFIG_UCLAMP_TASK */ 756 757 union rcu_special { 758 struct { 759 u8 blocked; 760 u8 need_qs; 761 u8 exp_hint; /* Hint for performance. */ 762 u8 need_mb; /* Readers need smp_mb(). */ 763 } b; /* Bits. */ 764 u32 s; /* Set of bits. */ 765 }; 766 767 enum perf_event_task_context { 768 perf_invalid_context = -1, 769 perf_hw_context = 0, 770 perf_sw_context, 771 perf_nr_task_contexts, 772 }; 773 774 /* 775 * Number of contexts where an event can trigger: 776 * task, softirq, hardirq, nmi. 777 */ 778 #define PERF_NR_CONTEXTS 4 779 780 struct wake_q_node { 781 struct wake_q_node *next; 782 }; 783 784 struct kmap_ctrl { 785 #ifdef CONFIG_KMAP_LOCAL 786 int idx; 787 pte_t pteval[KM_MAX_IDX]; 788 #endif 789 }; 790 791 struct task_struct { 792 #ifdef CONFIG_THREAD_INFO_IN_TASK 793 /* 794 * For reasons of header soup (see current_thread_info()), this 795 * must be the first element of task_struct. 796 */ 797 struct thread_info thread_info; 798 #endif 799 unsigned int __state; 800 801 /* saved state for "spinlock sleepers" */ 802 unsigned int saved_state; 803 804 /* 805 * This begins the randomizable portion of task_struct. Only 806 * scheduling-critical items should be added above here. 807 */ 808 randomized_struct_fields_start 809 810 void *stack; 811 refcount_t usage; 812 /* Per task flags (PF_*), defined further below: */ 813 unsigned int flags; 814 unsigned int ptrace; 815 816 #ifdef CONFIG_MEM_ALLOC_PROFILING 817 struct alloc_tag *alloc_tag; 818 #endif 819 820 #ifdef CONFIG_SMP 821 int on_cpu; 822 struct __call_single_node wake_entry; 823 unsigned int wakee_flips; 824 unsigned long wakee_flip_decay_ts; 825 struct task_struct *last_wakee; 826 827 /* 828 * recent_used_cpu is initially set as the last CPU used by a task 829 * that wakes affine another task. Waker/wakee relationships can 830 * push tasks around a CPU where each wakeup moves to the next one. 831 * Tracking a recently used CPU allows a quick search for a recently 832 * used CPU that may be idle. 833 */ 834 int recent_used_cpu; 835 int wake_cpu; 836 #endif 837 int on_rq; 838 839 int prio; 840 int static_prio; 841 int normal_prio; 842 unsigned int rt_priority; 843 844 struct sched_entity se; 845 struct sched_rt_entity rt; 846 struct sched_dl_entity dl; 847 struct sched_dl_entity *dl_server; 848 #ifdef CONFIG_SCHED_CLASS_EXT 849 struct sched_ext_entity scx; 850 #endif 851 const struct sched_class *sched_class; 852 853 #ifdef CONFIG_SCHED_CORE 854 struct rb_node core_node; 855 unsigned long core_cookie; 856 unsigned int core_occupation; 857 #endif 858 859 #ifdef CONFIG_CGROUP_SCHED 860 struct task_group *sched_task_group; 861 #endif 862 863 864 #ifdef CONFIG_UCLAMP_TASK 865 /* 866 * Clamp values requested for a scheduling entity. 867 * Must be updated with task_rq_lock() held. 868 */ 869 struct uclamp_se uclamp_req[UCLAMP_CNT]; 870 /* 871 * Effective clamp values used for a scheduling entity. 872 * Must be updated with task_rq_lock() held. 873 */ 874 struct uclamp_se uclamp[UCLAMP_CNT]; 875 #endif 876 877 struct sched_statistics stats; 878 879 #ifdef CONFIG_PREEMPT_NOTIFIERS 880 /* List of struct preempt_notifier: */ 881 struct hlist_head preempt_notifiers; 882 #endif 883 884 #ifdef CONFIG_BLK_DEV_IO_TRACE 885 unsigned int btrace_seq; 886 #endif 887 888 unsigned int policy; 889 unsigned long max_allowed_capacity; 890 int nr_cpus_allowed; 891 const cpumask_t *cpus_ptr; 892 cpumask_t *user_cpus_ptr; 893 cpumask_t cpus_mask; 894 void *migration_pending; 895 #ifdef CONFIG_SMP 896 unsigned short migration_disabled; 897 #endif 898 unsigned short migration_flags; 899 900 #ifdef CONFIG_PREEMPT_RCU 901 int rcu_read_lock_nesting; 902 union rcu_special rcu_read_unlock_special; 903 struct list_head rcu_node_entry; 904 struct rcu_node *rcu_blocked_node; 905 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 906 907 #ifdef CONFIG_TASKS_RCU 908 unsigned long rcu_tasks_nvcsw; 909 u8 rcu_tasks_holdout; 910 u8 rcu_tasks_idx; 911 int rcu_tasks_idle_cpu; 912 struct list_head rcu_tasks_holdout_list; 913 int rcu_tasks_exit_cpu; 914 struct list_head rcu_tasks_exit_list; 915 #endif /* #ifdef CONFIG_TASKS_RCU */ 916 917 #ifdef CONFIG_TASKS_TRACE_RCU 918 int trc_reader_nesting; 919 int trc_ipi_to_cpu; 920 union rcu_special trc_reader_special; 921 struct list_head trc_holdout_list; 922 struct list_head trc_blkd_node; 923 int trc_blkd_cpu; 924 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 925 926 struct sched_info sched_info; 927 928 struct list_head tasks; 929 #ifdef CONFIG_SMP 930 struct plist_node pushable_tasks; 931 struct rb_node pushable_dl_tasks; 932 #endif 933 934 struct mm_struct *mm; 935 struct mm_struct *active_mm; 936 struct address_space *faults_disabled_mapping; 937 938 int exit_state; 939 int exit_code; 940 int exit_signal; 941 /* The signal sent when the parent dies: */ 942 int pdeath_signal; 943 /* JOBCTL_*, siglock protected: */ 944 unsigned long jobctl; 945 946 /* Used for emulating ABI behavior of previous Linux versions: */ 947 unsigned int personality; 948 949 /* Scheduler bits, serialized by scheduler locks: */ 950 unsigned sched_reset_on_fork:1; 951 unsigned sched_contributes_to_load:1; 952 unsigned sched_migrated:1; 953 954 /* Force alignment to the next boundary: */ 955 unsigned :0; 956 957 /* Unserialized, strictly 'current' */ 958 959 /* 960 * This field must not be in the scheduler word above due to wakelist 961 * queueing no longer being serialized by p->on_cpu. However: 962 * 963 * p->XXX = X; ttwu() 964 * schedule() if (p->on_rq && ..) // false 965 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 966 * deactivate_task() ttwu_queue_wakelist()) 967 * p->on_rq = 0; p->sched_remote_wakeup = Y; 968 * 969 * guarantees all stores of 'current' are visible before 970 * ->sched_remote_wakeup gets used, so it can be in this word. 971 */ 972 unsigned sched_remote_wakeup:1; 973 #ifdef CONFIG_RT_MUTEXES 974 unsigned sched_rt_mutex:1; 975 #endif 976 977 /* Bit to tell TOMOYO we're in execve(): */ 978 unsigned in_execve:1; 979 unsigned in_iowait:1; 980 #ifndef TIF_RESTORE_SIGMASK 981 unsigned restore_sigmask:1; 982 #endif 983 #ifdef CONFIG_MEMCG_V1 984 unsigned in_user_fault:1; 985 #endif 986 #ifdef CONFIG_LRU_GEN 987 /* whether the LRU algorithm may apply to this access */ 988 unsigned in_lru_fault:1; 989 #endif 990 #ifdef CONFIG_COMPAT_BRK 991 unsigned brk_randomized:1; 992 #endif 993 #ifdef CONFIG_CGROUPS 994 /* disallow userland-initiated cgroup migration */ 995 unsigned no_cgroup_migration:1; 996 /* task is frozen/stopped (used by the cgroup freezer) */ 997 unsigned frozen:1; 998 #endif 999 #ifdef CONFIG_BLK_CGROUP 1000 unsigned use_memdelay:1; 1001 #endif 1002 #ifdef CONFIG_PSI 1003 /* Stalled due to lack of memory */ 1004 unsigned in_memstall:1; 1005 #endif 1006 #ifdef CONFIG_PAGE_OWNER 1007 /* Used by page_owner=on to detect recursion in page tracking. */ 1008 unsigned in_page_owner:1; 1009 #endif 1010 #ifdef CONFIG_EVENTFD 1011 /* Recursion prevention for eventfd_signal() */ 1012 unsigned in_eventfd:1; 1013 #endif 1014 #ifdef CONFIG_ARCH_HAS_CPU_PASID 1015 unsigned pasid_activated:1; 1016 #endif 1017 #ifdef CONFIG_X86_BUS_LOCK_DETECT 1018 unsigned reported_split_lock:1; 1019 #endif 1020 #ifdef CONFIG_TASK_DELAY_ACCT 1021 /* delay due to memory thrashing */ 1022 unsigned in_thrashing:1; 1023 #endif 1024 #ifdef CONFIG_PREEMPT_RT 1025 struct netdev_xmit net_xmit; 1026 #endif 1027 unsigned long atomic_flags; /* Flags requiring atomic access. */ 1028 1029 struct restart_block restart_block; 1030 1031 pid_t pid; 1032 pid_t tgid; 1033 1034 #ifdef CONFIG_STACKPROTECTOR 1035 /* Canary value for the -fstack-protector GCC feature: */ 1036 unsigned long stack_canary; 1037 #endif 1038 /* 1039 * Pointers to the (original) parent process, youngest child, younger sibling, 1040 * older sibling, respectively. (p->father can be replaced with 1041 * p->real_parent->pid) 1042 */ 1043 1044 /* Real parent process: */ 1045 struct task_struct __rcu *real_parent; 1046 1047 /* Recipient of SIGCHLD, wait4() reports: */ 1048 struct task_struct __rcu *parent; 1049 1050 /* 1051 * Children/sibling form the list of natural children: 1052 */ 1053 struct list_head children; 1054 struct list_head sibling; 1055 struct task_struct *group_leader; 1056 1057 /* 1058 * 'ptraced' is the list of tasks this task is using ptrace() on. 1059 * 1060 * This includes both natural children and PTRACE_ATTACH targets. 1061 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1062 */ 1063 struct list_head ptraced; 1064 struct list_head ptrace_entry; 1065 1066 /* PID/PID hash table linkage. */ 1067 struct pid *thread_pid; 1068 struct hlist_node pid_links[PIDTYPE_MAX]; 1069 struct list_head thread_node; 1070 1071 struct completion *vfork_done; 1072 1073 /* CLONE_CHILD_SETTID: */ 1074 int __user *set_child_tid; 1075 1076 /* CLONE_CHILD_CLEARTID: */ 1077 int __user *clear_child_tid; 1078 1079 /* PF_KTHREAD | PF_IO_WORKER */ 1080 void *worker_private; 1081 1082 u64 utime; 1083 u64 stime; 1084 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1085 u64 utimescaled; 1086 u64 stimescaled; 1087 #endif 1088 u64 gtime; 1089 struct prev_cputime prev_cputime; 1090 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1091 struct vtime vtime; 1092 #endif 1093 1094 #ifdef CONFIG_NO_HZ_FULL 1095 atomic_t tick_dep_mask; 1096 #endif 1097 /* Context switch counts: */ 1098 unsigned long nvcsw; 1099 unsigned long nivcsw; 1100 1101 /* Monotonic time in nsecs: */ 1102 u64 start_time; 1103 1104 /* Boot based time in nsecs: */ 1105 u64 start_boottime; 1106 1107 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1108 unsigned long min_flt; 1109 unsigned long maj_flt; 1110 1111 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1112 struct posix_cputimers posix_cputimers; 1113 1114 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1115 struct posix_cputimers_work posix_cputimers_work; 1116 #endif 1117 1118 /* Process credentials: */ 1119 1120 /* Tracer's credentials at attach: */ 1121 const struct cred __rcu *ptracer_cred; 1122 1123 /* Objective and real subjective task credentials (COW): */ 1124 const struct cred __rcu *real_cred; 1125 1126 /* Effective (overridable) subjective task credentials (COW): */ 1127 const struct cred __rcu *cred; 1128 1129 #ifdef CONFIG_KEYS 1130 /* Cached requested key. */ 1131 struct key *cached_requested_key; 1132 #endif 1133 1134 /* 1135 * executable name, excluding path. 1136 * 1137 * - normally initialized begin_new_exec() 1138 * - set it with set_task_comm() 1139 * - strscpy_pad() to ensure it is always NUL-terminated and 1140 * zero-padded 1141 * - task_lock() to ensure the operation is atomic and the name is 1142 * fully updated. 1143 */ 1144 char comm[TASK_COMM_LEN]; 1145 1146 struct nameidata *nameidata; 1147 1148 #ifdef CONFIG_SYSVIPC 1149 struct sysv_sem sysvsem; 1150 struct sysv_shm sysvshm; 1151 #endif 1152 #ifdef CONFIG_DETECT_HUNG_TASK 1153 unsigned long last_switch_count; 1154 unsigned long last_switch_time; 1155 #endif 1156 /* Filesystem information: */ 1157 struct fs_struct *fs; 1158 1159 /* Open file information: */ 1160 struct files_struct *files; 1161 1162 #ifdef CONFIG_IO_URING 1163 struct io_uring_task *io_uring; 1164 #endif 1165 1166 /* Namespaces: */ 1167 struct nsproxy *nsproxy; 1168 1169 /* Signal handlers: */ 1170 struct signal_struct *signal; 1171 struct sighand_struct __rcu *sighand; 1172 sigset_t blocked; 1173 sigset_t real_blocked; 1174 /* Restored if set_restore_sigmask() was used: */ 1175 sigset_t saved_sigmask; 1176 struct sigpending pending; 1177 unsigned long sas_ss_sp; 1178 size_t sas_ss_size; 1179 unsigned int sas_ss_flags; 1180 1181 struct callback_head *task_works; 1182 1183 #ifdef CONFIG_AUDIT 1184 #ifdef CONFIG_AUDITSYSCALL 1185 struct audit_context *audit_context; 1186 #endif 1187 kuid_t loginuid; 1188 unsigned int sessionid; 1189 #endif 1190 struct seccomp seccomp; 1191 struct syscall_user_dispatch syscall_dispatch; 1192 1193 /* Thread group tracking: */ 1194 u64 parent_exec_id; 1195 u64 self_exec_id; 1196 1197 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1198 spinlock_t alloc_lock; 1199 1200 /* Protection of the PI data structures: */ 1201 raw_spinlock_t pi_lock; 1202 1203 struct wake_q_node wake_q; 1204 1205 #ifdef CONFIG_RT_MUTEXES 1206 /* PI waiters blocked on a rt_mutex held by this task: */ 1207 struct rb_root_cached pi_waiters; 1208 /* Updated under owner's pi_lock and rq lock */ 1209 struct task_struct *pi_top_task; 1210 /* Deadlock detection and priority inheritance handling: */ 1211 struct rt_mutex_waiter *pi_blocked_on; 1212 #endif 1213 1214 #ifdef CONFIG_DEBUG_MUTEXES 1215 /* Mutex deadlock detection: */ 1216 struct mutex_waiter *blocked_on; 1217 #endif 1218 1219 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1220 int non_block_count; 1221 #endif 1222 1223 #ifdef CONFIG_TRACE_IRQFLAGS 1224 struct irqtrace_events irqtrace; 1225 unsigned int hardirq_threaded; 1226 u64 hardirq_chain_key; 1227 int softirqs_enabled; 1228 int softirq_context; 1229 int irq_config; 1230 #endif 1231 #ifdef CONFIG_PREEMPT_RT 1232 int softirq_disable_cnt; 1233 #endif 1234 1235 #ifdef CONFIG_LOCKDEP 1236 # define MAX_LOCK_DEPTH 48UL 1237 u64 curr_chain_key; 1238 int lockdep_depth; 1239 unsigned int lockdep_recursion; 1240 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1241 #endif 1242 1243 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1244 unsigned int in_ubsan; 1245 #endif 1246 1247 /* Journalling filesystem info: */ 1248 void *journal_info; 1249 1250 /* Stacked block device info: */ 1251 struct bio_list *bio_list; 1252 1253 /* Stack plugging: */ 1254 struct blk_plug *plug; 1255 1256 /* VM state: */ 1257 struct reclaim_state *reclaim_state; 1258 1259 struct io_context *io_context; 1260 1261 #ifdef CONFIG_COMPACTION 1262 struct capture_control *capture_control; 1263 #endif 1264 /* Ptrace state: */ 1265 unsigned long ptrace_message; 1266 kernel_siginfo_t *last_siginfo; 1267 1268 struct task_io_accounting ioac; 1269 #ifdef CONFIG_PSI 1270 /* Pressure stall state */ 1271 unsigned int psi_flags; 1272 #endif 1273 #ifdef CONFIG_TASK_XACCT 1274 /* Accumulated RSS usage: */ 1275 u64 acct_rss_mem1; 1276 /* Accumulated virtual memory usage: */ 1277 u64 acct_vm_mem1; 1278 /* stime + utime since last update: */ 1279 u64 acct_timexpd; 1280 #endif 1281 #ifdef CONFIG_CPUSETS 1282 /* Protected by ->alloc_lock: */ 1283 nodemask_t mems_allowed; 1284 /* Sequence number to catch updates: */ 1285 seqcount_spinlock_t mems_allowed_seq; 1286 int cpuset_mem_spread_rotor; 1287 #endif 1288 #ifdef CONFIG_CGROUPS 1289 /* Control Group info protected by css_set_lock: */ 1290 struct css_set __rcu *cgroups; 1291 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1292 struct list_head cg_list; 1293 #endif 1294 #ifdef CONFIG_X86_CPU_RESCTRL 1295 u32 closid; 1296 u32 rmid; 1297 #endif 1298 #ifdef CONFIG_FUTEX 1299 struct robust_list_head __user *robust_list; 1300 #ifdef CONFIG_COMPAT 1301 struct compat_robust_list_head __user *compat_robust_list; 1302 #endif 1303 struct list_head pi_state_list; 1304 struct futex_pi_state *pi_state_cache; 1305 struct mutex futex_exit_mutex; 1306 unsigned int futex_state; 1307 #endif 1308 #ifdef CONFIG_PERF_EVENTS 1309 u8 perf_recursion[PERF_NR_CONTEXTS]; 1310 struct perf_event_context *perf_event_ctxp; 1311 struct mutex perf_event_mutex; 1312 struct list_head perf_event_list; 1313 #endif 1314 #ifdef CONFIG_DEBUG_PREEMPT 1315 unsigned long preempt_disable_ip; 1316 #endif 1317 #ifdef CONFIG_NUMA 1318 /* Protected by alloc_lock: */ 1319 struct mempolicy *mempolicy; 1320 short il_prev; 1321 u8 il_weight; 1322 short pref_node_fork; 1323 #endif 1324 #ifdef CONFIG_NUMA_BALANCING 1325 int numa_scan_seq; 1326 unsigned int numa_scan_period; 1327 unsigned int numa_scan_period_max; 1328 int numa_preferred_nid; 1329 unsigned long numa_migrate_retry; 1330 /* Migration stamp: */ 1331 u64 node_stamp; 1332 u64 last_task_numa_placement; 1333 u64 last_sum_exec_runtime; 1334 struct callback_head numa_work; 1335 1336 /* 1337 * This pointer is only modified for current in syscall and 1338 * pagefault context (and for tasks being destroyed), so it can be read 1339 * from any of the following contexts: 1340 * - RCU read-side critical section 1341 * - current->numa_group from everywhere 1342 * - task's runqueue locked, task not running 1343 */ 1344 struct numa_group __rcu *numa_group; 1345 1346 /* 1347 * numa_faults is an array split into four regions: 1348 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1349 * in this precise order. 1350 * 1351 * faults_memory: Exponential decaying average of faults on a per-node 1352 * basis. Scheduling placement decisions are made based on these 1353 * counts. The values remain static for the duration of a PTE scan. 1354 * faults_cpu: Track the nodes the process was running on when a NUMA 1355 * hinting fault was incurred. 1356 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1357 * during the current scan window. When the scan completes, the counts 1358 * in faults_memory and faults_cpu decay and these values are copied. 1359 */ 1360 unsigned long *numa_faults; 1361 unsigned long total_numa_faults; 1362 1363 /* 1364 * numa_faults_locality tracks if faults recorded during the last 1365 * scan window were remote/local or failed to migrate. The task scan 1366 * period is adapted based on the locality of the faults with different 1367 * weights depending on whether they were shared or private faults 1368 */ 1369 unsigned long numa_faults_locality[3]; 1370 1371 unsigned long numa_pages_migrated; 1372 #endif /* CONFIG_NUMA_BALANCING */ 1373 1374 #ifdef CONFIG_RSEQ 1375 struct rseq __user *rseq; 1376 u32 rseq_len; 1377 u32 rseq_sig; 1378 /* 1379 * RmW on rseq_event_mask must be performed atomically 1380 * with respect to preemption. 1381 */ 1382 unsigned long rseq_event_mask; 1383 #endif 1384 1385 #ifdef CONFIG_SCHED_MM_CID 1386 int mm_cid; /* Current cid in mm */ 1387 int last_mm_cid; /* Most recent cid in mm */ 1388 int migrate_from_cpu; 1389 int mm_cid_active; /* Whether cid bitmap is active */ 1390 struct callback_head cid_work; 1391 #endif 1392 1393 struct tlbflush_unmap_batch tlb_ubc; 1394 1395 /* Cache last used pipe for splice(): */ 1396 struct pipe_inode_info *splice_pipe; 1397 1398 struct page_frag task_frag; 1399 1400 #ifdef CONFIG_TASK_DELAY_ACCT 1401 struct task_delay_info *delays; 1402 #endif 1403 1404 #ifdef CONFIG_FAULT_INJECTION 1405 int make_it_fail; 1406 unsigned int fail_nth; 1407 #endif 1408 /* 1409 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1410 * balance_dirty_pages() for a dirty throttling pause: 1411 */ 1412 int nr_dirtied; 1413 int nr_dirtied_pause; 1414 /* Start of a write-and-pause period: */ 1415 unsigned long dirty_paused_when; 1416 1417 #ifdef CONFIG_LATENCYTOP 1418 int latency_record_count; 1419 struct latency_record latency_record[LT_SAVECOUNT]; 1420 #endif 1421 /* 1422 * Time slack values; these are used to round up poll() and 1423 * select() etc timeout values. These are in nanoseconds. 1424 */ 1425 u64 timer_slack_ns; 1426 u64 default_timer_slack_ns; 1427 1428 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1429 unsigned int kasan_depth; 1430 #endif 1431 1432 #ifdef CONFIG_KCSAN 1433 struct kcsan_ctx kcsan_ctx; 1434 #ifdef CONFIG_TRACE_IRQFLAGS 1435 struct irqtrace_events kcsan_save_irqtrace; 1436 #endif 1437 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1438 int kcsan_stack_depth; 1439 #endif 1440 #endif 1441 1442 #ifdef CONFIG_KMSAN 1443 struct kmsan_ctx kmsan_ctx; 1444 #endif 1445 1446 #if IS_ENABLED(CONFIG_KUNIT) 1447 struct kunit *kunit_test; 1448 #endif 1449 1450 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1451 /* Index of current stored address in ret_stack: */ 1452 int curr_ret_stack; 1453 int curr_ret_depth; 1454 1455 /* Stack of return addresses for return function tracing: */ 1456 unsigned long *ret_stack; 1457 1458 /* Timestamp for last schedule: */ 1459 unsigned long long ftrace_timestamp; 1460 unsigned long long ftrace_sleeptime; 1461 1462 /* 1463 * Number of functions that haven't been traced 1464 * because of depth overrun: 1465 */ 1466 atomic_t trace_overrun; 1467 1468 /* Pause tracing: */ 1469 atomic_t tracing_graph_pause; 1470 #endif 1471 1472 #ifdef CONFIG_TRACING 1473 /* Bitmask and counter of trace recursion: */ 1474 unsigned long trace_recursion; 1475 #endif /* CONFIG_TRACING */ 1476 1477 #ifdef CONFIG_KCOV 1478 /* See kernel/kcov.c for more details. */ 1479 1480 /* Coverage collection mode enabled for this task (0 if disabled): */ 1481 unsigned int kcov_mode; 1482 1483 /* Size of the kcov_area: */ 1484 unsigned int kcov_size; 1485 1486 /* Buffer for coverage collection: */ 1487 void *kcov_area; 1488 1489 /* KCOV descriptor wired with this task or NULL: */ 1490 struct kcov *kcov; 1491 1492 /* KCOV common handle for remote coverage collection: */ 1493 u64 kcov_handle; 1494 1495 /* KCOV sequence number: */ 1496 int kcov_sequence; 1497 1498 /* Collect coverage from softirq context: */ 1499 unsigned int kcov_softirq; 1500 #endif 1501 1502 #ifdef CONFIG_MEMCG_V1 1503 struct mem_cgroup *memcg_in_oom; 1504 #endif 1505 1506 #ifdef CONFIG_MEMCG 1507 /* Number of pages to reclaim on returning to userland: */ 1508 unsigned int memcg_nr_pages_over_high; 1509 1510 /* Used by memcontrol for targeted memcg charge: */ 1511 struct mem_cgroup *active_memcg; 1512 1513 /* Cache for current->cgroups->memcg->objcg lookups: */ 1514 struct obj_cgroup *objcg; 1515 #endif 1516 1517 #ifdef CONFIG_BLK_CGROUP 1518 struct gendisk *throttle_disk; 1519 #endif 1520 1521 #ifdef CONFIG_UPROBES 1522 struct uprobe_task *utask; 1523 #endif 1524 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1525 unsigned int sequential_io; 1526 unsigned int sequential_io_avg; 1527 #endif 1528 struct kmap_ctrl kmap_ctrl; 1529 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1530 unsigned long task_state_change; 1531 # ifdef CONFIG_PREEMPT_RT 1532 unsigned long saved_state_change; 1533 # endif 1534 #endif 1535 struct rcu_head rcu; 1536 refcount_t rcu_users; 1537 int pagefault_disabled; 1538 #ifdef CONFIG_MMU 1539 struct task_struct *oom_reaper_list; 1540 struct timer_list oom_reaper_timer; 1541 #endif 1542 #ifdef CONFIG_VMAP_STACK 1543 struct vm_struct *stack_vm_area; 1544 #endif 1545 #ifdef CONFIG_THREAD_INFO_IN_TASK 1546 /* A live task holds one reference: */ 1547 refcount_t stack_refcount; 1548 #endif 1549 #ifdef CONFIG_LIVEPATCH 1550 int patch_state; 1551 #endif 1552 #ifdef CONFIG_SECURITY 1553 /* Used by LSM modules for access restriction: */ 1554 void *security; 1555 #endif 1556 #ifdef CONFIG_BPF_SYSCALL 1557 /* Used by BPF task local storage */ 1558 struct bpf_local_storage __rcu *bpf_storage; 1559 /* Used for BPF run context */ 1560 struct bpf_run_ctx *bpf_ctx; 1561 #endif 1562 /* Used by BPF for per-TASK xdp storage */ 1563 struct bpf_net_context *bpf_net_context; 1564 1565 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1566 unsigned long lowest_stack; 1567 unsigned long prev_lowest_stack; 1568 #endif 1569 1570 #ifdef CONFIG_X86_MCE 1571 void __user *mce_vaddr; 1572 __u64 mce_kflags; 1573 u64 mce_addr; 1574 __u64 mce_ripv : 1, 1575 mce_whole_page : 1, 1576 __mce_reserved : 62; 1577 struct callback_head mce_kill_me; 1578 int mce_count; 1579 #endif 1580 1581 #ifdef CONFIG_KRETPROBES 1582 struct llist_head kretprobe_instances; 1583 #endif 1584 #ifdef CONFIG_RETHOOK 1585 struct llist_head rethooks; 1586 #endif 1587 1588 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1589 /* 1590 * If L1D flush is supported on mm context switch 1591 * then we use this callback head to queue kill work 1592 * to kill tasks that are not running on SMT disabled 1593 * cores 1594 */ 1595 struct callback_head l1d_flush_kill; 1596 #endif 1597 1598 #ifdef CONFIG_RV 1599 /* 1600 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS. 1601 * If we find justification for more monitors, we can think 1602 * about adding more or developing a dynamic method. So far, 1603 * none of these are justified. 1604 */ 1605 union rv_task_monitor rv[RV_PER_TASK_MONITORS]; 1606 #endif 1607 1608 #ifdef CONFIG_USER_EVENTS 1609 struct user_event_mm *user_event_mm; 1610 #endif 1611 1612 /* 1613 * New fields for task_struct should be added above here, so that 1614 * they are included in the randomized portion of task_struct. 1615 */ 1616 randomized_struct_fields_end 1617 1618 /* CPU-specific state of this task: */ 1619 struct thread_struct thread; 1620 1621 /* 1622 * WARNING: on x86, 'thread_struct' contains a variable-sized 1623 * structure. It *MUST* be at the end of 'task_struct'. 1624 * 1625 * Do not put anything below here! 1626 */ 1627 }; 1628 1629 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1630 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1631 1632 static inline unsigned int __task_state_index(unsigned int tsk_state, 1633 unsigned int tsk_exit_state) 1634 { 1635 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1636 1637 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1638 1639 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1640 state = TASK_REPORT_IDLE; 1641 1642 /* 1643 * We're lying here, but rather than expose a completely new task state 1644 * to userspace, we can make this appear as if the task has gone through 1645 * a regular rt_mutex_lock() call. 1646 * Report frozen tasks as uninterruptible. 1647 */ 1648 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN)) 1649 state = TASK_UNINTERRUPTIBLE; 1650 1651 return fls(state); 1652 } 1653 1654 static inline unsigned int task_state_index(struct task_struct *tsk) 1655 { 1656 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1657 } 1658 1659 static inline char task_index_to_char(unsigned int state) 1660 { 1661 static const char state_char[] = "RSDTtXZPI"; 1662 1663 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1)); 1664 1665 return state_char[state]; 1666 } 1667 1668 static inline char task_state_to_char(struct task_struct *tsk) 1669 { 1670 return task_index_to_char(task_state_index(tsk)); 1671 } 1672 1673 extern struct pid *cad_pid; 1674 1675 /* 1676 * Per process flags 1677 */ 1678 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1679 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1680 #define PF_EXITING 0x00000004 /* Getting shut down */ 1681 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1682 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1683 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1684 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1685 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1686 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1687 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1688 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1689 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1690 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1691 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1692 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1693 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1694 #define PF__HOLE__00010000 0x00010000 1695 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1696 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1697 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1698 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1699 * I am cleaning dirty pages from some other bdi. */ 1700 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1701 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1702 #define PF__HOLE__00800000 0x00800000 1703 #define PF__HOLE__01000000 0x01000000 1704 #define PF__HOLE__02000000 0x02000000 1705 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1706 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1707 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1708 * See memalloc_pin_save() */ 1709 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1710 #define PF__HOLE__40000000 0x40000000 1711 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1712 1713 /* 1714 * Only the _current_ task can read/write to tsk->flags, but other 1715 * tasks can access tsk->flags in readonly mode for example 1716 * with tsk_used_math (like during threaded core dumping). 1717 * There is however an exception to this rule during ptrace 1718 * or during fork: the ptracer task is allowed to write to the 1719 * child->flags of its traced child (same goes for fork, the parent 1720 * can write to the child->flags), because we're guaranteed the 1721 * child is not running and in turn not changing child->flags 1722 * at the same time the parent does it. 1723 */ 1724 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1725 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1726 #define clear_used_math() clear_stopped_child_used_math(current) 1727 #define set_used_math() set_stopped_child_used_math(current) 1728 1729 #define conditional_stopped_child_used_math(condition, child) \ 1730 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1731 1732 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1733 1734 #define copy_to_stopped_child_used_math(child) \ 1735 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1736 1737 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1738 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1739 #define used_math() tsk_used_math(current) 1740 1741 static __always_inline bool is_percpu_thread(void) 1742 { 1743 #ifdef CONFIG_SMP 1744 return (current->flags & PF_NO_SETAFFINITY) && 1745 (current->nr_cpus_allowed == 1); 1746 #else 1747 return true; 1748 #endif 1749 } 1750 1751 /* Per-process atomic flags. */ 1752 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1753 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1754 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1755 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1756 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1757 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1758 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1759 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1760 1761 #define TASK_PFA_TEST(name, func) \ 1762 static inline bool task_##func(struct task_struct *p) \ 1763 { return test_bit(PFA_##name, &p->atomic_flags); } 1764 1765 #define TASK_PFA_SET(name, func) \ 1766 static inline void task_set_##func(struct task_struct *p) \ 1767 { set_bit(PFA_##name, &p->atomic_flags); } 1768 1769 #define TASK_PFA_CLEAR(name, func) \ 1770 static inline void task_clear_##func(struct task_struct *p) \ 1771 { clear_bit(PFA_##name, &p->atomic_flags); } 1772 1773 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1774 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1775 1776 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1777 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1778 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1779 1780 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1781 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1782 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1783 1784 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1785 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1786 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1787 1788 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1789 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1790 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1791 1792 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1793 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1794 1795 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1796 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1797 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1798 1799 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1800 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1801 1802 static inline void 1803 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1804 { 1805 current->flags &= ~flags; 1806 current->flags |= orig_flags & flags; 1807 } 1808 1809 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1810 extern int task_can_attach(struct task_struct *p); 1811 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1812 extern void dl_bw_free(int cpu, u64 dl_bw); 1813 #ifdef CONFIG_SMP 1814 1815 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1816 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1817 1818 /** 1819 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1820 * @p: the task 1821 * @new_mask: CPU affinity mask 1822 * 1823 * Return: zero if successful, or a negative error code 1824 */ 1825 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1826 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1827 extern void release_user_cpus_ptr(struct task_struct *p); 1828 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1829 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1830 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1831 #else 1832 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1833 { 1834 } 1835 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1836 { 1837 /* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */ 1838 if ((*cpumask_bits(new_mask) & 1) == 0) 1839 return -EINVAL; 1840 return 0; 1841 } 1842 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node) 1843 { 1844 if (src->user_cpus_ptr) 1845 return -EINVAL; 1846 return 0; 1847 } 1848 static inline void release_user_cpus_ptr(struct task_struct *p) 1849 { 1850 WARN_ON(p->user_cpus_ptr); 1851 } 1852 1853 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) 1854 { 1855 return 0; 1856 } 1857 #endif 1858 1859 extern int yield_to(struct task_struct *p, bool preempt); 1860 extern void set_user_nice(struct task_struct *p, long nice); 1861 extern int task_prio(const struct task_struct *p); 1862 1863 /** 1864 * task_nice - return the nice value of a given task. 1865 * @p: the task in question. 1866 * 1867 * Return: The nice value [ -20 ... 0 ... 19 ]. 1868 */ 1869 static inline int task_nice(const struct task_struct *p) 1870 { 1871 return PRIO_TO_NICE((p)->static_prio); 1872 } 1873 1874 extern int can_nice(const struct task_struct *p, const int nice); 1875 extern int task_curr(const struct task_struct *p); 1876 extern int idle_cpu(int cpu); 1877 extern int available_idle_cpu(int cpu); 1878 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1879 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1880 extern void sched_set_fifo(struct task_struct *p); 1881 extern void sched_set_fifo_low(struct task_struct *p); 1882 extern void sched_set_normal(struct task_struct *p, int nice); 1883 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1884 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1885 extern struct task_struct *idle_task(int cpu); 1886 1887 /** 1888 * is_idle_task - is the specified task an idle task? 1889 * @p: the task in question. 1890 * 1891 * Return: 1 if @p is an idle task. 0 otherwise. 1892 */ 1893 static __always_inline bool is_idle_task(const struct task_struct *p) 1894 { 1895 return !!(p->flags & PF_IDLE); 1896 } 1897 1898 extern struct task_struct *curr_task(int cpu); 1899 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1900 1901 void yield(void); 1902 1903 union thread_union { 1904 struct task_struct task; 1905 #ifndef CONFIG_THREAD_INFO_IN_TASK 1906 struct thread_info thread_info; 1907 #endif 1908 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1909 }; 1910 1911 #ifndef CONFIG_THREAD_INFO_IN_TASK 1912 extern struct thread_info init_thread_info; 1913 #endif 1914 1915 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1916 1917 #ifdef CONFIG_THREAD_INFO_IN_TASK 1918 # define task_thread_info(task) (&(task)->thread_info) 1919 #else 1920 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1921 #endif 1922 1923 /* 1924 * find a task by one of its numerical ids 1925 * 1926 * find_task_by_pid_ns(): 1927 * finds a task by its pid in the specified namespace 1928 * find_task_by_vpid(): 1929 * finds a task by its virtual pid 1930 * 1931 * see also find_vpid() etc in include/linux/pid.h 1932 */ 1933 1934 extern struct task_struct *find_task_by_vpid(pid_t nr); 1935 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1936 1937 /* 1938 * find a task by its virtual pid and get the task struct 1939 */ 1940 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1941 1942 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1943 extern int wake_up_process(struct task_struct *tsk); 1944 extern void wake_up_new_task(struct task_struct *tsk); 1945 1946 #ifdef CONFIG_SMP 1947 extern void kick_process(struct task_struct *tsk); 1948 #else 1949 static inline void kick_process(struct task_struct *tsk) { } 1950 #endif 1951 1952 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1953 1954 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1955 { 1956 __set_task_comm(tsk, from, false); 1957 } 1958 1959 /* 1960 * - Why not use task_lock()? 1961 * User space can randomly change their names anyway, so locking for readers 1962 * doesn't make sense. For writers, locking is probably necessary, as a race 1963 * condition could lead to long-term mixed results. 1964 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is 1965 * always NUL-terminated and zero-padded. Therefore the race condition between 1966 * reader and writer is not an issue. 1967 * 1968 * - BUILD_BUG_ON() can help prevent the buf from being truncated. 1969 * Since the callers don't perform any return value checks, this safeguard is 1970 * necessary. 1971 */ 1972 #define get_task_comm(buf, tsk) ({ \ 1973 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \ 1974 strscpy_pad(buf, (tsk)->comm); \ 1975 buf; \ 1976 }) 1977 1978 #ifdef CONFIG_SMP 1979 static __always_inline void scheduler_ipi(void) 1980 { 1981 /* 1982 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1983 * TIF_NEED_RESCHED remotely (for the first time) will also send 1984 * this IPI. 1985 */ 1986 preempt_fold_need_resched(); 1987 } 1988 #else 1989 static inline void scheduler_ipi(void) { } 1990 #endif 1991 1992 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 1993 1994 /* 1995 * Set thread flags in other task's structures. 1996 * See asm/thread_info.h for TIF_xxxx flags available: 1997 */ 1998 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1999 { 2000 set_ti_thread_flag(task_thread_info(tsk), flag); 2001 } 2002 2003 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2004 { 2005 clear_ti_thread_flag(task_thread_info(tsk), flag); 2006 } 2007 2008 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 2009 bool value) 2010 { 2011 update_ti_thread_flag(task_thread_info(tsk), flag, value); 2012 } 2013 2014 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2015 { 2016 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2017 } 2018 2019 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2020 { 2021 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2022 } 2023 2024 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2025 { 2026 return test_ti_thread_flag(task_thread_info(tsk), flag); 2027 } 2028 2029 static inline void set_tsk_need_resched(struct task_struct *tsk) 2030 { 2031 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2032 } 2033 2034 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2035 { 2036 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY, 2037 (atomic_long_t *)&task_thread_info(tsk)->flags); 2038 } 2039 2040 static inline int test_tsk_need_resched(struct task_struct *tsk) 2041 { 2042 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2043 } 2044 2045 /* 2046 * cond_resched() and cond_resched_lock(): latency reduction via 2047 * explicit rescheduling in places that are safe. The return 2048 * value indicates whether a reschedule was done in fact. 2049 * cond_resched_lock() will drop the spinlock before scheduling, 2050 */ 2051 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2052 extern int __cond_resched(void); 2053 2054 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2055 2056 void sched_dynamic_klp_enable(void); 2057 void sched_dynamic_klp_disable(void); 2058 2059 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2060 2061 static __always_inline int _cond_resched(void) 2062 { 2063 return static_call_mod(cond_resched)(); 2064 } 2065 2066 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2067 2068 extern int dynamic_cond_resched(void); 2069 2070 static __always_inline int _cond_resched(void) 2071 { 2072 return dynamic_cond_resched(); 2073 } 2074 2075 #else /* !CONFIG_PREEMPTION */ 2076 2077 static inline int _cond_resched(void) 2078 { 2079 klp_sched_try_switch(); 2080 return __cond_resched(); 2081 } 2082 2083 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2084 2085 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2086 2087 static inline int _cond_resched(void) 2088 { 2089 klp_sched_try_switch(); 2090 return 0; 2091 } 2092 2093 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2094 2095 #define cond_resched() ({ \ 2096 __might_resched(__FILE__, __LINE__, 0); \ 2097 _cond_resched(); \ 2098 }) 2099 2100 extern int __cond_resched_lock(spinlock_t *lock); 2101 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2102 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2103 2104 #define MIGHT_RESCHED_RCU_SHIFT 8 2105 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2106 2107 #ifndef CONFIG_PREEMPT_RT 2108 /* 2109 * Non RT kernels have an elevated preempt count due to the held lock, 2110 * but are not allowed to be inside a RCU read side critical section 2111 */ 2112 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2113 #else 2114 /* 2115 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2116 * cond_resched*lock() has to take that into account because it checks for 2117 * preempt_count() and rcu_preempt_depth(). 2118 */ 2119 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2120 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2121 #endif 2122 2123 #define cond_resched_lock(lock) ({ \ 2124 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2125 __cond_resched_lock(lock); \ 2126 }) 2127 2128 #define cond_resched_rwlock_read(lock) ({ \ 2129 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2130 __cond_resched_rwlock_read(lock); \ 2131 }) 2132 2133 #define cond_resched_rwlock_write(lock) ({ \ 2134 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2135 __cond_resched_rwlock_write(lock); \ 2136 }) 2137 2138 static __always_inline bool need_resched(void) 2139 { 2140 return unlikely(tif_need_resched()); 2141 } 2142 2143 /* 2144 * Wrappers for p->thread_info->cpu access. No-op on UP. 2145 */ 2146 #ifdef CONFIG_SMP 2147 2148 static inline unsigned int task_cpu(const struct task_struct *p) 2149 { 2150 return READ_ONCE(task_thread_info(p)->cpu); 2151 } 2152 2153 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2154 2155 #else 2156 2157 static inline unsigned int task_cpu(const struct task_struct *p) 2158 { 2159 return 0; 2160 } 2161 2162 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2163 { 2164 } 2165 2166 #endif /* CONFIG_SMP */ 2167 2168 static inline bool task_is_runnable(struct task_struct *p) 2169 { 2170 return p->on_rq && !p->se.sched_delayed; 2171 } 2172 2173 extern bool sched_task_on_rq(struct task_struct *p); 2174 extern unsigned long get_wchan(struct task_struct *p); 2175 extern struct task_struct *cpu_curr_snapshot(int cpu); 2176 2177 #include <linux/spinlock.h> 2178 2179 /* 2180 * In order to reduce various lock holder preemption latencies provide an 2181 * interface to see if a vCPU is currently running or not. 2182 * 2183 * This allows us to terminate optimistic spin loops and block, analogous to 2184 * the native optimistic spin heuristic of testing if the lock owner task is 2185 * running or not. 2186 */ 2187 #ifndef vcpu_is_preempted 2188 static inline bool vcpu_is_preempted(int cpu) 2189 { 2190 return false; 2191 } 2192 #endif 2193 2194 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2195 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2196 2197 #ifndef TASK_SIZE_OF 2198 #define TASK_SIZE_OF(tsk) TASK_SIZE 2199 #endif 2200 2201 #ifdef CONFIG_SMP 2202 static inline bool owner_on_cpu(struct task_struct *owner) 2203 { 2204 /* 2205 * As lock holder preemption issue, we both skip spinning if 2206 * task is not on cpu or its cpu is preempted 2207 */ 2208 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2209 } 2210 2211 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2212 unsigned long sched_cpu_util(int cpu); 2213 #endif /* CONFIG_SMP */ 2214 2215 #ifdef CONFIG_SCHED_CORE 2216 extern void sched_core_free(struct task_struct *tsk); 2217 extern void sched_core_fork(struct task_struct *p); 2218 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2219 unsigned long uaddr); 2220 extern int sched_core_idle_cpu(int cpu); 2221 #else 2222 static inline void sched_core_free(struct task_struct *tsk) { } 2223 static inline void sched_core_fork(struct task_struct *p) { } 2224 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2225 #endif 2226 2227 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2228 2229 #ifdef CONFIG_MEM_ALLOC_PROFILING 2230 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2231 { 2232 swap(current->alloc_tag, tag); 2233 return tag; 2234 } 2235 2236 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2237 { 2238 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2239 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2240 #endif 2241 current->alloc_tag = old; 2242 } 2243 #else 2244 #define alloc_tag_save(_tag) NULL 2245 #define alloc_tag_restore(_tag, _old) do {} while (0) 2246 #endif 2247 2248 #endif 2249