1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_SCHED_H 3 #define _LINUX_SCHED_H 4 5 /* 6 * Define 'struct task_struct' and provide the main scheduler 7 * APIs (schedule(), wakeup variants, etc.) 8 */ 9 10 #include <uapi/linux/sched.h> 11 12 #include <asm/current.h> 13 #include <asm/processor.h> 14 #include <linux/thread_info.h> 15 #include <linux/preempt.h> 16 #include <linux/cpumask_types.h> 17 18 #include <linux/cache.h> 19 #include <linux/irqflags_types.h> 20 #include <linux/smp_types.h> 21 #include <linux/pid_types.h> 22 #include <linux/sem_types.h> 23 #include <linux/shm.h> 24 #include <linux/kmsan_types.h> 25 #include <linux/mutex_types.h> 26 #include <linux/plist_types.h> 27 #include <linux/hrtimer_types.h> 28 #include <linux/timer_types.h> 29 #include <linux/seccomp_types.h> 30 #include <linux/nodemask_types.h> 31 #include <linux/refcount_types.h> 32 #include <linux/resource.h> 33 #include <linux/latencytop.h> 34 #include <linux/sched/prio.h> 35 #include <linux/sched/types.h> 36 #include <linux/signal_types.h> 37 #include <linux/spinlock.h> 38 #include <linux/syscall_user_dispatch_types.h> 39 #include <linux/mm_types_task.h> 40 #include <linux/netdevice_xmit.h> 41 #include <linux/task_io_accounting.h> 42 #include <linux/posix-timers_types.h> 43 #include <linux/restart_block.h> 44 #include <uapi/linux/rseq.h> 45 #include <linux/seqlock_types.h> 46 #include <linux/kcsan.h> 47 #include <linux/rv.h> 48 #include <linux/uidgid_types.h> 49 #include <linux/tracepoint-defs.h> 50 #include <asm/kmap_size.h> 51 52 /* task_struct member predeclarations (sorted alphabetically): */ 53 struct audit_context; 54 struct bio_list; 55 struct blk_plug; 56 struct bpf_local_storage; 57 struct bpf_run_ctx; 58 struct bpf_net_context; 59 struct capture_control; 60 struct cfs_rq; 61 struct fs_struct; 62 struct futex_pi_state; 63 struct io_context; 64 struct io_uring_task; 65 struct mempolicy; 66 struct nameidata; 67 struct nsproxy; 68 struct perf_event_context; 69 struct perf_ctx_data; 70 struct pid_namespace; 71 struct pipe_inode_info; 72 struct rcu_node; 73 struct reclaim_state; 74 struct robust_list_head; 75 struct root_domain; 76 struct rq; 77 struct sched_attr; 78 struct sched_dl_entity; 79 struct seq_file; 80 struct sighand_struct; 81 struct signal_struct; 82 struct task_delay_info; 83 struct task_group; 84 struct task_struct; 85 struct user_event_mm; 86 87 #include <linux/sched/ext.h> 88 89 /* 90 * Task state bitmask. NOTE! These bits are also 91 * encoded in fs/proc/array.c: get_task_state(). 92 * 93 * We have two separate sets of flags: task->__state 94 * is about runnability, while task->exit_state are 95 * about the task exiting. Confusing, but this way 96 * modifying one set can't modify the other one by 97 * mistake. 98 */ 99 100 /* Used in tsk->__state: */ 101 #define TASK_RUNNING 0x00000000 102 #define TASK_INTERRUPTIBLE 0x00000001 103 #define TASK_UNINTERRUPTIBLE 0x00000002 104 #define __TASK_STOPPED 0x00000004 105 #define __TASK_TRACED 0x00000008 106 /* Used in tsk->exit_state: */ 107 #define EXIT_DEAD 0x00000010 108 #define EXIT_ZOMBIE 0x00000020 109 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 110 /* Used in tsk->__state again: */ 111 #define TASK_PARKED 0x00000040 112 #define TASK_DEAD 0x00000080 113 #define TASK_WAKEKILL 0x00000100 114 #define TASK_WAKING 0x00000200 115 #define TASK_NOLOAD 0x00000400 116 #define TASK_NEW 0x00000800 117 #define TASK_RTLOCK_WAIT 0x00001000 118 #define TASK_FREEZABLE 0x00002000 119 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP)) 120 #define TASK_FROZEN 0x00008000 121 #define TASK_STATE_MAX 0x00010000 122 123 #define TASK_ANY (TASK_STATE_MAX-1) 124 125 /* 126 * DO NOT ADD ANY NEW USERS ! 127 */ 128 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE) 129 130 /* Convenience macros for the sake of set_current_state: */ 131 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 132 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 133 #define TASK_TRACED __TASK_TRACED 134 135 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 136 137 /* Convenience macros for the sake of wake_up(): */ 138 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 139 140 /* get_task_state(): */ 141 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 142 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 143 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 144 TASK_PARKED) 145 146 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 147 148 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 149 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 150 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 151 152 /* 153 * Special states are those that do not use the normal wait-loop pattern. See 154 * the comment with set_special_state(). 155 */ 156 #define is_special_task_state(state) \ 157 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \ 158 TASK_DEAD | TASK_FROZEN)) 159 160 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 161 # define debug_normal_state_change(state_value) \ 162 do { \ 163 WARN_ON_ONCE(is_special_task_state(state_value)); \ 164 current->task_state_change = _THIS_IP_; \ 165 } while (0) 166 167 # define debug_special_state_change(state_value) \ 168 do { \ 169 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 170 current->task_state_change = _THIS_IP_; \ 171 } while (0) 172 173 # define debug_rtlock_wait_set_state() \ 174 do { \ 175 current->saved_state_change = current->task_state_change;\ 176 current->task_state_change = _THIS_IP_; \ 177 } while (0) 178 179 # define debug_rtlock_wait_restore_state() \ 180 do { \ 181 current->task_state_change = current->saved_state_change;\ 182 } while (0) 183 184 #else 185 # define debug_normal_state_change(cond) do { } while (0) 186 # define debug_special_state_change(cond) do { } while (0) 187 # define debug_rtlock_wait_set_state() do { } while (0) 188 # define debug_rtlock_wait_restore_state() do { } while (0) 189 #endif 190 191 #define trace_set_current_state(state_value) \ 192 do { \ 193 if (tracepoint_enabled(sched_set_state_tp)) \ 194 __trace_set_current_state(state_value); \ 195 } while (0) 196 197 /* 198 * set_current_state() includes a barrier so that the write of current->__state 199 * is correctly serialised wrt the caller's subsequent test of whether to 200 * actually sleep: 201 * 202 * for (;;) { 203 * set_current_state(TASK_UNINTERRUPTIBLE); 204 * if (CONDITION) 205 * break; 206 * 207 * schedule(); 208 * } 209 * __set_current_state(TASK_RUNNING); 210 * 211 * If the caller does not need such serialisation (because, for instance, the 212 * CONDITION test and condition change and wakeup are under the same lock) then 213 * use __set_current_state(). 214 * 215 * The above is typically ordered against the wakeup, which does: 216 * 217 * CONDITION = 1; 218 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 219 * 220 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 221 * accessing p->__state. 222 * 223 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is, 224 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 225 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 226 * 227 * However, with slightly different timing the wakeup TASK_RUNNING store can 228 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 229 * a problem either because that will result in one extra go around the loop 230 * and our @cond test will save the day. 231 * 232 * Also see the comments of try_to_wake_up(). 233 */ 234 #define __set_current_state(state_value) \ 235 do { \ 236 debug_normal_state_change((state_value)); \ 237 trace_set_current_state(state_value); \ 238 WRITE_ONCE(current->__state, (state_value)); \ 239 } while (0) 240 241 #define set_current_state(state_value) \ 242 do { \ 243 debug_normal_state_change((state_value)); \ 244 trace_set_current_state(state_value); \ 245 smp_store_mb(current->__state, (state_value)); \ 246 } while (0) 247 248 /* 249 * set_special_state() should be used for those states when the blocking task 250 * can not use the regular condition based wait-loop. In that case we must 251 * serialize against wakeups such that any possible in-flight TASK_RUNNING 252 * stores will not collide with our state change. 253 */ 254 #define set_special_state(state_value) \ 255 do { \ 256 unsigned long flags; /* may shadow */ \ 257 \ 258 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 259 debug_special_state_change((state_value)); \ 260 trace_set_current_state(state_value); \ 261 WRITE_ONCE(current->__state, (state_value)); \ 262 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 263 } while (0) 264 265 /* 266 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 267 * 268 * RT's spin/rwlock substitutions are state preserving. The state of the 269 * task when blocking on the lock is saved in task_struct::saved_state and 270 * restored after the lock has been acquired. These operations are 271 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 272 * lock related wakeups while the task is blocked on the lock are 273 * redirected to operate on task_struct::saved_state to ensure that these 274 * are not dropped. On restore task_struct::saved_state is set to 275 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 276 * 277 * The lock operation looks like this: 278 * 279 * current_save_and_set_rtlock_wait_state(); 280 * for (;;) { 281 * if (try_lock()) 282 * break; 283 * raw_spin_unlock_irq(&lock->wait_lock); 284 * schedule_rtlock(); 285 * raw_spin_lock_irq(&lock->wait_lock); 286 * set_current_state(TASK_RTLOCK_WAIT); 287 * } 288 * current_restore_rtlock_saved_state(); 289 */ 290 #define current_save_and_set_rtlock_wait_state() \ 291 do { \ 292 lockdep_assert_irqs_disabled(); \ 293 raw_spin_lock(¤t->pi_lock); \ 294 current->saved_state = current->__state; \ 295 debug_rtlock_wait_set_state(); \ 296 trace_set_current_state(TASK_RTLOCK_WAIT); \ 297 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 298 raw_spin_unlock(¤t->pi_lock); \ 299 } while (0); 300 301 #define current_restore_rtlock_saved_state() \ 302 do { \ 303 lockdep_assert_irqs_disabled(); \ 304 raw_spin_lock(¤t->pi_lock); \ 305 debug_rtlock_wait_restore_state(); \ 306 trace_set_current_state(current->saved_state); \ 307 WRITE_ONCE(current->__state, current->saved_state); \ 308 current->saved_state = TASK_RUNNING; \ 309 raw_spin_unlock(¤t->pi_lock); \ 310 } while (0); 311 312 #define get_current_state() READ_ONCE(current->__state) 313 314 /* 315 * Define the task command name length as enum, then it can be visible to 316 * BPF programs. 317 */ 318 enum { 319 TASK_COMM_LEN = 16, 320 }; 321 322 extern void sched_tick(void); 323 324 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 325 326 extern long schedule_timeout(long timeout); 327 extern long schedule_timeout_interruptible(long timeout); 328 extern long schedule_timeout_killable(long timeout); 329 extern long schedule_timeout_uninterruptible(long timeout); 330 extern long schedule_timeout_idle(long timeout); 331 asmlinkage void schedule(void); 332 extern void schedule_preempt_disabled(void); 333 asmlinkage void preempt_schedule_irq(void); 334 #ifdef CONFIG_PREEMPT_RT 335 extern void schedule_rtlock(void); 336 #endif 337 338 extern int __must_check io_schedule_prepare(void); 339 extern void io_schedule_finish(int token); 340 extern long io_schedule_timeout(long timeout); 341 extern void io_schedule(void); 342 343 /* wrapper functions to trace from this header file */ 344 DECLARE_TRACEPOINT(sched_set_state_tp); 345 extern void __trace_set_current_state(int state_value); 346 DECLARE_TRACEPOINT(sched_set_need_resched_tp); 347 extern void __trace_set_need_resched(struct task_struct *curr, int tif); 348 349 /** 350 * struct prev_cputime - snapshot of system and user cputime 351 * @utime: time spent in user mode 352 * @stime: time spent in system mode 353 * @lock: protects the above two fields 354 * 355 * Stores previous user/system time values such that we can guarantee 356 * monotonicity. 357 */ 358 struct prev_cputime { 359 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 360 u64 utime; 361 u64 stime; 362 raw_spinlock_t lock; 363 #endif 364 }; 365 366 enum vtime_state { 367 /* Task is sleeping or running in a CPU with VTIME inactive: */ 368 VTIME_INACTIVE = 0, 369 /* Task is idle */ 370 VTIME_IDLE, 371 /* Task runs in kernelspace in a CPU with VTIME active: */ 372 VTIME_SYS, 373 /* Task runs in userspace in a CPU with VTIME active: */ 374 VTIME_USER, 375 /* Task runs as guests in a CPU with VTIME active: */ 376 VTIME_GUEST, 377 }; 378 379 struct vtime { 380 seqcount_t seqcount; 381 unsigned long long starttime; 382 enum vtime_state state; 383 unsigned int cpu; 384 u64 utime; 385 u64 stime; 386 u64 gtime; 387 }; 388 389 /* 390 * Utilization clamp constraints. 391 * @UCLAMP_MIN: Minimum utilization 392 * @UCLAMP_MAX: Maximum utilization 393 * @UCLAMP_CNT: Utilization clamp constraints count 394 */ 395 enum uclamp_id { 396 UCLAMP_MIN = 0, 397 UCLAMP_MAX, 398 UCLAMP_CNT 399 }; 400 401 extern struct root_domain def_root_domain; 402 extern struct mutex sched_domains_mutex; 403 extern void sched_domains_mutex_lock(void); 404 extern void sched_domains_mutex_unlock(void); 405 406 struct sched_param { 407 int sched_priority; 408 }; 409 410 struct sched_info { 411 #ifdef CONFIG_SCHED_INFO 412 /* Cumulative counters: */ 413 414 /* # of times we have run on this CPU: */ 415 unsigned long pcount; 416 417 /* Time spent waiting on a runqueue: */ 418 unsigned long long run_delay; 419 420 /* Max time spent waiting on a runqueue: */ 421 unsigned long long max_run_delay; 422 423 /* Min time spent waiting on a runqueue: */ 424 unsigned long long min_run_delay; 425 426 /* Timestamps: */ 427 428 /* When did we last run on a CPU? */ 429 unsigned long long last_arrival; 430 431 /* When were we last queued to run? */ 432 unsigned long long last_queued; 433 434 #endif /* CONFIG_SCHED_INFO */ 435 }; 436 437 /* 438 * Integer metrics need fixed point arithmetic, e.g., sched/fair 439 * has a few: load, load_avg, util_avg, freq, and capacity. 440 * 441 * We define a basic fixed point arithmetic range, and then formalize 442 * all these metrics based on that basic range. 443 */ 444 # define SCHED_FIXEDPOINT_SHIFT 10 445 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 446 447 /* Increase resolution of cpu_capacity calculations */ 448 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 449 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 450 451 struct load_weight { 452 unsigned long weight; 453 u32 inv_weight; 454 }; 455 456 /* 457 * The load/runnable/util_avg accumulates an infinite geometric series 458 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 459 * 460 * [load_avg definition] 461 * 462 * load_avg = runnable% * scale_load_down(load) 463 * 464 * [runnable_avg definition] 465 * 466 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 467 * 468 * [util_avg definition] 469 * 470 * util_avg = running% * SCHED_CAPACITY_SCALE 471 * 472 * where runnable% is the time ratio that a sched_entity is runnable and 473 * running% the time ratio that a sched_entity is running. 474 * 475 * For cfs_rq, they are the aggregated values of all runnable and blocked 476 * sched_entities. 477 * 478 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 479 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 480 * for computing those signals (see update_rq_clock_pelt()) 481 * 482 * N.B., the above ratios (runnable% and running%) themselves are in the 483 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 484 * to as large a range as necessary. This is for example reflected by 485 * util_avg's SCHED_CAPACITY_SCALE. 486 * 487 * [Overflow issue] 488 * 489 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 490 * with the highest load (=88761), always runnable on a single cfs_rq, 491 * and should not overflow as the number already hits PID_MAX_LIMIT. 492 * 493 * For all other cases (including 32-bit kernels), struct load_weight's 494 * weight will overflow first before we do, because: 495 * 496 * Max(load_avg) <= Max(load.weight) 497 * 498 * Then it is the load_weight's responsibility to consider overflow 499 * issues. 500 */ 501 struct sched_avg { 502 u64 last_update_time; 503 u64 load_sum; 504 u64 runnable_sum; 505 u32 util_sum; 506 u32 period_contrib; 507 unsigned long load_avg; 508 unsigned long runnable_avg; 509 unsigned long util_avg; 510 unsigned int util_est; 511 } ____cacheline_aligned; 512 513 /* 514 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 515 * updates. When a task is dequeued, its util_est should not be updated if its 516 * util_avg has not been updated in the meantime. 517 * This information is mapped into the MSB bit of util_est at dequeue time. 518 * Since max value of util_est for a task is 1024 (PELT util_avg for a task) 519 * it is safe to use MSB. 520 */ 521 #define UTIL_EST_WEIGHT_SHIFT 2 522 #define UTIL_AVG_UNCHANGED 0x80000000 523 524 struct sched_statistics { 525 #ifdef CONFIG_SCHEDSTATS 526 u64 wait_start; 527 u64 wait_max; 528 u64 wait_count; 529 u64 wait_sum; 530 u64 iowait_count; 531 u64 iowait_sum; 532 533 u64 sleep_start; 534 u64 sleep_max; 535 s64 sum_sleep_runtime; 536 537 u64 block_start; 538 u64 block_max; 539 s64 sum_block_runtime; 540 541 s64 exec_max; 542 u64 slice_max; 543 544 u64 nr_migrations_cold; 545 u64 nr_failed_migrations_affine; 546 u64 nr_failed_migrations_running; 547 u64 nr_failed_migrations_hot; 548 u64 nr_forced_migrations; 549 550 u64 nr_wakeups; 551 u64 nr_wakeups_sync; 552 u64 nr_wakeups_migrate; 553 u64 nr_wakeups_local; 554 u64 nr_wakeups_remote; 555 u64 nr_wakeups_affine; 556 u64 nr_wakeups_affine_attempts; 557 u64 nr_wakeups_passive; 558 u64 nr_wakeups_idle; 559 560 #ifdef CONFIG_SCHED_CORE 561 u64 core_forceidle_sum; 562 #endif 563 #endif /* CONFIG_SCHEDSTATS */ 564 } ____cacheline_aligned; 565 566 struct sched_entity { 567 /* For load-balancing: */ 568 struct load_weight load; 569 struct rb_node run_node; 570 u64 deadline; 571 u64 min_vruntime; 572 u64 min_slice; 573 574 struct list_head group_node; 575 unsigned char on_rq; 576 unsigned char sched_delayed; 577 unsigned char rel_deadline; 578 unsigned char custom_slice; 579 /* hole */ 580 581 u64 exec_start; 582 u64 sum_exec_runtime; 583 u64 prev_sum_exec_runtime; 584 u64 vruntime; 585 union { 586 /* 587 * When !@on_rq this field is vlag. 588 * When cfs_rq->curr == se (which implies @on_rq) 589 * this field is vprot. See protect_slice(). 590 */ 591 s64 vlag; 592 u64 vprot; 593 }; 594 u64 slice; 595 596 u64 nr_migrations; 597 598 #ifdef CONFIG_FAIR_GROUP_SCHED 599 int depth; 600 struct sched_entity *parent; 601 /* rq on which this entity is (to be) queued: */ 602 struct cfs_rq *cfs_rq; 603 /* rq "owned" by this entity/group: */ 604 struct cfs_rq *my_q; 605 /* cached value of my_q->h_nr_running */ 606 unsigned long runnable_weight; 607 #endif 608 609 /* 610 * Per entity load average tracking. 611 * 612 * Put into separate cache line so it does not 613 * collide with read-mostly values above. 614 */ 615 struct sched_avg avg; 616 }; 617 618 struct sched_rt_entity { 619 struct list_head run_list; 620 unsigned long timeout; 621 unsigned long watchdog_stamp; 622 unsigned int time_slice; 623 unsigned short on_rq; 624 unsigned short on_list; 625 626 struct sched_rt_entity *back; 627 #ifdef CONFIG_RT_GROUP_SCHED 628 struct sched_rt_entity *parent; 629 /* rq on which this entity is (to be) queued: */ 630 struct rt_rq *rt_rq; 631 /* rq "owned" by this entity/group: */ 632 struct rt_rq *my_q; 633 #endif 634 } __randomize_layout; 635 636 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *); 637 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *); 638 639 struct sched_dl_entity { 640 struct rb_node rb_node; 641 642 /* 643 * Original scheduling parameters. Copied here from sched_attr 644 * during sched_setattr(), they will remain the same until 645 * the next sched_setattr(). 646 */ 647 u64 dl_runtime; /* Maximum runtime for each instance */ 648 u64 dl_deadline; /* Relative deadline of each instance */ 649 u64 dl_period; /* Separation of two instances (period) */ 650 u64 dl_bw; /* dl_runtime / dl_period */ 651 u64 dl_density; /* dl_runtime / dl_deadline */ 652 653 /* 654 * Actual scheduling parameters. Initialized with the values above, 655 * they are continuously updated during task execution. Note that 656 * the remaining runtime could be < 0 in case we are in overrun. 657 */ 658 s64 runtime; /* Remaining runtime for this instance */ 659 u64 deadline; /* Absolute deadline for this instance */ 660 unsigned int flags; /* Specifying the scheduler behaviour */ 661 662 /* 663 * Some bool flags: 664 * 665 * @dl_throttled tells if we exhausted the runtime. If so, the 666 * task has to wait for a replenishment to be performed at the 667 * next firing of dl_timer. 668 * 669 * @dl_yielded tells if task gave up the CPU before consuming 670 * all its available runtime during the last job. 671 * 672 * @dl_non_contending tells if the task is inactive while still 673 * contributing to the active utilization. In other words, it 674 * indicates if the inactive timer has been armed and its handler 675 * has not been executed yet. This flag is useful to avoid race 676 * conditions between the inactive timer handler and the wakeup 677 * code. 678 * 679 * @dl_overrun tells if the task asked to be informed about runtime 680 * overruns. 681 * 682 * @dl_server tells if this is a server entity. 683 * 684 * @dl_defer tells if this is a deferred or regular server. For 685 * now only defer server exists. 686 * 687 * @dl_defer_armed tells if the deferrable server is waiting 688 * for the replenishment timer to activate it. 689 * 690 * @dl_server_active tells if the dlserver is active(started). 691 * dlserver is started on first cfs enqueue on an idle runqueue 692 * and is stopped when a dequeue results in 0 cfs tasks on the 693 * runqueue. In other words, dlserver is active only when cpu's 694 * runqueue has atleast one cfs task. 695 * 696 * @dl_defer_running tells if the deferrable server is actually 697 * running, skipping the defer phase. 698 */ 699 unsigned int dl_throttled : 1; 700 unsigned int dl_yielded : 1; 701 unsigned int dl_non_contending : 1; 702 unsigned int dl_overrun : 1; 703 unsigned int dl_server : 1; 704 unsigned int dl_server_active : 1; 705 unsigned int dl_defer : 1; 706 unsigned int dl_defer_armed : 1; 707 unsigned int dl_defer_running : 1; 708 unsigned int dl_server_idle : 1; 709 710 /* 711 * Bandwidth enforcement timer. Each -deadline task has its 712 * own bandwidth to be enforced, thus we need one timer per task. 713 */ 714 struct hrtimer dl_timer; 715 716 /* 717 * Inactive timer, responsible for decreasing the active utilization 718 * at the "0-lag time". When a -deadline task blocks, it contributes 719 * to GRUB's active utilization until the "0-lag time", hence a 720 * timer is needed to decrease the active utilization at the correct 721 * time. 722 */ 723 struct hrtimer inactive_timer; 724 725 /* 726 * Bits for DL-server functionality. Also see the comment near 727 * dl_server_update(). 728 * 729 * @rq the runqueue this server is for 730 * 731 * @server_has_tasks() returns true if @server_pick return a 732 * runnable task. 733 */ 734 struct rq *rq; 735 dl_server_has_tasks_f server_has_tasks; 736 dl_server_pick_f server_pick_task; 737 738 #ifdef CONFIG_RT_MUTEXES 739 /* 740 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 741 * pi_se points to the donor, otherwise points to the dl_se it belongs 742 * to (the original one/itself). 743 */ 744 struct sched_dl_entity *pi_se; 745 #endif 746 }; 747 748 #ifdef CONFIG_UCLAMP_TASK 749 /* Number of utilization clamp buckets (shorter alias) */ 750 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 751 752 /* 753 * Utilization clamp for a scheduling entity 754 * @value: clamp value "assigned" to a se 755 * @bucket_id: bucket index corresponding to the "assigned" value 756 * @active: the se is currently refcounted in a rq's bucket 757 * @user_defined: the requested clamp value comes from user-space 758 * 759 * The bucket_id is the index of the clamp bucket matching the clamp value 760 * which is pre-computed and stored to avoid expensive integer divisions from 761 * the fast path. 762 * 763 * The active bit is set whenever a task has got an "effective" value assigned, 764 * which can be different from the clamp value "requested" from user-space. 765 * This allows to know a task is refcounted in the rq's bucket corresponding 766 * to the "effective" bucket_id. 767 * 768 * The user_defined bit is set whenever a task has got a task-specific clamp 769 * value requested from userspace, i.e. the system defaults apply to this task 770 * just as a restriction. This allows to relax default clamps when a less 771 * restrictive task-specific value has been requested, thus allowing to 772 * implement a "nice" semantic. For example, a task running with a 20% 773 * default boost can still drop its own boosting to 0%. 774 */ 775 struct uclamp_se { 776 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 777 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 778 unsigned int active : 1; 779 unsigned int user_defined : 1; 780 }; 781 #endif /* CONFIG_UCLAMP_TASK */ 782 783 union rcu_special { 784 struct { 785 u8 blocked; 786 u8 need_qs; 787 u8 exp_hint; /* Hint for performance. */ 788 u8 need_mb; /* Readers need smp_mb(). */ 789 } b; /* Bits. */ 790 u32 s; /* Set of bits. */ 791 }; 792 793 enum perf_event_task_context { 794 perf_invalid_context = -1, 795 perf_hw_context = 0, 796 perf_sw_context, 797 perf_nr_task_contexts, 798 }; 799 800 /* 801 * Number of contexts where an event can trigger: 802 * task, softirq, hardirq, nmi. 803 */ 804 #define PERF_NR_CONTEXTS 4 805 806 struct wake_q_node { 807 struct wake_q_node *next; 808 }; 809 810 struct kmap_ctrl { 811 #ifdef CONFIG_KMAP_LOCAL 812 int idx; 813 pte_t pteval[KM_MAX_IDX]; 814 #endif 815 }; 816 817 struct task_struct { 818 #ifdef CONFIG_THREAD_INFO_IN_TASK 819 /* 820 * For reasons of header soup (see current_thread_info()), this 821 * must be the first element of task_struct. 822 */ 823 struct thread_info thread_info; 824 #endif 825 unsigned int __state; 826 827 /* saved state for "spinlock sleepers" */ 828 unsigned int saved_state; 829 830 /* 831 * This begins the randomizable portion of task_struct. Only 832 * scheduling-critical items should be added above here. 833 */ 834 randomized_struct_fields_start 835 836 void *stack; 837 refcount_t usage; 838 /* Per task flags (PF_*), defined further below: */ 839 unsigned int flags; 840 unsigned int ptrace; 841 842 #ifdef CONFIG_MEM_ALLOC_PROFILING 843 struct alloc_tag *alloc_tag; 844 #endif 845 846 int on_cpu; 847 struct __call_single_node wake_entry; 848 unsigned int wakee_flips; 849 unsigned long wakee_flip_decay_ts; 850 struct task_struct *last_wakee; 851 852 /* 853 * recent_used_cpu is initially set as the last CPU used by a task 854 * that wakes affine another task. Waker/wakee relationships can 855 * push tasks around a CPU where each wakeup moves to the next one. 856 * Tracking a recently used CPU allows a quick search for a recently 857 * used CPU that may be idle. 858 */ 859 int recent_used_cpu; 860 int wake_cpu; 861 int on_rq; 862 863 int prio; 864 int static_prio; 865 int normal_prio; 866 unsigned int rt_priority; 867 868 struct sched_entity se; 869 struct sched_rt_entity rt; 870 struct sched_dl_entity dl; 871 struct sched_dl_entity *dl_server; 872 #ifdef CONFIG_SCHED_CLASS_EXT 873 struct sched_ext_entity scx; 874 #endif 875 const struct sched_class *sched_class; 876 877 #ifdef CONFIG_SCHED_CORE 878 struct rb_node core_node; 879 unsigned long core_cookie; 880 unsigned int core_occupation; 881 #endif 882 883 #ifdef CONFIG_CGROUP_SCHED 884 struct task_group *sched_task_group; 885 #endif 886 887 888 #ifdef CONFIG_UCLAMP_TASK 889 /* 890 * Clamp values requested for a scheduling entity. 891 * Must be updated with task_rq_lock() held. 892 */ 893 struct uclamp_se uclamp_req[UCLAMP_CNT]; 894 /* 895 * Effective clamp values used for a scheduling entity. 896 * Must be updated with task_rq_lock() held. 897 */ 898 struct uclamp_se uclamp[UCLAMP_CNT]; 899 #endif 900 901 struct sched_statistics stats; 902 903 #ifdef CONFIG_PREEMPT_NOTIFIERS 904 /* List of struct preempt_notifier: */ 905 struct hlist_head preempt_notifiers; 906 #endif 907 908 #ifdef CONFIG_BLK_DEV_IO_TRACE 909 unsigned int btrace_seq; 910 #endif 911 912 unsigned int policy; 913 unsigned long max_allowed_capacity; 914 int nr_cpus_allowed; 915 const cpumask_t *cpus_ptr; 916 cpumask_t *user_cpus_ptr; 917 cpumask_t cpus_mask; 918 void *migration_pending; 919 unsigned short migration_disabled; 920 unsigned short migration_flags; 921 922 #ifdef CONFIG_PREEMPT_RCU 923 int rcu_read_lock_nesting; 924 union rcu_special rcu_read_unlock_special; 925 struct list_head rcu_node_entry; 926 struct rcu_node *rcu_blocked_node; 927 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 928 929 #ifdef CONFIG_TASKS_RCU 930 unsigned long rcu_tasks_nvcsw; 931 u8 rcu_tasks_holdout; 932 u8 rcu_tasks_idx; 933 int rcu_tasks_idle_cpu; 934 struct list_head rcu_tasks_holdout_list; 935 int rcu_tasks_exit_cpu; 936 struct list_head rcu_tasks_exit_list; 937 #endif /* #ifdef CONFIG_TASKS_RCU */ 938 939 #ifdef CONFIG_TASKS_TRACE_RCU 940 int trc_reader_nesting; 941 int trc_ipi_to_cpu; 942 union rcu_special trc_reader_special; 943 struct list_head trc_holdout_list; 944 struct list_head trc_blkd_node; 945 int trc_blkd_cpu; 946 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 947 948 struct sched_info sched_info; 949 950 struct list_head tasks; 951 struct plist_node pushable_tasks; 952 struct rb_node pushable_dl_tasks; 953 954 struct mm_struct *mm; 955 struct mm_struct *active_mm; 956 struct address_space *faults_disabled_mapping; 957 958 int exit_state; 959 int exit_code; 960 int exit_signal; 961 /* The signal sent when the parent dies: */ 962 int pdeath_signal; 963 /* JOBCTL_*, siglock protected: */ 964 unsigned long jobctl; 965 966 /* Used for emulating ABI behavior of previous Linux versions: */ 967 unsigned int personality; 968 969 /* Scheduler bits, serialized by scheduler locks: */ 970 unsigned sched_reset_on_fork:1; 971 unsigned sched_contributes_to_load:1; 972 unsigned sched_migrated:1; 973 unsigned sched_task_hot:1; 974 975 /* Force alignment to the next boundary: */ 976 unsigned :0; 977 978 /* Unserialized, strictly 'current' */ 979 980 /* 981 * This field must not be in the scheduler word above due to wakelist 982 * queueing no longer being serialized by p->on_cpu. However: 983 * 984 * p->XXX = X; ttwu() 985 * schedule() if (p->on_rq && ..) // false 986 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 987 * deactivate_task() ttwu_queue_wakelist()) 988 * p->on_rq = 0; p->sched_remote_wakeup = Y; 989 * 990 * guarantees all stores of 'current' are visible before 991 * ->sched_remote_wakeup gets used, so it can be in this word. 992 */ 993 unsigned sched_remote_wakeup:1; 994 #ifdef CONFIG_RT_MUTEXES 995 unsigned sched_rt_mutex:1; 996 #endif 997 998 /* Bit to tell TOMOYO we're in execve(): */ 999 unsigned in_execve:1; 1000 unsigned in_iowait:1; 1001 #ifndef TIF_RESTORE_SIGMASK 1002 unsigned restore_sigmask:1; 1003 #endif 1004 #ifdef CONFIG_MEMCG_V1 1005 unsigned in_user_fault:1; 1006 #endif 1007 #ifdef CONFIG_LRU_GEN 1008 /* whether the LRU algorithm may apply to this access */ 1009 unsigned in_lru_fault:1; 1010 #endif 1011 #ifdef CONFIG_COMPAT_BRK 1012 unsigned brk_randomized:1; 1013 #endif 1014 #ifdef CONFIG_CGROUPS 1015 /* disallow userland-initiated cgroup migration */ 1016 unsigned no_cgroup_migration:1; 1017 /* task is frozen/stopped (used by the cgroup freezer) */ 1018 unsigned frozen:1; 1019 #endif 1020 #ifdef CONFIG_BLK_CGROUP 1021 unsigned use_memdelay:1; 1022 #endif 1023 #ifdef CONFIG_PSI 1024 /* Stalled due to lack of memory */ 1025 unsigned in_memstall:1; 1026 #endif 1027 #ifdef CONFIG_PAGE_OWNER 1028 /* Used by page_owner=on to detect recursion in page tracking. */ 1029 unsigned in_page_owner:1; 1030 #endif 1031 #ifdef CONFIG_EVENTFD 1032 /* Recursion prevention for eventfd_signal() */ 1033 unsigned in_eventfd:1; 1034 #endif 1035 #ifdef CONFIG_ARCH_HAS_CPU_PASID 1036 unsigned pasid_activated:1; 1037 #endif 1038 #ifdef CONFIG_X86_BUS_LOCK_DETECT 1039 unsigned reported_split_lock:1; 1040 #endif 1041 #ifdef CONFIG_TASK_DELAY_ACCT 1042 /* delay due to memory thrashing */ 1043 unsigned in_thrashing:1; 1044 #endif 1045 unsigned in_nf_duplicate:1; 1046 #ifdef CONFIG_PREEMPT_RT 1047 struct netdev_xmit net_xmit; 1048 #endif 1049 unsigned long atomic_flags; /* Flags requiring atomic access. */ 1050 1051 struct restart_block restart_block; 1052 1053 pid_t pid; 1054 pid_t tgid; 1055 1056 #ifdef CONFIG_STACKPROTECTOR 1057 /* Canary value for the -fstack-protector GCC feature: */ 1058 unsigned long stack_canary; 1059 #endif 1060 /* 1061 * Pointers to the (original) parent process, youngest child, younger sibling, 1062 * older sibling, respectively. (p->father can be replaced with 1063 * p->real_parent->pid) 1064 */ 1065 1066 /* Real parent process: */ 1067 struct task_struct __rcu *real_parent; 1068 1069 /* Recipient of SIGCHLD, wait4() reports: */ 1070 struct task_struct __rcu *parent; 1071 1072 /* 1073 * Children/sibling form the list of natural children: 1074 */ 1075 struct list_head children; 1076 struct list_head sibling; 1077 struct task_struct *group_leader; 1078 1079 /* 1080 * 'ptraced' is the list of tasks this task is using ptrace() on. 1081 * 1082 * This includes both natural children and PTRACE_ATTACH targets. 1083 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1084 */ 1085 struct list_head ptraced; 1086 struct list_head ptrace_entry; 1087 1088 /* PID/PID hash table linkage. */ 1089 struct pid *thread_pid; 1090 struct hlist_node pid_links[PIDTYPE_MAX]; 1091 struct list_head thread_node; 1092 1093 struct completion *vfork_done; 1094 1095 /* CLONE_CHILD_SETTID: */ 1096 int __user *set_child_tid; 1097 1098 /* CLONE_CHILD_CLEARTID: */ 1099 int __user *clear_child_tid; 1100 1101 /* PF_KTHREAD | PF_IO_WORKER */ 1102 void *worker_private; 1103 1104 u64 utime; 1105 u64 stime; 1106 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1107 u64 utimescaled; 1108 u64 stimescaled; 1109 #endif 1110 u64 gtime; 1111 struct prev_cputime prev_cputime; 1112 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1113 struct vtime vtime; 1114 #endif 1115 1116 #ifdef CONFIG_NO_HZ_FULL 1117 atomic_t tick_dep_mask; 1118 #endif 1119 /* Context switch counts: */ 1120 unsigned long nvcsw; 1121 unsigned long nivcsw; 1122 1123 /* Monotonic time in nsecs: */ 1124 u64 start_time; 1125 1126 /* Boot based time in nsecs: */ 1127 u64 start_boottime; 1128 1129 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1130 unsigned long min_flt; 1131 unsigned long maj_flt; 1132 1133 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1134 struct posix_cputimers posix_cputimers; 1135 1136 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1137 struct posix_cputimers_work posix_cputimers_work; 1138 #endif 1139 1140 /* Process credentials: */ 1141 1142 /* Tracer's credentials at attach: */ 1143 const struct cred __rcu *ptracer_cred; 1144 1145 /* Objective and real subjective task credentials (COW): */ 1146 const struct cred __rcu *real_cred; 1147 1148 /* Effective (overridable) subjective task credentials (COW): */ 1149 const struct cred __rcu *cred; 1150 1151 #ifdef CONFIG_KEYS 1152 /* Cached requested key. */ 1153 struct key *cached_requested_key; 1154 #endif 1155 1156 /* 1157 * executable name, excluding path. 1158 * 1159 * - normally initialized begin_new_exec() 1160 * - set it with set_task_comm() 1161 * - strscpy_pad() to ensure it is always NUL-terminated and 1162 * zero-padded 1163 * - task_lock() to ensure the operation is atomic and the name is 1164 * fully updated. 1165 */ 1166 char comm[TASK_COMM_LEN]; 1167 1168 struct nameidata *nameidata; 1169 1170 #ifdef CONFIG_SYSVIPC 1171 struct sysv_sem sysvsem; 1172 struct sysv_shm sysvshm; 1173 #endif 1174 #ifdef CONFIG_DETECT_HUNG_TASK 1175 unsigned long last_switch_count; 1176 unsigned long last_switch_time; 1177 #endif 1178 /* Filesystem information: */ 1179 struct fs_struct *fs; 1180 1181 /* Open file information: */ 1182 struct files_struct *files; 1183 1184 #ifdef CONFIG_IO_URING 1185 struct io_uring_task *io_uring; 1186 #endif 1187 1188 /* Namespaces: */ 1189 struct nsproxy *nsproxy; 1190 1191 /* Signal handlers: */ 1192 struct signal_struct *signal; 1193 struct sighand_struct __rcu *sighand; 1194 sigset_t blocked; 1195 sigset_t real_blocked; 1196 /* Restored if set_restore_sigmask() was used: */ 1197 sigset_t saved_sigmask; 1198 struct sigpending pending; 1199 unsigned long sas_ss_sp; 1200 size_t sas_ss_size; 1201 unsigned int sas_ss_flags; 1202 1203 struct callback_head *task_works; 1204 1205 #ifdef CONFIG_AUDIT 1206 #ifdef CONFIG_AUDITSYSCALL 1207 struct audit_context *audit_context; 1208 #endif 1209 kuid_t loginuid; 1210 unsigned int sessionid; 1211 #endif 1212 struct seccomp seccomp; 1213 struct syscall_user_dispatch syscall_dispatch; 1214 1215 /* Thread group tracking: */ 1216 u64 parent_exec_id; 1217 u64 self_exec_id; 1218 1219 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1220 spinlock_t alloc_lock; 1221 1222 /* Protection of the PI data structures: */ 1223 raw_spinlock_t pi_lock; 1224 1225 struct wake_q_node wake_q; 1226 1227 #ifdef CONFIG_RT_MUTEXES 1228 /* PI waiters blocked on a rt_mutex held by this task: */ 1229 struct rb_root_cached pi_waiters; 1230 /* Updated under owner's pi_lock and rq lock */ 1231 struct task_struct *pi_top_task; 1232 /* Deadlock detection and priority inheritance handling: */ 1233 struct rt_mutex_waiter *pi_blocked_on; 1234 #endif 1235 1236 struct mutex *blocked_on; /* lock we're blocked on */ 1237 1238 #ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER 1239 /* 1240 * Encoded lock address causing task block (lower 2 bits = type from 1241 * <linux/hung_task.h>). Accessed via hung_task_*() helpers. 1242 */ 1243 unsigned long blocker; 1244 #endif 1245 1246 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1247 int non_block_count; 1248 #endif 1249 1250 #ifdef CONFIG_TRACE_IRQFLAGS 1251 struct irqtrace_events irqtrace; 1252 unsigned int hardirq_threaded; 1253 u64 hardirq_chain_key; 1254 int softirqs_enabled; 1255 int softirq_context; 1256 int irq_config; 1257 #endif 1258 #ifdef CONFIG_PREEMPT_RT 1259 int softirq_disable_cnt; 1260 #endif 1261 1262 #ifdef CONFIG_LOCKDEP 1263 # define MAX_LOCK_DEPTH 48UL 1264 u64 curr_chain_key; 1265 int lockdep_depth; 1266 unsigned int lockdep_recursion; 1267 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1268 #endif 1269 1270 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1271 unsigned int in_ubsan; 1272 #endif 1273 1274 /* Journalling filesystem info: */ 1275 void *journal_info; 1276 1277 /* Stacked block device info: */ 1278 struct bio_list *bio_list; 1279 1280 /* Stack plugging: */ 1281 struct blk_plug *plug; 1282 1283 /* VM state: */ 1284 struct reclaim_state *reclaim_state; 1285 1286 struct io_context *io_context; 1287 1288 #ifdef CONFIG_COMPACTION 1289 struct capture_control *capture_control; 1290 #endif 1291 /* Ptrace state: */ 1292 unsigned long ptrace_message; 1293 kernel_siginfo_t *last_siginfo; 1294 1295 struct task_io_accounting ioac; 1296 #ifdef CONFIG_PSI 1297 /* Pressure stall state */ 1298 unsigned int psi_flags; 1299 #endif 1300 #ifdef CONFIG_TASK_XACCT 1301 /* Accumulated RSS usage: */ 1302 u64 acct_rss_mem1; 1303 /* Accumulated virtual memory usage: */ 1304 u64 acct_vm_mem1; 1305 /* stime + utime since last update: */ 1306 u64 acct_timexpd; 1307 #endif 1308 #ifdef CONFIG_CPUSETS 1309 /* Protected by ->alloc_lock: */ 1310 nodemask_t mems_allowed; 1311 /* Sequence number to catch updates: */ 1312 seqcount_spinlock_t mems_allowed_seq; 1313 int cpuset_mem_spread_rotor; 1314 #endif 1315 #ifdef CONFIG_CGROUPS 1316 /* Control Group info protected by css_set_lock: */ 1317 struct css_set __rcu *cgroups; 1318 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1319 struct list_head cg_list; 1320 #endif 1321 #ifdef CONFIG_X86_CPU_RESCTRL 1322 u32 closid; 1323 u32 rmid; 1324 #endif 1325 #ifdef CONFIG_FUTEX 1326 struct robust_list_head __user *robust_list; 1327 #ifdef CONFIG_COMPAT 1328 struct compat_robust_list_head __user *compat_robust_list; 1329 #endif 1330 struct list_head pi_state_list; 1331 struct futex_pi_state *pi_state_cache; 1332 struct mutex futex_exit_mutex; 1333 unsigned int futex_state; 1334 #endif 1335 #ifdef CONFIG_PERF_EVENTS 1336 u8 perf_recursion[PERF_NR_CONTEXTS]; 1337 struct perf_event_context *perf_event_ctxp; 1338 struct mutex perf_event_mutex; 1339 struct list_head perf_event_list; 1340 struct perf_ctx_data __rcu *perf_ctx_data; 1341 #endif 1342 #ifdef CONFIG_DEBUG_PREEMPT 1343 unsigned long preempt_disable_ip; 1344 #endif 1345 #ifdef CONFIG_NUMA 1346 /* Protected by alloc_lock: */ 1347 struct mempolicy *mempolicy; 1348 short il_prev; 1349 u8 il_weight; 1350 short pref_node_fork; 1351 #endif 1352 #ifdef CONFIG_NUMA_BALANCING 1353 int numa_scan_seq; 1354 unsigned int numa_scan_period; 1355 unsigned int numa_scan_period_max; 1356 int numa_preferred_nid; 1357 unsigned long numa_migrate_retry; 1358 /* Migration stamp: */ 1359 u64 node_stamp; 1360 u64 last_task_numa_placement; 1361 u64 last_sum_exec_runtime; 1362 struct callback_head numa_work; 1363 1364 /* 1365 * This pointer is only modified for current in syscall and 1366 * pagefault context (and for tasks being destroyed), so it can be read 1367 * from any of the following contexts: 1368 * - RCU read-side critical section 1369 * - current->numa_group from everywhere 1370 * - task's runqueue locked, task not running 1371 */ 1372 struct numa_group __rcu *numa_group; 1373 1374 /* 1375 * numa_faults is an array split into four regions: 1376 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1377 * in this precise order. 1378 * 1379 * faults_memory: Exponential decaying average of faults on a per-node 1380 * basis. Scheduling placement decisions are made based on these 1381 * counts. The values remain static for the duration of a PTE scan. 1382 * faults_cpu: Track the nodes the process was running on when a NUMA 1383 * hinting fault was incurred. 1384 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1385 * during the current scan window. When the scan completes, the counts 1386 * in faults_memory and faults_cpu decay and these values are copied. 1387 */ 1388 unsigned long *numa_faults; 1389 unsigned long total_numa_faults; 1390 1391 /* 1392 * numa_faults_locality tracks if faults recorded during the last 1393 * scan window were remote/local or failed to migrate. The task scan 1394 * period is adapted based on the locality of the faults with different 1395 * weights depending on whether they were shared or private faults 1396 */ 1397 unsigned long numa_faults_locality[3]; 1398 1399 unsigned long numa_pages_migrated; 1400 #endif /* CONFIG_NUMA_BALANCING */ 1401 1402 #ifdef CONFIG_RSEQ 1403 struct rseq __user *rseq; 1404 u32 rseq_len; 1405 u32 rseq_sig; 1406 /* 1407 * RmW on rseq_event_mask must be performed atomically 1408 * with respect to preemption. 1409 */ 1410 unsigned long rseq_event_mask; 1411 # ifdef CONFIG_DEBUG_RSEQ 1412 /* 1413 * This is a place holder to save a copy of the rseq fields for 1414 * validation of read-only fields. The struct rseq has a 1415 * variable-length array at the end, so it cannot be used 1416 * directly. Reserve a size large enough for the known fields. 1417 */ 1418 char rseq_fields[sizeof(struct rseq)]; 1419 # endif 1420 #endif 1421 1422 #ifdef CONFIG_SCHED_MM_CID 1423 int mm_cid; /* Current cid in mm */ 1424 int last_mm_cid; /* Most recent cid in mm */ 1425 int migrate_from_cpu; 1426 int mm_cid_active; /* Whether cid bitmap is active */ 1427 struct callback_head cid_work; 1428 #endif 1429 1430 struct tlbflush_unmap_batch tlb_ubc; 1431 1432 /* Cache last used pipe for splice(): */ 1433 struct pipe_inode_info *splice_pipe; 1434 1435 struct page_frag task_frag; 1436 1437 #ifdef CONFIG_TASK_DELAY_ACCT 1438 struct task_delay_info *delays; 1439 #endif 1440 1441 #ifdef CONFIG_FAULT_INJECTION 1442 int make_it_fail; 1443 unsigned int fail_nth; 1444 #endif 1445 /* 1446 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1447 * balance_dirty_pages() for a dirty throttling pause: 1448 */ 1449 int nr_dirtied; 1450 int nr_dirtied_pause; 1451 /* Start of a write-and-pause period: */ 1452 unsigned long dirty_paused_when; 1453 1454 #ifdef CONFIG_LATENCYTOP 1455 int latency_record_count; 1456 struct latency_record latency_record[LT_SAVECOUNT]; 1457 #endif 1458 /* 1459 * Time slack values; these are used to round up poll() and 1460 * select() etc timeout values. These are in nanoseconds. 1461 */ 1462 u64 timer_slack_ns; 1463 u64 default_timer_slack_ns; 1464 1465 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1466 unsigned int kasan_depth; 1467 #endif 1468 1469 #ifdef CONFIG_KCSAN 1470 struct kcsan_ctx kcsan_ctx; 1471 #ifdef CONFIG_TRACE_IRQFLAGS 1472 struct irqtrace_events kcsan_save_irqtrace; 1473 #endif 1474 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1475 int kcsan_stack_depth; 1476 #endif 1477 #endif 1478 1479 #ifdef CONFIG_KMSAN 1480 struct kmsan_ctx kmsan_ctx; 1481 #endif 1482 1483 #if IS_ENABLED(CONFIG_KUNIT) 1484 struct kunit *kunit_test; 1485 #endif 1486 1487 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1488 /* Index of current stored address in ret_stack: */ 1489 int curr_ret_stack; 1490 int curr_ret_depth; 1491 1492 /* Stack of return addresses for return function tracing: */ 1493 unsigned long *ret_stack; 1494 1495 /* Timestamp for last schedule: */ 1496 unsigned long long ftrace_timestamp; 1497 unsigned long long ftrace_sleeptime; 1498 1499 /* 1500 * Number of functions that haven't been traced 1501 * because of depth overrun: 1502 */ 1503 atomic_t trace_overrun; 1504 1505 /* Pause tracing: */ 1506 atomic_t tracing_graph_pause; 1507 #endif 1508 1509 #ifdef CONFIG_TRACING 1510 /* Bitmask and counter of trace recursion: */ 1511 unsigned long trace_recursion; 1512 #endif /* CONFIG_TRACING */ 1513 1514 #ifdef CONFIG_KCOV 1515 /* See kernel/kcov.c for more details. */ 1516 1517 /* Coverage collection mode enabled for this task (0 if disabled): */ 1518 unsigned int kcov_mode; 1519 1520 /* Size of the kcov_area: */ 1521 unsigned int kcov_size; 1522 1523 /* Buffer for coverage collection: */ 1524 void *kcov_area; 1525 1526 /* KCOV descriptor wired with this task or NULL: */ 1527 struct kcov *kcov; 1528 1529 /* KCOV common handle for remote coverage collection: */ 1530 u64 kcov_handle; 1531 1532 /* KCOV sequence number: */ 1533 int kcov_sequence; 1534 1535 /* Collect coverage from softirq context: */ 1536 unsigned int kcov_softirq; 1537 #endif 1538 1539 #ifdef CONFIG_MEMCG_V1 1540 struct mem_cgroup *memcg_in_oom; 1541 #endif 1542 1543 #ifdef CONFIG_MEMCG 1544 /* Number of pages to reclaim on returning to userland: */ 1545 unsigned int memcg_nr_pages_over_high; 1546 1547 /* Used by memcontrol for targeted memcg charge: */ 1548 struct mem_cgroup *active_memcg; 1549 1550 /* Cache for current->cgroups->memcg->objcg lookups: */ 1551 struct obj_cgroup *objcg; 1552 #endif 1553 1554 #ifdef CONFIG_BLK_CGROUP 1555 struct gendisk *throttle_disk; 1556 #endif 1557 1558 #ifdef CONFIG_UPROBES 1559 struct uprobe_task *utask; 1560 #endif 1561 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1562 unsigned int sequential_io; 1563 unsigned int sequential_io_avg; 1564 #endif 1565 struct kmap_ctrl kmap_ctrl; 1566 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1567 unsigned long task_state_change; 1568 # ifdef CONFIG_PREEMPT_RT 1569 unsigned long saved_state_change; 1570 # endif 1571 #endif 1572 struct rcu_head rcu; 1573 refcount_t rcu_users; 1574 int pagefault_disabled; 1575 #ifdef CONFIG_MMU 1576 struct task_struct *oom_reaper_list; 1577 struct timer_list oom_reaper_timer; 1578 #endif 1579 #ifdef CONFIG_VMAP_STACK 1580 struct vm_struct *stack_vm_area; 1581 #endif 1582 #ifdef CONFIG_THREAD_INFO_IN_TASK 1583 /* A live task holds one reference: */ 1584 refcount_t stack_refcount; 1585 #endif 1586 #ifdef CONFIG_LIVEPATCH 1587 int patch_state; 1588 #endif 1589 #ifdef CONFIG_SECURITY 1590 /* Used by LSM modules for access restriction: */ 1591 void *security; 1592 #endif 1593 #ifdef CONFIG_BPF_SYSCALL 1594 /* Used by BPF task local storage */ 1595 struct bpf_local_storage __rcu *bpf_storage; 1596 /* Used for BPF run context */ 1597 struct bpf_run_ctx *bpf_ctx; 1598 #endif 1599 /* Used by BPF for per-TASK xdp storage */ 1600 struct bpf_net_context *bpf_net_context; 1601 1602 #ifdef CONFIG_KSTACK_ERASE 1603 unsigned long lowest_stack; 1604 #endif 1605 #ifdef CONFIG_KSTACK_ERASE_METRICS 1606 unsigned long prev_lowest_stack; 1607 #endif 1608 1609 #ifdef CONFIG_X86_MCE 1610 void __user *mce_vaddr; 1611 __u64 mce_kflags; 1612 u64 mce_addr; 1613 __u64 mce_ripv : 1, 1614 mce_whole_page : 1, 1615 __mce_reserved : 62; 1616 struct callback_head mce_kill_me; 1617 int mce_count; 1618 #endif 1619 1620 #ifdef CONFIG_KRETPROBES 1621 struct llist_head kretprobe_instances; 1622 #endif 1623 #ifdef CONFIG_RETHOOK 1624 struct llist_head rethooks; 1625 #endif 1626 1627 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1628 /* 1629 * If L1D flush is supported on mm context switch 1630 * then we use this callback head to queue kill work 1631 * to kill tasks that are not running on SMT disabled 1632 * cores 1633 */ 1634 struct callback_head l1d_flush_kill; 1635 #endif 1636 1637 #ifdef CONFIG_RV 1638 /* 1639 * Per-task RV monitor, fixed in CONFIG_RV_PER_TASK_MONITORS. 1640 * If memory becomes a concern, we can think about a dynamic method. 1641 */ 1642 union rv_task_monitor rv[CONFIG_RV_PER_TASK_MONITORS]; 1643 #endif 1644 1645 #ifdef CONFIG_USER_EVENTS 1646 struct user_event_mm *user_event_mm; 1647 #endif 1648 1649 /* CPU-specific state of this task: */ 1650 struct thread_struct thread; 1651 1652 /* 1653 * New fields for task_struct should be added above here, so that 1654 * they are included in the randomized portion of task_struct. 1655 */ 1656 randomized_struct_fields_end 1657 } __attribute__ ((aligned (64))); 1658 1659 #ifdef CONFIG_SCHED_PROXY_EXEC 1660 DECLARE_STATIC_KEY_TRUE(__sched_proxy_exec); 1661 static inline bool sched_proxy_exec(void) 1662 { 1663 return static_branch_likely(&__sched_proxy_exec); 1664 } 1665 #else 1666 static inline bool sched_proxy_exec(void) 1667 { 1668 return false; 1669 } 1670 #endif 1671 1672 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1673 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1674 1675 static inline unsigned int __task_state_index(unsigned int tsk_state, 1676 unsigned int tsk_exit_state) 1677 { 1678 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1679 1680 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1681 1682 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1683 state = TASK_REPORT_IDLE; 1684 1685 /* 1686 * We're lying here, but rather than expose a completely new task state 1687 * to userspace, we can make this appear as if the task has gone through 1688 * a regular rt_mutex_lock() call. 1689 * Report frozen tasks as uninterruptible. 1690 */ 1691 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN)) 1692 state = TASK_UNINTERRUPTIBLE; 1693 1694 return fls(state); 1695 } 1696 1697 static inline unsigned int task_state_index(struct task_struct *tsk) 1698 { 1699 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1700 } 1701 1702 static inline char task_index_to_char(unsigned int state) 1703 { 1704 static const char state_char[] = "RSDTtXZPI"; 1705 1706 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1)); 1707 1708 return state_char[state]; 1709 } 1710 1711 static inline char task_state_to_char(struct task_struct *tsk) 1712 { 1713 return task_index_to_char(task_state_index(tsk)); 1714 } 1715 1716 extern struct pid *cad_pid; 1717 1718 /* 1719 * Per process flags 1720 */ 1721 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1722 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1723 #define PF_EXITING 0x00000004 /* Getting shut down */ 1724 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1725 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1726 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1727 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1728 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1729 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1730 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1731 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1732 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1733 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1734 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1735 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1736 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1737 #define PF_KCOMPACTD 0x00010000 /* I am kcompactd */ 1738 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1739 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1740 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1741 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1742 * I am cleaning dirty pages from some other bdi. */ 1743 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1744 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1745 #define PF__HOLE__00800000 0x00800000 1746 #define PF__HOLE__01000000 0x01000000 1747 #define PF__HOLE__02000000 0x02000000 1748 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1749 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1750 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1751 * See memalloc_pin_save() */ 1752 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1753 #define PF__HOLE__40000000 0x40000000 1754 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1755 1756 /* 1757 * Only the _current_ task can read/write to tsk->flags, but other 1758 * tasks can access tsk->flags in readonly mode for example 1759 * with tsk_used_math (like during threaded core dumping). 1760 * There is however an exception to this rule during ptrace 1761 * or during fork: the ptracer task is allowed to write to the 1762 * child->flags of its traced child (same goes for fork, the parent 1763 * can write to the child->flags), because we're guaranteed the 1764 * child is not running and in turn not changing child->flags 1765 * at the same time the parent does it. 1766 */ 1767 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1768 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1769 #define clear_used_math() clear_stopped_child_used_math(current) 1770 #define set_used_math() set_stopped_child_used_math(current) 1771 1772 #define conditional_stopped_child_used_math(condition, child) \ 1773 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1774 1775 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1776 1777 #define copy_to_stopped_child_used_math(child) \ 1778 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1779 1780 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1781 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1782 #define used_math() tsk_used_math(current) 1783 1784 static __always_inline bool is_percpu_thread(void) 1785 { 1786 return (current->flags & PF_NO_SETAFFINITY) && 1787 (current->nr_cpus_allowed == 1); 1788 } 1789 1790 /* Per-process atomic flags. */ 1791 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1792 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1793 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1794 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1795 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1796 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1797 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1798 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1799 1800 #define TASK_PFA_TEST(name, func) \ 1801 static inline bool task_##func(struct task_struct *p) \ 1802 { return test_bit(PFA_##name, &p->atomic_flags); } 1803 1804 #define TASK_PFA_SET(name, func) \ 1805 static inline void task_set_##func(struct task_struct *p) \ 1806 { set_bit(PFA_##name, &p->atomic_flags); } 1807 1808 #define TASK_PFA_CLEAR(name, func) \ 1809 static inline void task_clear_##func(struct task_struct *p) \ 1810 { clear_bit(PFA_##name, &p->atomic_flags); } 1811 1812 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1813 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1814 1815 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1816 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1817 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1818 1819 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1820 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1821 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1822 1823 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1824 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1825 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1826 1827 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1828 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1829 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1830 1831 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1832 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1833 1834 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1835 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1836 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1837 1838 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1839 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1840 1841 static inline void 1842 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1843 { 1844 current->flags &= ~flags; 1845 current->flags |= orig_flags & flags; 1846 } 1847 1848 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1849 extern int task_can_attach(struct task_struct *p); 1850 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1851 extern void dl_bw_free(int cpu, u64 dl_bw); 1852 1853 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1854 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1855 1856 /** 1857 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1858 * @p: the task 1859 * @new_mask: CPU affinity mask 1860 * 1861 * Return: zero if successful, or a negative error code 1862 */ 1863 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1864 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1865 extern void release_user_cpus_ptr(struct task_struct *p); 1866 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1867 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1868 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1869 1870 extern int yield_to(struct task_struct *p, bool preempt); 1871 extern void set_user_nice(struct task_struct *p, long nice); 1872 extern int task_prio(const struct task_struct *p); 1873 1874 /** 1875 * task_nice - return the nice value of a given task. 1876 * @p: the task in question. 1877 * 1878 * Return: The nice value [ -20 ... 0 ... 19 ]. 1879 */ 1880 static inline int task_nice(const struct task_struct *p) 1881 { 1882 return PRIO_TO_NICE((p)->static_prio); 1883 } 1884 1885 extern int can_nice(const struct task_struct *p, const int nice); 1886 extern int task_curr(const struct task_struct *p); 1887 extern int idle_cpu(int cpu); 1888 extern int available_idle_cpu(int cpu); 1889 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1890 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1891 extern void sched_set_fifo(struct task_struct *p); 1892 extern void sched_set_fifo_low(struct task_struct *p); 1893 extern void sched_set_normal(struct task_struct *p, int nice); 1894 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1895 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1896 extern struct task_struct *idle_task(int cpu); 1897 1898 /** 1899 * is_idle_task - is the specified task an idle task? 1900 * @p: the task in question. 1901 * 1902 * Return: 1 if @p is an idle task. 0 otherwise. 1903 */ 1904 static __always_inline bool is_idle_task(const struct task_struct *p) 1905 { 1906 return !!(p->flags & PF_IDLE); 1907 } 1908 1909 extern struct task_struct *curr_task(int cpu); 1910 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1911 1912 void yield(void); 1913 1914 union thread_union { 1915 struct task_struct task; 1916 #ifndef CONFIG_THREAD_INFO_IN_TASK 1917 struct thread_info thread_info; 1918 #endif 1919 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1920 }; 1921 1922 #ifndef CONFIG_THREAD_INFO_IN_TASK 1923 extern struct thread_info init_thread_info; 1924 #endif 1925 1926 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1927 1928 #ifdef CONFIG_THREAD_INFO_IN_TASK 1929 # define task_thread_info(task) (&(task)->thread_info) 1930 #else 1931 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1932 #endif 1933 1934 /* 1935 * find a task by one of its numerical ids 1936 * 1937 * find_task_by_pid_ns(): 1938 * finds a task by its pid in the specified namespace 1939 * find_task_by_vpid(): 1940 * finds a task by its virtual pid 1941 * 1942 * see also find_vpid() etc in include/linux/pid.h 1943 */ 1944 1945 extern struct task_struct *find_task_by_vpid(pid_t nr); 1946 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1947 1948 /* 1949 * find a task by its virtual pid and get the task struct 1950 */ 1951 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1952 1953 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1954 extern int wake_up_process(struct task_struct *tsk); 1955 extern void wake_up_new_task(struct task_struct *tsk); 1956 1957 extern void kick_process(struct task_struct *tsk); 1958 1959 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1960 #define set_task_comm(tsk, from) ({ \ 1961 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \ 1962 __set_task_comm(tsk, from, false); \ 1963 }) 1964 1965 /* 1966 * - Why not use task_lock()? 1967 * User space can randomly change their names anyway, so locking for readers 1968 * doesn't make sense. For writers, locking is probably necessary, as a race 1969 * condition could lead to long-term mixed results. 1970 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is 1971 * always NUL-terminated and zero-padded. Therefore the race condition between 1972 * reader and writer is not an issue. 1973 * 1974 * - BUILD_BUG_ON() can help prevent the buf from being truncated. 1975 * Since the callers don't perform any return value checks, this safeguard is 1976 * necessary. 1977 */ 1978 #define get_task_comm(buf, tsk) ({ \ 1979 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \ 1980 strscpy_pad(buf, (tsk)->comm); \ 1981 buf; \ 1982 }) 1983 1984 static __always_inline void scheduler_ipi(void) 1985 { 1986 /* 1987 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1988 * TIF_NEED_RESCHED remotely (for the first time) will also send 1989 * this IPI. 1990 */ 1991 preempt_fold_need_resched(); 1992 } 1993 1994 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 1995 1996 /* 1997 * Set thread flags in other task's structures. 1998 * See asm/thread_info.h for TIF_xxxx flags available: 1999 */ 2000 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 2001 { 2002 set_ti_thread_flag(task_thread_info(tsk), flag); 2003 } 2004 2005 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2006 { 2007 clear_ti_thread_flag(task_thread_info(tsk), flag); 2008 } 2009 2010 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 2011 bool value) 2012 { 2013 update_ti_thread_flag(task_thread_info(tsk), flag, value); 2014 } 2015 2016 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2017 { 2018 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2019 } 2020 2021 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2022 { 2023 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2024 } 2025 2026 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2027 { 2028 return test_ti_thread_flag(task_thread_info(tsk), flag); 2029 } 2030 2031 static inline void set_tsk_need_resched(struct task_struct *tsk) 2032 { 2033 if (tracepoint_enabled(sched_set_need_resched_tp) && 2034 !test_tsk_thread_flag(tsk, TIF_NEED_RESCHED)) 2035 __trace_set_need_resched(tsk, TIF_NEED_RESCHED); 2036 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2037 } 2038 2039 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2040 { 2041 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY, 2042 (atomic_long_t *)&task_thread_info(tsk)->flags); 2043 } 2044 2045 static inline int test_tsk_need_resched(struct task_struct *tsk) 2046 { 2047 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2048 } 2049 2050 /* 2051 * cond_resched() and cond_resched_lock(): latency reduction via 2052 * explicit rescheduling in places that are safe. The return 2053 * value indicates whether a reschedule was done in fact. 2054 * cond_resched_lock() will drop the spinlock before scheduling, 2055 */ 2056 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2057 extern int __cond_resched(void); 2058 2059 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2060 2061 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2062 2063 static __always_inline int _cond_resched(void) 2064 { 2065 return static_call_mod(cond_resched)(); 2066 } 2067 2068 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2069 2070 extern int dynamic_cond_resched(void); 2071 2072 static __always_inline int _cond_resched(void) 2073 { 2074 return dynamic_cond_resched(); 2075 } 2076 2077 #else /* !CONFIG_PREEMPTION */ 2078 2079 static inline int _cond_resched(void) 2080 { 2081 return __cond_resched(); 2082 } 2083 2084 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2085 2086 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2087 2088 static inline int _cond_resched(void) 2089 { 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 #ifndef CONFIG_PREEMPT_RT 2139 static inline struct mutex *__get_task_blocked_on(struct task_struct *p) 2140 { 2141 struct mutex *m = p->blocked_on; 2142 2143 if (m) 2144 lockdep_assert_held_once(&m->wait_lock); 2145 return m; 2146 } 2147 2148 static inline void __set_task_blocked_on(struct task_struct *p, struct mutex *m) 2149 { 2150 WARN_ON_ONCE(!m); 2151 /* The task should only be setting itself as blocked */ 2152 WARN_ON_ONCE(p != current); 2153 /* Currently we serialize blocked_on under the mutex::wait_lock */ 2154 lockdep_assert_held_once(&m->wait_lock); 2155 /* 2156 * Check ensure we don't overwrite existing mutex value 2157 * with a different mutex. Note, setting it to the same 2158 * lock repeatedly is ok. 2159 */ 2160 WARN_ON_ONCE(p->blocked_on && p->blocked_on != m); 2161 p->blocked_on = m; 2162 } 2163 2164 static inline void set_task_blocked_on(struct task_struct *p, struct mutex *m) 2165 { 2166 guard(raw_spinlock_irqsave)(&m->wait_lock); 2167 __set_task_blocked_on(p, m); 2168 } 2169 2170 static inline void __clear_task_blocked_on(struct task_struct *p, struct mutex *m) 2171 { 2172 WARN_ON_ONCE(!m); 2173 /* Currently we serialize blocked_on under the mutex::wait_lock */ 2174 lockdep_assert_held_once(&m->wait_lock); 2175 /* 2176 * There may be cases where we re-clear already cleared 2177 * blocked_on relationships, but make sure we are not 2178 * clearing the relationship with a different lock. 2179 */ 2180 WARN_ON_ONCE(m && p->blocked_on && p->blocked_on != m); 2181 p->blocked_on = NULL; 2182 } 2183 2184 static inline void clear_task_blocked_on(struct task_struct *p, struct mutex *m) 2185 { 2186 guard(raw_spinlock_irqsave)(&m->wait_lock); 2187 __clear_task_blocked_on(p, m); 2188 } 2189 #else 2190 static inline void __clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m) 2191 { 2192 } 2193 2194 static inline void clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m) 2195 { 2196 } 2197 #endif /* !CONFIG_PREEMPT_RT */ 2198 2199 static __always_inline bool need_resched(void) 2200 { 2201 return unlikely(tif_need_resched()); 2202 } 2203 2204 /* 2205 * Wrappers for p->thread_info->cpu access. No-op on UP. 2206 */ 2207 #ifdef CONFIG_SMP 2208 2209 static inline unsigned int task_cpu(const struct task_struct *p) 2210 { 2211 return READ_ONCE(task_thread_info(p)->cpu); 2212 } 2213 2214 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2215 2216 #else 2217 2218 static inline unsigned int task_cpu(const struct task_struct *p) 2219 { 2220 return 0; 2221 } 2222 2223 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2224 { 2225 } 2226 2227 #endif /* CONFIG_SMP */ 2228 2229 static inline bool task_is_runnable(struct task_struct *p) 2230 { 2231 return p->on_rq && !p->se.sched_delayed; 2232 } 2233 2234 extern bool sched_task_on_rq(struct task_struct *p); 2235 extern unsigned long get_wchan(struct task_struct *p); 2236 extern struct task_struct *cpu_curr_snapshot(int cpu); 2237 2238 /* 2239 * In order to reduce various lock holder preemption latencies provide an 2240 * interface to see if a vCPU is currently running or not. 2241 * 2242 * This allows us to terminate optimistic spin loops and block, analogous to 2243 * the native optimistic spin heuristic of testing if the lock owner task is 2244 * running or not. 2245 */ 2246 #ifndef vcpu_is_preempted 2247 static inline bool vcpu_is_preempted(int cpu) 2248 { 2249 return false; 2250 } 2251 #endif 2252 2253 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2254 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2255 2256 #ifndef TASK_SIZE_OF 2257 #define TASK_SIZE_OF(tsk) TASK_SIZE 2258 #endif 2259 2260 static inline bool owner_on_cpu(struct task_struct *owner) 2261 { 2262 /* 2263 * As lock holder preemption issue, we both skip spinning if 2264 * task is not on cpu or its cpu is preempted 2265 */ 2266 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2267 } 2268 2269 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2270 unsigned long sched_cpu_util(int cpu); 2271 2272 #ifdef CONFIG_SCHED_CORE 2273 extern void sched_core_free(struct task_struct *tsk); 2274 extern void sched_core_fork(struct task_struct *p); 2275 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2276 unsigned long uaddr); 2277 extern int sched_core_idle_cpu(int cpu); 2278 #else 2279 static inline void sched_core_free(struct task_struct *tsk) { } 2280 static inline void sched_core_fork(struct task_struct *p) { } 2281 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2282 #endif 2283 2284 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2285 2286 #ifdef CONFIG_MEM_ALLOC_PROFILING 2287 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2288 { 2289 swap(current->alloc_tag, tag); 2290 return tag; 2291 } 2292 2293 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2294 { 2295 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2296 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2297 #endif 2298 current->alloc_tag = old; 2299 } 2300 #else 2301 #define alloc_tag_save(_tag) NULL 2302 #define alloc_tag_restore(_tag, _old) do {} while (0) 2303 #endif 2304 2305 #endif 2306