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