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 <linux/rseq_types.h> 45 #include <linux/seqlock_types.h> 46 #include <linux/kcsan.h> 47 #include <linux/rv.h> 48 #include <linux/uidgid_types.h> 49 #include <linux/tracepoint-defs.h> 50 #include <linux/unwind_deferred_types.h> 51 #include <asm/kmap_size.h> 52 #ifndef COMPILE_OFFSETS 53 #include <generated/rq-offsets.h> 54 #endif 55 56 /* task_struct member predeclarations (sorted alphabetically): */ 57 struct audit_context; 58 struct bio_list; 59 struct blk_plug; 60 struct bpf_local_storage; 61 struct bpf_run_ctx; 62 struct bpf_net_context; 63 struct capture_control; 64 struct cfs_rq; 65 struct fs_struct; 66 struct futex_pi_state; 67 struct io_context; 68 struct io_uring_task; 69 struct mempolicy; 70 struct nameidata; 71 struct nsproxy; 72 struct perf_event_context; 73 struct perf_ctx_data; 74 struct pid_namespace; 75 struct pipe_inode_info; 76 struct rcu_node; 77 struct reclaim_state; 78 struct robust_list_head; 79 struct root_domain; 80 struct rq; 81 struct sched_attr; 82 struct sched_dl_entity; 83 struct seq_file; 84 struct sighand_struct; 85 struct signal_struct; 86 struct task_delay_info; 87 struct task_group; 88 struct task_struct; 89 struct user_event_mm; 90 91 #include <linux/sched/ext.h> 92 93 /* 94 * Task state bitmask. NOTE! These bits are also 95 * encoded in fs/proc/array.c: get_task_state(). 96 * 97 * We have two separate sets of flags: task->__state 98 * is about runnability, while task->exit_state are 99 * about the task exiting. Confusing, but this way 100 * modifying one set can't modify the other one by 101 * mistake. 102 */ 103 104 /* Used in tsk->__state: */ 105 #define TASK_RUNNING 0x00000000 106 #define TASK_INTERRUPTIBLE 0x00000001 107 #define TASK_UNINTERRUPTIBLE 0x00000002 108 #define __TASK_STOPPED 0x00000004 109 #define __TASK_TRACED 0x00000008 110 /* Used in tsk->exit_state: */ 111 #define EXIT_DEAD 0x00000010 112 #define EXIT_ZOMBIE 0x00000020 113 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 114 /* Used in tsk->__state again: */ 115 #define TASK_PARKED 0x00000040 116 #define TASK_DEAD 0x00000080 117 #define TASK_WAKEKILL 0x00000100 118 #define TASK_WAKING 0x00000200 119 #define TASK_NOLOAD 0x00000400 120 #define TASK_NEW 0x00000800 121 #define TASK_RTLOCK_WAIT 0x00001000 122 #define TASK_FREEZABLE 0x00002000 123 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP)) 124 #define TASK_FROZEN 0x00008000 125 #define TASK_STATE_MAX 0x00010000 126 127 #define TASK_ANY (TASK_STATE_MAX-1) 128 129 /* 130 * DO NOT ADD ANY NEW USERS ! 131 */ 132 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE) 133 134 /* Convenience macros for the sake of set_current_state: */ 135 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 136 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 137 #define TASK_TRACED __TASK_TRACED 138 139 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 140 141 /* Convenience macros for the sake of wake_up(): */ 142 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 143 144 /* get_task_state(): */ 145 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 146 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 147 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 148 TASK_PARKED) 149 150 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 151 152 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 153 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 154 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 155 156 /* 157 * Special states are those that do not use the normal wait-loop pattern. See 158 * the comment with set_special_state(). 159 */ 160 #define is_special_task_state(state) \ 161 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \ 162 TASK_DEAD | TASK_FROZEN)) 163 164 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 165 # define debug_normal_state_change(state_value) \ 166 do { \ 167 WARN_ON_ONCE(is_special_task_state(state_value)); \ 168 current->task_state_change = _THIS_IP_; \ 169 } while (0) 170 171 # define debug_special_state_change(state_value) \ 172 do { \ 173 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 174 current->task_state_change = _THIS_IP_; \ 175 } while (0) 176 177 # define debug_rtlock_wait_set_state() \ 178 do { \ 179 current->saved_state_change = current->task_state_change;\ 180 current->task_state_change = _THIS_IP_; \ 181 } while (0) 182 183 # define debug_rtlock_wait_restore_state() \ 184 do { \ 185 current->task_state_change = current->saved_state_change;\ 186 } while (0) 187 188 #else 189 # define debug_normal_state_change(cond) do { } while (0) 190 # define debug_special_state_change(cond) do { } while (0) 191 # define debug_rtlock_wait_set_state() do { } while (0) 192 # define debug_rtlock_wait_restore_state() do { } while (0) 193 #endif 194 195 #define trace_set_current_state(state_value) \ 196 do { \ 197 if (tracepoint_enabled(sched_set_state_tp)) \ 198 __trace_set_current_state(state_value); \ 199 } while (0) 200 201 /* 202 * set_current_state() includes a barrier so that the write of current->__state 203 * is correctly serialised wrt the caller's subsequent test of whether to 204 * actually sleep: 205 * 206 * for (;;) { 207 * set_current_state(TASK_UNINTERRUPTIBLE); 208 * if (CONDITION) 209 * break; 210 * 211 * schedule(); 212 * } 213 * __set_current_state(TASK_RUNNING); 214 * 215 * If the caller does not need such serialisation (because, for instance, the 216 * CONDITION test and condition change and wakeup are under the same lock) then 217 * use __set_current_state(). 218 * 219 * The above is typically ordered against the wakeup, which does: 220 * 221 * CONDITION = 1; 222 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 223 * 224 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 225 * accessing p->__state. 226 * 227 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is, 228 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 229 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 230 * 231 * However, with slightly different timing the wakeup TASK_RUNNING store can 232 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 233 * a problem either because that will result in one extra go around the loop 234 * and our @cond test will save the day. 235 * 236 * Also see the comments of try_to_wake_up(). 237 */ 238 #define __set_current_state(state_value) \ 239 do { \ 240 debug_normal_state_change((state_value)); \ 241 trace_set_current_state(state_value); \ 242 WRITE_ONCE(current->__state, (state_value)); \ 243 } while (0) 244 245 #define set_current_state(state_value) \ 246 do { \ 247 debug_normal_state_change((state_value)); \ 248 trace_set_current_state(state_value); \ 249 smp_store_mb(current->__state, (state_value)); \ 250 } while (0) 251 252 /* 253 * set_special_state() should be used for those states when the blocking task 254 * can not use the regular condition based wait-loop. In that case we must 255 * serialize against wakeups such that any possible in-flight TASK_RUNNING 256 * stores will not collide with our state change. 257 */ 258 #define set_special_state(state_value) \ 259 do { \ 260 unsigned long flags; /* may shadow */ \ 261 \ 262 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 263 debug_special_state_change((state_value)); \ 264 trace_set_current_state(state_value); \ 265 WRITE_ONCE(current->__state, (state_value)); \ 266 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 267 } while (0) 268 269 /* 270 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 271 * 272 * RT's spin/rwlock substitutions are state preserving. The state of the 273 * task when blocking on the lock is saved in task_struct::saved_state and 274 * restored after the lock has been acquired. These operations are 275 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 276 * lock related wakeups while the task is blocked on the lock are 277 * redirected to operate on task_struct::saved_state to ensure that these 278 * are not dropped. On restore task_struct::saved_state is set to 279 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 280 * 281 * The lock operation looks like this: 282 * 283 * current_save_and_set_rtlock_wait_state(); 284 * for (;;) { 285 * if (try_lock()) 286 * break; 287 * raw_spin_unlock_irq(&lock->wait_lock); 288 * schedule_rtlock(); 289 * raw_spin_lock_irq(&lock->wait_lock); 290 * set_current_state(TASK_RTLOCK_WAIT); 291 * } 292 * current_restore_rtlock_saved_state(); 293 */ 294 #define current_save_and_set_rtlock_wait_state() \ 295 do { \ 296 lockdep_assert_irqs_disabled(); \ 297 raw_spin_lock(¤t->pi_lock); \ 298 current->saved_state = current->__state; \ 299 debug_rtlock_wait_set_state(); \ 300 trace_set_current_state(TASK_RTLOCK_WAIT); \ 301 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 302 raw_spin_unlock(¤t->pi_lock); \ 303 } while (0); 304 305 #define current_restore_rtlock_saved_state() \ 306 do { \ 307 lockdep_assert_irqs_disabled(); \ 308 raw_spin_lock(¤t->pi_lock); \ 309 debug_rtlock_wait_restore_state(); \ 310 trace_set_current_state(current->saved_state); \ 311 WRITE_ONCE(current->__state, current->saved_state); \ 312 current->saved_state = TASK_RUNNING; \ 313 raw_spin_unlock(¤t->pi_lock); \ 314 } while (0); 315 316 #define get_current_state() READ_ONCE(current->__state) 317 318 /* 319 * Define the task command name length as enum, then it can be visible to 320 * BPF programs. 321 */ 322 enum { 323 TASK_COMM_LEN = 16, 324 }; 325 326 extern void sched_tick(void); 327 328 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 329 330 extern long schedule_timeout(long timeout); 331 extern long schedule_timeout_interruptible(long timeout); 332 extern long schedule_timeout_killable(long timeout); 333 extern long schedule_timeout_uninterruptible(long timeout); 334 extern long schedule_timeout_idle(long timeout); 335 asmlinkage void schedule(void); 336 extern void schedule_preempt_disabled(void); 337 asmlinkage void preempt_schedule_irq(void); 338 #ifdef CONFIG_PREEMPT_RT 339 extern void schedule_rtlock(void); 340 #endif 341 342 extern int __must_check io_schedule_prepare(void); 343 extern void io_schedule_finish(int token); 344 extern long io_schedule_timeout(long timeout); 345 extern void io_schedule(void); 346 347 /* wrapper functions to trace from this header file */ 348 DECLARE_TRACEPOINT(sched_set_state_tp); 349 extern void __trace_set_current_state(int state_value); 350 DECLARE_TRACEPOINT(sched_set_need_resched_tp); 351 extern void __trace_set_need_resched(struct task_struct *curr, int tif); 352 353 /** 354 * struct prev_cputime - snapshot of system and user cputime 355 * @utime: time spent in user mode 356 * @stime: time spent in system mode 357 * @lock: protects the above two fields 358 * 359 * Stores previous user/system time values such that we can guarantee 360 * monotonicity. 361 */ 362 struct prev_cputime { 363 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 364 u64 utime; 365 u64 stime; 366 raw_spinlock_t lock; 367 #endif 368 }; 369 370 enum vtime_state { 371 /* Task is sleeping or running in a CPU with VTIME inactive: */ 372 VTIME_INACTIVE = 0, 373 /* Task is idle */ 374 VTIME_IDLE, 375 /* Task runs in kernelspace in a CPU with VTIME active: */ 376 VTIME_SYS, 377 /* Task runs in userspace in a CPU with VTIME active: */ 378 VTIME_USER, 379 /* Task runs as guests in a CPU with VTIME active: */ 380 VTIME_GUEST, 381 }; 382 383 struct vtime { 384 seqcount_t seqcount; 385 unsigned long long starttime; 386 enum vtime_state state; 387 unsigned int cpu; 388 u64 utime; 389 u64 stime; 390 u64 gtime; 391 }; 392 393 /* 394 * Utilization clamp constraints. 395 * @UCLAMP_MIN: Minimum utilization 396 * @UCLAMP_MAX: Maximum utilization 397 * @UCLAMP_CNT: Utilization clamp constraints count 398 */ 399 enum uclamp_id { 400 UCLAMP_MIN = 0, 401 UCLAMP_MAX, 402 UCLAMP_CNT 403 }; 404 405 extern struct root_domain def_root_domain; 406 extern struct mutex sched_domains_mutex; 407 extern void sched_domains_mutex_lock(void); 408 extern void sched_domains_mutex_unlock(void); 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; 553 554 u64 nr_wakeups; 555 u64 nr_wakeups_sync; 556 u64 nr_wakeups_migrate; 557 u64 nr_wakeups_local; 558 u64 nr_wakeups_remote; 559 u64 nr_wakeups_affine; 560 u64 nr_wakeups_affine_attempts; 561 u64 nr_wakeups_passive; 562 u64 nr_wakeups_idle; 563 564 #ifdef CONFIG_SCHED_CORE 565 u64 core_forceidle_sum; 566 #endif 567 #endif /* CONFIG_SCHEDSTATS */ 568 } ____cacheline_aligned; 569 570 struct sched_entity { 571 /* For load-balancing: */ 572 struct load_weight load; 573 struct rb_node run_node; 574 u64 deadline; 575 u64 min_vruntime; 576 u64 min_slice; 577 578 struct list_head group_node; 579 unsigned char on_rq; 580 unsigned char sched_delayed; 581 unsigned char rel_deadline; 582 unsigned char custom_slice; 583 /* hole */ 584 585 u64 exec_start; 586 u64 sum_exec_runtime; 587 u64 prev_sum_exec_runtime; 588 u64 vruntime; 589 /* Approximated virtual lag: */ 590 s64 vlag; 591 /* 'Protected' deadline, to give out minimum quantums: */ 592 u64 vprot; 593 u64 slice; 594 595 u64 nr_migrations; 596 597 #ifdef CONFIG_FAIR_GROUP_SCHED 598 int depth; 599 struct sched_entity *parent; 600 /* rq on which this entity is (to be) queued: */ 601 struct cfs_rq *cfs_rq; 602 /* rq "owned" by this entity/group: */ 603 struct cfs_rq *my_q; 604 /* cached value of my_q->h_nr_running */ 605 unsigned long runnable_weight; 606 #endif 607 608 /* 609 * Per entity load average tracking. 610 * 611 * Put into separate cache line so it does not 612 * collide with read-mostly values above. 613 */ 614 struct sched_avg avg; 615 }; 616 617 struct sched_rt_entity { 618 struct list_head run_list; 619 unsigned long timeout; 620 unsigned long watchdog_stamp; 621 unsigned int time_slice; 622 unsigned short on_rq; 623 unsigned short on_list; 624 625 struct sched_rt_entity *back; 626 #ifdef CONFIG_RT_GROUP_SCHED 627 struct sched_rt_entity *parent; 628 /* rq on which this entity is (to be) queued: */ 629 struct rt_rq *rt_rq; 630 /* rq "owned" by this entity/group: */ 631 struct rt_rq *my_q; 632 #endif 633 } __randomize_layout; 634 635 struct rq_flags; 636 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *, struct rq_flags *rf); 637 638 struct sched_dl_entity { 639 struct rb_node rb_node; 640 641 /* 642 * Original scheduling parameters. Copied here from sched_attr 643 * during sched_setattr(), they will remain the same until 644 * the next sched_setattr(). 645 */ 646 u64 dl_runtime; /* Maximum runtime for each instance */ 647 u64 dl_deadline; /* Relative deadline of each instance */ 648 u64 dl_period; /* Separation of two instances (period) */ 649 u64 dl_bw; /* dl_runtime / dl_period */ 650 u64 dl_density; /* dl_runtime / dl_deadline */ 651 652 /* 653 * Actual scheduling parameters. Initialized with the values above, 654 * they are continuously updated during task execution. Note that 655 * the remaining runtime could be < 0 in case we are in overrun. 656 */ 657 s64 runtime; /* Remaining runtime for this instance */ 658 u64 deadline; /* Absolute deadline for this instance */ 659 unsigned int flags; /* Specifying the scheduler behaviour */ 660 661 /* 662 * Some bool flags: 663 * 664 * @dl_throttled tells if we exhausted the runtime. If so, the 665 * task has to wait for a replenishment to be performed at the 666 * next firing of dl_timer. 667 * 668 * @dl_yielded tells if task gave up the CPU before consuming 669 * all its available runtime during the last job. 670 * 671 * @dl_non_contending tells if the task is inactive while still 672 * contributing to the active utilization. In other words, it 673 * indicates if the inactive timer has been armed and its handler 674 * has not been executed yet. This flag is useful to avoid race 675 * conditions between the inactive timer handler and the wakeup 676 * code. 677 * 678 * @dl_overrun tells if the task asked to be informed about runtime 679 * overruns. 680 * 681 * @dl_server tells if this is a server entity. 682 * 683 * @dl_server_active tells if the dlserver is active(started). 684 * dlserver is started on first cfs enqueue on an idle runqueue 685 * and is stopped when a dequeue results in 0 cfs tasks on the 686 * runqueue. In other words, dlserver is active only when cpu's 687 * runqueue has atleast one cfs task. 688 * 689 * @dl_defer tells if this is a deferred or regular server. For 690 * now only defer server exists. 691 * 692 * @dl_defer_armed tells if the deferrable server is waiting 693 * for the replenishment timer to activate it. 694 * 695 * @dl_defer_running tells if the deferrable server is actually 696 * running, skipping the defer phase. 697 * 698 * @dl_defer_idle tracks idle state 699 */ 700 unsigned int dl_throttled : 1; 701 unsigned int dl_yielded : 1; 702 unsigned int dl_non_contending : 1; 703 unsigned int dl_overrun : 1; 704 unsigned int dl_server : 1; 705 unsigned int dl_server_active : 1; 706 unsigned int dl_defer : 1; 707 unsigned int dl_defer_armed : 1; 708 unsigned int dl_defer_running : 1; 709 unsigned int dl_defer_idle : 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 struct rq *rq; 733 dl_server_pick_f server_pick_task; 734 735 #ifdef CONFIG_RT_MUTEXES 736 /* 737 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 738 * pi_se points to the donor, otherwise points to the dl_se it belongs 739 * to (the original one/itself). 740 */ 741 struct sched_dl_entity *pi_se; 742 #endif 743 }; 744 745 #ifdef CONFIG_UCLAMP_TASK 746 /* Number of utilization clamp buckets (shorter alias) */ 747 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 748 749 /* 750 * Utilization clamp for a scheduling entity 751 * @value: clamp value "assigned" to a se 752 * @bucket_id: bucket index corresponding to the "assigned" value 753 * @active: the se is currently refcounted in a rq's bucket 754 * @user_defined: the requested clamp value comes from user-space 755 * 756 * The bucket_id is the index of the clamp bucket matching the clamp value 757 * which is pre-computed and stored to avoid expensive integer divisions from 758 * the fast path. 759 * 760 * The active bit is set whenever a task has got an "effective" value assigned, 761 * which can be different from the clamp value "requested" from user-space. 762 * This allows to know a task is refcounted in the rq's bucket corresponding 763 * to the "effective" bucket_id. 764 * 765 * The user_defined bit is set whenever a task has got a task-specific clamp 766 * value requested from userspace, i.e. the system defaults apply to this task 767 * just as a restriction. This allows to relax default clamps when a less 768 * restrictive task-specific value has been requested, thus allowing to 769 * implement a "nice" semantic. For example, a task running with a 20% 770 * default boost can still drop its own boosting to 0%. 771 */ 772 struct uclamp_se { 773 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 774 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 775 unsigned int active : 1; 776 unsigned int user_defined : 1; 777 }; 778 #endif /* CONFIG_UCLAMP_TASK */ 779 780 union rcu_special { 781 struct { 782 u8 blocked; 783 u8 need_qs; 784 u8 exp_hint; /* Hint for performance. */ 785 u8 need_mb; /* Readers need smp_mb(). */ 786 } b; /* Bits. */ 787 u32 s; /* Set of bits. */ 788 }; 789 790 enum perf_event_task_context { 791 perf_invalid_context = -1, 792 perf_hw_context = 0, 793 perf_sw_context, 794 perf_nr_task_contexts, 795 }; 796 797 /* 798 * Number of contexts where an event can trigger: 799 * task, softirq, hardirq, nmi. 800 */ 801 #define PERF_NR_CONTEXTS 4 802 803 struct wake_q_node { 804 struct wake_q_node *next; 805 }; 806 807 struct kmap_ctrl { 808 #ifdef CONFIG_KMAP_LOCAL 809 int idx; 810 pte_t pteval[KM_MAX_IDX]; 811 #endif 812 }; 813 814 struct task_struct { 815 #ifdef CONFIG_THREAD_INFO_IN_TASK 816 /* 817 * For reasons of header soup (see current_thread_info()), this 818 * must be the first element of task_struct. 819 */ 820 struct thread_info thread_info; 821 #endif 822 unsigned int __state; 823 824 /* saved state for "spinlock sleepers" */ 825 unsigned int saved_state; 826 827 /* 828 * This begins the randomizable portion of task_struct. Only 829 * scheduling-critical items should be added above here. 830 */ 831 randomized_struct_fields_start 832 833 void *stack; 834 refcount_t usage; 835 /* Per task flags (PF_*), defined further below: */ 836 unsigned int flags; 837 unsigned int ptrace; 838 839 #ifdef CONFIG_MEM_ALLOC_PROFILING 840 struct alloc_tag *alloc_tag; 841 #endif 842 843 int on_cpu; 844 struct __call_single_node wake_entry; 845 unsigned int wakee_flips; 846 unsigned long wakee_flip_decay_ts; 847 struct task_struct *last_wakee; 848 849 /* 850 * recent_used_cpu is initially set as the last CPU used by a task 851 * that wakes affine another task. Waker/wakee relationships can 852 * push tasks around a CPU where each wakeup moves to the next one. 853 * Tracking a recently used CPU allows a quick search for a recently 854 * used CPU that may be idle. 855 */ 856 int recent_used_cpu; 857 int wake_cpu; 858 int on_rq; 859 860 int prio; 861 int static_prio; 862 int normal_prio; 863 unsigned int rt_priority; 864 865 struct sched_entity se; 866 struct sched_rt_entity rt; 867 struct sched_dl_entity dl; 868 struct sched_dl_entity *dl_server; 869 #ifdef CONFIG_SCHED_CLASS_EXT 870 struct sched_ext_entity scx; 871 #endif 872 const struct sched_class *sched_class; 873 874 #ifdef CONFIG_SCHED_CORE 875 struct rb_node core_node; 876 unsigned long core_cookie; 877 unsigned int core_occupation; 878 #endif 879 880 #ifdef CONFIG_CGROUP_SCHED 881 struct task_group *sched_task_group; 882 #ifdef CONFIG_CFS_BANDWIDTH 883 struct callback_head sched_throttle_work; 884 struct list_head throttle_node; 885 bool throttled; 886 #endif 887 #endif 888 889 890 #ifdef CONFIG_UCLAMP_TASK 891 /* 892 * Clamp values requested for a scheduling entity. 893 * Must be updated with task_rq_lock() held. 894 */ 895 struct uclamp_se uclamp_req[UCLAMP_CNT]; 896 /* 897 * Effective clamp values used for a scheduling entity. 898 * Must be updated with task_rq_lock() held. 899 */ 900 struct uclamp_se uclamp[UCLAMP_CNT]; 901 #endif 902 903 struct sched_statistics stats; 904 905 #ifdef CONFIG_PREEMPT_NOTIFIERS 906 /* List of struct preempt_notifier: */ 907 struct hlist_head preempt_notifiers; 908 #endif 909 910 #ifdef CONFIG_BLK_DEV_IO_TRACE 911 unsigned int btrace_seq; 912 #endif 913 914 unsigned int policy; 915 unsigned long max_allowed_capacity; 916 int nr_cpus_allowed; 917 const cpumask_t *cpus_ptr; 918 cpumask_t *user_cpus_ptr; 919 cpumask_t cpus_mask; 920 void *migration_pending; 921 unsigned short migration_disabled; 922 unsigned short migration_flags; 923 924 #ifdef CONFIG_PREEMPT_RCU 925 int rcu_read_lock_nesting; 926 union rcu_special rcu_read_unlock_special; 927 struct list_head rcu_node_entry; 928 struct rcu_node *rcu_blocked_node; 929 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 930 931 #ifdef CONFIG_TASKS_RCU 932 unsigned long rcu_tasks_nvcsw; 933 u8 rcu_tasks_holdout; 934 u8 rcu_tasks_idx; 935 int rcu_tasks_idle_cpu; 936 struct list_head rcu_tasks_holdout_list; 937 int rcu_tasks_exit_cpu; 938 struct list_head rcu_tasks_exit_list; 939 #endif /* #ifdef CONFIG_TASKS_RCU */ 940 941 #ifdef CONFIG_TASKS_TRACE_RCU 942 int trc_reader_nesting; 943 struct srcu_ctr __percpu *trc_reader_scp; 944 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 945 946 struct sched_info sched_info; 947 948 struct list_head tasks; 949 struct plist_node pushable_tasks; 950 struct rb_node pushable_dl_tasks; 951 952 struct mm_struct *mm; 953 struct mm_struct *active_mm; 954 955 int exit_state; 956 int exit_code; 957 int exit_signal; 958 /* The signal sent when the parent dies: */ 959 int pdeath_signal; 960 /* JOBCTL_*, siglock protected: */ 961 unsigned long jobctl; 962 963 /* Used for emulating ABI behavior of previous Linux versions: */ 964 unsigned int personality; 965 966 /* Scheduler bits, serialized by scheduler locks: */ 967 unsigned sched_reset_on_fork:1; 968 unsigned sched_contributes_to_load:1; 969 unsigned sched_migrated:1; 970 unsigned sched_task_hot:1; 971 972 /* Force alignment to the next boundary: */ 973 unsigned :0; 974 975 /* Unserialized, strictly 'current' */ 976 977 /* 978 * This field must not be in the scheduler word above due to wakelist 979 * queueing no longer being serialized by p->on_cpu. However: 980 * 981 * p->XXX = X; ttwu() 982 * schedule() if (p->on_rq && ..) // false 983 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 984 * deactivate_task() ttwu_queue_wakelist()) 985 * p->on_rq = 0; p->sched_remote_wakeup = Y; 986 * 987 * guarantees all stores of 'current' are visible before 988 * ->sched_remote_wakeup gets used, so it can be in this word. 989 */ 990 unsigned sched_remote_wakeup:1; 991 #ifdef CONFIG_RT_MUTEXES 992 unsigned sched_rt_mutex:1; 993 #endif 994 995 /* Bit to tell TOMOYO we're in execve(): */ 996 unsigned in_execve:1; 997 unsigned in_iowait:1; 998 #ifndef TIF_RESTORE_SIGMASK 999 unsigned restore_sigmask:1; 1000 #endif 1001 #ifdef CONFIG_MEMCG_V1 1002 unsigned in_user_fault:1; 1003 #endif 1004 #ifdef CONFIG_LRU_GEN 1005 /* whether the LRU algorithm may apply to this access */ 1006 unsigned in_lru_fault:1; 1007 #endif 1008 #ifdef CONFIG_COMPAT_BRK 1009 unsigned brk_randomized:1; 1010 #endif 1011 #ifdef CONFIG_CGROUPS 1012 /* disallow userland-initiated cgroup migration */ 1013 unsigned no_cgroup_migration:1; 1014 /* task is frozen/stopped (used by the cgroup freezer) */ 1015 unsigned frozen:1; 1016 #endif 1017 #ifdef CONFIG_BLK_CGROUP 1018 unsigned use_memdelay:1; 1019 #endif 1020 #ifdef CONFIG_PSI 1021 /* Stalled due to lack of memory */ 1022 unsigned in_memstall:1; 1023 #endif 1024 #ifdef CONFIG_PAGE_OWNER 1025 /* Used by page_owner=on to detect recursion in page tracking. */ 1026 unsigned in_page_owner:1; 1027 #endif 1028 #ifdef CONFIG_EVENTFD 1029 /* Recursion prevention for eventfd_signal() */ 1030 unsigned in_eventfd:1; 1031 #endif 1032 #ifdef CONFIG_ARCH_HAS_CPU_PASID 1033 unsigned pasid_activated:1; 1034 #endif 1035 #ifdef CONFIG_X86_BUS_LOCK_DETECT 1036 unsigned reported_split_lock:1; 1037 #endif 1038 #ifdef CONFIG_TASK_DELAY_ACCT 1039 /* delay due to memory thrashing */ 1040 unsigned in_thrashing:1; 1041 #endif 1042 unsigned in_nf_duplicate:1; 1043 #ifdef CONFIG_PREEMPT_RT 1044 struct netdev_xmit net_xmit; 1045 #endif 1046 unsigned long atomic_flags; /* Flags requiring atomic access. */ 1047 1048 struct restart_block restart_block; 1049 1050 pid_t pid; 1051 pid_t tgid; 1052 1053 #ifdef CONFIG_STACKPROTECTOR 1054 /* Canary value for the -fstack-protector GCC feature: */ 1055 unsigned long stack_canary; 1056 #endif 1057 /* 1058 * Pointers to the (original) parent process, youngest child, younger sibling, 1059 * older sibling, respectively. (p->father can be replaced with 1060 * p->real_parent->pid) 1061 */ 1062 1063 /* Real parent process: */ 1064 struct task_struct __rcu *real_parent; 1065 1066 /* Recipient of SIGCHLD, wait4() reports: */ 1067 struct task_struct __rcu *parent; 1068 1069 /* 1070 * Children/sibling form the list of natural children: 1071 */ 1072 struct list_head children; 1073 struct list_head sibling; 1074 struct task_struct *group_leader; 1075 1076 /* 1077 * 'ptraced' is the list of tasks this task is using ptrace() on. 1078 * 1079 * This includes both natural children and PTRACE_ATTACH targets. 1080 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1081 */ 1082 struct list_head ptraced; 1083 struct list_head ptrace_entry; 1084 1085 /* PID/PID hash table linkage. */ 1086 struct pid *thread_pid; 1087 struct hlist_node pid_links[PIDTYPE_MAX]; 1088 struct list_head thread_node; 1089 1090 struct completion *vfork_done; 1091 1092 /* CLONE_CHILD_SETTID: */ 1093 int __user *set_child_tid; 1094 1095 /* CLONE_CHILD_CLEARTID: */ 1096 int __user *clear_child_tid; 1097 1098 /* PF_KTHREAD | PF_IO_WORKER */ 1099 void *worker_private; 1100 1101 u64 utime; 1102 u64 stime; 1103 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1104 u64 utimescaled; 1105 u64 stimescaled; 1106 #endif 1107 u64 gtime; 1108 struct prev_cputime prev_cputime; 1109 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1110 struct vtime vtime; 1111 #endif 1112 1113 #ifdef CONFIG_NO_HZ_FULL 1114 atomic_t tick_dep_mask; 1115 #endif 1116 /* Context switch counts: */ 1117 unsigned long nvcsw; 1118 unsigned long nivcsw; 1119 1120 /* Monotonic time in nsecs: */ 1121 u64 start_time; 1122 1123 /* Boot based time in nsecs: */ 1124 u64 start_boottime; 1125 1126 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1127 unsigned long min_flt; 1128 unsigned long maj_flt; 1129 1130 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1131 struct posix_cputimers posix_cputimers; 1132 1133 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1134 struct posix_cputimers_work posix_cputimers_work; 1135 #endif 1136 1137 /* Process credentials: */ 1138 1139 /* Tracer's credentials at attach: */ 1140 const struct cred __rcu *ptracer_cred; 1141 1142 /* Objective and real subjective task credentials (COW): */ 1143 const struct cred __rcu *real_cred; 1144 1145 /* Effective (overridable) subjective task credentials (COW): */ 1146 const struct cred __rcu *cred; 1147 1148 #ifdef CONFIG_KEYS 1149 /* Cached requested key. */ 1150 struct key *cached_requested_key; 1151 #endif 1152 1153 /* 1154 * executable name, excluding path. 1155 * 1156 * - normally initialized begin_new_exec() 1157 * - set it with set_task_comm() 1158 * - strscpy_pad() to ensure it is always NUL-terminated and 1159 * zero-padded 1160 * - task_lock() to ensure the operation is atomic and the name is 1161 * fully updated. 1162 */ 1163 char comm[TASK_COMM_LEN]; 1164 1165 struct nameidata *nameidata; 1166 1167 #ifdef CONFIG_SYSVIPC 1168 struct sysv_sem sysvsem; 1169 struct sysv_shm sysvshm; 1170 #endif 1171 #ifdef CONFIG_DETECT_HUNG_TASK 1172 unsigned long last_switch_count; 1173 unsigned long last_switch_time; 1174 #endif 1175 /* Filesystem information: */ 1176 struct fs_struct *fs; 1177 1178 /* Open file information: */ 1179 struct files_struct *files; 1180 1181 #ifdef CONFIG_IO_URING 1182 struct io_uring_task *io_uring; 1183 struct io_restriction *io_uring_restrict; 1184 #endif 1185 1186 /* Namespaces: */ 1187 struct nsproxy *nsproxy; 1188 1189 /* Signal handlers: */ 1190 struct signal_struct *signal; 1191 struct sighand_struct __rcu *sighand; 1192 sigset_t blocked; 1193 sigset_t real_blocked; 1194 /* Restored if set_restore_sigmask() was used: */ 1195 sigset_t saved_sigmask; 1196 struct sigpending pending; 1197 unsigned long sas_ss_sp; 1198 size_t sas_ss_size; 1199 unsigned int sas_ss_flags; 1200 1201 struct callback_head *task_works; 1202 1203 #ifdef CONFIG_AUDIT 1204 #ifdef CONFIG_AUDITSYSCALL 1205 struct audit_context *audit_context; 1206 #endif 1207 kuid_t loginuid; 1208 unsigned int sessionid; 1209 #endif 1210 struct seccomp seccomp; 1211 struct syscall_user_dispatch syscall_dispatch; 1212 1213 /* Thread group tracking: */ 1214 u64 parent_exec_id; 1215 u64 self_exec_id; 1216 1217 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1218 spinlock_t alloc_lock; 1219 1220 /* Protection of the PI data structures: */ 1221 raw_spinlock_t pi_lock; 1222 1223 struct wake_q_node wake_q; 1224 1225 #ifdef CONFIG_RT_MUTEXES 1226 /* PI waiters blocked on a rt_mutex held by this task: */ 1227 struct rb_root_cached pi_waiters; 1228 /* Updated under owner's pi_lock and rq lock */ 1229 struct task_struct *pi_top_task; 1230 /* Deadlock detection and priority inheritance handling: */ 1231 struct rt_mutex_waiter *pi_blocked_on; 1232 #endif 1233 1234 struct mutex *blocked_on; /* lock we're blocked on */ 1235 1236 #ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER 1237 /* 1238 * Encoded lock address causing task block (lower 2 bits = type from 1239 * <linux/hung_task.h>). Accessed via hung_task_*() helpers. 1240 */ 1241 unsigned long blocker; 1242 #endif 1243 1244 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1245 int non_block_count; 1246 #endif 1247 1248 #ifdef CONFIG_TRACE_IRQFLAGS 1249 struct irqtrace_events irqtrace; 1250 unsigned int hardirq_threaded; 1251 u64 hardirq_chain_key; 1252 int softirqs_enabled; 1253 int softirq_context; 1254 int irq_config; 1255 #endif 1256 #ifdef CONFIG_PREEMPT_RT 1257 int softirq_disable_cnt; 1258 #endif 1259 1260 #ifdef CONFIG_LOCKDEP 1261 # define MAX_LOCK_DEPTH 48UL 1262 u64 curr_chain_key; 1263 int lockdep_depth; 1264 unsigned int lockdep_recursion; 1265 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1266 #endif 1267 1268 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1269 unsigned int in_ubsan; 1270 #endif 1271 1272 /* Journalling filesystem info: */ 1273 void *journal_info; 1274 1275 /* Stacked block device info: */ 1276 struct bio_list *bio_list; 1277 1278 /* Stack plugging: */ 1279 struct blk_plug *plug; 1280 1281 /* VM state: */ 1282 struct reclaim_state *reclaim_state; 1283 1284 struct io_context *io_context; 1285 1286 #ifdef CONFIG_COMPACTION 1287 struct capture_control *capture_control; 1288 #endif 1289 /* Ptrace state: */ 1290 unsigned long ptrace_message; 1291 kernel_siginfo_t *last_siginfo; 1292 1293 struct task_io_accounting ioac; 1294 #ifdef CONFIG_PSI 1295 /* Pressure stall state */ 1296 unsigned int psi_flags; 1297 #endif 1298 #ifdef CONFIG_TASK_XACCT 1299 /* Accumulated RSS usage: */ 1300 u64 acct_rss_mem1; 1301 /* Accumulated virtual memory usage: */ 1302 u64 acct_vm_mem1; 1303 /* stime + utime since last update: */ 1304 u64 acct_timexpd; 1305 #endif 1306 #ifdef CONFIG_CPUSETS 1307 /* Protected by ->alloc_lock: */ 1308 nodemask_t mems_allowed; 1309 /* Sequence number to catch updates: */ 1310 seqcount_spinlock_t mems_allowed_seq; 1311 int cpuset_mem_spread_rotor; 1312 #endif 1313 #ifdef CONFIG_CGROUPS 1314 /* Control Group info protected by css_set_lock: */ 1315 struct css_set __rcu *cgroups; 1316 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1317 struct list_head cg_list; 1318 #ifdef CONFIG_PREEMPT_RT 1319 struct llist_node cg_dead_lnode; 1320 #endif /* CONFIG_PREEMPT_RT */ 1321 #endif /* CONFIG_CGROUPS */ 1322 #ifdef CONFIG_X86_CPU_RESCTRL 1323 u32 closid; 1324 u32 rmid; 1325 #endif 1326 #ifdef CONFIG_FUTEX 1327 struct robust_list_head __user *robust_list; 1328 #ifdef CONFIG_COMPAT 1329 struct compat_robust_list_head __user *compat_robust_list; 1330 #endif 1331 struct list_head pi_state_list; 1332 struct futex_pi_state *pi_state_cache; 1333 struct mutex futex_exit_mutex; 1334 unsigned int futex_state; 1335 #endif 1336 #ifdef CONFIG_PERF_EVENTS 1337 u8 perf_recursion[PERF_NR_CONTEXTS]; 1338 struct perf_event_context *perf_event_ctxp; 1339 struct mutex perf_event_mutex; 1340 struct list_head perf_event_list; 1341 struct perf_ctx_data __rcu *perf_ctx_data; 1342 #endif 1343 #ifdef CONFIG_DEBUG_PREEMPT 1344 unsigned long preempt_disable_ip; 1345 #endif 1346 #ifdef CONFIG_NUMA 1347 /* Protected by alloc_lock: */ 1348 struct mempolicy *mempolicy; 1349 short il_prev; 1350 u8 il_weight; 1351 short pref_node_fork; 1352 #endif 1353 #ifdef CONFIG_NUMA_BALANCING 1354 int numa_scan_seq; 1355 unsigned int numa_scan_period; 1356 unsigned int numa_scan_period_max; 1357 int numa_preferred_nid; 1358 unsigned long numa_migrate_retry; 1359 /* Migration stamp: */ 1360 u64 node_stamp; 1361 u64 last_task_numa_placement; 1362 u64 last_sum_exec_runtime; 1363 struct callback_head numa_work; 1364 1365 /* 1366 * This pointer is only modified for current in syscall and 1367 * pagefault context (and for tasks being destroyed), so it can be read 1368 * from any of the following contexts: 1369 * - RCU read-side critical section 1370 * - current->numa_group from everywhere 1371 * - task's runqueue locked, task not running 1372 */ 1373 struct numa_group __rcu *numa_group; 1374 1375 /* 1376 * numa_faults is an array split into four regions: 1377 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1378 * in this precise order. 1379 * 1380 * faults_memory: Exponential decaying average of faults on a per-node 1381 * basis. Scheduling placement decisions are made based on these 1382 * counts. The values remain static for the duration of a PTE scan. 1383 * faults_cpu: Track the nodes the process was running on when a NUMA 1384 * hinting fault was incurred. 1385 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1386 * during the current scan window. When the scan completes, the counts 1387 * in faults_memory and faults_cpu decay and these values are copied. 1388 */ 1389 unsigned long *numa_faults; 1390 unsigned long total_numa_faults; 1391 1392 /* 1393 * numa_faults_locality tracks if faults recorded during the last 1394 * scan window were remote/local or failed to migrate. The task scan 1395 * period is adapted based on the locality of the faults with different 1396 * weights depending on whether they were shared or private faults 1397 */ 1398 unsigned long numa_faults_locality[3]; 1399 1400 unsigned long numa_pages_migrated; 1401 #endif /* CONFIG_NUMA_BALANCING */ 1402 1403 struct rseq_data rseq; 1404 struct sched_mm_cid mm_cid; 1405 1406 struct tlbflush_unmap_batch tlb_ubc; 1407 1408 /* Cache last used pipe for splice(): */ 1409 struct pipe_inode_info *splice_pipe; 1410 1411 struct page_frag task_frag; 1412 1413 #ifdef CONFIG_TASK_DELAY_ACCT 1414 struct task_delay_info *delays; 1415 #endif 1416 1417 #ifdef CONFIG_FAULT_INJECTION 1418 int make_it_fail; 1419 unsigned int fail_nth; 1420 #endif 1421 /* 1422 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1423 * balance_dirty_pages() for a dirty throttling pause: 1424 */ 1425 int nr_dirtied; 1426 int nr_dirtied_pause; 1427 /* Start of a write-and-pause period: */ 1428 unsigned long dirty_paused_when; 1429 1430 #ifdef CONFIG_LATENCYTOP 1431 int latency_record_count; 1432 struct latency_record latency_record[LT_SAVECOUNT]; 1433 #endif 1434 /* 1435 * Time slack values; these are used to round up poll() and 1436 * select() etc timeout values. These are in nanoseconds. 1437 */ 1438 u64 timer_slack_ns; 1439 u64 default_timer_slack_ns; 1440 1441 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1442 unsigned int kasan_depth; 1443 #endif 1444 1445 #ifdef CONFIG_KCSAN 1446 struct kcsan_ctx kcsan_ctx; 1447 #ifdef CONFIG_TRACE_IRQFLAGS 1448 struct irqtrace_events kcsan_save_irqtrace; 1449 #endif 1450 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1451 int kcsan_stack_depth; 1452 #endif 1453 #endif 1454 1455 #ifdef CONFIG_KMSAN 1456 struct kmsan_ctx kmsan_ctx; 1457 #endif 1458 1459 #if IS_ENABLED(CONFIG_KUNIT) 1460 struct kunit *kunit_test; 1461 #endif 1462 1463 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1464 /* Index of current stored address in ret_stack: */ 1465 int curr_ret_stack; 1466 int curr_ret_depth; 1467 1468 /* Stack of return addresses for return function tracing: */ 1469 unsigned long *ret_stack; 1470 1471 /* Timestamp for last schedule: */ 1472 unsigned long long ftrace_timestamp; 1473 unsigned long long ftrace_sleeptime; 1474 1475 /* 1476 * Number of functions that haven't been traced 1477 * because of depth overrun: 1478 */ 1479 atomic_t trace_overrun; 1480 1481 /* Pause tracing: */ 1482 atomic_t tracing_graph_pause; 1483 #endif 1484 1485 #ifdef CONFIG_TRACING 1486 /* Bitmask and counter of trace recursion: */ 1487 unsigned long trace_recursion; 1488 #endif /* CONFIG_TRACING */ 1489 1490 #ifdef CONFIG_KCOV 1491 /* See kernel/kcov.c for more details. */ 1492 1493 /* Coverage collection mode enabled for this task (0 if disabled): */ 1494 unsigned int kcov_mode; 1495 1496 /* Size of the kcov_area: */ 1497 unsigned int kcov_size; 1498 1499 /* Buffer for coverage collection: */ 1500 void *kcov_area; 1501 1502 /* KCOV descriptor wired with this task or NULL: */ 1503 struct kcov *kcov; 1504 1505 /* KCOV common handle for remote coverage collection: */ 1506 u64 kcov_handle; 1507 1508 /* KCOV sequence number: */ 1509 int kcov_sequence; 1510 1511 /* Collect coverage from softirq context: */ 1512 unsigned int kcov_softirq; 1513 #endif 1514 1515 #ifdef CONFIG_MEMCG_V1 1516 struct mem_cgroup *memcg_in_oom; 1517 #endif 1518 1519 #ifdef CONFIG_MEMCG 1520 /* Number of pages to reclaim on returning to userland: */ 1521 unsigned int memcg_nr_pages_over_high; 1522 1523 /* Used by memcontrol for targeted memcg charge: */ 1524 struct mem_cgroup *active_memcg; 1525 1526 /* Cache for current->cgroups->memcg->objcg lookups: */ 1527 struct obj_cgroup *objcg; 1528 #endif 1529 1530 #ifdef CONFIG_BLK_CGROUP 1531 struct gendisk *throttle_disk; 1532 #endif 1533 1534 #ifdef CONFIG_UPROBES 1535 struct uprobe_task *utask; 1536 #endif 1537 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1538 unsigned int sequential_io; 1539 unsigned int sequential_io_avg; 1540 #endif 1541 struct kmap_ctrl kmap_ctrl; 1542 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1543 unsigned long task_state_change; 1544 # ifdef CONFIG_PREEMPT_RT 1545 unsigned long saved_state_change; 1546 # endif 1547 #endif 1548 struct rcu_head rcu; 1549 refcount_t rcu_users; 1550 int pagefault_disabled; 1551 #ifdef CONFIG_MMU 1552 struct task_struct *oom_reaper_list; 1553 struct timer_list oom_reaper_timer; 1554 #endif 1555 #ifdef CONFIG_VMAP_STACK 1556 struct vm_struct *stack_vm_area; 1557 #endif 1558 #ifdef CONFIG_THREAD_INFO_IN_TASK 1559 /* A live task holds one reference: */ 1560 refcount_t stack_refcount; 1561 #endif 1562 #ifdef CONFIG_LIVEPATCH 1563 int patch_state; 1564 #endif 1565 #ifdef CONFIG_SECURITY 1566 /* Used by LSM modules for access restriction: */ 1567 void *security; 1568 #endif 1569 #ifdef CONFIG_BPF_SYSCALL 1570 /* Used by BPF task local storage */ 1571 struct bpf_local_storage __rcu *bpf_storage; 1572 /* Used for BPF run context */ 1573 struct bpf_run_ctx *bpf_ctx; 1574 #endif 1575 /* Used by BPF for per-TASK xdp storage */ 1576 struct bpf_net_context *bpf_net_context; 1577 1578 #ifdef CONFIG_KSTACK_ERASE 1579 unsigned long lowest_stack; 1580 #endif 1581 #ifdef CONFIG_KSTACK_ERASE_METRICS 1582 unsigned long prev_lowest_stack; 1583 #endif 1584 1585 #ifdef CONFIG_X86_MCE 1586 void __user *mce_vaddr; 1587 __u64 mce_kflags; 1588 u64 mce_addr; 1589 __u64 mce_ripv : 1, 1590 mce_whole_page : 1, 1591 __mce_reserved : 62; 1592 struct callback_head mce_kill_me; 1593 int mce_count; 1594 #endif 1595 1596 #ifdef CONFIG_KRETPROBES 1597 struct llist_head kretprobe_instances; 1598 #endif 1599 #ifdef CONFIG_RETHOOK 1600 struct llist_head rethooks; 1601 #endif 1602 1603 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1604 /* 1605 * If L1D flush is supported on mm context switch 1606 * then we use this callback head to queue kill work 1607 * to kill tasks that are not running on SMT disabled 1608 * cores 1609 */ 1610 struct callback_head l1d_flush_kill; 1611 #endif 1612 1613 #ifdef CONFIG_RV 1614 /* 1615 * Per-task RV monitor, fixed in CONFIG_RV_PER_TASK_MONITORS. 1616 * If memory becomes a concern, we can think about a dynamic method. 1617 */ 1618 union rv_task_monitor rv[CONFIG_RV_PER_TASK_MONITORS]; 1619 #endif 1620 1621 #ifdef CONFIG_USER_EVENTS 1622 struct user_event_mm *user_event_mm; 1623 #endif 1624 1625 #ifdef CONFIG_UNWIND_USER 1626 struct unwind_task_info unwind_info; 1627 #endif 1628 1629 /* CPU-specific state of this task: */ 1630 struct thread_struct thread; 1631 1632 /* 1633 * New fields for task_struct should be added above here, so that 1634 * they are included in the randomized portion of task_struct. 1635 */ 1636 randomized_struct_fields_end 1637 } __attribute__ ((aligned (64))); 1638 1639 #ifdef CONFIG_SCHED_PROXY_EXEC 1640 DECLARE_STATIC_KEY_TRUE(__sched_proxy_exec); 1641 static inline bool sched_proxy_exec(void) 1642 { 1643 return static_branch_likely(&__sched_proxy_exec); 1644 } 1645 #else 1646 static inline bool sched_proxy_exec(void) 1647 { 1648 return false; 1649 } 1650 #endif 1651 1652 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1653 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1654 1655 static inline unsigned int __task_state_index(unsigned int tsk_state, 1656 unsigned int tsk_exit_state) 1657 { 1658 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1659 1660 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1661 1662 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1663 state = TASK_REPORT_IDLE; 1664 1665 /* 1666 * We're lying here, but rather than expose a completely new task state 1667 * to userspace, we can make this appear as if the task has gone through 1668 * a regular rt_mutex_lock() call. 1669 * Report frozen tasks as uninterruptible. 1670 */ 1671 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN)) 1672 state = TASK_UNINTERRUPTIBLE; 1673 1674 return fls(state); 1675 } 1676 1677 static inline unsigned int task_state_index(struct task_struct *tsk) 1678 { 1679 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1680 } 1681 1682 static inline char task_index_to_char(unsigned int state) 1683 { 1684 static const char state_char[] = "RSDTtXZPI"; 1685 1686 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1)); 1687 1688 return state_char[state]; 1689 } 1690 1691 static inline char task_state_to_char(struct task_struct *tsk) 1692 { 1693 return task_index_to_char(task_state_index(tsk)); 1694 } 1695 1696 extern struct pid *cad_pid; 1697 1698 /* 1699 * Per process flags 1700 */ 1701 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1702 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1703 #define PF_EXITING 0x00000004 /* Getting shut down */ 1704 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1705 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1706 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1707 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1708 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1709 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1710 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1711 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1712 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1713 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1714 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1715 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1716 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1717 #define PF_KCOMPACTD 0x00010000 /* I am kcompactd */ 1718 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1719 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1720 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1721 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1722 * I am cleaning dirty pages from some other bdi. */ 1723 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1724 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1725 #define PF__HOLE__00800000 0x00800000 1726 #define PF__HOLE__01000000 0x01000000 1727 #define PF__HOLE__02000000 0x02000000 1728 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1729 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1730 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1731 * See memalloc_pin_save() */ 1732 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1733 #define PF__HOLE__40000000 0x40000000 1734 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1735 1736 /* 1737 * Only the _current_ task can read/write to tsk->flags, but other 1738 * tasks can access tsk->flags in readonly mode for example 1739 * with tsk_used_math (like during threaded core dumping). 1740 * There is however an exception to this rule during ptrace 1741 * or during fork: the ptracer task is allowed to write to the 1742 * child->flags of its traced child (same goes for fork, the parent 1743 * can write to the child->flags), because we're guaranteed the 1744 * child is not running and in turn not changing child->flags 1745 * at the same time the parent does it. 1746 */ 1747 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1748 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1749 #define clear_used_math() clear_stopped_child_used_math(current) 1750 #define set_used_math() set_stopped_child_used_math(current) 1751 1752 #define conditional_stopped_child_used_math(condition, child) \ 1753 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1754 1755 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1756 1757 #define copy_to_stopped_child_used_math(child) \ 1758 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1759 1760 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1761 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1762 #define used_math() tsk_used_math(current) 1763 1764 static __always_inline bool is_percpu_thread(void) 1765 { 1766 return (current->flags & PF_NO_SETAFFINITY) && 1767 (current->nr_cpus_allowed == 1); 1768 } 1769 1770 static __always_inline bool is_user_task(struct task_struct *task) 1771 { 1772 return task->mm && !(task->flags & (PF_KTHREAD | PF_USER_WORKER)); 1773 } 1774 1775 /* Per-process atomic flags. */ 1776 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1777 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1778 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1779 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1780 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1781 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1782 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1783 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1784 1785 #define TASK_PFA_TEST(name, func) \ 1786 static inline bool task_##func(struct task_struct *p) \ 1787 { return test_bit(PFA_##name, &p->atomic_flags); } 1788 1789 #define TASK_PFA_SET(name, func) \ 1790 static inline void task_set_##func(struct task_struct *p) \ 1791 { set_bit(PFA_##name, &p->atomic_flags); } 1792 1793 #define TASK_PFA_CLEAR(name, func) \ 1794 static inline void task_clear_##func(struct task_struct *p) \ 1795 { clear_bit(PFA_##name, &p->atomic_flags); } 1796 1797 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1798 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1799 1800 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1801 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1802 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1803 1804 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1805 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1806 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1807 1808 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1809 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1810 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1811 1812 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1813 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1814 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1815 1816 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1817 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1818 1819 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1820 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1821 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1822 1823 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1824 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1825 1826 static inline void 1827 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1828 { 1829 current->flags &= ~flags; 1830 current->flags |= orig_flags & flags; 1831 } 1832 1833 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1834 extern int task_can_attach(struct task_struct *p); 1835 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1836 extern void dl_bw_free(int cpu, u64 dl_bw); 1837 1838 /* set_cpus_allowed_force() - consider using set_cpus_allowed_ptr() instead */ 1839 extern void set_cpus_allowed_force(struct task_struct *p, const struct cpumask *new_mask); 1840 1841 /** 1842 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1843 * @p: the task 1844 * @new_mask: CPU affinity mask 1845 * 1846 * Return: zero if successful, or a negative error code 1847 */ 1848 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1849 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1850 extern void release_user_cpus_ptr(struct task_struct *p); 1851 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1852 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1853 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1854 1855 extern int yield_to(struct task_struct *p, bool preempt); 1856 extern void set_user_nice(struct task_struct *p, long nice); 1857 extern int task_prio(const struct task_struct *p); 1858 1859 /** 1860 * task_nice - return the nice value of a given task. 1861 * @p: the task in question. 1862 * 1863 * Return: The nice value [ -20 ... 0 ... 19 ]. 1864 */ 1865 static inline int task_nice(const struct task_struct *p) 1866 { 1867 return PRIO_TO_NICE((p)->static_prio); 1868 } 1869 1870 extern int can_nice(const struct task_struct *p, const int nice); 1871 extern int task_curr(const struct task_struct *p); 1872 extern int idle_cpu(int cpu); 1873 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1874 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1875 extern void sched_set_fifo(struct task_struct *p); 1876 extern void sched_set_fifo_low(struct task_struct *p); 1877 extern void sched_set_fifo_secondary(struct task_struct *p); 1878 extern void sched_set_normal(struct task_struct *p, int nice); 1879 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1880 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1881 extern struct task_struct *idle_task(int cpu); 1882 1883 /** 1884 * is_idle_task - is the specified task an idle task? 1885 * @p: the task in question. 1886 * 1887 * Return: 1 if @p is an idle task. 0 otherwise. 1888 */ 1889 static __always_inline bool is_idle_task(const struct task_struct *p) 1890 { 1891 return !!(p->flags & PF_IDLE); 1892 } 1893 1894 extern struct task_struct *curr_task(int cpu); 1895 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1896 1897 void yield(void); 1898 1899 union thread_union { 1900 struct task_struct task; 1901 #ifndef CONFIG_THREAD_INFO_IN_TASK 1902 struct thread_info thread_info; 1903 #endif 1904 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1905 }; 1906 1907 #ifndef CONFIG_THREAD_INFO_IN_TASK 1908 extern struct thread_info init_thread_info; 1909 #endif 1910 1911 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1912 1913 #ifdef CONFIG_THREAD_INFO_IN_TASK 1914 # define task_thread_info(task) (&(task)->thread_info) 1915 #else 1916 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1917 #endif 1918 1919 /* 1920 * find a task by one of its numerical ids 1921 * 1922 * find_task_by_pid_ns(): 1923 * finds a task by its pid in the specified namespace 1924 * find_task_by_vpid(): 1925 * finds a task by its virtual pid 1926 * 1927 * see also find_vpid() etc in include/linux/pid.h 1928 */ 1929 1930 extern struct task_struct *find_task_by_vpid(pid_t nr); 1931 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1932 1933 /* 1934 * find a task by its virtual pid and get the task struct 1935 */ 1936 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1937 1938 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1939 extern int wake_up_process(struct task_struct *tsk); 1940 extern void wake_up_new_task(struct task_struct *tsk); 1941 1942 extern void kick_process(struct task_struct *tsk); 1943 1944 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1945 #define set_task_comm(tsk, from) ({ \ 1946 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \ 1947 __set_task_comm(tsk, from, false); \ 1948 }) 1949 1950 /* 1951 * - Why not use task_lock()? 1952 * User space can randomly change their names anyway, so locking for readers 1953 * doesn't make sense. For writers, locking is probably necessary, as a race 1954 * condition could lead to long-term mixed results. 1955 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is 1956 * always NUL-terminated and zero-padded. Therefore the race condition between 1957 * reader and writer is not an issue. 1958 * 1959 * - BUILD_BUG_ON() can help prevent the buf from being truncated. 1960 * Since the callers don't perform any return value checks, this safeguard is 1961 * necessary. 1962 */ 1963 #define get_task_comm(buf, tsk) ({ \ 1964 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \ 1965 strscpy_pad(buf, (tsk)->comm); \ 1966 buf; \ 1967 }) 1968 1969 static __always_inline void scheduler_ipi(void) 1970 { 1971 /* 1972 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1973 * TIF_NEED_RESCHED remotely (for the first time) will also send 1974 * this IPI. 1975 */ 1976 preempt_fold_need_resched(); 1977 } 1978 1979 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 1980 1981 /* 1982 * Set thread flags in other task's structures. 1983 * See asm/thread_info.h for TIF_xxxx flags available: 1984 */ 1985 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1986 { 1987 set_ti_thread_flag(task_thread_info(tsk), flag); 1988 } 1989 1990 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1991 { 1992 clear_ti_thread_flag(task_thread_info(tsk), flag); 1993 } 1994 1995 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1996 bool value) 1997 { 1998 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1999 } 2000 2001 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2002 { 2003 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2004 } 2005 2006 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2007 { 2008 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2009 } 2010 2011 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2012 { 2013 return test_ti_thread_flag(task_thread_info(tsk), flag); 2014 } 2015 2016 static inline void set_tsk_need_resched(struct task_struct *tsk) 2017 { 2018 if (tracepoint_enabled(sched_set_need_resched_tp) && 2019 !test_tsk_thread_flag(tsk, TIF_NEED_RESCHED)) 2020 __trace_set_need_resched(tsk, TIF_NEED_RESCHED); 2021 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2022 } 2023 2024 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2025 { 2026 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY, 2027 (atomic_long_t *)&task_thread_info(tsk)->flags); 2028 } 2029 2030 static inline int test_tsk_need_resched(struct task_struct *tsk) 2031 { 2032 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2033 } 2034 2035 static inline void set_need_resched_current(void) 2036 { 2037 lockdep_assert_irqs_disabled(); 2038 set_tsk_need_resched(current); 2039 set_preempt_need_resched(); 2040 } 2041 2042 /* 2043 * cond_resched() and cond_resched_lock(): latency reduction via 2044 * explicit rescheduling in places that are safe. The return 2045 * value indicates whether a reschedule was done in fact. 2046 * cond_resched_lock() will drop the spinlock before scheduling, 2047 */ 2048 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2049 extern int __cond_resched(void); 2050 2051 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2052 2053 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2054 2055 static __always_inline int _cond_resched(void) 2056 { 2057 return static_call_mod(cond_resched)(); 2058 } 2059 2060 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2061 2062 extern int dynamic_cond_resched(void); 2063 2064 static __always_inline int _cond_resched(void) 2065 { 2066 return dynamic_cond_resched(); 2067 } 2068 2069 #else /* !CONFIG_PREEMPTION */ 2070 2071 static inline int _cond_resched(void) 2072 { 2073 return __cond_resched(); 2074 } 2075 2076 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2077 2078 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2079 2080 static inline int _cond_resched(void) 2081 { 2082 return 0; 2083 } 2084 2085 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2086 2087 #define cond_resched() ({ \ 2088 __might_resched(__FILE__, __LINE__, 0); \ 2089 _cond_resched(); \ 2090 }) 2091 2092 extern int __cond_resched_lock(spinlock_t *lock) __must_hold(lock); 2093 extern int __cond_resched_rwlock_read(rwlock_t *lock) __must_hold_shared(lock); 2094 extern int __cond_resched_rwlock_write(rwlock_t *lock) __must_hold(lock); 2095 2096 #define MIGHT_RESCHED_RCU_SHIFT 8 2097 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2098 2099 #ifndef CONFIG_PREEMPT_RT 2100 /* 2101 * Non RT kernels have an elevated preempt count due to the held lock, 2102 * but are not allowed to be inside a RCU read side critical section 2103 */ 2104 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2105 #else 2106 /* 2107 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2108 * cond_resched*lock() has to take that into account because it checks for 2109 * preempt_count() and rcu_preempt_depth(). 2110 */ 2111 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2112 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2113 #endif 2114 2115 #define cond_resched_lock(lock) ({ \ 2116 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2117 __cond_resched_lock(lock); \ 2118 }) 2119 2120 #define cond_resched_rwlock_read(lock) ({ \ 2121 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2122 __cond_resched_rwlock_read(lock); \ 2123 }) 2124 2125 #define cond_resched_rwlock_write(lock) ({ \ 2126 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2127 __cond_resched_rwlock_write(lock); \ 2128 }) 2129 2130 #ifndef CONFIG_PREEMPT_RT 2131 static inline struct mutex *__get_task_blocked_on(struct task_struct *p) 2132 { 2133 struct mutex *m = p->blocked_on; 2134 2135 if (m) 2136 lockdep_assert_held_once(&m->wait_lock); 2137 return m; 2138 } 2139 2140 static inline void __set_task_blocked_on(struct task_struct *p, struct mutex *m) 2141 { 2142 struct mutex *blocked_on = READ_ONCE(p->blocked_on); 2143 2144 WARN_ON_ONCE(!m); 2145 /* The task should only be setting itself as blocked */ 2146 WARN_ON_ONCE(p != current); 2147 /* Currently we serialize blocked_on under the mutex::wait_lock */ 2148 lockdep_assert_held_once(&m->wait_lock); 2149 /* 2150 * Check ensure we don't overwrite existing mutex value 2151 * with a different mutex. Note, setting it to the same 2152 * lock repeatedly is ok. 2153 */ 2154 WARN_ON_ONCE(blocked_on && blocked_on != m); 2155 WRITE_ONCE(p->blocked_on, m); 2156 } 2157 2158 static inline void set_task_blocked_on(struct task_struct *p, struct mutex *m) 2159 { 2160 guard(raw_spinlock_irqsave)(&m->wait_lock); 2161 __set_task_blocked_on(p, m); 2162 } 2163 2164 static inline void __clear_task_blocked_on(struct task_struct *p, struct mutex *m) 2165 { 2166 if (m) { 2167 struct mutex *blocked_on = READ_ONCE(p->blocked_on); 2168 2169 /* Currently we serialize blocked_on under the mutex::wait_lock */ 2170 lockdep_assert_held_once(&m->wait_lock); 2171 /* 2172 * There may be cases where we re-clear already cleared 2173 * blocked_on relationships, but make sure we are not 2174 * clearing the relationship with a different lock. 2175 */ 2176 WARN_ON_ONCE(blocked_on && blocked_on != m); 2177 } 2178 WRITE_ONCE(p->blocked_on, NULL); 2179 } 2180 2181 static inline void clear_task_blocked_on(struct task_struct *p, struct mutex *m) 2182 { 2183 guard(raw_spinlock_irqsave)(&m->wait_lock); 2184 __clear_task_blocked_on(p, m); 2185 } 2186 #else 2187 static inline void __clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m) 2188 { 2189 } 2190 2191 static inline void clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m) 2192 { 2193 } 2194 #endif /* !CONFIG_PREEMPT_RT */ 2195 2196 static __always_inline bool need_resched(void) 2197 { 2198 return unlikely(tif_need_resched()); 2199 } 2200 2201 /* 2202 * Wrappers for p->thread_info->cpu access. No-op on UP. 2203 */ 2204 #ifdef CONFIG_SMP 2205 2206 static inline unsigned int task_cpu(const struct task_struct *p) 2207 { 2208 return READ_ONCE(task_thread_info(p)->cpu); 2209 } 2210 2211 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2212 2213 #else 2214 2215 static inline unsigned int task_cpu(const struct task_struct *p) 2216 { 2217 return 0; 2218 } 2219 2220 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2221 { 2222 } 2223 2224 #endif /* CONFIG_SMP */ 2225 2226 static inline bool task_is_runnable(struct task_struct *p) 2227 { 2228 return p->on_rq && !p->se.sched_delayed; 2229 } 2230 2231 extern bool sched_task_on_rq(struct task_struct *p); 2232 extern unsigned long get_wchan(struct task_struct *p); 2233 extern struct task_struct *cpu_curr_snapshot(int cpu); 2234 2235 /* 2236 * In order to reduce various lock holder preemption latencies provide an 2237 * interface to see if a vCPU is currently running or not. 2238 * 2239 * This allows us to terminate optimistic spin loops and block, analogous to 2240 * the native optimistic spin heuristic of testing if the lock owner task is 2241 * running or not. 2242 */ 2243 #ifndef vcpu_is_preempted 2244 static inline bool vcpu_is_preempted(int cpu) 2245 { 2246 return false; 2247 } 2248 #endif 2249 2250 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2251 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2252 2253 #ifndef TASK_SIZE_OF 2254 #define TASK_SIZE_OF(tsk) TASK_SIZE 2255 #endif 2256 2257 static inline bool owner_on_cpu(struct task_struct *owner) 2258 { 2259 /* 2260 * As lock holder preemption issue, we both skip spinning if 2261 * task is not on cpu or its cpu is preempted 2262 */ 2263 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2264 } 2265 2266 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2267 unsigned long sched_cpu_util(int cpu); 2268 2269 #ifdef CONFIG_SCHED_CORE 2270 extern void sched_core_free(struct task_struct *tsk); 2271 extern void sched_core_fork(struct task_struct *p); 2272 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2273 unsigned long uaddr); 2274 extern int sched_core_idle_cpu(int cpu); 2275 #else 2276 static inline void sched_core_free(struct task_struct *tsk) { } 2277 static inline void sched_core_fork(struct task_struct *p) { } 2278 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2279 #endif 2280 2281 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2282 2283 #ifdef CONFIG_MEM_ALLOC_PROFILING 2284 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2285 { 2286 swap(current->alloc_tag, tag); 2287 return tag; 2288 } 2289 2290 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2291 { 2292 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2293 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2294 #endif 2295 current->alloc_tag = old; 2296 } 2297 #else 2298 #define alloc_tag_save(_tag) NULL 2299 #define alloc_tag_restore(_tag, _old) do {} while (0) 2300 #endif 2301 2302 /* Avoids recursive inclusion hell */ 2303 #ifdef CONFIG_SCHED_MM_CID 2304 void sched_mm_cid_before_execve(struct task_struct *t); 2305 void sched_mm_cid_after_execve(struct task_struct *t); 2306 void sched_mm_cid_fork(struct task_struct *t); 2307 void sched_mm_cid_exit(struct task_struct *t); 2308 static __always_inline int task_mm_cid(struct task_struct *t) 2309 { 2310 return t->mm_cid.cid & ~(MM_CID_ONCPU | MM_CID_TRANSIT); 2311 } 2312 #else 2313 static inline void sched_mm_cid_before_execve(struct task_struct *t) { } 2314 static inline void sched_mm_cid_after_execve(struct task_struct *t) { } 2315 static inline void sched_mm_cid_fork(struct task_struct *t) { } 2316 static inline void sched_mm_cid_exit(struct task_struct *t) { } 2317 static __always_inline int task_mm_cid(struct task_struct *t) 2318 { 2319 /* 2320 * Use the processor id as a fall-back when the mm cid feature is 2321 * disabled. This provides functional per-cpu data structure accesses 2322 * in user-space, althrough it won't provide the memory usage benefits. 2323 */ 2324 return task_cpu(t); 2325 } 2326 #endif 2327 2328 #ifndef MODULE 2329 #ifndef COMPILE_OFFSETS 2330 2331 extern void ___migrate_enable(void); 2332 2333 struct rq; 2334 DECLARE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues); 2335 2336 /* 2337 * The "struct rq" is not available here, so we can't access the 2338 * "runqueues" with this_cpu_ptr(), as the compilation will fail in 2339 * this_cpu_ptr() -> raw_cpu_ptr() -> __verify_pcpu_ptr(): 2340 * typeof((ptr) + 0) 2341 * 2342 * So use arch_raw_cpu_ptr()/PERCPU_PTR() directly here. 2343 */ 2344 #ifdef CONFIG_SMP 2345 #define this_rq_raw() arch_raw_cpu_ptr(&runqueues) 2346 #else 2347 #define this_rq_raw() PERCPU_PTR(&runqueues) 2348 #endif 2349 #define this_rq_pinned() (*(unsigned int *)((void *)this_rq_raw() + RQ_nr_pinned)) 2350 2351 static inline void __migrate_enable(void) 2352 { 2353 struct task_struct *p = current; 2354 2355 #ifdef CONFIG_DEBUG_PREEMPT 2356 /* 2357 * Check both overflow from migrate_disable() and superfluous 2358 * migrate_enable(). 2359 */ 2360 if (WARN_ON_ONCE((s16)p->migration_disabled <= 0)) 2361 return; 2362 #endif 2363 2364 if (p->migration_disabled > 1) { 2365 p->migration_disabled--; 2366 return; 2367 } 2368 2369 /* 2370 * Ensure stop_task runs either before or after this, and that 2371 * __set_cpus_allowed_ptr(SCA_MIGRATE_ENABLE) doesn't schedule(). 2372 */ 2373 guard(preempt)(); 2374 if (unlikely(p->cpus_ptr != &p->cpus_mask)) 2375 ___migrate_enable(); 2376 /* 2377 * Mustn't clear migration_disabled() until cpus_ptr points back at the 2378 * regular cpus_mask, otherwise things that race (eg. 2379 * select_fallback_rq) get confused. 2380 */ 2381 barrier(); 2382 p->migration_disabled = 0; 2383 this_rq_pinned()--; 2384 } 2385 2386 static inline void __migrate_disable(void) 2387 { 2388 struct task_struct *p = current; 2389 2390 if (p->migration_disabled) { 2391 #ifdef CONFIG_DEBUG_PREEMPT 2392 /* 2393 *Warn about overflow half-way through the range. 2394 */ 2395 WARN_ON_ONCE((s16)p->migration_disabled < 0); 2396 #endif 2397 p->migration_disabled++; 2398 return; 2399 } 2400 2401 guard(preempt)(); 2402 this_rq_pinned()++; 2403 p->migration_disabled = 1; 2404 } 2405 #else /* !COMPILE_OFFSETS */ 2406 static inline void __migrate_disable(void) { } 2407 static inline void __migrate_enable(void) { } 2408 #endif /* !COMPILE_OFFSETS */ 2409 2410 /* 2411 * So that it is possible to not export the runqueues variable, define and 2412 * export migrate_enable/migrate_disable in kernel/sched/core.c too, and use 2413 * them for the modules. The macro "INSTANTIATE_EXPORTED_MIGRATE_DISABLE" will 2414 * be defined in kernel/sched/core.c. 2415 */ 2416 #ifndef INSTANTIATE_EXPORTED_MIGRATE_DISABLE 2417 static __always_inline void migrate_disable(void) 2418 { 2419 __migrate_disable(); 2420 } 2421 2422 static __always_inline void migrate_enable(void) 2423 { 2424 __migrate_enable(); 2425 } 2426 #else /* INSTANTIATE_EXPORTED_MIGRATE_DISABLE */ 2427 extern void migrate_disable(void); 2428 extern void migrate_enable(void); 2429 #endif /* INSTANTIATE_EXPORTED_MIGRATE_DISABLE */ 2430 2431 #else /* MODULE */ 2432 extern void migrate_disable(void); 2433 extern void migrate_enable(void); 2434 #endif /* MODULE */ 2435 2436 DEFINE_LOCK_GUARD_0(migrate, migrate_disable(), migrate_enable()) 2437 2438 #endif 2439