1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_SCHED_H 3 #define _LINUX_SCHED_H 4 5 /* 6 * Define 'struct task_struct' and provide the main scheduler 7 * APIs (schedule(), wakeup variants, etc.) 8 */ 9 10 #include <uapi/linux/sched.h> 11 12 #include <asm/current.h> 13 #include <asm/processor.h> 14 #include <linux/thread_info.h> 15 #include <linux/preempt.h> 16 #include <linux/cpumask_types.h> 17 18 #include <linux/cache.h> 19 #include <linux/irqflags_types.h> 20 #include <linux/smp_types.h> 21 #include <linux/pid_types.h> 22 #include <linux/sem_types.h> 23 #include <linux/shm.h> 24 #include <linux/kmsan_types.h> 25 #include <linux/mutex_types.h> 26 #include <linux/plist_types.h> 27 #include <linux/hrtimer_types.h> 28 #include <linux/timer_types.h> 29 #include <linux/seccomp_types.h> 30 #include <linux/nodemask_types.h> 31 #include <linux/refcount_types.h> 32 #include <linux/resource.h> 33 #include <linux/latencytop.h> 34 #include <linux/sched/prio.h> 35 #include <linux/sched/types.h> 36 #include <linux/signal_types.h> 37 #include <linux/syscall_user_dispatch_types.h> 38 #include <linux/mm_types_task.h> 39 #include <linux/netdevice_xmit.h> 40 #include <linux/task_io_accounting.h> 41 #include <linux/posix-timers_types.h> 42 #include <linux/restart_block.h> 43 #include <uapi/linux/rseq.h> 44 #include <linux/seqlock_types.h> 45 #include <linux/kcsan.h> 46 #include <linux/rv.h> 47 #include <linux/livepatch_sched.h> 48 #include <linux/uidgid_types.h> 49 #include <asm/kmap_size.h> 50 51 /* task_struct member predeclarations (sorted alphabetically): */ 52 struct audit_context; 53 struct bio_list; 54 struct blk_plug; 55 struct bpf_local_storage; 56 struct bpf_run_ctx; 57 struct bpf_net_context; 58 struct capture_control; 59 struct cfs_rq; 60 struct fs_struct; 61 struct futex_pi_state; 62 struct io_context; 63 struct io_uring_task; 64 struct mempolicy; 65 struct nameidata; 66 struct nsproxy; 67 struct perf_event_context; 68 struct pid_namespace; 69 struct pipe_inode_info; 70 struct rcu_node; 71 struct reclaim_state; 72 struct robust_list_head; 73 struct root_domain; 74 struct rq; 75 struct sched_attr; 76 struct sched_dl_entity; 77 struct seq_file; 78 struct sighand_struct; 79 struct signal_struct; 80 struct task_delay_info; 81 struct task_group; 82 struct task_struct; 83 struct user_event_mm; 84 85 #include <linux/sched/ext.h> 86 87 /* 88 * Task state bitmask. NOTE! These bits are also 89 * encoded in fs/proc/array.c: get_task_state(). 90 * 91 * We have two separate sets of flags: task->__state 92 * is about runnability, while task->exit_state are 93 * about the task exiting. Confusing, but this way 94 * modifying one set can't modify the other one by 95 * mistake. 96 */ 97 98 /* Used in tsk->__state: */ 99 #define TASK_RUNNING 0x00000000 100 #define TASK_INTERRUPTIBLE 0x00000001 101 #define TASK_UNINTERRUPTIBLE 0x00000002 102 #define __TASK_STOPPED 0x00000004 103 #define __TASK_TRACED 0x00000008 104 /* Used in tsk->exit_state: */ 105 #define EXIT_DEAD 0x00000010 106 #define EXIT_ZOMBIE 0x00000020 107 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 108 /* Used in tsk->__state again: */ 109 #define TASK_PARKED 0x00000040 110 #define TASK_DEAD 0x00000080 111 #define TASK_WAKEKILL 0x00000100 112 #define TASK_WAKING 0x00000200 113 #define TASK_NOLOAD 0x00000400 114 #define TASK_NEW 0x00000800 115 #define TASK_RTLOCK_WAIT 0x00001000 116 #define TASK_FREEZABLE 0x00002000 117 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP)) 118 #define TASK_FROZEN 0x00008000 119 #define TASK_STATE_MAX 0x00010000 120 121 #define TASK_ANY (TASK_STATE_MAX-1) 122 123 /* 124 * DO NOT ADD ANY NEW USERS ! 125 */ 126 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE) 127 128 /* Convenience macros for the sake of set_current_state: */ 129 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 130 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 131 #define TASK_TRACED __TASK_TRACED 132 133 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 134 135 /* Convenience macros for the sake of wake_up(): */ 136 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 137 138 /* get_task_state(): */ 139 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 140 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 141 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 142 TASK_PARKED) 143 144 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 145 146 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 147 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 148 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 149 150 /* 151 * Special states are those that do not use the normal wait-loop pattern. See 152 * the comment with set_special_state(). 153 */ 154 #define is_special_task_state(state) \ 155 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \ 156 TASK_DEAD | TASK_FROZEN)) 157 158 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 159 # define debug_normal_state_change(state_value) \ 160 do { \ 161 WARN_ON_ONCE(is_special_task_state(state_value)); \ 162 current->task_state_change = _THIS_IP_; \ 163 } while (0) 164 165 # define debug_special_state_change(state_value) \ 166 do { \ 167 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 168 current->task_state_change = _THIS_IP_; \ 169 } while (0) 170 171 # define debug_rtlock_wait_set_state() \ 172 do { \ 173 current->saved_state_change = current->task_state_change;\ 174 current->task_state_change = _THIS_IP_; \ 175 } while (0) 176 177 # define debug_rtlock_wait_restore_state() \ 178 do { \ 179 current->task_state_change = current->saved_state_change;\ 180 } while (0) 181 182 #else 183 # define debug_normal_state_change(cond) do { } while (0) 184 # define debug_special_state_change(cond) do { } while (0) 185 # define debug_rtlock_wait_set_state() do { } while (0) 186 # define debug_rtlock_wait_restore_state() do { } while (0) 187 #endif 188 189 /* 190 * set_current_state() includes a barrier so that the write of current->__state 191 * is correctly serialised wrt the caller's subsequent test of whether to 192 * actually sleep: 193 * 194 * for (;;) { 195 * set_current_state(TASK_UNINTERRUPTIBLE); 196 * if (CONDITION) 197 * break; 198 * 199 * schedule(); 200 * } 201 * __set_current_state(TASK_RUNNING); 202 * 203 * If the caller does not need such serialisation (because, for instance, the 204 * CONDITION test and condition change and wakeup are under the same lock) then 205 * use __set_current_state(). 206 * 207 * The above is typically ordered against the wakeup, which does: 208 * 209 * CONDITION = 1; 210 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 211 * 212 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 213 * accessing p->__state. 214 * 215 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is, 216 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 217 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 218 * 219 * However, with slightly different timing the wakeup TASK_RUNNING store can 220 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 221 * a problem either because that will result in one extra go around the loop 222 * and our @cond test will save the day. 223 * 224 * Also see the comments of try_to_wake_up(). 225 */ 226 #define __set_current_state(state_value) \ 227 do { \ 228 debug_normal_state_change((state_value)); \ 229 WRITE_ONCE(current->__state, (state_value)); \ 230 } while (0) 231 232 #define set_current_state(state_value) \ 233 do { \ 234 debug_normal_state_change((state_value)); \ 235 smp_store_mb(current->__state, (state_value)); \ 236 } while (0) 237 238 /* 239 * set_special_state() should be used for those states when the blocking task 240 * can not use the regular condition based wait-loop. In that case we must 241 * serialize against wakeups such that any possible in-flight TASK_RUNNING 242 * stores will not collide with our state change. 243 */ 244 #define set_special_state(state_value) \ 245 do { \ 246 unsigned long flags; /* may shadow */ \ 247 \ 248 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 249 debug_special_state_change((state_value)); \ 250 WRITE_ONCE(current->__state, (state_value)); \ 251 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 252 } while (0) 253 254 /* 255 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 256 * 257 * RT's spin/rwlock substitutions are state preserving. The state of the 258 * task when blocking on the lock is saved in task_struct::saved_state and 259 * restored after the lock has been acquired. These operations are 260 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 261 * lock related wakeups while the task is blocked on the lock are 262 * redirected to operate on task_struct::saved_state to ensure that these 263 * are not dropped. On restore task_struct::saved_state is set to 264 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 265 * 266 * The lock operation looks like this: 267 * 268 * current_save_and_set_rtlock_wait_state(); 269 * for (;;) { 270 * if (try_lock()) 271 * break; 272 * raw_spin_unlock_irq(&lock->wait_lock); 273 * schedule_rtlock(); 274 * raw_spin_lock_irq(&lock->wait_lock); 275 * set_current_state(TASK_RTLOCK_WAIT); 276 * } 277 * current_restore_rtlock_saved_state(); 278 */ 279 #define current_save_and_set_rtlock_wait_state() \ 280 do { \ 281 lockdep_assert_irqs_disabled(); \ 282 raw_spin_lock(¤t->pi_lock); \ 283 current->saved_state = current->__state; \ 284 debug_rtlock_wait_set_state(); \ 285 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 286 raw_spin_unlock(¤t->pi_lock); \ 287 } while (0); 288 289 #define current_restore_rtlock_saved_state() \ 290 do { \ 291 lockdep_assert_irqs_disabled(); \ 292 raw_spin_lock(¤t->pi_lock); \ 293 debug_rtlock_wait_restore_state(); \ 294 WRITE_ONCE(current->__state, current->saved_state); \ 295 current->saved_state = TASK_RUNNING; \ 296 raw_spin_unlock(¤t->pi_lock); \ 297 } while (0); 298 299 #define get_current_state() READ_ONCE(current->__state) 300 301 /* 302 * Define the task command name length as enum, then it can be visible to 303 * BPF programs. 304 */ 305 enum { 306 TASK_COMM_LEN = 16, 307 }; 308 309 extern void sched_tick(void); 310 311 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 312 313 extern long schedule_timeout(long timeout); 314 extern long schedule_timeout_interruptible(long timeout); 315 extern long schedule_timeout_killable(long timeout); 316 extern long schedule_timeout_uninterruptible(long timeout); 317 extern long schedule_timeout_idle(long timeout); 318 asmlinkage void schedule(void); 319 extern void schedule_preempt_disabled(void); 320 asmlinkage void preempt_schedule_irq(void); 321 #ifdef CONFIG_PREEMPT_RT 322 extern void schedule_rtlock(void); 323 #endif 324 325 extern int __must_check io_schedule_prepare(void); 326 extern void io_schedule_finish(int token); 327 extern long io_schedule_timeout(long timeout); 328 extern void io_schedule(void); 329 330 /** 331 * struct prev_cputime - snapshot of system and user cputime 332 * @utime: time spent in user mode 333 * @stime: time spent in system mode 334 * @lock: protects the above two fields 335 * 336 * Stores previous user/system time values such that we can guarantee 337 * monotonicity. 338 */ 339 struct prev_cputime { 340 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 341 u64 utime; 342 u64 stime; 343 raw_spinlock_t lock; 344 #endif 345 }; 346 347 enum vtime_state { 348 /* Task is sleeping or running in a CPU with VTIME inactive: */ 349 VTIME_INACTIVE = 0, 350 /* Task is idle */ 351 VTIME_IDLE, 352 /* Task runs in kernelspace in a CPU with VTIME active: */ 353 VTIME_SYS, 354 /* Task runs in userspace in a CPU with VTIME active: */ 355 VTIME_USER, 356 /* Task runs as guests in a CPU with VTIME active: */ 357 VTIME_GUEST, 358 }; 359 360 struct vtime { 361 seqcount_t seqcount; 362 unsigned long long starttime; 363 enum vtime_state state; 364 unsigned int cpu; 365 u64 utime; 366 u64 stime; 367 u64 gtime; 368 }; 369 370 /* 371 * Utilization clamp constraints. 372 * @UCLAMP_MIN: Minimum utilization 373 * @UCLAMP_MAX: Maximum utilization 374 * @UCLAMP_CNT: Utilization clamp constraints count 375 */ 376 enum uclamp_id { 377 UCLAMP_MIN = 0, 378 UCLAMP_MAX, 379 UCLAMP_CNT 380 }; 381 382 #ifdef CONFIG_SMP 383 extern struct root_domain def_root_domain; 384 extern struct mutex sched_domains_mutex; 385 #endif 386 387 struct sched_param { 388 int sched_priority; 389 }; 390 391 struct sched_info { 392 #ifdef CONFIG_SCHED_INFO 393 /* Cumulative counters: */ 394 395 /* # of times we have run on this CPU: */ 396 unsigned long pcount; 397 398 /* Time spent waiting on a runqueue: */ 399 unsigned long long run_delay; 400 401 /* Max time spent waiting on a runqueue: */ 402 unsigned long long max_run_delay; 403 404 /* Timestamps: */ 405 406 /* When did we last run on a CPU? */ 407 unsigned long long last_arrival; 408 409 /* When were we last queued to run? */ 410 unsigned long long last_queued; 411 412 #endif /* CONFIG_SCHED_INFO */ 413 }; 414 415 /* 416 * Integer metrics need fixed point arithmetic, e.g., sched/fair 417 * has a few: load, load_avg, util_avg, freq, and capacity. 418 * 419 * We define a basic fixed point arithmetic range, and then formalize 420 * all these metrics based on that basic range. 421 */ 422 # define SCHED_FIXEDPOINT_SHIFT 10 423 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 424 425 /* Increase resolution of cpu_capacity calculations */ 426 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 427 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 428 429 struct load_weight { 430 unsigned long weight; 431 u32 inv_weight; 432 }; 433 434 /* 435 * The load/runnable/util_avg accumulates an infinite geometric series 436 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 437 * 438 * [load_avg definition] 439 * 440 * load_avg = runnable% * scale_load_down(load) 441 * 442 * [runnable_avg definition] 443 * 444 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 445 * 446 * [util_avg definition] 447 * 448 * util_avg = running% * SCHED_CAPACITY_SCALE 449 * 450 * where runnable% is the time ratio that a sched_entity is runnable and 451 * running% the time ratio that a sched_entity is running. 452 * 453 * For cfs_rq, they are the aggregated values of all runnable and blocked 454 * sched_entities. 455 * 456 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 457 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 458 * for computing those signals (see update_rq_clock_pelt()) 459 * 460 * N.B., the above ratios (runnable% and running%) themselves are in the 461 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 462 * to as large a range as necessary. This is for example reflected by 463 * util_avg's SCHED_CAPACITY_SCALE. 464 * 465 * [Overflow issue] 466 * 467 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 468 * with the highest load (=88761), always runnable on a single cfs_rq, 469 * and should not overflow as the number already hits PID_MAX_LIMIT. 470 * 471 * For all other cases (including 32-bit kernels), struct load_weight's 472 * weight will overflow first before we do, because: 473 * 474 * Max(load_avg) <= Max(load.weight) 475 * 476 * Then it is the load_weight's responsibility to consider overflow 477 * issues. 478 */ 479 struct sched_avg { 480 u64 last_update_time; 481 u64 load_sum; 482 u64 runnable_sum; 483 u32 util_sum; 484 u32 period_contrib; 485 unsigned long load_avg; 486 unsigned long runnable_avg; 487 unsigned long util_avg; 488 unsigned int util_est; 489 } ____cacheline_aligned; 490 491 /* 492 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 493 * updates. When a task is dequeued, its util_est should not be updated if its 494 * util_avg has not been updated in the meantime. 495 * This information is mapped into the MSB bit of util_est at dequeue time. 496 * Since max value of util_est for a task is 1024 (PELT util_avg for a task) 497 * it is safe to use MSB. 498 */ 499 #define UTIL_EST_WEIGHT_SHIFT 2 500 #define UTIL_AVG_UNCHANGED 0x80000000 501 502 struct sched_statistics { 503 #ifdef CONFIG_SCHEDSTATS 504 u64 wait_start; 505 u64 wait_max; 506 u64 wait_count; 507 u64 wait_sum; 508 u64 iowait_count; 509 u64 iowait_sum; 510 511 u64 sleep_start; 512 u64 sleep_max; 513 s64 sum_sleep_runtime; 514 515 u64 block_start; 516 u64 block_max; 517 s64 sum_block_runtime; 518 519 s64 exec_max; 520 u64 slice_max; 521 522 u64 nr_migrations_cold; 523 u64 nr_failed_migrations_affine; 524 u64 nr_failed_migrations_running; 525 u64 nr_failed_migrations_hot; 526 u64 nr_forced_migrations; 527 528 u64 nr_wakeups; 529 u64 nr_wakeups_sync; 530 u64 nr_wakeups_migrate; 531 u64 nr_wakeups_local; 532 u64 nr_wakeups_remote; 533 u64 nr_wakeups_affine; 534 u64 nr_wakeups_affine_attempts; 535 u64 nr_wakeups_passive; 536 u64 nr_wakeups_idle; 537 538 #ifdef CONFIG_SCHED_CORE 539 u64 core_forceidle_sum; 540 #endif 541 #endif /* CONFIG_SCHEDSTATS */ 542 } ____cacheline_aligned; 543 544 struct sched_entity { 545 /* For load-balancing: */ 546 struct load_weight load; 547 struct rb_node run_node; 548 u64 deadline; 549 u64 min_vruntime; 550 u64 min_slice; 551 552 struct list_head group_node; 553 unsigned char on_rq; 554 unsigned char sched_delayed; 555 unsigned char rel_deadline; 556 unsigned char custom_slice; 557 /* hole */ 558 559 u64 exec_start; 560 u64 sum_exec_runtime; 561 u64 prev_sum_exec_runtime; 562 u64 vruntime; 563 s64 vlag; 564 u64 slice; 565 566 u64 nr_migrations; 567 568 #ifdef CONFIG_FAIR_GROUP_SCHED 569 int depth; 570 struct sched_entity *parent; 571 /* rq on which this entity is (to be) queued: */ 572 struct cfs_rq *cfs_rq; 573 /* rq "owned" by this entity/group: */ 574 struct cfs_rq *my_q; 575 /* cached value of my_q->h_nr_running */ 576 unsigned long runnable_weight; 577 #endif 578 579 #ifdef CONFIG_SMP 580 /* 581 * Per entity load average tracking. 582 * 583 * Put into separate cache line so it does not 584 * collide with read-mostly values above. 585 */ 586 struct sched_avg avg; 587 #endif 588 }; 589 590 struct sched_rt_entity { 591 struct list_head run_list; 592 unsigned long timeout; 593 unsigned long watchdog_stamp; 594 unsigned int time_slice; 595 unsigned short on_rq; 596 unsigned short on_list; 597 598 struct sched_rt_entity *back; 599 #ifdef CONFIG_RT_GROUP_SCHED 600 struct sched_rt_entity *parent; 601 /* rq on which this entity is (to be) queued: */ 602 struct rt_rq *rt_rq; 603 /* rq "owned" by this entity/group: */ 604 struct rt_rq *my_q; 605 #endif 606 } __randomize_layout; 607 608 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *); 609 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *); 610 611 struct sched_dl_entity { 612 struct rb_node rb_node; 613 614 /* 615 * Original scheduling parameters. Copied here from sched_attr 616 * during sched_setattr(), they will remain the same until 617 * the next sched_setattr(). 618 */ 619 u64 dl_runtime; /* Maximum runtime for each instance */ 620 u64 dl_deadline; /* Relative deadline of each instance */ 621 u64 dl_period; /* Separation of two instances (period) */ 622 u64 dl_bw; /* dl_runtime / dl_period */ 623 u64 dl_density; /* dl_runtime / dl_deadline */ 624 625 /* 626 * Actual scheduling parameters. Initialized with the values above, 627 * they are continuously updated during task execution. Note that 628 * the remaining runtime could be < 0 in case we are in overrun. 629 */ 630 s64 runtime; /* Remaining runtime for this instance */ 631 u64 deadline; /* Absolute deadline for this instance */ 632 unsigned int flags; /* Specifying the scheduler behaviour */ 633 634 /* 635 * Some bool flags: 636 * 637 * @dl_throttled tells if we exhausted the runtime. If so, the 638 * task has to wait for a replenishment to be performed at the 639 * next firing of dl_timer. 640 * 641 * @dl_yielded tells if task gave up the CPU before consuming 642 * all its available runtime during the last job. 643 * 644 * @dl_non_contending tells if the task is inactive while still 645 * contributing to the active utilization. In other words, it 646 * indicates if the inactive timer has been armed and its handler 647 * has not been executed yet. This flag is useful to avoid race 648 * conditions between the inactive timer handler and the wakeup 649 * code. 650 * 651 * @dl_overrun tells if the task asked to be informed about runtime 652 * overruns. 653 * 654 * @dl_server tells if this is a server entity. 655 * 656 * @dl_defer tells if this is a deferred or regular server. For 657 * now only defer server exists. 658 * 659 * @dl_defer_armed tells if the deferrable server is waiting 660 * for the replenishment timer to activate it. 661 * 662 * @dl_server_active tells if the dlserver is active(started). 663 * dlserver is started on first cfs enqueue on an idle runqueue 664 * and is stopped when a dequeue results in 0 cfs tasks on the 665 * runqueue. In other words, dlserver is active only when cpu's 666 * runqueue has atleast one cfs task. 667 * 668 * @dl_defer_running tells if the deferrable server is actually 669 * running, skipping the defer phase. 670 */ 671 unsigned int dl_throttled : 1; 672 unsigned int dl_yielded : 1; 673 unsigned int dl_non_contending : 1; 674 unsigned int dl_overrun : 1; 675 unsigned int dl_server : 1; 676 unsigned int dl_server_active : 1; 677 unsigned int dl_defer : 1; 678 unsigned int dl_defer_armed : 1; 679 unsigned int dl_defer_running : 1; 680 681 /* 682 * Bandwidth enforcement timer. Each -deadline task has its 683 * own bandwidth to be enforced, thus we need one timer per task. 684 */ 685 struct hrtimer dl_timer; 686 687 /* 688 * Inactive timer, responsible for decreasing the active utilization 689 * at the "0-lag time". When a -deadline task blocks, it contributes 690 * to GRUB's active utilization until the "0-lag time", hence a 691 * timer is needed to decrease the active utilization at the correct 692 * time. 693 */ 694 struct hrtimer inactive_timer; 695 696 /* 697 * Bits for DL-server functionality. Also see the comment near 698 * dl_server_update(). 699 * 700 * @rq the runqueue this server is for 701 * 702 * @server_has_tasks() returns true if @server_pick return a 703 * runnable task. 704 */ 705 struct rq *rq; 706 dl_server_has_tasks_f server_has_tasks; 707 dl_server_pick_f server_pick_task; 708 709 #ifdef CONFIG_RT_MUTEXES 710 /* 711 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 712 * pi_se points to the donor, otherwise points to the dl_se it belongs 713 * to (the original one/itself). 714 */ 715 struct sched_dl_entity *pi_se; 716 #endif 717 }; 718 719 #ifdef CONFIG_UCLAMP_TASK 720 /* Number of utilization clamp buckets (shorter alias) */ 721 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 722 723 /* 724 * Utilization clamp for a scheduling entity 725 * @value: clamp value "assigned" to a se 726 * @bucket_id: bucket index corresponding to the "assigned" value 727 * @active: the se is currently refcounted in a rq's bucket 728 * @user_defined: the requested clamp value comes from user-space 729 * 730 * The bucket_id is the index of the clamp bucket matching the clamp value 731 * which is pre-computed and stored to avoid expensive integer divisions from 732 * the fast path. 733 * 734 * The active bit is set whenever a task has got an "effective" value assigned, 735 * which can be different from the clamp value "requested" from user-space. 736 * This allows to know a task is refcounted in the rq's bucket corresponding 737 * to the "effective" bucket_id. 738 * 739 * The user_defined bit is set whenever a task has got a task-specific clamp 740 * value requested from userspace, i.e. the system defaults apply to this task 741 * just as a restriction. This allows to relax default clamps when a less 742 * restrictive task-specific value has been requested, thus allowing to 743 * implement a "nice" semantic. For example, a task running with a 20% 744 * default boost can still drop its own boosting to 0%. 745 */ 746 struct uclamp_se { 747 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 748 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 749 unsigned int active : 1; 750 unsigned int user_defined : 1; 751 }; 752 #endif /* CONFIG_UCLAMP_TASK */ 753 754 union rcu_special { 755 struct { 756 u8 blocked; 757 u8 need_qs; 758 u8 exp_hint; /* Hint for performance. */ 759 u8 need_mb; /* Readers need smp_mb(). */ 760 } b; /* Bits. */ 761 u32 s; /* Set of bits. */ 762 }; 763 764 enum perf_event_task_context { 765 perf_invalid_context = -1, 766 perf_hw_context = 0, 767 perf_sw_context, 768 perf_nr_task_contexts, 769 }; 770 771 /* 772 * Number of contexts where an event can trigger: 773 * task, softirq, hardirq, nmi. 774 */ 775 #define PERF_NR_CONTEXTS 4 776 777 struct wake_q_node { 778 struct wake_q_node *next; 779 }; 780 781 struct kmap_ctrl { 782 #ifdef CONFIG_KMAP_LOCAL 783 int idx; 784 pte_t pteval[KM_MAX_IDX]; 785 #endif 786 }; 787 788 struct task_struct { 789 #ifdef CONFIG_THREAD_INFO_IN_TASK 790 /* 791 * For reasons of header soup (see current_thread_info()), this 792 * must be the first element of task_struct. 793 */ 794 struct thread_info thread_info; 795 #endif 796 unsigned int __state; 797 798 /* saved state for "spinlock sleepers" */ 799 unsigned int saved_state; 800 801 /* 802 * This begins the randomizable portion of task_struct. Only 803 * scheduling-critical items should be added above here. 804 */ 805 randomized_struct_fields_start 806 807 void *stack; 808 refcount_t usage; 809 /* Per task flags (PF_*), defined further below: */ 810 unsigned int flags; 811 unsigned int ptrace; 812 813 #ifdef CONFIG_MEM_ALLOC_PROFILING 814 struct alloc_tag *alloc_tag; 815 #endif 816 817 #ifdef CONFIG_SMP 818 int on_cpu; 819 struct __call_single_node wake_entry; 820 unsigned int wakee_flips; 821 unsigned long wakee_flip_decay_ts; 822 struct task_struct *last_wakee; 823 824 /* 825 * recent_used_cpu is initially set as the last CPU used by a task 826 * that wakes affine another task. Waker/wakee relationships can 827 * push tasks around a CPU where each wakeup moves to the next one. 828 * Tracking a recently used CPU allows a quick search for a recently 829 * used CPU that may be idle. 830 */ 831 int recent_used_cpu; 832 int wake_cpu; 833 #endif 834 int on_rq; 835 836 int prio; 837 int static_prio; 838 int normal_prio; 839 unsigned int rt_priority; 840 841 struct sched_entity se; 842 struct sched_rt_entity rt; 843 struct sched_dl_entity dl; 844 struct sched_dl_entity *dl_server; 845 #ifdef CONFIG_SCHED_CLASS_EXT 846 struct sched_ext_entity scx; 847 #endif 848 const struct sched_class *sched_class; 849 850 #ifdef CONFIG_SCHED_CORE 851 struct rb_node core_node; 852 unsigned long core_cookie; 853 unsigned int core_occupation; 854 #endif 855 856 #ifdef CONFIG_CGROUP_SCHED 857 struct task_group *sched_task_group; 858 #endif 859 860 861 #ifdef CONFIG_UCLAMP_TASK 862 /* 863 * Clamp values requested for a scheduling entity. 864 * Must be updated with task_rq_lock() held. 865 */ 866 struct uclamp_se uclamp_req[UCLAMP_CNT]; 867 /* 868 * Effective clamp values used for a scheduling entity. 869 * Must be updated with task_rq_lock() held. 870 */ 871 struct uclamp_se uclamp[UCLAMP_CNT]; 872 #endif 873 874 struct sched_statistics stats; 875 876 #ifdef CONFIG_PREEMPT_NOTIFIERS 877 /* List of struct preempt_notifier: */ 878 struct hlist_head preempt_notifiers; 879 #endif 880 881 #ifdef CONFIG_BLK_DEV_IO_TRACE 882 unsigned int btrace_seq; 883 #endif 884 885 unsigned int policy; 886 unsigned long max_allowed_capacity; 887 int nr_cpus_allowed; 888 const cpumask_t *cpus_ptr; 889 cpumask_t *user_cpus_ptr; 890 cpumask_t cpus_mask; 891 void *migration_pending; 892 #ifdef CONFIG_SMP 893 unsigned short migration_disabled; 894 #endif 895 unsigned short migration_flags; 896 897 #ifdef CONFIG_PREEMPT_RCU 898 int rcu_read_lock_nesting; 899 union rcu_special rcu_read_unlock_special; 900 struct list_head rcu_node_entry; 901 struct rcu_node *rcu_blocked_node; 902 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 903 904 #ifdef CONFIG_TASKS_RCU 905 unsigned long rcu_tasks_nvcsw; 906 u8 rcu_tasks_holdout; 907 u8 rcu_tasks_idx; 908 int rcu_tasks_idle_cpu; 909 struct list_head rcu_tasks_holdout_list; 910 int rcu_tasks_exit_cpu; 911 struct list_head rcu_tasks_exit_list; 912 #endif /* #ifdef CONFIG_TASKS_RCU */ 913 914 #ifdef CONFIG_TASKS_TRACE_RCU 915 int trc_reader_nesting; 916 int trc_ipi_to_cpu; 917 union rcu_special trc_reader_special; 918 struct list_head trc_holdout_list; 919 struct list_head trc_blkd_node; 920 int trc_blkd_cpu; 921 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 922 923 struct sched_info sched_info; 924 925 struct list_head tasks; 926 #ifdef CONFIG_SMP 927 struct plist_node pushable_tasks; 928 struct rb_node pushable_dl_tasks; 929 #endif 930 931 struct mm_struct *mm; 932 struct mm_struct *active_mm; 933 struct address_space *faults_disabled_mapping; 934 935 int exit_state; 936 int exit_code; 937 int exit_signal; 938 /* The signal sent when the parent dies: */ 939 int pdeath_signal; 940 /* JOBCTL_*, siglock protected: */ 941 unsigned long jobctl; 942 943 /* Used for emulating ABI behavior of previous Linux versions: */ 944 unsigned int personality; 945 946 /* Scheduler bits, serialized by scheduler locks: */ 947 unsigned sched_reset_on_fork:1; 948 unsigned sched_contributes_to_load:1; 949 unsigned sched_migrated:1; 950 951 /* Force alignment to the next boundary: */ 952 unsigned :0; 953 954 /* Unserialized, strictly 'current' */ 955 956 /* 957 * This field must not be in the scheduler word above due to wakelist 958 * queueing no longer being serialized by p->on_cpu. However: 959 * 960 * p->XXX = X; ttwu() 961 * schedule() if (p->on_rq && ..) // false 962 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 963 * deactivate_task() ttwu_queue_wakelist()) 964 * p->on_rq = 0; p->sched_remote_wakeup = Y; 965 * 966 * guarantees all stores of 'current' are visible before 967 * ->sched_remote_wakeup gets used, so it can be in this word. 968 */ 969 unsigned sched_remote_wakeup:1; 970 #ifdef CONFIG_RT_MUTEXES 971 unsigned sched_rt_mutex:1; 972 #endif 973 974 /* Bit to tell TOMOYO we're in execve(): */ 975 unsigned in_execve:1; 976 unsigned in_iowait:1; 977 #ifndef TIF_RESTORE_SIGMASK 978 unsigned restore_sigmask:1; 979 #endif 980 #ifdef CONFIG_MEMCG_V1 981 unsigned in_user_fault:1; 982 #endif 983 #ifdef CONFIG_LRU_GEN 984 /* whether the LRU algorithm may apply to this access */ 985 unsigned in_lru_fault:1; 986 #endif 987 #ifdef CONFIG_COMPAT_BRK 988 unsigned brk_randomized:1; 989 #endif 990 #ifdef CONFIG_CGROUPS 991 /* disallow userland-initiated cgroup migration */ 992 unsigned no_cgroup_migration:1; 993 /* task is frozen/stopped (used by the cgroup freezer) */ 994 unsigned frozen:1; 995 #endif 996 #ifdef CONFIG_BLK_CGROUP 997 unsigned use_memdelay:1; 998 #endif 999 #ifdef CONFIG_PSI 1000 /* Stalled due to lack of memory */ 1001 unsigned in_memstall:1; 1002 #endif 1003 #ifdef CONFIG_PAGE_OWNER 1004 /* Used by page_owner=on to detect recursion in page tracking. */ 1005 unsigned in_page_owner:1; 1006 #endif 1007 #ifdef CONFIG_EVENTFD 1008 /* Recursion prevention for eventfd_signal() */ 1009 unsigned in_eventfd:1; 1010 #endif 1011 #ifdef CONFIG_ARCH_HAS_CPU_PASID 1012 unsigned pasid_activated:1; 1013 #endif 1014 #ifdef CONFIG_X86_BUS_LOCK_DETECT 1015 unsigned reported_split_lock:1; 1016 #endif 1017 #ifdef CONFIG_TASK_DELAY_ACCT 1018 /* delay due to memory thrashing */ 1019 unsigned in_thrashing:1; 1020 #endif 1021 #ifdef CONFIG_PREEMPT_RT 1022 struct netdev_xmit net_xmit; 1023 #endif 1024 unsigned long atomic_flags; /* Flags requiring atomic access. */ 1025 1026 struct restart_block restart_block; 1027 1028 pid_t pid; 1029 pid_t tgid; 1030 1031 #ifdef CONFIG_STACKPROTECTOR 1032 /* Canary value for the -fstack-protector GCC feature: */ 1033 unsigned long stack_canary; 1034 #endif 1035 /* 1036 * Pointers to the (original) parent process, youngest child, younger sibling, 1037 * older sibling, respectively. (p->father can be replaced with 1038 * p->real_parent->pid) 1039 */ 1040 1041 /* Real parent process: */ 1042 struct task_struct __rcu *real_parent; 1043 1044 /* Recipient of SIGCHLD, wait4() reports: */ 1045 struct task_struct __rcu *parent; 1046 1047 /* 1048 * Children/sibling form the list of natural children: 1049 */ 1050 struct list_head children; 1051 struct list_head sibling; 1052 struct task_struct *group_leader; 1053 1054 /* 1055 * 'ptraced' is the list of tasks this task is using ptrace() on. 1056 * 1057 * This includes both natural children and PTRACE_ATTACH targets. 1058 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1059 */ 1060 struct list_head ptraced; 1061 struct list_head ptrace_entry; 1062 1063 /* PID/PID hash table linkage. */ 1064 struct pid *thread_pid; 1065 struct hlist_node pid_links[PIDTYPE_MAX]; 1066 struct list_head thread_node; 1067 1068 struct completion *vfork_done; 1069 1070 /* CLONE_CHILD_SETTID: */ 1071 int __user *set_child_tid; 1072 1073 /* CLONE_CHILD_CLEARTID: */ 1074 int __user *clear_child_tid; 1075 1076 /* PF_KTHREAD | PF_IO_WORKER */ 1077 void *worker_private; 1078 1079 u64 utime; 1080 u64 stime; 1081 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1082 u64 utimescaled; 1083 u64 stimescaled; 1084 #endif 1085 u64 gtime; 1086 struct prev_cputime prev_cputime; 1087 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1088 struct vtime vtime; 1089 #endif 1090 1091 #ifdef CONFIG_NO_HZ_FULL 1092 atomic_t tick_dep_mask; 1093 #endif 1094 /* Context switch counts: */ 1095 unsigned long nvcsw; 1096 unsigned long nivcsw; 1097 1098 /* Monotonic time in nsecs: */ 1099 u64 start_time; 1100 1101 /* Boot based time in nsecs: */ 1102 u64 start_boottime; 1103 1104 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1105 unsigned long min_flt; 1106 unsigned long maj_flt; 1107 1108 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1109 struct posix_cputimers posix_cputimers; 1110 1111 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1112 struct posix_cputimers_work posix_cputimers_work; 1113 #endif 1114 1115 /* Process credentials: */ 1116 1117 /* Tracer's credentials at attach: */ 1118 const struct cred __rcu *ptracer_cred; 1119 1120 /* Objective and real subjective task credentials (COW): */ 1121 const struct cred __rcu *real_cred; 1122 1123 /* Effective (overridable) subjective task credentials (COW): */ 1124 const struct cred __rcu *cred; 1125 1126 #ifdef CONFIG_KEYS 1127 /* Cached requested key. */ 1128 struct key *cached_requested_key; 1129 #endif 1130 1131 /* 1132 * executable name, excluding path. 1133 * 1134 * - normally initialized begin_new_exec() 1135 * - set it with set_task_comm() 1136 * - strscpy_pad() to ensure it is always NUL-terminated and 1137 * zero-padded 1138 * - task_lock() to ensure the operation is atomic and the name is 1139 * fully updated. 1140 */ 1141 char comm[TASK_COMM_LEN]; 1142 1143 struct nameidata *nameidata; 1144 1145 #ifdef CONFIG_SYSVIPC 1146 struct sysv_sem sysvsem; 1147 struct sysv_shm sysvshm; 1148 #endif 1149 #ifdef CONFIG_DETECT_HUNG_TASK 1150 unsigned long last_switch_count; 1151 unsigned long last_switch_time; 1152 #endif 1153 /* Filesystem information: */ 1154 struct fs_struct *fs; 1155 1156 /* Open file information: */ 1157 struct files_struct *files; 1158 1159 #ifdef CONFIG_IO_URING 1160 struct io_uring_task *io_uring; 1161 #endif 1162 1163 /* Namespaces: */ 1164 struct nsproxy *nsproxy; 1165 1166 /* Signal handlers: */ 1167 struct signal_struct *signal; 1168 struct sighand_struct __rcu *sighand; 1169 sigset_t blocked; 1170 sigset_t real_blocked; 1171 /* Restored if set_restore_sigmask() was used: */ 1172 sigset_t saved_sigmask; 1173 struct sigpending pending; 1174 unsigned long sas_ss_sp; 1175 size_t sas_ss_size; 1176 unsigned int sas_ss_flags; 1177 1178 struct callback_head *task_works; 1179 1180 #ifdef CONFIG_AUDIT 1181 #ifdef CONFIG_AUDITSYSCALL 1182 struct audit_context *audit_context; 1183 #endif 1184 kuid_t loginuid; 1185 unsigned int sessionid; 1186 #endif 1187 struct seccomp seccomp; 1188 struct syscall_user_dispatch syscall_dispatch; 1189 1190 /* Thread group tracking: */ 1191 u64 parent_exec_id; 1192 u64 self_exec_id; 1193 1194 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1195 spinlock_t alloc_lock; 1196 1197 /* Protection of the PI data structures: */ 1198 raw_spinlock_t pi_lock; 1199 1200 struct wake_q_node wake_q; 1201 1202 #ifdef CONFIG_RT_MUTEXES 1203 /* PI waiters blocked on a rt_mutex held by this task: */ 1204 struct rb_root_cached pi_waiters; 1205 /* Updated under owner's pi_lock and rq lock */ 1206 struct task_struct *pi_top_task; 1207 /* Deadlock detection and priority inheritance handling: */ 1208 struct rt_mutex_waiter *pi_blocked_on; 1209 #endif 1210 1211 #ifdef CONFIG_DEBUG_MUTEXES 1212 /* Mutex deadlock detection: */ 1213 struct mutex_waiter *blocked_on; 1214 #endif 1215 1216 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1217 int non_block_count; 1218 #endif 1219 1220 #ifdef CONFIG_TRACE_IRQFLAGS 1221 struct irqtrace_events irqtrace; 1222 unsigned int hardirq_threaded; 1223 u64 hardirq_chain_key; 1224 int softirqs_enabled; 1225 int softirq_context; 1226 int irq_config; 1227 #endif 1228 #ifdef CONFIG_PREEMPT_RT 1229 int softirq_disable_cnt; 1230 #endif 1231 1232 #ifdef CONFIG_LOCKDEP 1233 # define MAX_LOCK_DEPTH 48UL 1234 u64 curr_chain_key; 1235 int lockdep_depth; 1236 unsigned int lockdep_recursion; 1237 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1238 #endif 1239 1240 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1241 unsigned int in_ubsan; 1242 #endif 1243 1244 /* Journalling filesystem info: */ 1245 void *journal_info; 1246 1247 /* Stacked block device info: */ 1248 struct bio_list *bio_list; 1249 1250 /* Stack plugging: */ 1251 struct blk_plug *plug; 1252 1253 /* VM state: */ 1254 struct reclaim_state *reclaim_state; 1255 1256 struct io_context *io_context; 1257 1258 #ifdef CONFIG_COMPACTION 1259 struct capture_control *capture_control; 1260 #endif 1261 /* Ptrace state: */ 1262 unsigned long ptrace_message; 1263 kernel_siginfo_t *last_siginfo; 1264 1265 struct task_io_accounting ioac; 1266 #ifdef CONFIG_PSI 1267 /* Pressure stall state */ 1268 unsigned int psi_flags; 1269 #endif 1270 #ifdef CONFIG_TASK_XACCT 1271 /* Accumulated RSS usage: */ 1272 u64 acct_rss_mem1; 1273 /* Accumulated virtual memory usage: */ 1274 u64 acct_vm_mem1; 1275 /* stime + utime since last update: */ 1276 u64 acct_timexpd; 1277 #endif 1278 #ifdef CONFIG_CPUSETS 1279 /* Protected by ->alloc_lock: */ 1280 nodemask_t mems_allowed; 1281 /* Sequence number to catch updates: */ 1282 seqcount_spinlock_t mems_allowed_seq; 1283 int cpuset_mem_spread_rotor; 1284 #endif 1285 #ifdef CONFIG_CGROUPS 1286 /* Control Group info protected by css_set_lock: */ 1287 struct css_set __rcu *cgroups; 1288 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1289 struct list_head cg_list; 1290 #endif 1291 #ifdef CONFIG_X86_CPU_RESCTRL 1292 u32 closid; 1293 u32 rmid; 1294 #endif 1295 #ifdef CONFIG_FUTEX 1296 struct robust_list_head __user *robust_list; 1297 #ifdef CONFIG_COMPAT 1298 struct compat_robust_list_head __user *compat_robust_list; 1299 #endif 1300 struct list_head pi_state_list; 1301 struct futex_pi_state *pi_state_cache; 1302 struct mutex futex_exit_mutex; 1303 unsigned int futex_state; 1304 #endif 1305 #ifdef CONFIG_PERF_EVENTS 1306 u8 perf_recursion[PERF_NR_CONTEXTS]; 1307 struct perf_event_context *perf_event_ctxp; 1308 struct mutex perf_event_mutex; 1309 struct list_head perf_event_list; 1310 #endif 1311 #ifdef CONFIG_DEBUG_PREEMPT 1312 unsigned long preempt_disable_ip; 1313 #endif 1314 #ifdef CONFIG_NUMA 1315 /* Protected by alloc_lock: */ 1316 struct mempolicy *mempolicy; 1317 short il_prev; 1318 u8 il_weight; 1319 short pref_node_fork; 1320 #endif 1321 #ifdef CONFIG_NUMA_BALANCING 1322 int numa_scan_seq; 1323 unsigned int numa_scan_period; 1324 unsigned int numa_scan_period_max; 1325 int numa_preferred_nid; 1326 unsigned long numa_migrate_retry; 1327 /* Migration stamp: */ 1328 u64 node_stamp; 1329 u64 last_task_numa_placement; 1330 u64 last_sum_exec_runtime; 1331 struct callback_head numa_work; 1332 1333 /* 1334 * This pointer is only modified for current in syscall and 1335 * pagefault context (and for tasks being destroyed), so it can be read 1336 * from any of the following contexts: 1337 * - RCU read-side critical section 1338 * - current->numa_group from everywhere 1339 * - task's runqueue locked, task not running 1340 */ 1341 struct numa_group __rcu *numa_group; 1342 1343 /* 1344 * numa_faults is an array split into four regions: 1345 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1346 * in this precise order. 1347 * 1348 * faults_memory: Exponential decaying average of faults on a per-node 1349 * basis. Scheduling placement decisions are made based on these 1350 * counts. The values remain static for the duration of a PTE scan. 1351 * faults_cpu: Track the nodes the process was running on when a NUMA 1352 * hinting fault was incurred. 1353 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1354 * during the current scan window. When the scan completes, the counts 1355 * in faults_memory and faults_cpu decay and these values are copied. 1356 */ 1357 unsigned long *numa_faults; 1358 unsigned long total_numa_faults; 1359 1360 /* 1361 * numa_faults_locality tracks if faults recorded during the last 1362 * scan window were remote/local or failed to migrate. The task scan 1363 * period is adapted based on the locality of the faults with different 1364 * weights depending on whether they were shared or private faults 1365 */ 1366 unsigned long numa_faults_locality[3]; 1367 1368 unsigned long numa_pages_migrated; 1369 #endif /* CONFIG_NUMA_BALANCING */ 1370 1371 #ifdef CONFIG_RSEQ 1372 struct rseq __user *rseq; 1373 u32 rseq_len; 1374 u32 rseq_sig; 1375 /* 1376 * RmW on rseq_event_mask must be performed atomically 1377 * with respect to preemption. 1378 */ 1379 unsigned long rseq_event_mask; 1380 #endif 1381 1382 #ifdef CONFIG_SCHED_MM_CID 1383 int mm_cid; /* Current cid in mm */ 1384 int last_mm_cid; /* Most recent cid in mm */ 1385 int migrate_from_cpu; 1386 int mm_cid_active; /* Whether cid bitmap is active */ 1387 struct callback_head cid_work; 1388 #endif 1389 1390 struct tlbflush_unmap_batch tlb_ubc; 1391 1392 /* Cache last used pipe for splice(): */ 1393 struct pipe_inode_info *splice_pipe; 1394 1395 struct page_frag task_frag; 1396 1397 #ifdef CONFIG_TASK_DELAY_ACCT 1398 struct task_delay_info *delays; 1399 #endif 1400 1401 #ifdef CONFIG_FAULT_INJECTION 1402 int make_it_fail; 1403 unsigned int fail_nth; 1404 #endif 1405 /* 1406 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1407 * balance_dirty_pages() for a dirty throttling pause: 1408 */ 1409 int nr_dirtied; 1410 int nr_dirtied_pause; 1411 /* Start of a write-and-pause period: */ 1412 unsigned long dirty_paused_when; 1413 1414 #ifdef CONFIG_LATENCYTOP 1415 int latency_record_count; 1416 struct latency_record latency_record[LT_SAVECOUNT]; 1417 #endif 1418 /* 1419 * Time slack values; these are used to round up poll() and 1420 * select() etc timeout values. These are in nanoseconds. 1421 */ 1422 u64 timer_slack_ns; 1423 u64 default_timer_slack_ns; 1424 1425 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1426 unsigned int kasan_depth; 1427 #endif 1428 1429 #ifdef CONFIG_KCSAN 1430 struct kcsan_ctx kcsan_ctx; 1431 #ifdef CONFIG_TRACE_IRQFLAGS 1432 struct irqtrace_events kcsan_save_irqtrace; 1433 #endif 1434 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1435 int kcsan_stack_depth; 1436 #endif 1437 #endif 1438 1439 #ifdef CONFIG_KMSAN 1440 struct kmsan_ctx kmsan_ctx; 1441 #endif 1442 1443 #if IS_ENABLED(CONFIG_KUNIT) 1444 struct kunit *kunit_test; 1445 #endif 1446 1447 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1448 /* Index of current stored address in ret_stack: */ 1449 int curr_ret_stack; 1450 int curr_ret_depth; 1451 1452 /* Stack of return addresses for return function tracing: */ 1453 unsigned long *ret_stack; 1454 1455 /* Timestamp for last schedule: */ 1456 unsigned long long ftrace_timestamp; 1457 unsigned long long ftrace_sleeptime; 1458 1459 /* 1460 * Number of functions that haven't been traced 1461 * because of depth overrun: 1462 */ 1463 atomic_t trace_overrun; 1464 1465 /* Pause tracing: */ 1466 atomic_t tracing_graph_pause; 1467 #endif 1468 1469 #ifdef CONFIG_TRACING 1470 /* Bitmask and counter of trace recursion: */ 1471 unsigned long trace_recursion; 1472 #endif /* CONFIG_TRACING */ 1473 1474 #ifdef CONFIG_KCOV 1475 /* See kernel/kcov.c for more details. */ 1476 1477 /* Coverage collection mode enabled for this task (0 if disabled): */ 1478 unsigned int kcov_mode; 1479 1480 /* Size of the kcov_area: */ 1481 unsigned int kcov_size; 1482 1483 /* Buffer for coverage collection: */ 1484 void *kcov_area; 1485 1486 /* KCOV descriptor wired with this task or NULL: */ 1487 struct kcov *kcov; 1488 1489 /* KCOV common handle for remote coverage collection: */ 1490 u64 kcov_handle; 1491 1492 /* KCOV sequence number: */ 1493 int kcov_sequence; 1494 1495 /* Collect coverage from softirq context: */ 1496 unsigned int kcov_softirq; 1497 #endif 1498 1499 #ifdef CONFIG_MEMCG_V1 1500 struct mem_cgroup *memcg_in_oom; 1501 #endif 1502 1503 #ifdef CONFIG_MEMCG 1504 /* Number of pages to reclaim on returning to userland: */ 1505 unsigned int memcg_nr_pages_over_high; 1506 1507 /* Used by memcontrol for targeted memcg charge: */ 1508 struct mem_cgroup *active_memcg; 1509 1510 /* Cache for current->cgroups->memcg->objcg lookups: */ 1511 struct obj_cgroup *objcg; 1512 #endif 1513 1514 #ifdef CONFIG_BLK_CGROUP 1515 struct gendisk *throttle_disk; 1516 #endif 1517 1518 #ifdef CONFIG_UPROBES 1519 struct uprobe_task *utask; 1520 #endif 1521 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1522 unsigned int sequential_io; 1523 unsigned int sequential_io_avg; 1524 #endif 1525 struct kmap_ctrl kmap_ctrl; 1526 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1527 unsigned long task_state_change; 1528 # ifdef CONFIG_PREEMPT_RT 1529 unsigned long saved_state_change; 1530 # endif 1531 #endif 1532 struct rcu_head rcu; 1533 refcount_t rcu_users; 1534 int pagefault_disabled; 1535 #ifdef CONFIG_MMU 1536 struct task_struct *oom_reaper_list; 1537 struct timer_list oom_reaper_timer; 1538 #endif 1539 #ifdef CONFIG_VMAP_STACK 1540 struct vm_struct *stack_vm_area; 1541 #endif 1542 #ifdef CONFIG_THREAD_INFO_IN_TASK 1543 /* A live task holds one reference: */ 1544 refcount_t stack_refcount; 1545 #endif 1546 #ifdef CONFIG_LIVEPATCH 1547 int patch_state; 1548 #endif 1549 #ifdef CONFIG_SECURITY 1550 /* Used by LSM modules for access restriction: */ 1551 void *security; 1552 #endif 1553 #ifdef CONFIG_BPF_SYSCALL 1554 /* Used by BPF task local storage */ 1555 struct bpf_local_storage __rcu *bpf_storage; 1556 /* Used for BPF run context */ 1557 struct bpf_run_ctx *bpf_ctx; 1558 #endif 1559 /* Used by BPF for per-TASK xdp storage */ 1560 struct bpf_net_context *bpf_net_context; 1561 1562 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1563 unsigned long lowest_stack; 1564 unsigned long prev_lowest_stack; 1565 #endif 1566 1567 #ifdef CONFIG_X86_MCE 1568 void __user *mce_vaddr; 1569 __u64 mce_kflags; 1570 u64 mce_addr; 1571 __u64 mce_ripv : 1, 1572 mce_whole_page : 1, 1573 __mce_reserved : 62; 1574 struct callback_head mce_kill_me; 1575 int mce_count; 1576 #endif 1577 1578 #ifdef CONFIG_KRETPROBES 1579 struct llist_head kretprobe_instances; 1580 #endif 1581 #ifdef CONFIG_RETHOOK 1582 struct llist_head rethooks; 1583 #endif 1584 1585 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1586 /* 1587 * If L1D flush is supported on mm context switch 1588 * then we use this callback head to queue kill work 1589 * to kill tasks that are not running on SMT disabled 1590 * cores 1591 */ 1592 struct callback_head l1d_flush_kill; 1593 #endif 1594 1595 #ifdef CONFIG_RV 1596 /* 1597 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS. 1598 * If we find justification for more monitors, we can think 1599 * about adding more or developing a dynamic method. So far, 1600 * none of these are justified. 1601 */ 1602 union rv_task_monitor rv[RV_PER_TASK_MONITORS]; 1603 #endif 1604 1605 #ifdef CONFIG_USER_EVENTS 1606 struct user_event_mm *user_event_mm; 1607 #endif 1608 1609 /* 1610 * New fields for task_struct should be added above here, so that 1611 * they are included in the randomized portion of task_struct. 1612 */ 1613 randomized_struct_fields_end 1614 1615 /* CPU-specific state of this task: */ 1616 struct thread_struct thread; 1617 1618 /* 1619 * WARNING: on x86, 'thread_struct' contains a variable-sized 1620 * structure. It *MUST* be at the end of 'task_struct'. 1621 * 1622 * Do not put anything below here! 1623 */ 1624 }; 1625 1626 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1627 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1628 1629 static inline unsigned int __task_state_index(unsigned int tsk_state, 1630 unsigned int tsk_exit_state) 1631 { 1632 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1633 1634 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1635 1636 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1637 state = TASK_REPORT_IDLE; 1638 1639 /* 1640 * We're lying here, but rather than expose a completely new task state 1641 * to userspace, we can make this appear as if the task has gone through 1642 * a regular rt_mutex_lock() call. 1643 * Report frozen tasks as uninterruptible. 1644 */ 1645 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN)) 1646 state = TASK_UNINTERRUPTIBLE; 1647 1648 return fls(state); 1649 } 1650 1651 static inline unsigned int task_state_index(struct task_struct *tsk) 1652 { 1653 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1654 } 1655 1656 static inline char task_index_to_char(unsigned int state) 1657 { 1658 static const char state_char[] = "RSDTtXZPI"; 1659 1660 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1)); 1661 1662 return state_char[state]; 1663 } 1664 1665 static inline char task_state_to_char(struct task_struct *tsk) 1666 { 1667 return task_index_to_char(task_state_index(tsk)); 1668 } 1669 1670 extern struct pid *cad_pid; 1671 1672 /* 1673 * Per process flags 1674 */ 1675 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1676 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1677 #define PF_EXITING 0x00000004 /* Getting shut down */ 1678 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1679 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1680 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1681 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1682 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1683 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1684 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1685 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1686 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1687 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1688 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1689 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1690 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1691 #define PF__HOLE__00010000 0x00010000 1692 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1693 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1694 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1695 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1696 * I am cleaning dirty pages from some other bdi. */ 1697 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1698 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1699 #define PF__HOLE__00800000 0x00800000 1700 #define PF__HOLE__01000000 0x01000000 1701 #define PF__HOLE__02000000 0x02000000 1702 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1703 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1704 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1705 * See memalloc_pin_save() */ 1706 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1707 #define PF__HOLE__40000000 0x40000000 1708 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1709 1710 /* 1711 * Only the _current_ task can read/write to tsk->flags, but other 1712 * tasks can access tsk->flags in readonly mode for example 1713 * with tsk_used_math (like during threaded core dumping). 1714 * There is however an exception to this rule during ptrace 1715 * or during fork: the ptracer task is allowed to write to the 1716 * child->flags of its traced child (same goes for fork, the parent 1717 * can write to the child->flags), because we're guaranteed the 1718 * child is not running and in turn not changing child->flags 1719 * at the same time the parent does it. 1720 */ 1721 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1722 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1723 #define clear_used_math() clear_stopped_child_used_math(current) 1724 #define set_used_math() set_stopped_child_used_math(current) 1725 1726 #define conditional_stopped_child_used_math(condition, child) \ 1727 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1728 1729 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1730 1731 #define copy_to_stopped_child_used_math(child) \ 1732 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1733 1734 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1735 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1736 #define used_math() tsk_used_math(current) 1737 1738 static __always_inline bool is_percpu_thread(void) 1739 { 1740 #ifdef CONFIG_SMP 1741 return (current->flags & PF_NO_SETAFFINITY) && 1742 (current->nr_cpus_allowed == 1); 1743 #else 1744 return true; 1745 #endif 1746 } 1747 1748 /* Per-process atomic flags. */ 1749 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1750 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1751 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1752 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1753 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1754 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1755 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1756 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1757 1758 #define TASK_PFA_TEST(name, func) \ 1759 static inline bool task_##func(struct task_struct *p) \ 1760 { return test_bit(PFA_##name, &p->atomic_flags); } 1761 1762 #define TASK_PFA_SET(name, func) \ 1763 static inline void task_set_##func(struct task_struct *p) \ 1764 { set_bit(PFA_##name, &p->atomic_flags); } 1765 1766 #define TASK_PFA_CLEAR(name, func) \ 1767 static inline void task_clear_##func(struct task_struct *p) \ 1768 { clear_bit(PFA_##name, &p->atomic_flags); } 1769 1770 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1771 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1772 1773 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1774 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1775 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1776 1777 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1778 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1779 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1780 1781 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1782 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1783 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1784 1785 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1786 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1787 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1788 1789 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1790 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1791 1792 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1793 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1794 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1795 1796 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1797 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1798 1799 static inline void 1800 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1801 { 1802 current->flags &= ~flags; 1803 current->flags |= orig_flags & flags; 1804 } 1805 1806 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1807 extern int task_can_attach(struct task_struct *p); 1808 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1809 extern void dl_bw_free(int cpu, u64 dl_bw); 1810 #ifdef CONFIG_SMP 1811 1812 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1813 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1814 1815 /** 1816 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1817 * @p: the task 1818 * @new_mask: CPU affinity mask 1819 * 1820 * Return: zero if successful, or a negative error code 1821 */ 1822 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1823 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1824 extern void release_user_cpus_ptr(struct task_struct *p); 1825 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1826 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1827 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1828 #else 1829 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1830 { 1831 } 1832 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1833 { 1834 /* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */ 1835 if ((*cpumask_bits(new_mask) & 1) == 0) 1836 return -EINVAL; 1837 return 0; 1838 } 1839 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node) 1840 { 1841 if (src->user_cpus_ptr) 1842 return -EINVAL; 1843 return 0; 1844 } 1845 static inline void release_user_cpus_ptr(struct task_struct *p) 1846 { 1847 WARN_ON(p->user_cpus_ptr); 1848 } 1849 1850 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) 1851 { 1852 return 0; 1853 } 1854 #endif 1855 1856 extern int yield_to(struct task_struct *p, bool preempt); 1857 extern void set_user_nice(struct task_struct *p, long nice); 1858 extern int task_prio(const struct task_struct *p); 1859 1860 /** 1861 * task_nice - return the nice value of a given task. 1862 * @p: the task in question. 1863 * 1864 * Return: The nice value [ -20 ... 0 ... 19 ]. 1865 */ 1866 static inline int task_nice(const struct task_struct *p) 1867 { 1868 return PRIO_TO_NICE((p)->static_prio); 1869 } 1870 1871 extern int can_nice(const struct task_struct *p, const int nice); 1872 extern int task_curr(const struct task_struct *p); 1873 extern int idle_cpu(int cpu); 1874 extern int available_idle_cpu(int cpu); 1875 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1876 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1877 extern void sched_set_fifo(struct task_struct *p); 1878 extern void sched_set_fifo_low(struct task_struct *p); 1879 extern void sched_set_normal(struct task_struct *p, int nice); 1880 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1881 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1882 extern struct task_struct *idle_task(int cpu); 1883 1884 /** 1885 * is_idle_task - is the specified task an idle task? 1886 * @p: the task in question. 1887 * 1888 * Return: 1 if @p is an idle task. 0 otherwise. 1889 */ 1890 static __always_inline bool is_idle_task(const struct task_struct *p) 1891 { 1892 return !!(p->flags & PF_IDLE); 1893 } 1894 1895 extern struct task_struct *curr_task(int cpu); 1896 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1897 1898 void yield(void); 1899 1900 union thread_union { 1901 struct task_struct task; 1902 #ifndef CONFIG_THREAD_INFO_IN_TASK 1903 struct thread_info thread_info; 1904 #endif 1905 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1906 }; 1907 1908 #ifndef CONFIG_THREAD_INFO_IN_TASK 1909 extern struct thread_info init_thread_info; 1910 #endif 1911 1912 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1913 1914 #ifdef CONFIG_THREAD_INFO_IN_TASK 1915 # define task_thread_info(task) (&(task)->thread_info) 1916 #else 1917 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1918 #endif 1919 1920 /* 1921 * find a task by one of its numerical ids 1922 * 1923 * find_task_by_pid_ns(): 1924 * finds a task by its pid in the specified namespace 1925 * find_task_by_vpid(): 1926 * finds a task by its virtual pid 1927 * 1928 * see also find_vpid() etc in include/linux/pid.h 1929 */ 1930 1931 extern struct task_struct *find_task_by_vpid(pid_t nr); 1932 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1933 1934 /* 1935 * find a task by its virtual pid and get the task struct 1936 */ 1937 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1938 1939 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1940 extern int wake_up_process(struct task_struct *tsk); 1941 extern void wake_up_new_task(struct task_struct *tsk); 1942 1943 #ifdef CONFIG_SMP 1944 extern void kick_process(struct task_struct *tsk); 1945 #else 1946 static inline void kick_process(struct task_struct *tsk) { } 1947 #endif 1948 1949 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1950 1951 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1952 { 1953 __set_task_comm(tsk, from, false); 1954 } 1955 1956 /* 1957 * - Why not use task_lock()? 1958 * User space can randomly change their names anyway, so locking for readers 1959 * doesn't make sense. For writers, locking is probably necessary, as a race 1960 * condition could lead to long-term mixed results. 1961 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is 1962 * always NUL-terminated and zero-padded. Therefore the race condition between 1963 * reader and writer is not an issue. 1964 * 1965 * - BUILD_BUG_ON() can help prevent the buf from being truncated. 1966 * Since the callers don't perform any return value checks, this safeguard is 1967 * necessary. 1968 */ 1969 #define get_task_comm(buf, tsk) ({ \ 1970 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \ 1971 strscpy_pad(buf, (tsk)->comm); \ 1972 buf; \ 1973 }) 1974 1975 #ifdef CONFIG_SMP 1976 static __always_inline void scheduler_ipi(void) 1977 { 1978 /* 1979 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1980 * TIF_NEED_RESCHED remotely (for the first time) will also send 1981 * this IPI. 1982 */ 1983 preempt_fold_need_resched(); 1984 } 1985 #else 1986 static inline void scheduler_ipi(void) { } 1987 #endif 1988 1989 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 1990 1991 /* 1992 * Set thread flags in other task's structures. 1993 * See asm/thread_info.h for TIF_xxxx flags available: 1994 */ 1995 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1996 { 1997 set_ti_thread_flag(task_thread_info(tsk), flag); 1998 } 1999 2000 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2001 { 2002 clear_ti_thread_flag(task_thread_info(tsk), flag); 2003 } 2004 2005 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 2006 bool value) 2007 { 2008 update_ti_thread_flag(task_thread_info(tsk), flag, value); 2009 } 2010 2011 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2012 { 2013 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2014 } 2015 2016 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2017 { 2018 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2019 } 2020 2021 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2022 { 2023 return test_ti_thread_flag(task_thread_info(tsk), flag); 2024 } 2025 2026 static inline void set_tsk_need_resched(struct task_struct *tsk) 2027 { 2028 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2029 } 2030 2031 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2032 { 2033 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY, 2034 (atomic_long_t *)&task_thread_info(tsk)->flags); 2035 } 2036 2037 static inline int test_tsk_need_resched(struct task_struct *tsk) 2038 { 2039 return unlikely(test_tsk_thread_flag(tsk,TIF_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 void sched_dynamic_klp_enable(void); 2054 void sched_dynamic_klp_disable(void); 2055 2056 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2057 2058 static __always_inline int _cond_resched(void) 2059 { 2060 return static_call_mod(cond_resched)(); 2061 } 2062 2063 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2064 2065 extern int dynamic_cond_resched(void); 2066 2067 static __always_inline int _cond_resched(void) 2068 { 2069 return dynamic_cond_resched(); 2070 } 2071 2072 #else /* !CONFIG_PREEMPTION */ 2073 2074 static inline int _cond_resched(void) 2075 { 2076 klp_sched_try_switch(); 2077 return __cond_resched(); 2078 } 2079 2080 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2081 2082 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2083 2084 static inline int _cond_resched(void) 2085 { 2086 klp_sched_try_switch(); 2087 return 0; 2088 } 2089 2090 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2091 2092 #define cond_resched() ({ \ 2093 __might_resched(__FILE__, __LINE__, 0); \ 2094 _cond_resched(); \ 2095 }) 2096 2097 extern int __cond_resched_lock(spinlock_t *lock); 2098 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2099 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2100 2101 #define MIGHT_RESCHED_RCU_SHIFT 8 2102 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2103 2104 #ifndef CONFIG_PREEMPT_RT 2105 /* 2106 * Non RT kernels have an elevated preempt count due to the held lock, 2107 * but are not allowed to be inside a RCU read side critical section 2108 */ 2109 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2110 #else 2111 /* 2112 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2113 * cond_resched*lock() has to take that into account because it checks for 2114 * preempt_count() and rcu_preempt_depth(). 2115 */ 2116 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2117 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2118 #endif 2119 2120 #define cond_resched_lock(lock) ({ \ 2121 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2122 __cond_resched_lock(lock); \ 2123 }) 2124 2125 #define cond_resched_rwlock_read(lock) ({ \ 2126 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2127 __cond_resched_rwlock_read(lock); \ 2128 }) 2129 2130 #define cond_resched_rwlock_write(lock) ({ \ 2131 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2132 __cond_resched_rwlock_write(lock); \ 2133 }) 2134 2135 static __always_inline bool need_resched(void) 2136 { 2137 return unlikely(tif_need_resched()); 2138 } 2139 2140 /* 2141 * Wrappers for p->thread_info->cpu access. No-op on UP. 2142 */ 2143 #ifdef CONFIG_SMP 2144 2145 static inline unsigned int task_cpu(const struct task_struct *p) 2146 { 2147 return READ_ONCE(task_thread_info(p)->cpu); 2148 } 2149 2150 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2151 2152 #else 2153 2154 static inline unsigned int task_cpu(const struct task_struct *p) 2155 { 2156 return 0; 2157 } 2158 2159 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2160 { 2161 } 2162 2163 #endif /* CONFIG_SMP */ 2164 2165 static inline bool task_is_runnable(struct task_struct *p) 2166 { 2167 return p->on_rq && !p->se.sched_delayed; 2168 } 2169 2170 extern bool sched_task_on_rq(struct task_struct *p); 2171 extern unsigned long get_wchan(struct task_struct *p); 2172 extern struct task_struct *cpu_curr_snapshot(int cpu); 2173 2174 #include <linux/spinlock.h> 2175 2176 /* 2177 * In order to reduce various lock holder preemption latencies provide an 2178 * interface to see if a vCPU is currently running or not. 2179 * 2180 * This allows us to terminate optimistic spin loops and block, analogous to 2181 * the native optimistic spin heuristic of testing if the lock owner task is 2182 * running or not. 2183 */ 2184 #ifndef vcpu_is_preempted 2185 static inline bool vcpu_is_preempted(int cpu) 2186 { 2187 return false; 2188 } 2189 #endif 2190 2191 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2192 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2193 2194 #ifndef TASK_SIZE_OF 2195 #define TASK_SIZE_OF(tsk) TASK_SIZE 2196 #endif 2197 2198 #ifdef CONFIG_SMP 2199 static inline bool owner_on_cpu(struct task_struct *owner) 2200 { 2201 /* 2202 * As lock holder preemption issue, we both skip spinning if 2203 * task is not on cpu or its cpu is preempted 2204 */ 2205 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2206 } 2207 2208 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2209 unsigned long sched_cpu_util(int cpu); 2210 #endif /* CONFIG_SMP */ 2211 2212 #ifdef CONFIG_SCHED_CORE 2213 extern void sched_core_free(struct task_struct *tsk); 2214 extern void sched_core_fork(struct task_struct *p); 2215 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2216 unsigned long uaddr); 2217 extern int sched_core_idle_cpu(int cpu); 2218 #else 2219 static inline void sched_core_free(struct task_struct *tsk) { } 2220 static inline void sched_core_fork(struct task_struct *p) { } 2221 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2222 #endif 2223 2224 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2225 2226 #ifdef CONFIG_MEM_ALLOC_PROFILING 2227 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2228 { 2229 swap(current->alloc_tag, tag); 2230 return tag; 2231 } 2232 2233 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2234 { 2235 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2236 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2237 #endif 2238 current->alloc_tag = old; 2239 } 2240 #else 2241 #define alloc_tag_save(_tag) NULL 2242 #define alloc_tag_restore(_tag, _old) do {} while (0) 2243 #endif 2244 2245 #endif 2246