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