1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_SCHED_H 3 #define _LINUX_SCHED_H 4 5 /* 6 * Define 'struct task_struct' and provide the main scheduler 7 * APIs (schedule(), wakeup variants, etc.) 8 */ 9 10 #include <uapi/linux/sched.h> 11 12 #include <asm/current.h> 13 #include <asm/processor.h> 14 #include <linux/thread_info.h> 15 #include <linux/preempt.h> 16 #include <linux/cpumask_types.h> 17 18 #include <linux/cache.h> 19 #include <linux/irqflags_types.h> 20 #include <linux/smp_types.h> 21 #include <linux/pid_types.h> 22 #include <linux/sem_types.h> 23 #include <linux/shm.h> 24 #include <linux/kmsan_types.h> 25 #include <linux/mutex_types.h> 26 #include <linux/plist_types.h> 27 #include <linux/hrtimer_types.h> 28 #include <linux/timer_types.h> 29 #include <linux/seccomp_types.h> 30 #include <linux/nodemask_types.h> 31 #include <linux/refcount_types.h> 32 #include <linux/resource.h> 33 #include <linux/latencytop.h> 34 #include <linux/sched/prio.h> 35 #include <linux/sched/types.h> 36 #include <linux/signal_types.h> 37 #include <linux/syscall_user_dispatch_types.h> 38 #include <linux/mm_types_task.h> 39 #include <linux/netdevice_xmit.h> 40 #include <linux/task_io_accounting.h> 41 #include <linux/posix-timers_types.h> 42 #include <linux/restart_block.h> 43 #include <uapi/linux/rseq.h> 44 #include <linux/seqlock_types.h> 45 #include <linux/kcsan.h> 46 #include <linux/rv.h> 47 #include <linux/uidgid_types.h> 48 #include <linux/tracepoint-defs.h> 49 #include <asm/kmap_size.h> 50 51 /* task_struct member predeclarations (sorted alphabetically): */ 52 struct audit_context; 53 struct bio_list; 54 struct blk_plug; 55 struct bpf_local_storage; 56 struct bpf_run_ctx; 57 struct bpf_net_context; 58 struct capture_control; 59 struct cfs_rq; 60 struct fs_struct; 61 struct futex_pi_state; 62 struct io_context; 63 struct io_uring_task; 64 struct mempolicy; 65 struct nameidata; 66 struct nsproxy; 67 struct perf_event_context; 68 struct perf_ctx_data; 69 struct pid_namespace; 70 struct pipe_inode_info; 71 struct rcu_node; 72 struct reclaim_state; 73 struct robust_list_head; 74 struct root_domain; 75 struct rq; 76 struct sched_attr; 77 struct sched_dl_entity; 78 struct seq_file; 79 struct sighand_struct; 80 struct signal_struct; 81 struct task_delay_info; 82 struct task_group; 83 struct task_struct; 84 struct user_event_mm; 85 86 #include <linux/sched/ext.h> 87 88 /* 89 * Task state bitmask. NOTE! These bits are also 90 * encoded in fs/proc/array.c: get_task_state(). 91 * 92 * We have two separate sets of flags: task->__state 93 * is about runnability, while task->exit_state are 94 * about the task exiting. Confusing, but this way 95 * modifying one set can't modify the other one by 96 * mistake. 97 */ 98 99 /* Used in tsk->__state: */ 100 #define TASK_RUNNING 0x00000000 101 #define TASK_INTERRUPTIBLE 0x00000001 102 #define TASK_UNINTERRUPTIBLE 0x00000002 103 #define __TASK_STOPPED 0x00000004 104 #define __TASK_TRACED 0x00000008 105 /* Used in tsk->exit_state: */ 106 #define EXIT_DEAD 0x00000010 107 #define EXIT_ZOMBIE 0x00000020 108 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 109 /* Used in tsk->__state again: */ 110 #define TASK_PARKED 0x00000040 111 #define TASK_DEAD 0x00000080 112 #define TASK_WAKEKILL 0x00000100 113 #define TASK_WAKING 0x00000200 114 #define TASK_NOLOAD 0x00000400 115 #define TASK_NEW 0x00000800 116 #define TASK_RTLOCK_WAIT 0x00001000 117 #define TASK_FREEZABLE 0x00002000 118 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP)) 119 #define TASK_FROZEN 0x00008000 120 #define TASK_STATE_MAX 0x00010000 121 122 #define TASK_ANY (TASK_STATE_MAX-1) 123 124 /* 125 * DO NOT ADD ANY NEW USERS ! 126 */ 127 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE) 128 129 /* Convenience macros for the sake of set_current_state: */ 130 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 131 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 132 #define TASK_TRACED __TASK_TRACED 133 134 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 135 136 /* Convenience macros for the sake of wake_up(): */ 137 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 138 139 /* get_task_state(): */ 140 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 141 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 142 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 143 TASK_PARKED) 144 145 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 146 147 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 148 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 149 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 150 151 /* 152 * Special states are those that do not use the normal wait-loop pattern. See 153 * the comment with set_special_state(). 154 */ 155 #define is_special_task_state(state) \ 156 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \ 157 TASK_DEAD | TASK_FROZEN)) 158 159 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 160 # define debug_normal_state_change(state_value) \ 161 do { \ 162 WARN_ON_ONCE(is_special_task_state(state_value)); \ 163 current->task_state_change = _THIS_IP_; \ 164 } while (0) 165 166 # define debug_special_state_change(state_value) \ 167 do { \ 168 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 169 current->task_state_change = _THIS_IP_; \ 170 } while (0) 171 172 # define debug_rtlock_wait_set_state() \ 173 do { \ 174 current->saved_state_change = current->task_state_change;\ 175 current->task_state_change = _THIS_IP_; \ 176 } while (0) 177 178 # define debug_rtlock_wait_restore_state() \ 179 do { \ 180 current->task_state_change = current->saved_state_change;\ 181 } while (0) 182 183 #else 184 # define debug_normal_state_change(cond) do { } while (0) 185 # define debug_special_state_change(cond) do { } while (0) 186 # define debug_rtlock_wait_set_state() do { } while (0) 187 # define debug_rtlock_wait_restore_state() do { } while (0) 188 #endif 189 190 #define trace_set_current_state(state_value) \ 191 do { \ 192 if (tracepoint_enabled(sched_set_state_tp)) \ 193 __trace_set_current_state(state_value); \ 194 } while (0) 195 196 /* 197 * set_current_state() includes a barrier so that the write of current->__state 198 * is correctly serialised wrt the caller's subsequent test of whether to 199 * actually sleep: 200 * 201 * for (;;) { 202 * set_current_state(TASK_UNINTERRUPTIBLE); 203 * if (CONDITION) 204 * break; 205 * 206 * schedule(); 207 * } 208 * __set_current_state(TASK_RUNNING); 209 * 210 * If the caller does not need such serialisation (because, for instance, the 211 * CONDITION test and condition change and wakeup are under the same lock) then 212 * use __set_current_state(). 213 * 214 * The above is typically ordered against the wakeup, which does: 215 * 216 * CONDITION = 1; 217 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 218 * 219 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 220 * accessing p->__state. 221 * 222 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is, 223 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 224 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 225 * 226 * However, with slightly different timing the wakeup TASK_RUNNING store can 227 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 228 * a problem either because that will result in one extra go around the loop 229 * and our @cond test will save the day. 230 * 231 * Also see the comments of try_to_wake_up(). 232 */ 233 #define __set_current_state(state_value) \ 234 do { \ 235 debug_normal_state_change((state_value)); \ 236 trace_set_current_state(state_value); \ 237 WRITE_ONCE(current->__state, (state_value)); \ 238 } while (0) 239 240 #define set_current_state(state_value) \ 241 do { \ 242 debug_normal_state_change((state_value)); \ 243 trace_set_current_state(state_value); \ 244 smp_store_mb(current->__state, (state_value)); \ 245 } while (0) 246 247 /* 248 * set_special_state() should be used for those states when the blocking task 249 * can not use the regular condition based wait-loop. In that case we must 250 * serialize against wakeups such that any possible in-flight TASK_RUNNING 251 * stores will not collide with our state change. 252 */ 253 #define set_special_state(state_value) \ 254 do { \ 255 unsigned long flags; /* may shadow */ \ 256 \ 257 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 258 debug_special_state_change((state_value)); \ 259 trace_set_current_state(state_value); \ 260 WRITE_ONCE(current->__state, (state_value)); \ 261 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 262 } while (0) 263 264 /* 265 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 266 * 267 * RT's spin/rwlock substitutions are state preserving. The state of the 268 * task when blocking on the lock is saved in task_struct::saved_state and 269 * restored after the lock has been acquired. These operations are 270 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 271 * lock related wakeups while the task is blocked on the lock are 272 * redirected to operate on task_struct::saved_state to ensure that these 273 * are not dropped. On restore task_struct::saved_state is set to 274 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 275 * 276 * The lock operation looks like this: 277 * 278 * current_save_and_set_rtlock_wait_state(); 279 * for (;;) { 280 * if (try_lock()) 281 * break; 282 * raw_spin_unlock_irq(&lock->wait_lock); 283 * schedule_rtlock(); 284 * raw_spin_lock_irq(&lock->wait_lock); 285 * set_current_state(TASK_RTLOCK_WAIT); 286 * } 287 * current_restore_rtlock_saved_state(); 288 */ 289 #define current_save_and_set_rtlock_wait_state() \ 290 do { \ 291 lockdep_assert_irqs_disabled(); \ 292 raw_spin_lock(¤t->pi_lock); \ 293 current->saved_state = current->__state; \ 294 debug_rtlock_wait_set_state(); \ 295 trace_set_current_state(TASK_RTLOCK_WAIT); \ 296 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 297 raw_spin_unlock(¤t->pi_lock); \ 298 } while (0); 299 300 #define current_restore_rtlock_saved_state() \ 301 do { \ 302 lockdep_assert_irqs_disabled(); \ 303 raw_spin_lock(¤t->pi_lock); \ 304 debug_rtlock_wait_restore_state(); \ 305 trace_set_current_state(current->saved_state); \ 306 WRITE_ONCE(current->__state, current->saved_state); \ 307 current->saved_state = TASK_RUNNING; \ 308 raw_spin_unlock(¤t->pi_lock); \ 309 } while (0); 310 311 #define get_current_state() READ_ONCE(current->__state) 312 313 /* 314 * Define the task command name length as enum, then it can be visible to 315 * BPF programs. 316 */ 317 enum { 318 TASK_COMM_LEN = 16, 319 }; 320 321 extern void sched_tick(void); 322 323 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 324 325 extern long schedule_timeout(long timeout); 326 extern long schedule_timeout_interruptible(long timeout); 327 extern long schedule_timeout_killable(long timeout); 328 extern long schedule_timeout_uninterruptible(long timeout); 329 extern long schedule_timeout_idle(long timeout); 330 asmlinkage void schedule(void); 331 extern void schedule_preempt_disabled(void); 332 asmlinkage void preempt_schedule_irq(void); 333 #ifdef CONFIG_PREEMPT_RT 334 extern void schedule_rtlock(void); 335 #endif 336 337 extern int __must_check io_schedule_prepare(void); 338 extern void io_schedule_finish(int token); 339 extern long io_schedule_timeout(long timeout); 340 extern void io_schedule(void); 341 342 /* wrapper function to trace from this header file */ 343 DECLARE_TRACEPOINT(sched_set_state_tp); 344 extern void __trace_set_current_state(int state_value); 345 346 /** 347 * struct prev_cputime - snapshot of system and user cputime 348 * @utime: time spent in user mode 349 * @stime: time spent in system mode 350 * @lock: protects the above two fields 351 * 352 * Stores previous user/system time values such that we can guarantee 353 * monotonicity. 354 */ 355 struct prev_cputime { 356 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 357 u64 utime; 358 u64 stime; 359 raw_spinlock_t lock; 360 #endif 361 }; 362 363 enum vtime_state { 364 /* Task is sleeping or running in a CPU with VTIME inactive: */ 365 VTIME_INACTIVE = 0, 366 /* Task is idle */ 367 VTIME_IDLE, 368 /* Task runs in kernelspace in a CPU with VTIME active: */ 369 VTIME_SYS, 370 /* Task runs in userspace in a CPU with VTIME active: */ 371 VTIME_USER, 372 /* Task runs as guests in a CPU with VTIME active: */ 373 VTIME_GUEST, 374 }; 375 376 struct vtime { 377 seqcount_t seqcount; 378 unsigned long long starttime; 379 enum vtime_state state; 380 unsigned int cpu; 381 u64 utime; 382 u64 stime; 383 u64 gtime; 384 }; 385 386 /* 387 * Utilization clamp constraints. 388 * @UCLAMP_MIN: Minimum utilization 389 * @UCLAMP_MAX: Maximum utilization 390 * @UCLAMP_CNT: Utilization clamp constraints count 391 */ 392 enum uclamp_id { 393 UCLAMP_MIN = 0, 394 UCLAMP_MAX, 395 UCLAMP_CNT 396 }; 397 398 extern struct root_domain def_root_domain; 399 extern struct mutex sched_domains_mutex; 400 extern void sched_domains_mutex_lock(void); 401 extern void sched_domains_mutex_unlock(void); 402 403 struct sched_param { 404 int sched_priority; 405 }; 406 407 struct sched_info { 408 #ifdef CONFIG_SCHED_INFO 409 /* Cumulative counters: */ 410 411 /* # of times we have run on this CPU: */ 412 unsigned long pcount; 413 414 /* Time spent waiting on a runqueue: */ 415 unsigned long long run_delay; 416 417 /* Max time spent waiting on a runqueue: */ 418 unsigned long long max_run_delay; 419 420 /* Min time spent waiting on a runqueue: */ 421 unsigned long long min_run_delay; 422 423 /* Timestamps: */ 424 425 /* When did we last run on a CPU? */ 426 unsigned long long last_arrival; 427 428 /* When were we last queued to run? */ 429 unsigned long long last_queued; 430 431 #endif /* CONFIG_SCHED_INFO */ 432 }; 433 434 /* 435 * Integer metrics need fixed point arithmetic, e.g., sched/fair 436 * has a few: load, load_avg, util_avg, freq, and capacity. 437 * 438 * We define a basic fixed point arithmetic range, and then formalize 439 * all these metrics based on that basic range. 440 */ 441 # define SCHED_FIXEDPOINT_SHIFT 10 442 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 443 444 /* Increase resolution of cpu_capacity calculations */ 445 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 446 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 447 448 struct load_weight { 449 unsigned long weight; 450 u32 inv_weight; 451 }; 452 453 /* 454 * The load/runnable/util_avg accumulates an infinite geometric series 455 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 456 * 457 * [load_avg definition] 458 * 459 * load_avg = runnable% * scale_load_down(load) 460 * 461 * [runnable_avg definition] 462 * 463 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 464 * 465 * [util_avg definition] 466 * 467 * util_avg = running% * SCHED_CAPACITY_SCALE 468 * 469 * where runnable% is the time ratio that a sched_entity is runnable and 470 * running% the time ratio that a sched_entity is running. 471 * 472 * For cfs_rq, they are the aggregated values of all runnable and blocked 473 * sched_entities. 474 * 475 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 476 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 477 * for computing those signals (see update_rq_clock_pelt()) 478 * 479 * N.B., the above ratios (runnable% and running%) themselves are in the 480 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 481 * to as large a range as necessary. This is for example reflected by 482 * util_avg's SCHED_CAPACITY_SCALE. 483 * 484 * [Overflow issue] 485 * 486 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 487 * with the highest load (=88761), always runnable on a single cfs_rq, 488 * and should not overflow as the number already hits PID_MAX_LIMIT. 489 * 490 * For all other cases (including 32-bit kernels), struct load_weight's 491 * weight will overflow first before we do, because: 492 * 493 * Max(load_avg) <= Max(load.weight) 494 * 495 * Then it is the load_weight's responsibility to consider overflow 496 * issues. 497 */ 498 struct sched_avg { 499 u64 last_update_time; 500 u64 load_sum; 501 u64 runnable_sum; 502 u32 util_sum; 503 u32 period_contrib; 504 unsigned long load_avg; 505 unsigned long runnable_avg; 506 unsigned long util_avg; 507 unsigned int util_est; 508 } ____cacheline_aligned; 509 510 /* 511 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 512 * updates. When a task is dequeued, its util_est should not be updated if its 513 * util_avg has not been updated in the meantime. 514 * This information is mapped into the MSB bit of util_est at dequeue time. 515 * Since max value of util_est for a task is 1024 (PELT util_avg for a task) 516 * it is safe to use MSB. 517 */ 518 #define UTIL_EST_WEIGHT_SHIFT 2 519 #define UTIL_AVG_UNCHANGED 0x80000000 520 521 struct sched_statistics { 522 #ifdef CONFIG_SCHEDSTATS 523 u64 wait_start; 524 u64 wait_max; 525 u64 wait_count; 526 u64 wait_sum; 527 u64 iowait_count; 528 u64 iowait_sum; 529 530 u64 sleep_start; 531 u64 sleep_max; 532 s64 sum_sleep_runtime; 533 534 u64 block_start; 535 u64 block_max; 536 s64 sum_block_runtime; 537 538 s64 exec_max; 539 u64 slice_max; 540 541 u64 nr_migrations_cold; 542 u64 nr_failed_migrations_affine; 543 u64 nr_failed_migrations_running; 544 u64 nr_failed_migrations_hot; 545 u64 nr_forced_migrations; 546 #ifdef CONFIG_NUMA_BALANCING 547 u64 numa_task_migrated; 548 u64 numa_task_swapped; 549 #endif 550 551 u64 nr_wakeups; 552 u64 nr_wakeups_sync; 553 u64 nr_wakeups_migrate; 554 u64 nr_wakeups_local; 555 u64 nr_wakeups_remote; 556 u64 nr_wakeups_affine; 557 u64 nr_wakeups_affine_attempts; 558 u64 nr_wakeups_passive; 559 u64 nr_wakeups_idle; 560 561 #ifdef CONFIG_SCHED_CORE 562 u64 core_forceidle_sum; 563 #endif 564 #endif /* CONFIG_SCHEDSTATS */ 565 } ____cacheline_aligned; 566 567 struct sched_entity { 568 /* For load-balancing: */ 569 struct load_weight load; 570 struct rb_node run_node; 571 u64 deadline; 572 u64 min_vruntime; 573 u64 min_slice; 574 575 struct list_head group_node; 576 unsigned char on_rq; 577 unsigned char sched_delayed; 578 unsigned char rel_deadline; 579 unsigned char custom_slice; 580 /* hole */ 581 582 u64 exec_start; 583 u64 sum_exec_runtime; 584 u64 prev_sum_exec_runtime; 585 u64 vruntime; 586 s64 vlag; 587 u64 slice; 588 589 u64 nr_migrations; 590 591 #ifdef CONFIG_FAIR_GROUP_SCHED 592 int depth; 593 struct sched_entity *parent; 594 /* rq on which this entity is (to be) queued: */ 595 struct cfs_rq *cfs_rq; 596 /* rq "owned" by this entity/group: */ 597 struct cfs_rq *my_q; 598 /* cached value of my_q->h_nr_running */ 599 unsigned long runnable_weight; 600 #endif 601 602 /* 603 * Per entity load average tracking. 604 * 605 * Put into separate cache line so it does not 606 * collide with read-mostly values above. 607 */ 608 struct sched_avg avg; 609 }; 610 611 struct sched_rt_entity { 612 struct list_head run_list; 613 unsigned long timeout; 614 unsigned long watchdog_stamp; 615 unsigned int time_slice; 616 unsigned short on_rq; 617 unsigned short on_list; 618 619 struct sched_rt_entity *back; 620 #ifdef CONFIG_RT_GROUP_SCHED 621 struct sched_rt_entity *parent; 622 /* rq on which this entity is (to be) queued: */ 623 struct rt_rq *rt_rq; 624 /* rq "owned" by this entity/group: */ 625 struct rt_rq *my_q; 626 #endif 627 } __randomize_layout; 628 629 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *); 630 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *); 631 632 struct sched_dl_entity { 633 struct rb_node rb_node; 634 635 /* 636 * Original scheduling parameters. Copied here from sched_attr 637 * during sched_setattr(), they will remain the same until 638 * the next sched_setattr(). 639 */ 640 u64 dl_runtime; /* Maximum runtime for each instance */ 641 u64 dl_deadline; /* Relative deadline of each instance */ 642 u64 dl_period; /* Separation of two instances (period) */ 643 u64 dl_bw; /* dl_runtime / dl_period */ 644 u64 dl_density; /* dl_runtime / dl_deadline */ 645 646 /* 647 * Actual scheduling parameters. Initialized with the values above, 648 * they are continuously updated during task execution. Note that 649 * the remaining runtime could be < 0 in case we are in overrun. 650 */ 651 s64 runtime; /* Remaining runtime for this instance */ 652 u64 deadline; /* Absolute deadline for this instance */ 653 unsigned int flags; /* Specifying the scheduler behaviour */ 654 655 /* 656 * Some bool flags: 657 * 658 * @dl_throttled tells if we exhausted the runtime. If so, the 659 * task has to wait for a replenishment to be performed at the 660 * next firing of dl_timer. 661 * 662 * @dl_yielded tells if task gave up the CPU before consuming 663 * all its available runtime during the last job. 664 * 665 * @dl_non_contending tells if the task is inactive while still 666 * contributing to the active utilization. In other words, it 667 * indicates if the inactive timer has been armed and its handler 668 * has not been executed yet. This flag is useful to avoid race 669 * conditions between the inactive timer handler and the wakeup 670 * code. 671 * 672 * @dl_overrun tells if the task asked to be informed about runtime 673 * overruns. 674 * 675 * @dl_server tells if this is a server entity. 676 * 677 * @dl_defer tells if this is a deferred or regular server. For 678 * now only defer server exists. 679 * 680 * @dl_defer_armed tells if the deferrable server is waiting 681 * for the replenishment timer to activate it. 682 * 683 * @dl_server_active tells if the dlserver is active(started). 684 * dlserver is started on first cfs enqueue on an idle runqueue 685 * and is stopped when a dequeue results in 0 cfs tasks on the 686 * runqueue. In other words, dlserver is active only when cpu's 687 * runqueue has atleast one cfs task. 688 * 689 * @dl_defer_running tells if the deferrable server is actually 690 * running, skipping the defer phase. 691 */ 692 unsigned int dl_throttled : 1; 693 unsigned int dl_yielded : 1; 694 unsigned int dl_non_contending : 1; 695 unsigned int dl_overrun : 1; 696 unsigned int dl_server : 1; 697 unsigned int dl_server_active : 1; 698 unsigned int dl_defer : 1; 699 unsigned int dl_defer_armed : 1; 700 unsigned int dl_defer_running : 1; 701 702 /* 703 * Bandwidth enforcement timer. Each -deadline task has its 704 * own bandwidth to be enforced, thus we need one timer per task. 705 */ 706 struct hrtimer dl_timer; 707 708 /* 709 * Inactive timer, responsible for decreasing the active utilization 710 * at the "0-lag time". When a -deadline task blocks, it contributes 711 * to GRUB's active utilization until the "0-lag time", hence a 712 * timer is needed to decrease the active utilization at the correct 713 * time. 714 */ 715 struct hrtimer inactive_timer; 716 717 /* 718 * Bits for DL-server functionality. Also see the comment near 719 * dl_server_update(). 720 * 721 * @rq the runqueue this server is for 722 * 723 * @server_has_tasks() returns true if @server_pick return a 724 * runnable task. 725 */ 726 struct rq *rq; 727 dl_server_has_tasks_f server_has_tasks; 728 dl_server_pick_f server_pick_task; 729 730 #ifdef CONFIG_RT_MUTEXES 731 /* 732 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 733 * pi_se points to the donor, otherwise points to the dl_se it belongs 734 * to (the original one/itself). 735 */ 736 struct sched_dl_entity *pi_se; 737 #endif 738 }; 739 740 #ifdef CONFIG_UCLAMP_TASK 741 /* Number of utilization clamp buckets (shorter alias) */ 742 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 743 744 /* 745 * Utilization clamp for a scheduling entity 746 * @value: clamp value "assigned" to a se 747 * @bucket_id: bucket index corresponding to the "assigned" value 748 * @active: the se is currently refcounted in a rq's bucket 749 * @user_defined: the requested clamp value comes from user-space 750 * 751 * The bucket_id is the index of the clamp bucket matching the clamp value 752 * which is pre-computed and stored to avoid expensive integer divisions from 753 * the fast path. 754 * 755 * The active bit is set whenever a task has got an "effective" value assigned, 756 * which can be different from the clamp value "requested" from user-space. 757 * This allows to know a task is refcounted in the rq's bucket corresponding 758 * to the "effective" bucket_id. 759 * 760 * The user_defined bit is set whenever a task has got a task-specific clamp 761 * value requested from userspace, i.e. the system defaults apply to this task 762 * just as a restriction. This allows to relax default clamps when a less 763 * restrictive task-specific value has been requested, thus allowing to 764 * implement a "nice" semantic. For example, a task running with a 20% 765 * default boost can still drop its own boosting to 0%. 766 */ 767 struct uclamp_se { 768 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 769 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 770 unsigned int active : 1; 771 unsigned int user_defined : 1; 772 }; 773 #endif /* CONFIG_UCLAMP_TASK */ 774 775 union rcu_special { 776 struct { 777 u8 blocked; 778 u8 need_qs; 779 u8 exp_hint; /* Hint for performance. */ 780 u8 need_mb; /* Readers need smp_mb(). */ 781 } b; /* Bits. */ 782 u32 s; /* Set of bits. */ 783 }; 784 785 enum perf_event_task_context { 786 perf_invalid_context = -1, 787 perf_hw_context = 0, 788 perf_sw_context, 789 perf_nr_task_contexts, 790 }; 791 792 /* 793 * Number of contexts where an event can trigger: 794 * task, softirq, hardirq, nmi. 795 */ 796 #define PERF_NR_CONTEXTS 4 797 798 struct wake_q_node { 799 struct wake_q_node *next; 800 }; 801 802 struct kmap_ctrl { 803 #ifdef CONFIG_KMAP_LOCAL 804 int idx; 805 pte_t pteval[KM_MAX_IDX]; 806 #endif 807 }; 808 809 struct task_struct { 810 #ifdef CONFIG_THREAD_INFO_IN_TASK 811 /* 812 * For reasons of header soup (see current_thread_info()), this 813 * must be the first element of task_struct. 814 */ 815 struct thread_info thread_info; 816 #endif 817 unsigned int __state; 818 819 /* saved state for "spinlock sleepers" */ 820 unsigned int saved_state; 821 822 /* 823 * This begins the randomizable portion of task_struct. Only 824 * scheduling-critical items should be added above here. 825 */ 826 randomized_struct_fields_start 827 828 void *stack; 829 refcount_t usage; 830 /* Per task flags (PF_*), defined further below: */ 831 unsigned int flags; 832 unsigned int ptrace; 833 834 #ifdef CONFIG_MEM_ALLOC_PROFILING 835 struct alloc_tag *alloc_tag; 836 #endif 837 838 int on_cpu; 839 struct __call_single_node wake_entry; 840 unsigned int wakee_flips; 841 unsigned long wakee_flip_decay_ts; 842 struct task_struct *last_wakee; 843 844 /* 845 * recent_used_cpu is initially set as the last CPU used by a task 846 * that wakes affine another task. Waker/wakee relationships can 847 * push tasks around a CPU where each wakeup moves to the next one. 848 * Tracking a recently used CPU allows a quick search for a recently 849 * used CPU that may be idle. 850 */ 851 int recent_used_cpu; 852 int wake_cpu; 853 int on_rq; 854 855 int prio; 856 int static_prio; 857 int normal_prio; 858 unsigned int rt_priority; 859 860 struct sched_entity se; 861 struct sched_rt_entity rt; 862 struct sched_dl_entity dl; 863 struct sched_dl_entity *dl_server; 864 #ifdef CONFIG_SCHED_CLASS_EXT 865 struct sched_ext_entity scx; 866 #endif 867 const struct sched_class *sched_class; 868 869 #ifdef CONFIG_SCHED_CORE 870 struct rb_node core_node; 871 unsigned long core_cookie; 872 unsigned int core_occupation; 873 #endif 874 875 #ifdef CONFIG_CGROUP_SCHED 876 struct task_group *sched_task_group; 877 #endif 878 879 880 #ifdef CONFIG_UCLAMP_TASK 881 /* 882 * Clamp values requested for a scheduling entity. 883 * Must be updated with task_rq_lock() held. 884 */ 885 struct uclamp_se uclamp_req[UCLAMP_CNT]; 886 /* 887 * Effective clamp values used for a scheduling entity. 888 * Must be updated with task_rq_lock() held. 889 */ 890 struct uclamp_se uclamp[UCLAMP_CNT]; 891 #endif 892 893 struct sched_statistics stats; 894 895 #ifdef CONFIG_PREEMPT_NOTIFIERS 896 /* List of struct preempt_notifier: */ 897 struct hlist_head preempt_notifiers; 898 #endif 899 900 #ifdef CONFIG_BLK_DEV_IO_TRACE 901 unsigned int btrace_seq; 902 #endif 903 904 unsigned int policy; 905 unsigned long max_allowed_capacity; 906 int nr_cpus_allowed; 907 const cpumask_t *cpus_ptr; 908 cpumask_t *user_cpus_ptr; 909 cpumask_t cpus_mask; 910 void *migration_pending; 911 unsigned short migration_disabled; 912 unsigned short migration_flags; 913 914 #ifdef CONFIG_PREEMPT_RCU 915 int rcu_read_lock_nesting; 916 union rcu_special rcu_read_unlock_special; 917 struct list_head rcu_node_entry; 918 struct rcu_node *rcu_blocked_node; 919 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 920 921 #ifdef CONFIG_TASKS_RCU 922 unsigned long rcu_tasks_nvcsw; 923 u8 rcu_tasks_holdout; 924 u8 rcu_tasks_idx; 925 int rcu_tasks_idle_cpu; 926 struct list_head rcu_tasks_holdout_list; 927 int rcu_tasks_exit_cpu; 928 struct list_head rcu_tasks_exit_list; 929 #endif /* #ifdef CONFIG_TASKS_RCU */ 930 931 #ifdef CONFIG_TASKS_TRACE_RCU 932 int trc_reader_nesting; 933 int trc_ipi_to_cpu; 934 union rcu_special trc_reader_special; 935 struct list_head trc_holdout_list; 936 struct list_head trc_blkd_node; 937 int trc_blkd_cpu; 938 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 939 940 struct sched_info sched_info; 941 942 struct list_head tasks; 943 struct plist_node pushable_tasks; 944 struct rb_node pushable_dl_tasks; 945 946 struct mm_struct *mm; 947 struct mm_struct *active_mm; 948 struct address_space *faults_disabled_mapping; 949 950 int exit_state; 951 int exit_code; 952 int exit_signal; 953 /* The signal sent when the parent dies: */ 954 int pdeath_signal; 955 /* JOBCTL_*, siglock protected: */ 956 unsigned long jobctl; 957 958 /* Used for emulating ABI behavior of previous Linux versions: */ 959 unsigned int personality; 960 961 /* Scheduler bits, serialized by scheduler locks: */ 962 unsigned sched_reset_on_fork:1; 963 unsigned sched_contributes_to_load:1; 964 unsigned sched_migrated:1; 965 unsigned sched_task_hot:1; 966 967 /* Force alignment to the next boundary: */ 968 unsigned :0; 969 970 /* Unserialized, strictly 'current' */ 971 972 /* 973 * This field must not be in the scheduler word above due to wakelist 974 * queueing no longer being serialized by p->on_cpu. However: 975 * 976 * p->XXX = X; ttwu() 977 * schedule() if (p->on_rq && ..) // false 978 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 979 * deactivate_task() ttwu_queue_wakelist()) 980 * p->on_rq = 0; p->sched_remote_wakeup = Y; 981 * 982 * guarantees all stores of 'current' are visible before 983 * ->sched_remote_wakeup gets used, so it can be in this word. 984 */ 985 unsigned sched_remote_wakeup:1; 986 #ifdef CONFIG_RT_MUTEXES 987 unsigned sched_rt_mutex:1; 988 #endif 989 990 /* Bit to tell TOMOYO we're in execve(): */ 991 unsigned in_execve:1; 992 unsigned in_iowait:1; 993 #ifndef TIF_RESTORE_SIGMASK 994 unsigned restore_sigmask:1; 995 #endif 996 #ifdef CONFIG_MEMCG_V1 997 unsigned in_user_fault:1; 998 #endif 999 #ifdef CONFIG_LRU_GEN 1000 /* whether the LRU algorithm may apply to this access */ 1001 unsigned in_lru_fault:1; 1002 #endif 1003 #ifdef CONFIG_COMPAT_BRK 1004 unsigned brk_randomized:1; 1005 #endif 1006 #ifdef CONFIG_CGROUPS 1007 /* disallow userland-initiated cgroup migration */ 1008 unsigned no_cgroup_migration:1; 1009 /* task is frozen/stopped (used by the cgroup freezer) */ 1010 unsigned frozen:1; 1011 #endif 1012 #ifdef CONFIG_BLK_CGROUP 1013 unsigned use_memdelay:1; 1014 #endif 1015 #ifdef CONFIG_PSI 1016 /* Stalled due to lack of memory */ 1017 unsigned in_memstall:1; 1018 #endif 1019 #ifdef CONFIG_PAGE_OWNER 1020 /* Used by page_owner=on to detect recursion in page tracking. */ 1021 unsigned in_page_owner:1; 1022 #endif 1023 #ifdef CONFIG_EVENTFD 1024 /* Recursion prevention for eventfd_signal() */ 1025 unsigned in_eventfd:1; 1026 #endif 1027 #ifdef CONFIG_ARCH_HAS_CPU_PASID 1028 unsigned pasid_activated:1; 1029 #endif 1030 #ifdef CONFIG_X86_BUS_LOCK_DETECT 1031 unsigned reported_split_lock:1; 1032 #endif 1033 #ifdef CONFIG_TASK_DELAY_ACCT 1034 /* delay due to memory thrashing */ 1035 unsigned in_thrashing:1; 1036 #endif 1037 unsigned in_nf_duplicate:1; 1038 #ifdef CONFIG_PREEMPT_RT 1039 struct netdev_xmit net_xmit; 1040 #endif 1041 unsigned long atomic_flags; /* Flags requiring atomic access. */ 1042 1043 struct restart_block restart_block; 1044 1045 pid_t pid; 1046 pid_t tgid; 1047 1048 #ifdef CONFIG_STACKPROTECTOR 1049 /* Canary value for the -fstack-protector GCC feature: */ 1050 unsigned long stack_canary; 1051 #endif 1052 /* 1053 * Pointers to the (original) parent process, youngest child, younger sibling, 1054 * older sibling, respectively. (p->father can be replaced with 1055 * p->real_parent->pid) 1056 */ 1057 1058 /* Real parent process: */ 1059 struct task_struct __rcu *real_parent; 1060 1061 /* Recipient of SIGCHLD, wait4() reports: */ 1062 struct task_struct __rcu *parent; 1063 1064 /* 1065 * Children/sibling form the list of natural children: 1066 */ 1067 struct list_head children; 1068 struct list_head sibling; 1069 struct task_struct *group_leader; 1070 1071 /* 1072 * 'ptraced' is the list of tasks this task is using ptrace() on. 1073 * 1074 * This includes both natural children and PTRACE_ATTACH targets. 1075 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1076 */ 1077 struct list_head ptraced; 1078 struct list_head ptrace_entry; 1079 1080 /* PID/PID hash table linkage. */ 1081 struct pid *thread_pid; 1082 struct hlist_node pid_links[PIDTYPE_MAX]; 1083 struct list_head thread_node; 1084 1085 struct completion *vfork_done; 1086 1087 /* CLONE_CHILD_SETTID: */ 1088 int __user *set_child_tid; 1089 1090 /* CLONE_CHILD_CLEARTID: */ 1091 int __user *clear_child_tid; 1092 1093 /* PF_KTHREAD | PF_IO_WORKER */ 1094 void *worker_private; 1095 1096 u64 utime; 1097 u64 stime; 1098 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1099 u64 utimescaled; 1100 u64 stimescaled; 1101 #endif 1102 u64 gtime; 1103 struct prev_cputime prev_cputime; 1104 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1105 struct vtime vtime; 1106 #endif 1107 1108 #ifdef CONFIG_NO_HZ_FULL 1109 atomic_t tick_dep_mask; 1110 #endif 1111 /* Context switch counts: */ 1112 unsigned long nvcsw; 1113 unsigned long nivcsw; 1114 1115 /* Monotonic time in nsecs: */ 1116 u64 start_time; 1117 1118 /* Boot based time in nsecs: */ 1119 u64 start_boottime; 1120 1121 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1122 unsigned long min_flt; 1123 unsigned long maj_flt; 1124 1125 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1126 struct posix_cputimers posix_cputimers; 1127 1128 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1129 struct posix_cputimers_work posix_cputimers_work; 1130 #endif 1131 1132 /* Process credentials: */ 1133 1134 /* Tracer's credentials at attach: */ 1135 const struct cred __rcu *ptracer_cred; 1136 1137 /* Objective and real subjective task credentials (COW): */ 1138 const struct cred __rcu *real_cred; 1139 1140 /* Effective (overridable) subjective task credentials (COW): */ 1141 const struct cred __rcu *cred; 1142 1143 #ifdef CONFIG_KEYS 1144 /* Cached requested key. */ 1145 struct key *cached_requested_key; 1146 #endif 1147 1148 /* 1149 * executable name, excluding path. 1150 * 1151 * - normally initialized begin_new_exec() 1152 * - set it with set_task_comm() 1153 * - strscpy_pad() to ensure it is always NUL-terminated and 1154 * zero-padded 1155 * - task_lock() to ensure the operation is atomic and the name is 1156 * fully updated. 1157 */ 1158 char comm[TASK_COMM_LEN]; 1159 1160 struct nameidata *nameidata; 1161 1162 #ifdef CONFIG_SYSVIPC 1163 struct sysv_sem sysvsem; 1164 struct sysv_shm sysvshm; 1165 #endif 1166 #ifdef CONFIG_DETECT_HUNG_TASK 1167 unsigned long last_switch_count; 1168 unsigned long last_switch_time; 1169 #endif 1170 /* Filesystem information: */ 1171 struct fs_struct *fs; 1172 1173 /* Open file information: */ 1174 struct files_struct *files; 1175 1176 #ifdef CONFIG_IO_URING 1177 struct io_uring_task *io_uring; 1178 #endif 1179 1180 /* Namespaces: */ 1181 struct nsproxy *nsproxy; 1182 1183 /* Signal handlers: */ 1184 struct signal_struct *signal; 1185 struct sighand_struct __rcu *sighand; 1186 sigset_t blocked; 1187 sigset_t real_blocked; 1188 /* Restored if set_restore_sigmask() was used: */ 1189 sigset_t saved_sigmask; 1190 struct sigpending pending; 1191 unsigned long sas_ss_sp; 1192 size_t sas_ss_size; 1193 unsigned int sas_ss_flags; 1194 1195 struct callback_head *task_works; 1196 1197 #ifdef CONFIG_AUDIT 1198 #ifdef CONFIG_AUDITSYSCALL 1199 struct audit_context *audit_context; 1200 #endif 1201 kuid_t loginuid; 1202 unsigned int sessionid; 1203 #endif 1204 struct seccomp seccomp; 1205 struct syscall_user_dispatch syscall_dispatch; 1206 1207 /* Thread group tracking: */ 1208 u64 parent_exec_id; 1209 u64 self_exec_id; 1210 1211 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1212 spinlock_t alloc_lock; 1213 1214 /* Protection of the PI data structures: */ 1215 raw_spinlock_t pi_lock; 1216 1217 struct wake_q_node wake_q; 1218 1219 #ifdef CONFIG_RT_MUTEXES 1220 /* PI waiters blocked on a rt_mutex held by this task: */ 1221 struct rb_root_cached pi_waiters; 1222 /* Updated under owner's pi_lock and rq lock */ 1223 struct task_struct *pi_top_task; 1224 /* Deadlock detection and priority inheritance handling: */ 1225 struct rt_mutex_waiter *pi_blocked_on; 1226 #endif 1227 1228 #ifdef CONFIG_DEBUG_MUTEXES 1229 /* Mutex deadlock detection: */ 1230 struct mutex_waiter *blocked_on; 1231 #endif 1232 1233 #ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER 1234 /* 1235 * Encoded lock address causing task block (lower 2 bits = type from 1236 * <linux/hung_task.h>). Accessed via hung_task_*() helpers. 1237 */ 1238 unsigned long blocker; 1239 #endif 1240 1241 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1242 int non_block_count; 1243 #endif 1244 1245 #ifdef CONFIG_TRACE_IRQFLAGS 1246 struct irqtrace_events irqtrace; 1247 unsigned int hardirq_threaded; 1248 u64 hardirq_chain_key; 1249 int softirqs_enabled; 1250 int softirq_context; 1251 int irq_config; 1252 #endif 1253 #ifdef CONFIG_PREEMPT_RT 1254 int softirq_disable_cnt; 1255 #endif 1256 1257 #ifdef CONFIG_LOCKDEP 1258 # define MAX_LOCK_DEPTH 48UL 1259 u64 curr_chain_key; 1260 int lockdep_depth; 1261 unsigned int lockdep_recursion; 1262 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1263 #endif 1264 1265 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1266 unsigned int in_ubsan; 1267 #endif 1268 1269 /* Journalling filesystem info: */ 1270 void *journal_info; 1271 1272 /* Stacked block device info: */ 1273 struct bio_list *bio_list; 1274 1275 /* Stack plugging: */ 1276 struct blk_plug *plug; 1277 1278 /* VM state: */ 1279 struct reclaim_state *reclaim_state; 1280 1281 struct io_context *io_context; 1282 1283 #ifdef CONFIG_COMPACTION 1284 struct capture_control *capture_control; 1285 #endif 1286 /* Ptrace state: */ 1287 unsigned long ptrace_message; 1288 kernel_siginfo_t *last_siginfo; 1289 1290 struct task_io_accounting ioac; 1291 #ifdef CONFIG_PSI 1292 /* Pressure stall state */ 1293 unsigned int psi_flags; 1294 #endif 1295 #ifdef CONFIG_TASK_XACCT 1296 /* Accumulated RSS usage: */ 1297 u64 acct_rss_mem1; 1298 /* Accumulated virtual memory usage: */ 1299 u64 acct_vm_mem1; 1300 /* stime + utime since last update: */ 1301 u64 acct_timexpd; 1302 #endif 1303 #ifdef CONFIG_CPUSETS 1304 /* Protected by ->alloc_lock: */ 1305 nodemask_t mems_allowed; 1306 /* Sequence number to catch updates: */ 1307 seqcount_spinlock_t mems_allowed_seq; 1308 int cpuset_mem_spread_rotor; 1309 #endif 1310 #ifdef CONFIG_CGROUPS 1311 /* Control Group info protected by css_set_lock: */ 1312 struct css_set __rcu *cgroups; 1313 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1314 struct list_head cg_list; 1315 #endif 1316 #ifdef CONFIG_X86_CPU_RESCTRL 1317 u32 closid; 1318 u32 rmid; 1319 #endif 1320 #ifdef CONFIG_FUTEX 1321 struct robust_list_head __user *robust_list; 1322 #ifdef CONFIG_COMPAT 1323 struct compat_robust_list_head __user *compat_robust_list; 1324 #endif 1325 struct list_head pi_state_list; 1326 struct futex_pi_state *pi_state_cache; 1327 struct mutex futex_exit_mutex; 1328 unsigned int futex_state; 1329 #endif 1330 #ifdef CONFIG_PERF_EVENTS 1331 u8 perf_recursion[PERF_NR_CONTEXTS]; 1332 struct perf_event_context *perf_event_ctxp; 1333 struct mutex perf_event_mutex; 1334 struct list_head perf_event_list; 1335 struct perf_ctx_data __rcu *perf_ctx_data; 1336 #endif 1337 #ifdef CONFIG_DEBUG_PREEMPT 1338 unsigned long preempt_disable_ip; 1339 #endif 1340 #ifdef CONFIG_NUMA 1341 /* Protected by alloc_lock: */ 1342 struct mempolicy *mempolicy; 1343 short il_prev; 1344 u8 il_weight; 1345 short pref_node_fork; 1346 #endif 1347 #ifdef CONFIG_NUMA_BALANCING 1348 int numa_scan_seq; 1349 unsigned int numa_scan_period; 1350 unsigned int numa_scan_period_max; 1351 int numa_preferred_nid; 1352 unsigned long numa_migrate_retry; 1353 /* Migration stamp: */ 1354 u64 node_stamp; 1355 u64 last_task_numa_placement; 1356 u64 last_sum_exec_runtime; 1357 struct callback_head numa_work; 1358 1359 /* 1360 * This pointer is only modified for current in syscall and 1361 * pagefault context (and for tasks being destroyed), so it can be read 1362 * from any of the following contexts: 1363 * - RCU read-side critical section 1364 * - current->numa_group from everywhere 1365 * - task's runqueue locked, task not running 1366 */ 1367 struct numa_group __rcu *numa_group; 1368 1369 /* 1370 * numa_faults is an array split into four regions: 1371 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1372 * in this precise order. 1373 * 1374 * faults_memory: Exponential decaying average of faults on a per-node 1375 * basis. Scheduling placement decisions are made based on these 1376 * counts. The values remain static for the duration of a PTE scan. 1377 * faults_cpu: Track the nodes the process was running on when a NUMA 1378 * hinting fault was incurred. 1379 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1380 * during the current scan window. When the scan completes, the counts 1381 * in faults_memory and faults_cpu decay and these values are copied. 1382 */ 1383 unsigned long *numa_faults; 1384 unsigned long total_numa_faults; 1385 1386 /* 1387 * numa_faults_locality tracks if faults recorded during the last 1388 * scan window were remote/local or failed to migrate. The task scan 1389 * period is adapted based on the locality of the faults with different 1390 * weights depending on whether they were shared or private faults 1391 */ 1392 unsigned long numa_faults_locality[3]; 1393 1394 unsigned long numa_pages_migrated; 1395 #endif /* CONFIG_NUMA_BALANCING */ 1396 1397 #ifdef CONFIG_RSEQ 1398 struct rseq __user *rseq; 1399 u32 rseq_len; 1400 u32 rseq_sig; 1401 /* 1402 * RmW on rseq_event_mask must be performed atomically 1403 * with respect to preemption. 1404 */ 1405 unsigned long rseq_event_mask; 1406 # ifdef CONFIG_DEBUG_RSEQ 1407 /* 1408 * This is a place holder to save a copy of the rseq fields for 1409 * validation of read-only fields. The struct rseq has a 1410 * variable-length array at the end, so it cannot be used 1411 * directly. Reserve a size large enough for the known fields. 1412 */ 1413 char rseq_fields[sizeof(struct rseq)]; 1414 # endif 1415 #endif 1416 1417 #ifdef CONFIG_SCHED_MM_CID 1418 int mm_cid; /* Current cid in mm */ 1419 int last_mm_cid; /* Most recent cid in mm */ 1420 int migrate_from_cpu; 1421 int mm_cid_active; /* Whether cid bitmap is active */ 1422 struct callback_head cid_work; 1423 #endif 1424 1425 struct tlbflush_unmap_batch tlb_ubc; 1426 1427 /* Cache last used pipe for splice(): */ 1428 struct pipe_inode_info *splice_pipe; 1429 1430 struct page_frag task_frag; 1431 1432 #ifdef CONFIG_TASK_DELAY_ACCT 1433 struct task_delay_info *delays; 1434 #endif 1435 1436 #ifdef CONFIG_FAULT_INJECTION 1437 int make_it_fail; 1438 unsigned int fail_nth; 1439 #endif 1440 /* 1441 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1442 * balance_dirty_pages() for a dirty throttling pause: 1443 */ 1444 int nr_dirtied; 1445 int nr_dirtied_pause; 1446 /* Start of a write-and-pause period: */ 1447 unsigned long dirty_paused_when; 1448 1449 #ifdef CONFIG_LATENCYTOP 1450 int latency_record_count; 1451 struct latency_record latency_record[LT_SAVECOUNT]; 1452 #endif 1453 /* 1454 * Time slack values; these are used to round up poll() and 1455 * select() etc timeout values. These are in nanoseconds. 1456 */ 1457 u64 timer_slack_ns; 1458 u64 default_timer_slack_ns; 1459 1460 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1461 unsigned int kasan_depth; 1462 #endif 1463 1464 #ifdef CONFIG_KCSAN 1465 struct kcsan_ctx kcsan_ctx; 1466 #ifdef CONFIG_TRACE_IRQFLAGS 1467 struct irqtrace_events kcsan_save_irqtrace; 1468 #endif 1469 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1470 int kcsan_stack_depth; 1471 #endif 1472 #endif 1473 1474 #ifdef CONFIG_KMSAN 1475 struct kmsan_ctx kmsan_ctx; 1476 #endif 1477 1478 #if IS_ENABLED(CONFIG_KUNIT) 1479 struct kunit *kunit_test; 1480 #endif 1481 1482 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1483 /* Index of current stored address in ret_stack: */ 1484 int curr_ret_stack; 1485 int curr_ret_depth; 1486 1487 /* Stack of return addresses for return function tracing: */ 1488 unsigned long *ret_stack; 1489 1490 /* Timestamp for last schedule: */ 1491 unsigned long long ftrace_timestamp; 1492 unsigned long long ftrace_sleeptime; 1493 1494 /* 1495 * Number of functions that haven't been traced 1496 * because of depth overrun: 1497 */ 1498 atomic_t trace_overrun; 1499 1500 /* Pause tracing: */ 1501 atomic_t tracing_graph_pause; 1502 #endif 1503 1504 #ifdef CONFIG_TRACING 1505 /* Bitmask and counter of trace recursion: */ 1506 unsigned long trace_recursion; 1507 #endif /* CONFIG_TRACING */ 1508 1509 #ifdef CONFIG_KCOV 1510 /* See kernel/kcov.c for more details. */ 1511 1512 /* Coverage collection mode enabled for this task (0 if disabled): */ 1513 unsigned int kcov_mode; 1514 1515 /* Size of the kcov_area: */ 1516 unsigned int kcov_size; 1517 1518 /* Buffer for coverage collection: */ 1519 void *kcov_area; 1520 1521 /* KCOV descriptor wired with this task or NULL: */ 1522 struct kcov *kcov; 1523 1524 /* KCOV common handle for remote coverage collection: */ 1525 u64 kcov_handle; 1526 1527 /* KCOV sequence number: */ 1528 int kcov_sequence; 1529 1530 /* Collect coverage from softirq context: */ 1531 unsigned int kcov_softirq; 1532 #endif 1533 1534 #ifdef CONFIG_MEMCG_V1 1535 struct mem_cgroup *memcg_in_oom; 1536 #endif 1537 1538 #ifdef CONFIG_MEMCG 1539 /* Number of pages to reclaim on returning to userland: */ 1540 unsigned int memcg_nr_pages_over_high; 1541 1542 /* Used by memcontrol for targeted memcg charge: */ 1543 struct mem_cgroup *active_memcg; 1544 1545 /* Cache for current->cgroups->memcg->objcg lookups: */ 1546 struct obj_cgroup *objcg; 1547 #endif 1548 1549 #ifdef CONFIG_BLK_CGROUP 1550 struct gendisk *throttle_disk; 1551 #endif 1552 1553 #ifdef CONFIG_UPROBES 1554 struct uprobe_task *utask; 1555 #endif 1556 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1557 unsigned int sequential_io; 1558 unsigned int sequential_io_avg; 1559 #endif 1560 struct kmap_ctrl kmap_ctrl; 1561 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1562 unsigned long task_state_change; 1563 # ifdef CONFIG_PREEMPT_RT 1564 unsigned long saved_state_change; 1565 # endif 1566 #endif 1567 struct rcu_head rcu; 1568 refcount_t rcu_users; 1569 int pagefault_disabled; 1570 #ifdef CONFIG_MMU 1571 struct task_struct *oom_reaper_list; 1572 struct timer_list oom_reaper_timer; 1573 #endif 1574 #ifdef CONFIG_VMAP_STACK 1575 struct vm_struct *stack_vm_area; 1576 #endif 1577 #ifdef CONFIG_THREAD_INFO_IN_TASK 1578 /* A live task holds one reference: */ 1579 refcount_t stack_refcount; 1580 #endif 1581 #ifdef CONFIG_LIVEPATCH 1582 int patch_state; 1583 #endif 1584 #ifdef CONFIG_SECURITY 1585 /* Used by LSM modules for access restriction: */ 1586 void *security; 1587 #endif 1588 #ifdef CONFIG_BPF_SYSCALL 1589 /* Used by BPF task local storage */ 1590 struct bpf_local_storage __rcu *bpf_storage; 1591 /* Used for BPF run context */ 1592 struct bpf_run_ctx *bpf_ctx; 1593 #endif 1594 /* Used by BPF for per-TASK xdp storage */ 1595 struct bpf_net_context *bpf_net_context; 1596 1597 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1598 unsigned long lowest_stack; 1599 unsigned long prev_lowest_stack; 1600 #endif 1601 1602 #ifdef CONFIG_X86_MCE 1603 void __user *mce_vaddr; 1604 __u64 mce_kflags; 1605 u64 mce_addr; 1606 __u64 mce_ripv : 1, 1607 mce_whole_page : 1, 1608 __mce_reserved : 62; 1609 struct callback_head mce_kill_me; 1610 int mce_count; 1611 #endif 1612 1613 #ifdef CONFIG_KRETPROBES 1614 struct llist_head kretprobe_instances; 1615 #endif 1616 #ifdef CONFIG_RETHOOK 1617 struct llist_head rethooks; 1618 #endif 1619 1620 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1621 /* 1622 * If L1D flush is supported on mm context switch 1623 * then we use this callback head to queue kill work 1624 * to kill tasks that are not running on SMT disabled 1625 * cores 1626 */ 1627 struct callback_head l1d_flush_kill; 1628 #endif 1629 1630 #ifdef CONFIG_RV 1631 /* 1632 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS. 1633 * If we find justification for more monitors, we can think 1634 * about adding more or developing a dynamic method. So far, 1635 * none of these are justified. 1636 */ 1637 union rv_task_monitor rv[RV_PER_TASK_MONITORS]; 1638 #endif 1639 1640 #ifdef CONFIG_USER_EVENTS 1641 struct user_event_mm *user_event_mm; 1642 #endif 1643 1644 /* CPU-specific state of this task: */ 1645 struct thread_struct thread; 1646 1647 /* 1648 * New fields for task_struct should be added above here, so that 1649 * they are included in the randomized portion of task_struct. 1650 */ 1651 randomized_struct_fields_end 1652 } __attribute__ ((aligned (64))); 1653 1654 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1655 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1656 1657 static inline unsigned int __task_state_index(unsigned int tsk_state, 1658 unsigned int tsk_exit_state) 1659 { 1660 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1661 1662 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1663 1664 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1665 state = TASK_REPORT_IDLE; 1666 1667 /* 1668 * We're lying here, but rather than expose a completely new task state 1669 * to userspace, we can make this appear as if the task has gone through 1670 * a regular rt_mutex_lock() call. 1671 * Report frozen tasks as uninterruptible. 1672 */ 1673 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN)) 1674 state = TASK_UNINTERRUPTIBLE; 1675 1676 return fls(state); 1677 } 1678 1679 static inline unsigned int task_state_index(struct task_struct *tsk) 1680 { 1681 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1682 } 1683 1684 static inline char task_index_to_char(unsigned int state) 1685 { 1686 static const char state_char[] = "RSDTtXZPI"; 1687 1688 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1)); 1689 1690 return state_char[state]; 1691 } 1692 1693 static inline char task_state_to_char(struct task_struct *tsk) 1694 { 1695 return task_index_to_char(task_state_index(tsk)); 1696 } 1697 1698 extern struct pid *cad_pid; 1699 1700 /* 1701 * Per process flags 1702 */ 1703 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1704 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1705 #define PF_EXITING 0x00000004 /* Getting shut down */ 1706 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1707 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1708 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1709 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1710 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1711 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1712 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1713 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1714 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1715 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1716 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1717 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1718 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1719 #define PF_KCOMPACTD 0x00010000 /* I am kcompactd */ 1720 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1721 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1722 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1723 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1724 * I am cleaning dirty pages from some other bdi. */ 1725 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1726 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1727 #define PF__HOLE__00800000 0x00800000 1728 #define PF__HOLE__01000000 0x01000000 1729 #define PF__HOLE__02000000 0x02000000 1730 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1731 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1732 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1733 * See memalloc_pin_save() */ 1734 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1735 #define PF__HOLE__40000000 0x40000000 1736 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1737 1738 /* 1739 * Only the _current_ task can read/write to tsk->flags, but other 1740 * tasks can access tsk->flags in readonly mode for example 1741 * with tsk_used_math (like during threaded core dumping). 1742 * There is however an exception to this rule during ptrace 1743 * or during fork: the ptracer task is allowed to write to the 1744 * child->flags of its traced child (same goes for fork, the parent 1745 * can write to the child->flags), because we're guaranteed the 1746 * child is not running and in turn not changing child->flags 1747 * at the same time the parent does it. 1748 */ 1749 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1750 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1751 #define clear_used_math() clear_stopped_child_used_math(current) 1752 #define set_used_math() set_stopped_child_used_math(current) 1753 1754 #define conditional_stopped_child_used_math(condition, child) \ 1755 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1756 1757 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1758 1759 #define copy_to_stopped_child_used_math(child) \ 1760 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1761 1762 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1763 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1764 #define used_math() tsk_used_math(current) 1765 1766 static __always_inline bool is_percpu_thread(void) 1767 { 1768 return (current->flags & PF_NO_SETAFFINITY) && 1769 (current->nr_cpus_allowed == 1); 1770 } 1771 1772 /* Per-process atomic flags. */ 1773 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1774 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1775 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1776 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1777 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1778 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1779 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1780 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1781 1782 #define TASK_PFA_TEST(name, func) \ 1783 static inline bool task_##func(struct task_struct *p) \ 1784 { return test_bit(PFA_##name, &p->atomic_flags); } 1785 1786 #define TASK_PFA_SET(name, func) \ 1787 static inline void task_set_##func(struct task_struct *p) \ 1788 { set_bit(PFA_##name, &p->atomic_flags); } 1789 1790 #define TASK_PFA_CLEAR(name, func) \ 1791 static inline void task_clear_##func(struct task_struct *p) \ 1792 { clear_bit(PFA_##name, &p->atomic_flags); } 1793 1794 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1795 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1796 1797 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1798 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1799 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1800 1801 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1802 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1803 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1804 1805 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1806 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1807 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1808 1809 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1810 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1811 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1812 1813 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1814 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1815 1816 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1817 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1818 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1819 1820 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1821 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1822 1823 static inline void 1824 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1825 { 1826 current->flags &= ~flags; 1827 current->flags |= orig_flags & flags; 1828 } 1829 1830 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1831 extern int task_can_attach(struct task_struct *p); 1832 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1833 extern void dl_bw_free(int cpu, u64 dl_bw); 1834 1835 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1836 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1837 1838 /** 1839 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1840 * @p: the task 1841 * @new_mask: CPU affinity mask 1842 * 1843 * Return: zero if successful, or a negative error code 1844 */ 1845 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1846 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1847 extern void release_user_cpus_ptr(struct task_struct *p); 1848 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1849 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1850 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1851 1852 extern int yield_to(struct task_struct *p, bool preempt); 1853 extern void set_user_nice(struct task_struct *p, long nice); 1854 extern int task_prio(const struct task_struct *p); 1855 1856 /** 1857 * task_nice - return the nice value of a given task. 1858 * @p: the task in question. 1859 * 1860 * Return: The nice value [ -20 ... 0 ... 19 ]. 1861 */ 1862 static inline int task_nice(const struct task_struct *p) 1863 { 1864 return PRIO_TO_NICE((p)->static_prio); 1865 } 1866 1867 extern int can_nice(const struct task_struct *p, const int nice); 1868 extern int task_curr(const struct task_struct *p); 1869 extern int idle_cpu(int cpu); 1870 extern int available_idle_cpu(int cpu); 1871 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1872 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1873 extern void sched_set_fifo(struct task_struct *p); 1874 extern void sched_set_fifo_low(struct task_struct *p); 1875 extern void sched_set_normal(struct task_struct *p, int nice); 1876 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1877 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1878 extern struct task_struct *idle_task(int cpu); 1879 1880 /** 1881 * is_idle_task - is the specified task an idle task? 1882 * @p: the task in question. 1883 * 1884 * Return: 1 if @p is an idle task. 0 otherwise. 1885 */ 1886 static __always_inline bool is_idle_task(const struct task_struct *p) 1887 { 1888 return !!(p->flags & PF_IDLE); 1889 } 1890 1891 extern struct task_struct *curr_task(int cpu); 1892 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1893 1894 void yield(void); 1895 1896 union thread_union { 1897 struct task_struct task; 1898 #ifndef CONFIG_THREAD_INFO_IN_TASK 1899 struct thread_info thread_info; 1900 #endif 1901 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1902 }; 1903 1904 #ifndef CONFIG_THREAD_INFO_IN_TASK 1905 extern struct thread_info init_thread_info; 1906 #endif 1907 1908 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1909 1910 #ifdef CONFIG_THREAD_INFO_IN_TASK 1911 # define task_thread_info(task) (&(task)->thread_info) 1912 #else 1913 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1914 #endif 1915 1916 /* 1917 * find a task by one of its numerical ids 1918 * 1919 * find_task_by_pid_ns(): 1920 * finds a task by its pid in the specified namespace 1921 * find_task_by_vpid(): 1922 * finds a task by its virtual pid 1923 * 1924 * see also find_vpid() etc in include/linux/pid.h 1925 */ 1926 1927 extern struct task_struct *find_task_by_vpid(pid_t nr); 1928 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1929 1930 /* 1931 * find a task by its virtual pid and get the task struct 1932 */ 1933 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1934 1935 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1936 extern int wake_up_process(struct task_struct *tsk); 1937 extern void wake_up_new_task(struct task_struct *tsk); 1938 1939 extern void kick_process(struct task_struct *tsk); 1940 1941 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1942 #define set_task_comm(tsk, from) ({ \ 1943 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \ 1944 __set_task_comm(tsk, from, false); \ 1945 }) 1946 1947 /* 1948 * - Why not use task_lock()? 1949 * User space can randomly change their names anyway, so locking for readers 1950 * doesn't make sense. For writers, locking is probably necessary, as a race 1951 * condition could lead to long-term mixed results. 1952 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is 1953 * always NUL-terminated and zero-padded. Therefore the race condition between 1954 * reader and writer is not an issue. 1955 * 1956 * - BUILD_BUG_ON() can help prevent the buf from being truncated. 1957 * Since the callers don't perform any return value checks, this safeguard is 1958 * necessary. 1959 */ 1960 #define get_task_comm(buf, tsk) ({ \ 1961 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \ 1962 strscpy_pad(buf, (tsk)->comm); \ 1963 buf; \ 1964 }) 1965 1966 static __always_inline void scheduler_ipi(void) 1967 { 1968 /* 1969 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1970 * TIF_NEED_RESCHED remotely (for the first time) will also send 1971 * this IPI. 1972 */ 1973 preempt_fold_need_resched(); 1974 } 1975 1976 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 1977 1978 /* 1979 * Set thread flags in other task's structures. 1980 * See asm/thread_info.h for TIF_xxxx flags available: 1981 */ 1982 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1983 { 1984 set_ti_thread_flag(task_thread_info(tsk), flag); 1985 } 1986 1987 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1988 { 1989 clear_ti_thread_flag(task_thread_info(tsk), flag); 1990 } 1991 1992 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1993 bool value) 1994 { 1995 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1996 } 1997 1998 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1999 { 2000 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2001 } 2002 2003 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2004 { 2005 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2006 } 2007 2008 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2009 { 2010 return test_ti_thread_flag(task_thread_info(tsk), flag); 2011 } 2012 2013 static inline void set_tsk_need_resched(struct task_struct *tsk) 2014 { 2015 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2016 } 2017 2018 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2019 { 2020 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY, 2021 (atomic_long_t *)&task_thread_info(tsk)->flags); 2022 } 2023 2024 static inline int test_tsk_need_resched(struct task_struct *tsk) 2025 { 2026 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2027 } 2028 2029 /* 2030 * cond_resched() and cond_resched_lock(): latency reduction via 2031 * explicit rescheduling in places that are safe. The return 2032 * value indicates whether a reschedule was done in fact. 2033 * cond_resched_lock() will drop the spinlock before scheduling, 2034 */ 2035 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2036 extern int __cond_resched(void); 2037 2038 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2039 2040 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2041 2042 static __always_inline int _cond_resched(void) 2043 { 2044 return static_call_mod(cond_resched)(); 2045 } 2046 2047 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2048 2049 extern int dynamic_cond_resched(void); 2050 2051 static __always_inline int _cond_resched(void) 2052 { 2053 return dynamic_cond_resched(); 2054 } 2055 2056 #else /* !CONFIG_PREEMPTION */ 2057 2058 static inline int _cond_resched(void) 2059 { 2060 return __cond_resched(); 2061 } 2062 2063 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2064 2065 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2066 2067 static inline int _cond_resched(void) 2068 { 2069 return 0; 2070 } 2071 2072 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2073 2074 #define cond_resched() ({ \ 2075 __might_resched(__FILE__, __LINE__, 0); \ 2076 _cond_resched(); \ 2077 }) 2078 2079 extern int __cond_resched_lock(spinlock_t *lock); 2080 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2081 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2082 2083 #define MIGHT_RESCHED_RCU_SHIFT 8 2084 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2085 2086 #ifndef CONFIG_PREEMPT_RT 2087 /* 2088 * Non RT kernels have an elevated preempt count due to the held lock, 2089 * but are not allowed to be inside a RCU read side critical section 2090 */ 2091 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2092 #else 2093 /* 2094 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2095 * cond_resched*lock() has to take that into account because it checks for 2096 * preempt_count() and rcu_preempt_depth(). 2097 */ 2098 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2099 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2100 #endif 2101 2102 #define cond_resched_lock(lock) ({ \ 2103 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2104 __cond_resched_lock(lock); \ 2105 }) 2106 2107 #define cond_resched_rwlock_read(lock) ({ \ 2108 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2109 __cond_resched_rwlock_read(lock); \ 2110 }) 2111 2112 #define cond_resched_rwlock_write(lock) ({ \ 2113 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2114 __cond_resched_rwlock_write(lock); \ 2115 }) 2116 2117 static __always_inline bool need_resched(void) 2118 { 2119 return unlikely(tif_need_resched()); 2120 } 2121 2122 /* 2123 * Wrappers for p->thread_info->cpu access. No-op on UP. 2124 */ 2125 #ifdef CONFIG_SMP 2126 2127 static inline unsigned int task_cpu(const struct task_struct *p) 2128 { 2129 return READ_ONCE(task_thread_info(p)->cpu); 2130 } 2131 2132 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2133 2134 #else 2135 2136 static inline unsigned int task_cpu(const struct task_struct *p) 2137 { 2138 return 0; 2139 } 2140 2141 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2142 { 2143 } 2144 2145 #endif /* CONFIG_SMP */ 2146 2147 static inline bool task_is_runnable(struct task_struct *p) 2148 { 2149 return p->on_rq && !p->se.sched_delayed; 2150 } 2151 2152 extern bool sched_task_on_rq(struct task_struct *p); 2153 extern unsigned long get_wchan(struct task_struct *p); 2154 extern struct task_struct *cpu_curr_snapshot(int cpu); 2155 2156 #include <linux/spinlock.h> 2157 2158 /* 2159 * In order to reduce various lock holder preemption latencies provide an 2160 * interface to see if a vCPU is currently running or not. 2161 * 2162 * This allows us to terminate optimistic spin loops and block, analogous to 2163 * the native optimistic spin heuristic of testing if the lock owner task is 2164 * running or not. 2165 */ 2166 #ifndef vcpu_is_preempted 2167 static inline bool vcpu_is_preempted(int cpu) 2168 { 2169 return false; 2170 } 2171 #endif 2172 2173 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2174 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2175 2176 #ifndef TASK_SIZE_OF 2177 #define TASK_SIZE_OF(tsk) TASK_SIZE 2178 #endif 2179 2180 static inline bool owner_on_cpu(struct task_struct *owner) 2181 { 2182 /* 2183 * As lock holder preemption issue, we both skip spinning if 2184 * task is not on cpu or its cpu is preempted 2185 */ 2186 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2187 } 2188 2189 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2190 unsigned long sched_cpu_util(int cpu); 2191 2192 #ifdef CONFIG_SCHED_CORE 2193 extern void sched_core_free(struct task_struct *tsk); 2194 extern void sched_core_fork(struct task_struct *p); 2195 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2196 unsigned long uaddr); 2197 extern int sched_core_idle_cpu(int cpu); 2198 #else 2199 static inline void sched_core_free(struct task_struct *tsk) { } 2200 static inline void sched_core_fork(struct task_struct *p) { } 2201 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2202 #endif 2203 2204 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2205 2206 #ifdef CONFIG_MEM_ALLOC_PROFILING 2207 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2208 { 2209 swap(current->alloc_tag, tag); 2210 return tag; 2211 } 2212 2213 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2214 { 2215 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2216 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2217 #endif 2218 current->alloc_tag = old; 2219 } 2220 #else 2221 #define alloc_tag_save(_tag) NULL 2222 #define alloc_tag_restore(_tag, _old) do {} while (0) 2223 #endif 2224 2225 #endif 2226