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 #ifdef CONFIG_SMP 399 extern struct root_domain def_root_domain; 400 extern struct mutex sched_domains_mutex; 401 extern void sched_domains_mutex_lock(void); 402 extern void sched_domains_mutex_unlock(void); 403 #else 404 static inline void sched_domains_mutex_lock(void) { } 405 static inline void sched_domains_mutex_unlock(void) { } 406 #endif 407 408 struct sched_param { 409 int sched_priority; 410 }; 411 412 struct sched_info { 413 #ifdef CONFIG_SCHED_INFO 414 /* Cumulative counters: */ 415 416 /* # of times we have run on this CPU: */ 417 unsigned long pcount; 418 419 /* Time spent waiting on a runqueue: */ 420 unsigned long long run_delay; 421 422 /* Max time spent waiting on a runqueue: */ 423 unsigned long long max_run_delay; 424 425 /* Min time spent waiting on a runqueue: */ 426 unsigned long long min_run_delay; 427 428 /* Timestamps: */ 429 430 /* When did we last run on a CPU? */ 431 unsigned long long last_arrival; 432 433 /* When were we last queued to run? */ 434 unsigned long long last_queued; 435 436 #endif /* CONFIG_SCHED_INFO */ 437 }; 438 439 /* 440 * Integer metrics need fixed point arithmetic, e.g., sched/fair 441 * has a few: load, load_avg, util_avg, freq, and capacity. 442 * 443 * We define a basic fixed point arithmetic range, and then formalize 444 * all these metrics based on that basic range. 445 */ 446 # define SCHED_FIXEDPOINT_SHIFT 10 447 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 448 449 /* Increase resolution of cpu_capacity calculations */ 450 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 451 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 452 453 struct load_weight { 454 unsigned long weight; 455 u32 inv_weight; 456 }; 457 458 /* 459 * The load/runnable/util_avg accumulates an infinite geometric series 460 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 461 * 462 * [load_avg definition] 463 * 464 * load_avg = runnable% * scale_load_down(load) 465 * 466 * [runnable_avg definition] 467 * 468 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 469 * 470 * [util_avg definition] 471 * 472 * util_avg = running% * SCHED_CAPACITY_SCALE 473 * 474 * where runnable% is the time ratio that a sched_entity is runnable and 475 * running% the time ratio that a sched_entity is running. 476 * 477 * For cfs_rq, they are the aggregated values of all runnable and blocked 478 * sched_entities. 479 * 480 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 481 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 482 * for computing those signals (see update_rq_clock_pelt()) 483 * 484 * N.B., the above ratios (runnable% and running%) themselves are in the 485 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 486 * to as large a range as necessary. This is for example reflected by 487 * util_avg's SCHED_CAPACITY_SCALE. 488 * 489 * [Overflow issue] 490 * 491 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 492 * with the highest load (=88761), always runnable on a single cfs_rq, 493 * and should not overflow as the number already hits PID_MAX_LIMIT. 494 * 495 * For all other cases (including 32-bit kernels), struct load_weight's 496 * weight will overflow first before we do, because: 497 * 498 * Max(load_avg) <= Max(load.weight) 499 * 500 * Then it is the load_weight's responsibility to consider overflow 501 * issues. 502 */ 503 struct sched_avg { 504 u64 last_update_time; 505 u64 load_sum; 506 u64 runnable_sum; 507 u32 util_sum; 508 u32 period_contrib; 509 unsigned long load_avg; 510 unsigned long runnable_avg; 511 unsigned long util_avg; 512 unsigned int util_est; 513 } ____cacheline_aligned; 514 515 /* 516 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 517 * updates. When a task is dequeued, its util_est should not be updated if its 518 * util_avg has not been updated in the meantime. 519 * This information is mapped into the MSB bit of util_est at dequeue time. 520 * Since max value of util_est for a task is 1024 (PELT util_avg for a task) 521 * it is safe to use MSB. 522 */ 523 #define UTIL_EST_WEIGHT_SHIFT 2 524 #define UTIL_AVG_UNCHANGED 0x80000000 525 526 struct sched_statistics { 527 #ifdef CONFIG_SCHEDSTATS 528 u64 wait_start; 529 u64 wait_max; 530 u64 wait_count; 531 u64 wait_sum; 532 u64 iowait_count; 533 u64 iowait_sum; 534 535 u64 sleep_start; 536 u64 sleep_max; 537 s64 sum_sleep_runtime; 538 539 u64 block_start; 540 u64 block_max; 541 s64 sum_block_runtime; 542 543 s64 exec_max; 544 u64 slice_max; 545 546 u64 nr_migrations_cold; 547 u64 nr_failed_migrations_affine; 548 u64 nr_failed_migrations_running; 549 u64 nr_failed_migrations_hot; 550 u64 nr_forced_migrations; 551 552 u64 nr_wakeups; 553 u64 nr_wakeups_sync; 554 u64 nr_wakeups_migrate; 555 u64 nr_wakeups_local; 556 u64 nr_wakeups_remote; 557 u64 nr_wakeups_affine; 558 u64 nr_wakeups_affine_attempts; 559 u64 nr_wakeups_passive; 560 u64 nr_wakeups_idle; 561 562 #ifdef CONFIG_SCHED_CORE 563 u64 core_forceidle_sum; 564 #endif 565 #endif /* CONFIG_SCHEDSTATS */ 566 } ____cacheline_aligned; 567 568 struct sched_entity { 569 /* For load-balancing: */ 570 struct load_weight load; 571 struct rb_node run_node; 572 u64 deadline; 573 u64 min_vruntime; 574 u64 min_slice; 575 576 struct list_head group_node; 577 unsigned char on_rq; 578 unsigned char sched_delayed; 579 unsigned char rel_deadline; 580 unsigned char custom_slice; 581 /* hole */ 582 583 u64 exec_start; 584 u64 sum_exec_runtime; 585 u64 prev_sum_exec_runtime; 586 u64 vruntime; 587 s64 vlag; 588 u64 slice; 589 590 u64 nr_migrations; 591 592 #ifdef CONFIG_FAIR_GROUP_SCHED 593 int depth; 594 struct sched_entity *parent; 595 /* rq on which this entity is (to be) queued: */ 596 struct cfs_rq *cfs_rq; 597 /* rq "owned" by this entity/group: */ 598 struct cfs_rq *my_q; 599 /* cached value of my_q->h_nr_running */ 600 unsigned long runnable_weight; 601 #endif 602 603 #ifdef CONFIG_SMP 604 /* 605 * Per entity load average tracking. 606 * 607 * Put into separate cache line so it does not 608 * collide with read-mostly values above. 609 */ 610 struct sched_avg avg; 611 #endif 612 }; 613 614 struct sched_rt_entity { 615 struct list_head run_list; 616 unsigned long timeout; 617 unsigned long watchdog_stamp; 618 unsigned int time_slice; 619 unsigned short on_rq; 620 unsigned short on_list; 621 622 struct sched_rt_entity *back; 623 #ifdef CONFIG_RT_GROUP_SCHED 624 struct sched_rt_entity *parent; 625 /* rq on which this entity is (to be) queued: */ 626 struct rt_rq *rt_rq; 627 /* rq "owned" by this entity/group: */ 628 struct rt_rq *my_q; 629 #endif 630 } __randomize_layout; 631 632 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *); 633 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *); 634 635 struct sched_dl_entity { 636 struct rb_node rb_node; 637 638 /* 639 * Original scheduling parameters. Copied here from sched_attr 640 * during sched_setattr(), they will remain the same until 641 * the next sched_setattr(). 642 */ 643 u64 dl_runtime; /* Maximum runtime for each instance */ 644 u64 dl_deadline; /* Relative deadline of each instance */ 645 u64 dl_period; /* Separation of two instances (period) */ 646 u64 dl_bw; /* dl_runtime / dl_period */ 647 u64 dl_density; /* dl_runtime / dl_deadline */ 648 649 /* 650 * Actual scheduling parameters. Initialized with the values above, 651 * they are continuously updated during task execution. Note that 652 * the remaining runtime could be < 0 in case we are in overrun. 653 */ 654 s64 runtime; /* Remaining runtime for this instance */ 655 u64 deadline; /* Absolute deadline for this instance */ 656 unsigned int flags; /* Specifying the scheduler behaviour */ 657 658 /* 659 * Some bool flags: 660 * 661 * @dl_throttled tells if we exhausted the runtime. If so, the 662 * task has to wait for a replenishment to be performed at the 663 * next firing of dl_timer. 664 * 665 * @dl_yielded tells if task gave up the CPU before consuming 666 * all its available runtime during the last job. 667 * 668 * @dl_non_contending tells if the task is inactive while still 669 * contributing to the active utilization. In other words, it 670 * indicates if the inactive timer has been armed and its handler 671 * has not been executed yet. This flag is useful to avoid race 672 * conditions between the inactive timer handler and the wakeup 673 * code. 674 * 675 * @dl_overrun tells if the task asked to be informed about runtime 676 * overruns. 677 * 678 * @dl_server tells if this is a server entity. 679 * 680 * @dl_defer tells if this is a deferred or regular server. For 681 * now only defer server exists. 682 * 683 * @dl_defer_armed tells if the deferrable server is waiting 684 * for the replenishment timer to activate it. 685 * 686 * @dl_server_active tells if the dlserver is active(started). 687 * dlserver is started on first cfs enqueue on an idle runqueue 688 * and is stopped when a dequeue results in 0 cfs tasks on the 689 * runqueue. In other words, dlserver is active only when cpu's 690 * runqueue has atleast one cfs task. 691 * 692 * @dl_defer_running tells if the deferrable server is actually 693 * running, skipping the defer phase. 694 */ 695 unsigned int dl_throttled : 1; 696 unsigned int dl_yielded : 1; 697 unsigned int dl_non_contending : 1; 698 unsigned int dl_overrun : 1; 699 unsigned int dl_server : 1; 700 unsigned int dl_server_active : 1; 701 unsigned int dl_defer : 1; 702 unsigned int dl_defer_armed : 1; 703 unsigned int dl_defer_running : 1; 704 705 /* 706 * Bandwidth enforcement timer. Each -deadline task has its 707 * own bandwidth to be enforced, thus we need one timer per task. 708 */ 709 struct hrtimer dl_timer; 710 711 /* 712 * Inactive timer, responsible for decreasing the active utilization 713 * at the "0-lag time". When a -deadline task blocks, it contributes 714 * to GRUB's active utilization until the "0-lag time", hence a 715 * timer is needed to decrease the active utilization at the correct 716 * time. 717 */ 718 struct hrtimer inactive_timer; 719 720 /* 721 * Bits for DL-server functionality. Also see the comment near 722 * dl_server_update(). 723 * 724 * @rq the runqueue this server is for 725 * 726 * @server_has_tasks() returns true if @server_pick return a 727 * runnable task. 728 */ 729 struct rq *rq; 730 dl_server_has_tasks_f server_has_tasks; 731 dl_server_pick_f server_pick_task; 732 733 #ifdef CONFIG_RT_MUTEXES 734 /* 735 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 736 * pi_se points to the donor, otherwise points to the dl_se it belongs 737 * to (the original one/itself). 738 */ 739 struct sched_dl_entity *pi_se; 740 #endif 741 }; 742 743 #ifdef CONFIG_UCLAMP_TASK 744 /* Number of utilization clamp buckets (shorter alias) */ 745 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 746 747 /* 748 * Utilization clamp for a scheduling entity 749 * @value: clamp value "assigned" to a se 750 * @bucket_id: bucket index corresponding to the "assigned" value 751 * @active: the se is currently refcounted in a rq's bucket 752 * @user_defined: the requested clamp value comes from user-space 753 * 754 * The bucket_id is the index of the clamp bucket matching the clamp value 755 * which is pre-computed and stored to avoid expensive integer divisions from 756 * the fast path. 757 * 758 * The active bit is set whenever a task has got an "effective" value assigned, 759 * which can be different from the clamp value "requested" from user-space. 760 * This allows to know a task is refcounted in the rq's bucket corresponding 761 * to the "effective" bucket_id. 762 * 763 * The user_defined bit is set whenever a task has got a task-specific clamp 764 * value requested from userspace, i.e. the system defaults apply to this task 765 * just as a restriction. This allows to relax default clamps when a less 766 * restrictive task-specific value has been requested, thus allowing to 767 * implement a "nice" semantic. For example, a task running with a 20% 768 * default boost can still drop its own boosting to 0%. 769 */ 770 struct uclamp_se { 771 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 772 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 773 unsigned int active : 1; 774 unsigned int user_defined : 1; 775 }; 776 #endif /* CONFIG_UCLAMP_TASK */ 777 778 union rcu_special { 779 struct { 780 u8 blocked; 781 u8 need_qs; 782 u8 exp_hint; /* Hint for performance. */ 783 u8 need_mb; /* Readers need smp_mb(). */ 784 } b; /* Bits. */ 785 u32 s; /* Set of bits. */ 786 }; 787 788 enum perf_event_task_context { 789 perf_invalid_context = -1, 790 perf_hw_context = 0, 791 perf_sw_context, 792 perf_nr_task_contexts, 793 }; 794 795 /* 796 * Number of contexts where an event can trigger: 797 * task, softirq, hardirq, nmi. 798 */ 799 #define PERF_NR_CONTEXTS 4 800 801 struct wake_q_node { 802 struct wake_q_node *next; 803 }; 804 805 struct kmap_ctrl { 806 #ifdef CONFIG_KMAP_LOCAL 807 int idx; 808 pte_t pteval[KM_MAX_IDX]; 809 #endif 810 }; 811 812 struct task_struct { 813 #ifdef CONFIG_THREAD_INFO_IN_TASK 814 /* 815 * For reasons of header soup (see current_thread_info()), this 816 * must be the first element of task_struct. 817 */ 818 struct thread_info thread_info; 819 #endif 820 unsigned int __state; 821 822 /* saved state for "spinlock sleepers" */ 823 unsigned int saved_state; 824 825 /* 826 * This begins the randomizable portion of task_struct. Only 827 * scheduling-critical items should be added above here. 828 */ 829 randomized_struct_fields_start 830 831 void *stack; 832 refcount_t usage; 833 /* Per task flags (PF_*), defined further below: */ 834 unsigned int flags; 835 unsigned int ptrace; 836 837 #ifdef CONFIG_MEM_ALLOC_PROFILING 838 struct alloc_tag *alloc_tag; 839 #endif 840 841 #ifdef CONFIG_SMP 842 int on_cpu; 843 struct __call_single_node wake_entry; 844 unsigned int wakee_flips; 845 unsigned long wakee_flip_decay_ts; 846 struct task_struct *last_wakee; 847 848 /* 849 * recent_used_cpu is initially set as the last CPU used by a task 850 * that wakes affine another task. Waker/wakee relationships can 851 * push tasks around a CPU where each wakeup moves to the next one. 852 * Tracking a recently used CPU allows a quick search for a recently 853 * used CPU that may be idle. 854 */ 855 int recent_used_cpu; 856 int wake_cpu; 857 #endif 858 int on_rq; 859 860 int prio; 861 int static_prio; 862 int normal_prio; 863 unsigned int rt_priority; 864 865 struct sched_entity se; 866 struct sched_rt_entity rt; 867 struct sched_dl_entity dl; 868 struct sched_dl_entity *dl_server; 869 #ifdef CONFIG_SCHED_CLASS_EXT 870 struct sched_ext_entity scx; 871 #endif 872 const struct sched_class *sched_class; 873 874 #ifdef CONFIG_SCHED_CORE 875 struct rb_node core_node; 876 unsigned long core_cookie; 877 unsigned int core_occupation; 878 #endif 879 880 #ifdef CONFIG_CGROUP_SCHED 881 struct task_group *sched_task_group; 882 #endif 883 884 885 #ifdef CONFIG_UCLAMP_TASK 886 /* 887 * Clamp values requested for a scheduling entity. 888 * Must be updated with task_rq_lock() held. 889 */ 890 struct uclamp_se uclamp_req[UCLAMP_CNT]; 891 /* 892 * Effective clamp values used for a scheduling entity. 893 * Must be updated with task_rq_lock() held. 894 */ 895 struct uclamp_se uclamp[UCLAMP_CNT]; 896 #endif 897 898 struct sched_statistics stats; 899 900 #ifdef CONFIG_PREEMPT_NOTIFIERS 901 /* List of struct preempt_notifier: */ 902 struct hlist_head preempt_notifiers; 903 #endif 904 905 #ifdef CONFIG_BLK_DEV_IO_TRACE 906 unsigned int btrace_seq; 907 #endif 908 909 unsigned int policy; 910 unsigned long max_allowed_capacity; 911 int nr_cpus_allowed; 912 const cpumask_t *cpus_ptr; 913 cpumask_t *user_cpus_ptr; 914 cpumask_t cpus_mask; 915 void *migration_pending; 916 #ifdef CONFIG_SMP 917 unsigned short migration_disabled; 918 #endif 919 unsigned short migration_flags; 920 921 #ifdef CONFIG_PREEMPT_RCU 922 int rcu_read_lock_nesting; 923 union rcu_special rcu_read_unlock_special; 924 struct list_head rcu_node_entry; 925 struct rcu_node *rcu_blocked_node; 926 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 927 928 #ifdef CONFIG_TASKS_RCU 929 unsigned long rcu_tasks_nvcsw; 930 u8 rcu_tasks_holdout; 931 u8 rcu_tasks_idx; 932 int rcu_tasks_idle_cpu; 933 struct list_head rcu_tasks_holdout_list; 934 int rcu_tasks_exit_cpu; 935 struct list_head rcu_tasks_exit_list; 936 #endif /* #ifdef CONFIG_TASKS_RCU */ 937 938 #ifdef CONFIG_TASKS_TRACE_RCU 939 int trc_reader_nesting; 940 int trc_ipi_to_cpu; 941 union rcu_special trc_reader_special; 942 struct list_head trc_holdout_list; 943 struct list_head trc_blkd_node; 944 int trc_blkd_cpu; 945 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 946 947 struct sched_info sched_info; 948 949 struct list_head tasks; 950 #ifdef CONFIG_SMP 951 struct plist_node pushable_tasks; 952 struct rb_node pushable_dl_tasks; 953 #endif 954 955 struct mm_struct *mm; 956 struct mm_struct *active_mm; 957 struct address_space *faults_disabled_mapping; 958 959 int exit_state; 960 int exit_code; 961 int exit_signal; 962 /* The signal sent when the parent dies: */ 963 int pdeath_signal; 964 /* JOBCTL_*, siglock protected: */ 965 unsigned long jobctl; 966 967 /* Used for emulating ABI behavior of previous Linux versions: */ 968 unsigned int personality; 969 970 /* Scheduler bits, serialized by scheduler locks: */ 971 unsigned sched_reset_on_fork:1; 972 unsigned sched_contributes_to_load:1; 973 unsigned sched_migrated:1; 974 unsigned sched_task_hot:1; 975 976 /* Force alignment to the next boundary: */ 977 unsigned :0; 978 979 /* Unserialized, strictly 'current' */ 980 981 /* 982 * This field must not be in the scheduler word above due to wakelist 983 * queueing no longer being serialized by p->on_cpu. However: 984 * 985 * p->XXX = X; ttwu() 986 * schedule() if (p->on_rq && ..) // false 987 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 988 * deactivate_task() ttwu_queue_wakelist()) 989 * p->on_rq = 0; p->sched_remote_wakeup = Y; 990 * 991 * guarantees all stores of 'current' are visible before 992 * ->sched_remote_wakeup gets used, so it can be in this word. 993 */ 994 unsigned sched_remote_wakeup:1; 995 #ifdef CONFIG_RT_MUTEXES 996 unsigned sched_rt_mutex:1; 997 #endif 998 999 /* Bit to tell TOMOYO we're in execve(): */ 1000 unsigned in_execve:1; 1001 unsigned in_iowait:1; 1002 #ifndef TIF_RESTORE_SIGMASK 1003 unsigned restore_sigmask:1; 1004 #endif 1005 #ifdef CONFIG_MEMCG_V1 1006 unsigned in_user_fault:1; 1007 #endif 1008 #ifdef CONFIG_LRU_GEN 1009 /* whether the LRU algorithm may apply to this access */ 1010 unsigned in_lru_fault:1; 1011 #endif 1012 #ifdef CONFIG_COMPAT_BRK 1013 unsigned brk_randomized:1; 1014 #endif 1015 #ifdef CONFIG_CGROUPS 1016 /* disallow userland-initiated cgroup migration */ 1017 unsigned no_cgroup_migration:1; 1018 /* task is frozen/stopped (used by the cgroup freezer) */ 1019 unsigned frozen:1; 1020 #endif 1021 #ifdef CONFIG_BLK_CGROUP 1022 unsigned use_memdelay:1; 1023 #endif 1024 #ifdef CONFIG_PSI 1025 /* Stalled due to lack of memory */ 1026 unsigned in_memstall:1; 1027 #endif 1028 #ifdef CONFIG_PAGE_OWNER 1029 /* Used by page_owner=on to detect recursion in page tracking. */ 1030 unsigned in_page_owner:1; 1031 #endif 1032 #ifdef CONFIG_EVENTFD 1033 /* Recursion prevention for eventfd_signal() */ 1034 unsigned in_eventfd:1; 1035 #endif 1036 #ifdef CONFIG_ARCH_HAS_CPU_PASID 1037 unsigned pasid_activated:1; 1038 #endif 1039 #ifdef CONFIG_X86_BUS_LOCK_DETECT 1040 unsigned reported_split_lock:1; 1041 #endif 1042 #ifdef CONFIG_TASK_DELAY_ACCT 1043 /* delay due to memory thrashing */ 1044 unsigned in_thrashing:1; 1045 #endif 1046 #ifdef CONFIG_PREEMPT_RT 1047 struct netdev_xmit net_xmit; 1048 #endif 1049 unsigned long atomic_flags; /* Flags requiring atomic access. */ 1050 1051 struct restart_block restart_block; 1052 1053 pid_t pid; 1054 pid_t tgid; 1055 1056 #ifdef CONFIG_STACKPROTECTOR 1057 /* Canary value for the -fstack-protector GCC feature: */ 1058 unsigned long stack_canary; 1059 #endif 1060 /* 1061 * Pointers to the (original) parent process, youngest child, younger sibling, 1062 * older sibling, respectively. (p->father can be replaced with 1063 * p->real_parent->pid) 1064 */ 1065 1066 /* Real parent process: */ 1067 struct task_struct __rcu *real_parent; 1068 1069 /* Recipient of SIGCHLD, wait4() reports: */ 1070 struct task_struct __rcu *parent; 1071 1072 /* 1073 * Children/sibling form the list of natural children: 1074 */ 1075 struct list_head children; 1076 struct list_head sibling; 1077 struct task_struct *group_leader; 1078 1079 /* 1080 * 'ptraced' is the list of tasks this task is using ptrace() on. 1081 * 1082 * This includes both natural children and PTRACE_ATTACH targets. 1083 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1084 */ 1085 struct list_head ptraced; 1086 struct list_head ptrace_entry; 1087 1088 /* PID/PID hash table linkage. */ 1089 struct pid *thread_pid; 1090 struct hlist_node pid_links[PIDTYPE_MAX]; 1091 struct list_head thread_node; 1092 1093 struct completion *vfork_done; 1094 1095 /* CLONE_CHILD_SETTID: */ 1096 int __user *set_child_tid; 1097 1098 /* CLONE_CHILD_CLEARTID: */ 1099 int __user *clear_child_tid; 1100 1101 /* PF_KTHREAD | PF_IO_WORKER */ 1102 void *worker_private; 1103 1104 u64 utime; 1105 u64 stime; 1106 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1107 u64 utimescaled; 1108 u64 stimescaled; 1109 #endif 1110 u64 gtime; 1111 struct prev_cputime prev_cputime; 1112 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1113 struct vtime vtime; 1114 #endif 1115 1116 #ifdef CONFIG_NO_HZ_FULL 1117 atomic_t tick_dep_mask; 1118 #endif 1119 /* Context switch counts: */ 1120 unsigned long nvcsw; 1121 unsigned long nivcsw; 1122 1123 /* Monotonic time in nsecs: */ 1124 u64 start_time; 1125 1126 /* Boot based time in nsecs: */ 1127 u64 start_boottime; 1128 1129 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1130 unsigned long min_flt; 1131 unsigned long maj_flt; 1132 1133 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1134 struct posix_cputimers posix_cputimers; 1135 1136 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1137 struct posix_cputimers_work posix_cputimers_work; 1138 #endif 1139 1140 /* Process credentials: */ 1141 1142 /* Tracer's credentials at attach: */ 1143 const struct cred __rcu *ptracer_cred; 1144 1145 /* Objective and real subjective task credentials (COW): */ 1146 const struct cred __rcu *real_cred; 1147 1148 /* Effective (overridable) subjective task credentials (COW): */ 1149 const struct cred __rcu *cred; 1150 1151 #ifdef CONFIG_KEYS 1152 /* Cached requested key. */ 1153 struct key *cached_requested_key; 1154 #endif 1155 1156 /* 1157 * executable name, excluding path. 1158 * 1159 * - normally initialized begin_new_exec() 1160 * - set it with set_task_comm() 1161 * - strscpy_pad() to ensure it is always NUL-terminated and 1162 * zero-padded 1163 * - task_lock() to ensure the operation is atomic and the name is 1164 * fully updated. 1165 */ 1166 char comm[TASK_COMM_LEN]; 1167 1168 struct nameidata *nameidata; 1169 1170 #ifdef CONFIG_SYSVIPC 1171 struct sysv_sem sysvsem; 1172 struct sysv_shm sysvshm; 1173 #endif 1174 #ifdef CONFIG_DETECT_HUNG_TASK 1175 unsigned long last_switch_count; 1176 unsigned long last_switch_time; 1177 #endif 1178 /* Filesystem information: */ 1179 struct fs_struct *fs; 1180 1181 /* Open file information: */ 1182 struct files_struct *files; 1183 1184 #ifdef CONFIG_IO_URING 1185 struct io_uring_task *io_uring; 1186 #endif 1187 1188 /* Namespaces: */ 1189 struct nsproxy *nsproxy; 1190 1191 /* Signal handlers: */ 1192 struct signal_struct *signal; 1193 struct sighand_struct __rcu *sighand; 1194 sigset_t blocked; 1195 sigset_t real_blocked; 1196 /* Restored if set_restore_sigmask() was used: */ 1197 sigset_t saved_sigmask; 1198 struct sigpending pending; 1199 unsigned long sas_ss_sp; 1200 size_t sas_ss_size; 1201 unsigned int sas_ss_flags; 1202 1203 struct callback_head *task_works; 1204 1205 #ifdef CONFIG_AUDIT 1206 #ifdef CONFIG_AUDITSYSCALL 1207 struct audit_context *audit_context; 1208 #endif 1209 kuid_t loginuid; 1210 unsigned int sessionid; 1211 #endif 1212 struct seccomp seccomp; 1213 struct syscall_user_dispatch syscall_dispatch; 1214 1215 /* Thread group tracking: */ 1216 u64 parent_exec_id; 1217 u64 self_exec_id; 1218 1219 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1220 spinlock_t alloc_lock; 1221 1222 /* Protection of the PI data structures: */ 1223 raw_spinlock_t pi_lock; 1224 1225 struct wake_q_node wake_q; 1226 1227 #ifdef CONFIG_RT_MUTEXES 1228 /* PI waiters blocked on a rt_mutex held by this task: */ 1229 struct rb_root_cached pi_waiters; 1230 /* Updated under owner's pi_lock and rq lock */ 1231 struct task_struct *pi_top_task; 1232 /* Deadlock detection and priority inheritance handling: */ 1233 struct rt_mutex_waiter *pi_blocked_on; 1234 #endif 1235 1236 #ifdef CONFIG_DEBUG_MUTEXES 1237 /* Mutex deadlock detection: */ 1238 struct mutex_waiter *blocked_on; 1239 #endif 1240 1241 #ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER 1242 struct mutex *blocker_mutex; 1243 #endif 1244 1245 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1246 int non_block_count; 1247 #endif 1248 1249 #ifdef CONFIG_TRACE_IRQFLAGS 1250 struct irqtrace_events irqtrace; 1251 unsigned int hardirq_threaded; 1252 u64 hardirq_chain_key; 1253 int softirqs_enabled; 1254 int softirq_context; 1255 int irq_config; 1256 #endif 1257 #ifdef CONFIG_PREEMPT_RT 1258 int softirq_disable_cnt; 1259 #endif 1260 1261 #ifdef CONFIG_LOCKDEP 1262 # define MAX_LOCK_DEPTH 48UL 1263 u64 curr_chain_key; 1264 int lockdep_depth; 1265 unsigned int lockdep_recursion; 1266 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1267 #endif 1268 1269 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1270 unsigned int in_ubsan; 1271 #endif 1272 1273 /* Journalling filesystem info: */ 1274 void *journal_info; 1275 1276 /* Stacked block device info: */ 1277 struct bio_list *bio_list; 1278 1279 /* Stack plugging: */ 1280 struct blk_plug *plug; 1281 1282 /* VM state: */ 1283 struct reclaim_state *reclaim_state; 1284 1285 struct io_context *io_context; 1286 1287 #ifdef CONFIG_COMPACTION 1288 struct capture_control *capture_control; 1289 #endif 1290 /* Ptrace state: */ 1291 unsigned long ptrace_message; 1292 kernel_siginfo_t *last_siginfo; 1293 1294 struct task_io_accounting ioac; 1295 #ifdef CONFIG_PSI 1296 /* Pressure stall state */ 1297 unsigned int psi_flags; 1298 #endif 1299 #ifdef CONFIG_TASK_XACCT 1300 /* Accumulated RSS usage: */ 1301 u64 acct_rss_mem1; 1302 /* Accumulated virtual memory usage: */ 1303 u64 acct_vm_mem1; 1304 /* stime + utime since last update: */ 1305 u64 acct_timexpd; 1306 #endif 1307 #ifdef CONFIG_CPUSETS 1308 /* Protected by ->alloc_lock: */ 1309 nodemask_t mems_allowed; 1310 /* Sequence number to catch updates: */ 1311 seqcount_spinlock_t mems_allowed_seq; 1312 int cpuset_mem_spread_rotor; 1313 #endif 1314 #ifdef CONFIG_CGROUPS 1315 /* Control Group info protected by css_set_lock: */ 1316 struct css_set __rcu *cgroups; 1317 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1318 struct list_head cg_list; 1319 #endif 1320 #ifdef CONFIG_X86_CPU_RESCTRL 1321 u32 closid; 1322 u32 rmid; 1323 #endif 1324 #ifdef CONFIG_FUTEX 1325 struct robust_list_head __user *robust_list; 1326 #ifdef CONFIG_COMPAT 1327 struct compat_robust_list_head __user *compat_robust_list; 1328 #endif 1329 struct list_head pi_state_list; 1330 struct futex_pi_state *pi_state_cache; 1331 struct mutex futex_exit_mutex; 1332 unsigned int futex_state; 1333 #endif 1334 #ifdef CONFIG_PERF_EVENTS 1335 u8 perf_recursion[PERF_NR_CONTEXTS]; 1336 struct perf_event_context *perf_event_ctxp; 1337 struct mutex perf_event_mutex; 1338 struct list_head perf_event_list; 1339 struct perf_ctx_data __rcu *perf_ctx_data; 1340 #endif 1341 #ifdef CONFIG_DEBUG_PREEMPT 1342 unsigned long preempt_disable_ip; 1343 #endif 1344 #ifdef CONFIG_NUMA 1345 /* Protected by alloc_lock: */ 1346 struct mempolicy *mempolicy; 1347 short il_prev; 1348 u8 il_weight; 1349 short pref_node_fork; 1350 #endif 1351 #ifdef CONFIG_NUMA_BALANCING 1352 int numa_scan_seq; 1353 unsigned int numa_scan_period; 1354 unsigned int numa_scan_period_max; 1355 int numa_preferred_nid; 1356 unsigned long numa_migrate_retry; 1357 /* Migration stamp: */ 1358 u64 node_stamp; 1359 u64 last_task_numa_placement; 1360 u64 last_sum_exec_runtime; 1361 struct callback_head numa_work; 1362 1363 /* 1364 * This pointer is only modified for current in syscall and 1365 * pagefault context (and for tasks being destroyed), so it can be read 1366 * from any of the following contexts: 1367 * - RCU read-side critical section 1368 * - current->numa_group from everywhere 1369 * - task's runqueue locked, task not running 1370 */ 1371 struct numa_group __rcu *numa_group; 1372 1373 /* 1374 * numa_faults is an array split into four regions: 1375 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1376 * in this precise order. 1377 * 1378 * faults_memory: Exponential decaying average of faults on a per-node 1379 * basis. Scheduling placement decisions are made based on these 1380 * counts. The values remain static for the duration of a PTE scan. 1381 * faults_cpu: Track the nodes the process was running on when a NUMA 1382 * hinting fault was incurred. 1383 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1384 * during the current scan window. When the scan completes, the counts 1385 * in faults_memory and faults_cpu decay and these values are copied. 1386 */ 1387 unsigned long *numa_faults; 1388 unsigned long total_numa_faults; 1389 1390 /* 1391 * numa_faults_locality tracks if faults recorded during the last 1392 * scan window were remote/local or failed to migrate. The task scan 1393 * period is adapted based on the locality of the faults with different 1394 * weights depending on whether they were shared or private faults 1395 */ 1396 unsigned long numa_faults_locality[3]; 1397 1398 unsigned long numa_pages_migrated; 1399 #endif /* CONFIG_NUMA_BALANCING */ 1400 1401 #ifdef CONFIG_RSEQ 1402 struct rseq __user *rseq; 1403 u32 rseq_len; 1404 u32 rseq_sig; 1405 /* 1406 * RmW on rseq_event_mask must be performed atomically 1407 * with respect to preemption. 1408 */ 1409 unsigned long rseq_event_mask; 1410 # ifdef CONFIG_DEBUG_RSEQ 1411 /* 1412 * This is a place holder to save a copy of the rseq fields for 1413 * validation of read-only fields. The struct rseq has a 1414 * variable-length array at the end, so it cannot be used 1415 * directly. Reserve a size large enough for the known fields. 1416 */ 1417 char rseq_fields[sizeof(struct rseq)]; 1418 # endif 1419 #endif 1420 1421 #ifdef CONFIG_SCHED_MM_CID 1422 int mm_cid; /* Current cid in mm */ 1423 int last_mm_cid; /* Most recent cid in mm */ 1424 int migrate_from_cpu; 1425 int mm_cid_active; /* Whether cid bitmap is active */ 1426 struct callback_head cid_work; 1427 #endif 1428 1429 struct tlbflush_unmap_batch tlb_ubc; 1430 1431 /* Cache last used pipe for splice(): */ 1432 struct pipe_inode_info *splice_pipe; 1433 1434 struct page_frag task_frag; 1435 1436 #ifdef CONFIG_TASK_DELAY_ACCT 1437 struct task_delay_info *delays; 1438 #endif 1439 1440 #ifdef CONFIG_FAULT_INJECTION 1441 int make_it_fail; 1442 unsigned int fail_nth; 1443 #endif 1444 /* 1445 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1446 * balance_dirty_pages() for a dirty throttling pause: 1447 */ 1448 int nr_dirtied; 1449 int nr_dirtied_pause; 1450 /* Start of a write-and-pause period: */ 1451 unsigned long dirty_paused_when; 1452 1453 #ifdef CONFIG_LATENCYTOP 1454 int latency_record_count; 1455 struct latency_record latency_record[LT_SAVECOUNT]; 1456 #endif 1457 /* 1458 * Time slack values; these are used to round up poll() and 1459 * select() etc timeout values. These are in nanoseconds. 1460 */ 1461 u64 timer_slack_ns; 1462 u64 default_timer_slack_ns; 1463 1464 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1465 unsigned int kasan_depth; 1466 #endif 1467 1468 #ifdef CONFIG_KCSAN 1469 struct kcsan_ctx kcsan_ctx; 1470 #ifdef CONFIG_TRACE_IRQFLAGS 1471 struct irqtrace_events kcsan_save_irqtrace; 1472 #endif 1473 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1474 int kcsan_stack_depth; 1475 #endif 1476 #endif 1477 1478 #ifdef CONFIG_KMSAN 1479 struct kmsan_ctx kmsan_ctx; 1480 #endif 1481 1482 #if IS_ENABLED(CONFIG_KUNIT) 1483 struct kunit *kunit_test; 1484 #endif 1485 1486 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1487 /* Index of current stored address in ret_stack: */ 1488 int curr_ret_stack; 1489 int curr_ret_depth; 1490 1491 /* Stack of return addresses for return function tracing: */ 1492 unsigned long *ret_stack; 1493 1494 /* Timestamp for last schedule: */ 1495 unsigned long long ftrace_timestamp; 1496 unsigned long long ftrace_sleeptime; 1497 1498 /* 1499 * Number of functions that haven't been traced 1500 * because of depth overrun: 1501 */ 1502 atomic_t trace_overrun; 1503 1504 /* Pause tracing: */ 1505 atomic_t tracing_graph_pause; 1506 #endif 1507 1508 #ifdef CONFIG_TRACING 1509 /* Bitmask and counter of trace recursion: */ 1510 unsigned long trace_recursion; 1511 #endif /* CONFIG_TRACING */ 1512 1513 #ifdef CONFIG_KCOV 1514 /* See kernel/kcov.c for more details. */ 1515 1516 /* Coverage collection mode enabled for this task (0 if disabled): */ 1517 unsigned int kcov_mode; 1518 1519 /* Size of the kcov_area: */ 1520 unsigned int kcov_size; 1521 1522 /* Buffer for coverage collection: */ 1523 void *kcov_area; 1524 1525 /* KCOV descriptor wired with this task or NULL: */ 1526 struct kcov *kcov; 1527 1528 /* KCOV common handle for remote coverage collection: */ 1529 u64 kcov_handle; 1530 1531 /* KCOV sequence number: */ 1532 int kcov_sequence; 1533 1534 /* Collect coverage from softirq context: */ 1535 unsigned int kcov_softirq; 1536 #endif 1537 1538 #ifdef CONFIG_MEMCG_V1 1539 struct mem_cgroup *memcg_in_oom; 1540 #endif 1541 1542 #ifdef CONFIG_MEMCG 1543 /* Number of pages to reclaim on returning to userland: */ 1544 unsigned int memcg_nr_pages_over_high; 1545 1546 /* Used by memcontrol for targeted memcg charge: */ 1547 struct mem_cgroup *active_memcg; 1548 1549 /* Cache for current->cgroups->memcg->objcg lookups: */ 1550 struct obj_cgroup *objcg; 1551 #endif 1552 1553 #ifdef CONFIG_BLK_CGROUP 1554 struct gendisk *throttle_disk; 1555 #endif 1556 1557 #ifdef CONFIG_UPROBES 1558 struct uprobe_task *utask; 1559 #endif 1560 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1561 unsigned int sequential_io; 1562 unsigned int sequential_io_avg; 1563 #endif 1564 struct kmap_ctrl kmap_ctrl; 1565 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1566 unsigned long task_state_change; 1567 # ifdef CONFIG_PREEMPT_RT 1568 unsigned long saved_state_change; 1569 # endif 1570 #endif 1571 struct rcu_head rcu; 1572 refcount_t rcu_users; 1573 int pagefault_disabled; 1574 #ifdef CONFIG_MMU 1575 struct task_struct *oom_reaper_list; 1576 struct timer_list oom_reaper_timer; 1577 #endif 1578 #ifdef CONFIG_VMAP_STACK 1579 struct vm_struct *stack_vm_area; 1580 #endif 1581 #ifdef CONFIG_THREAD_INFO_IN_TASK 1582 /* A live task holds one reference: */ 1583 refcount_t stack_refcount; 1584 #endif 1585 #ifdef CONFIG_LIVEPATCH 1586 int patch_state; 1587 #endif 1588 #ifdef CONFIG_SECURITY 1589 /* Used by LSM modules for access restriction: */ 1590 void *security; 1591 #endif 1592 #ifdef CONFIG_BPF_SYSCALL 1593 /* Used by BPF task local storage */ 1594 struct bpf_local_storage __rcu *bpf_storage; 1595 /* Used for BPF run context */ 1596 struct bpf_run_ctx *bpf_ctx; 1597 #endif 1598 /* Used by BPF for per-TASK xdp storage */ 1599 struct bpf_net_context *bpf_net_context; 1600 1601 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1602 unsigned long lowest_stack; 1603 unsigned long prev_lowest_stack; 1604 #endif 1605 1606 #ifdef CONFIG_X86_MCE 1607 void __user *mce_vaddr; 1608 __u64 mce_kflags; 1609 u64 mce_addr; 1610 __u64 mce_ripv : 1, 1611 mce_whole_page : 1, 1612 __mce_reserved : 62; 1613 struct callback_head mce_kill_me; 1614 int mce_count; 1615 #endif 1616 1617 #ifdef CONFIG_KRETPROBES 1618 struct llist_head kretprobe_instances; 1619 #endif 1620 #ifdef CONFIG_RETHOOK 1621 struct llist_head rethooks; 1622 #endif 1623 1624 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1625 /* 1626 * If L1D flush is supported on mm context switch 1627 * then we use this callback head to queue kill work 1628 * to kill tasks that are not running on SMT disabled 1629 * cores 1630 */ 1631 struct callback_head l1d_flush_kill; 1632 #endif 1633 1634 #ifdef CONFIG_RV 1635 /* 1636 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS. 1637 * If we find justification for more monitors, we can think 1638 * about adding more or developing a dynamic method. So far, 1639 * none of these are justified. 1640 */ 1641 union rv_task_monitor rv[RV_PER_TASK_MONITORS]; 1642 #endif 1643 1644 #ifdef CONFIG_USER_EVENTS 1645 struct user_event_mm *user_event_mm; 1646 #endif 1647 1648 /* CPU-specific state of this task: */ 1649 struct thread_struct thread; 1650 1651 /* 1652 * New fields for task_struct should be added above here, so that 1653 * they are included in the randomized portion of task_struct. 1654 */ 1655 randomized_struct_fields_end 1656 } __attribute__ ((aligned (64))); 1657 1658 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1659 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1660 1661 static inline unsigned int __task_state_index(unsigned int tsk_state, 1662 unsigned int tsk_exit_state) 1663 { 1664 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1665 1666 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1667 1668 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1669 state = TASK_REPORT_IDLE; 1670 1671 /* 1672 * We're lying here, but rather than expose a completely new task state 1673 * to userspace, we can make this appear as if the task has gone through 1674 * a regular rt_mutex_lock() call. 1675 * Report frozen tasks as uninterruptible. 1676 */ 1677 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN)) 1678 state = TASK_UNINTERRUPTIBLE; 1679 1680 return fls(state); 1681 } 1682 1683 static inline unsigned int task_state_index(struct task_struct *tsk) 1684 { 1685 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1686 } 1687 1688 static inline char task_index_to_char(unsigned int state) 1689 { 1690 static const char state_char[] = "RSDTtXZPI"; 1691 1692 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1)); 1693 1694 return state_char[state]; 1695 } 1696 1697 static inline char task_state_to_char(struct task_struct *tsk) 1698 { 1699 return task_index_to_char(task_state_index(tsk)); 1700 } 1701 1702 extern struct pid *cad_pid; 1703 1704 /* 1705 * Per process flags 1706 */ 1707 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1708 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1709 #define PF_EXITING 0x00000004 /* Getting shut down */ 1710 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1711 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1712 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1713 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1714 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1715 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1716 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1717 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1718 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1719 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1720 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1721 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1722 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1723 #define PF_KCOMPACTD 0x00010000 /* I am kcompactd */ 1724 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1725 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1726 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1727 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1728 * I am cleaning dirty pages from some other bdi. */ 1729 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1730 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1731 #define PF__HOLE__00800000 0x00800000 1732 #define PF__HOLE__01000000 0x01000000 1733 #define PF__HOLE__02000000 0x02000000 1734 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1735 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1736 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1737 * See memalloc_pin_save() */ 1738 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1739 #define PF__HOLE__40000000 0x40000000 1740 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1741 1742 /* 1743 * Only the _current_ task can read/write to tsk->flags, but other 1744 * tasks can access tsk->flags in readonly mode for example 1745 * with tsk_used_math (like during threaded core dumping). 1746 * There is however an exception to this rule during ptrace 1747 * or during fork: the ptracer task is allowed to write to the 1748 * child->flags of its traced child (same goes for fork, the parent 1749 * can write to the child->flags), because we're guaranteed the 1750 * child is not running and in turn not changing child->flags 1751 * at the same time the parent does it. 1752 */ 1753 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1754 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1755 #define clear_used_math() clear_stopped_child_used_math(current) 1756 #define set_used_math() set_stopped_child_used_math(current) 1757 1758 #define conditional_stopped_child_used_math(condition, child) \ 1759 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1760 1761 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1762 1763 #define copy_to_stopped_child_used_math(child) \ 1764 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1765 1766 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1767 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1768 #define used_math() tsk_used_math(current) 1769 1770 static __always_inline bool is_percpu_thread(void) 1771 { 1772 #ifdef CONFIG_SMP 1773 return (current->flags & PF_NO_SETAFFINITY) && 1774 (current->nr_cpus_allowed == 1); 1775 #else 1776 return true; 1777 #endif 1778 } 1779 1780 /* Per-process atomic flags. */ 1781 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1782 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1783 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1784 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1785 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1786 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1787 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1788 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1789 1790 #define TASK_PFA_TEST(name, func) \ 1791 static inline bool task_##func(struct task_struct *p) \ 1792 { return test_bit(PFA_##name, &p->atomic_flags); } 1793 1794 #define TASK_PFA_SET(name, func) \ 1795 static inline void task_set_##func(struct task_struct *p) \ 1796 { set_bit(PFA_##name, &p->atomic_flags); } 1797 1798 #define TASK_PFA_CLEAR(name, func) \ 1799 static inline void task_clear_##func(struct task_struct *p) \ 1800 { clear_bit(PFA_##name, &p->atomic_flags); } 1801 1802 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1803 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1804 1805 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1806 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1807 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1808 1809 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1810 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1811 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1812 1813 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1814 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1815 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1816 1817 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1818 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1819 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1820 1821 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1822 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1823 1824 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1825 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1826 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1827 1828 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1829 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1830 1831 static inline void 1832 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1833 { 1834 current->flags &= ~flags; 1835 current->flags |= orig_flags & flags; 1836 } 1837 1838 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1839 extern int task_can_attach(struct task_struct *p); 1840 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1841 extern void dl_bw_free(int cpu, u64 dl_bw); 1842 #ifdef CONFIG_SMP 1843 1844 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1845 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1846 1847 /** 1848 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1849 * @p: the task 1850 * @new_mask: CPU affinity mask 1851 * 1852 * Return: zero if successful, or a negative error code 1853 */ 1854 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1855 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1856 extern void release_user_cpus_ptr(struct task_struct *p); 1857 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1858 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1859 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1860 #else 1861 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1862 { 1863 } 1864 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1865 { 1866 /* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */ 1867 if ((*cpumask_bits(new_mask) & 1) == 0) 1868 return -EINVAL; 1869 return 0; 1870 } 1871 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node) 1872 { 1873 if (src->user_cpus_ptr) 1874 return -EINVAL; 1875 return 0; 1876 } 1877 static inline void release_user_cpus_ptr(struct task_struct *p) 1878 { 1879 WARN_ON(p->user_cpus_ptr); 1880 } 1881 1882 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) 1883 { 1884 return 0; 1885 } 1886 #endif 1887 1888 extern int yield_to(struct task_struct *p, bool preempt); 1889 extern void set_user_nice(struct task_struct *p, long nice); 1890 extern int task_prio(const struct task_struct *p); 1891 1892 /** 1893 * task_nice - return the nice value of a given task. 1894 * @p: the task in question. 1895 * 1896 * Return: The nice value [ -20 ... 0 ... 19 ]. 1897 */ 1898 static inline int task_nice(const struct task_struct *p) 1899 { 1900 return PRIO_TO_NICE((p)->static_prio); 1901 } 1902 1903 extern int can_nice(const struct task_struct *p, const int nice); 1904 extern int task_curr(const struct task_struct *p); 1905 extern int idle_cpu(int cpu); 1906 extern int available_idle_cpu(int cpu); 1907 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1908 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1909 extern void sched_set_fifo(struct task_struct *p); 1910 extern void sched_set_fifo_low(struct task_struct *p); 1911 extern void sched_set_normal(struct task_struct *p, int nice); 1912 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1913 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1914 extern struct task_struct *idle_task(int cpu); 1915 1916 /** 1917 * is_idle_task - is the specified task an idle task? 1918 * @p: the task in question. 1919 * 1920 * Return: 1 if @p is an idle task. 0 otherwise. 1921 */ 1922 static __always_inline bool is_idle_task(const struct task_struct *p) 1923 { 1924 return !!(p->flags & PF_IDLE); 1925 } 1926 1927 extern struct task_struct *curr_task(int cpu); 1928 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1929 1930 void yield(void); 1931 1932 union thread_union { 1933 struct task_struct task; 1934 #ifndef CONFIG_THREAD_INFO_IN_TASK 1935 struct thread_info thread_info; 1936 #endif 1937 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1938 }; 1939 1940 #ifndef CONFIG_THREAD_INFO_IN_TASK 1941 extern struct thread_info init_thread_info; 1942 #endif 1943 1944 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1945 1946 #ifdef CONFIG_THREAD_INFO_IN_TASK 1947 # define task_thread_info(task) (&(task)->thread_info) 1948 #else 1949 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1950 #endif 1951 1952 /* 1953 * find a task by one of its numerical ids 1954 * 1955 * find_task_by_pid_ns(): 1956 * finds a task by its pid in the specified namespace 1957 * find_task_by_vpid(): 1958 * finds a task by its virtual pid 1959 * 1960 * see also find_vpid() etc in include/linux/pid.h 1961 */ 1962 1963 extern struct task_struct *find_task_by_vpid(pid_t nr); 1964 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1965 1966 /* 1967 * find a task by its virtual pid and get the task struct 1968 */ 1969 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1970 1971 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1972 extern int wake_up_process(struct task_struct *tsk); 1973 extern void wake_up_new_task(struct task_struct *tsk); 1974 1975 #ifdef CONFIG_SMP 1976 extern void kick_process(struct task_struct *tsk); 1977 #else 1978 static inline void kick_process(struct task_struct *tsk) { } 1979 #endif 1980 1981 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1982 #define set_task_comm(tsk, from) ({ \ 1983 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \ 1984 __set_task_comm(tsk, from, false); \ 1985 }) 1986 1987 /* 1988 * - Why not use task_lock()? 1989 * User space can randomly change their names anyway, so locking for readers 1990 * doesn't make sense. For writers, locking is probably necessary, as a race 1991 * condition could lead to long-term mixed results. 1992 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is 1993 * always NUL-terminated and zero-padded. Therefore the race condition between 1994 * reader and writer is not an issue. 1995 * 1996 * - BUILD_BUG_ON() can help prevent the buf from being truncated. 1997 * Since the callers don't perform any return value checks, this safeguard is 1998 * necessary. 1999 */ 2000 #define get_task_comm(buf, tsk) ({ \ 2001 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \ 2002 strscpy_pad(buf, (tsk)->comm); \ 2003 buf; \ 2004 }) 2005 2006 #ifdef CONFIG_SMP 2007 static __always_inline void scheduler_ipi(void) 2008 { 2009 /* 2010 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 2011 * TIF_NEED_RESCHED remotely (for the first time) will also send 2012 * this IPI. 2013 */ 2014 preempt_fold_need_resched(); 2015 } 2016 #else 2017 static inline void scheduler_ipi(void) { } 2018 #endif 2019 2020 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 2021 2022 /* 2023 * Set thread flags in other task's structures. 2024 * See asm/thread_info.h for TIF_xxxx flags available: 2025 */ 2026 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 2027 { 2028 set_ti_thread_flag(task_thread_info(tsk), flag); 2029 } 2030 2031 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2032 { 2033 clear_ti_thread_flag(task_thread_info(tsk), flag); 2034 } 2035 2036 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 2037 bool value) 2038 { 2039 update_ti_thread_flag(task_thread_info(tsk), flag, value); 2040 } 2041 2042 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2043 { 2044 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2045 } 2046 2047 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2048 { 2049 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2050 } 2051 2052 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2053 { 2054 return test_ti_thread_flag(task_thread_info(tsk), flag); 2055 } 2056 2057 static inline void set_tsk_need_resched(struct task_struct *tsk) 2058 { 2059 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2060 } 2061 2062 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2063 { 2064 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY, 2065 (atomic_long_t *)&task_thread_info(tsk)->flags); 2066 } 2067 2068 static inline int test_tsk_need_resched(struct task_struct *tsk) 2069 { 2070 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2071 } 2072 2073 /* 2074 * cond_resched() and cond_resched_lock(): latency reduction via 2075 * explicit rescheduling in places that are safe. The return 2076 * value indicates whether a reschedule was done in fact. 2077 * cond_resched_lock() will drop the spinlock before scheduling, 2078 */ 2079 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2080 extern int __cond_resched(void); 2081 2082 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2083 2084 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2085 2086 static __always_inline int _cond_resched(void) 2087 { 2088 return static_call_mod(cond_resched)(); 2089 } 2090 2091 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2092 2093 extern int dynamic_cond_resched(void); 2094 2095 static __always_inline int _cond_resched(void) 2096 { 2097 return dynamic_cond_resched(); 2098 } 2099 2100 #else /* !CONFIG_PREEMPTION */ 2101 2102 static inline int _cond_resched(void) 2103 { 2104 return __cond_resched(); 2105 } 2106 2107 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2108 2109 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2110 2111 static inline int _cond_resched(void) 2112 { 2113 return 0; 2114 } 2115 2116 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2117 2118 #define cond_resched() ({ \ 2119 __might_resched(__FILE__, __LINE__, 0); \ 2120 _cond_resched(); \ 2121 }) 2122 2123 extern int __cond_resched_lock(spinlock_t *lock); 2124 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2125 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2126 2127 #define MIGHT_RESCHED_RCU_SHIFT 8 2128 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2129 2130 #ifndef CONFIG_PREEMPT_RT 2131 /* 2132 * Non RT kernels have an elevated preempt count due to the held lock, 2133 * but are not allowed to be inside a RCU read side critical section 2134 */ 2135 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2136 #else 2137 /* 2138 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2139 * cond_resched*lock() has to take that into account because it checks for 2140 * preempt_count() and rcu_preempt_depth(). 2141 */ 2142 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2143 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2144 #endif 2145 2146 #define cond_resched_lock(lock) ({ \ 2147 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2148 __cond_resched_lock(lock); \ 2149 }) 2150 2151 #define cond_resched_rwlock_read(lock) ({ \ 2152 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2153 __cond_resched_rwlock_read(lock); \ 2154 }) 2155 2156 #define cond_resched_rwlock_write(lock) ({ \ 2157 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2158 __cond_resched_rwlock_write(lock); \ 2159 }) 2160 2161 static __always_inline bool need_resched(void) 2162 { 2163 return unlikely(tif_need_resched()); 2164 } 2165 2166 /* 2167 * Wrappers for p->thread_info->cpu access. No-op on UP. 2168 */ 2169 #ifdef CONFIG_SMP 2170 2171 static inline unsigned int task_cpu(const struct task_struct *p) 2172 { 2173 return READ_ONCE(task_thread_info(p)->cpu); 2174 } 2175 2176 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2177 2178 #else 2179 2180 static inline unsigned int task_cpu(const struct task_struct *p) 2181 { 2182 return 0; 2183 } 2184 2185 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2186 { 2187 } 2188 2189 #endif /* CONFIG_SMP */ 2190 2191 static inline bool task_is_runnable(struct task_struct *p) 2192 { 2193 return p->on_rq && !p->se.sched_delayed; 2194 } 2195 2196 extern bool sched_task_on_rq(struct task_struct *p); 2197 extern unsigned long get_wchan(struct task_struct *p); 2198 extern struct task_struct *cpu_curr_snapshot(int cpu); 2199 2200 #include <linux/spinlock.h> 2201 2202 /* 2203 * In order to reduce various lock holder preemption latencies provide an 2204 * interface to see if a vCPU is currently running or not. 2205 * 2206 * This allows us to terminate optimistic spin loops and block, analogous to 2207 * the native optimistic spin heuristic of testing if the lock owner task is 2208 * running or not. 2209 */ 2210 #ifndef vcpu_is_preempted 2211 static inline bool vcpu_is_preempted(int cpu) 2212 { 2213 return false; 2214 } 2215 #endif 2216 2217 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2218 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2219 2220 #ifndef TASK_SIZE_OF 2221 #define TASK_SIZE_OF(tsk) TASK_SIZE 2222 #endif 2223 2224 #ifdef CONFIG_SMP 2225 static inline bool owner_on_cpu(struct task_struct *owner) 2226 { 2227 /* 2228 * As lock holder preemption issue, we both skip spinning if 2229 * task is not on cpu or its cpu is preempted 2230 */ 2231 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2232 } 2233 2234 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2235 unsigned long sched_cpu_util(int cpu); 2236 #endif /* CONFIG_SMP */ 2237 2238 #ifdef CONFIG_SCHED_CORE 2239 extern void sched_core_free(struct task_struct *tsk); 2240 extern void sched_core_fork(struct task_struct *p); 2241 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2242 unsigned long uaddr); 2243 extern int sched_core_idle_cpu(int cpu); 2244 #else 2245 static inline void sched_core_free(struct task_struct *tsk) { } 2246 static inline void sched_core_fork(struct task_struct *p) { } 2247 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2248 #endif 2249 2250 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2251 2252 #ifdef CONFIG_MEM_ALLOC_PROFILING 2253 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2254 { 2255 swap(current->alloc_tag, tag); 2256 return tag; 2257 } 2258 2259 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2260 { 2261 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2262 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2263 #endif 2264 current->alloc_tag = old; 2265 } 2266 #else 2267 #define alloc_tag_save(_tag) NULL 2268 #define alloc_tag_restore(_tag, _old) do {} while (0) 2269 #endif 2270 2271 #endif 2272