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.h> 17 18 #include <linux/cache.h> 19 #include <linux/irqflags_types.h> 20 #include <linux/smp_types.h> 21 #include <linux/pid_types.h> 22 #include <linux/sem_types.h> 23 #include <linux/shm.h> 24 #include <linux/kmsan_types.h> 25 #include <linux/mutex_types.h> 26 #include <linux/plist_types.h> 27 #include <linux/hrtimer_types.h> 28 #include <linux/timer_types.h> 29 #include <linux/seccomp_types.h> 30 #include <linux/nodemask_types.h> 31 #include <linux/refcount_types.h> 32 #include <linux/resource.h> 33 #include <linux/latencytop.h> 34 #include <linux/sched/prio.h> 35 #include <linux/sched/types.h> 36 #include <linux/signal_types.h> 37 #include <linux/syscall_user_dispatch_types.h> 38 #include <linux/mm_types_task.h> 39 #include <linux/netdevice_xmit.h> 40 #include <linux/task_io_accounting.h> 41 #include <linux/posix-timers_types.h> 42 #include <linux/restart_block.h> 43 #include <uapi/linux/rseq.h> 44 #include <linux/seqlock_types.h> 45 #include <linux/kcsan.h> 46 #include <linux/rv.h> 47 #include <linux/livepatch_sched.h> 48 #include <linux/uidgid_types.h> 49 #include <asm/kmap_size.h> 50 51 /* task_struct member predeclarations (sorted alphabetically): */ 52 struct audit_context; 53 struct bio_list; 54 struct blk_plug; 55 struct bpf_local_storage; 56 struct bpf_run_ctx; 57 struct bpf_net_context; 58 struct capture_control; 59 struct cfs_rq; 60 struct fs_struct; 61 struct futex_pi_state; 62 struct io_context; 63 struct io_uring_task; 64 struct mempolicy; 65 struct nameidata; 66 struct nsproxy; 67 struct perf_event_context; 68 struct pid_namespace; 69 struct pipe_inode_info; 70 struct rcu_node; 71 struct reclaim_state; 72 struct robust_list_head; 73 struct root_domain; 74 struct rq; 75 struct sched_attr; 76 struct sched_dl_entity; 77 struct seq_file; 78 struct sighand_struct; 79 struct signal_struct; 80 struct task_delay_info; 81 struct task_group; 82 struct task_struct; 83 struct user_event_mm; 84 85 /* 86 * Task state bitmask. NOTE! These bits are also 87 * encoded in fs/proc/array.c: get_task_state(). 88 * 89 * We have two separate sets of flags: task->__state 90 * is about runnability, while task->exit_state are 91 * about the task exiting. Confusing, but this way 92 * modifying one set can't modify the other one by 93 * mistake. 94 */ 95 96 /* Used in tsk->__state: */ 97 #define TASK_RUNNING 0x00000000 98 #define TASK_INTERRUPTIBLE 0x00000001 99 #define TASK_UNINTERRUPTIBLE 0x00000002 100 #define __TASK_STOPPED 0x00000004 101 #define __TASK_TRACED 0x00000008 102 /* Used in tsk->exit_state: */ 103 #define EXIT_DEAD 0x00000010 104 #define EXIT_ZOMBIE 0x00000020 105 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 106 /* Used in tsk->__state again: */ 107 #define TASK_PARKED 0x00000040 108 #define TASK_DEAD 0x00000080 109 #define TASK_WAKEKILL 0x00000100 110 #define TASK_WAKING 0x00000200 111 #define TASK_NOLOAD 0x00000400 112 #define TASK_NEW 0x00000800 113 #define TASK_RTLOCK_WAIT 0x00001000 114 #define TASK_FREEZABLE 0x00002000 115 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP)) 116 #define TASK_FROZEN 0x00008000 117 #define TASK_STATE_MAX 0x00010000 118 119 #define TASK_ANY (TASK_STATE_MAX-1) 120 121 /* 122 * DO NOT ADD ANY NEW USERS ! 123 */ 124 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE) 125 126 /* Convenience macros for the sake of set_current_state: */ 127 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 128 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 129 #define TASK_TRACED __TASK_TRACED 130 131 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 132 133 /* Convenience macros for the sake of wake_up(): */ 134 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 135 136 /* get_task_state(): */ 137 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 138 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 139 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 140 TASK_PARKED) 141 142 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 143 144 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 145 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 146 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 147 148 /* 149 * Special states are those that do not use the normal wait-loop pattern. See 150 * the comment with set_special_state(). 151 */ 152 #define is_special_task_state(state) \ 153 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD)) 154 155 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 156 # define debug_normal_state_change(state_value) \ 157 do { \ 158 WARN_ON_ONCE(is_special_task_state(state_value)); \ 159 current->task_state_change = _THIS_IP_; \ 160 } while (0) 161 162 # define debug_special_state_change(state_value) \ 163 do { \ 164 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 165 current->task_state_change = _THIS_IP_; \ 166 } while (0) 167 168 # define debug_rtlock_wait_set_state() \ 169 do { \ 170 current->saved_state_change = current->task_state_change;\ 171 current->task_state_change = _THIS_IP_; \ 172 } while (0) 173 174 # define debug_rtlock_wait_restore_state() \ 175 do { \ 176 current->task_state_change = current->saved_state_change;\ 177 } while (0) 178 179 #else 180 # define debug_normal_state_change(cond) do { } while (0) 181 # define debug_special_state_change(cond) do { } while (0) 182 # define debug_rtlock_wait_set_state() do { } while (0) 183 # define debug_rtlock_wait_restore_state() do { } while (0) 184 #endif 185 186 /* 187 * set_current_state() includes a barrier so that the write of current->__state 188 * is correctly serialised wrt the caller's subsequent test of whether to 189 * actually sleep: 190 * 191 * for (;;) { 192 * set_current_state(TASK_UNINTERRUPTIBLE); 193 * if (CONDITION) 194 * break; 195 * 196 * schedule(); 197 * } 198 * __set_current_state(TASK_RUNNING); 199 * 200 * If the caller does not need such serialisation (because, for instance, the 201 * CONDITION test and condition change and wakeup are under the same lock) then 202 * use __set_current_state(). 203 * 204 * The above is typically ordered against the wakeup, which does: 205 * 206 * CONDITION = 1; 207 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 208 * 209 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 210 * accessing p->__state. 211 * 212 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is, 213 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 214 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 215 * 216 * However, with slightly different timing the wakeup TASK_RUNNING store can 217 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 218 * a problem either because that will result in one extra go around the loop 219 * and our @cond test will save the day. 220 * 221 * Also see the comments of try_to_wake_up(). 222 */ 223 #define __set_current_state(state_value) \ 224 do { \ 225 debug_normal_state_change((state_value)); \ 226 WRITE_ONCE(current->__state, (state_value)); \ 227 } while (0) 228 229 #define set_current_state(state_value) \ 230 do { \ 231 debug_normal_state_change((state_value)); \ 232 smp_store_mb(current->__state, (state_value)); \ 233 } while (0) 234 235 /* 236 * set_special_state() should be used for those states when the blocking task 237 * can not use the regular condition based wait-loop. In that case we must 238 * serialize against wakeups such that any possible in-flight TASK_RUNNING 239 * stores will not collide with our state change. 240 */ 241 #define set_special_state(state_value) \ 242 do { \ 243 unsigned long flags; /* may shadow */ \ 244 \ 245 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 246 debug_special_state_change((state_value)); \ 247 WRITE_ONCE(current->__state, (state_value)); \ 248 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 249 } while (0) 250 251 /* 252 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 253 * 254 * RT's spin/rwlock substitutions are state preserving. The state of the 255 * task when blocking on the lock is saved in task_struct::saved_state and 256 * restored after the lock has been acquired. These operations are 257 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 258 * lock related wakeups while the task is blocked on the lock are 259 * redirected to operate on task_struct::saved_state to ensure that these 260 * are not dropped. On restore task_struct::saved_state is set to 261 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 262 * 263 * The lock operation looks like this: 264 * 265 * current_save_and_set_rtlock_wait_state(); 266 * for (;;) { 267 * if (try_lock()) 268 * break; 269 * raw_spin_unlock_irq(&lock->wait_lock); 270 * schedule_rtlock(); 271 * raw_spin_lock_irq(&lock->wait_lock); 272 * set_current_state(TASK_RTLOCK_WAIT); 273 * } 274 * current_restore_rtlock_saved_state(); 275 */ 276 #define current_save_and_set_rtlock_wait_state() \ 277 do { \ 278 lockdep_assert_irqs_disabled(); \ 279 raw_spin_lock(¤t->pi_lock); \ 280 current->saved_state = current->__state; \ 281 debug_rtlock_wait_set_state(); \ 282 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 283 raw_spin_unlock(¤t->pi_lock); \ 284 } while (0); 285 286 #define current_restore_rtlock_saved_state() \ 287 do { \ 288 lockdep_assert_irqs_disabled(); \ 289 raw_spin_lock(¤t->pi_lock); \ 290 debug_rtlock_wait_restore_state(); \ 291 WRITE_ONCE(current->__state, current->saved_state); \ 292 current->saved_state = TASK_RUNNING; \ 293 raw_spin_unlock(¤t->pi_lock); \ 294 } while (0); 295 296 #define get_current_state() READ_ONCE(current->__state) 297 298 /* 299 * Define the task command name length as enum, then it can be visible to 300 * BPF programs. 301 */ 302 enum { 303 TASK_COMM_LEN = 16, 304 }; 305 306 extern void sched_tick(void); 307 308 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 309 310 extern long schedule_timeout(long timeout); 311 extern long schedule_timeout_interruptible(long timeout); 312 extern long schedule_timeout_killable(long timeout); 313 extern long schedule_timeout_uninterruptible(long timeout); 314 extern long schedule_timeout_idle(long timeout); 315 asmlinkage void schedule(void); 316 extern void schedule_preempt_disabled(void); 317 asmlinkage void preempt_schedule_irq(void); 318 #ifdef CONFIG_PREEMPT_RT 319 extern void schedule_rtlock(void); 320 #endif 321 322 extern int __must_check io_schedule_prepare(void); 323 extern void io_schedule_finish(int token); 324 extern long io_schedule_timeout(long timeout); 325 extern void io_schedule(void); 326 327 /** 328 * struct prev_cputime - snapshot of system and user cputime 329 * @utime: time spent in user mode 330 * @stime: time spent in system mode 331 * @lock: protects the above two fields 332 * 333 * Stores previous user/system time values such that we can guarantee 334 * monotonicity. 335 */ 336 struct prev_cputime { 337 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 338 u64 utime; 339 u64 stime; 340 raw_spinlock_t lock; 341 #endif 342 }; 343 344 enum vtime_state { 345 /* Task is sleeping or running in a CPU with VTIME inactive: */ 346 VTIME_INACTIVE = 0, 347 /* Task is idle */ 348 VTIME_IDLE, 349 /* Task runs in kernelspace in a CPU with VTIME active: */ 350 VTIME_SYS, 351 /* Task runs in userspace in a CPU with VTIME active: */ 352 VTIME_USER, 353 /* Task runs as guests in a CPU with VTIME active: */ 354 VTIME_GUEST, 355 }; 356 357 struct vtime { 358 seqcount_t seqcount; 359 unsigned long long starttime; 360 enum vtime_state state; 361 unsigned int cpu; 362 u64 utime; 363 u64 stime; 364 u64 gtime; 365 }; 366 367 /* 368 * Utilization clamp constraints. 369 * @UCLAMP_MIN: Minimum utilization 370 * @UCLAMP_MAX: Maximum utilization 371 * @UCLAMP_CNT: Utilization clamp constraints count 372 */ 373 enum uclamp_id { 374 UCLAMP_MIN = 0, 375 UCLAMP_MAX, 376 UCLAMP_CNT 377 }; 378 379 #ifdef CONFIG_SMP 380 extern struct root_domain def_root_domain; 381 extern struct mutex sched_domains_mutex; 382 #endif 383 384 struct sched_param { 385 int sched_priority; 386 }; 387 388 struct sched_info { 389 #ifdef CONFIG_SCHED_INFO 390 /* Cumulative counters: */ 391 392 /* # of times we have run on this CPU: */ 393 unsigned long pcount; 394 395 /* Time spent waiting on a runqueue: */ 396 unsigned long long run_delay; 397 398 /* Timestamps: */ 399 400 /* When did we last run on a CPU? */ 401 unsigned long long last_arrival; 402 403 /* When were we last queued to run? */ 404 unsigned long long last_queued; 405 406 #endif /* CONFIG_SCHED_INFO */ 407 }; 408 409 /* 410 * Integer metrics need fixed point arithmetic, e.g., sched/fair 411 * has a few: load, load_avg, util_avg, freq, and capacity. 412 * 413 * We define a basic fixed point arithmetic range, and then formalize 414 * all these metrics based on that basic range. 415 */ 416 # define SCHED_FIXEDPOINT_SHIFT 10 417 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 418 419 /* Increase resolution of cpu_capacity calculations */ 420 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 421 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 422 423 struct load_weight { 424 unsigned long weight; 425 u32 inv_weight; 426 }; 427 428 /* 429 * The load/runnable/util_avg accumulates an infinite geometric series 430 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 431 * 432 * [load_avg definition] 433 * 434 * load_avg = runnable% * scale_load_down(load) 435 * 436 * [runnable_avg definition] 437 * 438 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 439 * 440 * [util_avg definition] 441 * 442 * util_avg = running% * SCHED_CAPACITY_SCALE 443 * 444 * where runnable% is the time ratio that a sched_entity is runnable and 445 * running% the time ratio that a sched_entity is running. 446 * 447 * For cfs_rq, they are the aggregated values of all runnable and blocked 448 * sched_entities. 449 * 450 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 451 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 452 * for computing those signals (see update_rq_clock_pelt()) 453 * 454 * N.B., the above ratios (runnable% and running%) themselves are in the 455 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 456 * to as large a range as necessary. This is for example reflected by 457 * util_avg's SCHED_CAPACITY_SCALE. 458 * 459 * [Overflow issue] 460 * 461 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 462 * with the highest load (=88761), always runnable on a single cfs_rq, 463 * and should not overflow as the number already hits PID_MAX_LIMIT. 464 * 465 * For all other cases (including 32-bit kernels), struct load_weight's 466 * weight will overflow first before we do, because: 467 * 468 * Max(load_avg) <= Max(load.weight) 469 * 470 * Then it is the load_weight's responsibility to consider overflow 471 * issues. 472 */ 473 struct sched_avg { 474 u64 last_update_time; 475 u64 load_sum; 476 u64 runnable_sum; 477 u32 util_sum; 478 u32 period_contrib; 479 unsigned long load_avg; 480 unsigned long runnable_avg; 481 unsigned long util_avg; 482 unsigned int util_est; 483 } ____cacheline_aligned; 484 485 /* 486 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 487 * updates. When a task is dequeued, its util_est should not be updated if its 488 * util_avg has not been updated in the meantime. 489 * This information is mapped into the MSB bit of util_est at dequeue time. 490 * Since max value of util_est for a task is 1024 (PELT util_avg for a task) 491 * it is safe to use MSB. 492 */ 493 #define UTIL_EST_WEIGHT_SHIFT 2 494 #define UTIL_AVG_UNCHANGED 0x80000000 495 496 struct sched_statistics { 497 #ifdef CONFIG_SCHEDSTATS 498 u64 wait_start; 499 u64 wait_max; 500 u64 wait_count; 501 u64 wait_sum; 502 u64 iowait_count; 503 u64 iowait_sum; 504 505 u64 sleep_start; 506 u64 sleep_max; 507 s64 sum_sleep_runtime; 508 509 u64 block_start; 510 u64 block_max; 511 s64 sum_block_runtime; 512 513 s64 exec_max; 514 u64 slice_max; 515 516 u64 nr_migrations_cold; 517 u64 nr_failed_migrations_affine; 518 u64 nr_failed_migrations_running; 519 u64 nr_failed_migrations_hot; 520 u64 nr_forced_migrations; 521 522 u64 nr_wakeups; 523 u64 nr_wakeups_sync; 524 u64 nr_wakeups_migrate; 525 u64 nr_wakeups_local; 526 u64 nr_wakeups_remote; 527 u64 nr_wakeups_affine; 528 u64 nr_wakeups_affine_attempts; 529 u64 nr_wakeups_passive; 530 u64 nr_wakeups_idle; 531 532 #ifdef CONFIG_SCHED_CORE 533 u64 core_forceidle_sum; 534 #endif 535 #endif /* CONFIG_SCHEDSTATS */ 536 } ____cacheline_aligned; 537 538 struct sched_entity { 539 /* For load-balancing: */ 540 struct load_weight load; 541 struct rb_node run_node; 542 u64 deadline; 543 u64 min_vruntime; 544 545 struct list_head group_node; 546 unsigned int on_rq; 547 548 u64 exec_start; 549 u64 sum_exec_runtime; 550 u64 prev_sum_exec_runtime; 551 u64 vruntime; 552 s64 vlag; 553 u64 slice; 554 555 u64 nr_migrations; 556 557 #ifdef CONFIG_FAIR_GROUP_SCHED 558 int depth; 559 struct sched_entity *parent; 560 /* rq on which this entity is (to be) queued: */ 561 struct cfs_rq *cfs_rq; 562 /* rq "owned" by this entity/group: */ 563 struct cfs_rq *my_q; 564 /* cached value of my_q->h_nr_running */ 565 unsigned long runnable_weight; 566 #endif 567 568 #ifdef CONFIG_SMP 569 /* 570 * Per entity load average tracking. 571 * 572 * Put into separate cache line so it does not 573 * collide with read-mostly values above. 574 */ 575 struct sched_avg avg; 576 #endif 577 }; 578 579 struct sched_rt_entity { 580 struct list_head run_list; 581 unsigned long timeout; 582 unsigned long watchdog_stamp; 583 unsigned int time_slice; 584 unsigned short on_rq; 585 unsigned short on_list; 586 587 struct sched_rt_entity *back; 588 #ifdef CONFIG_RT_GROUP_SCHED 589 struct sched_rt_entity *parent; 590 /* rq on which this entity is (to be) queued: */ 591 struct rt_rq *rt_rq; 592 /* rq "owned" by this entity/group: */ 593 struct rt_rq *my_q; 594 #endif 595 } __randomize_layout; 596 597 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *); 598 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *); 599 600 struct sched_dl_entity { 601 struct rb_node rb_node; 602 603 /* 604 * Original scheduling parameters. Copied here from sched_attr 605 * during sched_setattr(), they will remain the same until 606 * the next sched_setattr(). 607 */ 608 u64 dl_runtime; /* Maximum runtime for each instance */ 609 u64 dl_deadline; /* Relative deadline of each instance */ 610 u64 dl_period; /* Separation of two instances (period) */ 611 u64 dl_bw; /* dl_runtime / dl_period */ 612 u64 dl_density; /* dl_runtime / dl_deadline */ 613 614 /* 615 * Actual scheduling parameters. Initialized with the values above, 616 * they are continuously updated during task execution. Note that 617 * the remaining runtime could be < 0 in case we are in overrun. 618 */ 619 s64 runtime; /* Remaining runtime for this instance */ 620 u64 deadline; /* Absolute deadline for this instance */ 621 unsigned int flags; /* Specifying the scheduler behaviour */ 622 623 /* 624 * Some bool flags: 625 * 626 * @dl_throttled tells if we exhausted the runtime. If so, the 627 * task has to wait for a replenishment to be performed at the 628 * next firing of dl_timer. 629 * 630 * @dl_yielded tells if task gave up the CPU before consuming 631 * all its available runtime during the last job. 632 * 633 * @dl_non_contending tells if the task is inactive while still 634 * contributing to the active utilization. In other words, it 635 * indicates if the inactive timer has been armed and its handler 636 * has not been executed yet. This flag is useful to avoid race 637 * conditions between the inactive timer handler and the wakeup 638 * code. 639 * 640 * @dl_overrun tells if the task asked to be informed about runtime 641 * overruns. 642 */ 643 unsigned int dl_throttled : 1; 644 unsigned int dl_yielded : 1; 645 unsigned int dl_non_contending : 1; 646 unsigned int dl_overrun : 1; 647 unsigned int dl_server : 1; 648 649 /* 650 * Bandwidth enforcement timer. Each -deadline task has its 651 * own bandwidth to be enforced, thus we need one timer per task. 652 */ 653 struct hrtimer dl_timer; 654 655 /* 656 * Inactive timer, responsible for decreasing the active utilization 657 * at the "0-lag time". When a -deadline task blocks, it contributes 658 * to GRUB's active utilization until the "0-lag time", hence a 659 * timer is needed to decrease the active utilization at the correct 660 * time. 661 */ 662 struct hrtimer inactive_timer; 663 664 /* 665 * Bits for DL-server functionality. Also see the comment near 666 * dl_server_update(). 667 * 668 * @rq the runqueue this server is for 669 * 670 * @server_has_tasks() returns true if @server_pick return a 671 * runnable task. 672 */ 673 struct rq *rq; 674 dl_server_has_tasks_f server_has_tasks; 675 dl_server_pick_f server_pick; 676 677 #ifdef CONFIG_RT_MUTEXES 678 /* 679 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 680 * pi_se points to the donor, otherwise points to the dl_se it belongs 681 * to (the original one/itself). 682 */ 683 struct sched_dl_entity *pi_se; 684 #endif 685 }; 686 687 #ifdef CONFIG_UCLAMP_TASK 688 /* Number of utilization clamp buckets (shorter alias) */ 689 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 690 691 /* 692 * Utilization clamp for a scheduling entity 693 * @value: clamp value "assigned" to a se 694 * @bucket_id: bucket index corresponding to the "assigned" value 695 * @active: the se is currently refcounted in a rq's bucket 696 * @user_defined: the requested clamp value comes from user-space 697 * 698 * The bucket_id is the index of the clamp bucket matching the clamp value 699 * which is pre-computed and stored to avoid expensive integer divisions from 700 * the fast path. 701 * 702 * The active bit is set whenever a task has got an "effective" value assigned, 703 * which can be different from the clamp value "requested" from user-space. 704 * This allows to know a task is refcounted in the rq's bucket corresponding 705 * to the "effective" bucket_id. 706 * 707 * The user_defined bit is set whenever a task has got a task-specific clamp 708 * value requested from userspace, i.e. the system defaults apply to this task 709 * just as a restriction. This allows to relax default clamps when a less 710 * restrictive task-specific value has been requested, thus allowing to 711 * implement a "nice" semantic. For example, a task running with a 20% 712 * default boost can still drop its own boosting to 0%. 713 */ 714 struct uclamp_se { 715 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 716 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 717 unsigned int active : 1; 718 unsigned int user_defined : 1; 719 }; 720 #endif /* CONFIG_UCLAMP_TASK */ 721 722 union rcu_special { 723 struct { 724 u8 blocked; 725 u8 need_qs; 726 u8 exp_hint; /* Hint for performance. */ 727 u8 need_mb; /* Readers need smp_mb(). */ 728 } b; /* Bits. */ 729 u32 s; /* Set of bits. */ 730 }; 731 732 enum perf_event_task_context { 733 perf_invalid_context = -1, 734 perf_hw_context = 0, 735 perf_sw_context, 736 perf_nr_task_contexts, 737 }; 738 739 struct wake_q_node { 740 struct wake_q_node *next; 741 }; 742 743 struct kmap_ctrl { 744 #ifdef CONFIG_KMAP_LOCAL 745 int idx; 746 pte_t pteval[KM_MAX_IDX]; 747 #endif 748 }; 749 750 struct task_struct { 751 #ifdef CONFIG_THREAD_INFO_IN_TASK 752 /* 753 * For reasons of header soup (see current_thread_info()), this 754 * must be the first element of task_struct. 755 */ 756 struct thread_info thread_info; 757 #endif 758 unsigned int __state; 759 760 /* saved state for "spinlock sleepers" */ 761 unsigned int saved_state; 762 763 /* 764 * This begins the randomizable portion of task_struct. Only 765 * scheduling-critical items should be added above here. 766 */ 767 randomized_struct_fields_start 768 769 void *stack; 770 refcount_t usage; 771 /* Per task flags (PF_*), defined further below: */ 772 unsigned int flags; 773 unsigned int ptrace; 774 775 #ifdef CONFIG_MEM_ALLOC_PROFILING 776 struct alloc_tag *alloc_tag; 777 #endif 778 779 #ifdef CONFIG_SMP 780 int on_cpu; 781 struct __call_single_node wake_entry; 782 unsigned int wakee_flips; 783 unsigned long wakee_flip_decay_ts; 784 struct task_struct *last_wakee; 785 786 /* 787 * recent_used_cpu is initially set as the last CPU used by a task 788 * that wakes affine another task. Waker/wakee relationships can 789 * push tasks around a CPU where each wakeup moves to the next one. 790 * Tracking a recently used CPU allows a quick search for a recently 791 * used CPU that may be idle. 792 */ 793 int recent_used_cpu; 794 int wake_cpu; 795 #endif 796 int on_rq; 797 798 int prio; 799 int static_prio; 800 int normal_prio; 801 unsigned int rt_priority; 802 803 struct sched_entity se; 804 struct sched_rt_entity rt; 805 struct sched_dl_entity dl; 806 struct sched_dl_entity *dl_server; 807 const struct sched_class *sched_class; 808 809 #ifdef CONFIG_SCHED_CORE 810 struct rb_node core_node; 811 unsigned long core_cookie; 812 unsigned int core_occupation; 813 #endif 814 815 #ifdef CONFIG_CGROUP_SCHED 816 struct task_group *sched_task_group; 817 #endif 818 819 820 #ifdef CONFIG_UCLAMP_TASK 821 /* 822 * Clamp values requested for a scheduling entity. 823 * Must be updated with task_rq_lock() held. 824 */ 825 struct uclamp_se uclamp_req[UCLAMP_CNT]; 826 /* 827 * Effective clamp values used for a scheduling entity. 828 * Must be updated with task_rq_lock() held. 829 */ 830 struct uclamp_se uclamp[UCLAMP_CNT]; 831 #endif 832 833 struct sched_statistics stats; 834 835 #ifdef CONFIG_PREEMPT_NOTIFIERS 836 /* List of struct preempt_notifier: */ 837 struct hlist_head preempt_notifiers; 838 #endif 839 840 #ifdef CONFIG_BLK_DEV_IO_TRACE 841 unsigned int btrace_seq; 842 #endif 843 844 unsigned int policy; 845 unsigned long max_allowed_capacity; 846 int nr_cpus_allowed; 847 const cpumask_t *cpus_ptr; 848 cpumask_t *user_cpus_ptr; 849 cpumask_t cpus_mask; 850 void *migration_pending; 851 #ifdef CONFIG_SMP 852 unsigned short migration_disabled; 853 #endif 854 unsigned short migration_flags; 855 856 #ifdef CONFIG_PREEMPT_RCU 857 int rcu_read_lock_nesting; 858 union rcu_special rcu_read_unlock_special; 859 struct list_head rcu_node_entry; 860 struct rcu_node *rcu_blocked_node; 861 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 862 863 #ifdef CONFIG_TASKS_RCU 864 unsigned long rcu_tasks_nvcsw; 865 u8 rcu_tasks_holdout; 866 u8 rcu_tasks_idx; 867 int rcu_tasks_idle_cpu; 868 struct list_head rcu_tasks_holdout_list; 869 int rcu_tasks_exit_cpu; 870 struct list_head rcu_tasks_exit_list; 871 #endif /* #ifdef CONFIG_TASKS_RCU */ 872 873 #ifdef CONFIG_TASKS_TRACE_RCU 874 int trc_reader_nesting; 875 int trc_ipi_to_cpu; 876 union rcu_special trc_reader_special; 877 struct list_head trc_holdout_list; 878 struct list_head trc_blkd_node; 879 int trc_blkd_cpu; 880 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 881 882 struct sched_info sched_info; 883 884 struct list_head tasks; 885 #ifdef CONFIG_SMP 886 struct plist_node pushable_tasks; 887 struct rb_node pushable_dl_tasks; 888 #endif 889 890 struct mm_struct *mm; 891 struct mm_struct *active_mm; 892 struct address_space *faults_disabled_mapping; 893 894 int exit_state; 895 int exit_code; 896 int exit_signal; 897 /* The signal sent when the parent dies: */ 898 int pdeath_signal; 899 /* JOBCTL_*, siglock protected: */ 900 unsigned long jobctl; 901 902 /* Used for emulating ABI behavior of previous Linux versions: */ 903 unsigned int personality; 904 905 /* Scheduler bits, serialized by scheduler locks: */ 906 unsigned sched_reset_on_fork:1; 907 unsigned sched_contributes_to_load:1; 908 unsigned sched_migrated:1; 909 910 /* Force alignment to the next boundary: */ 911 unsigned :0; 912 913 /* Unserialized, strictly 'current' */ 914 915 /* 916 * This field must not be in the scheduler word above due to wakelist 917 * queueing no longer being serialized by p->on_cpu. However: 918 * 919 * p->XXX = X; ttwu() 920 * schedule() if (p->on_rq && ..) // false 921 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 922 * deactivate_task() ttwu_queue_wakelist()) 923 * p->on_rq = 0; p->sched_remote_wakeup = Y; 924 * 925 * guarantees all stores of 'current' are visible before 926 * ->sched_remote_wakeup gets used, so it can be in this word. 927 */ 928 unsigned sched_remote_wakeup:1; 929 #ifdef CONFIG_RT_MUTEXES 930 unsigned sched_rt_mutex:1; 931 #endif 932 933 /* Bit to tell TOMOYO we're in execve(): */ 934 unsigned in_execve:1; 935 unsigned in_iowait:1; 936 #ifndef TIF_RESTORE_SIGMASK 937 unsigned restore_sigmask:1; 938 #endif 939 #ifdef CONFIG_MEMCG 940 unsigned in_user_fault:1; 941 #endif 942 #ifdef CONFIG_LRU_GEN 943 /* whether the LRU algorithm may apply to this access */ 944 unsigned in_lru_fault:1; 945 #endif 946 #ifdef CONFIG_COMPAT_BRK 947 unsigned brk_randomized:1; 948 #endif 949 #ifdef CONFIG_CGROUPS 950 /* disallow userland-initiated cgroup migration */ 951 unsigned no_cgroup_migration:1; 952 /* task is frozen/stopped (used by the cgroup freezer) */ 953 unsigned frozen:1; 954 #endif 955 #ifdef CONFIG_BLK_CGROUP 956 unsigned use_memdelay:1; 957 #endif 958 #ifdef CONFIG_PSI 959 /* Stalled due to lack of memory */ 960 unsigned in_memstall:1; 961 #endif 962 #ifdef CONFIG_PAGE_OWNER 963 /* Used by page_owner=on to detect recursion in page tracking. */ 964 unsigned in_page_owner:1; 965 #endif 966 #ifdef CONFIG_EVENTFD 967 /* Recursion prevention for eventfd_signal() */ 968 unsigned in_eventfd:1; 969 #endif 970 #ifdef CONFIG_ARCH_HAS_CPU_PASID 971 unsigned pasid_activated:1; 972 #endif 973 #ifdef CONFIG_CPU_SUP_INTEL 974 unsigned reported_split_lock:1; 975 #endif 976 #ifdef CONFIG_TASK_DELAY_ACCT 977 /* delay due to memory thrashing */ 978 unsigned in_thrashing:1; 979 #endif 980 #ifdef CONFIG_PREEMPT_RT 981 struct netdev_xmit net_xmit; 982 #endif 983 unsigned long atomic_flags; /* Flags requiring atomic access. */ 984 985 struct restart_block restart_block; 986 987 pid_t pid; 988 pid_t tgid; 989 990 #ifdef CONFIG_STACKPROTECTOR 991 /* Canary value for the -fstack-protector GCC feature: */ 992 unsigned long stack_canary; 993 #endif 994 /* 995 * Pointers to the (original) parent process, youngest child, younger sibling, 996 * older sibling, respectively. (p->father can be replaced with 997 * p->real_parent->pid) 998 */ 999 1000 /* Real parent process: */ 1001 struct task_struct __rcu *real_parent; 1002 1003 /* Recipient of SIGCHLD, wait4() reports: */ 1004 struct task_struct __rcu *parent; 1005 1006 /* 1007 * Children/sibling form the list of natural children: 1008 */ 1009 struct list_head children; 1010 struct list_head sibling; 1011 struct task_struct *group_leader; 1012 1013 /* 1014 * 'ptraced' is the list of tasks this task is using ptrace() on. 1015 * 1016 * This includes both natural children and PTRACE_ATTACH targets. 1017 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 1018 */ 1019 struct list_head ptraced; 1020 struct list_head ptrace_entry; 1021 1022 /* PID/PID hash table linkage. */ 1023 struct pid *thread_pid; 1024 struct hlist_node pid_links[PIDTYPE_MAX]; 1025 struct list_head thread_node; 1026 1027 struct completion *vfork_done; 1028 1029 /* CLONE_CHILD_SETTID: */ 1030 int __user *set_child_tid; 1031 1032 /* CLONE_CHILD_CLEARTID: */ 1033 int __user *clear_child_tid; 1034 1035 /* PF_KTHREAD | PF_IO_WORKER */ 1036 void *worker_private; 1037 1038 u64 utime; 1039 u64 stime; 1040 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1041 u64 utimescaled; 1042 u64 stimescaled; 1043 #endif 1044 u64 gtime; 1045 struct prev_cputime prev_cputime; 1046 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1047 struct vtime vtime; 1048 #endif 1049 1050 #ifdef CONFIG_NO_HZ_FULL 1051 atomic_t tick_dep_mask; 1052 #endif 1053 /* Context switch counts: */ 1054 unsigned long nvcsw; 1055 unsigned long nivcsw; 1056 1057 /* Monotonic time in nsecs: */ 1058 u64 start_time; 1059 1060 /* Boot based time in nsecs: */ 1061 u64 start_boottime; 1062 1063 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1064 unsigned long min_flt; 1065 unsigned long maj_flt; 1066 1067 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1068 struct posix_cputimers posix_cputimers; 1069 1070 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1071 struct posix_cputimers_work posix_cputimers_work; 1072 #endif 1073 1074 /* Process credentials: */ 1075 1076 /* Tracer's credentials at attach: */ 1077 const struct cred __rcu *ptracer_cred; 1078 1079 /* Objective and real subjective task credentials (COW): */ 1080 const struct cred __rcu *real_cred; 1081 1082 /* Effective (overridable) subjective task credentials (COW): */ 1083 const struct cred __rcu *cred; 1084 1085 #ifdef CONFIG_KEYS 1086 /* Cached requested key. */ 1087 struct key *cached_requested_key; 1088 #endif 1089 1090 /* 1091 * executable name, excluding path. 1092 * 1093 * - normally initialized setup_new_exec() 1094 * - access it with [gs]et_task_comm() 1095 * - lock it with task_lock() 1096 */ 1097 char comm[TASK_COMM_LEN]; 1098 1099 struct nameidata *nameidata; 1100 1101 #ifdef CONFIG_SYSVIPC 1102 struct sysv_sem sysvsem; 1103 struct sysv_shm sysvshm; 1104 #endif 1105 #ifdef CONFIG_DETECT_HUNG_TASK 1106 unsigned long last_switch_count; 1107 unsigned long last_switch_time; 1108 #endif 1109 /* Filesystem information: */ 1110 struct fs_struct *fs; 1111 1112 /* Open file information: */ 1113 struct files_struct *files; 1114 1115 #ifdef CONFIG_IO_URING 1116 struct io_uring_task *io_uring; 1117 #endif 1118 1119 /* Namespaces: */ 1120 struct nsproxy *nsproxy; 1121 1122 /* Signal handlers: */ 1123 struct signal_struct *signal; 1124 struct sighand_struct __rcu *sighand; 1125 sigset_t blocked; 1126 sigset_t real_blocked; 1127 /* Restored if set_restore_sigmask() was used: */ 1128 sigset_t saved_sigmask; 1129 struct sigpending pending; 1130 unsigned long sas_ss_sp; 1131 size_t sas_ss_size; 1132 unsigned int sas_ss_flags; 1133 1134 struct callback_head *task_works; 1135 1136 #ifdef CONFIG_AUDIT 1137 #ifdef CONFIG_AUDITSYSCALL 1138 struct audit_context *audit_context; 1139 #endif 1140 kuid_t loginuid; 1141 unsigned int sessionid; 1142 #endif 1143 struct seccomp seccomp; 1144 struct syscall_user_dispatch syscall_dispatch; 1145 1146 /* Thread group tracking: */ 1147 u64 parent_exec_id; 1148 u64 self_exec_id; 1149 1150 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1151 spinlock_t alloc_lock; 1152 1153 /* Protection of the PI data structures: */ 1154 raw_spinlock_t pi_lock; 1155 1156 struct wake_q_node wake_q; 1157 1158 #ifdef CONFIG_RT_MUTEXES 1159 /* PI waiters blocked on a rt_mutex held by this task: */ 1160 struct rb_root_cached pi_waiters; 1161 /* Updated under owner's pi_lock and rq lock */ 1162 struct task_struct *pi_top_task; 1163 /* Deadlock detection and priority inheritance handling: */ 1164 struct rt_mutex_waiter *pi_blocked_on; 1165 #endif 1166 1167 #ifdef CONFIG_DEBUG_MUTEXES 1168 /* Mutex deadlock detection: */ 1169 struct mutex_waiter *blocked_on; 1170 #endif 1171 1172 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1173 int non_block_count; 1174 #endif 1175 1176 #ifdef CONFIG_TRACE_IRQFLAGS 1177 struct irqtrace_events irqtrace; 1178 unsigned int hardirq_threaded; 1179 u64 hardirq_chain_key; 1180 int softirqs_enabled; 1181 int softirq_context; 1182 int irq_config; 1183 #endif 1184 #ifdef CONFIG_PREEMPT_RT 1185 int softirq_disable_cnt; 1186 #endif 1187 1188 #ifdef CONFIG_LOCKDEP 1189 # define MAX_LOCK_DEPTH 48UL 1190 u64 curr_chain_key; 1191 int lockdep_depth; 1192 unsigned int lockdep_recursion; 1193 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1194 #endif 1195 1196 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1197 unsigned int in_ubsan; 1198 #endif 1199 1200 /* Journalling filesystem info: */ 1201 void *journal_info; 1202 1203 /* Stacked block device info: */ 1204 struct bio_list *bio_list; 1205 1206 /* Stack plugging: */ 1207 struct blk_plug *plug; 1208 1209 /* VM state: */ 1210 struct reclaim_state *reclaim_state; 1211 1212 struct io_context *io_context; 1213 1214 #ifdef CONFIG_COMPACTION 1215 struct capture_control *capture_control; 1216 #endif 1217 /* Ptrace state: */ 1218 unsigned long ptrace_message; 1219 kernel_siginfo_t *last_siginfo; 1220 1221 struct task_io_accounting ioac; 1222 #ifdef CONFIG_PSI 1223 /* Pressure stall state */ 1224 unsigned int psi_flags; 1225 #endif 1226 #ifdef CONFIG_TASK_XACCT 1227 /* Accumulated RSS usage: */ 1228 u64 acct_rss_mem1; 1229 /* Accumulated virtual memory usage: */ 1230 u64 acct_vm_mem1; 1231 /* stime + utime since last update: */ 1232 u64 acct_timexpd; 1233 #endif 1234 #ifdef CONFIG_CPUSETS 1235 /* Protected by ->alloc_lock: */ 1236 nodemask_t mems_allowed; 1237 /* Sequence number to catch updates: */ 1238 seqcount_spinlock_t mems_allowed_seq; 1239 int cpuset_mem_spread_rotor; 1240 int cpuset_slab_spread_rotor; 1241 #endif 1242 #ifdef CONFIG_CGROUPS 1243 /* Control Group info protected by css_set_lock: */ 1244 struct css_set __rcu *cgroups; 1245 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1246 struct list_head cg_list; 1247 #endif 1248 #ifdef CONFIG_X86_CPU_RESCTRL 1249 u32 closid; 1250 u32 rmid; 1251 #endif 1252 #ifdef CONFIG_FUTEX 1253 struct robust_list_head __user *robust_list; 1254 #ifdef CONFIG_COMPAT 1255 struct compat_robust_list_head __user *compat_robust_list; 1256 #endif 1257 struct list_head pi_state_list; 1258 struct futex_pi_state *pi_state_cache; 1259 struct mutex futex_exit_mutex; 1260 unsigned int futex_state; 1261 #endif 1262 #ifdef CONFIG_PERF_EVENTS 1263 struct perf_event_context *perf_event_ctxp; 1264 struct mutex perf_event_mutex; 1265 struct list_head perf_event_list; 1266 #endif 1267 #ifdef CONFIG_DEBUG_PREEMPT 1268 unsigned long preempt_disable_ip; 1269 #endif 1270 #ifdef CONFIG_NUMA 1271 /* Protected by alloc_lock: */ 1272 struct mempolicy *mempolicy; 1273 short il_prev; 1274 u8 il_weight; 1275 short pref_node_fork; 1276 #endif 1277 #ifdef CONFIG_NUMA_BALANCING 1278 int numa_scan_seq; 1279 unsigned int numa_scan_period; 1280 unsigned int numa_scan_period_max; 1281 int numa_preferred_nid; 1282 unsigned long numa_migrate_retry; 1283 /* Migration stamp: */ 1284 u64 node_stamp; 1285 u64 last_task_numa_placement; 1286 u64 last_sum_exec_runtime; 1287 struct callback_head numa_work; 1288 1289 /* 1290 * This pointer is only modified for current in syscall and 1291 * pagefault context (and for tasks being destroyed), so it can be read 1292 * from any of the following contexts: 1293 * - RCU read-side critical section 1294 * - current->numa_group from everywhere 1295 * - task's runqueue locked, task not running 1296 */ 1297 struct numa_group __rcu *numa_group; 1298 1299 /* 1300 * numa_faults is an array split into four regions: 1301 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1302 * in this precise order. 1303 * 1304 * faults_memory: Exponential decaying average of faults on a per-node 1305 * basis. Scheduling placement decisions are made based on these 1306 * counts. The values remain static for the duration of a PTE scan. 1307 * faults_cpu: Track the nodes the process was running on when a NUMA 1308 * hinting fault was incurred. 1309 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1310 * during the current scan window. When the scan completes, the counts 1311 * in faults_memory and faults_cpu decay and these values are copied. 1312 */ 1313 unsigned long *numa_faults; 1314 unsigned long total_numa_faults; 1315 1316 /* 1317 * numa_faults_locality tracks if faults recorded during the last 1318 * scan window were remote/local or failed to migrate. The task scan 1319 * period is adapted based on the locality of the faults with different 1320 * weights depending on whether they were shared or private faults 1321 */ 1322 unsigned long numa_faults_locality[3]; 1323 1324 unsigned long numa_pages_migrated; 1325 #endif /* CONFIG_NUMA_BALANCING */ 1326 1327 #ifdef CONFIG_RSEQ 1328 struct rseq __user *rseq; 1329 u32 rseq_len; 1330 u32 rseq_sig; 1331 /* 1332 * RmW on rseq_event_mask must be performed atomically 1333 * with respect to preemption. 1334 */ 1335 unsigned long rseq_event_mask; 1336 #endif 1337 1338 #ifdef CONFIG_SCHED_MM_CID 1339 int mm_cid; /* Current cid in mm */ 1340 int last_mm_cid; /* Most recent cid in mm */ 1341 int migrate_from_cpu; 1342 int mm_cid_active; /* Whether cid bitmap is active */ 1343 struct callback_head cid_work; 1344 #endif 1345 1346 struct tlbflush_unmap_batch tlb_ubc; 1347 1348 /* Cache last used pipe for splice(): */ 1349 struct pipe_inode_info *splice_pipe; 1350 1351 struct page_frag task_frag; 1352 1353 #ifdef CONFIG_TASK_DELAY_ACCT 1354 struct task_delay_info *delays; 1355 #endif 1356 1357 #ifdef CONFIG_FAULT_INJECTION 1358 int make_it_fail; 1359 unsigned int fail_nth; 1360 #endif 1361 /* 1362 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1363 * balance_dirty_pages() for a dirty throttling pause: 1364 */ 1365 int nr_dirtied; 1366 int nr_dirtied_pause; 1367 /* Start of a write-and-pause period: */ 1368 unsigned long dirty_paused_when; 1369 1370 #ifdef CONFIG_LATENCYTOP 1371 int latency_record_count; 1372 struct latency_record latency_record[LT_SAVECOUNT]; 1373 #endif 1374 /* 1375 * Time slack values; these are used to round up poll() and 1376 * select() etc timeout values. These are in nanoseconds. 1377 */ 1378 u64 timer_slack_ns; 1379 u64 default_timer_slack_ns; 1380 1381 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1382 unsigned int kasan_depth; 1383 #endif 1384 1385 #ifdef CONFIG_KCSAN 1386 struct kcsan_ctx kcsan_ctx; 1387 #ifdef CONFIG_TRACE_IRQFLAGS 1388 struct irqtrace_events kcsan_save_irqtrace; 1389 #endif 1390 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1391 int kcsan_stack_depth; 1392 #endif 1393 #endif 1394 1395 #ifdef CONFIG_KMSAN 1396 struct kmsan_ctx kmsan_ctx; 1397 #endif 1398 1399 #if IS_ENABLED(CONFIG_KUNIT) 1400 struct kunit *kunit_test; 1401 #endif 1402 1403 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1404 /* Index of current stored address in ret_stack: */ 1405 int curr_ret_stack; 1406 int curr_ret_depth; 1407 1408 /* Stack of return addresses for return function tracing: */ 1409 struct ftrace_ret_stack *ret_stack; 1410 1411 /* Timestamp for last schedule: */ 1412 unsigned long long ftrace_timestamp; 1413 1414 /* 1415 * Number of functions that haven't been traced 1416 * because of depth overrun: 1417 */ 1418 atomic_t trace_overrun; 1419 1420 /* Pause tracing: */ 1421 atomic_t tracing_graph_pause; 1422 #endif 1423 1424 #ifdef CONFIG_TRACING 1425 /* Bitmask and counter of trace recursion: */ 1426 unsigned long trace_recursion; 1427 #endif /* CONFIG_TRACING */ 1428 1429 #ifdef CONFIG_KCOV 1430 /* See kernel/kcov.c for more details. */ 1431 1432 /* Coverage collection mode enabled for this task (0 if disabled): */ 1433 unsigned int kcov_mode; 1434 1435 /* Size of the kcov_area: */ 1436 unsigned int kcov_size; 1437 1438 /* Buffer for coverage collection: */ 1439 void *kcov_area; 1440 1441 /* KCOV descriptor wired with this task or NULL: */ 1442 struct kcov *kcov; 1443 1444 /* KCOV common handle for remote coverage collection: */ 1445 u64 kcov_handle; 1446 1447 /* KCOV sequence number: */ 1448 int kcov_sequence; 1449 1450 /* Collect coverage from softirq context: */ 1451 unsigned int kcov_softirq; 1452 #endif 1453 1454 #ifdef CONFIG_MEMCG 1455 struct mem_cgroup *memcg_in_oom; 1456 1457 /* Number of pages to reclaim on returning to userland: */ 1458 unsigned int memcg_nr_pages_over_high; 1459 1460 /* Used by memcontrol for targeted memcg charge: */ 1461 struct mem_cgroup *active_memcg; 1462 #endif 1463 1464 #ifdef CONFIG_MEMCG_KMEM 1465 struct obj_cgroup *objcg; 1466 #endif 1467 1468 #ifdef CONFIG_BLK_CGROUP 1469 struct gendisk *throttle_disk; 1470 #endif 1471 1472 #ifdef CONFIG_UPROBES 1473 struct uprobe_task *utask; 1474 #endif 1475 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1476 unsigned int sequential_io; 1477 unsigned int sequential_io_avg; 1478 #endif 1479 struct kmap_ctrl kmap_ctrl; 1480 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1481 unsigned long task_state_change; 1482 # ifdef CONFIG_PREEMPT_RT 1483 unsigned long saved_state_change; 1484 # endif 1485 #endif 1486 struct rcu_head rcu; 1487 refcount_t rcu_users; 1488 int pagefault_disabled; 1489 #ifdef CONFIG_MMU 1490 struct task_struct *oom_reaper_list; 1491 struct timer_list oom_reaper_timer; 1492 #endif 1493 #ifdef CONFIG_VMAP_STACK 1494 struct vm_struct *stack_vm_area; 1495 #endif 1496 #ifdef CONFIG_THREAD_INFO_IN_TASK 1497 /* A live task holds one reference: */ 1498 refcount_t stack_refcount; 1499 #endif 1500 #ifdef CONFIG_LIVEPATCH 1501 int patch_state; 1502 #endif 1503 #ifdef CONFIG_SECURITY 1504 /* Used by LSM modules for access restriction: */ 1505 void *security; 1506 #endif 1507 #ifdef CONFIG_BPF_SYSCALL 1508 /* Used by BPF task local storage */ 1509 struct bpf_local_storage __rcu *bpf_storage; 1510 /* Used for BPF run context */ 1511 struct bpf_run_ctx *bpf_ctx; 1512 #endif 1513 /* Used by BPF for per-TASK xdp storage */ 1514 struct bpf_net_context *bpf_net_context; 1515 1516 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1517 unsigned long lowest_stack; 1518 unsigned long prev_lowest_stack; 1519 #endif 1520 1521 #ifdef CONFIG_X86_MCE 1522 void __user *mce_vaddr; 1523 __u64 mce_kflags; 1524 u64 mce_addr; 1525 __u64 mce_ripv : 1, 1526 mce_whole_page : 1, 1527 __mce_reserved : 62; 1528 struct callback_head mce_kill_me; 1529 int mce_count; 1530 #endif 1531 1532 #ifdef CONFIG_KRETPROBES 1533 struct llist_head kretprobe_instances; 1534 #endif 1535 #ifdef CONFIG_RETHOOK 1536 struct llist_head rethooks; 1537 #endif 1538 1539 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1540 /* 1541 * If L1D flush is supported on mm context switch 1542 * then we use this callback head to queue kill work 1543 * to kill tasks that are not running on SMT disabled 1544 * cores 1545 */ 1546 struct callback_head l1d_flush_kill; 1547 #endif 1548 1549 #ifdef CONFIG_RV 1550 /* 1551 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS. 1552 * If we find justification for more monitors, we can think 1553 * about adding more or developing a dynamic method. So far, 1554 * none of these are justified. 1555 */ 1556 union rv_task_monitor rv[RV_PER_TASK_MONITORS]; 1557 #endif 1558 1559 #ifdef CONFIG_USER_EVENTS 1560 struct user_event_mm *user_event_mm; 1561 #endif 1562 1563 /* 1564 * New fields for task_struct should be added above here, so that 1565 * they are included in the randomized portion of task_struct. 1566 */ 1567 randomized_struct_fields_end 1568 1569 /* CPU-specific state of this task: */ 1570 struct thread_struct thread; 1571 1572 /* 1573 * WARNING: on x86, 'thread_struct' contains a variable-sized 1574 * structure. It *MUST* be at the end of 'task_struct'. 1575 * 1576 * Do not put anything below here! 1577 */ 1578 }; 1579 1580 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1581 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1582 1583 static inline unsigned int __task_state_index(unsigned int tsk_state, 1584 unsigned int tsk_exit_state) 1585 { 1586 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1587 1588 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1589 1590 if ((tsk_state & TASK_IDLE) == TASK_IDLE) 1591 state = TASK_REPORT_IDLE; 1592 1593 /* 1594 * We're lying here, but rather than expose a completely new task state 1595 * to userspace, we can make this appear as if the task has gone through 1596 * a regular rt_mutex_lock() call. 1597 */ 1598 if (tsk_state & TASK_RTLOCK_WAIT) 1599 state = TASK_UNINTERRUPTIBLE; 1600 1601 return fls(state); 1602 } 1603 1604 static inline unsigned int task_state_index(struct task_struct *tsk) 1605 { 1606 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1607 } 1608 1609 static inline char task_index_to_char(unsigned int state) 1610 { 1611 static const char state_char[] = "RSDTtXZPI"; 1612 1613 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1614 1615 return state_char[state]; 1616 } 1617 1618 static inline char task_state_to_char(struct task_struct *tsk) 1619 { 1620 return task_index_to_char(task_state_index(tsk)); 1621 } 1622 1623 extern struct pid *cad_pid; 1624 1625 /* 1626 * Per process flags 1627 */ 1628 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1629 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1630 #define PF_EXITING 0x00000004 /* Getting shut down */ 1631 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1632 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1633 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1634 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1635 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1636 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1637 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1638 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1639 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */ 1640 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1641 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1642 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */ 1643 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1644 #define PF__HOLE__00010000 0x00010000 1645 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1646 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */ 1647 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */ 1648 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1649 * I am cleaning dirty pages from some other bdi. */ 1650 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1651 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1652 #define PF_MEMALLOC_NORECLAIM 0x00800000 /* All allocation requests will clear __GFP_DIRECT_RECLAIM */ 1653 #define PF_MEMALLOC_NOWARN 0x01000000 /* All allocation requests will inherit __GFP_NOWARN */ 1654 #define PF__HOLE__02000000 0x02000000 1655 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1656 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1657 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning. 1658 * See memalloc_pin_save() */ 1659 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */ 1660 #define PF__HOLE__40000000 0x40000000 1661 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1662 1663 /* 1664 * Only the _current_ task can read/write to tsk->flags, but other 1665 * tasks can access tsk->flags in readonly mode for example 1666 * with tsk_used_math (like during threaded core dumping). 1667 * There is however an exception to this rule during ptrace 1668 * or during fork: the ptracer task is allowed to write to the 1669 * child->flags of its traced child (same goes for fork, the parent 1670 * can write to the child->flags), because we're guaranteed the 1671 * child is not running and in turn not changing child->flags 1672 * at the same time the parent does it. 1673 */ 1674 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1675 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1676 #define clear_used_math() clear_stopped_child_used_math(current) 1677 #define set_used_math() set_stopped_child_used_math(current) 1678 1679 #define conditional_stopped_child_used_math(condition, child) \ 1680 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1681 1682 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1683 1684 #define copy_to_stopped_child_used_math(child) \ 1685 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1686 1687 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1688 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1689 #define used_math() tsk_used_math(current) 1690 1691 static __always_inline bool is_percpu_thread(void) 1692 { 1693 #ifdef CONFIG_SMP 1694 return (current->flags & PF_NO_SETAFFINITY) && 1695 (current->nr_cpus_allowed == 1); 1696 #else 1697 return true; 1698 #endif 1699 } 1700 1701 /* Per-process atomic flags. */ 1702 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1703 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1704 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1705 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1706 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1707 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1708 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1709 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1710 1711 #define TASK_PFA_TEST(name, func) \ 1712 static inline bool task_##func(struct task_struct *p) \ 1713 { return test_bit(PFA_##name, &p->atomic_flags); } 1714 1715 #define TASK_PFA_SET(name, func) \ 1716 static inline void task_set_##func(struct task_struct *p) \ 1717 { set_bit(PFA_##name, &p->atomic_flags); } 1718 1719 #define TASK_PFA_CLEAR(name, func) \ 1720 static inline void task_clear_##func(struct task_struct *p) \ 1721 { clear_bit(PFA_##name, &p->atomic_flags); } 1722 1723 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1724 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1725 1726 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1727 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1728 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1729 1730 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1731 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1732 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1733 1734 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1735 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1736 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1737 1738 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1739 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1740 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1741 1742 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1743 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1744 1745 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1746 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1747 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1748 1749 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1750 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1751 1752 static inline void 1753 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1754 { 1755 current->flags &= ~flags; 1756 current->flags |= orig_flags & flags; 1757 } 1758 1759 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1760 extern int task_can_attach(struct task_struct *p); 1761 extern int dl_bw_alloc(int cpu, u64 dl_bw); 1762 extern void dl_bw_free(int cpu, u64 dl_bw); 1763 #ifdef CONFIG_SMP 1764 1765 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */ 1766 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1767 1768 /** 1769 * set_cpus_allowed_ptr - set CPU affinity mask of a task 1770 * @p: the task 1771 * @new_mask: CPU affinity mask 1772 * 1773 * Return: zero if successful, or a negative error code 1774 */ 1775 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1776 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1777 extern void release_user_cpus_ptr(struct task_struct *p); 1778 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1779 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1780 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1781 #else 1782 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1783 { 1784 } 1785 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1786 { 1787 if (!cpumask_test_cpu(0, new_mask)) 1788 return -EINVAL; 1789 return 0; 1790 } 1791 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node) 1792 { 1793 if (src->user_cpus_ptr) 1794 return -EINVAL; 1795 return 0; 1796 } 1797 static inline void release_user_cpus_ptr(struct task_struct *p) 1798 { 1799 WARN_ON(p->user_cpus_ptr); 1800 } 1801 1802 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) 1803 { 1804 return 0; 1805 } 1806 #endif 1807 1808 extern int yield_to(struct task_struct *p, bool preempt); 1809 extern void set_user_nice(struct task_struct *p, long nice); 1810 extern int task_prio(const struct task_struct *p); 1811 1812 /** 1813 * task_nice - return the nice value of a given task. 1814 * @p: the task in question. 1815 * 1816 * Return: The nice value [ -20 ... 0 ... 19 ]. 1817 */ 1818 static inline int task_nice(const struct task_struct *p) 1819 { 1820 return PRIO_TO_NICE((p)->static_prio); 1821 } 1822 1823 extern int can_nice(const struct task_struct *p, const int nice); 1824 extern int task_curr(const struct task_struct *p); 1825 extern int idle_cpu(int cpu); 1826 extern int available_idle_cpu(int cpu); 1827 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1828 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1829 extern void sched_set_fifo(struct task_struct *p); 1830 extern void sched_set_fifo_low(struct task_struct *p); 1831 extern void sched_set_normal(struct task_struct *p, int nice); 1832 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1833 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1834 extern struct task_struct *idle_task(int cpu); 1835 1836 /** 1837 * is_idle_task - is the specified task an idle task? 1838 * @p: the task in question. 1839 * 1840 * Return: 1 if @p is an idle task. 0 otherwise. 1841 */ 1842 static __always_inline bool is_idle_task(const struct task_struct *p) 1843 { 1844 return !!(p->flags & PF_IDLE); 1845 } 1846 1847 extern struct task_struct *curr_task(int cpu); 1848 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1849 1850 void yield(void); 1851 1852 union thread_union { 1853 struct task_struct task; 1854 #ifndef CONFIG_THREAD_INFO_IN_TASK 1855 struct thread_info thread_info; 1856 #endif 1857 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1858 }; 1859 1860 #ifndef CONFIG_THREAD_INFO_IN_TASK 1861 extern struct thread_info init_thread_info; 1862 #endif 1863 1864 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1865 1866 #ifdef CONFIG_THREAD_INFO_IN_TASK 1867 # define task_thread_info(task) (&(task)->thread_info) 1868 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1869 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1870 #endif 1871 1872 /* 1873 * find a task by one of its numerical ids 1874 * 1875 * find_task_by_pid_ns(): 1876 * finds a task by its pid in the specified namespace 1877 * find_task_by_vpid(): 1878 * finds a task by its virtual pid 1879 * 1880 * see also find_vpid() etc in include/linux/pid.h 1881 */ 1882 1883 extern struct task_struct *find_task_by_vpid(pid_t nr); 1884 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1885 1886 /* 1887 * find a task by its virtual pid and get the task struct 1888 */ 1889 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1890 1891 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1892 extern int wake_up_process(struct task_struct *tsk); 1893 extern void wake_up_new_task(struct task_struct *tsk); 1894 1895 #ifdef CONFIG_SMP 1896 extern void kick_process(struct task_struct *tsk); 1897 #else 1898 static inline void kick_process(struct task_struct *tsk) { } 1899 #endif 1900 1901 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1902 1903 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1904 { 1905 __set_task_comm(tsk, from, false); 1906 } 1907 1908 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1909 #define get_task_comm(buf, tsk) ({ \ 1910 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1911 __get_task_comm(buf, sizeof(buf), tsk); \ 1912 }) 1913 1914 #ifdef CONFIG_SMP 1915 static __always_inline void scheduler_ipi(void) 1916 { 1917 /* 1918 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1919 * TIF_NEED_RESCHED remotely (for the first time) will also send 1920 * this IPI. 1921 */ 1922 preempt_fold_need_resched(); 1923 } 1924 #else 1925 static inline void scheduler_ipi(void) { } 1926 #endif 1927 1928 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 1929 1930 /* 1931 * Set thread flags in other task's structures. 1932 * See asm/thread_info.h for TIF_xxxx flags available: 1933 */ 1934 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1935 { 1936 set_ti_thread_flag(task_thread_info(tsk), flag); 1937 } 1938 1939 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1940 { 1941 clear_ti_thread_flag(task_thread_info(tsk), flag); 1942 } 1943 1944 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1945 bool value) 1946 { 1947 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1948 } 1949 1950 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1951 { 1952 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1953 } 1954 1955 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1956 { 1957 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1958 } 1959 1960 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1961 { 1962 return test_ti_thread_flag(task_thread_info(tsk), flag); 1963 } 1964 1965 static inline void set_tsk_need_resched(struct task_struct *tsk) 1966 { 1967 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1968 } 1969 1970 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1971 { 1972 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1973 } 1974 1975 static inline int test_tsk_need_resched(struct task_struct *tsk) 1976 { 1977 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1978 } 1979 1980 /* 1981 * cond_resched() and cond_resched_lock(): latency reduction via 1982 * explicit rescheduling in places that are safe. The return 1983 * value indicates whether a reschedule was done in fact. 1984 * cond_resched_lock() will drop the spinlock before scheduling, 1985 */ 1986 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 1987 extern int __cond_resched(void); 1988 1989 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 1990 1991 void sched_dynamic_klp_enable(void); 1992 void sched_dynamic_klp_disable(void); 1993 1994 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 1995 1996 static __always_inline int _cond_resched(void) 1997 { 1998 return static_call_mod(cond_resched)(); 1999 } 2000 2001 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2002 2003 extern int dynamic_cond_resched(void); 2004 2005 static __always_inline int _cond_resched(void) 2006 { 2007 return dynamic_cond_resched(); 2008 } 2009 2010 #else /* !CONFIG_PREEMPTION */ 2011 2012 static inline int _cond_resched(void) 2013 { 2014 klp_sched_try_switch(); 2015 return __cond_resched(); 2016 } 2017 2018 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */ 2019 2020 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */ 2021 2022 static inline int _cond_resched(void) 2023 { 2024 klp_sched_try_switch(); 2025 return 0; 2026 } 2027 2028 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */ 2029 2030 #define cond_resched() ({ \ 2031 __might_resched(__FILE__, __LINE__, 0); \ 2032 _cond_resched(); \ 2033 }) 2034 2035 extern int __cond_resched_lock(spinlock_t *lock); 2036 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2037 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2038 2039 #define MIGHT_RESCHED_RCU_SHIFT 8 2040 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2041 2042 #ifndef CONFIG_PREEMPT_RT 2043 /* 2044 * Non RT kernels have an elevated preempt count due to the held lock, 2045 * but are not allowed to be inside a RCU read side critical section 2046 */ 2047 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2048 #else 2049 /* 2050 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2051 * cond_resched*lock() has to take that into account because it checks for 2052 * preempt_count() and rcu_preempt_depth(). 2053 */ 2054 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2055 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2056 #endif 2057 2058 #define cond_resched_lock(lock) ({ \ 2059 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2060 __cond_resched_lock(lock); \ 2061 }) 2062 2063 #define cond_resched_rwlock_read(lock) ({ \ 2064 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2065 __cond_resched_rwlock_read(lock); \ 2066 }) 2067 2068 #define cond_resched_rwlock_write(lock) ({ \ 2069 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2070 __cond_resched_rwlock_write(lock); \ 2071 }) 2072 2073 #ifdef CONFIG_PREEMPT_DYNAMIC 2074 2075 extern bool preempt_model_none(void); 2076 extern bool preempt_model_voluntary(void); 2077 extern bool preempt_model_full(void); 2078 2079 #else 2080 2081 static inline bool preempt_model_none(void) 2082 { 2083 return IS_ENABLED(CONFIG_PREEMPT_NONE); 2084 } 2085 static inline bool preempt_model_voluntary(void) 2086 { 2087 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY); 2088 } 2089 static inline bool preempt_model_full(void) 2090 { 2091 return IS_ENABLED(CONFIG_PREEMPT); 2092 } 2093 2094 #endif 2095 2096 static inline bool preempt_model_rt(void) 2097 { 2098 return IS_ENABLED(CONFIG_PREEMPT_RT); 2099 } 2100 2101 /* 2102 * Does the preemption model allow non-cooperative preemption? 2103 * 2104 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with 2105 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the 2106 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the 2107 * PREEMPT_NONE model. 2108 */ 2109 static inline bool preempt_model_preemptible(void) 2110 { 2111 return preempt_model_full() || preempt_model_rt(); 2112 } 2113 2114 static __always_inline bool need_resched(void) 2115 { 2116 return unlikely(tif_need_resched()); 2117 } 2118 2119 /* 2120 * Wrappers for p->thread_info->cpu access. No-op on UP. 2121 */ 2122 #ifdef CONFIG_SMP 2123 2124 static inline unsigned int task_cpu(const struct task_struct *p) 2125 { 2126 return READ_ONCE(task_thread_info(p)->cpu); 2127 } 2128 2129 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2130 2131 #else 2132 2133 static inline unsigned int task_cpu(const struct task_struct *p) 2134 { 2135 return 0; 2136 } 2137 2138 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2139 { 2140 } 2141 2142 #endif /* CONFIG_SMP */ 2143 2144 extern bool sched_task_on_rq(struct task_struct *p); 2145 extern unsigned long get_wchan(struct task_struct *p); 2146 extern struct task_struct *cpu_curr_snapshot(int cpu); 2147 2148 #include <linux/spinlock.h> 2149 2150 /* 2151 * In order to reduce various lock holder preemption latencies provide an 2152 * interface to see if a vCPU is currently running or not. 2153 * 2154 * This allows us to terminate optimistic spin loops and block, analogous to 2155 * the native optimistic spin heuristic of testing if the lock owner task is 2156 * running or not. 2157 */ 2158 #ifndef vcpu_is_preempted 2159 static inline bool vcpu_is_preempted(int cpu) 2160 { 2161 return false; 2162 } 2163 #endif 2164 2165 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2166 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2167 2168 #ifndef TASK_SIZE_OF 2169 #define TASK_SIZE_OF(tsk) TASK_SIZE 2170 #endif 2171 2172 #ifdef CONFIG_SMP 2173 static inline bool owner_on_cpu(struct task_struct *owner) 2174 { 2175 /* 2176 * As lock holder preemption issue, we both skip spinning if 2177 * task is not on cpu or its cpu is preempted 2178 */ 2179 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2180 } 2181 2182 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2183 unsigned long sched_cpu_util(int cpu); 2184 #endif /* CONFIG_SMP */ 2185 2186 #ifdef CONFIG_SCHED_CORE 2187 extern void sched_core_free(struct task_struct *tsk); 2188 extern void sched_core_fork(struct task_struct *p); 2189 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2190 unsigned long uaddr); 2191 extern int sched_core_idle_cpu(int cpu); 2192 #else 2193 static inline void sched_core_free(struct task_struct *tsk) { } 2194 static inline void sched_core_fork(struct task_struct *p) { } 2195 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); } 2196 #endif 2197 2198 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2199 2200 #ifdef CONFIG_MEM_ALLOC_PROFILING 2201 static inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag) 2202 { 2203 swap(current->alloc_tag, tag); 2204 return tag; 2205 } 2206 2207 static inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old) 2208 { 2209 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG 2210 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n"); 2211 #endif 2212 current->alloc_tag = old; 2213 } 2214 #else 2215 #define alloc_tag_save(_tag) NULL 2216 #define alloc_tag_restore(_tag, _old) do {} while (0) 2217 #endif 2218 2219 #endif 2220