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 14 #include <linux/pid.h> 15 #include <linux/sem.h> 16 #include <linux/shm.h> 17 #include <linux/mutex.h> 18 #include <linux/plist.h> 19 #include <linux/hrtimer.h> 20 #include <linux/irqflags.h> 21 #include <linux/seccomp.h> 22 #include <linux/nodemask.h> 23 #include <linux/rcupdate.h> 24 #include <linux/refcount.h> 25 #include <linux/resource.h> 26 #include <linux/latencytop.h> 27 #include <linux/sched/prio.h> 28 #include <linux/sched/types.h> 29 #include <linux/signal_types.h> 30 #include <linux/syscall_user_dispatch.h> 31 #include <linux/mm_types_task.h> 32 #include <linux/task_io_accounting.h> 33 #include <linux/posix-timers.h> 34 #include <linux/rseq.h> 35 #include <linux/seqlock.h> 36 #include <linux/kcsan.h> 37 #include <asm/kmap_size.h> 38 39 /* task_struct member predeclarations (sorted alphabetically): */ 40 struct audit_context; 41 struct backing_dev_info; 42 struct bio_list; 43 struct blk_plug; 44 struct bpf_local_storage; 45 struct bpf_run_ctx; 46 struct capture_control; 47 struct cfs_rq; 48 struct fs_struct; 49 struct futex_pi_state; 50 struct io_context; 51 struct io_uring_task; 52 struct mempolicy; 53 struct nameidata; 54 struct nsproxy; 55 struct perf_event_context; 56 struct pid_namespace; 57 struct pipe_inode_info; 58 struct rcu_node; 59 struct reclaim_state; 60 struct robust_list_head; 61 struct root_domain; 62 struct rq; 63 struct sched_attr; 64 struct sched_param; 65 struct seq_file; 66 struct sighand_struct; 67 struct signal_struct; 68 struct task_delay_info; 69 struct task_group; 70 71 /* 72 * Task state bitmask. NOTE! These bits are also 73 * encoded in fs/proc/array.c: get_task_state(). 74 * 75 * We have two separate sets of flags: task->state 76 * is about runnability, while task->exit_state are 77 * about the task exiting. Confusing, but this way 78 * modifying one set can't modify the other one by 79 * mistake. 80 */ 81 82 /* Used in tsk->state: */ 83 #define TASK_RUNNING 0x0000 84 #define TASK_INTERRUPTIBLE 0x0001 85 #define TASK_UNINTERRUPTIBLE 0x0002 86 #define __TASK_STOPPED 0x0004 87 #define __TASK_TRACED 0x0008 88 /* Used in tsk->exit_state: */ 89 #define EXIT_DEAD 0x0010 90 #define EXIT_ZOMBIE 0x0020 91 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 92 /* Used in tsk->state again: */ 93 #define TASK_PARKED 0x0040 94 #define TASK_DEAD 0x0080 95 #define TASK_WAKEKILL 0x0100 96 #define TASK_WAKING 0x0200 97 #define TASK_NOLOAD 0x0400 98 #define TASK_NEW 0x0800 99 /* RT specific auxilliary flag to mark RT lock waiters */ 100 #define TASK_RTLOCK_WAIT 0x1000 101 #define TASK_STATE_MAX 0x2000 102 103 /* Convenience macros for the sake of set_current_state: */ 104 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 105 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 106 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) 107 108 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 109 110 /* Convenience macros for the sake of wake_up(): */ 111 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 112 113 /* get_task_state(): */ 114 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 115 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 116 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 117 TASK_PARKED) 118 119 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 120 121 #define task_is_traced(task) ((READ_ONCE(task->__state) & __TASK_TRACED) != 0) 122 123 #define task_is_stopped(task) ((READ_ONCE(task->__state) & __TASK_STOPPED) != 0) 124 125 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->__state) & (__TASK_STOPPED | __TASK_TRACED)) != 0) 126 127 /* 128 * Special states are those that do not use the normal wait-loop pattern. See 129 * the comment with set_special_state(). 130 */ 131 #define is_special_task_state(state) \ 132 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD)) 133 134 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 135 # define debug_normal_state_change(state_value) \ 136 do { \ 137 WARN_ON_ONCE(is_special_task_state(state_value)); \ 138 current->task_state_change = _THIS_IP_; \ 139 } while (0) 140 141 # define debug_special_state_change(state_value) \ 142 do { \ 143 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 144 current->task_state_change = _THIS_IP_; \ 145 } while (0) 146 147 # define debug_rtlock_wait_set_state() \ 148 do { \ 149 current->saved_state_change = current->task_state_change;\ 150 current->task_state_change = _THIS_IP_; \ 151 } while (0) 152 153 # define debug_rtlock_wait_restore_state() \ 154 do { \ 155 current->task_state_change = current->saved_state_change;\ 156 } while (0) 157 158 #else 159 # define debug_normal_state_change(cond) do { } while (0) 160 # define debug_special_state_change(cond) do { } while (0) 161 # define debug_rtlock_wait_set_state() do { } while (0) 162 # define debug_rtlock_wait_restore_state() do { } while (0) 163 #endif 164 165 /* 166 * set_current_state() includes a barrier so that the write of current->state 167 * is correctly serialised wrt the caller's subsequent test of whether to 168 * actually sleep: 169 * 170 * for (;;) { 171 * set_current_state(TASK_UNINTERRUPTIBLE); 172 * if (CONDITION) 173 * break; 174 * 175 * schedule(); 176 * } 177 * __set_current_state(TASK_RUNNING); 178 * 179 * If the caller does not need such serialisation (because, for instance, the 180 * CONDITION test and condition change and wakeup are under the same lock) then 181 * use __set_current_state(). 182 * 183 * The above is typically ordered against the wakeup, which does: 184 * 185 * CONDITION = 1; 186 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 187 * 188 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 189 * accessing p->state. 190 * 191 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 192 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 193 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 194 * 195 * However, with slightly different timing the wakeup TASK_RUNNING store can 196 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 197 * a problem either because that will result in one extra go around the loop 198 * and our @cond test will save the day. 199 * 200 * Also see the comments of try_to_wake_up(). 201 */ 202 #define __set_current_state(state_value) \ 203 do { \ 204 debug_normal_state_change((state_value)); \ 205 WRITE_ONCE(current->__state, (state_value)); \ 206 } while (0) 207 208 #define set_current_state(state_value) \ 209 do { \ 210 debug_normal_state_change((state_value)); \ 211 smp_store_mb(current->__state, (state_value)); \ 212 } while (0) 213 214 /* 215 * set_special_state() should be used for those states when the blocking task 216 * can not use the regular condition based wait-loop. In that case we must 217 * serialize against wakeups such that any possible in-flight TASK_RUNNING 218 * stores will not collide with our state change. 219 */ 220 #define set_special_state(state_value) \ 221 do { \ 222 unsigned long flags; /* may shadow */ \ 223 \ 224 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 225 debug_special_state_change((state_value)); \ 226 WRITE_ONCE(current->__state, (state_value)); \ 227 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 228 } while (0) 229 230 /* 231 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 232 * 233 * RT's spin/rwlock substitutions are state preserving. The state of the 234 * task when blocking on the lock is saved in task_struct::saved_state and 235 * restored after the lock has been acquired. These operations are 236 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 237 * lock related wakeups while the task is blocked on the lock are 238 * redirected to operate on task_struct::saved_state to ensure that these 239 * are not dropped. On restore task_struct::saved_state is set to 240 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 241 * 242 * The lock operation looks like this: 243 * 244 * current_save_and_set_rtlock_wait_state(); 245 * for (;;) { 246 * if (try_lock()) 247 * break; 248 * raw_spin_unlock_irq(&lock->wait_lock); 249 * schedule_rtlock(); 250 * raw_spin_lock_irq(&lock->wait_lock); 251 * set_current_state(TASK_RTLOCK_WAIT); 252 * } 253 * current_restore_rtlock_saved_state(); 254 */ 255 #define current_save_and_set_rtlock_wait_state() \ 256 do { \ 257 lockdep_assert_irqs_disabled(); \ 258 raw_spin_lock(¤t->pi_lock); \ 259 current->saved_state = current->__state; \ 260 debug_rtlock_wait_set_state(); \ 261 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 262 raw_spin_unlock(¤t->pi_lock); \ 263 } while (0); 264 265 #define current_restore_rtlock_saved_state() \ 266 do { \ 267 lockdep_assert_irqs_disabled(); \ 268 raw_spin_lock(¤t->pi_lock); \ 269 debug_rtlock_wait_restore_state(); \ 270 WRITE_ONCE(current->__state, current->saved_state); \ 271 current->saved_state = TASK_RUNNING; \ 272 raw_spin_unlock(¤t->pi_lock); \ 273 } while (0); 274 275 #define get_current_state() READ_ONCE(current->__state) 276 277 /* Task command name length: */ 278 #define TASK_COMM_LEN 16 279 280 extern void scheduler_tick(void); 281 282 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 283 284 extern long schedule_timeout(long timeout); 285 extern long schedule_timeout_interruptible(long timeout); 286 extern long schedule_timeout_killable(long timeout); 287 extern long schedule_timeout_uninterruptible(long timeout); 288 extern long schedule_timeout_idle(long timeout); 289 asmlinkage void schedule(void); 290 extern void schedule_preempt_disabled(void); 291 asmlinkage void preempt_schedule_irq(void); 292 #ifdef CONFIG_PREEMPT_RT 293 extern void schedule_rtlock(void); 294 #endif 295 296 extern int __must_check io_schedule_prepare(void); 297 extern void io_schedule_finish(int token); 298 extern long io_schedule_timeout(long timeout); 299 extern void io_schedule(void); 300 301 /** 302 * struct prev_cputime - snapshot of system and user cputime 303 * @utime: time spent in user mode 304 * @stime: time spent in system mode 305 * @lock: protects the above two fields 306 * 307 * Stores previous user/system time values such that we can guarantee 308 * monotonicity. 309 */ 310 struct prev_cputime { 311 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 312 u64 utime; 313 u64 stime; 314 raw_spinlock_t lock; 315 #endif 316 }; 317 318 enum vtime_state { 319 /* Task is sleeping or running in a CPU with VTIME inactive: */ 320 VTIME_INACTIVE = 0, 321 /* Task is idle */ 322 VTIME_IDLE, 323 /* Task runs in kernelspace in a CPU with VTIME active: */ 324 VTIME_SYS, 325 /* Task runs in userspace in a CPU with VTIME active: */ 326 VTIME_USER, 327 /* Task runs as guests in a CPU with VTIME active: */ 328 VTIME_GUEST, 329 }; 330 331 struct vtime { 332 seqcount_t seqcount; 333 unsigned long long starttime; 334 enum vtime_state state; 335 unsigned int cpu; 336 u64 utime; 337 u64 stime; 338 u64 gtime; 339 }; 340 341 /* 342 * Utilization clamp constraints. 343 * @UCLAMP_MIN: Minimum utilization 344 * @UCLAMP_MAX: Maximum utilization 345 * @UCLAMP_CNT: Utilization clamp constraints count 346 */ 347 enum uclamp_id { 348 UCLAMP_MIN = 0, 349 UCLAMP_MAX, 350 UCLAMP_CNT 351 }; 352 353 #ifdef CONFIG_SMP 354 extern struct root_domain def_root_domain; 355 extern struct mutex sched_domains_mutex; 356 #endif 357 358 struct sched_info { 359 #ifdef CONFIG_SCHED_INFO 360 /* Cumulative counters: */ 361 362 /* # of times we have run on this CPU: */ 363 unsigned long pcount; 364 365 /* Time spent waiting on a runqueue: */ 366 unsigned long long run_delay; 367 368 /* Timestamps: */ 369 370 /* When did we last run on a CPU? */ 371 unsigned long long last_arrival; 372 373 /* When were we last queued to run? */ 374 unsigned long long last_queued; 375 376 #endif /* CONFIG_SCHED_INFO */ 377 }; 378 379 /* 380 * Integer metrics need fixed point arithmetic, e.g., sched/fair 381 * has a few: load, load_avg, util_avg, freq, and capacity. 382 * 383 * We define a basic fixed point arithmetic range, and then formalize 384 * all these metrics based on that basic range. 385 */ 386 # define SCHED_FIXEDPOINT_SHIFT 10 387 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 388 389 /* Increase resolution of cpu_capacity calculations */ 390 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 391 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 392 393 struct load_weight { 394 unsigned long weight; 395 u32 inv_weight; 396 }; 397 398 /** 399 * struct util_est - Estimation utilization of FAIR tasks 400 * @enqueued: instantaneous estimated utilization of a task/cpu 401 * @ewma: the Exponential Weighted Moving Average (EWMA) 402 * utilization of a task 403 * 404 * Support data structure to track an Exponential Weighted Moving Average 405 * (EWMA) of a FAIR task's utilization. New samples are added to the moving 406 * average each time a task completes an activation. Sample's weight is chosen 407 * so that the EWMA will be relatively insensitive to transient changes to the 408 * task's workload. 409 * 410 * The enqueued attribute has a slightly different meaning for tasks and cpus: 411 * - task: the task's util_avg at last task dequeue time 412 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU 413 * Thus, the util_est.enqueued of a task represents the contribution on the 414 * estimated utilization of the CPU where that task is currently enqueued. 415 * 416 * Only for tasks we track a moving average of the past instantaneous 417 * estimated utilization. This allows to absorb sporadic drops in utilization 418 * of an otherwise almost periodic task. 419 * 420 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 421 * updates. When a task is dequeued, its util_est should not be updated if its 422 * util_avg has not been updated in the meantime. 423 * This information is mapped into the MSB bit of util_est.enqueued at dequeue 424 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg 425 * for a task) it is safe to use MSB. 426 */ 427 struct util_est { 428 unsigned int enqueued; 429 unsigned int ewma; 430 #define UTIL_EST_WEIGHT_SHIFT 2 431 #define UTIL_AVG_UNCHANGED 0x80000000 432 } __attribute__((__aligned__(sizeof(u64)))); 433 434 /* 435 * The load/runnable/util_avg accumulates an infinite geometric series 436 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 437 * 438 * [load_avg definition] 439 * 440 * load_avg = runnable% * scale_load_down(load) 441 * 442 * [runnable_avg definition] 443 * 444 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 445 * 446 * [util_avg definition] 447 * 448 * util_avg = running% * SCHED_CAPACITY_SCALE 449 * 450 * where runnable% is the time ratio that a sched_entity is runnable and 451 * running% the time ratio that a sched_entity is running. 452 * 453 * For cfs_rq, they are the aggregated values of all runnable and blocked 454 * sched_entities. 455 * 456 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 457 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 458 * for computing those signals (see update_rq_clock_pelt()) 459 * 460 * N.B., the above ratios (runnable% and running%) themselves are in the 461 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 462 * to as large a range as necessary. This is for example reflected by 463 * util_avg's SCHED_CAPACITY_SCALE. 464 * 465 * [Overflow issue] 466 * 467 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 468 * with the highest load (=88761), always runnable on a single cfs_rq, 469 * and should not overflow as the number already hits PID_MAX_LIMIT. 470 * 471 * For all other cases (including 32-bit kernels), struct load_weight's 472 * weight will overflow first before we do, because: 473 * 474 * Max(load_avg) <= Max(load.weight) 475 * 476 * Then it is the load_weight's responsibility to consider overflow 477 * issues. 478 */ 479 struct sched_avg { 480 u64 last_update_time; 481 u64 load_sum; 482 u64 runnable_sum; 483 u32 util_sum; 484 u32 period_contrib; 485 unsigned long load_avg; 486 unsigned long runnable_avg; 487 unsigned long util_avg; 488 struct util_est util_est; 489 } ____cacheline_aligned; 490 491 struct sched_statistics { 492 #ifdef CONFIG_SCHEDSTATS 493 u64 wait_start; 494 u64 wait_max; 495 u64 wait_count; 496 u64 wait_sum; 497 u64 iowait_count; 498 u64 iowait_sum; 499 500 u64 sleep_start; 501 u64 sleep_max; 502 s64 sum_sleep_runtime; 503 504 u64 block_start; 505 u64 block_max; 506 s64 sum_block_runtime; 507 508 u64 exec_max; 509 u64 slice_max; 510 511 u64 nr_migrations_cold; 512 u64 nr_failed_migrations_affine; 513 u64 nr_failed_migrations_running; 514 u64 nr_failed_migrations_hot; 515 u64 nr_forced_migrations; 516 517 u64 nr_wakeups; 518 u64 nr_wakeups_sync; 519 u64 nr_wakeups_migrate; 520 u64 nr_wakeups_local; 521 u64 nr_wakeups_remote; 522 u64 nr_wakeups_affine; 523 u64 nr_wakeups_affine_attempts; 524 u64 nr_wakeups_passive; 525 u64 nr_wakeups_idle; 526 527 #ifdef CONFIG_SCHED_CORE 528 u64 core_forceidle_sum; 529 #endif 530 #endif /* CONFIG_SCHEDSTATS */ 531 } ____cacheline_aligned; 532 533 struct sched_entity { 534 /* For load-balancing: */ 535 struct load_weight load; 536 struct rb_node run_node; 537 struct list_head group_node; 538 unsigned int on_rq; 539 540 u64 exec_start; 541 u64 sum_exec_runtime; 542 u64 vruntime; 543 u64 prev_sum_exec_runtime; 544 545 u64 nr_migrations; 546 547 #ifdef CONFIG_FAIR_GROUP_SCHED 548 int depth; 549 struct sched_entity *parent; 550 /* rq on which this entity is (to be) queued: */ 551 struct cfs_rq *cfs_rq; 552 /* rq "owned" by this entity/group: */ 553 struct cfs_rq *my_q; 554 /* cached value of my_q->h_nr_running */ 555 unsigned long runnable_weight; 556 #endif 557 558 #ifdef CONFIG_SMP 559 /* 560 * Per entity load average tracking. 561 * 562 * Put into separate cache line so it does not 563 * collide with read-mostly values above. 564 */ 565 struct sched_avg avg; 566 #endif 567 }; 568 569 struct sched_rt_entity { 570 struct list_head run_list; 571 unsigned long timeout; 572 unsigned long watchdog_stamp; 573 unsigned int time_slice; 574 unsigned short on_rq; 575 unsigned short on_list; 576 577 struct sched_rt_entity *back; 578 #ifdef CONFIG_RT_GROUP_SCHED 579 struct sched_rt_entity *parent; 580 /* rq on which this entity is (to be) queued: */ 581 struct rt_rq *rt_rq; 582 /* rq "owned" by this entity/group: */ 583 struct rt_rq *my_q; 584 #endif 585 } __randomize_layout; 586 587 struct sched_dl_entity { 588 struct rb_node rb_node; 589 590 /* 591 * Original scheduling parameters. Copied here from sched_attr 592 * during sched_setattr(), they will remain the same until 593 * the next sched_setattr(). 594 */ 595 u64 dl_runtime; /* Maximum runtime for each instance */ 596 u64 dl_deadline; /* Relative deadline of each instance */ 597 u64 dl_period; /* Separation of two instances (period) */ 598 u64 dl_bw; /* dl_runtime / dl_period */ 599 u64 dl_density; /* dl_runtime / dl_deadline */ 600 601 /* 602 * Actual scheduling parameters. Initialized with the values above, 603 * they are continuously updated during task execution. Note that 604 * the remaining runtime could be < 0 in case we are in overrun. 605 */ 606 s64 runtime; /* Remaining runtime for this instance */ 607 u64 deadline; /* Absolute deadline for this instance */ 608 unsigned int flags; /* Specifying the scheduler behaviour */ 609 610 /* 611 * Some bool flags: 612 * 613 * @dl_throttled tells if we exhausted the runtime. If so, the 614 * task has to wait for a replenishment to be performed at the 615 * next firing of dl_timer. 616 * 617 * @dl_boosted tells if we are boosted due to DI. If so we are 618 * outside bandwidth enforcement mechanism (but only until we 619 * exit the critical section); 620 * 621 * @dl_yielded tells if task gave up the CPU before consuming 622 * all its available runtime during the last job. 623 * 624 * @dl_non_contending tells if the task is inactive while still 625 * contributing to the active utilization. In other words, it 626 * indicates if the inactive timer has been armed and its handler 627 * has not been executed yet. This flag is useful to avoid race 628 * conditions between the inactive timer handler and the wakeup 629 * code. 630 * 631 * @dl_overrun tells if the task asked to be informed about runtime 632 * overruns. 633 */ 634 unsigned int dl_throttled : 1; 635 unsigned int dl_yielded : 1; 636 unsigned int dl_non_contending : 1; 637 unsigned int dl_overrun : 1; 638 639 /* 640 * Bandwidth enforcement timer. Each -deadline task has its 641 * own bandwidth to be enforced, thus we need one timer per task. 642 */ 643 struct hrtimer dl_timer; 644 645 /* 646 * Inactive timer, responsible for decreasing the active utilization 647 * at the "0-lag time". When a -deadline task blocks, it contributes 648 * to GRUB's active utilization until the "0-lag time", hence a 649 * timer is needed to decrease the active utilization at the correct 650 * time. 651 */ 652 struct hrtimer inactive_timer; 653 654 #ifdef CONFIG_RT_MUTEXES 655 /* 656 * Priority Inheritance. When a DEADLINE scheduling entity is boosted 657 * pi_se points to the donor, otherwise points to the dl_se it belongs 658 * to (the original one/itself). 659 */ 660 struct sched_dl_entity *pi_se; 661 #endif 662 }; 663 664 #ifdef CONFIG_UCLAMP_TASK 665 /* Number of utilization clamp buckets (shorter alias) */ 666 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT 667 668 /* 669 * Utilization clamp for a scheduling entity 670 * @value: clamp value "assigned" to a se 671 * @bucket_id: bucket index corresponding to the "assigned" value 672 * @active: the se is currently refcounted in a rq's bucket 673 * @user_defined: the requested clamp value comes from user-space 674 * 675 * The bucket_id is the index of the clamp bucket matching the clamp value 676 * which is pre-computed and stored to avoid expensive integer divisions from 677 * the fast path. 678 * 679 * The active bit is set whenever a task has got an "effective" value assigned, 680 * which can be different from the clamp value "requested" from user-space. 681 * This allows to know a task is refcounted in the rq's bucket corresponding 682 * to the "effective" bucket_id. 683 * 684 * The user_defined bit is set whenever a task has got a task-specific clamp 685 * value requested from userspace, i.e. the system defaults apply to this task 686 * just as a restriction. This allows to relax default clamps when a less 687 * restrictive task-specific value has been requested, thus allowing to 688 * implement a "nice" semantic. For example, a task running with a 20% 689 * default boost can still drop its own boosting to 0%. 690 */ 691 struct uclamp_se { 692 unsigned int value : bits_per(SCHED_CAPACITY_SCALE); 693 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS); 694 unsigned int active : 1; 695 unsigned int user_defined : 1; 696 }; 697 #endif /* CONFIG_UCLAMP_TASK */ 698 699 union rcu_special { 700 struct { 701 u8 blocked; 702 u8 need_qs; 703 u8 exp_hint; /* Hint for performance. */ 704 u8 need_mb; /* Readers need smp_mb(). */ 705 } b; /* Bits. */ 706 u32 s; /* Set of bits. */ 707 }; 708 709 enum perf_event_task_context { 710 perf_invalid_context = -1, 711 perf_hw_context = 0, 712 perf_sw_context, 713 perf_nr_task_contexts, 714 }; 715 716 struct wake_q_node { 717 struct wake_q_node *next; 718 }; 719 720 struct kmap_ctrl { 721 #ifdef CONFIG_KMAP_LOCAL 722 int idx; 723 pte_t pteval[KM_MAX_IDX]; 724 #endif 725 }; 726 727 struct task_struct { 728 #ifdef CONFIG_THREAD_INFO_IN_TASK 729 /* 730 * For reasons of header soup (see current_thread_info()), this 731 * must be the first element of task_struct. 732 */ 733 struct thread_info thread_info; 734 #endif 735 unsigned int __state; 736 737 #ifdef CONFIG_PREEMPT_RT 738 /* saved state for "spinlock sleepers" */ 739 unsigned int saved_state; 740 #endif 741 742 /* 743 * This begins the randomizable portion of task_struct. Only 744 * scheduling-critical items should be added above here. 745 */ 746 randomized_struct_fields_start 747 748 void *stack; 749 refcount_t usage; 750 /* Per task flags (PF_*), defined further below: */ 751 unsigned int flags; 752 unsigned int ptrace; 753 754 #ifdef CONFIG_SMP 755 int on_cpu; 756 struct __call_single_node wake_entry; 757 unsigned int wakee_flips; 758 unsigned long wakee_flip_decay_ts; 759 struct task_struct *last_wakee; 760 761 /* 762 * recent_used_cpu is initially set as the last CPU used by a task 763 * that wakes affine another task. Waker/wakee relationships can 764 * push tasks around a CPU where each wakeup moves to the next one. 765 * Tracking a recently used CPU allows a quick search for a recently 766 * used CPU that may be idle. 767 */ 768 int recent_used_cpu; 769 int wake_cpu; 770 #endif 771 int on_rq; 772 773 int prio; 774 int static_prio; 775 int normal_prio; 776 unsigned int rt_priority; 777 778 struct sched_entity se; 779 struct sched_rt_entity rt; 780 struct sched_dl_entity dl; 781 const struct sched_class *sched_class; 782 783 #ifdef CONFIG_SCHED_CORE 784 struct rb_node core_node; 785 unsigned long core_cookie; 786 unsigned int core_occupation; 787 #endif 788 789 #ifdef CONFIG_CGROUP_SCHED 790 struct task_group *sched_task_group; 791 #endif 792 793 #ifdef CONFIG_UCLAMP_TASK 794 /* 795 * Clamp values requested for a scheduling entity. 796 * Must be updated with task_rq_lock() held. 797 */ 798 struct uclamp_se uclamp_req[UCLAMP_CNT]; 799 /* 800 * Effective clamp values used for a scheduling entity. 801 * Must be updated with task_rq_lock() held. 802 */ 803 struct uclamp_se uclamp[UCLAMP_CNT]; 804 #endif 805 806 struct sched_statistics stats; 807 808 #ifdef CONFIG_PREEMPT_NOTIFIERS 809 /* List of struct preempt_notifier: */ 810 struct hlist_head preempt_notifiers; 811 #endif 812 813 #ifdef CONFIG_BLK_DEV_IO_TRACE 814 unsigned int btrace_seq; 815 #endif 816 817 unsigned int policy; 818 int nr_cpus_allowed; 819 const cpumask_t *cpus_ptr; 820 cpumask_t *user_cpus_ptr; 821 cpumask_t cpus_mask; 822 void *migration_pending; 823 #ifdef CONFIG_SMP 824 unsigned short migration_disabled; 825 #endif 826 unsigned short migration_flags; 827 828 #ifdef CONFIG_PREEMPT_RCU 829 int rcu_read_lock_nesting; 830 union rcu_special rcu_read_unlock_special; 831 struct list_head rcu_node_entry; 832 struct rcu_node *rcu_blocked_node; 833 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 834 835 #ifdef CONFIG_TASKS_RCU 836 unsigned long rcu_tasks_nvcsw; 837 u8 rcu_tasks_holdout; 838 u8 rcu_tasks_idx; 839 int rcu_tasks_idle_cpu; 840 struct list_head rcu_tasks_holdout_list; 841 #endif /* #ifdef CONFIG_TASKS_RCU */ 842 843 #ifdef CONFIG_TASKS_TRACE_RCU 844 int trc_reader_nesting; 845 int trc_ipi_to_cpu; 846 union rcu_special trc_reader_special; 847 bool trc_reader_checked; 848 struct list_head trc_holdout_list; 849 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 850 851 struct sched_info sched_info; 852 853 struct list_head tasks; 854 #ifdef CONFIG_SMP 855 struct plist_node pushable_tasks; 856 struct rb_node pushable_dl_tasks; 857 #endif 858 859 struct mm_struct *mm; 860 struct mm_struct *active_mm; 861 862 /* Per-thread vma caching: */ 863 struct vmacache vmacache; 864 865 #ifdef SPLIT_RSS_COUNTING 866 struct task_rss_stat rss_stat; 867 #endif 868 int exit_state; 869 int exit_code; 870 int exit_signal; 871 /* The signal sent when the parent dies: */ 872 int pdeath_signal; 873 /* JOBCTL_*, siglock protected: */ 874 unsigned long jobctl; 875 876 /* Used for emulating ABI behavior of previous Linux versions: */ 877 unsigned int personality; 878 879 /* Scheduler bits, serialized by scheduler locks: */ 880 unsigned sched_reset_on_fork:1; 881 unsigned sched_contributes_to_load:1; 882 unsigned sched_migrated:1; 883 #ifdef CONFIG_PSI 884 unsigned sched_psi_wake_requeue:1; 885 #endif 886 887 /* Force alignment to the next boundary: */ 888 unsigned :0; 889 890 /* Unserialized, strictly 'current' */ 891 892 /* 893 * This field must not be in the scheduler word above due to wakelist 894 * queueing no longer being serialized by p->on_cpu. However: 895 * 896 * p->XXX = X; ttwu() 897 * schedule() if (p->on_rq && ..) // false 898 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 899 * deactivate_task() ttwu_queue_wakelist()) 900 * p->on_rq = 0; p->sched_remote_wakeup = Y; 901 * 902 * guarantees all stores of 'current' are visible before 903 * ->sched_remote_wakeup gets used, so it can be in this word. 904 */ 905 unsigned sched_remote_wakeup:1; 906 907 /* Bit to tell LSMs we're in execve(): */ 908 unsigned in_execve:1; 909 unsigned in_iowait:1; 910 #ifndef TIF_RESTORE_SIGMASK 911 unsigned restore_sigmask:1; 912 #endif 913 #ifdef CONFIG_MEMCG 914 unsigned in_user_fault:1; 915 #endif 916 #ifdef CONFIG_COMPAT_BRK 917 unsigned brk_randomized:1; 918 #endif 919 #ifdef CONFIG_CGROUPS 920 /* disallow userland-initiated cgroup migration */ 921 unsigned no_cgroup_migration:1; 922 /* task is frozen/stopped (used by the cgroup freezer) */ 923 unsigned frozen:1; 924 #endif 925 #ifdef CONFIG_BLK_CGROUP 926 unsigned use_memdelay:1; 927 #endif 928 #ifdef CONFIG_PSI 929 /* Stalled due to lack of memory */ 930 unsigned in_memstall:1; 931 #endif 932 #ifdef CONFIG_PAGE_OWNER 933 /* Used by page_owner=on to detect recursion in page tracking. */ 934 unsigned in_page_owner:1; 935 #endif 936 #ifdef CONFIG_EVENTFD 937 /* Recursion prevention for eventfd_signal() */ 938 unsigned in_eventfd_signal:1; 939 #endif 940 941 unsigned long atomic_flags; /* Flags requiring atomic access. */ 942 943 struct restart_block restart_block; 944 945 pid_t pid; 946 pid_t tgid; 947 948 #ifdef CONFIG_STACKPROTECTOR 949 /* Canary value for the -fstack-protector GCC feature: */ 950 unsigned long stack_canary; 951 #endif 952 /* 953 * Pointers to the (original) parent process, youngest child, younger sibling, 954 * older sibling, respectively. (p->father can be replaced with 955 * p->real_parent->pid) 956 */ 957 958 /* Real parent process: */ 959 struct task_struct __rcu *real_parent; 960 961 /* Recipient of SIGCHLD, wait4() reports: */ 962 struct task_struct __rcu *parent; 963 964 /* 965 * Children/sibling form the list of natural children: 966 */ 967 struct list_head children; 968 struct list_head sibling; 969 struct task_struct *group_leader; 970 971 /* 972 * 'ptraced' is the list of tasks this task is using ptrace() on. 973 * 974 * This includes both natural children and PTRACE_ATTACH targets. 975 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 976 */ 977 struct list_head ptraced; 978 struct list_head ptrace_entry; 979 980 /* PID/PID hash table linkage. */ 981 struct pid *thread_pid; 982 struct hlist_node pid_links[PIDTYPE_MAX]; 983 struct list_head thread_group; 984 struct list_head thread_node; 985 986 struct completion *vfork_done; 987 988 /* CLONE_CHILD_SETTID: */ 989 int __user *set_child_tid; 990 991 /* CLONE_CHILD_CLEARTID: */ 992 int __user *clear_child_tid; 993 994 /* PF_KTHREAD | PF_IO_WORKER */ 995 void *worker_private; 996 997 u64 utime; 998 u64 stime; 999 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1000 u64 utimescaled; 1001 u64 stimescaled; 1002 #endif 1003 u64 gtime; 1004 struct prev_cputime prev_cputime; 1005 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1006 struct vtime vtime; 1007 #endif 1008 1009 #ifdef CONFIG_NO_HZ_FULL 1010 atomic_t tick_dep_mask; 1011 #endif 1012 /* Context switch counts: */ 1013 unsigned long nvcsw; 1014 unsigned long nivcsw; 1015 1016 /* Monotonic time in nsecs: */ 1017 u64 start_time; 1018 1019 /* Boot based time in nsecs: */ 1020 u64 start_boottime; 1021 1022 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1023 unsigned long min_flt; 1024 unsigned long maj_flt; 1025 1026 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1027 struct posix_cputimers posix_cputimers; 1028 1029 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1030 struct posix_cputimers_work posix_cputimers_work; 1031 #endif 1032 1033 /* Process credentials: */ 1034 1035 /* Tracer's credentials at attach: */ 1036 const struct cred __rcu *ptracer_cred; 1037 1038 /* Objective and real subjective task credentials (COW): */ 1039 const struct cred __rcu *real_cred; 1040 1041 /* Effective (overridable) subjective task credentials (COW): */ 1042 const struct cred __rcu *cred; 1043 1044 #ifdef CONFIG_KEYS 1045 /* Cached requested key. */ 1046 struct key *cached_requested_key; 1047 #endif 1048 1049 /* 1050 * executable name, excluding path. 1051 * 1052 * - normally initialized setup_new_exec() 1053 * - access it with [gs]et_task_comm() 1054 * - lock it with task_lock() 1055 */ 1056 char comm[TASK_COMM_LEN]; 1057 1058 struct nameidata *nameidata; 1059 1060 #ifdef CONFIG_SYSVIPC 1061 struct sysv_sem sysvsem; 1062 struct sysv_shm sysvshm; 1063 #endif 1064 #ifdef CONFIG_DETECT_HUNG_TASK 1065 unsigned long last_switch_count; 1066 unsigned long last_switch_time; 1067 #endif 1068 /* Filesystem information: */ 1069 struct fs_struct *fs; 1070 1071 /* Open file information: */ 1072 struct files_struct *files; 1073 1074 #ifdef CONFIG_IO_URING 1075 struct io_uring_task *io_uring; 1076 #endif 1077 1078 /* Namespaces: */ 1079 struct nsproxy *nsproxy; 1080 1081 /* Signal handlers: */ 1082 struct signal_struct *signal; 1083 struct sighand_struct __rcu *sighand; 1084 sigset_t blocked; 1085 sigset_t real_blocked; 1086 /* Restored if set_restore_sigmask() was used: */ 1087 sigset_t saved_sigmask; 1088 struct sigpending pending; 1089 unsigned long sas_ss_sp; 1090 size_t sas_ss_size; 1091 unsigned int sas_ss_flags; 1092 1093 struct callback_head *task_works; 1094 1095 #ifdef CONFIG_AUDIT 1096 #ifdef CONFIG_AUDITSYSCALL 1097 struct audit_context *audit_context; 1098 #endif 1099 kuid_t loginuid; 1100 unsigned int sessionid; 1101 #endif 1102 struct seccomp seccomp; 1103 struct syscall_user_dispatch syscall_dispatch; 1104 1105 /* Thread group tracking: */ 1106 u64 parent_exec_id; 1107 u64 self_exec_id; 1108 1109 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1110 spinlock_t alloc_lock; 1111 1112 /* Protection of the PI data structures: */ 1113 raw_spinlock_t pi_lock; 1114 1115 struct wake_q_node wake_q; 1116 1117 #ifdef CONFIG_RT_MUTEXES 1118 /* PI waiters blocked on a rt_mutex held by this task: */ 1119 struct rb_root_cached pi_waiters; 1120 /* Updated under owner's pi_lock and rq lock */ 1121 struct task_struct *pi_top_task; 1122 /* Deadlock detection and priority inheritance handling: */ 1123 struct rt_mutex_waiter *pi_blocked_on; 1124 #endif 1125 1126 #ifdef CONFIG_DEBUG_MUTEXES 1127 /* Mutex deadlock detection: */ 1128 struct mutex_waiter *blocked_on; 1129 #endif 1130 1131 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1132 int non_block_count; 1133 #endif 1134 1135 #ifdef CONFIG_TRACE_IRQFLAGS 1136 struct irqtrace_events irqtrace; 1137 unsigned int hardirq_threaded; 1138 u64 hardirq_chain_key; 1139 int softirqs_enabled; 1140 int softirq_context; 1141 int irq_config; 1142 #endif 1143 #ifdef CONFIG_PREEMPT_RT 1144 int softirq_disable_cnt; 1145 #endif 1146 1147 #ifdef CONFIG_LOCKDEP 1148 # define MAX_LOCK_DEPTH 48UL 1149 u64 curr_chain_key; 1150 int lockdep_depth; 1151 unsigned int lockdep_recursion; 1152 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1153 #endif 1154 1155 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1156 unsigned int in_ubsan; 1157 #endif 1158 1159 /* Journalling filesystem info: */ 1160 void *journal_info; 1161 1162 /* Stacked block device info: */ 1163 struct bio_list *bio_list; 1164 1165 /* Stack plugging: */ 1166 struct blk_plug *plug; 1167 1168 /* VM state: */ 1169 struct reclaim_state *reclaim_state; 1170 1171 struct backing_dev_info *backing_dev_info; 1172 1173 struct io_context *io_context; 1174 1175 #ifdef CONFIG_COMPACTION 1176 struct capture_control *capture_control; 1177 #endif 1178 /* Ptrace state: */ 1179 unsigned long ptrace_message; 1180 kernel_siginfo_t *last_siginfo; 1181 1182 struct task_io_accounting ioac; 1183 #ifdef CONFIG_PSI 1184 /* Pressure stall state */ 1185 unsigned int psi_flags; 1186 #endif 1187 #ifdef CONFIG_TASK_XACCT 1188 /* Accumulated RSS usage: */ 1189 u64 acct_rss_mem1; 1190 /* Accumulated virtual memory usage: */ 1191 u64 acct_vm_mem1; 1192 /* stime + utime since last update: */ 1193 u64 acct_timexpd; 1194 #endif 1195 #ifdef CONFIG_CPUSETS 1196 /* Protected by ->alloc_lock: */ 1197 nodemask_t mems_allowed; 1198 /* Sequence number to catch updates: */ 1199 seqcount_spinlock_t mems_allowed_seq; 1200 int cpuset_mem_spread_rotor; 1201 int cpuset_slab_spread_rotor; 1202 #endif 1203 #ifdef CONFIG_CGROUPS 1204 /* Control Group info protected by css_set_lock: */ 1205 struct css_set __rcu *cgroups; 1206 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1207 struct list_head cg_list; 1208 #endif 1209 #ifdef CONFIG_X86_CPU_RESCTRL 1210 u32 closid; 1211 u32 rmid; 1212 #endif 1213 #ifdef CONFIG_FUTEX 1214 struct robust_list_head __user *robust_list; 1215 #ifdef CONFIG_COMPAT 1216 struct compat_robust_list_head __user *compat_robust_list; 1217 #endif 1218 struct list_head pi_state_list; 1219 struct futex_pi_state *pi_state_cache; 1220 struct mutex futex_exit_mutex; 1221 unsigned int futex_state; 1222 #endif 1223 #ifdef CONFIG_PERF_EVENTS 1224 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 1225 struct mutex perf_event_mutex; 1226 struct list_head perf_event_list; 1227 #endif 1228 #ifdef CONFIG_DEBUG_PREEMPT 1229 unsigned long preempt_disable_ip; 1230 #endif 1231 #ifdef CONFIG_NUMA 1232 /* Protected by alloc_lock: */ 1233 struct mempolicy *mempolicy; 1234 short il_prev; 1235 short pref_node_fork; 1236 #endif 1237 #ifdef CONFIG_NUMA_BALANCING 1238 int numa_scan_seq; 1239 unsigned int numa_scan_period; 1240 unsigned int numa_scan_period_max; 1241 int numa_preferred_nid; 1242 unsigned long numa_migrate_retry; 1243 /* Migration stamp: */ 1244 u64 node_stamp; 1245 u64 last_task_numa_placement; 1246 u64 last_sum_exec_runtime; 1247 struct callback_head numa_work; 1248 1249 /* 1250 * This pointer is only modified for current in syscall and 1251 * pagefault context (and for tasks being destroyed), so it can be read 1252 * from any of the following contexts: 1253 * - RCU read-side critical section 1254 * - current->numa_group from everywhere 1255 * - task's runqueue locked, task not running 1256 */ 1257 struct numa_group __rcu *numa_group; 1258 1259 /* 1260 * numa_faults is an array split into four regions: 1261 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1262 * in this precise order. 1263 * 1264 * faults_memory: Exponential decaying average of faults on a per-node 1265 * basis. Scheduling placement decisions are made based on these 1266 * counts. The values remain static for the duration of a PTE scan. 1267 * faults_cpu: Track the nodes the process was running on when a NUMA 1268 * hinting fault was incurred. 1269 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1270 * during the current scan window. When the scan completes, the counts 1271 * in faults_memory and faults_cpu decay and these values are copied. 1272 */ 1273 unsigned long *numa_faults; 1274 unsigned long total_numa_faults; 1275 1276 /* 1277 * numa_faults_locality tracks if faults recorded during the last 1278 * scan window were remote/local or failed to migrate. The task scan 1279 * period is adapted based on the locality of the faults with different 1280 * weights depending on whether they were shared or private faults 1281 */ 1282 unsigned long numa_faults_locality[3]; 1283 1284 unsigned long numa_pages_migrated; 1285 #endif /* CONFIG_NUMA_BALANCING */ 1286 1287 #ifdef CONFIG_RSEQ 1288 struct rseq __user *rseq; 1289 u32 rseq_sig; 1290 /* 1291 * RmW on rseq_event_mask must be performed atomically 1292 * with respect to preemption. 1293 */ 1294 unsigned long rseq_event_mask; 1295 #endif 1296 1297 struct tlbflush_unmap_batch tlb_ubc; 1298 1299 union { 1300 refcount_t rcu_users; 1301 struct rcu_head rcu; 1302 }; 1303 1304 /* Cache last used pipe for splice(): */ 1305 struct pipe_inode_info *splice_pipe; 1306 1307 struct page_frag task_frag; 1308 1309 #ifdef CONFIG_TASK_DELAY_ACCT 1310 struct task_delay_info *delays; 1311 #endif 1312 1313 #ifdef CONFIG_FAULT_INJECTION 1314 int make_it_fail; 1315 unsigned int fail_nth; 1316 #endif 1317 /* 1318 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1319 * balance_dirty_pages() for a dirty throttling pause: 1320 */ 1321 int nr_dirtied; 1322 int nr_dirtied_pause; 1323 /* Start of a write-and-pause period: */ 1324 unsigned long dirty_paused_when; 1325 1326 #ifdef CONFIG_LATENCYTOP 1327 int latency_record_count; 1328 struct latency_record latency_record[LT_SAVECOUNT]; 1329 #endif 1330 /* 1331 * Time slack values; these are used to round up poll() and 1332 * select() etc timeout values. These are in nanoseconds. 1333 */ 1334 u64 timer_slack_ns; 1335 u64 default_timer_slack_ns; 1336 1337 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1338 unsigned int kasan_depth; 1339 #endif 1340 1341 #ifdef CONFIG_KCSAN 1342 struct kcsan_ctx kcsan_ctx; 1343 #ifdef CONFIG_TRACE_IRQFLAGS 1344 struct irqtrace_events kcsan_save_irqtrace; 1345 #endif 1346 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1347 int kcsan_stack_depth; 1348 #endif 1349 #endif 1350 1351 #if IS_ENABLED(CONFIG_KUNIT) 1352 struct kunit *kunit_test; 1353 #endif 1354 1355 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1356 /* Index of current stored address in ret_stack: */ 1357 int curr_ret_stack; 1358 int curr_ret_depth; 1359 1360 /* Stack of return addresses for return function tracing: */ 1361 struct ftrace_ret_stack *ret_stack; 1362 1363 /* Timestamp for last schedule: */ 1364 unsigned long long ftrace_timestamp; 1365 1366 /* 1367 * Number of functions that haven't been traced 1368 * because of depth overrun: 1369 */ 1370 atomic_t trace_overrun; 1371 1372 /* Pause tracing: */ 1373 atomic_t tracing_graph_pause; 1374 #endif 1375 1376 #ifdef CONFIG_TRACING 1377 /* State flags for use by tracers: */ 1378 unsigned long trace; 1379 1380 /* Bitmask and counter of trace recursion: */ 1381 unsigned long trace_recursion; 1382 #endif /* CONFIG_TRACING */ 1383 1384 #ifdef CONFIG_KCOV 1385 /* See kernel/kcov.c for more details. */ 1386 1387 /* Coverage collection mode enabled for this task (0 if disabled): */ 1388 unsigned int kcov_mode; 1389 1390 /* Size of the kcov_area: */ 1391 unsigned int kcov_size; 1392 1393 /* Buffer for coverage collection: */ 1394 void *kcov_area; 1395 1396 /* KCOV descriptor wired with this task or NULL: */ 1397 struct kcov *kcov; 1398 1399 /* KCOV common handle for remote coverage collection: */ 1400 u64 kcov_handle; 1401 1402 /* KCOV sequence number: */ 1403 int kcov_sequence; 1404 1405 /* Collect coverage from softirq context: */ 1406 unsigned int kcov_softirq; 1407 #endif 1408 1409 #ifdef CONFIG_MEMCG 1410 struct mem_cgroup *memcg_in_oom; 1411 gfp_t memcg_oom_gfp_mask; 1412 int memcg_oom_order; 1413 1414 /* Number of pages to reclaim on returning to userland: */ 1415 unsigned int memcg_nr_pages_over_high; 1416 1417 /* Used by memcontrol for targeted memcg charge: */ 1418 struct mem_cgroup *active_memcg; 1419 #endif 1420 1421 #ifdef CONFIG_BLK_CGROUP 1422 struct request_queue *throttle_queue; 1423 #endif 1424 1425 #ifdef CONFIG_UPROBES 1426 struct uprobe_task *utask; 1427 #endif 1428 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1429 unsigned int sequential_io; 1430 unsigned int sequential_io_avg; 1431 #endif 1432 struct kmap_ctrl kmap_ctrl; 1433 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1434 unsigned long task_state_change; 1435 # ifdef CONFIG_PREEMPT_RT 1436 unsigned long saved_state_change; 1437 # endif 1438 #endif 1439 int pagefault_disabled; 1440 #ifdef CONFIG_MMU 1441 struct task_struct *oom_reaper_list; 1442 #endif 1443 #ifdef CONFIG_VMAP_STACK 1444 struct vm_struct *stack_vm_area; 1445 #endif 1446 #ifdef CONFIG_THREAD_INFO_IN_TASK 1447 /* A live task holds one reference: */ 1448 refcount_t stack_refcount; 1449 #endif 1450 #ifdef CONFIG_LIVEPATCH 1451 int patch_state; 1452 #endif 1453 #ifdef CONFIG_SECURITY 1454 /* Used by LSM modules for access restriction: */ 1455 void *security; 1456 #endif 1457 #ifdef CONFIG_BPF_SYSCALL 1458 /* Used by BPF task local storage */ 1459 struct bpf_local_storage __rcu *bpf_storage; 1460 /* Used for BPF run context */ 1461 struct bpf_run_ctx *bpf_ctx; 1462 #endif 1463 1464 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1465 unsigned long lowest_stack; 1466 unsigned long prev_lowest_stack; 1467 #endif 1468 1469 #ifdef CONFIG_X86_MCE 1470 void __user *mce_vaddr; 1471 __u64 mce_kflags; 1472 u64 mce_addr; 1473 __u64 mce_ripv : 1, 1474 mce_whole_page : 1, 1475 __mce_reserved : 62; 1476 struct callback_head mce_kill_me; 1477 int mce_count; 1478 #endif 1479 1480 #ifdef CONFIG_KRETPROBES 1481 struct llist_head kretprobe_instances; 1482 #endif 1483 1484 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1485 /* 1486 * If L1D flush is supported on mm context switch 1487 * then we use this callback head to queue kill work 1488 * to kill tasks that are not running on SMT disabled 1489 * cores 1490 */ 1491 struct callback_head l1d_flush_kill; 1492 #endif 1493 1494 /* 1495 * New fields for task_struct should be added above here, so that 1496 * they are included in the randomized portion of task_struct. 1497 */ 1498 randomized_struct_fields_end 1499 1500 /* CPU-specific state of this task: */ 1501 struct thread_struct thread; 1502 1503 /* 1504 * WARNING: on x86, 'thread_struct' contains a variable-sized 1505 * structure. It *MUST* be at the end of 'task_struct'. 1506 * 1507 * Do not put anything below here! 1508 */ 1509 }; 1510 1511 static inline struct pid *task_pid(struct task_struct *task) 1512 { 1513 return task->thread_pid; 1514 } 1515 1516 /* 1517 * the helpers to get the task's different pids as they are seen 1518 * from various namespaces 1519 * 1520 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1521 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1522 * current. 1523 * task_xid_nr_ns() : id seen from the ns specified; 1524 * 1525 * see also pid_nr() etc in include/linux/pid.h 1526 */ 1527 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1528 1529 static inline pid_t task_pid_nr(struct task_struct *tsk) 1530 { 1531 return tsk->pid; 1532 } 1533 1534 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1535 { 1536 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1537 } 1538 1539 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1540 { 1541 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1542 } 1543 1544 1545 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1546 { 1547 return tsk->tgid; 1548 } 1549 1550 /** 1551 * pid_alive - check that a task structure is not stale 1552 * @p: Task structure to be checked. 1553 * 1554 * Test if a process is not yet dead (at most zombie state) 1555 * If pid_alive fails, then pointers within the task structure 1556 * can be stale and must not be dereferenced. 1557 * 1558 * Return: 1 if the process is alive. 0 otherwise. 1559 */ 1560 static inline int pid_alive(const struct task_struct *p) 1561 { 1562 return p->thread_pid != NULL; 1563 } 1564 1565 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1566 { 1567 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1568 } 1569 1570 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1571 { 1572 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1573 } 1574 1575 1576 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1577 { 1578 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1579 } 1580 1581 static inline pid_t task_session_vnr(struct task_struct *tsk) 1582 { 1583 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1584 } 1585 1586 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1587 { 1588 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns); 1589 } 1590 1591 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1592 { 1593 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL); 1594 } 1595 1596 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1597 { 1598 pid_t pid = 0; 1599 1600 rcu_read_lock(); 1601 if (pid_alive(tsk)) 1602 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1603 rcu_read_unlock(); 1604 1605 return pid; 1606 } 1607 1608 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1609 { 1610 return task_ppid_nr_ns(tsk, &init_pid_ns); 1611 } 1612 1613 /* Obsolete, do not use: */ 1614 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1615 { 1616 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1617 } 1618 1619 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1620 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1621 1622 static inline unsigned int task_state_index(struct task_struct *tsk) 1623 { 1624 unsigned int tsk_state = READ_ONCE(tsk->__state); 1625 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT; 1626 1627 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1628 1629 if (tsk_state == TASK_IDLE) 1630 state = TASK_REPORT_IDLE; 1631 1632 return fls(state); 1633 } 1634 1635 static inline char task_index_to_char(unsigned int state) 1636 { 1637 static const char state_char[] = "RSDTtXZPI"; 1638 1639 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1640 1641 return state_char[state]; 1642 } 1643 1644 static inline char task_state_to_char(struct task_struct *tsk) 1645 { 1646 return task_index_to_char(task_state_index(tsk)); 1647 } 1648 1649 /** 1650 * is_global_init - check if a task structure is init. Since init 1651 * is free to have sub-threads we need to check tgid. 1652 * @tsk: Task structure to be checked. 1653 * 1654 * Check if a task structure is the first user space task the kernel created. 1655 * 1656 * Return: 1 if the task structure is init. 0 otherwise. 1657 */ 1658 static inline int is_global_init(struct task_struct *tsk) 1659 { 1660 return task_tgid_nr(tsk) == 1; 1661 } 1662 1663 extern struct pid *cad_pid; 1664 1665 /* 1666 * Per process flags 1667 */ 1668 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1669 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1670 #define PF_EXITING 0x00000004 /* Getting shut down */ 1671 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1672 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1673 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1674 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1675 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1676 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1677 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1678 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1679 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1680 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1681 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1682 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1683 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1684 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1685 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1686 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1687 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1688 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1689 * I am cleaning dirty pages from some other bdi. */ 1690 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1691 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1692 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1693 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1694 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1695 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */ 1696 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1697 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1698 1699 /* 1700 * Only the _current_ task can read/write to tsk->flags, but other 1701 * tasks can access tsk->flags in readonly mode for example 1702 * with tsk_used_math (like during threaded core dumping). 1703 * There is however an exception to this rule during ptrace 1704 * or during fork: the ptracer task is allowed to write to the 1705 * child->flags of its traced child (same goes for fork, the parent 1706 * can write to the child->flags), because we're guaranteed the 1707 * child is not running and in turn not changing child->flags 1708 * at the same time the parent does it. 1709 */ 1710 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1711 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1712 #define clear_used_math() clear_stopped_child_used_math(current) 1713 #define set_used_math() set_stopped_child_used_math(current) 1714 1715 #define conditional_stopped_child_used_math(condition, child) \ 1716 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1717 1718 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1719 1720 #define copy_to_stopped_child_used_math(child) \ 1721 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1722 1723 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1724 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1725 #define used_math() tsk_used_math(current) 1726 1727 static __always_inline bool is_percpu_thread(void) 1728 { 1729 #ifdef CONFIG_SMP 1730 return (current->flags & PF_NO_SETAFFINITY) && 1731 (current->nr_cpus_allowed == 1); 1732 #else 1733 return true; 1734 #endif 1735 } 1736 1737 /* Per-process atomic flags. */ 1738 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1739 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1740 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1741 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1742 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1743 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1744 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1745 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1746 1747 #define TASK_PFA_TEST(name, func) \ 1748 static inline bool task_##func(struct task_struct *p) \ 1749 { return test_bit(PFA_##name, &p->atomic_flags); } 1750 1751 #define TASK_PFA_SET(name, func) \ 1752 static inline void task_set_##func(struct task_struct *p) \ 1753 { set_bit(PFA_##name, &p->atomic_flags); } 1754 1755 #define TASK_PFA_CLEAR(name, func) \ 1756 static inline void task_clear_##func(struct task_struct *p) \ 1757 { clear_bit(PFA_##name, &p->atomic_flags); } 1758 1759 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1760 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1761 1762 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1763 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1764 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1765 1766 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1767 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1768 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1769 1770 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1771 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1772 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1773 1774 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1775 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1776 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1777 1778 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1779 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1780 1781 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1782 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1783 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1784 1785 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1786 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1787 1788 static inline void 1789 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1790 { 1791 current->flags &= ~flags; 1792 current->flags |= orig_flags & flags; 1793 } 1794 1795 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1796 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1797 #ifdef CONFIG_SMP 1798 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1799 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1800 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1801 extern void release_user_cpus_ptr(struct task_struct *p); 1802 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1803 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1804 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1805 #else 1806 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1807 { 1808 } 1809 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1810 { 1811 if (!cpumask_test_cpu(0, new_mask)) 1812 return -EINVAL; 1813 return 0; 1814 } 1815 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node) 1816 { 1817 if (src->user_cpus_ptr) 1818 return -EINVAL; 1819 return 0; 1820 } 1821 static inline void release_user_cpus_ptr(struct task_struct *p) 1822 { 1823 WARN_ON(p->user_cpus_ptr); 1824 } 1825 1826 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) 1827 { 1828 return 0; 1829 } 1830 #endif 1831 1832 extern int yield_to(struct task_struct *p, bool preempt); 1833 extern void set_user_nice(struct task_struct *p, long nice); 1834 extern int task_prio(const struct task_struct *p); 1835 1836 /** 1837 * task_nice - return the nice value of a given task. 1838 * @p: the task in question. 1839 * 1840 * Return: The nice value [ -20 ... 0 ... 19 ]. 1841 */ 1842 static inline int task_nice(const struct task_struct *p) 1843 { 1844 return PRIO_TO_NICE((p)->static_prio); 1845 } 1846 1847 extern int can_nice(const struct task_struct *p, const int nice); 1848 extern int task_curr(const struct task_struct *p); 1849 extern int idle_cpu(int cpu); 1850 extern int available_idle_cpu(int cpu); 1851 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1852 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1853 extern void sched_set_fifo(struct task_struct *p); 1854 extern void sched_set_fifo_low(struct task_struct *p); 1855 extern void sched_set_normal(struct task_struct *p, int nice); 1856 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1857 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1858 extern struct task_struct *idle_task(int cpu); 1859 1860 /** 1861 * is_idle_task - is the specified task an idle task? 1862 * @p: the task in question. 1863 * 1864 * Return: 1 if @p is an idle task. 0 otherwise. 1865 */ 1866 static __always_inline bool is_idle_task(const struct task_struct *p) 1867 { 1868 return !!(p->flags & PF_IDLE); 1869 } 1870 1871 extern struct task_struct *curr_task(int cpu); 1872 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1873 1874 void yield(void); 1875 1876 union thread_union { 1877 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1878 struct task_struct task; 1879 #endif 1880 #ifndef CONFIG_THREAD_INFO_IN_TASK 1881 struct thread_info thread_info; 1882 #endif 1883 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1884 }; 1885 1886 #ifndef CONFIG_THREAD_INFO_IN_TASK 1887 extern struct thread_info init_thread_info; 1888 #endif 1889 1890 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1891 1892 #ifdef CONFIG_THREAD_INFO_IN_TASK 1893 # define task_thread_info(task) (&(task)->thread_info) 1894 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1895 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1896 #endif 1897 1898 /* 1899 * find a task by one of its numerical ids 1900 * 1901 * find_task_by_pid_ns(): 1902 * finds a task by its pid in the specified namespace 1903 * find_task_by_vpid(): 1904 * finds a task by its virtual pid 1905 * 1906 * see also find_vpid() etc in include/linux/pid.h 1907 */ 1908 1909 extern struct task_struct *find_task_by_vpid(pid_t nr); 1910 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1911 1912 /* 1913 * find a task by its virtual pid and get the task struct 1914 */ 1915 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1916 1917 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1918 extern int wake_up_process(struct task_struct *tsk); 1919 extern void wake_up_new_task(struct task_struct *tsk); 1920 1921 #ifdef CONFIG_SMP 1922 extern void kick_process(struct task_struct *tsk); 1923 #else 1924 static inline void kick_process(struct task_struct *tsk) { } 1925 #endif 1926 1927 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1928 1929 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1930 { 1931 __set_task_comm(tsk, from, false); 1932 } 1933 1934 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1935 #define get_task_comm(buf, tsk) ({ \ 1936 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1937 __get_task_comm(buf, sizeof(buf), tsk); \ 1938 }) 1939 1940 #ifdef CONFIG_SMP 1941 static __always_inline void scheduler_ipi(void) 1942 { 1943 /* 1944 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1945 * TIF_NEED_RESCHED remotely (for the first time) will also send 1946 * this IPI. 1947 */ 1948 preempt_fold_need_resched(); 1949 } 1950 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 1951 #else 1952 static inline void scheduler_ipi(void) { } 1953 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state) 1954 { 1955 return 1; 1956 } 1957 #endif 1958 1959 /* 1960 * Set thread flags in other task's structures. 1961 * See asm/thread_info.h for TIF_xxxx flags available: 1962 */ 1963 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1964 { 1965 set_ti_thread_flag(task_thread_info(tsk), flag); 1966 } 1967 1968 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1969 { 1970 clear_ti_thread_flag(task_thread_info(tsk), flag); 1971 } 1972 1973 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 1974 bool value) 1975 { 1976 update_ti_thread_flag(task_thread_info(tsk), flag, value); 1977 } 1978 1979 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1980 { 1981 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1982 } 1983 1984 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1985 { 1986 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1987 } 1988 1989 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1990 { 1991 return test_ti_thread_flag(task_thread_info(tsk), flag); 1992 } 1993 1994 static inline void set_tsk_need_resched(struct task_struct *tsk) 1995 { 1996 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1997 } 1998 1999 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2000 { 2001 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2002 } 2003 2004 static inline int test_tsk_need_resched(struct task_struct *tsk) 2005 { 2006 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2007 } 2008 2009 /* 2010 * cond_resched() and cond_resched_lock(): latency reduction via 2011 * explicit rescheduling in places that are safe. The return 2012 * value indicates whether a reschedule was done in fact. 2013 * cond_resched_lock() will drop the spinlock before scheduling, 2014 */ 2015 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2016 extern int __cond_resched(void); 2017 2018 #ifdef CONFIG_PREEMPT_DYNAMIC 2019 2020 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2021 2022 static __always_inline int _cond_resched(void) 2023 { 2024 return static_call_mod(cond_resched)(); 2025 } 2026 2027 #else 2028 2029 static inline int _cond_resched(void) 2030 { 2031 return __cond_resched(); 2032 } 2033 2034 #endif /* CONFIG_PREEMPT_DYNAMIC */ 2035 2036 #else 2037 2038 static inline int _cond_resched(void) { return 0; } 2039 2040 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */ 2041 2042 #define cond_resched() ({ \ 2043 __might_resched(__FILE__, __LINE__, 0); \ 2044 _cond_resched(); \ 2045 }) 2046 2047 extern int __cond_resched_lock(spinlock_t *lock); 2048 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2049 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2050 2051 #define MIGHT_RESCHED_RCU_SHIFT 8 2052 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2053 2054 #ifndef CONFIG_PREEMPT_RT 2055 /* 2056 * Non RT kernels have an elevated preempt count due to the held lock, 2057 * but are not allowed to be inside a RCU read side critical section 2058 */ 2059 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2060 #else 2061 /* 2062 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2063 * cond_resched*lock() has to take that into account because it checks for 2064 * preempt_count() and rcu_preempt_depth(). 2065 */ 2066 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2067 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2068 #endif 2069 2070 #define cond_resched_lock(lock) ({ \ 2071 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2072 __cond_resched_lock(lock); \ 2073 }) 2074 2075 #define cond_resched_rwlock_read(lock) ({ \ 2076 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2077 __cond_resched_rwlock_read(lock); \ 2078 }) 2079 2080 #define cond_resched_rwlock_write(lock) ({ \ 2081 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2082 __cond_resched_rwlock_write(lock); \ 2083 }) 2084 2085 static inline void cond_resched_rcu(void) 2086 { 2087 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 2088 rcu_read_unlock(); 2089 cond_resched(); 2090 rcu_read_lock(); 2091 #endif 2092 } 2093 2094 /* 2095 * Does a critical section need to be broken due to another 2096 * task waiting?: (technically does not depend on CONFIG_PREEMPTION, 2097 * but a general need for low latency) 2098 */ 2099 static inline int spin_needbreak(spinlock_t *lock) 2100 { 2101 #ifdef CONFIG_PREEMPTION 2102 return spin_is_contended(lock); 2103 #else 2104 return 0; 2105 #endif 2106 } 2107 2108 /* 2109 * Check if a rwlock is contended. 2110 * Returns non-zero if there is another task waiting on the rwlock. 2111 * Returns zero if the lock is not contended or the system / underlying 2112 * rwlock implementation does not support contention detection. 2113 * Technically does not depend on CONFIG_PREEMPTION, but a general need 2114 * for low latency. 2115 */ 2116 static inline int rwlock_needbreak(rwlock_t *lock) 2117 { 2118 #ifdef CONFIG_PREEMPTION 2119 return rwlock_is_contended(lock); 2120 #else 2121 return 0; 2122 #endif 2123 } 2124 2125 static __always_inline bool need_resched(void) 2126 { 2127 return unlikely(tif_need_resched()); 2128 } 2129 2130 /* 2131 * Wrappers for p->thread_info->cpu access. No-op on UP. 2132 */ 2133 #ifdef CONFIG_SMP 2134 2135 static inline unsigned int task_cpu(const struct task_struct *p) 2136 { 2137 return READ_ONCE(task_thread_info(p)->cpu); 2138 } 2139 2140 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2141 2142 #else 2143 2144 static inline unsigned int task_cpu(const struct task_struct *p) 2145 { 2146 return 0; 2147 } 2148 2149 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2150 { 2151 } 2152 2153 #endif /* CONFIG_SMP */ 2154 2155 extern bool sched_task_on_rq(struct task_struct *p); 2156 extern unsigned long get_wchan(struct task_struct *p); 2157 2158 /* 2159 * In order to reduce various lock holder preemption latencies provide an 2160 * interface to see if a vCPU is currently running or not. 2161 * 2162 * This allows us to terminate optimistic spin loops and block, analogous to 2163 * the native optimistic spin heuristic of testing if the lock owner task is 2164 * running or not. 2165 */ 2166 #ifndef vcpu_is_preempted 2167 static inline bool vcpu_is_preempted(int cpu) 2168 { 2169 return false; 2170 } 2171 #endif 2172 2173 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2174 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2175 2176 #ifndef TASK_SIZE_OF 2177 #define TASK_SIZE_OF(tsk) TASK_SIZE 2178 #endif 2179 2180 #ifdef CONFIG_SMP 2181 static inline bool owner_on_cpu(struct task_struct *owner) 2182 { 2183 /* 2184 * As lock holder preemption issue, we both skip spinning if 2185 * task is not on cpu or its cpu is preempted 2186 */ 2187 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2188 } 2189 2190 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2191 unsigned long sched_cpu_util(int cpu, unsigned long max); 2192 #endif /* CONFIG_SMP */ 2193 2194 #ifdef CONFIG_RSEQ 2195 2196 /* 2197 * Map the event mask on the user-space ABI enum rseq_cs_flags 2198 * for direct mask checks. 2199 */ 2200 enum rseq_event_mask_bits { 2201 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT, 2202 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT, 2203 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT, 2204 }; 2205 2206 enum rseq_event_mask { 2207 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT), 2208 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT), 2209 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT), 2210 }; 2211 2212 static inline void rseq_set_notify_resume(struct task_struct *t) 2213 { 2214 if (t->rseq) 2215 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME); 2216 } 2217 2218 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs); 2219 2220 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2221 struct pt_regs *regs) 2222 { 2223 if (current->rseq) 2224 __rseq_handle_notify_resume(ksig, regs); 2225 } 2226 2227 static inline void rseq_signal_deliver(struct ksignal *ksig, 2228 struct pt_regs *regs) 2229 { 2230 preempt_disable(); 2231 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask); 2232 preempt_enable(); 2233 rseq_handle_notify_resume(ksig, regs); 2234 } 2235 2236 /* rseq_preempt() requires preemption to be disabled. */ 2237 static inline void rseq_preempt(struct task_struct *t) 2238 { 2239 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask); 2240 rseq_set_notify_resume(t); 2241 } 2242 2243 /* rseq_migrate() requires preemption to be disabled. */ 2244 static inline void rseq_migrate(struct task_struct *t) 2245 { 2246 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask); 2247 rseq_set_notify_resume(t); 2248 } 2249 2250 /* 2251 * If parent process has a registered restartable sequences area, the 2252 * child inherits. Unregister rseq for a clone with CLONE_VM set. 2253 */ 2254 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2255 { 2256 if (clone_flags & CLONE_VM) { 2257 t->rseq = NULL; 2258 t->rseq_sig = 0; 2259 t->rseq_event_mask = 0; 2260 } else { 2261 t->rseq = current->rseq; 2262 t->rseq_sig = current->rseq_sig; 2263 t->rseq_event_mask = current->rseq_event_mask; 2264 } 2265 } 2266 2267 static inline void rseq_execve(struct task_struct *t) 2268 { 2269 t->rseq = NULL; 2270 t->rseq_sig = 0; 2271 t->rseq_event_mask = 0; 2272 } 2273 2274 #else 2275 2276 static inline void rseq_set_notify_resume(struct task_struct *t) 2277 { 2278 } 2279 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2280 struct pt_regs *regs) 2281 { 2282 } 2283 static inline void rseq_signal_deliver(struct ksignal *ksig, 2284 struct pt_regs *regs) 2285 { 2286 } 2287 static inline void rseq_preempt(struct task_struct *t) 2288 { 2289 } 2290 static inline void rseq_migrate(struct task_struct *t) 2291 { 2292 } 2293 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2294 { 2295 } 2296 static inline void rseq_execve(struct task_struct *t) 2297 { 2298 } 2299 2300 #endif 2301 2302 #ifdef CONFIG_DEBUG_RSEQ 2303 2304 void rseq_syscall(struct pt_regs *regs); 2305 2306 #else 2307 2308 static inline void rseq_syscall(struct pt_regs *regs) 2309 { 2310 } 2311 2312 #endif 2313 2314 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq); 2315 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len); 2316 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq); 2317 2318 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq); 2319 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq); 2320 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq); 2321 2322 int sched_trace_rq_cpu(struct rq *rq); 2323 int sched_trace_rq_cpu_capacity(struct rq *rq); 2324 int sched_trace_rq_nr_running(struct rq *rq); 2325 2326 const struct cpumask *sched_trace_rd_span(struct root_domain *rd); 2327 2328 #ifdef CONFIG_SCHED_CORE 2329 extern void sched_core_free(struct task_struct *tsk); 2330 extern void sched_core_fork(struct task_struct *p); 2331 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2332 unsigned long uaddr); 2333 #else 2334 static inline void sched_core_free(struct task_struct *tsk) { } 2335 static inline void sched_core_fork(struct task_struct *p) { } 2336 #endif 2337 2338 #endif 2339