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 <linux/rv.h> 38 #include <asm/kmap_size.h> 39 40 /* task_struct member predeclarations (sorted alphabetically): */ 41 struct audit_context; 42 struct backing_dev_info; 43 struct bio_list; 44 struct blk_plug; 45 struct bpf_local_storage; 46 struct bpf_run_ctx; 47 struct capture_control; 48 struct cfs_rq; 49 struct fs_struct; 50 struct futex_pi_state; 51 struct io_context; 52 struct io_uring_task; 53 struct mempolicy; 54 struct nameidata; 55 struct nsproxy; 56 struct perf_event_context; 57 struct pid_namespace; 58 struct pipe_inode_info; 59 struct rcu_node; 60 struct reclaim_state; 61 struct robust_list_head; 62 struct root_domain; 63 struct rq; 64 struct sched_attr; 65 struct sched_param; 66 struct seq_file; 67 struct sighand_struct; 68 struct signal_struct; 69 struct task_delay_info; 70 struct task_group; 71 72 /* 73 * Task state bitmask. NOTE! These bits are also 74 * encoded in fs/proc/array.c: get_task_state(). 75 * 76 * We have two separate sets of flags: task->state 77 * is about runnability, while task->exit_state are 78 * about the task exiting. Confusing, but this way 79 * modifying one set can't modify the other one by 80 * mistake. 81 */ 82 83 /* Used in tsk->state: */ 84 #define TASK_RUNNING 0x0000 85 #define TASK_INTERRUPTIBLE 0x0001 86 #define TASK_UNINTERRUPTIBLE 0x0002 87 #define __TASK_STOPPED 0x0004 88 #define __TASK_TRACED 0x0008 89 /* Used in tsk->exit_state: */ 90 #define EXIT_DEAD 0x0010 91 #define EXIT_ZOMBIE 0x0020 92 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 93 /* Used in tsk->state again: */ 94 #define TASK_PARKED 0x0040 95 #define TASK_DEAD 0x0080 96 #define TASK_WAKEKILL 0x0100 97 #define TASK_WAKING 0x0200 98 #define TASK_NOLOAD 0x0400 99 #define TASK_NEW 0x0800 100 /* RT specific auxilliary flag to mark RT lock waiters */ 101 #define TASK_RTLOCK_WAIT 0x1000 102 #define TASK_STATE_MAX 0x2000 103 104 /* Convenience macros for the sake of set_current_state: */ 105 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 106 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 107 #define TASK_TRACED __TASK_TRACED 108 109 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 110 111 /* Convenience macros for the sake of wake_up(): */ 112 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 113 114 /* get_task_state(): */ 115 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 116 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 117 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 118 TASK_PARKED) 119 120 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING) 121 122 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0) 123 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0) 124 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0) 125 126 /* 127 * Special states are those that do not use the normal wait-loop pattern. See 128 * the comment with set_special_state(). 129 */ 130 #define is_special_task_state(state) \ 131 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD)) 132 133 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 134 # define debug_normal_state_change(state_value) \ 135 do { \ 136 WARN_ON_ONCE(is_special_task_state(state_value)); \ 137 current->task_state_change = _THIS_IP_; \ 138 } while (0) 139 140 # define debug_special_state_change(state_value) \ 141 do { \ 142 WARN_ON_ONCE(!is_special_task_state(state_value)); \ 143 current->task_state_change = _THIS_IP_; \ 144 } while (0) 145 146 # define debug_rtlock_wait_set_state() \ 147 do { \ 148 current->saved_state_change = current->task_state_change;\ 149 current->task_state_change = _THIS_IP_; \ 150 } while (0) 151 152 # define debug_rtlock_wait_restore_state() \ 153 do { \ 154 current->task_state_change = current->saved_state_change;\ 155 } while (0) 156 157 #else 158 # define debug_normal_state_change(cond) do { } while (0) 159 # define debug_special_state_change(cond) do { } while (0) 160 # define debug_rtlock_wait_set_state() do { } while (0) 161 # define debug_rtlock_wait_restore_state() do { } while (0) 162 #endif 163 164 /* 165 * set_current_state() includes a barrier so that the write of current->state 166 * is correctly serialised wrt the caller's subsequent test of whether to 167 * actually sleep: 168 * 169 * for (;;) { 170 * set_current_state(TASK_UNINTERRUPTIBLE); 171 * if (CONDITION) 172 * break; 173 * 174 * schedule(); 175 * } 176 * __set_current_state(TASK_RUNNING); 177 * 178 * If the caller does not need such serialisation (because, for instance, the 179 * CONDITION test and condition change and wakeup are under the same lock) then 180 * use __set_current_state(). 181 * 182 * The above is typically ordered against the wakeup, which does: 183 * 184 * CONDITION = 1; 185 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 186 * 187 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before 188 * accessing p->state. 189 * 190 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 191 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 192 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 193 * 194 * However, with slightly different timing the wakeup TASK_RUNNING store can 195 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not 196 * a problem either because that will result in one extra go around the loop 197 * and our @cond test will save the day. 198 * 199 * Also see the comments of try_to_wake_up(). 200 */ 201 #define __set_current_state(state_value) \ 202 do { \ 203 debug_normal_state_change((state_value)); \ 204 WRITE_ONCE(current->__state, (state_value)); \ 205 } while (0) 206 207 #define set_current_state(state_value) \ 208 do { \ 209 debug_normal_state_change((state_value)); \ 210 smp_store_mb(current->__state, (state_value)); \ 211 } while (0) 212 213 /* 214 * set_special_state() should be used for those states when the blocking task 215 * can not use the regular condition based wait-loop. In that case we must 216 * serialize against wakeups such that any possible in-flight TASK_RUNNING 217 * stores will not collide with our state change. 218 */ 219 #define set_special_state(state_value) \ 220 do { \ 221 unsigned long flags; /* may shadow */ \ 222 \ 223 raw_spin_lock_irqsave(¤t->pi_lock, flags); \ 224 debug_special_state_change((state_value)); \ 225 WRITE_ONCE(current->__state, (state_value)); \ 226 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \ 227 } while (0) 228 229 /* 230 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks 231 * 232 * RT's spin/rwlock substitutions are state preserving. The state of the 233 * task when blocking on the lock is saved in task_struct::saved_state and 234 * restored after the lock has been acquired. These operations are 235 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT 236 * lock related wakeups while the task is blocked on the lock are 237 * redirected to operate on task_struct::saved_state to ensure that these 238 * are not dropped. On restore task_struct::saved_state is set to 239 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail. 240 * 241 * The lock operation looks like this: 242 * 243 * current_save_and_set_rtlock_wait_state(); 244 * for (;;) { 245 * if (try_lock()) 246 * break; 247 * raw_spin_unlock_irq(&lock->wait_lock); 248 * schedule_rtlock(); 249 * raw_spin_lock_irq(&lock->wait_lock); 250 * set_current_state(TASK_RTLOCK_WAIT); 251 * } 252 * current_restore_rtlock_saved_state(); 253 */ 254 #define current_save_and_set_rtlock_wait_state() \ 255 do { \ 256 lockdep_assert_irqs_disabled(); \ 257 raw_spin_lock(¤t->pi_lock); \ 258 current->saved_state = current->__state; \ 259 debug_rtlock_wait_set_state(); \ 260 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \ 261 raw_spin_unlock(¤t->pi_lock); \ 262 } while (0); 263 264 #define current_restore_rtlock_saved_state() \ 265 do { \ 266 lockdep_assert_irqs_disabled(); \ 267 raw_spin_lock(¤t->pi_lock); \ 268 debug_rtlock_wait_restore_state(); \ 269 WRITE_ONCE(current->__state, current->saved_state); \ 270 current->saved_state = TASK_RUNNING; \ 271 raw_spin_unlock(¤t->pi_lock); \ 272 } while (0); 273 274 #define get_current_state() READ_ONCE(current->__state) 275 276 /* 277 * Define the task command name length as enum, then it can be visible to 278 * BPF programs. 279 */ 280 enum { 281 TASK_COMM_LEN = 16, 282 }; 283 284 extern void scheduler_tick(void); 285 286 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 287 288 extern long schedule_timeout(long timeout); 289 extern long schedule_timeout_interruptible(long timeout); 290 extern long schedule_timeout_killable(long timeout); 291 extern long schedule_timeout_uninterruptible(long timeout); 292 extern long schedule_timeout_idle(long timeout); 293 asmlinkage void schedule(void); 294 extern void schedule_preempt_disabled(void); 295 asmlinkage void preempt_schedule_irq(void); 296 #ifdef CONFIG_PREEMPT_RT 297 extern void schedule_rtlock(void); 298 #endif 299 300 extern int __must_check io_schedule_prepare(void); 301 extern void io_schedule_finish(int token); 302 extern long io_schedule_timeout(long timeout); 303 extern void io_schedule(void); 304 305 /** 306 * struct prev_cputime - snapshot of system and user cputime 307 * @utime: time spent in user mode 308 * @stime: time spent in system mode 309 * @lock: protects the above two fields 310 * 311 * Stores previous user/system time values such that we can guarantee 312 * monotonicity. 313 */ 314 struct prev_cputime { 315 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 316 u64 utime; 317 u64 stime; 318 raw_spinlock_t lock; 319 #endif 320 }; 321 322 enum vtime_state { 323 /* Task is sleeping or running in a CPU with VTIME inactive: */ 324 VTIME_INACTIVE = 0, 325 /* Task is idle */ 326 VTIME_IDLE, 327 /* Task runs in kernelspace in a CPU with VTIME active: */ 328 VTIME_SYS, 329 /* Task runs in userspace in a CPU with VTIME active: */ 330 VTIME_USER, 331 /* Task runs as guests in a CPU with VTIME active: */ 332 VTIME_GUEST, 333 }; 334 335 struct vtime { 336 seqcount_t seqcount; 337 unsigned long long starttime; 338 enum vtime_state state; 339 unsigned int cpu; 340 u64 utime; 341 u64 stime; 342 u64 gtime; 343 }; 344 345 /* 346 * Utilization clamp constraints. 347 * @UCLAMP_MIN: Minimum utilization 348 * @UCLAMP_MAX: Maximum utilization 349 * @UCLAMP_CNT: Utilization clamp constraints count 350 */ 351 enum uclamp_id { 352 UCLAMP_MIN = 0, 353 UCLAMP_MAX, 354 UCLAMP_CNT 355 }; 356 357 #ifdef CONFIG_SMP 358 extern struct root_domain def_root_domain; 359 extern struct mutex sched_domains_mutex; 360 #endif 361 362 struct sched_info { 363 #ifdef CONFIG_SCHED_INFO 364 /* Cumulative counters: */ 365 366 /* # of times we have run on this CPU: */ 367 unsigned long pcount; 368 369 /* Time spent waiting on a runqueue: */ 370 unsigned long long run_delay; 371 372 /* Timestamps: */ 373 374 /* When did we last run on a CPU? */ 375 unsigned long long last_arrival; 376 377 /* When were we last queued to run? */ 378 unsigned long long last_queued; 379 380 #endif /* CONFIG_SCHED_INFO */ 381 }; 382 383 /* 384 * Integer metrics need fixed point arithmetic, e.g., sched/fair 385 * has a few: load, load_avg, util_avg, freq, and capacity. 386 * 387 * We define a basic fixed point arithmetic range, and then formalize 388 * all these metrics based on that basic range. 389 */ 390 # define SCHED_FIXEDPOINT_SHIFT 10 391 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 392 393 /* Increase resolution of cpu_capacity calculations */ 394 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT 395 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT) 396 397 struct load_weight { 398 unsigned long weight; 399 u32 inv_weight; 400 }; 401 402 /** 403 * struct util_est - Estimation utilization of FAIR tasks 404 * @enqueued: instantaneous estimated utilization of a task/cpu 405 * @ewma: the Exponential Weighted Moving Average (EWMA) 406 * utilization of a task 407 * 408 * Support data structure to track an Exponential Weighted Moving Average 409 * (EWMA) of a FAIR task's utilization. New samples are added to the moving 410 * average each time a task completes an activation. Sample's weight is chosen 411 * so that the EWMA will be relatively insensitive to transient changes to the 412 * task's workload. 413 * 414 * The enqueued attribute has a slightly different meaning for tasks and cpus: 415 * - task: the task's util_avg at last task dequeue time 416 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU 417 * Thus, the util_est.enqueued of a task represents the contribution on the 418 * estimated utilization of the CPU where that task is currently enqueued. 419 * 420 * Only for tasks we track a moving average of the past instantaneous 421 * estimated utilization. This allows to absorb sporadic drops in utilization 422 * of an otherwise almost periodic task. 423 * 424 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg 425 * updates. When a task is dequeued, its util_est should not be updated if its 426 * util_avg has not been updated in the meantime. 427 * This information is mapped into the MSB bit of util_est.enqueued at dequeue 428 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg 429 * for a task) it is safe to use MSB. 430 */ 431 struct util_est { 432 unsigned int enqueued; 433 unsigned int ewma; 434 #define UTIL_EST_WEIGHT_SHIFT 2 435 #define UTIL_AVG_UNCHANGED 0x80000000 436 } __attribute__((__aligned__(sizeof(u64)))); 437 438 /* 439 * The load/runnable/util_avg accumulates an infinite geometric series 440 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c). 441 * 442 * [load_avg definition] 443 * 444 * load_avg = runnable% * scale_load_down(load) 445 * 446 * [runnable_avg definition] 447 * 448 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE 449 * 450 * [util_avg definition] 451 * 452 * util_avg = running% * SCHED_CAPACITY_SCALE 453 * 454 * where runnable% is the time ratio that a sched_entity is runnable and 455 * running% the time ratio that a sched_entity is running. 456 * 457 * For cfs_rq, they are the aggregated values of all runnable and blocked 458 * sched_entities. 459 * 460 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU 461 * capacity scaling. The scaling is done through the rq_clock_pelt that is used 462 * for computing those signals (see update_rq_clock_pelt()) 463 * 464 * N.B., the above ratios (runnable% and running%) themselves are in the 465 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them 466 * to as large a range as necessary. This is for example reflected by 467 * util_avg's SCHED_CAPACITY_SCALE. 468 * 469 * [Overflow issue] 470 * 471 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 472 * with the highest load (=88761), always runnable on a single cfs_rq, 473 * and should not overflow as the number already hits PID_MAX_LIMIT. 474 * 475 * For all other cases (including 32-bit kernels), struct load_weight's 476 * weight will overflow first before we do, because: 477 * 478 * Max(load_avg) <= Max(load.weight) 479 * 480 * Then it is the load_weight's responsibility to consider overflow 481 * issues. 482 */ 483 struct sched_avg { 484 u64 last_update_time; 485 u64 load_sum; 486 u64 runnable_sum; 487 u32 util_sum; 488 u32 period_contrib; 489 unsigned long load_avg; 490 unsigned long runnable_avg; 491 unsigned long util_avg; 492 struct util_est util_est; 493 } ____cacheline_aligned; 494 495 struct sched_statistics { 496 #ifdef CONFIG_SCHEDSTATS 497 u64 wait_start; 498 u64 wait_max; 499 u64 wait_count; 500 u64 wait_sum; 501 u64 iowait_count; 502 u64 iowait_sum; 503 504 u64 sleep_start; 505 u64 sleep_max; 506 s64 sum_sleep_runtime; 507 508 u64 block_start; 509 u64 block_max; 510 s64 sum_block_runtime; 511 512 u64 exec_max; 513 u64 slice_max; 514 515 u64 nr_migrations_cold; 516 u64 nr_failed_migrations_affine; 517 u64 nr_failed_migrations_running; 518 u64 nr_failed_migrations_hot; 519 u64 nr_forced_migrations; 520 521 u64 nr_wakeups; 522 u64 nr_wakeups_sync; 523 u64 nr_wakeups_migrate; 524 u64 nr_wakeups_local; 525 u64 nr_wakeups_remote; 526 u64 nr_wakeups_affine; 527 u64 nr_wakeups_affine_attempts; 528 u64 nr_wakeups_passive; 529 u64 nr_wakeups_idle; 530 531 #ifdef CONFIG_SCHED_CORE 532 u64 core_forceidle_sum; 533 #endif 534 #endif /* CONFIG_SCHEDSTATS */ 535 } ____cacheline_aligned; 536 537 struct sched_entity { 538 /* For load-balancing: */ 539 struct load_weight load; 540 struct rb_node run_node; 541 struct list_head group_node; 542 unsigned int on_rq; 543 544 u64 exec_start; 545 u64 sum_exec_runtime; 546 u64 vruntime; 547 u64 prev_sum_exec_runtime; 548 549 u64 nr_migrations; 550 551 #ifdef CONFIG_FAIR_GROUP_SCHED 552 int depth; 553 struct sched_entity *parent; 554 /* rq on which this entity is (to be) queued: */ 555 struct cfs_rq *cfs_rq; 556 /* rq "owned" by this entity/group: */ 557 struct cfs_rq *my_q; 558 /* cached value of my_q->h_nr_running */ 559 unsigned long runnable_weight; 560 #endif 561 562 #ifdef CONFIG_SMP 563 /* 564 * Per entity load average tracking. 565 * 566 * Put into separate cache line so it does not 567 * collide with read-mostly values above. 568 */ 569 struct sched_avg avg; 570 #endif 571 }; 572 573 struct sched_rt_entity { 574 struct list_head run_list; 575 unsigned long timeout; 576 unsigned long watchdog_stamp; 577 unsigned int time_slice; 578 unsigned short on_rq; 579 unsigned short on_list; 580 581 struct sched_rt_entity *back; 582 #ifdef CONFIG_RT_GROUP_SCHED 583 struct sched_rt_entity *parent; 584 /* rq on which this entity is (to be) queued: */ 585 struct rt_rq *rt_rq; 586 /* rq "owned" by this entity/group: */ 587 struct rt_rq *my_q; 588 #endif 589 } __randomize_layout; 590 591 struct sched_dl_entity { 592 struct rb_node rb_node; 593 594 /* 595 * Original scheduling parameters. Copied here from sched_attr 596 * during sched_setattr(), they will remain the same until 597 * the next sched_setattr(). 598 */ 599 u64 dl_runtime; /* Maximum runtime for each instance */ 600 u64 dl_deadline; /* Relative deadline of each instance */ 601 u64 dl_period; /* Separation of two instances (period) */ 602 u64 dl_bw; /* dl_runtime / dl_period */ 603 u64 dl_density; /* dl_runtime / dl_deadline */ 604 605 /* 606 * Actual scheduling parameters. Initialized with the values above, 607 * they are continuously updated during task execution. Note that 608 * the remaining runtime could be < 0 in case we are in overrun. 609 */ 610 s64 runtime; /* Remaining runtime for this instance */ 611 u64 deadline; /* Absolute deadline for this instance */ 612 unsigned int flags; /* Specifying the scheduler behaviour */ 613 614 /* 615 * Some bool flags: 616 * 617 * @dl_throttled tells if we exhausted the runtime. If so, the 618 * task has to wait for a replenishment to be performed at the 619 * next firing of dl_timer. 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 struct list_head trc_holdout_list; 848 struct list_head trc_blkd_node; 849 int trc_blkd_cpu; 850 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ 851 852 struct sched_info sched_info; 853 854 struct list_head tasks; 855 #ifdef CONFIG_SMP 856 struct plist_node pushable_tasks; 857 struct rb_node pushable_dl_tasks; 858 #endif 859 860 struct mm_struct *mm; 861 struct mm_struct *active_mm; 862 863 /* Per-thread vma caching: */ 864 struct vmacache vmacache; 865 866 #ifdef SPLIT_RSS_COUNTING 867 struct task_rss_stat rss_stat; 868 #endif 869 int exit_state; 870 int exit_code; 871 int exit_signal; 872 /* The signal sent when the parent dies: */ 873 int pdeath_signal; 874 /* JOBCTL_*, siglock protected: */ 875 unsigned long jobctl; 876 877 /* Used for emulating ABI behavior of previous Linux versions: */ 878 unsigned int personality; 879 880 /* Scheduler bits, serialized by scheduler locks: */ 881 unsigned sched_reset_on_fork:1; 882 unsigned sched_contributes_to_load:1; 883 unsigned sched_migrated:1; 884 #ifdef CONFIG_PSI 885 unsigned sched_psi_wake_requeue:1; 886 #endif 887 888 /* Force alignment to the next boundary: */ 889 unsigned :0; 890 891 /* Unserialized, strictly 'current' */ 892 893 /* 894 * This field must not be in the scheduler word above due to wakelist 895 * queueing no longer being serialized by p->on_cpu. However: 896 * 897 * p->XXX = X; ttwu() 898 * schedule() if (p->on_rq && ..) // false 899 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true 900 * deactivate_task() ttwu_queue_wakelist()) 901 * p->on_rq = 0; p->sched_remote_wakeup = Y; 902 * 903 * guarantees all stores of 'current' are visible before 904 * ->sched_remote_wakeup gets used, so it can be in this word. 905 */ 906 unsigned sched_remote_wakeup:1; 907 908 /* Bit to tell LSMs we're in execve(): */ 909 unsigned in_execve:1; 910 unsigned in_iowait:1; 911 #ifndef TIF_RESTORE_SIGMASK 912 unsigned restore_sigmask:1; 913 #endif 914 #ifdef CONFIG_MEMCG 915 unsigned in_user_fault:1; 916 #endif 917 #ifdef CONFIG_LRU_GEN 918 /* whether the LRU algorithm may apply to this access */ 919 unsigned in_lru_fault:1; 920 #endif 921 #ifdef CONFIG_COMPAT_BRK 922 unsigned brk_randomized:1; 923 #endif 924 #ifdef CONFIG_CGROUPS 925 /* disallow userland-initiated cgroup migration */ 926 unsigned no_cgroup_migration:1; 927 /* task is frozen/stopped (used by the cgroup freezer) */ 928 unsigned frozen:1; 929 #endif 930 #ifdef CONFIG_BLK_CGROUP 931 unsigned use_memdelay:1; 932 #endif 933 #ifdef CONFIG_PSI 934 /* Stalled due to lack of memory */ 935 unsigned in_memstall:1; 936 #endif 937 #ifdef CONFIG_PAGE_OWNER 938 /* Used by page_owner=on to detect recursion in page tracking. */ 939 unsigned in_page_owner:1; 940 #endif 941 #ifdef CONFIG_EVENTFD 942 /* Recursion prevention for eventfd_signal() */ 943 unsigned in_eventfd_signal:1; 944 #endif 945 #ifdef CONFIG_IOMMU_SVA 946 unsigned pasid_activated:1; 947 #endif 948 #ifdef CONFIG_CPU_SUP_INTEL 949 unsigned reported_split_lock:1; 950 #endif 951 #ifdef CONFIG_TASK_DELAY_ACCT 952 /* delay due to memory thrashing */ 953 unsigned in_thrashing:1; 954 #endif 955 956 unsigned long atomic_flags; /* Flags requiring atomic access. */ 957 958 struct restart_block restart_block; 959 960 pid_t pid; 961 pid_t tgid; 962 963 #ifdef CONFIG_STACKPROTECTOR 964 /* Canary value for the -fstack-protector GCC feature: */ 965 unsigned long stack_canary; 966 #endif 967 /* 968 * Pointers to the (original) parent process, youngest child, younger sibling, 969 * older sibling, respectively. (p->father can be replaced with 970 * p->real_parent->pid) 971 */ 972 973 /* Real parent process: */ 974 struct task_struct __rcu *real_parent; 975 976 /* Recipient of SIGCHLD, wait4() reports: */ 977 struct task_struct __rcu *parent; 978 979 /* 980 * Children/sibling form the list of natural children: 981 */ 982 struct list_head children; 983 struct list_head sibling; 984 struct task_struct *group_leader; 985 986 /* 987 * 'ptraced' is the list of tasks this task is using ptrace() on. 988 * 989 * This includes both natural children and PTRACE_ATTACH targets. 990 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 991 */ 992 struct list_head ptraced; 993 struct list_head ptrace_entry; 994 995 /* PID/PID hash table linkage. */ 996 struct pid *thread_pid; 997 struct hlist_node pid_links[PIDTYPE_MAX]; 998 struct list_head thread_group; 999 struct list_head thread_node; 1000 1001 struct completion *vfork_done; 1002 1003 /* CLONE_CHILD_SETTID: */ 1004 int __user *set_child_tid; 1005 1006 /* CLONE_CHILD_CLEARTID: */ 1007 int __user *clear_child_tid; 1008 1009 /* PF_KTHREAD | PF_IO_WORKER */ 1010 void *worker_private; 1011 1012 u64 utime; 1013 u64 stime; 1014 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 1015 u64 utimescaled; 1016 u64 stimescaled; 1017 #endif 1018 u64 gtime; 1019 struct prev_cputime prev_cputime; 1020 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 1021 struct vtime vtime; 1022 #endif 1023 1024 #ifdef CONFIG_NO_HZ_FULL 1025 atomic_t tick_dep_mask; 1026 #endif 1027 /* Context switch counts: */ 1028 unsigned long nvcsw; 1029 unsigned long nivcsw; 1030 1031 /* Monotonic time in nsecs: */ 1032 u64 start_time; 1033 1034 /* Boot based time in nsecs: */ 1035 u64 start_boottime; 1036 1037 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 1038 unsigned long min_flt; 1039 unsigned long maj_flt; 1040 1041 /* Empty if CONFIG_POSIX_CPUTIMERS=n */ 1042 struct posix_cputimers posix_cputimers; 1043 1044 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK 1045 struct posix_cputimers_work posix_cputimers_work; 1046 #endif 1047 1048 /* Process credentials: */ 1049 1050 /* Tracer's credentials at attach: */ 1051 const struct cred __rcu *ptracer_cred; 1052 1053 /* Objective and real subjective task credentials (COW): */ 1054 const struct cred __rcu *real_cred; 1055 1056 /* Effective (overridable) subjective task credentials (COW): */ 1057 const struct cred __rcu *cred; 1058 1059 #ifdef CONFIG_KEYS 1060 /* Cached requested key. */ 1061 struct key *cached_requested_key; 1062 #endif 1063 1064 /* 1065 * executable name, excluding path. 1066 * 1067 * - normally initialized setup_new_exec() 1068 * - access it with [gs]et_task_comm() 1069 * - lock it with task_lock() 1070 */ 1071 char comm[TASK_COMM_LEN]; 1072 1073 struct nameidata *nameidata; 1074 1075 #ifdef CONFIG_SYSVIPC 1076 struct sysv_sem sysvsem; 1077 struct sysv_shm sysvshm; 1078 #endif 1079 #ifdef CONFIG_DETECT_HUNG_TASK 1080 unsigned long last_switch_count; 1081 unsigned long last_switch_time; 1082 #endif 1083 /* Filesystem information: */ 1084 struct fs_struct *fs; 1085 1086 /* Open file information: */ 1087 struct files_struct *files; 1088 1089 #ifdef CONFIG_IO_URING 1090 struct io_uring_task *io_uring; 1091 #endif 1092 1093 /* Namespaces: */ 1094 struct nsproxy *nsproxy; 1095 1096 /* Signal handlers: */ 1097 struct signal_struct *signal; 1098 struct sighand_struct __rcu *sighand; 1099 sigset_t blocked; 1100 sigset_t real_blocked; 1101 /* Restored if set_restore_sigmask() was used: */ 1102 sigset_t saved_sigmask; 1103 struct sigpending pending; 1104 unsigned long sas_ss_sp; 1105 size_t sas_ss_size; 1106 unsigned int sas_ss_flags; 1107 1108 struct callback_head *task_works; 1109 1110 #ifdef CONFIG_AUDIT 1111 #ifdef CONFIG_AUDITSYSCALL 1112 struct audit_context *audit_context; 1113 #endif 1114 kuid_t loginuid; 1115 unsigned int sessionid; 1116 #endif 1117 struct seccomp seccomp; 1118 struct syscall_user_dispatch syscall_dispatch; 1119 1120 /* Thread group tracking: */ 1121 u64 parent_exec_id; 1122 u64 self_exec_id; 1123 1124 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 1125 spinlock_t alloc_lock; 1126 1127 /* Protection of the PI data structures: */ 1128 raw_spinlock_t pi_lock; 1129 1130 struct wake_q_node wake_q; 1131 1132 #ifdef CONFIG_RT_MUTEXES 1133 /* PI waiters blocked on a rt_mutex held by this task: */ 1134 struct rb_root_cached pi_waiters; 1135 /* Updated under owner's pi_lock and rq lock */ 1136 struct task_struct *pi_top_task; 1137 /* Deadlock detection and priority inheritance handling: */ 1138 struct rt_mutex_waiter *pi_blocked_on; 1139 #endif 1140 1141 #ifdef CONFIG_DEBUG_MUTEXES 1142 /* Mutex deadlock detection: */ 1143 struct mutex_waiter *blocked_on; 1144 #endif 1145 1146 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1147 int non_block_count; 1148 #endif 1149 1150 #ifdef CONFIG_TRACE_IRQFLAGS 1151 struct irqtrace_events irqtrace; 1152 unsigned int hardirq_threaded; 1153 u64 hardirq_chain_key; 1154 int softirqs_enabled; 1155 int softirq_context; 1156 int irq_config; 1157 #endif 1158 #ifdef CONFIG_PREEMPT_RT 1159 int softirq_disable_cnt; 1160 #endif 1161 1162 #ifdef CONFIG_LOCKDEP 1163 # define MAX_LOCK_DEPTH 48UL 1164 u64 curr_chain_key; 1165 int lockdep_depth; 1166 unsigned int lockdep_recursion; 1167 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1168 #endif 1169 1170 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP) 1171 unsigned int in_ubsan; 1172 #endif 1173 1174 /* Journalling filesystem info: */ 1175 void *journal_info; 1176 1177 /* Stacked block device info: */ 1178 struct bio_list *bio_list; 1179 1180 /* Stack plugging: */ 1181 struct blk_plug *plug; 1182 1183 /* VM state: */ 1184 struct reclaim_state *reclaim_state; 1185 1186 struct backing_dev_info *backing_dev_info; 1187 1188 struct io_context *io_context; 1189 1190 #ifdef CONFIG_COMPACTION 1191 struct capture_control *capture_control; 1192 #endif 1193 /* Ptrace state: */ 1194 unsigned long ptrace_message; 1195 kernel_siginfo_t *last_siginfo; 1196 1197 struct task_io_accounting ioac; 1198 #ifdef CONFIG_PSI 1199 /* Pressure stall state */ 1200 unsigned int psi_flags; 1201 #endif 1202 #ifdef CONFIG_TASK_XACCT 1203 /* Accumulated RSS usage: */ 1204 u64 acct_rss_mem1; 1205 /* Accumulated virtual memory usage: */ 1206 u64 acct_vm_mem1; 1207 /* stime + utime since last update: */ 1208 u64 acct_timexpd; 1209 #endif 1210 #ifdef CONFIG_CPUSETS 1211 /* Protected by ->alloc_lock: */ 1212 nodemask_t mems_allowed; 1213 /* Sequence number to catch updates: */ 1214 seqcount_spinlock_t mems_allowed_seq; 1215 int cpuset_mem_spread_rotor; 1216 int cpuset_slab_spread_rotor; 1217 #endif 1218 #ifdef CONFIG_CGROUPS 1219 /* Control Group info protected by css_set_lock: */ 1220 struct css_set __rcu *cgroups; 1221 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 1222 struct list_head cg_list; 1223 #endif 1224 #ifdef CONFIG_X86_CPU_RESCTRL 1225 u32 closid; 1226 u32 rmid; 1227 #endif 1228 #ifdef CONFIG_FUTEX 1229 struct robust_list_head __user *robust_list; 1230 #ifdef CONFIG_COMPAT 1231 struct compat_robust_list_head __user *compat_robust_list; 1232 #endif 1233 struct list_head pi_state_list; 1234 struct futex_pi_state *pi_state_cache; 1235 struct mutex futex_exit_mutex; 1236 unsigned int futex_state; 1237 #endif 1238 #ifdef CONFIG_PERF_EVENTS 1239 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 1240 struct mutex perf_event_mutex; 1241 struct list_head perf_event_list; 1242 #endif 1243 #ifdef CONFIG_DEBUG_PREEMPT 1244 unsigned long preempt_disable_ip; 1245 #endif 1246 #ifdef CONFIG_NUMA 1247 /* Protected by alloc_lock: */ 1248 struct mempolicy *mempolicy; 1249 short il_prev; 1250 short pref_node_fork; 1251 #endif 1252 #ifdef CONFIG_NUMA_BALANCING 1253 int numa_scan_seq; 1254 unsigned int numa_scan_period; 1255 unsigned int numa_scan_period_max; 1256 int numa_preferred_nid; 1257 unsigned long numa_migrate_retry; 1258 /* Migration stamp: */ 1259 u64 node_stamp; 1260 u64 last_task_numa_placement; 1261 u64 last_sum_exec_runtime; 1262 struct callback_head numa_work; 1263 1264 /* 1265 * This pointer is only modified for current in syscall and 1266 * pagefault context (and for tasks being destroyed), so it can be read 1267 * from any of the following contexts: 1268 * - RCU read-side critical section 1269 * - current->numa_group from everywhere 1270 * - task's runqueue locked, task not running 1271 */ 1272 struct numa_group __rcu *numa_group; 1273 1274 /* 1275 * numa_faults is an array split into four regions: 1276 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 1277 * in this precise order. 1278 * 1279 * faults_memory: Exponential decaying average of faults on a per-node 1280 * basis. Scheduling placement decisions are made based on these 1281 * counts. The values remain static for the duration of a PTE scan. 1282 * faults_cpu: Track the nodes the process was running on when a NUMA 1283 * hinting fault was incurred. 1284 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 1285 * during the current scan window. When the scan completes, the counts 1286 * in faults_memory and faults_cpu decay and these values are copied. 1287 */ 1288 unsigned long *numa_faults; 1289 unsigned long total_numa_faults; 1290 1291 /* 1292 * numa_faults_locality tracks if faults recorded during the last 1293 * scan window were remote/local or failed to migrate. The task scan 1294 * period is adapted based on the locality of the faults with different 1295 * weights depending on whether they were shared or private faults 1296 */ 1297 unsigned long numa_faults_locality[3]; 1298 1299 unsigned long numa_pages_migrated; 1300 #endif /* CONFIG_NUMA_BALANCING */ 1301 1302 #ifdef CONFIG_RSEQ 1303 struct rseq __user *rseq; 1304 u32 rseq_sig; 1305 /* 1306 * RmW on rseq_event_mask must be performed atomically 1307 * with respect to preemption. 1308 */ 1309 unsigned long rseq_event_mask; 1310 #endif 1311 1312 struct tlbflush_unmap_batch tlb_ubc; 1313 1314 union { 1315 refcount_t rcu_users; 1316 struct rcu_head rcu; 1317 }; 1318 1319 /* Cache last used pipe for splice(): */ 1320 struct pipe_inode_info *splice_pipe; 1321 1322 struct page_frag task_frag; 1323 1324 #ifdef CONFIG_TASK_DELAY_ACCT 1325 struct task_delay_info *delays; 1326 #endif 1327 1328 #ifdef CONFIG_FAULT_INJECTION 1329 int make_it_fail; 1330 unsigned int fail_nth; 1331 #endif 1332 /* 1333 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 1334 * balance_dirty_pages() for a dirty throttling pause: 1335 */ 1336 int nr_dirtied; 1337 int nr_dirtied_pause; 1338 /* Start of a write-and-pause period: */ 1339 unsigned long dirty_paused_when; 1340 1341 #ifdef CONFIG_LATENCYTOP 1342 int latency_record_count; 1343 struct latency_record latency_record[LT_SAVECOUNT]; 1344 #endif 1345 /* 1346 * Time slack values; these are used to round up poll() and 1347 * select() etc timeout values. These are in nanoseconds. 1348 */ 1349 u64 timer_slack_ns; 1350 u64 default_timer_slack_ns; 1351 1352 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 1353 unsigned int kasan_depth; 1354 #endif 1355 1356 #ifdef CONFIG_KCSAN 1357 struct kcsan_ctx kcsan_ctx; 1358 #ifdef CONFIG_TRACE_IRQFLAGS 1359 struct irqtrace_events kcsan_save_irqtrace; 1360 #endif 1361 #ifdef CONFIG_KCSAN_WEAK_MEMORY 1362 int kcsan_stack_depth; 1363 #endif 1364 #endif 1365 1366 #if IS_ENABLED(CONFIG_KUNIT) 1367 struct kunit *kunit_test; 1368 #endif 1369 1370 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1371 /* Index of current stored address in ret_stack: */ 1372 int curr_ret_stack; 1373 int curr_ret_depth; 1374 1375 /* Stack of return addresses for return function tracing: */ 1376 struct ftrace_ret_stack *ret_stack; 1377 1378 /* Timestamp for last schedule: */ 1379 unsigned long long ftrace_timestamp; 1380 1381 /* 1382 * Number of functions that haven't been traced 1383 * because of depth overrun: 1384 */ 1385 atomic_t trace_overrun; 1386 1387 /* Pause tracing: */ 1388 atomic_t tracing_graph_pause; 1389 #endif 1390 1391 #ifdef CONFIG_TRACING 1392 /* State flags for use by tracers: */ 1393 unsigned long trace; 1394 1395 /* Bitmask and counter of trace recursion: */ 1396 unsigned long trace_recursion; 1397 #endif /* CONFIG_TRACING */ 1398 1399 #ifdef CONFIG_KCOV 1400 /* See kernel/kcov.c for more details. */ 1401 1402 /* Coverage collection mode enabled for this task (0 if disabled): */ 1403 unsigned int kcov_mode; 1404 1405 /* Size of the kcov_area: */ 1406 unsigned int kcov_size; 1407 1408 /* Buffer for coverage collection: */ 1409 void *kcov_area; 1410 1411 /* KCOV descriptor wired with this task or NULL: */ 1412 struct kcov *kcov; 1413 1414 /* KCOV common handle for remote coverage collection: */ 1415 u64 kcov_handle; 1416 1417 /* KCOV sequence number: */ 1418 int kcov_sequence; 1419 1420 /* Collect coverage from softirq context: */ 1421 unsigned int kcov_softirq; 1422 #endif 1423 1424 #ifdef CONFIG_MEMCG 1425 struct mem_cgroup *memcg_in_oom; 1426 gfp_t memcg_oom_gfp_mask; 1427 int memcg_oom_order; 1428 1429 /* Number of pages to reclaim on returning to userland: */ 1430 unsigned int memcg_nr_pages_over_high; 1431 1432 /* Used by memcontrol for targeted memcg charge: */ 1433 struct mem_cgroup *active_memcg; 1434 #endif 1435 1436 #ifdef CONFIG_BLK_CGROUP 1437 struct request_queue *throttle_queue; 1438 #endif 1439 1440 #ifdef CONFIG_UPROBES 1441 struct uprobe_task *utask; 1442 #endif 1443 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1444 unsigned int sequential_io; 1445 unsigned int sequential_io_avg; 1446 #endif 1447 struct kmap_ctrl kmap_ctrl; 1448 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1449 unsigned long task_state_change; 1450 # ifdef CONFIG_PREEMPT_RT 1451 unsigned long saved_state_change; 1452 # endif 1453 #endif 1454 int pagefault_disabled; 1455 #ifdef CONFIG_MMU 1456 struct task_struct *oom_reaper_list; 1457 struct timer_list oom_reaper_timer; 1458 #endif 1459 #ifdef CONFIG_VMAP_STACK 1460 struct vm_struct *stack_vm_area; 1461 #endif 1462 #ifdef CONFIG_THREAD_INFO_IN_TASK 1463 /* A live task holds one reference: */ 1464 refcount_t stack_refcount; 1465 #endif 1466 #ifdef CONFIG_LIVEPATCH 1467 int patch_state; 1468 #endif 1469 #ifdef CONFIG_SECURITY 1470 /* Used by LSM modules for access restriction: */ 1471 void *security; 1472 #endif 1473 #ifdef CONFIG_BPF_SYSCALL 1474 /* Used by BPF task local storage */ 1475 struct bpf_local_storage __rcu *bpf_storage; 1476 /* Used for BPF run context */ 1477 struct bpf_run_ctx *bpf_ctx; 1478 #endif 1479 1480 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK 1481 unsigned long lowest_stack; 1482 unsigned long prev_lowest_stack; 1483 #endif 1484 1485 #ifdef CONFIG_X86_MCE 1486 void __user *mce_vaddr; 1487 __u64 mce_kflags; 1488 u64 mce_addr; 1489 __u64 mce_ripv : 1, 1490 mce_whole_page : 1, 1491 __mce_reserved : 62; 1492 struct callback_head mce_kill_me; 1493 int mce_count; 1494 #endif 1495 1496 #ifdef CONFIG_KRETPROBES 1497 struct llist_head kretprobe_instances; 1498 #endif 1499 #ifdef CONFIG_RETHOOK 1500 struct llist_head rethooks; 1501 #endif 1502 1503 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH 1504 /* 1505 * If L1D flush is supported on mm context switch 1506 * then we use this callback head to queue kill work 1507 * to kill tasks that are not running on SMT disabled 1508 * cores 1509 */ 1510 struct callback_head l1d_flush_kill; 1511 #endif 1512 1513 #ifdef CONFIG_RV 1514 /* 1515 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS. 1516 * If we find justification for more monitors, we can think 1517 * about adding more or developing a dynamic method. So far, 1518 * none of these are justified. 1519 */ 1520 union rv_task_monitor rv[RV_PER_TASK_MONITORS]; 1521 #endif 1522 1523 /* 1524 * New fields for task_struct should be added above here, so that 1525 * they are included in the randomized portion of task_struct. 1526 */ 1527 randomized_struct_fields_end 1528 1529 /* CPU-specific state of this task: */ 1530 struct thread_struct thread; 1531 1532 /* 1533 * WARNING: on x86, 'thread_struct' contains a variable-sized 1534 * structure. It *MUST* be at the end of 'task_struct'. 1535 * 1536 * Do not put anything below here! 1537 */ 1538 }; 1539 1540 static inline struct pid *task_pid(struct task_struct *task) 1541 { 1542 return task->thread_pid; 1543 } 1544 1545 /* 1546 * the helpers to get the task's different pids as they are seen 1547 * from various namespaces 1548 * 1549 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1550 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1551 * current. 1552 * task_xid_nr_ns() : id seen from the ns specified; 1553 * 1554 * see also pid_nr() etc in include/linux/pid.h 1555 */ 1556 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1557 1558 static inline pid_t task_pid_nr(struct task_struct *tsk) 1559 { 1560 return tsk->pid; 1561 } 1562 1563 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1564 { 1565 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1566 } 1567 1568 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1569 { 1570 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1571 } 1572 1573 1574 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1575 { 1576 return tsk->tgid; 1577 } 1578 1579 /** 1580 * pid_alive - check that a task structure is not stale 1581 * @p: Task structure to be checked. 1582 * 1583 * Test if a process is not yet dead (at most zombie state) 1584 * If pid_alive fails, then pointers within the task structure 1585 * can be stale and must not be dereferenced. 1586 * 1587 * Return: 1 if the process is alive. 0 otherwise. 1588 */ 1589 static inline int pid_alive(const struct task_struct *p) 1590 { 1591 return p->thread_pid != NULL; 1592 } 1593 1594 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1595 { 1596 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1597 } 1598 1599 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1600 { 1601 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1602 } 1603 1604 1605 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1606 { 1607 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1608 } 1609 1610 static inline pid_t task_session_vnr(struct task_struct *tsk) 1611 { 1612 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1613 } 1614 1615 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1616 { 1617 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns); 1618 } 1619 1620 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1621 { 1622 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL); 1623 } 1624 1625 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1626 { 1627 pid_t pid = 0; 1628 1629 rcu_read_lock(); 1630 if (pid_alive(tsk)) 1631 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1632 rcu_read_unlock(); 1633 1634 return pid; 1635 } 1636 1637 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1638 { 1639 return task_ppid_nr_ns(tsk, &init_pid_ns); 1640 } 1641 1642 /* Obsolete, do not use: */ 1643 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1644 { 1645 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1646 } 1647 1648 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1649 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1650 1651 static inline unsigned int __task_state_index(unsigned int tsk_state, 1652 unsigned int tsk_exit_state) 1653 { 1654 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT; 1655 1656 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1657 1658 if (tsk_state == TASK_IDLE) 1659 state = TASK_REPORT_IDLE; 1660 1661 /* 1662 * We're lying here, but rather than expose a completely new task state 1663 * to userspace, we can make this appear as if the task has gone through 1664 * a regular rt_mutex_lock() call. 1665 */ 1666 if (tsk_state == TASK_RTLOCK_WAIT) 1667 state = TASK_UNINTERRUPTIBLE; 1668 1669 return fls(state); 1670 } 1671 1672 static inline unsigned int task_state_index(struct task_struct *tsk) 1673 { 1674 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state); 1675 } 1676 1677 static inline char task_index_to_char(unsigned int state) 1678 { 1679 static const char state_char[] = "RSDTtXZPI"; 1680 1681 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1682 1683 return state_char[state]; 1684 } 1685 1686 static inline char task_state_to_char(struct task_struct *tsk) 1687 { 1688 return task_index_to_char(task_state_index(tsk)); 1689 } 1690 1691 /** 1692 * is_global_init - check if a task structure is init. Since init 1693 * is free to have sub-threads we need to check tgid. 1694 * @tsk: Task structure to be checked. 1695 * 1696 * Check if a task structure is the first user space task the kernel created. 1697 * 1698 * Return: 1 if the task structure is init. 0 otherwise. 1699 */ 1700 static inline int is_global_init(struct task_struct *tsk) 1701 { 1702 return task_tgid_nr(tsk) == 1; 1703 } 1704 1705 extern struct pid *cad_pid; 1706 1707 /* 1708 * Per process flags 1709 */ 1710 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */ 1711 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1712 #define PF_EXITING 0x00000004 /* Getting shut down */ 1713 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */ 1714 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */ 1715 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1716 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1717 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1718 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1719 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1720 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1721 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1722 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1723 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1724 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1725 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1726 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1727 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1728 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1729 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to, 1730 * I am cleaning dirty pages from some other bdi. */ 1731 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1732 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1733 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */ 1734 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1735 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */ 1736 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1737 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1738 1739 /* 1740 * Only the _current_ task can read/write to tsk->flags, but other 1741 * tasks can access tsk->flags in readonly mode for example 1742 * with tsk_used_math (like during threaded core dumping). 1743 * There is however an exception to this rule during ptrace 1744 * or during fork: the ptracer task is allowed to write to the 1745 * child->flags of its traced child (same goes for fork, the parent 1746 * can write to the child->flags), because we're guaranteed the 1747 * child is not running and in turn not changing child->flags 1748 * at the same time the parent does it. 1749 */ 1750 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1751 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1752 #define clear_used_math() clear_stopped_child_used_math(current) 1753 #define set_used_math() set_stopped_child_used_math(current) 1754 1755 #define conditional_stopped_child_used_math(condition, child) \ 1756 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1757 1758 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1759 1760 #define copy_to_stopped_child_used_math(child) \ 1761 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1762 1763 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1764 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1765 #define used_math() tsk_used_math(current) 1766 1767 static __always_inline bool is_percpu_thread(void) 1768 { 1769 #ifdef CONFIG_SMP 1770 return (current->flags & PF_NO_SETAFFINITY) && 1771 (current->nr_cpus_allowed == 1); 1772 #else 1773 return true; 1774 #endif 1775 } 1776 1777 /* Per-process atomic flags. */ 1778 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1779 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1780 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1781 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */ 1782 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/ 1783 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */ 1784 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */ 1785 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */ 1786 1787 #define TASK_PFA_TEST(name, func) \ 1788 static inline bool task_##func(struct task_struct *p) \ 1789 { return test_bit(PFA_##name, &p->atomic_flags); } 1790 1791 #define TASK_PFA_SET(name, func) \ 1792 static inline void task_set_##func(struct task_struct *p) \ 1793 { set_bit(PFA_##name, &p->atomic_flags); } 1794 1795 #define TASK_PFA_CLEAR(name, func) \ 1796 static inline void task_clear_##func(struct task_struct *p) \ 1797 { clear_bit(PFA_##name, &p->atomic_flags); } 1798 1799 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1800 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1801 1802 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1803 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1804 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1805 1806 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1807 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1808 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1809 1810 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable) 1811 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable) 1812 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable) 1813 1814 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1815 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1816 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec) 1817 1818 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1819 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable) 1820 1821 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable) 1822 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable) 1823 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable) 1824 1825 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1826 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable) 1827 1828 static inline void 1829 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1830 { 1831 current->flags &= ~flags; 1832 current->flags |= orig_flags & flags; 1833 } 1834 1835 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1836 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_effective_cpus); 1837 #ifdef CONFIG_SMP 1838 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1839 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1840 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node); 1841 extern void release_user_cpus_ptr(struct task_struct *p); 1842 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask); 1843 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p); 1844 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p); 1845 #else 1846 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1847 { 1848 } 1849 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1850 { 1851 if (!cpumask_test_cpu(0, new_mask)) 1852 return -EINVAL; 1853 return 0; 1854 } 1855 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node) 1856 { 1857 if (src->user_cpus_ptr) 1858 return -EINVAL; 1859 return 0; 1860 } 1861 static inline void release_user_cpus_ptr(struct task_struct *p) 1862 { 1863 WARN_ON(p->user_cpus_ptr); 1864 } 1865 1866 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) 1867 { 1868 return 0; 1869 } 1870 #endif 1871 1872 extern int yield_to(struct task_struct *p, bool preempt); 1873 extern void set_user_nice(struct task_struct *p, long nice); 1874 extern int task_prio(const struct task_struct *p); 1875 1876 /** 1877 * task_nice - return the nice value of a given task. 1878 * @p: the task in question. 1879 * 1880 * Return: The nice value [ -20 ... 0 ... 19 ]. 1881 */ 1882 static inline int task_nice(const struct task_struct *p) 1883 { 1884 return PRIO_TO_NICE((p)->static_prio); 1885 } 1886 1887 extern int can_nice(const struct task_struct *p, const int nice); 1888 extern int task_curr(const struct task_struct *p); 1889 extern int idle_cpu(int cpu); 1890 extern int available_idle_cpu(int cpu); 1891 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1892 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1893 extern void sched_set_fifo(struct task_struct *p); 1894 extern void sched_set_fifo_low(struct task_struct *p); 1895 extern void sched_set_normal(struct task_struct *p, int nice); 1896 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1897 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1898 extern struct task_struct *idle_task(int cpu); 1899 1900 /** 1901 * is_idle_task - is the specified task an idle task? 1902 * @p: the task in question. 1903 * 1904 * Return: 1 if @p is an idle task. 0 otherwise. 1905 */ 1906 static __always_inline bool is_idle_task(const struct task_struct *p) 1907 { 1908 return !!(p->flags & PF_IDLE); 1909 } 1910 1911 extern struct task_struct *curr_task(int cpu); 1912 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1913 1914 void yield(void); 1915 1916 union thread_union { 1917 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1918 struct task_struct task; 1919 #endif 1920 #ifndef CONFIG_THREAD_INFO_IN_TASK 1921 struct thread_info thread_info; 1922 #endif 1923 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1924 }; 1925 1926 #ifndef CONFIG_THREAD_INFO_IN_TASK 1927 extern struct thread_info init_thread_info; 1928 #endif 1929 1930 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1931 1932 #ifdef CONFIG_THREAD_INFO_IN_TASK 1933 # define task_thread_info(task) (&(task)->thread_info) 1934 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1935 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1936 #endif 1937 1938 /* 1939 * find a task by one of its numerical ids 1940 * 1941 * find_task_by_pid_ns(): 1942 * finds a task by its pid in the specified namespace 1943 * find_task_by_vpid(): 1944 * finds a task by its virtual pid 1945 * 1946 * see also find_vpid() etc in include/linux/pid.h 1947 */ 1948 1949 extern struct task_struct *find_task_by_vpid(pid_t nr); 1950 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1951 1952 /* 1953 * find a task by its virtual pid and get the task struct 1954 */ 1955 extern struct task_struct *find_get_task_by_vpid(pid_t nr); 1956 1957 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1958 extern int wake_up_process(struct task_struct *tsk); 1959 extern void wake_up_new_task(struct task_struct *tsk); 1960 1961 #ifdef CONFIG_SMP 1962 extern void kick_process(struct task_struct *tsk); 1963 #else 1964 static inline void kick_process(struct task_struct *tsk) { } 1965 #endif 1966 1967 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1968 1969 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1970 { 1971 __set_task_comm(tsk, from, false); 1972 } 1973 1974 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1975 #define get_task_comm(buf, tsk) ({ \ 1976 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1977 __get_task_comm(buf, sizeof(buf), tsk); \ 1978 }) 1979 1980 #ifdef CONFIG_SMP 1981 static __always_inline void scheduler_ipi(void) 1982 { 1983 /* 1984 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting 1985 * TIF_NEED_RESCHED remotely (for the first time) will also send 1986 * this IPI. 1987 */ 1988 preempt_fold_need_resched(); 1989 } 1990 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state); 1991 #else 1992 static inline void scheduler_ipi(void) { } 1993 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state) 1994 { 1995 return 1; 1996 } 1997 #endif 1998 1999 /* 2000 * Set thread flags in other task's structures. 2001 * See asm/thread_info.h for TIF_xxxx flags available: 2002 */ 2003 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 2004 { 2005 set_ti_thread_flag(task_thread_info(tsk), flag); 2006 } 2007 2008 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2009 { 2010 clear_ti_thread_flag(task_thread_info(tsk), flag); 2011 } 2012 2013 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, 2014 bool value) 2015 { 2016 update_ti_thread_flag(task_thread_info(tsk), flag, value); 2017 } 2018 2019 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2020 { 2021 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2022 } 2023 2024 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2025 { 2026 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2027 } 2028 2029 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2030 { 2031 return test_ti_thread_flag(task_thread_info(tsk), flag); 2032 } 2033 2034 static inline void set_tsk_need_resched(struct task_struct *tsk) 2035 { 2036 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2037 } 2038 2039 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2040 { 2041 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2042 } 2043 2044 static inline int test_tsk_need_resched(struct task_struct *tsk) 2045 { 2046 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2047 } 2048 2049 /* 2050 * cond_resched() and cond_resched_lock(): latency reduction via 2051 * explicit rescheduling in places that are safe. The return 2052 * value indicates whether a reschedule was done in fact. 2053 * cond_resched_lock() will drop the spinlock before scheduling, 2054 */ 2055 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) 2056 extern int __cond_resched(void); 2057 2058 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL) 2059 2060 DECLARE_STATIC_CALL(cond_resched, __cond_resched); 2061 2062 static __always_inline int _cond_resched(void) 2063 { 2064 return static_call_mod(cond_resched)(); 2065 } 2066 2067 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY) 2068 extern int dynamic_cond_resched(void); 2069 2070 static __always_inline int _cond_resched(void) 2071 { 2072 return dynamic_cond_resched(); 2073 } 2074 2075 #else 2076 2077 static inline int _cond_resched(void) 2078 { 2079 return __cond_resched(); 2080 } 2081 2082 #endif /* CONFIG_PREEMPT_DYNAMIC */ 2083 2084 #else 2085 2086 static inline int _cond_resched(void) { return 0; } 2087 2088 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */ 2089 2090 #define cond_resched() ({ \ 2091 __might_resched(__FILE__, __LINE__, 0); \ 2092 _cond_resched(); \ 2093 }) 2094 2095 extern int __cond_resched_lock(spinlock_t *lock); 2096 extern int __cond_resched_rwlock_read(rwlock_t *lock); 2097 extern int __cond_resched_rwlock_write(rwlock_t *lock); 2098 2099 #define MIGHT_RESCHED_RCU_SHIFT 8 2100 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1) 2101 2102 #ifndef CONFIG_PREEMPT_RT 2103 /* 2104 * Non RT kernels have an elevated preempt count due to the held lock, 2105 * but are not allowed to be inside a RCU read side critical section 2106 */ 2107 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET 2108 #else 2109 /* 2110 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in 2111 * cond_resched*lock() has to take that into account because it checks for 2112 * preempt_count() and rcu_preempt_depth(). 2113 */ 2114 # define PREEMPT_LOCK_RESCHED_OFFSETS \ 2115 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT)) 2116 #endif 2117 2118 #define cond_resched_lock(lock) ({ \ 2119 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2120 __cond_resched_lock(lock); \ 2121 }) 2122 2123 #define cond_resched_rwlock_read(lock) ({ \ 2124 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2125 __cond_resched_rwlock_read(lock); \ 2126 }) 2127 2128 #define cond_resched_rwlock_write(lock) ({ \ 2129 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \ 2130 __cond_resched_rwlock_write(lock); \ 2131 }) 2132 2133 static inline void cond_resched_rcu(void) 2134 { 2135 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 2136 rcu_read_unlock(); 2137 cond_resched(); 2138 rcu_read_lock(); 2139 #endif 2140 } 2141 2142 #ifdef CONFIG_PREEMPT_DYNAMIC 2143 2144 extern bool preempt_model_none(void); 2145 extern bool preempt_model_voluntary(void); 2146 extern bool preempt_model_full(void); 2147 2148 #else 2149 2150 static inline bool preempt_model_none(void) 2151 { 2152 return IS_ENABLED(CONFIG_PREEMPT_NONE); 2153 } 2154 static inline bool preempt_model_voluntary(void) 2155 { 2156 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY); 2157 } 2158 static inline bool preempt_model_full(void) 2159 { 2160 return IS_ENABLED(CONFIG_PREEMPT); 2161 } 2162 2163 #endif 2164 2165 static inline bool preempt_model_rt(void) 2166 { 2167 return IS_ENABLED(CONFIG_PREEMPT_RT); 2168 } 2169 2170 /* 2171 * Does the preemption model allow non-cooperative preemption? 2172 * 2173 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with 2174 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the 2175 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the 2176 * PREEMPT_NONE model. 2177 */ 2178 static inline bool preempt_model_preemptible(void) 2179 { 2180 return preempt_model_full() || preempt_model_rt(); 2181 } 2182 2183 /* 2184 * Does a critical section need to be broken due to another 2185 * task waiting?: (technically does not depend on CONFIG_PREEMPTION, 2186 * but a general need for low latency) 2187 */ 2188 static inline int spin_needbreak(spinlock_t *lock) 2189 { 2190 #ifdef CONFIG_PREEMPTION 2191 return spin_is_contended(lock); 2192 #else 2193 return 0; 2194 #endif 2195 } 2196 2197 /* 2198 * Check if a rwlock is contended. 2199 * Returns non-zero if there is another task waiting on the rwlock. 2200 * Returns zero if the lock is not contended or the system / underlying 2201 * rwlock implementation does not support contention detection. 2202 * Technically does not depend on CONFIG_PREEMPTION, but a general need 2203 * for low latency. 2204 */ 2205 static inline int rwlock_needbreak(rwlock_t *lock) 2206 { 2207 #ifdef CONFIG_PREEMPTION 2208 return rwlock_is_contended(lock); 2209 #else 2210 return 0; 2211 #endif 2212 } 2213 2214 static __always_inline bool need_resched(void) 2215 { 2216 return unlikely(tif_need_resched()); 2217 } 2218 2219 /* 2220 * Wrappers for p->thread_info->cpu access. No-op on UP. 2221 */ 2222 #ifdef CONFIG_SMP 2223 2224 static inline unsigned int task_cpu(const struct task_struct *p) 2225 { 2226 return READ_ONCE(task_thread_info(p)->cpu); 2227 } 2228 2229 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2230 2231 #else 2232 2233 static inline unsigned int task_cpu(const struct task_struct *p) 2234 { 2235 return 0; 2236 } 2237 2238 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2239 { 2240 } 2241 2242 #endif /* CONFIG_SMP */ 2243 2244 extern bool sched_task_on_rq(struct task_struct *p); 2245 extern unsigned long get_wchan(struct task_struct *p); 2246 extern struct task_struct *cpu_curr_snapshot(int cpu); 2247 2248 /* 2249 * In order to reduce various lock holder preemption latencies provide an 2250 * interface to see if a vCPU is currently running or not. 2251 * 2252 * This allows us to terminate optimistic spin loops and block, analogous to 2253 * the native optimistic spin heuristic of testing if the lock owner task is 2254 * running or not. 2255 */ 2256 #ifndef vcpu_is_preempted 2257 static inline bool vcpu_is_preempted(int cpu) 2258 { 2259 return false; 2260 } 2261 #endif 2262 2263 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2264 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2265 2266 #ifndef TASK_SIZE_OF 2267 #define TASK_SIZE_OF(tsk) TASK_SIZE 2268 #endif 2269 2270 #ifdef CONFIG_SMP 2271 static inline bool owner_on_cpu(struct task_struct *owner) 2272 { 2273 /* 2274 * As lock holder preemption issue, we both skip spinning if 2275 * task is not on cpu or its cpu is preempted 2276 */ 2277 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner)); 2278 } 2279 2280 /* Returns effective CPU energy utilization, as seen by the scheduler */ 2281 unsigned long sched_cpu_util(int cpu); 2282 #endif /* CONFIG_SMP */ 2283 2284 #ifdef CONFIG_RSEQ 2285 2286 /* 2287 * Map the event mask on the user-space ABI enum rseq_cs_flags 2288 * for direct mask checks. 2289 */ 2290 enum rseq_event_mask_bits { 2291 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT, 2292 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT, 2293 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT, 2294 }; 2295 2296 enum rseq_event_mask { 2297 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT), 2298 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT), 2299 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT), 2300 }; 2301 2302 static inline void rseq_set_notify_resume(struct task_struct *t) 2303 { 2304 if (t->rseq) 2305 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME); 2306 } 2307 2308 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs); 2309 2310 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2311 struct pt_regs *regs) 2312 { 2313 if (current->rseq) 2314 __rseq_handle_notify_resume(ksig, regs); 2315 } 2316 2317 static inline void rseq_signal_deliver(struct ksignal *ksig, 2318 struct pt_regs *regs) 2319 { 2320 preempt_disable(); 2321 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask); 2322 preempt_enable(); 2323 rseq_handle_notify_resume(ksig, regs); 2324 } 2325 2326 /* rseq_preempt() requires preemption to be disabled. */ 2327 static inline void rseq_preempt(struct task_struct *t) 2328 { 2329 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask); 2330 rseq_set_notify_resume(t); 2331 } 2332 2333 /* rseq_migrate() requires preemption to be disabled. */ 2334 static inline void rseq_migrate(struct task_struct *t) 2335 { 2336 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask); 2337 rseq_set_notify_resume(t); 2338 } 2339 2340 /* 2341 * If parent process has a registered restartable sequences area, the 2342 * child inherits. Unregister rseq for a clone with CLONE_VM set. 2343 */ 2344 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2345 { 2346 if (clone_flags & CLONE_VM) { 2347 t->rseq = NULL; 2348 t->rseq_sig = 0; 2349 t->rseq_event_mask = 0; 2350 } else { 2351 t->rseq = current->rseq; 2352 t->rseq_sig = current->rseq_sig; 2353 t->rseq_event_mask = current->rseq_event_mask; 2354 } 2355 } 2356 2357 static inline void rseq_execve(struct task_struct *t) 2358 { 2359 t->rseq = NULL; 2360 t->rseq_sig = 0; 2361 t->rseq_event_mask = 0; 2362 } 2363 2364 #else 2365 2366 static inline void rseq_set_notify_resume(struct task_struct *t) 2367 { 2368 } 2369 static inline void rseq_handle_notify_resume(struct ksignal *ksig, 2370 struct pt_regs *regs) 2371 { 2372 } 2373 static inline void rseq_signal_deliver(struct ksignal *ksig, 2374 struct pt_regs *regs) 2375 { 2376 } 2377 static inline void rseq_preempt(struct task_struct *t) 2378 { 2379 } 2380 static inline void rseq_migrate(struct task_struct *t) 2381 { 2382 } 2383 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags) 2384 { 2385 } 2386 static inline void rseq_execve(struct task_struct *t) 2387 { 2388 } 2389 2390 #endif 2391 2392 #ifdef CONFIG_DEBUG_RSEQ 2393 2394 void rseq_syscall(struct pt_regs *regs); 2395 2396 #else 2397 2398 static inline void rseq_syscall(struct pt_regs *regs) 2399 { 2400 } 2401 2402 #endif 2403 2404 #ifdef CONFIG_SCHED_CORE 2405 extern void sched_core_free(struct task_struct *tsk); 2406 extern void sched_core_fork(struct task_struct *p); 2407 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type, 2408 unsigned long uaddr); 2409 #else 2410 static inline void sched_core_free(struct task_struct *tsk) { } 2411 static inline void sched_core_fork(struct task_struct *p) { } 2412 #endif 2413 2414 extern void sched_set_stop_task(int cpu, struct task_struct *stop); 2415 2416 #endif 2417