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/kcov.h> 18 #include <linux/mutex.h> 19 #include <linux/plist.h> 20 #include <linux/hrtimer.h> 21 #include <linux/seccomp.h> 22 #include <linux/nodemask.h> 23 #include <linux/rcupdate.h> 24 #include <linux/resource.h> 25 #include <linux/latencytop.h> 26 #include <linux/sched/prio.h> 27 #include <linux/signal_types.h> 28 #include <linux/mm_types_task.h> 29 #include <linux/task_io_accounting.h> 30 31 /* task_struct member predeclarations (sorted alphabetically): */ 32 struct audit_context; 33 struct backing_dev_info; 34 struct bio_list; 35 struct blk_plug; 36 struct cfs_rq; 37 struct fs_struct; 38 struct futex_pi_state; 39 struct io_context; 40 struct mempolicy; 41 struct nameidata; 42 struct nsproxy; 43 struct perf_event_context; 44 struct pid_namespace; 45 struct pipe_inode_info; 46 struct rcu_node; 47 struct reclaim_state; 48 struct robust_list_head; 49 struct sched_attr; 50 struct sched_param; 51 struct seq_file; 52 struct sighand_struct; 53 struct signal_struct; 54 struct task_delay_info; 55 struct task_group; 56 57 /* 58 * Task state bitmask. NOTE! These bits are also 59 * encoded in fs/proc/array.c: get_task_state(). 60 * 61 * We have two separate sets of flags: task->state 62 * is about runnability, while task->exit_state are 63 * about the task exiting. Confusing, but this way 64 * modifying one set can't modify the other one by 65 * mistake. 66 */ 67 68 /* Used in tsk->state: */ 69 #define TASK_RUNNING 0x0000 70 #define TASK_INTERRUPTIBLE 0x0001 71 #define TASK_UNINTERRUPTIBLE 0x0002 72 #define __TASK_STOPPED 0x0004 73 #define __TASK_TRACED 0x0008 74 /* Used in tsk->exit_state: */ 75 #define EXIT_DEAD 0x0010 76 #define EXIT_ZOMBIE 0x0020 77 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD) 78 /* Used in tsk->state again: */ 79 #define TASK_PARKED 0x0040 80 #define TASK_DEAD 0x0080 81 #define TASK_WAKEKILL 0x0100 82 #define TASK_WAKING 0x0200 83 #define TASK_NOLOAD 0x0400 84 #define TASK_NEW 0x0800 85 #define TASK_STATE_MAX 0x1000 86 87 /* Convenience macros for the sake of set_current_state: */ 88 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 89 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 90 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) 91 92 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD) 93 94 /* Convenience macros for the sake of wake_up(): */ 95 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 96 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED) 97 98 /* get_task_state(): */ 99 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 100 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 101 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \ 102 TASK_PARKED) 103 104 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0) 105 106 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0) 107 108 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) 109 110 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \ 111 (task->flags & PF_FROZEN) == 0 && \ 112 (task->state & TASK_NOLOAD) == 0) 113 114 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 115 116 #define __set_current_state(state_value) \ 117 do { \ 118 current->task_state_change = _THIS_IP_; \ 119 current->state = (state_value); \ 120 } while (0) 121 #define set_current_state(state_value) \ 122 do { \ 123 current->task_state_change = _THIS_IP_; \ 124 smp_store_mb(current->state, (state_value)); \ 125 } while (0) 126 127 #else 128 /* 129 * set_current_state() includes a barrier so that the write of current->state 130 * is correctly serialised wrt the caller's subsequent test of whether to 131 * actually sleep: 132 * 133 * for (;;) { 134 * set_current_state(TASK_UNINTERRUPTIBLE); 135 * if (!need_sleep) 136 * break; 137 * 138 * schedule(); 139 * } 140 * __set_current_state(TASK_RUNNING); 141 * 142 * If the caller does not need such serialisation (because, for instance, the 143 * condition test and condition change and wakeup are under the same lock) then 144 * use __set_current_state(). 145 * 146 * The above is typically ordered against the wakeup, which does: 147 * 148 * need_sleep = false; 149 * wake_up_state(p, TASK_UNINTERRUPTIBLE); 150 * 151 * Where wake_up_state() (and all other wakeup primitives) imply enough 152 * barriers to order the store of the variable against wakeup. 153 * 154 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is, 155 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a 156 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING). 157 * 158 * This is obviously fine, since they both store the exact same value. 159 * 160 * Also see the comments of try_to_wake_up(). 161 */ 162 #define __set_current_state(state_value) do { current->state = (state_value); } while (0) 163 #define set_current_state(state_value) smp_store_mb(current->state, (state_value)) 164 #endif 165 166 /* Task command name length: */ 167 #define TASK_COMM_LEN 16 168 169 extern void scheduler_tick(void); 170 171 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 172 173 extern long schedule_timeout(long timeout); 174 extern long schedule_timeout_interruptible(long timeout); 175 extern long schedule_timeout_killable(long timeout); 176 extern long schedule_timeout_uninterruptible(long timeout); 177 extern long schedule_timeout_idle(long timeout); 178 asmlinkage void schedule(void); 179 extern void schedule_preempt_disabled(void); 180 181 extern int __must_check io_schedule_prepare(void); 182 extern void io_schedule_finish(int token); 183 extern long io_schedule_timeout(long timeout); 184 extern void io_schedule(void); 185 186 /** 187 * struct prev_cputime - snapshot of system and user cputime 188 * @utime: time spent in user mode 189 * @stime: time spent in system mode 190 * @lock: protects the above two fields 191 * 192 * Stores previous user/system time values such that we can guarantee 193 * monotonicity. 194 */ 195 struct prev_cputime { 196 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE 197 u64 utime; 198 u64 stime; 199 raw_spinlock_t lock; 200 #endif 201 }; 202 203 /** 204 * struct task_cputime - collected CPU time counts 205 * @utime: time spent in user mode, in nanoseconds 206 * @stime: time spent in kernel mode, in nanoseconds 207 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds 208 * 209 * This structure groups together three kinds of CPU time that are tracked for 210 * threads and thread groups. Most things considering CPU time want to group 211 * these counts together and treat all three of them in parallel. 212 */ 213 struct task_cputime { 214 u64 utime; 215 u64 stime; 216 unsigned long long sum_exec_runtime; 217 }; 218 219 /* Alternate field names when used on cache expirations: */ 220 #define virt_exp utime 221 #define prof_exp stime 222 #define sched_exp sum_exec_runtime 223 224 enum vtime_state { 225 /* Task is sleeping or running in a CPU with VTIME inactive: */ 226 VTIME_INACTIVE = 0, 227 /* Task runs in userspace in a CPU with VTIME active: */ 228 VTIME_USER, 229 /* Task runs in kernelspace in a CPU with VTIME active: */ 230 VTIME_SYS, 231 }; 232 233 struct vtime { 234 seqcount_t seqcount; 235 unsigned long long starttime; 236 enum vtime_state state; 237 u64 utime; 238 u64 stime; 239 u64 gtime; 240 }; 241 242 struct sched_info { 243 #ifdef CONFIG_SCHED_INFO 244 /* Cumulative counters: */ 245 246 /* # of times we have run on this CPU: */ 247 unsigned long pcount; 248 249 /* Time spent waiting on a runqueue: */ 250 unsigned long long run_delay; 251 252 /* Timestamps: */ 253 254 /* When did we last run on a CPU? */ 255 unsigned long long last_arrival; 256 257 /* When were we last queued to run? */ 258 unsigned long long last_queued; 259 260 #endif /* CONFIG_SCHED_INFO */ 261 }; 262 263 /* 264 * Integer metrics need fixed point arithmetic, e.g., sched/fair 265 * has a few: load, load_avg, util_avg, freq, and capacity. 266 * 267 * We define a basic fixed point arithmetic range, and then formalize 268 * all these metrics based on that basic range. 269 */ 270 # define SCHED_FIXEDPOINT_SHIFT 10 271 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT) 272 273 struct load_weight { 274 unsigned long weight; 275 u32 inv_weight; 276 }; 277 278 /* 279 * The load_avg/util_avg accumulates an infinite geometric series 280 * (see __update_load_avg() in kernel/sched/fair.c). 281 * 282 * [load_avg definition] 283 * 284 * load_avg = runnable% * scale_load_down(load) 285 * 286 * where runnable% is the time ratio that a sched_entity is runnable. 287 * For cfs_rq, it is the aggregated load_avg of all runnable and 288 * blocked sched_entities. 289 * 290 * load_avg may also take frequency scaling into account: 291 * 292 * load_avg = runnable% * scale_load_down(load) * freq% 293 * 294 * where freq% is the CPU frequency normalized to the highest frequency. 295 * 296 * [util_avg definition] 297 * 298 * util_avg = running% * SCHED_CAPACITY_SCALE 299 * 300 * where running% is the time ratio that a sched_entity is running on 301 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable 302 * and blocked sched_entities. 303 * 304 * util_avg may also factor frequency scaling and CPU capacity scaling: 305 * 306 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity% 307 * 308 * where freq% is the same as above, and capacity% is the CPU capacity 309 * normalized to the greatest capacity (due to uarch differences, etc). 310 * 311 * N.B., the above ratios (runnable%, running%, freq%, and capacity%) 312 * themselves are in the range of [0, 1]. To do fixed point arithmetics, 313 * we therefore scale them to as large a range as necessary. This is for 314 * example reflected by util_avg's SCHED_CAPACITY_SCALE. 315 * 316 * [Overflow issue] 317 * 318 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities 319 * with the highest load (=88761), always runnable on a single cfs_rq, 320 * and should not overflow as the number already hits PID_MAX_LIMIT. 321 * 322 * For all other cases (including 32-bit kernels), struct load_weight's 323 * weight will overflow first before we do, because: 324 * 325 * Max(load_avg) <= Max(load.weight) 326 * 327 * Then it is the load_weight's responsibility to consider overflow 328 * issues. 329 */ 330 struct sched_avg { 331 u64 last_update_time; 332 u64 load_sum; 333 u64 runnable_load_sum; 334 u32 util_sum; 335 u32 period_contrib; 336 unsigned long load_avg; 337 unsigned long runnable_load_avg; 338 unsigned long util_avg; 339 }; 340 341 struct sched_statistics { 342 #ifdef CONFIG_SCHEDSTATS 343 u64 wait_start; 344 u64 wait_max; 345 u64 wait_count; 346 u64 wait_sum; 347 u64 iowait_count; 348 u64 iowait_sum; 349 350 u64 sleep_start; 351 u64 sleep_max; 352 s64 sum_sleep_runtime; 353 354 u64 block_start; 355 u64 block_max; 356 u64 exec_max; 357 u64 slice_max; 358 359 u64 nr_migrations_cold; 360 u64 nr_failed_migrations_affine; 361 u64 nr_failed_migrations_running; 362 u64 nr_failed_migrations_hot; 363 u64 nr_forced_migrations; 364 365 u64 nr_wakeups; 366 u64 nr_wakeups_sync; 367 u64 nr_wakeups_migrate; 368 u64 nr_wakeups_local; 369 u64 nr_wakeups_remote; 370 u64 nr_wakeups_affine; 371 u64 nr_wakeups_affine_attempts; 372 u64 nr_wakeups_passive; 373 u64 nr_wakeups_idle; 374 #endif 375 }; 376 377 struct sched_entity { 378 /* For load-balancing: */ 379 struct load_weight load; 380 unsigned long runnable_weight; 381 struct rb_node run_node; 382 struct list_head group_node; 383 unsigned int on_rq; 384 385 u64 exec_start; 386 u64 sum_exec_runtime; 387 u64 vruntime; 388 u64 prev_sum_exec_runtime; 389 390 u64 nr_migrations; 391 392 struct sched_statistics statistics; 393 394 #ifdef CONFIG_FAIR_GROUP_SCHED 395 int depth; 396 struct sched_entity *parent; 397 /* rq on which this entity is (to be) queued: */ 398 struct cfs_rq *cfs_rq; 399 /* rq "owned" by this entity/group: */ 400 struct cfs_rq *my_q; 401 #endif 402 403 #ifdef CONFIG_SMP 404 /* 405 * Per entity load average tracking. 406 * 407 * Put into separate cache line so it does not 408 * collide with read-mostly values above. 409 */ 410 struct sched_avg avg ____cacheline_aligned_in_smp; 411 #endif 412 }; 413 414 struct sched_rt_entity { 415 struct list_head run_list; 416 unsigned long timeout; 417 unsigned long watchdog_stamp; 418 unsigned int time_slice; 419 unsigned short on_rq; 420 unsigned short on_list; 421 422 struct sched_rt_entity *back; 423 #ifdef CONFIG_RT_GROUP_SCHED 424 struct sched_rt_entity *parent; 425 /* rq on which this entity is (to be) queued: */ 426 struct rt_rq *rt_rq; 427 /* rq "owned" by this entity/group: */ 428 struct rt_rq *my_q; 429 #endif 430 } __randomize_layout; 431 432 struct sched_dl_entity { 433 struct rb_node rb_node; 434 435 /* 436 * Original scheduling parameters. Copied here from sched_attr 437 * during sched_setattr(), they will remain the same until 438 * the next sched_setattr(). 439 */ 440 u64 dl_runtime; /* Maximum runtime for each instance */ 441 u64 dl_deadline; /* Relative deadline of each instance */ 442 u64 dl_period; /* Separation of two instances (period) */ 443 u64 dl_bw; /* dl_runtime / dl_period */ 444 u64 dl_density; /* dl_runtime / dl_deadline */ 445 446 /* 447 * Actual scheduling parameters. Initialized with the values above, 448 * they are continously updated during task execution. Note that 449 * the remaining runtime could be < 0 in case we are in overrun. 450 */ 451 s64 runtime; /* Remaining runtime for this instance */ 452 u64 deadline; /* Absolute deadline for this instance */ 453 unsigned int flags; /* Specifying the scheduler behaviour */ 454 455 /* 456 * Some bool flags: 457 * 458 * @dl_throttled tells if we exhausted the runtime. If so, the 459 * task has to wait for a replenishment to be performed at the 460 * next firing of dl_timer. 461 * 462 * @dl_boosted tells if we are boosted due to DI. If so we are 463 * outside bandwidth enforcement mechanism (but only until we 464 * exit the critical section); 465 * 466 * @dl_yielded tells if task gave up the CPU before consuming 467 * all its available runtime during the last job. 468 * 469 * @dl_non_contending tells if the task is inactive while still 470 * contributing to the active utilization. In other words, it 471 * indicates if the inactive timer has been armed and its handler 472 * has not been executed yet. This flag is useful to avoid race 473 * conditions between the inactive timer handler and the wakeup 474 * code. 475 * 476 * @dl_overrun tells if the task asked to be informed about runtime 477 * overruns. 478 */ 479 unsigned int dl_throttled : 1; 480 unsigned int dl_boosted : 1; 481 unsigned int dl_yielded : 1; 482 unsigned int dl_non_contending : 1; 483 unsigned int dl_overrun : 1; 484 485 /* 486 * Bandwidth enforcement timer. Each -deadline task has its 487 * own bandwidth to be enforced, thus we need one timer per task. 488 */ 489 struct hrtimer dl_timer; 490 491 /* 492 * Inactive timer, responsible for decreasing the active utilization 493 * at the "0-lag time". When a -deadline task blocks, it contributes 494 * to GRUB's active utilization until the "0-lag time", hence a 495 * timer is needed to decrease the active utilization at the correct 496 * time. 497 */ 498 struct hrtimer inactive_timer; 499 }; 500 501 union rcu_special { 502 struct { 503 u8 blocked; 504 u8 need_qs; 505 u8 exp_need_qs; 506 507 /* Otherwise the compiler can store garbage here: */ 508 u8 pad; 509 } b; /* Bits. */ 510 u32 s; /* Set of bits. */ 511 }; 512 513 enum perf_event_task_context { 514 perf_invalid_context = -1, 515 perf_hw_context = 0, 516 perf_sw_context, 517 perf_nr_task_contexts, 518 }; 519 520 struct wake_q_node { 521 struct wake_q_node *next; 522 }; 523 524 struct task_struct { 525 #ifdef CONFIG_THREAD_INFO_IN_TASK 526 /* 527 * For reasons of header soup (see current_thread_info()), this 528 * must be the first element of task_struct. 529 */ 530 struct thread_info thread_info; 531 #endif 532 /* -1 unrunnable, 0 runnable, >0 stopped: */ 533 volatile long state; 534 535 /* 536 * This begins the randomizable portion of task_struct. Only 537 * scheduling-critical items should be added above here. 538 */ 539 randomized_struct_fields_start 540 541 void *stack; 542 atomic_t usage; 543 /* Per task flags (PF_*), defined further below: */ 544 unsigned int flags; 545 unsigned int ptrace; 546 547 #ifdef CONFIG_SMP 548 struct llist_node wake_entry; 549 int on_cpu; 550 #ifdef CONFIG_THREAD_INFO_IN_TASK 551 /* Current CPU: */ 552 unsigned int cpu; 553 #endif 554 unsigned int wakee_flips; 555 unsigned long wakee_flip_decay_ts; 556 struct task_struct *last_wakee; 557 558 int wake_cpu; 559 #endif 560 int on_rq; 561 562 int prio; 563 int static_prio; 564 int normal_prio; 565 unsigned int rt_priority; 566 567 const struct sched_class *sched_class; 568 struct sched_entity se; 569 struct sched_rt_entity rt; 570 #ifdef CONFIG_CGROUP_SCHED 571 struct task_group *sched_task_group; 572 #endif 573 struct sched_dl_entity dl; 574 575 #ifdef CONFIG_PREEMPT_NOTIFIERS 576 /* List of struct preempt_notifier: */ 577 struct hlist_head preempt_notifiers; 578 #endif 579 580 #ifdef CONFIG_BLK_DEV_IO_TRACE 581 unsigned int btrace_seq; 582 #endif 583 584 unsigned int policy; 585 int nr_cpus_allowed; 586 cpumask_t cpus_allowed; 587 588 #ifdef CONFIG_PREEMPT_RCU 589 int rcu_read_lock_nesting; 590 union rcu_special rcu_read_unlock_special; 591 struct list_head rcu_node_entry; 592 struct rcu_node *rcu_blocked_node; 593 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 594 595 #ifdef CONFIG_TASKS_RCU 596 unsigned long rcu_tasks_nvcsw; 597 u8 rcu_tasks_holdout; 598 u8 rcu_tasks_idx; 599 int rcu_tasks_idle_cpu; 600 struct list_head rcu_tasks_holdout_list; 601 #endif /* #ifdef CONFIG_TASKS_RCU */ 602 603 struct sched_info sched_info; 604 605 struct list_head tasks; 606 #ifdef CONFIG_SMP 607 struct plist_node pushable_tasks; 608 struct rb_node pushable_dl_tasks; 609 #endif 610 611 struct mm_struct *mm; 612 struct mm_struct *active_mm; 613 614 /* Per-thread vma caching: */ 615 struct vmacache vmacache; 616 617 #ifdef SPLIT_RSS_COUNTING 618 struct task_rss_stat rss_stat; 619 #endif 620 int exit_state; 621 int exit_code; 622 int exit_signal; 623 /* The signal sent when the parent dies: */ 624 int pdeath_signal; 625 /* JOBCTL_*, siglock protected: */ 626 unsigned long jobctl; 627 628 /* Used for emulating ABI behavior of previous Linux versions: */ 629 unsigned int personality; 630 631 /* Scheduler bits, serialized by scheduler locks: */ 632 unsigned sched_reset_on_fork:1; 633 unsigned sched_contributes_to_load:1; 634 unsigned sched_migrated:1; 635 unsigned sched_remote_wakeup:1; 636 /* Force alignment to the next boundary: */ 637 unsigned :0; 638 639 /* Unserialized, strictly 'current' */ 640 641 /* Bit to tell LSMs we're in execve(): */ 642 unsigned in_execve:1; 643 unsigned in_iowait:1; 644 #ifndef TIF_RESTORE_SIGMASK 645 unsigned restore_sigmask:1; 646 #endif 647 #ifdef CONFIG_MEMCG 648 unsigned memcg_may_oom:1; 649 #ifndef CONFIG_SLOB 650 unsigned memcg_kmem_skip_account:1; 651 #endif 652 #endif 653 #ifdef CONFIG_COMPAT_BRK 654 unsigned brk_randomized:1; 655 #endif 656 #ifdef CONFIG_CGROUPS 657 /* disallow userland-initiated cgroup migration */ 658 unsigned no_cgroup_migration:1; 659 #endif 660 661 unsigned long atomic_flags; /* Flags requiring atomic access. */ 662 663 struct restart_block restart_block; 664 665 pid_t pid; 666 pid_t tgid; 667 668 #ifdef CONFIG_CC_STACKPROTECTOR 669 /* Canary value for the -fstack-protector GCC feature: */ 670 unsigned long stack_canary; 671 #endif 672 /* 673 * Pointers to the (original) parent process, youngest child, younger sibling, 674 * older sibling, respectively. (p->father can be replaced with 675 * p->real_parent->pid) 676 */ 677 678 /* Real parent process: */ 679 struct task_struct __rcu *real_parent; 680 681 /* Recipient of SIGCHLD, wait4() reports: */ 682 struct task_struct __rcu *parent; 683 684 /* 685 * Children/sibling form the list of natural children: 686 */ 687 struct list_head children; 688 struct list_head sibling; 689 struct task_struct *group_leader; 690 691 /* 692 * 'ptraced' is the list of tasks this task is using ptrace() on. 693 * 694 * This includes both natural children and PTRACE_ATTACH targets. 695 * 'ptrace_entry' is this task's link on the p->parent->ptraced list. 696 */ 697 struct list_head ptraced; 698 struct list_head ptrace_entry; 699 700 /* PID/PID hash table linkage. */ 701 struct pid_link pids[PIDTYPE_MAX]; 702 struct list_head thread_group; 703 struct list_head thread_node; 704 705 struct completion *vfork_done; 706 707 /* CLONE_CHILD_SETTID: */ 708 int __user *set_child_tid; 709 710 /* CLONE_CHILD_CLEARTID: */ 711 int __user *clear_child_tid; 712 713 u64 utime; 714 u64 stime; 715 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME 716 u64 utimescaled; 717 u64 stimescaled; 718 #endif 719 u64 gtime; 720 struct prev_cputime prev_cputime; 721 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN 722 struct vtime vtime; 723 #endif 724 725 #ifdef CONFIG_NO_HZ_FULL 726 atomic_t tick_dep_mask; 727 #endif 728 /* Context switch counts: */ 729 unsigned long nvcsw; 730 unsigned long nivcsw; 731 732 /* Monotonic time in nsecs: */ 733 u64 start_time; 734 735 /* Boot based time in nsecs: */ 736 u64 real_start_time; 737 738 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */ 739 unsigned long min_flt; 740 unsigned long maj_flt; 741 742 #ifdef CONFIG_POSIX_TIMERS 743 struct task_cputime cputime_expires; 744 struct list_head cpu_timers[3]; 745 #endif 746 747 /* Process credentials: */ 748 749 /* Tracer's credentials at attach: */ 750 const struct cred __rcu *ptracer_cred; 751 752 /* Objective and real subjective task credentials (COW): */ 753 const struct cred __rcu *real_cred; 754 755 /* Effective (overridable) subjective task credentials (COW): */ 756 const struct cred __rcu *cred; 757 758 /* 759 * executable name, excluding path. 760 * 761 * - normally initialized setup_new_exec() 762 * - access it with [gs]et_task_comm() 763 * - lock it with task_lock() 764 */ 765 char comm[TASK_COMM_LEN]; 766 767 struct nameidata *nameidata; 768 769 #ifdef CONFIG_SYSVIPC 770 struct sysv_sem sysvsem; 771 struct sysv_shm sysvshm; 772 #endif 773 #ifdef CONFIG_DETECT_HUNG_TASK 774 unsigned long last_switch_count; 775 #endif 776 /* Filesystem information: */ 777 struct fs_struct *fs; 778 779 /* Open file information: */ 780 struct files_struct *files; 781 782 /* Namespaces: */ 783 struct nsproxy *nsproxy; 784 785 /* Signal handlers: */ 786 struct signal_struct *signal; 787 struct sighand_struct *sighand; 788 sigset_t blocked; 789 sigset_t real_blocked; 790 /* Restored if set_restore_sigmask() was used: */ 791 sigset_t saved_sigmask; 792 struct sigpending pending; 793 unsigned long sas_ss_sp; 794 size_t sas_ss_size; 795 unsigned int sas_ss_flags; 796 797 struct callback_head *task_works; 798 799 struct audit_context *audit_context; 800 #ifdef CONFIG_AUDITSYSCALL 801 kuid_t loginuid; 802 unsigned int sessionid; 803 #endif 804 struct seccomp seccomp; 805 806 /* Thread group tracking: */ 807 u32 parent_exec_id; 808 u32 self_exec_id; 809 810 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */ 811 spinlock_t alloc_lock; 812 813 /* Protection of the PI data structures: */ 814 raw_spinlock_t pi_lock; 815 816 struct wake_q_node wake_q; 817 818 #ifdef CONFIG_RT_MUTEXES 819 /* PI waiters blocked on a rt_mutex held by this task: */ 820 struct rb_root_cached pi_waiters; 821 /* Updated under owner's pi_lock and rq lock */ 822 struct task_struct *pi_top_task; 823 /* Deadlock detection and priority inheritance handling: */ 824 struct rt_mutex_waiter *pi_blocked_on; 825 #endif 826 827 #ifdef CONFIG_DEBUG_MUTEXES 828 /* Mutex deadlock detection: */ 829 struct mutex_waiter *blocked_on; 830 #endif 831 832 #ifdef CONFIG_TRACE_IRQFLAGS 833 unsigned int irq_events; 834 unsigned long hardirq_enable_ip; 835 unsigned long hardirq_disable_ip; 836 unsigned int hardirq_enable_event; 837 unsigned int hardirq_disable_event; 838 int hardirqs_enabled; 839 int hardirq_context; 840 unsigned long softirq_disable_ip; 841 unsigned long softirq_enable_ip; 842 unsigned int softirq_disable_event; 843 unsigned int softirq_enable_event; 844 int softirqs_enabled; 845 int softirq_context; 846 #endif 847 848 #ifdef CONFIG_LOCKDEP 849 # define MAX_LOCK_DEPTH 48UL 850 u64 curr_chain_key; 851 int lockdep_depth; 852 unsigned int lockdep_recursion; 853 struct held_lock held_locks[MAX_LOCK_DEPTH]; 854 #endif 855 856 #ifdef CONFIG_UBSAN 857 unsigned int in_ubsan; 858 #endif 859 860 /* Journalling filesystem info: */ 861 void *journal_info; 862 863 /* Stacked block device info: */ 864 struct bio_list *bio_list; 865 866 #ifdef CONFIG_BLOCK 867 /* Stack plugging: */ 868 struct blk_plug *plug; 869 #endif 870 871 /* VM state: */ 872 struct reclaim_state *reclaim_state; 873 874 struct backing_dev_info *backing_dev_info; 875 876 struct io_context *io_context; 877 878 /* Ptrace state: */ 879 unsigned long ptrace_message; 880 siginfo_t *last_siginfo; 881 882 struct task_io_accounting ioac; 883 #ifdef CONFIG_TASK_XACCT 884 /* Accumulated RSS usage: */ 885 u64 acct_rss_mem1; 886 /* Accumulated virtual memory usage: */ 887 u64 acct_vm_mem1; 888 /* stime + utime since last update: */ 889 u64 acct_timexpd; 890 #endif 891 #ifdef CONFIG_CPUSETS 892 /* Protected by ->alloc_lock: */ 893 nodemask_t mems_allowed; 894 /* Seqence number to catch updates: */ 895 seqcount_t mems_allowed_seq; 896 int cpuset_mem_spread_rotor; 897 int cpuset_slab_spread_rotor; 898 #endif 899 #ifdef CONFIG_CGROUPS 900 /* Control Group info protected by css_set_lock: */ 901 struct css_set __rcu *cgroups; 902 /* cg_list protected by css_set_lock and tsk->alloc_lock: */ 903 struct list_head cg_list; 904 #endif 905 #ifdef CONFIG_INTEL_RDT 906 u32 closid; 907 u32 rmid; 908 #endif 909 #ifdef CONFIG_FUTEX 910 struct robust_list_head __user *robust_list; 911 #ifdef CONFIG_COMPAT 912 struct compat_robust_list_head __user *compat_robust_list; 913 #endif 914 struct list_head pi_state_list; 915 struct futex_pi_state *pi_state_cache; 916 #endif 917 #ifdef CONFIG_PERF_EVENTS 918 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 919 struct mutex perf_event_mutex; 920 struct list_head perf_event_list; 921 #endif 922 #ifdef CONFIG_DEBUG_PREEMPT 923 unsigned long preempt_disable_ip; 924 #endif 925 #ifdef CONFIG_NUMA 926 /* Protected by alloc_lock: */ 927 struct mempolicy *mempolicy; 928 short il_prev; 929 short pref_node_fork; 930 #endif 931 #ifdef CONFIG_NUMA_BALANCING 932 int numa_scan_seq; 933 unsigned int numa_scan_period; 934 unsigned int numa_scan_period_max; 935 int numa_preferred_nid; 936 unsigned long numa_migrate_retry; 937 /* Migration stamp: */ 938 u64 node_stamp; 939 u64 last_task_numa_placement; 940 u64 last_sum_exec_runtime; 941 struct callback_head numa_work; 942 943 struct list_head numa_entry; 944 struct numa_group *numa_group; 945 946 /* 947 * numa_faults is an array split into four regions: 948 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer 949 * in this precise order. 950 * 951 * faults_memory: Exponential decaying average of faults on a per-node 952 * basis. Scheduling placement decisions are made based on these 953 * counts. The values remain static for the duration of a PTE scan. 954 * faults_cpu: Track the nodes the process was running on when a NUMA 955 * hinting fault was incurred. 956 * faults_memory_buffer and faults_cpu_buffer: Record faults per node 957 * during the current scan window. When the scan completes, the counts 958 * in faults_memory and faults_cpu decay and these values are copied. 959 */ 960 unsigned long *numa_faults; 961 unsigned long total_numa_faults; 962 963 /* 964 * numa_faults_locality tracks if faults recorded during the last 965 * scan window were remote/local or failed to migrate. The task scan 966 * period is adapted based on the locality of the faults with different 967 * weights depending on whether they were shared or private faults 968 */ 969 unsigned long numa_faults_locality[3]; 970 971 unsigned long numa_pages_migrated; 972 #endif /* CONFIG_NUMA_BALANCING */ 973 974 struct tlbflush_unmap_batch tlb_ubc; 975 976 struct rcu_head rcu; 977 978 /* Cache last used pipe for splice(): */ 979 struct pipe_inode_info *splice_pipe; 980 981 struct page_frag task_frag; 982 983 #ifdef CONFIG_TASK_DELAY_ACCT 984 struct task_delay_info *delays; 985 #endif 986 987 #ifdef CONFIG_FAULT_INJECTION 988 int make_it_fail; 989 unsigned int fail_nth; 990 #endif 991 /* 992 * When (nr_dirtied >= nr_dirtied_pause), it's time to call 993 * balance_dirty_pages() for a dirty throttling pause: 994 */ 995 int nr_dirtied; 996 int nr_dirtied_pause; 997 /* Start of a write-and-pause period: */ 998 unsigned long dirty_paused_when; 999 1000 #ifdef CONFIG_LATENCYTOP 1001 int latency_record_count; 1002 struct latency_record latency_record[LT_SAVECOUNT]; 1003 #endif 1004 /* 1005 * Time slack values; these are used to round up poll() and 1006 * select() etc timeout values. These are in nanoseconds. 1007 */ 1008 u64 timer_slack_ns; 1009 u64 default_timer_slack_ns; 1010 1011 #ifdef CONFIG_KASAN 1012 unsigned int kasan_depth; 1013 #endif 1014 1015 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1016 /* Index of current stored address in ret_stack: */ 1017 int curr_ret_stack; 1018 1019 /* Stack of return addresses for return function tracing: */ 1020 struct ftrace_ret_stack *ret_stack; 1021 1022 /* Timestamp for last schedule: */ 1023 unsigned long long ftrace_timestamp; 1024 1025 /* 1026 * Number of functions that haven't been traced 1027 * because of depth overrun: 1028 */ 1029 atomic_t trace_overrun; 1030 1031 /* Pause tracing: */ 1032 atomic_t tracing_graph_pause; 1033 #endif 1034 1035 #ifdef CONFIG_TRACING 1036 /* State flags for use by tracers: */ 1037 unsigned long trace; 1038 1039 /* Bitmask and counter of trace recursion: */ 1040 unsigned long trace_recursion; 1041 #endif /* CONFIG_TRACING */ 1042 1043 #ifdef CONFIG_KCOV 1044 /* Coverage collection mode enabled for this task (0 if disabled): */ 1045 enum kcov_mode kcov_mode; 1046 1047 /* Size of the kcov_area: */ 1048 unsigned int kcov_size; 1049 1050 /* Buffer for coverage collection: */ 1051 void *kcov_area; 1052 1053 /* KCOV descriptor wired with this task or NULL: */ 1054 struct kcov *kcov; 1055 #endif 1056 1057 #ifdef CONFIG_MEMCG 1058 struct mem_cgroup *memcg_in_oom; 1059 gfp_t memcg_oom_gfp_mask; 1060 int memcg_oom_order; 1061 1062 /* Number of pages to reclaim on returning to userland: */ 1063 unsigned int memcg_nr_pages_over_high; 1064 #endif 1065 1066 #ifdef CONFIG_UPROBES 1067 struct uprobe_task *utask; 1068 #endif 1069 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE) 1070 unsigned int sequential_io; 1071 unsigned int sequential_io_avg; 1072 #endif 1073 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP 1074 unsigned long task_state_change; 1075 #endif 1076 int pagefault_disabled; 1077 #ifdef CONFIG_MMU 1078 struct task_struct *oom_reaper_list; 1079 #endif 1080 #ifdef CONFIG_VMAP_STACK 1081 struct vm_struct *stack_vm_area; 1082 #endif 1083 #ifdef CONFIG_THREAD_INFO_IN_TASK 1084 /* A live task holds one reference: */ 1085 atomic_t stack_refcount; 1086 #endif 1087 #ifdef CONFIG_LIVEPATCH 1088 int patch_state; 1089 #endif 1090 #ifdef CONFIG_SECURITY 1091 /* Used by LSM modules for access restriction: */ 1092 void *security; 1093 #endif 1094 1095 /* 1096 * New fields for task_struct should be added above here, so that 1097 * they are included in the randomized portion of task_struct. 1098 */ 1099 randomized_struct_fields_end 1100 1101 /* CPU-specific state of this task: */ 1102 struct thread_struct thread; 1103 1104 /* 1105 * WARNING: on x86, 'thread_struct' contains a variable-sized 1106 * structure. It *MUST* be at the end of 'task_struct'. 1107 * 1108 * Do not put anything below here! 1109 */ 1110 }; 1111 1112 static inline struct pid *task_pid(struct task_struct *task) 1113 { 1114 return task->pids[PIDTYPE_PID].pid; 1115 } 1116 1117 static inline struct pid *task_tgid(struct task_struct *task) 1118 { 1119 return task->group_leader->pids[PIDTYPE_PID].pid; 1120 } 1121 1122 /* 1123 * Without tasklist or RCU lock it is not safe to dereference 1124 * the result of task_pgrp/task_session even if task == current, 1125 * we can race with another thread doing sys_setsid/sys_setpgid. 1126 */ 1127 static inline struct pid *task_pgrp(struct task_struct *task) 1128 { 1129 return task->group_leader->pids[PIDTYPE_PGID].pid; 1130 } 1131 1132 static inline struct pid *task_session(struct task_struct *task) 1133 { 1134 return task->group_leader->pids[PIDTYPE_SID].pid; 1135 } 1136 1137 /* 1138 * the helpers to get the task's different pids as they are seen 1139 * from various namespaces 1140 * 1141 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1142 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1143 * current. 1144 * task_xid_nr_ns() : id seen from the ns specified; 1145 * 1146 * see also pid_nr() etc in include/linux/pid.h 1147 */ 1148 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns); 1149 1150 static inline pid_t task_pid_nr(struct task_struct *tsk) 1151 { 1152 return tsk->pid; 1153 } 1154 1155 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1156 { 1157 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1158 } 1159 1160 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1161 { 1162 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1163 } 1164 1165 1166 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1167 { 1168 return tsk->tgid; 1169 } 1170 1171 /** 1172 * pid_alive - check that a task structure is not stale 1173 * @p: Task structure to be checked. 1174 * 1175 * Test if a process is not yet dead (at most zombie state) 1176 * If pid_alive fails, then pointers within the task structure 1177 * can be stale and must not be dereferenced. 1178 * 1179 * Return: 1 if the process is alive. 0 otherwise. 1180 */ 1181 static inline int pid_alive(const struct task_struct *p) 1182 { 1183 return p->pids[PIDTYPE_PID].pid != NULL; 1184 } 1185 1186 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1187 { 1188 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1189 } 1190 1191 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1192 { 1193 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1194 } 1195 1196 1197 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1198 { 1199 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1200 } 1201 1202 static inline pid_t task_session_vnr(struct task_struct *tsk) 1203 { 1204 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1205 } 1206 1207 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) 1208 { 1209 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns); 1210 } 1211 1212 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1213 { 1214 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL); 1215 } 1216 1217 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns) 1218 { 1219 pid_t pid = 0; 1220 1221 rcu_read_lock(); 1222 if (pid_alive(tsk)) 1223 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns); 1224 rcu_read_unlock(); 1225 1226 return pid; 1227 } 1228 1229 static inline pid_t task_ppid_nr(const struct task_struct *tsk) 1230 { 1231 return task_ppid_nr_ns(tsk, &init_pid_ns); 1232 } 1233 1234 /* Obsolete, do not use: */ 1235 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1236 { 1237 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1238 } 1239 1240 #define TASK_REPORT_IDLE (TASK_REPORT + 1) 1241 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1) 1242 1243 static inline unsigned int task_state_index(struct task_struct *tsk) 1244 { 1245 unsigned int tsk_state = READ_ONCE(tsk->state); 1246 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT; 1247 1248 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX); 1249 1250 if (tsk_state == TASK_IDLE) 1251 state = TASK_REPORT_IDLE; 1252 1253 return fls(state); 1254 } 1255 1256 static inline char task_index_to_char(unsigned int state) 1257 { 1258 static const char state_char[] = "RSDTtXZPI"; 1259 1260 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1); 1261 1262 return state_char[state]; 1263 } 1264 1265 static inline char task_state_to_char(struct task_struct *tsk) 1266 { 1267 return task_index_to_char(task_state_index(tsk)); 1268 } 1269 1270 /** 1271 * is_global_init - check if a task structure is init. Since init 1272 * is free to have sub-threads we need to check tgid. 1273 * @tsk: Task structure to be checked. 1274 * 1275 * Check if a task structure is the first user space task the kernel created. 1276 * 1277 * Return: 1 if the task structure is init. 0 otherwise. 1278 */ 1279 static inline int is_global_init(struct task_struct *tsk) 1280 { 1281 return task_tgid_nr(tsk) == 1; 1282 } 1283 1284 extern struct pid *cad_pid; 1285 1286 /* 1287 * Per process flags 1288 */ 1289 #define PF_IDLE 0x00000002 /* I am an IDLE thread */ 1290 #define PF_EXITING 0x00000004 /* Getting shut down */ 1291 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */ 1292 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */ 1293 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1294 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */ 1295 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */ 1296 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */ 1297 #define PF_DUMPCORE 0x00000200 /* Dumped core */ 1298 #define PF_SIGNALED 0x00000400 /* Killed by a signal */ 1299 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1300 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */ 1301 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */ 1302 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */ 1303 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */ 1304 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */ 1305 #define PF_KSWAPD 0x00020000 /* I am kswapd */ 1306 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */ 1307 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */ 1308 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ 1309 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1310 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */ 1311 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1312 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */ 1313 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1314 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */ 1315 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1316 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */ 1317 1318 /* 1319 * Only the _current_ task can read/write to tsk->flags, but other 1320 * tasks can access tsk->flags in readonly mode for example 1321 * with tsk_used_math (like during threaded core dumping). 1322 * There is however an exception to this rule during ptrace 1323 * or during fork: the ptracer task is allowed to write to the 1324 * child->flags of its traced child (same goes for fork, the parent 1325 * can write to the child->flags), because we're guaranteed the 1326 * child is not running and in turn not changing child->flags 1327 * at the same time the parent does it. 1328 */ 1329 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1330 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1331 #define clear_used_math() clear_stopped_child_used_math(current) 1332 #define set_used_math() set_stopped_child_used_math(current) 1333 1334 #define conditional_stopped_child_used_math(condition, child) \ 1335 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1336 1337 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current) 1338 1339 #define copy_to_stopped_child_used_math(child) \ 1340 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1341 1342 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1343 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1344 #define used_math() tsk_used_math(current) 1345 1346 static inline bool is_percpu_thread(void) 1347 { 1348 #ifdef CONFIG_SMP 1349 return (current->flags & PF_NO_SETAFFINITY) && 1350 (current->nr_cpus_allowed == 1); 1351 #else 1352 return true; 1353 #endif 1354 } 1355 1356 /* Per-process atomic flags. */ 1357 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ 1358 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ 1359 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ 1360 1361 1362 #define TASK_PFA_TEST(name, func) \ 1363 static inline bool task_##func(struct task_struct *p) \ 1364 { return test_bit(PFA_##name, &p->atomic_flags); } 1365 1366 #define TASK_PFA_SET(name, func) \ 1367 static inline void task_set_##func(struct task_struct *p) \ 1368 { set_bit(PFA_##name, &p->atomic_flags); } 1369 1370 #define TASK_PFA_CLEAR(name, func) \ 1371 static inline void task_clear_##func(struct task_struct *p) \ 1372 { clear_bit(PFA_##name, &p->atomic_flags); } 1373 1374 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs) 1375 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs) 1376 1377 TASK_PFA_TEST(SPREAD_PAGE, spread_page) 1378 TASK_PFA_SET(SPREAD_PAGE, spread_page) 1379 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page) 1380 1381 TASK_PFA_TEST(SPREAD_SLAB, spread_slab) 1382 TASK_PFA_SET(SPREAD_SLAB, spread_slab) 1383 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) 1384 1385 static inline void 1386 current_restore_flags(unsigned long orig_flags, unsigned long flags) 1387 { 1388 current->flags &= ~flags; 1389 current->flags |= orig_flags & flags; 1390 } 1391 1392 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial); 1393 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed); 1394 #ifdef CONFIG_SMP 1395 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask); 1396 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask); 1397 #else 1398 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask) 1399 { 1400 } 1401 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) 1402 { 1403 if (!cpumask_test_cpu(0, new_mask)) 1404 return -EINVAL; 1405 return 0; 1406 } 1407 #endif 1408 1409 #ifndef cpu_relax_yield 1410 #define cpu_relax_yield() cpu_relax() 1411 #endif 1412 1413 extern int yield_to(struct task_struct *p, bool preempt); 1414 extern void set_user_nice(struct task_struct *p, long nice); 1415 extern int task_prio(const struct task_struct *p); 1416 1417 /** 1418 * task_nice - return the nice value of a given task. 1419 * @p: the task in question. 1420 * 1421 * Return: The nice value [ -20 ... 0 ... 19 ]. 1422 */ 1423 static inline int task_nice(const struct task_struct *p) 1424 { 1425 return PRIO_TO_NICE((p)->static_prio); 1426 } 1427 1428 extern int can_nice(const struct task_struct *p, const int nice); 1429 extern int task_curr(const struct task_struct *p); 1430 extern int idle_cpu(int cpu); 1431 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *); 1432 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *); 1433 extern int sched_setattr(struct task_struct *, const struct sched_attr *); 1434 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *); 1435 extern struct task_struct *idle_task(int cpu); 1436 1437 /** 1438 * is_idle_task - is the specified task an idle task? 1439 * @p: the task in question. 1440 * 1441 * Return: 1 if @p is an idle task. 0 otherwise. 1442 */ 1443 static inline bool is_idle_task(const struct task_struct *p) 1444 { 1445 return !!(p->flags & PF_IDLE); 1446 } 1447 1448 extern struct task_struct *curr_task(int cpu); 1449 extern void ia64_set_curr_task(int cpu, struct task_struct *p); 1450 1451 void yield(void); 1452 1453 union thread_union { 1454 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK 1455 struct task_struct task; 1456 #endif 1457 #ifndef CONFIG_THREAD_INFO_IN_TASK 1458 struct thread_info thread_info; 1459 #endif 1460 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1461 }; 1462 1463 #ifndef CONFIG_THREAD_INFO_IN_TASK 1464 extern struct thread_info init_thread_info; 1465 #endif 1466 1467 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)]; 1468 1469 #ifdef CONFIG_THREAD_INFO_IN_TASK 1470 static inline struct thread_info *task_thread_info(struct task_struct *task) 1471 { 1472 return &task->thread_info; 1473 } 1474 #elif !defined(__HAVE_THREAD_FUNCTIONS) 1475 # define task_thread_info(task) ((struct thread_info *)(task)->stack) 1476 #endif 1477 1478 /* 1479 * find a task by one of its numerical ids 1480 * 1481 * find_task_by_pid_ns(): 1482 * finds a task by its pid in the specified namespace 1483 * find_task_by_vpid(): 1484 * finds a task by its virtual pid 1485 * 1486 * see also find_vpid() etc in include/linux/pid.h 1487 */ 1488 1489 extern struct task_struct *find_task_by_vpid(pid_t nr); 1490 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); 1491 1492 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1493 extern int wake_up_process(struct task_struct *tsk); 1494 extern void wake_up_new_task(struct task_struct *tsk); 1495 1496 #ifdef CONFIG_SMP 1497 extern void kick_process(struct task_struct *tsk); 1498 #else 1499 static inline void kick_process(struct task_struct *tsk) { } 1500 #endif 1501 1502 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec); 1503 1504 static inline void set_task_comm(struct task_struct *tsk, const char *from) 1505 { 1506 __set_task_comm(tsk, from, false); 1507 } 1508 1509 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk); 1510 #define get_task_comm(buf, tsk) ({ \ 1511 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \ 1512 __get_task_comm(buf, sizeof(buf), tsk); \ 1513 }) 1514 1515 #ifdef CONFIG_SMP 1516 void scheduler_ipi(void); 1517 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 1518 #else 1519 static inline void scheduler_ipi(void) { } 1520 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) 1521 { 1522 return 1; 1523 } 1524 #endif 1525 1526 /* 1527 * Set thread flags in other task's structures. 1528 * See asm/thread_info.h for TIF_xxxx flags available: 1529 */ 1530 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1531 { 1532 set_ti_thread_flag(task_thread_info(tsk), flag); 1533 } 1534 1535 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1536 { 1537 clear_ti_thread_flag(task_thread_info(tsk), flag); 1538 } 1539 1540 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1541 { 1542 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1543 } 1544 1545 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1546 { 1547 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1548 } 1549 1550 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1551 { 1552 return test_ti_thread_flag(task_thread_info(tsk), flag); 1553 } 1554 1555 static inline void set_tsk_need_resched(struct task_struct *tsk) 1556 { 1557 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1558 } 1559 1560 static inline void clear_tsk_need_resched(struct task_struct *tsk) 1561 { 1562 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1563 } 1564 1565 static inline int test_tsk_need_resched(struct task_struct *tsk) 1566 { 1567 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 1568 } 1569 1570 /* 1571 * cond_resched() and cond_resched_lock(): latency reduction via 1572 * explicit rescheduling in places that are safe. The return 1573 * value indicates whether a reschedule was done in fact. 1574 * cond_resched_lock() will drop the spinlock before scheduling, 1575 * cond_resched_softirq() will enable bhs before scheduling. 1576 */ 1577 #ifndef CONFIG_PREEMPT 1578 extern int _cond_resched(void); 1579 #else 1580 static inline int _cond_resched(void) { return 0; } 1581 #endif 1582 1583 #define cond_resched() ({ \ 1584 ___might_sleep(__FILE__, __LINE__, 0); \ 1585 _cond_resched(); \ 1586 }) 1587 1588 extern int __cond_resched_lock(spinlock_t *lock); 1589 1590 #define cond_resched_lock(lock) ({ \ 1591 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\ 1592 __cond_resched_lock(lock); \ 1593 }) 1594 1595 extern int __cond_resched_softirq(void); 1596 1597 #define cond_resched_softirq() ({ \ 1598 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \ 1599 __cond_resched_softirq(); \ 1600 }) 1601 1602 static inline void cond_resched_rcu(void) 1603 { 1604 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU) 1605 rcu_read_unlock(); 1606 cond_resched(); 1607 rcu_read_lock(); 1608 #endif 1609 } 1610 1611 /* 1612 * Does a critical section need to be broken due to another 1613 * task waiting?: (technically does not depend on CONFIG_PREEMPT, 1614 * but a general need for low latency) 1615 */ 1616 static inline int spin_needbreak(spinlock_t *lock) 1617 { 1618 #ifdef CONFIG_PREEMPT 1619 return spin_is_contended(lock); 1620 #else 1621 return 0; 1622 #endif 1623 } 1624 1625 static __always_inline bool need_resched(void) 1626 { 1627 return unlikely(tif_need_resched()); 1628 } 1629 1630 /* 1631 * Wrappers for p->thread_info->cpu access. No-op on UP. 1632 */ 1633 #ifdef CONFIG_SMP 1634 1635 static inline unsigned int task_cpu(const struct task_struct *p) 1636 { 1637 #ifdef CONFIG_THREAD_INFO_IN_TASK 1638 return p->cpu; 1639 #else 1640 return task_thread_info(p)->cpu; 1641 #endif 1642 } 1643 1644 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 1645 1646 #else 1647 1648 static inline unsigned int task_cpu(const struct task_struct *p) 1649 { 1650 return 0; 1651 } 1652 1653 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 1654 { 1655 } 1656 1657 #endif /* CONFIG_SMP */ 1658 1659 /* 1660 * In order to reduce various lock holder preemption latencies provide an 1661 * interface to see if a vCPU is currently running or not. 1662 * 1663 * This allows us to terminate optimistic spin loops and block, analogous to 1664 * the native optimistic spin heuristic of testing if the lock owner task is 1665 * running or not. 1666 */ 1667 #ifndef vcpu_is_preempted 1668 # define vcpu_is_preempted(cpu) false 1669 #endif 1670 1671 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 1672 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 1673 1674 #ifndef TASK_SIZE_OF 1675 #define TASK_SIZE_OF(tsk) TASK_SIZE 1676 #endif 1677 1678 #endif 1679