sched.h revision 669c55e9f99b90e46eaa0f98a67ec53d46dc969a
1#ifndef _LINUX_SCHED_H 2#define _LINUX_SCHED_H 3 4/* 5 * cloning flags: 6 */ 7#define CSIGNAL 0x000000ff /* signal mask to be sent at exit */ 8#define CLONE_VM 0x00000100 /* set if VM shared between processes */ 9#define CLONE_FS 0x00000200 /* set if fs info shared between processes */ 10#define CLONE_FILES 0x00000400 /* set if open files shared between processes */ 11#define CLONE_SIGHAND 0x00000800 /* set if signal handlers and blocked signals shared */ 12#define CLONE_PTRACE 0x00002000 /* set if we want to let tracing continue on the child too */ 13#define CLONE_VFORK 0x00004000 /* set if the parent wants the child to wake it up on mm_release */ 14#define CLONE_PARENT 0x00008000 /* set if we want to have the same parent as the cloner */ 15#define CLONE_THREAD 0x00010000 /* Same thread group? */ 16#define CLONE_NEWNS 0x00020000 /* New namespace group? */ 17#define CLONE_SYSVSEM 0x00040000 /* share system V SEM_UNDO semantics */ 18#define CLONE_SETTLS 0x00080000 /* create a new TLS for the child */ 19#define CLONE_PARENT_SETTID 0x00100000 /* set the TID in the parent */ 20#define CLONE_CHILD_CLEARTID 0x00200000 /* clear the TID in the child */ 21#define CLONE_DETACHED 0x00400000 /* Unused, ignored */ 22#define CLONE_UNTRACED 0x00800000 /* set if the tracing process can't force CLONE_PTRACE on this clone */ 23#define CLONE_CHILD_SETTID 0x01000000 /* set the TID in the child */ 24#define CLONE_STOPPED 0x02000000 /* Start in stopped state */ 25#define CLONE_NEWUTS 0x04000000 /* New utsname group? */ 26#define CLONE_NEWIPC 0x08000000 /* New ipcs */ 27#define CLONE_NEWUSER 0x10000000 /* New user namespace */ 28#define CLONE_NEWPID 0x20000000 /* New pid namespace */ 29#define CLONE_NEWNET 0x40000000 /* New network namespace */ 30#define CLONE_IO 0x80000000 /* Clone io context */ 31 32/* 33 * Scheduling policies 34 */ 35#define SCHED_NORMAL 0 36#define SCHED_FIFO 1 37#define SCHED_RR 2 38#define SCHED_BATCH 3 39/* SCHED_ISO: reserved but not implemented yet */ 40#define SCHED_IDLE 5 41/* Can be ORed in to make sure the process is reverted back to SCHED_NORMAL on fork */ 42#define SCHED_RESET_ON_FORK 0x40000000 43 44#ifdef __KERNEL__ 45 46struct sched_param { 47 int sched_priority; 48}; 49 50#include <asm/param.h> /* for HZ */ 51 52#include <linux/capability.h> 53#include <linux/threads.h> 54#include <linux/kernel.h> 55#include <linux/types.h> 56#include <linux/timex.h> 57#include <linux/jiffies.h> 58#include <linux/rbtree.h> 59#include <linux/thread_info.h> 60#include <linux/cpumask.h> 61#include <linux/errno.h> 62#include <linux/nodemask.h> 63#include <linux/mm_types.h> 64 65#include <asm/system.h> 66#include <asm/page.h> 67#include <asm/ptrace.h> 68#include <asm/cputime.h> 69 70#include <linux/smp.h> 71#include <linux/sem.h> 72#include <linux/signal.h> 73#include <linux/path.h> 74#include <linux/compiler.h> 75#include <linux/completion.h> 76#include <linux/pid.h> 77#include <linux/percpu.h> 78#include <linux/topology.h> 79#include <linux/proportions.h> 80#include <linux/seccomp.h> 81#include <linux/rcupdate.h> 82#include <linux/rculist.h> 83#include <linux/rtmutex.h> 84 85#include <linux/time.h> 86#include <linux/param.h> 87#include <linux/resource.h> 88#include <linux/timer.h> 89#include <linux/hrtimer.h> 90#include <linux/task_io_accounting.h> 91#include <linux/kobject.h> 92#include <linux/latencytop.h> 93#include <linux/cred.h> 94 95#include <asm/processor.h> 96 97struct exec_domain; 98struct futex_pi_state; 99struct robust_list_head; 100struct bio_list; 101struct fs_struct; 102struct bts_context; 103struct perf_event_context; 104 105/* 106 * List of flags we want to share for kernel threads, 107 * if only because they are not used by them anyway. 108 */ 109#define CLONE_KERNEL (CLONE_FS | CLONE_FILES | CLONE_SIGHAND) 110 111/* 112 * These are the constant used to fake the fixed-point load-average 113 * counting. Some notes: 114 * - 11 bit fractions expand to 22 bits by the multiplies: this gives 115 * a load-average precision of 10 bits integer + 11 bits fractional 116 * - if you want to count load-averages more often, you need more 117 * precision, or rounding will get you. With 2-second counting freq, 118 * the EXP_n values would be 1981, 2034 and 2043 if still using only 119 * 11 bit fractions. 120 */ 121extern unsigned long avenrun[]; /* Load averages */ 122extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift); 123 124#define FSHIFT 11 /* nr of bits of precision */ 125#define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */ 126#define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */ 127#define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */ 128#define EXP_5 2014 /* 1/exp(5sec/5min) */ 129#define EXP_15 2037 /* 1/exp(5sec/15min) */ 130 131#define CALC_LOAD(load,exp,n) \ 132 load *= exp; \ 133 load += n*(FIXED_1-exp); \ 134 load >>= FSHIFT; 135 136extern unsigned long total_forks; 137extern int nr_threads; 138DECLARE_PER_CPU(unsigned long, process_counts); 139extern int nr_processes(void); 140extern unsigned long nr_running(void); 141extern unsigned long nr_uninterruptible(void); 142extern unsigned long nr_iowait(void); 143extern unsigned long nr_iowait_cpu(void); 144extern unsigned long this_cpu_load(void); 145 146 147extern void calc_global_load(void); 148 149extern unsigned long get_parent_ip(unsigned long addr); 150 151struct seq_file; 152struct cfs_rq; 153struct task_group; 154#ifdef CONFIG_SCHED_DEBUG 155extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m); 156extern void proc_sched_set_task(struct task_struct *p); 157extern void 158print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq); 159#else 160static inline void 161proc_sched_show_task(struct task_struct *p, struct seq_file *m) 162{ 163} 164static inline void proc_sched_set_task(struct task_struct *p) 165{ 166} 167static inline void 168print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq) 169{ 170} 171#endif 172 173/* 174 * Task state bitmask. NOTE! These bits are also 175 * encoded in fs/proc/array.c: get_task_state(). 176 * 177 * We have two separate sets of flags: task->state 178 * is about runnability, while task->exit_state are 179 * about the task exiting. Confusing, but this way 180 * modifying one set can't modify the other one by 181 * mistake. 182 */ 183#define TASK_RUNNING 0 184#define TASK_INTERRUPTIBLE 1 185#define TASK_UNINTERRUPTIBLE 2 186#define __TASK_STOPPED 4 187#define __TASK_TRACED 8 188/* in tsk->exit_state */ 189#define EXIT_ZOMBIE 16 190#define EXIT_DEAD 32 191/* in tsk->state again */ 192#define TASK_DEAD 64 193#define TASK_WAKEKILL 128 194#define TASK_WAKING 256 195#define TASK_STATE_MAX 512 196 197#define TASK_STATE_TO_CHAR_STR "RSDTtZXxKW" 198 199extern char ___assert_task_state[1 - 2*!!( 200 sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)]; 201 202/* Convenience macros for the sake of set_task_state */ 203#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE) 204#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED) 205#define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED) 206 207/* Convenience macros for the sake of wake_up */ 208#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE) 209#define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED) 210 211/* get_task_state() */ 212#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \ 213 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \ 214 __TASK_TRACED) 215 216#define task_is_traced(task) ((task->state & __TASK_TRACED) != 0) 217#define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0) 218#define task_is_stopped_or_traced(task) \ 219 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) 220#define task_contributes_to_load(task) \ 221 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \ 222 (task->flags & PF_FREEZING) == 0) 223 224#define __set_task_state(tsk, state_value) \ 225 do { (tsk)->state = (state_value); } while (0) 226#define set_task_state(tsk, state_value) \ 227 set_mb((tsk)->state, (state_value)) 228 229/* 230 * set_current_state() includes a barrier so that the write of current->state 231 * is correctly serialised wrt the caller's subsequent test of whether to 232 * actually sleep: 233 * 234 * set_current_state(TASK_UNINTERRUPTIBLE); 235 * if (do_i_need_to_sleep()) 236 * schedule(); 237 * 238 * If the caller does not need such serialisation then use __set_current_state() 239 */ 240#define __set_current_state(state_value) \ 241 do { current->state = (state_value); } while (0) 242#define set_current_state(state_value) \ 243 set_mb(current->state, (state_value)) 244 245/* Task command name length */ 246#define TASK_COMM_LEN 16 247 248#include <linux/spinlock.h> 249 250/* 251 * This serializes "schedule()" and also protects 252 * the run-queue from deletions/modifications (but 253 * _adding_ to the beginning of the run-queue has 254 * a separate lock). 255 */ 256extern rwlock_t tasklist_lock; 257extern spinlock_t mmlist_lock; 258 259struct task_struct; 260 261#ifdef CONFIG_PROVE_RCU 262extern int lockdep_tasklist_lock_is_held(void); 263#endif /* #ifdef CONFIG_PROVE_RCU */ 264 265extern void sched_init(void); 266extern void sched_init_smp(void); 267extern asmlinkage void schedule_tail(struct task_struct *prev); 268extern void init_idle(struct task_struct *idle, int cpu); 269extern void init_idle_bootup_task(struct task_struct *idle); 270 271extern int runqueue_is_locked(int cpu); 272extern void task_rq_unlock_wait(struct task_struct *p); 273 274extern cpumask_var_t nohz_cpu_mask; 275#if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ) 276extern int select_nohz_load_balancer(int cpu); 277extern int get_nohz_load_balancer(void); 278extern int nohz_ratelimit(int cpu); 279#else 280static inline int select_nohz_load_balancer(int cpu) 281{ 282 return 0; 283} 284 285static inline int nohz_ratelimit(int cpu) 286{ 287 return 0; 288} 289#endif 290 291/* 292 * Only dump TASK_* tasks. (0 for all tasks) 293 */ 294extern void show_state_filter(unsigned long state_filter); 295 296static inline void show_state(void) 297{ 298 show_state_filter(0); 299} 300 301extern void show_regs(struct pt_regs *); 302 303/* 304 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current 305 * task), SP is the stack pointer of the first frame that should be shown in the back 306 * trace (or NULL if the entire call-chain of the task should be shown). 307 */ 308extern void show_stack(struct task_struct *task, unsigned long *sp); 309 310void io_schedule(void); 311long io_schedule_timeout(long timeout); 312 313extern void cpu_init (void); 314extern void trap_init(void); 315extern void update_process_times(int user); 316extern void scheduler_tick(void); 317 318extern void sched_show_task(struct task_struct *p); 319 320#ifdef CONFIG_DETECT_SOFTLOCKUP 321extern void softlockup_tick(void); 322extern void touch_softlockup_watchdog(void); 323extern void touch_softlockup_watchdog_sync(void); 324extern void touch_all_softlockup_watchdogs(void); 325extern int proc_dosoftlockup_thresh(struct ctl_table *table, int write, 326 void __user *buffer, 327 size_t *lenp, loff_t *ppos); 328extern unsigned int softlockup_panic; 329extern int softlockup_thresh; 330#else 331static inline void softlockup_tick(void) 332{ 333} 334static inline void touch_softlockup_watchdog(void) 335{ 336} 337static inline void touch_softlockup_watchdog_sync(void) 338{ 339} 340static inline void touch_all_softlockup_watchdogs(void) 341{ 342} 343#endif 344 345#ifdef CONFIG_DETECT_HUNG_TASK 346extern unsigned int sysctl_hung_task_panic; 347extern unsigned long sysctl_hung_task_check_count; 348extern unsigned long sysctl_hung_task_timeout_secs; 349extern unsigned long sysctl_hung_task_warnings; 350extern int proc_dohung_task_timeout_secs(struct ctl_table *table, int write, 351 void __user *buffer, 352 size_t *lenp, loff_t *ppos); 353#endif 354 355/* Attach to any functions which should be ignored in wchan output. */ 356#define __sched __attribute__((__section__(".sched.text"))) 357 358/* Linker adds these: start and end of __sched functions */ 359extern char __sched_text_start[], __sched_text_end[]; 360 361/* Is this address in the __sched functions? */ 362extern int in_sched_functions(unsigned long addr); 363 364#define MAX_SCHEDULE_TIMEOUT LONG_MAX 365extern signed long schedule_timeout(signed long timeout); 366extern signed long schedule_timeout_interruptible(signed long timeout); 367extern signed long schedule_timeout_killable(signed long timeout); 368extern signed long schedule_timeout_uninterruptible(signed long timeout); 369asmlinkage void schedule(void); 370extern int mutex_spin_on_owner(struct mutex *lock, struct thread_info *owner); 371 372struct nsproxy; 373struct user_namespace; 374 375/* 376 * Default maximum number of active map areas, this limits the number of vmas 377 * per mm struct. Users can overwrite this number by sysctl but there is a 378 * problem. 379 * 380 * When a program's coredump is generated as ELF format, a section is created 381 * per a vma. In ELF, the number of sections is represented in unsigned short. 382 * This means the number of sections should be smaller than 65535 at coredump. 383 * Because the kernel adds some informative sections to a image of program at 384 * generating coredump, we need some margin. The number of extra sections is 385 * 1-3 now and depends on arch. We use "5" as safe margin, here. 386 */ 387#define MAPCOUNT_ELF_CORE_MARGIN (5) 388#define DEFAULT_MAX_MAP_COUNT (USHORT_MAX - MAPCOUNT_ELF_CORE_MARGIN) 389 390extern int sysctl_max_map_count; 391 392#include <linux/aio.h> 393 394#ifdef CONFIG_MMU 395extern void arch_pick_mmap_layout(struct mm_struct *mm); 396extern unsigned long 397arch_get_unmapped_area(struct file *, unsigned long, unsigned long, 398 unsigned long, unsigned long); 399extern unsigned long 400arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr, 401 unsigned long len, unsigned long pgoff, 402 unsigned long flags); 403extern void arch_unmap_area(struct mm_struct *, unsigned long); 404extern void arch_unmap_area_topdown(struct mm_struct *, unsigned long); 405#else 406static inline void arch_pick_mmap_layout(struct mm_struct *mm) {} 407#endif 408 409 410extern void set_dumpable(struct mm_struct *mm, int value); 411extern int get_dumpable(struct mm_struct *mm); 412 413/* mm flags */ 414/* dumpable bits */ 415#define MMF_DUMPABLE 0 /* core dump is permitted */ 416#define MMF_DUMP_SECURELY 1 /* core file is readable only by root */ 417 418#define MMF_DUMPABLE_BITS 2 419#define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1) 420 421/* coredump filter bits */ 422#define MMF_DUMP_ANON_PRIVATE 2 423#define MMF_DUMP_ANON_SHARED 3 424#define MMF_DUMP_MAPPED_PRIVATE 4 425#define MMF_DUMP_MAPPED_SHARED 5 426#define MMF_DUMP_ELF_HEADERS 6 427#define MMF_DUMP_HUGETLB_PRIVATE 7 428#define MMF_DUMP_HUGETLB_SHARED 8 429 430#define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS 431#define MMF_DUMP_FILTER_BITS 7 432#define MMF_DUMP_FILTER_MASK \ 433 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT) 434#define MMF_DUMP_FILTER_DEFAULT \ 435 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\ 436 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF) 437 438#ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS 439# define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS) 440#else 441# define MMF_DUMP_MASK_DEFAULT_ELF 0 442#endif 443 /* leave room for more dump flags */ 444#define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */ 445 446#define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK) 447 448struct sighand_struct { 449 atomic_t count; 450 struct k_sigaction action[_NSIG]; 451 spinlock_t siglock; 452 wait_queue_head_t signalfd_wqh; 453}; 454 455struct pacct_struct { 456 int ac_flag; 457 long ac_exitcode; 458 unsigned long ac_mem; 459 cputime_t ac_utime, ac_stime; 460 unsigned long ac_minflt, ac_majflt; 461}; 462 463struct cpu_itimer { 464 cputime_t expires; 465 cputime_t incr; 466 u32 error; 467 u32 incr_error; 468}; 469 470/** 471 * struct task_cputime - collected CPU time counts 472 * @utime: time spent in user mode, in &cputime_t units 473 * @stime: time spent in kernel mode, in &cputime_t units 474 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds 475 * 476 * This structure groups together three kinds of CPU time that are 477 * tracked for threads and thread groups. Most things considering 478 * CPU time want to group these counts together and treat all three 479 * of them in parallel. 480 */ 481struct task_cputime { 482 cputime_t utime; 483 cputime_t stime; 484 unsigned long long sum_exec_runtime; 485}; 486/* Alternate field names when used to cache expirations. */ 487#define prof_exp stime 488#define virt_exp utime 489#define sched_exp sum_exec_runtime 490 491#define INIT_CPUTIME \ 492 (struct task_cputime) { \ 493 .utime = cputime_zero, \ 494 .stime = cputime_zero, \ 495 .sum_exec_runtime = 0, \ 496 } 497 498/* 499 * Disable preemption until the scheduler is running. 500 * Reset by start_kernel()->sched_init()->init_idle(). 501 * 502 * We include PREEMPT_ACTIVE to avoid cond_resched() from working 503 * before the scheduler is active -- see should_resched(). 504 */ 505#define INIT_PREEMPT_COUNT (1 + PREEMPT_ACTIVE) 506 507/** 508 * struct thread_group_cputimer - thread group interval timer counts 509 * @cputime: thread group interval timers. 510 * @running: non-zero when there are timers running and 511 * @cputime receives updates. 512 * @lock: lock for fields in this struct. 513 * 514 * This structure contains the version of task_cputime, above, that is 515 * used for thread group CPU timer calculations. 516 */ 517struct thread_group_cputimer { 518 struct task_cputime cputime; 519 int running; 520 spinlock_t lock; 521}; 522 523/* 524 * NOTE! "signal_struct" does not have it's own 525 * locking, because a shared signal_struct always 526 * implies a shared sighand_struct, so locking 527 * sighand_struct is always a proper superset of 528 * the locking of signal_struct. 529 */ 530struct signal_struct { 531 atomic_t count; 532 atomic_t live; 533 534 wait_queue_head_t wait_chldexit; /* for wait4() */ 535 536 /* current thread group signal load-balancing target: */ 537 struct task_struct *curr_target; 538 539 /* shared signal handling: */ 540 struct sigpending shared_pending; 541 542 /* thread group exit support */ 543 int group_exit_code; 544 /* overloaded: 545 * - notify group_exit_task when ->count is equal to notify_count 546 * - everyone except group_exit_task is stopped during signal delivery 547 * of fatal signals, group_exit_task processes the signal. 548 */ 549 int notify_count; 550 struct task_struct *group_exit_task; 551 552 /* thread group stop support, overloads group_exit_code too */ 553 int group_stop_count; 554 unsigned int flags; /* see SIGNAL_* flags below */ 555 556 /* POSIX.1b Interval Timers */ 557 struct list_head posix_timers; 558 559 /* ITIMER_REAL timer for the process */ 560 struct hrtimer real_timer; 561 struct pid *leader_pid; 562 ktime_t it_real_incr; 563 564 /* 565 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use 566 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these 567 * values are defined to 0 and 1 respectively 568 */ 569 struct cpu_itimer it[2]; 570 571 /* 572 * Thread group totals for process CPU timers. 573 * See thread_group_cputimer(), et al, for details. 574 */ 575 struct thread_group_cputimer cputimer; 576 577 /* Earliest-expiration cache. */ 578 struct task_cputime cputime_expires; 579 580 struct list_head cpu_timers[3]; 581 582 struct pid *tty_old_pgrp; 583 584 /* boolean value for session group leader */ 585 int leader; 586 587 struct tty_struct *tty; /* NULL if no tty */ 588 589 /* 590 * Cumulative resource counters for dead threads in the group, 591 * and for reaped dead child processes forked by this group. 592 * Live threads maintain their own counters and add to these 593 * in __exit_signal, except for the group leader. 594 */ 595 cputime_t utime, stime, cutime, cstime; 596 cputime_t gtime; 597 cputime_t cgtime; 598#ifndef CONFIG_VIRT_CPU_ACCOUNTING 599 cputime_t prev_utime, prev_stime; 600#endif 601 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw; 602 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt; 603 unsigned long inblock, oublock, cinblock, coublock; 604 unsigned long maxrss, cmaxrss; 605 struct task_io_accounting ioac; 606 607 /* 608 * Cumulative ns of schedule CPU time fo dead threads in the 609 * group, not including a zombie group leader, (This only differs 610 * from jiffies_to_ns(utime + stime) if sched_clock uses something 611 * other than jiffies.) 612 */ 613 unsigned long long sum_sched_runtime; 614 615 /* 616 * We don't bother to synchronize most readers of this at all, 617 * because there is no reader checking a limit that actually needs 618 * to get both rlim_cur and rlim_max atomically, and either one 619 * alone is a single word that can safely be read normally. 620 * getrlimit/setrlimit use task_lock(current->group_leader) to 621 * protect this instead of the siglock, because they really 622 * have no need to disable irqs. 623 */ 624 struct rlimit rlim[RLIM_NLIMITS]; 625 626#ifdef CONFIG_BSD_PROCESS_ACCT 627 struct pacct_struct pacct; /* per-process accounting information */ 628#endif 629#ifdef CONFIG_TASKSTATS 630 struct taskstats *stats; 631#endif 632#ifdef CONFIG_AUDIT 633 unsigned audit_tty; 634 struct tty_audit_buf *tty_audit_buf; 635#endif 636 637 int oom_adj; /* OOM kill score adjustment (bit shift) */ 638}; 639 640/* Context switch must be unlocked if interrupts are to be enabled */ 641#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW 642# define __ARCH_WANT_UNLOCKED_CTXSW 643#endif 644 645/* 646 * Bits in flags field of signal_struct. 647 */ 648#define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */ 649#define SIGNAL_STOP_DEQUEUED 0x00000002 /* stop signal dequeued */ 650#define SIGNAL_STOP_CONTINUED 0x00000004 /* SIGCONT since WCONTINUED reap */ 651#define SIGNAL_GROUP_EXIT 0x00000008 /* group exit in progress */ 652/* 653 * Pending notifications to parent. 654 */ 655#define SIGNAL_CLD_STOPPED 0x00000010 656#define SIGNAL_CLD_CONTINUED 0x00000020 657#define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED) 658 659#define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */ 660 661/* If true, all threads except ->group_exit_task have pending SIGKILL */ 662static inline int signal_group_exit(const struct signal_struct *sig) 663{ 664 return (sig->flags & SIGNAL_GROUP_EXIT) || 665 (sig->group_exit_task != NULL); 666} 667 668/* 669 * Some day this will be a full-fledged user tracking system.. 670 */ 671struct user_struct { 672 atomic_t __count; /* reference count */ 673 atomic_t processes; /* How many processes does this user have? */ 674 atomic_t files; /* How many open files does this user have? */ 675 atomic_t sigpending; /* How many pending signals does this user have? */ 676#ifdef CONFIG_INOTIFY_USER 677 atomic_t inotify_watches; /* How many inotify watches does this user have? */ 678 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */ 679#endif 680#ifdef CONFIG_EPOLL 681 atomic_t epoll_watches; /* The number of file descriptors currently watched */ 682#endif 683#ifdef CONFIG_POSIX_MQUEUE 684 /* protected by mq_lock */ 685 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */ 686#endif 687 unsigned long locked_shm; /* How many pages of mlocked shm ? */ 688 689#ifdef CONFIG_KEYS 690 struct key *uid_keyring; /* UID specific keyring */ 691 struct key *session_keyring; /* UID's default session keyring */ 692#endif 693 694 /* Hash table maintenance information */ 695 struct hlist_node uidhash_node; 696 uid_t uid; 697 struct user_namespace *user_ns; 698 699#ifdef CONFIG_PERF_EVENTS 700 atomic_long_t locked_vm; 701#endif 702}; 703 704extern int uids_sysfs_init(void); 705 706extern struct user_struct *find_user(uid_t); 707 708extern struct user_struct root_user; 709#define INIT_USER (&root_user) 710 711 712struct backing_dev_info; 713struct reclaim_state; 714 715#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) 716struct sched_info { 717 /* cumulative counters */ 718 unsigned long pcount; /* # of times run on this cpu */ 719 unsigned long long run_delay; /* time spent waiting on a runqueue */ 720 721 /* timestamps */ 722 unsigned long long last_arrival,/* when we last ran on a cpu */ 723 last_queued; /* when we were last queued to run */ 724#ifdef CONFIG_SCHEDSTATS 725 /* BKL stats */ 726 unsigned int bkl_count; 727#endif 728}; 729#endif /* defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) */ 730 731#ifdef CONFIG_TASK_DELAY_ACCT 732struct task_delay_info { 733 spinlock_t lock; 734 unsigned int flags; /* Private per-task flags */ 735 736 /* For each stat XXX, add following, aligned appropriately 737 * 738 * struct timespec XXX_start, XXX_end; 739 * u64 XXX_delay; 740 * u32 XXX_count; 741 * 742 * Atomicity of updates to XXX_delay, XXX_count protected by 743 * single lock above (split into XXX_lock if contention is an issue). 744 */ 745 746 /* 747 * XXX_count is incremented on every XXX operation, the delay 748 * associated with the operation is added to XXX_delay. 749 * XXX_delay contains the accumulated delay time in nanoseconds. 750 */ 751 struct timespec blkio_start, blkio_end; /* Shared by blkio, swapin */ 752 u64 blkio_delay; /* wait for sync block io completion */ 753 u64 swapin_delay; /* wait for swapin block io completion */ 754 u32 blkio_count; /* total count of the number of sync block */ 755 /* io operations performed */ 756 u32 swapin_count; /* total count of the number of swapin block */ 757 /* io operations performed */ 758 759 struct timespec freepages_start, freepages_end; 760 u64 freepages_delay; /* wait for memory reclaim */ 761 u32 freepages_count; /* total count of memory reclaim */ 762}; 763#endif /* CONFIG_TASK_DELAY_ACCT */ 764 765static inline int sched_info_on(void) 766{ 767#ifdef CONFIG_SCHEDSTATS 768 return 1; 769#elif defined(CONFIG_TASK_DELAY_ACCT) 770 extern int delayacct_on; 771 return delayacct_on; 772#else 773 return 0; 774#endif 775} 776 777enum cpu_idle_type { 778 CPU_IDLE, 779 CPU_NOT_IDLE, 780 CPU_NEWLY_IDLE, 781 CPU_MAX_IDLE_TYPES 782}; 783 784/* 785 * sched-domains (multiprocessor balancing) declarations: 786 */ 787 788/* 789 * Increase resolution of nice-level calculations: 790 */ 791#define SCHED_LOAD_SHIFT 10 792#define SCHED_LOAD_SCALE (1L << SCHED_LOAD_SHIFT) 793 794#define SCHED_LOAD_SCALE_FUZZ SCHED_LOAD_SCALE 795 796#ifdef CONFIG_SMP 797#define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */ 798#define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */ 799#define SD_BALANCE_EXEC 0x0004 /* Balance on exec */ 800#define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */ 801#define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */ 802#define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */ 803#define SD_PREFER_LOCAL 0x0040 /* Prefer to keep tasks local to this domain */ 804#define SD_SHARE_CPUPOWER 0x0080 /* Domain members share cpu power */ 805#define SD_POWERSAVINGS_BALANCE 0x0100 /* Balance for power savings */ 806#define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */ 807#define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */ 808 809#define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */ 810 811enum powersavings_balance_level { 812 POWERSAVINGS_BALANCE_NONE = 0, /* No power saving load balance */ 813 POWERSAVINGS_BALANCE_BASIC, /* Fill one thread/core/package 814 * first for long running threads 815 */ 816 POWERSAVINGS_BALANCE_WAKEUP, /* Also bias task wakeups to semi-idle 817 * cpu package for power savings 818 */ 819 MAX_POWERSAVINGS_BALANCE_LEVELS 820}; 821 822extern int sched_mc_power_savings, sched_smt_power_savings; 823 824static inline int sd_balance_for_mc_power(void) 825{ 826 if (sched_smt_power_savings) 827 return SD_POWERSAVINGS_BALANCE; 828 829 if (!sched_mc_power_savings) 830 return SD_PREFER_SIBLING; 831 832 return 0; 833} 834 835static inline int sd_balance_for_package_power(void) 836{ 837 if (sched_mc_power_savings | sched_smt_power_savings) 838 return SD_POWERSAVINGS_BALANCE; 839 840 return SD_PREFER_SIBLING; 841} 842 843/* 844 * Optimise SD flags for power savings: 845 * SD_BALANCE_NEWIDLE helps agressive task consolidation and power savings. 846 * Keep default SD flags if sched_{smt,mc}_power_saving=0 847 */ 848 849static inline int sd_power_saving_flags(void) 850{ 851 if (sched_mc_power_savings | sched_smt_power_savings) 852 return SD_BALANCE_NEWIDLE; 853 854 return 0; 855} 856 857struct sched_group { 858 struct sched_group *next; /* Must be a circular list */ 859 860 /* 861 * CPU power of this group, SCHED_LOAD_SCALE being max power for a 862 * single CPU. 863 */ 864 unsigned int cpu_power; 865 866 /* 867 * The CPUs this group covers. 868 * 869 * NOTE: this field is variable length. (Allocated dynamically 870 * by attaching extra space to the end of the structure, 871 * depending on how many CPUs the kernel has booted up with) 872 * 873 * It is also be embedded into static data structures at build 874 * time. (See 'struct static_sched_group' in kernel/sched.c) 875 */ 876 unsigned long cpumask[0]; 877}; 878 879static inline struct cpumask *sched_group_cpus(struct sched_group *sg) 880{ 881 return to_cpumask(sg->cpumask); 882} 883 884enum sched_domain_level { 885 SD_LV_NONE = 0, 886 SD_LV_SIBLING, 887 SD_LV_MC, 888 SD_LV_CPU, 889 SD_LV_NODE, 890 SD_LV_ALLNODES, 891 SD_LV_MAX 892}; 893 894struct sched_domain_attr { 895 int relax_domain_level; 896}; 897 898#define SD_ATTR_INIT (struct sched_domain_attr) { \ 899 .relax_domain_level = -1, \ 900} 901 902struct sched_domain { 903 /* These fields must be setup */ 904 struct sched_domain *parent; /* top domain must be null terminated */ 905 struct sched_domain *child; /* bottom domain must be null terminated */ 906 struct sched_group *groups; /* the balancing groups of the domain */ 907 unsigned long min_interval; /* Minimum balance interval ms */ 908 unsigned long max_interval; /* Maximum balance interval ms */ 909 unsigned int busy_factor; /* less balancing by factor if busy */ 910 unsigned int imbalance_pct; /* No balance until over watermark */ 911 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */ 912 unsigned int busy_idx; 913 unsigned int idle_idx; 914 unsigned int newidle_idx; 915 unsigned int wake_idx; 916 unsigned int forkexec_idx; 917 unsigned int smt_gain; 918 int flags; /* See SD_* */ 919 enum sched_domain_level level; 920 921 /* Runtime fields. */ 922 unsigned long last_balance; /* init to jiffies. units in jiffies */ 923 unsigned int balance_interval; /* initialise to 1. units in ms. */ 924 unsigned int nr_balance_failed; /* initialise to 0 */ 925 926 u64 last_update; 927 928#ifdef CONFIG_SCHEDSTATS 929 /* load_balance() stats */ 930 unsigned int lb_count[CPU_MAX_IDLE_TYPES]; 931 unsigned int lb_failed[CPU_MAX_IDLE_TYPES]; 932 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES]; 933 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES]; 934 unsigned int lb_gained[CPU_MAX_IDLE_TYPES]; 935 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES]; 936 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES]; 937 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES]; 938 939 /* Active load balancing */ 940 unsigned int alb_count; 941 unsigned int alb_failed; 942 unsigned int alb_pushed; 943 944 /* SD_BALANCE_EXEC stats */ 945 unsigned int sbe_count; 946 unsigned int sbe_balanced; 947 unsigned int sbe_pushed; 948 949 /* SD_BALANCE_FORK stats */ 950 unsigned int sbf_count; 951 unsigned int sbf_balanced; 952 unsigned int sbf_pushed; 953 954 /* try_to_wake_up() stats */ 955 unsigned int ttwu_wake_remote; 956 unsigned int ttwu_move_affine; 957 unsigned int ttwu_move_balance; 958#endif 959#ifdef CONFIG_SCHED_DEBUG 960 char *name; 961#endif 962 963 unsigned int span_weight; 964 /* 965 * Span of all CPUs in this domain. 966 * 967 * NOTE: this field is variable length. (Allocated dynamically 968 * by attaching extra space to the end of the structure, 969 * depending on how many CPUs the kernel has booted up with) 970 * 971 * It is also be embedded into static data structures at build 972 * time. (See 'struct static_sched_domain' in kernel/sched.c) 973 */ 974 unsigned long span[0]; 975}; 976 977static inline struct cpumask *sched_domain_span(struct sched_domain *sd) 978{ 979 return to_cpumask(sd->span); 980} 981 982extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[], 983 struct sched_domain_attr *dattr_new); 984 985/* Allocate an array of sched domains, for partition_sched_domains(). */ 986cpumask_var_t *alloc_sched_domains(unsigned int ndoms); 987void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms); 988 989/* Test a flag in parent sched domain */ 990static inline int test_sd_parent(struct sched_domain *sd, int flag) 991{ 992 if (sd->parent && (sd->parent->flags & flag)) 993 return 1; 994 995 return 0; 996} 997 998unsigned long default_scale_freq_power(struct sched_domain *sd, int cpu); 999unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu); 1000 1001#else /* CONFIG_SMP */ 1002 1003struct sched_domain_attr; 1004 1005static inline void 1006partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[], 1007 struct sched_domain_attr *dattr_new) 1008{ 1009} 1010#endif /* !CONFIG_SMP */ 1011 1012 1013struct io_context; /* See blkdev.h */ 1014 1015 1016#ifdef ARCH_HAS_PREFETCH_SWITCH_STACK 1017extern void prefetch_stack(struct task_struct *t); 1018#else 1019static inline void prefetch_stack(struct task_struct *t) { } 1020#endif 1021 1022struct audit_context; /* See audit.c */ 1023struct mempolicy; 1024struct pipe_inode_info; 1025struct uts_namespace; 1026 1027struct rq; 1028struct sched_domain; 1029 1030/* 1031 * wake flags 1032 */ 1033#define WF_SYNC 0x01 /* waker goes to sleep after wakup */ 1034#define WF_FORK 0x02 /* child wakeup after fork */ 1035 1036#define ENQUEUE_WAKEUP 1 1037#define ENQUEUE_WAKING 2 1038#define ENQUEUE_HEAD 4 1039 1040#define DEQUEUE_SLEEP 1 1041 1042struct sched_class { 1043 const struct sched_class *next; 1044 1045 void (*enqueue_task) (struct rq *rq, struct task_struct *p, int flags); 1046 void (*dequeue_task) (struct rq *rq, struct task_struct *p, int flags); 1047 void (*yield_task) (struct rq *rq); 1048 1049 void (*check_preempt_curr) (struct rq *rq, struct task_struct *p, int flags); 1050 1051 struct task_struct * (*pick_next_task) (struct rq *rq); 1052 void (*put_prev_task) (struct rq *rq, struct task_struct *p); 1053 1054#ifdef CONFIG_SMP 1055 int (*select_task_rq)(struct rq *rq, struct task_struct *p, 1056 int sd_flag, int flags); 1057 1058 void (*pre_schedule) (struct rq *this_rq, struct task_struct *task); 1059 void (*post_schedule) (struct rq *this_rq); 1060 void (*task_waking) (struct rq *this_rq, struct task_struct *task); 1061 void (*task_woken) (struct rq *this_rq, struct task_struct *task); 1062 1063 void (*set_cpus_allowed)(struct task_struct *p, 1064 const struct cpumask *newmask); 1065 1066 void (*rq_online)(struct rq *rq); 1067 void (*rq_offline)(struct rq *rq); 1068#endif 1069 1070 void (*set_curr_task) (struct rq *rq); 1071 void (*task_tick) (struct rq *rq, struct task_struct *p, int queued); 1072 void (*task_fork) (struct task_struct *p); 1073 1074 void (*switched_from) (struct rq *this_rq, struct task_struct *task, 1075 int running); 1076 void (*switched_to) (struct rq *this_rq, struct task_struct *task, 1077 int running); 1078 void (*prio_changed) (struct rq *this_rq, struct task_struct *task, 1079 int oldprio, int running); 1080 1081 unsigned int (*get_rr_interval) (struct rq *rq, 1082 struct task_struct *task); 1083 1084#ifdef CONFIG_FAIR_GROUP_SCHED 1085 void (*moved_group) (struct task_struct *p, int on_rq); 1086#endif 1087}; 1088 1089struct load_weight { 1090 unsigned long weight, inv_weight; 1091}; 1092 1093#ifdef CONFIG_SCHEDSTATS 1094struct sched_statistics { 1095 u64 wait_start; 1096 u64 wait_max; 1097 u64 wait_count; 1098 u64 wait_sum; 1099 u64 iowait_count; 1100 u64 iowait_sum; 1101 1102 u64 sleep_start; 1103 u64 sleep_max; 1104 s64 sum_sleep_runtime; 1105 1106 u64 block_start; 1107 u64 block_max; 1108 u64 exec_max; 1109 u64 slice_max; 1110 1111 u64 nr_migrations_cold; 1112 u64 nr_failed_migrations_affine; 1113 u64 nr_failed_migrations_running; 1114 u64 nr_failed_migrations_hot; 1115 u64 nr_forced_migrations; 1116 1117 u64 nr_wakeups; 1118 u64 nr_wakeups_sync; 1119 u64 nr_wakeups_migrate; 1120 u64 nr_wakeups_local; 1121 u64 nr_wakeups_remote; 1122 u64 nr_wakeups_affine; 1123 u64 nr_wakeups_affine_attempts; 1124 u64 nr_wakeups_passive; 1125 u64 nr_wakeups_idle; 1126}; 1127#endif 1128 1129struct sched_entity { 1130 struct load_weight load; /* for load-balancing */ 1131 struct rb_node run_node; 1132 struct list_head group_node; 1133 unsigned int on_rq; 1134 1135 u64 exec_start; 1136 u64 sum_exec_runtime; 1137 u64 vruntime; 1138 u64 prev_sum_exec_runtime; 1139 1140 u64 nr_migrations; 1141 1142#ifdef CONFIG_SCHEDSTATS 1143 struct sched_statistics statistics; 1144#endif 1145 1146#ifdef CONFIG_FAIR_GROUP_SCHED 1147 struct sched_entity *parent; 1148 /* rq on which this entity is (to be) queued: */ 1149 struct cfs_rq *cfs_rq; 1150 /* rq "owned" by this entity/group: */ 1151 struct cfs_rq *my_q; 1152#endif 1153}; 1154 1155struct sched_rt_entity { 1156 struct list_head run_list; 1157 unsigned long timeout; 1158 unsigned int time_slice; 1159 int nr_cpus_allowed; 1160 1161 struct sched_rt_entity *back; 1162#ifdef CONFIG_RT_GROUP_SCHED 1163 struct sched_rt_entity *parent; 1164 /* rq on which this entity is (to be) queued: */ 1165 struct rt_rq *rt_rq; 1166 /* rq "owned" by this entity/group: */ 1167 struct rt_rq *my_q; 1168#endif 1169}; 1170 1171struct rcu_node; 1172 1173struct task_struct { 1174 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */ 1175 void *stack; 1176 atomic_t usage; 1177 unsigned int flags; /* per process flags, defined below */ 1178 unsigned int ptrace; 1179 1180 int lock_depth; /* BKL lock depth */ 1181 1182#ifdef CONFIG_SMP 1183#ifdef __ARCH_WANT_UNLOCKED_CTXSW 1184 int oncpu; 1185#endif 1186#endif 1187 1188 int prio, static_prio, normal_prio; 1189 unsigned int rt_priority; 1190 const struct sched_class *sched_class; 1191 struct sched_entity se; 1192 struct sched_rt_entity rt; 1193 1194#ifdef CONFIG_PREEMPT_NOTIFIERS 1195 /* list of struct preempt_notifier: */ 1196 struct hlist_head preempt_notifiers; 1197#endif 1198 1199 /* 1200 * fpu_counter contains the number of consecutive context switches 1201 * that the FPU is used. If this is over a threshold, the lazy fpu 1202 * saving becomes unlazy to save the trap. This is an unsigned char 1203 * so that after 256 times the counter wraps and the behavior turns 1204 * lazy again; this to deal with bursty apps that only use FPU for 1205 * a short time 1206 */ 1207 unsigned char fpu_counter; 1208#ifdef CONFIG_BLK_DEV_IO_TRACE 1209 unsigned int btrace_seq; 1210#endif 1211 1212 unsigned int policy; 1213 cpumask_t cpus_allowed; 1214 1215#ifdef CONFIG_TREE_PREEMPT_RCU 1216 int rcu_read_lock_nesting; 1217 char rcu_read_unlock_special; 1218 struct rcu_node *rcu_blocked_node; 1219 struct list_head rcu_node_entry; 1220#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */ 1221 1222#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) 1223 struct sched_info sched_info; 1224#endif 1225 1226 struct list_head tasks; 1227 struct plist_node pushable_tasks; 1228 1229 struct mm_struct *mm, *active_mm; 1230#if defined(SPLIT_RSS_COUNTING) 1231 struct task_rss_stat rss_stat; 1232#endif 1233/* task state */ 1234 int exit_state; 1235 int exit_code, exit_signal; 1236 int pdeath_signal; /* The signal sent when the parent dies */ 1237 /* ??? */ 1238 unsigned int personality; 1239 unsigned did_exec:1; 1240 unsigned in_execve:1; /* Tell the LSMs that the process is doing an 1241 * execve */ 1242 unsigned in_iowait:1; 1243 1244 1245 /* Revert to default priority/policy when forking */ 1246 unsigned sched_reset_on_fork:1; 1247 1248 pid_t pid; 1249 pid_t tgid; 1250 1251#ifdef CONFIG_CC_STACKPROTECTOR 1252 /* Canary value for the -fstack-protector gcc feature */ 1253 unsigned long stack_canary; 1254#endif 1255 1256 /* 1257 * pointers to (original) parent process, youngest child, younger sibling, 1258 * older sibling, respectively. (p->father can be replaced with 1259 * p->real_parent->pid) 1260 */ 1261 struct task_struct *real_parent; /* real parent process */ 1262 struct task_struct *parent; /* recipient of SIGCHLD, wait4() reports */ 1263 /* 1264 * children/sibling forms the list of my natural children 1265 */ 1266 struct list_head children; /* list of my children */ 1267 struct list_head sibling; /* linkage in my parent's children list */ 1268 struct task_struct *group_leader; /* threadgroup leader */ 1269 1270 /* 1271 * ptraced is the list of tasks this task is using ptrace on. 1272 * This includes both natural children and PTRACE_ATTACH targets. 1273 * p->ptrace_entry is p's link on the p->parent->ptraced list. 1274 */ 1275 struct list_head ptraced; 1276 struct list_head ptrace_entry; 1277 1278 /* 1279 * This is the tracer handle for the ptrace BTS extension. 1280 * This field actually belongs to the ptracer task. 1281 */ 1282 struct bts_context *bts; 1283 1284 /* PID/PID hash table linkage. */ 1285 struct pid_link pids[PIDTYPE_MAX]; 1286 struct list_head thread_group; 1287 1288 struct completion *vfork_done; /* for vfork() */ 1289 int __user *set_child_tid; /* CLONE_CHILD_SETTID */ 1290 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */ 1291 1292 cputime_t utime, stime, utimescaled, stimescaled; 1293 cputime_t gtime; 1294#ifndef CONFIG_VIRT_CPU_ACCOUNTING 1295 cputime_t prev_utime, prev_stime; 1296#endif 1297 unsigned long nvcsw, nivcsw; /* context switch counts */ 1298 struct timespec start_time; /* monotonic time */ 1299 struct timespec real_start_time; /* boot based time */ 1300/* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */ 1301 unsigned long min_flt, maj_flt; 1302 1303 struct task_cputime cputime_expires; 1304 struct list_head cpu_timers[3]; 1305 1306/* process credentials */ 1307 const struct cred *real_cred; /* objective and real subjective task 1308 * credentials (COW) */ 1309 const struct cred *cred; /* effective (overridable) subjective task 1310 * credentials (COW) */ 1311 struct mutex cred_guard_mutex; /* guard against foreign influences on 1312 * credential calculations 1313 * (notably. ptrace) */ 1314 struct cred *replacement_session_keyring; /* for KEYCTL_SESSION_TO_PARENT */ 1315 1316 char comm[TASK_COMM_LEN]; /* executable name excluding path 1317 - access with [gs]et_task_comm (which lock 1318 it with task_lock()) 1319 - initialized normally by setup_new_exec */ 1320/* file system info */ 1321 int link_count, total_link_count; 1322#ifdef CONFIG_SYSVIPC 1323/* ipc stuff */ 1324 struct sysv_sem sysvsem; 1325#endif 1326#ifdef CONFIG_DETECT_HUNG_TASK 1327/* hung task detection */ 1328 unsigned long last_switch_count; 1329#endif 1330/* CPU-specific state of this task */ 1331 struct thread_struct thread; 1332/* filesystem information */ 1333 struct fs_struct *fs; 1334/* open file information */ 1335 struct files_struct *files; 1336/* namespaces */ 1337 struct nsproxy *nsproxy; 1338/* signal handlers */ 1339 struct signal_struct *signal; 1340 struct sighand_struct *sighand; 1341 1342 sigset_t blocked, real_blocked; 1343 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */ 1344 struct sigpending pending; 1345 1346 unsigned long sas_ss_sp; 1347 size_t sas_ss_size; 1348 int (*notifier)(void *priv); 1349 void *notifier_data; 1350 sigset_t *notifier_mask; 1351 struct audit_context *audit_context; 1352#ifdef CONFIG_AUDITSYSCALL 1353 uid_t loginuid; 1354 unsigned int sessionid; 1355#endif 1356 seccomp_t seccomp; 1357 1358/* Thread group tracking */ 1359 u32 parent_exec_id; 1360 u32 self_exec_id; 1361/* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, 1362 * mempolicy */ 1363 spinlock_t alloc_lock; 1364 1365#ifdef CONFIG_GENERIC_HARDIRQS 1366 /* IRQ handler threads */ 1367 struct irqaction *irqaction; 1368#endif 1369 1370 /* Protection of the PI data structures: */ 1371 raw_spinlock_t pi_lock; 1372 1373#ifdef CONFIG_RT_MUTEXES 1374 /* PI waiters blocked on a rt_mutex held by this task */ 1375 struct plist_head pi_waiters; 1376 /* Deadlock detection and priority inheritance handling */ 1377 struct rt_mutex_waiter *pi_blocked_on; 1378#endif 1379 1380#ifdef CONFIG_DEBUG_MUTEXES 1381 /* mutex deadlock detection */ 1382 struct mutex_waiter *blocked_on; 1383#endif 1384#ifdef CONFIG_TRACE_IRQFLAGS 1385 unsigned int irq_events; 1386 unsigned long hardirq_enable_ip; 1387 unsigned long hardirq_disable_ip; 1388 unsigned int hardirq_enable_event; 1389 unsigned int hardirq_disable_event; 1390 int hardirqs_enabled; 1391 int hardirq_context; 1392 unsigned long softirq_disable_ip; 1393 unsigned long softirq_enable_ip; 1394 unsigned int softirq_disable_event; 1395 unsigned int softirq_enable_event; 1396 int softirqs_enabled; 1397 int softirq_context; 1398#endif 1399#ifdef CONFIG_LOCKDEP 1400# define MAX_LOCK_DEPTH 48UL 1401 u64 curr_chain_key; 1402 int lockdep_depth; 1403 unsigned int lockdep_recursion; 1404 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1405 gfp_t lockdep_reclaim_gfp; 1406#endif 1407 1408/* journalling filesystem info */ 1409 void *journal_info; 1410 1411/* stacked block device info */ 1412 struct bio_list *bio_list; 1413 1414/* VM state */ 1415 struct reclaim_state *reclaim_state; 1416 1417 struct backing_dev_info *backing_dev_info; 1418 1419 struct io_context *io_context; 1420 1421 unsigned long ptrace_message; 1422 siginfo_t *last_siginfo; /* For ptrace use. */ 1423 struct task_io_accounting ioac; 1424#if defined(CONFIG_TASK_XACCT) 1425 u64 acct_rss_mem1; /* accumulated rss usage */ 1426 u64 acct_vm_mem1; /* accumulated virtual memory usage */ 1427 cputime_t acct_timexpd; /* stime + utime since last update */ 1428#endif 1429#ifdef CONFIG_CPUSETS 1430 nodemask_t mems_allowed; /* Protected by alloc_lock */ 1431 int cpuset_mem_spread_rotor; 1432#endif 1433#ifdef CONFIG_CGROUPS 1434 /* Control Group info protected by css_set_lock */ 1435 struct css_set *cgroups; 1436 /* cg_list protected by css_set_lock and tsk->alloc_lock */ 1437 struct list_head cg_list; 1438#endif 1439#ifdef CONFIG_FUTEX 1440 struct robust_list_head __user *robust_list; 1441#ifdef CONFIG_COMPAT 1442 struct compat_robust_list_head __user *compat_robust_list; 1443#endif 1444 struct list_head pi_state_list; 1445 struct futex_pi_state *pi_state_cache; 1446#endif 1447#ifdef CONFIG_PERF_EVENTS 1448 struct perf_event_context *perf_event_ctxp; 1449 struct mutex perf_event_mutex; 1450 struct list_head perf_event_list; 1451#endif 1452#ifdef CONFIG_NUMA 1453 struct mempolicy *mempolicy; /* Protected by alloc_lock */ 1454 short il_next; 1455#endif 1456 atomic_t fs_excl; /* holding fs exclusive resources */ 1457 struct rcu_head rcu; 1458 1459 /* 1460 * cache last used pipe for splice 1461 */ 1462 struct pipe_inode_info *splice_pipe; 1463#ifdef CONFIG_TASK_DELAY_ACCT 1464 struct task_delay_info *delays; 1465#endif 1466#ifdef CONFIG_FAULT_INJECTION 1467 int make_it_fail; 1468#endif 1469 struct prop_local_single dirties; 1470#ifdef CONFIG_LATENCYTOP 1471 int latency_record_count; 1472 struct latency_record latency_record[LT_SAVECOUNT]; 1473#endif 1474 /* 1475 * time slack values; these are used to round up poll() and 1476 * select() etc timeout values. These are in nanoseconds. 1477 */ 1478 unsigned long timer_slack_ns; 1479 unsigned long default_timer_slack_ns; 1480 1481 struct list_head *scm_work_list; 1482#ifdef CONFIG_FUNCTION_GRAPH_TRACER 1483 /* Index of current stored address in ret_stack */ 1484 int curr_ret_stack; 1485 /* Stack of return addresses for return function tracing */ 1486 struct ftrace_ret_stack *ret_stack; 1487 /* time stamp for last schedule */ 1488 unsigned long long ftrace_timestamp; 1489 /* 1490 * Number of functions that haven't been traced 1491 * because of depth overrun. 1492 */ 1493 atomic_t trace_overrun; 1494 /* Pause for the tracing */ 1495 atomic_t tracing_graph_pause; 1496#endif 1497#ifdef CONFIG_TRACING 1498 /* state flags for use by tracers */ 1499 unsigned long trace; 1500 /* bitmask of trace recursion */ 1501 unsigned long trace_recursion; 1502#endif /* CONFIG_TRACING */ 1503 unsigned long stack_start; 1504#ifdef CONFIG_CGROUP_MEM_RES_CTLR /* memcg uses this to do batch job */ 1505 struct memcg_batch_info { 1506 int do_batch; /* incremented when batch uncharge started */ 1507 struct mem_cgroup *memcg; /* target memcg of uncharge */ 1508 unsigned long bytes; /* uncharged usage */ 1509 unsigned long memsw_bytes; /* uncharged mem+swap usage */ 1510 } memcg_batch; 1511#endif 1512}; 1513 1514/* Future-safe accessor for struct task_struct's cpus_allowed. */ 1515#define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed) 1516 1517/* 1518 * Priority of a process goes from 0..MAX_PRIO-1, valid RT 1519 * priority is 0..MAX_RT_PRIO-1, and SCHED_NORMAL/SCHED_BATCH 1520 * tasks are in the range MAX_RT_PRIO..MAX_PRIO-1. Priority 1521 * values are inverted: lower p->prio value means higher priority. 1522 * 1523 * The MAX_USER_RT_PRIO value allows the actual maximum 1524 * RT priority to be separate from the value exported to 1525 * user-space. This allows kernel threads to set their 1526 * priority to a value higher than any user task. Note: 1527 * MAX_RT_PRIO must not be smaller than MAX_USER_RT_PRIO. 1528 */ 1529 1530#define MAX_USER_RT_PRIO 100 1531#define MAX_RT_PRIO MAX_USER_RT_PRIO 1532 1533#define MAX_PRIO (MAX_RT_PRIO + 40) 1534#define DEFAULT_PRIO (MAX_RT_PRIO + 20) 1535 1536static inline int rt_prio(int prio) 1537{ 1538 if (unlikely(prio < MAX_RT_PRIO)) 1539 return 1; 1540 return 0; 1541} 1542 1543static inline int rt_task(struct task_struct *p) 1544{ 1545 return rt_prio(p->prio); 1546} 1547 1548static inline struct pid *task_pid(struct task_struct *task) 1549{ 1550 return task->pids[PIDTYPE_PID].pid; 1551} 1552 1553static inline struct pid *task_tgid(struct task_struct *task) 1554{ 1555 return task->group_leader->pids[PIDTYPE_PID].pid; 1556} 1557 1558/* 1559 * Without tasklist or rcu lock it is not safe to dereference 1560 * the result of task_pgrp/task_session even if task == current, 1561 * we can race with another thread doing sys_setsid/sys_setpgid. 1562 */ 1563static inline struct pid *task_pgrp(struct task_struct *task) 1564{ 1565 return task->group_leader->pids[PIDTYPE_PGID].pid; 1566} 1567 1568static inline struct pid *task_session(struct task_struct *task) 1569{ 1570 return task->group_leader->pids[PIDTYPE_SID].pid; 1571} 1572 1573struct pid_namespace; 1574 1575/* 1576 * the helpers to get the task's different pids as they are seen 1577 * from various namespaces 1578 * 1579 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1580 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1581 * current. 1582 * task_xid_nr_ns() : id seen from the ns specified; 1583 * 1584 * set_task_vxid() : assigns a virtual id to a task; 1585 * 1586 * see also pid_nr() etc in include/linux/pid.h 1587 */ 1588pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, 1589 struct pid_namespace *ns); 1590 1591static inline pid_t task_pid_nr(struct task_struct *tsk) 1592{ 1593 return tsk->pid; 1594} 1595 1596static inline pid_t task_pid_nr_ns(struct task_struct *tsk, 1597 struct pid_namespace *ns) 1598{ 1599 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1600} 1601 1602static inline pid_t task_pid_vnr(struct task_struct *tsk) 1603{ 1604 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1605} 1606 1607 1608static inline pid_t task_tgid_nr(struct task_struct *tsk) 1609{ 1610 return tsk->tgid; 1611} 1612 1613pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns); 1614 1615static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1616{ 1617 return pid_vnr(task_tgid(tsk)); 1618} 1619 1620 1621static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, 1622 struct pid_namespace *ns) 1623{ 1624 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1625} 1626 1627static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1628{ 1629 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1630} 1631 1632 1633static inline pid_t task_session_nr_ns(struct task_struct *tsk, 1634 struct pid_namespace *ns) 1635{ 1636 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1637} 1638 1639static inline pid_t task_session_vnr(struct task_struct *tsk) 1640{ 1641 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1642} 1643 1644/* obsolete, do not use */ 1645static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1646{ 1647 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1648} 1649 1650/** 1651 * pid_alive - check that a task structure is not stale 1652 * @p: Task structure to be checked. 1653 * 1654 * Test if a process is not yet dead (at most zombie state) 1655 * If pid_alive fails, then pointers within the task structure 1656 * can be stale and must not be dereferenced. 1657 */ 1658static inline int pid_alive(struct task_struct *p) 1659{ 1660 return p->pids[PIDTYPE_PID].pid != NULL; 1661} 1662 1663/** 1664 * is_global_init - check if a task structure is init 1665 * @tsk: Task structure to be checked. 1666 * 1667 * Check if a task structure is the first user space task the kernel created. 1668 */ 1669static inline int is_global_init(struct task_struct *tsk) 1670{ 1671 return tsk->pid == 1; 1672} 1673 1674/* 1675 * is_container_init: 1676 * check whether in the task is init in its own pid namespace. 1677 */ 1678extern int is_container_init(struct task_struct *tsk); 1679 1680extern struct pid *cad_pid; 1681 1682extern void free_task(struct task_struct *tsk); 1683#define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0) 1684 1685extern void __put_task_struct(struct task_struct *t); 1686 1687static inline void put_task_struct(struct task_struct *t) 1688{ 1689 if (atomic_dec_and_test(&t->usage)) 1690 __put_task_struct(t); 1691} 1692 1693extern void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st); 1694extern void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st); 1695 1696/* 1697 * Per process flags 1698 */ 1699#define PF_ALIGNWARN 0x00000001 /* Print alignment warning msgs */ 1700 /* Not implemented yet, only for 486*/ 1701#define PF_STARTING 0x00000002 /* being created */ 1702#define PF_EXITING 0x00000004 /* getting shut down */ 1703#define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */ 1704#define PF_VCPU 0x00000010 /* I'm a virtual CPU */ 1705#define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */ 1706#define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */ 1707#define PF_SUPERPRIV 0x00000100 /* used super-user privileges */ 1708#define PF_DUMPCORE 0x00000200 /* dumped core */ 1709#define PF_SIGNALED 0x00000400 /* killed by a signal */ 1710#define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1711#define PF_FLUSHER 0x00001000 /* responsible for disk writeback */ 1712#define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */ 1713#define PF_FREEZING 0x00004000 /* freeze in progress. do not account to load */ 1714#define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */ 1715#define PF_FROZEN 0x00010000 /* frozen for system suspend */ 1716#define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */ 1717#define PF_KSWAPD 0x00040000 /* I am kswapd */ 1718#define PF_OOM_ORIGIN 0x00080000 /* Allocating much memory to others */ 1719#define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ 1720#define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1721#define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */ 1722#define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1723#define PF_SPREAD_PAGE 0x01000000 /* Spread page cache over cpuset */ 1724#define PF_SPREAD_SLAB 0x02000000 /* Spread some slab caches over cpuset */ 1725#define PF_THREAD_BOUND 0x04000000 /* Thread bound to specific cpu */ 1726#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1727#define PF_MEMPOLICY 0x10000000 /* Non-default NUMA mempolicy */ 1728#define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */ 1729#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezeable */ 1730#define PF_FREEZER_NOSIG 0x80000000 /* Freezer won't send signals to it */ 1731 1732/* 1733 * Only the _current_ task can read/write to tsk->flags, but other 1734 * tasks can access tsk->flags in readonly mode for example 1735 * with tsk_used_math (like during threaded core dumping). 1736 * There is however an exception to this rule during ptrace 1737 * or during fork: the ptracer task is allowed to write to the 1738 * child->flags of its traced child (same goes for fork, the parent 1739 * can write to the child->flags), because we're guaranteed the 1740 * child is not running and in turn not changing child->flags 1741 * at the same time the parent does it. 1742 */ 1743#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1744#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1745#define clear_used_math() clear_stopped_child_used_math(current) 1746#define set_used_math() set_stopped_child_used_math(current) 1747#define conditional_stopped_child_used_math(condition, child) \ 1748 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1749#define conditional_used_math(condition) \ 1750 conditional_stopped_child_used_math(condition, current) 1751#define copy_to_stopped_child_used_math(child) \ 1752 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1753/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1754#define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1755#define used_math() tsk_used_math(current) 1756 1757#ifdef CONFIG_TREE_PREEMPT_RCU 1758 1759#define RCU_READ_UNLOCK_BLOCKED (1 << 0) /* blocked while in RCU read-side. */ 1760#define RCU_READ_UNLOCK_NEED_QS (1 << 1) /* RCU core needs CPU response. */ 1761 1762static inline void rcu_copy_process(struct task_struct *p) 1763{ 1764 p->rcu_read_lock_nesting = 0; 1765 p->rcu_read_unlock_special = 0; 1766 p->rcu_blocked_node = NULL; 1767 INIT_LIST_HEAD(&p->rcu_node_entry); 1768} 1769 1770#else 1771 1772static inline void rcu_copy_process(struct task_struct *p) 1773{ 1774} 1775 1776#endif 1777 1778#ifdef CONFIG_SMP 1779extern int set_cpus_allowed_ptr(struct task_struct *p, 1780 const struct cpumask *new_mask); 1781#else 1782static inline int set_cpus_allowed_ptr(struct task_struct *p, 1783 const struct cpumask *new_mask) 1784{ 1785 if (!cpumask_test_cpu(0, new_mask)) 1786 return -EINVAL; 1787 return 0; 1788} 1789#endif 1790 1791#ifndef CONFIG_CPUMASK_OFFSTACK 1792static inline int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask) 1793{ 1794 return set_cpus_allowed_ptr(p, &new_mask); 1795} 1796#endif 1797 1798/* 1799 * Architectures can set this to 1 if they have specified 1800 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig, 1801 * but then during bootup it turns out that sched_clock() 1802 * is reliable after all: 1803 */ 1804#ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK 1805extern int sched_clock_stable; 1806#endif 1807 1808/* ftrace calls sched_clock() directly */ 1809extern unsigned long long notrace sched_clock(void); 1810 1811extern void sched_clock_init(void); 1812extern u64 sched_clock_cpu(int cpu); 1813 1814#ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK 1815static inline void sched_clock_tick(void) 1816{ 1817} 1818 1819static inline void sched_clock_idle_sleep_event(void) 1820{ 1821} 1822 1823static inline void sched_clock_idle_wakeup_event(u64 delta_ns) 1824{ 1825} 1826#else 1827extern void sched_clock_tick(void); 1828extern void sched_clock_idle_sleep_event(void); 1829extern void sched_clock_idle_wakeup_event(u64 delta_ns); 1830#endif 1831 1832/* 1833 * For kernel-internal use: high-speed (but slightly incorrect) per-cpu 1834 * clock constructed from sched_clock(): 1835 */ 1836extern unsigned long long cpu_clock(int cpu); 1837 1838extern unsigned long long 1839task_sched_runtime(struct task_struct *task); 1840extern unsigned long long thread_group_sched_runtime(struct task_struct *task); 1841 1842/* sched_exec is called by processes performing an exec */ 1843#ifdef CONFIG_SMP 1844extern void sched_exec(void); 1845#else 1846#define sched_exec() {} 1847#endif 1848 1849extern void sched_clock_idle_sleep_event(void); 1850extern void sched_clock_idle_wakeup_event(u64 delta_ns); 1851 1852#ifdef CONFIG_HOTPLUG_CPU 1853extern void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p); 1854extern void idle_task_exit(void); 1855#else 1856static inline void idle_task_exit(void) {} 1857#endif 1858 1859extern void sched_idle_next(void); 1860 1861#if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP) 1862extern void wake_up_idle_cpu(int cpu); 1863#else 1864static inline void wake_up_idle_cpu(int cpu) { } 1865#endif 1866 1867extern unsigned int sysctl_sched_latency; 1868extern unsigned int sysctl_sched_min_granularity; 1869extern unsigned int sysctl_sched_wakeup_granularity; 1870extern unsigned int sysctl_sched_shares_ratelimit; 1871extern unsigned int sysctl_sched_shares_thresh; 1872extern unsigned int sysctl_sched_child_runs_first; 1873 1874enum sched_tunable_scaling { 1875 SCHED_TUNABLESCALING_NONE, 1876 SCHED_TUNABLESCALING_LOG, 1877 SCHED_TUNABLESCALING_LINEAR, 1878 SCHED_TUNABLESCALING_END, 1879}; 1880extern enum sched_tunable_scaling sysctl_sched_tunable_scaling; 1881 1882#ifdef CONFIG_SCHED_DEBUG 1883extern unsigned int sysctl_sched_migration_cost; 1884extern unsigned int sysctl_sched_nr_migrate; 1885extern unsigned int sysctl_sched_time_avg; 1886extern unsigned int sysctl_timer_migration; 1887 1888int sched_proc_update_handler(struct ctl_table *table, int write, 1889 void __user *buffer, size_t *length, 1890 loff_t *ppos); 1891#endif 1892#ifdef CONFIG_SCHED_DEBUG 1893static inline unsigned int get_sysctl_timer_migration(void) 1894{ 1895 return sysctl_timer_migration; 1896} 1897#else 1898static inline unsigned int get_sysctl_timer_migration(void) 1899{ 1900 return 1; 1901} 1902#endif 1903extern unsigned int sysctl_sched_rt_period; 1904extern int sysctl_sched_rt_runtime; 1905 1906int sched_rt_handler(struct ctl_table *table, int write, 1907 void __user *buffer, size_t *lenp, 1908 loff_t *ppos); 1909 1910extern unsigned int sysctl_sched_compat_yield; 1911 1912#ifdef CONFIG_RT_MUTEXES 1913extern int rt_mutex_getprio(struct task_struct *p); 1914extern void rt_mutex_setprio(struct task_struct *p, int prio); 1915extern void rt_mutex_adjust_pi(struct task_struct *p); 1916#else 1917static inline int rt_mutex_getprio(struct task_struct *p) 1918{ 1919 return p->normal_prio; 1920} 1921# define rt_mutex_adjust_pi(p) do { } while (0) 1922#endif 1923 1924extern void set_user_nice(struct task_struct *p, long nice); 1925extern int task_prio(const struct task_struct *p); 1926extern int task_nice(const struct task_struct *p); 1927extern int can_nice(const struct task_struct *p, const int nice); 1928extern int task_curr(const struct task_struct *p); 1929extern int idle_cpu(int cpu); 1930extern int sched_setscheduler(struct task_struct *, int, struct sched_param *); 1931extern int sched_setscheduler_nocheck(struct task_struct *, int, 1932 struct sched_param *); 1933extern struct task_struct *idle_task(int cpu); 1934extern struct task_struct *curr_task(int cpu); 1935extern void set_curr_task(int cpu, struct task_struct *p); 1936 1937void yield(void); 1938 1939/* 1940 * The default (Linux) execution domain. 1941 */ 1942extern struct exec_domain default_exec_domain; 1943 1944union thread_union { 1945 struct thread_info thread_info; 1946 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1947}; 1948 1949#ifndef __HAVE_ARCH_KSTACK_END 1950static inline int kstack_end(void *addr) 1951{ 1952 /* Reliable end of stack detection: 1953 * Some APM bios versions misalign the stack 1954 */ 1955 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*))); 1956} 1957#endif 1958 1959extern union thread_union init_thread_union; 1960extern struct task_struct init_task; 1961 1962extern struct mm_struct init_mm; 1963 1964extern struct pid_namespace init_pid_ns; 1965 1966/* 1967 * find a task by one of its numerical ids 1968 * 1969 * find_task_by_pid_ns(): 1970 * finds a task by its pid in the specified namespace 1971 * find_task_by_vpid(): 1972 * finds a task by its virtual pid 1973 * 1974 * see also find_vpid() etc in include/linux/pid.h 1975 */ 1976 1977extern struct task_struct *find_task_by_vpid(pid_t nr); 1978extern struct task_struct *find_task_by_pid_ns(pid_t nr, 1979 struct pid_namespace *ns); 1980 1981extern void __set_special_pids(struct pid *pid); 1982 1983/* per-UID process charging. */ 1984extern struct user_struct * alloc_uid(struct user_namespace *, uid_t); 1985static inline struct user_struct *get_uid(struct user_struct *u) 1986{ 1987 atomic_inc(&u->__count); 1988 return u; 1989} 1990extern void free_uid(struct user_struct *); 1991extern void release_uids(struct user_namespace *ns); 1992 1993#include <asm/current.h> 1994 1995extern void do_timer(unsigned long ticks); 1996 1997extern int wake_up_state(struct task_struct *tsk, unsigned int state); 1998extern int wake_up_process(struct task_struct *tsk); 1999extern void wake_up_new_task(struct task_struct *tsk, 2000 unsigned long clone_flags); 2001#ifdef CONFIG_SMP 2002 extern void kick_process(struct task_struct *tsk); 2003#else 2004 static inline void kick_process(struct task_struct *tsk) { } 2005#endif 2006extern void sched_fork(struct task_struct *p, int clone_flags); 2007extern void sched_dead(struct task_struct *p); 2008 2009extern void proc_caches_init(void); 2010extern void flush_signals(struct task_struct *); 2011extern void __flush_signals(struct task_struct *); 2012extern void ignore_signals(struct task_struct *); 2013extern void flush_signal_handlers(struct task_struct *, int force_default); 2014extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info); 2015 2016static inline int dequeue_signal_lock(struct task_struct *tsk, sigset_t *mask, siginfo_t *info) 2017{ 2018 unsigned long flags; 2019 int ret; 2020 2021 spin_lock_irqsave(&tsk->sighand->siglock, flags); 2022 ret = dequeue_signal(tsk, mask, info); 2023 spin_unlock_irqrestore(&tsk->sighand->siglock, flags); 2024 2025 return ret; 2026} 2027 2028extern void block_all_signals(int (*notifier)(void *priv), void *priv, 2029 sigset_t *mask); 2030extern void unblock_all_signals(void); 2031extern void release_task(struct task_struct * p); 2032extern int send_sig_info(int, struct siginfo *, struct task_struct *); 2033extern int force_sigsegv(int, struct task_struct *); 2034extern int force_sig_info(int, struct siginfo *, struct task_struct *); 2035extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp); 2036extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid); 2037extern int kill_pid_info_as_uid(int, struct siginfo *, struct pid *, uid_t, uid_t, u32); 2038extern int kill_pgrp(struct pid *pid, int sig, int priv); 2039extern int kill_pid(struct pid *pid, int sig, int priv); 2040extern int kill_proc_info(int, struct siginfo *, pid_t); 2041extern int do_notify_parent(struct task_struct *, int); 2042extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent); 2043extern void force_sig(int, struct task_struct *); 2044extern int send_sig(int, struct task_struct *, int); 2045extern void zap_other_threads(struct task_struct *p); 2046extern struct sigqueue *sigqueue_alloc(void); 2047extern void sigqueue_free(struct sigqueue *); 2048extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group); 2049extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *); 2050extern int do_sigaltstack(const stack_t __user *, stack_t __user *, unsigned long); 2051 2052static inline int kill_cad_pid(int sig, int priv) 2053{ 2054 return kill_pid(cad_pid, sig, priv); 2055} 2056 2057/* These can be the second arg to send_sig_info/send_group_sig_info. */ 2058#define SEND_SIG_NOINFO ((struct siginfo *) 0) 2059#define SEND_SIG_PRIV ((struct siginfo *) 1) 2060#define SEND_SIG_FORCED ((struct siginfo *) 2) 2061 2062/* 2063 * True if we are on the alternate signal stack. 2064 */ 2065static inline int on_sig_stack(unsigned long sp) 2066{ 2067#ifdef CONFIG_STACK_GROWSUP 2068 return sp >= current->sas_ss_sp && 2069 sp - current->sas_ss_sp < current->sas_ss_size; 2070#else 2071 return sp > current->sas_ss_sp && 2072 sp - current->sas_ss_sp <= current->sas_ss_size; 2073#endif 2074} 2075 2076static inline int sas_ss_flags(unsigned long sp) 2077{ 2078 return (current->sas_ss_size == 0 ? SS_DISABLE 2079 : on_sig_stack(sp) ? SS_ONSTACK : 0); 2080} 2081 2082/* 2083 * Routines for handling mm_structs 2084 */ 2085extern struct mm_struct * mm_alloc(void); 2086 2087/* mmdrop drops the mm and the page tables */ 2088extern void __mmdrop(struct mm_struct *); 2089static inline void mmdrop(struct mm_struct * mm) 2090{ 2091 if (unlikely(atomic_dec_and_test(&mm->mm_count))) 2092 __mmdrop(mm); 2093} 2094 2095/* mmput gets rid of the mappings and all user-space */ 2096extern void mmput(struct mm_struct *); 2097/* Grab a reference to a task's mm, if it is not already going away */ 2098extern struct mm_struct *get_task_mm(struct task_struct *task); 2099/* Remove the current tasks stale references to the old mm_struct */ 2100extern void mm_release(struct task_struct *, struct mm_struct *); 2101/* Allocate a new mm structure and copy contents from tsk->mm */ 2102extern struct mm_struct *dup_mm(struct task_struct *tsk); 2103 2104extern int copy_thread(unsigned long, unsigned long, unsigned long, 2105 struct task_struct *, struct pt_regs *); 2106extern void flush_thread(void); 2107extern void exit_thread(void); 2108 2109extern void exit_files(struct task_struct *); 2110extern void __cleanup_signal(struct signal_struct *); 2111extern void __cleanup_sighand(struct sighand_struct *); 2112 2113extern void exit_itimers(struct signal_struct *); 2114extern void flush_itimer_signals(void); 2115 2116extern NORET_TYPE void do_group_exit(int); 2117 2118extern void daemonize(const char *, ...); 2119extern int allow_signal(int); 2120extern int disallow_signal(int); 2121 2122extern int do_execve(char *, char __user * __user *, char __user * __user *, struct pt_regs *); 2123extern long do_fork(unsigned long, unsigned long, struct pt_regs *, unsigned long, int __user *, int __user *); 2124struct task_struct *fork_idle(int); 2125 2126extern void set_task_comm(struct task_struct *tsk, char *from); 2127extern char *get_task_comm(char *to, struct task_struct *tsk); 2128 2129#ifdef CONFIG_SMP 2130extern void wait_task_context_switch(struct task_struct *p); 2131extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 2132#else 2133static inline void wait_task_context_switch(struct task_struct *p) {} 2134static inline unsigned long wait_task_inactive(struct task_struct *p, 2135 long match_state) 2136{ 2137 return 1; 2138} 2139#endif 2140 2141#define next_task(p) \ 2142 list_entry_rcu((p)->tasks.next, struct task_struct, tasks) 2143 2144#define for_each_process(p) \ 2145 for (p = &init_task ; (p = next_task(p)) != &init_task ; ) 2146 2147extern bool current_is_single_threaded(void); 2148 2149/* 2150 * Careful: do_each_thread/while_each_thread is a double loop so 2151 * 'break' will not work as expected - use goto instead. 2152 */ 2153#define do_each_thread(g, t) \ 2154 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do 2155 2156#define while_each_thread(g, t) \ 2157 while ((t = next_thread(t)) != g) 2158 2159/* de_thread depends on thread_group_leader not being a pid based check */ 2160#define thread_group_leader(p) (p == p->group_leader) 2161 2162/* Do to the insanities of de_thread it is possible for a process 2163 * to have the pid of the thread group leader without actually being 2164 * the thread group leader. For iteration through the pids in proc 2165 * all we care about is that we have a task with the appropriate 2166 * pid, we don't actually care if we have the right task. 2167 */ 2168static inline int has_group_leader_pid(struct task_struct *p) 2169{ 2170 return p->pid == p->tgid; 2171} 2172 2173static inline 2174int same_thread_group(struct task_struct *p1, struct task_struct *p2) 2175{ 2176 return p1->tgid == p2->tgid; 2177} 2178 2179static inline struct task_struct *next_thread(const struct task_struct *p) 2180{ 2181 return list_entry_rcu(p->thread_group.next, 2182 struct task_struct, thread_group); 2183} 2184 2185static inline int thread_group_empty(struct task_struct *p) 2186{ 2187 return list_empty(&p->thread_group); 2188} 2189 2190#define delay_group_leader(p) \ 2191 (thread_group_leader(p) && !thread_group_empty(p)) 2192 2193static inline int task_detached(struct task_struct *p) 2194{ 2195 return p->exit_signal == -1; 2196} 2197 2198/* 2199 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring 2200 * subscriptions and synchronises with wait4(). Also used in procfs. Also 2201 * pins the final release of task.io_context. Also protects ->cpuset and 2202 * ->cgroup.subsys[]. 2203 * 2204 * Nests both inside and outside of read_lock(&tasklist_lock). 2205 * It must not be nested with write_lock_irq(&tasklist_lock), 2206 * neither inside nor outside. 2207 */ 2208static inline void task_lock(struct task_struct *p) 2209{ 2210 spin_lock(&p->alloc_lock); 2211} 2212 2213static inline void task_unlock(struct task_struct *p) 2214{ 2215 spin_unlock(&p->alloc_lock); 2216} 2217 2218extern struct sighand_struct *lock_task_sighand(struct task_struct *tsk, 2219 unsigned long *flags); 2220 2221static inline void unlock_task_sighand(struct task_struct *tsk, 2222 unsigned long *flags) 2223{ 2224 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags); 2225} 2226 2227#ifndef __HAVE_THREAD_FUNCTIONS 2228 2229#define task_thread_info(task) ((struct thread_info *)(task)->stack) 2230#define task_stack_page(task) ((task)->stack) 2231 2232static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org) 2233{ 2234 *task_thread_info(p) = *task_thread_info(org); 2235 task_thread_info(p)->task = p; 2236} 2237 2238static inline unsigned long *end_of_stack(struct task_struct *p) 2239{ 2240 return (unsigned long *)(task_thread_info(p) + 1); 2241} 2242 2243#endif 2244 2245static inline int object_is_on_stack(void *obj) 2246{ 2247 void *stack = task_stack_page(current); 2248 2249 return (obj >= stack) && (obj < (stack + THREAD_SIZE)); 2250} 2251 2252extern void thread_info_cache_init(void); 2253 2254#ifdef CONFIG_DEBUG_STACK_USAGE 2255static inline unsigned long stack_not_used(struct task_struct *p) 2256{ 2257 unsigned long *n = end_of_stack(p); 2258 2259 do { /* Skip over canary */ 2260 n++; 2261 } while (!*n); 2262 2263 return (unsigned long)n - (unsigned long)end_of_stack(p); 2264} 2265#endif 2266 2267/* set thread flags in other task's structures 2268 * - see asm/thread_info.h for TIF_xxxx flags available 2269 */ 2270static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 2271{ 2272 set_ti_thread_flag(task_thread_info(tsk), flag); 2273} 2274 2275static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2276{ 2277 clear_ti_thread_flag(task_thread_info(tsk), flag); 2278} 2279 2280static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2281{ 2282 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2283} 2284 2285static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2286{ 2287 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2288} 2289 2290static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2291{ 2292 return test_ti_thread_flag(task_thread_info(tsk), flag); 2293} 2294 2295static inline void set_tsk_need_resched(struct task_struct *tsk) 2296{ 2297 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2298} 2299 2300static inline void clear_tsk_need_resched(struct task_struct *tsk) 2301{ 2302 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2303} 2304 2305static inline int test_tsk_need_resched(struct task_struct *tsk) 2306{ 2307 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2308} 2309 2310static inline int restart_syscall(void) 2311{ 2312 set_tsk_thread_flag(current, TIF_SIGPENDING); 2313 return -ERESTARTNOINTR; 2314} 2315 2316static inline int signal_pending(struct task_struct *p) 2317{ 2318 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING)); 2319} 2320 2321static inline int __fatal_signal_pending(struct task_struct *p) 2322{ 2323 return unlikely(sigismember(&p->pending.signal, SIGKILL)); 2324} 2325 2326static inline int fatal_signal_pending(struct task_struct *p) 2327{ 2328 return signal_pending(p) && __fatal_signal_pending(p); 2329} 2330 2331static inline int signal_pending_state(long state, struct task_struct *p) 2332{ 2333 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL))) 2334 return 0; 2335 if (!signal_pending(p)) 2336 return 0; 2337 2338 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p); 2339} 2340 2341static inline int need_resched(void) 2342{ 2343 return unlikely(test_thread_flag(TIF_NEED_RESCHED)); 2344} 2345 2346/* 2347 * cond_resched() and cond_resched_lock(): latency reduction via 2348 * explicit rescheduling in places that are safe. The return 2349 * value indicates whether a reschedule was done in fact. 2350 * cond_resched_lock() will drop the spinlock before scheduling, 2351 * cond_resched_softirq() will enable bhs before scheduling. 2352 */ 2353extern int _cond_resched(void); 2354 2355#define cond_resched() ({ \ 2356 __might_sleep(__FILE__, __LINE__, 0); \ 2357 _cond_resched(); \ 2358}) 2359 2360extern int __cond_resched_lock(spinlock_t *lock); 2361 2362#ifdef CONFIG_PREEMPT 2363#define PREEMPT_LOCK_OFFSET PREEMPT_OFFSET 2364#else 2365#define PREEMPT_LOCK_OFFSET 0 2366#endif 2367 2368#define cond_resched_lock(lock) ({ \ 2369 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \ 2370 __cond_resched_lock(lock); \ 2371}) 2372 2373extern int __cond_resched_softirq(void); 2374 2375#define cond_resched_softirq() ({ \ 2376 __might_sleep(__FILE__, __LINE__, SOFTIRQ_OFFSET); \ 2377 __cond_resched_softirq(); \ 2378}) 2379 2380/* 2381 * Does a critical section need to be broken due to another 2382 * task waiting?: (technically does not depend on CONFIG_PREEMPT, 2383 * but a general need for low latency) 2384 */ 2385static inline int spin_needbreak(spinlock_t *lock) 2386{ 2387#ifdef CONFIG_PREEMPT 2388 return spin_is_contended(lock); 2389#else 2390 return 0; 2391#endif 2392} 2393 2394/* 2395 * Thread group CPU time accounting. 2396 */ 2397void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times); 2398void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times); 2399 2400static inline void thread_group_cputime_init(struct signal_struct *sig) 2401{ 2402 spin_lock_init(&sig->cputimer.lock); 2403} 2404 2405static inline void thread_group_cputime_free(struct signal_struct *sig) 2406{ 2407} 2408 2409/* 2410 * Reevaluate whether the task has signals pending delivery. 2411 * Wake the task if so. 2412 * This is required every time the blocked sigset_t changes. 2413 * callers must hold sighand->siglock. 2414 */ 2415extern void recalc_sigpending_and_wake(struct task_struct *t); 2416extern void recalc_sigpending(void); 2417 2418extern void signal_wake_up(struct task_struct *t, int resume_stopped); 2419 2420/* 2421 * Wrappers for p->thread_info->cpu access. No-op on UP. 2422 */ 2423#ifdef CONFIG_SMP 2424 2425static inline unsigned int task_cpu(const struct task_struct *p) 2426{ 2427 return task_thread_info(p)->cpu; 2428} 2429 2430extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2431 2432#else 2433 2434static inline unsigned int task_cpu(const struct task_struct *p) 2435{ 2436 return 0; 2437} 2438 2439static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2440{ 2441} 2442 2443#endif /* CONFIG_SMP */ 2444 2445#ifdef CONFIG_TRACING 2446extern void 2447__trace_special(void *__tr, void *__data, 2448 unsigned long arg1, unsigned long arg2, unsigned long arg3); 2449#else 2450static inline void 2451__trace_special(void *__tr, void *__data, 2452 unsigned long arg1, unsigned long arg2, unsigned long arg3) 2453{ 2454} 2455#endif 2456 2457extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2458extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2459 2460extern void normalize_rt_tasks(void); 2461 2462#ifdef CONFIG_CGROUP_SCHED 2463 2464extern struct task_group init_task_group; 2465 2466extern struct task_group *sched_create_group(struct task_group *parent); 2467extern void sched_destroy_group(struct task_group *tg); 2468extern void sched_move_task(struct task_struct *tsk); 2469#ifdef CONFIG_FAIR_GROUP_SCHED 2470extern int sched_group_set_shares(struct task_group *tg, unsigned long shares); 2471extern unsigned long sched_group_shares(struct task_group *tg); 2472#endif 2473#ifdef CONFIG_RT_GROUP_SCHED 2474extern int sched_group_set_rt_runtime(struct task_group *tg, 2475 long rt_runtime_us); 2476extern long sched_group_rt_runtime(struct task_group *tg); 2477extern int sched_group_set_rt_period(struct task_group *tg, 2478 long rt_period_us); 2479extern long sched_group_rt_period(struct task_group *tg); 2480extern int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk); 2481#endif 2482#endif 2483 2484extern int task_can_switch_user(struct user_struct *up, 2485 struct task_struct *tsk); 2486 2487#ifdef CONFIG_TASK_XACCT 2488static inline void add_rchar(struct task_struct *tsk, ssize_t amt) 2489{ 2490 tsk->ioac.rchar += amt; 2491} 2492 2493static inline void add_wchar(struct task_struct *tsk, ssize_t amt) 2494{ 2495 tsk->ioac.wchar += amt; 2496} 2497 2498static inline void inc_syscr(struct task_struct *tsk) 2499{ 2500 tsk->ioac.syscr++; 2501} 2502 2503static inline void inc_syscw(struct task_struct *tsk) 2504{ 2505 tsk->ioac.syscw++; 2506} 2507#else 2508static inline void add_rchar(struct task_struct *tsk, ssize_t amt) 2509{ 2510} 2511 2512static inline void add_wchar(struct task_struct *tsk, ssize_t amt) 2513{ 2514} 2515 2516static inline void inc_syscr(struct task_struct *tsk) 2517{ 2518} 2519 2520static inline void inc_syscw(struct task_struct *tsk) 2521{ 2522} 2523#endif 2524 2525#ifndef TASK_SIZE_OF 2526#define TASK_SIZE_OF(tsk) TASK_SIZE 2527#endif 2528 2529/* 2530 * Call the function if the target task is executing on a CPU right now: 2531 */ 2532extern void task_oncpu_function_call(struct task_struct *p, 2533 void (*func) (void *info), void *info); 2534 2535 2536#ifdef CONFIG_MM_OWNER 2537extern void mm_update_next_owner(struct mm_struct *mm); 2538extern void mm_init_owner(struct mm_struct *mm, struct task_struct *p); 2539#else 2540static inline void mm_update_next_owner(struct mm_struct *mm) 2541{ 2542} 2543 2544static inline void mm_init_owner(struct mm_struct *mm, struct task_struct *p) 2545{ 2546} 2547#endif /* CONFIG_MM_OWNER */ 2548 2549static inline unsigned long task_rlimit(const struct task_struct *tsk, 2550 unsigned int limit) 2551{ 2552 return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_cur); 2553} 2554 2555static inline unsigned long task_rlimit_max(const struct task_struct *tsk, 2556 unsigned int limit) 2557{ 2558 return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_max); 2559} 2560 2561static inline unsigned long rlimit(unsigned int limit) 2562{ 2563 return task_rlimit(current, limit); 2564} 2565 2566static inline unsigned long rlimit_max(unsigned int limit) 2567{ 2568 return task_rlimit_max(current, limit); 2569} 2570 2571#endif /* __KERNEL__ */ 2572 2573#endif 2574