sched.h revision c66f08be7e3ad0a28bcd9a0aef766fdf08ea0ec6
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_NEWNET 0x40000000 /* New network namespace */ 29 30/* 31 * Scheduling policies 32 */ 33#define SCHED_NORMAL 0 34#define SCHED_FIFO 1 35#define SCHED_RR 2 36#define SCHED_BATCH 3 37/* SCHED_ISO: reserved but not implemented yet */ 38#define SCHED_IDLE 5 39 40#ifdef __KERNEL__ 41 42struct sched_param { 43 int sched_priority; 44}; 45 46#include <asm/param.h> /* for HZ */ 47 48#include <linux/capability.h> 49#include <linux/threads.h> 50#include <linux/kernel.h> 51#include <linux/types.h> 52#include <linux/timex.h> 53#include <linux/jiffies.h> 54#include <linux/rbtree.h> 55#include <linux/thread_info.h> 56#include <linux/cpumask.h> 57#include <linux/errno.h> 58#include <linux/nodemask.h> 59#include <linux/mm_types.h> 60 61#include <asm/system.h> 62#include <asm/semaphore.h> 63#include <asm/page.h> 64#include <asm/ptrace.h> 65#include <asm/cputime.h> 66 67#include <linux/smp.h> 68#include <linux/sem.h> 69#include <linux/signal.h> 70#include <linux/securebits.h> 71#include <linux/fs_struct.h> 72#include <linux/compiler.h> 73#include <linux/completion.h> 74#include <linux/pid.h> 75#include <linux/percpu.h> 76#include <linux/topology.h> 77#include <linux/proportions.h> 78#include <linux/seccomp.h> 79#include <linux/rcupdate.h> 80#include <linux/futex.h> 81#include <linux/rtmutex.h> 82 83#include <linux/time.h> 84#include <linux/param.h> 85#include <linux/resource.h> 86#include <linux/timer.h> 87#include <linux/hrtimer.h> 88#include <linux/task_io_accounting.h> 89#include <linux/kobject.h> 90 91#include <asm/processor.h> 92 93struct exec_domain; 94struct futex_pi_state; 95struct bio; 96 97/* 98 * List of flags we want to share for kernel threads, 99 * if only because they are not used by them anyway. 100 */ 101#define CLONE_KERNEL (CLONE_FS | CLONE_FILES | CLONE_SIGHAND) 102 103/* 104 * These are the constant used to fake the fixed-point load-average 105 * counting. Some notes: 106 * - 11 bit fractions expand to 22 bits by the multiplies: this gives 107 * a load-average precision of 10 bits integer + 11 bits fractional 108 * - if you want to count load-averages more often, you need more 109 * precision, or rounding will get you. With 2-second counting freq, 110 * the EXP_n values would be 1981, 2034 and 2043 if still using only 111 * 11 bit fractions. 112 */ 113extern unsigned long avenrun[]; /* Load averages */ 114 115#define FSHIFT 11 /* nr of bits of precision */ 116#define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */ 117#define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */ 118#define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */ 119#define EXP_5 2014 /* 1/exp(5sec/5min) */ 120#define EXP_15 2037 /* 1/exp(5sec/15min) */ 121 122#define CALC_LOAD(load,exp,n) \ 123 load *= exp; \ 124 load += n*(FIXED_1-exp); \ 125 load >>= FSHIFT; 126 127extern unsigned long total_forks; 128extern int nr_threads; 129DECLARE_PER_CPU(unsigned long, process_counts); 130extern int nr_processes(void); 131extern unsigned long nr_running(void); 132extern unsigned long nr_uninterruptible(void); 133extern unsigned long nr_active(void); 134extern unsigned long nr_iowait(void); 135extern unsigned long weighted_cpuload(const int cpu); 136 137struct seq_file; 138struct cfs_rq; 139struct task_group; 140#ifdef CONFIG_SCHED_DEBUG 141extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m); 142extern void proc_sched_set_task(struct task_struct *p); 143extern void 144print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq); 145#else 146static inline void 147proc_sched_show_task(struct task_struct *p, struct seq_file *m) 148{ 149} 150static inline void proc_sched_set_task(struct task_struct *p) 151{ 152} 153static inline void 154print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq) 155{ 156} 157#endif 158 159/* 160 * Task state bitmask. NOTE! These bits are also 161 * encoded in fs/proc/array.c: get_task_state(). 162 * 163 * We have two separate sets of flags: task->state 164 * is about runnability, while task->exit_state are 165 * about the task exiting. Confusing, but this way 166 * modifying one set can't modify the other one by 167 * mistake. 168 */ 169#define TASK_RUNNING 0 170#define TASK_INTERRUPTIBLE 1 171#define TASK_UNINTERRUPTIBLE 2 172#define TASK_STOPPED 4 173#define TASK_TRACED 8 174/* in tsk->exit_state */ 175#define EXIT_ZOMBIE 16 176#define EXIT_DEAD 32 177/* in tsk->state again */ 178#define TASK_DEAD 64 179 180#define __set_task_state(tsk, state_value) \ 181 do { (tsk)->state = (state_value); } while (0) 182#define set_task_state(tsk, state_value) \ 183 set_mb((tsk)->state, (state_value)) 184 185/* 186 * set_current_state() includes a barrier so that the write of current->state 187 * is correctly serialised wrt the caller's subsequent test of whether to 188 * actually sleep: 189 * 190 * set_current_state(TASK_UNINTERRUPTIBLE); 191 * if (do_i_need_to_sleep()) 192 * schedule(); 193 * 194 * If the caller does not need such serialisation then use __set_current_state() 195 */ 196#define __set_current_state(state_value) \ 197 do { current->state = (state_value); } while (0) 198#define set_current_state(state_value) \ 199 set_mb(current->state, (state_value)) 200 201/* Task command name length */ 202#define TASK_COMM_LEN 16 203 204#include <linux/spinlock.h> 205 206/* 207 * This serializes "schedule()" and also protects 208 * the run-queue from deletions/modifications (but 209 * _adding_ to the beginning of the run-queue has 210 * a separate lock). 211 */ 212extern rwlock_t tasklist_lock; 213extern spinlock_t mmlist_lock; 214 215struct task_struct; 216 217extern void sched_init(void); 218extern void sched_init_smp(void); 219extern void init_idle(struct task_struct *idle, int cpu); 220extern void init_idle_bootup_task(struct task_struct *idle); 221 222extern cpumask_t nohz_cpu_mask; 223#if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ) 224extern int select_nohz_load_balancer(int cpu); 225#else 226static inline int select_nohz_load_balancer(int cpu) 227{ 228 return 0; 229} 230#endif 231 232/* 233 * Only dump TASK_* tasks. (0 for all tasks) 234 */ 235extern void show_state_filter(unsigned long state_filter); 236 237static inline void show_state(void) 238{ 239 show_state_filter(0); 240} 241 242extern void show_regs(struct pt_regs *); 243 244/* 245 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current 246 * task), SP is the stack pointer of the first frame that should be shown in the back 247 * trace (or NULL if the entire call-chain of the task should be shown). 248 */ 249extern void show_stack(struct task_struct *task, unsigned long *sp); 250 251void io_schedule(void); 252long io_schedule_timeout(long timeout); 253 254extern void cpu_init (void); 255extern void trap_init(void); 256extern void update_process_times(int user); 257extern void scheduler_tick(void); 258 259#ifdef CONFIG_DETECT_SOFTLOCKUP 260extern void softlockup_tick(void); 261extern void spawn_softlockup_task(void); 262extern void touch_softlockup_watchdog(void); 263extern void touch_all_softlockup_watchdogs(void); 264extern int softlockup_thresh; 265#else 266static inline void softlockup_tick(void) 267{ 268} 269static inline void spawn_softlockup_task(void) 270{ 271} 272static inline void touch_softlockup_watchdog(void) 273{ 274} 275static inline void touch_all_softlockup_watchdogs(void) 276{ 277} 278#endif 279 280 281/* Attach to any functions which should be ignored in wchan output. */ 282#define __sched __attribute__((__section__(".sched.text"))) 283/* Is this address in the __sched functions? */ 284extern int in_sched_functions(unsigned long addr); 285 286#define MAX_SCHEDULE_TIMEOUT LONG_MAX 287extern signed long FASTCALL(schedule_timeout(signed long timeout)); 288extern signed long schedule_timeout_interruptible(signed long timeout); 289extern signed long schedule_timeout_uninterruptible(signed long timeout); 290asmlinkage void schedule(void); 291 292struct nsproxy; 293struct user_namespace; 294 295/* Maximum number of active map areas.. This is a random (large) number */ 296#define DEFAULT_MAX_MAP_COUNT 65536 297 298extern int sysctl_max_map_count; 299 300#include <linux/aio.h> 301 302extern unsigned long 303arch_get_unmapped_area(struct file *, unsigned long, unsigned long, 304 unsigned long, unsigned long); 305extern unsigned long 306arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr, 307 unsigned long len, unsigned long pgoff, 308 unsigned long flags); 309extern void arch_unmap_area(struct mm_struct *, unsigned long); 310extern void arch_unmap_area_topdown(struct mm_struct *, unsigned long); 311 312#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS 313/* 314 * The mm counters are not protected by its page_table_lock, 315 * so must be incremented atomically. 316 */ 317#define set_mm_counter(mm, member, value) atomic_long_set(&(mm)->_##member, value) 318#define get_mm_counter(mm, member) ((unsigned long)atomic_long_read(&(mm)->_##member)) 319#define add_mm_counter(mm, member, value) atomic_long_add(value, &(mm)->_##member) 320#define inc_mm_counter(mm, member) atomic_long_inc(&(mm)->_##member) 321#define dec_mm_counter(mm, member) atomic_long_dec(&(mm)->_##member) 322 323#else /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */ 324/* 325 * The mm counters are protected by its page_table_lock, 326 * so can be incremented directly. 327 */ 328#define set_mm_counter(mm, member, value) (mm)->_##member = (value) 329#define get_mm_counter(mm, member) ((mm)->_##member) 330#define add_mm_counter(mm, member, value) (mm)->_##member += (value) 331#define inc_mm_counter(mm, member) (mm)->_##member++ 332#define dec_mm_counter(mm, member) (mm)->_##member-- 333 334#endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */ 335 336#define get_mm_rss(mm) \ 337 (get_mm_counter(mm, file_rss) + get_mm_counter(mm, anon_rss)) 338#define update_hiwater_rss(mm) do { \ 339 unsigned long _rss = get_mm_rss(mm); \ 340 if ((mm)->hiwater_rss < _rss) \ 341 (mm)->hiwater_rss = _rss; \ 342} while (0) 343#define update_hiwater_vm(mm) do { \ 344 if ((mm)->hiwater_vm < (mm)->total_vm) \ 345 (mm)->hiwater_vm = (mm)->total_vm; \ 346} while (0) 347 348extern void set_dumpable(struct mm_struct *mm, int value); 349extern int get_dumpable(struct mm_struct *mm); 350 351/* mm flags */ 352/* dumpable bits */ 353#define MMF_DUMPABLE 0 /* core dump is permitted */ 354#define MMF_DUMP_SECURELY 1 /* core file is readable only by root */ 355#define MMF_DUMPABLE_BITS 2 356 357/* coredump filter bits */ 358#define MMF_DUMP_ANON_PRIVATE 2 359#define MMF_DUMP_ANON_SHARED 3 360#define MMF_DUMP_MAPPED_PRIVATE 4 361#define MMF_DUMP_MAPPED_SHARED 5 362#define MMF_DUMP_ELF_HEADERS 6 363#define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS 364#define MMF_DUMP_FILTER_BITS 5 365#define MMF_DUMP_FILTER_MASK \ 366 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT) 367#define MMF_DUMP_FILTER_DEFAULT \ 368 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED)) 369 370struct sighand_struct { 371 atomic_t count; 372 struct k_sigaction action[_NSIG]; 373 spinlock_t siglock; 374 wait_queue_head_t signalfd_wqh; 375}; 376 377struct pacct_struct { 378 int ac_flag; 379 long ac_exitcode; 380 unsigned long ac_mem; 381 cputime_t ac_utime, ac_stime; 382 unsigned long ac_minflt, ac_majflt; 383}; 384 385/* 386 * NOTE! "signal_struct" does not have it's own 387 * locking, because a shared signal_struct always 388 * implies a shared sighand_struct, so locking 389 * sighand_struct is always a proper superset of 390 * the locking of signal_struct. 391 */ 392struct signal_struct { 393 atomic_t count; 394 atomic_t live; 395 396 wait_queue_head_t wait_chldexit; /* for wait4() */ 397 398 /* current thread group signal load-balancing target: */ 399 struct task_struct *curr_target; 400 401 /* shared signal handling: */ 402 struct sigpending shared_pending; 403 404 /* thread group exit support */ 405 int group_exit_code; 406 /* overloaded: 407 * - notify group_exit_task when ->count is equal to notify_count 408 * - everyone except group_exit_task is stopped during signal delivery 409 * of fatal signals, group_exit_task processes the signal. 410 */ 411 struct task_struct *group_exit_task; 412 int notify_count; 413 414 /* thread group stop support, overloads group_exit_code too */ 415 int group_stop_count; 416 unsigned int flags; /* see SIGNAL_* flags below */ 417 418 /* POSIX.1b Interval Timers */ 419 struct list_head posix_timers; 420 421 /* ITIMER_REAL timer for the process */ 422 struct hrtimer real_timer; 423 struct task_struct *tsk; 424 ktime_t it_real_incr; 425 426 /* ITIMER_PROF and ITIMER_VIRTUAL timers for the process */ 427 cputime_t it_prof_expires, it_virt_expires; 428 cputime_t it_prof_incr, it_virt_incr; 429 430 /* job control IDs */ 431 pid_t pgrp; 432 struct pid *tty_old_pgrp; 433 434 union { 435 pid_t session __deprecated; 436 pid_t __session; 437 }; 438 439 /* boolean value for session group leader */ 440 int leader; 441 442 struct tty_struct *tty; /* NULL if no tty */ 443 444 /* 445 * Cumulative resource counters for dead threads in the group, 446 * and for reaped dead child processes forked by this group. 447 * Live threads maintain their own counters and add to these 448 * in __exit_signal, except for the group leader. 449 */ 450 cputime_t utime, stime, cutime, cstime; 451 cputime_t gtime; 452 cputime_t cgtime; 453 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw; 454 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt; 455 unsigned long inblock, oublock, cinblock, coublock; 456 457 /* 458 * Cumulative ns of scheduled CPU time for dead threads in the 459 * group, not including a zombie group leader. (This only differs 460 * from jiffies_to_ns(utime + stime) if sched_clock uses something 461 * other than jiffies.) 462 */ 463 unsigned long long sum_sched_runtime; 464 465 /* 466 * We don't bother to synchronize most readers of this at all, 467 * because there is no reader checking a limit that actually needs 468 * to get both rlim_cur and rlim_max atomically, and either one 469 * alone is a single word that can safely be read normally. 470 * getrlimit/setrlimit use task_lock(current->group_leader) to 471 * protect this instead of the siglock, because they really 472 * have no need to disable irqs. 473 */ 474 struct rlimit rlim[RLIM_NLIMITS]; 475 476 struct list_head cpu_timers[3]; 477 478 /* keep the process-shared keyrings here so that they do the right 479 * thing in threads created with CLONE_THREAD */ 480#ifdef CONFIG_KEYS 481 struct key *session_keyring; /* keyring inherited over fork */ 482 struct key *process_keyring; /* keyring private to this process */ 483#endif 484#ifdef CONFIG_BSD_PROCESS_ACCT 485 struct pacct_struct pacct; /* per-process accounting information */ 486#endif 487#ifdef CONFIG_TASKSTATS 488 struct taskstats *stats; 489#endif 490#ifdef CONFIG_AUDIT 491 unsigned audit_tty; 492 struct tty_audit_buf *tty_audit_buf; 493#endif 494}; 495 496/* Context switch must be unlocked if interrupts are to be enabled */ 497#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW 498# define __ARCH_WANT_UNLOCKED_CTXSW 499#endif 500 501/* 502 * Bits in flags field of signal_struct. 503 */ 504#define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */ 505#define SIGNAL_STOP_DEQUEUED 0x00000002 /* stop signal dequeued */ 506#define SIGNAL_STOP_CONTINUED 0x00000004 /* SIGCONT since WCONTINUED reap */ 507#define SIGNAL_GROUP_EXIT 0x00000008 /* group exit in progress */ 508 509/* 510 * Some day this will be a full-fledged user tracking system.. 511 */ 512struct user_struct { 513 atomic_t __count; /* reference count */ 514 atomic_t processes; /* How many processes does this user have? */ 515 atomic_t files; /* How many open files does this user have? */ 516 atomic_t sigpending; /* How many pending signals does this user have? */ 517#ifdef CONFIG_INOTIFY_USER 518 atomic_t inotify_watches; /* How many inotify watches does this user have? */ 519 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */ 520#endif 521#ifdef CONFIG_POSIX_MQUEUE 522 /* protected by mq_lock */ 523 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */ 524#endif 525 unsigned long locked_shm; /* How many pages of mlocked shm ? */ 526 527#ifdef CONFIG_KEYS 528 struct key *uid_keyring; /* UID specific keyring */ 529 struct key *session_keyring; /* UID's default session keyring */ 530#endif 531 532 /* Hash table maintenance information */ 533 struct hlist_node uidhash_node; 534 uid_t uid; 535 536#ifdef CONFIG_FAIR_USER_SCHED 537 struct task_group *tg; 538#ifdef CONFIG_SYSFS 539 struct kset kset; 540 struct subsys_attribute user_attr; 541 struct work_struct work; 542#endif 543#endif 544}; 545 546#ifdef CONFIG_FAIR_USER_SCHED 547extern int uids_kobject_init(void); 548#else 549static inline int uids_kobject_init(void) { return 0; } 550#endif 551 552extern struct user_struct *find_user(uid_t); 553 554extern struct user_struct root_user; 555#define INIT_USER (&root_user) 556 557struct backing_dev_info; 558struct reclaim_state; 559 560#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) 561struct sched_info { 562 /* cumulative counters */ 563 unsigned long pcount; /* # of times run on this cpu */ 564 unsigned long long cpu_time, /* time spent on the cpu */ 565 run_delay; /* time spent waiting on a runqueue */ 566 567 /* timestamps */ 568 unsigned long long last_arrival,/* when we last ran on a cpu */ 569 last_queued; /* when we were last queued to run */ 570#ifdef CONFIG_SCHEDSTATS 571 /* BKL stats */ 572 unsigned long bkl_count; 573#endif 574}; 575#endif /* defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) */ 576 577#ifdef CONFIG_SCHEDSTATS 578extern const struct file_operations proc_schedstat_operations; 579#endif /* CONFIG_SCHEDSTATS */ 580 581#ifdef CONFIG_TASK_DELAY_ACCT 582struct task_delay_info { 583 spinlock_t lock; 584 unsigned int flags; /* Private per-task flags */ 585 586 /* For each stat XXX, add following, aligned appropriately 587 * 588 * struct timespec XXX_start, XXX_end; 589 * u64 XXX_delay; 590 * u32 XXX_count; 591 * 592 * Atomicity of updates to XXX_delay, XXX_count protected by 593 * single lock above (split into XXX_lock if contention is an issue). 594 */ 595 596 /* 597 * XXX_count is incremented on every XXX operation, the delay 598 * associated with the operation is added to XXX_delay. 599 * XXX_delay contains the accumulated delay time in nanoseconds. 600 */ 601 struct timespec blkio_start, blkio_end; /* Shared by blkio, swapin */ 602 u64 blkio_delay; /* wait for sync block io completion */ 603 u64 swapin_delay; /* wait for swapin block io completion */ 604 u32 blkio_count; /* total count of the number of sync block */ 605 /* io operations performed */ 606 u32 swapin_count; /* total count of the number of swapin block */ 607 /* io operations performed */ 608}; 609#endif /* CONFIG_TASK_DELAY_ACCT */ 610 611static inline int sched_info_on(void) 612{ 613#ifdef CONFIG_SCHEDSTATS 614 return 1; 615#elif defined(CONFIG_TASK_DELAY_ACCT) 616 extern int delayacct_on; 617 return delayacct_on; 618#else 619 return 0; 620#endif 621} 622 623enum cpu_idle_type { 624 CPU_IDLE, 625 CPU_NOT_IDLE, 626 CPU_NEWLY_IDLE, 627 CPU_MAX_IDLE_TYPES 628}; 629 630/* 631 * sched-domains (multiprocessor balancing) declarations: 632 */ 633 634/* 635 * Increase resolution of nice-level calculations: 636 */ 637#define SCHED_LOAD_SHIFT 10 638#define SCHED_LOAD_SCALE (1L << SCHED_LOAD_SHIFT) 639 640#define SCHED_LOAD_SCALE_FUZZ SCHED_LOAD_SCALE 641 642#ifdef CONFIG_SMP 643#define SD_LOAD_BALANCE 1 /* Do load balancing on this domain. */ 644#define SD_BALANCE_NEWIDLE 2 /* Balance when about to become idle */ 645#define SD_BALANCE_EXEC 4 /* Balance on exec */ 646#define SD_BALANCE_FORK 8 /* Balance on fork, clone */ 647#define SD_WAKE_IDLE 16 /* Wake to idle CPU on task wakeup */ 648#define SD_WAKE_AFFINE 32 /* Wake task to waking CPU */ 649#define SD_WAKE_BALANCE 64 /* Perform balancing at task wakeup */ 650#define SD_SHARE_CPUPOWER 128 /* Domain members share cpu power */ 651#define SD_POWERSAVINGS_BALANCE 256 /* Balance for power savings */ 652#define SD_SHARE_PKG_RESOURCES 512 /* Domain members share cpu pkg resources */ 653#define SD_SERIALIZE 1024 /* Only a single load balancing instance */ 654 655#define BALANCE_FOR_MC_POWER \ 656 (sched_smt_power_savings ? SD_POWERSAVINGS_BALANCE : 0) 657 658#define BALANCE_FOR_PKG_POWER \ 659 ((sched_mc_power_savings || sched_smt_power_savings) ? \ 660 SD_POWERSAVINGS_BALANCE : 0) 661 662#define test_sd_parent(sd, flag) ((sd->parent && \ 663 (sd->parent->flags & flag)) ? 1 : 0) 664 665 666struct sched_group { 667 struct sched_group *next; /* Must be a circular list */ 668 cpumask_t cpumask; 669 670 /* 671 * CPU power of this group, SCHED_LOAD_SCALE being max power for a 672 * single CPU. This is read only (except for setup, hotplug CPU). 673 * Note : Never change cpu_power without recompute its reciprocal 674 */ 675 unsigned int __cpu_power; 676 /* 677 * reciprocal value of cpu_power to avoid expensive divides 678 * (see include/linux/reciprocal_div.h) 679 */ 680 u32 reciprocal_cpu_power; 681}; 682 683struct sched_domain { 684 /* These fields must be setup */ 685 struct sched_domain *parent; /* top domain must be null terminated */ 686 struct sched_domain *child; /* bottom domain must be null terminated */ 687 struct sched_group *groups; /* the balancing groups of the domain */ 688 cpumask_t span; /* span of all CPUs in this domain */ 689 unsigned long min_interval; /* Minimum balance interval ms */ 690 unsigned long max_interval; /* Maximum balance interval ms */ 691 unsigned int busy_factor; /* less balancing by factor if busy */ 692 unsigned int imbalance_pct; /* No balance until over watermark */ 693 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */ 694 unsigned int busy_idx; 695 unsigned int idle_idx; 696 unsigned int newidle_idx; 697 unsigned int wake_idx; 698 unsigned int forkexec_idx; 699 int flags; /* See SD_* */ 700 701 /* Runtime fields. */ 702 unsigned long last_balance; /* init to jiffies. units in jiffies */ 703 unsigned int balance_interval; /* initialise to 1. units in ms. */ 704 unsigned int nr_balance_failed; /* initialise to 0 */ 705 706#ifdef CONFIG_SCHEDSTATS 707 /* load_balance() stats */ 708 unsigned long lb_count[CPU_MAX_IDLE_TYPES]; 709 unsigned long lb_failed[CPU_MAX_IDLE_TYPES]; 710 unsigned long lb_balanced[CPU_MAX_IDLE_TYPES]; 711 unsigned long lb_imbalance[CPU_MAX_IDLE_TYPES]; 712 unsigned long lb_gained[CPU_MAX_IDLE_TYPES]; 713 unsigned long lb_hot_gained[CPU_MAX_IDLE_TYPES]; 714 unsigned long lb_nobusyg[CPU_MAX_IDLE_TYPES]; 715 unsigned long lb_nobusyq[CPU_MAX_IDLE_TYPES]; 716 717 /* Active load balancing */ 718 unsigned long alb_count; 719 unsigned long alb_failed; 720 unsigned long alb_pushed; 721 722 /* SD_BALANCE_EXEC stats */ 723 unsigned long sbe_count; 724 unsigned long sbe_balanced; 725 unsigned long sbe_pushed; 726 727 /* SD_BALANCE_FORK stats */ 728 unsigned long sbf_count; 729 unsigned long sbf_balanced; 730 unsigned long sbf_pushed; 731 732 /* try_to_wake_up() stats */ 733 unsigned long ttwu_wake_remote; 734 unsigned long ttwu_move_affine; 735 unsigned long ttwu_move_balance; 736#endif 737}; 738 739#endif /* CONFIG_SMP */ 740 741/* 742 * A runqueue laden with a single nice 0 task scores a weighted_cpuload of 743 * SCHED_LOAD_SCALE. This function returns 1 if any cpu is laden with a 744 * task of nice 0 or enough lower priority tasks to bring up the 745 * weighted_cpuload 746 */ 747static inline int above_background_load(void) 748{ 749 unsigned long cpu; 750 751 for_each_online_cpu(cpu) { 752 if (weighted_cpuload(cpu) >= SCHED_LOAD_SCALE) 753 return 1; 754 } 755 return 0; 756} 757 758struct io_context; /* See blkdev.h */ 759struct cpuset; 760 761#define NGROUPS_SMALL 32 762#define NGROUPS_PER_BLOCK ((int)(PAGE_SIZE / sizeof(gid_t))) 763struct group_info { 764 int ngroups; 765 atomic_t usage; 766 gid_t small_block[NGROUPS_SMALL]; 767 int nblocks; 768 gid_t *blocks[0]; 769}; 770 771/* 772 * get_group_info() must be called with the owning task locked (via task_lock()) 773 * when task != current. The reason being that the vast majority of callers are 774 * looking at current->group_info, which can not be changed except by the 775 * current task. Changing current->group_info requires the task lock, too. 776 */ 777#define get_group_info(group_info) do { \ 778 atomic_inc(&(group_info)->usage); \ 779} while (0) 780 781#define put_group_info(group_info) do { \ 782 if (atomic_dec_and_test(&(group_info)->usage)) \ 783 groups_free(group_info); \ 784} while (0) 785 786extern struct group_info *groups_alloc(int gidsetsize); 787extern void groups_free(struct group_info *group_info); 788extern int set_current_groups(struct group_info *group_info); 789extern int groups_search(struct group_info *group_info, gid_t grp); 790/* access the groups "array" with this macro */ 791#define GROUP_AT(gi, i) \ 792 ((gi)->blocks[(i)/NGROUPS_PER_BLOCK][(i)%NGROUPS_PER_BLOCK]) 793 794#ifdef ARCH_HAS_PREFETCH_SWITCH_STACK 795extern void prefetch_stack(struct task_struct *t); 796#else 797static inline void prefetch_stack(struct task_struct *t) { } 798#endif 799 800struct audit_context; /* See audit.c */ 801struct mempolicy; 802struct pipe_inode_info; 803struct uts_namespace; 804 805struct rq; 806struct sched_domain; 807 808struct sched_class { 809 const struct sched_class *next; 810 811 void (*enqueue_task) (struct rq *rq, struct task_struct *p, int wakeup); 812 void (*dequeue_task) (struct rq *rq, struct task_struct *p, int sleep); 813 void (*yield_task) (struct rq *rq); 814 815 void (*check_preempt_curr) (struct rq *rq, struct task_struct *p); 816 817 struct task_struct * (*pick_next_task) (struct rq *rq); 818 void (*put_prev_task) (struct rq *rq, struct task_struct *p); 819 820 unsigned long (*load_balance) (struct rq *this_rq, int this_cpu, 821 struct rq *busiest, 822 unsigned long max_nr_move, unsigned long max_load_move, 823 struct sched_domain *sd, enum cpu_idle_type idle, 824 int *all_pinned, int *this_best_prio); 825 826 void (*set_curr_task) (struct rq *rq); 827 void (*task_tick) (struct rq *rq, struct task_struct *p); 828 void (*task_new) (struct rq *rq, struct task_struct *p); 829}; 830 831struct load_weight { 832 unsigned long weight, inv_weight; 833}; 834 835/* 836 * CFS stats for a schedulable entity (task, task-group etc) 837 * 838 * Current field usage histogram: 839 * 840 * 4 se->block_start 841 * 4 se->run_node 842 * 4 se->sleep_start 843 * 6 se->load.weight 844 */ 845struct sched_entity { 846 struct load_weight load; /* for load-balancing */ 847 struct rb_node run_node; 848 unsigned int on_rq; 849 int peer_preempt; 850 851 u64 exec_start; 852 u64 sum_exec_runtime; 853 u64 vruntime; 854 u64 prev_sum_exec_runtime; 855 856#ifdef CONFIG_SCHEDSTATS 857 u64 wait_start; 858 u64 wait_max; 859 860 u64 sleep_start; 861 u64 sleep_max; 862 s64 sum_sleep_runtime; 863 864 u64 block_start; 865 u64 block_max; 866 u64 exec_max; 867 u64 slice_max; 868 869 u64 nr_migrations; 870 u64 nr_migrations_cold; 871 u64 nr_failed_migrations_affine; 872 u64 nr_failed_migrations_running; 873 u64 nr_failed_migrations_hot; 874 u64 nr_forced_migrations; 875 u64 nr_forced2_migrations; 876 877 u64 nr_wakeups; 878 u64 nr_wakeups_sync; 879 u64 nr_wakeups_migrate; 880 u64 nr_wakeups_local; 881 u64 nr_wakeups_remote; 882 u64 nr_wakeups_affine; 883 u64 nr_wakeups_affine_attempts; 884 u64 nr_wakeups_passive; 885 u64 nr_wakeups_idle; 886#endif 887 888#ifdef CONFIG_FAIR_GROUP_SCHED 889 struct sched_entity *parent; 890 /* rq on which this entity is (to be) queued: */ 891 struct cfs_rq *cfs_rq; 892 /* rq "owned" by this entity/group: */ 893 struct cfs_rq *my_q; 894#endif 895}; 896 897struct task_struct { 898 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */ 899 void *stack; 900 atomic_t usage; 901 unsigned int flags; /* per process flags, defined below */ 902 unsigned int ptrace; 903 904 int lock_depth; /* BKL lock depth */ 905 906#ifdef CONFIG_SMP 907#ifdef __ARCH_WANT_UNLOCKED_CTXSW 908 int oncpu; 909#endif 910#endif 911 912 int prio, static_prio, normal_prio; 913 struct list_head run_list; 914 const struct sched_class *sched_class; 915 struct sched_entity se; 916 917#ifdef CONFIG_PREEMPT_NOTIFIERS 918 /* list of struct preempt_notifier: */ 919 struct hlist_head preempt_notifiers; 920#endif 921 922 unsigned short ioprio; 923 /* 924 * fpu_counter contains the number of consecutive context switches 925 * that the FPU is used. If this is over a threshold, the lazy fpu 926 * saving becomes unlazy to save the trap. This is an unsigned char 927 * so that after 256 times the counter wraps and the behavior turns 928 * lazy again; this to deal with bursty apps that only use FPU for 929 * a short time 930 */ 931 unsigned char fpu_counter; 932 s8 oomkilladj; /* OOM kill score adjustment (bit shift). */ 933#ifdef CONFIG_BLK_DEV_IO_TRACE 934 unsigned int btrace_seq; 935#endif 936 937 unsigned int policy; 938 cpumask_t cpus_allowed; 939 unsigned int time_slice; 940 941#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) 942 struct sched_info sched_info; 943#endif 944 945 struct list_head tasks; 946 /* 947 * ptrace_list/ptrace_children forms the list of my children 948 * that were stolen by a ptracer. 949 */ 950 struct list_head ptrace_children; 951 struct list_head ptrace_list; 952 953 struct mm_struct *mm, *active_mm; 954 955/* task state */ 956 struct linux_binfmt *binfmt; 957 int exit_state; 958 int exit_code, exit_signal; 959 int pdeath_signal; /* The signal sent when the parent dies */ 960 /* ??? */ 961 unsigned int personality; 962 unsigned did_exec:1; 963 pid_t pid; 964 pid_t tgid; 965 966#ifdef CONFIG_CC_STACKPROTECTOR 967 /* Canary value for the -fstack-protector gcc feature */ 968 unsigned long stack_canary; 969#endif 970 /* 971 * pointers to (original) parent process, youngest child, younger sibling, 972 * older sibling, respectively. (p->father can be replaced with 973 * p->parent->pid) 974 */ 975 struct task_struct *real_parent; /* real parent process (when being debugged) */ 976 struct task_struct *parent; /* parent process */ 977 /* 978 * children/sibling forms the list of my children plus the 979 * tasks I'm ptracing. 980 */ 981 struct list_head children; /* list of my children */ 982 struct list_head sibling; /* linkage in my parent's children list */ 983 struct task_struct *group_leader; /* threadgroup leader */ 984 985 /* PID/PID hash table linkage. */ 986 struct pid_link pids[PIDTYPE_MAX]; 987 struct list_head thread_group; 988 989 struct completion *vfork_done; /* for vfork() */ 990 int __user *set_child_tid; /* CLONE_CHILD_SETTID */ 991 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */ 992 993 unsigned int rt_priority; 994 cputime_t utime, stime, utimescaled, stimescaled; 995 cputime_t gtime; 996 unsigned long nvcsw, nivcsw; /* context switch counts */ 997 struct timespec start_time; /* monotonic time */ 998 struct timespec real_start_time; /* boot based time */ 999/* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */ 1000 unsigned long min_flt, maj_flt; 1001 1002 cputime_t it_prof_expires, it_virt_expires; 1003 unsigned long long it_sched_expires; 1004 struct list_head cpu_timers[3]; 1005 1006/* process credentials */ 1007 uid_t uid,euid,suid,fsuid; 1008 gid_t gid,egid,sgid,fsgid; 1009 struct group_info *group_info; 1010 kernel_cap_t cap_effective, cap_inheritable, cap_permitted; 1011 unsigned keep_capabilities:1; 1012 struct user_struct *user; 1013#ifdef CONFIG_KEYS 1014 struct key *request_key_auth; /* assumed request_key authority */ 1015 struct key *thread_keyring; /* keyring private to this thread */ 1016 unsigned char jit_keyring; /* default keyring to attach requested keys to */ 1017#endif 1018 char comm[TASK_COMM_LEN]; /* executable name excluding path 1019 - access with [gs]et_task_comm (which lock 1020 it with task_lock()) 1021 - initialized normally by flush_old_exec */ 1022/* file system info */ 1023 int link_count, total_link_count; 1024#ifdef CONFIG_SYSVIPC 1025/* ipc stuff */ 1026 struct sysv_sem sysvsem; 1027#endif 1028/* CPU-specific state of this task */ 1029 struct thread_struct thread; 1030/* filesystem information */ 1031 struct fs_struct *fs; 1032/* open file information */ 1033 struct files_struct *files; 1034/* namespaces */ 1035 struct nsproxy *nsproxy; 1036/* signal handlers */ 1037 struct signal_struct *signal; 1038 struct sighand_struct *sighand; 1039 1040 sigset_t blocked, real_blocked; 1041 sigset_t saved_sigmask; /* To be restored with TIF_RESTORE_SIGMASK */ 1042 struct sigpending pending; 1043 1044 unsigned long sas_ss_sp; 1045 size_t sas_ss_size; 1046 int (*notifier)(void *priv); 1047 void *notifier_data; 1048 sigset_t *notifier_mask; 1049#ifdef CONFIG_SECURITY 1050 void *security; 1051#endif 1052 struct audit_context *audit_context; 1053 seccomp_t seccomp; 1054 1055/* Thread group tracking */ 1056 u32 parent_exec_id; 1057 u32 self_exec_id; 1058/* Protection of (de-)allocation: mm, files, fs, tty, keyrings */ 1059 spinlock_t alloc_lock; 1060 1061 /* Protection of the PI data structures: */ 1062 spinlock_t pi_lock; 1063 1064#ifdef CONFIG_RT_MUTEXES 1065 /* PI waiters blocked on a rt_mutex held by this task */ 1066 struct plist_head pi_waiters; 1067 /* Deadlock detection and priority inheritance handling */ 1068 struct rt_mutex_waiter *pi_blocked_on; 1069#endif 1070 1071#ifdef CONFIG_DEBUG_MUTEXES 1072 /* mutex deadlock detection */ 1073 struct mutex_waiter *blocked_on; 1074#endif 1075#ifdef CONFIG_TRACE_IRQFLAGS 1076 unsigned int irq_events; 1077 int hardirqs_enabled; 1078 unsigned long hardirq_enable_ip; 1079 unsigned int hardirq_enable_event; 1080 unsigned long hardirq_disable_ip; 1081 unsigned int hardirq_disable_event; 1082 int softirqs_enabled; 1083 unsigned long softirq_disable_ip; 1084 unsigned int softirq_disable_event; 1085 unsigned long softirq_enable_ip; 1086 unsigned int softirq_enable_event; 1087 int hardirq_context; 1088 int softirq_context; 1089#endif 1090#ifdef CONFIG_LOCKDEP 1091# define MAX_LOCK_DEPTH 30UL 1092 u64 curr_chain_key; 1093 int lockdep_depth; 1094 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1095 unsigned int lockdep_recursion; 1096#endif 1097 1098/* journalling filesystem info */ 1099 void *journal_info; 1100 1101/* stacked block device info */ 1102 struct bio *bio_list, **bio_tail; 1103 1104/* VM state */ 1105 struct reclaim_state *reclaim_state; 1106 1107 struct backing_dev_info *backing_dev_info; 1108 1109 struct io_context *io_context; 1110 1111 unsigned long ptrace_message; 1112 siginfo_t *last_siginfo; /* For ptrace use. */ 1113#ifdef CONFIG_TASK_XACCT 1114/* i/o counters(bytes read/written, #syscalls */ 1115 u64 rchar, wchar, syscr, syscw; 1116#endif 1117 struct task_io_accounting ioac; 1118#if defined(CONFIG_TASK_XACCT) 1119 u64 acct_rss_mem1; /* accumulated rss usage */ 1120 u64 acct_vm_mem1; /* accumulated virtual memory usage */ 1121 cputime_t acct_stimexpd;/* stime since last update */ 1122#endif 1123#ifdef CONFIG_NUMA 1124 struct mempolicy *mempolicy; 1125 short il_next; 1126#endif 1127#ifdef CONFIG_CPUSETS 1128 struct cpuset *cpuset; 1129 nodemask_t mems_allowed; 1130 int cpuset_mems_generation; 1131 int cpuset_mem_spread_rotor; 1132#endif 1133#ifdef CONFIG_FUTEX 1134 struct robust_list_head __user *robust_list; 1135#ifdef CONFIG_COMPAT 1136 struct compat_robust_list_head __user *compat_robust_list; 1137#endif 1138 struct list_head pi_state_list; 1139 struct futex_pi_state *pi_state_cache; 1140#endif 1141 atomic_t fs_excl; /* holding fs exclusive resources */ 1142 struct rcu_head rcu; 1143 1144 /* 1145 * cache last used pipe for splice 1146 */ 1147 struct pipe_inode_info *splice_pipe; 1148#ifdef CONFIG_TASK_DELAY_ACCT 1149 struct task_delay_info *delays; 1150#endif 1151#ifdef CONFIG_FAULT_INJECTION 1152 int make_it_fail; 1153#endif 1154 struct prop_local_single dirties; 1155}; 1156 1157/* 1158 * Priority of a process goes from 0..MAX_PRIO-1, valid RT 1159 * priority is 0..MAX_RT_PRIO-1, and SCHED_NORMAL/SCHED_BATCH 1160 * tasks are in the range MAX_RT_PRIO..MAX_PRIO-1. Priority 1161 * values are inverted: lower p->prio value means higher priority. 1162 * 1163 * The MAX_USER_RT_PRIO value allows the actual maximum 1164 * RT priority to be separate from the value exported to 1165 * user-space. This allows kernel threads to set their 1166 * priority to a value higher than any user task. Note: 1167 * MAX_RT_PRIO must not be smaller than MAX_USER_RT_PRIO. 1168 */ 1169 1170#define MAX_USER_RT_PRIO 100 1171#define MAX_RT_PRIO MAX_USER_RT_PRIO 1172 1173#define MAX_PRIO (MAX_RT_PRIO + 40) 1174#define DEFAULT_PRIO (MAX_RT_PRIO + 20) 1175 1176static inline int rt_prio(int prio) 1177{ 1178 if (unlikely(prio < MAX_RT_PRIO)) 1179 return 1; 1180 return 0; 1181} 1182 1183static inline int rt_task(struct task_struct *p) 1184{ 1185 return rt_prio(p->prio); 1186} 1187 1188static inline pid_t process_group(struct task_struct *tsk) 1189{ 1190 return tsk->signal->pgrp; 1191} 1192 1193static inline pid_t signal_session(struct signal_struct *sig) 1194{ 1195 return sig->__session; 1196} 1197 1198static inline pid_t process_session(struct task_struct *tsk) 1199{ 1200 return signal_session(tsk->signal); 1201} 1202 1203static inline void set_signal_session(struct signal_struct *sig, pid_t session) 1204{ 1205 sig->__session = session; 1206} 1207 1208static inline struct pid *task_pid(struct task_struct *task) 1209{ 1210 return task->pids[PIDTYPE_PID].pid; 1211} 1212 1213static inline struct pid *task_tgid(struct task_struct *task) 1214{ 1215 return task->group_leader->pids[PIDTYPE_PID].pid; 1216} 1217 1218static inline struct pid *task_pgrp(struct task_struct *task) 1219{ 1220 return task->group_leader->pids[PIDTYPE_PGID].pid; 1221} 1222 1223static inline struct pid *task_session(struct task_struct *task) 1224{ 1225 return task->group_leader->pids[PIDTYPE_SID].pid; 1226} 1227 1228/** 1229 * pid_alive - check that a task structure is not stale 1230 * @p: Task structure to be checked. 1231 * 1232 * Test if a process is not yet dead (at most zombie state) 1233 * If pid_alive fails, then pointers within the task structure 1234 * can be stale and must not be dereferenced. 1235 */ 1236static inline int pid_alive(struct task_struct *p) 1237{ 1238 return p->pids[PIDTYPE_PID].pid != NULL; 1239} 1240 1241/** 1242 * is_init - check if a task structure is init 1243 * @tsk: Task structure to be checked. 1244 * 1245 * Check if a task structure is the first user space task the kernel created. 1246 */ 1247static inline int is_init(struct task_struct *tsk) 1248{ 1249 return tsk->pid == 1; 1250} 1251 1252extern struct pid *cad_pid; 1253 1254extern void free_task(struct task_struct *tsk); 1255#define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0) 1256 1257extern void __put_task_struct(struct task_struct *t); 1258 1259static inline void put_task_struct(struct task_struct *t) 1260{ 1261 if (atomic_dec_and_test(&t->usage)) 1262 __put_task_struct(t); 1263} 1264 1265/* 1266 * Per process flags 1267 */ 1268#define PF_ALIGNWARN 0x00000001 /* Print alignment warning msgs */ 1269 /* Not implemented yet, only for 486*/ 1270#define PF_STARTING 0x00000002 /* being created */ 1271#define PF_EXITING 0x00000004 /* getting shut down */ 1272#define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */ 1273#define PF_VCPU 0x00000010 /* I'm a virtual CPU */ 1274#define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */ 1275#define PF_SUPERPRIV 0x00000100 /* used super-user privileges */ 1276#define PF_DUMPCORE 0x00000200 /* dumped core */ 1277#define PF_SIGNALED 0x00000400 /* killed by a signal */ 1278#define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1279#define PF_FLUSHER 0x00001000 /* responsible for disk writeback */ 1280#define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */ 1281#define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */ 1282#define PF_FROZEN 0x00010000 /* frozen for system suspend */ 1283#define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */ 1284#define PF_KSWAPD 0x00040000 /* I am kswapd */ 1285#define PF_SWAPOFF 0x00080000 /* I am in swapoff */ 1286#define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ 1287#define PF_BORROWED_MM 0x00200000 /* I am a kthread doing use_mm */ 1288#define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */ 1289#define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1290#define PF_SPREAD_PAGE 0x01000000 /* Spread page cache over cpuset */ 1291#define PF_SPREAD_SLAB 0x02000000 /* Spread some slab caches over cpuset */ 1292#define PF_MEMPOLICY 0x10000000 /* Non-default NUMA mempolicy */ 1293#define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */ 1294#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezeable */ 1295 1296/* 1297 * Only the _current_ task can read/write to tsk->flags, but other 1298 * tasks can access tsk->flags in readonly mode for example 1299 * with tsk_used_math (like during threaded core dumping). 1300 * There is however an exception to this rule during ptrace 1301 * or during fork: the ptracer task is allowed to write to the 1302 * child->flags of its traced child (same goes for fork, the parent 1303 * can write to the child->flags), because we're guaranteed the 1304 * child is not running and in turn not changing child->flags 1305 * at the same time the parent does it. 1306 */ 1307#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1308#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1309#define clear_used_math() clear_stopped_child_used_math(current) 1310#define set_used_math() set_stopped_child_used_math(current) 1311#define conditional_stopped_child_used_math(condition, child) \ 1312 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1313#define conditional_used_math(condition) \ 1314 conditional_stopped_child_used_math(condition, current) 1315#define copy_to_stopped_child_used_math(child) \ 1316 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1317/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1318#define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1319#define used_math() tsk_used_math(current) 1320 1321#ifdef CONFIG_SMP 1322extern int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask); 1323#else 1324static inline int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask) 1325{ 1326 if (!cpu_isset(0, new_mask)) 1327 return -EINVAL; 1328 return 0; 1329} 1330#endif 1331 1332extern unsigned long long sched_clock(void); 1333 1334/* 1335 * For kernel-internal use: high-speed (but slightly incorrect) per-cpu 1336 * clock constructed from sched_clock(): 1337 */ 1338extern unsigned long long cpu_clock(int cpu); 1339 1340extern unsigned long long 1341task_sched_runtime(struct task_struct *task); 1342 1343/* sched_exec is called by processes performing an exec */ 1344#ifdef CONFIG_SMP 1345extern void sched_exec(void); 1346#else 1347#define sched_exec() {} 1348#endif 1349 1350extern void sched_clock_idle_sleep_event(void); 1351extern void sched_clock_idle_wakeup_event(u64 delta_ns); 1352 1353#ifdef CONFIG_HOTPLUG_CPU 1354extern void idle_task_exit(void); 1355#else 1356static inline void idle_task_exit(void) {} 1357#endif 1358 1359extern void sched_idle_next(void); 1360 1361#ifdef CONFIG_SCHED_DEBUG 1362extern unsigned int sysctl_sched_latency; 1363extern unsigned int sysctl_sched_nr_latency; 1364extern unsigned int sysctl_sched_wakeup_granularity; 1365extern unsigned int sysctl_sched_batch_wakeup_granularity; 1366extern unsigned int sysctl_sched_child_runs_first; 1367extern unsigned int sysctl_sched_features; 1368extern unsigned int sysctl_sched_migration_cost; 1369#endif 1370 1371extern unsigned int sysctl_sched_compat_yield; 1372 1373#ifdef CONFIG_RT_MUTEXES 1374extern int rt_mutex_getprio(struct task_struct *p); 1375extern void rt_mutex_setprio(struct task_struct *p, int prio); 1376extern void rt_mutex_adjust_pi(struct task_struct *p); 1377#else 1378static inline int rt_mutex_getprio(struct task_struct *p) 1379{ 1380 return p->normal_prio; 1381} 1382# define rt_mutex_adjust_pi(p) do { } while (0) 1383#endif 1384 1385extern void set_user_nice(struct task_struct *p, long nice); 1386extern int task_prio(const struct task_struct *p); 1387extern int task_nice(const struct task_struct *p); 1388extern int can_nice(const struct task_struct *p, const int nice); 1389extern int task_curr(const struct task_struct *p); 1390extern int idle_cpu(int cpu); 1391extern int sched_setscheduler(struct task_struct *, int, struct sched_param *); 1392extern struct task_struct *idle_task(int cpu); 1393extern struct task_struct *curr_task(int cpu); 1394extern void set_curr_task(int cpu, struct task_struct *p); 1395 1396void yield(void); 1397 1398/* 1399 * The default (Linux) execution domain. 1400 */ 1401extern struct exec_domain default_exec_domain; 1402 1403union thread_union { 1404 struct thread_info thread_info; 1405 unsigned long stack[THREAD_SIZE/sizeof(long)]; 1406}; 1407 1408#ifndef __HAVE_ARCH_KSTACK_END 1409static inline int kstack_end(void *addr) 1410{ 1411 /* Reliable end of stack detection: 1412 * Some APM bios versions misalign the stack 1413 */ 1414 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*))); 1415} 1416#endif 1417 1418extern union thread_union init_thread_union; 1419extern struct task_struct init_task; 1420 1421extern struct mm_struct init_mm; 1422 1423#define find_task_by_pid(nr) find_task_by_pid_type(PIDTYPE_PID, nr) 1424extern struct task_struct *find_task_by_pid_type(int type, int pid); 1425extern void __set_special_pids(pid_t session, pid_t pgrp); 1426 1427/* per-UID process charging. */ 1428extern struct user_struct * alloc_uid(struct user_namespace *, uid_t); 1429static inline struct user_struct *get_uid(struct user_struct *u) 1430{ 1431 atomic_inc(&u->__count); 1432 return u; 1433} 1434extern void free_uid(struct user_struct *); 1435extern void switch_uid(struct user_struct *); 1436extern void release_uids(struct user_namespace *ns); 1437 1438#include <asm/current.h> 1439 1440extern void do_timer(unsigned long ticks); 1441 1442extern int FASTCALL(wake_up_state(struct task_struct * tsk, unsigned int state)); 1443extern int FASTCALL(wake_up_process(struct task_struct * tsk)); 1444extern void FASTCALL(wake_up_new_task(struct task_struct * tsk, 1445 unsigned long clone_flags)); 1446#ifdef CONFIG_SMP 1447 extern void kick_process(struct task_struct *tsk); 1448#else 1449 static inline void kick_process(struct task_struct *tsk) { } 1450#endif 1451extern void sched_fork(struct task_struct *p, int clone_flags); 1452extern void sched_dead(struct task_struct *p); 1453 1454extern int in_group_p(gid_t); 1455extern int in_egroup_p(gid_t); 1456 1457extern void proc_caches_init(void); 1458extern void flush_signals(struct task_struct *); 1459extern void ignore_signals(struct task_struct *); 1460extern void flush_signal_handlers(struct task_struct *, int force_default); 1461extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info); 1462 1463static inline int dequeue_signal_lock(struct task_struct *tsk, sigset_t *mask, siginfo_t *info) 1464{ 1465 unsigned long flags; 1466 int ret; 1467 1468 spin_lock_irqsave(&tsk->sighand->siglock, flags); 1469 ret = dequeue_signal(tsk, mask, info); 1470 spin_unlock_irqrestore(&tsk->sighand->siglock, flags); 1471 1472 return ret; 1473} 1474 1475extern void block_all_signals(int (*notifier)(void *priv), void *priv, 1476 sigset_t *mask); 1477extern void unblock_all_signals(void); 1478extern void release_task(struct task_struct * p); 1479extern int send_sig_info(int, struct siginfo *, struct task_struct *); 1480extern int send_group_sig_info(int, struct siginfo *, struct task_struct *); 1481extern int force_sigsegv(int, struct task_struct *); 1482extern int force_sig_info(int, struct siginfo *, struct task_struct *); 1483extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp); 1484extern int kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp); 1485extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid); 1486extern int kill_pid_info_as_uid(int, struct siginfo *, struct pid *, uid_t, uid_t, u32); 1487extern int kill_pgrp(struct pid *pid, int sig, int priv); 1488extern int kill_pid(struct pid *pid, int sig, int priv); 1489extern int kill_proc_info(int, struct siginfo *, pid_t); 1490extern void do_notify_parent(struct task_struct *, int); 1491extern void force_sig(int, struct task_struct *); 1492extern void force_sig_specific(int, struct task_struct *); 1493extern int send_sig(int, struct task_struct *, int); 1494extern void zap_other_threads(struct task_struct *p); 1495extern int kill_proc(pid_t, int, int); 1496extern struct sigqueue *sigqueue_alloc(void); 1497extern void sigqueue_free(struct sigqueue *); 1498extern int send_sigqueue(int, struct sigqueue *, struct task_struct *); 1499extern int send_group_sigqueue(int, struct sigqueue *, struct task_struct *); 1500extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *); 1501extern int do_sigaltstack(const stack_t __user *, stack_t __user *, unsigned long); 1502 1503static inline int kill_cad_pid(int sig, int priv) 1504{ 1505 return kill_pid(cad_pid, sig, priv); 1506} 1507 1508/* These can be the second arg to send_sig_info/send_group_sig_info. */ 1509#define SEND_SIG_NOINFO ((struct siginfo *) 0) 1510#define SEND_SIG_PRIV ((struct siginfo *) 1) 1511#define SEND_SIG_FORCED ((struct siginfo *) 2) 1512 1513static inline int is_si_special(const struct siginfo *info) 1514{ 1515 return info <= SEND_SIG_FORCED; 1516} 1517 1518/* True if we are on the alternate signal stack. */ 1519 1520static inline int on_sig_stack(unsigned long sp) 1521{ 1522 return (sp - current->sas_ss_sp < current->sas_ss_size); 1523} 1524 1525static inline int sas_ss_flags(unsigned long sp) 1526{ 1527 return (current->sas_ss_size == 0 ? SS_DISABLE 1528 : on_sig_stack(sp) ? SS_ONSTACK : 0); 1529} 1530 1531/* 1532 * Routines for handling mm_structs 1533 */ 1534extern struct mm_struct * mm_alloc(void); 1535 1536/* mmdrop drops the mm and the page tables */ 1537extern void FASTCALL(__mmdrop(struct mm_struct *)); 1538static inline void mmdrop(struct mm_struct * mm) 1539{ 1540 if (unlikely(atomic_dec_and_test(&mm->mm_count))) 1541 __mmdrop(mm); 1542} 1543 1544/* mmput gets rid of the mappings and all user-space */ 1545extern void mmput(struct mm_struct *); 1546/* Grab a reference to a task's mm, if it is not already going away */ 1547extern struct mm_struct *get_task_mm(struct task_struct *task); 1548/* Remove the current tasks stale references to the old mm_struct */ 1549extern void mm_release(struct task_struct *, struct mm_struct *); 1550 1551extern int copy_thread(int, unsigned long, unsigned long, unsigned long, struct task_struct *, struct pt_regs *); 1552extern void flush_thread(void); 1553extern void exit_thread(void); 1554 1555extern void exit_files(struct task_struct *); 1556extern void __cleanup_signal(struct signal_struct *); 1557extern void __cleanup_sighand(struct sighand_struct *); 1558extern void exit_itimers(struct signal_struct *); 1559 1560extern NORET_TYPE void do_group_exit(int); 1561 1562extern void daemonize(const char *, ...); 1563extern int allow_signal(int); 1564extern int disallow_signal(int); 1565 1566extern int do_execve(char *, char __user * __user *, char __user * __user *, struct pt_regs *); 1567extern long do_fork(unsigned long, unsigned long, struct pt_regs *, unsigned long, int __user *, int __user *); 1568struct task_struct *fork_idle(int); 1569 1570extern void set_task_comm(struct task_struct *tsk, char *from); 1571extern void get_task_comm(char *to, struct task_struct *tsk); 1572 1573#ifdef CONFIG_SMP 1574extern void wait_task_inactive(struct task_struct * p); 1575#else 1576#define wait_task_inactive(p) do { } while (0) 1577#endif 1578 1579#define remove_parent(p) list_del_init(&(p)->sibling) 1580#define add_parent(p) list_add_tail(&(p)->sibling,&(p)->parent->children) 1581 1582#define next_task(p) list_entry(rcu_dereference((p)->tasks.next), struct task_struct, tasks) 1583 1584#define for_each_process(p) \ 1585 for (p = &init_task ; (p = next_task(p)) != &init_task ; ) 1586 1587/* 1588 * Careful: do_each_thread/while_each_thread is a double loop so 1589 * 'break' will not work as expected - use goto instead. 1590 */ 1591#define do_each_thread(g, t) \ 1592 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do 1593 1594#define while_each_thread(g, t) \ 1595 while ((t = next_thread(t)) != g) 1596 1597/* de_thread depends on thread_group_leader not being a pid based check */ 1598#define thread_group_leader(p) (p == p->group_leader) 1599 1600/* Do to the insanities of de_thread it is possible for a process 1601 * to have the pid of the thread group leader without actually being 1602 * the thread group leader. For iteration through the pids in proc 1603 * all we care about is that we have a task with the appropriate 1604 * pid, we don't actually care if we have the right task. 1605 */ 1606static inline int has_group_leader_pid(struct task_struct *p) 1607{ 1608 return p->pid == p->tgid; 1609} 1610 1611static inline struct task_struct *next_thread(const struct task_struct *p) 1612{ 1613 return list_entry(rcu_dereference(p->thread_group.next), 1614 struct task_struct, thread_group); 1615} 1616 1617static inline int thread_group_empty(struct task_struct *p) 1618{ 1619 return list_empty(&p->thread_group); 1620} 1621 1622#define delay_group_leader(p) \ 1623 (thread_group_leader(p) && !thread_group_empty(p)) 1624 1625/* 1626 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring 1627 * subscriptions and synchronises with wait4(). Also used in procfs. Also 1628 * pins the final release of task.io_context. Also protects ->cpuset. 1629 * 1630 * Nests both inside and outside of read_lock(&tasklist_lock). 1631 * It must not be nested with write_lock_irq(&tasklist_lock), 1632 * neither inside nor outside. 1633 */ 1634static inline void task_lock(struct task_struct *p) 1635{ 1636 spin_lock(&p->alloc_lock); 1637} 1638 1639static inline void task_unlock(struct task_struct *p) 1640{ 1641 spin_unlock(&p->alloc_lock); 1642} 1643 1644extern struct sighand_struct *lock_task_sighand(struct task_struct *tsk, 1645 unsigned long *flags); 1646 1647static inline void unlock_task_sighand(struct task_struct *tsk, 1648 unsigned long *flags) 1649{ 1650 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags); 1651} 1652 1653#ifndef __HAVE_THREAD_FUNCTIONS 1654 1655#define task_thread_info(task) ((struct thread_info *)(task)->stack) 1656#define task_stack_page(task) ((task)->stack) 1657 1658static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org) 1659{ 1660 *task_thread_info(p) = *task_thread_info(org); 1661 task_thread_info(p)->task = p; 1662} 1663 1664static inline unsigned long *end_of_stack(struct task_struct *p) 1665{ 1666 return (unsigned long *)(task_thread_info(p) + 1); 1667} 1668 1669#endif 1670 1671/* set thread flags in other task's structures 1672 * - see asm/thread_info.h for TIF_xxxx flags available 1673 */ 1674static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 1675{ 1676 set_ti_thread_flag(task_thread_info(tsk), flag); 1677} 1678 1679static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1680{ 1681 clear_ti_thread_flag(task_thread_info(tsk), flag); 1682} 1683 1684static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 1685{ 1686 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 1687} 1688 1689static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 1690{ 1691 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 1692} 1693 1694static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 1695{ 1696 return test_ti_thread_flag(task_thread_info(tsk), flag); 1697} 1698 1699static inline void set_tsk_need_resched(struct task_struct *tsk) 1700{ 1701 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1702} 1703 1704static inline void clear_tsk_need_resched(struct task_struct *tsk) 1705{ 1706 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 1707} 1708 1709static inline int signal_pending(struct task_struct *p) 1710{ 1711 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING)); 1712} 1713 1714static inline int need_resched(void) 1715{ 1716 return unlikely(test_thread_flag(TIF_NEED_RESCHED)); 1717} 1718 1719/* 1720 * cond_resched() and cond_resched_lock(): latency reduction via 1721 * explicit rescheduling in places that are safe. The return 1722 * value indicates whether a reschedule was done in fact. 1723 * cond_resched_lock() will drop the spinlock before scheduling, 1724 * cond_resched_softirq() will enable bhs before scheduling. 1725 */ 1726extern int cond_resched(void); 1727extern int cond_resched_lock(spinlock_t * lock); 1728extern int cond_resched_softirq(void); 1729 1730/* 1731 * Does a critical section need to be broken due to another 1732 * task waiting?: 1733 */ 1734#if defined(CONFIG_PREEMPT) && defined(CONFIG_SMP) 1735# define need_lockbreak(lock) ((lock)->break_lock) 1736#else 1737# define need_lockbreak(lock) 0 1738#endif 1739 1740/* 1741 * Does a critical section need to be broken due to another 1742 * task waiting or preemption being signalled: 1743 */ 1744static inline int lock_need_resched(spinlock_t *lock) 1745{ 1746 if (need_lockbreak(lock) || need_resched()) 1747 return 1; 1748 return 0; 1749} 1750 1751/* 1752 * Reevaluate whether the task has signals pending delivery. 1753 * Wake the task if so. 1754 * This is required every time the blocked sigset_t changes. 1755 * callers must hold sighand->siglock. 1756 */ 1757extern void recalc_sigpending_and_wake(struct task_struct *t); 1758extern void recalc_sigpending(void); 1759 1760extern void signal_wake_up(struct task_struct *t, int resume_stopped); 1761 1762/* 1763 * Wrappers for p->thread_info->cpu access. No-op on UP. 1764 */ 1765#ifdef CONFIG_SMP 1766 1767static inline unsigned int task_cpu(const struct task_struct *p) 1768{ 1769 return task_thread_info(p)->cpu; 1770} 1771 1772extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 1773 1774#else 1775 1776static inline unsigned int task_cpu(const struct task_struct *p) 1777{ 1778 return 0; 1779} 1780 1781static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 1782{ 1783} 1784 1785#endif /* CONFIG_SMP */ 1786 1787#ifdef HAVE_ARCH_PICK_MMAP_LAYOUT 1788extern void arch_pick_mmap_layout(struct mm_struct *mm); 1789#else 1790static inline void arch_pick_mmap_layout(struct mm_struct *mm) 1791{ 1792 mm->mmap_base = TASK_UNMAPPED_BASE; 1793 mm->get_unmapped_area = arch_get_unmapped_area; 1794 mm->unmap_area = arch_unmap_area; 1795} 1796#endif 1797 1798extern long sched_setaffinity(pid_t pid, cpumask_t new_mask); 1799extern long sched_getaffinity(pid_t pid, cpumask_t *mask); 1800 1801extern int sched_mc_power_savings, sched_smt_power_savings; 1802 1803extern void normalize_rt_tasks(void); 1804 1805#ifdef CONFIG_FAIR_GROUP_SCHED 1806 1807extern struct task_group init_task_group; 1808 1809extern struct task_group *sched_create_group(void); 1810extern void sched_destroy_group(struct task_group *tg); 1811extern void sched_move_task(struct task_struct *tsk); 1812extern int sched_group_set_shares(struct task_group *tg, unsigned long shares); 1813extern unsigned long sched_group_shares(struct task_group *tg); 1814 1815#endif 1816 1817#ifdef CONFIG_TASK_XACCT 1818static inline void add_rchar(struct task_struct *tsk, ssize_t amt) 1819{ 1820 tsk->rchar += amt; 1821} 1822 1823static inline void add_wchar(struct task_struct *tsk, ssize_t amt) 1824{ 1825 tsk->wchar += amt; 1826} 1827 1828static inline void inc_syscr(struct task_struct *tsk) 1829{ 1830 tsk->syscr++; 1831} 1832 1833static inline void inc_syscw(struct task_struct *tsk) 1834{ 1835 tsk->syscw++; 1836} 1837#else 1838static inline void add_rchar(struct task_struct *tsk, ssize_t amt) 1839{ 1840} 1841 1842static inline void add_wchar(struct task_struct *tsk, ssize_t amt) 1843{ 1844} 1845 1846static inline void inc_syscr(struct task_struct *tsk) 1847{ 1848} 1849 1850static inline void inc_syscw(struct task_struct *tsk) 1851{ 1852} 1853#endif 1854 1855#endif /* __KERNEL__ */ 1856 1857#endif 1858