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 /* 0x02000000 was previously the unused CLONE_STOPPED (Start in stopped state) 25 and is now available for re-use. */ 26 #define CLONE_NEWUTS 0x04000000 /* New utsname group? */ 27 #define CLONE_NEWIPC 0x08000000 /* New ipcs */ 28 #define CLONE_NEWUSER 0x10000000 /* New user namespace */ 29 #define CLONE_NEWPID 0x20000000 /* New pid namespace */ 30 #define CLONE_NEWNET 0x40000000 /* New network namespace */ 31 #define CLONE_IO 0x80000000 /* Clone io context */ 32 33 /* 34 * Scheduling policies 35 */ 36 #define SCHED_NORMAL 0 37 #define SCHED_FIFO 1 38 #define SCHED_RR 2 39 #define SCHED_BATCH 3 40 /* SCHED_ISO: reserved but not implemented yet */ 41 #define SCHED_IDLE 5 42 /* Can be ORed in to make sure the process is reverted back to SCHED_NORMAL on fork */ 43 #define SCHED_RESET_ON_FORK 0x40000000 44 45 #ifdef __KERNEL__ 46 47 struct sched_param { 48 int sched_priority; 49 }; 50 51 #include <asm/param.h> /* for HZ */ 52 53 #include <linux/capability.h> 54 #include <linux/threads.h> 55 #include <linux/kernel.h> 56 #include <linux/types.h> 57 #include <linux/timex.h> 58 #include <linux/jiffies.h> 59 #include <linux/rbtree.h> 60 #include <linux/thread_info.h> 61 #include <linux/cpumask.h> 62 #include <linux/errno.h> 63 #include <linux/nodemask.h> 64 #include <linux/mm_types.h> 65 66 #include <asm/system.h> 67 #include <asm/page.h> 68 #include <asm/ptrace.h> 69 #include <asm/cputime.h> 70 71 #include <linux/smp.h> 72 #include <linux/sem.h> 73 #include <linux/signal.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/latencytop.h> 92 #include <linux/cred.h> 93 #include <linux/llist.h> 94 95 #include <asm/processor.h> 96 97 struct exec_domain; 98 struct futex_pi_state; 99 struct robust_list_head; 100 struct bio_list; 101 struct fs_struct; 102 struct perf_event_context; 103 struct blk_plug; 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 */ 121 extern unsigned long avenrun[]; /* Load averages */ 122 extern 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; 135 136 extern unsigned long total_forks; 137 extern int nr_threads; 138 DECLARE_PER_CPU(unsigned long, process_counts); 139 extern int nr_processes(void); 140 extern unsigned long nr_running(void); 141 extern unsigned long nr_uninterruptible(void); 142 extern unsigned long nr_iowait(void); 143 extern unsigned long nr_iowait_cpu(int cpu); 144 extern unsigned long this_cpu_load(void); 145 146 147 extern void calc_global_load(unsigned long ticks); 148 149 extern unsigned long get_parent_ip(unsigned long addr); 150 151 struct seq_file; 152 struct cfs_rq; 153 struct task_group; 154 #ifdef CONFIG_SCHED_DEBUG 155 extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m); 156 extern void proc_sched_set_task(struct task_struct *p); 157 extern void 158 print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq); 159 #else 160 static inline void 161 proc_sched_show_task(struct task_struct *p, struct seq_file *m) 162 { 163 } 164 static inline void proc_sched_set_task(struct task_struct *p) 165 { 166 } 167 static inline void 168 print_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" 198 199 extern 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_dead(task) ((task)->exit_state != 0) 219 #define task_is_stopped_or_traced(task) \ 220 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0) 221 #define task_contributes_to_load(task) \ 222 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \ 223 (task->flags & PF_FROZEN) == 0) 224 225 #define __set_task_state(tsk, state_value) \ 226 do { (tsk)->state = (state_value); } while (0) 227 #define set_task_state(tsk, state_value) \ 228 set_mb((tsk)->state, (state_value)) 229 230 /* 231 * set_current_state() includes a barrier so that the write of current->state 232 * is correctly serialised wrt the caller's subsequent test of whether to 233 * actually sleep: 234 * 235 * set_current_state(TASK_UNINTERRUPTIBLE); 236 * if (do_i_need_to_sleep()) 237 * schedule(); 238 * 239 * If the caller does not need such serialisation then use __set_current_state() 240 */ 241 #define __set_current_state(state_value) \ 242 do { current->state = (state_value); } while (0) 243 #define set_current_state(state_value) \ 244 set_mb(current->state, (state_value)) 245 246 /* Task command name length */ 247 #define TASK_COMM_LEN 16 248 249 #include <linux/spinlock.h> 250 251 /* 252 * This serializes "schedule()" and also protects 253 * the run-queue from deletions/modifications (but 254 * _adding_ to the beginning of the run-queue has 255 * a separate lock). 256 */ 257 extern rwlock_t tasklist_lock; 258 extern spinlock_t mmlist_lock; 259 260 struct task_struct; 261 262 #ifdef CONFIG_PROVE_RCU 263 extern int lockdep_tasklist_lock_is_held(void); 264 #endif /* #ifdef CONFIG_PROVE_RCU */ 265 266 extern void sched_init(void); 267 extern void sched_init_smp(void); 268 extern asmlinkage void schedule_tail(struct task_struct *prev); 269 extern void init_idle(struct task_struct *idle, int cpu); 270 extern void init_idle_bootup_task(struct task_struct *idle); 271 272 extern int runqueue_is_locked(int cpu); 273 274 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ) 275 extern void select_nohz_load_balancer(int stop_tick); 276 extern void set_cpu_sd_state_idle(void); 277 extern int get_nohz_timer_target(void); 278 #else 279 static inline void select_nohz_load_balancer(int stop_tick) { } 280 static inline void set_cpu_sd_state_idle(void) { } 281 #endif 282 283 /* 284 * Only dump TASK_* tasks. (0 for all tasks) 285 */ 286 extern void show_state_filter(unsigned long state_filter); 287 288 static inline void show_state(void) 289 { 290 show_state_filter(0); 291 } 292 293 extern void show_regs(struct pt_regs *); 294 295 /* 296 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current 297 * task), SP is the stack pointer of the first frame that should be shown in the back 298 * trace (or NULL if the entire call-chain of the task should be shown). 299 */ 300 extern void show_stack(struct task_struct *task, unsigned long *sp); 301 302 void io_schedule(void); 303 long io_schedule_timeout(long timeout); 304 305 extern void cpu_init (void); 306 extern void trap_init(void); 307 extern void update_process_times(int user); 308 extern void scheduler_tick(void); 309 310 extern void sched_show_task(struct task_struct *p); 311 312 #ifdef CONFIG_LOCKUP_DETECTOR 313 extern void touch_softlockup_watchdog(void); 314 extern void touch_softlockup_watchdog_sync(void); 315 extern void touch_all_softlockup_watchdogs(void); 316 extern int proc_dowatchdog_thresh(struct ctl_table *table, int write, 317 void __user *buffer, 318 size_t *lenp, loff_t *ppos); 319 extern unsigned int softlockup_panic; 320 void lockup_detector_init(void); 321 #else 322 static inline void touch_softlockup_watchdog(void) 323 { 324 } 325 static inline void touch_softlockup_watchdog_sync(void) 326 { 327 } 328 static inline void touch_all_softlockup_watchdogs(void) 329 { 330 } 331 static inline void lockup_detector_init(void) 332 { 333 } 334 #endif 335 336 #ifdef CONFIG_DETECT_HUNG_TASK 337 extern unsigned int sysctl_hung_task_panic; 338 extern unsigned long sysctl_hung_task_check_count; 339 extern unsigned long sysctl_hung_task_timeout_secs; 340 extern unsigned long sysctl_hung_task_warnings; 341 extern int proc_dohung_task_timeout_secs(struct ctl_table *table, int write, 342 void __user *buffer, 343 size_t *lenp, loff_t *ppos); 344 #else 345 /* Avoid need for ifdefs elsewhere in the code */ 346 enum { sysctl_hung_task_timeout_secs = 0 }; 347 #endif 348 349 /* Attach to any functions which should be ignored in wchan output. */ 350 #define __sched __attribute__((__section__(".sched.text"))) 351 352 /* Linker adds these: start and end of __sched functions */ 353 extern char __sched_text_start[], __sched_text_end[]; 354 355 /* Is this address in the __sched functions? */ 356 extern int in_sched_functions(unsigned long addr); 357 358 #define MAX_SCHEDULE_TIMEOUT LONG_MAX 359 extern signed long schedule_timeout(signed long timeout); 360 extern signed long schedule_timeout_interruptible(signed long timeout); 361 extern signed long schedule_timeout_killable(signed long timeout); 362 extern signed long schedule_timeout_uninterruptible(signed long timeout); 363 asmlinkage void schedule(void); 364 extern int mutex_spin_on_owner(struct mutex *lock, struct task_struct *owner); 365 366 struct nsproxy; 367 struct user_namespace; 368 369 /* 370 * Default maximum number of active map areas, this limits the number of vmas 371 * per mm struct. Users can overwrite this number by sysctl but there is a 372 * problem. 373 * 374 * When a program's coredump is generated as ELF format, a section is created 375 * per a vma. In ELF, the number of sections is represented in unsigned short. 376 * This means the number of sections should be smaller than 65535 at coredump. 377 * Because the kernel adds some informative sections to a image of program at 378 * generating coredump, we need some margin. The number of extra sections is 379 * 1-3 now and depends on arch. We use "5" as safe margin, here. 380 */ 381 #define MAPCOUNT_ELF_CORE_MARGIN (5) 382 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN) 383 384 extern int sysctl_max_map_count; 385 386 #include <linux/aio.h> 387 388 #ifdef CONFIG_MMU 389 extern void arch_pick_mmap_layout(struct mm_struct *mm); 390 extern unsigned long 391 arch_get_unmapped_area(struct file *, unsigned long, unsigned long, 392 unsigned long, unsigned long); 393 extern unsigned long 394 arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr, 395 unsigned long len, unsigned long pgoff, 396 unsigned long flags); 397 extern void arch_unmap_area(struct mm_struct *, unsigned long); 398 extern void arch_unmap_area_topdown(struct mm_struct *, unsigned long); 399 #else 400 static inline void arch_pick_mmap_layout(struct mm_struct *mm) {} 401 #endif 402 403 404 extern void set_dumpable(struct mm_struct *mm, int value); 405 extern int get_dumpable(struct mm_struct *mm); 406 407 /* mm flags */ 408 /* dumpable bits */ 409 #define MMF_DUMPABLE 0 /* core dump is permitted */ 410 #define MMF_DUMP_SECURELY 1 /* core file is readable only by root */ 411 412 #define MMF_DUMPABLE_BITS 2 413 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1) 414 415 /* coredump filter bits */ 416 #define MMF_DUMP_ANON_PRIVATE 2 417 #define MMF_DUMP_ANON_SHARED 3 418 #define MMF_DUMP_MAPPED_PRIVATE 4 419 #define MMF_DUMP_MAPPED_SHARED 5 420 #define MMF_DUMP_ELF_HEADERS 6 421 #define MMF_DUMP_HUGETLB_PRIVATE 7 422 #define MMF_DUMP_HUGETLB_SHARED 8 423 424 #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS 425 #define MMF_DUMP_FILTER_BITS 7 426 #define MMF_DUMP_FILTER_MASK \ 427 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT) 428 #define MMF_DUMP_FILTER_DEFAULT \ 429 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\ 430 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF) 431 432 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS 433 # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS) 434 #else 435 # define MMF_DUMP_MASK_DEFAULT_ELF 0 436 #endif 437 /* leave room for more dump flags */ 438 #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */ 439 #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */ 440 441 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK) 442 443 struct sighand_struct { 444 atomic_t count; 445 struct k_sigaction action[_NSIG]; 446 spinlock_t siglock; 447 wait_queue_head_t signalfd_wqh; 448 }; 449 450 struct pacct_struct { 451 int ac_flag; 452 long ac_exitcode; 453 unsigned long ac_mem; 454 cputime_t ac_utime, ac_stime; 455 unsigned long ac_minflt, ac_majflt; 456 }; 457 458 struct cpu_itimer { 459 cputime_t expires; 460 cputime_t incr; 461 u32 error; 462 u32 incr_error; 463 }; 464 465 /** 466 * struct task_cputime - collected CPU time counts 467 * @utime: time spent in user mode, in &cputime_t units 468 * @stime: time spent in kernel mode, in &cputime_t units 469 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds 470 * 471 * This structure groups together three kinds of CPU time that are 472 * tracked for threads and thread groups. Most things considering 473 * CPU time want to group these counts together and treat all three 474 * of them in parallel. 475 */ 476 struct task_cputime { 477 cputime_t utime; 478 cputime_t stime; 479 unsigned long long sum_exec_runtime; 480 }; 481 /* Alternate field names when used to cache expirations. */ 482 #define prof_exp stime 483 #define virt_exp utime 484 #define sched_exp sum_exec_runtime 485 486 #define INIT_CPUTIME \ 487 (struct task_cputime) { \ 488 .utime = 0, \ 489 .stime = 0, \ 490 .sum_exec_runtime = 0, \ 491 } 492 493 /* 494 * Disable preemption until the scheduler is running. 495 * Reset by start_kernel()->sched_init()->init_idle(). 496 * 497 * We include PREEMPT_ACTIVE to avoid cond_resched() from working 498 * before the scheduler is active -- see should_resched(). 499 */ 500 #define INIT_PREEMPT_COUNT (1 + PREEMPT_ACTIVE) 501 502 /** 503 * struct thread_group_cputimer - thread group interval timer counts 504 * @cputime: thread group interval timers. 505 * @running: non-zero when there are timers running and 506 * @cputime receives updates. 507 * @lock: lock for fields in this struct. 508 * 509 * This structure contains the version of task_cputime, above, that is 510 * used for thread group CPU timer calculations. 511 */ 512 struct thread_group_cputimer { 513 struct task_cputime cputime; 514 int running; 515 raw_spinlock_t lock; 516 }; 517 518 #include <linux/rwsem.h> 519 struct autogroup; 520 521 /* 522 * NOTE! "signal_struct" does not have its own 523 * locking, because a shared signal_struct always 524 * implies a shared sighand_struct, so locking 525 * sighand_struct is always a proper superset of 526 * the locking of signal_struct. 527 */ 528 struct signal_struct { 529 atomic_t sigcnt; 530 atomic_t live; 531 int nr_threads; 532 533 wait_queue_head_t wait_chldexit; /* for wait4() */ 534 535 /* current thread group signal load-balancing target: */ 536 struct task_struct *curr_target; 537 538 /* shared signal handling: */ 539 struct sigpending shared_pending; 540 541 /* thread group exit support */ 542 int group_exit_code; 543 /* overloaded: 544 * - notify group_exit_task when ->count is equal to notify_count 545 * - everyone except group_exit_task is stopped during signal delivery 546 * of fatal signals, group_exit_task processes the signal. 547 */ 548 int notify_count; 549 struct task_struct *group_exit_task; 550 551 /* thread group stop support, overloads group_exit_code too */ 552 int group_stop_count; 553 unsigned int flags; /* see SIGNAL_* flags below */ 554 555 /* POSIX.1b Interval Timers */ 556 struct list_head posix_timers; 557 558 /* ITIMER_REAL timer for the process */ 559 struct hrtimer real_timer; 560 struct pid *leader_pid; 561 ktime_t it_real_incr; 562 563 /* 564 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use 565 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these 566 * values are defined to 0 and 1 respectively 567 */ 568 struct cpu_itimer it[2]; 569 570 /* 571 * Thread group totals for process CPU timers. 572 * See thread_group_cputimer(), et al, for details. 573 */ 574 struct thread_group_cputimer cputimer; 575 576 /* Earliest-expiration cache. */ 577 struct task_cputime cputime_expires; 578 579 struct list_head cpu_timers[3]; 580 581 struct pid *tty_old_pgrp; 582 583 /* boolean value for session group leader */ 584 int leader; 585 586 struct tty_struct *tty; /* NULL if no tty */ 587 588 #ifdef CONFIG_SCHED_AUTOGROUP 589 struct autogroup *autogroup; 590 #endif 591 /* 592 * Cumulative resource counters for dead threads in the group, 593 * and for reaped dead child processes forked by this group. 594 * Live threads maintain their own counters and add to these 595 * in __exit_signal, except for the group leader. 596 */ 597 cputime_t utime, stime, cutime, cstime; 598 cputime_t gtime; 599 cputime_t cgtime; 600 #ifndef CONFIG_VIRT_CPU_ACCOUNTING 601 cputime_t prev_utime, prev_stime; 602 #endif 603 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw; 604 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt; 605 unsigned long inblock, oublock, cinblock, coublock; 606 unsigned long maxrss, cmaxrss; 607 struct task_io_accounting ioac; 608 609 /* 610 * Cumulative ns of schedule CPU time fo dead threads in the 611 * group, not including a zombie group leader, (This only differs 612 * from jiffies_to_ns(utime + stime) if sched_clock uses something 613 * other than jiffies.) 614 */ 615 unsigned long long sum_sched_runtime; 616 617 /* 618 * We don't bother to synchronize most readers of this at all, 619 * because there is no reader checking a limit that actually needs 620 * to get both rlim_cur and rlim_max atomically, and either one 621 * alone is a single word that can safely be read normally. 622 * getrlimit/setrlimit use task_lock(current->group_leader) to 623 * protect this instead of the siglock, because they really 624 * have no need to disable irqs. 625 */ 626 struct rlimit rlim[RLIM_NLIMITS]; 627 628 #ifdef CONFIG_BSD_PROCESS_ACCT 629 struct pacct_struct pacct; /* per-process accounting information */ 630 #endif 631 #ifdef CONFIG_TASKSTATS 632 struct taskstats *stats; 633 #endif 634 #ifdef CONFIG_AUDIT 635 unsigned audit_tty; 636 struct tty_audit_buf *tty_audit_buf; 637 #endif 638 #ifdef CONFIG_CGROUPS 639 /* 640 * group_rwsem prevents new tasks from entering the threadgroup and 641 * member tasks from exiting,a more specifically, setting of 642 * PF_EXITING. fork and exit paths are protected with this rwsem 643 * using threadgroup_change_begin/end(). Users which require 644 * threadgroup to remain stable should use threadgroup_[un]lock() 645 * which also takes care of exec path. Currently, cgroup is the 646 * only user. 647 */ 648 struct rw_semaphore group_rwsem; 649 #endif 650 651 int oom_adj; /* OOM kill score adjustment (bit shift) */ 652 int oom_score_adj; /* OOM kill score adjustment */ 653 int oom_score_adj_min; /* OOM kill score adjustment minimum value. 654 * Only settable by CAP_SYS_RESOURCE. */ 655 656 struct mutex cred_guard_mutex; /* guard against foreign influences on 657 * credential calculations 658 * (notably. ptrace) */ 659 }; 660 661 /* Context switch must be unlocked if interrupts are to be enabled */ 662 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW 663 # define __ARCH_WANT_UNLOCKED_CTXSW 664 #endif 665 666 /* 667 * Bits in flags field of signal_struct. 668 */ 669 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */ 670 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */ 671 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */ 672 /* 673 * Pending notifications to parent. 674 */ 675 #define SIGNAL_CLD_STOPPED 0x00000010 676 #define SIGNAL_CLD_CONTINUED 0x00000020 677 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED) 678 679 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */ 680 681 /* If true, all threads except ->group_exit_task have pending SIGKILL */ 682 static inline int signal_group_exit(const struct signal_struct *sig) 683 { 684 return (sig->flags & SIGNAL_GROUP_EXIT) || 685 (sig->group_exit_task != NULL); 686 } 687 688 /* 689 * Some day this will be a full-fledged user tracking system.. 690 */ 691 struct user_struct { 692 atomic_t __count; /* reference count */ 693 atomic_t processes; /* How many processes does this user have? */ 694 atomic_t files; /* How many open files does this user have? */ 695 atomic_t sigpending; /* How many pending signals does this user have? */ 696 #ifdef CONFIG_INOTIFY_USER 697 atomic_t inotify_watches; /* How many inotify watches does this user have? */ 698 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */ 699 #endif 700 #ifdef CONFIG_FANOTIFY 701 atomic_t fanotify_listeners; 702 #endif 703 #ifdef CONFIG_EPOLL 704 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */ 705 #endif 706 #ifdef CONFIG_POSIX_MQUEUE 707 /* protected by mq_lock */ 708 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */ 709 #endif 710 unsigned long locked_shm; /* How many pages of mlocked shm ? */ 711 712 #ifdef CONFIG_KEYS 713 struct key *uid_keyring; /* UID specific keyring */ 714 struct key *session_keyring; /* UID's default session keyring */ 715 #endif 716 717 /* Hash table maintenance information */ 718 struct hlist_node uidhash_node; 719 uid_t uid; 720 struct user_namespace *user_ns; 721 722 #ifdef CONFIG_PERF_EVENTS 723 atomic_long_t locked_vm; 724 #endif 725 }; 726 727 extern int uids_sysfs_init(void); 728 729 extern struct user_struct *find_user(uid_t); 730 731 extern struct user_struct root_user; 732 #define INIT_USER (&root_user) 733 734 735 struct backing_dev_info; 736 struct reclaim_state; 737 738 #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) 739 struct sched_info { 740 /* cumulative counters */ 741 unsigned long pcount; /* # of times run on this cpu */ 742 unsigned long long run_delay; /* time spent waiting on a runqueue */ 743 744 /* timestamps */ 745 unsigned long long last_arrival,/* when we last ran on a cpu */ 746 last_queued; /* when we were last queued to run */ 747 }; 748 #endif /* defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) */ 749 750 #ifdef CONFIG_TASK_DELAY_ACCT 751 struct task_delay_info { 752 spinlock_t lock; 753 unsigned int flags; /* Private per-task flags */ 754 755 /* For each stat XXX, add following, aligned appropriately 756 * 757 * struct timespec XXX_start, XXX_end; 758 * u64 XXX_delay; 759 * u32 XXX_count; 760 * 761 * Atomicity of updates to XXX_delay, XXX_count protected by 762 * single lock above (split into XXX_lock if contention is an issue). 763 */ 764 765 /* 766 * XXX_count is incremented on every XXX operation, the delay 767 * associated with the operation is added to XXX_delay. 768 * XXX_delay contains the accumulated delay time in nanoseconds. 769 */ 770 struct timespec blkio_start, blkio_end; /* Shared by blkio, swapin */ 771 u64 blkio_delay; /* wait for sync block io completion */ 772 u64 swapin_delay; /* wait for swapin block io completion */ 773 u32 blkio_count; /* total count of the number of sync block */ 774 /* io operations performed */ 775 u32 swapin_count; /* total count of the number of swapin block */ 776 /* io operations performed */ 777 778 struct timespec freepages_start, freepages_end; 779 u64 freepages_delay; /* wait for memory reclaim */ 780 u32 freepages_count; /* total count of memory reclaim */ 781 }; 782 #endif /* CONFIG_TASK_DELAY_ACCT */ 783 784 static inline int sched_info_on(void) 785 { 786 #ifdef CONFIG_SCHEDSTATS 787 return 1; 788 #elif defined(CONFIG_TASK_DELAY_ACCT) 789 extern int delayacct_on; 790 return delayacct_on; 791 #else 792 return 0; 793 #endif 794 } 795 796 enum cpu_idle_type { 797 CPU_IDLE, 798 CPU_NOT_IDLE, 799 CPU_NEWLY_IDLE, 800 CPU_MAX_IDLE_TYPES 801 }; 802 803 /* 804 * Increase resolution of nice-level calculations for 64-bit architectures. 805 * The extra resolution improves shares distribution and load balancing of 806 * low-weight task groups (eg. nice +19 on an autogroup), deeper taskgroup 807 * hierarchies, especially on larger systems. This is not a user-visible change 808 * and does not change the user-interface for setting shares/weights. 809 * 810 * We increase resolution only if we have enough bits to allow this increased 811 * resolution (i.e. BITS_PER_LONG > 32). The costs for increasing resolution 812 * when BITS_PER_LONG <= 32 are pretty high and the returns do not justify the 813 * increased costs. 814 */ 815 #if 0 /* BITS_PER_LONG > 32 -- currently broken: it increases power usage under light load */ 816 # define SCHED_LOAD_RESOLUTION 10 817 # define scale_load(w) ((w) << SCHED_LOAD_RESOLUTION) 818 # define scale_load_down(w) ((w) >> SCHED_LOAD_RESOLUTION) 819 #else 820 # define SCHED_LOAD_RESOLUTION 0 821 # define scale_load(w) (w) 822 # define scale_load_down(w) (w) 823 #endif 824 825 #define SCHED_LOAD_SHIFT (10 + SCHED_LOAD_RESOLUTION) 826 #define SCHED_LOAD_SCALE (1L << SCHED_LOAD_SHIFT) 827 828 /* 829 * Increase resolution of cpu_power calculations 830 */ 831 #define SCHED_POWER_SHIFT 10 832 #define SCHED_POWER_SCALE (1L << SCHED_POWER_SHIFT) 833 834 /* 835 * sched-domains (multiprocessor balancing) declarations: 836 */ 837 #ifdef CONFIG_SMP 838 #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */ 839 #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */ 840 #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */ 841 #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */ 842 #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */ 843 #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */ 844 #define SD_PREFER_LOCAL 0x0040 /* Prefer to keep tasks local to this domain */ 845 #define SD_SHARE_CPUPOWER 0x0080 /* Domain members share cpu power */ 846 #define SD_POWERSAVINGS_BALANCE 0x0100 /* Balance for power savings */ 847 #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */ 848 #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */ 849 #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */ 850 #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */ 851 #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */ 852 853 enum powersavings_balance_level { 854 POWERSAVINGS_BALANCE_NONE = 0, /* No power saving load balance */ 855 POWERSAVINGS_BALANCE_BASIC, /* Fill one thread/core/package 856 * first for long running threads 857 */ 858 POWERSAVINGS_BALANCE_WAKEUP, /* Also bias task wakeups to semi-idle 859 * cpu package for power savings 860 */ 861 MAX_POWERSAVINGS_BALANCE_LEVELS 862 }; 863 864 extern int sched_mc_power_savings, sched_smt_power_savings; 865 866 static inline int sd_balance_for_mc_power(void) 867 { 868 if (sched_smt_power_savings) 869 return SD_POWERSAVINGS_BALANCE; 870 871 if (!sched_mc_power_savings) 872 return SD_PREFER_SIBLING; 873 874 return 0; 875 } 876 877 static inline int sd_balance_for_package_power(void) 878 { 879 if (sched_mc_power_savings | sched_smt_power_savings) 880 return SD_POWERSAVINGS_BALANCE; 881 882 return SD_PREFER_SIBLING; 883 } 884 885 extern int __weak arch_sd_sibiling_asym_packing(void); 886 887 /* 888 * Optimise SD flags for power savings: 889 * SD_BALANCE_NEWIDLE helps aggressive task consolidation and power savings. 890 * Keep default SD flags if sched_{smt,mc}_power_saving=0 891 */ 892 893 static inline int sd_power_saving_flags(void) 894 { 895 if (sched_mc_power_savings | sched_smt_power_savings) 896 return SD_BALANCE_NEWIDLE; 897 898 return 0; 899 } 900 901 struct sched_group_power { 902 atomic_t ref; 903 /* 904 * CPU power of this group, SCHED_LOAD_SCALE being max power for a 905 * single CPU. 906 */ 907 unsigned int power, power_orig; 908 /* 909 * Number of busy cpus in this group. 910 */ 911 atomic_t nr_busy_cpus; 912 }; 913 914 struct sched_group { 915 struct sched_group *next; /* Must be a circular list */ 916 atomic_t ref; 917 918 unsigned int group_weight; 919 struct sched_group_power *sgp; 920 921 /* 922 * The CPUs this group covers. 923 * 924 * NOTE: this field is variable length. (Allocated dynamically 925 * by attaching extra space to the end of the structure, 926 * depending on how many CPUs the kernel has booted up with) 927 */ 928 unsigned long cpumask[0]; 929 }; 930 931 static inline struct cpumask *sched_group_cpus(struct sched_group *sg) 932 { 933 return to_cpumask(sg->cpumask); 934 } 935 936 /** 937 * group_first_cpu - Returns the first cpu in the cpumask of a sched_group. 938 * @group: The group whose first cpu is to be returned. 939 */ 940 static inline unsigned int group_first_cpu(struct sched_group *group) 941 { 942 return cpumask_first(sched_group_cpus(group)); 943 } 944 945 struct sched_domain_attr { 946 int relax_domain_level; 947 }; 948 949 #define SD_ATTR_INIT (struct sched_domain_attr) { \ 950 .relax_domain_level = -1, \ 951 } 952 953 extern int sched_domain_level_max; 954 955 struct sched_domain { 956 /* These fields must be setup */ 957 struct sched_domain *parent; /* top domain must be null terminated */ 958 struct sched_domain *child; /* bottom domain must be null terminated */ 959 struct sched_group *groups; /* the balancing groups of the domain */ 960 unsigned long min_interval; /* Minimum balance interval ms */ 961 unsigned long max_interval; /* Maximum balance interval ms */ 962 unsigned int busy_factor; /* less balancing by factor if busy */ 963 unsigned int imbalance_pct; /* No balance until over watermark */ 964 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */ 965 unsigned int busy_idx; 966 unsigned int idle_idx; 967 unsigned int newidle_idx; 968 unsigned int wake_idx; 969 unsigned int forkexec_idx; 970 unsigned int smt_gain; 971 int flags; /* See SD_* */ 972 int level; 973 974 /* Runtime fields. */ 975 unsigned long last_balance; /* init to jiffies. units in jiffies */ 976 unsigned int balance_interval; /* initialise to 1. units in ms. */ 977 unsigned int nr_balance_failed; /* initialise to 0 */ 978 979 u64 last_update; 980 981 #ifdef CONFIG_SCHEDSTATS 982 /* load_balance() stats */ 983 unsigned int lb_count[CPU_MAX_IDLE_TYPES]; 984 unsigned int lb_failed[CPU_MAX_IDLE_TYPES]; 985 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES]; 986 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES]; 987 unsigned int lb_gained[CPU_MAX_IDLE_TYPES]; 988 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES]; 989 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES]; 990 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES]; 991 992 /* Active load balancing */ 993 unsigned int alb_count; 994 unsigned int alb_failed; 995 unsigned int alb_pushed; 996 997 /* SD_BALANCE_EXEC stats */ 998 unsigned int sbe_count; 999 unsigned int sbe_balanced; 1000 unsigned int sbe_pushed; 1001 1002 /* SD_BALANCE_FORK stats */ 1003 unsigned int sbf_count; 1004 unsigned int sbf_balanced; 1005 unsigned int sbf_pushed; 1006 1007 /* try_to_wake_up() stats */ 1008 unsigned int ttwu_wake_remote; 1009 unsigned int ttwu_move_affine; 1010 unsigned int ttwu_move_balance; 1011 #endif 1012 #ifdef CONFIG_SCHED_DEBUG 1013 char *name; 1014 #endif 1015 union { 1016 void *private; /* used during construction */ 1017 struct rcu_head rcu; /* used during destruction */ 1018 }; 1019 1020 unsigned int span_weight; 1021 /* 1022 * Span of all CPUs in this domain. 1023 * 1024 * NOTE: this field is variable length. (Allocated dynamically 1025 * by attaching extra space to the end of the structure, 1026 * depending on how many CPUs the kernel has booted up with) 1027 */ 1028 unsigned long span[0]; 1029 }; 1030 1031 static inline struct cpumask *sched_domain_span(struct sched_domain *sd) 1032 { 1033 return to_cpumask(sd->span); 1034 } 1035 1036 extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[], 1037 struct sched_domain_attr *dattr_new); 1038 1039 /* Allocate an array of sched domains, for partition_sched_domains(). */ 1040 cpumask_var_t *alloc_sched_domains(unsigned int ndoms); 1041 void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms); 1042 1043 /* Test a flag in parent sched domain */ 1044 static inline int test_sd_parent(struct sched_domain *sd, int flag) 1045 { 1046 if (sd->parent && (sd->parent->flags & flag)) 1047 return 1; 1048 1049 return 0; 1050 } 1051 1052 unsigned long default_scale_freq_power(struct sched_domain *sd, int cpu); 1053 unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu); 1054 1055 #else /* CONFIG_SMP */ 1056 1057 struct sched_domain_attr; 1058 1059 static inline void 1060 partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[], 1061 struct sched_domain_attr *dattr_new) 1062 { 1063 } 1064 #endif /* !CONFIG_SMP */ 1065 1066 1067 struct io_context; /* See blkdev.h */ 1068 1069 1070 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK 1071 extern void prefetch_stack(struct task_struct *t); 1072 #else 1073 static inline void prefetch_stack(struct task_struct *t) { } 1074 #endif 1075 1076 struct audit_context; /* See audit.c */ 1077 struct mempolicy; 1078 struct pipe_inode_info; 1079 struct uts_namespace; 1080 1081 struct rq; 1082 struct sched_domain; 1083 1084 /* 1085 * wake flags 1086 */ 1087 #define WF_SYNC 0x01 /* waker goes to sleep after wakup */ 1088 #define WF_FORK 0x02 /* child wakeup after fork */ 1089 #define WF_MIGRATED 0x04 /* internal use, task got migrated */ 1090 1091 #define ENQUEUE_WAKEUP 1 1092 #define ENQUEUE_HEAD 2 1093 #ifdef CONFIG_SMP 1094 #define ENQUEUE_WAKING 4 /* sched_class::task_waking was called */ 1095 #else 1096 #define ENQUEUE_WAKING 0 1097 #endif 1098 1099 #define DEQUEUE_SLEEP 1 1100 1101 struct sched_class { 1102 const struct sched_class *next; 1103 1104 void (*enqueue_task) (struct rq *rq, struct task_struct *p, int flags); 1105 void (*dequeue_task) (struct rq *rq, struct task_struct *p, int flags); 1106 void (*yield_task) (struct rq *rq); 1107 bool (*yield_to_task) (struct rq *rq, struct task_struct *p, bool preempt); 1108 1109 void (*check_preempt_curr) (struct rq *rq, struct task_struct *p, int flags); 1110 1111 struct task_struct * (*pick_next_task) (struct rq *rq); 1112 void (*put_prev_task) (struct rq *rq, struct task_struct *p); 1113 1114 #ifdef CONFIG_SMP 1115 int (*select_task_rq)(struct task_struct *p, int sd_flag, int flags); 1116 1117 void (*pre_schedule) (struct rq *this_rq, struct task_struct *task); 1118 void (*post_schedule) (struct rq *this_rq); 1119 void (*task_waking) (struct task_struct *task); 1120 void (*task_woken) (struct rq *this_rq, struct task_struct *task); 1121 1122 void (*set_cpus_allowed)(struct task_struct *p, 1123 const struct cpumask *newmask); 1124 1125 void (*rq_online)(struct rq *rq); 1126 void (*rq_offline)(struct rq *rq); 1127 #endif 1128 1129 void (*set_curr_task) (struct rq *rq); 1130 void (*task_tick) (struct rq *rq, struct task_struct *p, int queued); 1131 void (*task_fork) (struct task_struct *p); 1132 1133 void (*switched_from) (struct rq *this_rq, struct task_struct *task); 1134 void (*switched_to) (struct rq *this_rq, struct task_struct *task); 1135 void (*prio_changed) (struct rq *this_rq, struct task_struct *task, 1136 int oldprio); 1137 1138 unsigned int (*get_rr_interval) (struct rq *rq, 1139 struct task_struct *task); 1140 1141 #ifdef CONFIG_FAIR_GROUP_SCHED 1142 void (*task_move_group) (struct task_struct *p, int on_rq); 1143 #endif 1144 }; 1145 1146 struct load_weight { 1147 unsigned long weight, inv_weight; 1148 }; 1149 1150 #ifdef CONFIG_SCHEDSTATS 1151 struct sched_statistics { 1152 u64 wait_start; 1153 u64 wait_max; 1154 u64 wait_count; 1155 u64 wait_sum; 1156 u64 iowait_count; 1157 u64 iowait_sum; 1158 1159 u64 sleep_start; 1160 u64 sleep_max; 1161 s64 sum_sleep_runtime; 1162 1163 u64 block_start; 1164 u64 block_max; 1165 u64 exec_max; 1166 u64 slice_max; 1167 1168 u64 nr_migrations_cold; 1169 u64 nr_failed_migrations_affine; 1170 u64 nr_failed_migrations_running; 1171 u64 nr_failed_migrations_hot; 1172 u64 nr_forced_migrations; 1173 1174 u64 nr_wakeups; 1175 u64 nr_wakeups_sync; 1176 u64 nr_wakeups_migrate; 1177 u64 nr_wakeups_local; 1178 u64 nr_wakeups_remote; 1179 u64 nr_wakeups_affine; 1180 u64 nr_wakeups_affine_attempts; 1181 u64 nr_wakeups_passive; 1182 u64 nr_wakeups_idle; 1183 }; 1184 #endif 1185 1186 struct sched_entity { 1187 struct load_weight load; /* for load-balancing */ 1188 struct rb_node run_node; 1189 struct list_head group_node; 1190 unsigned int on_rq; 1191 1192 u64 exec_start; 1193 u64 sum_exec_runtime; 1194 u64 vruntime; 1195 u64 prev_sum_exec_runtime; 1196 1197 u64 nr_migrations; 1198 1199 #ifdef CONFIG_SCHEDSTATS 1200 struct sched_statistics statistics; 1201 #endif 1202 1203 #ifdef CONFIG_FAIR_GROUP_SCHED 1204 struct sched_entity *parent; 1205 /* rq on which this entity is (to be) queued: */ 1206 struct cfs_rq *cfs_rq; 1207 /* rq "owned" by this entity/group: */ 1208 struct cfs_rq *my_q; 1209 #endif 1210 }; 1211 1212 struct sched_rt_entity { 1213 struct list_head run_list; 1214 unsigned long timeout; 1215 unsigned int time_slice; 1216 int nr_cpus_allowed; 1217 1218 struct sched_rt_entity *back; 1219 #ifdef CONFIG_RT_GROUP_SCHED 1220 struct sched_rt_entity *parent; 1221 /* rq on which this entity is (to be) queued: */ 1222 struct rt_rq *rt_rq; 1223 /* rq "owned" by this entity/group: */ 1224 struct rt_rq *my_q; 1225 #endif 1226 }; 1227 1228 struct rcu_node; 1229 1230 enum perf_event_task_context { 1231 perf_invalid_context = -1, 1232 perf_hw_context = 0, 1233 perf_sw_context, 1234 perf_nr_task_contexts, 1235 }; 1236 1237 struct task_struct { 1238 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */ 1239 void *stack; 1240 atomic_t usage; 1241 unsigned int flags; /* per process flags, defined below */ 1242 unsigned int ptrace; 1243 1244 #ifdef CONFIG_SMP 1245 struct llist_node wake_entry; 1246 int on_cpu; 1247 #endif 1248 int on_rq; 1249 1250 int prio, static_prio, normal_prio; 1251 unsigned int rt_priority; 1252 const struct sched_class *sched_class; 1253 struct sched_entity se; 1254 struct sched_rt_entity rt; 1255 1256 #ifdef CONFIG_PREEMPT_NOTIFIERS 1257 /* list of struct preempt_notifier: */ 1258 struct hlist_head preempt_notifiers; 1259 #endif 1260 1261 /* 1262 * fpu_counter contains the number of consecutive context switches 1263 * that the FPU is used. If this is over a threshold, the lazy fpu 1264 * saving becomes unlazy to save the trap. This is an unsigned char 1265 * so that after 256 times the counter wraps and the behavior turns 1266 * lazy again; this to deal with bursty apps that only use FPU for 1267 * a short time 1268 */ 1269 unsigned char fpu_counter; 1270 #ifdef CONFIG_BLK_DEV_IO_TRACE 1271 unsigned int btrace_seq; 1272 #endif 1273 1274 unsigned int policy; 1275 cpumask_t cpus_allowed; 1276 1277 #ifdef CONFIG_PREEMPT_RCU 1278 int rcu_read_lock_nesting; 1279 char rcu_read_unlock_special; 1280 struct list_head rcu_node_entry; 1281 #endif /* #ifdef CONFIG_PREEMPT_RCU */ 1282 #ifdef CONFIG_TREE_PREEMPT_RCU 1283 struct rcu_node *rcu_blocked_node; 1284 #endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */ 1285 #ifdef CONFIG_RCU_BOOST 1286 struct rt_mutex *rcu_boost_mutex; 1287 #endif /* #ifdef CONFIG_RCU_BOOST */ 1288 1289 #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) 1290 struct sched_info sched_info; 1291 #endif 1292 1293 struct list_head tasks; 1294 #ifdef CONFIG_SMP 1295 struct plist_node pushable_tasks; 1296 #endif 1297 1298 struct mm_struct *mm, *active_mm; 1299 #ifdef CONFIG_COMPAT_BRK 1300 unsigned brk_randomized:1; 1301 #endif 1302 #if defined(SPLIT_RSS_COUNTING) 1303 struct task_rss_stat rss_stat; 1304 #endif 1305 /* task state */ 1306 int exit_state; 1307 int exit_code, exit_signal; 1308 int pdeath_signal; /* The signal sent when the parent dies */ 1309 unsigned int jobctl; /* JOBCTL_*, siglock protected */ 1310 /* ??? */ 1311 unsigned int personality; 1312 unsigned did_exec:1; 1313 unsigned in_execve:1; /* Tell the LSMs that the process is doing an 1314 * execve */ 1315 unsigned in_iowait:1; 1316 1317 1318 /* Revert to default priority/policy when forking */ 1319 unsigned sched_reset_on_fork:1; 1320 unsigned sched_contributes_to_load:1; 1321 1322 pid_t pid; 1323 pid_t tgid; 1324 1325 #ifdef CONFIG_CC_STACKPROTECTOR 1326 /* Canary value for the -fstack-protector gcc feature */ 1327 unsigned long stack_canary; 1328 #endif 1329 1330 /* 1331 * pointers to (original) parent process, youngest child, younger sibling, 1332 * older sibling, respectively. (p->father can be replaced with 1333 * p->real_parent->pid) 1334 */ 1335 struct task_struct __rcu *real_parent; /* real parent process */ 1336 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */ 1337 /* 1338 * children/sibling forms the list of my natural children 1339 */ 1340 struct list_head children; /* list of my children */ 1341 struct list_head sibling; /* linkage in my parent's children list */ 1342 struct task_struct *group_leader; /* threadgroup leader */ 1343 1344 /* 1345 * ptraced is the list of tasks this task is using ptrace on. 1346 * This includes both natural children and PTRACE_ATTACH targets. 1347 * p->ptrace_entry is p's link on the p->parent->ptraced list. 1348 */ 1349 struct list_head ptraced; 1350 struct list_head ptrace_entry; 1351 1352 /* PID/PID hash table linkage. */ 1353 struct pid_link pids[PIDTYPE_MAX]; 1354 struct list_head thread_group; 1355 1356 struct completion *vfork_done; /* for vfork() */ 1357 int __user *set_child_tid; /* CLONE_CHILD_SETTID */ 1358 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */ 1359 1360 cputime_t utime, stime, utimescaled, stimescaled; 1361 cputime_t gtime; 1362 #ifndef CONFIG_VIRT_CPU_ACCOUNTING 1363 cputime_t prev_utime, prev_stime; 1364 #endif 1365 unsigned long nvcsw, nivcsw; /* context switch counts */ 1366 struct timespec start_time; /* monotonic time */ 1367 struct timespec real_start_time; /* boot based time */ 1368 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */ 1369 unsigned long min_flt, maj_flt; 1370 1371 struct task_cputime cputime_expires; 1372 struct list_head cpu_timers[3]; 1373 1374 /* process credentials */ 1375 const struct cred __rcu *real_cred; /* objective and real subjective task 1376 * credentials (COW) */ 1377 const struct cred __rcu *cred; /* effective (overridable) subjective task 1378 * credentials (COW) */ 1379 struct cred *replacement_session_keyring; /* for KEYCTL_SESSION_TO_PARENT */ 1380 1381 char comm[TASK_COMM_LEN]; /* executable name excluding path 1382 - access with [gs]et_task_comm (which lock 1383 it with task_lock()) 1384 - initialized normally by setup_new_exec */ 1385 /* file system info */ 1386 int link_count, total_link_count; 1387 #ifdef CONFIG_SYSVIPC 1388 /* ipc stuff */ 1389 struct sysv_sem sysvsem; 1390 #endif 1391 #ifdef CONFIG_DETECT_HUNG_TASK 1392 /* hung task detection */ 1393 unsigned long last_switch_count; 1394 #endif 1395 /* CPU-specific state of this task */ 1396 struct thread_struct thread; 1397 /* filesystem information */ 1398 struct fs_struct *fs; 1399 /* open file information */ 1400 struct files_struct *files; 1401 /* namespaces */ 1402 struct nsproxy *nsproxy; 1403 /* signal handlers */ 1404 struct signal_struct *signal; 1405 struct sighand_struct *sighand; 1406 1407 sigset_t blocked, real_blocked; 1408 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */ 1409 struct sigpending pending; 1410 1411 unsigned long sas_ss_sp; 1412 size_t sas_ss_size; 1413 int (*notifier)(void *priv); 1414 void *notifier_data; 1415 sigset_t *notifier_mask; 1416 struct audit_context *audit_context; 1417 #ifdef CONFIG_AUDITSYSCALL 1418 uid_t loginuid; 1419 unsigned int sessionid; 1420 #endif 1421 seccomp_t seccomp; 1422 1423 /* Thread group tracking */ 1424 u32 parent_exec_id; 1425 u32 self_exec_id; 1426 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, 1427 * mempolicy */ 1428 spinlock_t alloc_lock; 1429 1430 #ifdef CONFIG_GENERIC_HARDIRQS 1431 /* IRQ handler threads */ 1432 struct irqaction *irqaction; 1433 #endif 1434 1435 /* Protection of the PI data structures: */ 1436 raw_spinlock_t pi_lock; 1437 1438 #ifdef CONFIG_RT_MUTEXES 1439 /* PI waiters blocked on a rt_mutex held by this task */ 1440 struct plist_head pi_waiters; 1441 /* Deadlock detection and priority inheritance handling */ 1442 struct rt_mutex_waiter *pi_blocked_on; 1443 #endif 1444 1445 #ifdef CONFIG_DEBUG_MUTEXES 1446 /* mutex deadlock detection */ 1447 struct mutex_waiter *blocked_on; 1448 #endif 1449 #ifdef CONFIG_TRACE_IRQFLAGS 1450 unsigned int irq_events; 1451 unsigned long hardirq_enable_ip; 1452 unsigned long hardirq_disable_ip; 1453 unsigned int hardirq_enable_event; 1454 unsigned int hardirq_disable_event; 1455 int hardirqs_enabled; 1456 int hardirq_context; 1457 unsigned long softirq_disable_ip; 1458 unsigned long softirq_enable_ip; 1459 unsigned int softirq_disable_event; 1460 unsigned int softirq_enable_event; 1461 int softirqs_enabled; 1462 int softirq_context; 1463 #endif 1464 #ifdef CONFIG_LOCKDEP 1465 # define MAX_LOCK_DEPTH 48UL 1466 u64 curr_chain_key; 1467 int lockdep_depth; 1468 unsigned int lockdep_recursion; 1469 struct held_lock held_locks[MAX_LOCK_DEPTH]; 1470 gfp_t lockdep_reclaim_gfp; 1471 #endif 1472 1473 /* journalling filesystem info */ 1474 void *journal_info; 1475 1476 /* stacked block device info */ 1477 struct bio_list *bio_list; 1478 1479 #ifdef CONFIG_BLOCK 1480 /* stack plugging */ 1481 struct blk_plug *plug; 1482 #endif 1483 1484 /* VM state */ 1485 struct reclaim_state *reclaim_state; 1486 1487 struct backing_dev_info *backing_dev_info; 1488 1489 struct io_context *io_context; 1490 1491 unsigned long ptrace_message; 1492 siginfo_t *last_siginfo; /* For ptrace use. */ 1493 struct task_io_accounting ioac; 1494 #if defined(CONFIG_TASK_XACCT) 1495 u64 acct_rss_mem1; /* accumulated rss usage */ 1496 u64 acct_vm_mem1; /* accumulated virtual memory usage */ 1497 cputime_t acct_timexpd; /* stime + utime since last update */ 1498 #endif 1499 #ifdef CONFIG_CPUSETS 1500 nodemask_t mems_allowed; /* Protected by alloc_lock */ 1501 int mems_allowed_change_disable; 1502 int cpuset_mem_spread_rotor; 1503 int cpuset_slab_spread_rotor; 1504 #endif 1505 #ifdef CONFIG_CGROUPS 1506 /* Control Group info protected by css_set_lock */ 1507 struct css_set __rcu *cgroups; 1508 /* cg_list protected by css_set_lock and tsk->alloc_lock */ 1509 struct list_head cg_list; 1510 #endif 1511 #ifdef CONFIG_FUTEX 1512 struct robust_list_head __user *robust_list; 1513 #ifdef CONFIG_COMPAT 1514 struct compat_robust_list_head __user *compat_robust_list; 1515 #endif 1516 struct list_head pi_state_list; 1517 struct futex_pi_state *pi_state_cache; 1518 #endif 1519 #ifdef CONFIG_PERF_EVENTS 1520 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts]; 1521 struct mutex perf_event_mutex; 1522 struct list_head perf_event_list; 1523 #endif 1524 #ifdef CONFIG_NUMA 1525 struct mempolicy *mempolicy; /* Protected by alloc_lock */ 1526 short il_next; 1527 short pref_node_fork; 1528 #endif 1529 struct rcu_head rcu; 1530 1531 /* 1532 * cache last used pipe for splice 1533 */ 1534 struct pipe_inode_info *splice_pipe; 1535 #ifdef CONFIG_TASK_DELAY_ACCT 1536 struct task_delay_info *delays; 1537 #endif 1538 #ifdef CONFIG_FAULT_INJECTION 1539 int make_it_fail; 1540 #endif 1541 /* 1542 * when (nr_dirtied >= nr_dirtied_pause), it's time to call 1543 * balance_dirty_pages() for some dirty throttling pause 1544 */ 1545 int nr_dirtied; 1546 int nr_dirtied_pause; 1547 unsigned long dirty_paused_when; /* start of a write-and-pause period */ 1548 1549 #ifdef CONFIG_LATENCYTOP 1550 int latency_record_count; 1551 struct latency_record latency_record[LT_SAVECOUNT]; 1552 #endif 1553 /* 1554 * time slack values; these are used to round up poll() and 1555 * select() etc timeout values. These are in nanoseconds. 1556 */ 1557 unsigned long timer_slack_ns; 1558 unsigned long default_timer_slack_ns; 1559 1560 struct list_head *scm_work_list; 1561 #ifdef CONFIG_FUNCTION_GRAPH_TRACER 1562 /* Index of current stored address in ret_stack */ 1563 int curr_ret_stack; 1564 /* Stack of return addresses for return function tracing */ 1565 struct ftrace_ret_stack *ret_stack; 1566 /* time stamp for last schedule */ 1567 unsigned long long ftrace_timestamp; 1568 /* 1569 * Number of functions that haven't been traced 1570 * because of depth overrun. 1571 */ 1572 atomic_t trace_overrun; 1573 /* Pause for the tracing */ 1574 atomic_t tracing_graph_pause; 1575 #endif 1576 #ifdef CONFIG_TRACING 1577 /* state flags for use by tracers */ 1578 unsigned long trace; 1579 /* bitmask and counter of trace recursion */ 1580 unsigned long trace_recursion; 1581 #endif /* CONFIG_TRACING */ 1582 #ifdef CONFIG_CGROUP_MEM_RES_CTLR /* memcg uses this to do batch job */ 1583 struct memcg_batch_info { 1584 int do_batch; /* incremented when batch uncharge started */ 1585 struct mem_cgroup *memcg; /* target memcg of uncharge */ 1586 unsigned long nr_pages; /* uncharged usage */ 1587 unsigned long memsw_nr_pages; /* uncharged mem+swap usage */ 1588 } memcg_batch; 1589 #endif 1590 #ifdef CONFIG_HAVE_HW_BREAKPOINT 1591 atomic_t ptrace_bp_refcnt; 1592 #endif 1593 }; 1594 1595 /* Future-safe accessor for struct task_struct's cpus_allowed. */ 1596 #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed) 1597 1598 /* 1599 * Priority of a process goes from 0..MAX_PRIO-1, valid RT 1600 * priority is 0..MAX_RT_PRIO-1, and SCHED_NORMAL/SCHED_BATCH 1601 * tasks are in the range MAX_RT_PRIO..MAX_PRIO-1. Priority 1602 * values are inverted: lower p->prio value means higher priority. 1603 * 1604 * The MAX_USER_RT_PRIO value allows the actual maximum 1605 * RT priority to be separate from the value exported to 1606 * user-space. This allows kernel threads to set their 1607 * priority to a value higher than any user task. Note: 1608 * MAX_RT_PRIO must not be smaller than MAX_USER_RT_PRIO. 1609 */ 1610 1611 #define MAX_USER_RT_PRIO 100 1612 #define MAX_RT_PRIO MAX_USER_RT_PRIO 1613 1614 #define MAX_PRIO (MAX_RT_PRIO + 40) 1615 #define DEFAULT_PRIO (MAX_RT_PRIO + 20) 1616 1617 static inline int rt_prio(int prio) 1618 { 1619 if (unlikely(prio < MAX_RT_PRIO)) 1620 return 1; 1621 return 0; 1622 } 1623 1624 static inline int rt_task(struct task_struct *p) 1625 { 1626 return rt_prio(p->prio); 1627 } 1628 1629 static inline struct pid *task_pid(struct task_struct *task) 1630 { 1631 return task->pids[PIDTYPE_PID].pid; 1632 } 1633 1634 static inline struct pid *task_tgid(struct task_struct *task) 1635 { 1636 return task->group_leader->pids[PIDTYPE_PID].pid; 1637 } 1638 1639 /* 1640 * Without tasklist or rcu lock it is not safe to dereference 1641 * the result of task_pgrp/task_session even if task == current, 1642 * we can race with another thread doing sys_setsid/sys_setpgid. 1643 */ 1644 static inline struct pid *task_pgrp(struct task_struct *task) 1645 { 1646 return task->group_leader->pids[PIDTYPE_PGID].pid; 1647 } 1648 1649 static inline struct pid *task_session(struct task_struct *task) 1650 { 1651 return task->group_leader->pids[PIDTYPE_SID].pid; 1652 } 1653 1654 struct pid_namespace; 1655 1656 /* 1657 * the helpers to get the task's different pids as they are seen 1658 * from various namespaces 1659 * 1660 * task_xid_nr() : global id, i.e. the id seen from the init namespace; 1661 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of 1662 * current. 1663 * task_xid_nr_ns() : id seen from the ns specified; 1664 * 1665 * set_task_vxid() : assigns a virtual id to a task; 1666 * 1667 * see also pid_nr() etc in include/linux/pid.h 1668 */ 1669 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, 1670 struct pid_namespace *ns); 1671 1672 static inline pid_t task_pid_nr(struct task_struct *tsk) 1673 { 1674 return tsk->pid; 1675 } 1676 1677 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, 1678 struct pid_namespace *ns) 1679 { 1680 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns); 1681 } 1682 1683 static inline pid_t task_pid_vnr(struct task_struct *tsk) 1684 { 1685 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL); 1686 } 1687 1688 1689 static inline pid_t task_tgid_nr(struct task_struct *tsk) 1690 { 1691 return tsk->tgid; 1692 } 1693 1694 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns); 1695 1696 static inline pid_t task_tgid_vnr(struct task_struct *tsk) 1697 { 1698 return pid_vnr(task_tgid(tsk)); 1699 } 1700 1701 1702 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, 1703 struct pid_namespace *ns) 1704 { 1705 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns); 1706 } 1707 1708 static inline pid_t task_pgrp_vnr(struct task_struct *tsk) 1709 { 1710 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL); 1711 } 1712 1713 1714 static inline pid_t task_session_nr_ns(struct task_struct *tsk, 1715 struct pid_namespace *ns) 1716 { 1717 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns); 1718 } 1719 1720 static inline pid_t task_session_vnr(struct task_struct *tsk) 1721 { 1722 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL); 1723 } 1724 1725 /* obsolete, do not use */ 1726 static inline pid_t task_pgrp_nr(struct task_struct *tsk) 1727 { 1728 return task_pgrp_nr_ns(tsk, &init_pid_ns); 1729 } 1730 1731 /** 1732 * pid_alive - check that a task structure is not stale 1733 * @p: Task structure to be checked. 1734 * 1735 * Test if a process is not yet dead (at most zombie state) 1736 * If pid_alive fails, then pointers within the task structure 1737 * can be stale and must not be dereferenced. 1738 */ 1739 static inline int pid_alive(struct task_struct *p) 1740 { 1741 return p->pids[PIDTYPE_PID].pid != NULL; 1742 } 1743 1744 /** 1745 * is_global_init - check if a task structure is init 1746 * @tsk: Task structure to be checked. 1747 * 1748 * Check if a task structure is the first user space task the kernel created. 1749 */ 1750 static inline int is_global_init(struct task_struct *tsk) 1751 { 1752 return tsk->pid == 1; 1753 } 1754 1755 /* 1756 * is_container_init: 1757 * check whether in the task is init in its own pid namespace. 1758 */ 1759 extern int is_container_init(struct task_struct *tsk); 1760 1761 extern struct pid *cad_pid; 1762 1763 extern void free_task(struct task_struct *tsk); 1764 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0) 1765 1766 extern void __put_task_struct(struct task_struct *t); 1767 1768 static inline void put_task_struct(struct task_struct *t) 1769 { 1770 if (atomic_dec_and_test(&t->usage)) 1771 __put_task_struct(t); 1772 } 1773 1774 extern void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st); 1775 extern void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st); 1776 1777 /* 1778 * Per process flags 1779 */ 1780 #define PF_STARTING 0x00000002 /* being created */ 1781 #define PF_EXITING 0x00000004 /* getting shut down */ 1782 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */ 1783 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */ 1784 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */ 1785 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */ 1786 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */ 1787 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */ 1788 #define PF_DUMPCORE 0x00000200 /* dumped core */ 1789 #define PF_SIGNALED 0x00000400 /* killed by a signal */ 1790 #define PF_MEMALLOC 0x00000800 /* Allocating memory */ 1791 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */ 1792 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */ 1793 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */ 1794 #define PF_FROZEN 0x00010000 /* frozen for system suspend */ 1795 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */ 1796 #define PF_KSWAPD 0x00040000 /* I am kswapd */ 1797 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */ 1798 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */ 1799 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */ 1800 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */ 1801 #define PF_SPREAD_PAGE 0x01000000 /* Spread page cache over cpuset */ 1802 #define PF_SPREAD_SLAB 0x02000000 /* Spread some slab caches over cpuset */ 1803 #define PF_THREAD_BOUND 0x04000000 /* Thread bound to specific cpu */ 1804 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */ 1805 #define PF_MEMPOLICY 0x10000000 /* Non-default NUMA mempolicy */ 1806 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */ 1807 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */ 1808 1809 /* 1810 * Only the _current_ task can read/write to tsk->flags, but other 1811 * tasks can access tsk->flags in readonly mode for example 1812 * with tsk_used_math (like during threaded core dumping). 1813 * There is however an exception to this rule during ptrace 1814 * or during fork: the ptracer task is allowed to write to the 1815 * child->flags of its traced child (same goes for fork, the parent 1816 * can write to the child->flags), because we're guaranteed the 1817 * child is not running and in turn not changing child->flags 1818 * at the same time the parent does it. 1819 */ 1820 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0) 1821 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0) 1822 #define clear_used_math() clear_stopped_child_used_math(current) 1823 #define set_used_math() set_stopped_child_used_math(current) 1824 #define conditional_stopped_child_used_math(condition, child) \ 1825 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0) 1826 #define conditional_used_math(condition) \ 1827 conditional_stopped_child_used_math(condition, current) 1828 #define copy_to_stopped_child_used_math(child) \ 1829 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0) 1830 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */ 1831 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH) 1832 #define used_math() tsk_used_math(current) 1833 1834 /* 1835 * task->jobctl flags 1836 */ 1837 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */ 1838 1839 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */ 1840 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */ 1841 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */ 1842 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */ 1843 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */ 1844 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */ 1845 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */ 1846 1847 #define JOBCTL_STOP_DEQUEUED (1 << JOBCTL_STOP_DEQUEUED_BIT) 1848 #define JOBCTL_STOP_PENDING (1 << JOBCTL_STOP_PENDING_BIT) 1849 #define JOBCTL_STOP_CONSUME (1 << JOBCTL_STOP_CONSUME_BIT) 1850 #define JOBCTL_TRAP_STOP (1 << JOBCTL_TRAP_STOP_BIT) 1851 #define JOBCTL_TRAP_NOTIFY (1 << JOBCTL_TRAP_NOTIFY_BIT) 1852 #define JOBCTL_TRAPPING (1 << JOBCTL_TRAPPING_BIT) 1853 #define JOBCTL_LISTENING (1 << JOBCTL_LISTENING_BIT) 1854 1855 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY) 1856 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK) 1857 1858 extern bool task_set_jobctl_pending(struct task_struct *task, 1859 unsigned int mask); 1860 extern void task_clear_jobctl_trapping(struct task_struct *task); 1861 extern void task_clear_jobctl_pending(struct task_struct *task, 1862 unsigned int mask); 1863 1864 #ifdef CONFIG_PREEMPT_RCU 1865 1866 #define RCU_READ_UNLOCK_BLOCKED (1 << 0) /* blocked while in RCU read-side. */ 1867 #define RCU_READ_UNLOCK_BOOSTED (1 << 1) /* boosted while in RCU read-side. */ 1868 #define RCU_READ_UNLOCK_NEED_QS (1 << 2) /* RCU core needs CPU response. */ 1869 1870 static inline void rcu_copy_process(struct task_struct *p) 1871 { 1872 p->rcu_read_lock_nesting = 0; 1873 p->rcu_read_unlock_special = 0; 1874 #ifdef CONFIG_TREE_PREEMPT_RCU 1875 p->rcu_blocked_node = NULL; 1876 #endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */ 1877 #ifdef CONFIG_RCU_BOOST 1878 p->rcu_boost_mutex = NULL; 1879 #endif /* #ifdef CONFIG_RCU_BOOST */ 1880 INIT_LIST_HEAD(&p->rcu_node_entry); 1881 } 1882 1883 #else 1884 1885 static inline void rcu_copy_process(struct task_struct *p) 1886 { 1887 } 1888 1889 #endif 1890 1891 #ifdef CONFIG_SMP 1892 extern void do_set_cpus_allowed(struct task_struct *p, 1893 const struct cpumask *new_mask); 1894 1895 extern int set_cpus_allowed_ptr(struct task_struct *p, 1896 const struct cpumask *new_mask); 1897 #else 1898 static inline void do_set_cpus_allowed(struct task_struct *p, 1899 const struct cpumask *new_mask) 1900 { 1901 } 1902 static inline int set_cpus_allowed_ptr(struct task_struct *p, 1903 const struct cpumask *new_mask) 1904 { 1905 if (!cpumask_test_cpu(0, new_mask)) 1906 return -EINVAL; 1907 return 0; 1908 } 1909 #endif 1910 1911 #ifndef CONFIG_CPUMASK_OFFSTACK 1912 static inline int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask) 1913 { 1914 return set_cpus_allowed_ptr(p, &new_mask); 1915 } 1916 #endif 1917 1918 /* 1919 * Do not use outside of architecture code which knows its limitations. 1920 * 1921 * sched_clock() has no promise of monotonicity or bounded drift between 1922 * CPUs, use (which you should not) requires disabling IRQs. 1923 * 1924 * Please use one of the three interfaces below. 1925 */ 1926 extern unsigned long long notrace sched_clock(void); 1927 /* 1928 * See the comment in kernel/sched_clock.c 1929 */ 1930 extern u64 cpu_clock(int cpu); 1931 extern u64 local_clock(void); 1932 extern u64 sched_clock_cpu(int cpu); 1933 1934 1935 extern void sched_clock_init(void); 1936 1937 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK 1938 static inline void sched_clock_tick(void) 1939 { 1940 } 1941 1942 static inline void sched_clock_idle_sleep_event(void) 1943 { 1944 } 1945 1946 static inline void sched_clock_idle_wakeup_event(u64 delta_ns) 1947 { 1948 } 1949 #else 1950 /* 1951 * Architectures can set this to 1 if they have specified 1952 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig, 1953 * but then during bootup it turns out that sched_clock() 1954 * is reliable after all: 1955 */ 1956 extern int sched_clock_stable; 1957 1958 extern void sched_clock_tick(void); 1959 extern void sched_clock_idle_sleep_event(void); 1960 extern void sched_clock_idle_wakeup_event(u64 delta_ns); 1961 #endif 1962 1963 #ifdef CONFIG_IRQ_TIME_ACCOUNTING 1964 /* 1965 * An i/f to runtime opt-in for irq time accounting based off of sched_clock. 1966 * The reason for this explicit opt-in is not to have perf penalty with 1967 * slow sched_clocks. 1968 */ 1969 extern void enable_sched_clock_irqtime(void); 1970 extern void disable_sched_clock_irqtime(void); 1971 #else 1972 static inline void enable_sched_clock_irqtime(void) {} 1973 static inline void disable_sched_clock_irqtime(void) {} 1974 #endif 1975 1976 extern unsigned long long 1977 task_sched_runtime(struct task_struct *task); 1978 1979 /* sched_exec is called by processes performing an exec */ 1980 #ifdef CONFIG_SMP 1981 extern void sched_exec(void); 1982 #else 1983 #define sched_exec() {} 1984 #endif 1985 1986 extern void sched_clock_idle_sleep_event(void); 1987 extern void sched_clock_idle_wakeup_event(u64 delta_ns); 1988 1989 #ifdef CONFIG_HOTPLUG_CPU 1990 extern void idle_task_exit(void); 1991 #else 1992 static inline void idle_task_exit(void) {} 1993 #endif 1994 1995 #if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP) 1996 extern void wake_up_idle_cpu(int cpu); 1997 #else 1998 static inline void wake_up_idle_cpu(int cpu) { } 1999 #endif 2000 2001 extern unsigned int sysctl_sched_latency; 2002 extern unsigned int sysctl_sched_min_granularity; 2003 extern unsigned int sysctl_sched_wakeup_granularity; 2004 extern unsigned int sysctl_sched_child_runs_first; 2005 2006 enum sched_tunable_scaling { 2007 SCHED_TUNABLESCALING_NONE, 2008 SCHED_TUNABLESCALING_LOG, 2009 SCHED_TUNABLESCALING_LINEAR, 2010 SCHED_TUNABLESCALING_END, 2011 }; 2012 extern enum sched_tunable_scaling sysctl_sched_tunable_scaling; 2013 2014 #ifdef CONFIG_SCHED_DEBUG 2015 extern unsigned int sysctl_sched_migration_cost; 2016 extern unsigned int sysctl_sched_nr_migrate; 2017 extern unsigned int sysctl_sched_time_avg; 2018 extern unsigned int sysctl_timer_migration; 2019 extern unsigned int sysctl_sched_shares_window; 2020 2021 int sched_proc_update_handler(struct ctl_table *table, int write, 2022 void __user *buffer, size_t *length, 2023 loff_t *ppos); 2024 #endif 2025 #ifdef CONFIG_SCHED_DEBUG 2026 static inline unsigned int get_sysctl_timer_migration(void) 2027 { 2028 return sysctl_timer_migration; 2029 } 2030 #else 2031 static inline unsigned int get_sysctl_timer_migration(void) 2032 { 2033 return 1; 2034 } 2035 #endif 2036 extern unsigned int sysctl_sched_rt_period; 2037 extern int sysctl_sched_rt_runtime; 2038 2039 int sched_rt_handler(struct ctl_table *table, int write, 2040 void __user *buffer, size_t *lenp, 2041 loff_t *ppos); 2042 2043 #ifdef CONFIG_SCHED_AUTOGROUP 2044 extern unsigned int sysctl_sched_autogroup_enabled; 2045 2046 extern void sched_autogroup_create_attach(struct task_struct *p); 2047 extern void sched_autogroup_detach(struct task_struct *p); 2048 extern void sched_autogroup_fork(struct signal_struct *sig); 2049 extern void sched_autogroup_exit(struct signal_struct *sig); 2050 #ifdef CONFIG_PROC_FS 2051 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m); 2052 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int *nice); 2053 #endif 2054 #else 2055 static inline void sched_autogroup_create_attach(struct task_struct *p) { } 2056 static inline void sched_autogroup_detach(struct task_struct *p) { } 2057 static inline void sched_autogroup_fork(struct signal_struct *sig) { } 2058 static inline void sched_autogroup_exit(struct signal_struct *sig) { } 2059 #endif 2060 2061 #ifdef CONFIG_CFS_BANDWIDTH 2062 extern unsigned int sysctl_sched_cfs_bandwidth_slice; 2063 #endif 2064 2065 #ifdef CONFIG_RT_MUTEXES 2066 extern int rt_mutex_getprio(struct task_struct *p); 2067 extern void rt_mutex_setprio(struct task_struct *p, int prio); 2068 extern void rt_mutex_adjust_pi(struct task_struct *p); 2069 #else 2070 static inline int rt_mutex_getprio(struct task_struct *p) 2071 { 2072 return p->normal_prio; 2073 } 2074 # define rt_mutex_adjust_pi(p) do { } while (0) 2075 #endif 2076 2077 extern bool yield_to(struct task_struct *p, bool preempt); 2078 extern void set_user_nice(struct task_struct *p, long nice); 2079 extern int task_prio(const struct task_struct *p); 2080 extern int task_nice(const struct task_struct *p); 2081 extern int can_nice(const struct task_struct *p, const int nice); 2082 extern int task_curr(const struct task_struct *p); 2083 extern int idle_cpu(int cpu); 2084 extern int sched_setscheduler(struct task_struct *, int, 2085 const struct sched_param *); 2086 extern int sched_setscheduler_nocheck(struct task_struct *, int, 2087 const struct sched_param *); 2088 extern struct task_struct *idle_task(int cpu); 2089 /** 2090 * is_idle_task - is the specified task an idle task? 2091 * @tsk: the task in question. 2092 */ 2093 static inline bool is_idle_task(struct task_struct *p) 2094 { 2095 return p->pid == 0; 2096 } 2097 extern struct task_struct *curr_task(int cpu); 2098 extern void set_curr_task(int cpu, struct task_struct *p); 2099 2100 void yield(void); 2101 2102 /* 2103 * The default (Linux) execution domain. 2104 */ 2105 extern struct exec_domain default_exec_domain; 2106 2107 union thread_union { 2108 struct thread_info thread_info; 2109 unsigned long stack[THREAD_SIZE/sizeof(long)]; 2110 }; 2111 2112 #ifndef __HAVE_ARCH_KSTACK_END 2113 static inline int kstack_end(void *addr) 2114 { 2115 /* Reliable end of stack detection: 2116 * Some APM bios versions misalign the stack 2117 */ 2118 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*))); 2119 } 2120 #endif 2121 2122 extern union thread_union init_thread_union; 2123 extern struct task_struct init_task; 2124 2125 extern struct mm_struct init_mm; 2126 2127 extern struct pid_namespace init_pid_ns; 2128 2129 /* 2130 * find a task by one of its numerical ids 2131 * 2132 * find_task_by_pid_ns(): 2133 * finds a task by its pid in the specified namespace 2134 * find_task_by_vpid(): 2135 * finds a task by its virtual pid 2136 * 2137 * see also find_vpid() etc in include/linux/pid.h 2138 */ 2139 2140 extern struct task_struct *find_task_by_vpid(pid_t nr); 2141 extern struct task_struct *find_task_by_pid_ns(pid_t nr, 2142 struct pid_namespace *ns); 2143 2144 extern void __set_special_pids(struct pid *pid); 2145 2146 /* per-UID process charging. */ 2147 extern struct user_struct * alloc_uid(struct user_namespace *, uid_t); 2148 static inline struct user_struct *get_uid(struct user_struct *u) 2149 { 2150 atomic_inc(&u->__count); 2151 return u; 2152 } 2153 extern void free_uid(struct user_struct *); 2154 extern void release_uids(struct user_namespace *ns); 2155 2156 #include <asm/current.h> 2157 2158 extern void xtime_update(unsigned long ticks); 2159 2160 extern int wake_up_state(struct task_struct *tsk, unsigned int state); 2161 extern int wake_up_process(struct task_struct *tsk); 2162 extern void wake_up_new_task(struct task_struct *tsk); 2163 #ifdef CONFIG_SMP 2164 extern void kick_process(struct task_struct *tsk); 2165 #else 2166 static inline void kick_process(struct task_struct *tsk) { } 2167 #endif 2168 extern void sched_fork(struct task_struct *p); 2169 extern void sched_dead(struct task_struct *p); 2170 2171 extern void proc_caches_init(void); 2172 extern void flush_signals(struct task_struct *); 2173 extern void __flush_signals(struct task_struct *); 2174 extern void ignore_signals(struct task_struct *); 2175 extern void flush_signal_handlers(struct task_struct *, int force_default); 2176 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info); 2177 2178 static inline int dequeue_signal_lock(struct task_struct *tsk, sigset_t *mask, siginfo_t *info) 2179 { 2180 unsigned long flags; 2181 int ret; 2182 2183 spin_lock_irqsave(&tsk->sighand->siglock, flags); 2184 ret = dequeue_signal(tsk, mask, info); 2185 spin_unlock_irqrestore(&tsk->sighand->siglock, flags); 2186 2187 return ret; 2188 } 2189 2190 extern void block_all_signals(int (*notifier)(void *priv), void *priv, 2191 sigset_t *mask); 2192 extern void unblock_all_signals(void); 2193 extern void release_task(struct task_struct * p); 2194 extern int send_sig_info(int, struct siginfo *, struct task_struct *); 2195 extern int force_sigsegv(int, struct task_struct *); 2196 extern int force_sig_info(int, struct siginfo *, struct task_struct *); 2197 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp); 2198 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid); 2199 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *, 2200 const struct cred *, u32); 2201 extern int kill_pgrp(struct pid *pid, int sig, int priv); 2202 extern int kill_pid(struct pid *pid, int sig, int priv); 2203 extern int kill_proc_info(int, struct siginfo *, pid_t); 2204 extern __must_check bool do_notify_parent(struct task_struct *, int); 2205 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent); 2206 extern void force_sig(int, struct task_struct *); 2207 extern int send_sig(int, struct task_struct *, int); 2208 extern int zap_other_threads(struct task_struct *p); 2209 extern struct sigqueue *sigqueue_alloc(void); 2210 extern void sigqueue_free(struct sigqueue *); 2211 extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group); 2212 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *); 2213 extern int do_sigaltstack(const stack_t __user *, stack_t __user *, unsigned long); 2214 2215 static inline int kill_cad_pid(int sig, int priv) 2216 { 2217 return kill_pid(cad_pid, sig, priv); 2218 } 2219 2220 /* These can be the second arg to send_sig_info/send_group_sig_info. */ 2221 #define SEND_SIG_NOINFO ((struct siginfo *) 0) 2222 #define SEND_SIG_PRIV ((struct siginfo *) 1) 2223 #define SEND_SIG_FORCED ((struct siginfo *) 2) 2224 2225 /* 2226 * True if we are on the alternate signal stack. 2227 */ 2228 static inline int on_sig_stack(unsigned long sp) 2229 { 2230 #ifdef CONFIG_STACK_GROWSUP 2231 return sp >= current->sas_ss_sp && 2232 sp - current->sas_ss_sp < current->sas_ss_size; 2233 #else 2234 return sp > current->sas_ss_sp && 2235 sp - current->sas_ss_sp <= current->sas_ss_size; 2236 #endif 2237 } 2238 2239 static inline int sas_ss_flags(unsigned long sp) 2240 { 2241 return (current->sas_ss_size == 0 ? SS_DISABLE 2242 : on_sig_stack(sp) ? SS_ONSTACK : 0); 2243 } 2244 2245 /* 2246 * Routines for handling mm_structs 2247 */ 2248 extern struct mm_struct * mm_alloc(void); 2249 2250 /* mmdrop drops the mm and the page tables */ 2251 extern void __mmdrop(struct mm_struct *); 2252 static inline void mmdrop(struct mm_struct * mm) 2253 { 2254 if (unlikely(atomic_dec_and_test(&mm->mm_count))) 2255 __mmdrop(mm); 2256 } 2257 2258 /* mmput gets rid of the mappings and all user-space */ 2259 extern void mmput(struct mm_struct *); 2260 /* Grab a reference to a task's mm, if it is not already going away */ 2261 extern struct mm_struct *get_task_mm(struct task_struct *task); 2262 /* Remove the current tasks stale references to the old mm_struct */ 2263 extern void mm_release(struct task_struct *, struct mm_struct *); 2264 /* Allocate a new mm structure and copy contents from tsk->mm */ 2265 extern struct mm_struct *dup_mm(struct task_struct *tsk); 2266 2267 extern int copy_thread(unsigned long, unsigned long, unsigned long, 2268 struct task_struct *, struct pt_regs *); 2269 extern void flush_thread(void); 2270 extern void exit_thread(void); 2271 2272 extern void exit_files(struct task_struct *); 2273 extern void __cleanup_sighand(struct sighand_struct *); 2274 2275 extern void exit_itimers(struct signal_struct *); 2276 extern void flush_itimer_signals(void); 2277 2278 extern void do_group_exit(int); 2279 2280 extern void daemonize(const char *, ...); 2281 extern int allow_signal(int); 2282 extern int disallow_signal(int); 2283 2284 extern int do_execve(const char *, 2285 const char __user * const __user *, 2286 const char __user * const __user *, struct pt_regs *); 2287 extern long do_fork(unsigned long, unsigned long, struct pt_regs *, unsigned long, int __user *, int __user *); 2288 struct task_struct *fork_idle(int); 2289 2290 extern void set_task_comm(struct task_struct *tsk, char *from); 2291 extern char *get_task_comm(char *to, struct task_struct *tsk); 2292 2293 #ifdef CONFIG_SMP 2294 void scheduler_ipi(void); 2295 extern unsigned long wait_task_inactive(struct task_struct *, long match_state); 2296 #else 2297 static inline void scheduler_ipi(void) { } 2298 static inline unsigned long wait_task_inactive(struct task_struct *p, 2299 long match_state) 2300 { 2301 return 1; 2302 } 2303 #endif 2304 2305 #define next_task(p) \ 2306 list_entry_rcu((p)->tasks.next, struct task_struct, tasks) 2307 2308 #define for_each_process(p) \ 2309 for (p = &init_task ; (p = next_task(p)) != &init_task ; ) 2310 2311 extern bool current_is_single_threaded(void); 2312 2313 /* 2314 * Careful: do_each_thread/while_each_thread is a double loop so 2315 * 'break' will not work as expected - use goto instead. 2316 */ 2317 #define do_each_thread(g, t) \ 2318 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do 2319 2320 #define while_each_thread(g, t) \ 2321 while ((t = next_thread(t)) != g) 2322 2323 static inline int get_nr_threads(struct task_struct *tsk) 2324 { 2325 return tsk->signal->nr_threads; 2326 } 2327 2328 static inline bool thread_group_leader(struct task_struct *p) 2329 { 2330 return p->exit_signal >= 0; 2331 } 2332 2333 /* Do to the insanities of de_thread it is possible for a process 2334 * to have the pid of the thread group leader without actually being 2335 * the thread group leader. For iteration through the pids in proc 2336 * all we care about is that we have a task with the appropriate 2337 * pid, we don't actually care if we have the right task. 2338 */ 2339 static inline int has_group_leader_pid(struct task_struct *p) 2340 { 2341 return p->pid == p->tgid; 2342 } 2343 2344 static inline 2345 int same_thread_group(struct task_struct *p1, struct task_struct *p2) 2346 { 2347 return p1->tgid == p2->tgid; 2348 } 2349 2350 static inline struct task_struct *next_thread(const struct task_struct *p) 2351 { 2352 return list_entry_rcu(p->thread_group.next, 2353 struct task_struct, thread_group); 2354 } 2355 2356 static inline int thread_group_empty(struct task_struct *p) 2357 { 2358 return list_empty(&p->thread_group); 2359 } 2360 2361 #define delay_group_leader(p) \ 2362 (thread_group_leader(p) && !thread_group_empty(p)) 2363 2364 /* 2365 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring 2366 * subscriptions and synchronises with wait4(). Also used in procfs. Also 2367 * pins the final release of task.io_context. Also protects ->cpuset and 2368 * ->cgroup.subsys[]. 2369 * 2370 * Nests both inside and outside of read_lock(&tasklist_lock). 2371 * It must not be nested with write_lock_irq(&tasklist_lock), 2372 * neither inside nor outside. 2373 */ 2374 static inline void task_lock(struct task_struct *p) 2375 { 2376 spin_lock(&p->alloc_lock); 2377 } 2378 2379 static inline void task_unlock(struct task_struct *p) 2380 { 2381 spin_unlock(&p->alloc_lock); 2382 } 2383 2384 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk, 2385 unsigned long *flags); 2386 2387 #define lock_task_sighand(tsk, flags) \ 2388 ({ struct sighand_struct *__ss; \ 2389 __cond_lock(&(tsk)->sighand->siglock, \ 2390 (__ss = __lock_task_sighand(tsk, flags))); \ 2391 __ss; \ 2392 }) \ 2393 2394 static inline void unlock_task_sighand(struct task_struct *tsk, 2395 unsigned long *flags) 2396 { 2397 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags); 2398 } 2399 2400 #ifdef CONFIG_CGROUPS 2401 static inline void threadgroup_change_begin(struct task_struct *tsk) 2402 { 2403 down_read(&tsk->signal->group_rwsem); 2404 } 2405 static inline void threadgroup_change_end(struct task_struct *tsk) 2406 { 2407 up_read(&tsk->signal->group_rwsem); 2408 } 2409 2410 /** 2411 * threadgroup_lock - lock threadgroup 2412 * @tsk: member task of the threadgroup to lock 2413 * 2414 * Lock the threadgroup @tsk belongs to. No new task is allowed to enter 2415 * and member tasks aren't allowed to exit (as indicated by PF_EXITING) or 2416 * perform exec. This is useful for cases where the threadgroup needs to 2417 * stay stable across blockable operations. 2418 * 2419 * fork and exit paths explicitly call threadgroup_change_{begin|end}() for 2420 * synchronization. While held, no new task will be added to threadgroup 2421 * and no existing live task will have its PF_EXITING set. 2422 * 2423 * During exec, a task goes and puts its thread group through unusual 2424 * changes. After de-threading, exclusive access is assumed to resources 2425 * which are usually shared by tasks in the same group - e.g. sighand may 2426 * be replaced with a new one. Also, the exec'ing task takes over group 2427 * leader role including its pid. Exclude these changes while locked by 2428 * grabbing cred_guard_mutex which is used to synchronize exec path. 2429 */ 2430 static inline void threadgroup_lock(struct task_struct *tsk) 2431 { 2432 /* 2433 * exec uses exit for de-threading nesting group_rwsem inside 2434 * cred_guard_mutex. Grab cred_guard_mutex first. 2435 */ 2436 mutex_lock(&tsk->signal->cred_guard_mutex); 2437 down_write(&tsk->signal->group_rwsem); 2438 } 2439 2440 /** 2441 * threadgroup_unlock - unlock threadgroup 2442 * @tsk: member task of the threadgroup to unlock 2443 * 2444 * Reverse threadgroup_lock(). 2445 */ 2446 static inline void threadgroup_unlock(struct task_struct *tsk) 2447 { 2448 up_write(&tsk->signal->group_rwsem); 2449 mutex_unlock(&tsk->signal->cred_guard_mutex); 2450 } 2451 #else 2452 static inline void threadgroup_change_begin(struct task_struct *tsk) {} 2453 static inline void threadgroup_change_end(struct task_struct *tsk) {} 2454 static inline void threadgroup_lock(struct task_struct *tsk) {} 2455 static inline void threadgroup_unlock(struct task_struct *tsk) {} 2456 #endif 2457 2458 #ifndef __HAVE_THREAD_FUNCTIONS 2459 2460 #define task_thread_info(task) ((struct thread_info *)(task)->stack) 2461 #define task_stack_page(task) ((task)->stack) 2462 2463 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org) 2464 { 2465 *task_thread_info(p) = *task_thread_info(org); 2466 task_thread_info(p)->task = p; 2467 } 2468 2469 static inline unsigned long *end_of_stack(struct task_struct *p) 2470 { 2471 return (unsigned long *)(task_thread_info(p) + 1); 2472 } 2473 2474 #endif 2475 2476 static inline int object_is_on_stack(void *obj) 2477 { 2478 void *stack = task_stack_page(current); 2479 2480 return (obj >= stack) && (obj < (stack + THREAD_SIZE)); 2481 } 2482 2483 extern void thread_info_cache_init(void); 2484 2485 #ifdef CONFIG_DEBUG_STACK_USAGE 2486 static inline unsigned long stack_not_used(struct task_struct *p) 2487 { 2488 unsigned long *n = end_of_stack(p); 2489 2490 do { /* Skip over canary */ 2491 n++; 2492 } while (!*n); 2493 2494 return (unsigned long)n - (unsigned long)end_of_stack(p); 2495 } 2496 #endif 2497 2498 /* set thread flags in other task's structures 2499 * - see asm/thread_info.h for TIF_xxxx flags available 2500 */ 2501 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag) 2502 { 2503 set_ti_thread_flag(task_thread_info(tsk), flag); 2504 } 2505 2506 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2507 { 2508 clear_ti_thread_flag(task_thread_info(tsk), flag); 2509 } 2510 2511 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag) 2512 { 2513 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag); 2514 } 2515 2516 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag) 2517 { 2518 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag); 2519 } 2520 2521 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag) 2522 { 2523 return test_ti_thread_flag(task_thread_info(tsk), flag); 2524 } 2525 2526 static inline void set_tsk_need_resched(struct task_struct *tsk) 2527 { 2528 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2529 } 2530 2531 static inline void clear_tsk_need_resched(struct task_struct *tsk) 2532 { 2533 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED); 2534 } 2535 2536 static inline int test_tsk_need_resched(struct task_struct *tsk) 2537 { 2538 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); 2539 } 2540 2541 static inline int restart_syscall(void) 2542 { 2543 set_tsk_thread_flag(current, TIF_SIGPENDING); 2544 return -ERESTARTNOINTR; 2545 } 2546 2547 static inline int signal_pending(struct task_struct *p) 2548 { 2549 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING)); 2550 } 2551 2552 static inline int __fatal_signal_pending(struct task_struct *p) 2553 { 2554 return unlikely(sigismember(&p->pending.signal, SIGKILL)); 2555 } 2556 2557 static inline int fatal_signal_pending(struct task_struct *p) 2558 { 2559 return signal_pending(p) && __fatal_signal_pending(p); 2560 } 2561 2562 static inline int signal_pending_state(long state, struct task_struct *p) 2563 { 2564 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL))) 2565 return 0; 2566 if (!signal_pending(p)) 2567 return 0; 2568 2569 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p); 2570 } 2571 2572 static inline int need_resched(void) 2573 { 2574 return unlikely(test_thread_flag(TIF_NEED_RESCHED)); 2575 } 2576 2577 /* 2578 * cond_resched() and cond_resched_lock(): latency reduction via 2579 * explicit rescheduling in places that are safe. The return 2580 * value indicates whether a reschedule was done in fact. 2581 * cond_resched_lock() will drop the spinlock before scheduling, 2582 * cond_resched_softirq() will enable bhs before scheduling. 2583 */ 2584 extern int _cond_resched(void); 2585 2586 #define cond_resched() ({ \ 2587 __might_sleep(__FILE__, __LINE__, 0); \ 2588 _cond_resched(); \ 2589 }) 2590 2591 extern int __cond_resched_lock(spinlock_t *lock); 2592 2593 #ifdef CONFIG_PREEMPT_COUNT 2594 #define PREEMPT_LOCK_OFFSET PREEMPT_OFFSET 2595 #else 2596 #define PREEMPT_LOCK_OFFSET 0 2597 #endif 2598 2599 #define cond_resched_lock(lock) ({ \ 2600 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \ 2601 __cond_resched_lock(lock); \ 2602 }) 2603 2604 extern int __cond_resched_softirq(void); 2605 2606 #define cond_resched_softirq() ({ \ 2607 __might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \ 2608 __cond_resched_softirq(); \ 2609 }) 2610 2611 /* 2612 * Does a critical section need to be broken due to another 2613 * task waiting?: (technically does not depend on CONFIG_PREEMPT, 2614 * but a general need for low latency) 2615 */ 2616 static inline int spin_needbreak(spinlock_t *lock) 2617 { 2618 #ifdef CONFIG_PREEMPT 2619 return spin_is_contended(lock); 2620 #else 2621 return 0; 2622 #endif 2623 } 2624 2625 /* 2626 * Thread group CPU time accounting. 2627 */ 2628 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times); 2629 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times); 2630 2631 static inline void thread_group_cputime_init(struct signal_struct *sig) 2632 { 2633 raw_spin_lock_init(&sig->cputimer.lock); 2634 } 2635 2636 /* 2637 * Reevaluate whether the task has signals pending delivery. 2638 * Wake the task if so. 2639 * This is required every time the blocked sigset_t changes. 2640 * callers must hold sighand->siglock. 2641 */ 2642 extern void recalc_sigpending_and_wake(struct task_struct *t); 2643 extern void recalc_sigpending(void); 2644 2645 extern void signal_wake_up(struct task_struct *t, int resume_stopped); 2646 2647 /* 2648 * Wrappers for p->thread_info->cpu access. No-op on UP. 2649 */ 2650 #ifdef CONFIG_SMP 2651 2652 static inline unsigned int task_cpu(const struct task_struct *p) 2653 { 2654 return task_thread_info(p)->cpu; 2655 } 2656 2657 extern void set_task_cpu(struct task_struct *p, unsigned int cpu); 2658 2659 #else 2660 2661 static inline unsigned int task_cpu(const struct task_struct *p) 2662 { 2663 return 0; 2664 } 2665 2666 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) 2667 { 2668 } 2669 2670 #endif /* CONFIG_SMP */ 2671 2672 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); 2673 extern long sched_getaffinity(pid_t pid, struct cpumask *mask); 2674 2675 extern void normalize_rt_tasks(void); 2676 2677 #ifdef CONFIG_CGROUP_SCHED 2678 2679 extern struct task_group root_task_group; 2680 2681 extern struct task_group *sched_create_group(struct task_group *parent); 2682 extern void sched_destroy_group(struct task_group *tg); 2683 extern void sched_move_task(struct task_struct *tsk); 2684 #ifdef CONFIG_FAIR_GROUP_SCHED 2685 extern int sched_group_set_shares(struct task_group *tg, unsigned long shares); 2686 extern unsigned long sched_group_shares(struct task_group *tg); 2687 #endif 2688 #ifdef CONFIG_RT_GROUP_SCHED 2689 extern int sched_group_set_rt_runtime(struct task_group *tg, 2690 long rt_runtime_us); 2691 extern long sched_group_rt_runtime(struct task_group *tg); 2692 extern int sched_group_set_rt_period(struct task_group *tg, 2693 long rt_period_us); 2694 extern long sched_group_rt_period(struct task_group *tg); 2695 extern int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk); 2696 #endif 2697 #endif 2698 2699 extern int task_can_switch_user(struct user_struct *up, 2700 struct task_struct *tsk); 2701 2702 #ifdef CONFIG_TASK_XACCT 2703 static inline void add_rchar(struct task_struct *tsk, ssize_t amt) 2704 { 2705 tsk->ioac.rchar += amt; 2706 } 2707 2708 static inline void add_wchar(struct task_struct *tsk, ssize_t amt) 2709 { 2710 tsk->ioac.wchar += amt; 2711 } 2712 2713 static inline void inc_syscr(struct task_struct *tsk) 2714 { 2715 tsk->ioac.syscr++; 2716 } 2717 2718 static inline void inc_syscw(struct task_struct *tsk) 2719 { 2720 tsk->ioac.syscw++; 2721 } 2722 #else 2723 static inline void add_rchar(struct task_struct *tsk, ssize_t amt) 2724 { 2725 } 2726 2727 static inline void add_wchar(struct task_struct *tsk, ssize_t amt) 2728 { 2729 } 2730 2731 static inline void inc_syscr(struct task_struct *tsk) 2732 { 2733 } 2734 2735 static inline void inc_syscw(struct task_struct *tsk) 2736 { 2737 } 2738 #endif 2739 2740 #ifndef TASK_SIZE_OF 2741 #define TASK_SIZE_OF(tsk) TASK_SIZE 2742 #endif 2743 2744 #ifdef CONFIG_MM_OWNER 2745 extern void mm_update_next_owner(struct mm_struct *mm); 2746 extern void mm_init_owner(struct mm_struct *mm, struct task_struct *p); 2747 #else 2748 static inline void mm_update_next_owner(struct mm_struct *mm) 2749 { 2750 } 2751 2752 static inline void mm_init_owner(struct mm_struct *mm, struct task_struct *p) 2753 { 2754 } 2755 #endif /* CONFIG_MM_OWNER */ 2756 2757 static inline unsigned long task_rlimit(const struct task_struct *tsk, 2758 unsigned int limit) 2759 { 2760 return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_cur); 2761 } 2762 2763 static inline unsigned long task_rlimit_max(const struct task_struct *tsk, 2764 unsigned int limit) 2765 { 2766 return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_max); 2767 } 2768 2769 static inline unsigned long rlimit(unsigned int limit) 2770 { 2771 return task_rlimit(current, limit); 2772 } 2773 2774 static inline unsigned long rlimit_max(unsigned int limit) 2775 { 2776 return task_rlimit_max(current, limit); 2777 } 2778 2779 #endif /* __KERNEL__ */ 2780 2781 #endif 2782