1 /* 2 * linux/kernel/exit.c 3 * 4 * Copyright (C) 1991, 1992 Linus Torvalds 5 */ 6 7 #include <linux/mm.h> 8 #include <linux/slab.h> 9 #include <linux/interrupt.h> 10 #include <linux/smp_lock.h> 11 #include <linux/module.h> 12 #include <linux/capability.h> 13 #include <linux/completion.h> 14 #include <linux/personality.h> 15 #include <linux/tty.h> 16 #include <linux/namespace.h> 17 #include <linux/key.h> 18 #include <linux/security.h> 19 #include <linux/cpu.h> 20 #include <linux/acct.h> 21 #include <linux/tsacct_kern.h> 22 #include <linux/file.h> 23 #include <linux/binfmts.h> 24 #include <linux/nsproxy.h> 25 #include <linux/ptrace.h> 26 #include <linux/profile.h> 27 #include <linux/mount.h> 28 #include <linux/proc_fs.h> 29 #include <linux/mempolicy.h> 30 #include <linux/taskstats_kern.h> 31 #include <linux/delayacct.h> 32 #include <linux/cpuset.h> 33 #include <linux/syscalls.h> 34 #include <linux/signal.h> 35 #include <linux/posix-timers.h> 36 #include <linux/cn_proc.h> 37 #include <linux/mutex.h> 38 #include <linux/futex.h> 39 #include <linux/compat.h> 40 #include <linux/pipe_fs_i.h> 41 #include <linux/audit.h> /* for audit_free() */ 42 #include <linux/resource.h> 43 #include <linux/blkdev.h> 44 45 #include <asm/uaccess.h> 46 #include <asm/unistd.h> 47 #include <asm/pgtable.h> 48 #include <asm/mmu_context.h> 49 50 extern void sem_exit (void); 51 extern struct task_struct *child_reaper; 52 53 static void exit_mm(struct task_struct * tsk); 54 55 static void __unhash_process(struct task_struct *p) 56 { 57 nr_threads--; 58 detach_pid(p, PIDTYPE_PID); 59 if (thread_group_leader(p)) { 60 detach_pid(p, PIDTYPE_PGID); 61 detach_pid(p, PIDTYPE_SID); 62 63 list_del_rcu(&p->tasks); 64 __get_cpu_var(process_counts)--; 65 } 66 list_del_rcu(&p->thread_group); 67 remove_parent(p); 68 } 69 70 /* 71 * This function expects the tasklist_lock write-locked. 72 */ 73 static void __exit_signal(struct task_struct *tsk) 74 { 75 struct signal_struct *sig = tsk->signal; 76 struct sighand_struct *sighand; 77 78 BUG_ON(!sig); 79 BUG_ON(!atomic_read(&sig->count)); 80 81 rcu_read_lock(); 82 sighand = rcu_dereference(tsk->sighand); 83 spin_lock(&sighand->siglock); 84 85 posix_cpu_timers_exit(tsk); 86 if (atomic_dec_and_test(&sig->count)) 87 posix_cpu_timers_exit_group(tsk); 88 else { 89 /* 90 * If there is any task waiting for the group exit 91 * then notify it: 92 */ 93 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) { 94 wake_up_process(sig->group_exit_task); 95 sig->group_exit_task = NULL; 96 } 97 if (tsk == sig->curr_target) 98 sig->curr_target = next_thread(tsk); 99 /* 100 * Accumulate here the counters for all threads but the 101 * group leader as they die, so they can be added into 102 * the process-wide totals when those are taken. 103 * The group leader stays around as a zombie as long 104 * as there are other threads. When it gets reaped, 105 * the exit.c code will add its counts into these totals. 106 * We won't ever get here for the group leader, since it 107 * will have been the last reference on the signal_struct. 108 */ 109 sig->utime = cputime_add(sig->utime, tsk->utime); 110 sig->stime = cputime_add(sig->stime, tsk->stime); 111 sig->min_flt += tsk->min_flt; 112 sig->maj_flt += tsk->maj_flt; 113 sig->nvcsw += tsk->nvcsw; 114 sig->nivcsw += tsk->nivcsw; 115 sig->sched_time += tsk->sched_time; 116 sig = NULL; /* Marker for below. */ 117 } 118 119 __unhash_process(tsk); 120 121 tsk->signal = NULL; 122 tsk->sighand = NULL; 123 spin_unlock(&sighand->siglock); 124 rcu_read_unlock(); 125 126 __cleanup_sighand(sighand); 127 clear_tsk_thread_flag(tsk,TIF_SIGPENDING); 128 flush_sigqueue(&tsk->pending); 129 if (sig) { 130 flush_sigqueue(&sig->shared_pending); 131 taskstats_tgid_free(sig); 132 __cleanup_signal(sig); 133 } 134 } 135 136 static void delayed_put_task_struct(struct rcu_head *rhp) 137 { 138 put_task_struct(container_of(rhp, struct task_struct, rcu)); 139 } 140 141 void release_task(struct task_struct * p) 142 { 143 struct task_struct *leader; 144 int zap_leader; 145 repeat: 146 atomic_dec(&p->user->processes); 147 write_lock_irq(&tasklist_lock); 148 ptrace_unlink(p); 149 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children)); 150 __exit_signal(p); 151 152 /* 153 * If we are the last non-leader member of the thread 154 * group, and the leader is zombie, then notify the 155 * group leader's parent process. (if it wants notification.) 156 */ 157 zap_leader = 0; 158 leader = p->group_leader; 159 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { 160 BUG_ON(leader->exit_signal == -1); 161 do_notify_parent(leader, leader->exit_signal); 162 /* 163 * If we were the last child thread and the leader has 164 * exited already, and the leader's parent ignores SIGCHLD, 165 * then we are the one who should release the leader. 166 * 167 * do_notify_parent() will have marked it self-reaping in 168 * that case. 169 */ 170 zap_leader = (leader->exit_signal == -1); 171 } 172 173 sched_exit(p); 174 write_unlock_irq(&tasklist_lock); 175 proc_flush_task(p); 176 release_thread(p); 177 call_rcu(&p->rcu, delayed_put_task_struct); 178 179 p = leader; 180 if (unlikely(zap_leader)) 181 goto repeat; 182 } 183 184 /* 185 * This checks not only the pgrp, but falls back on the pid if no 186 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly 187 * without this... 188 */ 189 int session_of_pgrp(int pgrp) 190 { 191 struct task_struct *p; 192 int sid = -1; 193 194 read_lock(&tasklist_lock); 195 do_each_task_pid(pgrp, PIDTYPE_PGID, p) { 196 if (p->signal->session > 0) { 197 sid = p->signal->session; 198 goto out; 199 } 200 } while_each_task_pid(pgrp, PIDTYPE_PGID, p); 201 p = find_task_by_pid(pgrp); 202 if (p) 203 sid = p->signal->session; 204 out: 205 read_unlock(&tasklist_lock); 206 207 return sid; 208 } 209 210 /* 211 * Determine if a process group is "orphaned", according to the POSIX 212 * definition in 2.2.2.52. Orphaned process groups are not to be affected 213 * by terminal-generated stop signals. Newly orphaned process groups are 214 * to receive a SIGHUP and a SIGCONT. 215 * 216 * "I ask you, have you ever known what it is to be an orphan?" 217 */ 218 static int will_become_orphaned_pgrp(int pgrp, struct task_struct *ignored_task) 219 { 220 struct task_struct *p; 221 int ret = 1; 222 223 do_each_task_pid(pgrp, PIDTYPE_PGID, p) { 224 if (p == ignored_task 225 || p->exit_state 226 || is_init(p->real_parent)) 227 continue; 228 if (process_group(p->real_parent) != pgrp 229 && p->real_parent->signal->session == p->signal->session) { 230 ret = 0; 231 break; 232 } 233 } while_each_task_pid(pgrp, PIDTYPE_PGID, p); 234 return ret; /* (sighing) "Often!" */ 235 } 236 237 int is_orphaned_pgrp(int pgrp) 238 { 239 int retval; 240 241 read_lock(&tasklist_lock); 242 retval = will_become_orphaned_pgrp(pgrp, NULL); 243 read_unlock(&tasklist_lock); 244 245 return retval; 246 } 247 248 static int has_stopped_jobs(int pgrp) 249 { 250 int retval = 0; 251 struct task_struct *p; 252 253 do_each_task_pid(pgrp, PIDTYPE_PGID, p) { 254 if (p->state != TASK_STOPPED) 255 continue; 256 retval = 1; 257 break; 258 } while_each_task_pid(pgrp, PIDTYPE_PGID, p); 259 return retval; 260 } 261 262 /** 263 * reparent_to_init - Reparent the calling kernel thread to the init task. 264 * 265 * If a kernel thread is launched as a result of a system call, or if 266 * it ever exits, it should generally reparent itself to init so that 267 * it is correctly cleaned up on exit. 268 * 269 * The various task state such as scheduling policy and priority may have 270 * been inherited from a user process, so we reset them to sane values here. 271 * 272 * NOTE that reparent_to_init() gives the caller full capabilities. 273 */ 274 static void reparent_to_init(void) 275 { 276 write_lock_irq(&tasklist_lock); 277 278 ptrace_unlink(current); 279 /* Reparent to init */ 280 remove_parent(current); 281 current->parent = child_reaper; 282 current->real_parent = child_reaper; 283 add_parent(current); 284 285 /* Set the exit signal to SIGCHLD so we signal init on exit */ 286 current->exit_signal = SIGCHLD; 287 288 if (!has_rt_policy(current) && (task_nice(current) < 0)) 289 set_user_nice(current, 0); 290 /* cpus_allowed? */ 291 /* rt_priority? */ 292 /* signals? */ 293 security_task_reparent_to_init(current); 294 memcpy(current->signal->rlim, init_task.signal->rlim, 295 sizeof(current->signal->rlim)); 296 atomic_inc(&(INIT_USER->__count)); 297 write_unlock_irq(&tasklist_lock); 298 switch_uid(INIT_USER); 299 } 300 301 void __set_special_pids(pid_t session, pid_t pgrp) 302 { 303 struct task_struct *curr = current->group_leader; 304 305 if (curr->signal->session != session) { 306 detach_pid(curr, PIDTYPE_SID); 307 curr->signal->session = session; 308 attach_pid(curr, PIDTYPE_SID, session); 309 } 310 if (process_group(curr) != pgrp) { 311 detach_pid(curr, PIDTYPE_PGID); 312 curr->signal->pgrp = pgrp; 313 attach_pid(curr, PIDTYPE_PGID, pgrp); 314 } 315 } 316 317 void set_special_pids(pid_t session, pid_t pgrp) 318 { 319 write_lock_irq(&tasklist_lock); 320 __set_special_pids(session, pgrp); 321 write_unlock_irq(&tasklist_lock); 322 } 323 324 /* 325 * Let kernel threads use this to say that they 326 * allow a certain signal (since daemonize() will 327 * have disabled all of them by default). 328 */ 329 int allow_signal(int sig) 330 { 331 if (!valid_signal(sig) || sig < 1) 332 return -EINVAL; 333 334 spin_lock_irq(¤t->sighand->siglock); 335 sigdelset(¤t->blocked, sig); 336 if (!current->mm) { 337 /* Kernel threads handle their own signals. 338 Let the signal code know it'll be handled, so 339 that they don't get converted to SIGKILL or 340 just silently dropped */ 341 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; 342 } 343 recalc_sigpending(); 344 spin_unlock_irq(¤t->sighand->siglock); 345 return 0; 346 } 347 348 EXPORT_SYMBOL(allow_signal); 349 350 int disallow_signal(int sig) 351 { 352 if (!valid_signal(sig) || sig < 1) 353 return -EINVAL; 354 355 spin_lock_irq(¤t->sighand->siglock); 356 sigaddset(¤t->blocked, sig); 357 recalc_sigpending(); 358 spin_unlock_irq(¤t->sighand->siglock); 359 return 0; 360 } 361 362 EXPORT_SYMBOL(disallow_signal); 363 364 /* 365 * Put all the gunge required to become a kernel thread without 366 * attached user resources in one place where it belongs. 367 */ 368 369 void daemonize(const char *name, ...) 370 { 371 va_list args; 372 struct fs_struct *fs; 373 sigset_t blocked; 374 375 va_start(args, name); 376 vsnprintf(current->comm, sizeof(current->comm), name, args); 377 va_end(args); 378 379 /* 380 * If we were started as result of loading a module, close all of the 381 * user space pages. We don't need them, and if we didn't close them 382 * they would be locked into memory. 383 */ 384 exit_mm(current); 385 386 set_special_pids(1, 1); 387 mutex_lock(&tty_mutex); 388 current->signal->tty = NULL; 389 mutex_unlock(&tty_mutex); 390 391 /* Block and flush all signals */ 392 sigfillset(&blocked); 393 sigprocmask(SIG_BLOCK, &blocked, NULL); 394 flush_signals(current); 395 396 /* Become as one with the init task */ 397 398 exit_fs(current); /* current->fs->count--; */ 399 fs = init_task.fs; 400 current->fs = fs; 401 atomic_inc(&fs->count); 402 403 exit_task_namespaces(current); 404 current->nsproxy = init_task.nsproxy; 405 get_task_namespaces(current); 406 407 exit_files(current); 408 current->files = init_task.files; 409 atomic_inc(¤t->files->count); 410 411 reparent_to_init(); 412 } 413 414 EXPORT_SYMBOL(daemonize); 415 416 static void close_files(struct files_struct * files) 417 { 418 int i, j; 419 struct fdtable *fdt; 420 421 j = 0; 422 423 /* 424 * It is safe to dereference the fd table without RCU or 425 * ->file_lock because this is the last reference to the 426 * files structure. 427 */ 428 fdt = files_fdtable(files); 429 for (;;) { 430 unsigned long set; 431 i = j * __NFDBITS; 432 if (i >= fdt->max_fdset || i >= fdt->max_fds) 433 break; 434 set = fdt->open_fds->fds_bits[j++]; 435 while (set) { 436 if (set & 1) { 437 struct file * file = xchg(&fdt->fd[i], NULL); 438 if (file) 439 filp_close(file, files); 440 } 441 i++; 442 set >>= 1; 443 } 444 } 445 } 446 447 struct files_struct *get_files_struct(struct task_struct *task) 448 { 449 struct files_struct *files; 450 451 task_lock(task); 452 files = task->files; 453 if (files) 454 atomic_inc(&files->count); 455 task_unlock(task); 456 457 return files; 458 } 459 460 void fastcall put_files_struct(struct files_struct *files) 461 { 462 struct fdtable *fdt; 463 464 if (atomic_dec_and_test(&files->count)) { 465 close_files(files); 466 /* 467 * Free the fd and fdset arrays if we expanded them. 468 * If the fdtable was embedded, pass files for freeing 469 * at the end of the RCU grace period. Otherwise, 470 * you can free files immediately. 471 */ 472 fdt = files_fdtable(files); 473 if (fdt == &files->fdtab) 474 fdt->free_files = files; 475 else 476 kmem_cache_free(files_cachep, files); 477 free_fdtable(fdt); 478 } 479 } 480 481 EXPORT_SYMBOL(put_files_struct); 482 483 void reset_files_struct(struct task_struct *tsk, struct files_struct *files) 484 { 485 struct files_struct *old; 486 487 old = tsk->files; 488 task_lock(tsk); 489 tsk->files = files; 490 task_unlock(tsk); 491 put_files_struct(old); 492 } 493 EXPORT_SYMBOL(reset_files_struct); 494 495 static inline void __exit_files(struct task_struct *tsk) 496 { 497 struct files_struct * files = tsk->files; 498 499 if (files) { 500 task_lock(tsk); 501 tsk->files = NULL; 502 task_unlock(tsk); 503 put_files_struct(files); 504 } 505 } 506 507 void exit_files(struct task_struct *tsk) 508 { 509 __exit_files(tsk); 510 } 511 512 static inline void __put_fs_struct(struct fs_struct *fs) 513 { 514 /* No need to hold fs->lock if we are killing it */ 515 if (atomic_dec_and_test(&fs->count)) { 516 dput(fs->root); 517 mntput(fs->rootmnt); 518 dput(fs->pwd); 519 mntput(fs->pwdmnt); 520 if (fs->altroot) { 521 dput(fs->altroot); 522 mntput(fs->altrootmnt); 523 } 524 kmem_cache_free(fs_cachep, fs); 525 } 526 } 527 528 void put_fs_struct(struct fs_struct *fs) 529 { 530 __put_fs_struct(fs); 531 } 532 533 static inline void __exit_fs(struct task_struct *tsk) 534 { 535 struct fs_struct * fs = tsk->fs; 536 537 if (fs) { 538 task_lock(tsk); 539 tsk->fs = NULL; 540 task_unlock(tsk); 541 __put_fs_struct(fs); 542 } 543 } 544 545 void exit_fs(struct task_struct *tsk) 546 { 547 __exit_fs(tsk); 548 } 549 550 EXPORT_SYMBOL_GPL(exit_fs); 551 552 /* 553 * Turn us into a lazy TLB process if we 554 * aren't already.. 555 */ 556 static void exit_mm(struct task_struct * tsk) 557 { 558 struct mm_struct *mm = tsk->mm; 559 560 mm_release(tsk, mm); 561 if (!mm) 562 return; 563 /* 564 * Serialize with any possible pending coredump. 565 * We must hold mmap_sem around checking core_waiters 566 * and clearing tsk->mm. The core-inducing thread 567 * will increment core_waiters for each thread in the 568 * group with ->mm != NULL. 569 */ 570 down_read(&mm->mmap_sem); 571 if (mm->core_waiters) { 572 up_read(&mm->mmap_sem); 573 down_write(&mm->mmap_sem); 574 if (!--mm->core_waiters) 575 complete(mm->core_startup_done); 576 up_write(&mm->mmap_sem); 577 578 wait_for_completion(&mm->core_done); 579 down_read(&mm->mmap_sem); 580 } 581 atomic_inc(&mm->mm_count); 582 BUG_ON(mm != tsk->active_mm); 583 /* more a memory barrier than a real lock */ 584 task_lock(tsk); 585 tsk->mm = NULL; 586 up_read(&mm->mmap_sem); 587 enter_lazy_tlb(mm, current); 588 task_unlock(tsk); 589 mmput(mm); 590 } 591 592 static inline void 593 choose_new_parent(struct task_struct *p, struct task_struct *reaper) 594 { 595 /* 596 * Make sure we're not reparenting to ourselves and that 597 * the parent is not a zombie. 598 */ 599 BUG_ON(p == reaper || reaper->exit_state); 600 p->real_parent = reaper; 601 } 602 603 static void 604 reparent_thread(struct task_struct *p, struct task_struct *father, int traced) 605 { 606 /* We don't want people slaying init. */ 607 if (p->exit_signal != -1) 608 p->exit_signal = SIGCHLD; 609 610 if (p->pdeath_signal) 611 /* We already hold the tasklist_lock here. */ 612 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); 613 614 /* Move the child from its dying parent to the new one. */ 615 if (unlikely(traced)) { 616 /* Preserve ptrace links if someone else is tracing this child. */ 617 list_del_init(&p->ptrace_list); 618 if (p->parent != p->real_parent) 619 list_add(&p->ptrace_list, &p->real_parent->ptrace_children); 620 } else { 621 /* If this child is being traced, then we're the one tracing it 622 * anyway, so let go of it. 623 */ 624 p->ptrace = 0; 625 remove_parent(p); 626 p->parent = p->real_parent; 627 add_parent(p); 628 629 /* If we'd notified the old parent about this child's death, 630 * also notify the new parent. 631 */ 632 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 && 633 thread_group_empty(p)) 634 do_notify_parent(p, p->exit_signal); 635 else if (p->state == TASK_TRACED) { 636 /* 637 * If it was at a trace stop, turn it into 638 * a normal stop since it's no longer being 639 * traced. 640 */ 641 ptrace_untrace(p); 642 } 643 } 644 645 /* 646 * process group orphan check 647 * Case ii: Our child is in a different pgrp 648 * than we are, and it was the only connection 649 * outside, so the child pgrp is now orphaned. 650 */ 651 if ((process_group(p) != process_group(father)) && 652 (p->signal->session == father->signal->session)) { 653 int pgrp = process_group(p); 654 655 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) { 656 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp); 657 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp); 658 } 659 } 660 } 661 662 /* 663 * When we die, we re-parent all our children. 664 * Try to give them to another thread in our thread 665 * group, and if no such member exists, give it to 666 * the global child reaper process (ie "init") 667 */ 668 static void 669 forget_original_parent(struct task_struct *father, struct list_head *to_release) 670 { 671 struct task_struct *p, *reaper = father; 672 struct list_head *_p, *_n; 673 674 do { 675 reaper = next_thread(reaper); 676 if (reaper == father) { 677 reaper = child_reaper; 678 break; 679 } 680 } while (reaper->exit_state); 681 682 /* 683 * There are only two places where our children can be: 684 * 685 * - in our child list 686 * - in our ptraced child list 687 * 688 * Search them and reparent children. 689 */ 690 list_for_each_safe(_p, _n, &father->children) { 691 int ptrace; 692 p = list_entry(_p, struct task_struct, sibling); 693 694 ptrace = p->ptrace; 695 696 /* if father isn't the real parent, then ptrace must be enabled */ 697 BUG_ON(father != p->real_parent && !ptrace); 698 699 if (father == p->real_parent) { 700 /* reparent with a reaper, real father it's us */ 701 choose_new_parent(p, reaper); 702 reparent_thread(p, father, 0); 703 } else { 704 /* reparent ptraced task to its real parent */ 705 __ptrace_unlink (p); 706 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 && 707 thread_group_empty(p)) 708 do_notify_parent(p, p->exit_signal); 709 } 710 711 /* 712 * if the ptraced child is a zombie with exit_signal == -1 713 * we must collect it before we exit, or it will remain 714 * zombie forever since we prevented it from self-reap itself 715 * while it was being traced by us, to be able to see it in wait4. 716 */ 717 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1)) 718 list_add(&p->ptrace_list, to_release); 719 } 720 list_for_each_safe(_p, _n, &father->ptrace_children) { 721 p = list_entry(_p, struct task_struct, ptrace_list); 722 choose_new_parent(p, reaper); 723 reparent_thread(p, father, 1); 724 } 725 } 726 727 /* 728 * Send signals to all our closest relatives so that they know 729 * to properly mourn us.. 730 */ 731 static void exit_notify(struct task_struct *tsk) 732 { 733 int state; 734 struct task_struct *t; 735 struct list_head ptrace_dead, *_p, *_n; 736 737 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT) 738 && !thread_group_empty(tsk)) { 739 /* 740 * This occurs when there was a race between our exit 741 * syscall and a group signal choosing us as the one to 742 * wake up. It could be that we are the only thread 743 * alerted to check for pending signals, but another thread 744 * should be woken now to take the signal since we will not. 745 * Now we'll wake all the threads in the group just to make 746 * sure someone gets all the pending signals. 747 */ 748 read_lock(&tasklist_lock); 749 spin_lock_irq(&tsk->sighand->siglock); 750 for (t = next_thread(tsk); t != tsk; t = next_thread(t)) 751 if (!signal_pending(t) && !(t->flags & PF_EXITING)) { 752 recalc_sigpending_tsk(t); 753 if (signal_pending(t)) 754 signal_wake_up(t, 0); 755 } 756 spin_unlock_irq(&tsk->sighand->siglock); 757 read_unlock(&tasklist_lock); 758 } 759 760 write_lock_irq(&tasklist_lock); 761 762 /* 763 * This does two things: 764 * 765 * A. Make init inherit all the child processes 766 * B. Check to see if any process groups have become orphaned 767 * as a result of our exiting, and if they have any stopped 768 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) 769 */ 770 771 INIT_LIST_HEAD(&ptrace_dead); 772 forget_original_parent(tsk, &ptrace_dead); 773 BUG_ON(!list_empty(&tsk->children)); 774 BUG_ON(!list_empty(&tsk->ptrace_children)); 775 776 /* 777 * Check to see if any process groups have become orphaned 778 * as a result of our exiting, and if they have any stopped 779 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) 780 * 781 * Case i: Our father is in a different pgrp than we are 782 * and we were the only connection outside, so our pgrp 783 * is about to become orphaned. 784 */ 785 786 t = tsk->real_parent; 787 788 if ((process_group(t) != process_group(tsk)) && 789 (t->signal->session == tsk->signal->session) && 790 will_become_orphaned_pgrp(process_group(tsk), tsk) && 791 has_stopped_jobs(process_group(tsk))) { 792 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk)); 793 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk)); 794 } 795 796 /* Let father know we died 797 * 798 * Thread signals are configurable, but you aren't going to use 799 * that to send signals to arbitary processes. 800 * That stops right now. 801 * 802 * If the parent exec id doesn't match the exec id we saved 803 * when we started then we know the parent has changed security 804 * domain. 805 * 806 * If our self_exec id doesn't match our parent_exec_id then 807 * we have changed execution domain as these two values started 808 * the same after a fork. 809 * 810 */ 811 812 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 && 813 ( tsk->parent_exec_id != t->self_exec_id || 814 tsk->self_exec_id != tsk->parent_exec_id) 815 && !capable(CAP_KILL)) 816 tsk->exit_signal = SIGCHLD; 817 818 819 /* If something other than our normal parent is ptracing us, then 820 * send it a SIGCHLD instead of honoring exit_signal. exit_signal 821 * only has special meaning to our real parent. 822 */ 823 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) { 824 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD; 825 do_notify_parent(tsk, signal); 826 } else if (tsk->ptrace) { 827 do_notify_parent(tsk, SIGCHLD); 828 } 829 830 state = EXIT_ZOMBIE; 831 if (tsk->exit_signal == -1 && 832 (likely(tsk->ptrace == 0) || 833 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT))) 834 state = EXIT_DEAD; 835 tsk->exit_state = state; 836 837 write_unlock_irq(&tasklist_lock); 838 839 list_for_each_safe(_p, _n, &ptrace_dead) { 840 list_del_init(_p); 841 t = list_entry(_p, struct task_struct, ptrace_list); 842 release_task(t); 843 } 844 845 /* If the process is dead, release it - nobody will wait for it */ 846 if (state == EXIT_DEAD) 847 release_task(tsk); 848 } 849 850 fastcall NORET_TYPE void do_exit(long code) 851 { 852 struct task_struct *tsk = current; 853 struct taskstats *tidstats; 854 int group_dead; 855 unsigned int mycpu; 856 857 profile_task_exit(tsk); 858 859 WARN_ON(atomic_read(&tsk->fs_excl)); 860 861 if (unlikely(in_interrupt())) 862 panic("Aiee, killing interrupt handler!"); 863 if (unlikely(!tsk->pid)) 864 panic("Attempted to kill the idle task!"); 865 if (unlikely(tsk == child_reaper)) 866 panic("Attempted to kill init!"); 867 868 if (unlikely(current->ptrace & PT_TRACE_EXIT)) { 869 current->ptrace_message = code; 870 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP); 871 } 872 873 /* 874 * We're taking recursive faults here in do_exit. Safest is to just 875 * leave this task alone and wait for reboot. 876 */ 877 if (unlikely(tsk->flags & PF_EXITING)) { 878 printk(KERN_ALERT 879 "Fixing recursive fault but reboot is needed!\n"); 880 if (tsk->io_context) 881 exit_io_context(); 882 set_current_state(TASK_UNINTERRUPTIBLE); 883 schedule(); 884 } 885 886 tsk->flags |= PF_EXITING; 887 888 if (unlikely(in_atomic())) 889 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", 890 current->comm, current->pid, 891 preempt_count()); 892 893 taskstats_exit_alloc(&tidstats, &mycpu); 894 895 acct_update_integrals(tsk); 896 if (tsk->mm) { 897 update_hiwater_rss(tsk->mm); 898 update_hiwater_vm(tsk->mm); 899 } 900 group_dead = atomic_dec_and_test(&tsk->signal->live); 901 if (group_dead) { 902 hrtimer_cancel(&tsk->signal->real_timer); 903 exit_itimers(tsk->signal); 904 } 905 acct_collect(code, group_dead); 906 if (unlikely(tsk->robust_list)) 907 exit_robust_list(tsk); 908 #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT) 909 if (unlikely(tsk->compat_robust_list)) 910 compat_exit_robust_list(tsk); 911 #endif 912 if (unlikely(tsk->audit_context)) 913 audit_free(tsk); 914 taskstats_exit_send(tsk, tidstats, group_dead, mycpu); 915 taskstats_exit_free(tidstats); 916 917 exit_mm(tsk); 918 919 if (group_dead) 920 acct_process(); 921 exit_sem(tsk); 922 __exit_files(tsk); 923 __exit_fs(tsk); 924 exit_thread(); 925 cpuset_exit(tsk); 926 exit_keys(tsk); 927 928 if (group_dead && tsk->signal->leader) 929 disassociate_ctty(1); 930 931 module_put(task_thread_info(tsk)->exec_domain->module); 932 if (tsk->binfmt) 933 module_put(tsk->binfmt->module); 934 935 tsk->exit_code = code; 936 proc_exit_connector(tsk); 937 exit_notify(tsk); 938 exit_task_namespaces(tsk); 939 #ifdef CONFIG_NUMA 940 mpol_free(tsk->mempolicy); 941 tsk->mempolicy = NULL; 942 #endif 943 /* 944 * This must happen late, after the PID is not 945 * hashed anymore: 946 */ 947 if (unlikely(!list_empty(&tsk->pi_state_list))) 948 exit_pi_state_list(tsk); 949 if (unlikely(current->pi_state_cache)) 950 kfree(current->pi_state_cache); 951 /* 952 * Make sure we are holding no locks: 953 */ 954 debug_check_no_locks_held(tsk); 955 956 if (tsk->io_context) 957 exit_io_context(); 958 959 if (tsk->splice_pipe) 960 __free_pipe_info(tsk->splice_pipe); 961 962 preempt_disable(); 963 /* causes final put_task_struct in finish_task_switch(). */ 964 tsk->state = TASK_DEAD; 965 966 schedule(); 967 BUG(); 968 /* Avoid "noreturn function does return". */ 969 for (;;) 970 cpu_relax(); /* For when BUG is null */ 971 } 972 973 EXPORT_SYMBOL_GPL(do_exit); 974 975 NORET_TYPE void complete_and_exit(struct completion *comp, long code) 976 { 977 if (comp) 978 complete(comp); 979 980 do_exit(code); 981 } 982 983 EXPORT_SYMBOL(complete_and_exit); 984 985 asmlinkage long sys_exit(int error_code) 986 { 987 do_exit((error_code&0xff)<<8); 988 } 989 990 /* 991 * Take down every thread in the group. This is called by fatal signals 992 * as well as by sys_exit_group (below). 993 */ 994 NORET_TYPE void 995 do_group_exit(int exit_code) 996 { 997 BUG_ON(exit_code & 0x80); /* core dumps don't get here */ 998 999 if (current->signal->flags & SIGNAL_GROUP_EXIT) 1000 exit_code = current->signal->group_exit_code; 1001 else if (!thread_group_empty(current)) { 1002 struct signal_struct *const sig = current->signal; 1003 struct sighand_struct *const sighand = current->sighand; 1004 spin_lock_irq(&sighand->siglock); 1005 if (sig->flags & SIGNAL_GROUP_EXIT) 1006 /* Another thread got here before we took the lock. */ 1007 exit_code = sig->group_exit_code; 1008 else { 1009 sig->group_exit_code = exit_code; 1010 zap_other_threads(current); 1011 } 1012 spin_unlock_irq(&sighand->siglock); 1013 } 1014 1015 do_exit(exit_code); 1016 /* NOTREACHED */ 1017 } 1018 1019 /* 1020 * this kills every thread in the thread group. Note that any externally 1021 * wait4()-ing process will get the correct exit code - even if this 1022 * thread is not the thread group leader. 1023 */ 1024 asmlinkage void sys_exit_group(int error_code) 1025 { 1026 do_group_exit((error_code & 0xff) << 8); 1027 } 1028 1029 static int eligible_child(pid_t pid, int options, struct task_struct *p) 1030 { 1031 if (pid > 0) { 1032 if (p->pid != pid) 1033 return 0; 1034 } else if (!pid) { 1035 if (process_group(p) != process_group(current)) 1036 return 0; 1037 } else if (pid != -1) { 1038 if (process_group(p) != -pid) 1039 return 0; 1040 } 1041 1042 /* 1043 * Do not consider detached threads that are 1044 * not ptraced: 1045 */ 1046 if (p->exit_signal == -1 && !p->ptrace) 1047 return 0; 1048 1049 /* Wait for all children (clone and not) if __WALL is set; 1050 * otherwise, wait for clone children *only* if __WCLONE is 1051 * set; otherwise, wait for non-clone children *only*. (Note: 1052 * A "clone" child here is one that reports to its parent 1053 * using a signal other than SIGCHLD.) */ 1054 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0)) 1055 && !(options & __WALL)) 1056 return 0; 1057 /* 1058 * Do not consider thread group leaders that are 1059 * in a non-empty thread group: 1060 */ 1061 if (delay_group_leader(p)) 1062 return 2; 1063 1064 if (security_task_wait(p)) 1065 return 0; 1066 1067 return 1; 1068 } 1069 1070 static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid, 1071 int why, int status, 1072 struct siginfo __user *infop, 1073 struct rusage __user *rusagep) 1074 { 1075 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0; 1076 1077 put_task_struct(p); 1078 if (!retval) 1079 retval = put_user(SIGCHLD, &infop->si_signo); 1080 if (!retval) 1081 retval = put_user(0, &infop->si_errno); 1082 if (!retval) 1083 retval = put_user((short)why, &infop->si_code); 1084 if (!retval) 1085 retval = put_user(pid, &infop->si_pid); 1086 if (!retval) 1087 retval = put_user(uid, &infop->si_uid); 1088 if (!retval) 1089 retval = put_user(status, &infop->si_status); 1090 if (!retval) 1091 retval = pid; 1092 return retval; 1093 } 1094 1095 /* 1096 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold 1097 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold 1098 * the lock and this task is uninteresting. If we return nonzero, we have 1099 * released the lock and the system call should return. 1100 */ 1101 static int wait_task_zombie(struct task_struct *p, int noreap, 1102 struct siginfo __user *infop, 1103 int __user *stat_addr, struct rusage __user *ru) 1104 { 1105 unsigned long state; 1106 int retval; 1107 int status; 1108 1109 if (unlikely(noreap)) { 1110 pid_t pid = p->pid; 1111 uid_t uid = p->uid; 1112 int exit_code = p->exit_code; 1113 int why, status; 1114 1115 if (unlikely(p->exit_state != EXIT_ZOMBIE)) 1116 return 0; 1117 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) 1118 return 0; 1119 get_task_struct(p); 1120 read_unlock(&tasklist_lock); 1121 if ((exit_code & 0x7f) == 0) { 1122 why = CLD_EXITED; 1123 status = exit_code >> 8; 1124 } else { 1125 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; 1126 status = exit_code & 0x7f; 1127 } 1128 return wait_noreap_copyout(p, pid, uid, why, 1129 status, infop, ru); 1130 } 1131 1132 /* 1133 * Try to move the task's state to DEAD 1134 * only one thread is allowed to do this: 1135 */ 1136 state = xchg(&p->exit_state, EXIT_DEAD); 1137 if (state != EXIT_ZOMBIE) { 1138 BUG_ON(state != EXIT_DEAD); 1139 return 0; 1140 } 1141 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) { 1142 /* 1143 * This can only happen in a race with a ptraced thread 1144 * dying on another processor. 1145 */ 1146 return 0; 1147 } 1148 1149 if (likely(p->real_parent == p->parent) && likely(p->signal)) { 1150 struct signal_struct *psig; 1151 struct signal_struct *sig; 1152 1153 /* 1154 * The resource counters for the group leader are in its 1155 * own task_struct. Those for dead threads in the group 1156 * are in its signal_struct, as are those for the child 1157 * processes it has previously reaped. All these 1158 * accumulate in the parent's signal_struct c* fields. 1159 * 1160 * We don't bother to take a lock here to protect these 1161 * p->signal fields, because they are only touched by 1162 * __exit_signal, which runs with tasklist_lock 1163 * write-locked anyway, and so is excluded here. We do 1164 * need to protect the access to p->parent->signal fields, 1165 * as other threads in the parent group can be right 1166 * here reaping other children at the same time. 1167 */ 1168 spin_lock_irq(&p->parent->sighand->siglock); 1169 psig = p->parent->signal; 1170 sig = p->signal; 1171 psig->cutime = 1172 cputime_add(psig->cutime, 1173 cputime_add(p->utime, 1174 cputime_add(sig->utime, 1175 sig->cutime))); 1176 psig->cstime = 1177 cputime_add(psig->cstime, 1178 cputime_add(p->stime, 1179 cputime_add(sig->stime, 1180 sig->cstime))); 1181 psig->cmin_flt += 1182 p->min_flt + sig->min_flt + sig->cmin_flt; 1183 psig->cmaj_flt += 1184 p->maj_flt + sig->maj_flt + sig->cmaj_flt; 1185 psig->cnvcsw += 1186 p->nvcsw + sig->nvcsw + sig->cnvcsw; 1187 psig->cnivcsw += 1188 p->nivcsw + sig->nivcsw + sig->cnivcsw; 1189 spin_unlock_irq(&p->parent->sighand->siglock); 1190 } 1191 1192 /* 1193 * Now we are sure this task is interesting, and no other 1194 * thread can reap it because we set its state to EXIT_DEAD. 1195 */ 1196 read_unlock(&tasklist_lock); 1197 1198 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; 1199 status = (p->signal->flags & SIGNAL_GROUP_EXIT) 1200 ? p->signal->group_exit_code : p->exit_code; 1201 if (!retval && stat_addr) 1202 retval = put_user(status, stat_addr); 1203 if (!retval && infop) 1204 retval = put_user(SIGCHLD, &infop->si_signo); 1205 if (!retval && infop) 1206 retval = put_user(0, &infop->si_errno); 1207 if (!retval && infop) { 1208 int why; 1209 1210 if ((status & 0x7f) == 0) { 1211 why = CLD_EXITED; 1212 status >>= 8; 1213 } else { 1214 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; 1215 status &= 0x7f; 1216 } 1217 retval = put_user((short)why, &infop->si_code); 1218 if (!retval) 1219 retval = put_user(status, &infop->si_status); 1220 } 1221 if (!retval && infop) 1222 retval = put_user(p->pid, &infop->si_pid); 1223 if (!retval && infop) 1224 retval = put_user(p->uid, &infop->si_uid); 1225 if (retval) { 1226 // TODO: is this safe? 1227 p->exit_state = EXIT_ZOMBIE; 1228 return retval; 1229 } 1230 retval = p->pid; 1231 if (p->real_parent != p->parent) { 1232 write_lock_irq(&tasklist_lock); 1233 /* Double-check with lock held. */ 1234 if (p->real_parent != p->parent) { 1235 __ptrace_unlink(p); 1236 // TODO: is this safe? 1237 p->exit_state = EXIT_ZOMBIE; 1238 /* 1239 * If this is not a detached task, notify the parent. 1240 * If it's still not detached after that, don't release 1241 * it now. 1242 */ 1243 if (p->exit_signal != -1) { 1244 do_notify_parent(p, p->exit_signal); 1245 if (p->exit_signal != -1) 1246 p = NULL; 1247 } 1248 } 1249 write_unlock_irq(&tasklist_lock); 1250 } 1251 if (p != NULL) 1252 release_task(p); 1253 BUG_ON(!retval); 1254 return retval; 1255 } 1256 1257 /* 1258 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold 1259 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold 1260 * the lock and this task is uninteresting. If we return nonzero, we have 1261 * released the lock and the system call should return. 1262 */ 1263 static int wait_task_stopped(struct task_struct *p, int delayed_group_leader, 1264 int noreap, struct siginfo __user *infop, 1265 int __user *stat_addr, struct rusage __user *ru) 1266 { 1267 int retval, exit_code; 1268 1269 if (!p->exit_code) 1270 return 0; 1271 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) && 1272 p->signal && p->signal->group_stop_count > 0) 1273 /* 1274 * A group stop is in progress and this is the group leader. 1275 * We won't report until all threads have stopped. 1276 */ 1277 return 0; 1278 1279 /* 1280 * Now we are pretty sure this task is interesting. 1281 * Make sure it doesn't get reaped out from under us while we 1282 * give up the lock and then examine it below. We don't want to 1283 * keep holding onto the tasklist_lock while we call getrusage and 1284 * possibly take page faults for user memory. 1285 */ 1286 get_task_struct(p); 1287 read_unlock(&tasklist_lock); 1288 1289 if (unlikely(noreap)) { 1290 pid_t pid = p->pid; 1291 uid_t uid = p->uid; 1292 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED; 1293 1294 exit_code = p->exit_code; 1295 if (unlikely(!exit_code) || 1296 unlikely(p->state & TASK_TRACED)) 1297 goto bail_ref; 1298 return wait_noreap_copyout(p, pid, uid, 1299 why, (exit_code << 8) | 0x7f, 1300 infop, ru); 1301 } 1302 1303 write_lock_irq(&tasklist_lock); 1304 1305 /* 1306 * This uses xchg to be atomic with the thread resuming and setting 1307 * it. It must also be done with the write lock held to prevent a 1308 * race with the EXIT_ZOMBIE case. 1309 */ 1310 exit_code = xchg(&p->exit_code, 0); 1311 if (unlikely(p->exit_state)) { 1312 /* 1313 * The task resumed and then died. Let the next iteration 1314 * catch it in EXIT_ZOMBIE. Note that exit_code might 1315 * already be zero here if it resumed and did _exit(0). 1316 * The task itself is dead and won't touch exit_code again; 1317 * other processors in this function are locked out. 1318 */ 1319 p->exit_code = exit_code; 1320 exit_code = 0; 1321 } 1322 if (unlikely(exit_code == 0)) { 1323 /* 1324 * Another thread in this function got to it first, or it 1325 * resumed, or it resumed and then died. 1326 */ 1327 write_unlock_irq(&tasklist_lock); 1328 bail_ref: 1329 put_task_struct(p); 1330 /* 1331 * We are returning to the wait loop without having successfully 1332 * removed the process and having released the lock. We cannot 1333 * continue, since the "p" task pointer is potentially stale. 1334 * 1335 * Return -EAGAIN, and do_wait() will restart the loop from the 1336 * beginning. Do _not_ re-acquire the lock. 1337 */ 1338 return -EAGAIN; 1339 } 1340 1341 /* move to end of parent's list to avoid starvation */ 1342 remove_parent(p); 1343 add_parent(p); 1344 1345 write_unlock_irq(&tasklist_lock); 1346 1347 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; 1348 if (!retval && stat_addr) 1349 retval = put_user((exit_code << 8) | 0x7f, stat_addr); 1350 if (!retval && infop) 1351 retval = put_user(SIGCHLD, &infop->si_signo); 1352 if (!retval && infop) 1353 retval = put_user(0, &infop->si_errno); 1354 if (!retval && infop) 1355 retval = put_user((short)((p->ptrace & PT_PTRACED) 1356 ? CLD_TRAPPED : CLD_STOPPED), 1357 &infop->si_code); 1358 if (!retval && infop) 1359 retval = put_user(exit_code, &infop->si_status); 1360 if (!retval && infop) 1361 retval = put_user(p->pid, &infop->si_pid); 1362 if (!retval && infop) 1363 retval = put_user(p->uid, &infop->si_uid); 1364 if (!retval) 1365 retval = p->pid; 1366 put_task_struct(p); 1367 1368 BUG_ON(!retval); 1369 return retval; 1370 } 1371 1372 /* 1373 * Handle do_wait work for one task in a live, non-stopped state. 1374 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold 1375 * the lock and this task is uninteresting. If we return nonzero, we have 1376 * released the lock and the system call should return. 1377 */ 1378 static int wait_task_continued(struct task_struct *p, int noreap, 1379 struct siginfo __user *infop, 1380 int __user *stat_addr, struct rusage __user *ru) 1381 { 1382 int retval; 1383 pid_t pid; 1384 uid_t uid; 1385 1386 if (unlikely(!p->signal)) 1387 return 0; 1388 1389 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) 1390 return 0; 1391 1392 spin_lock_irq(&p->sighand->siglock); 1393 /* Re-check with the lock held. */ 1394 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { 1395 spin_unlock_irq(&p->sighand->siglock); 1396 return 0; 1397 } 1398 if (!noreap) 1399 p->signal->flags &= ~SIGNAL_STOP_CONTINUED; 1400 spin_unlock_irq(&p->sighand->siglock); 1401 1402 pid = p->pid; 1403 uid = p->uid; 1404 get_task_struct(p); 1405 read_unlock(&tasklist_lock); 1406 1407 if (!infop) { 1408 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; 1409 put_task_struct(p); 1410 if (!retval && stat_addr) 1411 retval = put_user(0xffff, stat_addr); 1412 if (!retval) 1413 retval = p->pid; 1414 } else { 1415 retval = wait_noreap_copyout(p, pid, uid, 1416 CLD_CONTINUED, SIGCONT, 1417 infop, ru); 1418 BUG_ON(retval == 0); 1419 } 1420 1421 return retval; 1422 } 1423 1424 1425 static inline int my_ptrace_child(struct task_struct *p) 1426 { 1427 if (!(p->ptrace & PT_PTRACED)) 1428 return 0; 1429 if (!(p->ptrace & PT_ATTACHED)) 1430 return 1; 1431 /* 1432 * This child was PTRACE_ATTACH'd. We should be seeing it only if 1433 * we are the attacher. If we are the real parent, this is a race 1434 * inside ptrace_attach. It is waiting for the tasklist_lock, 1435 * which we have to switch the parent links, but has already set 1436 * the flags in p->ptrace. 1437 */ 1438 return (p->parent != p->real_parent); 1439 } 1440 1441 static long do_wait(pid_t pid, int options, struct siginfo __user *infop, 1442 int __user *stat_addr, struct rusage __user *ru) 1443 { 1444 DECLARE_WAITQUEUE(wait, current); 1445 struct task_struct *tsk; 1446 int flag, retval; 1447 1448 add_wait_queue(¤t->signal->wait_chldexit,&wait); 1449 repeat: 1450 /* 1451 * We will set this flag if we see any child that might later 1452 * match our criteria, even if we are not able to reap it yet. 1453 */ 1454 flag = 0; 1455 current->state = TASK_INTERRUPTIBLE; 1456 read_lock(&tasklist_lock); 1457 tsk = current; 1458 do { 1459 struct task_struct *p; 1460 struct list_head *_p; 1461 int ret; 1462 1463 list_for_each(_p,&tsk->children) { 1464 p = list_entry(_p, struct task_struct, sibling); 1465 1466 ret = eligible_child(pid, options, p); 1467 if (!ret) 1468 continue; 1469 1470 switch (p->state) { 1471 case TASK_TRACED: 1472 /* 1473 * When we hit the race with PTRACE_ATTACH, 1474 * we will not report this child. But the 1475 * race means it has not yet been moved to 1476 * our ptrace_children list, so we need to 1477 * set the flag here to avoid a spurious ECHILD 1478 * when the race happens with the only child. 1479 */ 1480 flag = 1; 1481 if (!my_ptrace_child(p)) 1482 continue; 1483 /*FALLTHROUGH*/ 1484 case TASK_STOPPED: 1485 /* 1486 * It's stopped now, so it might later 1487 * continue, exit, or stop again. 1488 */ 1489 flag = 1; 1490 if (!(options & WUNTRACED) && 1491 !my_ptrace_child(p)) 1492 continue; 1493 retval = wait_task_stopped(p, ret == 2, 1494 (options & WNOWAIT), 1495 infop, 1496 stat_addr, ru); 1497 if (retval == -EAGAIN) 1498 goto repeat; 1499 if (retval != 0) /* He released the lock. */ 1500 goto end; 1501 break; 1502 default: 1503 // case EXIT_DEAD: 1504 if (p->exit_state == EXIT_DEAD) 1505 continue; 1506 // case EXIT_ZOMBIE: 1507 if (p->exit_state == EXIT_ZOMBIE) { 1508 /* 1509 * Eligible but we cannot release 1510 * it yet: 1511 */ 1512 if (ret == 2) 1513 goto check_continued; 1514 if (!likely(options & WEXITED)) 1515 continue; 1516 retval = wait_task_zombie( 1517 p, (options & WNOWAIT), 1518 infop, stat_addr, ru); 1519 /* He released the lock. */ 1520 if (retval != 0) 1521 goto end; 1522 break; 1523 } 1524 check_continued: 1525 /* 1526 * It's running now, so it might later 1527 * exit, stop, or stop and then continue. 1528 */ 1529 flag = 1; 1530 if (!unlikely(options & WCONTINUED)) 1531 continue; 1532 retval = wait_task_continued( 1533 p, (options & WNOWAIT), 1534 infop, stat_addr, ru); 1535 if (retval != 0) /* He released the lock. */ 1536 goto end; 1537 break; 1538 } 1539 } 1540 if (!flag) { 1541 list_for_each(_p, &tsk->ptrace_children) { 1542 p = list_entry(_p, struct task_struct, 1543 ptrace_list); 1544 if (!eligible_child(pid, options, p)) 1545 continue; 1546 flag = 1; 1547 break; 1548 } 1549 } 1550 if (options & __WNOTHREAD) 1551 break; 1552 tsk = next_thread(tsk); 1553 BUG_ON(tsk->signal != current->signal); 1554 } while (tsk != current); 1555 1556 read_unlock(&tasklist_lock); 1557 if (flag) { 1558 retval = 0; 1559 if (options & WNOHANG) 1560 goto end; 1561 retval = -ERESTARTSYS; 1562 if (signal_pending(current)) 1563 goto end; 1564 schedule(); 1565 goto repeat; 1566 } 1567 retval = -ECHILD; 1568 end: 1569 current->state = TASK_RUNNING; 1570 remove_wait_queue(¤t->signal->wait_chldexit,&wait); 1571 if (infop) { 1572 if (retval > 0) 1573 retval = 0; 1574 else { 1575 /* 1576 * For a WNOHANG return, clear out all the fields 1577 * we would set so the user can easily tell the 1578 * difference. 1579 */ 1580 if (!retval) 1581 retval = put_user(0, &infop->si_signo); 1582 if (!retval) 1583 retval = put_user(0, &infop->si_errno); 1584 if (!retval) 1585 retval = put_user(0, &infop->si_code); 1586 if (!retval) 1587 retval = put_user(0, &infop->si_pid); 1588 if (!retval) 1589 retval = put_user(0, &infop->si_uid); 1590 if (!retval) 1591 retval = put_user(0, &infop->si_status); 1592 } 1593 } 1594 return retval; 1595 } 1596 1597 asmlinkage long sys_waitid(int which, pid_t pid, 1598 struct siginfo __user *infop, int options, 1599 struct rusage __user *ru) 1600 { 1601 long ret; 1602 1603 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) 1604 return -EINVAL; 1605 if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) 1606 return -EINVAL; 1607 1608 switch (which) { 1609 case P_ALL: 1610 pid = -1; 1611 break; 1612 case P_PID: 1613 if (pid <= 0) 1614 return -EINVAL; 1615 break; 1616 case P_PGID: 1617 if (pid <= 0) 1618 return -EINVAL; 1619 pid = -pid; 1620 break; 1621 default: 1622 return -EINVAL; 1623 } 1624 1625 ret = do_wait(pid, options, infop, NULL, ru); 1626 1627 /* avoid REGPARM breakage on x86: */ 1628 prevent_tail_call(ret); 1629 return ret; 1630 } 1631 1632 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr, 1633 int options, struct rusage __user *ru) 1634 { 1635 long ret; 1636 1637 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| 1638 __WNOTHREAD|__WCLONE|__WALL)) 1639 return -EINVAL; 1640 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru); 1641 1642 /* avoid REGPARM breakage on x86: */ 1643 prevent_tail_call(ret); 1644 return ret; 1645 } 1646 1647 #ifdef __ARCH_WANT_SYS_WAITPID 1648 1649 /* 1650 * sys_waitpid() remains for compatibility. waitpid() should be 1651 * implemented by calling sys_wait4() from libc.a. 1652 */ 1653 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options) 1654 { 1655 return sys_wait4(pid, stat_addr, options, NULL); 1656 } 1657 1658 #endif 1659