]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - kernel/exit.c
802722814925d6391bb021d99106314c7c6b085e
[mirror_ubuntu-zesty-kernel.git] / kernel / exit.c
1 /*
2 * linux/kernel/exit.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7 #include <linux/config.h>
8 #include <linux/mm.h>
9 #include <linux/slab.h>
10 #include <linux/interrupt.h>
11 #include <linux/smp_lock.h>
12 #include <linux/module.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/file.h>
22 #include <linux/binfmts.h>
23 #include <linux/ptrace.h>
24 #include <linux/profile.h>
25 #include <linux/mount.h>
26 #include <linux/proc_fs.h>
27 #include <linux/mempolicy.h>
28 #include <linux/cpuset.h>
29 #include <linux/syscalls.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/mutex.h>
33
34 #include <asm/uaccess.h>
35 #include <asm/unistd.h>
36 #include <asm/pgtable.h>
37 #include <asm/mmu_context.h>
38
39 extern void sem_exit (void);
40 extern struct task_struct *child_reaper;
41
42 int getrusage(struct task_struct *, int, struct rusage __user *);
43
44 static void exit_mm(struct task_struct * tsk);
45
46 static void __unhash_process(struct task_struct *p)
47 {
48 nr_threads--;
49 detach_pid(p, PIDTYPE_PID);
50 detach_pid(p, PIDTYPE_TGID);
51 if (thread_group_leader(p)) {
52 detach_pid(p, PIDTYPE_PGID);
53 detach_pid(p, PIDTYPE_SID);
54 if (p->pid)
55 __get_cpu_var(process_counts)--;
56 }
57
58 REMOVE_LINKS(p);
59 }
60
61 void release_task(struct task_struct * p)
62 {
63 int zap_leader;
64 task_t *leader;
65 struct dentry *proc_dentry;
66
67 repeat:
68 atomic_dec(&p->user->processes);
69 spin_lock(&p->proc_lock);
70 proc_dentry = proc_pid_unhash(p);
71 write_lock_irq(&tasklist_lock);
72 if (unlikely(p->ptrace))
73 __ptrace_unlink(p);
74 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
75 __exit_signal(p);
76 /*
77 * Note that the fastpath in sys_times depends on __exit_signal having
78 * updated the counters before a task is removed from the tasklist of
79 * the process by __unhash_process.
80 */
81 __unhash_process(p);
82
83 /*
84 * If we are the last non-leader member of the thread
85 * group, and the leader is zombie, then notify the
86 * group leader's parent process. (if it wants notification.)
87 */
88 zap_leader = 0;
89 leader = p->group_leader;
90 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
91 BUG_ON(leader->exit_signal == -1);
92 do_notify_parent(leader, leader->exit_signal);
93 /*
94 * If we were the last child thread and the leader has
95 * exited already, and the leader's parent ignores SIGCHLD,
96 * then we are the one who should release the leader.
97 *
98 * do_notify_parent() will have marked it self-reaping in
99 * that case.
100 */
101 zap_leader = (leader->exit_signal == -1);
102 }
103
104 sched_exit(p);
105 write_unlock_irq(&tasklist_lock);
106 spin_unlock(&p->proc_lock);
107 proc_pid_flush(proc_dentry);
108 release_thread(p);
109 put_task_struct(p);
110
111 p = leader;
112 if (unlikely(zap_leader))
113 goto repeat;
114 }
115
116 /* we are using it only for SMP init */
117
118 void unhash_process(struct task_struct *p)
119 {
120 struct dentry *proc_dentry;
121
122 spin_lock(&p->proc_lock);
123 proc_dentry = proc_pid_unhash(p);
124 write_lock_irq(&tasklist_lock);
125 __unhash_process(p);
126 write_unlock_irq(&tasklist_lock);
127 spin_unlock(&p->proc_lock);
128 proc_pid_flush(proc_dentry);
129 }
130
131 /*
132 * This checks not only the pgrp, but falls back on the pid if no
133 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
134 * without this...
135 */
136 int session_of_pgrp(int pgrp)
137 {
138 struct task_struct *p;
139 int sid = -1;
140
141 read_lock(&tasklist_lock);
142 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
143 if (p->signal->session > 0) {
144 sid = p->signal->session;
145 goto out;
146 }
147 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
148 p = find_task_by_pid(pgrp);
149 if (p)
150 sid = p->signal->session;
151 out:
152 read_unlock(&tasklist_lock);
153
154 return sid;
155 }
156
157 /*
158 * Determine if a process group is "orphaned", according to the POSIX
159 * definition in 2.2.2.52. Orphaned process groups are not to be affected
160 * by terminal-generated stop signals. Newly orphaned process groups are
161 * to receive a SIGHUP and a SIGCONT.
162 *
163 * "I ask you, have you ever known what it is to be an orphan?"
164 */
165 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
166 {
167 struct task_struct *p;
168 int ret = 1;
169
170 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
171 if (p == ignored_task
172 || p->exit_state
173 || p->real_parent->pid == 1)
174 continue;
175 if (process_group(p->real_parent) != pgrp
176 && p->real_parent->signal->session == p->signal->session) {
177 ret = 0;
178 break;
179 }
180 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
181 return ret; /* (sighing) "Often!" */
182 }
183
184 int is_orphaned_pgrp(int pgrp)
185 {
186 int retval;
187
188 read_lock(&tasklist_lock);
189 retval = will_become_orphaned_pgrp(pgrp, NULL);
190 read_unlock(&tasklist_lock);
191
192 return retval;
193 }
194
195 static inline int has_stopped_jobs(int pgrp)
196 {
197 int retval = 0;
198 struct task_struct *p;
199
200 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
201 if (p->state != TASK_STOPPED)
202 continue;
203
204 /* If p is stopped by a debugger on a signal that won't
205 stop it, then don't count p as stopped. This isn't
206 perfect but it's a good approximation. */
207 if (unlikely (p->ptrace)
208 && p->exit_code != SIGSTOP
209 && p->exit_code != SIGTSTP
210 && p->exit_code != SIGTTOU
211 && p->exit_code != SIGTTIN)
212 continue;
213
214 retval = 1;
215 break;
216 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
217 return retval;
218 }
219
220 /**
221 * reparent_to_init - Reparent the calling kernel thread to the init task.
222 *
223 * If a kernel thread is launched as a result of a system call, or if
224 * it ever exits, it should generally reparent itself to init so that
225 * it is correctly cleaned up on exit.
226 *
227 * The various task state such as scheduling policy and priority may have
228 * been inherited from a user process, so we reset them to sane values here.
229 *
230 * NOTE that reparent_to_init() gives the caller full capabilities.
231 */
232 static inline void reparent_to_init(void)
233 {
234 write_lock_irq(&tasklist_lock);
235
236 ptrace_unlink(current);
237 /* Reparent to init */
238 REMOVE_LINKS(current);
239 current->parent = child_reaper;
240 current->real_parent = child_reaper;
241 SET_LINKS(current);
242
243 /* Set the exit signal to SIGCHLD so we signal init on exit */
244 current->exit_signal = SIGCHLD;
245
246 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
247 set_user_nice(current, 0);
248 /* cpus_allowed? */
249 /* rt_priority? */
250 /* signals? */
251 security_task_reparent_to_init(current);
252 memcpy(current->signal->rlim, init_task.signal->rlim,
253 sizeof(current->signal->rlim));
254 atomic_inc(&(INIT_USER->__count));
255 write_unlock_irq(&tasklist_lock);
256 switch_uid(INIT_USER);
257 }
258
259 void __set_special_pids(pid_t session, pid_t pgrp)
260 {
261 struct task_struct *curr = current->group_leader;
262
263 if (curr->signal->session != session) {
264 detach_pid(curr, PIDTYPE_SID);
265 curr->signal->session = session;
266 attach_pid(curr, PIDTYPE_SID, session);
267 }
268 if (process_group(curr) != pgrp) {
269 detach_pid(curr, PIDTYPE_PGID);
270 curr->signal->pgrp = pgrp;
271 attach_pid(curr, PIDTYPE_PGID, pgrp);
272 }
273 }
274
275 void set_special_pids(pid_t session, pid_t pgrp)
276 {
277 write_lock_irq(&tasklist_lock);
278 __set_special_pids(session, pgrp);
279 write_unlock_irq(&tasklist_lock);
280 }
281
282 /*
283 * Let kernel threads use this to say that they
284 * allow a certain signal (since daemonize() will
285 * have disabled all of them by default).
286 */
287 int allow_signal(int sig)
288 {
289 if (!valid_signal(sig) || sig < 1)
290 return -EINVAL;
291
292 spin_lock_irq(&current->sighand->siglock);
293 sigdelset(&current->blocked, sig);
294 if (!current->mm) {
295 /* Kernel threads handle their own signals.
296 Let the signal code know it'll be handled, so
297 that they don't get converted to SIGKILL or
298 just silently dropped */
299 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
300 }
301 recalc_sigpending();
302 spin_unlock_irq(&current->sighand->siglock);
303 return 0;
304 }
305
306 EXPORT_SYMBOL(allow_signal);
307
308 int disallow_signal(int sig)
309 {
310 if (!valid_signal(sig) || sig < 1)
311 return -EINVAL;
312
313 spin_lock_irq(&current->sighand->siglock);
314 sigaddset(&current->blocked, sig);
315 recalc_sigpending();
316 spin_unlock_irq(&current->sighand->siglock);
317 return 0;
318 }
319
320 EXPORT_SYMBOL(disallow_signal);
321
322 /*
323 * Put all the gunge required to become a kernel thread without
324 * attached user resources in one place where it belongs.
325 */
326
327 void daemonize(const char *name, ...)
328 {
329 va_list args;
330 struct fs_struct *fs;
331 sigset_t blocked;
332
333 va_start(args, name);
334 vsnprintf(current->comm, sizeof(current->comm), name, args);
335 va_end(args);
336
337 /*
338 * If we were started as result of loading a module, close all of the
339 * user space pages. We don't need them, and if we didn't close them
340 * they would be locked into memory.
341 */
342 exit_mm(current);
343
344 set_special_pids(1, 1);
345 down(&tty_sem);
346 current->signal->tty = NULL;
347 up(&tty_sem);
348
349 /* Block and flush all signals */
350 sigfillset(&blocked);
351 sigprocmask(SIG_BLOCK, &blocked, NULL);
352 flush_signals(current);
353
354 /* Become as one with the init task */
355
356 exit_fs(current); /* current->fs->count--; */
357 fs = init_task.fs;
358 current->fs = fs;
359 atomic_inc(&fs->count);
360 exit_files(current);
361 current->files = init_task.files;
362 atomic_inc(&current->files->count);
363
364 reparent_to_init();
365 }
366
367 EXPORT_SYMBOL(daemonize);
368
369 static inline void close_files(struct files_struct * files)
370 {
371 int i, j;
372 struct fdtable *fdt;
373
374 j = 0;
375
376 /*
377 * It is safe to dereference the fd table without RCU or
378 * ->file_lock because this is the last reference to the
379 * files structure.
380 */
381 fdt = files_fdtable(files);
382 for (;;) {
383 unsigned long set;
384 i = j * __NFDBITS;
385 if (i >= fdt->max_fdset || i >= fdt->max_fds)
386 break;
387 set = fdt->open_fds->fds_bits[j++];
388 while (set) {
389 if (set & 1) {
390 struct file * file = xchg(&fdt->fd[i], NULL);
391 if (file)
392 filp_close(file, files);
393 }
394 i++;
395 set >>= 1;
396 }
397 }
398 }
399
400 struct files_struct *get_files_struct(struct task_struct *task)
401 {
402 struct files_struct *files;
403
404 task_lock(task);
405 files = task->files;
406 if (files)
407 atomic_inc(&files->count);
408 task_unlock(task);
409
410 return files;
411 }
412
413 void fastcall put_files_struct(struct files_struct *files)
414 {
415 struct fdtable *fdt;
416
417 if (atomic_dec_and_test(&files->count)) {
418 close_files(files);
419 /*
420 * Free the fd and fdset arrays if we expanded them.
421 * If the fdtable was embedded, pass files for freeing
422 * at the end of the RCU grace period. Otherwise,
423 * you can free files immediately.
424 */
425 fdt = files_fdtable(files);
426 if (fdt == &files->fdtab)
427 fdt->free_files = files;
428 else
429 kmem_cache_free(files_cachep, files);
430 free_fdtable(fdt);
431 }
432 }
433
434 EXPORT_SYMBOL(put_files_struct);
435
436 static inline void __exit_files(struct task_struct *tsk)
437 {
438 struct files_struct * files = tsk->files;
439
440 if (files) {
441 task_lock(tsk);
442 tsk->files = NULL;
443 task_unlock(tsk);
444 put_files_struct(files);
445 }
446 }
447
448 void exit_files(struct task_struct *tsk)
449 {
450 __exit_files(tsk);
451 }
452
453 static inline void __put_fs_struct(struct fs_struct *fs)
454 {
455 /* No need to hold fs->lock if we are killing it */
456 if (atomic_dec_and_test(&fs->count)) {
457 dput(fs->root);
458 mntput(fs->rootmnt);
459 dput(fs->pwd);
460 mntput(fs->pwdmnt);
461 if (fs->altroot) {
462 dput(fs->altroot);
463 mntput(fs->altrootmnt);
464 }
465 kmem_cache_free(fs_cachep, fs);
466 }
467 }
468
469 void put_fs_struct(struct fs_struct *fs)
470 {
471 __put_fs_struct(fs);
472 }
473
474 static inline void __exit_fs(struct task_struct *tsk)
475 {
476 struct fs_struct * fs = tsk->fs;
477
478 if (fs) {
479 task_lock(tsk);
480 tsk->fs = NULL;
481 task_unlock(tsk);
482 __put_fs_struct(fs);
483 }
484 }
485
486 void exit_fs(struct task_struct *tsk)
487 {
488 __exit_fs(tsk);
489 }
490
491 EXPORT_SYMBOL_GPL(exit_fs);
492
493 /*
494 * Turn us into a lazy TLB process if we
495 * aren't already..
496 */
497 static void exit_mm(struct task_struct * tsk)
498 {
499 struct mm_struct *mm = tsk->mm;
500
501 mm_release(tsk, mm);
502 if (!mm)
503 return;
504 /*
505 * Serialize with any possible pending coredump.
506 * We must hold mmap_sem around checking core_waiters
507 * and clearing tsk->mm. The core-inducing thread
508 * will increment core_waiters for each thread in the
509 * group with ->mm != NULL.
510 */
511 down_read(&mm->mmap_sem);
512 if (mm->core_waiters) {
513 up_read(&mm->mmap_sem);
514 down_write(&mm->mmap_sem);
515 if (!--mm->core_waiters)
516 complete(mm->core_startup_done);
517 up_write(&mm->mmap_sem);
518
519 wait_for_completion(&mm->core_done);
520 down_read(&mm->mmap_sem);
521 }
522 atomic_inc(&mm->mm_count);
523 if (mm != tsk->active_mm) BUG();
524 /* more a memory barrier than a real lock */
525 task_lock(tsk);
526 tsk->mm = NULL;
527 up_read(&mm->mmap_sem);
528 enter_lazy_tlb(mm, current);
529 task_unlock(tsk);
530 mmput(mm);
531 }
532
533 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
534 {
535 /*
536 * Make sure we're not reparenting to ourselves and that
537 * the parent is not a zombie.
538 */
539 BUG_ON(p == reaper || reaper->exit_state >= EXIT_ZOMBIE);
540 p->real_parent = reaper;
541 }
542
543 static inline void reparent_thread(task_t *p, task_t *father, int traced)
544 {
545 /* We don't want people slaying init. */
546 if (p->exit_signal != -1)
547 p->exit_signal = SIGCHLD;
548
549 if (p->pdeath_signal)
550 /* We already hold the tasklist_lock here. */
551 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
552
553 /* Move the child from its dying parent to the new one. */
554 if (unlikely(traced)) {
555 /* Preserve ptrace links if someone else is tracing this child. */
556 list_del_init(&p->ptrace_list);
557 if (p->parent != p->real_parent)
558 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
559 } else {
560 /* If this child is being traced, then we're the one tracing it
561 * anyway, so let go of it.
562 */
563 p->ptrace = 0;
564 list_del_init(&p->sibling);
565 p->parent = p->real_parent;
566 list_add_tail(&p->sibling, &p->parent->children);
567
568 /* If we'd notified the old parent about this child's death,
569 * also notify the new parent.
570 */
571 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
572 thread_group_empty(p))
573 do_notify_parent(p, p->exit_signal);
574 else if (p->state == TASK_TRACED) {
575 /*
576 * If it was at a trace stop, turn it into
577 * a normal stop since it's no longer being
578 * traced.
579 */
580 ptrace_untrace(p);
581 }
582 }
583
584 /*
585 * process group orphan check
586 * Case ii: Our child is in a different pgrp
587 * than we are, and it was the only connection
588 * outside, so the child pgrp is now orphaned.
589 */
590 if ((process_group(p) != process_group(father)) &&
591 (p->signal->session == father->signal->session)) {
592 int pgrp = process_group(p);
593
594 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
595 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
596 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
597 }
598 }
599 }
600
601 /*
602 * When we die, we re-parent all our children.
603 * Try to give them to another thread in our thread
604 * group, and if no such member exists, give it to
605 * the global child reaper process (ie "init")
606 */
607 static inline void forget_original_parent(struct task_struct * father,
608 struct list_head *to_release)
609 {
610 struct task_struct *p, *reaper = father;
611 struct list_head *_p, *_n;
612
613 do {
614 reaper = next_thread(reaper);
615 if (reaper == father) {
616 reaper = child_reaper;
617 break;
618 }
619 } while (reaper->exit_state);
620
621 /*
622 * There are only two places where our children can be:
623 *
624 * - in our child list
625 * - in our ptraced child list
626 *
627 * Search them and reparent children.
628 */
629 list_for_each_safe(_p, _n, &father->children) {
630 int ptrace;
631 p = list_entry(_p,struct task_struct,sibling);
632
633 ptrace = p->ptrace;
634
635 /* if father isn't the real parent, then ptrace must be enabled */
636 BUG_ON(father != p->real_parent && !ptrace);
637
638 if (father == p->real_parent) {
639 /* reparent with a reaper, real father it's us */
640 choose_new_parent(p, reaper, child_reaper);
641 reparent_thread(p, father, 0);
642 } else {
643 /* reparent ptraced task to its real parent */
644 __ptrace_unlink (p);
645 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
646 thread_group_empty(p))
647 do_notify_parent(p, p->exit_signal);
648 }
649
650 /*
651 * if the ptraced child is a zombie with exit_signal == -1
652 * we must collect it before we exit, or it will remain
653 * zombie forever since we prevented it from self-reap itself
654 * while it was being traced by us, to be able to see it in wait4.
655 */
656 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
657 list_add(&p->ptrace_list, to_release);
658 }
659 list_for_each_safe(_p, _n, &father->ptrace_children) {
660 p = list_entry(_p,struct task_struct,ptrace_list);
661 choose_new_parent(p, reaper, child_reaper);
662 reparent_thread(p, father, 1);
663 }
664 }
665
666 /*
667 * Send signals to all our closest relatives so that they know
668 * to properly mourn us..
669 */
670 static void exit_notify(struct task_struct *tsk)
671 {
672 int state;
673 struct task_struct *t;
674 struct list_head ptrace_dead, *_p, *_n;
675
676 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
677 && !thread_group_empty(tsk)) {
678 /*
679 * This occurs when there was a race between our exit
680 * syscall and a group signal choosing us as the one to
681 * wake up. It could be that we are the only thread
682 * alerted to check for pending signals, but another thread
683 * should be woken now to take the signal since we will not.
684 * Now we'll wake all the threads in the group just to make
685 * sure someone gets all the pending signals.
686 */
687 read_lock(&tasklist_lock);
688 spin_lock_irq(&tsk->sighand->siglock);
689 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
690 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
691 recalc_sigpending_tsk(t);
692 if (signal_pending(t))
693 signal_wake_up(t, 0);
694 }
695 spin_unlock_irq(&tsk->sighand->siglock);
696 read_unlock(&tasklist_lock);
697 }
698
699 write_lock_irq(&tasklist_lock);
700
701 /*
702 * This does two things:
703 *
704 * A. Make init inherit all the child processes
705 * B. Check to see if any process groups have become orphaned
706 * as a result of our exiting, and if they have any stopped
707 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
708 */
709
710 INIT_LIST_HEAD(&ptrace_dead);
711 forget_original_parent(tsk, &ptrace_dead);
712 BUG_ON(!list_empty(&tsk->children));
713 BUG_ON(!list_empty(&tsk->ptrace_children));
714
715 /*
716 * Check to see if any process groups have become orphaned
717 * as a result of our exiting, and if they have any stopped
718 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
719 *
720 * Case i: Our father is in a different pgrp than we are
721 * and we were the only connection outside, so our pgrp
722 * is about to become orphaned.
723 */
724
725 t = tsk->real_parent;
726
727 if ((process_group(t) != process_group(tsk)) &&
728 (t->signal->session == tsk->signal->session) &&
729 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
730 has_stopped_jobs(process_group(tsk))) {
731 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
732 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
733 }
734
735 /* Let father know we died
736 *
737 * Thread signals are configurable, but you aren't going to use
738 * that to send signals to arbitary processes.
739 * That stops right now.
740 *
741 * If the parent exec id doesn't match the exec id we saved
742 * when we started then we know the parent has changed security
743 * domain.
744 *
745 * If our self_exec id doesn't match our parent_exec_id then
746 * we have changed execution domain as these two values started
747 * the same after a fork.
748 *
749 */
750
751 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
752 ( tsk->parent_exec_id != t->self_exec_id ||
753 tsk->self_exec_id != tsk->parent_exec_id)
754 && !capable(CAP_KILL))
755 tsk->exit_signal = SIGCHLD;
756
757
758 /* If something other than our normal parent is ptracing us, then
759 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
760 * only has special meaning to our real parent.
761 */
762 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
763 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
764 do_notify_parent(tsk, signal);
765 } else if (tsk->ptrace) {
766 do_notify_parent(tsk, SIGCHLD);
767 }
768
769 state = EXIT_ZOMBIE;
770 if (tsk->exit_signal == -1 &&
771 (likely(tsk->ptrace == 0) ||
772 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
773 state = EXIT_DEAD;
774 tsk->exit_state = state;
775
776 write_unlock_irq(&tasklist_lock);
777
778 list_for_each_safe(_p, _n, &ptrace_dead) {
779 list_del_init(_p);
780 t = list_entry(_p,struct task_struct,ptrace_list);
781 release_task(t);
782 }
783
784 /* If the process is dead, release it - nobody will wait for it */
785 if (state == EXIT_DEAD)
786 release_task(tsk);
787 }
788
789 fastcall NORET_TYPE void do_exit(long code)
790 {
791 struct task_struct *tsk = current;
792 int group_dead;
793
794 profile_task_exit(tsk);
795
796 WARN_ON(atomic_read(&tsk->fs_excl));
797
798 if (unlikely(in_interrupt()))
799 panic("Aiee, killing interrupt handler!");
800 if (unlikely(!tsk->pid))
801 panic("Attempted to kill the idle task!");
802 if (unlikely(tsk->pid == 1))
803 panic("Attempted to kill init!");
804 if (tsk->io_context)
805 exit_io_context();
806
807 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
808 current->ptrace_message = code;
809 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
810 }
811
812 /*
813 * We're taking recursive faults here in do_exit. Safest is to just
814 * leave this task alone and wait for reboot.
815 */
816 if (unlikely(tsk->flags & PF_EXITING)) {
817 printk(KERN_ALERT
818 "Fixing recursive fault but reboot is needed!\n");
819 set_current_state(TASK_UNINTERRUPTIBLE);
820 schedule();
821 }
822
823 tsk->flags |= PF_EXITING;
824
825 /*
826 * Make sure we don't try to process any timer firings
827 * while we are already exiting.
828 */
829 tsk->it_virt_expires = cputime_zero;
830 tsk->it_prof_expires = cputime_zero;
831 tsk->it_sched_expires = 0;
832
833 if (unlikely(in_atomic()))
834 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
835 current->comm, current->pid,
836 preempt_count());
837
838 acct_update_integrals(tsk);
839 if (tsk->mm) {
840 update_hiwater_rss(tsk->mm);
841 update_hiwater_vm(tsk->mm);
842 }
843 group_dead = atomic_dec_and_test(&tsk->signal->live);
844 if (group_dead) {
845 hrtimer_cancel(&tsk->signal->real_timer);
846 exit_itimers(tsk->signal);
847 acct_process(code);
848 }
849 exit_mm(tsk);
850
851 exit_sem(tsk);
852 __exit_files(tsk);
853 __exit_fs(tsk);
854 exit_namespace(tsk);
855 exit_thread();
856 cpuset_exit(tsk);
857 exit_keys(tsk);
858
859 if (group_dead && tsk->signal->leader)
860 disassociate_ctty(1);
861
862 module_put(task_thread_info(tsk)->exec_domain->module);
863 if (tsk->binfmt)
864 module_put(tsk->binfmt->module);
865
866 tsk->exit_code = code;
867 proc_exit_connector(tsk);
868 exit_notify(tsk);
869 #ifdef CONFIG_NUMA
870 mpol_free(tsk->mempolicy);
871 tsk->mempolicy = NULL;
872 #endif
873 /*
874 * If DEBUG_MUTEXES is on, make sure we are holding no locks:
875 */
876 mutex_debug_check_no_locks_held(tsk);
877
878 /* PF_DEAD causes final put_task_struct after we schedule. */
879 preempt_disable();
880 BUG_ON(tsk->flags & PF_DEAD);
881 tsk->flags |= PF_DEAD;
882
883 schedule();
884 BUG();
885 /* Avoid "noreturn function does return". */
886 for (;;) ;
887 }
888
889 EXPORT_SYMBOL_GPL(do_exit);
890
891 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
892 {
893 if (comp)
894 complete(comp);
895
896 do_exit(code);
897 }
898
899 EXPORT_SYMBOL(complete_and_exit);
900
901 asmlinkage long sys_exit(int error_code)
902 {
903 do_exit((error_code&0xff)<<8);
904 }
905
906 task_t fastcall *next_thread(const task_t *p)
907 {
908 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
909 }
910
911 EXPORT_SYMBOL(next_thread);
912
913 /*
914 * Take down every thread in the group. This is called by fatal signals
915 * as well as by sys_exit_group (below).
916 */
917 NORET_TYPE void
918 do_group_exit(int exit_code)
919 {
920 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
921
922 if (current->signal->flags & SIGNAL_GROUP_EXIT)
923 exit_code = current->signal->group_exit_code;
924 else if (!thread_group_empty(current)) {
925 struct signal_struct *const sig = current->signal;
926 struct sighand_struct *const sighand = current->sighand;
927 read_lock(&tasklist_lock);
928 spin_lock_irq(&sighand->siglock);
929 if (sig->flags & SIGNAL_GROUP_EXIT)
930 /* Another thread got here before we took the lock. */
931 exit_code = sig->group_exit_code;
932 else {
933 sig->group_exit_code = exit_code;
934 zap_other_threads(current);
935 }
936 spin_unlock_irq(&sighand->siglock);
937 read_unlock(&tasklist_lock);
938 }
939
940 do_exit(exit_code);
941 /* NOTREACHED */
942 }
943
944 /*
945 * this kills every thread in the thread group. Note that any externally
946 * wait4()-ing process will get the correct exit code - even if this
947 * thread is not the thread group leader.
948 */
949 asmlinkage void sys_exit_group(int error_code)
950 {
951 do_group_exit((error_code & 0xff) << 8);
952 }
953
954 static int eligible_child(pid_t pid, int options, task_t *p)
955 {
956 if (pid > 0) {
957 if (p->pid != pid)
958 return 0;
959 } else if (!pid) {
960 if (process_group(p) != process_group(current))
961 return 0;
962 } else if (pid != -1) {
963 if (process_group(p) != -pid)
964 return 0;
965 }
966
967 /*
968 * Do not consider detached threads that are
969 * not ptraced:
970 */
971 if (p->exit_signal == -1 && !p->ptrace)
972 return 0;
973
974 /* Wait for all children (clone and not) if __WALL is set;
975 * otherwise, wait for clone children *only* if __WCLONE is
976 * set; otherwise, wait for non-clone children *only*. (Note:
977 * A "clone" child here is one that reports to its parent
978 * using a signal other than SIGCHLD.) */
979 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
980 && !(options & __WALL))
981 return 0;
982 /*
983 * Do not consider thread group leaders that are
984 * in a non-empty thread group:
985 */
986 if (current->tgid != p->tgid && delay_group_leader(p))
987 return 2;
988
989 if (security_task_wait(p))
990 return 0;
991
992 return 1;
993 }
994
995 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
996 int why, int status,
997 struct siginfo __user *infop,
998 struct rusage __user *rusagep)
999 {
1000 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1001 put_task_struct(p);
1002 if (!retval)
1003 retval = put_user(SIGCHLD, &infop->si_signo);
1004 if (!retval)
1005 retval = put_user(0, &infop->si_errno);
1006 if (!retval)
1007 retval = put_user((short)why, &infop->si_code);
1008 if (!retval)
1009 retval = put_user(pid, &infop->si_pid);
1010 if (!retval)
1011 retval = put_user(uid, &infop->si_uid);
1012 if (!retval)
1013 retval = put_user(status, &infop->si_status);
1014 if (!retval)
1015 retval = pid;
1016 return retval;
1017 }
1018
1019 /*
1020 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1021 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1022 * the lock and this task is uninteresting. If we return nonzero, we have
1023 * released the lock and the system call should return.
1024 */
1025 static int wait_task_zombie(task_t *p, int noreap,
1026 struct siginfo __user *infop,
1027 int __user *stat_addr, struct rusage __user *ru)
1028 {
1029 unsigned long state;
1030 int retval;
1031 int status;
1032
1033 if (unlikely(noreap)) {
1034 pid_t pid = p->pid;
1035 uid_t uid = p->uid;
1036 int exit_code = p->exit_code;
1037 int why, status;
1038
1039 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1040 return 0;
1041 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1042 return 0;
1043 get_task_struct(p);
1044 read_unlock(&tasklist_lock);
1045 if ((exit_code & 0x7f) == 0) {
1046 why = CLD_EXITED;
1047 status = exit_code >> 8;
1048 } else {
1049 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1050 status = exit_code & 0x7f;
1051 }
1052 return wait_noreap_copyout(p, pid, uid, why,
1053 status, infop, ru);
1054 }
1055
1056 /*
1057 * Try to move the task's state to DEAD
1058 * only one thread is allowed to do this:
1059 */
1060 state = xchg(&p->exit_state, EXIT_DEAD);
1061 if (state != EXIT_ZOMBIE) {
1062 BUG_ON(state != EXIT_DEAD);
1063 return 0;
1064 }
1065 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1066 /*
1067 * This can only happen in a race with a ptraced thread
1068 * dying on another processor.
1069 */
1070 return 0;
1071 }
1072
1073 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1074 struct signal_struct *psig;
1075 struct signal_struct *sig;
1076
1077 /*
1078 * The resource counters for the group leader are in its
1079 * own task_struct. Those for dead threads in the group
1080 * are in its signal_struct, as are those for the child
1081 * processes it has previously reaped. All these
1082 * accumulate in the parent's signal_struct c* fields.
1083 *
1084 * We don't bother to take a lock here to protect these
1085 * p->signal fields, because they are only touched by
1086 * __exit_signal, which runs with tasklist_lock
1087 * write-locked anyway, and so is excluded here. We do
1088 * need to protect the access to p->parent->signal fields,
1089 * as other threads in the parent group can be right
1090 * here reaping other children at the same time.
1091 */
1092 spin_lock_irq(&p->parent->sighand->siglock);
1093 psig = p->parent->signal;
1094 sig = p->signal;
1095 psig->cutime =
1096 cputime_add(psig->cutime,
1097 cputime_add(p->utime,
1098 cputime_add(sig->utime,
1099 sig->cutime)));
1100 psig->cstime =
1101 cputime_add(psig->cstime,
1102 cputime_add(p->stime,
1103 cputime_add(sig->stime,
1104 sig->cstime)));
1105 psig->cmin_flt +=
1106 p->min_flt + sig->min_flt + sig->cmin_flt;
1107 psig->cmaj_flt +=
1108 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1109 psig->cnvcsw +=
1110 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1111 psig->cnivcsw +=
1112 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1113 spin_unlock_irq(&p->parent->sighand->siglock);
1114 }
1115
1116 /*
1117 * Now we are sure this task is interesting, and no other
1118 * thread can reap it because we set its state to EXIT_DEAD.
1119 */
1120 read_unlock(&tasklist_lock);
1121
1122 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1123 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1124 ? p->signal->group_exit_code : p->exit_code;
1125 if (!retval && stat_addr)
1126 retval = put_user(status, stat_addr);
1127 if (!retval && infop)
1128 retval = put_user(SIGCHLD, &infop->si_signo);
1129 if (!retval && infop)
1130 retval = put_user(0, &infop->si_errno);
1131 if (!retval && infop) {
1132 int why;
1133
1134 if ((status & 0x7f) == 0) {
1135 why = CLD_EXITED;
1136 status >>= 8;
1137 } else {
1138 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1139 status &= 0x7f;
1140 }
1141 retval = put_user((short)why, &infop->si_code);
1142 if (!retval)
1143 retval = put_user(status, &infop->si_status);
1144 }
1145 if (!retval && infop)
1146 retval = put_user(p->pid, &infop->si_pid);
1147 if (!retval && infop)
1148 retval = put_user(p->uid, &infop->si_uid);
1149 if (retval) {
1150 // TODO: is this safe?
1151 p->exit_state = EXIT_ZOMBIE;
1152 return retval;
1153 }
1154 retval = p->pid;
1155 if (p->real_parent != p->parent) {
1156 write_lock_irq(&tasklist_lock);
1157 /* Double-check with lock held. */
1158 if (p->real_parent != p->parent) {
1159 __ptrace_unlink(p);
1160 // TODO: is this safe?
1161 p->exit_state = EXIT_ZOMBIE;
1162 /*
1163 * If this is not a detached task, notify the parent.
1164 * If it's still not detached after that, don't release
1165 * it now.
1166 */
1167 if (p->exit_signal != -1) {
1168 do_notify_parent(p, p->exit_signal);
1169 if (p->exit_signal != -1)
1170 p = NULL;
1171 }
1172 }
1173 write_unlock_irq(&tasklist_lock);
1174 }
1175 if (p != NULL)
1176 release_task(p);
1177 BUG_ON(!retval);
1178 return retval;
1179 }
1180
1181 /*
1182 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1183 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1184 * the lock and this task is uninteresting. If we return nonzero, we have
1185 * released the lock and the system call should return.
1186 */
1187 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1188 struct siginfo __user *infop,
1189 int __user *stat_addr, struct rusage __user *ru)
1190 {
1191 int retval, exit_code;
1192
1193 if (!p->exit_code)
1194 return 0;
1195 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1196 p->signal && p->signal->group_stop_count > 0)
1197 /*
1198 * A group stop is in progress and this is the group leader.
1199 * We won't report until all threads have stopped.
1200 */
1201 return 0;
1202
1203 /*
1204 * Now we are pretty sure this task is interesting.
1205 * Make sure it doesn't get reaped out from under us while we
1206 * give up the lock and then examine it below. We don't want to
1207 * keep holding onto the tasklist_lock while we call getrusage and
1208 * possibly take page faults for user memory.
1209 */
1210 get_task_struct(p);
1211 read_unlock(&tasklist_lock);
1212
1213 if (unlikely(noreap)) {
1214 pid_t pid = p->pid;
1215 uid_t uid = p->uid;
1216 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1217
1218 exit_code = p->exit_code;
1219 if (unlikely(!exit_code) ||
1220 unlikely(p->state & TASK_TRACED))
1221 goto bail_ref;
1222 return wait_noreap_copyout(p, pid, uid,
1223 why, (exit_code << 8) | 0x7f,
1224 infop, ru);
1225 }
1226
1227 write_lock_irq(&tasklist_lock);
1228
1229 /*
1230 * This uses xchg to be atomic with the thread resuming and setting
1231 * it. It must also be done with the write lock held to prevent a
1232 * race with the EXIT_ZOMBIE case.
1233 */
1234 exit_code = xchg(&p->exit_code, 0);
1235 if (unlikely(p->exit_state)) {
1236 /*
1237 * The task resumed and then died. Let the next iteration
1238 * catch it in EXIT_ZOMBIE. Note that exit_code might
1239 * already be zero here if it resumed and did _exit(0).
1240 * The task itself is dead and won't touch exit_code again;
1241 * other processors in this function are locked out.
1242 */
1243 p->exit_code = exit_code;
1244 exit_code = 0;
1245 }
1246 if (unlikely(exit_code == 0)) {
1247 /*
1248 * Another thread in this function got to it first, or it
1249 * resumed, or it resumed and then died.
1250 */
1251 write_unlock_irq(&tasklist_lock);
1252 bail_ref:
1253 put_task_struct(p);
1254 /*
1255 * We are returning to the wait loop without having successfully
1256 * removed the process and having released the lock. We cannot
1257 * continue, since the "p" task pointer is potentially stale.
1258 *
1259 * Return -EAGAIN, and do_wait() will restart the loop from the
1260 * beginning. Do _not_ re-acquire the lock.
1261 */
1262 return -EAGAIN;
1263 }
1264
1265 /* move to end of parent's list to avoid starvation */
1266 remove_parent(p);
1267 add_parent(p, p->parent);
1268
1269 write_unlock_irq(&tasklist_lock);
1270
1271 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1272 if (!retval && stat_addr)
1273 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1274 if (!retval && infop)
1275 retval = put_user(SIGCHLD, &infop->si_signo);
1276 if (!retval && infop)
1277 retval = put_user(0, &infop->si_errno);
1278 if (!retval && infop)
1279 retval = put_user((short)((p->ptrace & PT_PTRACED)
1280 ? CLD_TRAPPED : CLD_STOPPED),
1281 &infop->si_code);
1282 if (!retval && infop)
1283 retval = put_user(exit_code, &infop->si_status);
1284 if (!retval && infop)
1285 retval = put_user(p->pid, &infop->si_pid);
1286 if (!retval && infop)
1287 retval = put_user(p->uid, &infop->si_uid);
1288 if (!retval)
1289 retval = p->pid;
1290 put_task_struct(p);
1291
1292 BUG_ON(!retval);
1293 return retval;
1294 }
1295
1296 /*
1297 * Handle do_wait work for one task in a live, non-stopped state.
1298 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1299 * the lock and this task is uninteresting. If we return nonzero, we have
1300 * released the lock and the system call should return.
1301 */
1302 static int wait_task_continued(task_t *p, int noreap,
1303 struct siginfo __user *infop,
1304 int __user *stat_addr, struct rusage __user *ru)
1305 {
1306 int retval;
1307 pid_t pid;
1308 uid_t uid;
1309
1310 if (unlikely(!p->signal))
1311 return 0;
1312
1313 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1314 return 0;
1315
1316 spin_lock_irq(&p->sighand->siglock);
1317 /* Re-check with the lock held. */
1318 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1319 spin_unlock_irq(&p->sighand->siglock);
1320 return 0;
1321 }
1322 if (!noreap)
1323 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1324 spin_unlock_irq(&p->sighand->siglock);
1325
1326 pid = p->pid;
1327 uid = p->uid;
1328 get_task_struct(p);
1329 read_unlock(&tasklist_lock);
1330
1331 if (!infop) {
1332 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1333 put_task_struct(p);
1334 if (!retval && stat_addr)
1335 retval = put_user(0xffff, stat_addr);
1336 if (!retval)
1337 retval = p->pid;
1338 } else {
1339 retval = wait_noreap_copyout(p, pid, uid,
1340 CLD_CONTINUED, SIGCONT,
1341 infop, ru);
1342 BUG_ON(retval == 0);
1343 }
1344
1345 return retval;
1346 }
1347
1348
1349 static inline int my_ptrace_child(struct task_struct *p)
1350 {
1351 if (!(p->ptrace & PT_PTRACED))
1352 return 0;
1353 if (!(p->ptrace & PT_ATTACHED))
1354 return 1;
1355 /*
1356 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1357 * we are the attacher. If we are the real parent, this is a race
1358 * inside ptrace_attach. It is waiting for the tasklist_lock,
1359 * which we have to switch the parent links, but has already set
1360 * the flags in p->ptrace.
1361 */
1362 return (p->parent != p->real_parent);
1363 }
1364
1365 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1366 int __user *stat_addr, struct rusage __user *ru)
1367 {
1368 DECLARE_WAITQUEUE(wait, current);
1369 struct task_struct *tsk;
1370 int flag, retval;
1371
1372 add_wait_queue(&current->signal->wait_chldexit,&wait);
1373 repeat:
1374 /*
1375 * We will set this flag if we see any child that might later
1376 * match our criteria, even if we are not able to reap it yet.
1377 */
1378 flag = 0;
1379 current->state = TASK_INTERRUPTIBLE;
1380 read_lock(&tasklist_lock);
1381 tsk = current;
1382 do {
1383 struct task_struct *p;
1384 struct list_head *_p;
1385 int ret;
1386
1387 list_for_each(_p,&tsk->children) {
1388 p = list_entry(_p,struct task_struct,sibling);
1389
1390 ret = eligible_child(pid, options, p);
1391 if (!ret)
1392 continue;
1393
1394 switch (p->state) {
1395 case TASK_TRACED:
1396 /*
1397 * When we hit the race with PTRACE_ATTACH,
1398 * we will not report this child. But the
1399 * race means it has not yet been moved to
1400 * our ptrace_children list, so we need to
1401 * set the flag here to avoid a spurious ECHILD
1402 * when the race happens with the only child.
1403 */
1404 flag = 1;
1405 if (!my_ptrace_child(p))
1406 continue;
1407 /*FALLTHROUGH*/
1408 case TASK_STOPPED:
1409 /*
1410 * It's stopped now, so it might later
1411 * continue, exit, or stop again.
1412 */
1413 flag = 1;
1414 if (!(options & WUNTRACED) &&
1415 !my_ptrace_child(p))
1416 continue;
1417 retval = wait_task_stopped(p, ret == 2,
1418 (options & WNOWAIT),
1419 infop,
1420 stat_addr, ru);
1421 if (retval == -EAGAIN)
1422 goto repeat;
1423 if (retval != 0) /* He released the lock. */
1424 goto end;
1425 break;
1426 default:
1427 // case EXIT_DEAD:
1428 if (p->exit_state == EXIT_DEAD)
1429 continue;
1430 // case EXIT_ZOMBIE:
1431 if (p->exit_state == EXIT_ZOMBIE) {
1432 /*
1433 * Eligible but we cannot release
1434 * it yet:
1435 */
1436 if (ret == 2)
1437 goto check_continued;
1438 if (!likely(options & WEXITED))
1439 continue;
1440 retval = wait_task_zombie(
1441 p, (options & WNOWAIT),
1442 infop, stat_addr, ru);
1443 /* He released the lock. */
1444 if (retval != 0)
1445 goto end;
1446 break;
1447 }
1448 check_continued:
1449 /*
1450 * It's running now, so it might later
1451 * exit, stop, or stop and then continue.
1452 */
1453 flag = 1;
1454 if (!unlikely(options & WCONTINUED))
1455 continue;
1456 retval = wait_task_continued(
1457 p, (options & WNOWAIT),
1458 infop, stat_addr, ru);
1459 if (retval != 0) /* He released the lock. */
1460 goto end;
1461 break;
1462 }
1463 }
1464 if (!flag) {
1465 list_for_each(_p, &tsk->ptrace_children) {
1466 p = list_entry(_p, struct task_struct,
1467 ptrace_list);
1468 if (!eligible_child(pid, options, p))
1469 continue;
1470 flag = 1;
1471 break;
1472 }
1473 }
1474 if (options & __WNOTHREAD)
1475 break;
1476 tsk = next_thread(tsk);
1477 if (tsk->signal != current->signal)
1478 BUG();
1479 } while (tsk != current);
1480
1481 read_unlock(&tasklist_lock);
1482 if (flag) {
1483 retval = 0;
1484 if (options & WNOHANG)
1485 goto end;
1486 retval = -ERESTARTSYS;
1487 if (signal_pending(current))
1488 goto end;
1489 schedule();
1490 goto repeat;
1491 }
1492 retval = -ECHILD;
1493 end:
1494 current->state = TASK_RUNNING;
1495 remove_wait_queue(&current->signal->wait_chldexit,&wait);
1496 if (infop) {
1497 if (retval > 0)
1498 retval = 0;
1499 else {
1500 /*
1501 * For a WNOHANG return, clear out all the fields
1502 * we would set so the user can easily tell the
1503 * difference.
1504 */
1505 if (!retval)
1506 retval = put_user(0, &infop->si_signo);
1507 if (!retval)
1508 retval = put_user(0, &infop->si_errno);
1509 if (!retval)
1510 retval = put_user(0, &infop->si_code);
1511 if (!retval)
1512 retval = put_user(0, &infop->si_pid);
1513 if (!retval)
1514 retval = put_user(0, &infop->si_uid);
1515 if (!retval)
1516 retval = put_user(0, &infop->si_status);
1517 }
1518 }
1519 return retval;
1520 }
1521
1522 asmlinkage long sys_waitid(int which, pid_t pid,
1523 struct siginfo __user *infop, int options,
1524 struct rusage __user *ru)
1525 {
1526 long ret;
1527
1528 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1529 return -EINVAL;
1530 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1531 return -EINVAL;
1532
1533 switch (which) {
1534 case P_ALL:
1535 pid = -1;
1536 break;
1537 case P_PID:
1538 if (pid <= 0)
1539 return -EINVAL;
1540 break;
1541 case P_PGID:
1542 if (pid <= 0)
1543 return -EINVAL;
1544 pid = -pid;
1545 break;
1546 default:
1547 return -EINVAL;
1548 }
1549
1550 ret = do_wait(pid, options, infop, NULL, ru);
1551
1552 /* avoid REGPARM breakage on x86: */
1553 prevent_tail_call(ret);
1554 return ret;
1555 }
1556
1557 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1558 int options, struct rusage __user *ru)
1559 {
1560 long ret;
1561
1562 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1563 __WNOTHREAD|__WCLONE|__WALL))
1564 return -EINVAL;
1565 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1566
1567 /* avoid REGPARM breakage on x86: */
1568 prevent_tail_call(ret);
1569 return ret;
1570 }
1571
1572 #ifdef __ARCH_WANT_SYS_WAITPID
1573
1574 /*
1575 * sys_waitpid() remains for compatibility. waitpid() should be
1576 * implemented by calling sys_wait4() from libc.a.
1577 */
1578 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1579 {
1580 return sys_wait4(pid, stat_addr, options, NULL);
1581 }
1582
1583 #endif