sh: Add support Renesas Solutions AP-325RXA board
[linux-2.6/pdupreez.git] / kernel / exit.c
blobeb4d6470d1d01f79a33a94df89833c3ba7d4387a
1 /*
2 * linux/kernel/exit.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
7 #include <linux/mm.h>
8 #include <linux/slab.h>
9 #include <linux/interrupt.h>
10 #include <linux/module.h>
11 #include <linux/capability.h>
12 #include <linux/completion.h>
13 #include <linux/personality.h>
14 #include <linux/tty.h>
15 #include <linux/mnt_namespace.h>
16 #include <linux/iocontext.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/fdtable.h>
24 #include <linux/binfmts.h>
25 #include <linux/nsproxy.h>
26 #include <linux/pid_namespace.h>
27 #include <linux/ptrace.h>
28 #include <linux/profile.h>
29 #include <linux/mount.h>
30 #include <linux/proc_fs.h>
31 #include <linux/kthread.h>
32 #include <linux/mempolicy.h>
33 #include <linux/taskstats_kern.h>
34 #include <linux/delayacct.h>
35 #include <linux/freezer.h>
36 #include <linux/cgroup.h>
37 #include <linux/syscalls.h>
38 #include <linux/signal.h>
39 #include <linux/posix-timers.h>
40 #include <linux/cn_proc.h>
41 #include <linux/mutex.h>
42 #include <linux/futex.h>
43 #include <linux/compat.h>
44 #include <linux/pipe_fs_i.h>
45 #include <linux/audit.h> /* for audit_free() */
46 #include <linux/resource.h>
47 #include <linux/blkdev.h>
48 #include <linux/task_io_accounting_ops.h>
49 #include <linux/tracehook.h>
51 #include <asm/uaccess.h>
52 #include <asm/unistd.h>
53 #include <asm/pgtable.h>
54 #include <asm/mmu_context.h>
56 static void exit_mm(struct task_struct * tsk);
58 static inline int task_detached(struct task_struct *p)
60 return p->exit_signal == -1;
63 static void __unhash_process(struct task_struct *p)
65 nr_threads--;
66 detach_pid(p, PIDTYPE_PID);
67 if (thread_group_leader(p)) {
68 detach_pid(p, PIDTYPE_PGID);
69 detach_pid(p, PIDTYPE_SID);
71 list_del_rcu(&p->tasks);
72 __get_cpu_var(process_counts)--;
74 list_del_rcu(&p->thread_group);
75 list_del_init(&p->sibling);
79 * This function expects the tasklist_lock write-locked.
81 static void __exit_signal(struct task_struct *tsk)
83 struct signal_struct *sig = tsk->signal;
84 struct sighand_struct *sighand;
86 BUG_ON(!sig);
87 BUG_ON(!atomic_read(&sig->count));
89 sighand = rcu_dereference(tsk->sighand);
90 spin_lock(&sighand->siglock);
92 posix_cpu_timers_exit(tsk);
93 if (atomic_dec_and_test(&sig->count))
94 posix_cpu_timers_exit_group(tsk);
95 else {
97 * If there is any task waiting for the group exit
98 * then notify it:
100 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count)
101 wake_up_process(sig->group_exit_task);
103 if (tsk == sig->curr_target)
104 sig->curr_target = next_thread(tsk);
106 * Accumulate here the counters for all threads but the
107 * group leader as they die, so they can be added into
108 * the process-wide totals when those are taken.
109 * The group leader stays around as a zombie as long
110 * as there are other threads. When it gets reaped,
111 * the exit.c code will add its counts into these totals.
112 * We won't ever get here for the group leader, since it
113 * will have been the last reference on the signal_struct.
115 sig->utime = cputime_add(sig->utime, tsk->utime);
116 sig->stime = cputime_add(sig->stime, tsk->stime);
117 sig->gtime = cputime_add(sig->gtime, tsk->gtime);
118 sig->min_flt += tsk->min_flt;
119 sig->maj_flt += tsk->maj_flt;
120 sig->nvcsw += tsk->nvcsw;
121 sig->nivcsw += tsk->nivcsw;
122 sig->inblock += task_io_get_inblock(tsk);
123 sig->oublock += task_io_get_oublock(tsk);
124 task_io_accounting_add(&sig->ioac, &tsk->ioac);
125 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
126 sig = NULL; /* Marker for below. */
129 __unhash_process(tsk);
132 * Do this under ->siglock, we can race with another thread
133 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
135 flush_sigqueue(&tsk->pending);
137 tsk->signal = NULL;
138 tsk->sighand = NULL;
139 spin_unlock(&sighand->siglock);
141 __cleanup_sighand(sighand);
142 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
143 if (sig) {
144 flush_sigqueue(&sig->shared_pending);
145 taskstats_tgid_free(sig);
146 __cleanup_signal(sig);
150 static void delayed_put_task_struct(struct rcu_head *rhp)
152 put_task_struct(container_of(rhp, struct task_struct, rcu));
156 void release_task(struct task_struct * p)
158 struct task_struct *leader;
159 int zap_leader;
160 repeat:
161 tracehook_prepare_release_task(p);
162 atomic_dec(&p->user->processes);
163 proc_flush_task(p);
164 write_lock_irq(&tasklist_lock);
165 tracehook_finish_release_task(p);
166 __exit_signal(p);
169 * If we are the last non-leader member of the thread
170 * group, and the leader is zombie, then notify the
171 * group leader's parent process. (if it wants notification.)
173 zap_leader = 0;
174 leader = p->group_leader;
175 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
176 BUG_ON(task_detached(leader));
177 do_notify_parent(leader, leader->exit_signal);
179 * If we were the last child thread and the leader has
180 * exited already, and the leader's parent ignores SIGCHLD,
181 * then we are the one who should release the leader.
183 * do_notify_parent() will have marked it self-reaping in
184 * that case.
186 zap_leader = task_detached(leader);
189 * This maintains the invariant that release_task()
190 * only runs on a task in EXIT_DEAD, just for sanity.
192 if (zap_leader)
193 leader->exit_state = EXIT_DEAD;
196 write_unlock_irq(&tasklist_lock);
197 release_thread(p);
198 call_rcu(&p->rcu, delayed_put_task_struct);
200 p = leader;
201 if (unlikely(zap_leader))
202 goto repeat;
206 * This checks not only the pgrp, but falls back on the pid if no
207 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
208 * without this...
210 * The caller must hold rcu lock or the tasklist lock.
212 struct pid *session_of_pgrp(struct pid *pgrp)
214 struct task_struct *p;
215 struct pid *sid = NULL;
217 p = pid_task(pgrp, PIDTYPE_PGID);
218 if (p == NULL)
219 p = pid_task(pgrp, PIDTYPE_PID);
220 if (p != NULL)
221 sid = task_session(p);
223 return sid;
227 * Determine if a process group is "orphaned", according to the POSIX
228 * definition in 2.2.2.52. Orphaned process groups are not to be affected
229 * by terminal-generated stop signals. Newly orphaned process groups are
230 * to receive a SIGHUP and a SIGCONT.
232 * "I ask you, have you ever known what it is to be an orphan?"
234 static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
236 struct task_struct *p;
238 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
239 if ((p == ignored_task) ||
240 (p->exit_state && thread_group_empty(p)) ||
241 is_global_init(p->real_parent))
242 continue;
244 if (task_pgrp(p->real_parent) != pgrp &&
245 task_session(p->real_parent) == task_session(p))
246 return 0;
247 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
249 return 1;
252 int is_current_pgrp_orphaned(void)
254 int retval;
256 read_lock(&tasklist_lock);
257 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
258 read_unlock(&tasklist_lock);
260 return retval;
263 static int has_stopped_jobs(struct pid *pgrp)
265 int retval = 0;
266 struct task_struct *p;
268 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
269 if (!task_is_stopped(p))
270 continue;
271 retval = 1;
272 break;
273 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
274 return retval;
278 * Check to see if any process groups have become orphaned as
279 * a result of our exiting, and if they have any stopped jobs,
280 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
282 static void
283 kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
285 struct pid *pgrp = task_pgrp(tsk);
286 struct task_struct *ignored_task = tsk;
288 if (!parent)
289 /* exit: our father is in a different pgrp than
290 * we are and we were the only connection outside.
292 parent = tsk->real_parent;
293 else
294 /* reparent: our child is in a different pgrp than
295 * we are, and it was the only connection outside.
297 ignored_task = NULL;
299 if (task_pgrp(parent) != pgrp &&
300 task_session(parent) == task_session(tsk) &&
301 will_become_orphaned_pgrp(pgrp, ignored_task) &&
302 has_stopped_jobs(pgrp)) {
303 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
304 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
309 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
311 * If a kernel thread is launched as a result of a system call, or if
312 * it ever exits, it should generally reparent itself to kthreadd so it
313 * isn't in the way of other processes and is correctly cleaned up on exit.
315 * The various task state such as scheduling policy and priority may have
316 * been inherited from a user process, so we reset them to sane values here.
318 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
320 static void reparent_to_kthreadd(void)
322 write_lock_irq(&tasklist_lock);
324 ptrace_unlink(current);
325 /* Reparent to init */
326 current->real_parent = current->parent = kthreadd_task;
327 list_move_tail(&current->sibling, &current->real_parent->children);
329 /* Set the exit signal to SIGCHLD so we signal init on exit */
330 current->exit_signal = SIGCHLD;
332 if (task_nice(current) < 0)
333 set_user_nice(current, 0);
334 /* cpus_allowed? */
335 /* rt_priority? */
336 /* signals? */
337 security_task_reparent_to_init(current);
338 memcpy(current->signal->rlim, init_task.signal->rlim,
339 sizeof(current->signal->rlim));
340 atomic_inc(&(INIT_USER->__count));
341 write_unlock_irq(&tasklist_lock);
342 switch_uid(INIT_USER);
345 void __set_special_pids(struct pid *pid)
347 struct task_struct *curr = current->group_leader;
348 pid_t nr = pid_nr(pid);
350 if (task_session(curr) != pid) {
351 change_pid(curr, PIDTYPE_SID, pid);
352 set_task_session(curr, nr);
354 if (task_pgrp(curr) != pid) {
355 change_pid(curr, PIDTYPE_PGID, pid);
356 set_task_pgrp(curr, nr);
360 static void set_special_pids(struct pid *pid)
362 write_lock_irq(&tasklist_lock);
363 __set_special_pids(pid);
364 write_unlock_irq(&tasklist_lock);
368 * Let kernel threads use this to say that they
369 * allow a certain signal (since daemonize() will
370 * have disabled all of them by default).
372 int allow_signal(int sig)
374 if (!valid_signal(sig) || sig < 1)
375 return -EINVAL;
377 spin_lock_irq(&current->sighand->siglock);
378 sigdelset(&current->blocked, sig);
379 if (!current->mm) {
380 /* Kernel threads handle their own signals.
381 Let the signal code know it'll be handled, so
382 that they don't get converted to SIGKILL or
383 just silently dropped */
384 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
386 recalc_sigpending();
387 spin_unlock_irq(&current->sighand->siglock);
388 return 0;
391 EXPORT_SYMBOL(allow_signal);
393 int disallow_signal(int sig)
395 if (!valid_signal(sig) || sig < 1)
396 return -EINVAL;
398 spin_lock_irq(&current->sighand->siglock);
399 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
400 recalc_sigpending();
401 spin_unlock_irq(&current->sighand->siglock);
402 return 0;
405 EXPORT_SYMBOL(disallow_signal);
408 * Put all the gunge required to become a kernel thread without
409 * attached user resources in one place where it belongs.
412 void daemonize(const char *name, ...)
414 va_list args;
415 struct fs_struct *fs;
416 sigset_t blocked;
418 va_start(args, name);
419 vsnprintf(current->comm, sizeof(current->comm), name, args);
420 va_end(args);
423 * If we were started as result of loading a module, close all of the
424 * user space pages. We don't need them, and if we didn't close them
425 * they would be locked into memory.
427 exit_mm(current);
429 * We don't want to have TIF_FREEZE set if the system-wide hibernation
430 * or suspend transition begins right now.
432 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
434 if (current->nsproxy != &init_nsproxy) {
435 get_nsproxy(&init_nsproxy);
436 switch_task_namespaces(current, &init_nsproxy);
438 set_special_pids(&init_struct_pid);
439 proc_clear_tty(current);
441 /* Block and flush all signals */
442 sigfillset(&blocked);
443 sigprocmask(SIG_BLOCK, &blocked, NULL);
444 flush_signals(current);
446 /* Become as one with the init task */
448 exit_fs(current); /* current->fs->count--; */
449 fs = init_task.fs;
450 current->fs = fs;
451 atomic_inc(&fs->count);
453 exit_files(current);
454 current->files = init_task.files;
455 atomic_inc(&current->files->count);
457 reparent_to_kthreadd();
460 EXPORT_SYMBOL(daemonize);
462 static void close_files(struct files_struct * files)
464 int i, j;
465 struct fdtable *fdt;
467 j = 0;
470 * It is safe to dereference the fd table without RCU or
471 * ->file_lock because this is the last reference to the
472 * files structure.
474 fdt = files_fdtable(files);
475 for (;;) {
476 unsigned long set;
477 i = j * __NFDBITS;
478 if (i >= fdt->max_fds)
479 break;
480 set = fdt->open_fds->fds_bits[j++];
481 while (set) {
482 if (set & 1) {
483 struct file * file = xchg(&fdt->fd[i], NULL);
484 if (file) {
485 filp_close(file, files);
486 cond_resched();
489 i++;
490 set >>= 1;
495 struct files_struct *get_files_struct(struct task_struct *task)
497 struct files_struct *files;
499 task_lock(task);
500 files = task->files;
501 if (files)
502 atomic_inc(&files->count);
503 task_unlock(task);
505 return files;
508 void put_files_struct(struct files_struct *files)
510 struct fdtable *fdt;
512 if (atomic_dec_and_test(&files->count)) {
513 close_files(files);
515 * Free the fd and fdset arrays if we expanded them.
516 * If the fdtable was embedded, pass files for freeing
517 * at the end of the RCU grace period. Otherwise,
518 * you can free files immediately.
520 fdt = files_fdtable(files);
521 if (fdt != &files->fdtab)
522 kmem_cache_free(files_cachep, files);
523 free_fdtable(fdt);
527 void reset_files_struct(struct files_struct *files)
529 struct task_struct *tsk = current;
530 struct files_struct *old;
532 old = tsk->files;
533 task_lock(tsk);
534 tsk->files = files;
535 task_unlock(tsk);
536 put_files_struct(old);
539 void exit_files(struct task_struct *tsk)
541 struct files_struct * files = tsk->files;
543 if (files) {
544 task_lock(tsk);
545 tsk->files = NULL;
546 task_unlock(tsk);
547 put_files_struct(files);
551 void put_fs_struct(struct fs_struct *fs)
553 /* No need to hold fs->lock if we are killing it */
554 if (atomic_dec_and_test(&fs->count)) {
555 path_put(&fs->root);
556 path_put(&fs->pwd);
557 kmem_cache_free(fs_cachep, fs);
561 void exit_fs(struct task_struct *tsk)
563 struct fs_struct * fs = tsk->fs;
565 if (fs) {
566 task_lock(tsk);
567 tsk->fs = NULL;
568 task_unlock(tsk);
569 put_fs_struct(fs);
573 EXPORT_SYMBOL_GPL(exit_fs);
575 #ifdef CONFIG_MM_OWNER
577 * Task p is exiting and it owned mm, lets find a new owner for it
579 static inline int
580 mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
583 * If there are other users of the mm and the owner (us) is exiting
584 * we need to find a new owner to take on the responsibility.
586 if (!mm)
587 return 0;
588 if (atomic_read(&mm->mm_users) <= 1)
589 return 0;
590 if (mm->owner != p)
591 return 0;
592 return 1;
595 void mm_update_next_owner(struct mm_struct *mm)
597 struct task_struct *c, *g, *p = current;
599 retry:
600 if (!mm_need_new_owner(mm, p))
601 return;
603 read_lock(&tasklist_lock);
605 * Search in the children
607 list_for_each_entry(c, &p->children, sibling) {
608 if (c->mm == mm)
609 goto assign_new_owner;
613 * Search in the siblings
615 list_for_each_entry(c, &p->parent->children, sibling) {
616 if (c->mm == mm)
617 goto assign_new_owner;
621 * Search through everything else. We should not get
622 * here often
624 do_each_thread(g, c) {
625 if (c->mm == mm)
626 goto assign_new_owner;
627 } while_each_thread(g, c);
629 read_unlock(&tasklist_lock);
630 return;
632 assign_new_owner:
633 BUG_ON(c == p);
634 get_task_struct(c);
636 * The task_lock protects c->mm from changing.
637 * We always want mm->owner->mm == mm
639 task_lock(c);
641 * Delay read_unlock() till we have the task_lock()
642 * to ensure that c does not slip away underneath us
644 read_unlock(&tasklist_lock);
645 if (c->mm != mm) {
646 task_unlock(c);
647 put_task_struct(c);
648 goto retry;
650 cgroup_mm_owner_callbacks(mm->owner, c);
651 mm->owner = c;
652 task_unlock(c);
653 put_task_struct(c);
655 #endif /* CONFIG_MM_OWNER */
658 * Turn us into a lazy TLB process if we
659 * aren't already..
661 static void exit_mm(struct task_struct * tsk)
663 struct mm_struct *mm = tsk->mm;
664 struct core_state *core_state;
666 mm_release(tsk, mm);
667 if (!mm)
668 return;
670 * Serialize with any possible pending coredump.
671 * We must hold mmap_sem around checking core_state
672 * and clearing tsk->mm. The core-inducing thread
673 * will increment ->nr_threads for each thread in the
674 * group with ->mm != NULL.
676 down_read(&mm->mmap_sem);
677 core_state = mm->core_state;
678 if (core_state) {
679 struct core_thread self;
680 up_read(&mm->mmap_sem);
682 self.task = tsk;
683 self.next = xchg(&core_state->dumper.next, &self);
685 * Implies mb(), the result of xchg() must be visible
686 * to core_state->dumper.
688 if (atomic_dec_and_test(&core_state->nr_threads))
689 complete(&core_state->startup);
691 for (;;) {
692 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
693 if (!self.task) /* see coredump_finish() */
694 break;
695 schedule();
697 __set_task_state(tsk, TASK_RUNNING);
698 down_read(&mm->mmap_sem);
700 atomic_inc(&mm->mm_count);
701 BUG_ON(mm != tsk->active_mm);
702 /* more a memory barrier than a real lock */
703 task_lock(tsk);
704 tsk->mm = NULL;
705 up_read(&mm->mmap_sem);
706 enter_lazy_tlb(mm, current);
707 /* We don't want this task to be frozen prematurely */
708 clear_freeze_flag(tsk);
709 task_unlock(tsk);
710 mm_update_next_owner(mm);
711 mmput(mm);
715 * Return nonzero if @parent's children should reap themselves.
717 * Called with write_lock_irq(&tasklist_lock) held.
719 static int ignoring_children(struct task_struct *parent)
721 int ret;
722 struct sighand_struct *psig = parent->sighand;
723 unsigned long flags;
724 spin_lock_irqsave(&psig->siglock, flags);
725 ret = (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
726 (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT));
727 spin_unlock_irqrestore(&psig->siglock, flags);
728 return ret;
732 * Detach all tasks we were using ptrace on.
733 * Any that need to be release_task'd are put on the @dead list.
735 * Called with write_lock(&tasklist_lock) held.
737 static void ptrace_exit(struct task_struct *parent, struct list_head *dead)
739 struct task_struct *p, *n;
740 int ign = -1;
742 list_for_each_entry_safe(p, n, &parent->ptraced, ptrace_entry) {
743 __ptrace_unlink(p);
745 if (p->exit_state != EXIT_ZOMBIE)
746 continue;
749 * If it's a zombie, our attachedness prevented normal
750 * parent notification or self-reaping. Do notification
751 * now if it would have happened earlier. If it should
752 * reap itself, add it to the @dead list. We can't call
753 * release_task() here because we already hold tasklist_lock.
755 * If it's our own child, there is no notification to do.
756 * But if our normal children self-reap, then this child
757 * was prevented by ptrace and we must reap it now.
759 if (!task_detached(p) && thread_group_empty(p)) {
760 if (!same_thread_group(p->real_parent, parent))
761 do_notify_parent(p, p->exit_signal);
762 else {
763 if (ign < 0)
764 ign = ignoring_children(parent);
765 if (ign)
766 p->exit_signal = -1;
770 if (task_detached(p)) {
772 * Mark it as in the process of being reaped.
774 p->exit_state = EXIT_DEAD;
775 list_add(&p->ptrace_entry, dead);
781 * Finish up exit-time ptrace cleanup.
783 * Called without locks.
785 static void ptrace_exit_finish(struct task_struct *parent,
786 struct list_head *dead)
788 struct task_struct *p, *n;
790 BUG_ON(!list_empty(&parent->ptraced));
792 list_for_each_entry_safe(p, n, dead, ptrace_entry) {
793 list_del_init(&p->ptrace_entry);
794 release_task(p);
798 static void reparent_thread(struct task_struct *p, struct task_struct *father)
800 if (p->pdeath_signal)
801 /* We already hold the tasklist_lock here. */
802 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
804 list_move_tail(&p->sibling, &p->real_parent->children);
806 /* If this is a threaded reparent there is no need to
807 * notify anyone anything has happened.
809 if (same_thread_group(p->real_parent, father))
810 return;
812 /* We don't want people slaying init. */
813 if (!task_detached(p))
814 p->exit_signal = SIGCHLD;
816 /* If we'd notified the old parent about this child's death,
817 * also notify the new parent.
819 if (!ptrace_reparented(p) &&
820 p->exit_state == EXIT_ZOMBIE &&
821 !task_detached(p) && thread_group_empty(p))
822 do_notify_parent(p, p->exit_signal);
824 kill_orphaned_pgrp(p, father);
828 * When we die, we re-parent all our children.
829 * Try to give them to another thread in our thread
830 * group, and if no such member exists, give it to
831 * the child reaper process (ie "init") in our pid
832 * space.
834 static void forget_original_parent(struct task_struct *father)
836 struct task_struct *p, *n, *reaper = father;
837 LIST_HEAD(ptrace_dead);
839 write_lock_irq(&tasklist_lock);
842 * First clean up ptrace if we were using it.
844 ptrace_exit(father, &ptrace_dead);
846 do {
847 reaper = next_thread(reaper);
848 if (reaper == father) {
849 reaper = task_child_reaper(father);
850 break;
852 } while (reaper->flags & PF_EXITING);
854 list_for_each_entry_safe(p, n, &father->children, sibling) {
855 p->real_parent = reaper;
856 if (p->parent == father) {
857 BUG_ON(p->ptrace);
858 p->parent = p->real_parent;
860 reparent_thread(p, father);
863 write_unlock_irq(&tasklist_lock);
864 BUG_ON(!list_empty(&father->children));
866 ptrace_exit_finish(father, &ptrace_dead);
870 * Send signals to all our closest relatives so that they know
871 * to properly mourn us..
873 static void exit_notify(struct task_struct *tsk, int group_dead)
875 int signal;
876 void *cookie;
879 * This does two things:
881 * A. Make init inherit all the child processes
882 * B. Check to see if any process groups have become orphaned
883 * as a result of our exiting, and if they have any stopped
884 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
886 forget_original_parent(tsk);
887 exit_task_namespaces(tsk);
889 write_lock_irq(&tasklist_lock);
890 if (group_dead)
891 kill_orphaned_pgrp(tsk->group_leader, NULL);
893 /* Let father know we died
895 * Thread signals are configurable, but you aren't going to use
896 * that to send signals to arbitary processes.
897 * That stops right now.
899 * If the parent exec id doesn't match the exec id we saved
900 * when we started then we know the parent has changed security
901 * domain.
903 * If our self_exec id doesn't match our parent_exec_id then
904 * we have changed execution domain as these two values started
905 * the same after a fork.
907 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
908 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
909 tsk->self_exec_id != tsk->parent_exec_id) &&
910 !capable(CAP_KILL))
911 tsk->exit_signal = SIGCHLD;
913 signal = tracehook_notify_death(tsk, &cookie, group_dead);
914 if (signal > 0)
915 signal = do_notify_parent(tsk, signal);
917 tsk->exit_state = signal < 0 ? EXIT_DEAD : EXIT_ZOMBIE;
919 /* mt-exec, de_thread() is waiting for us */
920 if (thread_group_leader(tsk) &&
921 tsk->signal->notify_count < 0 &&
922 tsk->signal->group_exit_task)
923 wake_up_process(tsk->signal->group_exit_task);
925 write_unlock_irq(&tasklist_lock);
927 tracehook_report_death(tsk, signal, cookie, group_dead);
929 /* If the process is dead, release it - nobody will wait for it */
930 if (signal < 0)
931 release_task(tsk);
934 #ifdef CONFIG_DEBUG_STACK_USAGE
935 static void check_stack_usage(void)
937 static DEFINE_SPINLOCK(low_water_lock);
938 static int lowest_to_date = THREAD_SIZE;
939 unsigned long *n = end_of_stack(current);
940 unsigned long free;
942 while (*n == 0)
943 n++;
944 free = (unsigned long)n - (unsigned long)end_of_stack(current);
946 if (free >= lowest_to_date)
947 return;
949 spin_lock(&low_water_lock);
950 if (free < lowest_to_date) {
951 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
952 "left\n",
953 current->comm, free);
954 lowest_to_date = free;
956 spin_unlock(&low_water_lock);
958 #else
959 static inline void check_stack_usage(void) {}
960 #endif
962 static inline void exit_child_reaper(struct task_struct *tsk)
964 if (likely(tsk->group_leader != task_child_reaper(tsk)))
965 return;
967 if (tsk->nsproxy->pid_ns == &init_pid_ns)
968 panic("Attempted to kill init!");
971 * @tsk is the last thread in the 'cgroup-init' and is exiting.
972 * Terminate all remaining processes in the namespace and reap them
973 * before exiting @tsk.
975 * Note that @tsk (last thread of cgroup-init) may not necessarily
976 * be the child-reaper (i.e main thread of cgroup-init) of the
977 * namespace i.e the child_reaper may have already exited.
979 * Even after a child_reaper exits, we let it inherit orphaned children,
980 * because, pid_ns->child_reaper remains valid as long as there is
981 * at least one living sub-thread in the cgroup init.
983 * This living sub-thread of the cgroup-init will be notified when
984 * a child inherited by the 'child-reaper' exits (do_notify_parent()
985 * uses __group_send_sig_info()). Further, when reaping child processes,
986 * do_wait() iterates over children of all living sub threads.
988 * i.e even though 'child_reaper' thread is listed as the parent of the
989 * orphaned children, any living sub-thread in the cgroup-init can
990 * perform the role of the child_reaper.
992 zap_pid_ns_processes(tsk->nsproxy->pid_ns);
995 NORET_TYPE void do_exit(long code)
997 struct task_struct *tsk = current;
998 int group_dead;
1000 profile_task_exit(tsk);
1002 WARN_ON(atomic_read(&tsk->fs_excl));
1004 if (unlikely(in_interrupt()))
1005 panic("Aiee, killing interrupt handler!");
1006 if (unlikely(!tsk->pid))
1007 panic("Attempted to kill the idle task!");
1009 tracehook_report_exit(&code);
1012 * We're taking recursive faults here in do_exit. Safest is to just
1013 * leave this task alone and wait for reboot.
1015 if (unlikely(tsk->flags & PF_EXITING)) {
1016 printk(KERN_ALERT
1017 "Fixing recursive fault but reboot is needed!\n");
1019 * We can do this unlocked here. The futex code uses
1020 * this flag just to verify whether the pi state
1021 * cleanup has been done or not. In the worst case it
1022 * loops once more. We pretend that the cleanup was
1023 * done as there is no way to return. Either the
1024 * OWNER_DIED bit is set by now or we push the blocked
1025 * task into the wait for ever nirwana as well.
1027 tsk->flags |= PF_EXITPIDONE;
1028 if (tsk->io_context)
1029 exit_io_context();
1030 set_current_state(TASK_UNINTERRUPTIBLE);
1031 schedule();
1034 exit_signals(tsk); /* sets PF_EXITING */
1036 * tsk->flags are checked in the futex code to protect against
1037 * an exiting task cleaning up the robust pi futexes.
1039 smp_mb();
1040 spin_unlock_wait(&tsk->pi_lock);
1042 if (unlikely(in_atomic()))
1043 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
1044 current->comm, task_pid_nr(current),
1045 preempt_count());
1047 acct_update_integrals(tsk);
1048 if (tsk->mm) {
1049 update_hiwater_rss(tsk->mm);
1050 update_hiwater_vm(tsk->mm);
1052 group_dead = atomic_dec_and_test(&tsk->signal->live);
1053 if (group_dead) {
1054 exit_child_reaper(tsk);
1055 hrtimer_cancel(&tsk->signal->real_timer);
1056 exit_itimers(tsk->signal);
1058 acct_collect(code, group_dead);
1059 #ifdef CONFIG_FUTEX
1060 if (unlikely(tsk->robust_list))
1061 exit_robust_list(tsk);
1062 #ifdef CONFIG_COMPAT
1063 if (unlikely(tsk->compat_robust_list))
1064 compat_exit_robust_list(tsk);
1065 #endif
1066 #endif
1067 if (group_dead)
1068 tty_audit_exit();
1069 if (unlikely(tsk->audit_context))
1070 audit_free(tsk);
1072 tsk->exit_code = code;
1073 taskstats_exit(tsk, group_dead);
1075 exit_mm(tsk);
1077 if (group_dead)
1078 acct_process();
1079 exit_sem(tsk);
1080 exit_files(tsk);
1081 exit_fs(tsk);
1082 check_stack_usage();
1083 exit_thread();
1084 cgroup_exit(tsk, 1);
1085 exit_keys(tsk);
1087 if (group_dead && tsk->signal->leader)
1088 disassociate_ctty(1);
1090 module_put(task_thread_info(tsk)->exec_domain->module);
1091 if (tsk->binfmt)
1092 module_put(tsk->binfmt->module);
1094 proc_exit_connector(tsk);
1095 exit_notify(tsk, group_dead);
1096 #ifdef CONFIG_NUMA
1097 mpol_put(tsk->mempolicy);
1098 tsk->mempolicy = NULL;
1099 #endif
1100 #ifdef CONFIG_FUTEX
1102 * This must happen late, after the PID is not
1103 * hashed anymore:
1105 if (unlikely(!list_empty(&tsk->pi_state_list)))
1106 exit_pi_state_list(tsk);
1107 if (unlikely(current->pi_state_cache))
1108 kfree(current->pi_state_cache);
1109 #endif
1111 * Make sure we are holding no locks:
1113 debug_check_no_locks_held(tsk);
1115 * We can do this unlocked here. The futex code uses this flag
1116 * just to verify whether the pi state cleanup has been done
1117 * or not. In the worst case it loops once more.
1119 tsk->flags |= PF_EXITPIDONE;
1121 if (tsk->io_context)
1122 exit_io_context();
1124 if (tsk->splice_pipe)
1125 __free_pipe_info(tsk->splice_pipe);
1127 preempt_disable();
1128 /* causes final put_task_struct in finish_task_switch(). */
1129 tsk->state = TASK_DEAD;
1131 schedule();
1132 BUG();
1133 /* Avoid "noreturn function does return". */
1134 for (;;)
1135 cpu_relax(); /* For when BUG is null */
1138 EXPORT_SYMBOL_GPL(do_exit);
1140 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1142 if (comp)
1143 complete(comp);
1145 do_exit(code);
1148 EXPORT_SYMBOL(complete_and_exit);
1150 asmlinkage long sys_exit(int error_code)
1152 do_exit((error_code&0xff)<<8);
1156 * Take down every thread in the group. This is called by fatal signals
1157 * as well as by sys_exit_group (below).
1159 NORET_TYPE void
1160 do_group_exit(int exit_code)
1162 struct signal_struct *sig = current->signal;
1164 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1166 if (signal_group_exit(sig))
1167 exit_code = sig->group_exit_code;
1168 else if (!thread_group_empty(current)) {
1169 struct sighand_struct *const sighand = current->sighand;
1170 spin_lock_irq(&sighand->siglock);
1171 if (signal_group_exit(sig))
1172 /* Another thread got here before we took the lock. */
1173 exit_code = sig->group_exit_code;
1174 else {
1175 sig->group_exit_code = exit_code;
1176 sig->flags = SIGNAL_GROUP_EXIT;
1177 zap_other_threads(current);
1179 spin_unlock_irq(&sighand->siglock);
1182 do_exit(exit_code);
1183 /* NOTREACHED */
1187 * this kills every thread in the thread group. Note that any externally
1188 * wait4()-ing process will get the correct exit code - even if this
1189 * thread is not the thread group leader.
1191 asmlinkage void sys_exit_group(int error_code)
1193 do_group_exit((error_code & 0xff) << 8);
1196 static struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1198 struct pid *pid = NULL;
1199 if (type == PIDTYPE_PID)
1200 pid = task->pids[type].pid;
1201 else if (type < PIDTYPE_MAX)
1202 pid = task->group_leader->pids[type].pid;
1203 return pid;
1206 static int eligible_child(enum pid_type type, struct pid *pid, int options,
1207 struct task_struct *p)
1209 int err;
1211 if (type < PIDTYPE_MAX) {
1212 if (task_pid_type(p, type) != pid)
1213 return 0;
1216 /* Wait for all children (clone and not) if __WALL is set;
1217 * otherwise, wait for clone children *only* if __WCLONE is
1218 * set; otherwise, wait for non-clone children *only*. (Note:
1219 * A "clone" child here is one that reports to its parent
1220 * using a signal other than SIGCHLD.) */
1221 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1222 && !(options & __WALL))
1223 return 0;
1225 err = security_task_wait(p);
1226 if (err)
1227 return err;
1229 return 1;
1232 static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
1233 int why, int status,
1234 struct siginfo __user *infop,
1235 struct rusage __user *rusagep)
1237 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1239 put_task_struct(p);
1240 if (!retval)
1241 retval = put_user(SIGCHLD, &infop->si_signo);
1242 if (!retval)
1243 retval = put_user(0, &infop->si_errno);
1244 if (!retval)
1245 retval = put_user((short)why, &infop->si_code);
1246 if (!retval)
1247 retval = put_user(pid, &infop->si_pid);
1248 if (!retval)
1249 retval = put_user(uid, &infop->si_uid);
1250 if (!retval)
1251 retval = put_user(status, &infop->si_status);
1252 if (!retval)
1253 retval = pid;
1254 return retval;
1258 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. 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.
1263 static int wait_task_zombie(struct task_struct *p, int options,
1264 struct siginfo __user *infop,
1265 int __user *stat_addr, struct rusage __user *ru)
1267 unsigned long state;
1268 int retval, status, traced;
1269 pid_t pid = task_pid_vnr(p);
1271 if (!likely(options & WEXITED))
1272 return 0;
1274 if (unlikely(options & WNOWAIT)) {
1275 uid_t uid = p->uid;
1276 int exit_code = p->exit_code;
1277 int why, status;
1279 get_task_struct(p);
1280 read_unlock(&tasklist_lock);
1281 if ((exit_code & 0x7f) == 0) {
1282 why = CLD_EXITED;
1283 status = exit_code >> 8;
1284 } else {
1285 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1286 status = exit_code & 0x7f;
1288 return wait_noreap_copyout(p, pid, uid, why,
1289 status, infop, ru);
1293 * Try to move the task's state to DEAD
1294 * only one thread is allowed to do this:
1296 state = xchg(&p->exit_state, EXIT_DEAD);
1297 if (state != EXIT_ZOMBIE) {
1298 BUG_ON(state != EXIT_DEAD);
1299 return 0;
1302 traced = ptrace_reparented(p);
1304 if (likely(!traced)) {
1305 struct signal_struct *psig;
1306 struct signal_struct *sig;
1309 * The resource counters for the group leader are in its
1310 * own task_struct. Those for dead threads in the group
1311 * are in its signal_struct, as are those for the child
1312 * processes it has previously reaped. All these
1313 * accumulate in the parent's signal_struct c* fields.
1315 * We don't bother to take a lock here to protect these
1316 * p->signal fields, because they are only touched by
1317 * __exit_signal, which runs with tasklist_lock
1318 * write-locked anyway, and so is excluded here. We do
1319 * need to protect the access to p->parent->signal fields,
1320 * as other threads in the parent group can be right
1321 * here reaping other children at the same time.
1323 spin_lock_irq(&p->parent->sighand->siglock);
1324 psig = p->parent->signal;
1325 sig = p->signal;
1326 psig->cutime =
1327 cputime_add(psig->cutime,
1328 cputime_add(p->utime,
1329 cputime_add(sig->utime,
1330 sig->cutime)));
1331 psig->cstime =
1332 cputime_add(psig->cstime,
1333 cputime_add(p->stime,
1334 cputime_add(sig->stime,
1335 sig->cstime)));
1336 psig->cgtime =
1337 cputime_add(psig->cgtime,
1338 cputime_add(p->gtime,
1339 cputime_add(sig->gtime,
1340 sig->cgtime)));
1341 psig->cmin_flt +=
1342 p->min_flt + sig->min_flt + sig->cmin_flt;
1343 psig->cmaj_flt +=
1344 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1345 psig->cnvcsw +=
1346 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1347 psig->cnivcsw +=
1348 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1349 psig->cinblock +=
1350 task_io_get_inblock(p) +
1351 sig->inblock + sig->cinblock;
1352 psig->coublock +=
1353 task_io_get_oublock(p) +
1354 sig->oublock + sig->coublock;
1355 task_io_accounting_add(&psig->ioac, &p->ioac);
1356 task_io_accounting_add(&psig->ioac, &sig->ioac);
1357 spin_unlock_irq(&p->parent->sighand->siglock);
1361 * Now we are sure this task is interesting, and no other
1362 * thread can reap it because we set its state to EXIT_DEAD.
1364 read_unlock(&tasklist_lock);
1366 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1367 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1368 ? p->signal->group_exit_code : p->exit_code;
1369 if (!retval && stat_addr)
1370 retval = put_user(status, stat_addr);
1371 if (!retval && infop)
1372 retval = put_user(SIGCHLD, &infop->si_signo);
1373 if (!retval && infop)
1374 retval = put_user(0, &infop->si_errno);
1375 if (!retval && infop) {
1376 int why;
1378 if ((status & 0x7f) == 0) {
1379 why = CLD_EXITED;
1380 status >>= 8;
1381 } else {
1382 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1383 status &= 0x7f;
1385 retval = put_user((short)why, &infop->si_code);
1386 if (!retval)
1387 retval = put_user(status, &infop->si_status);
1389 if (!retval && infop)
1390 retval = put_user(pid, &infop->si_pid);
1391 if (!retval && infop)
1392 retval = put_user(p->uid, &infop->si_uid);
1393 if (!retval)
1394 retval = pid;
1396 if (traced) {
1397 write_lock_irq(&tasklist_lock);
1398 /* We dropped tasklist, ptracer could die and untrace */
1399 ptrace_unlink(p);
1401 * If this is not a detached task, notify the parent.
1402 * If it's still not detached after that, don't release
1403 * it now.
1405 if (!task_detached(p)) {
1406 do_notify_parent(p, p->exit_signal);
1407 if (!task_detached(p)) {
1408 p->exit_state = EXIT_ZOMBIE;
1409 p = NULL;
1412 write_unlock_irq(&tasklist_lock);
1414 if (p != NULL)
1415 release_task(p);
1417 return retval;
1421 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1422 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1423 * the lock and this task is uninteresting. If we return nonzero, we have
1424 * released the lock and the system call should return.
1426 static int wait_task_stopped(int ptrace, struct task_struct *p,
1427 int options, struct siginfo __user *infop,
1428 int __user *stat_addr, struct rusage __user *ru)
1430 int retval, exit_code, why;
1431 uid_t uid = 0; /* unneeded, required by compiler */
1432 pid_t pid;
1434 if (!(options & WUNTRACED))
1435 return 0;
1437 exit_code = 0;
1438 spin_lock_irq(&p->sighand->siglock);
1440 if (unlikely(!task_is_stopped_or_traced(p)))
1441 goto unlock_sig;
1443 if (!ptrace && p->signal->group_stop_count > 0)
1445 * A group stop is in progress and this is the group leader.
1446 * We won't report until all threads have stopped.
1448 goto unlock_sig;
1450 exit_code = p->exit_code;
1451 if (!exit_code)
1452 goto unlock_sig;
1454 if (!unlikely(options & WNOWAIT))
1455 p->exit_code = 0;
1457 uid = p->uid;
1458 unlock_sig:
1459 spin_unlock_irq(&p->sighand->siglock);
1460 if (!exit_code)
1461 return 0;
1464 * Now we are pretty sure this task is interesting.
1465 * Make sure it doesn't get reaped out from under us while we
1466 * give up the lock and then examine it below. We don't want to
1467 * keep holding onto the tasklist_lock while we call getrusage and
1468 * possibly take page faults for user memory.
1470 get_task_struct(p);
1471 pid = task_pid_vnr(p);
1472 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1473 read_unlock(&tasklist_lock);
1475 if (unlikely(options & WNOWAIT))
1476 return wait_noreap_copyout(p, pid, uid,
1477 why, exit_code,
1478 infop, ru);
1480 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1481 if (!retval && stat_addr)
1482 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1483 if (!retval && infop)
1484 retval = put_user(SIGCHLD, &infop->si_signo);
1485 if (!retval && infop)
1486 retval = put_user(0, &infop->si_errno);
1487 if (!retval && infop)
1488 retval = put_user((short)why, &infop->si_code);
1489 if (!retval && infop)
1490 retval = put_user(exit_code, &infop->si_status);
1491 if (!retval && infop)
1492 retval = put_user(pid, &infop->si_pid);
1493 if (!retval && infop)
1494 retval = put_user(uid, &infop->si_uid);
1495 if (!retval)
1496 retval = pid;
1497 put_task_struct(p);
1499 BUG_ON(!retval);
1500 return retval;
1504 * Handle do_wait work for one task in a live, non-stopped state.
1505 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1506 * the lock and this task is uninteresting. If we return nonzero, we have
1507 * released the lock and the system call should return.
1509 static int wait_task_continued(struct task_struct *p, int options,
1510 struct siginfo __user *infop,
1511 int __user *stat_addr, struct rusage __user *ru)
1513 int retval;
1514 pid_t pid;
1515 uid_t uid;
1517 if (!unlikely(options & WCONTINUED))
1518 return 0;
1520 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1521 return 0;
1523 spin_lock_irq(&p->sighand->siglock);
1524 /* Re-check with the lock held. */
1525 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1526 spin_unlock_irq(&p->sighand->siglock);
1527 return 0;
1529 if (!unlikely(options & WNOWAIT))
1530 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1531 spin_unlock_irq(&p->sighand->siglock);
1533 pid = task_pid_vnr(p);
1534 uid = p->uid;
1535 get_task_struct(p);
1536 read_unlock(&tasklist_lock);
1538 if (!infop) {
1539 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1540 put_task_struct(p);
1541 if (!retval && stat_addr)
1542 retval = put_user(0xffff, stat_addr);
1543 if (!retval)
1544 retval = pid;
1545 } else {
1546 retval = wait_noreap_copyout(p, pid, uid,
1547 CLD_CONTINUED, SIGCONT,
1548 infop, ru);
1549 BUG_ON(retval == 0);
1552 return retval;
1556 * Consider @p for a wait by @parent.
1558 * -ECHILD should be in *@notask_error before the first call.
1559 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1560 * Returns zero if the search for a child should continue;
1561 * then *@notask_error is 0 if @p is an eligible child,
1562 * or another error from security_task_wait(), or still -ECHILD.
1564 static int wait_consider_task(struct task_struct *parent, int ptrace,
1565 struct task_struct *p, int *notask_error,
1566 enum pid_type type, struct pid *pid, int options,
1567 struct siginfo __user *infop,
1568 int __user *stat_addr, struct rusage __user *ru)
1570 int ret = eligible_child(type, pid, options, p);
1571 if (!ret)
1572 return ret;
1574 if (unlikely(ret < 0)) {
1576 * If we have not yet seen any eligible child,
1577 * then let this error code replace -ECHILD.
1578 * A permission error will give the user a clue
1579 * to look for security policy problems, rather
1580 * than for mysterious wait bugs.
1582 if (*notask_error)
1583 *notask_error = ret;
1586 if (likely(!ptrace) && unlikely(p->ptrace)) {
1588 * This child is hidden by ptrace.
1589 * We aren't allowed to see it now, but eventually we will.
1591 *notask_error = 0;
1592 return 0;
1595 if (p->exit_state == EXIT_DEAD)
1596 return 0;
1599 * We don't reap group leaders with subthreads.
1601 if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
1602 return wait_task_zombie(p, options, infop, stat_addr, ru);
1605 * It's stopped or running now, so it might
1606 * later continue, exit, or stop again.
1608 *notask_error = 0;
1610 if (task_is_stopped_or_traced(p))
1611 return wait_task_stopped(ptrace, p, options,
1612 infop, stat_addr, ru);
1614 return wait_task_continued(p, options, infop, stat_addr, ru);
1618 * Do the work of do_wait() for one thread in the group, @tsk.
1620 * -ECHILD should be in *@notask_error before the first call.
1621 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1622 * Returns zero if the search for a child should continue; then
1623 * *@notask_error is 0 if there were any eligible children,
1624 * or another error from security_task_wait(), or still -ECHILD.
1626 static int do_wait_thread(struct task_struct *tsk, int *notask_error,
1627 enum pid_type type, struct pid *pid, int options,
1628 struct siginfo __user *infop, int __user *stat_addr,
1629 struct rusage __user *ru)
1631 struct task_struct *p;
1633 list_for_each_entry(p, &tsk->children, sibling) {
1635 * Do not consider detached threads.
1637 if (!task_detached(p)) {
1638 int ret = wait_consider_task(tsk, 0, p, notask_error,
1639 type, pid, options,
1640 infop, stat_addr, ru);
1641 if (ret)
1642 return ret;
1646 return 0;
1649 static int ptrace_do_wait(struct task_struct *tsk, int *notask_error,
1650 enum pid_type type, struct pid *pid, int options,
1651 struct siginfo __user *infop, int __user *stat_addr,
1652 struct rusage __user *ru)
1654 struct task_struct *p;
1657 * Traditionally we see ptrace'd stopped tasks regardless of options.
1659 options |= WUNTRACED;
1661 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1662 int ret = wait_consider_task(tsk, 1, p, notask_error,
1663 type, pid, options,
1664 infop, stat_addr, ru);
1665 if (ret)
1666 return ret;
1669 return 0;
1672 static long do_wait(enum pid_type type, struct pid *pid, int options,
1673 struct siginfo __user *infop, int __user *stat_addr,
1674 struct rusage __user *ru)
1676 DECLARE_WAITQUEUE(wait, current);
1677 struct task_struct *tsk;
1678 int retval;
1680 add_wait_queue(&current->signal->wait_chldexit,&wait);
1681 repeat:
1683 * If there is nothing that can match our critiera just get out.
1684 * We will clear @retval to zero if we see any child that might later
1685 * match our criteria, even if we are not able to reap it yet.
1687 retval = -ECHILD;
1688 if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type])))
1689 goto end;
1691 current->state = TASK_INTERRUPTIBLE;
1692 read_lock(&tasklist_lock);
1693 tsk = current;
1694 do {
1695 int tsk_result = do_wait_thread(tsk, &retval,
1696 type, pid, options,
1697 infop, stat_addr, ru);
1698 if (!tsk_result)
1699 tsk_result = ptrace_do_wait(tsk, &retval,
1700 type, pid, options,
1701 infop, stat_addr, ru);
1702 if (tsk_result) {
1704 * tasklist_lock is unlocked and we have a final result.
1706 retval = tsk_result;
1707 goto end;
1710 if (options & __WNOTHREAD)
1711 break;
1712 tsk = next_thread(tsk);
1713 BUG_ON(tsk->signal != current->signal);
1714 } while (tsk != current);
1715 read_unlock(&tasklist_lock);
1717 if (!retval && !(options & WNOHANG)) {
1718 retval = -ERESTARTSYS;
1719 if (!signal_pending(current)) {
1720 schedule();
1721 goto repeat;
1725 end:
1726 current->state = TASK_RUNNING;
1727 remove_wait_queue(&current->signal->wait_chldexit,&wait);
1728 if (infop) {
1729 if (retval > 0)
1730 retval = 0;
1731 else {
1733 * For a WNOHANG return, clear out all the fields
1734 * we would set so the user can easily tell the
1735 * difference.
1737 if (!retval)
1738 retval = put_user(0, &infop->si_signo);
1739 if (!retval)
1740 retval = put_user(0, &infop->si_errno);
1741 if (!retval)
1742 retval = put_user(0, &infop->si_code);
1743 if (!retval)
1744 retval = put_user(0, &infop->si_pid);
1745 if (!retval)
1746 retval = put_user(0, &infop->si_uid);
1747 if (!retval)
1748 retval = put_user(0, &infop->si_status);
1751 return retval;
1754 asmlinkage long sys_waitid(int which, pid_t upid,
1755 struct siginfo __user *infop, int options,
1756 struct rusage __user *ru)
1758 struct pid *pid = NULL;
1759 enum pid_type type;
1760 long ret;
1762 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1763 return -EINVAL;
1764 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1765 return -EINVAL;
1767 switch (which) {
1768 case P_ALL:
1769 type = PIDTYPE_MAX;
1770 break;
1771 case P_PID:
1772 type = PIDTYPE_PID;
1773 if (upid <= 0)
1774 return -EINVAL;
1775 break;
1776 case P_PGID:
1777 type = PIDTYPE_PGID;
1778 if (upid <= 0)
1779 return -EINVAL;
1780 break;
1781 default:
1782 return -EINVAL;
1785 if (type < PIDTYPE_MAX)
1786 pid = find_get_pid(upid);
1787 ret = do_wait(type, pid, options, infop, NULL, ru);
1788 put_pid(pid);
1790 /* avoid REGPARM breakage on x86: */
1791 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
1792 return ret;
1795 asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr,
1796 int options, struct rusage __user *ru)
1798 struct pid *pid = NULL;
1799 enum pid_type type;
1800 long ret;
1802 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1803 __WNOTHREAD|__WCLONE|__WALL))
1804 return -EINVAL;
1806 if (upid == -1)
1807 type = PIDTYPE_MAX;
1808 else if (upid < 0) {
1809 type = PIDTYPE_PGID;
1810 pid = find_get_pid(-upid);
1811 } else if (upid == 0) {
1812 type = PIDTYPE_PGID;
1813 pid = get_pid(task_pgrp(current));
1814 } else /* upid > 0 */ {
1815 type = PIDTYPE_PID;
1816 pid = find_get_pid(upid);
1819 ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru);
1820 put_pid(pid);
1822 /* avoid REGPARM breakage on x86: */
1823 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
1824 return ret;
1827 #ifdef __ARCH_WANT_SYS_WAITPID
1830 * sys_waitpid() remains for compatibility. waitpid() should be
1831 * implemented by calling sys_wait4() from libc.a.
1833 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1835 return sys_wait4(pid, stat_addr, options, NULL);
1838 #endif