4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/config.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/security.h>
18 #include <linux/acct.h>
19 #include <linux/file.h>
20 #include <linux/binfmts.h>
21 #include <linux/ptrace.h>
22 #include <linux/profile.h>
23 #include <linux/mount.h>
24 #include <linux/proc_fs.h>
25 #include <linux/mempolicy.h>
26 #include <linux/ckrm.h>
27 #include <linux/ckrm_tsk.h>
28 #include <linux/vs_limit.h>
30 #include <asm/uaccess.h>
31 #include <asm/unistd.h>
32 #include <asm/pgtable.h>
33 #include <asm/mmu_context.h>
35 extern void sem_exit (void);
36 extern struct task_struct *child_reaper;
38 int getrusage(struct task_struct *, int, struct rusage __user *);
40 static void __unhash_process(struct task_struct *p)
43 /* tasklist_lock is held, is this sufficient? */
45 atomic_dec(&p->vx_info->cacct.nr_threads);
46 atomic_dec(&p->vx_info->limit.rcur[RLIMIT_NPROC]);
48 detach_pid(p, PIDTYPE_PID);
49 detach_pid(p, PIDTYPE_TGID);
50 if (thread_group_leader(p)) {
51 detach_pid(p, PIDTYPE_PGID);
52 detach_pid(p, PIDTYPE_SID);
54 __get_cpu_var(process_counts)--;
60 void release_task(struct task_struct * p)
64 struct dentry *proc_dentry;
67 BUG_ON(p->state < TASK_ZOMBIE);
69 atomic_dec(&p->user->processes);
70 spin_lock(&p->proc_lock);
71 proc_dentry = proc_pid_unhash(p);
72 write_lock_irq(&tasklist_lock);
73 if (unlikely(p->ptrace))
75 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
81 * If we are the last non-leader member of the thread
82 * group, and the leader is zombie, then notify the
83 * group leader's parent process. (if it wants notification.)
86 leader = p->group_leader;
87 if (leader != p && thread_group_empty(leader) && leader->state == TASK_ZOMBIE) {
88 BUG_ON(leader->exit_signal == -1);
89 do_notify_parent(leader, leader->exit_signal);
91 * If we were the last child thread and the leader has
92 * exited already, and the leader's parent ignores SIGCHLD,
93 * then we are the one who should release the leader.
95 * do_notify_parent() will have marked it self-reaping in
98 zap_leader = (leader->exit_signal == -1);
101 p->parent->cutime += p->utime + p->cutime;
102 p->parent->cstime += p->stime + p->cstime;
103 p->parent->cmin_flt += p->min_flt + p->cmin_flt;
104 p->parent->cmaj_flt += p->maj_flt + p->cmaj_flt;
105 p->parent->cnvcsw += p->nvcsw + p->cnvcsw;
106 p->parent->cnivcsw += p->nivcsw + p->cnivcsw;
108 write_unlock_irq(&tasklist_lock);
109 spin_unlock(&p->proc_lock);
115 if (unlikely(zap_leader))
119 /* we are using it only for SMP init */
121 void unhash_process(struct task_struct *p)
123 struct dentry *proc_dentry;
125 spin_lock(&p->proc_lock);
126 proc_dentry = proc_pid_unhash(p);
127 write_lock_irq(&tasklist_lock);
129 write_unlock_irq(&tasklist_lock);
130 spin_unlock(&p->proc_lock);
131 proc_pid_flush(proc_dentry);
135 * This checks not only the pgrp, but falls back on the pid if no
136 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
139 int session_of_pgrp(int pgrp)
141 struct task_struct *p;
146 read_lock(&tasklist_lock);
147 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid)
148 if (p->signal->session > 0) {
149 sid = p->signal->session;
152 p = find_task_by_pid(pgrp);
154 sid = p->signal->session;
156 read_unlock(&tasklist_lock);
162 * Determine if a process group is "orphaned", according to the POSIX
163 * definition in 2.2.2.52. Orphaned process groups are not to be affected
164 * by terminal-generated stop signals. Newly orphaned process groups are
165 * to receive a SIGHUP and a SIGCONT.
167 * "I ask you, have you ever known what it is to be an orphan?"
169 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
171 struct task_struct *p;
176 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
177 if (p == ignored_task
178 || p->state >= TASK_ZOMBIE
179 || p->real_parent->pid == 1)
181 if (process_group(p->real_parent) != pgrp
182 && p->real_parent->signal->session == p->signal->session) {
187 return ret; /* (sighing) "Often!" */
190 int is_orphaned_pgrp(int pgrp)
194 read_lock(&tasklist_lock);
195 retval = will_become_orphaned_pgrp(pgrp, NULL);
196 read_unlock(&tasklist_lock);
201 static inline int has_stopped_jobs(int pgrp)
204 struct task_struct *p;
208 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
209 if (p->state != TASK_STOPPED)
212 /* If p is stopped by a debugger on a signal that won't
213 stop it, then don't count p as stopped. This isn't
214 perfect but it's a good approximation. */
215 if (unlikely (p->ptrace)
216 && p->exit_code != SIGSTOP
217 && p->exit_code != SIGTSTP
218 && p->exit_code != SIGTTOU
219 && p->exit_code != SIGTTIN)
229 * reparent_to_init() - Reparent the calling kernel thread to the init task.
231 * If a kernel thread is launched as a result of a system call, or if
232 * it ever exits, it should generally reparent itself to init so that
233 * it is correctly cleaned up on exit.
235 * The various task state such as scheduling policy and priority may have
236 * been inherited from a user process, so we reset them to sane values here.
238 * NOTE that reparent_to_init() gives the caller full capabilities.
240 void reparent_to_init(void)
242 write_lock_irq(&tasklist_lock);
244 ptrace_unlink(current);
245 /* Reparent to init */
246 REMOVE_LINKS(current);
247 /* FIXME handle vchild_reaper/initpid */
248 current->parent = child_reaper;
249 current->real_parent = child_reaper;
252 /* Set the exit signal to SIGCHLD so we signal init on exit */
253 current->exit_signal = SIGCHLD;
255 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
256 set_user_nice(current, 0);
260 security_task_reparent_to_init(current);
261 memcpy(current->rlim, init_task.rlim, sizeof(*(current->rlim)));
262 atomic_inc(&(INIT_USER->__count));
263 switch_uid(INIT_USER);
265 write_unlock_irq(&tasklist_lock);
268 void __set_special_pids(pid_t session, pid_t pgrp)
270 struct task_struct *curr = current;
272 if (curr->signal->session != session) {
273 detach_pid(curr, PIDTYPE_SID);
274 curr->signal->session = session;
275 attach_pid(curr, PIDTYPE_SID, session);
277 if (process_group(curr) != pgrp) {
278 detach_pid(curr, PIDTYPE_PGID);
279 curr->signal->pgrp = pgrp;
280 attach_pid(curr, PIDTYPE_PGID, pgrp);
284 void set_special_pids(pid_t session, pid_t pgrp)
286 write_lock_irq(&tasklist_lock);
287 __set_special_pids(session, pgrp);
288 write_unlock_irq(&tasklist_lock);
292 * Let kernel threads use this to say that they
293 * allow a certain signal (since daemonize() will
294 * have disabled all of them by default).
296 int allow_signal(int sig)
298 if (sig < 1 || sig > _NSIG)
301 spin_lock_irq(¤t->sighand->siglock);
302 sigdelset(¤t->blocked, sig);
304 /* Kernel threads handle their own signals.
305 Let the signal code know it'll be handled, so
306 that they don't get converted to SIGKILL or
307 just silently dropped */
308 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
311 spin_unlock_irq(¤t->sighand->siglock);
315 EXPORT_SYMBOL(allow_signal);
317 int disallow_signal(int sig)
319 if (sig < 1 || sig > _NSIG)
322 spin_lock_irq(¤t->sighand->siglock);
323 sigaddset(¤t->blocked, sig);
325 spin_unlock_irq(¤t->sighand->siglock);
329 EXPORT_SYMBOL(disallow_signal);
332 * Put all the gunge required to become a kernel thread without
333 * attached user resources in one place where it belongs.
336 void daemonize(const char *name, ...)
339 struct fs_struct *fs;
342 va_start(args, name);
343 vsnprintf(current->comm, sizeof(current->comm), name, args);
347 * If we were started as result of loading a module, close all of the
348 * user space pages. We don't need them, and if we didn't close them
349 * they would be locked into memory.
353 set_special_pids(1, 1);
354 current->signal->tty = NULL;
356 /* Block and flush all signals */
357 sigfillset(&blocked);
358 sigprocmask(SIG_BLOCK, &blocked, NULL);
359 flush_signals(current);
361 /* Become as one with the init task */
363 exit_fs(current); /* current->fs->count--; */
366 atomic_inc(&fs->count);
368 current->files = init_task.files;
369 atomic_inc(¤t->files->count);
374 EXPORT_SYMBOL(daemonize);
376 static inline void close_files(struct files_struct * files)
384 if (i >= files->max_fdset || i >= files->max_fds)
386 set = files->open_fds->fds_bits[j++];
389 struct file * file = xchg(&files->fd[i], NULL);
391 filp_close(file, files);
394 // vx_openfd_dec(fd);
402 struct files_struct *get_files_struct(struct task_struct *task)
404 struct files_struct *files;
409 atomic_inc(&files->count);
415 void fastcall put_files_struct(struct files_struct *files)
417 if (atomic_dec_and_test(&files->count)) {
420 * Free the fd and fdset arrays if we expanded them.
422 if (files->fd != &files->fd_array[0])
423 free_fd_array(files->fd, files->max_fds);
424 if (files->max_fdset > __FD_SETSIZE) {
425 free_fdset(files->open_fds, files->max_fdset);
426 free_fdset(files->close_on_exec, files->max_fdset);
428 kmem_cache_free(files_cachep, files);
432 EXPORT_SYMBOL(put_files_struct);
434 static inline void __exit_files(struct task_struct *tsk)
436 struct files_struct * files = tsk->files;
442 put_files_struct(files);
446 void exit_files(struct task_struct *tsk)
451 static inline void __put_fs_struct(struct fs_struct *fs)
453 /* No need to hold fs->lock if we are killing it */
454 if (atomic_dec_and_test(&fs->count)) {
461 mntput(fs->altrootmnt);
463 kmem_cache_free(fs_cachep, fs);
467 void put_fs_struct(struct fs_struct *fs)
472 static inline void __exit_fs(struct task_struct *tsk)
474 struct fs_struct * fs = tsk->fs;
484 void exit_fs(struct task_struct *tsk)
489 EXPORT_SYMBOL_GPL(exit_fs);
492 * Turn us into a lazy TLB process if we
495 static inline void __exit_mm(struct task_struct * tsk)
497 struct mm_struct *mm = tsk->mm;
503 * Serialize with any possible pending coredump.
504 * We must hold mmap_sem around checking core_waiters
505 * and clearing tsk->mm. The core-inducing thread
506 * will increment core_waiters for each thread in the
507 * group with ->mm != NULL.
509 down_read(&mm->mmap_sem);
510 if (mm->core_waiters) {
511 up_read(&mm->mmap_sem);
512 down_write(&mm->mmap_sem);
513 if (!--mm->core_waiters)
514 complete(mm->core_startup_done);
515 up_write(&mm->mmap_sem);
517 wait_for_completion(&mm->core_done);
518 down_read(&mm->mmap_sem);
520 atomic_inc(&mm->mm_count);
521 if (mm != tsk->active_mm) BUG();
522 /* more a memory barrier than a real lock */
525 up_read(&mm->mmap_sem);
526 enter_lazy_tlb(mm, current);
531 void exit_mm(struct task_struct *tsk)
536 EXPORT_SYMBOL(exit_mm);
538 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
541 * Make sure we're not reparenting to ourselves and that
542 * the parent is not a zombie.
544 if (p == reaper || reaper->state >= TASK_ZOMBIE)
545 p->real_parent = child_reaper;
547 p->real_parent = reaper;
548 if (p->parent == p->real_parent)
552 static inline void reparent_thread(task_t *p, task_t *father, int traced)
554 /* We don't want people slaying init. */
555 if (p->exit_signal != -1)
556 p->exit_signal = SIGCHLD;
559 if (p->pdeath_signal)
560 /* We already hold the tasklist_lock here. */
561 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
563 /* Move the child from its dying parent to the new one. */
564 if (unlikely(traced)) {
565 /* Preserve ptrace links if someone else is tracing this child. */
566 list_del_init(&p->ptrace_list);
567 if (p->parent != p->real_parent)
568 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
570 /* If this child is being traced, then we're the one tracing it
571 * anyway, so let go of it.
574 list_del_init(&p->sibling);
575 p->parent = p->real_parent;
576 list_add_tail(&p->sibling, &p->parent->children);
578 /* If we'd notified the old parent about this child's death,
579 * also notify the new parent.
581 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
582 thread_group_empty(p))
583 do_notify_parent(p, p->exit_signal);
587 * process group orphan check
588 * Case ii: Our child is in a different pgrp
589 * than we are, and it was the only connection
590 * outside, so the child pgrp is now orphaned.
592 if ((process_group(p) != process_group(father)) &&
593 (p->signal->session == father->signal->session)) {
594 int pgrp = process_group(p);
596 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
597 __kill_pg_info(SIGHUP, (void *)1, pgrp);
598 __kill_pg_info(SIGCONT, (void *)1, pgrp);
604 * When we die, we re-parent all our children.
605 * Try to give them to another thread in our thread
606 * group, and if no such member exists, give it to
607 * the global child reaper process (ie "init")
609 static inline void forget_original_parent(struct task_struct * father,
610 struct list_head *to_release)
612 struct task_struct *p, *reaper = father;
613 struct list_head *_p, *_n;
615 /* FIXME handle vchild_reaper/initpid */
616 reaper = father->group_leader;
617 if (reaper == father)
618 reaper = child_reaper;
621 * There are only two places where our children can be:
623 * - in our child list
624 * - in our ptraced child list
626 * Search them and reparent children.
628 list_for_each_safe(_p, _n, &father->children) {
630 p = list_entry(_p,struct task_struct,sibling);
634 /* if father isn't the real parent, then ptrace must be enabled */
635 BUG_ON(father != p->real_parent && !ptrace);
637 if (father == p->real_parent) {
638 /* reparent with a reaper, real father it's us */
639 choose_new_parent(p, reaper, child_reaper);
640 reparent_thread(p, father, 0);
642 /* reparent ptraced task to its real parent */
644 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
645 thread_group_empty(p))
646 do_notify_parent(p, p->exit_signal);
650 * if the ptraced child is a zombie with exit_signal == -1
651 * we must collect it before we exit, or it will remain
652 * zombie forever since we prevented it from self-reap itself
653 * while it was being traced by us, to be able to see it in wait4.
655 if (unlikely(ptrace && p->state == TASK_ZOMBIE && p->exit_signal == -1))
656 list_add(&p->ptrace_list, to_release);
658 list_for_each_safe(_p, _n, &father->ptrace_children) {
659 p = list_entry(_p,struct task_struct,ptrace_list);
660 choose_new_parent(p, reaper, child_reaper);
661 reparent_thread(p, father, 1);
666 * Send signals to all our closest relatives so that they know
667 * to properly mourn us..
669 static void exit_notify(struct task_struct *tsk)
672 struct task_struct *t;
673 struct list_head ptrace_dead, *_p, *_n;
677 if (signal_pending(tsk) && !tsk->signal->group_exit
678 && !thread_group_empty(tsk)) {
680 * This occurs when there was a race between our exit
681 * syscall and a group signal choosing us as the one to
682 * wake up. It could be that we are the only thread
683 * alerted to check for pending signals, but another thread
684 * should be woken now to take the signal since we will not.
685 * Now we'll wake all the threads in the group just to make
686 * sure someone gets all the pending signals.
688 read_lock(&tasklist_lock);
689 spin_lock_irq(&tsk->sighand->siglock);
690 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
691 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
692 recalc_sigpending_tsk(t);
693 if (signal_pending(t))
694 signal_wake_up(t, 0);
696 spin_unlock_irq(&tsk->sighand->siglock);
697 read_unlock(&tasklist_lock);
700 write_lock_irq(&tasklist_lock);
703 * This does two things:
705 * A. Make init inherit all the child processes
706 * B. Check to see if any process groups have become orphaned
707 * as a result of our exiting, and if they have any stopped
708 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
711 INIT_LIST_HEAD(&ptrace_dead);
712 forget_original_parent(tsk, &ptrace_dead);
713 BUG_ON(!list_empty(&tsk->children));
714 BUG_ON(!list_empty(&tsk->ptrace_children));
717 * Check to see if any process groups have become orphaned
718 * as a result of our exiting, and if they have any stopped
719 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
721 * Case i: Our father is in a different pgrp than we are
722 * and we were the only connection outside, so our pgrp
723 * is about to become orphaned.
726 t = tsk->real_parent;
728 if ((process_group(t) != process_group(tsk)) &&
729 (t->signal->session == tsk->signal->session) &&
730 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
731 has_stopped_jobs(process_group(tsk))) {
732 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
733 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
736 /* Let father know we died
738 * Thread signals are configurable, but you aren't going to use
739 * that to send signals to arbitary processes.
740 * That stops right now.
742 * If the parent exec id doesn't match the exec id we saved
743 * when we started then we know the parent has changed security
746 * If our self_exec id doesn't match our parent_exec_id then
747 * we have changed execution domain as these two values started
748 * the same after a fork.
752 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
753 ( tsk->parent_exec_id != t->self_exec_id ||
754 tsk->self_exec_id != tsk->parent_exec_id)
755 && !capable(CAP_KILL))
756 tsk->exit_signal = SIGCHLD;
759 /* If something other than our normal parent is ptracing us, then
760 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
761 * only has special meaning to our real parent.
763 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
764 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
765 do_notify_parent(tsk, signal);
766 } else if (tsk->ptrace) {
767 do_notify_parent(tsk, SIGCHLD);
771 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
774 tsk->flags |= PF_DEAD;
777 * Clear these here so that update_process_times() won't try to deliver
778 * itimer, profile or rlimit signals to this task while it is in late exit.
780 tsk->it_virt_value = 0;
781 tsk->it_prof_value = 0;
782 tsk->rlim[RLIMIT_CPU].rlim_cur = RLIM_INFINITY;
785 * In the preemption case it must be impossible for the task
786 * to get runnable again, so use "_raw_" unlock to keep
787 * preempt_count elevated until we schedule().
789 * To avoid deadlock on SMP, interrupts must be unmasked. If we
790 * don't, subsequently called functions (e.g, wait_task_inactive()
791 * via release_task()) will spin, with interrupt flags
792 * unwittingly blocked, until the other task sleeps. That task
793 * may itself be waiting for smp_call_function() to answer and
794 * complete, and with interrupts blocked that will never happen.
796 _raw_write_unlock(&tasklist_lock);
799 list_for_each_safe(_p, _n, &ptrace_dead) {
801 t = list_entry(_p,struct task_struct,ptrace_list);
805 /* If the process is dead, release it - nobody will wait for it */
806 if (state == TASK_DEAD)
811 asmlinkage NORET_TYPE void do_exit(long code)
813 struct task_struct *tsk = current;
815 if (unlikely(in_interrupt()))
816 panic("Aiee, killing interrupt handler!");
817 if (unlikely(!tsk->pid))
818 panic("Attempted to kill the idle task!");
819 if (unlikely(tsk->pid == 1))
820 panic("Attempted to kill init!");
823 tsk->flags |= PF_EXITING;
824 del_timer_sync(&tsk->real_timer);
826 if (unlikely(in_atomic()))
827 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
828 current->comm, current->pid,
831 profile_exit_task(tsk);
833 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
834 current->ptrace_message = code;
835 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
839 if (current->tux_info) {
840 #ifdef CONFIG_TUX_DEBUG
841 printk("Possibly unexpected TUX-thread exit(%ld) at %p?\n",
842 code, __builtin_return_address(0));
854 if (tsk->signal->leader)
855 disassociate_ctty(1);
857 module_put(tsk->thread_info->exec_domain->module);
859 module_put(tsk->binfmt->module);
861 tsk->exit_code = code;
862 #ifdef CONFIG_CKRM_TYPE_TASKCLASS
863 numtasks_put_ref(tsk->taskclass);
867 mpol_free(tsk->mempolicy);
868 tsk->mempolicy = NULL;
872 /* Avoid "noreturn function does return". */
876 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
884 EXPORT_SYMBOL(complete_and_exit);
886 asmlinkage long sys_exit(int error_code)
888 do_exit((error_code&0xff)<<8);
891 task_t fastcall *next_thread(const task_t *p)
893 const struct pid_link *link = p->pids + PIDTYPE_TGID;
894 const struct list_head *tmp, *head = &link->pidptr->task_list;
899 if (!spin_is_locked(&p->sighand->siglock) &&
900 !rwlock_is_locked(&tasklist_lock))
903 tmp = link->pid_chain.next;
907 return pid_task(tmp, PIDTYPE_TGID);
910 EXPORT_SYMBOL(next_thread);
913 * Take down every thread in the group. This is called by fatal signals
914 * as well as by sys_exit_group (below).
917 do_group_exit(int exit_code)
919 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
921 if (current->signal->group_exit)
922 exit_code = current->signal->group_exit_code;
923 else if (!thread_group_empty(current)) {
924 struct signal_struct *const sig = current->signal;
925 struct sighand_struct *const sighand = current->sighand;
926 read_lock(&tasklist_lock);
927 spin_lock_irq(&sighand->siglock);
929 /* Another thread got here before we took the lock. */
930 exit_code = sig->group_exit_code;
933 sig->group_exit_code = exit_code;
934 zap_other_threads(current);
936 spin_unlock_irq(&sighand->siglock);
937 read_unlock(&tasklist_lock);
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.
949 asmlinkage void sys_exit_group(int error_code)
951 do_group_exit((error_code & 0xff) << 8);
954 static int eligible_child(pid_t pid, int options, task_t *p)
960 if (process_group(p) != process_group(current))
962 } else if (pid != -1) {
963 if (process_group(p) != -pid)
968 * Do not consider detached threads that are
971 if (p->exit_signal == -1 && !p->ptrace)
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))
983 * Do not consider thread group leaders that are
984 * in a non-empty thread group:
986 if (current->tgid != p->tgid && delay_group_leader(p))
989 if (security_task_wait(p))
996 * Handle sys_wait4 work for one task in state TASK_ZOMBIE. We hold
997 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
998 * the lock and this task is uninteresting. If we return nonzero, we have
999 * released the lock and the system call should return.
1001 static int wait_task_zombie(task_t *p, unsigned int __user *stat_addr, struct rusage __user *ru)
1003 unsigned long state;
1007 * Try to move the task's state to DEAD
1008 * only one thread is allowed to do this:
1010 state = xchg(&p->state, TASK_DEAD);
1011 if (state != TASK_ZOMBIE) {
1012 BUG_ON(state != TASK_DEAD);
1015 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1017 * This can only happen in a race with a ptraced thread
1018 * dying on another processor.
1023 * Now we are sure this task is interesting, and no other
1024 * thread can reap it because we set its state to TASK_DEAD.
1026 read_unlock(&tasklist_lock);
1028 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1029 if (!retval && stat_addr) {
1030 if (p->signal->group_exit)
1031 retval = put_user(p->signal->group_exit_code, stat_addr);
1033 retval = put_user(p->exit_code, stat_addr);
1036 p->state = TASK_ZOMBIE;
1040 if (p->real_parent != p->parent) {
1041 write_lock_irq(&tasklist_lock);
1042 /* Double-check with lock held. */
1043 if (p->real_parent != p->parent) {
1045 p->state = TASK_ZOMBIE;
1047 * If this is not a detached task, notify the parent. If it's
1048 * still not detached after that, don't release it now.
1050 if (p->exit_signal != -1) {
1051 do_notify_parent(p, p->exit_signal);
1052 if (p->exit_signal != -1)
1056 write_unlock_irq(&tasklist_lock);
1065 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1066 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1067 * the lock and this task is uninteresting. If we return nonzero, we have
1068 * released the lock and the system call should return.
1070 static int wait_task_stopped(task_t *p, int delayed_group_leader,
1071 unsigned int __user *stat_addr,
1072 struct rusage __user *ru)
1074 int retval, exit_code;
1078 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1079 p->signal && p->signal->group_stop_count > 0)
1081 * A group stop is in progress and this is the group leader.
1082 * We won't report until all threads have stopped.
1087 * Now we are pretty sure this task is interesting.
1088 * Make sure it doesn't get reaped out from under us while we
1089 * give up the lock and then examine it below. We don't want to
1090 * keep holding onto the tasklist_lock while we call getrusage and
1091 * possibly take page faults for user memory.
1094 read_unlock(&tasklist_lock);
1095 write_lock_irq(&tasklist_lock);
1098 * This uses xchg to be atomic with the thread resuming and setting
1099 * it. It must also be done with the write lock held to prevent a
1100 * race with the TASK_ZOMBIE case.
1102 exit_code = xchg(&p->exit_code, 0);
1103 if (unlikely(p->state > TASK_STOPPED)) {
1105 * The task resumed and then died. Let the next iteration
1106 * catch it in TASK_ZOMBIE. Note that exit_code might
1107 * already be zero here if it resumed and did _exit(0).
1108 * The task itself is dead and won't touch exit_code again;
1109 * other processors in this function are locked out.
1111 p->exit_code = exit_code;
1114 if (unlikely(exit_code == 0)) {
1116 * Another thread in this function got to it first, or it
1117 * resumed, or it resumed and then died.
1119 write_unlock_irq(&tasklist_lock);
1121 read_lock(&tasklist_lock);
1125 /* move to end of parent's list to avoid starvation */
1127 add_parent(p, p->parent);
1129 write_unlock_irq(&tasklist_lock);
1131 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1132 if (!retval && stat_addr)
1133 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1142 asmlinkage long sys_wait4(pid_t pid,unsigned int __user *stat_addr, int options, struct rusage __user *ru)
1144 DECLARE_WAITQUEUE(wait, current);
1145 struct task_struct *tsk;
1148 if (options & ~(WNOHANG|WUNTRACED|__WNOTHREAD|__WCLONE|__WALL))
1151 add_wait_queue(¤t->wait_chldexit,&wait);
1154 current->state = TASK_INTERRUPTIBLE;
1155 read_lock(&tasklist_lock);
1158 struct task_struct *p;
1159 struct list_head *_p;
1162 list_for_each(_p,&tsk->children) {
1163 p = list_entry(_p,struct task_struct,sibling);
1165 ret = eligible_child(pid, options, p);
1172 if (!(options & WUNTRACED) &&
1173 !(p->ptrace & PT_PTRACED))
1175 retval = wait_task_stopped(p, ret == 2,
1177 if (retval != 0) /* He released the lock. */
1182 * Eligible but we cannot release it yet:
1186 retval = wait_task_zombie(p, stat_addr, ru);
1187 if (retval != 0) /* He released the lock. */
1193 list_for_each (_p,&tsk->ptrace_children) {
1194 p = list_entry(_p,struct task_struct,ptrace_list);
1195 if (!eligible_child(pid, options, p))
1201 if (options & __WNOTHREAD)
1203 tsk = next_thread(tsk);
1204 if (tsk->signal != current->signal)
1206 } while (tsk != current);
1207 read_unlock(&tasklist_lock);
1210 if (options & WNOHANG)
1212 retval = -ERESTARTSYS;
1213 if (signal_pending(current))
1220 current->state = TASK_RUNNING;
1221 remove_wait_queue(¤t->wait_chldexit,&wait);
1225 #ifdef __ARCH_WANT_SYS_WAITPID
1228 * sys_waitpid() remains for compatibility. waitpid() should be
1229 * implemented by calling sys_wait4() from libc.a.
1231 asmlinkage long sys_waitpid(pid_t pid, unsigned __user *stat_addr, int options)
1233 return sys_wait4(pid, stat_addr, options, NULL);