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>
27 #include <asm/uaccess.h>
28 #include <asm/unistd.h>
29 #include <asm/pgtable.h>
30 #include <asm/mmu_context.h>
32 extern void sem_exit (void);
33 extern struct task_struct *child_reaper;
35 int getrusage(struct task_struct *, int, struct rusage __user *);
37 static void __unhash_process(struct task_struct *p)
40 /* tasklist_lock is held, is this sufficient? */
42 atomic_dec(&p->vx_info->cacct.nr_threads);
43 atomic_dec(&p->vx_info->limit.res[RLIMIT_NPROC]);
45 detach_pid(p, PIDTYPE_PID);
46 detach_pid(p, PIDTYPE_TGID);
47 if (thread_group_leader(p)) {
48 detach_pid(p, PIDTYPE_PGID);
49 detach_pid(p, PIDTYPE_SID);
51 __get_cpu_var(process_counts)--;
57 void release_task(struct task_struct * p)
61 struct dentry *proc_dentry;
64 BUG_ON(p->state < TASK_ZOMBIE);
66 atomic_dec(&p->user->processes);
67 spin_lock(&p->proc_lock);
68 proc_dentry = proc_pid_unhash(p);
69 write_lock_irq(&tasklist_lock);
70 if (unlikely(p->ptrace))
72 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
78 * If we are the last non-leader member of the thread
79 * group, and the leader is zombie, then notify the
80 * group leader's parent process. (if it wants notification.)
83 leader = p->group_leader;
84 if (leader != p && thread_group_empty(leader) && leader->state == TASK_ZOMBIE) {
85 BUG_ON(leader->exit_signal == -1);
86 do_notify_parent(leader, leader->exit_signal);
88 * If we were the last child thread and the leader has
89 * exited already, and the leader's parent ignores SIGCHLD,
90 * then we are the one who should release the leader.
92 * do_notify_parent() will have marked it self-reaping in
95 zap_leader = (leader->exit_signal == -1);
98 p->parent->cutime += p->utime + p->cutime;
99 p->parent->cstime += p->stime + p->cstime;
100 p->parent->cmin_flt += p->min_flt + p->cmin_flt;
101 p->parent->cmaj_flt += p->maj_flt + p->cmaj_flt;
102 p->parent->cnvcsw += p->nvcsw + p->cnvcsw;
103 p->parent->cnivcsw += p->nivcsw + p->cnivcsw;
105 write_unlock_irq(&tasklist_lock);
106 spin_unlock(&p->proc_lock);
107 proc_pid_flush(proc_dentry);
112 if (unlikely(zap_leader))
116 /* we are using it only for SMP init */
118 void unhash_process(struct task_struct *p)
120 struct dentry *proc_dentry;
122 spin_lock(&p->proc_lock);
123 proc_dentry = proc_pid_unhash(p);
124 write_lock_irq(&tasklist_lock);
126 write_unlock_irq(&tasklist_lock);
127 spin_unlock(&p->proc_lock);
128 proc_pid_flush(proc_dentry);
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
136 int session_of_pgrp(int pgrp)
138 struct task_struct *p;
143 read_lock(&tasklist_lock);
144 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid)
145 if (p->signal->session > 0) {
146 sid = p->signal->session;
149 p = find_task_by_pid(pgrp);
151 sid = p->signal->session;
153 read_unlock(&tasklist_lock);
159 * Determine if a process group is "orphaned", according to the POSIX
160 * definition in 2.2.2.52. Orphaned process groups are not to be affected
161 * by terminal-generated stop signals. Newly orphaned process groups are
162 * to receive a SIGHUP and a SIGCONT.
164 * "I ask you, have you ever known what it is to be an orphan?"
166 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
168 struct task_struct *p;
173 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
174 if (p == ignored_task
175 || p->state >= TASK_ZOMBIE
176 || p->real_parent->pid == 1)
178 if (process_group(p->real_parent) != pgrp
179 && p->real_parent->signal->session == p->signal->session) {
184 return ret; /* (sighing) "Often!" */
187 int is_orphaned_pgrp(int pgrp)
191 read_lock(&tasklist_lock);
192 retval = will_become_orphaned_pgrp(pgrp, NULL);
193 read_unlock(&tasklist_lock);
198 static inline int has_stopped_jobs(int pgrp)
201 struct task_struct *p;
205 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
206 if (p->state != TASK_STOPPED)
209 /* If p is stopped by a debugger on a signal that won't
210 stop it, then don't count p as stopped. This isn't
211 perfect but it's a good approximation. */
212 if (unlikely (p->ptrace)
213 && p->exit_code != SIGSTOP
214 && p->exit_code != SIGTSTP
215 && p->exit_code != SIGTTOU
216 && p->exit_code != SIGTTIN)
226 * reparent_to_init() - Reparent the calling kernel thread to the init task.
228 * If a kernel thread is launched as a result of a system call, or if
229 * it ever exits, it should generally reparent itself to init so that
230 * it is correctly cleaned up on exit.
232 * The various task state such as scheduling policy and priority may have
233 * been inherited from a user process, so we reset them to sane values here.
235 * NOTE that reparent_to_init() gives the caller full capabilities.
237 void reparent_to_init(void)
239 write_lock_irq(&tasklist_lock);
241 ptrace_unlink(current);
242 /* Reparent to init */
243 REMOVE_LINKS(current);
244 /* FIXME handle vchild_reaper/initpid */
245 current->parent = child_reaper;
246 current->real_parent = child_reaper;
249 /* Set the exit signal to SIGCHLD so we signal init on exit */
250 current->exit_signal = SIGCHLD;
252 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
253 set_user_nice(current, 0);
257 security_task_reparent_to_init(current);
258 memcpy(current->rlim, init_task.rlim, sizeof(*(current->rlim)));
259 atomic_inc(&(INIT_USER->__count));
260 switch_uid(INIT_USER);
262 write_unlock_irq(&tasklist_lock);
265 void __set_special_pids(pid_t session, pid_t pgrp)
267 struct task_struct *curr = current;
269 if (curr->signal->session != session) {
270 detach_pid(curr, PIDTYPE_SID);
271 curr->signal->session = session;
272 attach_pid(curr, PIDTYPE_SID, session);
274 if (process_group(curr) != pgrp) {
275 detach_pid(curr, PIDTYPE_PGID);
276 curr->signal->pgrp = pgrp;
277 attach_pid(curr, PIDTYPE_PGID, pgrp);
281 void set_special_pids(pid_t session, pid_t pgrp)
283 write_lock_irq(&tasklist_lock);
284 __set_special_pids(session, pgrp);
285 write_unlock_irq(&tasklist_lock);
289 * Let kernel threads use this to say that they
290 * allow a certain signal (since daemonize() will
291 * have disabled all of them by default).
293 int allow_signal(int sig)
295 if (sig < 1 || sig > _NSIG)
298 spin_lock_irq(¤t->sighand->siglock);
299 sigdelset(¤t->blocked, sig);
301 /* Kernel threads handle their own signals.
302 Let the signal code know it'll be handled, so
303 that they don't get converted to SIGKILL or
304 just silently dropped */
305 current->sighand->action[(sig)-1].sa.sa_handler = (void *)2;
308 spin_unlock_irq(¤t->sighand->siglock);
312 EXPORT_SYMBOL(allow_signal);
314 int disallow_signal(int sig)
316 if (sig < 1 || sig > _NSIG)
319 spin_lock_irq(¤t->sighand->siglock);
320 sigaddset(¤t->blocked, sig);
322 spin_unlock_irq(¤t->sighand->siglock);
326 EXPORT_SYMBOL(disallow_signal);
329 * Put all the gunge required to become a kernel thread without
330 * attached user resources in one place where it belongs.
333 void daemonize(const char *name, ...)
336 struct fs_struct *fs;
339 va_start(args, name);
340 vsnprintf(current->comm, sizeof(current->comm), name, args);
344 * If we were started as result of loading a module, close all of the
345 * user space pages. We don't need them, and if we didn't close them
346 * they would be locked into memory.
350 set_special_pids(1, 1);
351 current->signal->tty = NULL;
353 /* Block and flush all signals */
354 sigfillset(&blocked);
355 sigprocmask(SIG_BLOCK, &blocked, NULL);
356 flush_signals(current);
358 /* Become as one with the init task */
360 exit_fs(current); /* current->fs->count--; */
363 atomic_inc(&fs->count);
365 current->files = init_task.files;
366 atomic_inc(¤t->files->count);
371 EXPORT_SYMBOL(daemonize);
373 static inline void close_files(struct files_struct * files)
381 if (i >= files->max_fdset || i >= files->max_fds)
383 set = files->open_fds->fds_bits[j++];
386 struct file * file = xchg(&files->fd[i], NULL);
388 filp_close(file, files);
397 struct files_struct *get_files_struct(struct task_struct *task)
399 struct files_struct *files;
404 atomic_inc(&files->count);
410 void fastcall put_files_struct(struct files_struct *files)
412 if (atomic_dec_and_test(&files->count)) {
415 * Free the fd and fdset arrays if we expanded them.
417 if (files->fd != &files->fd_array[0])
418 free_fd_array(files->fd, files->max_fds);
419 if (files->max_fdset > __FD_SETSIZE) {
420 free_fdset(files->open_fds, files->max_fdset);
421 free_fdset(files->close_on_exec, files->max_fdset);
423 kmem_cache_free(files_cachep, files);
427 EXPORT_SYMBOL(put_files_struct);
429 static inline void __exit_files(struct task_struct *tsk)
431 struct files_struct * files = tsk->files;
437 put_files_struct(files);
441 void exit_files(struct task_struct *tsk)
446 static inline void __put_fs_struct(struct fs_struct *fs)
448 /* No need to hold fs->lock if we are killing it */
449 if (atomic_dec_and_test(&fs->count)) {
456 mntput(fs->altrootmnt);
458 kmem_cache_free(fs_cachep, fs);
462 void put_fs_struct(struct fs_struct *fs)
467 static inline void __exit_fs(struct task_struct *tsk)
469 struct fs_struct * fs = tsk->fs;
479 void exit_fs(struct task_struct *tsk)
484 EXPORT_SYMBOL_GPL(exit_fs);
487 * Turn us into a lazy TLB process if we
490 static inline void __exit_mm(struct task_struct * tsk)
492 struct mm_struct *mm = tsk->mm;
498 * Serialize with any possible pending coredump.
499 * We must hold mmap_sem around checking core_waiters
500 * and clearing tsk->mm. The core-inducing thread
501 * will increment core_waiters for each thread in the
502 * group with ->mm != NULL.
504 down_read(&mm->mmap_sem);
505 if (mm->core_waiters) {
506 up_read(&mm->mmap_sem);
507 down_write(&mm->mmap_sem);
508 if (!--mm->core_waiters)
509 complete(mm->core_startup_done);
510 up_write(&mm->mmap_sem);
512 wait_for_completion(&mm->core_done);
513 down_read(&mm->mmap_sem);
515 atomic_inc(&mm->mm_count);
516 if (mm != tsk->active_mm) BUG();
517 /* more a memory barrier than a real lock */
520 up_read(&mm->mmap_sem);
521 enter_lazy_tlb(mm, current);
526 void exit_mm(struct task_struct *tsk)
531 EXPORT_SYMBOL(exit_mm);
533 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
536 * Make sure we're not reparenting to ourselves and that
537 * the parent is not a zombie.
539 if (p == reaper || reaper->state >= TASK_ZOMBIE)
540 p->real_parent = child_reaper;
542 p->real_parent = reaper;
543 if (p->parent == p->real_parent)
547 static inline void reparent_thread(task_t *p, task_t *father, int traced)
549 /* We don't want people slaying init. */
550 if (p->exit_signal != -1)
551 p->exit_signal = SIGCHLD;
554 if (p->pdeath_signal)
555 /* We already hold the tasklist_lock here. */
556 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
558 /* Move the child from its dying parent to the new one. */
559 if (unlikely(traced)) {
560 /* Preserve ptrace links if someone else is tracing this child. */
561 list_del_init(&p->ptrace_list);
562 if (p->parent != p->real_parent)
563 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
565 /* If this child is being traced, then we're the one tracing it
566 * anyway, so let go of it.
569 list_del_init(&p->sibling);
570 p->parent = p->real_parent;
571 list_add_tail(&p->sibling, &p->parent->children);
573 /* If we'd notified the old parent about this child's death,
574 * also notify the new parent.
576 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
577 thread_group_empty(p))
578 do_notify_parent(p, p->exit_signal);
582 * process group orphan check
583 * Case ii: Our child is in a different pgrp
584 * than we are, and it was the only connection
585 * outside, so the child pgrp is now orphaned.
587 if ((process_group(p) != process_group(father)) &&
588 (p->signal->session == father->signal->session)) {
589 int pgrp = process_group(p);
591 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
592 __kill_pg_info(SIGHUP, (void *)1, pgrp);
593 __kill_pg_info(SIGCONT, (void *)1, pgrp);
599 * When we die, we re-parent all our children.
600 * Try to give them to another thread in our thread
601 * group, and if no such member exists, give it to
602 * the global child reaper process (ie "init")
604 static inline void forget_original_parent(struct task_struct * father)
606 struct task_struct *p, *reaper = father;
607 struct list_head *_p, *_n;
609 /* FIXME handle vchild_reaper/initpid */
610 reaper = father->group_leader;
611 if (reaper == father)
612 reaper = child_reaper;
615 * There are only two places where our children can be:
617 * - in our child list
618 * - in our ptraced child list
620 * Search them and reparent children.
622 list_for_each_safe(_p, _n, &father->children) {
623 p = list_entry(_p,struct task_struct,sibling);
624 if (father == p->real_parent) {
625 choose_new_parent(p, reaper, child_reaper);
626 reparent_thread(p, father, 0);
629 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
630 thread_group_empty(p))
631 do_notify_parent(p, p->exit_signal);
634 list_for_each_safe(_p, _n, &father->ptrace_children) {
635 p = list_entry(_p,struct task_struct,ptrace_list);
636 choose_new_parent(p, reaper, child_reaper);
637 reparent_thread(p, father, 1);
642 * Send signals to all our closest relatives so that they know
643 * to properly mourn us..
645 static void exit_notify(struct task_struct *tsk)
648 struct task_struct *t;
650 if (signal_pending(tsk) && !tsk->signal->group_exit
651 && !thread_group_empty(tsk)) {
653 * This occurs when there was a race between our exit
654 * syscall and a group signal choosing us as the one to
655 * wake up. It could be that we are the only thread
656 * alerted to check for pending signals, but another thread
657 * should be woken now to take the signal since we will not.
658 * Now we'll wake all the threads in the group just to make
659 * sure someone gets all the pending signals.
661 read_lock(&tasklist_lock);
662 spin_lock_irq(&tsk->sighand->siglock);
663 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
664 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
665 recalc_sigpending_tsk(t);
666 if (signal_pending(t))
667 signal_wake_up(t, 0);
669 spin_unlock_irq(&tsk->sighand->siglock);
670 read_unlock(&tasklist_lock);
673 write_lock_irq(&tasklist_lock);
676 * This does two things:
678 * A. Make init inherit all the child processes
679 * B. Check to see if any process groups have become orphaned
680 * as a result of our exiting, and if they have any stopped
681 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
684 forget_original_parent(tsk);
685 BUG_ON(!list_empty(&tsk->children));
688 * Check to see if any process groups have become orphaned
689 * as a result of our exiting, and if they have any stopped
690 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
692 * Case i: Our father is in a different pgrp than we are
693 * and we were the only connection outside, so our pgrp
694 * is about to become orphaned.
697 t = tsk->real_parent;
699 if ((process_group(t) != process_group(tsk)) &&
700 (t->signal->session == tsk->signal->session) &&
701 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
702 has_stopped_jobs(process_group(tsk))) {
703 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
704 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
707 /* Let father know we died
709 * Thread signals are configurable, but you aren't going to use
710 * that to send signals to arbitary processes.
711 * That stops right now.
713 * If the parent exec id doesn't match the exec id we saved
714 * when we started then we know the parent has changed security
717 * If our self_exec id doesn't match our parent_exec_id then
718 * we have changed execution domain as these two values started
719 * the same after a fork.
723 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
724 ( tsk->parent_exec_id != t->self_exec_id ||
725 tsk->self_exec_id != tsk->parent_exec_id)
726 && !capable(CAP_KILL))
727 tsk->exit_signal = SIGCHLD;
730 /* If something other than our normal parent is ptracing us, then
731 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
732 * only has special meaning to our real parent.
734 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
735 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
736 do_notify_parent(tsk, signal);
737 } else if (tsk->ptrace) {
738 do_notify_parent(tsk, SIGCHLD);
742 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
745 tsk->flags |= PF_DEAD;
748 * In the preemption case it must be impossible for the task
749 * to get runnable again, so use "_raw_" unlock to keep
750 * preempt_count elevated until we schedule().
752 * To avoid deadlock on SMP, interrupts must be unmasked. If we
753 * don't, subsequently called functions (e.g, wait_task_inactive()
754 * via release_task()) will spin, with interrupt flags
755 * unwittingly blocked, until the other task sleeps. That task
756 * may itself be waiting for smp_call_function() to answer and
757 * complete, and with interrupts blocked that will never happen.
759 _raw_write_unlock(&tasklist_lock);
762 /* If the process is dead, release it - nobody will wait for it */
763 if (state == TASK_DEAD)
768 asmlinkage NORET_TYPE void do_exit(long code)
770 struct task_struct *tsk = current;
772 if (unlikely(in_interrupt()))
773 panic("Aiee, killing interrupt handler!");
774 if (unlikely(!tsk->pid))
775 panic("Attempted to kill the idle task!");
776 if (unlikely(tsk->pid == 1))
777 panic("Attempted to kill init!");
780 tsk->flags |= PF_EXITING;
781 del_timer_sync(&tsk->real_timer);
783 if (unlikely(in_atomic()))
784 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
785 current->comm, current->pid,
788 profile_exit_task(tsk);
790 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
791 current->ptrace_message = code;
792 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
796 if (current->tux_info) {
797 #ifdef CONFIG_TUX_DEBUG
798 printk("Possibly unexpected TUX-thread exit(%ld) at %p?\n",
799 code, __builtin_return_address(0));
811 mpol_free(tsk->mempolicy);
814 if (tsk->signal->leader)
815 disassociate_ctty(1);
817 module_put(tsk->thread_info->exec_domain->module);
819 module_put(tsk->binfmt->module);
821 tsk->exit_code = code;
825 /* Avoid "noreturn function does return". */
829 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
837 EXPORT_SYMBOL(complete_and_exit);
839 asmlinkage long sys_exit(int error_code)
841 do_exit((error_code&0xff)<<8);
844 task_t fastcall *next_thread(task_t *p)
846 struct pid_link *link = p->pids + PIDTYPE_TGID;
847 struct list_head *tmp, *head = &link->pidptr->task_list;
852 if (!spin_is_locked(&p->sighand->siglock) &&
853 !rwlock_is_locked(&tasklist_lock))
856 tmp = link->pid_chain.next;
860 return pid_task(tmp, PIDTYPE_TGID);
863 EXPORT_SYMBOL(next_thread);
866 * Take down every thread in the group. This is called by fatal signals
867 * as well as by sys_exit_group (below).
870 do_group_exit(int exit_code)
872 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
874 if (current->signal->group_exit)
875 exit_code = current->signal->group_exit_code;
876 else if (!thread_group_empty(current)) {
877 struct signal_struct *const sig = current->signal;
878 struct sighand_struct *const sighand = current->sighand;
879 read_lock(&tasklist_lock);
880 spin_lock_irq(&sighand->siglock);
882 /* Another thread got here before we took the lock. */
883 exit_code = sig->group_exit_code;
886 sig->group_exit_code = exit_code;
887 zap_other_threads(current);
889 spin_unlock_irq(&sighand->siglock);
890 read_unlock(&tasklist_lock);
898 * this kills every thread in the thread group. Note that any externally
899 * wait4()-ing process will get the correct exit code - even if this
900 * thread is not the thread group leader.
902 asmlinkage void sys_exit_group(int error_code)
904 do_group_exit((error_code & 0xff) << 8);
907 static int eligible_child(pid_t pid, int options, task_t *p)
913 if (process_group(p) != process_group(current))
915 } else if (pid != -1) {
916 if (process_group(p) != -pid)
921 * Do not consider detached threads that are
924 if (p->exit_signal == -1 && !p->ptrace)
927 /* Wait for all children (clone and not) if __WALL is set;
928 * otherwise, wait for clone children *only* if __WCLONE is
929 * set; otherwise, wait for non-clone children *only*. (Note:
930 * A "clone" child here is one that reports to its parent
931 * using a signal other than SIGCHLD.) */
932 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
933 && !(options & __WALL))
936 * Do not consider thread group leaders that are
937 * in a non-empty thread group:
939 if (current->tgid != p->tgid && delay_group_leader(p))
942 if (security_task_wait(p))
949 * Handle sys_wait4 work for one task in state TASK_ZOMBIE. We hold
950 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
951 * the lock and this task is uninteresting. If we return nonzero, we have
952 * released the lock and the system call should return.
954 static int wait_task_zombie(task_t *p, unsigned int __user *stat_addr, struct rusage __user *ru)
960 * Try to move the task's state to DEAD
961 * only one thread is allowed to do this:
963 state = xchg(&p->state, TASK_DEAD);
964 if (state != TASK_ZOMBIE) {
965 BUG_ON(state != TASK_DEAD);
968 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
970 * This can only happen in a race with a ptraced thread
971 * dying on another processor.
976 * Now we are sure this task is interesting, and no other
977 * thread can reap it because we set its state to TASK_DEAD.
979 read_unlock(&tasklist_lock);
981 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
982 if (!retval && stat_addr) {
983 if (p->signal->group_exit)
984 retval = put_user(p->signal->group_exit_code, stat_addr);
986 retval = put_user(p->exit_code, stat_addr);
989 p->state = TASK_ZOMBIE;
993 if (p->real_parent != p->parent) {
994 write_lock_irq(&tasklist_lock);
995 /* Double-check with lock held. */
996 if (p->real_parent != p->parent) {
998 p->state = TASK_ZOMBIE;
999 /* If this is a detached thread, this is where it goes away. */
1000 if (p->exit_signal == -1) {
1001 /* release_task takes the lock itself. */
1002 write_unlock_irq(&tasklist_lock);
1006 do_notify_parent(p, p->exit_signal);
1007 write_unlock_irq(&tasklist_lock);
1012 write_unlock_irq(&tasklist_lock);
1021 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1022 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1023 * the lock and this task is uninteresting. If we return nonzero, we have
1024 * released the lock and the system call should return.
1026 static int wait_task_stopped(task_t *p, int delayed_group_leader,
1027 unsigned int __user *stat_addr,
1028 struct rusage __user *ru)
1030 int retval, exit_code;
1034 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1035 p->signal && p->signal->group_stop_count > 0)
1037 * A group stop is in progress and this is the group leader.
1038 * We won't report until all threads have stopped.
1043 * Now we are pretty sure this task is interesting.
1044 * Make sure it doesn't get reaped out from under us while we
1045 * give up the lock and then examine it below. We don't want to
1046 * keep holding onto the tasklist_lock while we call getrusage and
1047 * possibly take page faults for user memory.
1050 read_unlock(&tasklist_lock);
1051 write_lock_irq(&tasklist_lock);
1054 * This uses xchg to be atomic with the thread resuming and setting
1055 * it. It must also be done with the write lock held to prevent a
1056 * race with the TASK_ZOMBIE case.
1058 exit_code = xchg(&p->exit_code, 0);
1059 if (unlikely(p->state > TASK_STOPPED)) {
1061 * The task resumed and then died. Let the next iteration
1062 * catch it in TASK_ZOMBIE. Note that exit_code might
1063 * already be zero here if it resumed and did _exit(0).
1064 * The task itself is dead and won't touch exit_code again;
1065 * other processors in this function are locked out.
1067 p->exit_code = exit_code;
1070 if (unlikely(exit_code == 0)) {
1072 * Another thread in this function got to it first, or it
1073 * resumed, or it resumed and then died.
1075 write_unlock_irq(&tasklist_lock);
1077 read_lock(&tasklist_lock);
1081 /* move to end of parent's list to avoid starvation */
1083 add_parent(p, p->parent);
1085 write_unlock_irq(&tasklist_lock);
1087 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1088 if (!retval && stat_addr)
1089 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1098 asmlinkage long sys_wait4(pid_t pid,unsigned int __user *stat_addr, int options, struct rusage __user *ru)
1100 DECLARE_WAITQUEUE(wait, current);
1101 struct task_struct *tsk;
1104 if (options & ~(WNOHANG|WUNTRACED|__WNOTHREAD|__WCLONE|__WALL))
1107 add_wait_queue(¤t->wait_chldexit,&wait);
1110 current->state = TASK_INTERRUPTIBLE;
1111 read_lock(&tasklist_lock);
1114 struct task_struct *p;
1115 struct list_head *_p;
1118 list_for_each(_p,&tsk->children) {
1119 p = list_entry(_p,struct task_struct,sibling);
1121 ret = eligible_child(pid, options, p);
1128 if (!(options & WUNTRACED) &&
1129 !(p->ptrace & PT_PTRACED))
1131 retval = wait_task_stopped(p, ret == 2,
1133 if (retval != 0) /* He released the lock. */
1138 * Eligible but we cannot release it yet:
1142 retval = wait_task_zombie(p, stat_addr, ru);
1143 if (retval != 0) /* He released the lock. */
1149 list_for_each (_p,&tsk->ptrace_children) {
1150 p = list_entry(_p,struct task_struct,ptrace_list);
1151 if (!eligible_child(pid, options, p))
1157 if (options & __WNOTHREAD)
1159 tsk = next_thread(tsk);
1160 if (tsk->signal != current->signal)
1162 } while (tsk != current);
1163 read_unlock(&tasklist_lock);
1166 if (options & WNOHANG)
1168 retval = -ERESTARTSYS;
1169 if (signal_pending(current))
1176 current->state = TASK_RUNNING;
1177 remove_wait_queue(¤t->wait_chldexit,&wait);
1181 #ifdef __ARCH_WANT_SYS_WAITPID
1184 * sys_waitpid() remains for compatibility. waitpid() should be
1185 * implemented by calling sys_wait4() from libc.a.
1187 asmlinkage long sys_waitpid(pid_t pid, unsigned __user *stat_addr, int options)
1189 return sys_wait4(pid, stat_addr, options, NULL);