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>
29 #include <asm/uaccess.h>
30 #include <asm/unistd.h>
31 #include <asm/pgtable.h>
32 #include <asm/mmu_context.h>
34 extern void sem_exit (void);
35 extern struct task_struct *child_reaper;
37 int getrusage(struct task_struct *, int, struct rusage __user *);
39 static void __unhash_process(struct task_struct *p)
42 detach_pid(p, PIDTYPE_PID);
43 detach_pid(p, PIDTYPE_TGID);
44 if (thread_group_leader(p)) {
45 detach_pid(p, PIDTYPE_PGID);
46 detach_pid(p, PIDTYPE_SID);
48 __get_cpu_var(process_counts)--;
54 void release_task(struct task_struct * p)
58 struct dentry *proc_dentry;
61 BUG_ON(p->state < TASK_ZOMBIE);
63 atomic_dec(&p->user->processes);
64 spin_lock(&p->proc_lock);
65 proc_dentry = proc_pid_unhash(p);
66 write_lock_irq(&tasklist_lock);
67 if (unlikely(p->ptrace))
69 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
75 * If we are the last non-leader member of the thread
76 * group, and the leader is zombie, then notify the
77 * group leader's parent process. (if it wants notification.)
80 leader = p->group_leader;
81 if (leader != p && thread_group_empty(leader) && leader->state == TASK_ZOMBIE) {
82 BUG_ON(leader->exit_signal == -1);
83 do_notify_parent(leader, leader->exit_signal);
85 * If we were the last child thread and the leader has
86 * exited already, and the leader's parent ignores SIGCHLD,
87 * then we are the one who should release the leader.
89 * do_notify_parent() will have marked it self-reaping in
92 zap_leader = (leader->exit_signal == -1);
95 p->parent->cutime += p->utime + p->cutime;
96 p->parent->cstime += p->stime + p->cstime;
97 p->parent->cmin_flt += p->min_flt + p->cmin_flt;
98 p->parent->cmaj_flt += p->maj_flt + p->cmaj_flt;
99 p->parent->cnvcsw += p->nvcsw + p->cnvcsw;
100 p->parent->cnivcsw += p->nivcsw + p->cnivcsw;
102 write_unlock_irq(&tasklist_lock);
103 spin_unlock(&p->proc_lock);
104 proc_pid_flush(proc_dentry);
109 if (unlikely(zap_leader))
113 /* we are using it only for SMP init */
115 void unhash_process(struct task_struct *p)
117 struct dentry *proc_dentry;
119 spin_lock(&p->proc_lock);
120 proc_dentry = proc_pid_unhash(p);
121 write_lock_irq(&tasklist_lock);
123 write_unlock_irq(&tasklist_lock);
124 spin_unlock(&p->proc_lock);
125 proc_pid_flush(proc_dentry);
129 * This checks not only the pgrp, but falls back on the pid if no
130 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
133 int session_of_pgrp(int pgrp)
135 struct task_struct *p;
140 read_lock(&tasklist_lock);
141 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid)
142 if (p->signal->session > 0) {
143 sid = p->signal->session;
146 p = find_task_by_pid(pgrp);
148 sid = p->signal->session;
150 read_unlock(&tasklist_lock);
156 * Determine if a process group is "orphaned", according to the POSIX
157 * definition in 2.2.2.52. Orphaned process groups are not to be affected
158 * by terminal-generated stop signals. Newly orphaned process groups are
159 * to receive a SIGHUP and a SIGCONT.
161 * "I ask you, have you ever known what it is to be an orphan?"
163 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
165 struct task_struct *p;
170 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
171 if (p == ignored_task
172 || p->state >= TASK_ZOMBIE
173 || p->real_parent->pid == 1)
175 if (process_group(p->real_parent) != pgrp
176 && p->real_parent->signal->session == p->signal->session) {
181 return ret; /* (sighing) "Often!" */
184 int is_orphaned_pgrp(int pgrp)
188 read_lock(&tasklist_lock);
189 retval = will_become_orphaned_pgrp(pgrp, NULL);
190 read_unlock(&tasklist_lock);
195 static inline int has_stopped_jobs(int pgrp)
198 struct task_struct *p;
202 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
203 if (p->state != TASK_STOPPED)
206 /* If p is stopped by a debugger on a signal that won't
207 stop it, then don't count p as stopped. This isn't
208 perfect but it's a good approximation. */
209 if (unlikely (p->ptrace)
210 && p->exit_code != SIGSTOP
211 && p->exit_code != SIGTSTP
212 && p->exit_code != SIGTTOU
213 && p->exit_code != SIGTTIN)
223 * reparent_to_init() - Reparent the calling kernel thread to the init task.
225 * If a kernel thread is launched as a result of a system call, or if
226 * it ever exits, it should generally reparent itself to init so that
227 * it is correctly cleaned up on exit.
229 * The various task state such as scheduling policy and priority may have
230 * been inherited from a user process, so we reset them to sane values here.
232 * NOTE that reparent_to_init() gives the caller full capabilities.
234 void reparent_to_init(void)
236 write_lock_irq(&tasklist_lock);
238 ptrace_unlink(current);
239 /* Reparent to init */
240 REMOVE_LINKS(current);
241 current->parent = child_reaper;
242 current->real_parent = child_reaper;
245 /* Set the exit signal to SIGCHLD so we signal init on exit */
246 current->exit_signal = SIGCHLD;
248 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
249 set_user_nice(current, 0);
253 security_task_reparent_to_init(current);
254 memcpy(current->rlim, init_task.rlim, sizeof(*(current->rlim)));
255 atomic_inc(&(INIT_USER->__count));
256 switch_uid(INIT_USER);
258 write_unlock_irq(&tasklist_lock);
261 void __set_special_pids(pid_t session, pid_t pgrp)
263 struct task_struct *curr = current;
265 if (curr->signal->session != session) {
266 detach_pid(curr, PIDTYPE_SID);
267 curr->signal->session = session;
268 attach_pid(curr, PIDTYPE_SID, session);
270 if (process_group(curr) != pgrp) {
271 detach_pid(curr, PIDTYPE_PGID);
272 curr->signal->pgrp = pgrp;
273 attach_pid(curr, PIDTYPE_PGID, pgrp);
277 void set_special_pids(pid_t session, pid_t pgrp)
279 write_lock_irq(&tasklist_lock);
280 __set_special_pids(session, pgrp);
281 write_unlock_irq(&tasklist_lock);
285 * Let kernel threads use this to say that they
286 * allow a certain signal (since daemonize() will
287 * have disabled all of them by default).
289 int allow_signal(int sig)
291 if (sig < 1 || sig > _NSIG)
294 spin_lock_irq(¤t->sighand->siglock);
295 sigdelset(¤t->blocked, sig);
297 /* Kernel threads handle their own signals.
298 Let the signal code know it'll be handled, so
299 that they don't get converted to SIGKILL or
300 just silently dropped */
301 current->sighand->action[(sig)-1].sa.sa_handler = (void *)2;
304 spin_unlock_irq(¤t->sighand->siglock);
308 EXPORT_SYMBOL(allow_signal);
310 int disallow_signal(int sig)
312 if (sig < 1 || sig > _NSIG)
315 spin_lock_irq(¤t->sighand->siglock);
316 sigaddset(¤t->blocked, sig);
318 spin_unlock_irq(¤t->sighand->siglock);
322 EXPORT_SYMBOL(disallow_signal);
325 * Put all the gunge required to become a kernel thread without
326 * attached user resources in one place where it belongs.
329 void daemonize(const char *name, ...)
332 struct fs_struct *fs;
335 va_start(args, name);
336 vsnprintf(current->comm, sizeof(current->comm), name, args);
340 * If we were started as result of loading a module, close all of the
341 * user space pages. We don't need them, and if we didn't close them
342 * they would be locked into memory.
346 set_special_pids(1, 1);
347 current->signal->tty = NULL;
349 /* Block and flush all signals */
350 sigfillset(&blocked);
351 sigprocmask(SIG_BLOCK, &blocked, NULL);
352 flush_signals(current);
354 /* Become as one with the init task */
356 exit_fs(current); /* current->fs->count--; */
359 atomic_inc(&fs->count);
361 current->files = init_task.files;
362 atomic_inc(¤t->files->count);
367 EXPORT_SYMBOL(daemonize);
369 static inline void close_files(struct files_struct * files)
377 if (i >= files->max_fdset || i >= files->max_fds)
379 set = files->open_fds->fds_bits[j++];
382 struct file * file = xchg(&files->fd[i], NULL);
384 filp_close(file, files);
392 struct files_struct *get_files_struct(struct task_struct *task)
394 struct files_struct *files;
399 atomic_inc(&files->count);
405 void fastcall put_files_struct(struct files_struct *files)
407 if (atomic_dec_and_test(&files->count)) {
410 * Free the fd and fdset arrays if we expanded them.
412 if (files->fd != &files->fd_array[0])
413 free_fd_array(files->fd, files->max_fds);
414 if (files->max_fdset > __FD_SETSIZE) {
415 free_fdset(files->open_fds, files->max_fdset);
416 free_fdset(files->close_on_exec, files->max_fdset);
418 kmem_cache_free(files_cachep, files);
422 EXPORT_SYMBOL(put_files_struct);
424 static inline void __exit_files(struct task_struct *tsk)
426 struct files_struct * files = tsk->files;
432 put_files_struct(files);
436 void exit_files(struct task_struct *tsk)
441 static inline void __put_fs_struct(struct fs_struct *fs)
443 /* No need to hold fs->lock if we are killing it */
444 if (atomic_dec_and_test(&fs->count)) {
451 mntput(fs->altrootmnt);
453 kmem_cache_free(fs_cachep, fs);
457 void put_fs_struct(struct fs_struct *fs)
462 static inline void __exit_fs(struct task_struct *tsk)
464 struct fs_struct * fs = tsk->fs;
474 void exit_fs(struct task_struct *tsk)
479 EXPORT_SYMBOL_GPL(exit_fs);
482 * Turn us into a lazy TLB process if we
485 static inline void __exit_mm(struct task_struct * tsk)
487 struct mm_struct *mm = tsk->mm;
493 * Serialize with any possible pending coredump.
494 * We must hold mmap_sem around checking core_waiters
495 * and clearing tsk->mm. The core-inducing thread
496 * will increment core_waiters for each thread in the
497 * group with ->mm != NULL.
499 down_read(&mm->mmap_sem);
500 if (mm->core_waiters) {
501 up_read(&mm->mmap_sem);
502 down_write(&mm->mmap_sem);
503 if (!--mm->core_waiters)
504 complete(mm->core_startup_done);
505 up_write(&mm->mmap_sem);
507 wait_for_completion(&mm->core_done);
508 down_read(&mm->mmap_sem);
510 atomic_inc(&mm->mm_count);
511 if (mm != tsk->active_mm) BUG();
512 /* more a memory barrier than a real lock */
515 up_read(&mm->mmap_sem);
516 enter_lazy_tlb(mm, current);
521 void exit_mm(struct task_struct *tsk)
526 EXPORT_SYMBOL(exit_mm);
528 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
531 * Make sure we're not reparenting to ourselves and that
532 * the parent is not a zombie.
534 if (p == reaper || reaper->state >= TASK_ZOMBIE)
535 p->real_parent = child_reaper;
537 p->real_parent = reaper;
538 if (p->parent == p->real_parent)
542 static inline void reparent_thread(task_t *p, task_t *father, int traced)
544 /* We don't want people slaying init. */
545 if (p->exit_signal != -1)
546 p->exit_signal = SIGCHLD;
549 if (p->pdeath_signal)
550 /* We already hold the tasklist_lock here. */
551 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
553 /* Move the child from its dying parent to the new one. */
554 if (unlikely(traced)) {
555 /* Preserve ptrace links if someone else is tracing this child. */
556 list_del_init(&p->ptrace_list);
557 if (p->parent != p->real_parent)
558 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
560 /* If this child is being traced, then we're the one tracing it
561 * anyway, so let go of it.
564 list_del_init(&p->sibling);
565 p->parent = p->real_parent;
566 list_add_tail(&p->sibling, &p->parent->children);
568 /* If we'd notified the old parent about this child's death,
569 * also notify the new parent.
571 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
572 thread_group_empty(p))
573 do_notify_parent(p, p->exit_signal);
577 * process group orphan check
578 * Case ii: Our child is in a different pgrp
579 * than we are, and it was the only connection
580 * outside, so the child pgrp is now orphaned.
582 if ((process_group(p) != process_group(father)) &&
583 (p->signal->session == father->signal->session)) {
584 int pgrp = process_group(p);
586 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
587 __kill_pg_info(SIGHUP, (void *)1, pgrp);
588 __kill_pg_info(SIGCONT, (void *)1, pgrp);
594 * When we die, we re-parent all our children.
595 * Try to give them to another thread in our thread
596 * group, and if no such member exists, give it to
597 * the global child reaper process (ie "init")
599 static inline void forget_original_parent(struct task_struct * father)
601 struct task_struct *p, *reaper = father;
602 struct list_head *_p, *_n;
604 reaper = father->group_leader;
605 if (reaper == father)
606 reaper = child_reaper;
609 * There are only two places where our children can be:
611 * - in our child list
612 * - in our ptraced child list
614 * Search them and reparent children.
616 list_for_each_safe(_p, _n, &father->children) {
617 p = list_entry(_p,struct task_struct,sibling);
618 if (father == p->real_parent) {
619 choose_new_parent(p, reaper, child_reaper);
620 reparent_thread(p, father, 0);
623 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
624 thread_group_empty(p))
625 do_notify_parent(p, p->exit_signal);
628 list_for_each_safe(_p, _n, &father->ptrace_children) {
629 p = list_entry(_p,struct task_struct,ptrace_list);
630 choose_new_parent(p, reaper, child_reaper);
631 reparent_thread(p, father, 1);
636 * Send signals to all our closest relatives so that they know
637 * to properly mourn us..
639 static void exit_notify(struct task_struct *tsk)
642 struct task_struct *t;
646 if (signal_pending(tsk) && !tsk->signal->group_exit
647 && !thread_group_empty(tsk)) {
649 * This occurs when there was a race between our exit
650 * syscall and a group signal choosing us as the one to
651 * wake up. It could be that we are the only thread
652 * alerted to check for pending signals, but another thread
653 * should be woken now to take the signal since we will not.
654 * Now we'll wake all the threads in the group just to make
655 * sure someone gets all the pending signals.
657 read_lock(&tasklist_lock);
658 spin_lock_irq(&tsk->sighand->siglock);
659 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
660 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
661 recalc_sigpending_tsk(t);
662 if (signal_pending(t))
663 signal_wake_up(t, 0);
665 spin_unlock_irq(&tsk->sighand->siglock);
666 read_unlock(&tasklist_lock);
669 write_lock_irq(&tasklist_lock);
672 * This does two things:
674 * A. Make init inherit all the child processes
675 * B. Check to see if any process groups have become orphaned
676 * as a result of our exiting, and if they have any stopped
677 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
680 forget_original_parent(tsk);
681 BUG_ON(!list_empty(&tsk->children));
684 * Check to see if any process groups have become orphaned
685 * as a result of our exiting, and if they have any stopped
686 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
688 * Case i: Our father is in a different pgrp than we are
689 * and we were the only connection outside, so our pgrp
690 * is about to become orphaned.
693 t = tsk->real_parent;
695 if ((process_group(t) != process_group(tsk)) &&
696 (t->signal->session == tsk->signal->session) &&
697 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
698 has_stopped_jobs(process_group(tsk))) {
699 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
700 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
703 /* Let father know we died
705 * Thread signals are configurable, but you aren't going to use
706 * that to send signals to arbitary processes.
707 * That stops right now.
709 * If the parent exec id doesn't match the exec id we saved
710 * when we started then we know the parent has changed security
713 * If our self_exec id doesn't match our parent_exec_id then
714 * we have changed execution domain as these two values started
715 * the same after a fork.
719 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
720 ( tsk->parent_exec_id != t->self_exec_id ||
721 tsk->self_exec_id != tsk->parent_exec_id)
722 && !capable(CAP_KILL))
723 tsk->exit_signal = SIGCHLD;
726 /* If something other than our normal parent is ptracing us, then
727 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
728 * only has special meaning to our real parent.
730 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
731 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
732 do_notify_parent(tsk, signal);
733 } else if (tsk->ptrace) {
734 do_notify_parent(tsk, SIGCHLD);
738 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
741 tsk->flags |= PF_DEAD;
744 * Clear these here so that update_process_times() won't try to deliver
745 * itimer, profile or rlimit signals to this task while it is in late exit.
747 tsk->it_virt_value = 0;
748 tsk->it_prof_value = 0;
749 tsk->rlim[RLIMIT_CPU].rlim_cur = RLIM_INFINITY;
752 * In the preemption case it must be impossible for the task
753 * to get runnable again, so use "_raw_" unlock to keep
754 * preempt_count elevated until we schedule().
756 * To avoid deadlock on SMP, interrupts must be unmasked. If we
757 * don't, subsequently called functions (e.g, wait_task_inactive()
758 * via release_task()) will spin, with interrupt flags
759 * unwittingly blocked, until the other task sleeps. That task
760 * may itself be waiting for smp_call_function() to answer and
761 * complete, and with interrupts blocked that will never happen.
763 _raw_write_unlock(&tasklist_lock);
766 /* If the process is dead, release it - nobody will wait for it */
767 if (state == TASK_DEAD)
772 asmlinkage NORET_TYPE void do_exit(long code)
774 struct task_struct *tsk = current;
776 if (unlikely(in_interrupt()))
777 panic("Aiee, killing interrupt handler!");
778 if (unlikely(!tsk->pid))
779 panic("Attempted to kill the idle task!");
780 if (unlikely(tsk->pid == 1))
781 panic("Attempted to kill init!");
784 tsk->flags |= PF_EXITING;
785 del_timer_sync(&tsk->real_timer);
787 if (unlikely(in_atomic()))
788 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
789 current->comm, current->pid,
792 profile_exit_task(tsk);
794 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
795 current->ptrace_message = code;
796 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
808 mpol_free(tsk->mempolicy);
811 if (tsk->signal->leader)
812 disassociate_ctty(1);
814 module_put(tsk->thread_info->exec_domain->module);
816 module_put(tsk->binfmt->module);
818 tsk->exit_code = code;
819 #ifdef CONFIG_CKRM_TYPE_TASKCLASS
820 numtasks_put_ref(tsk->taskclass);
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);