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/vs_limit.h>
28 #include <asm/uaccess.h>
29 #include <asm/unistd.h>
30 #include <asm/pgtable.h>
31 #include <asm/mmu_context.h>
33 extern void sem_exit (void);
34 extern struct task_struct *child_reaper;
36 int getrusage(struct task_struct *, int, struct rusage __user *);
38 static void __unhash_process(struct task_struct *p)
41 /* tasklist_lock is held, is this sufficient? */
43 atomic_dec(&p->vx_info->cacct.nr_threads);
44 atomic_dec(&p->vx_info->limit.rcur[RLIMIT_NPROC]);
46 detach_pid(p, PIDTYPE_PID);
47 detach_pid(p, PIDTYPE_TGID);
48 if (thread_group_leader(p)) {
49 detach_pid(p, PIDTYPE_PGID);
50 detach_pid(p, PIDTYPE_SID);
52 __get_cpu_var(process_counts)--;
58 void release_task(struct task_struct * p)
62 struct dentry *proc_dentry;
65 BUG_ON(p->state < TASK_ZOMBIE);
67 atomic_dec(&p->user->processes);
68 spin_lock(&p->proc_lock);
69 proc_dentry = proc_pid_unhash(p);
70 write_lock_irq(&tasklist_lock);
71 if (unlikely(p->ptrace))
73 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
79 * If we are the last non-leader member of the thread
80 * group, and the leader is zombie, then notify the
81 * group leader's parent process. (if it wants notification.)
84 leader = p->group_leader;
85 if (leader != p && thread_group_empty(leader) && leader->state == TASK_ZOMBIE) {
86 BUG_ON(leader->exit_signal == -1);
87 do_notify_parent(leader, leader->exit_signal);
89 * If we were the last child thread and the leader has
90 * exited already, and the leader's parent ignores SIGCHLD,
91 * then we are the one who should release the leader.
93 * do_notify_parent() will have marked it self-reaping in
96 zap_leader = (leader->exit_signal == -1);
99 p->parent->cutime += p->utime + p->cutime;
100 p->parent->cstime += p->stime + p->cstime;
101 p->parent->cmin_flt += p->min_flt + p->cmin_flt;
102 p->parent->cmaj_flt += p->maj_flt + p->cmaj_flt;
103 p->parent->cnvcsw += p->nvcsw + p->cnvcsw;
104 p->parent->cnivcsw += p->nivcsw + p->cnivcsw;
106 write_unlock_irq(&tasklist_lock);
107 spin_unlock(&p->proc_lock);
108 proc_pid_flush(proc_dentry);
113 if (unlikely(zap_leader))
117 /* we are using it only for SMP init */
119 void unhash_process(struct task_struct *p)
121 struct dentry *proc_dentry;
123 spin_lock(&p->proc_lock);
124 proc_dentry = proc_pid_unhash(p);
125 write_lock_irq(&tasklist_lock);
127 write_unlock_irq(&tasklist_lock);
128 spin_unlock(&p->proc_lock);
129 proc_pid_flush(proc_dentry);
133 * This checks not only the pgrp, but falls back on the pid if no
134 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
137 int session_of_pgrp(int pgrp)
139 struct task_struct *p;
144 read_lock(&tasklist_lock);
145 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid)
146 if (p->signal->session > 0) {
147 sid = p->signal->session;
150 p = find_task_by_pid(pgrp);
152 sid = p->signal->session;
154 read_unlock(&tasklist_lock);
160 * Determine if a process group is "orphaned", according to the POSIX
161 * definition in 2.2.2.52. Orphaned process groups are not to be affected
162 * by terminal-generated stop signals. Newly orphaned process groups are
163 * to receive a SIGHUP and a SIGCONT.
165 * "I ask you, have you ever known what it is to be an orphan?"
167 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
169 struct task_struct *p;
174 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
175 if (p == ignored_task
176 || p->state >= TASK_ZOMBIE
177 || p->real_parent->pid == 1)
179 if (process_group(p->real_parent) != pgrp
180 && p->real_parent->signal->session == p->signal->session) {
185 return ret; /* (sighing) "Often!" */
188 int is_orphaned_pgrp(int pgrp)
192 read_lock(&tasklist_lock);
193 retval = will_become_orphaned_pgrp(pgrp, NULL);
194 read_unlock(&tasklist_lock);
199 static inline int has_stopped_jobs(int pgrp)
202 struct task_struct *p;
206 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
207 if (p->state != TASK_STOPPED)
210 /* If p is stopped by a debugger on a signal that won't
211 stop it, then don't count p as stopped. This isn't
212 perfect but it's a good approximation. */
213 if (unlikely (p->ptrace)
214 && p->exit_code != SIGSTOP
215 && p->exit_code != SIGTSTP
216 && p->exit_code != SIGTTOU
217 && p->exit_code != SIGTTIN)
227 * reparent_to_init() - Reparent the calling kernel thread to the init task.
229 * If a kernel thread is launched as a result of a system call, or if
230 * it ever exits, it should generally reparent itself to init so that
231 * it is correctly cleaned up on exit.
233 * The various task state such as scheduling policy and priority may have
234 * been inherited from a user process, so we reset them to sane values here.
236 * NOTE that reparent_to_init() gives the caller full capabilities.
238 void reparent_to_init(void)
240 write_lock_irq(&tasklist_lock);
242 ptrace_unlink(current);
243 /* Reparent to init */
244 REMOVE_LINKS(current);
245 /* FIXME handle vchild_reaper/initpid */
246 current->parent = child_reaper;
247 current->real_parent = child_reaper;
250 /* Set the exit signal to SIGCHLD so we signal init on exit */
251 current->exit_signal = SIGCHLD;
253 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
254 set_user_nice(current, 0);
258 security_task_reparent_to_init(current);
259 memcpy(current->rlim, init_task.rlim, sizeof(*(current->rlim)));
260 atomic_inc(&(INIT_USER->__count));
261 switch_uid(INIT_USER);
263 write_unlock_irq(&tasklist_lock);
266 void __set_special_pids(pid_t session, pid_t pgrp)
268 struct task_struct *curr = current;
270 if (curr->signal->session != session) {
271 detach_pid(curr, PIDTYPE_SID);
272 curr->signal->session = session;
273 attach_pid(curr, PIDTYPE_SID, session);
275 if (process_group(curr) != pgrp) {
276 detach_pid(curr, PIDTYPE_PGID);
277 curr->signal->pgrp = pgrp;
278 attach_pid(curr, PIDTYPE_PGID, pgrp);
282 void set_special_pids(pid_t session, pid_t pgrp)
284 write_lock_irq(&tasklist_lock);
285 __set_special_pids(session, pgrp);
286 write_unlock_irq(&tasklist_lock);
290 * Let kernel threads use this to say that they
291 * allow a certain signal (since daemonize() will
292 * have disabled all of them by default).
294 int allow_signal(int sig)
296 if (sig < 1 || sig > _NSIG)
299 spin_lock_irq(¤t->sighand->siglock);
300 sigdelset(¤t->blocked, sig);
302 /* Kernel threads handle their own signals.
303 Let the signal code know it'll be handled, so
304 that they don't get converted to SIGKILL or
305 just silently dropped */
306 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
309 spin_unlock_irq(¤t->sighand->siglock);
313 EXPORT_SYMBOL(allow_signal);
315 int disallow_signal(int sig)
317 if (sig < 1 || sig > _NSIG)
320 spin_lock_irq(¤t->sighand->siglock);
321 sigaddset(¤t->blocked, sig);
323 spin_unlock_irq(¤t->sighand->siglock);
327 EXPORT_SYMBOL(disallow_signal);
330 * Put all the gunge required to become a kernel thread without
331 * attached user resources in one place where it belongs.
334 void daemonize(const char *name, ...)
337 struct fs_struct *fs;
340 va_start(args, name);
341 vsnprintf(current->comm, sizeof(current->comm), name, args);
345 * If we were started as result of loading a module, close all of the
346 * user space pages. We don't need them, and if we didn't close them
347 * they would be locked into memory.
351 set_special_pids(1, 1);
352 current->signal->tty = NULL;
354 /* Block and flush all signals */
355 sigfillset(&blocked);
356 sigprocmask(SIG_BLOCK, &blocked, NULL);
357 flush_signals(current);
359 /* Become as one with the init task */
361 exit_fs(current); /* current->fs->count--; */
364 atomic_inc(&fs->count);
366 current->files = init_task.files;
367 atomic_inc(¤t->files->count);
372 EXPORT_SYMBOL(daemonize);
374 static inline void close_files(struct files_struct * files)
382 if (i >= files->max_fdset || i >= files->max_fds)
384 set = files->open_fds->fds_bits[j++];
387 struct file * file = xchg(&files->fd[i], NULL);
389 filp_close(file, files);
398 struct files_struct *get_files_struct(struct task_struct *task)
400 struct files_struct *files;
405 atomic_inc(&files->count);
411 void fastcall put_files_struct(struct files_struct *files)
413 if (atomic_dec_and_test(&files->count)) {
416 * Free the fd and fdset arrays if we expanded them.
418 if (files->fd != &files->fd_array[0])
419 free_fd_array(files->fd, files->max_fds);
420 if (files->max_fdset > __FD_SETSIZE) {
421 free_fdset(files->open_fds, files->max_fdset);
422 free_fdset(files->close_on_exec, files->max_fdset);
424 kmem_cache_free(files_cachep, files);
428 EXPORT_SYMBOL(put_files_struct);
430 static inline void __exit_files(struct task_struct *tsk)
432 struct files_struct * files = tsk->files;
438 put_files_struct(files);
442 void exit_files(struct task_struct *tsk)
447 static inline void __put_fs_struct(struct fs_struct *fs)
449 /* No need to hold fs->lock if we are killing it */
450 if (atomic_dec_and_test(&fs->count)) {
457 mntput(fs->altrootmnt);
459 kmem_cache_free(fs_cachep, fs);
463 void put_fs_struct(struct fs_struct *fs)
468 static inline void __exit_fs(struct task_struct *tsk)
470 struct fs_struct * fs = tsk->fs;
480 void exit_fs(struct task_struct *tsk)
485 EXPORT_SYMBOL_GPL(exit_fs);
488 * Turn us into a lazy TLB process if we
491 static inline void __exit_mm(struct task_struct * tsk)
493 struct mm_struct *mm = tsk->mm;
499 * Serialize with any possible pending coredump.
500 * We must hold mmap_sem around checking core_waiters
501 * and clearing tsk->mm. The core-inducing thread
502 * will increment core_waiters for each thread in the
503 * group with ->mm != NULL.
505 down_read(&mm->mmap_sem);
506 if (mm->core_waiters) {
507 up_read(&mm->mmap_sem);
508 down_write(&mm->mmap_sem);
509 if (!--mm->core_waiters)
510 complete(mm->core_startup_done);
511 up_write(&mm->mmap_sem);
513 wait_for_completion(&mm->core_done);
514 down_read(&mm->mmap_sem);
516 atomic_inc(&mm->mm_count);
517 if (mm != tsk->active_mm) BUG();
518 /* more a memory barrier than a real lock */
521 up_read(&mm->mmap_sem);
522 enter_lazy_tlb(mm, current);
527 void exit_mm(struct task_struct *tsk)
532 EXPORT_SYMBOL(exit_mm);
534 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
537 * Make sure we're not reparenting to ourselves and that
538 * the parent is not a zombie.
540 if (p == reaper || reaper->state >= TASK_ZOMBIE)
541 p->real_parent = child_reaper;
543 p->real_parent = reaper;
544 if (p->parent == p->real_parent)
548 static inline void reparent_thread(task_t *p, task_t *father, int traced)
550 /* We don't want people slaying init. */
551 if (p->exit_signal != -1)
552 p->exit_signal = SIGCHLD;
555 if (p->pdeath_signal)
556 /* We already hold the tasklist_lock here. */
557 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
559 /* Move the child from its dying parent to the new one. */
560 if (unlikely(traced)) {
561 /* Preserve ptrace links if someone else is tracing this child. */
562 list_del_init(&p->ptrace_list);
563 if (p->parent != p->real_parent)
564 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
566 /* If this child is being traced, then we're the one tracing it
567 * anyway, so let go of it.
570 list_del_init(&p->sibling);
571 p->parent = p->real_parent;
572 list_add_tail(&p->sibling, &p->parent->children);
574 /* If we'd notified the old parent about this child's death,
575 * also notify the new parent.
577 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
578 thread_group_empty(p))
579 do_notify_parent(p, p->exit_signal);
583 * process group orphan check
584 * Case ii: Our child is in a different pgrp
585 * than we are, and it was the only connection
586 * outside, so the child pgrp is now orphaned.
588 if ((process_group(p) != process_group(father)) &&
589 (p->signal->session == father->signal->session)) {
590 int pgrp = process_group(p);
592 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
593 __kill_pg_info(SIGHUP, (void *)1, pgrp);
594 __kill_pg_info(SIGCONT, (void *)1, pgrp);
600 * When we die, we re-parent all our children.
601 * Try to give them to another thread in our thread
602 * group, and if no such member exists, give it to
603 * the global child reaper process (ie "init")
605 static inline void forget_original_parent(struct task_struct * father,
606 struct list_head *to_release)
608 struct task_struct *p, *reaper = father;
609 struct list_head *_p, *_n;
611 /* FIXME handle vchild_reaper/initpid */
612 reaper = father->group_leader;
613 if (reaper == father)
614 reaper = child_reaper;
617 * There are only two places where our children can be:
619 * - in our child list
620 * - in our ptraced child list
622 * Search them and reparent children.
624 list_for_each_safe(_p, _n, &father->children) {
626 p = list_entry(_p,struct task_struct,sibling);
630 /* if father isn't the real parent, then ptrace must be enabled */
631 BUG_ON(father != p->real_parent && !ptrace);
633 if (father == p->real_parent) {
634 /* reparent with a reaper, real father it's us */
635 choose_new_parent(p, reaper, child_reaper);
636 reparent_thread(p, father, 0);
638 /* reparent ptraced task to its real parent */
640 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
641 thread_group_empty(p))
642 do_notify_parent(p, p->exit_signal);
646 * if the ptraced child is a zombie with exit_signal == -1
647 * we must collect it before we exit, or it will remain
648 * zombie forever since we prevented it from self-reap itself
649 * while it was being traced by us, to be able to see it in wait4.
651 if (unlikely(ptrace && p->state == TASK_ZOMBIE && p->exit_signal == -1))
652 list_add(&p->ptrace_list, to_release);
654 list_for_each_safe(_p, _n, &father->ptrace_children) {
655 p = list_entry(_p,struct task_struct,ptrace_list);
656 choose_new_parent(p, reaper, child_reaper);
657 reparent_thread(p, father, 1);
662 * Send signals to all our closest relatives so that they know
663 * to properly mourn us..
665 static void exit_notify(struct task_struct *tsk)
668 struct task_struct *t;
669 struct list_head ptrace_dead, *_p, *_n;
671 if (signal_pending(tsk) && !tsk->signal->group_exit
672 && !thread_group_empty(tsk)) {
674 * This occurs when there was a race between our exit
675 * syscall and a group signal choosing us as the one to
676 * wake up. It could be that we are the only thread
677 * alerted to check for pending signals, but another thread
678 * should be woken now to take the signal since we will not.
679 * Now we'll wake all the threads in the group just to make
680 * sure someone gets all the pending signals.
682 read_lock(&tasklist_lock);
683 spin_lock_irq(&tsk->sighand->siglock);
684 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
685 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
686 recalc_sigpending_tsk(t);
687 if (signal_pending(t))
688 signal_wake_up(t, 0);
690 spin_unlock_irq(&tsk->sighand->siglock);
691 read_unlock(&tasklist_lock);
694 write_lock_irq(&tasklist_lock);
697 * This does two things:
699 * A. Make init inherit all the child processes
700 * B. Check to see if any process groups have become orphaned
701 * as a result of our exiting, and if they have any stopped
702 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
705 INIT_LIST_HEAD(&ptrace_dead);
706 forget_original_parent(tsk, &ptrace_dead);
707 BUG_ON(!list_empty(&tsk->children));
708 BUG_ON(!list_empty(&tsk->ptrace_children));
711 * Check to see if any process groups have become orphaned
712 * as a result of our exiting, and if they have any stopped
713 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
715 * Case i: Our father is in a different pgrp than we are
716 * and we were the only connection outside, so our pgrp
717 * is about to become orphaned.
720 t = tsk->real_parent;
722 if ((process_group(t) != process_group(tsk)) &&
723 (t->signal->session == tsk->signal->session) &&
724 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
725 has_stopped_jobs(process_group(tsk))) {
726 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
727 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
730 /* Let father know we died
732 * Thread signals are configurable, but you aren't going to use
733 * that to send signals to arbitary processes.
734 * That stops right now.
736 * If the parent exec id doesn't match the exec id we saved
737 * when we started then we know the parent has changed security
740 * If our self_exec id doesn't match our parent_exec_id then
741 * we have changed execution domain as these two values started
742 * the same after a fork.
746 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
747 ( tsk->parent_exec_id != t->self_exec_id ||
748 tsk->self_exec_id != tsk->parent_exec_id)
749 && !capable(CAP_KILL))
750 tsk->exit_signal = SIGCHLD;
753 /* If something other than our normal parent is ptracing us, then
754 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
755 * only has special meaning to our real parent.
757 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
758 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
759 do_notify_parent(tsk, signal);
760 } else if (tsk->ptrace) {
761 do_notify_parent(tsk, SIGCHLD);
765 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
768 tsk->flags |= PF_DEAD;
771 * Clear these here so that update_process_times() won't try to deliver
772 * itimer, profile or rlimit signals to this task while it is in late exit.
774 tsk->it_virt_value = 0;
775 tsk->it_prof_value = 0;
776 tsk->rlim[RLIMIT_CPU].rlim_cur = RLIM_INFINITY;
779 * In the preemption case it must be impossible for the task
780 * to get runnable again, so use "_raw_" unlock to keep
781 * preempt_count elevated until we schedule().
783 * To avoid deadlock on SMP, interrupts must be unmasked. If we
784 * don't, subsequently called functions (e.g, wait_task_inactive()
785 * via release_task()) will spin, with interrupt flags
786 * unwittingly blocked, until the other task sleeps. That task
787 * may itself be waiting for smp_call_function() to answer and
788 * complete, and with interrupts blocked that will never happen.
790 _raw_write_unlock(&tasklist_lock);
793 list_for_each_safe(_p, _n, &ptrace_dead) {
795 t = list_entry(_p,struct task_struct,ptrace_list);
799 /* If the process is dead, release it - nobody will wait for it */
800 if (state == TASK_DEAD)
805 asmlinkage NORET_TYPE void do_exit(long code)
807 struct task_struct *tsk = current;
809 if (unlikely(in_interrupt()))
810 panic("Aiee, killing interrupt handler!");
811 if (unlikely(!tsk->pid))
812 panic("Attempted to kill the idle task!");
813 if (unlikely(tsk->pid == 1))
814 panic("Attempted to kill init!");
817 tsk->flags |= PF_EXITING;
818 del_timer_sync(&tsk->real_timer);
820 if (unlikely(in_atomic()))
821 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
822 current->comm, current->pid,
825 profile_exit_task(tsk);
827 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
828 current->ptrace_message = code;
829 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
841 if (tsk->signal->leader)
842 disassociate_ctty(1);
844 module_put(tsk->thread_info->exec_domain->module);
846 module_put(tsk->binfmt->module);
848 tsk->exit_code = code;
851 mpol_free(tsk->mempolicy);
852 tsk->mempolicy = NULL;
856 /* Avoid "noreturn function does return". */
860 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
868 EXPORT_SYMBOL(complete_and_exit);
870 asmlinkage long sys_exit(int error_code)
872 do_exit((error_code&0xff)<<8);
875 task_t fastcall *next_thread(const task_t *p)
877 const struct pid_link *link = p->pids + PIDTYPE_TGID;
878 const struct list_head *tmp, *head = &link->pidptr->task_list;
883 if (!spin_is_locked(&p->sighand->siglock) &&
884 !rwlock_is_locked(&tasklist_lock))
887 tmp = link->pid_chain.next;
891 return pid_task(tmp, PIDTYPE_TGID);
894 EXPORT_SYMBOL(next_thread);
897 * Take down every thread in the group. This is called by fatal signals
898 * as well as by sys_exit_group (below).
901 do_group_exit(int exit_code)
903 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
905 if (current->signal->group_exit)
906 exit_code = current->signal->group_exit_code;
907 else if (!thread_group_empty(current)) {
908 struct signal_struct *const sig = current->signal;
909 struct sighand_struct *const sighand = current->sighand;
910 read_lock(&tasklist_lock);
911 spin_lock_irq(&sighand->siglock);
913 /* Another thread got here before we took the lock. */
914 exit_code = sig->group_exit_code;
917 sig->group_exit_code = exit_code;
918 zap_other_threads(current);
920 spin_unlock_irq(&sighand->siglock);
921 read_unlock(&tasklist_lock);
929 * this kills every thread in the thread group. Note that any externally
930 * wait4()-ing process will get the correct exit code - even if this
931 * thread is not the thread group leader.
933 asmlinkage void sys_exit_group(int error_code)
935 do_group_exit((error_code & 0xff) << 8);
938 static int eligible_child(pid_t pid, int options, task_t *p)
944 if (process_group(p) != process_group(current))
946 } else if (pid != -1) {
947 if (process_group(p) != -pid)
952 * Do not consider detached threads that are
955 if (p->exit_signal == -1 && !p->ptrace)
958 /* Wait for all children (clone and not) if __WALL is set;
959 * otherwise, wait for clone children *only* if __WCLONE is
960 * set; otherwise, wait for non-clone children *only*. (Note:
961 * A "clone" child here is one that reports to its parent
962 * using a signal other than SIGCHLD.) */
963 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
964 && !(options & __WALL))
967 * Do not consider thread group leaders that are
968 * in a non-empty thread group:
970 if (current->tgid != p->tgid && delay_group_leader(p))
973 if (security_task_wait(p))
980 * Handle sys_wait4 work for one task in state TASK_ZOMBIE. We hold
981 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
982 * the lock and this task is uninteresting. If we return nonzero, we have
983 * released the lock and the system call should return.
985 static int wait_task_zombie(task_t *p, unsigned int __user *stat_addr, struct rusage __user *ru)
991 * Try to move the task's state to DEAD
992 * only one thread is allowed to do this:
994 state = xchg(&p->state, TASK_DEAD);
995 if (state != TASK_ZOMBIE) {
996 BUG_ON(state != TASK_DEAD);
999 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1001 * This can only happen in a race with a ptraced thread
1002 * dying on another processor.
1007 * Now we are sure this task is interesting, and no other
1008 * thread can reap it because we set its state to TASK_DEAD.
1010 read_unlock(&tasklist_lock);
1012 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1013 if (!retval && stat_addr) {
1014 if (p->signal->group_exit)
1015 retval = put_user(p->signal->group_exit_code, stat_addr);
1017 retval = put_user(p->exit_code, stat_addr);
1020 p->state = TASK_ZOMBIE;
1024 if (p->real_parent != p->parent) {
1025 write_lock_irq(&tasklist_lock);
1026 /* Double-check with lock held. */
1027 if (p->real_parent != p->parent) {
1029 p->state = TASK_ZOMBIE;
1031 * If this is not a detached task, notify the parent. If it's
1032 * still not detached after that, don't release it now.
1034 if (p->exit_signal != -1) {
1035 do_notify_parent(p, p->exit_signal);
1036 if (p->exit_signal != -1)
1040 write_unlock_irq(&tasklist_lock);
1049 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1050 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1051 * the lock and this task is uninteresting. If we return nonzero, we have
1052 * released the lock and the system call should return.
1054 static int wait_task_stopped(task_t *p, int delayed_group_leader,
1055 unsigned int __user *stat_addr,
1056 struct rusage __user *ru)
1058 int retval, exit_code;
1062 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1063 p->signal && p->signal->group_stop_count > 0)
1065 * A group stop is in progress and this is the group leader.
1066 * We won't report until all threads have stopped.
1071 * Now we are pretty sure this task is interesting.
1072 * Make sure it doesn't get reaped out from under us while we
1073 * give up the lock and then examine it below. We don't want to
1074 * keep holding onto the tasklist_lock while we call getrusage and
1075 * possibly take page faults for user memory.
1078 read_unlock(&tasklist_lock);
1079 write_lock_irq(&tasklist_lock);
1082 * This uses xchg to be atomic with the thread resuming and setting
1083 * it. It must also be done with the write lock held to prevent a
1084 * race with the TASK_ZOMBIE case.
1086 exit_code = xchg(&p->exit_code, 0);
1087 if (unlikely(p->state > TASK_STOPPED)) {
1089 * The task resumed and then died. Let the next iteration
1090 * catch it in TASK_ZOMBIE. Note that exit_code might
1091 * already be zero here if it resumed and did _exit(0).
1092 * The task itself is dead and won't touch exit_code again;
1093 * other processors in this function are locked out.
1095 p->exit_code = exit_code;
1098 if (unlikely(exit_code == 0)) {
1100 * Another thread in this function got to it first, or it
1101 * resumed, or it resumed and then died.
1103 write_unlock_irq(&tasklist_lock);
1105 read_lock(&tasklist_lock);
1109 /* move to end of parent's list to avoid starvation */
1111 add_parent(p, p->parent);
1113 write_unlock_irq(&tasklist_lock);
1115 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1116 if (!retval && stat_addr)
1117 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1126 asmlinkage long sys_wait4(pid_t pid,unsigned int __user *stat_addr, int options, struct rusage __user *ru)
1128 DECLARE_WAITQUEUE(wait, current);
1129 struct task_struct *tsk;
1132 if (options & ~(WNOHANG|WUNTRACED|__WNOTHREAD|__WCLONE|__WALL))
1135 add_wait_queue(¤t->wait_chldexit,&wait);
1138 current->state = TASK_INTERRUPTIBLE;
1139 read_lock(&tasklist_lock);
1142 struct task_struct *p;
1143 struct list_head *_p;
1146 list_for_each(_p,&tsk->children) {
1147 p = list_entry(_p,struct task_struct,sibling);
1149 ret = eligible_child(pid, options, p);
1156 if (!(options & WUNTRACED) &&
1157 !(p->ptrace & PT_PTRACED))
1159 retval = wait_task_stopped(p, ret == 2,
1161 if (retval != 0) /* He released the lock. */
1166 * Eligible but we cannot release it yet:
1170 retval = wait_task_zombie(p, stat_addr, ru);
1171 if (retval != 0) /* He released the lock. */
1177 list_for_each (_p,&tsk->ptrace_children) {
1178 p = list_entry(_p,struct task_struct,ptrace_list);
1179 if (!eligible_child(pid, options, p))
1185 if (options & __WNOTHREAD)
1187 tsk = next_thread(tsk);
1188 if (tsk->signal != current->signal)
1190 } while (tsk != current);
1191 read_unlock(&tasklist_lock);
1194 if (options & WNOHANG)
1196 retval = -ERESTARTSYS;
1197 if (signal_pending(current))
1204 current->state = TASK_RUNNING;
1205 remove_wait_queue(¤t->wait_chldexit,&wait);
1209 #ifdef __ARCH_WANT_SYS_WAITPID
1212 * sys_waitpid() remains for compatibility. waitpid() should be
1213 * implemented by calling sys_wait4() from libc.a.
1215 asmlinkage long sys_waitpid(pid_t pid, unsigned __user *stat_addr, int options)
1217 return sys_wait4(pid, stat_addr, options, NULL);