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 vx_nproc_dec(p->vx_info);
45 // atomic_dec(&p->vx_info->limit.res[RLIMIT_NPROC]);
47 detach_pid(p, PIDTYPE_PID);
48 detach_pid(p, PIDTYPE_TGID);
49 if (thread_group_leader(p)) {
50 detach_pid(p, PIDTYPE_PGID);
51 detach_pid(p, PIDTYPE_SID);
53 __get_cpu_var(process_counts)--;
59 void release_task(struct task_struct * p)
63 struct dentry *proc_dentry;
66 BUG_ON(p->state < TASK_ZOMBIE);
68 atomic_dec(&p->user->processes);
69 spin_lock(&p->proc_lock);
70 proc_dentry = proc_pid_unhash(p);
71 write_lock_irq(&tasklist_lock);
72 if (unlikely(p->ptrace))
74 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
80 * If we are the last non-leader member of the thread
81 * group, and the leader is zombie, then notify the
82 * group leader's parent process. (if it wants notification.)
85 leader = p->group_leader;
86 if (leader != p && thread_group_empty(leader) && leader->state == TASK_ZOMBIE) {
87 BUG_ON(leader->exit_signal == -1);
88 do_notify_parent(leader, leader->exit_signal);
90 * If we were the last child thread and the leader has
91 * exited already, and the leader's parent ignores SIGCHLD,
92 * then we are the one who should release the leader.
94 * do_notify_parent() will have marked it self-reaping in
97 zap_leader = (leader->exit_signal == -1);
100 p->parent->cutime += p->utime + p->cutime;
101 p->parent->cstime += p->stime + p->cstime;
102 p->parent->cmin_flt += p->min_flt + p->cmin_flt;
103 p->parent->cmaj_flt += p->maj_flt + p->cmaj_flt;
104 p->parent->cnvcsw += p->nvcsw + p->cnvcsw;
105 p->parent->cnivcsw += p->nivcsw + p->cnivcsw;
107 write_unlock_irq(&tasklist_lock);
108 spin_unlock(&p->proc_lock);
109 proc_pid_flush(proc_dentry);
114 if (unlikely(zap_leader))
118 /* we are using it only for SMP init */
120 void unhash_process(struct task_struct *p)
122 struct dentry *proc_dentry;
124 spin_lock(&p->proc_lock);
125 proc_dentry = proc_pid_unhash(p);
126 write_lock_irq(&tasklist_lock);
128 write_unlock_irq(&tasklist_lock);
129 spin_unlock(&p->proc_lock);
130 proc_pid_flush(proc_dentry);
134 * This checks not only the pgrp, but falls back on the pid if no
135 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
138 int session_of_pgrp(int pgrp)
140 struct task_struct *p;
145 read_lock(&tasklist_lock);
146 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid)
147 if (p->signal->session > 0) {
148 sid = p->signal->session;
151 p = find_task_by_pid(pgrp);
153 sid = p->signal->session;
155 read_unlock(&tasklist_lock);
161 * Determine if a process group is "orphaned", according to the POSIX
162 * definition in 2.2.2.52. Orphaned process groups are not to be affected
163 * by terminal-generated stop signals. Newly orphaned process groups are
164 * to receive a SIGHUP and a SIGCONT.
166 * "I ask you, have you ever known what it is to be an orphan?"
168 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
170 struct task_struct *p;
175 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
176 if (p == ignored_task
177 || p->state >= TASK_ZOMBIE
178 || p->real_parent->pid == 1)
180 if (process_group(p->real_parent) != pgrp
181 && p->real_parent->signal->session == p->signal->session) {
186 return ret; /* (sighing) "Often!" */
189 int is_orphaned_pgrp(int pgrp)
193 read_lock(&tasklist_lock);
194 retval = will_become_orphaned_pgrp(pgrp, NULL);
195 read_unlock(&tasklist_lock);
200 static inline int has_stopped_jobs(int pgrp)
203 struct task_struct *p;
207 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
208 if (p->state != TASK_STOPPED)
211 /* If p is stopped by a debugger on a signal that won't
212 stop it, then don't count p as stopped. This isn't
213 perfect but it's a good approximation. */
214 if (unlikely (p->ptrace)
215 && p->exit_code != SIGSTOP
216 && p->exit_code != SIGTSTP
217 && p->exit_code != SIGTTOU
218 && p->exit_code != SIGTTIN)
228 * reparent_to_init() - Reparent the calling kernel thread to the init task.
230 * If a kernel thread is launched as a result of a system call, or if
231 * it ever exits, it should generally reparent itself to init so that
232 * it is correctly cleaned up on exit.
234 * The various task state such as scheduling policy and priority may have
235 * been inherited from a user process, so we reset them to sane values here.
237 * NOTE that reparent_to_init() gives the caller full capabilities.
239 void reparent_to_init(void)
241 write_lock_irq(&tasklist_lock);
243 ptrace_unlink(current);
244 /* Reparent to init */
245 REMOVE_LINKS(current);
246 /* FIXME handle vchild_reaper/initpid */
247 current->parent = child_reaper;
248 current->real_parent = child_reaper;
251 /* Set the exit signal to SIGCHLD so we signal init on exit */
252 current->exit_signal = SIGCHLD;
254 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
255 set_user_nice(current, 0);
259 security_task_reparent_to_init(current);
260 memcpy(current->rlim, init_task.rlim, sizeof(*(current->rlim)));
261 atomic_inc(&(INIT_USER->__count));
262 switch_uid(INIT_USER);
264 write_unlock_irq(&tasklist_lock);
267 void __set_special_pids(pid_t session, pid_t pgrp)
269 struct task_struct *curr = current;
271 if (curr->signal->session != session) {
272 detach_pid(curr, PIDTYPE_SID);
273 curr->signal->session = session;
274 attach_pid(curr, PIDTYPE_SID, session);
276 if (process_group(curr) != pgrp) {
277 detach_pid(curr, PIDTYPE_PGID);
278 curr->signal->pgrp = pgrp;
279 attach_pid(curr, PIDTYPE_PGID, pgrp);
283 void set_special_pids(pid_t session, pid_t pgrp)
285 write_lock_irq(&tasklist_lock);
286 __set_special_pids(session, pgrp);
287 write_unlock_irq(&tasklist_lock);
291 * Let kernel threads use this to say that they
292 * allow a certain signal (since daemonize() will
293 * have disabled all of them by default).
295 int allow_signal(int sig)
297 if (sig < 1 || sig > _NSIG)
300 spin_lock_irq(¤t->sighand->siglock);
301 sigdelset(¤t->blocked, sig);
303 /* Kernel threads handle their own signals.
304 Let the signal code know it'll be handled, so
305 that they don't get converted to SIGKILL or
306 just silently dropped */
307 current->sighand->action[(sig)-1].sa.sa_handler = (void *)2;
310 spin_unlock_irq(¤t->sighand->siglock);
314 EXPORT_SYMBOL(allow_signal);
316 int disallow_signal(int sig)
318 if (sig < 1 || sig > _NSIG)
321 spin_lock_irq(¤t->sighand->siglock);
322 sigaddset(¤t->blocked, sig);
324 spin_unlock_irq(¤t->sighand->siglock);
328 EXPORT_SYMBOL(disallow_signal);
331 * Put all the gunge required to become a kernel thread without
332 * attached user resources in one place where it belongs.
335 void daemonize(const char *name, ...)
338 struct fs_struct *fs;
341 va_start(args, name);
342 vsnprintf(current->comm, sizeof(current->comm), name, args);
346 * If we were started as result of loading a module, close all of the
347 * user space pages. We don't need them, and if we didn't close them
348 * they would be locked into memory.
352 set_special_pids(1, 1);
353 current->signal->tty = NULL;
355 /* Block and flush all signals */
356 sigfillset(&blocked);
357 sigprocmask(SIG_BLOCK, &blocked, NULL);
358 flush_signals(current);
360 /* Become as one with the init task */
362 exit_fs(current); /* current->fs->count--; */
365 atomic_inc(&fs->count);
367 current->files = init_task.files;
368 atomic_inc(¤t->files->count);
373 EXPORT_SYMBOL(daemonize);
375 static inline void close_files(struct files_struct * files)
383 if (i >= files->max_fdset || i >= files->max_fds)
385 set = files->open_fds->fds_bits[j++];
388 struct file * file = xchg(&files->fd[i], NULL);
390 filp_close(file, files);
399 struct files_struct *get_files_struct(struct task_struct *task)
401 struct files_struct *files;
406 atomic_inc(&files->count);
412 void fastcall put_files_struct(struct files_struct *files)
414 if (atomic_dec_and_test(&files->count)) {
417 * Free the fd and fdset arrays if we expanded them.
419 if (files->fd != &files->fd_array[0])
420 free_fd_array(files->fd, files->max_fds);
421 if (files->max_fdset > __FD_SETSIZE) {
422 free_fdset(files->open_fds, files->max_fdset);
423 free_fdset(files->close_on_exec, files->max_fdset);
425 kmem_cache_free(files_cachep, files);
429 EXPORT_SYMBOL(put_files_struct);
431 static inline void __exit_files(struct task_struct *tsk)
433 struct files_struct * files = tsk->files;
439 put_files_struct(files);
443 void exit_files(struct task_struct *tsk)
448 static inline void __put_fs_struct(struct fs_struct *fs)
450 /* No need to hold fs->lock if we are killing it */
451 if (atomic_dec_and_test(&fs->count)) {
458 mntput(fs->altrootmnt);
460 kmem_cache_free(fs_cachep, fs);
464 void put_fs_struct(struct fs_struct *fs)
469 static inline void __exit_fs(struct task_struct *tsk)
471 struct fs_struct * fs = tsk->fs;
481 void exit_fs(struct task_struct *tsk)
486 EXPORT_SYMBOL_GPL(exit_fs);
489 * Turn us into a lazy TLB process if we
492 static inline void __exit_mm(struct task_struct * tsk)
494 struct mm_struct *mm = tsk->mm;
500 * Serialize with any possible pending coredump.
501 * We must hold mmap_sem around checking core_waiters
502 * and clearing tsk->mm. The core-inducing thread
503 * will increment core_waiters for each thread in the
504 * group with ->mm != NULL.
506 down_read(&mm->mmap_sem);
507 if (mm->core_waiters) {
508 up_read(&mm->mmap_sem);
509 down_write(&mm->mmap_sem);
510 if (!--mm->core_waiters)
511 complete(mm->core_startup_done);
512 up_write(&mm->mmap_sem);
514 wait_for_completion(&mm->core_done);
515 down_read(&mm->mmap_sem);
517 atomic_inc(&mm->mm_count);
518 if (mm != tsk->active_mm) BUG();
519 /* more a memory barrier than a real lock */
522 up_read(&mm->mmap_sem);
523 enter_lazy_tlb(mm, current);
528 void exit_mm(struct task_struct *tsk)
533 EXPORT_SYMBOL(exit_mm);
535 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
538 * Make sure we're not reparenting to ourselves and that
539 * the parent is not a zombie.
541 if (p == reaper || reaper->state >= TASK_ZOMBIE)
542 p->real_parent = child_reaper;
544 p->real_parent = reaper;
545 if (p->parent == p->real_parent)
549 static inline void reparent_thread(task_t *p, task_t *father, int traced)
551 /* We don't want people slaying init. */
552 if (p->exit_signal != -1)
553 p->exit_signal = SIGCHLD;
556 if (p->pdeath_signal)
557 /* We already hold the tasklist_lock here. */
558 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
560 /* Move the child from its dying parent to the new one. */
561 if (unlikely(traced)) {
562 /* Preserve ptrace links if someone else is tracing this child. */
563 list_del_init(&p->ptrace_list);
564 if (p->parent != p->real_parent)
565 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
567 /* If this child is being traced, then we're the one tracing it
568 * anyway, so let go of it.
571 list_del_init(&p->sibling);
572 p->parent = p->real_parent;
573 list_add_tail(&p->sibling, &p->parent->children);
575 /* If we'd notified the old parent about this child's death,
576 * also notify the new parent.
578 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
579 thread_group_empty(p))
580 do_notify_parent(p, p->exit_signal);
584 * process group orphan check
585 * Case ii: Our child is in a different pgrp
586 * than we are, and it was the only connection
587 * outside, so the child pgrp is now orphaned.
589 if ((process_group(p) != process_group(father)) &&
590 (p->signal->session == father->signal->session)) {
591 int pgrp = process_group(p);
593 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
594 __kill_pg_info(SIGHUP, (void *)1, pgrp);
595 __kill_pg_info(SIGCONT, (void *)1, pgrp);
601 * When we die, we re-parent all our children.
602 * Try to give them to another thread in our thread
603 * group, and if no such member exists, give it to
604 * the global child reaper process (ie "init")
606 static inline void forget_original_parent(struct task_struct * father)
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) {
625 p = list_entry(_p,struct task_struct,sibling);
626 if (father == p->real_parent) {
627 choose_new_parent(p, reaper, child_reaper);
628 reparent_thread(p, father, 0);
631 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
632 thread_group_empty(p))
633 do_notify_parent(p, p->exit_signal);
636 list_for_each_safe(_p, _n, &father->ptrace_children) {
637 p = list_entry(_p,struct task_struct,ptrace_list);
638 choose_new_parent(p, reaper, child_reaper);
639 reparent_thread(p, father, 1);
644 * Send signals to all our closest relatives so that they know
645 * to properly mourn us..
647 static void exit_notify(struct task_struct *tsk)
650 struct task_struct *t;
652 if (signal_pending(tsk) && !tsk->signal->group_exit
653 && !thread_group_empty(tsk)) {
655 * This occurs when there was a race between our exit
656 * syscall and a group signal choosing us as the one to
657 * wake up. It could be that we are the only thread
658 * alerted to check for pending signals, but another thread
659 * should be woken now to take the signal since we will not.
660 * Now we'll wake all the threads in the group just to make
661 * sure someone gets all the pending signals.
663 read_lock(&tasklist_lock);
664 spin_lock_irq(&tsk->sighand->siglock);
665 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
666 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
667 recalc_sigpending_tsk(t);
668 if (signal_pending(t))
669 signal_wake_up(t, 0);
671 spin_unlock_irq(&tsk->sighand->siglock);
672 read_unlock(&tasklist_lock);
675 write_lock_irq(&tasklist_lock);
678 * This does two things:
680 * A. Make init inherit all the child processes
681 * B. Check to see if any process groups have become orphaned
682 * as a result of our exiting, and if they have any stopped
683 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
686 forget_original_parent(tsk);
687 BUG_ON(!list_empty(&tsk->children));
690 * Check to see if any process groups have become orphaned
691 * as a result of our exiting, and if they have any stopped
692 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
694 * Case i: Our father is in a different pgrp than we are
695 * and we were the only connection outside, so our pgrp
696 * is about to become orphaned.
699 t = tsk->real_parent;
701 if ((process_group(t) != process_group(tsk)) &&
702 (t->signal->session == tsk->signal->session) &&
703 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
704 has_stopped_jobs(process_group(tsk))) {
705 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
706 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
709 /* Let father know we died
711 * Thread signals are configurable, but you aren't going to use
712 * that to send signals to arbitary processes.
713 * That stops right now.
715 * If the parent exec id doesn't match the exec id we saved
716 * when we started then we know the parent has changed security
719 * If our self_exec id doesn't match our parent_exec_id then
720 * we have changed execution domain as these two values started
721 * the same after a fork.
725 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
726 ( tsk->parent_exec_id != t->self_exec_id ||
727 tsk->self_exec_id != tsk->parent_exec_id)
728 && !capable(CAP_KILL))
729 tsk->exit_signal = SIGCHLD;
732 /* If something other than our normal parent is ptracing us, then
733 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
734 * only has special meaning to our real parent.
736 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
737 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
738 do_notify_parent(tsk, signal);
739 } else if (tsk->ptrace) {
740 do_notify_parent(tsk, SIGCHLD);
744 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
747 tsk->flags |= PF_DEAD;
750 * Clear these here so that update_process_times() won't try to deliver
751 * itimer, profile or rlimit signals to this task while it is in late exit.
753 tsk->it_virt_value = 0;
754 tsk->it_prof_value = 0;
755 tsk->rlim[RLIMIT_CPU].rlim_cur = RLIM_INFINITY;
758 * In the preemption case it must be impossible for the task
759 * to get runnable again, so use "_raw_" unlock to keep
760 * preempt_count elevated until we schedule().
762 * To avoid deadlock on SMP, interrupts must be unmasked. If we
763 * don't, subsequently called functions (e.g, wait_task_inactive()
764 * via release_task()) will spin, with interrupt flags
765 * unwittingly blocked, until the other task sleeps. That task
766 * may itself be waiting for smp_call_function() to answer and
767 * complete, and with interrupts blocked that will never happen.
769 _raw_write_unlock(&tasklist_lock);
772 /* If the process is dead, release it - nobody will wait for it */
773 if (state == TASK_DEAD)
778 asmlinkage NORET_TYPE void do_exit(long code)
780 struct task_struct *tsk = current;
782 if (unlikely(in_interrupt()))
783 panic("Aiee, killing interrupt handler!");
784 if (unlikely(!tsk->pid))
785 panic("Attempted to kill the idle task!");
786 if (unlikely(tsk->pid == 1))
787 panic("Attempted to kill init!");
790 tsk->flags |= PF_EXITING;
791 del_timer_sync(&tsk->real_timer);
793 if (unlikely(in_atomic()))
794 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
795 current->comm, current->pid,
798 profile_exit_task(tsk);
800 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
801 current->ptrace_message = code;
802 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
814 mpol_free(tsk->mempolicy);
817 if (tsk->signal->leader)
818 disassociate_ctty(1);
820 module_put(tsk->thread_info->exec_domain->module);
822 module_put(tsk->binfmt->module);
824 tsk->exit_code = code;
828 /* Avoid "noreturn function does return". */
832 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
840 EXPORT_SYMBOL(complete_and_exit);
842 asmlinkage long sys_exit(int error_code)
844 do_exit((error_code&0xff)<<8);
847 task_t fastcall *next_thread(task_t *p)
849 struct pid_link *link = p->pids + PIDTYPE_TGID;
850 struct list_head *tmp, *head = &link->pidptr->task_list;
855 if (!spin_is_locked(&p->sighand->siglock) &&
856 !rwlock_is_locked(&tasklist_lock))
859 tmp = link->pid_chain.next;
863 return pid_task(tmp, PIDTYPE_TGID);
866 EXPORT_SYMBOL(next_thread);
869 * Take down every thread in the group. This is called by fatal signals
870 * as well as by sys_exit_group (below).
873 do_group_exit(int exit_code)
875 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
877 if (current->signal->group_exit)
878 exit_code = current->signal->group_exit_code;
879 else if (!thread_group_empty(current)) {
880 struct signal_struct *const sig = current->signal;
881 struct sighand_struct *const sighand = current->sighand;
882 read_lock(&tasklist_lock);
883 spin_lock_irq(&sighand->siglock);
885 /* Another thread got here before we took the lock. */
886 exit_code = sig->group_exit_code;
889 sig->group_exit_code = exit_code;
890 zap_other_threads(current);
892 spin_unlock_irq(&sighand->siglock);
893 read_unlock(&tasklist_lock);
901 * this kills every thread in the thread group. Note that any externally
902 * wait4()-ing process will get the correct exit code - even if this
903 * thread is not the thread group leader.
905 asmlinkage void sys_exit_group(int error_code)
907 do_group_exit((error_code & 0xff) << 8);
910 static int eligible_child(pid_t pid, int options, task_t *p)
916 if (process_group(p) != process_group(current))
918 } else if (pid != -1) {
919 if (process_group(p) != -pid)
924 * Do not consider detached threads that are
927 if (p->exit_signal == -1 && !p->ptrace)
930 /* Wait for all children (clone and not) if __WALL is set;
931 * otherwise, wait for clone children *only* if __WCLONE is
932 * set; otherwise, wait for non-clone children *only*. (Note:
933 * A "clone" child here is one that reports to its parent
934 * using a signal other than SIGCHLD.) */
935 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
936 && !(options & __WALL))
939 * Do not consider thread group leaders that are
940 * in a non-empty thread group:
942 if (current->tgid != p->tgid && delay_group_leader(p))
945 if (security_task_wait(p))
952 * Handle sys_wait4 work for one task in state TASK_ZOMBIE. We hold
953 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
954 * the lock and this task is uninteresting. If we return nonzero, we have
955 * released the lock and the system call should return.
957 static int wait_task_zombie(task_t *p, unsigned int __user *stat_addr, struct rusage __user *ru)
963 * Try to move the task's state to DEAD
964 * only one thread is allowed to do this:
966 state = xchg(&p->state, TASK_DEAD);
967 if (state != TASK_ZOMBIE) {
968 BUG_ON(state != TASK_DEAD);
971 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
973 * This can only happen in a race with a ptraced thread
974 * dying on another processor.
979 * Now we are sure this task is interesting, and no other
980 * thread can reap it because we set its state to TASK_DEAD.
982 read_unlock(&tasklist_lock);
984 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
985 if (!retval && stat_addr) {
986 if (p->signal->group_exit)
987 retval = put_user(p->signal->group_exit_code, stat_addr);
989 retval = put_user(p->exit_code, stat_addr);
992 p->state = TASK_ZOMBIE;
996 if (p->real_parent != p->parent) {
997 write_lock_irq(&tasklist_lock);
998 /* Double-check with lock held. */
999 if (p->real_parent != p->parent) {
1001 p->state = TASK_ZOMBIE;
1002 /* If this is a detached thread, this is where it goes away. */
1003 if (p->exit_signal == -1) {
1004 /* release_task takes the lock itself. */
1005 write_unlock_irq(&tasklist_lock);
1009 do_notify_parent(p, p->exit_signal);
1010 write_unlock_irq(&tasklist_lock);
1015 write_unlock_irq(&tasklist_lock);
1024 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1025 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1026 * the lock and this task is uninteresting. If we return nonzero, we have
1027 * released the lock and the system call should return.
1029 static int wait_task_stopped(task_t *p, int delayed_group_leader,
1030 unsigned int __user *stat_addr,
1031 struct rusage __user *ru)
1033 int retval, exit_code;
1037 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1038 p->signal && p->signal->group_stop_count > 0)
1040 * A group stop is in progress and this is the group leader.
1041 * We won't report until all threads have stopped.
1046 * Now we are pretty sure this task is interesting.
1047 * Make sure it doesn't get reaped out from under us while we
1048 * give up the lock and then examine it below. We don't want to
1049 * keep holding onto the tasklist_lock while we call getrusage and
1050 * possibly take page faults for user memory.
1053 read_unlock(&tasklist_lock);
1054 write_lock_irq(&tasklist_lock);
1057 * This uses xchg to be atomic with the thread resuming and setting
1058 * it. It must also be done with the write lock held to prevent a
1059 * race with the TASK_ZOMBIE case.
1061 exit_code = xchg(&p->exit_code, 0);
1062 if (unlikely(p->state > TASK_STOPPED)) {
1064 * The task resumed and then died. Let the next iteration
1065 * catch it in TASK_ZOMBIE. Note that exit_code might
1066 * already be zero here if it resumed and did _exit(0).
1067 * The task itself is dead and won't touch exit_code again;
1068 * other processors in this function are locked out.
1070 p->exit_code = exit_code;
1073 if (unlikely(exit_code == 0)) {
1075 * Another thread in this function got to it first, or it
1076 * resumed, or it resumed and then died.
1078 write_unlock_irq(&tasklist_lock);
1080 read_lock(&tasklist_lock);
1084 /* move to end of parent's list to avoid starvation */
1086 add_parent(p, p->parent);
1088 write_unlock_irq(&tasklist_lock);
1090 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1091 if (!retval && stat_addr)
1092 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1101 asmlinkage long sys_wait4(pid_t pid,unsigned int __user *stat_addr, int options, struct rusage __user *ru)
1103 DECLARE_WAITQUEUE(wait, current);
1104 struct task_struct *tsk;
1107 if (options & ~(WNOHANG|WUNTRACED|__WNOTHREAD|__WCLONE|__WALL))
1110 add_wait_queue(¤t->wait_chldexit,&wait);
1113 current->state = TASK_INTERRUPTIBLE;
1114 read_lock(&tasklist_lock);
1117 struct task_struct *p;
1118 struct list_head *_p;
1121 list_for_each(_p,&tsk->children) {
1122 p = list_entry(_p,struct task_struct,sibling);
1124 ret = eligible_child(pid, options, p);
1131 if (!(options & WUNTRACED) &&
1132 !(p->ptrace & PT_PTRACED))
1134 retval = wait_task_stopped(p, ret == 2,
1136 if (retval != 0) /* He released the lock. */
1141 * Eligible but we cannot release it yet:
1145 retval = wait_task_zombie(p, stat_addr, ru);
1146 if (retval != 0) /* He released the lock. */
1152 list_for_each (_p,&tsk->ptrace_children) {
1153 p = list_entry(_p,struct task_struct,ptrace_list);
1154 if (!eligible_child(pid, options, p))
1160 if (options & __WNOTHREAD)
1162 tsk = next_thread(tsk);
1163 if (tsk->signal != current->signal)
1165 } while (tsk != current);
1166 read_unlock(&tasklist_lock);
1169 if (options & WNOHANG)
1171 retval = -ERESTARTSYS;
1172 if (signal_pending(current))
1179 current->state = TASK_RUNNING;
1180 remove_wait_queue(¤t->wait_chldexit,&wait);
1184 #ifdef __ARCH_WANT_SYS_WAITPID
1187 * sys_waitpid() remains for compatibility. waitpid() should be
1188 * implemented by calling sys_wait4() from libc.a.
1190 asmlinkage long sys_waitpid(pid_t pid, unsigned __user *stat_addr, int options)
1192 return sys_wait4(pid, stat_addr, options, NULL);