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/capability.h>
14 #include <linux/completion.h>
15 #include <linux/personality.h>
16 #include <linux/tty.h>
17 #include <linux/namespace.h>
18 #include <linux/key.h>
19 #include <linux/security.h>
20 #include <linux/cpu.h>
21 #include <linux/acct.h>
22 #include <linux/file.h>
23 #include <linux/binfmts.h>
24 #include <linux/ptrace.h>
25 #include <linux/profile.h>
26 #include <linux/mount.h>
27 #include <linux/proc_fs.h>
28 #include <linux/mempolicy.h>
29 #include <linux/cpuset.h>
30 #include <linux/syscalls.h>
31 #include <linux/signal.h>
32 #include <linux/posix-timers.h>
33 #include <linux/cn_proc.h>
34 #include <linux/mutex.h>
35 #include <linux/futex.h>
36 #include <linux/compat.h>
38 #include <asm/uaccess.h>
39 #include <asm/unistd.h>
40 #include <asm/pgtable.h>
41 #include <asm/mmu_context.h>
43 extern void sem_exit (void);
44 extern struct task_struct *child_reaper;
46 int getrusage(struct task_struct *, int, struct rusage __user *);
48 static void exit_mm(struct task_struct * tsk);
50 static void __unhash_process(struct task_struct *p)
53 detach_pid(p, PIDTYPE_PID);
54 if (thread_group_leader(p)) {
55 detach_pid(p, PIDTYPE_PGID);
56 detach_pid(p, PIDTYPE_SID);
58 list_del_init(&p->tasks);
59 __get_cpu_var(process_counts)--;
61 list_del_rcu(&p->thread_group);
66 * This function expects the tasklist_lock write-locked.
68 static void __exit_signal(struct task_struct *tsk)
70 struct signal_struct *sig = tsk->signal;
71 struct sighand_struct *sighand;
74 BUG_ON(!atomic_read(&sig->count));
77 sighand = rcu_dereference(tsk->sighand);
78 spin_lock(&sighand->siglock);
80 posix_cpu_timers_exit(tsk);
81 if (atomic_dec_and_test(&sig->count))
82 posix_cpu_timers_exit_group(tsk);
85 * If there is any task waiting for the group exit
88 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
89 wake_up_process(sig->group_exit_task);
90 sig->group_exit_task = NULL;
92 if (tsk == sig->curr_target)
93 sig->curr_target = next_thread(tsk);
95 * Accumulate here the counters for all threads but the
96 * group leader as they die, so they can be added into
97 * the process-wide totals when those are taken.
98 * The group leader stays around as a zombie as long
99 * as there are other threads. When it gets reaped,
100 * the exit.c code will add its counts into these totals.
101 * We won't ever get here for the group leader, since it
102 * will have been the last reference on the signal_struct.
104 sig->utime = cputime_add(sig->utime, tsk->utime);
105 sig->stime = cputime_add(sig->stime, tsk->stime);
106 sig->min_flt += tsk->min_flt;
107 sig->maj_flt += tsk->maj_flt;
108 sig->nvcsw += tsk->nvcsw;
109 sig->nivcsw += tsk->nivcsw;
110 sig->sched_time += tsk->sched_time;
111 sig = NULL; /* Marker for below. */
114 __unhash_process(tsk);
118 spin_unlock(&sighand->siglock);
121 __cleanup_sighand(sighand);
122 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
123 flush_sigqueue(&tsk->pending);
125 flush_sigqueue(&sig->shared_pending);
126 __cleanup_signal(sig);
130 static void delayed_put_task_struct(struct rcu_head *rhp)
132 put_task_struct(container_of(rhp, struct task_struct, rcu));
135 void release_task(struct task_struct * p)
139 struct dentry *proc_dentry;
142 atomic_dec(&p->user->processes);
143 spin_lock(&p->proc_lock);
144 proc_dentry = proc_pid_unhash(p);
145 write_lock_irq(&tasklist_lock);
147 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
151 * If we are the last non-leader member of the thread
152 * group, and the leader is zombie, then notify the
153 * group leader's parent process. (if it wants notification.)
156 leader = p->group_leader;
157 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
158 BUG_ON(leader->exit_signal == -1);
159 do_notify_parent(leader, leader->exit_signal);
161 * If we were the last child thread and the leader has
162 * exited already, and the leader's parent ignores SIGCHLD,
163 * then we are the one who should release the leader.
165 * do_notify_parent() will have marked it self-reaping in
168 zap_leader = (leader->exit_signal == -1);
172 write_unlock_irq(&tasklist_lock);
173 spin_unlock(&p->proc_lock);
174 proc_pid_flush(proc_dentry);
176 call_rcu(&p->rcu, delayed_put_task_struct);
179 if (unlikely(zap_leader))
184 * This checks not only the pgrp, but falls back on the pid if no
185 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
188 int session_of_pgrp(int pgrp)
190 struct task_struct *p;
193 read_lock(&tasklist_lock);
194 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
195 if (p->signal->session > 0) {
196 sid = p->signal->session;
199 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
200 p = find_task_by_pid(pgrp);
202 sid = p->signal->session;
204 read_unlock(&tasklist_lock);
210 * Determine if a process group is "orphaned", according to the POSIX
211 * definition in 2.2.2.52. Orphaned process groups are not to be affected
212 * by terminal-generated stop signals. Newly orphaned process groups are
213 * to receive a SIGHUP and a SIGCONT.
215 * "I ask you, have you ever known what it is to be an orphan?"
217 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
219 struct task_struct *p;
222 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
223 if (p == ignored_task
225 || p->real_parent->pid == 1)
227 if (process_group(p->real_parent) != pgrp
228 && p->real_parent->signal->session == p->signal->session) {
232 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
233 return ret; /* (sighing) "Often!" */
236 int is_orphaned_pgrp(int pgrp)
240 read_lock(&tasklist_lock);
241 retval = will_become_orphaned_pgrp(pgrp, NULL);
242 read_unlock(&tasklist_lock);
247 static int has_stopped_jobs(int pgrp)
250 struct task_struct *p;
252 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
253 if (p->state != TASK_STOPPED)
256 /* If p is stopped by a debugger on a signal that won't
257 stop it, then don't count p as stopped. This isn't
258 perfect but it's a good approximation. */
259 if (unlikely (p->ptrace)
260 && p->exit_code != SIGSTOP
261 && p->exit_code != SIGTSTP
262 && p->exit_code != SIGTTOU
263 && p->exit_code != SIGTTIN)
268 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
273 * reparent_to_init - Reparent the calling kernel thread to the init task.
275 * If a kernel thread is launched as a result of a system call, or if
276 * it ever exits, it should generally reparent itself to init so that
277 * it is correctly cleaned up on exit.
279 * The various task state such as scheduling policy and priority may have
280 * been inherited from a user process, so we reset them to sane values here.
282 * NOTE that reparent_to_init() gives the caller full capabilities.
284 static void reparent_to_init(void)
286 write_lock_irq(&tasklist_lock);
288 ptrace_unlink(current);
289 /* Reparent to init */
290 remove_parent(current);
291 current->parent = child_reaper;
292 current->real_parent = child_reaper;
295 /* Set the exit signal to SIGCHLD so we signal init on exit */
296 current->exit_signal = SIGCHLD;
298 if ((current->policy == SCHED_NORMAL ||
299 current->policy == SCHED_BATCH)
300 && (task_nice(current) < 0))
301 set_user_nice(current, 0);
305 security_task_reparent_to_init(current);
306 memcpy(current->signal->rlim, init_task.signal->rlim,
307 sizeof(current->signal->rlim));
308 atomic_inc(&(INIT_USER->__count));
309 write_unlock_irq(&tasklist_lock);
310 switch_uid(INIT_USER);
313 void __set_special_pids(pid_t session, pid_t pgrp)
315 struct task_struct *curr = current->group_leader;
317 if (curr->signal->session != session) {
318 detach_pid(curr, PIDTYPE_SID);
319 curr->signal->session = session;
320 attach_pid(curr, PIDTYPE_SID, session);
322 if (process_group(curr) != pgrp) {
323 detach_pid(curr, PIDTYPE_PGID);
324 curr->signal->pgrp = pgrp;
325 attach_pid(curr, PIDTYPE_PGID, pgrp);
329 void set_special_pids(pid_t session, pid_t pgrp)
331 write_lock_irq(&tasklist_lock);
332 __set_special_pids(session, pgrp);
333 write_unlock_irq(&tasklist_lock);
337 * Let kernel threads use this to say that they
338 * allow a certain signal (since daemonize() will
339 * have disabled all of them by default).
341 int allow_signal(int sig)
343 if (!valid_signal(sig) || sig < 1)
346 spin_lock_irq(¤t->sighand->siglock);
347 sigdelset(¤t->blocked, sig);
349 /* Kernel threads handle their own signals.
350 Let the signal code know it'll be handled, so
351 that they don't get converted to SIGKILL or
352 just silently dropped */
353 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
356 spin_unlock_irq(¤t->sighand->siglock);
360 EXPORT_SYMBOL(allow_signal);
362 int disallow_signal(int sig)
364 if (!valid_signal(sig) || sig < 1)
367 spin_lock_irq(¤t->sighand->siglock);
368 sigaddset(¤t->blocked, sig);
370 spin_unlock_irq(¤t->sighand->siglock);
374 EXPORT_SYMBOL(disallow_signal);
377 * Put all the gunge required to become a kernel thread without
378 * attached user resources in one place where it belongs.
381 void daemonize(const char *name, ...)
384 struct fs_struct *fs;
387 va_start(args, name);
388 vsnprintf(current->comm, sizeof(current->comm), name, args);
392 * If we were started as result of loading a module, close all of the
393 * user space pages. We don't need them, and if we didn't close them
394 * they would be locked into memory.
398 set_special_pids(1, 1);
399 mutex_lock(&tty_mutex);
400 current->signal->tty = NULL;
401 mutex_unlock(&tty_mutex);
403 /* Block and flush all signals */
404 sigfillset(&blocked);
405 sigprocmask(SIG_BLOCK, &blocked, NULL);
406 flush_signals(current);
408 /* Become as one with the init task */
410 exit_fs(current); /* current->fs->count--; */
413 atomic_inc(&fs->count);
414 exit_namespace(current);
415 current->namespace = init_task.namespace;
416 get_namespace(current->namespace);
418 current->files = init_task.files;
419 atomic_inc(¤t->files->count);
424 EXPORT_SYMBOL(daemonize);
426 static void close_files(struct files_struct * files)
434 * It is safe to dereference the fd table without RCU or
435 * ->file_lock because this is the last reference to the
438 fdt = files_fdtable(files);
442 if (i >= fdt->max_fdset || i >= fdt->max_fds)
444 set = fdt->open_fds->fds_bits[j++];
447 struct file * file = xchg(&fdt->fd[i], NULL);
449 filp_close(file, files);
457 struct files_struct *get_files_struct(struct task_struct *task)
459 struct files_struct *files;
464 atomic_inc(&files->count);
470 void fastcall put_files_struct(struct files_struct *files)
474 if (atomic_dec_and_test(&files->count)) {
477 * Free the fd and fdset arrays if we expanded them.
478 * If the fdtable was embedded, pass files for freeing
479 * at the end of the RCU grace period. Otherwise,
480 * you can free files immediately.
482 fdt = files_fdtable(files);
483 if (fdt == &files->fdtab)
484 fdt->free_files = files;
486 kmem_cache_free(files_cachep, files);
491 EXPORT_SYMBOL(put_files_struct);
493 static inline void __exit_files(struct task_struct *tsk)
495 struct files_struct * files = tsk->files;
501 put_files_struct(files);
505 void exit_files(struct task_struct *tsk)
510 static inline void __put_fs_struct(struct fs_struct *fs)
512 /* No need to hold fs->lock if we are killing it */
513 if (atomic_dec_and_test(&fs->count)) {
520 mntput(fs->altrootmnt);
522 kmem_cache_free(fs_cachep, fs);
526 void put_fs_struct(struct fs_struct *fs)
531 static inline void __exit_fs(struct task_struct *tsk)
533 struct fs_struct * fs = tsk->fs;
543 void exit_fs(struct task_struct *tsk)
548 EXPORT_SYMBOL_GPL(exit_fs);
551 * Turn us into a lazy TLB process if we
554 static void exit_mm(struct task_struct * tsk)
556 struct mm_struct *mm = tsk->mm;
562 * Serialize with any possible pending coredump.
563 * We must hold mmap_sem around checking core_waiters
564 * and clearing tsk->mm. The core-inducing thread
565 * will increment core_waiters for each thread in the
566 * group with ->mm != NULL.
568 down_read(&mm->mmap_sem);
569 if (mm->core_waiters) {
570 up_read(&mm->mmap_sem);
571 down_write(&mm->mmap_sem);
572 if (!--mm->core_waiters)
573 complete(mm->core_startup_done);
574 up_write(&mm->mmap_sem);
576 wait_for_completion(&mm->core_done);
577 down_read(&mm->mmap_sem);
579 atomic_inc(&mm->mm_count);
580 if (mm != tsk->active_mm) BUG();
581 /* more a memory barrier than a real lock */
584 up_read(&mm->mmap_sem);
585 enter_lazy_tlb(mm, current);
590 static inline void choose_new_parent(task_t *p, task_t *reaper)
593 * Make sure we're not reparenting to ourselves and that
594 * the parent is not a zombie.
596 BUG_ON(p == reaper || reaper->exit_state);
597 p->real_parent = reaper;
600 static void reparent_thread(task_t *p, task_t *father, int traced)
602 /* We don't want people slaying init. */
603 if (p->exit_signal != -1)
604 p->exit_signal = SIGCHLD;
606 if (p->pdeath_signal)
607 /* We already hold the tasklist_lock here. */
608 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
610 /* Move the child from its dying parent to the new one. */
611 if (unlikely(traced)) {
612 /* Preserve ptrace links if someone else is tracing this child. */
613 list_del_init(&p->ptrace_list);
614 if (p->parent != p->real_parent)
615 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
617 /* If this child is being traced, then we're the one tracing it
618 * anyway, so let go of it.
622 p->parent = p->real_parent;
625 /* If we'd notified the old parent about this child's death,
626 * also notify the new parent.
628 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
629 thread_group_empty(p))
630 do_notify_parent(p, p->exit_signal);
631 else if (p->state == TASK_TRACED) {
633 * If it was at a trace stop, turn it into
634 * a normal stop since it's no longer being
642 * process group orphan check
643 * Case ii: Our child is in a different pgrp
644 * than we are, and it was the only connection
645 * outside, so the child pgrp is now orphaned.
647 if ((process_group(p) != process_group(father)) &&
648 (p->signal->session == father->signal->session)) {
649 int pgrp = process_group(p);
651 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
652 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
653 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
659 * When we die, we re-parent all our children.
660 * Try to give them to another thread in our thread
661 * group, and if no such member exists, give it to
662 * the global child reaper process (ie "init")
664 static void forget_original_parent(struct task_struct * father,
665 struct list_head *to_release)
667 struct task_struct *p, *reaper = father;
668 struct list_head *_p, *_n;
671 reaper = next_thread(reaper);
672 if (reaper == father) {
673 reaper = child_reaper;
676 } while (reaper->exit_state);
679 * There are only two places where our children can be:
681 * - in our child list
682 * - in our ptraced child list
684 * Search them and reparent children.
686 list_for_each_safe(_p, _n, &father->children) {
688 p = list_entry(_p,struct task_struct,sibling);
692 /* if father isn't the real parent, then ptrace must be enabled */
693 BUG_ON(father != p->real_parent && !ptrace);
695 if (father == p->real_parent) {
696 /* reparent with a reaper, real father it's us */
697 choose_new_parent(p, reaper);
698 reparent_thread(p, father, 0);
700 /* reparent ptraced task to its real parent */
702 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
703 thread_group_empty(p))
704 do_notify_parent(p, p->exit_signal);
708 * if the ptraced child is a zombie with exit_signal == -1
709 * we must collect it before we exit, or it will remain
710 * zombie forever since we prevented it from self-reap itself
711 * while it was being traced by us, to be able to see it in wait4.
713 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
714 list_add(&p->ptrace_list, to_release);
716 list_for_each_safe(_p, _n, &father->ptrace_children) {
717 p = list_entry(_p,struct task_struct,ptrace_list);
718 choose_new_parent(p, reaper);
719 reparent_thread(p, father, 1);
724 * Send signals to all our closest relatives so that they know
725 * to properly mourn us..
727 static void exit_notify(struct task_struct *tsk)
730 struct task_struct *t;
731 struct list_head ptrace_dead, *_p, *_n;
733 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
734 && !thread_group_empty(tsk)) {
736 * This occurs when there was a race between our exit
737 * syscall and a group signal choosing us as the one to
738 * wake up. It could be that we are the only thread
739 * alerted to check for pending signals, but another thread
740 * should be woken now to take the signal since we will not.
741 * Now we'll wake all the threads in the group just to make
742 * sure someone gets all the pending signals.
744 read_lock(&tasklist_lock);
745 spin_lock_irq(&tsk->sighand->siglock);
746 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
747 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
748 recalc_sigpending_tsk(t);
749 if (signal_pending(t))
750 signal_wake_up(t, 0);
752 spin_unlock_irq(&tsk->sighand->siglock);
753 read_unlock(&tasklist_lock);
756 write_lock_irq(&tasklist_lock);
759 * This does two things:
761 * A. Make init inherit all the child processes
762 * B. Check to see if any process groups have become orphaned
763 * as a result of our exiting, and if they have any stopped
764 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
767 INIT_LIST_HEAD(&ptrace_dead);
768 forget_original_parent(tsk, &ptrace_dead);
769 BUG_ON(!list_empty(&tsk->children));
770 BUG_ON(!list_empty(&tsk->ptrace_children));
773 * Check to see if any process groups have become orphaned
774 * as a result of our exiting, and if they have any stopped
775 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
777 * Case i: Our father is in a different pgrp than we are
778 * and we were the only connection outside, so our pgrp
779 * is about to become orphaned.
782 t = tsk->real_parent;
784 if ((process_group(t) != process_group(tsk)) &&
785 (t->signal->session == tsk->signal->session) &&
786 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
787 has_stopped_jobs(process_group(tsk))) {
788 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
789 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
792 /* Let father know we died
794 * Thread signals are configurable, but you aren't going to use
795 * that to send signals to arbitary processes.
796 * That stops right now.
798 * If the parent exec id doesn't match the exec id we saved
799 * when we started then we know the parent has changed security
802 * If our self_exec id doesn't match our parent_exec_id then
803 * we have changed execution domain as these two values started
804 * the same after a fork.
808 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
809 ( tsk->parent_exec_id != t->self_exec_id ||
810 tsk->self_exec_id != tsk->parent_exec_id)
811 && !capable(CAP_KILL))
812 tsk->exit_signal = SIGCHLD;
815 /* If something other than our normal parent is ptracing us, then
816 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
817 * only has special meaning to our real parent.
819 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
820 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
821 do_notify_parent(tsk, signal);
822 } else if (tsk->ptrace) {
823 do_notify_parent(tsk, SIGCHLD);
827 if (tsk->exit_signal == -1 &&
828 (likely(tsk->ptrace == 0) ||
829 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
831 tsk->exit_state = state;
833 write_unlock_irq(&tasklist_lock);
835 list_for_each_safe(_p, _n, &ptrace_dead) {
837 t = list_entry(_p,struct task_struct,ptrace_list);
841 /* If the process is dead, release it - nobody will wait for it */
842 if (state == EXIT_DEAD)
846 fastcall NORET_TYPE void do_exit(long code)
848 struct task_struct *tsk = current;
851 profile_task_exit(tsk);
853 WARN_ON(atomic_read(&tsk->fs_excl));
855 if (unlikely(in_interrupt()))
856 panic("Aiee, killing interrupt handler!");
857 if (unlikely(!tsk->pid))
858 panic("Attempted to kill the idle task!");
859 if (unlikely(tsk == child_reaper))
860 panic("Attempted to kill init!");
862 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
863 current->ptrace_message = code;
864 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
868 * We're taking recursive faults here in do_exit. Safest is to just
869 * leave this task alone and wait for reboot.
871 if (unlikely(tsk->flags & PF_EXITING)) {
873 "Fixing recursive fault but reboot is needed!\n");
876 set_current_state(TASK_UNINTERRUPTIBLE);
880 tsk->flags |= PF_EXITING;
883 * Make sure we don't try to process any timer firings
884 * while we are already exiting.
886 tsk->it_virt_expires = cputime_zero;
887 tsk->it_prof_expires = cputime_zero;
888 tsk->it_sched_expires = 0;
890 if (unlikely(in_atomic()))
891 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
892 current->comm, current->pid,
895 acct_update_integrals(tsk);
897 update_hiwater_rss(tsk->mm);
898 update_hiwater_vm(tsk->mm);
900 group_dead = atomic_dec_and_test(&tsk->signal->live);
902 hrtimer_cancel(&tsk->signal->real_timer);
903 exit_itimers(tsk->signal);
906 if (unlikely(tsk->robust_list))
907 exit_robust_list(tsk);
909 if (unlikely(tsk->compat_robust_list))
910 compat_exit_robust_list(tsk);
922 if (group_dead && tsk->signal->leader)
923 disassociate_ctty(1);
925 module_put(task_thread_info(tsk)->exec_domain->module);
927 module_put(tsk->binfmt->module);
929 tsk->exit_code = code;
930 proc_exit_connector(tsk);
933 mpol_free(tsk->mempolicy);
934 tsk->mempolicy = NULL;
937 * If DEBUG_MUTEXES is on, make sure we are holding no locks:
939 mutex_debug_check_no_locks_held(tsk);
944 /* PF_DEAD causes final put_task_struct after we schedule. */
946 BUG_ON(tsk->flags & PF_DEAD);
947 tsk->flags |= PF_DEAD;
951 /* Avoid "noreturn function does return". */
955 EXPORT_SYMBOL_GPL(do_exit);
957 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
965 EXPORT_SYMBOL(complete_and_exit);
967 asmlinkage long sys_exit(int error_code)
969 do_exit((error_code&0xff)<<8);
973 * Take down every thread in the group. This is called by fatal signals
974 * as well as by sys_exit_group (below).
977 do_group_exit(int exit_code)
979 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
981 if (current->signal->flags & SIGNAL_GROUP_EXIT)
982 exit_code = current->signal->group_exit_code;
983 else if (!thread_group_empty(current)) {
984 struct signal_struct *const sig = current->signal;
985 struct sighand_struct *const sighand = current->sighand;
986 spin_lock_irq(&sighand->siglock);
987 if (sig->flags & SIGNAL_GROUP_EXIT)
988 /* Another thread got here before we took the lock. */
989 exit_code = sig->group_exit_code;
991 sig->group_exit_code = exit_code;
992 zap_other_threads(current);
994 spin_unlock_irq(&sighand->siglock);
1002 * this kills every thread in the thread group. Note that any externally
1003 * wait4()-ing process will get the correct exit code - even if this
1004 * thread is not the thread group leader.
1006 asmlinkage void sys_exit_group(int error_code)
1008 do_group_exit((error_code & 0xff) << 8);
1011 static int eligible_child(pid_t pid, int options, task_t *p)
1017 if (process_group(p) != process_group(current))
1019 } else if (pid != -1) {
1020 if (process_group(p) != -pid)
1025 * Do not consider detached threads that are
1028 if (p->exit_signal == -1 && !p->ptrace)
1031 /* Wait for all children (clone and not) if __WALL is set;
1032 * otherwise, wait for clone children *only* if __WCLONE is
1033 * set; otherwise, wait for non-clone children *only*. (Note:
1034 * A "clone" child here is one that reports to its parent
1035 * using a signal other than SIGCHLD.) */
1036 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1037 && !(options & __WALL))
1040 * Do not consider thread group leaders that are
1041 * in a non-empty thread group:
1043 if (current->tgid != p->tgid && delay_group_leader(p))
1046 if (security_task_wait(p))
1052 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
1053 int why, int status,
1054 struct siginfo __user *infop,
1055 struct rusage __user *rusagep)
1057 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1060 retval = put_user(SIGCHLD, &infop->si_signo);
1062 retval = put_user(0, &infop->si_errno);
1064 retval = put_user((short)why, &infop->si_code);
1066 retval = put_user(pid, &infop->si_pid);
1068 retval = put_user(uid, &infop->si_uid);
1070 retval = put_user(status, &infop->si_status);
1077 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1078 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1079 * the lock and this task is uninteresting. If we return nonzero, we have
1080 * released the lock and the system call should return.
1082 static int wait_task_zombie(task_t *p, int noreap,
1083 struct siginfo __user *infop,
1084 int __user *stat_addr, struct rusage __user *ru)
1086 unsigned long state;
1090 if (unlikely(noreap)) {
1093 int exit_code = p->exit_code;
1096 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1098 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1101 read_unlock(&tasklist_lock);
1102 if ((exit_code & 0x7f) == 0) {
1104 status = exit_code >> 8;
1106 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1107 status = exit_code & 0x7f;
1109 return wait_noreap_copyout(p, pid, uid, why,
1114 * Try to move the task's state to DEAD
1115 * only one thread is allowed to do this:
1117 state = xchg(&p->exit_state, EXIT_DEAD);
1118 if (state != EXIT_ZOMBIE) {
1119 BUG_ON(state != EXIT_DEAD);
1122 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1124 * This can only happen in a race with a ptraced thread
1125 * dying on another processor.
1130 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1131 struct signal_struct *psig;
1132 struct signal_struct *sig;
1135 * The resource counters for the group leader are in its
1136 * own task_struct. Those for dead threads in the group
1137 * are in its signal_struct, as are those for the child
1138 * processes it has previously reaped. All these
1139 * accumulate in the parent's signal_struct c* fields.
1141 * We don't bother to take a lock here to protect these
1142 * p->signal fields, because they are only touched by
1143 * __exit_signal, which runs with tasklist_lock
1144 * write-locked anyway, and so is excluded here. We do
1145 * need to protect the access to p->parent->signal fields,
1146 * as other threads in the parent group can be right
1147 * here reaping other children at the same time.
1149 spin_lock_irq(&p->parent->sighand->siglock);
1150 psig = p->parent->signal;
1153 cputime_add(psig->cutime,
1154 cputime_add(p->utime,
1155 cputime_add(sig->utime,
1158 cputime_add(psig->cstime,
1159 cputime_add(p->stime,
1160 cputime_add(sig->stime,
1163 p->min_flt + sig->min_flt + sig->cmin_flt;
1165 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1167 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1169 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1170 spin_unlock_irq(&p->parent->sighand->siglock);
1174 * Now we are sure this task is interesting, and no other
1175 * thread can reap it because we set its state to EXIT_DEAD.
1177 read_unlock(&tasklist_lock);
1179 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1180 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1181 ? p->signal->group_exit_code : p->exit_code;
1182 if (!retval && stat_addr)
1183 retval = put_user(status, stat_addr);
1184 if (!retval && infop)
1185 retval = put_user(SIGCHLD, &infop->si_signo);
1186 if (!retval && infop)
1187 retval = put_user(0, &infop->si_errno);
1188 if (!retval && infop) {
1191 if ((status & 0x7f) == 0) {
1195 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1198 retval = put_user((short)why, &infop->si_code);
1200 retval = put_user(status, &infop->si_status);
1202 if (!retval && infop)
1203 retval = put_user(p->pid, &infop->si_pid);
1204 if (!retval && infop)
1205 retval = put_user(p->uid, &infop->si_uid);
1207 // TODO: is this safe?
1208 p->exit_state = EXIT_ZOMBIE;
1212 if (p->real_parent != p->parent) {
1213 write_lock_irq(&tasklist_lock);
1214 /* Double-check with lock held. */
1215 if (p->real_parent != p->parent) {
1217 // TODO: is this safe?
1218 p->exit_state = EXIT_ZOMBIE;
1220 * If this is not a detached task, notify the parent.
1221 * If it's still not detached after that, don't release
1224 if (p->exit_signal != -1) {
1225 do_notify_parent(p, p->exit_signal);
1226 if (p->exit_signal != -1)
1230 write_unlock_irq(&tasklist_lock);
1239 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1240 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1241 * the lock and this task is uninteresting. If we return nonzero, we have
1242 * released the lock and the system call should return.
1244 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1245 struct siginfo __user *infop,
1246 int __user *stat_addr, struct rusage __user *ru)
1248 int retval, exit_code;
1252 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1253 p->signal && p->signal->group_stop_count > 0)
1255 * A group stop is in progress and this is the group leader.
1256 * We won't report until all threads have stopped.
1261 * Now we are pretty sure this task is interesting.
1262 * Make sure it doesn't get reaped out from under us while we
1263 * give up the lock and then examine it below. We don't want to
1264 * keep holding onto the tasklist_lock while we call getrusage and
1265 * possibly take page faults for user memory.
1268 read_unlock(&tasklist_lock);
1270 if (unlikely(noreap)) {
1273 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1275 exit_code = p->exit_code;
1276 if (unlikely(!exit_code) ||
1277 unlikely(p->state & TASK_TRACED))
1279 return wait_noreap_copyout(p, pid, uid,
1280 why, (exit_code << 8) | 0x7f,
1284 write_lock_irq(&tasklist_lock);
1287 * This uses xchg to be atomic with the thread resuming and setting
1288 * it. It must also be done with the write lock held to prevent a
1289 * race with the EXIT_ZOMBIE case.
1291 exit_code = xchg(&p->exit_code, 0);
1292 if (unlikely(p->exit_state)) {
1294 * The task resumed and then died. Let the next iteration
1295 * catch it in EXIT_ZOMBIE. Note that exit_code might
1296 * already be zero here if it resumed and did _exit(0).
1297 * The task itself is dead and won't touch exit_code again;
1298 * other processors in this function are locked out.
1300 p->exit_code = exit_code;
1303 if (unlikely(exit_code == 0)) {
1305 * Another thread in this function got to it first, or it
1306 * resumed, or it resumed and then died.
1308 write_unlock_irq(&tasklist_lock);
1312 * We are returning to the wait loop without having successfully
1313 * removed the process and having released the lock. We cannot
1314 * continue, since the "p" task pointer is potentially stale.
1316 * Return -EAGAIN, and do_wait() will restart the loop from the
1317 * beginning. Do _not_ re-acquire the lock.
1322 /* move to end of parent's list to avoid starvation */
1326 write_unlock_irq(&tasklist_lock);
1328 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1329 if (!retval && stat_addr)
1330 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1331 if (!retval && infop)
1332 retval = put_user(SIGCHLD, &infop->si_signo);
1333 if (!retval && infop)
1334 retval = put_user(0, &infop->si_errno);
1335 if (!retval && infop)
1336 retval = put_user((short)((p->ptrace & PT_PTRACED)
1337 ? CLD_TRAPPED : CLD_STOPPED),
1339 if (!retval && infop)
1340 retval = put_user(exit_code, &infop->si_status);
1341 if (!retval && infop)
1342 retval = put_user(p->pid, &infop->si_pid);
1343 if (!retval && infop)
1344 retval = put_user(p->uid, &infop->si_uid);
1354 * Handle do_wait work for one task in a live, non-stopped state.
1355 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1356 * the lock and this task is uninteresting. If we return nonzero, we have
1357 * released the lock and the system call should return.
1359 static int wait_task_continued(task_t *p, int noreap,
1360 struct siginfo __user *infop,
1361 int __user *stat_addr, struct rusage __user *ru)
1367 if (unlikely(!p->signal))
1370 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1373 spin_lock_irq(&p->sighand->siglock);
1374 /* Re-check with the lock held. */
1375 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1376 spin_unlock_irq(&p->sighand->siglock);
1380 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1381 spin_unlock_irq(&p->sighand->siglock);
1386 read_unlock(&tasklist_lock);
1389 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1391 if (!retval && stat_addr)
1392 retval = put_user(0xffff, stat_addr);
1396 retval = wait_noreap_copyout(p, pid, uid,
1397 CLD_CONTINUED, SIGCONT,
1399 BUG_ON(retval == 0);
1406 static inline int my_ptrace_child(struct task_struct *p)
1408 if (!(p->ptrace & PT_PTRACED))
1410 if (!(p->ptrace & PT_ATTACHED))
1413 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1414 * we are the attacher. If we are the real parent, this is a race
1415 * inside ptrace_attach. It is waiting for the tasklist_lock,
1416 * which we have to switch the parent links, but has already set
1417 * the flags in p->ptrace.
1419 return (p->parent != p->real_parent);
1422 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1423 int __user *stat_addr, struct rusage __user *ru)
1425 DECLARE_WAITQUEUE(wait, current);
1426 struct task_struct *tsk;
1429 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1432 * We will set this flag if we see any child that might later
1433 * match our criteria, even if we are not able to reap it yet.
1436 current->state = TASK_INTERRUPTIBLE;
1437 read_lock(&tasklist_lock);
1440 struct task_struct *p;
1441 struct list_head *_p;
1444 list_for_each(_p,&tsk->children) {
1445 p = list_entry(_p,struct task_struct,sibling);
1447 ret = eligible_child(pid, options, p);
1454 * When we hit the race with PTRACE_ATTACH,
1455 * we will not report this child. But the
1456 * race means it has not yet been moved to
1457 * our ptrace_children list, so we need to
1458 * set the flag here to avoid a spurious ECHILD
1459 * when the race happens with the only child.
1462 if (!my_ptrace_child(p))
1467 * It's stopped now, so it might later
1468 * continue, exit, or stop again.
1471 if (!(options & WUNTRACED) &&
1472 !my_ptrace_child(p))
1474 retval = wait_task_stopped(p, ret == 2,
1475 (options & WNOWAIT),
1478 if (retval == -EAGAIN)
1480 if (retval != 0) /* He released the lock. */
1485 if (p->exit_state == EXIT_DEAD)
1487 // case EXIT_ZOMBIE:
1488 if (p->exit_state == EXIT_ZOMBIE) {
1490 * Eligible but we cannot release
1494 goto check_continued;
1495 if (!likely(options & WEXITED))
1497 retval = wait_task_zombie(
1498 p, (options & WNOWAIT),
1499 infop, stat_addr, ru);
1500 /* He released the lock. */
1507 * It's running now, so it might later
1508 * exit, stop, or stop and then continue.
1511 if (!unlikely(options & WCONTINUED))
1513 retval = wait_task_continued(
1514 p, (options & WNOWAIT),
1515 infop, stat_addr, ru);
1516 if (retval != 0) /* He released the lock. */
1522 list_for_each(_p, &tsk->ptrace_children) {
1523 p = list_entry(_p, struct task_struct,
1525 if (!eligible_child(pid, options, p))
1531 if (options & __WNOTHREAD)
1533 tsk = next_thread(tsk);
1534 if (tsk->signal != current->signal)
1536 } while (tsk != current);
1538 read_unlock(&tasklist_lock);
1541 if (options & WNOHANG)
1543 retval = -ERESTARTSYS;
1544 if (signal_pending(current))
1551 current->state = TASK_RUNNING;
1552 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1558 * For a WNOHANG return, clear out all the fields
1559 * we would set so the user can easily tell the
1563 retval = put_user(0, &infop->si_signo);
1565 retval = put_user(0, &infop->si_errno);
1567 retval = put_user(0, &infop->si_code);
1569 retval = put_user(0, &infop->si_pid);
1571 retval = put_user(0, &infop->si_uid);
1573 retval = put_user(0, &infop->si_status);
1579 asmlinkage long sys_waitid(int which, pid_t pid,
1580 struct siginfo __user *infop, int options,
1581 struct rusage __user *ru)
1585 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1587 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1607 ret = do_wait(pid, options, infop, NULL, ru);
1609 /* avoid REGPARM breakage on x86: */
1610 prevent_tail_call(ret);
1614 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1615 int options, struct rusage __user *ru)
1619 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1620 __WNOTHREAD|__WCLONE|__WALL))
1622 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1624 /* avoid REGPARM breakage on x86: */
1625 prevent_tail_call(ret);
1629 #ifdef __ARCH_WANT_SYS_WAITPID
1632 * sys_waitpid() remains for compatibility. waitpid() should be
1633 * implemented by calling sys_wait4() from libc.a.
1635 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1637 return sys_wait4(pid, stat_addr, options, NULL);