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>
37 #include <linux/pipe_fs_i.h>
38 #include <linux/audit.h> /* for audit_free() */
39 #include <linux/vs_limit.h>
40 #include <linux/vs_context.h>
41 #include <linux/vs_network.h>
42 #include <linux/vs_cvirt.h>
44 #include <asm/uaccess.h>
45 #include <asm/unistd.h>
46 #include <asm/pgtable.h>
47 #include <asm/mmu_context.h>
49 extern void sem_exit (void);
50 extern struct task_struct *child_reaper;
52 int getrusage(struct task_struct *, int, struct rusage __user *);
54 static void exit_mm(struct task_struct * tsk);
56 static void __unhash_process(struct task_struct *p)
59 detach_pid(p, PIDTYPE_PID);
60 if (thread_group_leader(p)) {
61 detach_pid(p, PIDTYPE_PGID);
62 detach_pid(p, PIDTYPE_SID);
64 list_del_rcu(&p->tasks);
65 __get_cpu_var(process_counts)--;
67 list_del_rcu(&p->thread_group);
72 * This function expects the tasklist_lock write-locked.
74 static void __exit_signal(struct task_struct *tsk)
76 struct signal_struct *sig = tsk->signal;
77 struct sighand_struct *sighand;
80 BUG_ON(!atomic_read(&sig->count));
83 sighand = rcu_dereference(tsk->sighand);
84 spin_lock(&sighand->siglock);
86 posix_cpu_timers_exit(tsk);
87 if (atomic_dec_and_test(&sig->count))
88 posix_cpu_timers_exit_group(tsk);
91 * If there is any task waiting for the group exit
94 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
95 wake_up_process(sig->group_exit_task);
96 sig->group_exit_task = NULL;
98 if (tsk == sig->curr_target)
99 sig->curr_target = next_thread(tsk);
101 * Accumulate here the counters for all threads but the
102 * group leader as they die, so they can be added into
103 * the process-wide totals when those are taken.
104 * The group leader stays around as a zombie as long
105 * as there are other threads. When it gets reaped,
106 * the exit.c code will add its counts into these totals.
107 * We won't ever get here for the group leader, since it
108 * will have been the last reference on the signal_struct.
110 sig->utime = cputime_add(sig->utime, tsk->utime);
111 sig->stime = cputime_add(sig->stime, tsk->stime);
112 sig->min_flt += tsk->min_flt;
113 sig->maj_flt += tsk->maj_flt;
114 sig->nvcsw += tsk->nvcsw;
115 sig->nivcsw += tsk->nivcsw;
116 sig->sched_time += tsk->sched_time;
117 sig = NULL; /* Marker for below. */
120 __unhash_process(tsk);
124 spin_unlock(&sighand->siglock);
127 __cleanup_sighand(sighand);
128 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
129 flush_sigqueue(&tsk->pending);
131 flush_sigqueue(&sig->shared_pending);
132 __cleanup_signal(sig);
136 static void delayed_put_task_struct(struct rcu_head *rhp)
138 put_task_struct(container_of(rhp, struct task_struct, rcu));
141 void release_task(struct task_struct * p)
145 struct dentry *proc_dentry;
148 atomic_dec(&p->user->processes);
149 spin_lock(&p->proc_lock);
150 proc_dentry = proc_pid_unhash(p);
151 write_lock_irq(&tasklist_lock);
153 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
157 * If we are the last non-leader member of the thread
158 * group, and the leader is zombie, then notify the
159 * group leader's parent process. (if it wants notification.)
162 leader = p->group_leader;
163 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
164 BUG_ON(leader->exit_signal == -1);
165 do_notify_parent(leader, leader->exit_signal);
167 * If we were the last child thread and the leader has
168 * exited already, and the leader's parent ignores SIGCHLD,
169 * then we are the one who should release the leader.
171 * do_notify_parent() will have marked it self-reaping in
174 zap_leader = (leader->exit_signal == -1);
178 write_unlock_irq(&tasklist_lock);
179 spin_unlock(&p->proc_lock);
180 proc_pid_flush(proc_dentry);
182 call_rcu(&p->rcu, delayed_put_task_struct);
185 if (unlikely(zap_leader))
190 * This checks not only the pgrp, but falls back on the pid if no
191 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
194 int session_of_pgrp(int pgrp)
196 struct task_struct *p;
199 read_lock(&tasklist_lock);
200 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
201 if (p->signal->session > 0) {
202 sid = p->signal->session;
205 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
206 p = find_task_by_pid(pgrp);
208 sid = p->signal->session;
210 read_unlock(&tasklist_lock);
216 * Determine if a process group is "orphaned", according to the POSIX
217 * definition in 2.2.2.52. Orphaned process groups are not to be affected
218 * by terminal-generated stop signals. Newly orphaned process groups are
219 * to receive a SIGHUP and a SIGCONT.
221 * "I ask you, have you ever known what it is to be an orphan?"
223 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
225 struct task_struct *p;
228 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
229 if (p == ignored_task
231 || p->real_parent->pid == 1)
233 if (process_group(p->real_parent) != pgrp
234 && p->real_parent->signal->session == p->signal->session) {
238 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
239 return ret; /* (sighing) "Often!" */
242 int is_orphaned_pgrp(int pgrp)
246 read_lock(&tasklist_lock);
247 retval = will_become_orphaned_pgrp(pgrp, NULL);
248 read_unlock(&tasklist_lock);
253 static int has_stopped_jobs(int pgrp)
256 struct task_struct *p;
258 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
259 if (p->state != TASK_STOPPED)
262 /* If p is stopped by a debugger on a signal that won't
263 stop it, then don't count p as stopped. This isn't
264 perfect but it's a good approximation. */
265 if (unlikely (p->ptrace)
266 && p->exit_code != SIGSTOP
267 && p->exit_code != SIGTSTP
268 && p->exit_code != SIGTTOU
269 && p->exit_code != SIGTTIN)
274 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
279 * reparent_to_init - Reparent the calling kernel thread to the init task.
281 * If a kernel thread is launched as a result of a system call, or if
282 * it ever exits, it should generally reparent itself to init so that
283 * it is correctly cleaned up on exit.
285 * The various task state such as scheduling policy and priority may have
286 * been inherited from a user process, so we reset them to sane values here.
288 * NOTE that reparent_to_init() gives the caller full capabilities.
290 static void reparent_to_init(void)
292 write_lock_irq(&tasklist_lock);
294 ptrace_unlink(current);
295 /* Reparent to init */
296 remove_parent(current);
297 /* FIXME: handle vchild_reaper/initpid */
298 current->parent = child_reaper;
299 current->real_parent = child_reaper;
302 /* Set the exit signal to SIGCHLD so we signal init on exit */
303 current->exit_signal = SIGCHLD;
305 if ((current->policy == SCHED_NORMAL ||
306 current->policy == SCHED_BATCH)
307 && (task_nice(current) < 0))
308 set_user_nice(current, 0);
312 security_task_reparent_to_init(current);
313 memcpy(current->signal->rlim, init_task.signal->rlim,
314 sizeof(current->signal->rlim));
315 atomic_inc(&(INIT_USER->__count));
316 write_unlock_irq(&tasklist_lock);
317 switch_uid(INIT_USER);
320 void __set_special_pids(pid_t session, pid_t pgrp)
322 struct task_struct *curr = current->group_leader;
324 if (curr->signal->session != session) {
325 detach_pid(curr, PIDTYPE_SID);
326 curr->signal->session = session;
327 attach_pid(curr, PIDTYPE_SID, session);
329 if (process_group(curr) != pgrp) {
330 detach_pid(curr, PIDTYPE_PGID);
331 curr->signal->pgrp = pgrp;
332 attach_pid(curr, PIDTYPE_PGID, pgrp);
336 void set_special_pids(pid_t session, pid_t pgrp)
338 write_lock_irq(&tasklist_lock);
339 __set_special_pids(session, pgrp);
340 write_unlock_irq(&tasklist_lock);
344 * Let kernel threads use this to say that they
345 * allow a certain signal (since daemonize() will
346 * have disabled all of them by default).
348 int allow_signal(int sig)
350 if (!valid_signal(sig) || sig < 1)
353 spin_lock_irq(¤t->sighand->siglock);
354 sigdelset(¤t->blocked, sig);
356 /* Kernel threads handle their own signals.
357 Let the signal code know it'll be handled, so
358 that they don't get converted to SIGKILL or
359 just silently dropped */
360 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
363 spin_unlock_irq(¤t->sighand->siglock);
367 EXPORT_SYMBOL(allow_signal);
369 int disallow_signal(int sig)
371 if (!valid_signal(sig) || sig < 1)
374 spin_lock_irq(¤t->sighand->siglock);
375 sigaddset(¤t->blocked, sig);
377 spin_unlock_irq(¤t->sighand->siglock);
381 EXPORT_SYMBOL(disallow_signal);
384 * Put all the gunge required to become a kernel thread without
385 * attached user resources in one place where it belongs.
388 void daemonize(const char *name, ...)
391 struct fs_struct *fs;
394 va_start(args, name);
395 vsnprintf(current->comm, sizeof(current->comm), name, args);
399 * If we were started as result of loading a module, close all of the
400 * user space pages. We don't need them, and if we didn't close them
401 * they would be locked into memory.
405 set_special_pids(1, 1);
406 mutex_lock(&tty_mutex);
407 current->signal->tty = NULL;
408 mutex_unlock(&tty_mutex);
410 /* Block and flush all signals */
411 sigfillset(&blocked);
412 sigprocmask(SIG_BLOCK, &blocked, NULL);
413 flush_signals(current);
415 /* Become as one with the init task */
417 exit_fs(current); /* current->fs->count--; */
420 atomic_inc(&fs->count);
421 exit_namespace(current);
422 current->namespace = init_task.namespace;
423 get_namespace(current->namespace);
425 current->files = init_task.files;
426 atomic_inc(¤t->files->count);
431 EXPORT_SYMBOL(daemonize);
433 static void close_files(struct files_struct * files)
441 * It is safe to dereference the fd table without RCU or
442 * ->file_lock because this is the last reference to the
445 fdt = files_fdtable(files);
449 if (i >= fdt->max_fdset || i >= fdt->max_fds)
451 set = fdt->open_fds->fds_bits[j++];
454 struct file * file = xchg(&fdt->fd[i], NULL);
456 filp_close(file, files);
465 struct files_struct *get_files_struct(struct task_struct *task)
467 struct files_struct *files;
472 atomic_inc(&files->count);
478 void fastcall put_files_struct(struct files_struct *files)
482 if (atomic_dec_and_test(&files->count)) {
485 * Free the fd and fdset arrays if we expanded them.
486 * If the fdtable was embedded, pass files for freeing
487 * at the end of the RCU grace period. Otherwise,
488 * you can free files immediately.
490 fdt = files_fdtable(files);
491 if (fdt == &files->fdtab)
492 fdt->free_files = files;
494 kmem_cache_free(files_cachep, files);
499 EXPORT_SYMBOL(put_files_struct);
501 static inline void __exit_files(struct task_struct *tsk)
503 struct files_struct * files = tsk->files;
509 put_files_struct(files);
513 void exit_files(struct task_struct *tsk)
518 static inline void __put_fs_struct(struct fs_struct *fs)
520 /* No need to hold fs->lock if we are killing it */
521 if (atomic_dec_and_test(&fs->count)) {
528 mntput(fs->altrootmnt);
530 kmem_cache_free(fs_cachep, fs);
534 void put_fs_struct(struct fs_struct *fs)
539 static inline void __exit_fs(struct task_struct *tsk)
541 struct fs_struct * fs = tsk->fs;
551 void exit_fs(struct task_struct *tsk)
556 EXPORT_SYMBOL_GPL(exit_fs);
559 * Turn us into a lazy TLB process if we
562 static void exit_mm(struct task_struct * tsk)
564 struct mm_struct *mm = tsk->mm;
570 * Serialize with any possible pending coredump.
571 * We must hold mmap_sem around checking core_waiters
572 * and clearing tsk->mm. The core-inducing thread
573 * will increment core_waiters for each thread in the
574 * group with ->mm != NULL.
576 down_read(&mm->mmap_sem);
577 if (mm->core_waiters) {
578 up_read(&mm->mmap_sem);
579 down_write(&mm->mmap_sem);
580 if (!--mm->core_waiters)
581 complete(mm->core_startup_done);
582 up_write(&mm->mmap_sem);
584 wait_for_completion(&mm->core_done);
585 down_read(&mm->mmap_sem);
587 atomic_inc(&mm->mm_count);
588 if (mm != tsk->active_mm) BUG();
589 /* more a memory barrier than a real lock */
592 up_read(&mm->mmap_sem);
593 enter_lazy_tlb(mm, current);
598 static inline void choose_new_parent(task_t *p, task_t *reaper)
600 /* check for reaper context */
601 vxwprintk((p->xid != reaper->xid) && (reaper != child_reaper),
602 "rogue reaper: %p[%d,#%u] <> %p[%d,#%u]",
603 p, p->pid, p->xid, reaper, reaper->pid, reaper->xid);
606 * Make sure we're not reparenting to ourselves and that
607 * the parent is not a zombie.
609 BUG_ON(p == reaper || reaper->exit_state);
610 p->real_parent = reaper;
613 static void reparent_thread(task_t *p, task_t *father, int traced)
615 /* We don't want people slaying init. */
616 if (p->exit_signal != -1)
617 p->exit_signal = SIGCHLD;
619 if (p->pdeath_signal)
620 /* We already hold the tasklist_lock here. */
621 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
623 /* Move the child from its dying parent to the new one. */
624 if (unlikely(traced)) {
625 /* Preserve ptrace links if someone else is tracing this child. */
626 list_del_init(&p->ptrace_list);
627 if (p->parent != p->real_parent)
628 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
630 /* If this child is being traced, then we're the one tracing it
631 * anyway, so let go of it.
635 p->parent = p->real_parent;
638 /* If we'd notified the old parent about this child's death,
639 * also notify the new parent.
641 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
642 thread_group_empty(p))
643 do_notify_parent(p, p->exit_signal);
644 else if (p->state == TASK_TRACED) {
646 * If it was at a trace stop, turn it into
647 * a normal stop since it's no longer being
655 * process group orphan check
656 * Case ii: Our child is in a different pgrp
657 * than we are, and it was the only connection
658 * outside, so the child pgrp is now orphaned.
660 if ((process_group(p) != process_group(father)) &&
661 (p->signal->session == father->signal->session)) {
662 int pgrp = process_group(p);
664 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
665 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
666 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
672 * When we die, we re-parent all our children.
673 * Try to give them to another thread in our thread
674 * group, and if no such member exists, give it to
675 * the global child reaper process (ie "init")
677 static void forget_original_parent(struct task_struct * father,
678 struct list_head *to_release)
680 struct task_struct *p, *reaper = father;
681 struct list_head *_p, *_n;
684 reaper = next_thread(reaper);
685 if (reaper == father) {
686 reaper = vx_child_reaper(father);
689 } while (reaper->exit_state);
692 * There are only two places where our children can be:
694 * - in our child list
695 * - in our ptraced child list
697 * Search them and reparent children.
699 list_for_each_safe(_p, _n, &father->children) {
701 p = list_entry(_p,struct task_struct,sibling);
705 /* if father isn't the real parent, then ptrace must be enabled */
706 BUG_ON(father != p->real_parent && !ptrace);
708 if (father == p->real_parent) {
709 /* reparent with a reaper, real father it's us */
710 choose_new_parent(p, vx_child_reaper(p));
711 reparent_thread(p, father, 0);
713 /* reparent ptraced task to its real parent */
715 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
716 thread_group_empty(p))
717 do_notify_parent(p, p->exit_signal);
721 * if the ptraced child is a zombie with exit_signal == -1
722 * we must collect it before we exit, or it will remain
723 * zombie forever since we prevented it from self-reap itself
724 * while it was being traced by us, to be able to see it in wait4.
726 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
727 list_add(&p->ptrace_list, to_release);
729 list_for_each_safe(_p, _n, &father->ptrace_children) {
730 p = list_entry(_p,struct task_struct,ptrace_list);
731 /* check for reaper context */
732 choose_new_parent(p, reaper);
733 reparent_thread(p, father, 1);
738 * Send signals to all our closest relatives so that they know
739 * to properly mourn us..
741 static void exit_notify(struct task_struct *tsk)
744 struct task_struct *t;
745 struct list_head ptrace_dead, *_p, *_n;
747 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
748 && !thread_group_empty(tsk)) {
750 * This occurs when there was a race between our exit
751 * syscall and a group signal choosing us as the one to
752 * wake up. It could be that we are the only thread
753 * alerted to check for pending signals, but another thread
754 * should be woken now to take the signal since we will not.
755 * Now we'll wake all the threads in the group just to make
756 * sure someone gets all the pending signals.
758 read_lock(&tasklist_lock);
759 spin_lock_irq(&tsk->sighand->siglock);
760 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
761 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
762 recalc_sigpending_tsk(t);
763 if (signal_pending(t))
764 signal_wake_up(t, 0);
766 spin_unlock_irq(&tsk->sighand->siglock);
767 read_unlock(&tasklist_lock);
770 write_lock_irq(&tasklist_lock);
773 * This does two things:
775 * A. Make init inherit all the child processes
776 * B. Check to see if any process groups have become orphaned
777 * as a result of our exiting, and if they have any stopped
778 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
781 INIT_LIST_HEAD(&ptrace_dead);
782 forget_original_parent(tsk, &ptrace_dead);
783 BUG_ON(!list_empty(&tsk->children));
784 BUG_ON(!list_empty(&tsk->ptrace_children));
787 * Check to see if any process groups have become orphaned
788 * as a result of our exiting, and if they have any stopped
789 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
791 * Case i: Our father is in a different pgrp than we are
792 * and we were the only connection outside, so our pgrp
793 * is about to become orphaned.
796 t = tsk->real_parent;
798 if ((process_group(t) != process_group(tsk)) &&
799 (t->signal->session == tsk->signal->session) &&
800 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
801 has_stopped_jobs(process_group(tsk))) {
802 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
803 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
806 /* Let father know we died
808 * Thread signals are configurable, but you aren't going to use
809 * that to send signals to arbitary processes.
810 * That stops right now.
812 * If the parent exec id doesn't match the exec id we saved
813 * when we started then we know the parent has changed security
816 * If our self_exec id doesn't match our parent_exec_id then
817 * we have changed execution domain as these two values started
818 * the same after a fork.
822 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
823 ( tsk->parent_exec_id != t->self_exec_id ||
824 tsk->self_exec_id != tsk->parent_exec_id)
825 && !capable(CAP_KILL))
826 tsk->exit_signal = SIGCHLD;
829 /* If something other than our normal parent is ptracing us, then
830 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
831 * only has special meaning to our real parent.
833 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
834 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
835 do_notify_parent(tsk, signal);
836 } else if (tsk->ptrace) {
837 do_notify_parent(tsk, SIGCHLD);
841 if (tsk->exit_signal == -1 &&
842 (likely(tsk->ptrace == 0) ||
843 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
845 tsk->exit_state = state;
847 write_unlock_irq(&tasklist_lock);
849 list_for_each_safe(_p, _n, &ptrace_dead) {
851 t = list_entry(_p,struct task_struct,ptrace_list);
855 /* If the process is dead, release it - nobody will wait for it */
856 if (state == EXIT_DEAD)
860 fastcall NORET_TYPE void do_exit(long code)
862 struct task_struct *tsk = current;
865 profile_task_exit(tsk);
867 WARN_ON(atomic_read(&tsk->fs_excl));
869 if (unlikely(in_interrupt()))
870 panic("Aiee, killing interrupt handler!");
871 if (unlikely(!tsk->pid))
872 panic("Attempted to kill the idle task!");
873 if (unlikely(tsk == child_reaper))
874 panic("Attempted to kill init!");
876 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
877 current->ptrace_message = code;
878 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
882 * We're taking recursive faults here in do_exit. Safest is to just
883 * leave this task alone and wait for reboot.
885 if (unlikely(tsk->flags & PF_EXITING)) {
887 "Fixing recursive fault but reboot is needed!\n");
890 set_current_state(TASK_UNINTERRUPTIBLE);
894 tsk->flags |= PF_EXITING;
896 if (unlikely(in_atomic()))
897 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
898 current->comm, current->pid,
901 acct_update_integrals(tsk);
903 update_hiwater_rss(tsk->mm);
904 update_hiwater_vm(tsk->mm);
906 group_dead = atomic_dec_and_test(&tsk->signal->live);
908 hrtimer_cancel(&tsk->signal->real_timer);
909 exit_itimers(tsk->signal);
911 if (current->tux_info) {
912 #ifdef CONFIG_TUX_DEBUG
913 printk("Possibly unexpected TUX-thread exit(%ld) at %p?\n",
914 code, __builtin_return_address(0));
919 if (unlikely(tsk->robust_list))
920 exit_robust_list(tsk);
921 #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
922 if (unlikely(tsk->compat_robust_list))
923 compat_exit_robust_list(tsk);
925 if (unlikely(tsk->audit_context))
933 exit_vx_info(tsk, code);
939 if (group_dead && tsk->signal->leader)
940 disassociate_ctty(1);
942 module_put(task_thread_info(tsk)->exec_domain->module);
944 module_put(tsk->binfmt->module);
946 tsk->exit_code = code;
947 proc_exit_connector(tsk);
950 mpol_free(tsk->mempolicy);
951 tsk->mempolicy = NULL;
954 * If DEBUG_MUTEXES is on, make sure we are holding no locks:
956 mutex_debug_check_no_locks_held(tsk);
961 if (tsk->splice_pipe)
962 __free_pipe_info(tsk->splice_pipe);
964 /* PF_DEAD causes final put_task_struct after we schedule. */
966 BUG_ON(tsk->flags & PF_DEAD);
967 tsk->flags |= PF_DEAD;
971 /* Avoid "noreturn function does return". */
975 EXPORT_SYMBOL_GPL(do_exit);
977 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
985 EXPORT_SYMBOL(complete_and_exit);
987 asmlinkage long sys_exit(int error_code)
989 do_exit((error_code&0xff)<<8);
993 * Take down every thread in the group. This is called by fatal signals
994 * as well as by sys_exit_group (below).
997 do_group_exit(int exit_code)
999 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1001 if (current->signal->flags & SIGNAL_GROUP_EXIT)
1002 exit_code = current->signal->group_exit_code;
1003 else if (!thread_group_empty(current)) {
1004 struct signal_struct *const sig = current->signal;
1005 struct sighand_struct *const sighand = current->sighand;
1006 spin_lock_irq(&sighand->siglock);
1007 if (sig->flags & SIGNAL_GROUP_EXIT)
1008 /* Another thread got here before we took the lock. */
1009 exit_code = sig->group_exit_code;
1011 sig->group_exit_code = exit_code;
1012 zap_other_threads(current);
1014 spin_unlock_irq(&sighand->siglock);
1022 * this kills every thread in the thread group. Note that any externally
1023 * wait4()-ing process will get the correct exit code - even if this
1024 * thread is not the thread group leader.
1026 asmlinkage void sys_exit_group(int error_code)
1028 do_group_exit((error_code & 0xff) << 8);
1031 static int eligible_child(pid_t pid, int options, task_t *p)
1037 if (process_group(p) != process_group(current))
1039 } else if (pid != -1) {
1040 if (process_group(p) != -pid)
1045 * Do not consider detached threads that are
1048 if (p->exit_signal == -1 && !p->ptrace)
1051 /* Wait for all children (clone and not) if __WALL is set;
1052 * otherwise, wait for clone children *only* if __WCLONE is
1053 * set; otherwise, wait for non-clone children *only*. (Note:
1054 * A "clone" child here is one that reports to its parent
1055 * using a signal other than SIGCHLD.) */
1056 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1057 && !(options & __WALL))
1060 * Do not consider thread group leaders that are
1061 * in a non-empty thread group:
1063 if (current->tgid != p->tgid && delay_group_leader(p))
1066 if (security_task_wait(p))
1072 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
1073 int why, int status,
1074 struct siginfo __user *infop,
1075 struct rusage __user *rusagep)
1077 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1080 retval = put_user(SIGCHLD, &infop->si_signo);
1082 retval = put_user(0, &infop->si_errno);
1084 retval = put_user((short)why, &infop->si_code);
1086 retval = put_user(pid, &infop->si_pid);
1088 retval = put_user(uid, &infop->si_uid);
1090 retval = put_user(status, &infop->si_status);
1097 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1098 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1099 * the lock and this task is uninteresting. If we return nonzero, we have
1100 * released the lock and the system call should return.
1102 static int wait_task_zombie(task_t *p, int noreap,
1103 struct siginfo __user *infop,
1104 int __user *stat_addr, struct rusage __user *ru)
1106 unsigned long state;
1110 if (unlikely(noreap)) {
1113 int exit_code = p->exit_code;
1116 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1118 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1121 read_unlock(&tasklist_lock);
1122 if ((exit_code & 0x7f) == 0) {
1124 status = exit_code >> 8;
1126 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1127 status = exit_code & 0x7f;
1129 return wait_noreap_copyout(p, pid, uid, why,
1134 * Try to move the task's state to DEAD
1135 * only one thread is allowed to do this:
1137 state = xchg(&p->exit_state, EXIT_DEAD);
1138 if (state != EXIT_ZOMBIE) {
1139 BUG_ON(state != EXIT_DEAD);
1142 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1144 * This can only happen in a race with a ptraced thread
1145 * dying on another processor.
1150 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1151 struct signal_struct *psig;
1152 struct signal_struct *sig;
1155 * The resource counters for the group leader are in its
1156 * own task_struct. Those for dead threads in the group
1157 * are in its signal_struct, as are those for the child
1158 * processes it has previously reaped. All these
1159 * accumulate in the parent's signal_struct c* fields.
1161 * We don't bother to take a lock here to protect these
1162 * p->signal fields, because they are only touched by
1163 * __exit_signal, which runs with tasklist_lock
1164 * write-locked anyway, and so is excluded here. We do
1165 * need to protect the access to p->parent->signal fields,
1166 * as other threads in the parent group can be right
1167 * here reaping other children at the same time.
1169 spin_lock_irq(&p->parent->sighand->siglock);
1170 psig = p->parent->signal;
1173 cputime_add(psig->cutime,
1174 cputime_add(p->utime,
1175 cputime_add(sig->utime,
1178 cputime_add(psig->cstime,
1179 cputime_add(p->stime,
1180 cputime_add(sig->stime,
1183 p->min_flt + sig->min_flt + sig->cmin_flt;
1185 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1187 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1189 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1190 spin_unlock_irq(&p->parent->sighand->siglock);
1194 * Now we are sure this task is interesting, and no other
1195 * thread can reap it because we set its state to EXIT_DEAD.
1197 read_unlock(&tasklist_lock);
1199 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1200 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1201 ? p->signal->group_exit_code : p->exit_code;
1202 if (!retval && stat_addr)
1203 retval = put_user(status, stat_addr);
1204 if (!retval && infop)
1205 retval = put_user(SIGCHLD, &infop->si_signo);
1206 if (!retval && infop)
1207 retval = put_user(0, &infop->si_errno);
1208 if (!retval && infop) {
1211 if ((status & 0x7f) == 0) {
1215 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1218 retval = put_user((short)why, &infop->si_code);
1220 retval = put_user(status, &infop->si_status);
1222 if (!retval && infop)
1223 retval = put_user(p->pid, &infop->si_pid);
1224 if (!retval && infop)
1225 retval = put_user(p->uid, &infop->si_uid);
1227 // TODO: is this safe?
1228 p->exit_state = EXIT_ZOMBIE;
1232 if (p->real_parent != p->parent) {
1233 write_lock_irq(&tasklist_lock);
1234 /* Double-check with lock held. */
1235 if (p->real_parent != p->parent) {
1237 // TODO: is this safe?
1238 p->exit_state = EXIT_ZOMBIE;
1240 * If this is not a detached task, notify the parent.
1241 * If it's still not detached after that, don't release
1244 if (p->exit_signal != -1) {
1245 do_notify_parent(p, p->exit_signal);
1246 if (p->exit_signal != -1)
1250 write_unlock_irq(&tasklist_lock);
1259 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1260 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1261 * the lock and this task is uninteresting. If we return nonzero, we have
1262 * released the lock and the system call should return.
1264 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1265 struct siginfo __user *infop,
1266 int __user *stat_addr, struct rusage __user *ru)
1268 int retval, exit_code;
1272 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1273 p->signal && p->signal->group_stop_count > 0)
1275 * A group stop is in progress and this is the group leader.
1276 * We won't report until all threads have stopped.
1281 * Now we are pretty sure this task is interesting.
1282 * Make sure it doesn't get reaped out from under us while we
1283 * give up the lock and then examine it below. We don't want to
1284 * keep holding onto the tasklist_lock while we call getrusage and
1285 * possibly take page faults for user memory.
1288 read_unlock(&tasklist_lock);
1290 if (unlikely(noreap)) {
1293 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1295 exit_code = p->exit_code;
1296 if (unlikely(!exit_code) ||
1297 unlikely(p->state & TASK_TRACED))
1299 return wait_noreap_copyout(p, pid, uid,
1300 why, (exit_code << 8) | 0x7f,
1304 write_lock_irq(&tasklist_lock);
1307 * This uses xchg to be atomic with the thread resuming and setting
1308 * it. It must also be done with the write lock held to prevent a
1309 * race with the EXIT_ZOMBIE case.
1311 exit_code = xchg(&p->exit_code, 0);
1312 if (unlikely(p->exit_state)) {
1314 * The task resumed and then died. Let the next iteration
1315 * catch it in EXIT_ZOMBIE. Note that exit_code might
1316 * already be zero here if it resumed and did _exit(0).
1317 * The task itself is dead and won't touch exit_code again;
1318 * other processors in this function are locked out.
1320 p->exit_code = exit_code;
1323 if (unlikely(exit_code == 0)) {
1325 * Another thread in this function got to it first, or it
1326 * resumed, or it resumed and then died.
1328 write_unlock_irq(&tasklist_lock);
1332 * We are returning to the wait loop without having successfully
1333 * removed the process and having released the lock. We cannot
1334 * continue, since the "p" task pointer is potentially stale.
1336 * Return -EAGAIN, and do_wait() will restart the loop from the
1337 * beginning. Do _not_ re-acquire the lock.
1342 /* move to end of parent's list to avoid starvation */
1346 write_unlock_irq(&tasklist_lock);
1348 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1349 if (!retval && stat_addr)
1350 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1351 if (!retval && infop)
1352 retval = put_user(SIGCHLD, &infop->si_signo);
1353 if (!retval && infop)
1354 retval = put_user(0, &infop->si_errno);
1355 if (!retval && infop)
1356 retval = put_user((short)((p->ptrace & PT_PTRACED)
1357 ? CLD_TRAPPED : CLD_STOPPED),
1359 if (!retval && infop)
1360 retval = put_user(exit_code, &infop->si_status);
1361 if (!retval && infop)
1362 retval = put_user(p->pid, &infop->si_pid);
1363 if (!retval && infop)
1364 retval = put_user(p->uid, &infop->si_uid);
1374 * Handle do_wait work for one task in a live, non-stopped state.
1375 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1376 * the lock and this task is uninteresting. If we return nonzero, we have
1377 * released the lock and the system call should return.
1379 static int wait_task_continued(task_t *p, int noreap,
1380 struct siginfo __user *infop,
1381 int __user *stat_addr, struct rusage __user *ru)
1387 if (unlikely(!p->signal))
1390 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1393 spin_lock_irq(&p->sighand->siglock);
1394 /* Re-check with the lock held. */
1395 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1396 spin_unlock_irq(&p->sighand->siglock);
1400 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1401 spin_unlock_irq(&p->sighand->siglock);
1406 read_unlock(&tasklist_lock);
1409 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1411 if (!retval && stat_addr)
1412 retval = put_user(0xffff, stat_addr);
1416 retval = wait_noreap_copyout(p, pid, uid,
1417 CLD_CONTINUED, SIGCONT,
1419 BUG_ON(retval == 0);
1426 static inline int my_ptrace_child(struct task_struct *p)
1428 if (!(p->ptrace & PT_PTRACED))
1430 if (!(p->ptrace & PT_ATTACHED))
1433 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1434 * we are the attacher. If we are the real parent, this is a race
1435 * inside ptrace_attach. It is waiting for the tasklist_lock,
1436 * which we have to switch the parent links, but has already set
1437 * the flags in p->ptrace.
1439 return (p->parent != p->real_parent);
1442 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1443 int __user *stat_addr, struct rusage __user *ru)
1445 DECLARE_WAITQUEUE(wait, current);
1446 struct task_struct *tsk;
1449 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1452 * We will set this flag if we see any child that might later
1453 * match our criteria, even if we are not able to reap it yet.
1456 current->state = TASK_INTERRUPTIBLE;
1457 read_lock(&tasklist_lock);
1460 struct task_struct *p;
1461 struct list_head *_p;
1464 list_for_each(_p,&tsk->children) {
1465 p = list_entry(_p,struct task_struct,sibling);
1467 ret = eligible_child(pid, options, p);
1474 * When we hit the race with PTRACE_ATTACH,
1475 * we will not report this child. But the
1476 * race means it has not yet been moved to
1477 * our ptrace_children list, so we need to
1478 * set the flag here to avoid a spurious ECHILD
1479 * when the race happens with the only child.
1482 if (!my_ptrace_child(p))
1487 * It's stopped now, so it might later
1488 * continue, exit, or stop again.
1491 if (!(options & WUNTRACED) &&
1492 !my_ptrace_child(p))
1494 retval = wait_task_stopped(p, ret == 2,
1495 (options & WNOWAIT),
1498 if (retval == -EAGAIN)
1500 if (retval != 0) /* He released the lock. */
1505 if (p->exit_state == EXIT_DEAD)
1507 // case EXIT_ZOMBIE:
1508 if (p->exit_state == EXIT_ZOMBIE) {
1510 * Eligible but we cannot release
1514 goto check_continued;
1515 if (!likely(options & WEXITED))
1517 retval = wait_task_zombie(
1518 p, (options & WNOWAIT),
1519 infop, stat_addr, ru);
1520 /* He released the lock. */
1527 * It's running now, so it might later
1528 * exit, stop, or stop and then continue.
1531 if (!unlikely(options & WCONTINUED))
1533 retval = wait_task_continued(
1534 p, (options & WNOWAIT),
1535 infop, stat_addr, ru);
1536 if (retval != 0) /* He released the lock. */
1542 list_for_each(_p, &tsk->ptrace_children) {
1543 p = list_entry(_p, struct task_struct,
1545 if (!eligible_child(pid, options, p))
1551 if (options & __WNOTHREAD)
1553 tsk = next_thread(tsk);
1554 if (tsk->signal != current->signal)
1556 } while (tsk != current);
1558 read_unlock(&tasklist_lock);
1561 if (options & WNOHANG)
1563 retval = -ERESTARTSYS;
1564 if (signal_pending(current))
1571 current->state = TASK_RUNNING;
1572 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1578 * For a WNOHANG return, clear out all the fields
1579 * we would set so the user can easily tell the
1583 retval = put_user(0, &infop->si_signo);
1585 retval = put_user(0, &infop->si_errno);
1587 retval = put_user(0, &infop->si_code);
1589 retval = put_user(0, &infop->si_pid);
1591 retval = put_user(0, &infop->si_uid);
1593 retval = put_user(0, &infop->si_status);
1599 asmlinkage long sys_waitid(int which, pid_t pid,
1600 struct siginfo __user *infop, int options,
1601 struct rusage __user *ru)
1605 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1607 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1627 ret = do_wait(pid, options, infop, NULL, ru);
1629 /* avoid REGPARM breakage on x86: */
1630 prevent_tail_call(ret);
1634 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1635 int options, struct rusage __user *ru)
1639 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1640 __WNOTHREAD|__WCLONE|__WALL))
1642 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1644 /* avoid REGPARM breakage on x86: */
1645 prevent_tail_call(ret);
1649 #ifdef __ARCH_WANT_SYS_WAITPID
1652 * sys_waitpid() remains for compatibility. waitpid() should be
1653 * implemented by calling sys_wait4() from libc.a.
1655 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1657 return sys_wait4(pid, stat_addr, options, NULL);