4 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/slab.h>
9 #include <linux/interrupt.h>
10 #include <linux/smp_lock.h>
11 #include <linux/module.h>
12 #include <linux/capability.h>
13 #include <linux/completion.h>
14 #include <linux/personality.h>
15 #include <linux/tty.h>
16 #include <linux/namespace.h>
17 #include <linux/key.h>
18 #include <linux/security.h>
19 #include <linux/cpu.h>
20 #include <linux/acct.h>
21 #include <linux/tsacct_kern.h>
22 #include <linux/file.h>
23 #include <linux/binfmts.h>
24 #include <linux/nsproxy.h>
25 #include <linux/ptrace.h>
26 #include <linux/profile.h>
27 #include <linux/mount.h>
28 #include <linux/proc_fs.h>
29 #include <linux/mempolicy.h>
30 #include <linux/taskstats_kern.h>
31 #include <linux/delayacct.h>
32 #include <linux/cpuset.h>
33 #include <linux/syscalls.h>
34 #include <linux/signal.h>
35 #include <linux/posix-timers.h>
36 #include <linux/cn_proc.h>
37 #include <linux/mutex.h>
38 #include <linux/futex.h>
39 #include <linux/compat.h>
40 #include <linux/pipe_fs_i.h>
41 #include <linux/audit.h> /* for audit_free() */
42 #include <linux/resource.h>
43 #include <linux/blkdev.h>
45 #include <asm/uaccess.h>
46 #include <asm/unistd.h>
47 #include <asm/pgtable.h>
48 #include <asm/mmu_context.h>
50 extern void sem_exit (void);
51 extern struct task_struct *child_reaper;
53 static void exit_mm(struct task_struct * tsk);
55 static void __unhash_process(struct task_struct *p)
58 detach_pid(p, PIDTYPE_PID);
59 if (thread_group_leader(p)) {
60 detach_pid(p, PIDTYPE_PGID);
61 detach_pid(p, PIDTYPE_SID);
63 list_del_rcu(&p->tasks);
64 __get_cpu_var(process_counts)--;
66 list_del_rcu(&p->thread_group);
71 * This function expects the tasklist_lock write-locked.
73 static void __exit_signal(struct task_struct *tsk)
75 struct signal_struct *sig = tsk->signal;
76 struct sighand_struct *sighand;
79 BUG_ON(!atomic_read(&sig->count));
82 sighand = rcu_dereference(tsk->sighand);
83 spin_lock(&sighand->siglock);
85 posix_cpu_timers_exit(tsk);
86 if (atomic_dec_and_test(&sig->count))
87 posix_cpu_timers_exit_group(tsk);
90 * If there is any task waiting for the group exit
93 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
94 wake_up_process(sig->group_exit_task);
95 sig->group_exit_task = NULL;
97 if (tsk == sig->curr_target)
98 sig->curr_target = next_thread(tsk);
100 * Accumulate here the counters for all threads but the
101 * group leader as they die, so they can be added into
102 * the process-wide totals when those are taken.
103 * The group leader stays around as a zombie as long
104 * as there are other threads. When it gets reaped,
105 * the exit.c code will add its counts into these totals.
106 * We won't ever get here for the group leader, since it
107 * will have been the last reference on the signal_struct.
109 sig->utime = cputime_add(sig->utime, tsk->utime);
110 sig->stime = cputime_add(sig->stime, tsk->stime);
111 sig->min_flt += tsk->min_flt;
112 sig->maj_flt += tsk->maj_flt;
113 sig->nvcsw += tsk->nvcsw;
114 sig->nivcsw += tsk->nivcsw;
115 sig->sched_time += tsk->sched_time;
116 sig = NULL; /* Marker for below. */
119 __unhash_process(tsk);
123 spin_unlock(&sighand->siglock);
126 __cleanup_sighand(sighand);
127 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
128 flush_sigqueue(&tsk->pending);
130 flush_sigqueue(&sig->shared_pending);
131 taskstats_tgid_free(sig);
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)
143 struct task_struct *leader;
146 atomic_dec(&p->user->processes);
147 write_lock_irq(&tasklist_lock);
149 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
153 * If we are the last non-leader member of the thread
154 * group, and the leader is zombie, then notify the
155 * group leader's parent process. (if it wants notification.)
158 leader = p->group_leader;
159 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
160 BUG_ON(leader->exit_signal == -1);
161 do_notify_parent(leader, leader->exit_signal);
163 * If we were the last child thread and the leader has
164 * exited already, and the leader's parent ignores SIGCHLD,
165 * then we are the one who should release the leader.
167 * do_notify_parent() will have marked it self-reaping in
170 zap_leader = (leader->exit_signal == -1);
174 write_unlock_irq(&tasklist_lock);
177 call_rcu(&p->rcu, delayed_put_task_struct);
180 if (unlikely(zap_leader))
185 * This checks not only the pgrp, but falls back on the pid if no
186 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
189 int session_of_pgrp(int pgrp)
191 struct task_struct *p;
194 read_lock(&tasklist_lock);
195 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
196 if (process_session(p) > 0) {
197 sid = process_session(p);
200 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
201 p = find_task_by_pid(pgrp);
203 sid = process_session(p);
205 read_unlock(&tasklist_lock);
211 * Determine if a process group is "orphaned", according to the POSIX
212 * definition in 2.2.2.52. Orphaned process groups are not to be affected
213 * by terminal-generated stop signals. Newly orphaned process groups are
214 * to receive a SIGHUP and a SIGCONT.
216 * "I ask you, have you ever known what it is to be an orphan?"
218 static int will_become_orphaned_pgrp(int pgrp, struct task_struct *ignored_task)
220 struct task_struct *p;
223 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
224 if (p == ignored_task
226 || is_init(p->real_parent))
228 if (process_group(p->real_parent) != pgrp &&
229 process_session(p->real_parent) == process_session(p)) {
233 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
234 return ret; /* (sighing) "Often!" */
237 int is_orphaned_pgrp(int pgrp)
241 read_lock(&tasklist_lock);
242 retval = will_become_orphaned_pgrp(pgrp, NULL);
243 read_unlock(&tasklist_lock);
248 static int has_stopped_jobs(int pgrp)
251 struct task_struct *p;
253 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
254 if (p->state != TASK_STOPPED)
258 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
263 * reparent_to_init - Reparent the calling kernel thread to the init task.
265 * If a kernel thread is launched as a result of a system call, or if
266 * it ever exits, it should generally reparent itself to init so that
267 * it is correctly cleaned up on exit.
269 * The various task state such as scheduling policy and priority may have
270 * been inherited from a user process, so we reset them to sane values here.
272 * NOTE that reparent_to_init() gives the caller full capabilities.
274 static void reparent_to_init(void)
276 write_lock_irq(&tasklist_lock);
278 ptrace_unlink(current);
279 /* Reparent to init */
280 remove_parent(current);
281 current->parent = child_reaper;
282 current->real_parent = child_reaper;
285 /* Set the exit signal to SIGCHLD so we signal init on exit */
286 current->exit_signal = SIGCHLD;
288 if (!has_rt_policy(current) && (task_nice(current) < 0))
289 set_user_nice(current, 0);
293 security_task_reparent_to_init(current);
294 memcpy(current->signal->rlim, init_task.signal->rlim,
295 sizeof(current->signal->rlim));
296 atomic_inc(&(INIT_USER->__count));
297 write_unlock_irq(&tasklist_lock);
298 switch_uid(INIT_USER);
301 void __set_special_pids(pid_t session, pid_t pgrp)
303 struct task_struct *curr = current->group_leader;
305 if (process_session(curr) != session) {
306 detach_pid(curr, PIDTYPE_SID);
307 set_signal_session(curr->signal, session);
308 attach_pid(curr, PIDTYPE_SID, session);
310 if (process_group(curr) != pgrp) {
311 detach_pid(curr, PIDTYPE_PGID);
312 curr->signal->pgrp = pgrp;
313 attach_pid(curr, PIDTYPE_PGID, pgrp);
317 static void set_special_pids(pid_t session, pid_t pgrp)
319 write_lock_irq(&tasklist_lock);
320 __set_special_pids(session, pgrp);
321 write_unlock_irq(&tasklist_lock);
325 * Let kernel threads use this to say that they
326 * allow a certain signal (since daemonize() will
327 * have disabled all of them by default).
329 int allow_signal(int sig)
331 if (!valid_signal(sig) || sig < 1)
334 spin_lock_irq(¤t->sighand->siglock);
335 sigdelset(¤t->blocked, sig);
337 /* Kernel threads handle their own signals.
338 Let the signal code know it'll be handled, so
339 that they don't get converted to SIGKILL or
340 just silently dropped */
341 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
344 spin_unlock_irq(¤t->sighand->siglock);
348 EXPORT_SYMBOL(allow_signal);
350 int disallow_signal(int sig)
352 if (!valid_signal(sig) || sig < 1)
355 spin_lock_irq(¤t->sighand->siglock);
356 sigaddset(¤t->blocked, sig);
358 spin_unlock_irq(¤t->sighand->siglock);
362 EXPORT_SYMBOL(disallow_signal);
365 * Put all the gunge required to become a kernel thread without
366 * attached user resources in one place where it belongs.
369 void daemonize(const char *name, ...)
372 struct fs_struct *fs;
375 va_start(args, name);
376 vsnprintf(current->comm, sizeof(current->comm), name, args);
380 * If we were started as result of loading a module, close all of the
381 * user space pages. We don't need them, and if we didn't close them
382 * they would be locked into memory.
386 set_special_pids(1, 1);
387 proc_clear_tty(current);
389 /* Block and flush all signals */
390 sigfillset(&blocked);
391 sigprocmask(SIG_BLOCK, &blocked, NULL);
392 flush_signals(current);
394 /* Become as one with the init task */
396 exit_fs(current); /* current->fs->count--; */
399 atomic_inc(&fs->count);
401 exit_task_namespaces(current);
402 current->nsproxy = init_task.nsproxy;
403 get_task_namespaces(current);
406 current->files = init_task.files;
407 atomic_inc(¤t->files->count);
412 EXPORT_SYMBOL(daemonize);
414 static void close_files(struct files_struct * files)
422 * It is safe to dereference the fd table without RCU or
423 * ->file_lock because this is the last reference to the
426 fdt = files_fdtable(files);
430 if (i >= fdt->max_fdset || i >= fdt->max_fds)
432 set = fdt->open_fds->fds_bits[j++];
435 struct file * file = xchg(&fdt->fd[i], NULL);
437 filp_close(file, files);
445 struct files_struct *get_files_struct(struct task_struct *task)
447 struct files_struct *files;
452 atomic_inc(&files->count);
458 void fastcall put_files_struct(struct files_struct *files)
462 if (atomic_dec_and_test(&files->count)) {
465 * Free the fd and fdset arrays if we expanded them.
466 * If the fdtable was embedded, pass files for freeing
467 * at the end of the RCU grace period. Otherwise,
468 * you can free files immediately.
470 fdt = files_fdtable(files);
471 if (fdt == &files->fdtab)
472 fdt->free_files = files;
474 kmem_cache_free(files_cachep, files);
479 EXPORT_SYMBOL(put_files_struct);
481 void reset_files_struct(struct task_struct *tsk, struct files_struct *files)
483 struct files_struct *old;
489 put_files_struct(old);
491 EXPORT_SYMBOL(reset_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 BUG_ON(mm != tsk->active_mm);
581 /* more a memory barrier than a real lock */
584 up_read(&mm->mmap_sem);
585 enter_lazy_tlb(mm, current);
591 choose_new_parent(struct task_struct *p, struct task_struct *reaper)
594 * Make sure we're not reparenting to ourselves and that
595 * the parent is not a zombie.
597 BUG_ON(p == reaper || reaper->exit_state);
598 p->real_parent = reaper;
602 reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
604 /* We don't want people slaying init. */
605 if (p->exit_signal != -1)
606 p->exit_signal = SIGCHLD;
608 if (p->pdeath_signal)
609 /* We already hold the tasklist_lock here. */
610 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
612 /* Move the child from its dying parent to the new one. */
613 if (unlikely(traced)) {
614 /* Preserve ptrace links if someone else is tracing this child. */
615 list_del_init(&p->ptrace_list);
616 if (p->parent != p->real_parent)
617 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
619 /* If this child is being traced, then we're the one tracing it
620 * anyway, so let go of it.
624 p->parent = p->real_parent;
627 /* If we'd notified the old parent about this child's death,
628 * also notify the new parent.
630 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
631 thread_group_empty(p))
632 do_notify_parent(p, p->exit_signal);
633 else if (p->state == TASK_TRACED) {
635 * If it was at a trace stop, turn it into
636 * a normal stop since it's no longer being
644 * process group orphan check
645 * Case ii: Our child is in a different pgrp
646 * than we are, and it was the only connection
647 * outside, so the child pgrp is now orphaned.
649 if ((process_group(p) != process_group(father)) &&
650 (process_session(p) == process_session(father))) {
651 int pgrp = process_group(p);
653 if (will_become_orphaned_pgrp(pgrp, NULL) &&
654 has_stopped_jobs(pgrp)) {
655 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
656 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
662 * When we die, we re-parent all our children.
663 * Try to give them to another thread in our thread
664 * group, and if no such member exists, give it to
665 * the global child reaper process (ie "init")
668 forget_original_parent(struct task_struct *father, struct list_head *to_release)
670 struct task_struct *p, *reaper = father;
671 struct list_head *_p, *_n;
674 reaper = next_thread(reaper);
675 if (reaper == father) {
676 reaper = child_reaper;
679 } while (reaper->exit_state);
682 * There are only two places where our children can be:
684 * - in our child list
685 * - in our ptraced child list
687 * Search them and reparent children.
689 list_for_each_safe(_p, _n, &father->children) {
691 p = list_entry(_p, struct task_struct, sibling);
695 /* if father isn't the real parent, then ptrace must be enabled */
696 BUG_ON(father != p->real_parent && !ptrace);
698 if (father == p->real_parent) {
699 /* reparent with a reaper, real father it's us */
700 choose_new_parent(p, reaper);
701 reparent_thread(p, father, 0);
703 /* reparent ptraced task to its real parent */
705 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
706 thread_group_empty(p))
707 do_notify_parent(p, p->exit_signal);
711 * if the ptraced child is a zombie with exit_signal == -1
712 * we must collect it before we exit, or it will remain
713 * zombie forever since we prevented it from self-reap itself
714 * while it was being traced by us, to be able to see it in wait4.
716 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
717 list_add(&p->ptrace_list, to_release);
719 list_for_each_safe(_p, _n, &father->ptrace_children) {
720 p = list_entry(_p, struct task_struct, ptrace_list);
721 choose_new_parent(p, reaper);
722 reparent_thread(p, father, 1);
727 * Send signals to all our closest relatives so that they know
728 * to properly mourn us..
730 static void exit_notify(struct task_struct *tsk)
733 struct task_struct *t;
734 struct list_head ptrace_dead, *_p, *_n;
736 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
737 && !thread_group_empty(tsk)) {
739 * This occurs when there was a race between our exit
740 * syscall and a group signal choosing us as the one to
741 * wake up. It could be that we are the only thread
742 * alerted to check for pending signals, but another thread
743 * should be woken now to take the signal since we will not.
744 * Now we'll wake all the threads in the group just to make
745 * sure someone gets all the pending signals.
747 read_lock(&tasklist_lock);
748 spin_lock_irq(&tsk->sighand->siglock);
749 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
750 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
751 recalc_sigpending_tsk(t);
752 if (signal_pending(t))
753 signal_wake_up(t, 0);
755 spin_unlock_irq(&tsk->sighand->siglock);
756 read_unlock(&tasklist_lock);
759 write_lock_irq(&tasklist_lock);
762 * This does two things:
764 * A. Make init inherit all the child processes
765 * B. Check to see if any process groups have become orphaned
766 * as a result of our exiting, and if they have any stopped
767 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
770 INIT_LIST_HEAD(&ptrace_dead);
771 forget_original_parent(tsk, &ptrace_dead);
772 BUG_ON(!list_empty(&tsk->children));
773 BUG_ON(!list_empty(&tsk->ptrace_children));
776 * 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)
780 * Case i: Our father is in a different pgrp than we are
781 * and we were the only connection outside, so our pgrp
782 * is about to become orphaned.
785 t = tsk->real_parent;
787 if ((process_group(t) != process_group(tsk)) &&
788 (process_session(t) == process_session(tsk)) &&
789 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
790 has_stopped_jobs(process_group(tsk))) {
791 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
792 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
795 /* Let father know we died
797 * Thread signals are configurable, but you aren't going to use
798 * that to send signals to arbitary processes.
799 * That stops right now.
801 * If the parent exec id doesn't match the exec id we saved
802 * when we started then we know the parent has changed security
805 * If our self_exec id doesn't match our parent_exec_id then
806 * we have changed execution domain as these two values started
807 * the same after a fork.
811 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
812 ( tsk->parent_exec_id != t->self_exec_id ||
813 tsk->self_exec_id != tsk->parent_exec_id)
814 && !capable(CAP_KILL))
815 tsk->exit_signal = SIGCHLD;
818 /* If something other than our normal parent is ptracing us, then
819 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
820 * only has special meaning to our real parent.
822 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
823 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
824 do_notify_parent(tsk, signal);
825 } else if (tsk->ptrace) {
826 do_notify_parent(tsk, SIGCHLD);
830 if (tsk->exit_signal == -1 &&
831 (likely(tsk->ptrace == 0) ||
832 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
834 tsk->exit_state = state;
836 write_unlock_irq(&tasklist_lock);
838 list_for_each_safe(_p, _n, &ptrace_dead) {
840 t = list_entry(_p, struct task_struct, ptrace_list);
844 /* If the process is dead, release it - nobody will wait for it */
845 if (state == EXIT_DEAD)
849 fastcall NORET_TYPE void do_exit(long code)
851 struct task_struct *tsk = current;
854 profile_task_exit(tsk);
856 WARN_ON(atomic_read(&tsk->fs_excl));
858 if (unlikely(in_interrupt()))
859 panic("Aiee, killing interrupt handler!");
860 if (unlikely(!tsk->pid))
861 panic("Attempted to kill the idle task!");
862 if (unlikely(tsk == child_reaper))
863 panic("Attempted to kill init!");
865 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
866 current->ptrace_message = code;
867 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
871 * We're taking recursive faults here in do_exit. Safest is to just
872 * leave this task alone and wait for reboot.
874 if (unlikely(tsk->flags & PF_EXITING)) {
876 "Fixing recursive fault but reboot is needed!\n");
879 set_current_state(TASK_UNINTERRUPTIBLE);
883 tsk->flags |= PF_EXITING;
885 if (unlikely(in_atomic()))
886 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
887 current->comm, current->pid,
890 acct_update_integrals(tsk);
892 update_hiwater_rss(tsk->mm);
893 update_hiwater_vm(tsk->mm);
895 group_dead = atomic_dec_and_test(&tsk->signal->live);
897 hrtimer_cancel(&tsk->signal->real_timer);
898 exit_itimers(tsk->signal);
900 acct_collect(code, group_dead);
901 if (unlikely(tsk->robust_list))
902 exit_robust_list(tsk);
903 #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
904 if (unlikely(tsk->compat_robust_list))
905 compat_exit_robust_list(tsk);
907 if (unlikely(tsk->audit_context))
910 taskstats_exit(tsk, group_dead);
923 if (group_dead && tsk->signal->leader)
924 disassociate_ctty(1);
926 module_put(task_thread_info(tsk)->exec_domain->module);
928 module_put(tsk->binfmt->module);
930 tsk->exit_code = code;
931 proc_exit_connector(tsk);
933 exit_task_namespaces(tsk);
935 mpol_free(tsk->mempolicy);
936 tsk->mempolicy = NULL;
939 * This must happen late, after the PID is not
942 if (unlikely(!list_empty(&tsk->pi_state_list)))
943 exit_pi_state_list(tsk);
944 if (unlikely(current->pi_state_cache))
945 kfree(current->pi_state_cache);
947 * Make sure we are holding no locks:
949 debug_check_no_locks_held(tsk);
954 if (tsk->splice_pipe)
955 __free_pipe_info(tsk->splice_pipe);
958 /* causes final put_task_struct in finish_task_switch(). */
959 tsk->state = TASK_DEAD;
963 /* Avoid "noreturn function does return". */
965 cpu_relax(); /* For when BUG is null */
968 EXPORT_SYMBOL_GPL(do_exit);
970 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
978 EXPORT_SYMBOL(complete_and_exit);
980 asmlinkage long sys_exit(int error_code)
982 do_exit((error_code&0xff)<<8);
986 * Take down every thread in the group. This is called by fatal signals
987 * as well as by sys_exit_group (below).
990 do_group_exit(int exit_code)
992 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
994 if (current->signal->flags & SIGNAL_GROUP_EXIT)
995 exit_code = current->signal->group_exit_code;
996 else if (!thread_group_empty(current)) {
997 struct signal_struct *const sig = current->signal;
998 struct sighand_struct *const sighand = current->sighand;
999 spin_lock_irq(&sighand->siglock);
1000 if (sig->flags & SIGNAL_GROUP_EXIT)
1001 /* Another thread got here before we took the lock. */
1002 exit_code = sig->group_exit_code;
1004 sig->group_exit_code = exit_code;
1005 zap_other_threads(current);
1007 spin_unlock_irq(&sighand->siglock);
1015 * this kills every thread in the thread group. Note that any externally
1016 * wait4()-ing process will get the correct exit code - even if this
1017 * thread is not the thread group leader.
1019 asmlinkage void sys_exit_group(int error_code)
1021 do_group_exit((error_code & 0xff) << 8);
1024 static int eligible_child(pid_t pid, int options, struct task_struct *p)
1030 if (process_group(p) != process_group(current))
1032 } else if (pid != -1) {
1033 if (process_group(p) != -pid)
1038 * Do not consider detached threads that are
1041 if (p->exit_signal == -1 && !p->ptrace)
1044 /* Wait for all children (clone and not) if __WALL is set;
1045 * otherwise, wait for clone children *only* if __WCLONE is
1046 * set; otherwise, wait for non-clone children *only*. (Note:
1047 * A "clone" child here is one that reports to its parent
1048 * using a signal other than SIGCHLD.) */
1049 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1050 && !(options & __WALL))
1053 * Do not consider thread group leaders that are
1054 * in a non-empty thread group:
1056 if (delay_group_leader(p))
1059 if (security_task_wait(p))
1065 static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
1066 int why, int status,
1067 struct siginfo __user *infop,
1068 struct rusage __user *rusagep)
1070 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1074 retval = put_user(SIGCHLD, &infop->si_signo);
1076 retval = put_user(0, &infop->si_errno);
1078 retval = put_user((short)why, &infop->si_code);
1080 retval = put_user(pid, &infop->si_pid);
1082 retval = put_user(uid, &infop->si_uid);
1084 retval = put_user(status, &infop->si_status);
1091 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1092 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1093 * the lock and this task is uninteresting. If we return nonzero, we have
1094 * released the lock and the system call should return.
1096 static int wait_task_zombie(struct task_struct *p, int noreap,
1097 struct siginfo __user *infop,
1098 int __user *stat_addr, struct rusage __user *ru)
1100 unsigned long state;
1104 if (unlikely(noreap)) {
1107 int exit_code = p->exit_code;
1110 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1112 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1115 read_unlock(&tasklist_lock);
1116 if ((exit_code & 0x7f) == 0) {
1118 status = exit_code >> 8;
1120 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1121 status = exit_code & 0x7f;
1123 return wait_noreap_copyout(p, pid, uid, why,
1128 * Try to move the task's state to DEAD
1129 * only one thread is allowed to do this:
1131 state = xchg(&p->exit_state, EXIT_DEAD);
1132 if (state != EXIT_ZOMBIE) {
1133 BUG_ON(state != EXIT_DEAD);
1136 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1138 * This can only happen in a race with a ptraced thread
1139 * dying on another processor.
1144 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1145 struct signal_struct *psig;
1146 struct signal_struct *sig;
1149 * The resource counters for the group leader are in its
1150 * own task_struct. Those for dead threads in the group
1151 * are in its signal_struct, as are those for the child
1152 * processes it has previously reaped. All these
1153 * accumulate in the parent's signal_struct c* fields.
1155 * We don't bother to take a lock here to protect these
1156 * p->signal fields, because they are only touched by
1157 * __exit_signal, which runs with tasklist_lock
1158 * write-locked anyway, and so is excluded here. We do
1159 * need to protect the access to p->parent->signal fields,
1160 * as other threads in the parent group can be right
1161 * here reaping other children at the same time.
1163 spin_lock_irq(&p->parent->sighand->siglock);
1164 psig = p->parent->signal;
1167 cputime_add(psig->cutime,
1168 cputime_add(p->utime,
1169 cputime_add(sig->utime,
1172 cputime_add(psig->cstime,
1173 cputime_add(p->stime,
1174 cputime_add(sig->stime,
1177 p->min_flt + sig->min_flt + sig->cmin_flt;
1179 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1181 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1183 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1184 spin_unlock_irq(&p->parent->sighand->siglock);
1188 * Now we are sure this task is interesting, and no other
1189 * thread can reap it because we set its state to EXIT_DEAD.
1191 read_unlock(&tasklist_lock);
1193 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1194 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1195 ? p->signal->group_exit_code : p->exit_code;
1196 if (!retval && stat_addr)
1197 retval = put_user(status, stat_addr);
1198 if (!retval && infop)
1199 retval = put_user(SIGCHLD, &infop->si_signo);
1200 if (!retval && infop)
1201 retval = put_user(0, &infop->si_errno);
1202 if (!retval && infop) {
1205 if ((status & 0x7f) == 0) {
1209 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1212 retval = put_user((short)why, &infop->si_code);
1214 retval = put_user(status, &infop->si_status);
1216 if (!retval && infop)
1217 retval = put_user(p->pid, &infop->si_pid);
1218 if (!retval && infop)
1219 retval = put_user(p->uid, &infop->si_uid);
1221 // TODO: is this safe?
1222 p->exit_state = EXIT_ZOMBIE;
1226 if (p->real_parent != p->parent) {
1227 write_lock_irq(&tasklist_lock);
1228 /* Double-check with lock held. */
1229 if (p->real_parent != p->parent) {
1231 // TODO: is this safe?
1232 p->exit_state = EXIT_ZOMBIE;
1234 * If this is not a detached task, notify the parent.
1235 * If it's still not detached after that, don't release
1238 if (p->exit_signal != -1) {
1239 do_notify_parent(p, p->exit_signal);
1240 if (p->exit_signal != -1)
1244 write_unlock_irq(&tasklist_lock);
1253 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1254 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1255 * the lock and this task is uninteresting. If we return nonzero, we have
1256 * released the lock and the system call should return.
1258 static int wait_task_stopped(struct task_struct *p, int delayed_group_leader,
1259 int noreap, struct siginfo __user *infop,
1260 int __user *stat_addr, struct rusage __user *ru)
1262 int retval, exit_code;
1266 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1267 p->signal && p->signal->group_stop_count > 0)
1269 * A group stop is in progress and this is the group leader.
1270 * We won't report until all threads have stopped.
1275 * Now we are pretty sure this task is interesting.
1276 * Make sure it doesn't get reaped out from under us while we
1277 * give up the lock and then examine it below. We don't want to
1278 * keep holding onto the tasklist_lock while we call getrusage and
1279 * possibly take page faults for user memory.
1282 read_unlock(&tasklist_lock);
1284 if (unlikely(noreap)) {
1287 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1289 exit_code = p->exit_code;
1290 if (unlikely(!exit_code) ||
1291 unlikely(p->state & TASK_TRACED))
1293 return wait_noreap_copyout(p, pid, uid,
1294 why, (exit_code << 8) | 0x7f,
1298 write_lock_irq(&tasklist_lock);
1301 * This uses xchg to be atomic with the thread resuming and setting
1302 * it. It must also be done with the write lock held to prevent a
1303 * race with the EXIT_ZOMBIE case.
1305 exit_code = xchg(&p->exit_code, 0);
1306 if (unlikely(p->exit_state)) {
1308 * The task resumed and then died. Let the next iteration
1309 * catch it in EXIT_ZOMBIE. Note that exit_code might
1310 * already be zero here if it resumed and did _exit(0).
1311 * The task itself is dead and won't touch exit_code again;
1312 * other processors in this function are locked out.
1314 p->exit_code = exit_code;
1317 if (unlikely(exit_code == 0)) {
1319 * Another thread in this function got to it first, or it
1320 * resumed, or it resumed and then died.
1322 write_unlock_irq(&tasklist_lock);
1326 * We are returning to the wait loop without having successfully
1327 * removed the process and having released the lock. We cannot
1328 * continue, since the "p" task pointer is potentially stale.
1330 * Return -EAGAIN, and do_wait() will restart the loop from the
1331 * beginning. Do _not_ re-acquire the lock.
1336 /* move to end of parent's list to avoid starvation */
1340 write_unlock_irq(&tasklist_lock);
1342 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1343 if (!retval && stat_addr)
1344 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1345 if (!retval && infop)
1346 retval = put_user(SIGCHLD, &infop->si_signo);
1347 if (!retval && infop)
1348 retval = put_user(0, &infop->si_errno);
1349 if (!retval && infop)
1350 retval = put_user((short)((p->ptrace & PT_PTRACED)
1351 ? CLD_TRAPPED : CLD_STOPPED),
1353 if (!retval && infop)
1354 retval = put_user(exit_code, &infop->si_status);
1355 if (!retval && infop)
1356 retval = put_user(p->pid, &infop->si_pid);
1357 if (!retval && infop)
1358 retval = put_user(p->uid, &infop->si_uid);
1368 * Handle do_wait work for one task in a live, non-stopped state.
1369 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1370 * the lock and this task is uninteresting. If we return nonzero, we have
1371 * released the lock and the system call should return.
1373 static int wait_task_continued(struct task_struct *p, int noreap,
1374 struct siginfo __user *infop,
1375 int __user *stat_addr, struct rusage __user *ru)
1381 if (unlikely(!p->signal))
1384 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1387 spin_lock_irq(&p->sighand->siglock);
1388 /* Re-check with the lock held. */
1389 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1390 spin_unlock_irq(&p->sighand->siglock);
1394 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1395 spin_unlock_irq(&p->sighand->siglock);
1400 read_unlock(&tasklist_lock);
1403 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1405 if (!retval && stat_addr)
1406 retval = put_user(0xffff, stat_addr);
1410 retval = wait_noreap_copyout(p, pid, uid,
1411 CLD_CONTINUED, SIGCONT,
1413 BUG_ON(retval == 0);
1420 static inline int my_ptrace_child(struct task_struct *p)
1422 if (!(p->ptrace & PT_PTRACED))
1424 if (!(p->ptrace & PT_ATTACHED))
1427 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1428 * we are the attacher. If we are the real parent, this is a race
1429 * inside ptrace_attach. It is waiting for the tasklist_lock,
1430 * which we have to switch the parent links, but has already set
1431 * the flags in p->ptrace.
1433 return (p->parent != p->real_parent);
1436 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1437 int __user *stat_addr, struct rusage __user *ru)
1439 DECLARE_WAITQUEUE(wait, current);
1440 struct task_struct *tsk;
1443 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1446 * We will set this flag if we see any child that might later
1447 * match our criteria, even if we are not able to reap it yet.
1450 current->state = TASK_INTERRUPTIBLE;
1451 read_lock(&tasklist_lock);
1454 struct task_struct *p;
1455 struct list_head *_p;
1458 list_for_each(_p,&tsk->children) {
1459 p = list_entry(_p, struct task_struct, sibling);
1461 ret = eligible_child(pid, options, p);
1468 * When we hit the race with PTRACE_ATTACH,
1469 * we will not report this child. But the
1470 * race means it has not yet been moved to
1471 * our ptrace_children list, so we need to
1472 * set the flag here to avoid a spurious ECHILD
1473 * when the race happens with the only child.
1476 if (!my_ptrace_child(p))
1481 * It's stopped now, so it might later
1482 * continue, exit, or stop again.
1485 if (!(options & WUNTRACED) &&
1486 !my_ptrace_child(p))
1488 retval = wait_task_stopped(p, ret == 2,
1489 (options & WNOWAIT),
1492 if (retval == -EAGAIN)
1494 if (retval != 0) /* He released the lock. */
1499 if (p->exit_state == EXIT_DEAD)
1501 // case EXIT_ZOMBIE:
1502 if (p->exit_state == EXIT_ZOMBIE) {
1504 * Eligible but we cannot release
1508 goto check_continued;
1509 if (!likely(options & WEXITED))
1511 retval = wait_task_zombie(
1512 p, (options & WNOWAIT),
1513 infop, stat_addr, ru);
1514 /* He released the lock. */
1521 * It's running now, so it might later
1522 * exit, stop, or stop and then continue.
1525 if (!unlikely(options & WCONTINUED))
1527 retval = wait_task_continued(
1528 p, (options & WNOWAIT),
1529 infop, stat_addr, ru);
1530 if (retval != 0) /* He released the lock. */
1536 list_for_each(_p, &tsk->ptrace_children) {
1537 p = list_entry(_p, struct task_struct,
1539 if (!eligible_child(pid, options, p))
1545 if (options & __WNOTHREAD)
1547 tsk = next_thread(tsk);
1548 BUG_ON(tsk->signal != current->signal);
1549 } while (tsk != current);
1551 read_unlock(&tasklist_lock);
1554 if (options & WNOHANG)
1556 retval = -ERESTARTSYS;
1557 if (signal_pending(current))
1564 current->state = TASK_RUNNING;
1565 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1571 * For a WNOHANG return, clear out all the fields
1572 * we would set so the user can easily tell the
1576 retval = put_user(0, &infop->si_signo);
1578 retval = put_user(0, &infop->si_errno);
1580 retval = put_user(0, &infop->si_code);
1582 retval = put_user(0, &infop->si_pid);
1584 retval = put_user(0, &infop->si_uid);
1586 retval = put_user(0, &infop->si_status);
1592 asmlinkage long sys_waitid(int which, pid_t pid,
1593 struct siginfo __user *infop, int options,
1594 struct rusage __user *ru)
1598 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1600 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1620 ret = do_wait(pid, options, infop, NULL, ru);
1622 /* avoid REGPARM breakage on x86: */
1623 prevent_tail_call(ret);
1627 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1628 int options, struct rusage __user *ru)
1632 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1633 __WNOTHREAD|__WCLONE|__WALL))
1635 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1637 /* avoid REGPARM breakage on x86: */
1638 prevent_tail_call(ret);
1642 #ifdef __ARCH_WANT_SYS_WAITPID
1645 * sys_waitpid() remains for compatibility. waitpid() should be
1646 * implemented by calling sys_wait4() from libc.a.
1648 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1650 return sys_wait4(pid, stat_addr, options, NULL);