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/file.h>
22 #include <linux/binfmts.h>
23 #include <linux/ptrace.h>
24 #include <linux/profile.h>
25 #include <linux/mount.h>
26 #include <linux/proc_fs.h>
27 #include <linux/mempolicy.h>
28 #include <linux/cpuset.h>
29 #include <linux/syscalls.h>
30 #include <linux/signal.h>
31 #include <linux/posix-timers.h>
32 #include <linux/cn_proc.h>
33 #include <linux/mutex.h>
34 #include <linux/futex.h>
35 #include <linux/compat.h>
36 #include <linux/pipe_fs_i.h>
37 #include <linux/audit.h> /* for audit_free() */
38 #include <linux/resource.h>
40 #include <asm/uaccess.h>
41 #include <asm/unistd.h>
42 #include <asm/pgtable.h>
43 #include <asm/mmu_context.h>
45 extern void sem_exit (void);
46 extern struct task_struct *child_reaper;
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_rcu(&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)
140 atomic_dec(&p->user->processes);
141 write_lock_irq(&tasklist_lock);
143 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
147 * If we are the last non-leader member of the thread
148 * group, and the leader is zombie, then notify the
149 * group leader's parent process. (if it wants notification.)
152 leader = p->group_leader;
153 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
154 BUG_ON(leader->exit_signal == -1);
155 do_notify_parent(leader, leader->exit_signal);
157 * If we were the last child thread and the leader has
158 * exited already, and the leader's parent ignores SIGCHLD,
159 * then we are the one who should release the leader.
161 * do_notify_parent() will have marked it self-reaping in
164 zap_leader = (leader->exit_signal == -1);
168 write_unlock_irq(&tasklist_lock);
171 call_rcu(&p->rcu, delayed_put_task_struct);
174 if (unlikely(zap_leader))
179 * This checks not only the pgrp, but falls back on the pid if no
180 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
183 int session_of_pgrp(int pgrp)
185 struct task_struct *p;
188 read_lock(&tasklist_lock);
189 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
190 if (p->signal->session > 0) {
191 sid = p->signal->session;
194 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
195 p = find_task_by_pid(pgrp);
197 sid = p->signal->session;
199 read_unlock(&tasklist_lock);
205 * Determine if a process group is "orphaned", according to the POSIX
206 * definition in 2.2.2.52. Orphaned process groups are not to be affected
207 * by terminal-generated stop signals. Newly orphaned process groups are
208 * to receive a SIGHUP and a SIGCONT.
210 * "I ask you, have you ever known what it is to be an orphan?"
212 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
214 struct task_struct *p;
217 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
218 if (p == ignored_task
220 || p->real_parent->pid == 1)
222 if (process_group(p->real_parent) != pgrp
223 && p->real_parent->signal->session == p->signal->session) {
227 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
228 return ret; /* (sighing) "Often!" */
231 int is_orphaned_pgrp(int pgrp)
235 read_lock(&tasklist_lock);
236 retval = will_become_orphaned_pgrp(pgrp, NULL);
237 read_unlock(&tasklist_lock);
242 static int has_stopped_jobs(int pgrp)
245 struct task_struct *p;
247 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
248 if (p->state != TASK_STOPPED)
251 /* If p is stopped by a debugger on a signal that won't
252 stop it, then don't count p as stopped. This isn't
253 perfect but it's a good approximation. */
254 if (unlikely (p->ptrace)
255 && p->exit_code != SIGSTOP
256 && p->exit_code != SIGTSTP
257 && p->exit_code != SIGTTOU
258 && p->exit_code != SIGTTIN)
263 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
268 * reparent_to_init - Reparent the calling kernel thread to the init task.
270 * If a kernel thread is launched as a result of a system call, or if
271 * it ever exits, it should generally reparent itself to init so that
272 * it is correctly cleaned up on exit.
274 * The various task state such as scheduling policy and priority may have
275 * been inherited from a user process, so we reset them to sane values here.
277 * NOTE that reparent_to_init() gives the caller full capabilities.
279 static void reparent_to_init(void)
281 write_lock_irq(&tasklist_lock);
283 ptrace_unlink(current);
284 /* Reparent to init */
285 remove_parent(current);
286 current->parent = child_reaper;
287 current->real_parent = child_reaper;
290 /* Set the exit signal to SIGCHLD so we signal init on exit */
291 current->exit_signal = SIGCHLD;
293 if ((current->policy == SCHED_NORMAL ||
294 current->policy == SCHED_BATCH)
295 && (task_nice(current) < 0))
296 set_user_nice(current, 0);
300 security_task_reparent_to_init(current);
301 memcpy(current->signal->rlim, init_task.signal->rlim,
302 sizeof(current->signal->rlim));
303 atomic_inc(&(INIT_USER->__count));
304 write_unlock_irq(&tasklist_lock);
305 switch_uid(INIT_USER);
308 void __set_special_pids(pid_t session, pid_t pgrp)
310 struct task_struct *curr = current->group_leader;
312 if (curr->signal->session != session) {
313 detach_pid(curr, PIDTYPE_SID);
314 curr->signal->session = session;
315 attach_pid(curr, PIDTYPE_SID, session);
317 if (process_group(curr) != pgrp) {
318 detach_pid(curr, PIDTYPE_PGID);
319 curr->signal->pgrp = pgrp;
320 attach_pid(curr, PIDTYPE_PGID, pgrp);
324 void set_special_pids(pid_t session, pid_t pgrp)
326 write_lock_irq(&tasklist_lock);
327 __set_special_pids(session, pgrp);
328 write_unlock_irq(&tasklist_lock);
332 * Let kernel threads use this to say that they
333 * allow a certain signal (since daemonize() will
334 * have disabled all of them by default).
336 int allow_signal(int sig)
338 if (!valid_signal(sig) || sig < 1)
341 spin_lock_irq(¤t->sighand->siglock);
342 sigdelset(¤t->blocked, sig);
344 /* Kernel threads handle their own signals.
345 Let the signal code know it'll be handled, so
346 that they don't get converted to SIGKILL or
347 just silently dropped */
348 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
351 spin_unlock_irq(¤t->sighand->siglock);
355 EXPORT_SYMBOL(allow_signal);
357 int disallow_signal(int sig)
359 if (!valid_signal(sig) || sig < 1)
362 spin_lock_irq(¤t->sighand->siglock);
363 sigaddset(¤t->blocked, sig);
365 spin_unlock_irq(¤t->sighand->siglock);
369 EXPORT_SYMBOL(disallow_signal);
372 * Put all the gunge required to become a kernel thread without
373 * attached user resources in one place where it belongs.
376 void daemonize(const char *name, ...)
379 struct fs_struct *fs;
382 va_start(args, name);
383 vsnprintf(current->comm, sizeof(current->comm), name, args);
387 * If we were started as result of loading a module, close all of the
388 * user space pages. We don't need them, and if we didn't close them
389 * they would be locked into memory.
393 set_special_pids(1, 1);
394 mutex_lock(&tty_mutex);
395 current->signal->tty = NULL;
396 mutex_unlock(&tty_mutex);
398 /* Block and flush all signals */
399 sigfillset(&blocked);
400 sigprocmask(SIG_BLOCK, &blocked, NULL);
401 flush_signals(current);
403 /* Become as one with the init task */
405 exit_fs(current); /* current->fs->count--; */
408 atomic_inc(&fs->count);
409 exit_namespace(current);
410 current->namespace = init_task.namespace;
411 get_namespace(current->namespace);
413 current->files = init_task.files;
414 atomic_inc(¤t->files->count);
419 EXPORT_SYMBOL(daemonize);
421 static void close_files(struct files_struct * files)
429 * It is safe to dereference the fd table without RCU or
430 * ->file_lock because this is the last reference to the
433 fdt = files_fdtable(files);
437 if (i >= fdt->max_fdset || i >= fdt->max_fds)
439 set = fdt->open_fds->fds_bits[j++];
442 struct file * file = xchg(&fdt->fd[i], NULL);
444 filp_close(file, files);
452 struct files_struct *get_files_struct(struct task_struct *task)
454 struct files_struct *files;
459 atomic_inc(&files->count);
465 void fastcall put_files_struct(struct files_struct *files)
469 if (atomic_dec_and_test(&files->count)) {
472 * Free the fd and fdset arrays if we expanded them.
473 * If the fdtable was embedded, pass files for freeing
474 * at the end of the RCU grace period. Otherwise,
475 * you can free files immediately.
477 fdt = files_fdtable(files);
478 if (fdt == &files->fdtab)
479 fdt->free_files = files;
481 kmem_cache_free(files_cachep, files);
486 EXPORT_SYMBOL(put_files_struct);
488 static inline void __exit_files(struct task_struct *tsk)
490 struct files_struct * files = tsk->files;
496 put_files_struct(files);
500 void exit_files(struct task_struct *tsk)
505 static inline void __put_fs_struct(struct fs_struct *fs)
507 /* No need to hold fs->lock if we are killing it */
508 if (atomic_dec_and_test(&fs->count)) {
515 mntput(fs->altrootmnt);
517 kmem_cache_free(fs_cachep, fs);
521 void put_fs_struct(struct fs_struct *fs)
526 static inline void __exit_fs(struct task_struct *tsk)
528 struct fs_struct * fs = tsk->fs;
538 void exit_fs(struct task_struct *tsk)
543 EXPORT_SYMBOL_GPL(exit_fs);
546 * Turn us into a lazy TLB process if we
549 static void exit_mm(struct task_struct * tsk)
551 struct mm_struct *mm = tsk->mm;
557 * Serialize with any possible pending coredump.
558 * We must hold mmap_sem around checking core_waiters
559 * and clearing tsk->mm. The core-inducing thread
560 * will increment core_waiters for each thread in the
561 * group with ->mm != NULL.
563 down_read(&mm->mmap_sem);
564 if (mm->core_waiters) {
565 up_read(&mm->mmap_sem);
566 down_write(&mm->mmap_sem);
567 if (!--mm->core_waiters)
568 complete(mm->core_startup_done);
569 up_write(&mm->mmap_sem);
571 wait_for_completion(&mm->core_done);
572 down_read(&mm->mmap_sem);
574 atomic_inc(&mm->mm_count);
575 BUG_ON(mm != tsk->active_mm);
576 /* more a memory barrier than a real lock */
579 up_read(&mm->mmap_sem);
580 enter_lazy_tlb(mm, current);
585 static inline void choose_new_parent(task_t *p, task_t *reaper)
588 * Make sure we're not reparenting to ourselves and that
589 * the parent is not a zombie.
591 BUG_ON(p == reaper || reaper->exit_state);
592 p->real_parent = reaper;
595 static void reparent_thread(task_t *p, task_t *father, int traced)
597 /* We don't want people slaying init. */
598 if (p->exit_signal != -1)
599 p->exit_signal = SIGCHLD;
601 if (p->pdeath_signal)
602 /* We already hold the tasklist_lock here. */
603 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
605 /* Move the child from its dying parent to the new one. */
606 if (unlikely(traced)) {
607 /* Preserve ptrace links if someone else is tracing this child. */
608 list_del_init(&p->ptrace_list);
609 if (p->parent != p->real_parent)
610 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
612 /* If this child is being traced, then we're the one tracing it
613 * anyway, so let go of it.
617 p->parent = p->real_parent;
620 /* If we'd notified the old parent about this child's death,
621 * also notify the new parent.
623 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
624 thread_group_empty(p))
625 do_notify_parent(p, p->exit_signal);
626 else if (p->state == TASK_TRACED) {
628 * If it was at a trace stop, turn it into
629 * a normal stop since it's no longer being
637 * process group orphan check
638 * Case ii: Our child is in a different pgrp
639 * than we are, and it was the only connection
640 * outside, so the child pgrp is now orphaned.
642 if ((process_group(p) != process_group(father)) &&
643 (p->signal->session == father->signal->session)) {
644 int pgrp = process_group(p);
646 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
647 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
648 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
654 * When we die, we re-parent all our children.
655 * Try to give them to another thread in our thread
656 * group, and if no such member exists, give it to
657 * the global child reaper process (ie "init")
659 static void forget_original_parent(struct task_struct * father,
660 struct list_head *to_release)
662 struct task_struct *p, *reaper = father;
663 struct list_head *_p, *_n;
666 reaper = next_thread(reaper);
667 if (reaper == father) {
668 reaper = child_reaper;
671 } while (reaper->exit_state);
674 * There are only two places where our children can be:
676 * - in our child list
677 * - in our ptraced child list
679 * Search them and reparent children.
681 list_for_each_safe(_p, _n, &father->children) {
683 p = list_entry(_p,struct task_struct,sibling);
687 /* if father isn't the real parent, then ptrace must be enabled */
688 BUG_ON(father != p->real_parent && !ptrace);
690 if (father == p->real_parent) {
691 /* reparent with a reaper, real father it's us */
692 choose_new_parent(p, reaper);
693 reparent_thread(p, father, 0);
695 /* reparent ptraced task to its real parent */
697 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
698 thread_group_empty(p))
699 do_notify_parent(p, p->exit_signal);
703 * if the ptraced child is a zombie with exit_signal == -1
704 * we must collect it before we exit, or it will remain
705 * zombie forever since we prevented it from self-reap itself
706 * while it was being traced by us, to be able to see it in wait4.
708 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
709 list_add(&p->ptrace_list, to_release);
711 list_for_each_safe(_p, _n, &father->ptrace_children) {
712 p = list_entry(_p,struct task_struct,ptrace_list);
713 choose_new_parent(p, reaper);
714 reparent_thread(p, father, 1);
719 * Send signals to all our closest relatives so that they know
720 * to properly mourn us..
722 static void exit_notify(struct task_struct *tsk)
725 struct task_struct *t;
726 struct list_head ptrace_dead, *_p, *_n;
728 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
729 && !thread_group_empty(tsk)) {
731 * This occurs when there was a race between our exit
732 * syscall and a group signal choosing us as the one to
733 * wake up. It could be that we are the only thread
734 * alerted to check for pending signals, but another thread
735 * should be woken now to take the signal since we will not.
736 * Now we'll wake all the threads in the group just to make
737 * sure someone gets all the pending signals.
739 read_lock(&tasklist_lock);
740 spin_lock_irq(&tsk->sighand->siglock);
741 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
742 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
743 recalc_sigpending_tsk(t);
744 if (signal_pending(t))
745 signal_wake_up(t, 0);
747 spin_unlock_irq(&tsk->sighand->siglock);
748 read_unlock(&tasklist_lock);
751 write_lock_irq(&tasklist_lock);
754 * This does two things:
756 * A. Make init inherit all the child processes
757 * B. Check to see if any process groups have become orphaned
758 * as a result of our exiting, and if they have any stopped
759 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
762 INIT_LIST_HEAD(&ptrace_dead);
763 forget_original_parent(tsk, &ptrace_dead);
764 BUG_ON(!list_empty(&tsk->children));
765 BUG_ON(!list_empty(&tsk->ptrace_children));
768 * Check to see if any process groups have become orphaned
769 * as a result of our exiting, and if they have any stopped
770 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
772 * Case i: Our father is in a different pgrp than we are
773 * and we were the only connection outside, so our pgrp
774 * is about to become orphaned.
777 t = tsk->real_parent;
779 if ((process_group(t) != process_group(tsk)) &&
780 (t->signal->session == tsk->signal->session) &&
781 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
782 has_stopped_jobs(process_group(tsk))) {
783 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
784 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
787 /* Let father know we died
789 * Thread signals are configurable, but you aren't going to use
790 * that to send signals to arbitary processes.
791 * That stops right now.
793 * If the parent exec id doesn't match the exec id we saved
794 * when we started then we know the parent has changed security
797 * If our self_exec id doesn't match our parent_exec_id then
798 * we have changed execution domain as these two values started
799 * the same after a fork.
803 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
804 ( tsk->parent_exec_id != t->self_exec_id ||
805 tsk->self_exec_id != tsk->parent_exec_id)
806 && !capable(CAP_KILL))
807 tsk->exit_signal = SIGCHLD;
810 /* If something other than our normal parent is ptracing us, then
811 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
812 * only has special meaning to our real parent.
814 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
815 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
816 do_notify_parent(tsk, signal);
817 } else if (tsk->ptrace) {
818 do_notify_parent(tsk, SIGCHLD);
822 if (tsk->exit_signal == -1 &&
823 (likely(tsk->ptrace == 0) ||
824 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
826 tsk->exit_state = state;
828 write_unlock_irq(&tasklist_lock);
830 list_for_each_safe(_p, _n, &ptrace_dead) {
832 t = list_entry(_p,struct task_struct,ptrace_list);
836 /* If the process is dead, release it - nobody will wait for it */
837 if (state == EXIT_DEAD)
841 fastcall NORET_TYPE void do_exit(long code)
843 struct task_struct *tsk = current;
846 profile_task_exit(tsk);
848 WARN_ON(atomic_read(&tsk->fs_excl));
850 if (unlikely(in_interrupt()))
851 panic("Aiee, killing interrupt handler!");
852 if (unlikely(!tsk->pid))
853 panic("Attempted to kill the idle task!");
854 if (unlikely(tsk == child_reaper))
855 panic("Attempted to kill init!");
857 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
858 current->ptrace_message = code;
859 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
863 * We're taking recursive faults here in do_exit. Safest is to just
864 * leave this task alone and wait for reboot.
866 if (unlikely(tsk->flags & PF_EXITING)) {
868 "Fixing recursive fault but reboot is needed!\n");
871 set_current_state(TASK_UNINTERRUPTIBLE);
875 tsk->flags |= PF_EXITING;
877 if (unlikely(in_atomic()))
878 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
879 current->comm, current->pid,
882 acct_update_integrals(tsk);
884 update_hiwater_rss(tsk->mm);
885 update_hiwater_vm(tsk->mm);
887 group_dead = atomic_dec_and_test(&tsk->signal->live);
889 hrtimer_cancel(&tsk->signal->real_timer);
890 exit_itimers(tsk->signal);
892 acct_collect(code, group_dead);
893 if (unlikely(tsk->robust_list))
894 exit_robust_list(tsk);
895 #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
896 if (unlikely(tsk->compat_robust_list))
897 compat_exit_robust_list(tsk);
899 if (unlikely(tsk->audit_context))
913 if (group_dead && tsk->signal->leader)
914 disassociate_ctty(1);
916 module_put(task_thread_info(tsk)->exec_domain->module);
918 module_put(tsk->binfmt->module);
920 tsk->exit_code = code;
921 proc_exit_connector(tsk);
924 mpol_free(tsk->mempolicy);
925 tsk->mempolicy = NULL;
928 * This must happen late, after the PID is not
931 if (unlikely(!list_empty(&tsk->pi_state_list)))
932 exit_pi_state_list(tsk);
933 if (unlikely(current->pi_state_cache))
934 kfree(current->pi_state_cache);
936 * If DEBUG_MUTEXES is on, make sure we are holding no locks:
938 mutex_debug_check_no_locks_held(tsk);
939 rt_mutex_debug_check_no_locks_held(tsk);
944 if (tsk->splice_pipe)
945 __free_pipe_info(tsk->splice_pipe);
947 /* PF_DEAD causes final put_task_struct after we schedule. */
949 BUG_ON(tsk->flags & PF_DEAD);
950 tsk->flags |= PF_DEAD;
954 /* Avoid "noreturn function does return". */
958 EXPORT_SYMBOL_GPL(do_exit);
960 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
968 EXPORT_SYMBOL(complete_and_exit);
970 asmlinkage long sys_exit(int error_code)
972 do_exit((error_code&0xff)<<8);
976 * Take down every thread in the group. This is called by fatal signals
977 * as well as by sys_exit_group (below).
980 do_group_exit(int exit_code)
982 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
984 if (current->signal->flags & SIGNAL_GROUP_EXIT)
985 exit_code = current->signal->group_exit_code;
986 else if (!thread_group_empty(current)) {
987 struct signal_struct *const sig = current->signal;
988 struct sighand_struct *const sighand = current->sighand;
989 spin_lock_irq(&sighand->siglock);
990 if (sig->flags & SIGNAL_GROUP_EXIT)
991 /* Another thread got here before we took the lock. */
992 exit_code = sig->group_exit_code;
994 sig->group_exit_code = exit_code;
995 zap_other_threads(current);
997 spin_unlock_irq(&sighand->siglock);
1005 * this kills every thread in the thread group. Note that any externally
1006 * wait4()-ing process will get the correct exit code - even if this
1007 * thread is not the thread group leader.
1009 asmlinkage void sys_exit_group(int error_code)
1011 do_group_exit((error_code & 0xff) << 8);
1014 static int eligible_child(pid_t pid, int options, task_t *p)
1020 if (process_group(p) != process_group(current))
1022 } else if (pid != -1) {
1023 if (process_group(p) != -pid)
1028 * Do not consider detached threads that are
1031 if (p->exit_signal == -1 && !p->ptrace)
1034 /* Wait for all children (clone and not) if __WALL is set;
1035 * otherwise, wait for clone children *only* if __WCLONE is
1036 * set; otherwise, wait for non-clone children *only*. (Note:
1037 * A "clone" child here is one that reports to its parent
1038 * using a signal other than SIGCHLD.) */
1039 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1040 && !(options & __WALL))
1043 * Do not consider thread group leaders that are
1044 * in a non-empty thread group:
1046 if (current->tgid != p->tgid && delay_group_leader(p))
1049 if (security_task_wait(p))
1055 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
1056 int why, int status,
1057 struct siginfo __user *infop,
1058 struct rusage __user *rusagep)
1060 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1063 retval = put_user(SIGCHLD, &infop->si_signo);
1065 retval = put_user(0, &infop->si_errno);
1067 retval = put_user((short)why, &infop->si_code);
1069 retval = put_user(pid, &infop->si_pid);
1071 retval = put_user(uid, &infop->si_uid);
1073 retval = put_user(status, &infop->si_status);
1080 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1081 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1082 * the lock and this task is uninteresting. If we return nonzero, we have
1083 * released the lock and the system call should return.
1085 static int wait_task_zombie(task_t *p, int noreap,
1086 struct siginfo __user *infop,
1087 int __user *stat_addr, struct rusage __user *ru)
1089 unsigned long state;
1093 if (unlikely(noreap)) {
1096 int exit_code = p->exit_code;
1099 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1101 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1104 read_unlock(&tasklist_lock);
1105 if ((exit_code & 0x7f) == 0) {
1107 status = exit_code >> 8;
1109 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1110 status = exit_code & 0x7f;
1112 return wait_noreap_copyout(p, pid, uid, why,
1117 * Try to move the task's state to DEAD
1118 * only one thread is allowed to do this:
1120 state = xchg(&p->exit_state, EXIT_DEAD);
1121 if (state != EXIT_ZOMBIE) {
1122 BUG_ON(state != EXIT_DEAD);
1125 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1127 * This can only happen in a race with a ptraced thread
1128 * dying on another processor.
1133 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1134 struct signal_struct *psig;
1135 struct signal_struct *sig;
1138 * The resource counters for the group leader are in its
1139 * own task_struct. Those for dead threads in the group
1140 * are in its signal_struct, as are those for the child
1141 * processes it has previously reaped. All these
1142 * accumulate in the parent's signal_struct c* fields.
1144 * We don't bother to take a lock here to protect these
1145 * p->signal fields, because they are only touched by
1146 * __exit_signal, which runs with tasklist_lock
1147 * write-locked anyway, and so is excluded here. We do
1148 * need to protect the access to p->parent->signal fields,
1149 * as other threads in the parent group can be right
1150 * here reaping other children at the same time.
1152 spin_lock_irq(&p->parent->sighand->siglock);
1153 psig = p->parent->signal;
1156 cputime_add(psig->cutime,
1157 cputime_add(p->utime,
1158 cputime_add(sig->utime,
1161 cputime_add(psig->cstime,
1162 cputime_add(p->stime,
1163 cputime_add(sig->stime,
1166 p->min_flt + sig->min_flt + sig->cmin_flt;
1168 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1170 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1172 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1173 spin_unlock_irq(&p->parent->sighand->siglock);
1177 * Now we are sure this task is interesting, and no other
1178 * thread can reap it because we set its state to EXIT_DEAD.
1180 read_unlock(&tasklist_lock);
1182 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1183 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1184 ? p->signal->group_exit_code : p->exit_code;
1185 if (!retval && stat_addr)
1186 retval = put_user(status, stat_addr);
1187 if (!retval && infop)
1188 retval = put_user(SIGCHLD, &infop->si_signo);
1189 if (!retval && infop)
1190 retval = put_user(0, &infop->si_errno);
1191 if (!retval && infop) {
1194 if ((status & 0x7f) == 0) {
1198 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1201 retval = put_user((short)why, &infop->si_code);
1203 retval = put_user(status, &infop->si_status);
1205 if (!retval && infop)
1206 retval = put_user(p->pid, &infop->si_pid);
1207 if (!retval && infop)
1208 retval = put_user(p->uid, &infop->si_uid);
1210 // TODO: is this safe?
1211 p->exit_state = EXIT_ZOMBIE;
1215 if (p->real_parent != p->parent) {
1216 write_lock_irq(&tasklist_lock);
1217 /* Double-check with lock held. */
1218 if (p->real_parent != p->parent) {
1220 // TODO: is this safe?
1221 p->exit_state = EXIT_ZOMBIE;
1223 * If this is not a detached task, notify the parent.
1224 * If it's still not detached after that, don't release
1227 if (p->exit_signal != -1) {
1228 do_notify_parent(p, p->exit_signal);
1229 if (p->exit_signal != -1)
1233 write_unlock_irq(&tasklist_lock);
1242 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1243 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1244 * the lock and this task is uninteresting. If we return nonzero, we have
1245 * released the lock and the system call should return.
1247 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1248 struct siginfo __user *infop,
1249 int __user *stat_addr, struct rusage __user *ru)
1251 int retval, exit_code;
1255 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1256 p->signal && p->signal->group_stop_count > 0)
1258 * A group stop is in progress and this is the group leader.
1259 * We won't report until all threads have stopped.
1264 * Now we are pretty sure this task is interesting.
1265 * Make sure it doesn't get reaped out from under us while we
1266 * give up the lock and then examine it below. We don't want to
1267 * keep holding onto the tasklist_lock while we call getrusage and
1268 * possibly take page faults for user memory.
1271 read_unlock(&tasklist_lock);
1273 if (unlikely(noreap)) {
1276 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1278 exit_code = p->exit_code;
1279 if (unlikely(!exit_code) ||
1280 unlikely(p->state & TASK_TRACED))
1282 return wait_noreap_copyout(p, pid, uid,
1283 why, (exit_code << 8) | 0x7f,
1287 write_lock_irq(&tasklist_lock);
1290 * This uses xchg to be atomic with the thread resuming and setting
1291 * it. It must also be done with the write lock held to prevent a
1292 * race with the EXIT_ZOMBIE case.
1294 exit_code = xchg(&p->exit_code, 0);
1295 if (unlikely(p->exit_state)) {
1297 * The task resumed and then died. Let the next iteration
1298 * catch it in EXIT_ZOMBIE. Note that exit_code might
1299 * already be zero here if it resumed and did _exit(0).
1300 * The task itself is dead and won't touch exit_code again;
1301 * other processors in this function are locked out.
1303 p->exit_code = exit_code;
1306 if (unlikely(exit_code == 0)) {
1308 * Another thread in this function got to it first, or it
1309 * resumed, or it resumed and then died.
1311 write_unlock_irq(&tasklist_lock);
1315 * We are returning to the wait loop without having successfully
1316 * removed the process and having released the lock. We cannot
1317 * continue, since the "p" task pointer is potentially stale.
1319 * Return -EAGAIN, and do_wait() will restart the loop from the
1320 * beginning. Do _not_ re-acquire the lock.
1325 /* move to end of parent's list to avoid starvation */
1329 write_unlock_irq(&tasklist_lock);
1331 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1332 if (!retval && stat_addr)
1333 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1334 if (!retval && infop)
1335 retval = put_user(SIGCHLD, &infop->si_signo);
1336 if (!retval && infop)
1337 retval = put_user(0, &infop->si_errno);
1338 if (!retval && infop)
1339 retval = put_user((short)((p->ptrace & PT_PTRACED)
1340 ? CLD_TRAPPED : CLD_STOPPED),
1342 if (!retval && infop)
1343 retval = put_user(exit_code, &infop->si_status);
1344 if (!retval && infop)
1345 retval = put_user(p->pid, &infop->si_pid);
1346 if (!retval && infop)
1347 retval = put_user(p->uid, &infop->si_uid);
1357 * Handle do_wait work for one task in a live, non-stopped state.
1358 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1359 * the lock and this task is uninteresting. If we return nonzero, we have
1360 * released the lock and the system call should return.
1362 static int wait_task_continued(task_t *p, int noreap,
1363 struct siginfo __user *infop,
1364 int __user *stat_addr, struct rusage __user *ru)
1370 if (unlikely(!p->signal))
1373 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1376 spin_lock_irq(&p->sighand->siglock);
1377 /* Re-check with the lock held. */
1378 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1379 spin_unlock_irq(&p->sighand->siglock);
1383 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1384 spin_unlock_irq(&p->sighand->siglock);
1389 read_unlock(&tasklist_lock);
1392 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1394 if (!retval && stat_addr)
1395 retval = put_user(0xffff, stat_addr);
1399 retval = wait_noreap_copyout(p, pid, uid,
1400 CLD_CONTINUED, SIGCONT,
1402 BUG_ON(retval == 0);
1409 static inline int my_ptrace_child(struct task_struct *p)
1411 if (!(p->ptrace & PT_PTRACED))
1413 if (!(p->ptrace & PT_ATTACHED))
1416 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1417 * we are the attacher. If we are the real parent, this is a race
1418 * inside ptrace_attach. It is waiting for the tasklist_lock,
1419 * which we have to switch the parent links, but has already set
1420 * the flags in p->ptrace.
1422 return (p->parent != p->real_parent);
1425 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1426 int __user *stat_addr, struct rusage __user *ru)
1428 DECLARE_WAITQUEUE(wait, current);
1429 struct task_struct *tsk;
1432 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1435 * We will set this flag if we see any child that might later
1436 * match our criteria, even if we are not able to reap it yet.
1439 current->state = TASK_INTERRUPTIBLE;
1440 read_lock(&tasklist_lock);
1443 struct task_struct *p;
1444 struct list_head *_p;
1447 list_for_each(_p,&tsk->children) {
1448 p = list_entry(_p,struct task_struct,sibling);
1450 ret = eligible_child(pid, options, p);
1457 * When we hit the race with PTRACE_ATTACH,
1458 * we will not report this child. But the
1459 * race means it has not yet been moved to
1460 * our ptrace_children list, so we need to
1461 * set the flag here to avoid a spurious ECHILD
1462 * when the race happens with the only child.
1465 if (!my_ptrace_child(p))
1470 * It's stopped now, so it might later
1471 * continue, exit, or stop again.
1474 if (!(options & WUNTRACED) &&
1475 !my_ptrace_child(p))
1477 retval = wait_task_stopped(p, ret == 2,
1478 (options & WNOWAIT),
1481 if (retval == -EAGAIN)
1483 if (retval != 0) /* He released the lock. */
1488 if (p->exit_state == EXIT_DEAD)
1490 // case EXIT_ZOMBIE:
1491 if (p->exit_state == EXIT_ZOMBIE) {
1493 * Eligible but we cannot release
1497 goto check_continued;
1498 if (!likely(options & WEXITED))
1500 retval = wait_task_zombie(
1501 p, (options & WNOWAIT),
1502 infop, stat_addr, ru);
1503 /* He released the lock. */
1510 * It's running now, so it might later
1511 * exit, stop, or stop and then continue.
1514 if (!unlikely(options & WCONTINUED))
1516 retval = wait_task_continued(
1517 p, (options & WNOWAIT),
1518 infop, stat_addr, ru);
1519 if (retval != 0) /* He released the lock. */
1525 list_for_each(_p, &tsk->ptrace_children) {
1526 p = list_entry(_p, struct task_struct,
1528 if (!eligible_child(pid, options, p))
1534 if (options & __WNOTHREAD)
1536 tsk = next_thread(tsk);
1537 BUG_ON(tsk->signal != current->signal);
1538 } while (tsk != current);
1540 read_unlock(&tasklist_lock);
1543 if (options & WNOHANG)
1545 retval = -ERESTARTSYS;
1546 if (signal_pending(current))
1553 current->state = TASK_RUNNING;
1554 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1560 * For a WNOHANG return, clear out all the fields
1561 * we would set so the user can easily tell the
1565 retval = put_user(0, &infop->si_signo);
1567 retval = put_user(0, &infop->si_errno);
1569 retval = put_user(0, &infop->si_code);
1571 retval = put_user(0, &infop->si_pid);
1573 retval = put_user(0, &infop->si_uid);
1575 retval = put_user(0, &infop->si_status);
1581 asmlinkage long sys_waitid(int which, pid_t pid,
1582 struct siginfo __user *infop, int options,
1583 struct rusage __user *ru)
1587 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1589 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1609 ret = do_wait(pid, options, infop, NULL, ru);
1611 /* avoid REGPARM breakage on x86: */
1612 prevent_tail_call(ret);
1616 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1617 int options, struct rusage __user *ru)
1621 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1622 __WNOTHREAD|__WCLONE|__WALL))
1624 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1626 /* avoid REGPARM breakage on x86: */
1627 prevent_tail_call(ret);
1631 #ifdef __ARCH_WANT_SYS_WAITPID
1634 * sys_waitpid() remains for compatibility. waitpid() should be
1635 * implemented by calling sys_wait4() from libc.a.
1637 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1639 return sys_wait4(pid, stat_addr, options, NULL);