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/taskstats_kern.h>
29 #include <linux/delayacct.h>
30 #include <linux/cpuset.h>
31 #include <linux/syscalls.h>
32 #include <linux/signal.h>
33 #include <linux/posix-timers.h>
34 #include <linux/cn_proc.h>
35 #include <linux/mutex.h>
36 #include <linux/futex.h>
37 #include <linux/compat.h>
38 #include <linux/pipe_fs_i.h>
39 #include <linux/audit.h> /* for audit_free() */
40 #include <linux/resource.h>
41 #include <linux/vs_base.h>
42 #include <linux/vs_context.h>
43 #include <linux/vs_network.h>
44 #include <linux/vs_limit.h>
46 #include <asm/uaccess.h>
47 #include <asm/unistd.h>
48 #include <asm/pgtable.h>
49 #include <asm/mmu_context.h>
51 extern void sem_exit (void);
52 extern struct task_struct *child_reaper;
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)
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 (p->signal->session > 0) {
197 sid = p->signal->session;
200 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
201 p = find_task_by_pid(pgrp);
203 sid = p->signal->session;
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 || p->real_parent->pid == 1)
228 if (process_group(p->real_parent) != pgrp
229 && p->real_parent->signal->session == p->signal->session) {
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)
257 /* If p is stopped by a debugger on a signal that won't
258 stop it, then don't count p as stopped. This isn't
259 perfect but it's a good approximation. */
260 if (unlikely (p->ptrace)
261 && p->exit_code != SIGSTOP
262 && p->exit_code != SIGTSTP
263 && p->exit_code != SIGTTOU
264 && p->exit_code != SIGTTIN)
269 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
274 * reparent_to_init - Reparent the calling kernel thread to the init task.
276 * If a kernel thread is launched as a result of a system call, or if
277 * it ever exits, it should generally reparent itself to init so that
278 * it is correctly cleaned up on exit.
280 * The various task state such as scheduling policy and priority may have
281 * been inherited from a user process, so we reset them to sane values here.
283 * NOTE that reparent_to_init() gives the caller full capabilities.
285 static void reparent_to_init(void)
287 write_lock_irq(&tasklist_lock);
289 ptrace_unlink(current);
290 /* Reparent to init */
291 remove_parent(current);
292 current->parent = child_reaper;
293 current->real_parent = child_reaper;
296 /* Set the exit signal to SIGCHLD so we signal init on exit */
297 current->exit_signal = SIGCHLD;
299 if ((current->policy == SCHED_NORMAL ||
300 current->policy == SCHED_BATCH)
301 && (task_nice(current) < 0))
302 set_user_nice(current, 0);
306 security_task_reparent_to_init(current);
307 memcpy(current->signal->rlim, init_task.signal->rlim,
308 sizeof(current->signal->rlim));
309 atomic_inc(&(INIT_USER->__count));
310 write_unlock_irq(&tasklist_lock);
311 switch_uid(INIT_USER);
314 void __set_special_pids(pid_t session, pid_t pgrp)
316 struct task_struct *curr = current->group_leader;
318 if (curr->signal->session != session) {
319 detach_pid(curr, PIDTYPE_SID);
320 curr->signal->session = session;
321 attach_pid(curr, PIDTYPE_SID, session);
323 if (process_group(curr) != pgrp) {
324 detach_pid(curr, PIDTYPE_PGID);
325 curr->signal->pgrp = pgrp;
326 attach_pid(curr, PIDTYPE_PGID, pgrp);
330 void set_special_pids(pid_t session, pid_t pgrp)
332 write_lock_irq(&tasklist_lock);
333 __set_special_pids(session, pgrp);
334 write_unlock_irq(&tasklist_lock);
338 * Let kernel threads use this to say that they
339 * allow a certain signal (since daemonize() will
340 * have disabled all of them by default).
342 int allow_signal(int sig)
344 if (!valid_signal(sig) || sig < 1)
347 spin_lock_irq(¤t->sighand->siglock);
348 sigdelset(¤t->blocked, sig);
350 /* Kernel threads handle their own signals.
351 Let the signal code know it'll be handled, so
352 that they don't get converted to SIGKILL or
353 just silently dropped */
354 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
357 spin_unlock_irq(¤t->sighand->siglock);
361 EXPORT_SYMBOL(allow_signal);
363 int disallow_signal(int sig)
365 if (!valid_signal(sig) || sig < 1)
368 spin_lock_irq(¤t->sighand->siglock);
369 sigaddset(¤t->blocked, sig);
371 spin_unlock_irq(¤t->sighand->siglock);
375 EXPORT_SYMBOL(disallow_signal);
378 * Put all the gunge required to become a kernel thread without
379 * attached user resources in one place where it belongs.
382 void daemonize(const char *name, ...)
385 struct fs_struct *fs;
388 va_start(args, name);
389 vsnprintf(current->comm, sizeof(current->comm), name, args);
393 * If we were started as result of loading a module, close all of the
394 * user space pages. We don't need them, and if we didn't close them
395 * they would be locked into memory.
399 set_special_pids(1, 1);
400 proc_clear_tty(current);
402 /* Block and flush all signals */
403 sigfillset(&blocked);
404 sigprocmask(SIG_BLOCK, &blocked, NULL);
405 flush_signals(current);
407 /* Become as one with the init task */
409 exit_fs(current); /* current->fs->count--; */
412 atomic_inc(&fs->count);
413 exit_namespace(current);
414 current->namespace = init_task.namespace;
415 get_namespace(current->namespace);
417 current->files = init_task.files;
418 atomic_inc(¤t->files->count);
423 EXPORT_SYMBOL(daemonize);
425 static void close_files(struct files_struct * files)
433 * It is safe to dereference the fd table without RCU or
434 * ->file_lock because this is the last reference to the
437 fdt = files_fdtable(files);
441 if (i >= fdt->max_fdset || i >= fdt->max_fds)
443 set = fdt->open_fds->fds_bits[j++];
446 struct file * file = xchg(&fdt->fd[i], NULL);
448 filp_close(file, files);
458 struct files_struct *get_files_struct(struct task_struct *task)
460 struct files_struct *files;
465 atomic_inc(&files->count);
471 void fastcall put_files_struct(struct files_struct *files)
475 if (atomic_dec_and_test(&files->count)) {
478 * Free the fd and fdset arrays if we expanded them.
479 * If the fdtable was embedded, pass files for freeing
480 * at the end of the RCU grace period. Otherwise,
481 * you can free files immediately.
483 fdt = files_fdtable(files);
484 if (fdt == &files->fdtab)
485 fdt->free_files = files;
487 kmem_cache_free(files_cachep, files);
492 EXPORT_SYMBOL(put_files_struct);
494 static inline void __exit_files(struct task_struct *tsk)
496 struct files_struct * files = tsk->files;
502 put_files_struct(files);
506 void exit_files(struct task_struct *tsk)
511 static inline void __put_fs_struct(struct fs_struct *fs)
513 /* No need to hold fs->lock if we are killing it */
514 if (atomic_dec_and_test(&fs->count)) {
521 mntput(fs->altrootmnt);
523 kmem_cache_free(fs_cachep, fs);
527 void put_fs_struct(struct fs_struct *fs)
532 static inline void __exit_fs(struct task_struct *tsk)
534 struct fs_struct * fs = tsk->fs;
544 void exit_fs(struct task_struct *tsk)
549 EXPORT_SYMBOL_GPL(exit_fs);
552 * Turn us into a lazy TLB process if we
555 static void exit_mm(struct task_struct * tsk)
557 struct mm_struct *mm = tsk->mm;
563 * Serialize with any possible pending coredump.
564 * We must hold mmap_sem around checking core_waiters
565 * and clearing tsk->mm. The core-inducing thread
566 * will increment core_waiters for each thread in the
567 * group with ->mm != NULL.
569 down_read(&mm->mmap_sem);
570 if (mm->core_waiters) {
571 up_read(&mm->mmap_sem);
572 down_write(&mm->mmap_sem);
573 if (!--mm->core_waiters)
574 complete(mm->core_startup_done);
575 up_write(&mm->mmap_sem);
577 wait_for_completion(&mm->core_done);
578 down_read(&mm->mmap_sem);
580 atomic_inc(&mm->mm_count);
581 BUG_ON(mm != tsk->active_mm);
582 /* more a memory barrier than a real lock */
585 up_read(&mm->mmap_sem);
586 enter_lazy_tlb(mm, current);
592 choose_new_parent(struct task_struct *p, struct task_struct *reaper)
594 /* check for reaper context */
595 vxwprintk((p->xid != reaper->xid) && (reaper != child_reaper),
596 "rogue reaper: %p[%d,#%u] <> %p[%d,#%u]",
597 p, p->pid, p->xid, reaper, reaper->pid, reaper->xid);
600 * Make sure we're not reparenting to ourselves and that
601 * the parent is not a zombie.
603 BUG_ON(p == reaper || reaper->exit_state);
604 p->real_parent = reaper;
608 reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
610 /* We don't want people slaying init. */
611 if (p->exit_signal != -1)
612 p->exit_signal = SIGCHLD;
614 if (p->pdeath_signal)
615 /* We already hold the tasklist_lock here. */
616 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
618 /* Move the child from its dying parent to the new one. */
619 if (unlikely(traced)) {
620 /* Preserve ptrace links if someone else is tracing this child. */
621 list_del_init(&p->ptrace_list);
622 if (p->parent != p->real_parent)
623 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
625 /* If this child is being traced, then we're the one tracing it
626 * anyway, so let go of it.
630 p->parent = p->real_parent;
633 /* If we'd notified the old parent about this child's death,
634 * also notify the new parent.
636 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
637 thread_group_empty(p))
638 do_notify_parent(p, p->exit_signal);
639 else if (p->state == TASK_TRACED) {
641 * If it was at a trace stop, turn it into
642 * a normal stop since it's no longer being
650 * process group orphan check
651 * Case ii: Our child is in a different pgrp
652 * than we are, and it was the only connection
653 * outside, so the child pgrp is now orphaned.
655 if ((process_group(p) != process_group(father)) &&
656 (p->signal->session == father->signal->session)) {
657 int pgrp = process_group(p);
659 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
660 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
661 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
667 * When we die, we re-parent all our children.
668 * Try to give them to another thread in our thread
669 * group, and if no such member exists, give it to
670 * the global child reaper process (ie "init")
673 forget_original_parent(struct task_struct *father, struct list_head *to_release)
675 struct task_struct *p, *reaper = father;
676 struct list_head *_p, *_n;
679 reaper = next_thread(reaper);
680 if (reaper == father) {
681 reaper = vx_child_reaper(father);
684 } while (reaper->exit_state);
687 * There are only two places where our children can be:
689 * - in our child list
690 * - in our ptraced child list
692 * Search them and reparent children.
694 list_for_each_safe(_p, _n, &father->children) {
696 p = list_entry(_p, struct task_struct, sibling);
700 /* if father isn't the real parent, then ptrace must be enabled */
701 BUG_ON(father != p->real_parent && !ptrace);
703 if (father == p->real_parent) {
704 /* reparent with a reaper, real father it's us */
705 choose_new_parent(p, vx_child_reaper(p));
706 reparent_thread(p, father, 0);
708 /* reparent ptraced task to its real parent */
710 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
711 thread_group_empty(p))
712 do_notify_parent(p, p->exit_signal);
716 * if the ptraced child is a zombie with exit_signal == -1
717 * we must collect it before we exit, or it will remain
718 * zombie forever since we prevented it from self-reap itself
719 * while it was being traced by us, to be able to see it in wait4.
721 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
722 list_add(&p->ptrace_list, to_release);
724 list_for_each_safe(_p, _n, &father->ptrace_children) {
725 p = list_entry(_p, struct task_struct, ptrace_list);
726 choose_new_parent(p, reaper);
727 reparent_thread(p, father, 1);
732 * Send signals to all our closest relatives so that they know
733 * to properly mourn us..
735 static void exit_notify(struct task_struct *tsk)
738 struct task_struct *t;
739 struct list_head ptrace_dead, *_p, *_n;
741 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
742 && !thread_group_empty(tsk)) {
744 * This occurs when there was a race between our exit
745 * syscall and a group signal choosing us as the one to
746 * wake up. It could be that we are the only thread
747 * alerted to check for pending signals, but another thread
748 * should be woken now to take the signal since we will not.
749 * Now we'll wake all the threads in the group just to make
750 * sure someone gets all the pending signals.
752 read_lock(&tasklist_lock);
753 spin_lock_irq(&tsk->sighand->siglock);
754 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
755 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
756 recalc_sigpending_tsk(t);
757 if (signal_pending(t))
758 signal_wake_up(t, 0);
760 spin_unlock_irq(&tsk->sighand->siglock);
761 read_unlock(&tasklist_lock);
764 write_lock_irq(&tasklist_lock);
767 * This does two things:
769 * A. Make init inherit all the child processes
770 * B. Check to see if any process groups have become orphaned
771 * as a result of our exiting, and if they have any stopped
772 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
775 INIT_LIST_HEAD(&ptrace_dead);
776 forget_original_parent(tsk, &ptrace_dead);
777 BUG_ON(!list_empty(&tsk->children));
778 BUG_ON(!list_empty(&tsk->ptrace_children));
781 * Check to see if any process groups have become orphaned
782 * as a result of our exiting, and if they have any stopped
783 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
785 * Case i: Our father is in a different pgrp than we are
786 * and we were the only connection outside, so our pgrp
787 * is about to become orphaned.
790 t = tsk->real_parent;
792 if ((process_group(t) != process_group(tsk)) &&
793 (t->signal->session == tsk->signal->session) &&
794 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
795 has_stopped_jobs(process_group(tsk))) {
796 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
797 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
800 /* Let father know we died
802 * Thread signals are configurable, but you aren't going to use
803 * that to send signals to arbitary processes.
804 * That stops right now.
806 * If the parent exec id doesn't match the exec id we saved
807 * when we started then we know the parent has changed security
810 * If our self_exec id doesn't match our parent_exec_id then
811 * we have changed execution domain as these two values started
812 * the same after a fork.
816 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
817 ( tsk->parent_exec_id != t->self_exec_id ||
818 tsk->self_exec_id != tsk->parent_exec_id)
819 && !capable(CAP_KILL))
820 tsk->exit_signal = SIGCHLD;
823 /* If something other than our normal parent is ptracing us, then
824 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
825 * only has special meaning to our real parent.
827 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
828 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
829 do_notify_parent(tsk, signal);
830 } else if (tsk->ptrace) {
831 do_notify_parent(tsk, SIGCHLD);
835 if (tsk->exit_signal == -1 &&
836 (likely(tsk->ptrace == 0) ||
837 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
839 tsk->exit_state = state;
841 write_unlock_irq(&tasklist_lock);
843 list_for_each_safe(_p, _n, &ptrace_dead) {
845 t = list_entry(_p, struct task_struct, ptrace_list);
849 /* If the process is dead, release it - nobody will wait for it */
850 if (state == EXIT_DEAD)
854 fastcall NORET_TYPE void do_exit(long code)
856 struct task_struct *tsk = current;
857 struct taskstats *tidstats;
861 profile_task_exit(tsk);
863 WARN_ON(atomic_read(&tsk->fs_excl));
865 if (unlikely(in_interrupt()))
866 panic("Aiee, killing interrupt handler!");
867 if (unlikely(!tsk->pid))
868 panic("Attempted to kill the idle task!");
869 if (unlikely(tsk == child_reaper))
870 panic("Attempted to kill init!");
872 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
873 current->ptrace_message = code;
874 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
878 * We're taking recursive faults here in do_exit. Safest is to just
879 * leave this task alone and wait for reboot.
881 if (unlikely(tsk->flags & PF_EXITING)) {
883 "Fixing recursive fault but reboot is needed!\n");
886 set_current_state(TASK_UNINTERRUPTIBLE);
890 tsk->flags |= PF_EXITING;
892 if (unlikely(in_atomic()))
893 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
894 current->comm, current->pid,
897 taskstats_exit_alloc(&tidstats, &mycpu);
899 acct_update_integrals(tsk);
901 update_hiwater_rss(tsk->mm);
902 update_hiwater_vm(tsk->mm);
904 group_dead = atomic_dec_and_test(&tsk->signal->live);
906 hrtimer_cancel(&tsk->signal->real_timer);
907 exit_itimers(tsk->signal);
910 if (current->tux_info) {
911 #ifdef CONFIG_TUX_DEBUG
912 printk("Possibly unexpected TUX-thread exit(%ld) at %p?\n",
913 code, __builtin_return_address(0));
918 acct_collect(code, group_dead);
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))
927 taskstats_exit_send(tsk, tidstats, group_dead, mycpu);
928 taskstats_exit_free(tidstats);
942 if (group_dead && tsk->signal->leader)
943 disassociate_ctty(1);
945 module_put(task_thread_info(tsk)->exec_domain->module);
947 module_put(tsk->binfmt->module);
949 tsk->exit_code = code;
950 proc_exit_connector(tsk);
951 /* needs to stay before exit_notify() */
952 exit_vx_info_early(tsk, code);
955 mpol_free(tsk->mempolicy);
956 tsk->mempolicy = NULL;
959 * This must happen late, after the PID is not
962 if (unlikely(!list_empty(&tsk->pi_state_list)))
963 exit_pi_state_list(tsk);
964 if (unlikely(current->pi_state_cache))
965 kfree(current->pi_state_cache);
967 * Make sure we are holding no locks:
969 debug_check_no_locks_held(tsk);
974 if (tsk->splice_pipe)
975 __free_pipe_info(tsk->splice_pipe);
977 /* needs to stay after exit_notify() */
978 exit_vx_info(tsk, code);
981 /* PF_DEAD causes final put_task_struct after we schedule. */
983 BUG_ON(tsk->flags & PF_DEAD);
984 tsk->flags |= PF_DEAD;
988 /* Avoid "noreturn function does return". */
992 EXPORT_SYMBOL_GPL(do_exit);
994 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1002 EXPORT_SYMBOL(complete_and_exit);
1004 asmlinkage long sys_exit(int error_code)
1006 do_exit((error_code&0xff)<<8);
1010 * Take down every thread in the group. This is called by fatal signals
1011 * as well as by sys_exit_group (below).
1014 do_group_exit(int exit_code)
1016 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1018 if (current->signal->flags & SIGNAL_GROUP_EXIT)
1019 exit_code = current->signal->group_exit_code;
1020 else if (!thread_group_empty(current)) {
1021 struct signal_struct *const sig = current->signal;
1022 struct sighand_struct *const sighand = current->sighand;
1023 spin_lock_irq(&sighand->siglock);
1024 if (sig->flags & SIGNAL_GROUP_EXIT)
1025 /* Another thread got here before we took the lock. */
1026 exit_code = sig->group_exit_code;
1028 sig->group_exit_code = exit_code;
1029 zap_other_threads(current);
1031 spin_unlock_irq(&sighand->siglock);
1039 * this kills every thread in the thread group. Note that any externally
1040 * wait4()-ing process will get the correct exit code - even if this
1041 * thread is not the thread group leader.
1043 asmlinkage void sys_exit_group(int error_code)
1045 do_group_exit((error_code & 0xff) << 8);
1048 static int eligible_child(pid_t pid, int options, struct task_struct *p)
1054 if (process_group(p) != process_group(current))
1056 } else if (pid != -1) {
1057 if (process_group(p) != -pid)
1062 * Do not consider detached threads that are
1065 if (p->exit_signal == -1 && !p->ptrace)
1068 /* Wait for all children (clone and not) if __WALL is set;
1069 * otherwise, wait for clone children *only* if __WCLONE is
1070 * set; otherwise, wait for non-clone children *only*. (Note:
1071 * A "clone" child here is one that reports to its parent
1072 * using a signal other than SIGCHLD.) */
1073 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1074 && !(options & __WALL))
1077 * Do not consider thread group leaders that are
1078 * in a non-empty thread group:
1080 if (delay_group_leader(p))
1083 if (security_task_wait(p))
1089 static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
1090 int why, int status,
1091 struct siginfo __user *infop,
1092 struct rusage __user *rusagep)
1094 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1098 retval = put_user(SIGCHLD, &infop->si_signo);
1100 retval = put_user(0, &infop->si_errno);
1102 retval = put_user((short)why, &infop->si_code);
1104 retval = put_user(pid, &infop->si_pid);
1106 retval = put_user(uid, &infop->si_uid);
1108 retval = put_user(status, &infop->si_status);
1115 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1116 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1117 * the lock and this task is uninteresting. If we return nonzero, we have
1118 * released the lock and the system call should return.
1120 static int wait_task_zombie(struct task_struct *p, int noreap,
1121 struct siginfo __user *infop,
1122 int __user *stat_addr, struct rusage __user *ru)
1124 unsigned long state;
1128 if (unlikely(noreap)) {
1131 int exit_code = p->exit_code;
1134 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1136 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1139 read_unlock(&tasklist_lock);
1140 if ((exit_code & 0x7f) == 0) {
1142 status = exit_code >> 8;
1144 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1145 status = exit_code & 0x7f;
1147 return wait_noreap_copyout(p, pid, uid, why,
1152 * Try to move the task's state to DEAD
1153 * only one thread is allowed to do this:
1155 state = xchg(&p->exit_state, EXIT_DEAD);
1156 if (state != EXIT_ZOMBIE) {
1157 BUG_ON(state != EXIT_DEAD);
1160 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1162 * This can only happen in a race with a ptraced thread
1163 * dying on another processor.
1168 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1169 struct signal_struct *psig;
1170 struct signal_struct *sig;
1173 * The resource counters for the group leader are in its
1174 * own task_struct. Those for dead threads in the group
1175 * are in its signal_struct, as are those for the child
1176 * processes it has previously reaped. All these
1177 * accumulate in the parent's signal_struct c* fields.
1179 * We don't bother to take a lock here to protect these
1180 * p->signal fields, because they are only touched by
1181 * __exit_signal, which runs with tasklist_lock
1182 * write-locked anyway, and so is excluded here. We do
1183 * need to protect the access to p->parent->signal fields,
1184 * as other threads in the parent group can be right
1185 * here reaping other children at the same time.
1187 spin_lock_irq(&p->parent->sighand->siglock);
1188 psig = p->parent->signal;
1191 cputime_add(psig->cutime,
1192 cputime_add(p->utime,
1193 cputime_add(sig->utime,
1196 cputime_add(psig->cstime,
1197 cputime_add(p->stime,
1198 cputime_add(sig->stime,
1201 p->min_flt + sig->min_flt + sig->cmin_flt;
1203 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1205 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1207 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1208 spin_unlock_irq(&p->parent->sighand->siglock);
1212 * Now we are sure this task is interesting, and no other
1213 * thread can reap it because we set its state to EXIT_DEAD.
1215 read_unlock(&tasklist_lock);
1217 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1218 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1219 ? p->signal->group_exit_code : p->exit_code;
1220 if (!retval && stat_addr)
1221 retval = put_user(status, stat_addr);
1222 if (!retval && infop)
1223 retval = put_user(SIGCHLD, &infop->si_signo);
1224 if (!retval && infop)
1225 retval = put_user(0, &infop->si_errno);
1226 if (!retval && infop) {
1229 if ((status & 0x7f) == 0) {
1233 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1236 retval = put_user((short)why, &infop->si_code);
1238 retval = put_user(status, &infop->si_status);
1240 if (!retval && infop)
1241 retval = put_user(p->pid, &infop->si_pid);
1242 if (!retval && infop)
1243 retval = put_user(p->uid, &infop->si_uid);
1245 // TODO: is this safe?
1246 p->exit_state = EXIT_ZOMBIE;
1250 if (p->real_parent != p->parent) {
1251 write_lock_irq(&tasklist_lock);
1252 /* Double-check with lock held. */
1253 if (p->real_parent != p->parent) {
1255 // TODO: is this safe?
1256 p->exit_state = EXIT_ZOMBIE;
1258 * If this is not a detached task, notify the parent.
1259 * If it's still not detached after that, don't release
1262 if (p->exit_signal != -1) {
1263 do_notify_parent(p, p->exit_signal);
1264 if (p->exit_signal != -1)
1268 write_unlock_irq(&tasklist_lock);
1277 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1278 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1279 * the lock and this task is uninteresting. If we return nonzero, we have
1280 * released the lock and the system call should return.
1282 static int wait_task_stopped(struct task_struct *p, int delayed_group_leader,
1283 int noreap, struct siginfo __user *infop,
1284 int __user *stat_addr, struct rusage __user *ru)
1286 int retval, exit_code;
1290 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1291 p->signal && p->signal->group_stop_count > 0)
1293 * A group stop is in progress and this is the group leader.
1294 * We won't report until all threads have stopped.
1299 * Now we are pretty sure this task is interesting.
1300 * Make sure it doesn't get reaped out from under us while we
1301 * give up the lock and then examine it below. We don't want to
1302 * keep holding onto the tasklist_lock while we call getrusage and
1303 * possibly take page faults for user memory.
1306 read_unlock(&tasklist_lock);
1308 if (unlikely(noreap)) {
1311 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1313 exit_code = p->exit_code;
1314 if (unlikely(!exit_code) ||
1315 unlikely(p->state & TASK_TRACED))
1317 return wait_noreap_copyout(p, pid, uid,
1318 why, (exit_code << 8) | 0x7f,
1322 write_lock_irq(&tasklist_lock);
1325 * This uses xchg to be atomic with the thread resuming and setting
1326 * it. It must also be done with the write lock held to prevent a
1327 * race with the EXIT_ZOMBIE case.
1329 exit_code = xchg(&p->exit_code, 0);
1330 if (unlikely(p->exit_state)) {
1332 * The task resumed and then died. Let the next iteration
1333 * catch it in EXIT_ZOMBIE. Note that exit_code might
1334 * already be zero here if it resumed and did _exit(0).
1335 * The task itself is dead and won't touch exit_code again;
1336 * other processors in this function are locked out.
1338 p->exit_code = exit_code;
1341 if (unlikely(exit_code == 0)) {
1343 * Another thread in this function got to it first, or it
1344 * resumed, or it resumed and then died.
1346 write_unlock_irq(&tasklist_lock);
1350 * We are returning to the wait loop without having successfully
1351 * removed the process and having released the lock. We cannot
1352 * continue, since the "p" task pointer is potentially stale.
1354 * Return -EAGAIN, and do_wait() will restart the loop from the
1355 * beginning. Do _not_ re-acquire the lock.
1360 /* move to end of parent's list to avoid starvation */
1364 write_unlock_irq(&tasklist_lock);
1366 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1367 if (!retval && stat_addr)
1368 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1369 if (!retval && infop)
1370 retval = put_user(SIGCHLD, &infop->si_signo);
1371 if (!retval && infop)
1372 retval = put_user(0, &infop->si_errno);
1373 if (!retval && infop)
1374 retval = put_user((short)((p->ptrace & PT_PTRACED)
1375 ? CLD_TRAPPED : CLD_STOPPED),
1377 if (!retval && infop)
1378 retval = put_user(exit_code, &infop->si_status);
1379 if (!retval && infop)
1380 retval = put_user(p->pid, &infop->si_pid);
1381 if (!retval && infop)
1382 retval = put_user(p->uid, &infop->si_uid);
1392 * Handle do_wait work for one task in a live, non-stopped state.
1393 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1394 * the lock and this task is uninteresting. If we return nonzero, we have
1395 * released the lock and the system call should return.
1397 static int wait_task_continued(struct task_struct *p, int noreap,
1398 struct siginfo __user *infop,
1399 int __user *stat_addr, struct rusage __user *ru)
1405 if (unlikely(!p->signal))
1408 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1411 spin_lock_irq(&p->sighand->siglock);
1412 /* Re-check with the lock held. */
1413 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1414 spin_unlock_irq(&p->sighand->siglock);
1418 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1419 spin_unlock_irq(&p->sighand->siglock);
1424 read_unlock(&tasklist_lock);
1427 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1429 if (!retval && stat_addr)
1430 retval = put_user(0xffff, stat_addr);
1434 retval = wait_noreap_copyout(p, pid, uid,
1435 CLD_CONTINUED, SIGCONT,
1437 BUG_ON(retval == 0);
1444 static inline int my_ptrace_child(struct task_struct *p)
1446 if (!(p->ptrace & PT_PTRACED))
1448 if (!(p->ptrace & PT_ATTACHED))
1451 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1452 * we are the attacher. If we are the real parent, this is a race
1453 * inside ptrace_attach. It is waiting for the tasklist_lock,
1454 * which we have to switch the parent links, but has already set
1455 * the flags in p->ptrace.
1457 return (p->parent != p->real_parent);
1460 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1461 int __user *stat_addr, struct rusage __user *ru)
1463 DECLARE_WAITQUEUE(wait, current);
1464 struct task_struct *tsk;
1467 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1470 * We will set this flag if we see any child that might later
1471 * match our criteria, even if we are not able to reap it yet.
1474 current->state = TASK_INTERRUPTIBLE;
1475 read_lock(&tasklist_lock);
1478 struct task_struct *p;
1479 struct list_head *_p;
1482 list_for_each(_p,&tsk->children) {
1483 p = list_entry(_p, struct task_struct, sibling);
1485 ret = eligible_child(pid, options, p);
1492 * When we hit the race with PTRACE_ATTACH,
1493 * we will not report this child. But the
1494 * race means it has not yet been moved to
1495 * our ptrace_children list, so we need to
1496 * set the flag here to avoid a spurious ECHILD
1497 * when the race happens with the only child.
1500 if (!my_ptrace_child(p))
1505 * It's stopped now, so it might later
1506 * continue, exit, or stop again.
1509 if (!(options & WUNTRACED) &&
1510 !my_ptrace_child(p))
1512 retval = wait_task_stopped(p, ret == 2,
1513 (options & WNOWAIT),
1516 if (retval == -EAGAIN)
1518 if (retval != 0) /* He released the lock. */
1523 if (p->exit_state == EXIT_DEAD)
1525 // case EXIT_ZOMBIE:
1526 if (p->exit_state == EXIT_ZOMBIE) {
1528 * Eligible but we cannot release
1532 goto check_continued;
1533 if (!likely(options & WEXITED))
1535 retval = wait_task_zombie(
1536 p, (options & WNOWAIT),
1537 infop, stat_addr, ru);
1538 /* He released the lock. */
1545 * It's running now, so it might later
1546 * exit, stop, or stop and then continue.
1549 if (!unlikely(options & WCONTINUED))
1551 retval = wait_task_continued(
1552 p, (options & WNOWAIT),
1553 infop, stat_addr, ru);
1554 if (retval != 0) /* He released the lock. */
1560 list_for_each(_p, &tsk->ptrace_children) {
1561 p = list_entry(_p, struct task_struct,
1563 if (!eligible_child(pid, options, p))
1569 if (options & __WNOTHREAD)
1571 tsk = next_thread(tsk);
1572 BUG_ON(tsk->signal != current->signal);
1573 } while (tsk != current);
1575 read_unlock(&tasklist_lock);
1578 if (options & WNOHANG)
1580 retval = -ERESTARTSYS;
1581 if (signal_pending(current))
1588 current->state = TASK_RUNNING;
1589 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1595 * For a WNOHANG return, clear out all the fields
1596 * we would set so the user can easily tell the
1600 retval = put_user(0, &infop->si_signo);
1602 retval = put_user(0, &infop->si_errno);
1604 retval = put_user(0, &infop->si_code);
1606 retval = put_user(0, &infop->si_pid);
1608 retval = put_user(0, &infop->si_uid);
1610 retval = put_user(0, &infop->si_status);
1616 asmlinkage long sys_waitid(int which, pid_t pid,
1617 struct siginfo __user *infop, int options,
1618 struct rusage __user *ru)
1622 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1624 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1644 ret = do_wait(pid, options, infop, NULL, ru);
1646 /* avoid REGPARM breakage on x86: */
1647 prevent_tail_call(ret);
1651 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1652 int options, struct rusage __user *ru)
1656 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1657 __WNOTHREAD|__WCLONE|__WALL))
1659 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1661 /* avoid REGPARM breakage on x86: */
1662 prevent_tail_call(ret);
1666 #ifdef __ARCH_WANT_SYS_WAITPID
1669 * sys_waitpid() remains for compatibility. waitpid() should be
1670 * implemented by calling sys_wait4() from libc.a.
1672 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1674 return sys_wait4(pid, stat_addr, options, NULL);