4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/config.h>
9 #include <linux/slab.h>
10 #include <linux/interrupt.h>
11 #include <linux/smp_lock.h>
12 #include <linux/module.h>
13 #include <linux/capability.h>
14 #include <linux/completion.h>
15 #include <linux/personality.h>
16 #include <linux/tty.h>
17 #include <linux/namespace.h>
18 #include <linux/key.h>
19 #include <linux/security.h>
20 #include <linux/cpu.h>
21 #include <linux/acct.h>
22 #include <linux/file.h>
23 #include <linux/binfmts.h>
24 #include <linux/ptrace.h>
25 #include <linux/profile.h>
26 #include <linux/mount.h>
27 #include <linux/proc_fs.h>
28 #include <linux/mempolicy.h>
29 #include <linux/cpuset.h>
30 #include <linux/syscalls.h>
31 #include <linux/signal.h>
32 #include <linux/posix-timers.h>
33 #include <linux/cn_proc.h>
34 #include <linux/mutex.h>
35 #include <linux/futex.h>
36 #include <linux/compat.h>
37 #include <linux/pipe_fs_i.h>
39 #include <asm/uaccess.h>
40 #include <asm/unistd.h>
41 #include <asm/pgtable.h>
42 #include <asm/mmu_context.h>
44 extern void sem_exit (void);
45 extern struct task_struct *child_reaper;
47 int getrusage(struct task_struct *, int, struct rusage __user *);
49 static void exit_mm(struct task_struct * tsk);
51 static void __unhash_process(struct task_struct *p)
54 detach_pid(p, PIDTYPE_PID);
55 if (thread_group_leader(p)) {
56 detach_pid(p, PIDTYPE_PGID);
57 detach_pid(p, PIDTYPE_SID);
59 list_del_rcu(&p->tasks);
60 __get_cpu_var(process_counts)--;
62 list_del_rcu(&p->thread_group);
67 * This function expects the tasklist_lock write-locked.
69 static void __exit_signal(struct task_struct *tsk)
71 struct signal_struct *sig = tsk->signal;
72 struct sighand_struct *sighand;
75 BUG_ON(!atomic_read(&sig->count));
78 sighand = rcu_dereference(tsk->sighand);
79 spin_lock(&sighand->siglock);
81 posix_cpu_timers_exit(tsk);
82 if (atomic_dec_and_test(&sig->count))
83 posix_cpu_timers_exit_group(tsk);
86 * If there is any task waiting for the group exit
89 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
90 wake_up_process(sig->group_exit_task);
91 sig->group_exit_task = NULL;
93 if (tsk == sig->curr_target)
94 sig->curr_target = next_thread(tsk);
96 * Accumulate here the counters for all threads but the
97 * group leader as they die, so they can be added into
98 * the process-wide totals when those are taken.
99 * The group leader stays around as a zombie as long
100 * as there are other threads. When it gets reaped,
101 * the exit.c code will add its counts into these totals.
102 * We won't ever get here for the group leader, since it
103 * will have been the last reference on the signal_struct.
105 sig->utime = cputime_add(sig->utime, tsk->utime);
106 sig->stime = cputime_add(sig->stime, tsk->stime);
107 sig->min_flt += tsk->min_flt;
108 sig->maj_flt += tsk->maj_flt;
109 sig->nvcsw += tsk->nvcsw;
110 sig->nivcsw += tsk->nivcsw;
111 sig->sched_time += tsk->sched_time;
112 sig = NULL; /* Marker for below. */
115 __unhash_process(tsk);
119 spin_unlock(&sighand->siglock);
122 __cleanup_sighand(sighand);
123 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
124 flush_sigqueue(&tsk->pending);
126 flush_sigqueue(&sig->shared_pending);
127 __cleanup_signal(sig);
131 static void delayed_put_task_struct(struct rcu_head *rhp)
133 put_task_struct(container_of(rhp, struct task_struct, rcu));
136 void release_task(struct task_struct * p)
140 struct dentry *proc_dentry;
143 atomic_dec(&p->user->processes);
144 spin_lock(&p->proc_lock);
145 proc_dentry = proc_pid_unhash(p);
146 write_lock_irq(&tasklist_lock);
148 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
152 * If we are the last non-leader member of the thread
153 * group, and the leader is zombie, then notify the
154 * group leader's parent process. (if it wants notification.)
157 leader = p->group_leader;
158 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
159 BUG_ON(leader->exit_signal == -1);
160 do_notify_parent(leader, leader->exit_signal);
162 * If we were the last child thread and the leader has
163 * exited already, and the leader's parent ignores SIGCHLD,
164 * then we are the one who should release the leader.
166 * do_notify_parent() will have marked it self-reaping in
169 zap_leader = (leader->exit_signal == -1);
173 write_unlock_irq(&tasklist_lock);
174 spin_unlock(&p->proc_lock);
175 proc_pid_flush(proc_dentry);
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, task_t *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 mutex_lock(&tty_mutex);
401 current->signal->tty = NULL;
402 mutex_unlock(&tty_mutex);
404 /* Block and flush all signals */
405 sigfillset(&blocked);
406 sigprocmask(SIG_BLOCK, &blocked, NULL);
407 flush_signals(current);
409 /* Become as one with the init task */
411 exit_fs(current); /* current->fs->count--; */
414 atomic_inc(&fs->count);
415 exit_namespace(current);
416 current->namespace = init_task.namespace;
417 get_namespace(current->namespace);
419 current->files = init_task.files;
420 atomic_inc(¤t->files->count);
425 EXPORT_SYMBOL(daemonize);
427 static void close_files(struct files_struct * files)
435 * It is safe to dereference the fd table without RCU or
436 * ->file_lock because this is the last reference to the
439 fdt = files_fdtable(files);
443 if (i >= fdt->max_fdset || i >= fdt->max_fds)
445 set = fdt->open_fds->fds_bits[j++];
448 struct file * file = xchg(&fdt->fd[i], NULL);
450 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 if (mm != tsk->active_mm) BUG();
582 /* more a memory barrier than a real lock */
585 up_read(&mm->mmap_sem);
586 enter_lazy_tlb(mm, current);
591 static inline void choose_new_parent(task_t *p, task_t *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;
601 static void reparent_thread(task_t *p, task_t *father, int traced)
603 /* We don't want people slaying init. */
604 if (p->exit_signal != -1)
605 p->exit_signal = SIGCHLD;
607 if (p->pdeath_signal)
608 /* We already hold the tasklist_lock here. */
609 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
611 /* Move the child from its dying parent to the new one. */
612 if (unlikely(traced)) {
613 /* Preserve ptrace links if someone else is tracing this child. */
614 list_del_init(&p->ptrace_list);
615 if (p->parent != p->real_parent)
616 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
618 /* If this child is being traced, then we're the one tracing it
619 * anyway, so let go of it.
623 p->parent = p->real_parent;
626 /* If we'd notified the old parent about this child's death,
627 * also notify the new parent.
629 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
630 thread_group_empty(p))
631 do_notify_parent(p, p->exit_signal);
632 else if (p->state == TASK_TRACED) {
634 * If it was at a trace stop, turn it into
635 * a normal stop since it's no longer being
643 * process group orphan check
644 * Case ii: Our child is in a different pgrp
645 * than we are, and it was the only connection
646 * outside, so the child pgrp is now orphaned.
648 if ((process_group(p) != process_group(father)) &&
649 (p->signal->session == father->signal->session)) {
650 int pgrp = process_group(p);
652 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
653 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
654 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
660 * When we die, we re-parent all our children.
661 * Try to give them to another thread in our thread
662 * group, and if no such member exists, give it to
663 * the global child reaper process (ie "init")
665 static void forget_original_parent(struct task_struct * father,
666 struct list_head *to_release)
668 struct task_struct *p, *reaper = father;
669 struct list_head *_p, *_n;
672 reaper = next_thread(reaper);
673 if (reaper == father) {
674 reaper = child_reaper;
677 } while (reaper->exit_state);
680 * There are only two places where our children can be:
682 * - in our child list
683 * - in our ptraced child list
685 * Search them and reparent children.
687 list_for_each_safe(_p, _n, &father->children) {
689 p = list_entry(_p,struct task_struct,sibling);
693 /* if father isn't the real parent, then ptrace must be enabled */
694 BUG_ON(father != p->real_parent && !ptrace);
696 if (father == p->real_parent) {
697 /* reparent with a reaper, real father it's us */
698 choose_new_parent(p, reaper);
699 reparent_thread(p, father, 0);
701 /* reparent ptraced task to its real parent */
703 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
704 thread_group_empty(p))
705 do_notify_parent(p, p->exit_signal);
709 * if the ptraced child is a zombie with exit_signal == -1
710 * we must collect it before we exit, or it will remain
711 * zombie forever since we prevented it from self-reap itself
712 * while it was being traced by us, to be able to see it in wait4.
714 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
715 list_add(&p->ptrace_list, to_release);
717 list_for_each_safe(_p, _n, &father->ptrace_children) {
718 p = list_entry(_p,struct task_struct,ptrace_list);
719 choose_new_parent(p, reaper);
720 reparent_thread(p, father, 1);
725 * Send signals to all our closest relatives so that they know
726 * to properly mourn us..
728 static void exit_notify(struct task_struct *tsk)
731 struct task_struct *t;
732 struct list_head ptrace_dead, *_p, *_n;
734 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
735 && !thread_group_empty(tsk)) {
737 * This occurs when there was a race between our exit
738 * syscall and a group signal choosing us as the one to
739 * wake up. It could be that we are the only thread
740 * alerted to check for pending signals, but another thread
741 * should be woken now to take the signal since we will not.
742 * Now we'll wake all the threads in the group just to make
743 * sure someone gets all the pending signals.
745 read_lock(&tasklist_lock);
746 spin_lock_irq(&tsk->sighand->siglock);
747 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
748 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
749 recalc_sigpending_tsk(t);
750 if (signal_pending(t))
751 signal_wake_up(t, 0);
753 spin_unlock_irq(&tsk->sighand->siglock);
754 read_unlock(&tasklist_lock);
757 write_lock_irq(&tasklist_lock);
760 * This does two things:
762 * A. Make init inherit all the child processes
763 * B. Check to see if any process groups have become orphaned
764 * as a result of our exiting, and if they have any stopped
765 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
768 INIT_LIST_HEAD(&ptrace_dead);
769 forget_original_parent(tsk, &ptrace_dead);
770 BUG_ON(!list_empty(&tsk->children));
771 BUG_ON(!list_empty(&tsk->ptrace_children));
774 * Check to see if any process groups have become orphaned
775 * as a result of our exiting, and if they have any stopped
776 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
778 * Case i: Our father is in a different pgrp than we are
779 * and we were the only connection outside, so our pgrp
780 * is about to become orphaned.
783 t = tsk->real_parent;
785 if ((process_group(t) != process_group(tsk)) &&
786 (t->signal->session == tsk->signal->session) &&
787 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
788 has_stopped_jobs(process_group(tsk))) {
789 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
790 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
793 /* Let father know we died
795 * Thread signals are configurable, but you aren't going to use
796 * that to send signals to arbitary processes.
797 * That stops right now.
799 * If the parent exec id doesn't match the exec id we saved
800 * when we started then we know the parent has changed security
803 * If our self_exec id doesn't match our parent_exec_id then
804 * we have changed execution domain as these two values started
805 * the same after a fork.
809 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
810 ( tsk->parent_exec_id != t->self_exec_id ||
811 tsk->self_exec_id != tsk->parent_exec_id)
812 && !capable(CAP_KILL))
813 tsk->exit_signal = SIGCHLD;
816 /* If something other than our normal parent is ptracing us, then
817 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
818 * only has special meaning to our real parent.
820 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
821 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
822 do_notify_parent(tsk, signal);
823 } else if (tsk->ptrace) {
824 do_notify_parent(tsk, SIGCHLD);
828 if (tsk->exit_signal == -1 &&
829 (likely(tsk->ptrace == 0) ||
830 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
832 tsk->exit_state = state;
834 write_unlock_irq(&tasklist_lock);
836 list_for_each_safe(_p, _n, &ptrace_dead) {
838 t = list_entry(_p,struct task_struct,ptrace_list);
842 /* If the process is dead, release it - nobody will wait for it */
843 if (state == EXIT_DEAD)
847 fastcall NORET_TYPE void do_exit(long code)
849 struct task_struct *tsk = current;
852 profile_task_exit(tsk);
854 WARN_ON(atomic_read(&tsk->fs_excl));
856 if (unlikely(in_interrupt()))
857 panic("Aiee, killing interrupt handler!");
858 if (unlikely(!tsk->pid))
859 panic("Attempted to kill the idle task!");
860 if (unlikely(tsk == child_reaper))
861 panic("Attempted to kill init!");
863 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
864 current->ptrace_message = code;
865 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
869 * We're taking recursive faults here in do_exit. Safest is to just
870 * leave this task alone and wait for reboot.
872 if (unlikely(tsk->flags & PF_EXITING)) {
874 "Fixing recursive fault but reboot is needed!\n");
877 set_current_state(TASK_UNINTERRUPTIBLE);
881 tsk->flags |= PF_EXITING;
884 * Make sure we don't try to process any timer firings
885 * while we are already exiting.
887 tsk->it_virt_expires = cputime_zero;
888 tsk->it_prof_expires = cputime_zero;
889 tsk->it_sched_expires = 0;
891 if (unlikely(in_atomic()))
892 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
893 current->comm, current->pid,
896 acct_update_integrals(tsk);
898 update_hiwater_rss(tsk->mm);
899 update_hiwater_vm(tsk->mm);
901 group_dead = atomic_dec_and_test(&tsk->signal->live);
903 hrtimer_cancel(&tsk->signal->real_timer);
904 exit_itimers(tsk->signal);
907 if (unlikely(tsk->robust_list))
908 exit_robust_list(tsk);
910 if (unlikely(tsk->compat_robust_list))
911 compat_exit_robust_list(tsk);
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);
934 mpol_free(tsk->mempolicy);
935 tsk->mempolicy = NULL;
938 * If DEBUG_MUTEXES is on, make sure we are holding no locks:
940 mutex_debug_check_no_locks_held(tsk);
945 if (tsk->splice_pipe)
946 __free_pipe_info(tsk->splice_pipe);
948 /* PF_DEAD causes final put_task_struct after we schedule. */
950 BUG_ON(tsk->flags & PF_DEAD);
951 tsk->flags |= PF_DEAD;
955 /* Avoid "noreturn function does return". */
959 EXPORT_SYMBOL_GPL(do_exit);
961 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
969 EXPORT_SYMBOL(complete_and_exit);
971 asmlinkage long sys_exit(int error_code)
973 do_exit((error_code&0xff)<<8);
977 * Take down every thread in the group. This is called by fatal signals
978 * as well as by sys_exit_group (below).
981 do_group_exit(int exit_code)
983 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
985 if (current->signal->flags & SIGNAL_GROUP_EXIT)
986 exit_code = current->signal->group_exit_code;
987 else if (!thread_group_empty(current)) {
988 struct signal_struct *const sig = current->signal;
989 struct sighand_struct *const sighand = current->sighand;
990 spin_lock_irq(&sighand->siglock);
991 if (sig->flags & SIGNAL_GROUP_EXIT)
992 /* Another thread got here before we took the lock. */
993 exit_code = sig->group_exit_code;
995 sig->group_exit_code = exit_code;
996 zap_other_threads(current);
998 spin_unlock_irq(&sighand->siglock);
1006 * this kills every thread in the thread group. Note that any externally
1007 * wait4()-ing process will get the correct exit code - even if this
1008 * thread is not the thread group leader.
1010 asmlinkage void sys_exit_group(int error_code)
1012 do_group_exit((error_code & 0xff) << 8);
1015 static int eligible_child(pid_t pid, int options, task_t *p)
1021 if (process_group(p) != process_group(current))
1023 } else if (pid != -1) {
1024 if (process_group(p) != -pid)
1029 * Do not consider detached threads that are
1032 if (p->exit_signal == -1 && !p->ptrace)
1035 /* Wait for all children (clone and not) if __WALL is set;
1036 * otherwise, wait for clone children *only* if __WCLONE is
1037 * set; otherwise, wait for non-clone children *only*. (Note:
1038 * A "clone" child here is one that reports to its parent
1039 * using a signal other than SIGCHLD.) */
1040 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1041 && !(options & __WALL))
1044 * Do not consider thread group leaders that are
1045 * in a non-empty thread group:
1047 if (current->tgid != p->tgid && delay_group_leader(p))
1050 if (security_task_wait(p))
1056 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
1057 int why, int status,
1058 struct siginfo __user *infop,
1059 struct rusage __user *rusagep)
1061 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1064 retval = put_user(SIGCHLD, &infop->si_signo);
1066 retval = put_user(0, &infop->si_errno);
1068 retval = put_user((short)why, &infop->si_code);
1070 retval = put_user(pid, &infop->si_pid);
1072 retval = put_user(uid, &infop->si_uid);
1074 retval = put_user(status, &infop->si_status);
1081 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1082 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1083 * the lock and this task is uninteresting. If we return nonzero, we have
1084 * released the lock and the system call should return.
1086 static int wait_task_zombie(task_t *p, int noreap,
1087 struct siginfo __user *infop,
1088 int __user *stat_addr, struct rusage __user *ru)
1090 unsigned long state;
1094 if (unlikely(noreap)) {
1097 int exit_code = p->exit_code;
1100 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1102 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1105 read_unlock(&tasklist_lock);
1106 if ((exit_code & 0x7f) == 0) {
1108 status = exit_code >> 8;
1110 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1111 status = exit_code & 0x7f;
1113 return wait_noreap_copyout(p, pid, uid, why,
1118 * Try to move the task's state to DEAD
1119 * only one thread is allowed to do this:
1121 state = xchg(&p->exit_state, EXIT_DEAD);
1122 if (state != EXIT_ZOMBIE) {
1123 BUG_ON(state != EXIT_DEAD);
1126 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1128 * This can only happen in a race with a ptraced thread
1129 * dying on another processor.
1134 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1135 struct signal_struct *psig;
1136 struct signal_struct *sig;
1139 * The resource counters for the group leader are in its
1140 * own task_struct. Those for dead threads in the group
1141 * are in its signal_struct, as are those for the child
1142 * processes it has previously reaped. All these
1143 * accumulate in the parent's signal_struct c* fields.
1145 * We don't bother to take a lock here to protect these
1146 * p->signal fields, because they are only touched by
1147 * __exit_signal, which runs with tasklist_lock
1148 * write-locked anyway, and so is excluded here. We do
1149 * need to protect the access to p->parent->signal fields,
1150 * as other threads in the parent group can be right
1151 * here reaping other children at the same time.
1153 spin_lock_irq(&p->parent->sighand->siglock);
1154 psig = p->parent->signal;
1157 cputime_add(psig->cutime,
1158 cputime_add(p->utime,
1159 cputime_add(sig->utime,
1162 cputime_add(psig->cstime,
1163 cputime_add(p->stime,
1164 cputime_add(sig->stime,
1167 p->min_flt + sig->min_flt + sig->cmin_flt;
1169 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1171 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1173 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1174 spin_unlock_irq(&p->parent->sighand->siglock);
1178 * Now we are sure this task is interesting, and no other
1179 * thread can reap it because we set its state to EXIT_DEAD.
1181 read_unlock(&tasklist_lock);
1183 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1184 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1185 ? p->signal->group_exit_code : p->exit_code;
1186 if (!retval && stat_addr)
1187 retval = put_user(status, stat_addr);
1188 if (!retval && infop)
1189 retval = put_user(SIGCHLD, &infop->si_signo);
1190 if (!retval && infop)
1191 retval = put_user(0, &infop->si_errno);
1192 if (!retval && infop) {
1195 if ((status & 0x7f) == 0) {
1199 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1202 retval = put_user((short)why, &infop->si_code);
1204 retval = put_user(status, &infop->si_status);
1206 if (!retval && infop)
1207 retval = put_user(p->pid, &infop->si_pid);
1208 if (!retval && infop)
1209 retval = put_user(p->uid, &infop->si_uid);
1211 // TODO: is this safe?
1212 p->exit_state = EXIT_ZOMBIE;
1216 if (p->real_parent != p->parent) {
1217 write_lock_irq(&tasklist_lock);
1218 /* Double-check with lock held. */
1219 if (p->real_parent != p->parent) {
1221 // TODO: is this safe?
1222 p->exit_state = EXIT_ZOMBIE;
1224 * If this is not a detached task, notify the parent.
1225 * If it's still not detached after that, don't release
1228 if (p->exit_signal != -1) {
1229 do_notify_parent(p, p->exit_signal);
1230 if (p->exit_signal != -1)
1234 write_unlock_irq(&tasklist_lock);
1243 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1244 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1245 * the lock and this task is uninteresting. If we return nonzero, we have
1246 * released the lock and the system call should return.
1248 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1249 struct siginfo __user *infop,
1250 int __user *stat_addr, struct rusage __user *ru)
1252 int retval, exit_code;
1256 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1257 p->signal && p->signal->group_stop_count > 0)
1259 * A group stop is in progress and this is the group leader.
1260 * We won't report until all threads have stopped.
1265 * Now we are pretty sure this task is interesting.
1266 * Make sure it doesn't get reaped out from under us while we
1267 * give up the lock and then examine it below. We don't want to
1268 * keep holding onto the tasklist_lock while we call getrusage and
1269 * possibly take page faults for user memory.
1272 read_unlock(&tasklist_lock);
1274 if (unlikely(noreap)) {
1277 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1279 exit_code = p->exit_code;
1280 if (unlikely(!exit_code) ||
1281 unlikely(p->state & TASK_TRACED))
1283 return wait_noreap_copyout(p, pid, uid,
1284 why, (exit_code << 8) | 0x7f,
1288 write_lock_irq(&tasklist_lock);
1291 * This uses xchg to be atomic with the thread resuming and setting
1292 * it. It must also be done with the write lock held to prevent a
1293 * race with the EXIT_ZOMBIE case.
1295 exit_code = xchg(&p->exit_code, 0);
1296 if (unlikely(p->exit_state)) {
1298 * The task resumed and then died. Let the next iteration
1299 * catch it in EXIT_ZOMBIE. Note that exit_code might
1300 * already be zero here if it resumed and did _exit(0).
1301 * The task itself is dead and won't touch exit_code again;
1302 * other processors in this function are locked out.
1304 p->exit_code = exit_code;
1307 if (unlikely(exit_code == 0)) {
1309 * Another thread in this function got to it first, or it
1310 * resumed, or it resumed and then died.
1312 write_unlock_irq(&tasklist_lock);
1316 * We are returning to the wait loop without having successfully
1317 * removed the process and having released the lock. We cannot
1318 * continue, since the "p" task pointer is potentially stale.
1320 * Return -EAGAIN, and do_wait() will restart the loop from the
1321 * beginning. Do _not_ re-acquire the lock.
1326 /* move to end of parent's list to avoid starvation */
1330 write_unlock_irq(&tasklist_lock);
1332 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1333 if (!retval && stat_addr)
1334 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1335 if (!retval && infop)
1336 retval = put_user(SIGCHLD, &infop->si_signo);
1337 if (!retval && infop)
1338 retval = put_user(0, &infop->si_errno);
1339 if (!retval && infop)
1340 retval = put_user((short)((p->ptrace & PT_PTRACED)
1341 ? CLD_TRAPPED : CLD_STOPPED),
1343 if (!retval && infop)
1344 retval = put_user(exit_code, &infop->si_status);
1345 if (!retval && infop)
1346 retval = put_user(p->pid, &infop->si_pid);
1347 if (!retval && infop)
1348 retval = put_user(p->uid, &infop->si_uid);
1358 * Handle do_wait work for one task in a live, non-stopped state.
1359 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1360 * the lock and this task is uninteresting. If we return nonzero, we have
1361 * released the lock and the system call should return.
1363 static int wait_task_continued(task_t *p, int noreap,
1364 struct siginfo __user *infop,
1365 int __user *stat_addr, struct rusage __user *ru)
1371 if (unlikely(!p->signal))
1374 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1377 spin_lock_irq(&p->sighand->siglock);
1378 /* Re-check with the lock held. */
1379 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1380 spin_unlock_irq(&p->sighand->siglock);
1384 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1385 spin_unlock_irq(&p->sighand->siglock);
1390 read_unlock(&tasklist_lock);
1393 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1395 if (!retval && stat_addr)
1396 retval = put_user(0xffff, stat_addr);
1400 retval = wait_noreap_copyout(p, pid, uid,
1401 CLD_CONTINUED, SIGCONT,
1403 BUG_ON(retval == 0);
1410 static inline int my_ptrace_child(struct task_struct *p)
1412 if (!(p->ptrace & PT_PTRACED))
1414 if (!(p->ptrace & PT_ATTACHED))
1417 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1418 * we are the attacher. If we are the real parent, this is a race
1419 * inside ptrace_attach. It is waiting for the tasklist_lock,
1420 * which we have to switch the parent links, but has already set
1421 * the flags in p->ptrace.
1423 return (p->parent != p->real_parent);
1426 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1427 int __user *stat_addr, struct rusage __user *ru)
1429 DECLARE_WAITQUEUE(wait, current);
1430 struct task_struct *tsk;
1433 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1436 * We will set this flag if we see any child that might later
1437 * match our criteria, even if we are not able to reap it yet.
1440 current->state = TASK_INTERRUPTIBLE;
1441 read_lock(&tasklist_lock);
1444 struct task_struct *p;
1445 struct list_head *_p;
1448 list_for_each(_p,&tsk->children) {
1449 p = list_entry(_p,struct task_struct,sibling);
1451 ret = eligible_child(pid, options, p);
1458 * When we hit the race with PTRACE_ATTACH,
1459 * we will not report this child. But the
1460 * race means it has not yet been moved to
1461 * our ptrace_children list, so we need to
1462 * set the flag here to avoid a spurious ECHILD
1463 * when the race happens with the only child.
1466 if (!my_ptrace_child(p))
1471 * It's stopped now, so it might later
1472 * continue, exit, or stop again.
1475 if (!(options & WUNTRACED) &&
1476 !my_ptrace_child(p))
1478 retval = wait_task_stopped(p, ret == 2,
1479 (options & WNOWAIT),
1482 if (retval == -EAGAIN)
1484 if (retval != 0) /* He released the lock. */
1489 if (p->exit_state == EXIT_DEAD)
1491 // case EXIT_ZOMBIE:
1492 if (p->exit_state == EXIT_ZOMBIE) {
1494 * Eligible but we cannot release
1498 goto check_continued;
1499 if (!likely(options & WEXITED))
1501 retval = wait_task_zombie(
1502 p, (options & WNOWAIT),
1503 infop, stat_addr, ru);
1504 /* He released the lock. */
1511 * It's running now, so it might later
1512 * exit, stop, or stop and then continue.
1515 if (!unlikely(options & WCONTINUED))
1517 retval = wait_task_continued(
1518 p, (options & WNOWAIT),
1519 infop, stat_addr, ru);
1520 if (retval != 0) /* He released the lock. */
1526 list_for_each(_p, &tsk->ptrace_children) {
1527 p = list_entry(_p, struct task_struct,
1529 if (!eligible_child(pid, options, p))
1535 if (options & __WNOTHREAD)
1537 tsk = next_thread(tsk);
1538 if (tsk->signal != current->signal)
1540 } while (tsk != current);
1542 read_unlock(&tasklist_lock);
1545 if (options & WNOHANG)
1547 retval = -ERESTARTSYS;
1548 if (signal_pending(current))
1555 current->state = TASK_RUNNING;
1556 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1562 * For a WNOHANG return, clear out all the fields
1563 * we would set so the user can easily tell the
1567 retval = put_user(0, &infop->si_signo);
1569 retval = put_user(0, &infop->si_errno);
1571 retval = put_user(0, &infop->si_code);
1573 retval = put_user(0, &infop->si_pid);
1575 retval = put_user(0, &infop->si_uid);
1577 retval = put_user(0, &infop->si_status);
1583 asmlinkage long sys_waitid(int which, pid_t pid,
1584 struct siginfo __user *infop, int options,
1585 struct rusage __user *ru)
1589 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1591 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1611 ret = do_wait(pid, options, infop, NULL, ru);
1613 /* avoid REGPARM breakage on x86: */
1614 prevent_tail_call(ret);
1618 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1619 int options, struct rusage __user *ru)
1623 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1624 __WNOTHREAD|__WCLONE|__WALL))
1626 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1628 /* avoid REGPARM breakage on x86: */
1629 prevent_tail_call(ret);
1633 #ifdef __ARCH_WANT_SYS_WAITPID
1636 * sys_waitpid() remains for compatibility. waitpid() should be
1637 * implemented by calling sys_wait4() from libc.a.
1639 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1641 return sys_wait4(pid, stat_addr, options, NULL);