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>
38 #include <asm/uaccess.h>
39 #include <asm/unistd.h>
40 #include <asm/pgtable.h>
41 #include <asm/mmu_context.h>
43 extern void sem_exit (void);
44 extern struct task_struct *child_reaper;
46 int getrusage(struct task_struct *, int, struct rusage __user *);
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 detach_pid(p, PIDTYPE_TGID);
55 if (thread_group_leader(p)) {
56 detach_pid(p, PIDTYPE_PGID);
57 detach_pid(p, PIDTYPE_SID);
59 list_del_init(&p->tasks);
60 __get_cpu_var(process_counts)--;
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. */
116 cleanup_sighand(tsk);
117 spin_unlock(&sighand->siglock);
120 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
121 flush_sigqueue(&tsk->pending);
123 flush_sigqueue(&sig->shared_pending);
124 __cleanup_signal(sig);
128 void release_task(struct task_struct * p)
132 struct dentry *proc_dentry;
135 atomic_dec(&p->user->processes);
136 spin_lock(&p->proc_lock);
137 proc_dentry = proc_pid_unhash(p);
138 write_lock_irq(&tasklist_lock);
140 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
143 * Note that the fastpath in sys_times depends on __exit_signal having
144 * updated the counters before a task is removed from the tasklist of
145 * the process by __unhash_process.
150 * If we are the last non-leader member of the thread
151 * group, and the leader is zombie, then notify the
152 * group leader's parent process. (if it wants notification.)
155 leader = p->group_leader;
156 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
157 BUG_ON(leader->exit_signal == -1);
158 do_notify_parent(leader, leader->exit_signal);
160 * If we were the last child thread and the leader has
161 * exited already, and the leader's parent ignores SIGCHLD,
162 * then we are the one who should release the leader.
164 * do_notify_parent() will have marked it self-reaping in
167 zap_leader = (leader->exit_signal == -1);
171 write_unlock_irq(&tasklist_lock);
172 spin_unlock(&p->proc_lock);
173 proc_pid_flush(proc_dentry);
178 if (unlikely(zap_leader))
183 * This checks not only the pgrp, but falls back on the pid if no
184 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
187 int session_of_pgrp(int pgrp)
189 struct task_struct *p;
192 read_lock(&tasklist_lock);
193 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
194 if (p->signal->session > 0) {
195 sid = p->signal->session;
198 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
199 p = find_task_by_pid(pgrp);
201 sid = p->signal->session;
203 read_unlock(&tasklist_lock);
209 * Determine if a process group is "orphaned", according to the POSIX
210 * definition in 2.2.2.52. Orphaned process groups are not to be affected
211 * by terminal-generated stop signals. Newly orphaned process groups are
212 * to receive a SIGHUP and a SIGCONT.
214 * "I ask you, have you ever known what it is to be an orphan?"
216 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
218 struct task_struct *p;
221 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
222 if (p == ignored_task
224 || p->real_parent->pid == 1)
226 if (process_group(p->real_parent) != pgrp
227 && p->real_parent->signal->session == p->signal->session) {
231 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
232 return ret; /* (sighing) "Often!" */
235 int is_orphaned_pgrp(int pgrp)
239 read_lock(&tasklist_lock);
240 retval = will_become_orphaned_pgrp(pgrp, NULL);
241 read_unlock(&tasklist_lock);
246 static int has_stopped_jobs(int pgrp)
249 struct task_struct *p;
251 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
252 if (p->state != TASK_STOPPED)
255 /* If p is stopped by a debugger on a signal that won't
256 stop it, then don't count p as stopped. This isn't
257 perfect but it's a good approximation. */
258 if (unlikely (p->ptrace)
259 && p->exit_code != SIGSTOP
260 && p->exit_code != SIGTSTP
261 && p->exit_code != SIGTTOU
262 && p->exit_code != SIGTTIN)
267 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
272 * reparent_to_init - Reparent the calling kernel thread to the init task.
274 * If a kernel thread is launched as a result of a system call, or if
275 * it ever exits, it should generally reparent itself to init so that
276 * it is correctly cleaned up on exit.
278 * The various task state such as scheduling policy and priority may have
279 * been inherited from a user process, so we reset them to sane values here.
281 * NOTE that reparent_to_init() gives the caller full capabilities.
283 static void reparent_to_init(void)
285 write_lock_irq(&tasklist_lock);
287 ptrace_unlink(current);
288 /* Reparent to init */
289 remove_parent(current);
290 current->parent = child_reaper;
291 current->real_parent = child_reaper;
294 /* Set the exit signal to SIGCHLD so we signal init on exit */
295 current->exit_signal = SIGCHLD;
297 if ((current->policy == SCHED_NORMAL ||
298 current->policy == SCHED_BATCH)
299 && (task_nice(current) < 0))
300 set_user_nice(current, 0);
304 security_task_reparent_to_init(current);
305 memcpy(current->signal->rlim, init_task.signal->rlim,
306 sizeof(current->signal->rlim));
307 atomic_inc(&(INIT_USER->__count));
308 write_unlock_irq(&tasklist_lock);
309 switch_uid(INIT_USER);
312 void __set_special_pids(pid_t session, pid_t pgrp)
314 struct task_struct *curr = current->group_leader;
316 if (curr->signal->session != session) {
317 detach_pid(curr, PIDTYPE_SID);
318 curr->signal->session = session;
319 attach_pid(curr, PIDTYPE_SID, session);
321 if (process_group(curr) != pgrp) {
322 detach_pid(curr, PIDTYPE_PGID);
323 curr->signal->pgrp = pgrp;
324 attach_pid(curr, PIDTYPE_PGID, pgrp);
328 void set_special_pids(pid_t session, pid_t pgrp)
330 write_lock_irq(&tasklist_lock);
331 __set_special_pids(session, pgrp);
332 write_unlock_irq(&tasklist_lock);
336 * Let kernel threads use this to say that they
337 * allow a certain signal (since daemonize() will
338 * have disabled all of them by default).
340 int allow_signal(int sig)
342 if (!valid_signal(sig) || sig < 1)
345 spin_lock_irq(¤t->sighand->siglock);
346 sigdelset(¤t->blocked, sig);
348 /* Kernel threads handle their own signals.
349 Let the signal code know it'll be handled, so
350 that they don't get converted to SIGKILL or
351 just silently dropped */
352 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
355 spin_unlock_irq(¤t->sighand->siglock);
359 EXPORT_SYMBOL(allow_signal);
361 int disallow_signal(int sig)
363 if (!valid_signal(sig) || sig < 1)
366 spin_lock_irq(¤t->sighand->siglock);
367 sigaddset(¤t->blocked, sig);
369 spin_unlock_irq(¤t->sighand->siglock);
373 EXPORT_SYMBOL(disallow_signal);
376 * Put all the gunge required to become a kernel thread without
377 * attached user resources in one place where it belongs.
380 void daemonize(const char *name, ...)
383 struct fs_struct *fs;
386 va_start(args, name);
387 vsnprintf(current->comm, sizeof(current->comm), name, args);
391 * If we were started as result of loading a module, close all of the
392 * user space pages. We don't need them, and if we didn't close them
393 * they would be locked into memory.
397 set_special_pids(1, 1);
398 mutex_lock(&tty_mutex);
399 current->signal->tty = NULL;
400 mutex_unlock(&tty_mutex);
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);
456 struct files_struct *get_files_struct(struct task_struct *task)
458 struct files_struct *files;
463 atomic_inc(&files->count);
469 void fastcall put_files_struct(struct files_struct *files)
473 if (atomic_dec_and_test(&files->count)) {
476 * Free the fd and fdset arrays if we expanded them.
477 * If the fdtable was embedded, pass files for freeing
478 * at the end of the RCU grace period. Otherwise,
479 * you can free files immediately.
481 fdt = files_fdtable(files);
482 if (fdt == &files->fdtab)
483 fdt->free_files = files;
485 kmem_cache_free(files_cachep, files);
490 EXPORT_SYMBOL(put_files_struct);
492 static inline void __exit_files(struct task_struct *tsk)
494 struct files_struct * files = tsk->files;
500 put_files_struct(files);
504 void exit_files(struct task_struct *tsk)
509 static inline void __put_fs_struct(struct fs_struct *fs)
511 /* No need to hold fs->lock if we are killing it */
512 if (atomic_dec_and_test(&fs->count)) {
519 mntput(fs->altrootmnt);
521 kmem_cache_free(fs_cachep, fs);
525 void put_fs_struct(struct fs_struct *fs)
530 static inline void __exit_fs(struct task_struct *tsk)
532 struct fs_struct * fs = tsk->fs;
542 void exit_fs(struct task_struct *tsk)
547 EXPORT_SYMBOL_GPL(exit_fs);
550 * Turn us into a lazy TLB process if we
553 static void exit_mm(struct task_struct * tsk)
555 struct mm_struct *mm = tsk->mm;
561 * Serialize with any possible pending coredump.
562 * We must hold mmap_sem around checking core_waiters
563 * and clearing tsk->mm. The core-inducing thread
564 * will increment core_waiters for each thread in the
565 * group with ->mm != NULL.
567 down_read(&mm->mmap_sem);
568 if (mm->core_waiters) {
569 up_read(&mm->mmap_sem);
570 down_write(&mm->mmap_sem);
571 if (!--mm->core_waiters)
572 complete(mm->core_startup_done);
573 up_write(&mm->mmap_sem);
575 wait_for_completion(&mm->core_done);
576 down_read(&mm->mmap_sem);
578 atomic_inc(&mm->mm_count);
579 if (mm != tsk->active_mm) BUG();
580 /* more a memory barrier than a real lock */
583 up_read(&mm->mmap_sem);
584 enter_lazy_tlb(mm, current);
589 static inline void choose_new_parent(task_t *p, task_t *reaper)
592 * Make sure we're not reparenting to ourselves and that
593 * the parent is not a zombie.
595 BUG_ON(p == reaper || reaper->exit_state);
596 p->real_parent = reaper;
599 static void reparent_thread(task_t *p, task_t *father, int traced)
601 /* We don't want people slaying init. */
602 if (p->exit_signal != -1)
603 p->exit_signal = SIGCHLD;
605 if (p->pdeath_signal)
606 /* We already hold the tasklist_lock here. */
607 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
609 /* Move the child from its dying parent to the new one. */
610 if (unlikely(traced)) {
611 /* Preserve ptrace links if someone else is tracing this child. */
612 list_del_init(&p->ptrace_list);
613 if (p->parent != p->real_parent)
614 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
616 /* If this child is being traced, then we're the one tracing it
617 * anyway, so let go of it.
621 p->parent = p->real_parent;
624 /* If we'd notified the old parent about this child's death,
625 * also notify the new parent.
627 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
628 thread_group_empty(p))
629 do_notify_parent(p, p->exit_signal);
630 else if (p->state == TASK_TRACED) {
632 * If it was at a trace stop, turn it into
633 * a normal stop since it's no longer being
641 * process group orphan check
642 * Case ii: Our child is in a different pgrp
643 * than we are, and it was the only connection
644 * outside, so the child pgrp is now orphaned.
646 if ((process_group(p) != process_group(father)) &&
647 (p->signal->session == father->signal->session)) {
648 int pgrp = process_group(p);
650 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
651 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
652 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
658 * When we die, we re-parent all our children.
659 * Try to give them to another thread in our thread
660 * group, and if no such member exists, give it to
661 * the global child reaper process (ie "init")
663 static void forget_original_parent(struct task_struct * father,
664 struct list_head *to_release)
666 struct task_struct *p, *reaper = father;
667 struct list_head *_p, *_n;
670 reaper = next_thread(reaper);
671 if (reaper == father) {
672 reaper = child_reaper;
675 } while (reaper->exit_state);
678 * There are only two places where our children can be:
680 * - in our child list
681 * - in our ptraced child list
683 * Search them and reparent children.
685 list_for_each_safe(_p, _n, &father->children) {
687 p = list_entry(_p,struct task_struct,sibling);
691 /* if father isn't the real parent, then ptrace must be enabled */
692 BUG_ON(father != p->real_parent && !ptrace);
694 if (father == p->real_parent) {
695 /* reparent with a reaper, real father it's us */
696 choose_new_parent(p, reaper);
697 reparent_thread(p, father, 0);
699 /* reparent ptraced task to its real parent */
701 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
702 thread_group_empty(p))
703 do_notify_parent(p, p->exit_signal);
707 * if the ptraced child is a zombie with exit_signal == -1
708 * we must collect it before we exit, or it will remain
709 * zombie forever since we prevented it from self-reap itself
710 * while it was being traced by us, to be able to see it in wait4.
712 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
713 list_add(&p->ptrace_list, to_release);
715 list_for_each_safe(_p, _n, &father->ptrace_children) {
716 p = list_entry(_p,struct task_struct,ptrace_list);
717 choose_new_parent(p, reaper);
718 reparent_thread(p, father, 1);
723 * Send signals to all our closest relatives so that they know
724 * to properly mourn us..
726 static void exit_notify(struct task_struct *tsk)
729 struct task_struct *t;
730 struct list_head ptrace_dead, *_p, *_n;
732 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
733 && !thread_group_empty(tsk)) {
735 * This occurs when there was a race between our exit
736 * syscall and a group signal choosing us as the one to
737 * wake up. It could be that we are the only thread
738 * alerted to check for pending signals, but another thread
739 * should be woken now to take the signal since we will not.
740 * Now we'll wake all the threads in the group just to make
741 * sure someone gets all the pending signals.
743 read_lock(&tasklist_lock);
744 spin_lock_irq(&tsk->sighand->siglock);
745 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
746 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
747 recalc_sigpending_tsk(t);
748 if (signal_pending(t))
749 signal_wake_up(t, 0);
751 spin_unlock_irq(&tsk->sighand->siglock);
752 read_unlock(&tasklist_lock);
755 write_lock_irq(&tasklist_lock);
758 * This does two things:
760 * A. Make init inherit all the child processes
761 * B. Check to see if any process groups have become orphaned
762 * as a result of our exiting, and if they have any stopped
763 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
766 INIT_LIST_HEAD(&ptrace_dead);
767 forget_original_parent(tsk, &ptrace_dead);
768 BUG_ON(!list_empty(&tsk->children));
769 BUG_ON(!list_empty(&tsk->ptrace_children));
772 * Check to see if any process groups have become orphaned
773 * as a result of our exiting, and if they have any stopped
774 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
776 * Case i: Our father is in a different pgrp than we are
777 * and we were the only connection outside, so our pgrp
778 * is about to become orphaned.
781 t = tsk->real_parent;
783 if ((process_group(t) != process_group(tsk)) &&
784 (t->signal->session == tsk->signal->session) &&
785 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
786 has_stopped_jobs(process_group(tsk))) {
787 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
788 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
791 /* Let father know we died
793 * Thread signals are configurable, but you aren't going to use
794 * that to send signals to arbitary processes.
795 * That stops right now.
797 * If the parent exec id doesn't match the exec id we saved
798 * when we started then we know the parent has changed security
801 * If our self_exec id doesn't match our parent_exec_id then
802 * we have changed execution domain as these two values started
803 * the same after a fork.
807 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
808 ( tsk->parent_exec_id != t->self_exec_id ||
809 tsk->self_exec_id != tsk->parent_exec_id)
810 && !capable(CAP_KILL))
811 tsk->exit_signal = SIGCHLD;
814 /* If something other than our normal parent is ptracing us, then
815 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
816 * only has special meaning to our real parent.
818 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
819 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
820 do_notify_parent(tsk, signal);
821 } else if (tsk->ptrace) {
822 do_notify_parent(tsk, SIGCHLD);
826 if (tsk->exit_signal == -1 &&
827 (likely(tsk->ptrace == 0) ||
828 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
830 tsk->exit_state = state;
832 write_unlock_irq(&tasklist_lock);
834 list_for_each_safe(_p, _n, &ptrace_dead) {
836 t = list_entry(_p,struct task_struct,ptrace_list);
840 /* If the process is dead, release it - nobody will wait for it */
841 if (state == EXIT_DEAD)
845 fastcall NORET_TYPE void do_exit(long code)
847 struct task_struct *tsk = current;
850 profile_task_exit(tsk);
852 WARN_ON(atomic_read(&tsk->fs_excl));
854 if (unlikely(in_interrupt()))
855 panic("Aiee, killing interrupt handler!");
856 if (unlikely(!tsk->pid))
857 panic("Attempted to kill the idle task!");
858 if (unlikely(tsk == child_reaper))
859 panic("Attempted to kill init!");
861 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
862 current->ptrace_message = code;
863 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
867 * We're taking recursive faults here in do_exit. Safest is to just
868 * leave this task alone and wait for reboot.
870 if (unlikely(tsk->flags & PF_EXITING)) {
872 "Fixing recursive fault but reboot is needed!\n");
875 set_current_state(TASK_UNINTERRUPTIBLE);
879 tsk->flags |= PF_EXITING;
882 * Make sure we don't try to process any timer firings
883 * while we are already exiting.
885 tsk->it_virt_expires = cputime_zero;
886 tsk->it_prof_expires = cputime_zero;
887 tsk->it_sched_expires = 0;
889 if (unlikely(in_atomic()))
890 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
891 current->comm, current->pid,
894 acct_update_integrals(tsk);
896 update_hiwater_rss(tsk->mm);
897 update_hiwater_vm(tsk->mm);
899 group_dead = atomic_dec_and_test(&tsk->signal->live);
901 hrtimer_cancel(&tsk->signal->real_timer);
902 exit_itimers(tsk->signal);
905 if (unlikely(tsk->robust_list))
906 exit_robust_list(tsk);
908 if (unlikely(tsk->compat_robust_list))
909 compat_exit_robust_list(tsk);
921 if (group_dead && tsk->signal->leader)
922 disassociate_ctty(1);
924 module_put(task_thread_info(tsk)->exec_domain->module);
926 module_put(tsk->binfmt->module);
928 tsk->exit_code = code;
929 proc_exit_connector(tsk);
932 mpol_free(tsk->mempolicy);
933 tsk->mempolicy = NULL;
936 * If DEBUG_MUTEXES is on, make sure we are holding no locks:
938 mutex_debug_check_no_locks_held(tsk);
943 /* PF_DEAD causes final put_task_struct after we schedule. */
945 BUG_ON(tsk->flags & PF_DEAD);
946 tsk->flags |= PF_DEAD;
950 /* Avoid "noreturn function does return". */
954 EXPORT_SYMBOL_GPL(do_exit);
956 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
964 EXPORT_SYMBOL(complete_and_exit);
966 asmlinkage long sys_exit(int error_code)
968 do_exit((error_code&0xff)<<8);
971 task_t fastcall *next_thread(const task_t *p)
973 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
976 EXPORT_SYMBOL(next_thread);
979 * Take down every thread in the group. This is called by fatal signals
980 * as well as by sys_exit_group (below).
983 do_group_exit(int exit_code)
985 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
987 if (current->signal->flags & SIGNAL_GROUP_EXIT)
988 exit_code = current->signal->group_exit_code;
989 else if (!thread_group_empty(current)) {
990 struct signal_struct *const sig = current->signal;
991 struct sighand_struct *const sighand = current->sighand;
992 read_lock(&tasklist_lock);
993 spin_lock_irq(&sighand->siglock);
994 if (sig->flags & SIGNAL_GROUP_EXIT)
995 /* Another thread got here before we took the lock. */
996 exit_code = sig->group_exit_code;
998 sig->group_exit_code = exit_code;
999 zap_other_threads(current);
1001 spin_unlock_irq(&sighand->siglock);
1002 read_unlock(&tasklist_lock);
1010 * this kills every thread in the thread group. Note that any externally
1011 * wait4()-ing process will get the correct exit code - even if this
1012 * thread is not the thread group leader.
1014 asmlinkage void sys_exit_group(int error_code)
1016 do_group_exit((error_code & 0xff) << 8);
1019 static int eligible_child(pid_t pid, int options, task_t *p)
1025 if (process_group(p) != process_group(current))
1027 } else if (pid != -1) {
1028 if (process_group(p) != -pid)
1033 * Do not consider detached threads that are
1036 if (p->exit_signal == -1 && !p->ptrace)
1039 /* Wait for all children (clone and not) if __WALL is set;
1040 * otherwise, wait for clone children *only* if __WCLONE is
1041 * set; otherwise, wait for non-clone children *only*. (Note:
1042 * A "clone" child here is one that reports to its parent
1043 * using a signal other than SIGCHLD.) */
1044 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1045 && !(options & __WALL))
1048 * Do not consider thread group leaders that are
1049 * in a non-empty thread group:
1051 if (current->tgid != p->tgid && delay_group_leader(p))
1054 if (security_task_wait(p))
1060 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
1061 int why, int status,
1062 struct siginfo __user *infop,
1063 struct rusage __user *rusagep)
1065 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1068 retval = put_user(SIGCHLD, &infop->si_signo);
1070 retval = put_user(0, &infop->si_errno);
1072 retval = put_user((short)why, &infop->si_code);
1074 retval = put_user(pid, &infop->si_pid);
1076 retval = put_user(uid, &infop->si_uid);
1078 retval = put_user(status, &infop->si_status);
1085 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1086 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1087 * the lock and this task is uninteresting. If we return nonzero, we have
1088 * released the lock and the system call should return.
1090 static int wait_task_zombie(task_t *p, int noreap,
1091 struct siginfo __user *infop,
1092 int __user *stat_addr, struct rusage __user *ru)
1094 unsigned long state;
1098 if (unlikely(noreap)) {
1101 int exit_code = p->exit_code;
1104 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1106 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1109 read_unlock(&tasklist_lock);
1110 if ((exit_code & 0x7f) == 0) {
1112 status = exit_code >> 8;
1114 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1115 status = exit_code & 0x7f;
1117 return wait_noreap_copyout(p, pid, uid, why,
1122 * Try to move the task's state to DEAD
1123 * only one thread is allowed to do this:
1125 state = xchg(&p->exit_state, EXIT_DEAD);
1126 if (state != EXIT_ZOMBIE) {
1127 BUG_ON(state != EXIT_DEAD);
1130 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1132 * This can only happen in a race with a ptraced thread
1133 * dying on another processor.
1138 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1139 struct signal_struct *psig;
1140 struct signal_struct *sig;
1143 * The resource counters for the group leader are in its
1144 * own task_struct. Those for dead threads in the group
1145 * are in its signal_struct, as are those for the child
1146 * processes it has previously reaped. All these
1147 * accumulate in the parent's signal_struct c* fields.
1149 * We don't bother to take a lock here to protect these
1150 * p->signal fields, because they are only touched by
1151 * __exit_signal, which runs with tasklist_lock
1152 * write-locked anyway, and so is excluded here. We do
1153 * need to protect the access to p->parent->signal fields,
1154 * as other threads in the parent group can be right
1155 * here reaping other children at the same time.
1157 spin_lock_irq(&p->parent->sighand->siglock);
1158 psig = p->parent->signal;
1161 cputime_add(psig->cutime,
1162 cputime_add(p->utime,
1163 cputime_add(sig->utime,
1166 cputime_add(psig->cstime,
1167 cputime_add(p->stime,
1168 cputime_add(sig->stime,
1171 p->min_flt + sig->min_flt + sig->cmin_flt;
1173 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1175 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1177 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1178 spin_unlock_irq(&p->parent->sighand->siglock);
1182 * Now we are sure this task is interesting, and no other
1183 * thread can reap it because we set its state to EXIT_DEAD.
1185 read_unlock(&tasklist_lock);
1187 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1188 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1189 ? p->signal->group_exit_code : p->exit_code;
1190 if (!retval && stat_addr)
1191 retval = put_user(status, stat_addr);
1192 if (!retval && infop)
1193 retval = put_user(SIGCHLD, &infop->si_signo);
1194 if (!retval && infop)
1195 retval = put_user(0, &infop->si_errno);
1196 if (!retval && infop) {
1199 if ((status & 0x7f) == 0) {
1203 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1206 retval = put_user((short)why, &infop->si_code);
1208 retval = put_user(status, &infop->si_status);
1210 if (!retval && infop)
1211 retval = put_user(p->pid, &infop->si_pid);
1212 if (!retval && infop)
1213 retval = put_user(p->uid, &infop->si_uid);
1215 // TODO: is this safe?
1216 p->exit_state = EXIT_ZOMBIE;
1220 if (p->real_parent != p->parent) {
1221 write_lock_irq(&tasklist_lock);
1222 /* Double-check with lock held. */
1223 if (p->real_parent != p->parent) {
1225 // TODO: is this safe?
1226 p->exit_state = EXIT_ZOMBIE;
1228 * If this is not a detached task, notify the parent.
1229 * If it's still not detached after that, don't release
1232 if (p->exit_signal != -1) {
1233 do_notify_parent(p, p->exit_signal);
1234 if (p->exit_signal != -1)
1238 write_unlock_irq(&tasklist_lock);
1247 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1248 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1249 * the lock and this task is uninteresting. If we return nonzero, we have
1250 * released the lock and the system call should return.
1252 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1253 struct siginfo __user *infop,
1254 int __user *stat_addr, struct rusage __user *ru)
1256 int retval, exit_code;
1260 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1261 p->signal && p->signal->group_stop_count > 0)
1263 * A group stop is in progress and this is the group leader.
1264 * We won't report until all threads have stopped.
1269 * Now we are pretty sure this task is interesting.
1270 * Make sure it doesn't get reaped out from under us while we
1271 * give up the lock and then examine it below. We don't want to
1272 * keep holding onto the tasklist_lock while we call getrusage and
1273 * possibly take page faults for user memory.
1276 read_unlock(&tasklist_lock);
1278 if (unlikely(noreap)) {
1281 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1283 exit_code = p->exit_code;
1284 if (unlikely(!exit_code) ||
1285 unlikely(p->state & TASK_TRACED))
1287 return wait_noreap_copyout(p, pid, uid,
1288 why, (exit_code << 8) | 0x7f,
1292 write_lock_irq(&tasklist_lock);
1295 * This uses xchg to be atomic with the thread resuming and setting
1296 * it. It must also be done with the write lock held to prevent a
1297 * race with the EXIT_ZOMBIE case.
1299 exit_code = xchg(&p->exit_code, 0);
1300 if (unlikely(p->exit_state)) {
1302 * The task resumed and then died. Let the next iteration
1303 * catch it in EXIT_ZOMBIE. Note that exit_code might
1304 * already be zero here if it resumed and did _exit(0).
1305 * The task itself is dead and won't touch exit_code again;
1306 * other processors in this function are locked out.
1308 p->exit_code = exit_code;
1311 if (unlikely(exit_code == 0)) {
1313 * Another thread in this function got to it first, or it
1314 * resumed, or it resumed and then died.
1316 write_unlock_irq(&tasklist_lock);
1320 * We are returning to the wait loop without having successfully
1321 * removed the process and having released the lock. We cannot
1322 * continue, since the "p" task pointer is potentially stale.
1324 * Return -EAGAIN, and do_wait() will restart the loop from the
1325 * beginning. Do _not_ re-acquire the lock.
1330 /* move to end of parent's list to avoid starvation */
1334 write_unlock_irq(&tasklist_lock);
1336 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1337 if (!retval && stat_addr)
1338 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1339 if (!retval && infop)
1340 retval = put_user(SIGCHLD, &infop->si_signo);
1341 if (!retval && infop)
1342 retval = put_user(0, &infop->si_errno);
1343 if (!retval && infop)
1344 retval = put_user((short)((p->ptrace & PT_PTRACED)
1345 ? CLD_TRAPPED : CLD_STOPPED),
1347 if (!retval && infop)
1348 retval = put_user(exit_code, &infop->si_status);
1349 if (!retval && infop)
1350 retval = put_user(p->pid, &infop->si_pid);
1351 if (!retval && infop)
1352 retval = put_user(p->uid, &infop->si_uid);
1362 * Handle do_wait work for one task in a live, non-stopped state.
1363 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1364 * the lock and this task is uninteresting. If we return nonzero, we have
1365 * released the lock and the system call should return.
1367 static int wait_task_continued(task_t *p, int noreap,
1368 struct siginfo __user *infop,
1369 int __user *stat_addr, struct rusage __user *ru)
1375 if (unlikely(!p->signal))
1378 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1381 spin_lock_irq(&p->sighand->siglock);
1382 /* Re-check with the lock held. */
1383 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1384 spin_unlock_irq(&p->sighand->siglock);
1388 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1389 spin_unlock_irq(&p->sighand->siglock);
1394 read_unlock(&tasklist_lock);
1397 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1399 if (!retval && stat_addr)
1400 retval = put_user(0xffff, stat_addr);
1404 retval = wait_noreap_copyout(p, pid, uid,
1405 CLD_CONTINUED, SIGCONT,
1407 BUG_ON(retval == 0);
1414 static inline int my_ptrace_child(struct task_struct *p)
1416 if (!(p->ptrace & PT_PTRACED))
1418 if (!(p->ptrace & PT_ATTACHED))
1421 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1422 * we are the attacher. If we are the real parent, this is a race
1423 * inside ptrace_attach. It is waiting for the tasklist_lock,
1424 * which we have to switch the parent links, but has already set
1425 * the flags in p->ptrace.
1427 return (p->parent != p->real_parent);
1430 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1431 int __user *stat_addr, struct rusage __user *ru)
1433 DECLARE_WAITQUEUE(wait, current);
1434 struct task_struct *tsk;
1437 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1440 * We will set this flag if we see any child that might later
1441 * match our criteria, even if we are not able to reap it yet.
1444 current->state = TASK_INTERRUPTIBLE;
1445 read_lock(&tasklist_lock);
1448 struct task_struct *p;
1449 struct list_head *_p;
1452 list_for_each(_p,&tsk->children) {
1453 p = list_entry(_p,struct task_struct,sibling);
1455 ret = eligible_child(pid, options, p);
1462 * When we hit the race with PTRACE_ATTACH,
1463 * we will not report this child. But the
1464 * race means it has not yet been moved to
1465 * our ptrace_children list, so we need to
1466 * set the flag here to avoid a spurious ECHILD
1467 * when the race happens with the only child.
1470 if (!my_ptrace_child(p))
1475 * It's stopped now, so it might later
1476 * continue, exit, or stop again.
1479 if (!(options & WUNTRACED) &&
1480 !my_ptrace_child(p))
1482 retval = wait_task_stopped(p, ret == 2,
1483 (options & WNOWAIT),
1486 if (retval == -EAGAIN)
1488 if (retval != 0) /* He released the lock. */
1493 if (p->exit_state == EXIT_DEAD)
1495 // case EXIT_ZOMBIE:
1496 if (p->exit_state == EXIT_ZOMBIE) {
1498 * Eligible but we cannot release
1502 goto check_continued;
1503 if (!likely(options & WEXITED))
1505 retval = wait_task_zombie(
1506 p, (options & WNOWAIT),
1507 infop, stat_addr, ru);
1508 /* He released the lock. */
1515 * It's running now, so it might later
1516 * exit, stop, or stop and then continue.
1519 if (!unlikely(options & WCONTINUED))
1521 retval = wait_task_continued(
1522 p, (options & WNOWAIT),
1523 infop, stat_addr, ru);
1524 if (retval != 0) /* He released the lock. */
1530 list_for_each(_p, &tsk->ptrace_children) {
1531 p = list_entry(_p, struct task_struct,
1533 if (!eligible_child(pid, options, p))
1539 if (options & __WNOTHREAD)
1541 tsk = next_thread(tsk);
1542 if (tsk->signal != current->signal)
1544 } while (tsk != current);
1546 read_unlock(&tasklist_lock);
1549 if (options & WNOHANG)
1551 retval = -ERESTARTSYS;
1552 if (signal_pending(current))
1559 current->state = TASK_RUNNING;
1560 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1566 * For a WNOHANG return, clear out all the fields
1567 * we would set so the user can easily tell the
1571 retval = put_user(0, &infop->si_signo);
1573 retval = put_user(0, &infop->si_errno);
1575 retval = put_user(0, &infop->si_code);
1577 retval = put_user(0, &infop->si_pid);
1579 retval = put_user(0, &infop->si_uid);
1581 retval = put_user(0, &infop->si_status);
1587 asmlinkage long sys_waitid(int which, pid_t pid,
1588 struct siginfo __user *infop, int options,
1589 struct rusage __user *ru)
1593 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1595 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1615 ret = do_wait(pid, options, infop, NULL, ru);
1617 /* avoid REGPARM breakage on x86: */
1618 prevent_tail_call(ret);
1622 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1623 int options, struct rusage __user *ru)
1627 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1628 __WNOTHREAD|__WCLONE|__WALL))
1630 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1632 /* avoid REGPARM breakage on x86: */
1633 prevent_tail_call(ret);
1637 #ifdef __ARCH_WANT_SYS_WAITPID
1640 * sys_waitpid() remains for compatibility. waitpid() should be
1641 * implemented by calling sys_wait4() from libc.a.
1643 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1645 return sys_wait4(pid, stat_addr, options, NULL);