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/cn_proc.h>
33 #include <linux/mutex.h>
34 #include <linux/futex.h>
35 #include <linux/compat.h>
37 #include <asm/uaccess.h>
38 #include <asm/unistd.h>
39 #include <asm/pgtable.h>
40 #include <asm/mmu_context.h>
42 extern void sem_exit (void);
43 extern struct task_struct *child_reaper;
45 int getrusage(struct task_struct *, int, struct rusage __user *);
47 static void exit_mm(struct task_struct * tsk);
49 static void __unhash_process(struct task_struct *p)
52 detach_pid(p, PIDTYPE_PID);
53 detach_pid(p, PIDTYPE_TGID);
54 if (thread_group_leader(p)) {
55 detach_pid(p, PIDTYPE_PGID);
56 detach_pid(p, PIDTYPE_SID);
58 list_del_init(&p->tasks);
59 __get_cpu_var(process_counts)--;
65 void release_task(struct task_struct * p)
69 struct dentry *proc_dentry;
72 atomic_dec(&p->user->processes);
73 spin_lock(&p->proc_lock);
74 proc_dentry = proc_pid_unhash(p);
75 write_lock_irq(&tasklist_lock);
76 if (unlikely(p->ptrace))
78 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
81 * Note that the fastpath in sys_times depends on __exit_signal having
82 * updated the counters before a task is removed from the tasklist of
83 * the process by __unhash_process.
88 * If we are the last non-leader member of the thread
89 * group, and the leader is zombie, then notify the
90 * group leader's parent process. (if it wants notification.)
93 leader = p->group_leader;
94 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
95 BUG_ON(leader->exit_signal == -1);
96 do_notify_parent(leader, leader->exit_signal);
98 * If we were the last child thread and the leader has
99 * exited already, and the leader's parent ignores SIGCHLD,
100 * then we are the one who should release the leader.
102 * do_notify_parent() will have marked it self-reaping in
105 zap_leader = (leader->exit_signal == -1);
109 write_unlock_irq(&tasklist_lock);
110 spin_unlock(&p->proc_lock);
111 proc_pid_flush(proc_dentry);
116 if (unlikely(zap_leader))
121 * This checks not only the pgrp, but falls back on the pid if no
122 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
125 int session_of_pgrp(int pgrp)
127 struct task_struct *p;
130 read_lock(&tasklist_lock);
131 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
132 if (p->signal->session > 0) {
133 sid = p->signal->session;
136 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
137 p = find_task_by_pid(pgrp);
139 sid = p->signal->session;
141 read_unlock(&tasklist_lock);
147 * Determine if a process group is "orphaned", according to the POSIX
148 * definition in 2.2.2.52. Orphaned process groups are not to be affected
149 * by terminal-generated stop signals. Newly orphaned process groups are
150 * to receive a SIGHUP and a SIGCONT.
152 * "I ask you, have you ever known what it is to be an orphan?"
154 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
156 struct task_struct *p;
159 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
160 if (p == ignored_task
162 || p->real_parent->pid == 1)
164 if (process_group(p->real_parent) != pgrp
165 && p->real_parent->signal->session == p->signal->session) {
169 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
170 return ret; /* (sighing) "Often!" */
173 int is_orphaned_pgrp(int pgrp)
177 read_lock(&tasklist_lock);
178 retval = will_become_orphaned_pgrp(pgrp, NULL);
179 read_unlock(&tasklist_lock);
184 static int has_stopped_jobs(int pgrp)
187 struct task_struct *p;
189 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
190 if (p->state != TASK_STOPPED)
193 /* If p is stopped by a debugger on a signal that won't
194 stop it, then don't count p as stopped. This isn't
195 perfect but it's a good approximation. */
196 if (unlikely (p->ptrace)
197 && p->exit_code != SIGSTOP
198 && p->exit_code != SIGTSTP
199 && p->exit_code != SIGTTOU
200 && p->exit_code != SIGTTIN)
205 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
210 * reparent_to_init - Reparent the calling kernel thread to the init task.
212 * If a kernel thread is launched as a result of a system call, or if
213 * it ever exits, it should generally reparent itself to init so that
214 * it is correctly cleaned up on exit.
216 * The various task state such as scheduling policy and priority may have
217 * been inherited from a user process, so we reset them to sane values here.
219 * NOTE that reparent_to_init() gives the caller full capabilities.
221 static void reparent_to_init(void)
223 write_lock_irq(&tasklist_lock);
225 ptrace_unlink(current);
226 /* Reparent to init */
227 remove_parent(current);
228 current->parent = child_reaper;
229 current->real_parent = child_reaper;
232 /* Set the exit signal to SIGCHLD so we signal init on exit */
233 current->exit_signal = SIGCHLD;
235 if ((current->policy == SCHED_NORMAL ||
236 current->policy == SCHED_BATCH)
237 && (task_nice(current) < 0))
238 set_user_nice(current, 0);
242 security_task_reparent_to_init(current);
243 memcpy(current->signal->rlim, init_task.signal->rlim,
244 sizeof(current->signal->rlim));
245 atomic_inc(&(INIT_USER->__count));
246 write_unlock_irq(&tasklist_lock);
247 switch_uid(INIT_USER);
250 void __set_special_pids(pid_t session, pid_t pgrp)
252 struct task_struct *curr = current->group_leader;
254 if (curr->signal->session != session) {
255 detach_pid(curr, PIDTYPE_SID);
256 curr->signal->session = session;
257 attach_pid(curr, PIDTYPE_SID, session);
259 if (process_group(curr) != pgrp) {
260 detach_pid(curr, PIDTYPE_PGID);
261 curr->signal->pgrp = pgrp;
262 attach_pid(curr, PIDTYPE_PGID, pgrp);
266 void set_special_pids(pid_t session, pid_t pgrp)
268 write_lock_irq(&tasklist_lock);
269 __set_special_pids(session, pgrp);
270 write_unlock_irq(&tasklist_lock);
274 * Let kernel threads use this to say that they
275 * allow a certain signal (since daemonize() will
276 * have disabled all of them by default).
278 int allow_signal(int sig)
280 if (!valid_signal(sig) || sig < 1)
283 spin_lock_irq(¤t->sighand->siglock);
284 sigdelset(¤t->blocked, sig);
286 /* Kernel threads handle their own signals.
287 Let the signal code know it'll be handled, so
288 that they don't get converted to SIGKILL or
289 just silently dropped */
290 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
293 spin_unlock_irq(¤t->sighand->siglock);
297 EXPORT_SYMBOL(allow_signal);
299 int disallow_signal(int sig)
301 if (!valid_signal(sig) || sig < 1)
304 spin_lock_irq(¤t->sighand->siglock);
305 sigaddset(¤t->blocked, sig);
307 spin_unlock_irq(¤t->sighand->siglock);
311 EXPORT_SYMBOL(disallow_signal);
314 * Put all the gunge required to become a kernel thread without
315 * attached user resources in one place where it belongs.
318 void daemonize(const char *name, ...)
321 struct fs_struct *fs;
324 va_start(args, name);
325 vsnprintf(current->comm, sizeof(current->comm), name, args);
329 * If we were started as result of loading a module, close all of the
330 * user space pages. We don't need them, and if we didn't close them
331 * they would be locked into memory.
335 set_special_pids(1, 1);
336 mutex_lock(&tty_mutex);
337 current->signal->tty = NULL;
338 mutex_unlock(&tty_mutex);
340 /* Block and flush all signals */
341 sigfillset(&blocked);
342 sigprocmask(SIG_BLOCK, &blocked, NULL);
343 flush_signals(current);
345 /* Become as one with the init task */
347 exit_fs(current); /* current->fs->count--; */
350 atomic_inc(&fs->count);
351 exit_namespace(current);
352 current->namespace = init_task.namespace;
353 get_namespace(current->namespace);
355 current->files = init_task.files;
356 atomic_inc(¤t->files->count);
361 EXPORT_SYMBOL(daemonize);
363 static void close_files(struct files_struct * files)
371 * It is safe to dereference the fd table without RCU or
372 * ->file_lock because this is the last reference to the
375 fdt = files_fdtable(files);
379 if (i >= fdt->max_fdset || i >= fdt->max_fds)
381 set = fdt->open_fds->fds_bits[j++];
384 struct file * file = xchg(&fdt->fd[i], NULL);
386 filp_close(file, files);
394 struct files_struct *get_files_struct(struct task_struct *task)
396 struct files_struct *files;
401 atomic_inc(&files->count);
407 void fastcall put_files_struct(struct files_struct *files)
411 if (atomic_dec_and_test(&files->count)) {
414 * Free the fd and fdset arrays if we expanded them.
415 * If the fdtable was embedded, pass files for freeing
416 * at the end of the RCU grace period. Otherwise,
417 * you can free files immediately.
419 fdt = files_fdtable(files);
420 if (fdt == &files->fdtab)
421 fdt->free_files = files;
423 kmem_cache_free(files_cachep, files);
428 EXPORT_SYMBOL(put_files_struct);
430 static inline void __exit_files(struct task_struct *tsk)
432 struct files_struct * files = tsk->files;
438 put_files_struct(files);
442 void exit_files(struct task_struct *tsk)
447 static inline void __put_fs_struct(struct fs_struct *fs)
449 /* No need to hold fs->lock if we are killing it */
450 if (atomic_dec_and_test(&fs->count)) {
457 mntput(fs->altrootmnt);
459 kmem_cache_free(fs_cachep, fs);
463 void put_fs_struct(struct fs_struct *fs)
468 static inline void __exit_fs(struct task_struct *tsk)
470 struct fs_struct * fs = tsk->fs;
480 void exit_fs(struct task_struct *tsk)
485 EXPORT_SYMBOL_GPL(exit_fs);
488 * Turn us into a lazy TLB process if we
491 static void exit_mm(struct task_struct * tsk)
493 struct mm_struct *mm = tsk->mm;
499 * Serialize with any possible pending coredump.
500 * We must hold mmap_sem around checking core_waiters
501 * and clearing tsk->mm. The core-inducing thread
502 * will increment core_waiters for each thread in the
503 * group with ->mm != NULL.
505 down_read(&mm->mmap_sem);
506 if (mm->core_waiters) {
507 up_read(&mm->mmap_sem);
508 down_write(&mm->mmap_sem);
509 if (!--mm->core_waiters)
510 complete(mm->core_startup_done);
511 up_write(&mm->mmap_sem);
513 wait_for_completion(&mm->core_done);
514 down_read(&mm->mmap_sem);
516 atomic_inc(&mm->mm_count);
517 if (mm != tsk->active_mm) BUG();
518 /* more a memory barrier than a real lock */
521 up_read(&mm->mmap_sem);
522 enter_lazy_tlb(mm, current);
527 static inline void choose_new_parent(task_t *p, task_t *reaper)
530 * Make sure we're not reparenting to ourselves and that
531 * the parent is not a zombie.
533 BUG_ON(p == reaper || reaper->exit_state);
534 p->real_parent = reaper;
537 static void reparent_thread(task_t *p, task_t *father, int traced)
539 /* We don't want people slaying init. */
540 if (p->exit_signal != -1)
541 p->exit_signal = SIGCHLD;
543 if (p->pdeath_signal)
544 /* We already hold the tasklist_lock here. */
545 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
547 /* Move the child from its dying parent to the new one. */
548 if (unlikely(traced)) {
549 /* Preserve ptrace links if someone else is tracing this child. */
550 list_del_init(&p->ptrace_list);
551 if (p->parent != p->real_parent)
552 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
554 /* If this child is being traced, then we're the one tracing it
555 * anyway, so let go of it.
559 p->parent = p->real_parent;
562 /* If we'd notified the old parent about this child's death,
563 * also notify the new parent.
565 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
566 thread_group_empty(p))
567 do_notify_parent(p, p->exit_signal);
568 else if (p->state == TASK_TRACED) {
570 * If it was at a trace stop, turn it into
571 * a normal stop since it's no longer being
579 * process group orphan check
580 * Case ii: Our child is in a different pgrp
581 * than we are, and it was the only connection
582 * outside, so the child pgrp is now orphaned.
584 if ((process_group(p) != process_group(father)) &&
585 (p->signal->session == father->signal->session)) {
586 int pgrp = process_group(p);
588 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
589 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
590 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
596 * When we die, we re-parent all our children.
597 * Try to give them to another thread in our thread
598 * group, and if no such member exists, give it to
599 * the global child reaper process (ie "init")
601 static void forget_original_parent(struct task_struct * father,
602 struct list_head *to_release)
604 struct task_struct *p, *reaper = father;
605 struct list_head *_p, *_n;
608 reaper = next_thread(reaper);
609 if (reaper == father) {
610 reaper = child_reaper;
613 } while (reaper->exit_state);
616 * There are only two places where our children can be:
618 * - in our child list
619 * - in our ptraced child list
621 * Search them and reparent children.
623 list_for_each_safe(_p, _n, &father->children) {
625 p = list_entry(_p,struct task_struct,sibling);
629 /* if father isn't the real parent, then ptrace must be enabled */
630 BUG_ON(father != p->real_parent && !ptrace);
632 if (father == p->real_parent) {
633 /* reparent with a reaper, real father it's us */
634 choose_new_parent(p, reaper);
635 reparent_thread(p, father, 0);
637 /* reparent ptraced task to its real parent */
639 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
640 thread_group_empty(p))
641 do_notify_parent(p, p->exit_signal);
645 * if the ptraced child is a zombie with exit_signal == -1
646 * we must collect it before we exit, or it will remain
647 * zombie forever since we prevented it from self-reap itself
648 * while it was being traced by us, to be able to see it in wait4.
650 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
651 list_add(&p->ptrace_list, to_release);
653 list_for_each_safe(_p, _n, &father->ptrace_children) {
654 p = list_entry(_p,struct task_struct,ptrace_list);
655 choose_new_parent(p, reaper);
656 reparent_thread(p, father, 1);
661 * Send signals to all our closest relatives so that they know
662 * to properly mourn us..
664 static void exit_notify(struct task_struct *tsk)
667 struct task_struct *t;
668 struct list_head ptrace_dead, *_p, *_n;
670 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
671 && !thread_group_empty(tsk)) {
673 * This occurs when there was a race between our exit
674 * syscall and a group signal choosing us as the one to
675 * wake up. It could be that we are the only thread
676 * alerted to check for pending signals, but another thread
677 * should be woken now to take the signal since we will not.
678 * Now we'll wake all the threads in the group just to make
679 * sure someone gets all the pending signals.
681 read_lock(&tasklist_lock);
682 spin_lock_irq(&tsk->sighand->siglock);
683 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
684 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
685 recalc_sigpending_tsk(t);
686 if (signal_pending(t))
687 signal_wake_up(t, 0);
689 spin_unlock_irq(&tsk->sighand->siglock);
690 read_unlock(&tasklist_lock);
693 write_lock_irq(&tasklist_lock);
696 * This does two things:
698 * A. Make init inherit all the child processes
699 * B. Check to see if any process groups have become orphaned
700 * as a result of our exiting, and if they have any stopped
701 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
704 INIT_LIST_HEAD(&ptrace_dead);
705 forget_original_parent(tsk, &ptrace_dead);
706 BUG_ON(!list_empty(&tsk->children));
707 BUG_ON(!list_empty(&tsk->ptrace_children));
710 * Check to see if any process groups have become orphaned
711 * as a result of our exiting, and if they have any stopped
712 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
714 * Case i: Our father is in a different pgrp than we are
715 * and we were the only connection outside, so our pgrp
716 * is about to become orphaned.
719 t = tsk->real_parent;
721 if ((process_group(t) != process_group(tsk)) &&
722 (t->signal->session == tsk->signal->session) &&
723 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
724 has_stopped_jobs(process_group(tsk))) {
725 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
726 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
729 /* Let father know we died
731 * Thread signals are configurable, but you aren't going to use
732 * that to send signals to arbitary processes.
733 * That stops right now.
735 * If the parent exec id doesn't match the exec id we saved
736 * when we started then we know the parent has changed security
739 * If our self_exec id doesn't match our parent_exec_id then
740 * we have changed execution domain as these two values started
741 * the same after a fork.
745 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
746 ( tsk->parent_exec_id != t->self_exec_id ||
747 tsk->self_exec_id != tsk->parent_exec_id)
748 && !capable(CAP_KILL))
749 tsk->exit_signal = SIGCHLD;
752 /* If something other than our normal parent is ptracing us, then
753 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
754 * only has special meaning to our real parent.
756 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
757 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
758 do_notify_parent(tsk, signal);
759 } else if (tsk->ptrace) {
760 do_notify_parent(tsk, SIGCHLD);
764 if (tsk->exit_signal == -1 &&
765 (likely(tsk->ptrace == 0) ||
766 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
768 tsk->exit_state = state;
770 write_unlock_irq(&tasklist_lock);
772 list_for_each_safe(_p, _n, &ptrace_dead) {
774 t = list_entry(_p,struct task_struct,ptrace_list);
778 /* If the process is dead, release it - nobody will wait for it */
779 if (state == EXIT_DEAD)
783 fastcall NORET_TYPE void do_exit(long code)
785 struct task_struct *tsk = current;
788 profile_task_exit(tsk);
790 WARN_ON(atomic_read(&tsk->fs_excl));
792 if (unlikely(in_interrupt()))
793 panic("Aiee, killing interrupt handler!");
794 if (unlikely(!tsk->pid))
795 panic("Attempted to kill the idle task!");
796 if (unlikely(tsk == child_reaper))
797 panic("Attempted to kill init!");
799 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
800 current->ptrace_message = code;
801 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
805 * We're taking recursive faults here in do_exit. Safest is to just
806 * leave this task alone and wait for reboot.
808 if (unlikely(tsk->flags & PF_EXITING)) {
810 "Fixing recursive fault but reboot is needed!\n");
813 set_current_state(TASK_UNINTERRUPTIBLE);
817 tsk->flags |= PF_EXITING;
820 * Make sure we don't try to process any timer firings
821 * while we are already exiting.
823 tsk->it_virt_expires = cputime_zero;
824 tsk->it_prof_expires = cputime_zero;
825 tsk->it_sched_expires = 0;
827 if (unlikely(in_atomic()))
828 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
829 current->comm, current->pid,
832 acct_update_integrals(tsk);
834 update_hiwater_rss(tsk->mm);
835 update_hiwater_vm(tsk->mm);
837 group_dead = atomic_dec_and_test(&tsk->signal->live);
839 hrtimer_cancel(&tsk->signal->real_timer);
840 exit_itimers(tsk->signal);
843 if (unlikely(tsk->robust_list))
844 exit_robust_list(tsk);
846 if (unlikely(tsk->compat_robust_list))
847 compat_exit_robust_list(tsk);
859 if (group_dead && tsk->signal->leader)
860 disassociate_ctty(1);
862 module_put(task_thread_info(tsk)->exec_domain->module);
864 module_put(tsk->binfmt->module);
866 tsk->exit_code = code;
867 proc_exit_connector(tsk);
870 mpol_free(tsk->mempolicy);
871 tsk->mempolicy = NULL;
874 * If DEBUG_MUTEXES is on, make sure we are holding no locks:
876 mutex_debug_check_no_locks_held(tsk);
881 /* PF_DEAD causes final put_task_struct after we schedule. */
883 BUG_ON(tsk->flags & PF_DEAD);
884 tsk->flags |= PF_DEAD;
888 /* Avoid "noreturn function does return". */
892 EXPORT_SYMBOL_GPL(do_exit);
894 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
902 EXPORT_SYMBOL(complete_and_exit);
904 asmlinkage long sys_exit(int error_code)
906 do_exit((error_code&0xff)<<8);
909 task_t fastcall *next_thread(const task_t *p)
911 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
914 EXPORT_SYMBOL(next_thread);
917 * Take down every thread in the group. This is called by fatal signals
918 * as well as by sys_exit_group (below).
921 do_group_exit(int exit_code)
923 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
925 if (current->signal->flags & SIGNAL_GROUP_EXIT)
926 exit_code = current->signal->group_exit_code;
927 else if (!thread_group_empty(current)) {
928 struct signal_struct *const sig = current->signal;
929 struct sighand_struct *const sighand = current->sighand;
930 read_lock(&tasklist_lock);
931 spin_lock_irq(&sighand->siglock);
932 if (sig->flags & SIGNAL_GROUP_EXIT)
933 /* Another thread got here before we took the lock. */
934 exit_code = sig->group_exit_code;
936 sig->group_exit_code = exit_code;
937 zap_other_threads(current);
939 spin_unlock_irq(&sighand->siglock);
940 read_unlock(&tasklist_lock);
948 * this kills every thread in the thread group. Note that any externally
949 * wait4()-ing process will get the correct exit code - even if this
950 * thread is not the thread group leader.
952 asmlinkage void sys_exit_group(int error_code)
954 do_group_exit((error_code & 0xff) << 8);
957 static int eligible_child(pid_t pid, int options, task_t *p)
963 if (process_group(p) != process_group(current))
965 } else if (pid != -1) {
966 if (process_group(p) != -pid)
971 * Do not consider detached threads that are
974 if (p->exit_signal == -1 && !p->ptrace)
977 /* Wait for all children (clone and not) if __WALL is set;
978 * otherwise, wait for clone children *only* if __WCLONE is
979 * set; otherwise, wait for non-clone children *only*. (Note:
980 * A "clone" child here is one that reports to its parent
981 * using a signal other than SIGCHLD.) */
982 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
983 && !(options & __WALL))
986 * Do not consider thread group leaders that are
987 * in a non-empty thread group:
989 if (current->tgid != p->tgid && delay_group_leader(p))
992 if (security_task_wait(p))
998 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
1000 struct siginfo __user *infop,
1001 struct rusage __user *rusagep)
1003 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1006 retval = put_user(SIGCHLD, &infop->si_signo);
1008 retval = put_user(0, &infop->si_errno);
1010 retval = put_user((short)why, &infop->si_code);
1012 retval = put_user(pid, &infop->si_pid);
1014 retval = put_user(uid, &infop->si_uid);
1016 retval = put_user(status, &infop->si_status);
1023 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1024 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1025 * the lock and this task is uninteresting. If we return nonzero, we have
1026 * released the lock and the system call should return.
1028 static int wait_task_zombie(task_t *p, int noreap,
1029 struct siginfo __user *infop,
1030 int __user *stat_addr, struct rusage __user *ru)
1032 unsigned long state;
1036 if (unlikely(noreap)) {
1039 int exit_code = p->exit_code;
1042 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1044 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1047 read_unlock(&tasklist_lock);
1048 if ((exit_code & 0x7f) == 0) {
1050 status = exit_code >> 8;
1052 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1053 status = exit_code & 0x7f;
1055 return wait_noreap_copyout(p, pid, uid, why,
1060 * Try to move the task's state to DEAD
1061 * only one thread is allowed to do this:
1063 state = xchg(&p->exit_state, EXIT_DEAD);
1064 if (state != EXIT_ZOMBIE) {
1065 BUG_ON(state != EXIT_DEAD);
1068 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1070 * This can only happen in a race with a ptraced thread
1071 * dying on another processor.
1076 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1077 struct signal_struct *psig;
1078 struct signal_struct *sig;
1081 * The resource counters for the group leader are in its
1082 * own task_struct. Those for dead threads in the group
1083 * are in its signal_struct, as are those for the child
1084 * processes it has previously reaped. All these
1085 * accumulate in the parent's signal_struct c* fields.
1087 * We don't bother to take a lock here to protect these
1088 * p->signal fields, because they are only touched by
1089 * __exit_signal, which runs with tasklist_lock
1090 * write-locked anyway, and so is excluded here. We do
1091 * need to protect the access to p->parent->signal fields,
1092 * as other threads in the parent group can be right
1093 * here reaping other children at the same time.
1095 spin_lock_irq(&p->parent->sighand->siglock);
1096 psig = p->parent->signal;
1099 cputime_add(psig->cutime,
1100 cputime_add(p->utime,
1101 cputime_add(sig->utime,
1104 cputime_add(psig->cstime,
1105 cputime_add(p->stime,
1106 cputime_add(sig->stime,
1109 p->min_flt + sig->min_flt + sig->cmin_flt;
1111 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1113 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1115 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1116 spin_unlock_irq(&p->parent->sighand->siglock);
1120 * Now we are sure this task is interesting, and no other
1121 * thread can reap it because we set its state to EXIT_DEAD.
1123 read_unlock(&tasklist_lock);
1125 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1126 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1127 ? p->signal->group_exit_code : p->exit_code;
1128 if (!retval && stat_addr)
1129 retval = put_user(status, stat_addr);
1130 if (!retval && infop)
1131 retval = put_user(SIGCHLD, &infop->si_signo);
1132 if (!retval && infop)
1133 retval = put_user(0, &infop->si_errno);
1134 if (!retval && infop) {
1137 if ((status & 0x7f) == 0) {
1141 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1144 retval = put_user((short)why, &infop->si_code);
1146 retval = put_user(status, &infop->si_status);
1148 if (!retval && infop)
1149 retval = put_user(p->pid, &infop->si_pid);
1150 if (!retval && infop)
1151 retval = put_user(p->uid, &infop->si_uid);
1153 // TODO: is this safe?
1154 p->exit_state = EXIT_ZOMBIE;
1158 if (p->real_parent != p->parent) {
1159 write_lock_irq(&tasklist_lock);
1160 /* Double-check with lock held. */
1161 if (p->real_parent != p->parent) {
1163 // TODO: is this safe?
1164 p->exit_state = EXIT_ZOMBIE;
1166 * If this is not a detached task, notify the parent.
1167 * If it's still not detached after that, don't release
1170 if (p->exit_signal != -1) {
1171 do_notify_parent(p, p->exit_signal);
1172 if (p->exit_signal != -1)
1176 write_unlock_irq(&tasklist_lock);
1185 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1186 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1187 * the lock and this task is uninteresting. If we return nonzero, we have
1188 * released the lock and the system call should return.
1190 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1191 struct siginfo __user *infop,
1192 int __user *stat_addr, struct rusage __user *ru)
1194 int retval, exit_code;
1198 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1199 p->signal && p->signal->group_stop_count > 0)
1201 * A group stop is in progress and this is the group leader.
1202 * We won't report until all threads have stopped.
1207 * Now we are pretty sure this task is interesting.
1208 * Make sure it doesn't get reaped out from under us while we
1209 * give up the lock and then examine it below. We don't want to
1210 * keep holding onto the tasklist_lock while we call getrusage and
1211 * possibly take page faults for user memory.
1214 read_unlock(&tasklist_lock);
1216 if (unlikely(noreap)) {
1219 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1221 exit_code = p->exit_code;
1222 if (unlikely(!exit_code) ||
1223 unlikely(p->state & TASK_TRACED))
1225 return wait_noreap_copyout(p, pid, uid,
1226 why, (exit_code << 8) | 0x7f,
1230 write_lock_irq(&tasklist_lock);
1233 * This uses xchg to be atomic with the thread resuming and setting
1234 * it. It must also be done with the write lock held to prevent a
1235 * race with the EXIT_ZOMBIE case.
1237 exit_code = xchg(&p->exit_code, 0);
1238 if (unlikely(p->exit_state)) {
1240 * The task resumed and then died. Let the next iteration
1241 * catch it in EXIT_ZOMBIE. Note that exit_code might
1242 * already be zero here if it resumed and did _exit(0).
1243 * The task itself is dead and won't touch exit_code again;
1244 * other processors in this function are locked out.
1246 p->exit_code = exit_code;
1249 if (unlikely(exit_code == 0)) {
1251 * Another thread in this function got to it first, or it
1252 * resumed, or it resumed and then died.
1254 write_unlock_irq(&tasklist_lock);
1258 * We are returning to the wait loop without having successfully
1259 * removed the process and having released the lock. We cannot
1260 * continue, since the "p" task pointer is potentially stale.
1262 * Return -EAGAIN, and do_wait() will restart the loop from the
1263 * beginning. Do _not_ re-acquire the lock.
1268 /* move to end of parent's list to avoid starvation */
1272 write_unlock_irq(&tasklist_lock);
1274 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1275 if (!retval && stat_addr)
1276 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1277 if (!retval && infop)
1278 retval = put_user(SIGCHLD, &infop->si_signo);
1279 if (!retval && infop)
1280 retval = put_user(0, &infop->si_errno);
1281 if (!retval && infop)
1282 retval = put_user((short)((p->ptrace & PT_PTRACED)
1283 ? CLD_TRAPPED : CLD_STOPPED),
1285 if (!retval && infop)
1286 retval = put_user(exit_code, &infop->si_status);
1287 if (!retval && infop)
1288 retval = put_user(p->pid, &infop->si_pid);
1289 if (!retval && infop)
1290 retval = put_user(p->uid, &infop->si_uid);
1300 * Handle do_wait work for one task in a live, non-stopped state.
1301 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1302 * the lock and this task is uninteresting. If we return nonzero, we have
1303 * released the lock and the system call should return.
1305 static int wait_task_continued(task_t *p, int noreap,
1306 struct siginfo __user *infop,
1307 int __user *stat_addr, struct rusage __user *ru)
1313 if (unlikely(!p->signal))
1316 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1319 spin_lock_irq(&p->sighand->siglock);
1320 /* Re-check with the lock held. */
1321 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1322 spin_unlock_irq(&p->sighand->siglock);
1326 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1327 spin_unlock_irq(&p->sighand->siglock);
1332 read_unlock(&tasklist_lock);
1335 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1337 if (!retval && stat_addr)
1338 retval = put_user(0xffff, stat_addr);
1342 retval = wait_noreap_copyout(p, pid, uid,
1343 CLD_CONTINUED, SIGCONT,
1345 BUG_ON(retval == 0);
1352 static inline int my_ptrace_child(struct task_struct *p)
1354 if (!(p->ptrace & PT_PTRACED))
1356 if (!(p->ptrace & PT_ATTACHED))
1359 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1360 * we are the attacher. If we are the real parent, this is a race
1361 * inside ptrace_attach. It is waiting for the tasklist_lock,
1362 * which we have to switch the parent links, but has already set
1363 * the flags in p->ptrace.
1365 return (p->parent != p->real_parent);
1368 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1369 int __user *stat_addr, struct rusage __user *ru)
1371 DECLARE_WAITQUEUE(wait, current);
1372 struct task_struct *tsk;
1375 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1378 * We will set this flag if we see any child that might later
1379 * match our criteria, even if we are not able to reap it yet.
1382 current->state = TASK_INTERRUPTIBLE;
1383 read_lock(&tasklist_lock);
1386 struct task_struct *p;
1387 struct list_head *_p;
1390 list_for_each(_p,&tsk->children) {
1391 p = list_entry(_p,struct task_struct,sibling);
1393 ret = eligible_child(pid, options, p);
1400 * When we hit the race with PTRACE_ATTACH,
1401 * we will not report this child. But the
1402 * race means it has not yet been moved to
1403 * our ptrace_children list, so we need to
1404 * set the flag here to avoid a spurious ECHILD
1405 * when the race happens with the only child.
1408 if (!my_ptrace_child(p))
1413 * It's stopped now, so it might later
1414 * continue, exit, or stop again.
1417 if (!(options & WUNTRACED) &&
1418 !my_ptrace_child(p))
1420 retval = wait_task_stopped(p, ret == 2,
1421 (options & WNOWAIT),
1424 if (retval == -EAGAIN)
1426 if (retval != 0) /* He released the lock. */
1431 if (p->exit_state == EXIT_DEAD)
1433 // case EXIT_ZOMBIE:
1434 if (p->exit_state == EXIT_ZOMBIE) {
1436 * Eligible but we cannot release
1440 goto check_continued;
1441 if (!likely(options & WEXITED))
1443 retval = wait_task_zombie(
1444 p, (options & WNOWAIT),
1445 infop, stat_addr, ru);
1446 /* He released the lock. */
1453 * It's running now, so it might later
1454 * exit, stop, or stop and then continue.
1457 if (!unlikely(options & WCONTINUED))
1459 retval = wait_task_continued(
1460 p, (options & WNOWAIT),
1461 infop, stat_addr, ru);
1462 if (retval != 0) /* He released the lock. */
1468 list_for_each(_p, &tsk->ptrace_children) {
1469 p = list_entry(_p, struct task_struct,
1471 if (!eligible_child(pid, options, p))
1477 if (options & __WNOTHREAD)
1479 tsk = next_thread(tsk);
1480 if (tsk->signal != current->signal)
1482 } while (tsk != current);
1484 read_unlock(&tasklist_lock);
1487 if (options & WNOHANG)
1489 retval = -ERESTARTSYS;
1490 if (signal_pending(current))
1497 current->state = TASK_RUNNING;
1498 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1504 * For a WNOHANG return, clear out all the fields
1505 * we would set so the user can easily tell the
1509 retval = put_user(0, &infop->si_signo);
1511 retval = put_user(0, &infop->si_errno);
1513 retval = put_user(0, &infop->si_code);
1515 retval = put_user(0, &infop->si_pid);
1517 retval = put_user(0, &infop->si_uid);
1519 retval = put_user(0, &infop->si_status);
1525 asmlinkage long sys_waitid(int which, pid_t pid,
1526 struct siginfo __user *infop, int options,
1527 struct rusage __user *ru)
1531 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1533 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1553 ret = do_wait(pid, options, infop, NULL, ru);
1555 /* avoid REGPARM breakage on x86: */
1556 prevent_tail_call(ret);
1560 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1561 int options, struct rusage __user *ru)
1565 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1566 __WNOTHREAD|__WCLONE|__WALL))
1568 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1570 /* avoid REGPARM breakage on x86: */
1571 prevent_tail_call(ret);
1575 #ifdef __ARCH_WANT_SYS_WAITPID
1578 * sys_waitpid() remains for compatibility. waitpid() should be
1579 * implemented by calling sys_wait4() from libc.a.
1581 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1583 return sys_wait4(pid, stat_addr, options, NULL);