4 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/slab.h>
9 #include <linux/interrupt.h>
10 #include <linux/smp_lock.h>
11 #include <linux/module.h>
12 #include <linux/capability.h>
13 #include <linux/completion.h>
14 #include <linux/personality.h>
15 #include <linux/tty.h>
16 #include <linux/namespace.h>
17 #include <linux/key.h>
18 #include <linux/security.h>
19 #include <linux/cpu.h>
20 #include <linux/acct.h>
21 #include <linux/file.h>
22 #include <linux/binfmts.h>
23 #include <linux/ptrace.h>
24 #include <linux/tracehook.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/taskstats_kern.h>
30 #include <linux/delayacct.h>
31 #include <linux/cpuset.h>
32 #include <linux/syscalls.h>
33 #include <linux/signal.h>
34 #include <linux/posix-timers.h>
35 #include <linux/cn_proc.h>
36 #include <linux/mutex.h>
37 #include <linux/futex.h>
38 #include <linux/compat.h>
39 #include <linux/pipe_fs_i.h>
40 #include <linux/audit.h> /* for audit_free() */
41 #include <linux/resource.h>
42 #include <linux/vs_limit.h>
43 #include <linux/vs_context.h>
44 #include <linux/vs_network.h>
46 #include <asm/uaccess.h>
47 #include <asm/unistd.h>
48 #include <asm/pgtable.h>
49 #include <asm/mmu_context.h>
51 extern void sem_exit (void);
52 extern struct task_struct *child_reaper;
54 static void exit_mm(struct task_struct * tsk);
56 static void __unhash_process(struct task_struct *p)
59 detach_pid(p, PIDTYPE_PID);
60 if (thread_group_leader(p)) {
61 detach_pid(p, PIDTYPE_PGID);
62 detach_pid(p, PIDTYPE_SID);
64 list_del_rcu(&p->tasks);
65 __get_cpu_var(process_counts)--;
67 list_del_rcu(&p->thread_group);
72 * This function expects the tasklist_lock write-locked.
74 static void __exit_signal(struct task_struct *tsk)
76 struct signal_struct *sig = tsk->signal;
77 struct sighand_struct *sighand;
80 BUG_ON(!atomic_read(&sig->count));
83 sighand = rcu_dereference(tsk->sighand);
84 spin_lock(&sighand->siglock);
86 posix_cpu_timers_exit(tsk);
87 if (atomic_dec_and_test(&sig->count))
88 posix_cpu_timers_exit_group(tsk);
91 * If there is any task waiting for the group exit
94 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
95 wake_up_process(sig->group_exit_task);
96 sig->group_exit_task = NULL;
98 if (tsk == sig->curr_target)
99 sig->curr_target = next_thread(tsk);
101 * Accumulate here the counters for all threads but the
102 * group leader as they die, so they can be added into
103 * the process-wide totals when those are taken.
104 * The group leader stays around as a zombie as long
105 * as there are other threads. When it gets reaped,
106 * the exit.c code will add its counts into these totals.
107 * We won't ever get here for the group leader, since it
108 * will have been the last reference on the signal_struct.
110 sig->utime = cputime_add(sig->utime, tsk->utime);
111 sig->stime = cputime_add(sig->stime, tsk->stime);
112 sig->min_flt += tsk->min_flt;
113 sig->maj_flt += tsk->maj_flt;
114 sig->nvcsw += tsk->nvcsw;
115 sig->nivcsw += tsk->nivcsw;
116 sig->sched_time += tsk->sched_time;
117 sig = NULL; /* Marker for below. */
120 __unhash_process(tsk);
124 spin_unlock(&sighand->siglock);
127 __cleanup_sighand(sighand);
128 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
129 flush_sigqueue(&tsk->pending);
131 flush_sigqueue(&sig->shared_pending);
132 __cleanup_signal(sig);
136 static void delayed_put_task_struct(struct rcu_head *rhp)
138 put_task_struct(container_of(rhp, struct task_struct, rcu));
141 void release_task(struct task_struct * p)
143 struct task_struct *leader;
146 tracehook_release_task(p);
147 atomic_dec(&p->user->processes);
148 write_lock_irq(&tasklist_lock);
149 BUG_ON(tracehook_check_released(p));
153 * If we are the last non-leader member of the thread
154 * group, and the leader is zombie, then notify the
155 * group leader's parent process. (if it wants notification.)
158 leader = p->group_leader;
159 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
160 BUG_ON(leader->exit_signal == -1);
161 do_notify_parent(leader, leader->exit_signal);
163 * If we were the last child thread and the leader has
164 * exited already, and the leader's parent ignores SIGCHLD,
165 * then we are the one who should release the leader.
167 * do_notify_parent() will have marked it self-reaping in
170 zap_leader = (leader->exit_signal == -1);
174 write_unlock_irq(&tasklist_lock);
177 call_rcu(&p->rcu, delayed_put_task_struct);
180 if (unlikely(zap_leader))
185 * This checks not only the pgrp, but falls back on the pid if no
186 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
189 int session_of_pgrp(int pgrp)
191 struct task_struct *p;
194 read_lock(&tasklist_lock);
195 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
196 if (p->signal->session > 0) {
197 sid = p->signal->session;
200 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
201 p = find_task_by_pid(pgrp);
203 sid = p->signal->session;
205 read_unlock(&tasklist_lock);
211 * Determine if a process group is "orphaned", according to the POSIX
212 * definition in 2.2.2.52. Orphaned process groups are not to be affected
213 * by terminal-generated stop signals. Newly orphaned process groups are
214 * to receive a SIGHUP and a SIGCONT.
216 * "I ask you, have you ever known what it is to be an orphan?"
218 static int will_become_orphaned_pgrp(int pgrp, struct task_struct *ignored_task)
220 struct task_struct *p;
223 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
224 if (p == ignored_task
226 || p->parent->pid == 1)
228 if (process_group(p->parent) != pgrp
229 && p->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)
259 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
264 * reparent_to_init - Reparent the calling kernel thread to the init task.
266 * If a kernel thread is launched as a result of a system call, or if
267 * it ever exits, it should generally reparent itself to init so that
268 * it is correctly cleaned up on exit.
270 * The various task state such as scheduling policy and priority may have
271 * been inherited from a user process, so we reset them to sane values here.
273 * NOTE that reparent_to_init() gives the caller full capabilities.
275 static void reparent_to_init(void)
277 write_lock_irq(&tasklist_lock);
279 /* Reparent to init */
280 remove_parent(current);
281 current->parent = child_reaper;
284 /* Set the exit signal to SIGCHLD so we signal init on exit */
285 current->exit_signal = SIGCHLD;
287 if ((current->policy == SCHED_NORMAL ||
288 current->policy == SCHED_BATCH)
289 && (task_nice(current) < 0))
290 set_user_nice(current, 0);
294 security_task_reparent_to_init(current);
295 memcpy(current->signal->rlim, init_task.signal->rlim,
296 sizeof(current->signal->rlim));
297 atomic_inc(&(INIT_USER->__count));
298 write_unlock_irq(&tasklist_lock);
299 switch_uid(INIT_USER);
302 void __set_special_pids(pid_t session, pid_t pgrp)
304 struct task_struct *curr = current->group_leader;
306 if (curr->signal->session != session) {
307 detach_pid(curr, PIDTYPE_SID);
308 curr->signal->session = session;
309 attach_pid(curr, PIDTYPE_SID, session);
311 if (process_group(curr) != pgrp) {
312 detach_pid(curr, PIDTYPE_PGID);
313 curr->signal->pgrp = pgrp;
314 attach_pid(curr, PIDTYPE_PGID, pgrp);
318 void set_special_pids(pid_t session, pid_t pgrp)
320 write_lock_irq(&tasklist_lock);
321 __set_special_pids(session, pgrp);
322 write_unlock_irq(&tasklist_lock);
326 * Let kernel threads use this to say that they
327 * allow a certain signal (since daemonize() will
328 * have disabled all of them by default).
330 int allow_signal(int sig)
332 if (!valid_signal(sig) || sig < 1)
335 spin_lock_irq(¤t->sighand->siglock);
336 sigdelset(¤t->blocked, sig);
338 /* Kernel threads handle their own signals.
339 Let the signal code know it'll be handled, so
340 that they don't get converted to SIGKILL or
341 just silently dropped */
342 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
345 spin_unlock_irq(¤t->sighand->siglock);
349 EXPORT_SYMBOL(allow_signal);
351 int disallow_signal(int sig)
353 if (!valid_signal(sig) || sig < 1)
356 spin_lock_irq(¤t->sighand->siglock);
357 sigaddset(¤t->blocked, sig);
359 spin_unlock_irq(¤t->sighand->siglock);
363 EXPORT_SYMBOL(disallow_signal);
366 * Put all the gunge required to become a kernel thread without
367 * attached user resources in one place where it belongs.
370 void daemonize(const char *name, ...)
373 struct fs_struct *fs;
376 va_start(args, name);
377 vsnprintf(current->comm, sizeof(current->comm), name, args);
381 * If we were started as result of loading a module, close all of the
382 * user space pages. We don't need them, and if we didn't close them
383 * they would be locked into memory.
387 set_special_pids(1, 1);
388 mutex_lock(&tty_mutex);
389 current->signal->tty = NULL;
390 mutex_unlock(&tty_mutex);
392 /* Block and flush all signals */
393 sigfillset(&blocked);
394 sigprocmask(SIG_BLOCK, &blocked, NULL);
395 flush_signals(current);
397 /* Become as one with the init task */
399 exit_fs(current); /* current->fs->count--; */
402 atomic_inc(&fs->count);
403 exit_namespace(current);
404 current->namespace = init_task.namespace;
405 get_namespace(current->namespace);
407 current->files = init_task.files;
408 atomic_inc(¤t->files->count);
413 EXPORT_SYMBOL(daemonize);
415 static void close_files(struct files_struct * files)
423 * It is safe to dereference the fd table without RCU or
424 * ->file_lock because this is the last reference to the
427 fdt = files_fdtable(files);
431 if (i >= fdt->max_fdset || i >= fdt->max_fds)
433 set = fdt->open_fds->fds_bits[j++];
436 struct file * file = xchg(&fdt->fd[i], NULL);
438 filp_close(file, files);
448 struct files_struct *get_files_struct(struct task_struct *task)
450 struct files_struct *files;
455 atomic_inc(&files->count);
461 void fastcall put_files_struct(struct files_struct *files)
465 if (atomic_dec_and_test(&files->count)) {
468 * Free the fd and fdset arrays if we expanded them.
469 * If the fdtable was embedded, pass files for freeing
470 * at the end of the RCU grace period. Otherwise,
471 * you can free files immediately.
473 fdt = files_fdtable(files);
474 if (fdt == &files->fdtab)
475 fdt->free_files = files;
477 kmem_cache_free(files_cachep, files);
482 EXPORT_SYMBOL(put_files_struct);
484 static inline void __exit_files(struct task_struct *tsk)
486 struct files_struct * files = tsk->files;
492 put_files_struct(files);
496 void exit_files(struct task_struct *tsk)
501 static inline void __put_fs_struct(struct fs_struct *fs)
503 /* No need to hold fs->lock if we are killing it */
504 if (atomic_dec_and_test(&fs->count)) {
511 mntput(fs->altrootmnt);
513 kmem_cache_free(fs_cachep, fs);
517 void put_fs_struct(struct fs_struct *fs)
522 static inline void __exit_fs(struct task_struct *tsk)
524 struct fs_struct * fs = tsk->fs;
534 void exit_fs(struct task_struct *tsk)
539 EXPORT_SYMBOL_GPL(exit_fs);
542 * Turn us into a lazy TLB process if we
545 static void exit_mm(struct task_struct * tsk)
547 struct mm_struct *mm = tsk->mm;
553 * Serialize with any possible pending coredump.
554 * We must hold mmap_sem around checking core_waiters
555 * and clearing tsk->mm. The core-inducing thread
556 * will increment core_waiters for each thread in the
557 * group with ->mm != NULL.
559 down_read(&mm->mmap_sem);
560 if (mm->core_waiters) {
561 up_read(&mm->mmap_sem);
562 down_write(&mm->mmap_sem);
563 if (!--mm->core_waiters)
564 complete(mm->core_startup_done);
565 up_write(&mm->mmap_sem);
567 wait_for_completion(&mm->core_done);
568 down_read(&mm->mmap_sem);
570 atomic_inc(&mm->mm_count);
571 BUG_ON(mm != tsk->active_mm);
572 /* more a memory barrier than a real lock */
575 up_read(&mm->mmap_sem);
576 enter_lazy_tlb(mm, current);
582 choose_new_parent(struct task_struct *p, struct task_struct *reaper)
584 /* check for reaper context */
585 vxwprintk((p->xid != reaper->xid) && (reaper != child_reaper),
586 "rogue reaper: %p[%d,#%u] <> %p[%d,#%u]",
587 p, p->pid, p->xid, reaper, reaper->pid, reaper->xid);
590 * Make sure we're not reparenting to ourselves and that
591 * the parent is not a zombie.
593 BUG_ON(p == reaper || reaper->exit_state);
598 reparent_thread(struct task_struct *p, struct task_struct *father)
600 /* We don't want people slaying init. */
601 if (p->exit_signal != -1)
602 p->exit_signal = SIGCHLD;
604 if (p->pdeath_signal)
605 /* We already hold the tasklist_lock here. */
606 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
608 /* Move the child from its dying parent to the new one. */
609 list_move_tail(&p->sibling, &p->parent->children);
611 /* If we'd notified the old parent about this child's death,
612 * also notify the new parent.
614 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
615 thread_group_empty(p))
616 do_notify_parent(p, p->exit_signal);
619 * process group orphan check
620 * Case ii: Our child is in a different pgrp
621 * than we are, and it was the only connection
622 * outside, so the child pgrp is now orphaned.
624 if ((process_group(p) != process_group(father)) &&
625 (p->signal->session == father->signal->session)) {
626 int pgrp = process_group(p);
628 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
629 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
630 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
636 * When we die, we re-parent all our children.
637 * Try to give them to another thread in our thread
638 * group, and if no such member exists, give it to
639 * the global child reaper process (ie "init")
642 forget_original_parent(struct task_struct *father)
644 struct task_struct *p, *reaper = father;
645 struct list_head *_p, *_n;
648 reaper = next_thread(reaper);
649 if (reaper == father) {
650 reaper = vx_child_reaper(father);
653 } while (reaper->exit_state);
655 list_for_each_safe(_p, _n, &father->children) {
656 p = list_entry(_p, struct task_struct, sibling);
657 choose_new_parent(p, vx_child_reaper(p));
658 reparent_thread(p, father);
663 * Send signals to all our closest relatives so that they know
664 * to properly mourn us..
666 static void exit_notify(struct task_struct *tsk)
669 struct task_struct *t;
673 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
674 && !thread_group_empty(tsk)) {
676 * This occurs when there was a race between our exit
677 * syscall and a group signal choosing us as the one to
678 * wake up. It could be that we are the only thread
679 * alerted to check for pending signals, but another thread
680 * should be woken now to take the signal since we will not.
681 * Now we'll wake all the threads in the group just to make
682 * sure someone gets all the pending signals.
684 read_lock(&tasklist_lock);
685 spin_lock_irq(&tsk->sighand->siglock);
686 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
687 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
688 recalc_sigpending_tsk(t);
689 if (signal_pending(t))
690 signal_wake_up(t, 0);
692 spin_unlock_irq(&tsk->sighand->siglock);
693 read_unlock(&tasklist_lock);
696 write_lock_irq(&tasklist_lock);
699 * This does two things:
701 * A. Make init inherit all the child processes
702 * B. Check to see if any process groups have become orphaned
703 * as a result of our exiting, and if they have any stopped
704 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
707 forget_original_parent(tsk);
708 BUG_ON(!list_empty(&tsk->children));
711 * Check to see if any process groups have become orphaned
712 * as a result of our exiting, and if they have any stopped
713 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
715 * Case i: Our father is in a different pgrp than we are
716 * and we were the only connection outside, so our pgrp
717 * is about to become orphaned.
722 if ((process_group(t) != process_group(tsk)) &&
723 (t->signal->session == tsk->signal->session) &&
724 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
725 has_stopped_jobs(process_group(tsk))) {
726 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
727 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
730 /* Let father know we died
732 * Thread signals are configurable, but you aren't going to use
733 * that to send signals to arbitary processes.
734 * That stops right now.
736 * If the parent exec id doesn't match the exec id we saved
737 * when we started then we know the parent has changed security
740 * If our self_exec id doesn't match our parent_exec_id then
741 * we have changed execution domain as these two values started
742 * the same after a fork.
746 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
747 ( tsk->parent_exec_id != t->self_exec_id ||
748 tsk->self_exec_id != tsk->parent_exec_id)
749 && !capable(CAP_KILL))
750 tsk->exit_signal = SIGCHLD;
752 if (!tracehook_notify_death(tsk, &noreap, &cookie)
753 && tsk->exit_signal != -1 && thread_group_empty(tsk))
754 do_notify_parent(tsk, tsk->exit_signal);
757 if (tsk->exit_signal == -1 && !noreap)
759 tsk->exit_state = state;
761 write_unlock_irq(&tasklist_lock);
763 tracehook_report_death(tsk, state, cookie);
765 /* If the process is dead, release it - nobody will wait for it */
766 if (state == EXIT_DEAD)
770 fastcall NORET_TYPE void do_exit(long code)
772 struct task_struct *tsk = current;
773 struct taskstats *tidstats;
777 profile_task_exit(tsk);
779 WARN_ON(atomic_read(&tsk->fs_excl));
781 if (unlikely(in_interrupt()))
782 panic("Aiee, killing interrupt handler!");
783 if (unlikely(!tsk->pid))
784 panic("Attempted to kill the idle task!");
785 if (unlikely(tsk == child_reaper))
786 panic("Attempted to kill init!");
788 tracehook_report_exit(&code);
791 * We're taking recursive faults here in do_exit. Safest is to just
792 * leave this task alone and wait for reboot.
794 if (unlikely(tsk->flags & PF_EXITING)) {
796 "Fixing recursive fault but reboot is needed!\n");
799 set_current_state(TASK_UNINTERRUPTIBLE);
803 tsk->flags |= PF_EXITING;
807 if (unlikely(in_atomic()))
808 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
809 current->comm, current->pid,
812 taskstats_exit_alloc(&tidstats, &mycpu);
814 acct_update_integrals(tsk);
816 update_hiwater_rss(tsk->mm);
817 update_hiwater_vm(tsk->mm);
819 group_dead = atomic_dec_and_test(&tsk->signal->live);
821 hrtimer_cancel(&tsk->signal->real_timer);
822 exit_itimers(tsk->signal);
825 if (current->tux_info) {
826 #ifdef CONFIG_TUX_DEBUG
827 printk("Possibly unexpected TUX-thread exit(%ld) at %p?\n",
828 code, __builtin_return_address(0));
833 acct_collect(code, group_dead);
834 if (unlikely(tsk->robust_list))
835 exit_robust_list(tsk);
836 #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
837 if (unlikely(tsk->compat_robust_list))
838 compat_exit_robust_list(tsk);
840 if (unlikely(tsk->audit_context))
842 taskstats_exit_send(tsk, tidstats, group_dead, mycpu);
843 taskstats_exit_free(tidstats);
857 if (group_dead && tsk->signal->leader)
858 disassociate_ctty(1);
860 module_put(task_thread_info(tsk)->exec_domain->module);
862 module_put(tsk->binfmt->module);
864 tsk->exit_code = code;
865 proc_exit_connector(tsk);
866 /* needs to stay before exit_notify() */
867 exit_vx_info_early(tsk, code);
870 mpol_free(tsk->mempolicy);
871 tsk->mempolicy = NULL;
874 * This must happen late, after the PID is not
877 if (unlikely(!list_empty(&tsk->pi_state_list)))
878 exit_pi_state_list(tsk);
879 if (unlikely(current->pi_state_cache))
880 kfree(current->pi_state_cache);
882 * Make sure we are holding no locks:
884 debug_check_no_locks_held(tsk);
889 if (tsk->splice_pipe)
890 __free_pipe_info(tsk->splice_pipe);
892 /* needs to stay after exit_notify() */
893 exit_vx_info(tsk, code);
896 /* PF_DEAD causes final put_task_struct after we schedule. */
898 BUG_ON(tsk->flags & PF_DEAD);
899 tsk->flags |= PF_DEAD;
903 /* Avoid "noreturn function does return". */
907 EXPORT_SYMBOL_GPL(do_exit);
909 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
917 EXPORT_SYMBOL(complete_and_exit);
919 asmlinkage long sys_exit(int error_code)
921 do_exit((error_code&0xff)<<8);
925 * Take down every thread in the group. This is called by fatal signals
926 * as well as by sys_exit_group (below).
929 do_group_exit(int exit_code)
931 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
933 if (current->signal->flags & SIGNAL_GROUP_EXIT)
934 exit_code = current->signal->group_exit_code;
935 else if (!thread_group_empty(current)) {
936 struct signal_struct *const sig = current->signal;
937 struct sighand_struct *const sighand = current->sighand;
938 spin_lock_irq(&sighand->siglock);
939 if (sig->flags & SIGNAL_GROUP_EXIT)
940 /* Another thread got here before we took the lock. */
941 exit_code = sig->group_exit_code;
943 sig->group_exit_code = exit_code;
944 zap_other_threads(current);
946 spin_unlock_irq(&sighand->siglock);
954 * this kills every thread in the thread group. Note that any externally
955 * wait4()-ing process will get the correct exit code - even if this
956 * thread is not the thread group leader.
958 asmlinkage void sys_exit_group(int error_code)
960 do_group_exit((error_code & 0xff) << 8);
963 static int eligible_child(pid_t pid, int options, struct task_struct *p)
969 if (process_group(p) != process_group(current))
971 } else if (pid != -1) {
972 if (process_group(p) != -pid)
977 * Do not consider detached threads.
979 if (p->exit_signal == -1)
982 /* Wait for all children (clone and not) if __WALL is set;
983 * otherwise, wait for clone children *only* if __WCLONE is
984 * set; otherwise, wait for non-clone children *only*. (Note:
985 * A "clone" child here is one that reports to its parent
986 * using a signal other than SIGCHLD.) */
987 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
988 && !(options & __WALL))
991 * Do not consider thread group leaders that are
992 * in a non-empty thread group:
994 if (delay_group_leader(p))
997 if (security_task_wait(p))
1003 static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
1004 int why, int status,
1005 struct siginfo __user *infop,
1006 struct rusage __user *rusagep)
1008 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1012 retval = put_user(SIGCHLD, &infop->si_signo);
1014 retval = put_user(0, &infop->si_errno);
1016 retval = put_user((short)why, &infop->si_code);
1018 retval = put_user(pid, &infop->si_pid);
1020 retval = put_user(uid, &infop->si_uid);
1022 retval = put_user(status, &infop->si_status);
1029 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1030 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1031 * the lock and this task is uninteresting. If we return nonzero, we have
1032 * released the lock and the system call should return.
1034 static int wait_task_zombie(struct task_struct *p, int noreap,
1035 struct siginfo __user *infop,
1036 int __user *stat_addr, struct rusage __user *ru)
1038 unsigned long state;
1042 if (unlikely(noreap)) {
1045 int exit_code = p->exit_code;
1048 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1050 if (unlikely(p->exit_signal == -1))
1053 read_unlock(&tasklist_lock);
1054 if ((exit_code & 0x7f) == 0) {
1056 status = exit_code >> 8;
1058 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1059 status = exit_code & 0x7f;
1061 return wait_noreap_copyout(p, pid, uid, why,
1066 * Try to move the task's state to DEAD
1067 * only one thread is allowed to do this:
1069 state = xchg(&p->exit_state, EXIT_DEAD);
1070 if (state != EXIT_ZOMBIE) {
1071 BUG_ON(state != EXIT_DEAD);
1074 BUG_ON(p->exit_signal == -1);
1076 if (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;
1165 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1166 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1167 * the lock and this task is uninteresting. If we return nonzero, we have
1168 * released the lock and the system call should return.
1170 static int wait_task_stopped(struct task_struct *p, int delayed_group_leader,
1171 int noreap, struct siginfo __user *infop,
1172 int __user *stat_addr, struct rusage __user *ru)
1174 int retval, exit_code;
1178 if (delayed_group_leader &&
1179 p->signal && p->signal->group_stop_count > 0)
1181 * A group stop is in progress and this is the group leader.
1182 * We won't report until all threads have stopped.
1187 * Now we are pretty sure this task is interesting.
1188 * Make sure it doesn't get reaped out from under us while we
1189 * give up the lock and then examine it below. We don't want to
1190 * keep holding onto the tasklist_lock while we call getrusage and
1191 * possibly take page faults for user memory.
1194 read_unlock(&tasklist_lock);
1196 if (unlikely(noreap)) {
1200 exit_code = p->exit_code;
1201 if (unlikely(!exit_code) ||
1202 unlikely(p->state & TASK_TRACED))
1204 return wait_noreap_copyout(p, pid, uid, CLD_STOPPED,
1205 (exit_code << 8) | 0x7f,
1209 write_lock_irq(&tasklist_lock);
1212 * This uses xchg to be atomic with the thread resuming and setting
1213 * it. It must also be done with the write lock held to prevent a
1214 * race with the EXIT_ZOMBIE case.
1216 exit_code = xchg(&p->exit_code, 0);
1217 if (unlikely(p->exit_state)) {
1219 * The task resumed and then died. Let the next iteration
1220 * catch it in EXIT_ZOMBIE. Note that exit_code might
1221 * already be zero here if it resumed and did _exit(0).
1222 * The task itself is dead and won't touch exit_code again;
1223 * other processors in this function are locked out.
1225 p->exit_code = exit_code;
1228 if (unlikely(exit_code == 0)) {
1230 * Another thread in this function got to it first, or it
1231 * resumed, or it resumed and then died.
1233 write_unlock_irq(&tasklist_lock);
1237 * We are returning to the wait loop without having successfully
1238 * removed the process and having released the lock. We cannot
1239 * continue, since the "p" task pointer is potentially stale.
1241 * Return -EAGAIN, and do_wait() will restart the loop from the
1242 * beginning. Do _not_ re-acquire the lock.
1247 /* move to end of parent's list to avoid starvation */
1251 write_unlock_irq(&tasklist_lock);
1253 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1254 if (!retval && stat_addr)
1255 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1256 if (!retval && infop)
1257 retval = put_user(SIGCHLD, &infop->si_signo);
1258 if (!retval && infop)
1259 retval = put_user(0, &infop->si_errno);
1260 if (!retval && infop)
1261 retval = put_user((short)CLD_STOPPED, &infop->si_code);
1262 if (!retval && infop)
1263 retval = put_user(exit_code, &infop->si_status);
1264 if (!retval && infop)
1265 retval = put_user(p->pid, &infop->si_pid);
1266 if (!retval && infop)
1267 retval = put_user(p->uid, &infop->si_uid);
1277 * Handle do_wait work for one task in a live, non-stopped state.
1278 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1279 * the lock and this task is uninteresting. If we return nonzero, we have
1280 * released the lock and the system call should return.
1282 static int wait_task_continued(struct task_struct *p, int noreap,
1283 struct siginfo __user *infop,
1284 int __user *stat_addr, struct rusage __user *ru)
1290 if (unlikely(!p->signal))
1293 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1296 spin_lock_irq(&p->sighand->siglock);
1297 /* Re-check with the lock held. */
1298 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1299 spin_unlock_irq(&p->sighand->siglock);
1303 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1304 spin_unlock_irq(&p->sighand->siglock);
1309 read_unlock(&tasklist_lock);
1312 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1314 if (!retval && stat_addr)
1315 retval = put_user(0xffff, stat_addr);
1319 retval = wait_noreap_copyout(p, pid, uid,
1320 CLD_CONTINUED, SIGCONT,
1322 BUG_ON(retval == 0);
1329 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1330 int __user *stat_addr, struct rusage __user *ru)
1332 DECLARE_WAITQUEUE(wait, current);
1333 struct task_struct *tsk;
1336 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1339 * We will set this flag if we see any child that might later
1340 * match our criteria, even if we are not able to reap it yet.
1343 current->state = TASK_INTERRUPTIBLE;
1344 read_lock(&tasklist_lock);
1347 struct task_struct *p;
1348 struct list_head *_p;
1351 list_for_each(_p,&tsk->children) {
1352 p = list_entry(_p, struct task_struct, sibling);
1354 ret = eligible_child(pid, options, p);
1364 * It's stopped now, so it might later
1365 * continue, exit, or stop again.
1368 if (!(options & WUNTRACED))
1370 if (tracehook_inhibit_wait_stopped(p))
1372 retval = wait_task_stopped(p, ret == 2,
1373 (options & WNOWAIT),
1376 if (retval == -EAGAIN)
1378 if (retval != 0) /* He released the lock. */
1383 if (p->exit_state == EXIT_DEAD)
1385 // case EXIT_ZOMBIE:
1386 if (p->exit_state == EXIT_ZOMBIE) {
1388 * Eligible but we cannot release
1392 goto check_continued;
1393 if (!likely(options & WEXITED))
1395 if (tracehook_inhibit_wait_zombie(p)) {
1399 retval = wait_task_zombie(
1400 p, (options & WNOWAIT),
1401 infop, stat_addr, ru);
1402 /* He released the lock. */
1409 * It's running now, so it might later
1410 * exit, stop, or stop and then continue.
1413 if (!unlikely(options & WCONTINUED))
1415 if (tracehook_inhibit_wait_continued(p))
1417 retval = wait_task_continued(
1418 p, (options & WNOWAIT),
1419 infop, stat_addr, ru);
1420 if (retval != 0) /* He released the lock. */
1426 retval = ptrace_do_wait(tsk, pid, options,
1427 infop, stat_addr, ru);
1428 if (retval != -ECHILD) {
1430 if (retval != 0) /* He released the lock. */
1434 if (options & __WNOTHREAD)
1436 tsk = next_thread(tsk);
1437 BUG_ON(tsk->signal != current->signal);
1438 } while (tsk != current);
1440 read_unlock(&tasklist_lock);
1443 if (options & WNOHANG)
1445 retval = -ERESTARTSYS;
1446 if (signal_pending(current))
1453 current->state = TASK_RUNNING;
1454 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1460 * For a WNOHANG return, clear out all the fields
1461 * we would set so the user can easily tell the
1465 retval = put_user(0, &infop->si_signo);
1467 retval = put_user(0, &infop->si_errno);
1469 retval = put_user(0, &infop->si_code);
1471 retval = put_user(0, &infop->si_pid);
1473 retval = put_user(0, &infop->si_uid);
1475 retval = put_user(0, &infop->si_status);
1481 asmlinkage long sys_waitid(int which, pid_t pid,
1482 struct siginfo __user *infop, int options,
1483 struct rusage __user *ru)
1487 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1489 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1509 ret = do_wait(pid, options, infop, NULL, ru);
1511 /* avoid REGPARM breakage on x86: */
1512 prevent_tail_call(ret);
1516 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1517 int options, struct rusage __user *ru)
1521 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1522 __WNOTHREAD|__WCLONE|__WALL))
1524 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1526 /* avoid REGPARM breakage on x86: */
1527 prevent_tail_call(ret);
1531 #ifdef __ARCH_WANT_SYS_WAITPID
1534 * sys_waitpid() remains for compatibility. waitpid() should be
1535 * implemented by calling sys_wait4() from libc.a.
1537 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1539 return sys_wait4(pid, stat_addr, options, NULL);