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 proc_clear_tty(current);
390 /* Block and flush all signals */
391 sigfillset(&blocked);
392 sigprocmask(SIG_BLOCK, &blocked, NULL);
393 flush_signals(current);
395 /* Become as one with the init task */
397 exit_fs(current); /* current->fs->count--; */
400 atomic_inc(&fs->count);
401 exit_namespace(current);
402 current->namespace = init_task.namespace;
403 get_namespace(current->namespace);
405 current->files = init_task.files;
406 atomic_inc(¤t->files->count);
411 EXPORT_SYMBOL(daemonize);
413 static void close_files(struct files_struct * files)
421 * It is safe to dereference the fd table without RCU or
422 * ->file_lock because this is the last reference to the
425 fdt = files_fdtable(files);
429 if (i >= fdt->max_fdset || i >= fdt->max_fds)
431 set = fdt->open_fds->fds_bits[j++];
434 struct file * file = xchg(&fdt->fd[i], NULL);
436 filp_close(file, files);
446 struct files_struct *get_files_struct(struct task_struct *task)
448 struct files_struct *files;
453 atomic_inc(&files->count);
459 void fastcall put_files_struct(struct files_struct *files)
463 if (atomic_dec_and_test(&files->count)) {
466 * Free the fd and fdset arrays if we expanded them.
467 * If the fdtable was embedded, pass files for freeing
468 * at the end of the RCU grace period. Otherwise,
469 * you can free files immediately.
471 fdt = files_fdtable(files);
472 if (fdt == &files->fdtab)
473 fdt->free_files = files;
475 kmem_cache_free(files_cachep, files);
480 EXPORT_SYMBOL(put_files_struct);
482 static inline void __exit_files(struct task_struct *tsk)
484 struct files_struct * files = tsk->files;
490 put_files_struct(files);
494 void exit_files(struct task_struct *tsk)
499 static inline void __put_fs_struct(struct fs_struct *fs)
501 /* No need to hold fs->lock if we are killing it */
502 if (atomic_dec_and_test(&fs->count)) {
509 mntput(fs->altrootmnt);
511 kmem_cache_free(fs_cachep, fs);
515 void put_fs_struct(struct fs_struct *fs)
520 static inline void __exit_fs(struct task_struct *tsk)
522 struct fs_struct * fs = tsk->fs;
532 void exit_fs(struct task_struct *tsk)
537 EXPORT_SYMBOL_GPL(exit_fs);
540 * Turn us into a lazy TLB process if we
543 static void exit_mm(struct task_struct * tsk)
545 struct mm_struct *mm = tsk->mm;
551 * Serialize with any possible pending coredump.
552 * We must hold mmap_sem around checking core_waiters
553 * and clearing tsk->mm. The core-inducing thread
554 * will increment core_waiters for each thread in the
555 * group with ->mm != NULL.
557 down_read(&mm->mmap_sem);
558 if (mm->core_waiters) {
559 up_read(&mm->mmap_sem);
560 down_write(&mm->mmap_sem);
561 if (!--mm->core_waiters)
562 complete(mm->core_startup_done);
563 up_write(&mm->mmap_sem);
565 wait_for_completion(&mm->core_done);
566 down_read(&mm->mmap_sem);
568 atomic_inc(&mm->mm_count);
569 BUG_ON(mm != tsk->active_mm);
570 /* more a memory barrier than a real lock */
573 up_read(&mm->mmap_sem);
574 enter_lazy_tlb(mm, current);
580 choose_new_parent(struct task_struct *p, struct task_struct *reaper)
582 /* check for reaper context */
583 vxwprintk((p->xid != reaper->xid) && (reaper != child_reaper),
584 "rogue reaper: %p[%d,#%u] <> %p[%d,#%u]",
585 p, p->pid, p->xid, reaper, reaper->pid, reaper->xid);
588 * Make sure we're not reparenting to ourselves and that
589 * the parent is not a zombie.
591 BUG_ON(p == reaper || reaper->exit_state);
596 reparent_thread(struct task_struct *p, struct task_struct *father)
598 /* We don't want people slaying init. */
599 if (p->exit_signal != -1)
600 p->exit_signal = SIGCHLD;
602 if (p->pdeath_signal)
603 /* We already hold the tasklist_lock here. */
604 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
606 /* Move the child from its dying parent to the new one. */
607 list_move_tail(&p->sibling, &p->parent->children);
609 /* If we'd notified the old parent about this child's death,
610 * also notify the new parent.
612 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
613 thread_group_empty(p))
614 do_notify_parent(p, p->exit_signal);
617 * process group orphan check
618 * Case ii: Our child is in a different pgrp
619 * than we are, and it was the only connection
620 * outside, so the child pgrp is now orphaned.
622 if ((process_group(p) != process_group(father)) &&
623 (p->signal->session == father->signal->session)) {
624 int pgrp = process_group(p);
626 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
627 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
628 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
634 * When we die, we re-parent all our children.
635 * Try to give them to another thread in our thread
636 * group, and if no such member exists, give it to
637 * the global child reaper process (ie "init")
640 forget_original_parent(struct task_struct *father)
642 struct task_struct *p, *reaper = father;
643 struct list_head *_p, *_n;
646 reaper = next_thread(reaper);
647 if (reaper == father) {
648 reaper = vx_child_reaper(father);
651 } while (reaper->exit_state);
653 list_for_each_safe(_p, _n, &father->children) {
654 p = list_entry(_p, struct task_struct, sibling);
655 choose_new_parent(p, vx_child_reaper(p));
656 reparent_thread(p, father);
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;
671 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
672 && !thread_group_empty(tsk)) {
674 * This occurs when there was a race between our exit
675 * syscall and a group signal choosing us as the one to
676 * wake up. It could be that we are the only thread
677 * alerted to check for pending signals, but another thread
678 * should be woken now to take the signal since we will not.
679 * Now we'll wake all the threads in the group just to make
680 * sure someone gets all the pending signals.
682 read_lock(&tasklist_lock);
683 spin_lock_irq(&tsk->sighand->siglock);
684 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
685 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
686 recalc_sigpending_tsk(t);
687 if (signal_pending(t))
688 signal_wake_up(t, 0);
690 spin_unlock_irq(&tsk->sighand->siglock);
691 read_unlock(&tasklist_lock);
694 write_lock_irq(&tasklist_lock);
697 * This does two things:
699 * A. Make init inherit all the child processes
700 * B. Check to see if any process groups have become orphaned
701 * as a result of our exiting, and if they have any stopped
702 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
705 forget_original_parent(tsk);
706 BUG_ON(!list_empty(&tsk->children));
709 * Check to see if any process groups have become orphaned
710 * as a result of our exiting, and if they have any stopped
711 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
713 * Case i: Our father is in a different pgrp than we are
714 * and we were the only connection outside, so our pgrp
715 * is about to become orphaned.
720 if ((process_group(t) != process_group(tsk)) &&
721 (t->signal->session == tsk->signal->session) &&
722 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
723 has_stopped_jobs(process_group(tsk))) {
724 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
725 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
728 /* Let father know we died
730 * Thread signals are configurable, but you aren't going to use
731 * that to send signals to arbitary processes.
732 * That stops right now.
734 * If the parent exec id doesn't match the exec id we saved
735 * when we started then we know the parent has changed security
738 * If our self_exec id doesn't match our parent_exec_id then
739 * we have changed execution domain as these two values started
740 * the same after a fork.
744 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
745 ( tsk->parent_exec_id != t->self_exec_id ||
746 tsk->self_exec_id != tsk->parent_exec_id)
747 && !capable(CAP_KILL))
748 tsk->exit_signal = SIGCHLD;
750 if (!tracehook_notify_death(tsk, &noreap, &cookie)
751 && tsk->exit_signal != -1 && thread_group_empty(tsk))
752 do_notify_parent(tsk, tsk->exit_signal);
755 if (tsk->exit_signal == -1 && !noreap)
757 tsk->exit_state = state;
759 write_unlock_irq(&tasklist_lock);
761 tracehook_report_death(tsk, state, cookie);
763 /* If the process is dead, release it - nobody will wait for it */
764 if (state == EXIT_DEAD)
768 fastcall NORET_TYPE void do_exit(long code)
770 struct task_struct *tsk = current;
771 struct taskstats *tidstats;
775 profile_task_exit(tsk);
777 WARN_ON(atomic_read(&tsk->fs_excl));
779 if (unlikely(in_interrupt()))
780 panic("Aiee, killing interrupt handler!");
781 if (unlikely(!tsk->pid))
782 panic("Attempted to kill the idle task!");
783 if (unlikely(tsk == child_reaper))
784 panic("Attempted to kill init!");
786 tracehook_report_exit(&code);
789 * We're taking recursive faults here in do_exit. Safest is to just
790 * leave this task alone and wait for reboot.
792 if (unlikely(tsk->flags & PF_EXITING)) {
794 "Fixing recursive fault but reboot is needed!\n");
797 set_current_state(TASK_UNINTERRUPTIBLE);
801 tsk->flags |= PF_EXITING;
805 if (unlikely(in_atomic()))
806 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
807 current->comm, current->pid,
810 taskstats_exit_alloc(&tidstats, &mycpu);
812 acct_update_integrals(tsk);
814 update_hiwater_rss(tsk->mm);
815 update_hiwater_vm(tsk->mm);
817 group_dead = atomic_dec_and_test(&tsk->signal->live);
819 hrtimer_cancel(&tsk->signal->real_timer);
820 exit_itimers(tsk->signal);
823 if (current->tux_info) {
824 #ifdef CONFIG_TUX_DEBUG
825 printk("Possibly unexpected TUX-thread exit(%ld) at %p?\n",
826 code, __builtin_return_address(0));
831 acct_collect(code, group_dead);
832 if (unlikely(tsk->robust_list))
833 exit_robust_list(tsk);
834 #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
835 if (unlikely(tsk->compat_robust_list))
836 compat_exit_robust_list(tsk);
838 if (unlikely(tsk->audit_context))
840 taskstats_exit_send(tsk, tidstats, group_dead, mycpu);
841 taskstats_exit_free(tidstats);
855 if (group_dead && tsk->signal->leader)
856 disassociate_ctty(1);
858 module_put(task_thread_info(tsk)->exec_domain->module);
860 module_put(tsk->binfmt->module);
862 tsk->exit_code = code;
863 proc_exit_connector(tsk);
864 /* needs to stay before exit_notify() */
865 exit_vx_info_early(tsk, code);
868 mpol_free(tsk->mempolicy);
869 tsk->mempolicy = NULL;
872 * This must happen late, after the PID is not
875 if (unlikely(!list_empty(&tsk->pi_state_list)))
876 exit_pi_state_list(tsk);
877 if (unlikely(current->pi_state_cache))
878 kfree(current->pi_state_cache);
880 * Make sure we are holding no locks:
882 debug_check_no_locks_held(tsk);
887 if (tsk->splice_pipe)
888 __free_pipe_info(tsk->splice_pipe);
890 /* needs to stay after exit_notify() */
891 exit_vx_info(tsk, code);
894 /* PF_DEAD causes final put_task_struct after we schedule. */
896 BUG_ON(tsk->flags & PF_DEAD);
897 tsk->flags |= PF_DEAD;
901 /* Avoid "noreturn function does return". */
905 EXPORT_SYMBOL_GPL(do_exit);
907 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
915 EXPORT_SYMBOL(complete_and_exit);
917 asmlinkage long sys_exit(int error_code)
919 do_exit((error_code&0xff)<<8);
923 * Take down every thread in the group. This is called by fatal signals
924 * as well as by sys_exit_group (below).
927 do_group_exit(int exit_code)
929 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
931 if (current->signal->flags & SIGNAL_GROUP_EXIT)
932 exit_code = current->signal->group_exit_code;
933 else if (!thread_group_empty(current)) {
934 struct signal_struct *const sig = current->signal;
935 struct sighand_struct *const sighand = current->sighand;
936 spin_lock_irq(&sighand->siglock);
937 if (sig->flags & SIGNAL_GROUP_EXIT)
938 /* Another thread got here before we took the lock. */
939 exit_code = sig->group_exit_code;
941 sig->group_exit_code = exit_code;
942 zap_other_threads(current);
944 spin_unlock_irq(&sighand->siglock);
952 * this kills every thread in the thread group. Note that any externally
953 * wait4()-ing process will get the correct exit code - even if this
954 * thread is not the thread group leader.
956 asmlinkage void sys_exit_group(int error_code)
958 do_group_exit((error_code & 0xff) << 8);
961 static int eligible_child(pid_t pid, int options, struct task_struct *p)
967 if (process_group(p) != process_group(current))
969 } else if (pid != -1) {
970 if (process_group(p) != -pid)
975 * Do not consider detached threads.
977 if (p->exit_signal == -1)
980 /* Wait for all children (clone and not) if __WALL is set;
981 * otherwise, wait for clone children *only* if __WCLONE is
982 * set; otherwise, wait for non-clone children *only*. (Note:
983 * A "clone" child here is one that reports to its parent
984 * using a signal other than SIGCHLD.) */
985 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
986 && !(options & __WALL))
989 * Do not consider thread group leaders that are
990 * in a non-empty thread group:
992 if (delay_group_leader(p))
995 if (security_task_wait(p))
1001 static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
1002 int why, int status,
1003 struct siginfo __user *infop,
1004 struct rusage __user *rusagep)
1006 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1010 retval = put_user(SIGCHLD, &infop->si_signo);
1012 retval = put_user(0, &infop->si_errno);
1014 retval = put_user((short)why, &infop->si_code);
1016 retval = put_user(pid, &infop->si_pid);
1018 retval = put_user(uid, &infop->si_uid);
1020 retval = put_user(status, &infop->si_status);
1027 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1028 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1029 * the lock and this task is uninteresting. If we return nonzero, we have
1030 * released the lock and the system call should return.
1032 static int wait_task_zombie(struct task_struct *p, int noreap,
1033 struct siginfo __user *infop,
1034 int __user *stat_addr, struct rusage __user *ru)
1036 unsigned long state;
1040 if (unlikely(noreap)) {
1043 int exit_code = p->exit_code;
1046 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1048 if (unlikely(p->exit_signal == -1))
1051 read_unlock(&tasklist_lock);
1052 if ((exit_code & 0x7f) == 0) {
1054 status = exit_code >> 8;
1056 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1057 status = exit_code & 0x7f;
1059 return wait_noreap_copyout(p, pid, uid, why,
1064 * Try to move the task's state to DEAD
1065 * only one thread is allowed to do this:
1067 state = xchg(&p->exit_state, EXIT_DEAD);
1068 if (state != EXIT_ZOMBIE) {
1069 BUG_ON(state != EXIT_DEAD);
1072 BUG_ON(p->exit_signal == -1);
1074 if (likely(p->signal)) {
1075 struct signal_struct *psig;
1076 struct signal_struct *sig;
1079 * The resource counters for the group leader are in its
1080 * own task_struct. Those for dead threads in the group
1081 * are in its signal_struct, as are those for the child
1082 * processes it has previously reaped. All these
1083 * accumulate in the parent's signal_struct c* fields.
1085 * We don't bother to take a lock here to protect these
1086 * p->signal fields, because they are only touched by
1087 * __exit_signal, which runs with tasklist_lock
1088 * write-locked anyway, and so is excluded here. We do
1089 * need to protect the access to p->parent->signal fields,
1090 * as other threads in the parent group can be right
1091 * here reaping other children at the same time.
1093 spin_lock_irq(&p->parent->sighand->siglock);
1094 psig = p->parent->signal;
1097 cputime_add(psig->cutime,
1098 cputime_add(p->utime,
1099 cputime_add(sig->utime,
1102 cputime_add(psig->cstime,
1103 cputime_add(p->stime,
1104 cputime_add(sig->stime,
1107 p->min_flt + sig->min_flt + sig->cmin_flt;
1109 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1111 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1113 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1114 spin_unlock_irq(&p->parent->sighand->siglock);
1118 * Now we are sure this task is interesting, and no other
1119 * thread can reap it because we set its state to EXIT_DEAD.
1121 read_unlock(&tasklist_lock);
1123 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1124 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1125 ? p->signal->group_exit_code : p->exit_code;
1126 if (!retval && stat_addr)
1127 retval = put_user(status, stat_addr);
1128 if (!retval && infop)
1129 retval = put_user(SIGCHLD, &infop->si_signo);
1130 if (!retval && infop)
1131 retval = put_user(0, &infop->si_errno);
1132 if (!retval && infop) {
1135 if ((status & 0x7f) == 0) {
1139 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1142 retval = put_user((short)why, &infop->si_code);
1144 retval = put_user(status, &infop->si_status);
1146 if (!retval && infop)
1147 retval = put_user(p->pid, &infop->si_pid);
1148 if (!retval && infop)
1149 retval = put_user(p->uid, &infop->si_uid);
1151 // TODO: is this safe?
1152 p->exit_state = EXIT_ZOMBIE;
1163 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1164 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1165 * the lock and this task is uninteresting. If we return nonzero, we have
1166 * released the lock and the system call should return.
1168 static int wait_task_stopped(struct task_struct *p, int delayed_group_leader,
1169 int noreap, struct siginfo __user *infop,
1170 int __user *stat_addr, struct rusage __user *ru)
1172 int retval, exit_code;
1176 if (delayed_group_leader &&
1177 p->signal && p->signal->group_stop_count > 0)
1179 * A group stop is in progress and this is the group leader.
1180 * We won't report until all threads have stopped.
1185 * Now we are pretty sure this task is interesting.
1186 * Make sure it doesn't get reaped out from under us while we
1187 * give up the lock and then examine it below. We don't want to
1188 * keep holding onto the tasklist_lock while we call getrusage and
1189 * possibly take page faults for user memory.
1192 read_unlock(&tasklist_lock);
1194 if (unlikely(noreap)) {
1198 exit_code = p->exit_code;
1199 if (unlikely(!exit_code) ||
1200 unlikely(p->state & TASK_TRACED))
1202 return wait_noreap_copyout(p, pid, uid, CLD_STOPPED,
1203 (exit_code << 8) | 0x7f,
1207 write_lock_irq(&tasklist_lock);
1210 * This uses xchg to be atomic with the thread resuming and setting
1211 * it. It must also be done with the write lock held to prevent a
1212 * race with the EXIT_ZOMBIE case.
1214 exit_code = xchg(&p->exit_code, 0);
1215 if (unlikely(p->exit_state)) {
1217 * The task resumed and then died. Let the next iteration
1218 * catch it in EXIT_ZOMBIE. Note that exit_code might
1219 * already be zero here if it resumed and did _exit(0).
1220 * The task itself is dead and won't touch exit_code again;
1221 * other processors in this function are locked out.
1223 p->exit_code = exit_code;
1226 if (unlikely(exit_code == 0)) {
1228 * Another thread in this function got to it first, or it
1229 * resumed, or it resumed and then died.
1231 write_unlock_irq(&tasklist_lock);
1235 * We are returning to the wait loop without having successfully
1236 * removed the process and having released the lock. We cannot
1237 * continue, since the "p" task pointer is potentially stale.
1239 * Return -EAGAIN, and do_wait() will restart the loop from the
1240 * beginning. Do _not_ re-acquire the lock.
1245 /* move to end of parent's list to avoid starvation */
1249 write_unlock_irq(&tasklist_lock);
1251 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1252 if (!retval && stat_addr)
1253 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1254 if (!retval && infop)
1255 retval = put_user(SIGCHLD, &infop->si_signo);
1256 if (!retval && infop)
1257 retval = put_user(0, &infop->si_errno);
1258 if (!retval && infop)
1259 retval = put_user((short)CLD_STOPPED, &infop->si_code);
1260 if (!retval && infop)
1261 retval = put_user(exit_code, &infop->si_status);
1262 if (!retval && infop)
1263 retval = put_user(p->pid, &infop->si_pid);
1264 if (!retval && infop)
1265 retval = put_user(p->uid, &infop->si_uid);
1275 * Handle do_wait work for one task in a live, non-stopped state.
1276 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1277 * the lock and this task is uninteresting. If we return nonzero, we have
1278 * released the lock and the system call should return.
1280 static int wait_task_continued(struct task_struct *p, int noreap,
1281 struct siginfo __user *infop,
1282 int __user *stat_addr, struct rusage __user *ru)
1288 if (unlikely(!p->signal))
1291 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1294 spin_lock_irq(&p->sighand->siglock);
1295 /* Re-check with the lock held. */
1296 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1297 spin_unlock_irq(&p->sighand->siglock);
1301 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1302 spin_unlock_irq(&p->sighand->siglock);
1307 read_unlock(&tasklist_lock);
1310 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1312 if (!retval && stat_addr)
1313 retval = put_user(0xffff, stat_addr);
1317 retval = wait_noreap_copyout(p, pid, uid,
1318 CLD_CONTINUED, SIGCONT,
1320 BUG_ON(retval == 0);
1327 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1328 int __user *stat_addr, struct rusage __user *ru)
1330 DECLARE_WAITQUEUE(wait, current);
1331 struct task_struct *tsk;
1334 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1337 * We will set this flag if we see any child that might later
1338 * match our criteria, even if we are not able to reap it yet.
1341 current->state = TASK_INTERRUPTIBLE;
1342 read_lock(&tasklist_lock);
1345 struct task_struct *p;
1346 struct list_head *_p;
1349 list_for_each(_p,&tsk->children) {
1350 p = list_entry(_p, struct task_struct, sibling);
1352 ret = eligible_child(pid, options, p);
1362 * It's stopped now, so it might later
1363 * continue, exit, or stop again.
1366 if (!(options & WUNTRACED))
1368 if (tracehook_inhibit_wait_stopped(p))
1370 retval = wait_task_stopped(p, ret == 2,
1371 (options & WNOWAIT),
1374 if (retval == -EAGAIN)
1376 if (retval != 0) /* He released the lock. */
1381 if (p->exit_state == EXIT_DEAD)
1383 // case EXIT_ZOMBIE:
1384 if (p->exit_state == EXIT_ZOMBIE) {
1386 * Eligible but we cannot release
1390 goto check_continued;
1391 if (!likely(options & WEXITED))
1393 if (tracehook_inhibit_wait_zombie(p)) {
1397 retval = wait_task_zombie(
1398 p, (options & WNOWAIT),
1399 infop, stat_addr, ru);
1400 /* He released the lock. */
1407 * It's running now, so it might later
1408 * exit, stop, or stop and then continue.
1411 if (!unlikely(options & WCONTINUED))
1413 if (tracehook_inhibit_wait_continued(p))
1415 retval = wait_task_continued(
1416 p, (options & WNOWAIT),
1417 infop, stat_addr, ru);
1418 if (retval != 0) /* He released the lock. */
1424 retval = ptrace_do_wait(tsk, pid, options,
1425 infop, stat_addr, ru);
1426 if (retval != -ECHILD) {
1428 if (retval != 0) /* He released the lock. */
1432 if (options & __WNOTHREAD)
1434 tsk = next_thread(tsk);
1435 BUG_ON(tsk->signal != current->signal);
1436 } while (tsk != current);
1438 read_unlock(&tasklist_lock);
1441 if (options & WNOHANG)
1443 retval = -ERESTARTSYS;
1444 if (signal_pending(current))
1451 current->state = TASK_RUNNING;
1452 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1458 * For a WNOHANG return, clear out all the fields
1459 * we would set so the user can easily tell the
1463 retval = put_user(0, &infop->si_signo);
1465 retval = put_user(0, &infop->si_errno);
1467 retval = put_user(0, &infop->si_code);
1469 retval = put_user(0, &infop->si_pid);
1471 retval = put_user(0, &infop->si_uid);
1473 retval = put_user(0, &infop->si_status);
1479 asmlinkage long sys_waitid(int which, pid_t pid,
1480 struct siginfo __user *infop, int options,
1481 struct rusage __user *ru)
1485 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1487 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1507 ret = do_wait(pid, options, infop, NULL, ru);
1509 /* avoid REGPARM breakage on x86: */
1510 prevent_tail_call(ret);
1514 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1515 int options, struct rusage __user *ru)
1519 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1520 __WNOTHREAD|__WCLONE|__WALL))
1522 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1524 /* avoid REGPARM breakage on x86: */
1525 prevent_tail_call(ret);
1529 #ifdef __ARCH_WANT_SYS_WAITPID
1532 * sys_waitpid() remains for compatibility. waitpid() should be
1533 * implemented by calling sys_wait4() from libc.a.
1535 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1537 return sys_wait4(pid, stat_addr, options, NULL);