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/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/profile.h>
25 #include <linux/mount.h>
26 #include <linux/proc_fs.h>
27 #include <linux/mempolicy.h>
28 #include <linux/syscalls.h>
29 #include <linux/vs_limit.h>
31 #include <asm/uaccess.h>
32 #include <asm/unistd.h>
33 #include <asm/pgtable.h>
34 #include <asm/mmu_context.h>
36 extern void sem_exit (void);
37 extern struct task_struct *child_reaper;
39 int getrusage(struct task_struct *, int, struct rusage __user *);
41 static void __unhash_process(struct task_struct *p)
44 /* tasklist_lock is held, is this sufficient? */
46 atomic_dec(&p->vx_info->cvirt.nr_threads);
49 detach_pid(p, PIDTYPE_PID);
50 detach_pid(p, PIDTYPE_TGID);
51 if (thread_group_leader(p)) {
52 detach_pid(p, PIDTYPE_PGID);
53 detach_pid(p, PIDTYPE_SID);
55 __get_cpu_var(process_counts)--;
61 void release_task(struct task_struct * p)
65 struct dentry *proc_dentry;
68 atomic_dec(&p->user->processes);
69 spin_lock(&p->proc_lock);
70 proc_dentry = proc_pid_unhash(p);
71 write_lock_irq(&tasklist_lock);
72 if (unlikely(p->ptrace))
74 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
80 * If we are the last non-leader member of the thread
81 * group, and the leader is zombie, then notify the
82 * group leader's parent process. (if it wants notification.)
85 leader = p->group_leader;
86 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
87 BUG_ON(leader->exit_signal == -1);
88 do_notify_parent(leader, leader->exit_signal);
90 * If we were the last child thread and the leader has
91 * exited already, and the leader's parent ignores SIGCHLD,
92 * then we are the one who should release the leader.
94 * do_notify_parent() will have marked it self-reaping in
97 zap_leader = (leader->exit_signal == -1);
101 write_unlock_irq(&tasklist_lock);
102 spin_unlock(&p->proc_lock);
103 proc_pid_flush(proc_dentry);
106 release_vx_info(p->vx_info, p);
110 if (unlikely(zap_leader))
114 /* we are using it only for SMP init */
116 void unhash_process(struct task_struct *p)
118 struct dentry *proc_dentry;
120 spin_lock(&p->proc_lock);
121 proc_dentry = proc_pid_unhash(p);
122 write_lock_irq(&tasklist_lock);
124 write_unlock_irq(&tasklist_lock);
125 spin_unlock(&p->proc_lock);
126 proc_pid_flush(proc_dentry);
130 * This checks not only the pgrp, but falls back on the pid if no
131 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
134 int session_of_pgrp(int pgrp)
136 struct task_struct *p;
139 read_lock(&tasklist_lock);
140 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
141 if (p->signal->session > 0) {
142 sid = p->signal->session;
145 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
146 p = find_task_by_pid(pgrp);
148 sid = p->signal->session;
150 read_unlock(&tasklist_lock);
156 * Determine if a process group is "orphaned", according to the POSIX
157 * definition in 2.2.2.52. Orphaned process groups are not to be affected
158 * by terminal-generated stop signals. Newly orphaned process groups are
159 * to receive a SIGHUP and a SIGCONT.
161 * "I ask you, have you ever known what it is to be an orphan?"
163 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
165 struct task_struct *p;
168 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
169 if (p == ignored_task
171 || p->real_parent->pid == 1)
173 if (process_group(p->real_parent) != pgrp
174 && p->real_parent->signal->session == p->signal->session) {
178 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
179 return ret; /* (sighing) "Often!" */
182 int is_orphaned_pgrp(int pgrp)
186 read_lock(&tasklist_lock);
187 retval = will_become_orphaned_pgrp(pgrp, NULL);
188 read_unlock(&tasklist_lock);
193 static inline int has_stopped_jobs(int pgrp)
196 struct task_struct *p;
198 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
199 if (p->state != TASK_STOPPED)
202 /* If p is stopped by a debugger on a signal that won't
203 stop it, then don't count p as stopped. This isn't
204 perfect but it's a good approximation. */
205 if (unlikely (p->ptrace)
206 && p->exit_code != SIGSTOP
207 && p->exit_code != SIGTSTP
208 && p->exit_code != SIGTTOU
209 && p->exit_code != SIGTTIN)
214 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
219 * reparent_to_init() - Reparent the calling kernel thread to the init task.
221 * If a kernel thread is launched as a result of a system call, or if
222 * it ever exits, it should generally reparent itself to init so that
223 * it is correctly cleaned up on exit.
225 * The various task state such as scheduling policy and priority may have
226 * been inherited from a user process, so we reset them to sane values here.
228 * NOTE that reparent_to_init() gives the caller full capabilities.
230 void reparent_to_init(void)
232 write_lock_irq(&tasklist_lock);
234 ptrace_unlink(current);
235 /* Reparent to init */
236 REMOVE_LINKS(current);
237 /* FIXME handle vchild_reaper/initpid */
238 current->parent = child_reaper;
239 current->real_parent = child_reaper;
242 /* Set the exit signal to SIGCHLD so we signal init on exit */
243 current->exit_signal = SIGCHLD;
245 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
246 set_user_nice(current, 0);
250 security_task_reparent_to_init(current);
251 memcpy(current->signal->rlim, init_task.signal->rlim,
252 sizeof(current->signal->rlim));
253 atomic_inc(&(INIT_USER->__count));
254 write_unlock_irq(&tasklist_lock);
255 switch_uid(INIT_USER);
258 void __set_special_pids(pid_t session, pid_t pgrp)
260 struct task_struct *curr = current;
262 if (curr->signal->session != session) {
263 detach_pid(curr, PIDTYPE_SID);
264 curr->signal->session = session;
265 attach_pid(curr, PIDTYPE_SID, session);
267 if (process_group(curr) != pgrp) {
268 detach_pid(curr, PIDTYPE_PGID);
269 curr->signal->pgrp = pgrp;
270 attach_pid(curr, PIDTYPE_PGID, pgrp);
274 void set_special_pids(pid_t session, pid_t pgrp)
276 write_lock_irq(&tasklist_lock);
277 __set_special_pids(session, pgrp);
278 write_unlock_irq(&tasklist_lock);
282 * Let kernel threads use this to say that they
283 * allow a certain signal (since daemonize() will
284 * have disabled all of them by default).
286 int allow_signal(int sig)
288 if (sig < 1 || sig > _NSIG)
291 spin_lock_irq(¤t->sighand->siglock);
292 sigdelset(¤t->blocked, sig);
294 /* Kernel threads handle their own signals.
295 Let the signal code know it'll be handled, so
296 that they don't get converted to SIGKILL or
297 just silently dropped */
298 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
301 spin_unlock_irq(¤t->sighand->siglock);
305 EXPORT_SYMBOL(allow_signal);
307 int disallow_signal(int sig)
309 if (sig < 1 || sig > _NSIG)
312 spin_lock_irq(¤t->sighand->siglock);
313 sigaddset(¤t->blocked, sig);
315 spin_unlock_irq(¤t->sighand->siglock);
319 EXPORT_SYMBOL(disallow_signal);
322 * Put all the gunge required to become a kernel thread without
323 * attached user resources in one place where it belongs.
326 void daemonize(const char *name, ...)
329 struct fs_struct *fs;
332 va_start(args, name);
333 vsnprintf(current->comm, sizeof(current->comm), name, args);
337 * If we were started as result of loading a module, close all of the
338 * user space pages. We don't need them, and if we didn't close them
339 * they would be locked into memory.
343 set_special_pids(1, 1);
345 current->signal->tty = NULL;
348 /* Block and flush all signals */
349 sigfillset(&blocked);
350 sigprocmask(SIG_BLOCK, &blocked, NULL);
351 flush_signals(current);
353 /* Become as one with the init task */
355 exit_fs(current); /* current->fs->count--; */
358 atomic_inc(&fs->count);
360 current->files = init_task.files;
361 atomic_inc(¤t->files->count);
366 EXPORT_SYMBOL(daemonize);
368 static inline void close_files(struct files_struct * files)
376 if (i >= files->max_fdset || i >= files->max_fds)
378 set = files->open_fds->fds_bits[j++];
381 struct file * file = xchg(&files->fd[i], NULL);
383 filp_close(file, files);
392 struct files_struct *get_files_struct(struct task_struct *task)
394 struct files_struct *files;
399 atomic_inc(&files->count);
405 void fastcall put_files_struct(struct files_struct *files)
407 if (atomic_dec_and_test(&files->count)) {
410 * Free the fd and fdset arrays if we expanded them.
412 if (files->fd != &files->fd_array[0])
413 free_fd_array(files->fd, files->max_fds);
414 if (files->max_fdset > __FD_SETSIZE) {
415 free_fdset(files->open_fds, files->max_fdset);
416 free_fdset(files->close_on_exec, files->max_fdset);
418 kmem_cache_free(files_cachep, files);
422 EXPORT_SYMBOL(put_files_struct);
424 static inline void __exit_files(struct task_struct *tsk)
426 struct files_struct * files = tsk->files;
432 put_files_struct(files);
436 void exit_files(struct task_struct *tsk)
441 static inline void __put_fs_struct(struct fs_struct *fs)
443 /* No need to hold fs->lock if we are killing it */
444 if (atomic_dec_and_test(&fs->count)) {
451 mntput(fs->altrootmnt);
453 kmem_cache_free(fs_cachep, fs);
457 void put_fs_struct(struct fs_struct *fs)
462 static inline void __exit_fs(struct task_struct *tsk)
464 struct fs_struct * fs = tsk->fs;
474 void exit_fs(struct task_struct *tsk)
479 EXPORT_SYMBOL_GPL(exit_fs);
482 * Turn us into a lazy TLB process if we
485 void exit_mm(struct task_struct * tsk)
487 struct mm_struct *mm = tsk->mm;
493 * Serialize with any possible pending coredump.
494 * We must hold mmap_sem around checking core_waiters
495 * and clearing tsk->mm. The core-inducing thread
496 * will increment core_waiters for each thread in the
497 * group with ->mm != NULL.
499 down_read(&mm->mmap_sem);
500 if (mm->core_waiters) {
501 up_read(&mm->mmap_sem);
502 down_write(&mm->mmap_sem);
503 if (!--mm->core_waiters)
504 complete(mm->core_startup_done);
505 up_write(&mm->mmap_sem);
507 wait_for_completion(&mm->core_done);
508 down_read(&mm->mmap_sem);
510 atomic_inc(&mm->mm_count);
511 if (mm != tsk->active_mm) BUG();
512 /* more a memory barrier than a real lock */
515 up_read(&mm->mmap_sem);
516 enter_lazy_tlb(mm, current);
521 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
524 * Make sure we're not reparenting to ourselves and that
525 * the parent is not a zombie.
527 BUG_ON(p == reaper || reaper->exit_state >= EXIT_ZOMBIE);
528 p->real_parent = reaper;
529 if (p->parent == p->real_parent)
533 static inline void reparent_thread(task_t *p, task_t *father, int traced)
535 /* We don't want people slaying init. */
536 if (p->exit_signal != -1)
537 p->exit_signal = SIGCHLD;
539 if (p->pdeath_signal)
540 /* We already hold the tasklist_lock here. */
541 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
543 /* Move the child from its dying parent to the new one. */
544 if (unlikely(traced)) {
545 /* Preserve ptrace links if someone else is tracing this child. */
546 list_del_init(&p->ptrace_list);
547 if (p->parent != p->real_parent)
548 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
550 /* If this child is being traced, then we're the one tracing it
551 * anyway, so let go of it.
554 list_del_init(&p->sibling);
555 p->parent = p->real_parent;
556 list_add_tail(&p->sibling, &p->parent->children);
558 /* If we'd notified the old parent about this child's death,
559 * also notify the new parent.
561 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
562 thread_group_empty(p))
563 do_notify_parent(p, p->exit_signal);
564 else if (p->state == TASK_TRACED) {
566 * If it was at a trace stop, turn it into
567 * a normal stop since it's no longer being
575 * process group orphan check
576 * Case ii: Our child is in a different pgrp
577 * than we are, and it was the only connection
578 * outside, so the child pgrp is now orphaned.
580 if ((process_group(p) != process_group(father)) &&
581 (p->signal->session == father->signal->session)) {
582 int pgrp = process_group(p);
584 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
585 __kill_pg_info(SIGHUP, (void *)1, pgrp);
586 __kill_pg_info(SIGCONT, (void *)1, pgrp);
592 * When we die, we re-parent all our children.
593 * Try to give them to another thread in our thread
594 * group, and if no such member exists, give it to
595 * the global child reaper process (ie "init")
597 static inline void forget_original_parent(struct task_struct * father,
598 struct list_head *to_release)
600 struct task_struct *p, *reaper = father;
601 struct list_head *_p, *_n;
603 /* FIXME handle vchild_reaper/initpid */
605 reaper = next_thread(reaper);
606 if (reaper == father) {
607 reaper = child_reaper;
610 } while (reaper->exit_state);
613 * There are only two places where our children can be:
615 * - in our child list
616 * - in our ptraced child list
618 * Search them and reparent children.
620 list_for_each_safe(_p, _n, &father->children) {
622 p = list_entry(_p,struct task_struct,sibling);
626 /* if father isn't the real parent, then ptrace must be enabled */
627 BUG_ON(father != p->real_parent && !ptrace);
629 if (father == p->real_parent) {
630 /* reparent with a reaper, real father it's us */
631 choose_new_parent(p, reaper, child_reaper);
632 reparent_thread(p, father, 0);
634 /* reparent ptraced task to its real parent */
636 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
637 thread_group_empty(p))
638 do_notify_parent(p, p->exit_signal);
642 * if the ptraced child is a zombie with exit_signal == -1
643 * we must collect it before we exit, or it will remain
644 * zombie forever since we prevented it from self-reap itself
645 * while it was being traced by us, to be able to see it in wait4.
647 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
648 list_add(&p->ptrace_list, to_release);
650 list_for_each_safe(_p, _n, &father->ptrace_children) {
651 p = list_entry(_p,struct task_struct,ptrace_list);
652 choose_new_parent(p, reaper, child_reaper);
653 reparent_thread(p, father, 1);
658 * Send signals to all our closest relatives so that they know
659 * to properly mourn us..
661 static void exit_notify(struct task_struct *tsk)
664 struct task_struct *t;
665 struct list_head ptrace_dead, *_p, *_n;
667 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
668 && !thread_group_empty(tsk)) {
670 * This occurs when there was a race between our exit
671 * syscall and a group signal choosing us as the one to
672 * wake up. It could be that we are the only thread
673 * alerted to check for pending signals, but another thread
674 * should be woken now to take the signal since we will not.
675 * Now we'll wake all the threads in the group just to make
676 * sure someone gets all the pending signals.
678 read_lock(&tasklist_lock);
679 spin_lock_irq(&tsk->sighand->siglock);
680 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
681 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
682 recalc_sigpending_tsk(t);
683 if (signal_pending(t))
684 signal_wake_up(t, 0);
686 spin_unlock_irq(&tsk->sighand->siglock);
687 read_unlock(&tasklist_lock);
690 write_lock_irq(&tasklist_lock);
693 * This does two things:
695 * A. Make init inherit all the child processes
696 * B. Check to see if any process groups have become orphaned
697 * as a result of our exiting, and if they have any stopped
698 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
701 INIT_LIST_HEAD(&ptrace_dead);
702 forget_original_parent(tsk, &ptrace_dead);
703 BUG_ON(!list_empty(&tsk->children));
704 BUG_ON(!list_empty(&tsk->ptrace_children));
707 * Check to see if any process groups have become orphaned
708 * as a result of our exiting, and if they have any stopped
709 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
711 * Case i: Our father is in a different pgrp than we are
712 * and we were the only connection outside, so our pgrp
713 * is about to become orphaned.
716 t = tsk->real_parent;
718 if ((process_group(t) != process_group(tsk)) &&
719 (t->signal->session == tsk->signal->session) &&
720 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
721 has_stopped_jobs(process_group(tsk))) {
722 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
723 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
726 /* Let father know we died
728 * Thread signals are configurable, but you aren't going to use
729 * that to send signals to arbitary processes.
730 * That stops right now.
732 * If the parent exec id doesn't match the exec id we saved
733 * when we started then we know the parent has changed security
736 * If our self_exec id doesn't match our parent_exec_id then
737 * we have changed execution domain as these two values started
738 * the same after a fork.
742 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
743 ( tsk->parent_exec_id != t->self_exec_id ||
744 tsk->self_exec_id != tsk->parent_exec_id)
745 && !capable(CAP_KILL))
746 tsk->exit_signal = SIGCHLD;
749 /* If something other than our normal parent is ptracing us, then
750 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
751 * only has special meaning to our real parent.
753 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
754 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
755 do_notify_parent(tsk, signal);
756 } else if (tsk->ptrace) {
757 do_notify_parent(tsk, SIGCHLD);
761 if (tsk->exit_signal == -1 &&
762 (likely(tsk->ptrace == 0) ||
763 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
765 tsk->exit_state = state;
768 * Clear these here so that update_process_times() won't try to deliver
769 * itimer, profile or rlimit signals to this task while it is in late exit.
771 tsk->it_virt_value = cputime_zero;
772 tsk->it_prof_value = cputime_zero;
774 write_unlock_irq(&tasklist_lock);
776 list_for_each_safe(_p, _n, &ptrace_dead) {
778 t = list_entry(_p,struct task_struct,ptrace_list);
782 /* If the process is dead, release it - nobody will wait for it */
783 if (state == EXIT_DEAD)
786 /* PF_DEAD causes final put_task_struct after we schedule. */
788 tsk->flags |= PF_DEAD;
791 fastcall NORET_TYPE void do_exit(long code)
793 struct task_struct *tsk = current;
796 profile_task_exit(tsk);
798 if (unlikely(in_interrupt()))
799 panic("Aiee, killing interrupt handler!");
800 if (unlikely(!tsk->pid))
801 panic("Attempted to kill the idle task!");
802 if (unlikely(tsk->pid == 1))
803 panic("Attempted to kill init!");
807 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
808 current->ptrace_message = code;
809 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
812 tsk->flags |= PF_EXITING;
813 del_timer_sync(&tsk->real_timer);
815 if (unlikely(in_atomic()))
816 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
817 current->comm, current->pid,
820 acct_update_integrals();
821 update_mem_hiwater();
822 group_dead = atomic_dec_and_test(&tsk->signal->live);
834 if (group_dead && tsk->signal->leader)
835 disassociate_ctty(1);
837 module_put(tsk->thread_info->exec_domain->module);
839 module_put(tsk->binfmt->module);
841 tsk->exit_code = code;
844 mpol_free(tsk->mempolicy);
845 tsk->mempolicy = NULL;
848 BUG_ON(!(current->flags & PF_DEAD));
851 /* Avoid "noreturn function does return". */
855 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
863 EXPORT_SYMBOL(complete_and_exit);
865 asmlinkage long sys_exit(int error_code)
867 do_exit((error_code&0xff)<<8);
870 task_t fastcall *next_thread(const task_t *p)
872 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
875 EXPORT_SYMBOL(next_thread);
878 * Take down every thread in the group. This is called by fatal signals
879 * as well as by sys_exit_group (below).
882 do_group_exit(int exit_code)
884 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
886 if (current->signal->flags & SIGNAL_GROUP_EXIT)
887 exit_code = current->signal->group_exit_code;
888 else if (!thread_group_empty(current)) {
889 struct signal_struct *const sig = current->signal;
890 struct sighand_struct *const sighand = current->sighand;
891 read_lock(&tasklist_lock);
892 spin_lock_irq(&sighand->siglock);
893 if (sig->flags & SIGNAL_GROUP_EXIT)
894 /* Another thread got here before we took the lock. */
895 exit_code = sig->group_exit_code;
897 sig->flags = SIGNAL_GROUP_EXIT;
898 sig->group_exit_code = exit_code;
899 zap_other_threads(current);
901 spin_unlock_irq(&sighand->siglock);
902 read_unlock(&tasklist_lock);
910 * this kills every thread in the thread group. Note that any externally
911 * wait4()-ing process will get the correct exit code - even if this
912 * thread is not the thread group leader.
914 asmlinkage void sys_exit_group(int error_code)
916 do_group_exit((error_code & 0xff) << 8);
919 static int eligible_child(pid_t pid, int options, task_t *p)
925 if (process_group(p) != process_group(current))
927 } else if (pid != -1) {
928 if (process_group(p) != -pid)
933 * Do not consider detached threads that are
936 if (p->exit_signal == -1 && !p->ptrace)
939 /* Wait for all children (clone and not) if __WALL is set;
940 * otherwise, wait for clone children *only* if __WCLONE is
941 * set; otherwise, wait for non-clone children *only*. (Note:
942 * A "clone" child here is one that reports to its parent
943 * using a signal other than SIGCHLD.) */
944 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
945 && !(options & __WALL))
948 * Do not consider thread group leaders that are
949 * in a non-empty thread group:
951 if (current->tgid != p->tgid && delay_group_leader(p))
954 if (security_task_wait(p))
960 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
962 struct siginfo __user *infop,
963 struct rusage __user *rusagep)
965 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
968 retval = put_user(SIGCHLD, &infop->si_signo);
970 retval = put_user(0, &infop->si_errno);
972 retval = put_user((short)why, &infop->si_code);
974 retval = put_user(pid, &infop->si_pid);
976 retval = put_user(uid, &infop->si_uid);
978 retval = put_user(status, &infop->si_status);
985 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
986 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
987 * the lock and this task is uninteresting. If we return nonzero, we have
988 * released the lock and the system call should return.
990 static int wait_task_zombie(task_t *p, int noreap,
991 struct siginfo __user *infop,
992 int __user *stat_addr, struct rusage __user *ru)
998 if (unlikely(noreap)) {
1001 int exit_code = p->exit_code;
1004 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1006 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1009 read_unlock(&tasklist_lock);
1010 if ((exit_code & 0x7f) == 0) {
1012 status = exit_code >> 8;
1014 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1015 status = exit_code & 0x7f;
1017 return wait_noreap_copyout(p, pid, uid, why,
1022 * Try to move the task's state to DEAD
1023 * only one thread is allowed to do this:
1025 state = xchg(&p->exit_state, EXIT_DEAD);
1026 if (state != EXIT_ZOMBIE) {
1027 BUG_ON(state != EXIT_DEAD);
1030 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1032 * This can only happen in a race with a ptraced thread
1033 * dying on another processor.
1038 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1040 * The resource counters for the group leader are in its
1041 * own task_struct. Those for dead threads in the group
1042 * are in its signal_struct, as are those for the child
1043 * processes it has previously reaped. All these
1044 * accumulate in the parent's signal_struct c* fields.
1046 * We don't bother to take a lock here to protect these
1047 * p->signal fields, because they are only touched by
1048 * __exit_signal, which runs with tasklist_lock
1049 * write-locked anyway, and so is excluded here. We do
1050 * need to protect the access to p->parent->signal fields,
1051 * as other threads in the parent group can be right
1052 * here reaping other children at the same time.
1054 spin_lock_irq(&p->parent->sighand->siglock);
1055 p->parent->signal->cutime =
1056 cputime_add(p->parent->signal->cutime,
1057 cputime_add(p->utime,
1058 cputime_add(p->signal->utime,
1059 p->signal->cutime)));
1060 p->parent->signal->cstime =
1061 cputime_add(p->parent->signal->cstime,
1062 cputime_add(p->stime,
1063 cputime_add(p->signal->stime,
1064 p->signal->cstime)));
1065 p->parent->signal->cmin_flt +=
1066 p->min_flt + p->signal->min_flt + p->signal->cmin_flt;
1067 p->parent->signal->cmaj_flt +=
1068 p->maj_flt + p->signal->maj_flt + p->signal->cmaj_flt;
1069 p->parent->signal->cnvcsw +=
1070 p->nvcsw + p->signal->nvcsw + p->signal->cnvcsw;
1071 p->parent->signal->cnivcsw +=
1072 p->nivcsw + p->signal->nivcsw + p->signal->cnivcsw;
1073 spin_unlock_irq(&p->parent->sighand->siglock);
1077 * Now we are sure this task is interesting, and no other
1078 * thread can reap it because we set its state to EXIT_DEAD.
1080 read_unlock(&tasklist_lock);
1082 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1083 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1084 ? p->signal->group_exit_code : p->exit_code;
1085 if (!retval && stat_addr)
1086 retval = put_user(status, stat_addr);
1087 if (!retval && infop)
1088 retval = put_user(SIGCHLD, &infop->si_signo);
1089 if (!retval && infop)
1090 retval = put_user(0, &infop->si_errno);
1091 if (!retval && infop) {
1094 if ((status & 0x7f) == 0) {
1098 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1101 retval = put_user((short)why, &infop->si_code);
1103 retval = put_user(status, &infop->si_status);
1105 if (!retval && infop)
1106 retval = put_user(p->pid, &infop->si_pid);
1107 if (!retval && infop)
1108 retval = put_user(p->uid, &infop->si_uid);
1110 // TODO: is this safe?
1111 p->exit_state = EXIT_ZOMBIE;
1115 if (p->real_parent != p->parent) {
1116 write_lock_irq(&tasklist_lock);
1117 /* Double-check with lock held. */
1118 if (p->real_parent != p->parent) {
1120 // TODO: is this safe?
1121 p->exit_state = EXIT_ZOMBIE;
1123 * If this is not a detached task, notify the parent.
1124 * If it's still not detached after that, don't release
1127 if (p->exit_signal != -1) {
1128 do_notify_parent(p, p->exit_signal);
1129 if (p->exit_signal != -1)
1133 write_unlock_irq(&tasklist_lock);
1142 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1143 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1144 * the lock and this task is uninteresting. If we return nonzero, we have
1145 * released the lock and the system call should return.
1147 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1148 struct siginfo __user *infop,
1149 int __user *stat_addr, struct rusage __user *ru)
1151 int retval, exit_code;
1155 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1156 p->signal && p->signal->group_stop_count > 0)
1158 * A group stop is in progress and this is the group leader.
1159 * We won't report until all threads have stopped.
1164 * Now we are pretty sure this task is interesting.
1165 * Make sure it doesn't get reaped out from under us while we
1166 * give up the lock and then examine it below. We don't want to
1167 * keep holding onto the tasklist_lock while we call getrusage and
1168 * possibly take page faults for user memory.
1171 read_unlock(&tasklist_lock);
1173 if (unlikely(noreap)) {
1176 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1178 exit_code = p->exit_code;
1179 if (unlikely(!exit_code) ||
1180 unlikely(p->state > TASK_STOPPED))
1182 return wait_noreap_copyout(p, pid, uid,
1183 why, (exit_code << 8) | 0x7f,
1187 write_lock_irq(&tasklist_lock);
1190 * This uses xchg to be atomic with the thread resuming and setting
1191 * it. It must also be done with the write lock held to prevent a
1192 * race with the EXIT_ZOMBIE case.
1194 exit_code = xchg(&p->exit_code, 0);
1195 if (unlikely(p->exit_state)) {
1197 * The task resumed and then died. Let the next iteration
1198 * catch it in EXIT_ZOMBIE. Note that exit_code might
1199 * already be zero here if it resumed and did _exit(0).
1200 * The task itself is dead and won't touch exit_code again;
1201 * other processors in this function are locked out.
1203 p->exit_code = exit_code;
1206 if (unlikely(exit_code == 0)) {
1208 * Another thread in this function got to it first, or it
1209 * resumed, or it resumed and then died.
1211 write_unlock_irq(&tasklist_lock);
1215 * We are returning to the wait loop without having successfully
1216 * removed the process and having released the lock. We cannot
1217 * continue, since the "p" task pointer is potentially stale.
1219 * Return -EAGAIN, and do_wait() will restart the loop from the
1220 * beginning. Do _not_ re-acquire the lock.
1225 /* move to end of parent's list to avoid starvation */
1227 add_parent(p, p->parent);
1229 write_unlock_irq(&tasklist_lock);
1231 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1232 if (!retval && stat_addr)
1233 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1234 if (!retval && infop)
1235 retval = put_user(SIGCHLD, &infop->si_signo);
1236 if (!retval && infop)
1237 retval = put_user(0, &infop->si_errno);
1238 if (!retval && infop)
1239 retval = put_user((short)((p->ptrace & PT_PTRACED)
1240 ? CLD_TRAPPED : CLD_STOPPED),
1242 if (!retval && infop)
1243 retval = put_user(exit_code, &infop->si_status);
1244 if (!retval && infop)
1245 retval = put_user(p->pid, &infop->si_pid);
1246 if (!retval && infop)
1247 retval = put_user(p->uid, &infop->si_uid);
1257 * Handle do_wait work for one task in a live, non-stopped state.
1258 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1259 * the lock and this task is uninteresting. If we return nonzero, we have
1260 * released the lock and the system call should return.
1262 static int wait_task_continued(task_t *p, int noreap,
1263 struct siginfo __user *infop,
1264 int __user *stat_addr, struct rusage __user *ru)
1270 if (unlikely(!p->signal))
1273 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1276 spin_lock_irq(&p->sighand->siglock);
1277 /* Re-check with the lock held. */
1278 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1279 spin_unlock_irq(&p->sighand->siglock);
1283 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1284 spin_unlock_irq(&p->sighand->siglock);
1289 read_unlock(&tasklist_lock);
1292 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1294 if (!retval && stat_addr)
1295 retval = put_user(0xffff, stat_addr);
1299 retval = wait_noreap_copyout(p, pid, uid,
1300 CLD_CONTINUED, SIGCONT,
1302 BUG_ON(retval == 0);
1309 static inline int my_ptrace_child(struct task_struct *p)
1311 if (!(p->ptrace & PT_PTRACED))
1313 if (!(p->ptrace & PT_ATTACHED))
1316 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1317 * we are the attacher. If we are the real parent, this is a race
1318 * inside ptrace_attach. It is waiting for the tasklist_lock,
1319 * which we have to switch the parent links, but has already set
1320 * the flags in p->ptrace.
1322 return (p->parent != p->real_parent);
1325 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1326 int __user *stat_addr, struct rusage __user *ru)
1328 DECLARE_WAITQUEUE(wait, current);
1329 struct task_struct *tsk;
1332 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1335 * We will set this flag if we see any child that might later
1336 * match our criteria, even if we are not able to reap it yet.
1339 current->state = TASK_INTERRUPTIBLE;
1340 read_lock(&tasklist_lock);
1343 struct task_struct *p;
1344 struct list_head *_p;
1347 list_for_each(_p,&tsk->children) {
1348 p = list_entry(_p,struct task_struct,sibling);
1350 ret = eligible_child(pid, options, p);
1356 if (!my_ptrace_child(p))
1361 * It's stopped now, so it might later
1362 * continue, exit, or stop again.
1365 if (!(options & WUNTRACED) &&
1366 !my_ptrace_child(p))
1368 retval = wait_task_stopped(p, ret == 2,
1369 (options & WNOWAIT),
1372 if (retval == -EAGAIN)
1374 if (retval != 0) /* He released the lock. */
1379 if (p->exit_state == EXIT_DEAD)
1381 // case EXIT_ZOMBIE:
1382 if (p->exit_state == EXIT_ZOMBIE) {
1384 * Eligible but we cannot release
1388 goto check_continued;
1389 if (!likely(options & WEXITED))
1391 retval = wait_task_zombie(
1392 p, (options & WNOWAIT),
1393 infop, stat_addr, ru);
1394 /* He released the lock. */
1401 * It's running now, so it might later
1402 * exit, stop, or stop and then continue.
1405 if (!unlikely(options & WCONTINUED))
1407 retval = wait_task_continued(
1408 p, (options & WNOWAIT),
1409 infop, stat_addr, ru);
1410 if (retval != 0) /* He released the lock. */
1416 list_for_each(_p, &tsk->ptrace_children) {
1417 p = list_entry(_p, struct task_struct,
1419 if (!eligible_child(pid, options, p))
1425 if (options & __WNOTHREAD)
1427 tsk = next_thread(tsk);
1428 if (tsk->signal != current->signal)
1430 } while (tsk != current);
1432 read_unlock(&tasklist_lock);
1435 if (options & WNOHANG)
1437 retval = -ERESTARTSYS;
1438 if (signal_pending(current))
1445 current->state = TASK_RUNNING;
1446 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1452 * For a WNOHANG return, clear out all the fields
1453 * we would set so the user can easily tell the
1457 retval = put_user(0, &infop->si_signo);
1459 retval = put_user(0, &infop->si_errno);
1461 retval = put_user(0, &infop->si_code);
1463 retval = put_user(0, &infop->si_pid);
1465 retval = put_user(0, &infop->si_uid);
1467 retval = put_user(0, &infop->si_status);
1473 asmlinkage long sys_waitid(int which, pid_t pid,
1474 struct siginfo __user *infop, int options,
1475 struct rusage __user *ru)
1479 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1481 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1501 ret = do_wait(pid, options, infop, NULL, ru);
1503 /* avoid REGPARM breakage on x86: */
1504 prevent_tail_call(ret);
1508 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1509 int options, struct rusage __user *ru)
1513 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1514 __WNOTHREAD|__WCLONE|__WALL))
1516 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1518 /* avoid REGPARM breakage on x86: */
1519 prevent_tail_call(ret);
1523 #ifdef __ARCH_WANT_SYS_WAITPID
1526 * sys_waitpid() remains for compatibility. waitpid() should be
1527 * implemented by calling sys_wait4() from libc.a.
1529 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1531 return sys_wait4(pid, stat_addr, options, NULL);