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/security.h>
18 #include <linux/cpu.h>
19 #include <linux/acct.h>
20 #include <linux/file.h>
21 #include <linux/binfmts.h>
22 #include <linux/ptrace.h>
23 #include <linux/profile.h>
24 #include <linux/mount.h>
25 #include <linux/proc_fs.h>
26 #include <linux/mempolicy.h>
27 #include <linux/vs_limit.h>
29 #include <asm/uaccess.h>
30 #include <asm/unistd.h>
31 #include <asm/pgtable.h>
32 #include <asm/mmu_context.h>
34 extern void sem_exit (void);
35 extern struct task_struct *child_reaper;
37 int getrusage(struct task_struct *, int, struct rusage __user *);
39 static void __unhash_process(struct task_struct *p)
42 /* tasklist_lock is held, is this sufficient? */
44 atomic_dec(&p->vx_info->cvirt.nr_threads);
47 detach_pid(p, PIDTYPE_PID);
48 detach_pid(p, PIDTYPE_TGID);
49 if (thread_group_leader(p)) {
50 detach_pid(p, PIDTYPE_PGID);
51 detach_pid(p, PIDTYPE_SID);
53 __get_cpu_var(process_counts)--;
59 void release_task(struct task_struct * p)
63 struct dentry *proc_dentry;
66 atomic_dec(&p->user->processes);
67 spin_lock(&p->proc_lock);
68 proc_dentry = proc_pid_unhash(p);
69 write_lock_irq(&tasklist_lock);
70 if (unlikely(p->ptrace))
72 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
78 * If we are the last non-leader member of the thread
79 * group, and the leader is zombie, then notify the
80 * group leader's parent process. (if it wants notification.)
83 leader = p->group_leader;
84 if (leader != p && thread_group_empty(leader) && leader->state == TASK_ZOMBIE) {
85 BUG_ON(leader->exit_signal == -1);
86 do_notify_parent(leader, leader->exit_signal);
88 * If we were the last child thread and the leader has
89 * exited already, and the leader's parent ignores SIGCHLD,
90 * then we are the one who should release the leader.
92 * do_notify_parent() will have marked it self-reaping in
95 zap_leader = (leader->exit_signal == -1);
99 write_unlock_irq(&tasklist_lock);
100 spin_unlock(&p->proc_lock);
101 proc_pid_flush(proc_dentry);
106 if (unlikely(zap_leader))
110 /* we are using it only for SMP init */
112 void unhash_process(struct task_struct *p)
114 struct dentry *proc_dentry;
116 spin_lock(&p->proc_lock);
117 proc_dentry = proc_pid_unhash(p);
118 write_lock_irq(&tasklist_lock);
120 write_unlock_irq(&tasklist_lock);
121 spin_unlock(&p->proc_lock);
122 proc_pid_flush(proc_dentry);
126 * This checks not only the pgrp, but falls back on the pid if no
127 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
130 int session_of_pgrp(int pgrp)
132 struct task_struct *p;
135 read_lock(&tasklist_lock);
136 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
137 if (p->signal->session > 0) {
138 sid = p->signal->session;
141 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
142 p = find_task_by_pid(pgrp);
144 sid = p->signal->session;
146 read_unlock(&tasklist_lock);
152 * Determine if a process group is "orphaned", according to the POSIX
153 * definition in 2.2.2.52. Orphaned process groups are not to be affected
154 * by terminal-generated stop signals. Newly orphaned process groups are
155 * to receive a SIGHUP and a SIGCONT.
157 * "I ask you, have you ever known what it is to be an orphan?"
159 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
161 struct task_struct *p;
164 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
165 if (p == ignored_task
166 || p->state >= TASK_ZOMBIE
167 || p->real_parent->pid == 1)
169 if (process_group(p->real_parent) != pgrp
170 && p->real_parent->signal->session == p->signal->session) {
174 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
175 return ret; /* (sighing) "Often!" */
178 int is_orphaned_pgrp(int pgrp)
182 read_lock(&tasklist_lock);
183 retval = will_become_orphaned_pgrp(pgrp, NULL);
184 read_unlock(&tasklist_lock);
189 static inline int has_stopped_jobs(int pgrp)
192 struct task_struct *p;
194 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
195 if (p->state != TASK_STOPPED)
198 /* If p is stopped by a debugger on a signal that won't
199 stop it, then don't count p as stopped. This isn't
200 perfect but it's a good approximation. */
201 if (unlikely (p->ptrace)
202 && p->exit_code != SIGSTOP
203 && p->exit_code != SIGTSTP
204 && p->exit_code != SIGTTOU
205 && p->exit_code != SIGTTIN)
210 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
215 * reparent_to_init() - Reparent the calling kernel thread to the init task.
217 * If a kernel thread is launched as a result of a system call, or if
218 * it ever exits, it should generally reparent itself to init so that
219 * it is correctly cleaned up on exit.
221 * The various task state such as scheduling policy and priority may have
222 * been inherited from a user process, so we reset them to sane values here.
224 * NOTE that reparent_to_init() gives the caller full capabilities.
226 void reparent_to_init(void)
228 write_lock_irq(&tasklist_lock);
230 ptrace_unlink(current);
231 /* Reparent to init */
232 REMOVE_LINKS(current);
233 /* FIXME handle vchild_reaper/initpid */
234 current->parent = child_reaper;
235 current->real_parent = child_reaper;
238 /* Set the exit signal to SIGCHLD so we signal init on exit */
239 current->exit_signal = SIGCHLD;
241 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
242 set_user_nice(current, 0);
246 security_task_reparent_to_init(current);
247 memcpy(current->rlim, init_task.rlim, sizeof(*(current->rlim)));
248 atomic_inc(&(INIT_USER->__count));
249 switch_uid(INIT_USER);
251 write_unlock_irq(&tasklist_lock);
254 void __set_special_pids(pid_t session, pid_t pgrp)
256 struct task_struct *curr = current;
258 if (curr->signal->session != session) {
259 detach_pid(curr, PIDTYPE_SID);
260 curr->signal->session = session;
261 attach_pid(curr, PIDTYPE_SID, session);
263 if (process_group(curr) != pgrp) {
264 detach_pid(curr, PIDTYPE_PGID);
265 curr->signal->pgrp = pgrp;
266 attach_pid(curr, PIDTYPE_PGID, pgrp);
270 void set_special_pids(pid_t session, pid_t pgrp)
272 write_lock_irq(&tasklist_lock);
273 __set_special_pids(session, pgrp);
274 write_unlock_irq(&tasklist_lock);
278 * Let kernel threads use this to say that they
279 * allow a certain signal (since daemonize() will
280 * have disabled all of them by default).
282 int allow_signal(int sig)
284 if (sig < 1 || sig > _NSIG)
287 spin_lock_irq(¤t->sighand->siglock);
288 sigdelset(¤t->blocked, sig);
290 /* Kernel threads handle their own signals.
291 Let the signal code know it'll be handled, so
292 that they don't get converted to SIGKILL or
293 just silently dropped */
294 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
297 spin_unlock_irq(¤t->sighand->siglock);
301 EXPORT_SYMBOL(allow_signal);
303 int disallow_signal(int sig)
305 if (sig < 1 || sig > _NSIG)
308 spin_lock_irq(¤t->sighand->siglock);
309 sigaddset(¤t->blocked, sig);
311 spin_unlock_irq(¤t->sighand->siglock);
315 EXPORT_SYMBOL(disallow_signal);
318 * Put all the gunge required to become a kernel thread without
319 * attached user resources in one place where it belongs.
322 void daemonize(const char *name, ...)
325 struct fs_struct *fs;
328 va_start(args, name);
329 vsnprintf(current->comm, sizeof(current->comm), name, args);
333 * If we were started as result of loading a module, close all of the
334 * user space pages. We don't need them, and if we didn't close them
335 * they would be locked into memory.
339 set_special_pids(1, 1);
340 current->signal->tty = NULL;
342 /* Block and flush all signals */
343 sigfillset(&blocked);
344 sigprocmask(SIG_BLOCK, &blocked, NULL);
345 flush_signals(current);
347 /* Become as one with the init task */
349 exit_fs(current); /* current->fs->count--; */
352 atomic_inc(&fs->count);
354 current->files = init_task.files;
355 atomic_inc(¤t->files->count);
360 EXPORT_SYMBOL(daemonize);
362 static inline void close_files(struct files_struct * files)
370 if (i >= files->max_fdset || i >= files->max_fds)
372 set = files->open_fds->fds_bits[j++];
375 struct file * file = xchg(&files->fd[i], NULL);
377 filp_close(file, files);
386 struct files_struct *get_files_struct(struct task_struct *task)
388 struct files_struct *files;
393 atomic_inc(&files->count);
399 void fastcall put_files_struct(struct files_struct *files)
401 if (atomic_dec_and_test(&files->count)) {
404 * Free the fd and fdset arrays if we expanded them.
406 if (files->fd != &files->fd_array[0])
407 free_fd_array(files->fd, files->max_fds);
408 if (files->max_fdset > __FD_SETSIZE) {
409 free_fdset(files->open_fds, files->max_fdset);
410 free_fdset(files->close_on_exec, files->max_fdset);
412 kmem_cache_free(files_cachep, files);
416 EXPORT_SYMBOL(put_files_struct);
418 static inline void __exit_files(struct task_struct *tsk)
420 struct files_struct * files = tsk->files;
426 put_files_struct(files);
430 void exit_files(struct task_struct *tsk)
435 static inline void __put_fs_struct(struct fs_struct *fs)
437 /* No need to hold fs->lock if we are killing it */
438 if (atomic_dec_and_test(&fs->count)) {
445 mntput(fs->altrootmnt);
447 kmem_cache_free(fs_cachep, fs);
451 void put_fs_struct(struct fs_struct *fs)
456 static inline void __exit_fs(struct task_struct *tsk)
458 struct fs_struct * fs = tsk->fs;
468 void exit_fs(struct task_struct *tsk)
473 EXPORT_SYMBOL_GPL(exit_fs);
476 * Turn us into a lazy TLB process if we
479 static inline void __exit_mm(struct task_struct * tsk)
481 struct mm_struct *mm = tsk->mm;
487 * Serialize with any possible pending coredump.
488 * We must hold mmap_sem around checking core_waiters
489 * and clearing tsk->mm. The core-inducing thread
490 * will increment core_waiters for each thread in the
491 * group with ->mm != NULL.
493 down_read(&mm->mmap_sem);
494 if (mm->core_waiters) {
495 up_read(&mm->mmap_sem);
496 down_write(&mm->mmap_sem);
497 if (!--mm->core_waiters)
498 complete(mm->core_startup_done);
499 up_write(&mm->mmap_sem);
501 wait_for_completion(&mm->core_done);
502 down_read(&mm->mmap_sem);
504 atomic_inc(&mm->mm_count);
505 if (mm != tsk->active_mm) BUG();
506 /* more a memory barrier than a real lock */
509 up_read(&mm->mmap_sem);
510 enter_lazy_tlb(mm, current);
515 void exit_mm(struct task_struct *tsk)
520 EXPORT_SYMBOL(exit_mm);
522 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
525 * Make sure we're not reparenting to ourselves and that
526 * the parent is not a zombie.
528 BUG_ON(p == reaper || reaper->state >= TASK_ZOMBIE);
529 p->real_parent = reaper;
530 if (p->parent == p->real_parent)
534 static inline void reparent_thread(task_t *p, task_t *father, int traced)
536 /* We don't want people slaying init. */
537 if (p->exit_signal != -1)
538 p->exit_signal = SIGCHLD;
541 if (p->pdeath_signal)
542 /* We already hold the tasklist_lock here. */
543 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
545 /* Move the child from its dying parent to the new one. */
546 if (unlikely(traced)) {
547 /* Preserve ptrace links if someone else is tracing this child. */
548 list_del_init(&p->ptrace_list);
549 if (p->parent != p->real_parent)
550 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
552 /* If this child is being traced, then we're the one tracing it
553 * anyway, so let go of it.
556 list_del_init(&p->sibling);
557 p->parent = p->real_parent;
558 list_add_tail(&p->sibling, &p->parent->children);
560 /* If we'd notified the old parent about this child's death,
561 * also notify the new parent.
563 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
564 thread_group_empty(p))
565 do_notify_parent(p, p->exit_signal);
566 else if (p->state == TASK_TRACED) {
568 * If it was at a trace stop, turn it into
569 * a normal stop since it's no longer being
572 p->state = TASK_STOPPED;
577 * process group orphan check
578 * Case ii: Our child is in a different pgrp
579 * than we are, and it was the only connection
580 * outside, so the child pgrp is now orphaned.
582 if ((process_group(p) != process_group(father)) &&
583 (p->signal->session == father->signal->session)) {
584 int pgrp = process_group(p);
586 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
587 __kill_pg_info(SIGHUP, (void *)1, pgrp);
588 __kill_pg_info(SIGCONT, (void *)1, pgrp);
594 * When we die, we re-parent all our children.
595 * Try to give them to another thread in our thread
596 * group, and if no such member exists, give it to
597 * the global child reaper process (ie "init")
599 static inline void forget_original_parent(struct task_struct * father,
600 struct list_head *to_release)
602 struct task_struct *p, *reaper = father;
603 struct list_head *_p, *_n;
605 /* FIXME handle vchild_reaper/initpid */
607 reaper = next_thread(reaper);
608 if (reaper == father) {
609 reaper = child_reaper;
612 } while (reaper->state >= TASK_ZOMBIE);
615 * There are only two places where our children can be:
617 * - in our child list
618 * - in our ptraced child list
620 * Search them and reparent children.
622 list_for_each_safe(_p, _n, &father->children) {
624 p = list_entry(_p,struct task_struct,sibling);
628 /* if father isn't the real parent, then ptrace must be enabled */
629 BUG_ON(father != p->real_parent && !ptrace);
631 if (father == p->real_parent) {
632 /* reparent with a reaper, real father it's us */
633 choose_new_parent(p, reaper, child_reaper);
634 reparent_thread(p, father, 0);
636 /* reparent ptraced task to its real parent */
638 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
639 thread_group_empty(p))
640 do_notify_parent(p, p->exit_signal);
644 * if the ptraced child is a zombie with exit_signal == -1
645 * we must collect it before we exit, or it will remain
646 * zombie forever since we prevented it from self-reap itself
647 * while it was being traced by us, to be able to see it in wait4.
649 if (unlikely(ptrace && p->state == TASK_ZOMBIE && p->exit_signal == -1))
650 list_add(&p->ptrace_list, to_release);
652 list_for_each_safe(_p, _n, &father->ptrace_children) {
653 p = list_entry(_p,struct task_struct,ptrace_list);
654 choose_new_parent(p, reaper, child_reaper);
655 reparent_thread(p, father, 1);
660 * Send signals to all our closest relatives so that they know
661 * to properly mourn us..
663 static void exit_notify(struct task_struct *tsk)
666 struct task_struct *t;
667 struct list_head ptrace_dead, *_p, *_n;
669 if (signal_pending(tsk) && !tsk->signal->group_exit
670 && !thread_group_empty(tsk)) {
672 * This occurs when there was a race between our exit
673 * syscall and a group signal choosing us as the one to
674 * wake up. It could be that we are the only thread
675 * alerted to check for pending signals, but another thread
676 * should be woken now to take the signal since we will not.
677 * Now we'll wake all the threads in the group just to make
678 * sure someone gets all the pending signals.
680 read_lock(&tasklist_lock);
681 spin_lock_irq(&tsk->sighand->siglock);
682 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
683 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
684 recalc_sigpending_tsk(t);
685 if (signal_pending(t))
686 signal_wake_up(t, 0);
688 spin_unlock_irq(&tsk->sighand->siglock);
689 read_unlock(&tasklist_lock);
692 write_lock_irq(&tasklist_lock);
695 * This does two things:
697 * A. Make init inherit all the child processes
698 * B. Check to see if any process groups have become orphaned
699 * as a result of our exiting, and if they have any stopped
700 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
703 INIT_LIST_HEAD(&ptrace_dead);
704 forget_original_parent(tsk, &ptrace_dead);
705 BUG_ON(!list_empty(&tsk->children));
706 BUG_ON(!list_empty(&tsk->ptrace_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.
718 t = tsk->real_parent;
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, (void *)1, process_group(tsk));
725 __kill_pg_info(SIGCONT, (void *)1, 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;
751 /* If something other than our normal parent is ptracing us, then
752 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
753 * only has special meaning to our real parent.
755 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
756 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
757 do_notify_parent(tsk, signal);
758 } else if (tsk->ptrace) {
759 do_notify_parent(tsk, SIGCHLD);
763 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
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 = 0;
772 tsk->it_prof_value = 0;
773 tsk->rlim[RLIMIT_CPU].rlim_cur = RLIM_INFINITY;
775 write_unlock_irq(&tasklist_lock);
777 list_for_each_safe(_p, _n, &ptrace_dead) {
779 t = list_entry(_p,struct task_struct,ptrace_list);
783 /* If the process is dead, release it - nobody will wait for it */
784 if (state == TASK_DEAD)
787 /* PF_DEAD causes final put_task_struct after we schedule. */
789 tsk->flags |= PF_DEAD;
792 asmlinkage NORET_TYPE void do_exit(long code)
794 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!");
806 tsk->flags |= PF_EXITING;
807 del_timer_sync(&tsk->real_timer);
809 if (unlikely(in_atomic()))
810 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
811 current->comm, current->pid,
814 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
815 current->ptrace_message = code;
816 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
828 if (tsk->signal->leader)
829 disassociate_ctty(1);
831 module_put(tsk->thread_info->exec_domain->module);
833 module_put(tsk->binfmt->module);
835 tsk->exit_code = code;
838 mpol_free(tsk->mempolicy);
839 tsk->mempolicy = NULL;
843 /* Avoid "noreturn function does return". */
847 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
855 EXPORT_SYMBOL(complete_and_exit);
857 asmlinkage long sys_exit(int error_code)
859 do_exit((error_code&0xff)<<8);
862 task_t fastcall *next_thread(const task_t *p)
867 if (!spin_is_locked(&p->sighand->siglock) &&
868 !rwlock_is_locked(&tasklist_lock))
871 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
874 EXPORT_SYMBOL(next_thread);
877 * Take down every thread in the group. This is called by fatal signals
878 * as well as by sys_exit_group (below).
881 do_group_exit(int exit_code)
883 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
885 if (current->signal->group_exit)
886 exit_code = current->signal->group_exit_code;
887 else if (!thread_group_empty(current)) {
888 struct signal_struct *const sig = current->signal;
889 struct sighand_struct *const sighand = current->sighand;
890 read_lock(&tasklist_lock);
891 spin_lock_irq(&sighand->siglock);
893 /* Another thread got here before we took the lock. */
894 exit_code = sig->group_exit_code;
897 sig->group_exit_code = exit_code;
898 zap_other_threads(current);
900 spin_unlock_irq(&sighand->siglock);
901 read_unlock(&tasklist_lock);
909 * this kills every thread in the thread group. Note that any externally
910 * wait4()-ing process will get the correct exit code - even if this
911 * thread is not the thread group leader.
913 asmlinkage void sys_exit_group(int error_code)
915 do_group_exit((error_code & 0xff) << 8);
918 static int eligible_child(pid_t pid, int options, task_t *p)
924 if (process_group(p) != process_group(current))
926 } else if (pid != -1) {
927 if (process_group(p) != -pid)
932 * Do not consider detached threads that are
935 if (p->exit_signal == -1 && !p->ptrace)
938 /* Wait for all children (clone and not) if __WALL is set;
939 * otherwise, wait for clone children *only* if __WCLONE is
940 * set; otherwise, wait for non-clone children *only*. (Note:
941 * A "clone" child here is one that reports to its parent
942 * using a signal other than SIGCHLD.) */
943 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
944 && !(options & __WALL))
947 * Do not consider thread group leaders that are
948 * in a non-empty thread group:
950 if (current->tgid != p->tgid && delay_group_leader(p))
953 if (security_task_wait(p))
959 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
961 struct siginfo __user *infop,
962 struct rusage __user *rusagep)
964 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
967 retval = put_user(SIGCHLD, &infop->si_signo);
969 retval = put_user(0, &infop->si_errno);
971 retval = put_user((short)why, &infop->si_code);
973 retval = put_user(pid, &infop->si_pid);
975 retval = put_user(uid, &infop->si_uid);
977 retval = put_user(status, &infop->si_status);
984 * Handle sys_wait4 work for one task in state TASK_ZOMBIE. We hold
985 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
986 * the lock and this task is uninteresting. If we return nonzero, we have
987 * released the lock and the system call should return.
989 static int wait_task_zombie(task_t *p, int noreap,
990 struct siginfo __user *infop,
991 int __user *stat_addr, struct rusage __user *ru)
997 if (unlikely(noreap)) {
1000 int exit_code = p->exit_code;
1003 if (unlikely(p->state != TASK_ZOMBIE))
1005 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1008 read_unlock(&tasklist_lock);
1009 if ((exit_code & 0x7f) == 0) {
1011 status = exit_code >> 8;
1013 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1014 status = exit_code & 0x7f;
1016 return wait_noreap_copyout(p, pid, uid, why,
1021 * Try to move the task's state to DEAD
1022 * only one thread is allowed to do this:
1024 state = xchg(&p->state, TASK_DEAD);
1025 if (state != TASK_ZOMBIE) {
1026 BUG_ON(state != TASK_DEAD);
1029 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1031 * This can only happen in a race with a ptraced thread
1032 * dying on another processor.
1037 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1039 * The resource counters for the group leader are in its
1040 * own task_struct. Those for dead threads in the group
1041 * are in its signal_struct, as are those for the child
1042 * processes it has previously reaped. All these
1043 * accumulate in the parent's signal_struct c* fields.
1045 * We don't bother to take a lock here to protect these
1046 * p->signal fields, because they are only touched by
1047 * __exit_signal, which runs with tasklist_lock
1048 * write-locked anyway, and so is excluded here. We do
1049 * need to protect the access to p->parent->signal fields,
1050 * as other threads in the parent group can be right
1051 * here reaping other children at the same time.
1053 spin_lock_irq(&p->parent->sighand->siglock);
1054 p->parent->signal->cutime +=
1055 p->utime + p->signal->utime + p->signal->cutime;
1056 p->parent->signal->cstime +=
1057 p->stime + p->signal->stime + p->signal->cstime;
1058 p->parent->signal->cmin_flt +=
1059 p->min_flt + p->signal->min_flt + p->signal->cmin_flt;
1060 p->parent->signal->cmaj_flt +=
1061 p->maj_flt + p->signal->maj_flt + p->signal->cmaj_flt;
1062 p->parent->signal->cnvcsw +=
1063 p->nvcsw + p->signal->nvcsw + p->signal->cnvcsw;
1064 p->parent->signal->cnivcsw +=
1065 p->nivcsw + p->signal->nivcsw + p->signal->cnivcsw;
1066 spin_unlock_irq(&p->parent->sighand->siglock);
1070 * Now we are sure this task is interesting, and no other
1071 * thread can reap it because we set its state to TASK_DEAD.
1073 read_unlock(&tasklist_lock);
1075 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1076 status = p->signal->group_exit
1077 ? p->signal->group_exit_code : p->exit_code;
1078 if (!retval && stat_addr)
1079 retval = put_user(status, stat_addr);
1080 if (!retval && infop)
1081 retval = put_user(SIGCHLD, &infop->si_signo);
1082 if (!retval && infop)
1083 retval = put_user(0, &infop->si_errno);
1084 if (!retval && infop) {
1087 if ((status & 0x7f) == 0) {
1091 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1094 retval = put_user((short)why, &infop->si_code);
1096 retval = put_user(status, &infop->si_status);
1098 if (!retval && infop)
1099 retval = put_user(p->pid, &infop->si_pid);
1100 if (!retval && infop)
1101 retval = put_user(p->uid, &infop->si_uid);
1103 p->state = TASK_ZOMBIE;
1107 if (p->real_parent != p->parent) {
1108 write_lock_irq(&tasklist_lock);
1109 /* Double-check with lock held. */
1110 if (p->real_parent != p->parent) {
1112 p->state = TASK_ZOMBIE;
1114 * If this is not a detached task, notify the parent.
1115 * If it's still not detached after that, don't release
1118 if (p->exit_signal != -1) {
1119 do_notify_parent(p, p->exit_signal);
1120 if (p->exit_signal != -1)
1124 write_unlock_irq(&tasklist_lock);
1133 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1134 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1135 * the lock and this task is uninteresting. If we return nonzero, we have
1136 * released the lock and the system call should return.
1138 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1139 struct siginfo __user *infop,
1140 int __user *stat_addr, struct rusage __user *ru)
1142 int retval, exit_code;
1146 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1147 p->signal && p->signal->group_stop_count > 0)
1149 * A group stop is in progress and this is the group leader.
1150 * We won't report until all threads have stopped.
1155 * Now we are pretty sure this task is interesting.
1156 * Make sure it doesn't get reaped out from under us while we
1157 * give up the lock and then examine it below. We don't want to
1158 * keep holding onto the tasklist_lock while we call getrusage and
1159 * possibly take page faults for user memory.
1162 read_unlock(&tasklist_lock);
1164 if (unlikely(noreap)) {
1167 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1169 exit_code = p->exit_code;
1170 if (unlikely(!exit_code) ||
1171 unlikely(p->state > TASK_STOPPED))
1173 return wait_noreap_copyout(p, pid, uid,
1174 why, (exit_code << 8) | 0x7f,
1178 write_lock_irq(&tasklist_lock);
1181 * This uses xchg to be atomic with the thread resuming and setting
1182 * it. It must also be done with the write lock held to prevent a
1183 * race with the TASK_ZOMBIE case.
1185 exit_code = xchg(&p->exit_code, 0);
1186 if (unlikely(p->state >= TASK_ZOMBIE)) {
1188 * The task resumed and then died. Let the next iteration
1189 * catch it in TASK_ZOMBIE. Note that exit_code might
1190 * already be zero here if it resumed and did _exit(0).
1191 * The task itself is dead and won't touch exit_code again;
1192 * other processors in this function are locked out.
1194 p->exit_code = exit_code;
1197 if (unlikely(exit_code == 0)) {
1199 * Another thread in this function got to it first, or it
1200 * resumed, or it resumed and then died.
1202 write_unlock_irq(&tasklist_lock);
1205 read_lock(&tasklist_lock);
1209 /* move to end of parent's list to avoid starvation */
1211 add_parent(p, p->parent);
1213 write_unlock_irq(&tasklist_lock);
1215 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1216 if (!retval && stat_addr)
1217 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1218 if (!retval && infop)
1219 retval = put_user(SIGCHLD, &infop->si_signo);
1220 if (!retval && infop)
1221 retval = put_user(0, &infop->si_errno);
1222 if (!retval && infop)
1223 retval = put_user((short)((p->ptrace & PT_PTRACED)
1224 ? CLD_TRAPPED : CLD_STOPPED),
1226 if (!retval && infop)
1227 retval = put_user(exit_code, &infop->si_status);
1228 if (!retval && infop)
1229 retval = put_user(p->pid, &infop->si_pid);
1230 if (!retval && infop)
1231 retval = put_user(p->uid, &infop->si_uid);
1240 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1241 int __user *stat_addr, struct rusage __user *ru)
1243 DECLARE_WAITQUEUE(wait, current);
1244 struct task_struct *tsk;
1247 add_wait_queue(¤t->wait_chldexit,&wait);
1250 current->state = TASK_INTERRUPTIBLE;
1251 read_lock(&tasklist_lock);
1254 struct task_struct *p;
1255 struct list_head *_p;
1258 list_for_each(_p,&tsk->children) {
1259 p = list_entry(_p,struct task_struct,sibling);
1261 ret = eligible_child(pid, options, p);
1268 if (!(p->ptrace & PT_PTRACED))
1272 if (!(options & WUNTRACED) &&
1273 !(p->ptrace & PT_PTRACED))
1275 retval = wait_task_stopped(p, ret == 2,
1276 (options & WNOWAIT),
1279 if (retval != 0) /* He released the lock. */
1284 * Eligible but we cannot release it yet:
1287 goto check_continued;
1288 if (!likely(options & WEXITED))
1290 retval = wait_task_zombie(
1291 p, (options & WNOWAIT),
1292 infop, stat_addr, ru);
1293 if (retval != 0) /* He released the lock. */
1300 if (!unlikely(options & WCONTINUED))
1302 if (unlikely(!p->signal))
1304 spin_lock_irq(&p->sighand->siglock);
1305 if (p->signal->stop_state < 0) {
1309 if (!(options & WNOWAIT))
1310 p->signal->stop_state = 0;
1311 spin_unlock_irq(&p->sighand->siglock);
1315 read_unlock(&tasklist_lock);
1316 retval = wait_noreap_copyout(p, pid,
1318 SIGCONT, infop, ru);
1319 BUG_ON(retval == 0);
1322 spin_unlock_irq(&p->sighand->siglock);
1327 list_for_each(_p, &tsk->ptrace_children) {
1328 p = list_entry(_p, struct task_struct,
1330 if (!eligible_child(pid, options, p))
1336 if (options & __WNOTHREAD)
1338 tsk = next_thread(tsk);
1339 if (tsk->signal != current->signal)
1341 } while (tsk != current);
1343 read_unlock(&tasklist_lock);
1346 if (options & WNOHANG)
1348 retval = -ERESTARTSYS;
1349 if (signal_pending(current))
1356 current->state = TASK_RUNNING;
1357 remove_wait_queue(¤t->wait_chldexit,&wait);
1363 * For a WNOHANG return, clear out all the fields
1364 * we would set so the user can easily tell the
1368 retval = put_user(0, &infop->si_signo);
1370 retval = put_user(0, &infop->si_errno);
1372 retval = put_user(0, &infop->si_code);
1374 retval = put_user(0, &infop->si_pid);
1376 retval = put_user(0, &infop->si_uid);
1378 retval = put_user(0, &infop->si_status);
1384 asmlinkage long sys_waitid(int which, pid_t pid,
1385 struct siginfo __user *infop, int options,
1386 struct rusage __user *ru)
1390 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1392 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1412 ret = do_wait(pid, options, infop, NULL, ru);
1414 /* avoid REGPARM breakage on x86: */
1415 prevent_tail_call(ret);
1419 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1420 int options, struct rusage __user *ru)
1424 if (options & ~(WNOHANG|WUNTRACED|__WNOTHREAD|__WCLONE|__WALL))
1426 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1428 /* avoid REGPARM breakage on x86: */
1429 prevent_tail_call(ret);
1433 #ifdef __ARCH_WANT_SYS_WAITPID
1436 * sys_waitpid() remains for compatibility. waitpid() should be
1437 * implemented by calling sys_wait4() from libc.a.
1439 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1441 return sys_wait4(pid, stat_addr, options, NULL);