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/acct.h>
19 #include <linux/file.h>
20 #include <linux/binfmts.h>
21 #include <linux/ptrace.h>
22 #include <linux/profile.h>
23 #include <linux/mount.h>
24 #include <linux/proc_fs.h>
25 #include <linux/mempolicy.h>
26 #include <linux/ckrm.h>
27 #include <linux/ckrm_tsk.h>
28 #include <linux/vs_limit.h>
29 #include <linux/ckrm_mem.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->cacct.nr_threads);
47 atomic_dec(&p->vx_info->limit.rcur[RLIMIT_NPROC]);
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 BUG_ON(p->state < TASK_ZOMBIE);
70 atomic_dec(&p->user->processes);
71 spin_lock(&p->proc_lock);
72 proc_dentry = proc_pid_unhash(p);
73 write_lock_irq(&tasklist_lock);
74 if (unlikely(p->ptrace))
76 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
82 * If we are the last non-leader member of the thread
83 * group, and the leader is zombie, then notify the
84 * group leader's parent process. (if it wants notification.)
87 leader = p->group_leader;
88 if (leader != p && thread_group_empty(leader) && leader->state == TASK_ZOMBIE) {
89 BUG_ON(leader->exit_signal == -1);
90 do_notify_parent(leader, leader->exit_signal);
92 * If we were the last child thread and the leader has
93 * exited already, and the leader's parent ignores SIGCHLD,
94 * then we are the one who should release the leader.
96 * do_notify_parent() will have marked it self-reaping in
99 zap_leader = (leader->exit_signal == -1);
102 p->parent->cutime += p->utime + p->cutime;
103 p->parent->cstime += p->stime + p->cstime;
104 p->parent->cmin_flt += p->min_flt + p->cmin_flt;
105 p->parent->cmaj_flt += p->maj_flt + p->cmaj_flt;
106 p->parent->cnvcsw += p->nvcsw + p->cnvcsw;
107 p->parent->cnivcsw += p->nivcsw + p->cnivcsw;
109 write_unlock_irq(&tasklist_lock);
110 spin_unlock(&p->proc_lock);
116 if (unlikely(zap_leader))
120 /* we are using it only for SMP init */
122 void unhash_process(struct task_struct *p)
124 struct dentry *proc_dentry;
126 spin_lock(&p->proc_lock);
127 proc_dentry = proc_pid_unhash(p);
128 write_lock_irq(&tasklist_lock);
130 write_unlock_irq(&tasklist_lock);
131 spin_unlock(&p->proc_lock);
132 proc_pid_flush(proc_dentry);
136 * This checks not only the pgrp, but falls back on the pid if no
137 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
140 int session_of_pgrp(int pgrp)
142 struct task_struct *p;
147 read_lock(&tasklist_lock);
148 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid)
149 if (p->signal->session > 0) {
150 sid = p->signal->session;
153 p = find_task_by_pid(pgrp);
155 sid = p->signal->session;
157 read_unlock(&tasklist_lock);
163 * Determine if a process group is "orphaned", according to the POSIX
164 * definition in 2.2.2.52. Orphaned process groups are not to be affected
165 * by terminal-generated stop signals. Newly orphaned process groups are
166 * to receive a SIGHUP and a SIGCONT.
168 * "I ask you, have you ever known what it is to be an orphan?"
170 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
172 struct task_struct *p;
177 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
178 if (p == ignored_task
179 || p->state >= TASK_ZOMBIE
180 || p->real_parent->pid == 1)
182 if (process_group(p->real_parent) != pgrp
183 && p->real_parent->signal->session == p->signal->session) {
188 return ret; /* (sighing) "Often!" */
191 int is_orphaned_pgrp(int pgrp)
195 read_lock(&tasklist_lock);
196 retval = will_become_orphaned_pgrp(pgrp, NULL);
197 read_unlock(&tasklist_lock);
202 static inline int has_stopped_jobs(int pgrp)
205 struct task_struct *p;
209 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
210 if (p->state != TASK_STOPPED)
213 /* If p is stopped by a debugger on a signal that won't
214 stop it, then don't count p as stopped. This isn't
215 perfect but it's a good approximation. */
216 if (unlikely (p->ptrace)
217 && p->exit_code != SIGSTOP
218 && p->exit_code != SIGTSTP
219 && p->exit_code != SIGTTOU
220 && p->exit_code != SIGTTIN)
230 * reparent_to_init() - Reparent the calling kernel thread to the init task.
232 * If a kernel thread is launched as a result of a system call, or if
233 * it ever exits, it should generally reparent itself to init so that
234 * it is correctly cleaned up on exit.
236 * The various task state such as scheduling policy and priority may have
237 * been inherited from a user process, so we reset them to sane values here.
239 * NOTE that reparent_to_init() gives the caller full capabilities.
241 void reparent_to_init(void)
243 write_lock_irq(&tasklist_lock);
245 ptrace_unlink(current);
246 /* Reparent to init */
247 REMOVE_LINKS(current);
248 /* FIXME handle vchild_reaper/initpid */
249 current->parent = child_reaper;
250 current->real_parent = child_reaper;
253 /* Set the exit signal to SIGCHLD so we signal init on exit */
254 current->exit_signal = SIGCHLD;
256 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
257 set_user_nice(current, 0);
261 security_task_reparent_to_init(current);
262 memcpy(current->rlim, init_task.rlim, sizeof(*(current->rlim)));
263 atomic_inc(&(INIT_USER->__count));
264 switch_uid(INIT_USER);
266 write_unlock_irq(&tasklist_lock);
269 void __set_special_pids(pid_t session, pid_t pgrp)
271 struct task_struct *curr = current;
273 if (curr->signal->session != session) {
274 detach_pid(curr, PIDTYPE_SID);
275 curr->signal->session = session;
276 attach_pid(curr, PIDTYPE_SID, session);
278 if (process_group(curr) != pgrp) {
279 detach_pid(curr, PIDTYPE_PGID);
280 curr->signal->pgrp = pgrp;
281 attach_pid(curr, PIDTYPE_PGID, pgrp);
285 void set_special_pids(pid_t session, pid_t pgrp)
287 write_lock_irq(&tasklist_lock);
288 __set_special_pids(session, pgrp);
289 write_unlock_irq(&tasklist_lock);
293 * Let kernel threads use this to say that they
294 * allow a certain signal (since daemonize() will
295 * have disabled all of them by default).
297 int allow_signal(int sig)
299 if (sig < 1 || sig > _NSIG)
302 spin_lock_irq(¤t->sighand->siglock);
303 sigdelset(¤t->blocked, sig);
305 /* Kernel threads handle their own signals.
306 Let the signal code know it'll be handled, so
307 that they don't get converted to SIGKILL or
308 just silently dropped */
309 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
312 spin_unlock_irq(¤t->sighand->siglock);
316 EXPORT_SYMBOL(allow_signal);
318 int disallow_signal(int sig)
320 if (sig < 1 || sig > _NSIG)
323 spin_lock_irq(¤t->sighand->siglock);
324 sigaddset(¤t->blocked, sig);
326 spin_unlock_irq(¤t->sighand->siglock);
330 EXPORT_SYMBOL(disallow_signal);
333 * Put all the gunge required to become a kernel thread without
334 * attached user resources in one place where it belongs.
337 void daemonize(const char *name, ...)
340 struct fs_struct *fs;
343 va_start(args, name);
344 vsnprintf(current->comm, sizeof(current->comm), name, args);
348 * If we were started as result of loading a module, close all of the
349 * user space pages. We don't need them, and if we didn't close them
350 * they would be locked into memory.
354 set_special_pids(1, 1);
355 current->signal->tty = NULL;
357 /* Block and flush all signals */
358 sigfillset(&blocked);
359 sigprocmask(SIG_BLOCK, &blocked, NULL);
360 flush_signals(current);
362 /* Become as one with the init task */
364 exit_fs(current); /* current->fs->count--; */
367 atomic_inc(&fs->count);
369 current->files = init_task.files;
370 atomic_inc(¤t->files->count);
375 EXPORT_SYMBOL(daemonize);
377 static inline void close_files(struct files_struct * files)
385 if (i >= files->max_fdset || i >= files->max_fds)
387 set = files->open_fds->fds_bits[j++];
390 struct file * file = xchg(&files->fd[i], NULL);
392 filp_close(file, files);
395 // vx_openfd_dec(fd);
403 struct files_struct *get_files_struct(struct task_struct *task)
405 struct files_struct *files;
410 atomic_inc(&files->count);
416 void fastcall put_files_struct(struct files_struct *files)
418 if (atomic_dec_and_test(&files->count)) {
421 * Free the fd and fdset arrays if we expanded them.
423 if (files->fd != &files->fd_array[0])
424 free_fd_array(files->fd, files->max_fds);
425 if (files->max_fdset > __FD_SETSIZE) {
426 free_fdset(files->open_fds, files->max_fdset);
427 free_fdset(files->close_on_exec, files->max_fdset);
429 kmem_cache_free(files_cachep, files);
433 EXPORT_SYMBOL(put_files_struct);
435 static inline void __exit_files(struct task_struct *tsk)
437 struct files_struct * files = tsk->files;
443 put_files_struct(files);
447 void exit_files(struct task_struct *tsk)
452 static inline void __put_fs_struct(struct fs_struct *fs)
454 /* No need to hold fs->lock if we are killing it */
455 if (atomic_dec_and_test(&fs->count)) {
462 mntput(fs->altrootmnt);
464 kmem_cache_free(fs_cachep, fs);
468 void put_fs_struct(struct fs_struct *fs)
473 static inline void __exit_fs(struct task_struct *tsk)
475 struct fs_struct * fs = tsk->fs;
485 void exit_fs(struct task_struct *tsk)
490 EXPORT_SYMBOL_GPL(exit_fs);
493 * Turn us into a lazy TLB process if we
496 static inline void __exit_mm(struct task_struct * tsk)
498 struct mm_struct *mm = tsk->mm;
504 * Serialize with any possible pending coredump.
505 * We must hold mmap_sem around checking core_waiters
506 * and clearing tsk->mm. The core-inducing thread
507 * will increment core_waiters for each thread in the
508 * group with ->mm != NULL.
510 down_read(&mm->mmap_sem);
511 if (mm->core_waiters) {
512 up_read(&mm->mmap_sem);
513 down_write(&mm->mmap_sem);
514 if (!--mm->core_waiters)
515 complete(mm->core_startup_done);
516 up_write(&mm->mmap_sem);
518 wait_for_completion(&mm->core_done);
519 down_read(&mm->mmap_sem);
521 atomic_inc(&mm->mm_count);
522 if (mm != tsk->active_mm) BUG();
523 /* more a memory barrier than a real lock */
526 up_read(&mm->mmap_sem);
527 #ifdef CONFIG_CKRM_RES_MEM
528 spin_lock(&mm->peertask_lock);
529 list_del_init(&tsk->mm_peers);
530 ckrm_mem_evaluate_mm(mm);
531 spin_unlock(&mm->peertask_lock);
533 enter_lazy_tlb(mm, current);
538 void exit_mm(struct task_struct *tsk)
543 EXPORT_SYMBOL(exit_mm);
545 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
548 * Make sure we're not reparenting to ourselves and that
549 * the parent is not a zombie.
551 if (p == reaper || reaper->state >= TASK_ZOMBIE)
552 p->real_parent = child_reaper;
554 p->real_parent = reaper;
555 if (p->parent == p->real_parent)
559 static inline void reparent_thread(task_t *p, task_t *father, int traced)
561 /* We don't want people slaying init. */
562 if (p->exit_signal != -1)
563 p->exit_signal = SIGCHLD;
566 if (p->pdeath_signal)
567 /* We already hold the tasklist_lock here. */
568 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
570 /* Move the child from its dying parent to the new one. */
571 if (unlikely(traced)) {
572 /* Preserve ptrace links if someone else is tracing this child. */
573 list_del_init(&p->ptrace_list);
574 if (p->parent != p->real_parent)
575 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
577 /* If this child is being traced, then we're the one tracing it
578 * anyway, so let go of it.
581 list_del_init(&p->sibling);
582 p->parent = p->real_parent;
583 list_add_tail(&p->sibling, &p->parent->children);
585 /* If we'd notified the old parent about this child's death,
586 * also notify the new parent.
588 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
589 thread_group_empty(p))
590 do_notify_parent(p, p->exit_signal);
594 * process group orphan check
595 * Case ii: Our child is in a different pgrp
596 * than we are, and it was the only connection
597 * outside, so the child pgrp is now orphaned.
599 if ((process_group(p) != process_group(father)) &&
600 (p->signal->session == father->signal->session)) {
601 int pgrp = process_group(p);
603 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
604 __kill_pg_info(SIGHUP, (void *)1, pgrp);
605 __kill_pg_info(SIGCONT, (void *)1, pgrp);
611 * When we die, we re-parent all our children.
612 * Try to give them to another thread in our thread
613 * group, and if no such member exists, give it to
614 * the global child reaper process (ie "init")
616 static inline void forget_original_parent(struct task_struct * father,
617 struct list_head *to_release)
619 struct task_struct *p, *reaper = father;
620 struct list_head *_p, *_n;
622 /* FIXME handle vchild_reaper/initpid */
623 reaper = father->group_leader;
624 if (reaper == father)
625 reaper = child_reaper;
628 * There are only two places where our children can be:
630 * - in our child list
631 * - in our ptraced child list
633 * Search them and reparent children.
635 list_for_each_safe(_p, _n, &father->children) {
637 p = list_entry(_p,struct task_struct,sibling);
641 /* if father isn't the real parent, then ptrace must be enabled */
642 BUG_ON(father != p->real_parent && !ptrace);
644 if (father == p->real_parent) {
645 /* reparent with a reaper, real father it's us */
646 choose_new_parent(p, reaper, child_reaper);
647 reparent_thread(p, father, 0);
649 /* reparent ptraced task to its real parent */
651 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
652 thread_group_empty(p))
653 do_notify_parent(p, p->exit_signal);
657 * if the ptraced child is a zombie with exit_signal == -1
658 * we must collect it before we exit, or it will remain
659 * zombie forever since we prevented it from self-reap itself
660 * while it was being traced by us, to be able to see it in wait4.
662 if (unlikely(ptrace && p->state == TASK_ZOMBIE && p->exit_signal == -1))
663 list_add(&p->ptrace_list, to_release);
665 list_for_each_safe(_p, _n, &father->ptrace_children) {
666 p = list_entry(_p,struct task_struct,ptrace_list);
667 choose_new_parent(p, reaper, child_reaper);
668 reparent_thread(p, father, 1);
673 * Send signals to all our closest relatives so that they know
674 * to properly mourn us..
676 static void exit_notify(struct task_struct *tsk)
679 struct task_struct *t;
680 struct list_head ptrace_dead, *_p, *_n;
684 if (signal_pending(tsk) && !tsk->signal->group_exit
685 && !thread_group_empty(tsk)) {
687 * This occurs when there was a race between our exit
688 * syscall and a group signal choosing us as the one to
689 * wake up. It could be that we are the only thread
690 * alerted to check for pending signals, but another thread
691 * should be woken now to take the signal since we will not.
692 * Now we'll wake all the threads in the group just to make
693 * sure someone gets all the pending signals.
695 read_lock(&tasklist_lock);
696 spin_lock_irq(&tsk->sighand->siglock);
697 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
698 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
699 recalc_sigpending_tsk(t);
700 if (signal_pending(t))
701 signal_wake_up(t, 0);
703 spin_unlock_irq(&tsk->sighand->siglock);
704 read_unlock(&tasklist_lock);
707 write_lock_irq(&tasklist_lock);
710 * This does two things:
712 * A. Make init inherit all the child processes
713 * B. Check to see if any process groups have become orphaned
714 * as a result of our exiting, and if they have any stopped
715 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
718 INIT_LIST_HEAD(&ptrace_dead);
719 forget_original_parent(tsk, &ptrace_dead);
720 BUG_ON(!list_empty(&tsk->children));
721 BUG_ON(!list_empty(&tsk->ptrace_children));
724 * Check to see if any process groups have become orphaned
725 * as a result of our exiting, and if they have any stopped
726 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
728 * Case i: Our father is in a different pgrp than we are
729 * and we were the only connection outside, so our pgrp
730 * is about to become orphaned.
733 t = tsk->real_parent;
735 if ((process_group(t) != process_group(tsk)) &&
736 (t->signal->session == tsk->signal->session) &&
737 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
738 has_stopped_jobs(process_group(tsk))) {
739 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
740 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
743 /* Let father know we died
745 * Thread signals are configurable, but you aren't going to use
746 * that to send signals to arbitary processes.
747 * That stops right now.
749 * If the parent exec id doesn't match the exec id we saved
750 * when we started then we know the parent has changed security
753 * If our self_exec id doesn't match our parent_exec_id then
754 * we have changed execution domain as these two values started
755 * the same after a fork.
759 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
760 ( tsk->parent_exec_id != t->self_exec_id ||
761 tsk->self_exec_id != tsk->parent_exec_id)
762 && !capable(CAP_KILL))
763 tsk->exit_signal = SIGCHLD;
766 /* If something other than our normal parent is ptracing us, then
767 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
768 * only has special meaning to our real parent.
770 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
771 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
772 do_notify_parent(tsk, signal);
773 } else if (tsk->ptrace) {
774 do_notify_parent(tsk, SIGCHLD);
778 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
781 tsk->flags |= PF_DEAD;
784 * Clear these here so that update_process_times() won't try to deliver
785 * itimer, profile or rlimit signals to this task while it is in late exit.
787 tsk->it_virt_value = 0;
788 tsk->it_prof_value = 0;
789 tsk->rlim[RLIMIT_CPU].rlim_cur = RLIM_INFINITY;
792 * In the preemption case it must be impossible for the task
793 * to get runnable again, so use "_raw_" unlock to keep
794 * preempt_count elevated until we schedule().
796 * To avoid deadlock on SMP, interrupts must be unmasked. If we
797 * don't, subsequently called functions (e.g, wait_task_inactive()
798 * via release_task()) will spin, with interrupt flags
799 * unwittingly blocked, until the other task sleeps. That task
800 * may itself be waiting for smp_call_function() to answer and
801 * complete, and with interrupts blocked that will never happen.
803 _raw_write_unlock(&tasklist_lock);
806 list_for_each_safe(_p, _n, &ptrace_dead) {
808 t = list_entry(_p,struct task_struct,ptrace_list);
812 /* If the process is dead, release it - nobody will wait for it */
813 if (state == TASK_DEAD)
818 asmlinkage NORET_TYPE void do_exit(long code)
820 struct task_struct *tsk = current;
822 if (unlikely(in_interrupt()))
823 panic("Aiee, killing interrupt handler!");
824 if (unlikely(!tsk->pid))
825 panic("Attempted to kill the idle task!");
826 if (unlikely(tsk->pid == 1))
827 panic("Attempted to kill init!");
830 tsk->flags |= PF_EXITING;
831 del_timer_sync(&tsk->real_timer);
833 if (unlikely(in_atomic()))
834 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
835 current->comm, current->pid,
838 profile_exit_task(tsk);
840 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
841 current->ptrace_message = code;
842 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
846 if (current->tux_info) {
847 #ifdef CONFIG_TUX_DEBUG
848 printk("Possibly unexpected TUX-thread exit(%ld) at %p?\n",
849 code, __builtin_return_address(0));
861 if (tsk->signal->leader)
862 disassociate_ctty(1);
864 module_put(tsk->thread_info->exec_domain->module);
866 module_put(tsk->binfmt->module);
868 tsk->exit_code = code;
871 mpol_free(tsk->mempolicy);
872 tsk->mempolicy = NULL;
876 /* Avoid "noreturn function does return". */
880 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
888 EXPORT_SYMBOL(complete_and_exit);
890 asmlinkage long sys_exit(int error_code)
892 do_exit((error_code&0xff)<<8);
895 task_t fastcall *next_thread(const task_t *p)
897 const struct pid_link *link = p->pids + PIDTYPE_TGID;
898 const struct list_head *tmp, *head = &link->pidptr->task_list;
903 if (!spin_is_locked(&p->sighand->siglock) &&
904 !rwlock_is_locked(&tasklist_lock))
907 tmp = link->pid_chain.next;
911 return pid_task(tmp, PIDTYPE_TGID);
914 EXPORT_SYMBOL(next_thread);
917 * Take down every thread in the group. This is called by fatal signals
918 * as well as by sys_exit_group (below).
921 do_group_exit(int exit_code)
923 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
925 if (current->signal->group_exit)
926 exit_code = current->signal->group_exit_code;
927 else if (!thread_group_empty(current)) {
928 struct signal_struct *const sig = current->signal;
929 struct sighand_struct *const sighand = current->sighand;
930 read_lock(&tasklist_lock);
931 spin_lock_irq(&sighand->siglock);
933 /* Another thread got here before we took the lock. */
934 exit_code = sig->group_exit_code;
937 sig->group_exit_code = exit_code;
938 zap_other_threads(current);
940 spin_unlock_irq(&sighand->siglock);
941 read_unlock(&tasklist_lock);
949 * this kills every thread in the thread group. Note that any externally
950 * wait4()-ing process will get the correct exit code - even if this
951 * thread is not the thread group leader.
953 asmlinkage void sys_exit_group(int error_code)
955 do_group_exit((error_code & 0xff) << 8);
958 static int eligible_child(pid_t pid, int options, task_t *p)
964 if (process_group(p) != process_group(current))
966 } else if (pid != -1) {
967 if (process_group(p) != -pid)
972 * Do not consider detached threads that are
975 if (p->exit_signal == -1 && !p->ptrace)
978 /* Wait for all children (clone and not) if __WALL is set;
979 * otherwise, wait for clone children *only* if __WCLONE is
980 * set; otherwise, wait for non-clone children *only*. (Note:
981 * A "clone" child here is one that reports to its parent
982 * using a signal other than SIGCHLD.) */
983 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
984 && !(options & __WALL))
987 * Do not consider thread group leaders that are
988 * in a non-empty thread group:
990 if (current->tgid != p->tgid && delay_group_leader(p))
993 if (security_task_wait(p))
1000 * Handle sys_wait4 work for one task in state TASK_ZOMBIE. We hold
1001 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1002 * the lock and this task is uninteresting. If we return nonzero, we have
1003 * released the lock and the system call should return.
1005 static int wait_task_zombie(task_t *p, unsigned int __user *stat_addr, struct rusage __user *ru)
1007 unsigned long state;
1011 * Try to move the task's state to DEAD
1012 * only one thread is allowed to do this:
1014 state = xchg(&p->state, TASK_DEAD);
1015 if (state != TASK_ZOMBIE) {
1016 BUG_ON(state != TASK_DEAD);
1019 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1021 * This can only happen in a race with a ptraced thread
1022 * dying on another processor.
1027 * Now we are sure this task is interesting, and no other
1028 * thread can reap it because we set its state to TASK_DEAD.
1030 read_unlock(&tasklist_lock);
1032 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1033 if (!retval && stat_addr) {
1034 if (p->signal->group_exit)
1035 retval = put_user(p->signal->group_exit_code, stat_addr);
1037 retval = put_user(p->exit_code, stat_addr);
1040 p->state = TASK_ZOMBIE;
1044 if (p->real_parent != p->parent) {
1045 write_lock_irq(&tasklist_lock);
1046 /* Double-check with lock held. */
1047 if (p->real_parent != p->parent) {
1049 p->state = TASK_ZOMBIE;
1051 * If this is not a detached task, notify the parent. If it's
1052 * still not detached after that, don't release it now.
1054 if (p->exit_signal != -1) {
1055 do_notify_parent(p, p->exit_signal);
1056 if (p->exit_signal != -1)
1060 write_unlock_irq(&tasklist_lock);
1069 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1070 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1071 * the lock and this task is uninteresting. If we return nonzero, we have
1072 * released the lock and the system call should return.
1074 static int wait_task_stopped(task_t *p, int delayed_group_leader,
1075 unsigned int __user *stat_addr,
1076 struct rusage __user *ru)
1078 int retval, exit_code;
1082 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1083 p->signal && p->signal->group_stop_count > 0)
1085 * A group stop is in progress and this is the group leader.
1086 * We won't report until all threads have stopped.
1091 * Now we are pretty sure this task is interesting.
1092 * Make sure it doesn't get reaped out from under us while we
1093 * give up the lock and then examine it below. We don't want to
1094 * keep holding onto the tasklist_lock while we call getrusage and
1095 * possibly take page faults for user memory.
1098 read_unlock(&tasklist_lock);
1099 write_lock_irq(&tasklist_lock);
1102 * This uses xchg to be atomic with the thread resuming and setting
1103 * it. It must also be done with the write lock held to prevent a
1104 * race with the TASK_ZOMBIE case.
1106 exit_code = xchg(&p->exit_code, 0);
1107 if (unlikely(p->state > TASK_STOPPED)) {
1109 * The task resumed and then died. Let the next iteration
1110 * catch it in TASK_ZOMBIE. Note that exit_code might
1111 * already be zero here if it resumed and did _exit(0).
1112 * The task itself is dead and won't touch exit_code again;
1113 * other processors in this function are locked out.
1115 p->exit_code = exit_code;
1118 if (unlikely(exit_code == 0)) {
1120 * Another thread in this function got to it first, or it
1121 * resumed, or it resumed and then died.
1123 write_unlock_irq(&tasklist_lock);
1125 read_lock(&tasklist_lock);
1129 /* move to end of parent's list to avoid starvation */
1131 add_parent(p, p->parent);
1133 write_unlock_irq(&tasklist_lock);
1135 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1136 if (!retval && stat_addr)
1137 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1146 asmlinkage long sys_wait4(pid_t pid,unsigned int __user *stat_addr, int options, struct rusage __user *ru)
1148 DECLARE_WAITQUEUE(wait, current);
1149 struct task_struct *tsk;
1152 if (options & ~(WNOHANG|WUNTRACED|__WNOTHREAD|__WCLONE|__WALL))
1155 add_wait_queue(¤t->wait_chldexit,&wait);
1158 current->state = TASK_INTERRUPTIBLE;
1159 read_lock(&tasklist_lock);
1162 struct task_struct *p;
1163 struct list_head *_p;
1166 list_for_each(_p,&tsk->children) {
1167 p = list_entry(_p,struct task_struct,sibling);
1169 ret = eligible_child(pid, options, p);
1176 if (!(options & WUNTRACED) &&
1177 !(p->ptrace & PT_PTRACED))
1179 retval = wait_task_stopped(p, ret == 2,
1181 if (retval != 0) /* He released the lock. */
1186 * Eligible but we cannot release it yet:
1190 retval = wait_task_zombie(p, stat_addr, ru);
1191 if (retval != 0) /* He released the lock. */
1197 list_for_each (_p,&tsk->ptrace_children) {
1198 p = list_entry(_p,struct task_struct,ptrace_list);
1199 if (!eligible_child(pid, options, p))
1205 if (options & __WNOTHREAD)
1207 tsk = next_thread(tsk);
1208 if (tsk->signal != current->signal)
1210 } while (tsk != current);
1211 read_unlock(&tasklist_lock);
1214 if (options & WNOHANG)
1216 retval = -ERESTARTSYS;
1217 if (signal_pending(current))
1224 current->state = TASK_RUNNING;
1225 remove_wait_queue(¤t->wait_chldexit,&wait);
1229 #ifdef __ARCH_WANT_SYS_WAITPID
1232 * sys_waitpid() remains for compatibility. waitpid() should be
1233 * implemented by calling sys_wait4() from libc.a.
1235 asmlinkage long sys_waitpid(pid_t pid, unsigned __user *stat_addr, int options)
1237 return sys_wait4(pid, stat_addr, options, NULL);