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>
26 #include <asm/uaccess.h>
27 #include <asm/pgtable.h>
28 #include <asm/mmu_context.h>
30 extern void sem_exit (void);
31 extern struct task_struct *child_reaper;
33 int getrusage(struct task_struct *, int, struct rusage *);
35 static void __unhash_process(struct task_struct *p)
38 /* tasklist_lock is held, is this sufficient? */
40 atomic_dec(&p->vx_info->cacct.nr_threads);
41 atomic_dec(&p->vx_info->limit.res[RLIMIT_NPROC]);
43 detach_pid(p, PIDTYPE_PID);
44 detach_pid(p, PIDTYPE_TGID);
45 if (thread_group_leader(p)) {
46 detach_pid(p, PIDTYPE_PGID);
47 detach_pid(p, PIDTYPE_SID);
49 __get_cpu_var(process_counts)--;
55 void release_task(struct task_struct * p)
59 struct dentry *proc_dentry;
62 BUG_ON(p->state < TASK_ZOMBIE);
64 atomic_dec(&p->user->processes);
65 spin_lock(&p->proc_lock);
66 proc_dentry = proc_pid_unhash(p);
67 write_lock_irq(&tasklist_lock);
68 if (unlikely(p->ptrace))
70 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
76 * If we are the last non-leader member of the thread
77 * group, and the leader is zombie, then notify the
78 * group leader's parent process. (if it wants notification.)
81 leader = p->group_leader;
82 if (leader != p && thread_group_empty(leader) && leader->state == TASK_ZOMBIE) {
83 BUG_ON(leader->exit_signal == -1);
84 do_notify_parent(leader, leader->exit_signal);
86 * If we were the last child thread and the leader has
87 * exited already, and the leader's parent ignores SIGCHLD,
88 * then we are the one who should release the leader.
90 * do_notify_parent() will have marked it self-reaping in
93 zap_leader = (leader->exit_signal == -1);
96 p->parent->cutime += p->utime + p->cutime;
97 p->parent->cstime += p->stime + p->cstime;
98 p->parent->cmin_flt += p->min_flt + p->cmin_flt;
99 p->parent->cmaj_flt += p->maj_flt + p->cmaj_flt;
100 p->parent->cnvcsw += p->nvcsw + p->cnvcsw;
101 p->parent->cnivcsw += p->nivcsw + p->cnivcsw;
103 write_unlock_irq(&tasklist_lock);
104 spin_unlock(&p->proc_lock);
105 proc_pid_flush(proc_dentry);
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;
141 read_lock(&tasklist_lock);
142 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid)
143 if (p->signal->session > 0) {
144 sid = p->signal->session;
147 p = find_task_by_pid(pgrp);
149 sid = p->signal->session;
151 read_unlock(&tasklist_lock);
157 * Determine if a process group is "orphaned", according to the POSIX
158 * definition in 2.2.2.52. Orphaned process groups are not to be affected
159 * by terminal-generated stop signals. Newly orphaned process groups are
160 * to receive a SIGHUP and a SIGCONT.
162 * "I ask you, have you ever known what it is to be an orphan?"
164 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
166 struct task_struct *p;
171 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
172 if (p == ignored_task
173 || p->state >= TASK_ZOMBIE
174 || p->real_parent->pid == 1)
176 if (process_group(p->real_parent) != pgrp
177 && p->real_parent->signal->session == p->signal->session) {
182 return ret; /* (sighing) "Often!" */
185 int is_orphaned_pgrp(int pgrp)
189 read_lock(&tasklist_lock);
190 retval = will_become_orphaned_pgrp(pgrp, NULL);
191 read_unlock(&tasklist_lock);
196 static inline int has_stopped_jobs(int pgrp)
199 struct task_struct *p;
203 for_each_task_pid(pgrp, PIDTYPE_PGID, p, l, pid) {
204 if (p->state != TASK_STOPPED)
207 /* If p is stopped by a debugger on a signal that won't
208 stop it, then don't count p as stopped. This isn't
209 perfect but it's a good approximation. */
210 if (unlikely (p->ptrace)
211 && p->exit_code != SIGSTOP
212 && p->exit_code != SIGTSTP
213 && p->exit_code != SIGTTOU
214 && p->exit_code != SIGTTIN)
224 * reparent_to_init() - Reparent the calling kernel thread to the init task.
226 * If a kernel thread is launched as a result of a system call, or if
227 * it ever exits, it should generally reparent itself to init so that
228 * it is correctly cleaned up on exit.
230 * The various task state such as scheduling policy and priority may have
231 * been inherited from a user process, so we reset them to sane values here.
233 * NOTE that reparent_to_init() gives the caller full capabilities.
235 void reparent_to_init(void)
237 write_lock_irq(&tasklist_lock);
239 ptrace_unlink(current);
240 /* Reparent to init */
241 REMOVE_LINKS(current);
242 /* FIXME handle vchild_reaper/initpid */
243 current->parent = child_reaper;
244 current->real_parent = child_reaper;
247 /* Set the exit signal to SIGCHLD so we signal init on exit */
248 current->exit_signal = SIGCHLD;
250 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
251 set_user_nice(current, 0);
255 security_task_reparent_to_init(current);
256 memcpy(current->rlim, init_task.rlim, sizeof(*(current->rlim)));
257 atomic_inc(&(INIT_USER->__count));
258 switch_uid(INIT_USER);
260 write_unlock_irq(&tasklist_lock);
263 void __set_special_pids(pid_t session, pid_t pgrp)
265 struct task_struct *curr = current;
267 if (curr->signal->session != session) {
268 detach_pid(curr, PIDTYPE_SID);
269 curr->signal->session = session;
270 attach_pid(curr, PIDTYPE_SID, session);
272 if (process_group(curr) != pgrp) {
273 detach_pid(curr, PIDTYPE_PGID);
274 curr->signal->pgrp = pgrp;
275 attach_pid(curr, PIDTYPE_PGID, pgrp);
279 void set_special_pids(pid_t session, pid_t pgrp)
281 write_lock_irq(&tasklist_lock);
282 __set_special_pids(session, pgrp);
283 write_unlock_irq(&tasklist_lock);
287 * Let kernel threads use this to say that they
288 * allow a certain signal (since daemonize() will
289 * have disabled all of them by default).
291 int allow_signal(int sig)
293 if (sig < 1 || sig > _NSIG)
296 spin_lock_irq(¤t->sighand->siglock);
297 sigdelset(¤t->blocked, sig);
299 /* Kernel threads handle their own signals.
300 Let the signal code know it'll be handled, so
301 that they don't get converted to SIGKILL or
302 just silently dropped */
303 current->sighand->action[(sig)-1].sa.sa_handler = (void *)2;
306 spin_unlock_irq(¤t->sighand->siglock);
310 EXPORT_SYMBOL(allow_signal);
312 int disallow_signal(int sig)
314 if (sig < 1 || sig > _NSIG)
317 spin_lock_irq(¤t->sighand->siglock);
318 sigaddset(¤t->blocked, sig);
320 spin_unlock_irq(¤t->sighand->siglock);
324 EXPORT_SYMBOL(disallow_signal);
327 * Put all the gunge required to become a kernel thread without
328 * attached user resources in one place where it belongs.
331 void daemonize(const char *name, ...)
334 struct fs_struct *fs;
337 va_start(args, name);
338 vsnprintf(current->comm, sizeof(current->comm), name, args);
342 * If we were started as result of loading a module, close all of the
343 * user space pages. We don't need them, and if we didn't close them
344 * they would be locked into memory.
348 set_special_pids(1, 1);
349 current->signal->tty = NULL;
351 /* Block and flush all signals */
352 sigfillset(&blocked);
353 sigprocmask(SIG_BLOCK, &blocked, NULL);
354 flush_signals(current);
356 /* Become as one with the init task */
358 exit_fs(current); /* current->fs->count--; */
361 atomic_inc(&fs->count);
363 current->files = init_task.files;
364 atomic_inc(¤t->files->count);
369 EXPORT_SYMBOL(daemonize);
371 static inline void close_files(struct files_struct * files)
379 if (i >= files->max_fdset || i >= files->max_fds)
381 set = files->open_fds->fds_bits[j++];
384 struct file * file = xchg(&files->fd[i], NULL);
386 filp_close(file, files);
395 struct files_struct *get_files_struct(struct task_struct *task)
397 struct files_struct *files;
402 atomic_inc(&files->count);
408 void fastcall put_files_struct(struct files_struct *files)
410 if (atomic_dec_and_test(&files->count)) {
413 * Free the fd and fdset arrays if we expanded them.
415 if (files->fd != &files->fd_array[0])
416 free_fd_array(files->fd, files->max_fds);
417 if (files->max_fdset > __FD_SETSIZE) {
418 free_fdset(files->open_fds, files->max_fdset);
419 free_fdset(files->close_on_exec, files->max_fdset);
421 kmem_cache_free(files_cachep, files);
425 EXPORT_SYMBOL(put_files_struct);
427 static inline void __exit_files(struct task_struct *tsk)
429 struct files_struct * files = tsk->files;
435 put_files_struct(files);
439 void exit_files(struct task_struct *tsk)
444 static inline void __put_fs_struct(struct fs_struct *fs)
446 /* No need to hold fs->lock if we are killing it */
447 if (atomic_dec_and_test(&fs->count)) {
454 mntput(fs->altrootmnt);
456 kmem_cache_free(fs_cachep, fs);
460 void put_fs_struct(struct fs_struct *fs)
465 static inline void __exit_fs(struct task_struct *tsk)
467 struct fs_struct * fs = tsk->fs;
477 void exit_fs(struct task_struct *tsk)
482 EXPORT_SYMBOL_GPL(exit_fs);
485 * Turn us into a lazy TLB process if we
488 static inline void __exit_mm(struct task_struct * tsk)
490 struct mm_struct *mm = tsk->mm;
496 * Serialize with any possible pending coredump.
497 * We must hold mmap_sem around checking core_waiters
498 * and clearing tsk->mm. The core-inducing thread
499 * will increment core_waiters for each thread in the
500 * group with ->mm != NULL.
502 down_read(&mm->mmap_sem);
503 if (mm->core_waiters) {
504 up_read(&mm->mmap_sem);
505 down_write(&mm->mmap_sem);
506 if (!--mm->core_waiters)
507 complete(mm->core_startup_done);
508 up_write(&mm->mmap_sem);
510 wait_for_completion(&mm->core_done);
511 down_read(&mm->mmap_sem);
513 atomic_inc(&mm->mm_count);
514 if (mm != tsk->active_mm) BUG();
515 /* more a memory barrier than a real lock */
518 up_read(&mm->mmap_sem);
519 enter_lazy_tlb(mm, current);
524 void exit_mm(struct task_struct *tsk)
529 EXPORT_SYMBOL(exit_mm);
531 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
534 * Make sure we're not reparenting to ourselves and that
535 * the parent is not a zombie.
537 if (p == reaper || reaper->state >= TASK_ZOMBIE)
538 p->real_parent = child_reaper;
540 p->real_parent = reaper;
541 if (p->parent == p->real_parent)
545 static inline void reparent_thread(task_t *p, task_t *father, int traced)
547 /* We don't want people slaying init. */
548 if (p->exit_signal != -1)
549 p->exit_signal = SIGCHLD;
552 if (p->pdeath_signal)
553 /* We already hold the tasklist_lock here. */
554 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
556 /* Move the child from its dying parent to the new one. */
557 if (unlikely(traced)) {
558 /* Preserve ptrace links if someone else is tracing this child. */
559 list_del_init(&p->ptrace_list);
560 if (p->parent != p->real_parent)
561 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
563 /* If this child is being traced, then we're the one tracing it
564 * anyway, so let go of it.
567 list_del_init(&p->sibling);
568 p->parent = p->real_parent;
569 list_add_tail(&p->sibling, &p->parent->children);
571 /* If we'd notified the old parent about this child's death,
572 * also notify the new parent.
574 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
575 thread_group_empty(p))
576 do_notify_parent(p, p->exit_signal);
580 * process group orphan check
581 * Case ii: Our child is in a different pgrp
582 * than we are, and it was the only connection
583 * outside, so the child pgrp is now orphaned.
585 if ((process_group(p) != process_group(father)) &&
586 (p->signal->session == father->signal->session)) {
587 int pgrp = process_group(p);
589 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
590 __kill_pg_info(SIGHUP, (void *)1, pgrp);
591 __kill_pg_info(SIGCONT, (void *)1, pgrp);
597 * When we die, we re-parent all our children.
598 * Try to give them to another thread in our thread
599 * group, and if no such member exists, give it to
600 * the global child reaper process (ie "init")
602 static inline void forget_original_parent(struct task_struct * father)
604 struct task_struct *p, *reaper = father;
605 struct list_head *_p, *_n;
607 /* FIXME handle vchild_reaper/initpid */
608 reaper = father->group_leader;
609 if (reaper == father)
610 reaper = child_reaper;
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) {
621 p = list_entry(_p,struct task_struct,sibling);
622 if (father == p->real_parent) {
623 choose_new_parent(p, reaper, child_reaper);
624 reparent_thread(p, father, 0);
627 if (p->state == TASK_ZOMBIE && p->exit_signal != -1 &&
628 thread_group_empty(p))
629 do_notify_parent(p, p->exit_signal);
632 list_for_each_safe(_p, _n, &father->ptrace_children) {
633 p = list_entry(_p,struct task_struct,ptrace_list);
634 choose_new_parent(p, reaper, child_reaper);
635 reparent_thread(p, father, 1);
640 * Send signals to all our closest relatives so that they know
641 * to properly mourn us..
643 static void exit_notify(struct task_struct *tsk)
646 struct task_struct *t;
648 if (signal_pending(tsk) && !tsk->signal->group_exit
649 && !thread_group_empty(tsk)) {
651 * This occurs when there was a race between our exit
652 * syscall and a group signal choosing us as the one to
653 * wake up. It could be that we are the only thread
654 * alerted to check for pending signals, but another thread
655 * should be woken now to take the signal since we will not.
656 * Now we'll wake all the threads in the group just to make
657 * sure someone gets all the pending signals.
659 read_lock(&tasklist_lock);
660 spin_lock_irq(&tsk->sighand->siglock);
661 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
662 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
663 recalc_sigpending_tsk(t);
664 if (signal_pending(t))
665 signal_wake_up(t, 0);
667 spin_unlock_irq(&tsk->sighand->siglock);
668 read_unlock(&tasklist_lock);
671 write_lock_irq(&tasklist_lock);
674 * This does two things:
676 * A. Make init inherit all the child processes
677 * B. Check to see if any process groups have become orphaned
678 * as a result of our exiting, and if they have any stopped
679 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
682 forget_original_parent(tsk);
683 BUG_ON(!list_empty(&tsk->children));
686 * Check to see if any process groups have become orphaned
687 * as a result of our exiting, and if they have any stopped
688 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
690 * Case i: Our father is in a different pgrp than we are
691 * and we were the only connection outside, so our pgrp
692 * is about to become orphaned.
695 t = tsk->real_parent;
697 if ((process_group(t) != process_group(tsk)) &&
698 (t->signal->session == tsk->signal->session) &&
699 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
700 has_stopped_jobs(process_group(tsk))) {
701 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
702 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
705 /* Let father know we died
707 * Thread signals are configurable, but you aren't going to use
708 * that to send signals to arbitary processes.
709 * That stops right now.
711 * If the parent exec id doesn't match the exec id we saved
712 * when we started then we know the parent has changed security
715 * If our self_exec id doesn't match our parent_exec_id then
716 * we have changed execution domain as these two values started
717 * the same after a fork.
721 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
722 ( tsk->parent_exec_id != t->self_exec_id ||
723 tsk->self_exec_id != tsk->parent_exec_id)
724 && !capable(CAP_KILL))
725 tsk->exit_signal = SIGCHLD;
728 /* If something other than our normal parent is ptracing us, then
729 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
730 * only has special meaning to our real parent.
732 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
733 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
734 do_notify_parent(tsk, signal);
735 } else if (tsk->ptrace) {
736 do_notify_parent(tsk, SIGCHLD);
740 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
743 tsk->flags |= PF_DEAD;
746 * In the preemption case it must be impossible for the task
747 * to get runnable again, so use "_raw_" unlock to keep
748 * preempt_count elevated until we schedule().
750 * To avoid deadlock on SMP, interrupts must be unmasked. If we
751 * don't, subsequently called functions (e.g, wait_task_inactive()
752 * via release_task()) will spin, with interrupt flags
753 * unwittingly blocked, until the other task sleeps. That task
754 * may itself be waiting for smp_call_function() to answer and
755 * complete, and with interrupts blocked that will never happen.
757 _raw_write_unlock(&tasklist_lock);
760 /* If the process is dead, release it - nobody will wait for it */
761 if (state == TASK_DEAD)
766 asmlinkage NORET_TYPE void do_exit(long code)
768 struct task_struct *tsk = current;
770 if (unlikely(in_interrupt()))
771 panic("Aiee, killing interrupt handler!");
772 if (unlikely(!tsk->pid))
773 panic("Attempted to kill the idle task!");
774 if (unlikely(tsk->pid == 1))
775 panic("Attempted to kill init!");
778 tsk->flags |= PF_EXITING;
779 del_timer_sync(&tsk->real_timer);
781 if (unlikely(in_atomic()))
782 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
783 current->comm, current->pid,
786 profile_exit_task(tsk);
788 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
789 current->ptrace_message = code;
790 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
802 if (tsk->signal->leader)
803 disassociate_ctty(1);
805 module_put(tsk->thread_info->exec_domain->module);
807 module_put(tsk->binfmt->module);
809 tsk->exit_code = code;
813 /* Avoid "noreturn function does return". */
817 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
825 EXPORT_SYMBOL(complete_and_exit);
827 asmlinkage long sys_exit(int error_code)
829 do_exit((error_code&0xff)<<8);
832 task_t fastcall *next_thread(task_t *p)
834 struct pid_link *link = p->pids + PIDTYPE_TGID;
835 struct list_head *tmp, *head = &link->pidptr->task_list;
840 if (!spin_is_locked(&p->sighand->siglock) &&
841 !rwlock_is_locked(&tasklist_lock))
844 tmp = link->pid_chain.next;
848 return pid_task(tmp, PIDTYPE_TGID);
851 EXPORT_SYMBOL(next_thread);
854 * Take down every thread in the group. This is called by fatal signals
855 * as well as by sys_exit_group (below).
858 do_group_exit(int exit_code)
860 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
862 if (current->signal->group_exit)
863 exit_code = current->signal->group_exit_code;
864 else if (!thread_group_empty(current)) {
865 struct signal_struct *const sig = current->signal;
866 struct sighand_struct *const sighand = current->sighand;
867 read_lock(&tasklist_lock);
868 spin_lock_irq(&sighand->siglock);
870 /* Another thread got here before we took the lock. */
871 exit_code = sig->group_exit_code;
874 sig->group_exit_code = exit_code;
875 zap_other_threads(current);
877 spin_unlock_irq(&sighand->siglock);
878 read_unlock(&tasklist_lock);
886 * this kills every thread in the thread group. Note that any externally
887 * wait4()-ing process will get the correct exit code - even if this
888 * thread is not the thread group leader.
890 asmlinkage void sys_exit_group(int error_code)
892 do_group_exit((error_code & 0xff) << 8);
895 static int eligible_child(pid_t pid, int options, task_t *p)
901 if (process_group(p) != process_group(current))
903 } else if (pid != -1) {
904 if (process_group(p) != -pid)
909 * Do not consider detached threads that are
912 if (p->exit_signal == -1 && !p->ptrace)
915 /* Wait for all children (clone and not) if __WALL is set;
916 * otherwise, wait for clone children *only* if __WCLONE is
917 * set; otherwise, wait for non-clone children *only*. (Note:
918 * A "clone" child here is one that reports to its parent
919 * using a signal other than SIGCHLD.) */
920 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
921 && !(options & __WALL))
924 * Do not consider thread group leaders that are
925 * in a non-empty thread group:
927 if (current->tgid != p->tgid && delay_group_leader(p))
930 if (security_task_wait(p))
937 * Handle sys_wait4 work for one task in state TASK_ZOMBIE. We hold
938 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
939 * the lock and this task is uninteresting. If we return nonzero, we have
940 * released the lock and the system call should return.
942 static int wait_task_zombie(task_t *p, unsigned int *stat_addr, struct rusage *ru)
948 * Try to move the task's state to DEAD
949 * only one thread is allowed to do this:
951 state = xchg(&p->state, TASK_DEAD);
952 if (state != TASK_ZOMBIE) {
953 BUG_ON(state != TASK_DEAD);
956 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
958 * This can only happen in a race with a ptraced thread
959 * dying on another processor.
964 * Now we are sure this task is interesting, and no other
965 * thread can reap it because we set its state to TASK_DEAD.
967 read_unlock(&tasklist_lock);
969 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
970 if (!retval && stat_addr) {
971 if (p->signal->group_exit)
972 retval = put_user(p->signal->group_exit_code, stat_addr);
974 retval = put_user(p->exit_code, stat_addr);
977 p->state = TASK_ZOMBIE;
981 if (p->real_parent != p->parent) {
982 write_lock_irq(&tasklist_lock);
983 /* Double-check with lock held. */
984 if (p->real_parent != p->parent) {
986 p->state = TASK_ZOMBIE;
987 /* If this is a detached thread, this is where it goes away. */
988 if (p->exit_signal == -1) {
989 /* release_task takes the lock itself. */
990 write_unlock_irq(&tasklist_lock);
994 do_notify_parent(p, p->exit_signal);
995 write_unlock_irq(&tasklist_lock);
1000 write_unlock_irq(&tasklist_lock);
1009 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1010 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1011 * the lock and this task is uninteresting. If we return nonzero, we have
1012 * released the lock and the system call should return.
1014 static int wait_task_stopped(task_t *p, int delayed_group_leader,
1015 unsigned int *stat_addr, struct rusage *ru)
1017 int retval, exit_code;
1021 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1022 p->signal && p->signal->group_stop_count > 0)
1024 * A group stop is in progress and this is the group leader.
1025 * We won't report until all threads have stopped.
1030 * Now we are pretty sure this task is interesting.
1031 * Make sure it doesn't get reaped out from under us while we
1032 * give up the lock and then examine it below. We don't want to
1033 * keep holding onto the tasklist_lock while we call getrusage and
1034 * possibly take page faults for user memory.
1037 read_unlock(&tasklist_lock);
1038 write_lock_irq(&tasklist_lock);
1041 * This uses xchg to be atomic with the thread resuming and setting
1042 * it. It must also be done with the write lock held to prevent a
1043 * race with the TASK_ZOMBIE case.
1045 exit_code = xchg(&p->exit_code, 0);
1046 if (unlikely(p->state > TASK_STOPPED)) {
1048 * The task resumed and then died. Let the next iteration
1049 * catch it in TASK_ZOMBIE. Note that exit_code might
1050 * already be zero here if it resumed and did _exit(0).
1051 * The task itself is dead and won't touch exit_code again;
1052 * other processors in this function are locked out.
1054 p->exit_code = exit_code;
1057 if (unlikely(exit_code == 0)) {
1059 * Another thread in this function got to it first, or it
1060 * resumed, or it resumed and then died.
1062 write_unlock_irq(&tasklist_lock);
1064 read_lock(&tasklist_lock);
1068 /* move to end of parent's list to avoid starvation */
1070 add_parent(p, p->parent);
1072 write_unlock_irq(&tasklist_lock);
1074 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1075 if (!retval && stat_addr)
1076 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1085 asmlinkage long sys_wait4(pid_t pid,unsigned int * stat_addr, int options, struct rusage * ru)
1087 DECLARE_WAITQUEUE(wait, current);
1088 struct task_struct *tsk;
1091 if (options & ~(WNOHANG|WUNTRACED|__WNOTHREAD|__WCLONE|__WALL))
1094 add_wait_queue(¤t->wait_chldexit,&wait);
1097 current->state = TASK_INTERRUPTIBLE;
1098 read_lock(&tasklist_lock);
1101 struct task_struct *p;
1102 struct list_head *_p;
1105 list_for_each(_p,&tsk->children) {
1106 p = list_entry(_p,struct task_struct,sibling);
1108 ret = eligible_child(pid, options, p);
1115 if (!(options & WUNTRACED) &&
1116 !(p->ptrace & PT_PTRACED))
1118 retval = wait_task_stopped(p, ret == 2,
1120 if (retval != 0) /* He released the lock. */
1125 * Eligible but we cannot release it yet:
1129 retval = wait_task_zombie(p, stat_addr, ru);
1130 if (retval != 0) /* He released the lock. */
1136 list_for_each (_p,&tsk->ptrace_children) {
1137 p = list_entry(_p,struct task_struct,ptrace_list);
1138 if (!eligible_child(pid, options, p))
1144 if (options & __WNOTHREAD)
1146 tsk = next_thread(tsk);
1147 if (tsk->signal != current->signal)
1149 } while (tsk != current);
1150 read_unlock(&tasklist_lock);
1153 if (options & WNOHANG)
1155 retval = -ERESTARTSYS;
1156 if (signal_pending(current))
1163 current->state = TASK_RUNNING;
1164 remove_wait_queue(¤t->wait_chldexit,&wait);
1168 #if !defined(__alpha__) && !defined(__ia64__) && \
1169 !defined(__arm__) && !defined(__s390__)
1172 * sys_waitpid() remains for compatibility. waitpid() should be
1173 * implemented by calling sys_wait4() from libc.a.
1175 asmlinkage long sys_waitpid(pid_t pid,unsigned int * stat_addr, int options)
1177 return sys_wait4(pid, stat_addr, options, NULL);