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/cpuset.h>
29 #include <linux/syscalls.h>
30 #include <linux/signal.h>
32 #include <asm/uaccess.h>
33 #include <asm/unistd.h>
34 #include <asm/pgtable.h>
35 #include <asm/mmu_context.h>
37 extern void sem_exit (void);
38 extern struct task_struct *child_reaper;
40 int getrusage(struct task_struct *, int, struct rusage __user *);
42 static void __unhash_process(struct task_struct *p)
45 detach_pid(p, PIDTYPE_PID);
46 detach_pid(p, PIDTYPE_TGID);
47 if (thread_group_leader(p)) {
48 detach_pid(p, PIDTYPE_PGID);
49 detach_pid(p, PIDTYPE_SID);
51 __get_cpu_var(process_counts)--;
57 void release_task(struct task_struct * p)
61 struct dentry *proc_dentry;
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->exit_state == EXIT_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);
97 write_unlock_irq(&tasklist_lock);
98 spin_unlock(&p->proc_lock);
99 proc_pid_flush(proc_dentry);
104 if (unlikely(zap_leader))
108 /* we are using it only for SMP init */
110 void unhash_process(struct task_struct *p)
112 struct dentry *proc_dentry;
114 spin_lock(&p->proc_lock);
115 proc_dentry = proc_pid_unhash(p);
116 write_lock_irq(&tasklist_lock);
118 write_unlock_irq(&tasklist_lock);
119 spin_unlock(&p->proc_lock);
120 proc_pid_flush(proc_dentry);
124 * This checks not only the pgrp, but falls back on the pid if no
125 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
128 int session_of_pgrp(int pgrp)
130 struct task_struct *p;
133 read_lock(&tasklist_lock);
134 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
135 if (p->signal->session > 0) {
136 sid = p->signal->session;
139 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
140 p = find_task_by_pid(pgrp);
142 sid = p->signal->session;
144 read_unlock(&tasklist_lock);
150 * Determine if a process group is "orphaned", according to the POSIX
151 * definition in 2.2.2.52. Orphaned process groups are not to be affected
152 * by terminal-generated stop signals. Newly orphaned process groups are
153 * to receive a SIGHUP and a SIGCONT.
155 * "I ask you, have you ever known what it is to be an orphan?"
157 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
159 struct task_struct *p;
162 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
163 if (p == ignored_task
165 || p->real_parent->pid == 1)
167 if (process_group(p->real_parent) != pgrp
168 && p->real_parent->signal->session == p->signal->session) {
172 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
173 return ret; /* (sighing) "Often!" */
176 int is_orphaned_pgrp(int pgrp)
180 read_lock(&tasklist_lock);
181 retval = will_become_orphaned_pgrp(pgrp, NULL);
182 read_unlock(&tasklist_lock);
187 static inline int has_stopped_jobs(int pgrp)
190 struct task_struct *p;
192 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
193 if (p->state != TASK_STOPPED)
196 /* If p is stopped by a debugger on a signal that won't
197 stop it, then don't count p as stopped. This isn't
198 perfect but it's a good approximation. */
199 if (unlikely (p->ptrace)
200 && p->exit_code != SIGSTOP
201 && p->exit_code != SIGTSTP
202 && p->exit_code != SIGTTOU
203 && p->exit_code != SIGTTIN)
208 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
213 * reparent_to_init - Reparent the calling kernel thread to the init task.
215 * If a kernel thread is launched as a result of a system call, or if
216 * it ever exits, it should generally reparent itself to init so that
217 * it is correctly cleaned up on exit.
219 * The various task state such as scheduling policy and priority may have
220 * been inherited from a user process, so we reset them to sane values here.
222 * NOTE that reparent_to_init() gives the caller full capabilities.
224 static inline void reparent_to_init(void)
226 write_lock_irq(&tasklist_lock);
228 ptrace_unlink(current);
229 /* Reparent to init */
230 REMOVE_LINKS(current);
231 current->parent = child_reaper;
232 current->real_parent = child_reaper;
235 /* Set the exit signal to SIGCHLD so we signal init on exit */
236 current->exit_signal = SIGCHLD;
238 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
239 set_user_nice(current, 0);
243 security_task_reparent_to_init(current);
244 memcpy(current->signal->rlim, init_task.signal->rlim,
245 sizeof(current->signal->rlim));
246 atomic_inc(&(INIT_USER->__count));
247 write_unlock_irq(&tasklist_lock);
248 switch_uid(INIT_USER);
251 void __set_special_pids(pid_t session, pid_t pgrp)
253 struct task_struct *curr = current;
255 if (curr->signal->session != session) {
256 detach_pid(curr, PIDTYPE_SID);
257 curr->signal->session = session;
258 attach_pid(curr, PIDTYPE_SID, session);
260 if (process_group(curr) != pgrp) {
261 detach_pid(curr, PIDTYPE_PGID);
262 curr->signal->pgrp = pgrp;
263 attach_pid(curr, PIDTYPE_PGID, pgrp);
267 void set_special_pids(pid_t session, pid_t pgrp)
269 write_lock_irq(&tasklist_lock);
270 __set_special_pids(session, pgrp);
271 write_unlock_irq(&tasklist_lock);
275 * Let kernel threads use this to say that they
276 * allow a certain signal (since daemonize() will
277 * have disabled all of them by default).
279 int allow_signal(int sig)
281 if (!valid_signal(sig) || sig < 1)
284 spin_lock_irq(¤t->sighand->siglock);
285 sigdelset(¤t->blocked, sig);
287 /* Kernel threads handle their own signals.
288 Let the signal code know it'll be handled, so
289 that they don't get converted to SIGKILL or
290 just silently dropped */
291 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
294 spin_unlock_irq(¤t->sighand->siglock);
298 EXPORT_SYMBOL(allow_signal);
300 int disallow_signal(int sig)
302 if (!valid_signal(sig) || sig < 1)
305 spin_lock_irq(¤t->sighand->siglock);
306 sigaddset(¤t->blocked, sig);
308 spin_unlock_irq(¤t->sighand->siglock);
312 EXPORT_SYMBOL(disallow_signal);
315 * Put all the gunge required to become a kernel thread without
316 * attached user resources in one place where it belongs.
319 void daemonize(const char *name, ...)
322 struct fs_struct *fs;
325 va_start(args, name);
326 vsnprintf(current->comm, sizeof(current->comm), name, args);
330 * If we were started as result of loading a module, close all of the
331 * user space pages. We don't need them, and if we didn't close them
332 * they would be locked into memory.
336 set_special_pids(1, 1);
338 current->signal->tty = NULL;
341 /* Block and flush all signals */
342 sigfillset(&blocked);
343 sigprocmask(SIG_BLOCK, &blocked, NULL);
344 flush_signals(current);
346 /* Become as one with the init task */
348 exit_fs(current); /* current->fs->count--; */
351 atomic_inc(&fs->count);
353 current->files = init_task.files;
354 atomic_inc(¤t->files->count);
359 EXPORT_SYMBOL(daemonize);
361 static inline void close_files(struct files_struct * files)
369 if (i >= files->max_fdset || i >= files->max_fds)
371 set = files->open_fds->fds_bits[j++];
374 struct file * file = xchg(&files->fd[i], NULL);
376 filp_close(file, files);
384 struct files_struct *get_files_struct(struct task_struct *task)
386 struct files_struct *files;
391 atomic_inc(&files->count);
397 void fastcall put_files_struct(struct files_struct *files)
399 if (atomic_dec_and_test(&files->count)) {
402 * Free the fd and fdset arrays if we expanded them.
404 if (files->fd != &files->fd_array[0])
405 free_fd_array(files->fd, files->max_fds);
406 if (files->max_fdset > __FD_SETSIZE) {
407 free_fdset(files->open_fds, files->max_fdset);
408 free_fdset(files->close_on_exec, files->max_fdset);
410 kmem_cache_free(files_cachep, files);
414 EXPORT_SYMBOL(put_files_struct);
416 static inline void __exit_files(struct task_struct *tsk)
418 struct files_struct * files = tsk->files;
424 put_files_struct(files);
428 void exit_files(struct task_struct *tsk)
433 static inline void __put_fs_struct(struct fs_struct *fs)
435 /* No need to hold fs->lock if we are killing it */
436 if (atomic_dec_and_test(&fs->count)) {
443 mntput(fs->altrootmnt);
445 kmem_cache_free(fs_cachep, fs);
449 void put_fs_struct(struct fs_struct *fs)
454 static inline void __exit_fs(struct task_struct *tsk)
456 struct fs_struct * fs = tsk->fs;
466 void exit_fs(struct task_struct *tsk)
471 EXPORT_SYMBOL_GPL(exit_fs);
474 * Turn us into a lazy TLB process if we
477 void exit_mm(struct task_struct * tsk)
479 struct mm_struct *mm = tsk->mm;
485 * Serialize with any possible pending coredump.
486 * We must hold mmap_sem around checking core_waiters
487 * and clearing tsk->mm. The core-inducing thread
488 * will increment core_waiters for each thread in the
489 * group with ->mm != NULL.
491 down_read(&mm->mmap_sem);
492 if (mm->core_waiters) {
493 up_read(&mm->mmap_sem);
494 down_write(&mm->mmap_sem);
495 if (!--mm->core_waiters)
496 complete(mm->core_startup_done);
497 up_write(&mm->mmap_sem);
499 wait_for_completion(&mm->core_done);
500 down_read(&mm->mmap_sem);
502 atomic_inc(&mm->mm_count);
503 if (mm != tsk->active_mm) BUG();
504 /* more a memory barrier than a real lock */
507 up_read(&mm->mmap_sem);
508 enter_lazy_tlb(mm, current);
513 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
516 * Make sure we're not reparenting to ourselves and that
517 * the parent is not a zombie.
519 BUG_ON(p == reaper || reaper->exit_state >= EXIT_ZOMBIE);
520 p->real_parent = reaper;
523 static inline void reparent_thread(task_t *p, task_t *father, int traced)
525 /* We don't want people slaying init. */
526 if (p->exit_signal != -1)
527 p->exit_signal = SIGCHLD;
529 if (p->pdeath_signal)
530 /* We already hold the tasklist_lock here. */
531 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
533 /* Move the child from its dying parent to the new one. */
534 if (unlikely(traced)) {
535 /* Preserve ptrace links if someone else is tracing this child. */
536 list_del_init(&p->ptrace_list);
537 if (p->parent != p->real_parent)
538 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
540 /* If this child is being traced, then we're the one tracing it
541 * anyway, so let go of it.
544 list_del_init(&p->sibling);
545 p->parent = p->real_parent;
546 list_add_tail(&p->sibling, &p->parent->children);
548 /* If we'd notified the old parent about this child's death,
549 * also notify the new parent.
551 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
552 thread_group_empty(p))
553 do_notify_parent(p, p->exit_signal);
554 else if (p->state == TASK_TRACED) {
556 * If it was at a trace stop, turn it into
557 * a normal stop since it's no longer being
565 * process group orphan check
566 * Case ii: Our child is in a different pgrp
567 * than we are, and it was the only connection
568 * outside, so the child pgrp is now orphaned.
570 if ((process_group(p) != process_group(father)) &&
571 (p->signal->session == father->signal->session)) {
572 int pgrp = process_group(p);
574 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
575 __kill_pg_info(SIGHUP, (void *)1, pgrp);
576 __kill_pg_info(SIGCONT, (void *)1, pgrp);
582 * When we die, we re-parent all our children.
583 * Try to give them to another thread in our thread
584 * group, and if no such member exists, give it to
585 * the global child reaper process (ie "init")
587 static inline void forget_original_parent(struct task_struct * father,
588 struct list_head *to_release)
590 struct task_struct *p, *reaper = father;
591 struct list_head *_p, *_n;
594 reaper = next_thread(reaper);
595 if (reaper == father) {
596 reaper = child_reaper;
599 } while (reaper->exit_state);
602 * There are only two places where our children can be:
604 * - in our child list
605 * - in our ptraced child list
607 * Search them and reparent children.
609 list_for_each_safe(_p, _n, &father->children) {
611 p = list_entry(_p,struct task_struct,sibling);
615 /* if father isn't the real parent, then ptrace must be enabled */
616 BUG_ON(father != p->real_parent && !ptrace);
618 if (father == p->real_parent) {
619 /* reparent with a reaper, real father it's us */
620 choose_new_parent(p, reaper, child_reaper);
621 reparent_thread(p, father, 0);
623 /* reparent ptraced task to its real parent */
625 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
626 thread_group_empty(p))
627 do_notify_parent(p, p->exit_signal);
631 * if the ptraced child is a zombie with exit_signal == -1
632 * we must collect it before we exit, or it will remain
633 * zombie forever since we prevented it from self-reap itself
634 * while it was being traced by us, to be able to see it in wait4.
636 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
637 list_add(&p->ptrace_list, to_release);
639 list_for_each_safe(_p, _n, &father->ptrace_children) {
640 p = list_entry(_p,struct task_struct,ptrace_list);
641 choose_new_parent(p, reaper, child_reaper);
642 reparent_thread(p, father, 1);
647 * Send signals to all our closest relatives so that they know
648 * to properly mourn us..
650 static void exit_notify(struct task_struct *tsk)
653 struct task_struct *t;
654 struct list_head ptrace_dead, *_p, *_n;
656 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
657 && !thread_group_empty(tsk)) {
659 * This occurs when there was a race between our exit
660 * syscall and a group signal choosing us as the one to
661 * wake up. It could be that we are the only thread
662 * alerted to check for pending signals, but another thread
663 * should be woken now to take the signal since we will not.
664 * Now we'll wake all the threads in the group just to make
665 * sure someone gets all the pending signals.
667 read_lock(&tasklist_lock);
668 spin_lock_irq(&tsk->sighand->siglock);
669 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
670 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
671 recalc_sigpending_tsk(t);
672 if (signal_pending(t))
673 signal_wake_up(t, 0);
675 spin_unlock_irq(&tsk->sighand->siglock);
676 read_unlock(&tasklist_lock);
679 write_lock_irq(&tasklist_lock);
682 * This does two things:
684 * A. Make init inherit all the child processes
685 * B. Check to see if any process groups have become orphaned
686 * as a result of our exiting, and if they have any stopped
687 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
690 INIT_LIST_HEAD(&ptrace_dead);
691 forget_original_parent(tsk, &ptrace_dead);
692 BUG_ON(!list_empty(&tsk->children));
693 BUG_ON(!list_empty(&tsk->ptrace_children));
696 * 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)
700 * Case i: Our father is in a different pgrp than we are
701 * and we were the only connection outside, so our pgrp
702 * is about to become orphaned.
705 t = tsk->real_parent;
707 if ((process_group(t) != process_group(tsk)) &&
708 (t->signal->session == tsk->signal->session) &&
709 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
710 has_stopped_jobs(process_group(tsk))) {
711 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
712 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
715 /* Let father know we died
717 * Thread signals are configurable, but you aren't going to use
718 * that to send signals to arbitary processes.
719 * That stops right now.
721 * If the parent exec id doesn't match the exec id we saved
722 * when we started then we know the parent has changed security
725 * If our self_exec id doesn't match our parent_exec_id then
726 * we have changed execution domain as these two values started
727 * the same after a fork.
731 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
732 ( tsk->parent_exec_id != t->self_exec_id ||
733 tsk->self_exec_id != tsk->parent_exec_id)
734 && !capable(CAP_KILL))
735 tsk->exit_signal = SIGCHLD;
738 /* If something other than our normal parent is ptracing us, then
739 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
740 * only has special meaning to our real parent.
742 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
743 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
744 do_notify_parent(tsk, signal);
745 } else if (tsk->ptrace) {
746 do_notify_parent(tsk, SIGCHLD);
750 if (tsk->exit_signal == -1 &&
751 (likely(tsk->ptrace == 0) ||
752 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
754 tsk->exit_state = state;
756 write_unlock_irq(&tasklist_lock);
758 list_for_each_safe(_p, _n, &ptrace_dead) {
760 t = list_entry(_p,struct task_struct,ptrace_list);
764 /* If the process is dead, release it - nobody will wait for it */
765 if (state == EXIT_DEAD)
768 /* PF_DEAD causes final put_task_struct after we schedule. */
770 tsk->flags |= PF_DEAD;
773 fastcall NORET_TYPE void do_exit(long code)
775 struct task_struct *tsk = current;
778 profile_task_exit(tsk);
780 if (unlikely(in_interrupt()))
781 panic("Aiee, killing interrupt handler!");
782 if (unlikely(!tsk->pid))
783 panic("Attempted to kill the idle task!");
784 if (unlikely(tsk->pid == 1))
785 panic("Attempted to kill init!");
789 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
790 current->ptrace_message = code;
791 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
794 tsk->flags |= PF_EXITING;
797 * Make sure we don't try to process any timer firings
798 * while we are already exiting.
800 tsk->it_virt_expires = cputime_zero;
801 tsk->it_prof_expires = cputime_zero;
802 tsk->it_sched_expires = 0;
804 if (unlikely(in_atomic()))
805 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
806 current->comm, current->pid,
809 acct_update_integrals(tsk);
810 update_mem_hiwater(tsk);
811 group_dead = atomic_dec_and_test(&tsk->signal->live);
813 del_timer_sync(&tsk->signal->real_timer);
826 if (group_dead && tsk->signal->leader)
827 disassociate_ctty(1);
829 module_put(tsk->thread_info->exec_domain->module);
831 module_put(tsk->binfmt->module);
833 tsk->exit_code = code;
836 mpol_free(tsk->mempolicy);
837 tsk->mempolicy = NULL;
840 BUG_ON(!(current->flags & PF_DEAD));
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)
864 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
867 EXPORT_SYMBOL(next_thread);
870 * Take down every thread in the group. This is called by fatal signals
871 * as well as by sys_exit_group (below).
874 do_group_exit(int exit_code)
876 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
878 if (current->signal->flags & SIGNAL_GROUP_EXIT)
879 exit_code = current->signal->group_exit_code;
880 else if (!thread_group_empty(current)) {
881 struct signal_struct *const sig = current->signal;
882 struct sighand_struct *const sighand = current->sighand;
883 read_lock(&tasklist_lock);
884 spin_lock_irq(&sighand->siglock);
885 if (sig->flags & SIGNAL_GROUP_EXIT)
886 /* Another thread got here before we took the lock. */
887 exit_code = sig->group_exit_code;
889 sig->flags = SIGNAL_GROUP_EXIT;
890 sig->group_exit_code = exit_code;
891 zap_other_threads(current);
893 spin_unlock_irq(&sighand->siglock);
894 read_unlock(&tasklist_lock);
902 * this kills every thread in the thread group. Note that any externally
903 * wait4()-ing process will get the correct exit code - even if this
904 * thread is not the thread group leader.
906 asmlinkage void sys_exit_group(int error_code)
908 do_group_exit((error_code & 0xff) << 8);
911 static int eligible_child(pid_t pid, int options, task_t *p)
917 if (process_group(p) != process_group(current))
919 } else if (pid != -1) {
920 if (process_group(p) != -pid)
925 * Do not consider detached threads that are
928 if (p->exit_signal == -1 && !p->ptrace)
931 /* Wait for all children (clone and not) if __WALL is set;
932 * otherwise, wait for clone children *only* if __WCLONE is
933 * set; otherwise, wait for non-clone children *only*. (Note:
934 * A "clone" child here is one that reports to its parent
935 * using a signal other than SIGCHLD.) */
936 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
937 && !(options & __WALL))
940 * Do not consider thread group leaders that are
941 * in a non-empty thread group:
943 if (current->tgid != p->tgid && delay_group_leader(p))
946 if (security_task_wait(p))
952 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
954 struct siginfo __user *infop,
955 struct rusage __user *rusagep)
957 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
960 retval = put_user(SIGCHLD, &infop->si_signo);
962 retval = put_user(0, &infop->si_errno);
964 retval = put_user((short)why, &infop->si_code);
966 retval = put_user(pid, &infop->si_pid);
968 retval = put_user(uid, &infop->si_uid);
970 retval = put_user(status, &infop->si_status);
977 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
978 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
979 * the lock and this task is uninteresting. If we return nonzero, we have
980 * released the lock and the system call should return.
982 static int wait_task_zombie(task_t *p, int noreap,
983 struct siginfo __user *infop,
984 int __user *stat_addr, struct rusage __user *ru)
990 if (unlikely(noreap)) {
993 int exit_code = p->exit_code;
996 if (unlikely(p->exit_state != EXIT_ZOMBIE))
998 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1001 read_unlock(&tasklist_lock);
1002 if ((exit_code & 0x7f) == 0) {
1004 status = exit_code >> 8;
1006 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1007 status = exit_code & 0x7f;
1009 return wait_noreap_copyout(p, pid, uid, why,
1014 * Try to move the task's state to DEAD
1015 * only one thread is allowed to do this:
1017 state = xchg(&p->exit_state, EXIT_DEAD);
1018 if (state != EXIT_ZOMBIE) {
1019 BUG_ON(state != EXIT_DEAD);
1022 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1024 * This can only happen in a race with a ptraced thread
1025 * dying on another processor.
1030 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1032 * The resource counters for the group leader are in its
1033 * own task_struct. Those for dead threads in the group
1034 * are in its signal_struct, as are those for the child
1035 * processes it has previously reaped. All these
1036 * accumulate in the parent's signal_struct c* fields.
1038 * We don't bother to take a lock here to protect these
1039 * p->signal fields, because they are only touched by
1040 * __exit_signal, which runs with tasklist_lock
1041 * write-locked anyway, and so is excluded here. We do
1042 * need to protect the access to p->parent->signal fields,
1043 * as other threads in the parent group can be right
1044 * here reaping other children at the same time.
1046 spin_lock_irq(&p->parent->sighand->siglock);
1047 p->parent->signal->cutime =
1048 cputime_add(p->parent->signal->cutime,
1049 cputime_add(p->utime,
1050 cputime_add(p->signal->utime,
1051 p->signal->cutime)));
1052 p->parent->signal->cstime =
1053 cputime_add(p->parent->signal->cstime,
1054 cputime_add(p->stime,
1055 cputime_add(p->signal->stime,
1056 p->signal->cstime)));
1057 p->parent->signal->cmin_flt +=
1058 p->min_flt + p->signal->min_flt + p->signal->cmin_flt;
1059 p->parent->signal->cmaj_flt +=
1060 p->maj_flt + p->signal->maj_flt + p->signal->cmaj_flt;
1061 p->parent->signal->cnvcsw +=
1062 p->nvcsw + p->signal->nvcsw + p->signal->cnvcsw;
1063 p->parent->signal->cnivcsw +=
1064 p->nivcsw + p->signal->nivcsw + p->signal->cnivcsw;
1065 spin_unlock_irq(&p->parent->sighand->siglock);
1069 * Now we are sure this task is interesting, and no other
1070 * thread can reap it because we set its state to EXIT_DEAD.
1072 read_unlock(&tasklist_lock);
1074 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1075 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1076 ? p->signal->group_exit_code : p->exit_code;
1077 if (!retval && stat_addr)
1078 retval = put_user(status, stat_addr);
1079 if (!retval && infop)
1080 retval = put_user(SIGCHLD, &infop->si_signo);
1081 if (!retval && infop)
1082 retval = put_user(0, &infop->si_errno);
1083 if (!retval && infop) {
1086 if ((status & 0x7f) == 0) {
1090 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1093 retval = put_user((short)why, &infop->si_code);
1095 retval = put_user(status, &infop->si_status);
1097 if (!retval && infop)
1098 retval = put_user(p->pid, &infop->si_pid);
1099 if (!retval && infop)
1100 retval = put_user(p->uid, &infop->si_uid);
1102 // TODO: is this safe?
1103 p->exit_state = EXIT_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 // TODO: is this safe?
1113 p->exit_state = EXIT_ZOMBIE;
1115 * If this is not a detached task, notify the parent.
1116 * If it's still not detached after that, don't release
1119 if (p->exit_signal != -1) {
1120 do_notify_parent(p, p->exit_signal);
1121 if (p->exit_signal != -1)
1125 write_unlock_irq(&tasklist_lock);
1134 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1135 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1136 * the lock and this task is uninteresting. If we return nonzero, we have
1137 * released the lock and the system call should return.
1139 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1140 struct siginfo __user *infop,
1141 int __user *stat_addr, struct rusage __user *ru)
1143 int retval, exit_code;
1147 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1148 p->signal && p->signal->group_stop_count > 0)
1150 * A group stop is in progress and this is the group leader.
1151 * We won't report until all threads have stopped.
1156 * Now we are pretty sure this task is interesting.
1157 * Make sure it doesn't get reaped out from under us while we
1158 * give up the lock and then examine it below. We don't want to
1159 * keep holding onto the tasklist_lock while we call getrusage and
1160 * possibly take page faults for user memory.
1163 read_unlock(&tasklist_lock);
1165 if (unlikely(noreap)) {
1168 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1170 exit_code = p->exit_code;
1171 if (unlikely(!exit_code) ||
1172 unlikely(p->state > TASK_STOPPED))
1174 return wait_noreap_copyout(p, pid, uid,
1175 why, (exit_code << 8) | 0x7f,
1179 write_lock_irq(&tasklist_lock);
1182 * This uses xchg to be atomic with the thread resuming and setting
1183 * it. It must also be done with the write lock held to prevent a
1184 * race with the EXIT_ZOMBIE case.
1186 exit_code = xchg(&p->exit_code, 0);
1187 if (unlikely(p->exit_state)) {
1189 * The task resumed and then died. Let the next iteration
1190 * catch it in EXIT_ZOMBIE. Note that exit_code might
1191 * already be zero here if it resumed and did _exit(0).
1192 * The task itself is dead and won't touch exit_code again;
1193 * other processors in this function are locked out.
1195 p->exit_code = exit_code;
1198 if (unlikely(exit_code == 0)) {
1200 * Another thread in this function got to it first, or it
1201 * resumed, or it resumed and then died.
1203 write_unlock_irq(&tasklist_lock);
1207 * We are returning to the wait loop without having successfully
1208 * removed the process and having released the lock. We cannot
1209 * continue, since the "p" task pointer is potentially stale.
1211 * Return -EAGAIN, and do_wait() will restart the loop from the
1212 * beginning. Do _not_ re-acquire the lock.
1217 /* move to end of parent's list to avoid starvation */
1219 add_parent(p, p->parent);
1221 write_unlock_irq(&tasklist_lock);
1223 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1224 if (!retval && stat_addr)
1225 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1226 if (!retval && infop)
1227 retval = put_user(SIGCHLD, &infop->si_signo);
1228 if (!retval && infop)
1229 retval = put_user(0, &infop->si_errno);
1230 if (!retval && infop)
1231 retval = put_user((short)((p->ptrace & PT_PTRACED)
1232 ? CLD_TRAPPED : CLD_STOPPED),
1234 if (!retval && infop)
1235 retval = put_user(exit_code, &infop->si_status);
1236 if (!retval && infop)
1237 retval = put_user(p->pid, &infop->si_pid);
1238 if (!retval && infop)
1239 retval = put_user(p->uid, &infop->si_uid);
1249 * Handle do_wait work for one task in a live, non-stopped state.
1250 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1251 * the lock and this task is uninteresting. If we return nonzero, we have
1252 * released the lock and the system call should return.
1254 static int wait_task_continued(task_t *p, int noreap,
1255 struct siginfo __user *infop,
1256 int __user *stat_addr, struct rusage __user *ru)
1262 if (unlikely(!p->signal))
1265 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1268 spin_lock_irq(&p->sighand->siglock);
1269 /* Re-check with the lock held. */
1270 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1271 spin_unlock_irq(&p->sighand->siglock);
1275 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1276 spin_unlock_irq(&p->sighand->siglock);
1281 read_unlock(&tasklist_lock);
1284 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1286 if (!retval && stat_addr)
1287 retval = put_user(0xffff, stat_addr);
1291 retval = wait_noreap_copyout(p, pid, uid,
1292 CLD_CONTINUED, SIGCONT,
1294 BUG_ON(retval == 0);
1301 static inline int my_ptrace_child(struct task_struct *p)
1303 if (!(p->ptrace & PT_PTRACED))
1305 if (!(p->ptrace & PT_ATTACHED))
1308 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1309 * we are the attacher. If we are the real parent, this is a race
1310 * inside ptrace_attach. It is waiting for the tasklist_lock,
1311 * which we have to switch the parent links, but has already set
1312 * the flags in p->ptrace.
1314 return (p->parent != p->real_parent);
1317 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1318 int __user *stat_addr, struct rusage __user *ru)
1320 DECLARE_WAITQUEUE(wait, current);
1321 struct task_struct *tsk;
1324 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1327 * We will set this flag if we see any child that might later
1328 * match our criteria, even if we are not able to reap it yet.
1331 current->state = TASK_INTERRUPTIBLE;
1332 read_lock(&tasklist_lock);
1335 struct task_struct *p;
1336 struct list_head *_p;
1339 list_for_each(_p,&tsk->children) {
1340 p = list_entry(_p,struct task_struct,sibling);
1342 ret = eligible_child(pid, options, p);
1348 if (!my_ptrace_child(p))
1353 * It's stopped now, so it might later
1354 * continue, exit, or stop again.
1357 if (!(options & WUNTRACED) &&
1358 !my_ptrace_child(p))
1360 retval = wait_task_stopped(p, ret == 2,
1361 (options & WNOWAIT),
1364 if (retval == -EAGAIN)
1366 if (retval != 0) /* He released the lock. */
1371 if (p->exit_state == EXIT_DEAD)
1373 // case EXIT_ZOMBIE:
1374 if (p->exit_state == EXIT_ZOMBIE) {
1376 * Eligible but we cannot release
1380 goto check_continued;
1381 if (!likely(options & WEXITED))
1383 retval = wait_task_zombie(
1384 p, (options & WNOWAIT),
1385 infop, stat_addr, ru);
1386 /* He released the lock. */
1393 * It's running now, so it might later
1394 * exit, stop, or stop and then continue.
1397 if (!unlikely(options & WCONTINUED))
1399 retval = wait_task_continued(
1400 p, (options & WNOWAIT),
1401 infop, stat_addr, ru);
1402 if (retval != 0) /* He released the lock. */
1408 list_for_each(_p, &tsk->ptrace_children) {
1409 p = list_entry(_p, struct task_struct,
1411 if (!eligible_child(pid, options, p))
1417 if (options & __WNOTHREAD)
1419 tsk = next_thread(tsk);
1420 if (tsk->signal != current->signal)
1422 } while (tsk != current);
1424 read_unlock(&tasklist_lock);
1427 if (options & WNOHANG)
1429 retval = -ERESTARTSYS;
1430 if (signal_pending(current))
1437 current->state = TASK_RUNNING;
1438 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1444 * For a WNOHANG return, clear out all the fields
1445 * we would set so the user can easily tell the
1449 retval = put_user(0, &infop->si_signo);
1451 retval = put_user(0, &infop->si_errno);
1453 retval = put_user(0, &infop->si_code);
1455 retval = put_user(0, &infop->si_pid);
1457 retval = put_user(0, &infop->si_uid);
1459 retval = put_user(0, &infop->si_status);
1465 asmlinkage long sys_waitid(int which, pid_t pid,
1466 struct siginfo __user *infop, int options,
1467 struct rusage __user *ru)
1471 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1473 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1493 ret = do_wait(pid, options, infop, NULL, ru);
1495 /* avoid REGPARM breakage on x86: */
1496 prevent_tail_call(ret);
1500 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1501 int options, struct rusage __user *ru)
1505 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1506 __WNOTHREAD|__WCLONE|__WALL))
1508 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1510 /* avoid REGPARM breakage on x86: */
1511 prevent_tail_call(ret);
1515 #ifdef __ARCH_WANT_SYS_WAITPID
1518 * sys_waitpid() remains for compatibility. waitpid() should be
1519 * implemented by calling sys_wait4() from libc.a.
1521 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1523 return sys_wait4(pid, stat_addr, options, NULL);