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>
31 #include <linux/cn_proc.h>
33 #include <asm/uaccess.h>
34 #include <asm/unistd.h>
35 #include <asm/pgtable.h>
36 #include <asm/mmu_context.h>
38 extern void sem_exit (void);
39 extern struct task_struct *child_reaper;
41 int getrusage(struct task_struct *, int, struct rusage __user *);
43 static void exit_mm(struct task_struct * tsk);
45 static void __unhash_process(struct task_struct *p)
48 detach_pid(p, PIDTYPE_PID);
49 detach_pid(p, PIDTYPE_TGID);
50 if (thread_group_leader(p)) {
51 detach_pid(p, PIDTYPE_PGID);
52 detach_pid(p, PIDTYPE_SID);
54 __get_cpu_var(process_counts)--;
60 void release_task(struct task_struct * p)
64 struct dentry *proc_dentry;
67 atomic_dec(&p->user->processes);
68 spin_lock(&p->proc_lock);
69 proc_dentry = proc_pid_unhash(p);
70 write_lock_irq(&tasklist_lock);
71 if (unlikely(p->ptrace))
73 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
76 * Note that the fastpath in sys_times depends on __exit_signal having
77 * updated the counters before a task is removed from the tasklist of
78 * the process by __unhash_process.
83 * If we are the last non-leader member of the thread
84 * group, and the leader is zombie, then notify the
85 * group leader's parent process. (if it wants notification.)
88 leader = p->group_leader;
89 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
90 BUG_ON(leader->exit_signal == -1);
91 do_notify_parent(leader, leader->exit_signal);
93 * If we were the last child thread and the leader has
94 * exited already, and the leader's parent ignores SIGCHLD,
95 * then we are the one who should release the leader.
97 * do_notify_parent() will have marked it self-reaping in
100 zap_leader = (leader->exit_signal == -1);
104 write_unlock_irq(&tasklist_lock);
105 spin_unlock(&p->proc_lock);
106 proc_pid_flush(proc_dentry);
111 if (unlikely(zap_leader))
115 /* we are using it only for SMP init */
117 void unhash_process(struct task_struct *p)
119 struct dentry *proc_dentry;
121 spin_lock(&p->proc_lock);
122 proc_dentry = proc_pid_unhash(p);
123 write_lock_irq(&tasklist_lock);
125 write_unlock_irq(&tasklist_lock);
126 spin_unlock(&p->proc_lock);
127 proc_pid_flush(proc_dentry);
131 * This checks not only the pgrp, but falls back on the pid if no
132 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
135 int session_of_pgrp(int pgrp)
137 struct task_struct *p;
140 read_lock(&tasklist_lock);
141 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
142 if (p->signal->session > 0) {
143 sid = p->signal->session;
146 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
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;
169 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
170 if (p == ignored_task
172 || p->real_parent->pid == 1)
174 if (process_group(p->real_parent) != pgrp
175 && p->real_parent->signal->session == p->signal->session) {
179 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
180 return ret; /* (sighing) "Often!" */
183 int is_orphaned_pgrp(int pgrp)
187 read_lock(&tasklist_lock);
188 retval = will_become_orphaned_pgrp(pgrp, NULL);
189 read_unlock(&tasklist_lock);
194 static inline int has_stopped_jobs(int pgrp)
197 struct task_struct *p;
199 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
200 if (p->state != TASK_STOPPED)
203 /* If p is stopped by a debugger on a signal that won't
204 stop it, then don't count p as stopped. This isn't
205 perfect but it's a good approximation. */
206 if (unlikely (p->ptrace)
207 && p->exit_code != SIGSTOP
208 && p->exit_code != SIGTSTP
209 && p->exit_code != SIGTTOU
210 && p->exit_code != SIGTTIN)
215 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
220 * reparent_to_init - Reparent the calling kernel thread to the init task.
222 * If a kernel thread is launched as a result of a system call, or if
223 * it ever exits, it should generally reparent itself to init so that
224 * it is correctly cleaned up on exit.
226 * The various task state such as scheduling policy and priority may have
227 * been inherited from a user process, so we reset them to sane values here.
229 * NOTE that reparent_to_init() gives the caller full capabilities.
231 static inline void reparent_to_init(void)
233 write_lock_irq(&tasklist_lock);
235 ptrace_unlink(current);
236 /* Reparent to init */
237 REMOVE_LINKS(current);
238 current->parent = child_reaper;
239 current->real_parent = child_reaper;
242 /* Set the exit signal to SIGCHLD so we signal init on exit */
243 current->exit_signal = SIGCHLD;
245 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
246 set_user_nice(current, 0);
250 security_task_reparent_to_init(current);
251 memcpy(current->signal->rlim, init_task.signal->rlim,
252 sizeof(current->signal->rlim));
253 atomic_inc(&(INIT_USER->__count));
254 write_unlock_irq(&tasklist_lock);
255 switch_uid(INIT_USER);
258 void __set_special_pids(pid_t session, pid_t pgrp)
260 struct task_struct *curr = current;
262 if (curr->signal->session != session) {
263 detach_pid(curr, PIDTYPE_SID);
264 curr->signal->session = session;
265 attach_pid(curr, PIDTYPE_SID, session);
267 if (process_group(curr) != pgrp) {
268 detach_pid(curr, PIDTYPE_PGID);
269 curr->signal->pgrp = pgrp;
270 attach_pid(curr, PIDTYPE_PGID, pgrp);
274 void set_special_pids(pid_t session, pid_t pgrp)
276 write_lock_irq(&tasklist_lock);
277 __set_special_pids(session, pgrp);
278 write_unlock_irq(&tasklist_lock);
282 * Let kernel threads use this to say that they
283 * allow a certain signal (since daemonize() will
284 * have disabled all of them by default).
286 int allow_signal(int sig)
288 if (!valid_signal(sig) || sig < 1)
291 spin_lock_irq(¤t->sighand->siglock);
292 sigdelset(¤t->blocked, sig);
294 /* Kernel threads handle their own signals.
295 Let the signal code know it'll be handled, so
296 that they don't get converted to SIGKILL or
297 just silently dropped */
298 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
301 spin_unlock_irq(¤t->sighand->siglock);
305 EXPORT_SYMBOL(allow_signal);
307 int disallow_signal(int sig)
309 if (!valid_signal(sig) || sig < 1)
312 spin_lock_irq(¤t->sighand->siglock);
313 sigaddset(¤t->blocked, sig);
315 spin_unlock_irq(¤t->sighand->siglock);
319 EXPORT_SYMBOL(disallow_signal);
322 * Put all the gunge required to become a kernel thread without
323 * attached user resources in one place where it belongs.
326 void daemonize(const char *name, ...)
329 struct fs_struct *fs;
332 va_start(args, name);
333 vsnprintf(current->comm, sizeof(current->comm), name, args);
337 * If we were started as result of loading a module, close all of the
338 * user space pages. We don't need them, and if we didn't close them
339 * they would be locked into memory.
343 set_special_pids(1, 1);
345 current->signal->tty = NULL;
348 /* Block and flush all signals */
349 sigfillset(&blocked);
350 sigprocmask(SIG_BLOCK, &blocked, NULL);
351 flush_signals(current);
353 /* Become as one with the init task */
355 exit_fs(current); /* current->fs->count--; */
358 atomic_inc(&fs->count);
360 current->files = init_task.files;
361 atomic_inc(¤t->files->count);
366 EXPORT_SYMBOL(daemonize);
368 static inline void close_files(struct files_struct * files)
376 * It is safe to dereference the fd table without RCU or
377 * ->file_lock because this is the last reference to the
380 fdt = files_fdtable(files);
384 if (i >= fdt->max_fdset || i >= fdt->max_fds)
386 set = fdt->open_fds->fds_bits[j++];
389 struct file * file = xchg(&fdt->fd[i], NULL);
391 filp_close(file, files);
399 struct files_struct *get_files_struct(struct task_struct *task)
401 struct files_struct *files;
406 atomic_inc(&files->count);
412 void fastcall put_files_struct(struct files_struct *files)
416 if (atomic_dec_and_test(&files->count)) {
419 * Free the fd and fdset arrays if we expanded them.
420 * If the fdtable was embedded, pass files for freeing
421 * at the end of the RCU grace period. Otherwise,
422 * you can free files immediately.
424 fdt = files_fdtable(files);
425 if (fdt == &files->fdtab)
426 fdt->free_files = files;
428 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 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 enter_lazy_tlb(mm, current);
532 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
535 * Make sure we're not reparenting to ourselves and that
536 * the parent is not a zombie.
538 BUG_ON(p == reaper || reaper->exit_state >= EXIT_ZOMBIE);
539 p->real_parent = reaper;
542 static inline void reparent_thread(task_t *p, task_t *father, int traced)
544 /* We don't want people slaying init. */
545 if (p->exit_signal != -1)
546 p->exit_signal = SIGCHLD;
548 if (p->pdeath_signal)
549 /* We already hold the tasklist_lock here. */
550 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
552 /* Move the child from its dying parent to the new one. */
553 if (unlikely(traced)) {
554 /* Preserve ptrace links if someone else is tracing this child. */
555 list_del_init(&p->ptrace_list);
556 if (p->parent != p->real_parent)
557 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
559 /* If this child is being traced, then we're the one tracing it
560 * anyway, so let go of it.
563 list_del_init(&p->sibling);
564 p->parent = p->real_parent;
565 list_add_tail(&p->sibling, &p->parent->children);
567 /* If we'd notified the old parent about this child's death,
568 * also notify the new parent.
570 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
571 thread_group_empty(p))
572 do_notify_parent(p, p->exit_signal);
573 else if (p->state == TASK_TRACED) {
575 * If it was at a trace stop, turn it into
576 * a normal stop since it's no longer being
584 * process group orphan check
585 * Case ii: Our child is in a different pgrp
586 * than we are, and it was the only connection
587 * outside, so the child pgrp is now orphaned.
589 if ((process_group(p) != process_group(father)) &&
590 (p->signal->session == father->signal->session)) {
591 int pgrp = process_group(p);
593 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
594 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
595 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
601 * When we die, we re-parent all our children.
602 * Try to give them to another thread in our thread
603 * group, and if no such member exists, give it to
604 * the global child reaper process (ie "init")
606 static inline void forget_original_parent(struct task_struct * father,
607 struct list_head *to_release)
609 struct task_struct *p, *reaper = father;
610 struct list_head *_p, *_n;
613 reaper = next_thread(reaper);
614 if (reaper == father) {
615 reaper = child_reaper;
618 } while (reaper->exit_state);
621 * There are only two places where our children can be:
623 * - in our child list
624 * - in our ptraced child list
626 * Search them and reparent children.
628 list_for_each_safe(_p, _n, &father->children) {
630 p = list_entry(_p,struct task_struct,sibling);
634 /* if father isn't the real parent, then ptrace must be enabled */
635 BUG_ON(father != p->real_parent && !ptrace);
637 if (father == p->real_parent) {
638 /* reparent with a reaper, real father it's us */
639 choose_new_parent(p, reaper, child_reaper);
640 reparent_thread(p, father, 0);
642 /* reparent ptraced task to its real parent */
644 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
645 thread_group_empty(p))
646 do_notify_parent(p, p->exit_signal);
650 * if the ptraced child is a zombie with exit_signal == -1
651 * we must collect it before we exit, or it will remain
652 * zombie forever since we prevented it from self-reap itself
653 * while it was being traced by us, to be able to see it in wait4.
655 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
656 list_add(&p->ptrace_list, to_release);
658 list_for_each_safe(_p, _n, &father->ptrace_children) {
659 p = list_entry(_p,struct task_struct,ptrace_list);
660 choose_new_parent(p, reaper, child_reaper);
661 reparent_thread(p, father, 1);
666 * Send signals to all our closest relatives so that they know
667 * to properly mourn us..
669 static void exit_notify(struct task_struct *tsk)
672 struct task_struct *t;
673 struct list_head ptrace_dead, *_p, *_n;
675 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
676 && !thread_group_empty(tsk)) {
678 * This occurs when there was a race between our exit
679 * syscall and a group signal choosing us as the one to
680 * wake up. It could be that we are the only thread
681 * alerted to check for pending signals, but another thread
682 * should be woken now to take the signal since we will not.
683 * Now we'll wake all the threads in the group just to make
684 * sure someone gets all the pending signals.
686 read_lock(&tasklist_lock);
687 spin_lock_irq(&tsk->sighand->siglock);
688 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
689 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
690 recalc_sigpending_tsk(t);
691 if (signal_pending(t))
692 signal_wake_up(t, 0);
694 spin_unlock_irq(&tsk->sighand->siglock);
695 read_unlock(&tasklist_lock);
698 write_lock_irq(&tasklist_lock);
701 * This does two things:
703 * A. Make init inherit all the child processes
704 * B. Check to see if any process groups have become orphaned
705 * as a result of our exiting, and if they have any stopped
706 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
709 INIT_LIST_HEAD(&ptrace_dead);
710 forget_original_parent(tsk, &ptrace_dead);
711 BUG_ON(!list_empty(&tsk->children));
712 BUG_ON(!list_empty(&tsk->ptrace_children));
715 * Check to see if any process groups have become orphaned
716 * as a result of our exiting, and if they have any stopped
717 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
719 * Case i: Our father is in a different pgrp than we are
720 * and we were the only connection outside, so our pgrp
721 * is about to become orphaned.
724 t = tsk->real_parent;
726 if ((process_group(t) != process_group(tsk)) &&
727 (t->signal->session == tsk->signal->session) &&
728 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
729 has_stopped_jobs(process_group(tsk))) {
730 __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
731 __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
734 /* Let father know we died
736 * Thread signals are configurable, but you aren't going to use
737 * that to send signals to arbitary processes.
738 * That stops right now.
740 * If the parent exec id doesn't match the exec id we saved
741 * when we started then we know the parent has changed security
744 * If our self_exec id doesn't match our parent_exec_id then
745 * we have changed execution domain as these two values started
746 * the same after a fork.
750 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
751 ( tsk->parent_exec_id != t->self_exec_id ||
752 tsk->self_exec_id != tsk->parent_exec_id)
753 && !capable(CAP_KILL))
754 tsk->exit_signal = SIGCHLD;
757 /* If something other than our normal parent is ptracing us, then
758 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
759 * only has special meaning to our real parent.
761 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
762 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
763 do_notify_parent(tsk, signal);
764 } else if (tsk->ptrace) {
765 do_notify_parent(tsk, SIGCHLD);
769 if (tsk->exit_signal == -1 &&
770 (likely(tsk->ptrace == 0) ||
771 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
773 tsk->exit_state = state;
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 == EXIT_DEAD)
788 fastcall NORET_TYPE void do_exit(long code)
790 struct task_struct *tsk = current;
793 profile_task_exit(tsk);
795 WARN_ON(atomic_read(&tsk->fs_excl));
797 if (unlikely(in_interrupt()))
798 panic("Aiee, killing interrupt handler!");
799 if (unlikely(!tsk->pid))
800 panic("Attempted to kill the idle task!");
801 if (unlikely(tsk->pid == 1))
802 panic("Attempted to kill init!");
806 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
807 current->ptrace_message = code;
808 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
812 * We're taking recursive faults here in do_exit. Safest is to just
813 * leave this task alone and wait for reboot.
815 if (unlikely(tsk->flags & PF_EXITING)) {
817 "Fixing recursive fault but reboot is needed!\n");
818 set_current_state(TASK_UNINTERRUPTIBLE);
822 tsk->flags |= PF_EXITING;
825 * Make sure we don't try to process any timer firings
826 * while we are already exiting.
828 tsk->it_virt_expires = cputime_zero;
829 tsk->it_prof_expires = cputime_zero;
830 tsk->it_sched_expires = 0;
832 if (unlikely(in_atomic()))
833 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
834 current->comm, current->pid,
837 acct_update_integrals(tsk);
839 update_hiwater_rss(tsk->mm);
840 update_hiwater_vm(tsk->mm);
842 group_dead = atomic_dec_and_test(&tsk->signal->live);
844 del_timer_sync(&tsk->signal->real_timer);
845 exit_itimers(tsk->signal);
858 if (group_dead && tsk->signal->leader)
859 disassociate_ctty(1);
861 module_put(task_thread_info(tsk)->exec_domain->module);
863 module_put(tsk->binfmt->module);
865 tsk->exit_code = code;
866 proc_exit_connector(tsk);
869 mpol_free(tsk->mempolicy);
870 tsk->mempolicy = NULL;
873 /* PF_DEAD causes final put_task_struct after we schedule. */
875 BUG_ON(tsk->flags & PF_DEAD);
876 tsk->flags |= PF_DEAD;
880 /* Avoid "noreturn function does return". */
884 EXPORT_SYMBOL_GPL(do_exit);
886 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
894 EXPORT_SYMBOL(complete_and_exit);
896 asmlinkage long sys_exit(int error_code)
898 do_exit((error_code&0xff)<<8);
901 task_t fastcall *next_thread(const task_t *p)
903 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
906 EXPORT_SYMBOL(next_thread);
909 * Take down every thread in the group. This is called by fatal signals
910 * as well as by sys_exit_group (below).
913 do_group_exit(int exit_code)
915 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
917 if (current->signal->flags & SIGNAL_GROUP_EXIT)
918 exit_code = current->signal->group_exit_code;
919 else if (!thread_group_empty(current)) {
920 struct signal_struct *const sig = current->signal;
921 struct sighand_struct *const sighand = current->sighand;
922 read_lock(&tasklist_lock);
923 spin_lock_irq(&sighand->siglock);
924 if (sig->flags & SIGNAL_GROUP_EXIT)
925 /* Another thread got here before we took the lock. */
926 exit_code = sig->group_exit_code;
928 sig->flags = SIGNAL_GROUP_EXIT;
929 sig->group_exit_code = exit_code;
930 zap_other_threads(current);
932 spin_unlock_irq(&sighand->siglock);
933 read_unlock(&tasklist_lock);
941 * this kills every thread in the thread group. Note that any externally
942 * wait4()-ing process will get the correct exit code - even if this
943 * thread is not the thread group leader.
945 asmlinkage void sys_exit_group(int error_code)
947 do_group_exit((error_code & 0xff) << 8);
950 static int eligible_child(pid_t pid, int options, task_t *p)
956 if (process_group(p) != process_group(current))
958 } else if (pid != -1) {
959 if (process_group(p) != -pid)
964 * Do not consider detached threads that are
967 if (p->exit_signal == -1 && !p->ptrace)
970 /* Wait for all children (clone and not) if __WALL is set;
971 * otherwise, wait for clone children *only* if __WCLONE is
972 * set; otherwise, wait for non-clone children *only*. (Note:
973 * A "clone" child here is one that reports to its parent
974 * using a signal other than SIGCHLD.) */
975 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
976 && !(options & __WALL))
979 * Do not consider thread group leaders that are
980 * in a non-empty thread group:
982 if (current->tgid != p->tgid && delay_group_leader(p))
985 if (security_task_wait(p))
991 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
993 struct siginfo __user *infop,
994 struct rusage __user *rusagep)
996 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
999 retval = put_user(SIGCHLD, &infop->si_signo);
1001 retval = put_user(0, &infop->si_errno);
1003 retval = put_user((short)why, &infop->si_code);
1005 retval = put_user(pid, &infop->si_pid);
1007 retval = put_user(uid, &infop->si_uid);
1009 retval = put_user(status, &infop->si_status);
1016 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1017 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1018 * the lock and this task is uninteresting. If we return nonzero, we have
1019 * released the lock and the system call should return.
1021 static int wait_task_zombie(task_t *p, int noreap,
1022 struct siginfo __user *infop,
1023 int __user *stat_addr, struct rusage __user *ru)
1025 unsigned long state;
1029 if (unlikely(noreap)) {
1032 int exit_code = p->exit_code;
1035 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1037 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1040 read_unlock(&tasklist_lock);
1041 if ((exit_code & 0x7f) == 0) {
1043 status = exit_code >> 8;
1045 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1046 status = exit_code & 0x7f;
1048 return wait_noreap_copyout(p, pid, uid, why,
1053 * Try to move the task's state to DEAD
1054 * only one thread is allowed to do this:
1056 state = xchg(&p->exit_state, EXIT_DEAD);
1057 if (state != EXIT_ZOMBIE) {
1058 BUG_ON(state != EXIT_DEAD);
1061 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1063 * This can only happen in a race with a ptraced thread
1064 * dying on another processor.
1069 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1071 * The resource counters for the group leader are in its
1072 * own task_struct. Those for dead threads in the group
1073 * are in its signal_struct, as are those for the child
1074 * processes it has previously reaped. All these
1075 * accumulate in the parent's signal_struct c* fields.
1077 * We don't bother to take a lock here to protect these
1078 * p->signal fields, because they are only touched by
1079 * __exit_signal, which runs with tasklist_lock
1080 * write-locked anyway, and so is excluded here. We do
1081 * need to protect the access to p->parent->signal fields,
1082 * as other threads in the parent group can be right
1083 * here reaping other children at the same time.
1085 spin_lock_irq(&p->parent->sighand->siglock);
1086 p->parent->signal->cutime =
1087 cputime_add(p->parent->signal->cutime,
1088 cputime_add(p->utime,
1089 cputime_add(p->signal->utime,
1090 p->signal->cutime)));
1091 p->parent->signal->cstime =
1092 cputime_add(p->parent->signal->cstime,
1093 cputime_add(p->stime,
1094 cputime_add(p->signal->stime,
1095 p->signal->cstime)));
1096 p->parent->signal->cmin_flt +=
1097 p->min_flt + p->signal->min_flt + p->signal->cmin_flt;
1098 p->parent->signal->cmaj_flt +=
1099 p->maj_flt + p->signal->maj_flt + p->signal->cmaj_flt;
1100 p->parent->signal->cnvcsw +=
1101 p->nvcsw + p->signal->nvcsw + p->signal->cnvcsw;
1102 p->parent->signal->cnivcsw +=
1103 p->nivcsw + p->signal->nivcsw + p->signal->cnivcsw;
1104 spin_unlock_irq(&p->parent->sighand->siglock);
1108 * Now we are sure this task is interesting, and no other
1109 * thread can reap it because we set its state to EXIT_DEAD.
1111 read_unlock(&tasklist_lock);
1113 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1114 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1115 ? p->signal->group_exit_code : p->exit_code;
1116 if (!retval && stat_addr)
1117 retval = put_user(status, stat_addr);
1118 if (!retval && infop)
1119 retval = put_user(SIGCHLD, &infop->si_signo);
1120 if (!retval && infop)
1121 retval = put_user(0, &infop->si_errno);
1122 if (!retval && infop) {
1125 if ((status & 0x7f) == 0) {
1129 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1132 retval = put_user((short)why, &infop->si_code);
1134 retval = put_user(status, &infop->si_status);
1136 if (!retval && infop)
1137 retval = put_user(p->pid, &infop->si_pid);
1138 if (!retval && infop)
1139 retval = put_user(p->uid, &infop->si_uid);
1141 // TODO: is this safe?
1142 p->exit_state = EXIT_ZOMBIE;
1146 if (p->real_parent != p->parent) {
1147 write_lock_irq(&tasklist_lock);
1148 /* Double-check with lock held. */
1149 if (p->real_parent != p->parent) {
1151 // TODO: is this safe?
1152 p->exit_state = EXIT_ZOMBIE;
1154 * If this is not a detached task, notify the parent.
1155 * If it's still not detached after that, don't release
1158 if (p->exit_signal != -1) {
1159 do_notify_parent(p, p->exit_signal);
1160 if (p->exit_signal != -1)
1164 write_unlock_irq(&tasklist_lock);
1173 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1174 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1175 * the lock and this task is uninteresting. If we return nonzero, we have
1176 * released the lock and the system call should return.
1178 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1179 struct siginfo __user *infop,
1180 int __user *stat_addr, struct rusage __user *ru)
1182 int retval, exit_code;
1186 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1187 p->signal && p->signal->group_stop_count > 0)
1189 * A group stop is in progress and this is the group leader.
1190 * We won't report until all threads have stopped.
1195 * Now we are pretty sure this task is interesting.
1196 * Make sure it doesn't get reaped out from under us while we
1197 * give up the lock and then examine it below. We don't want to
1198 * keep holding onto the tasklist_lock while we call getrusage and
1199 * possibly take page faults for user memory.
1202 read_unlock(&tasklist_lock);
1204 if (unlikely(noreap)) {
1207 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1209 exit_code = p->exit_code;
1210 if (unlikely(!exit_code) ||
1211 unlikely(p->state & TASK_TRACED))
1213 return wait_noreap_copyout(p, pid, uid,
1214 why, (exit_code << 8) | 0x7f,
1218 write_lock_irq(&tasklist_lock);
1221 * This uses xchg to be atomic with the thread resuming and setting
1222 * it. It must also be done with the write lock held to prevent a
1223 * race with the EXIT_ZOMBIE case.
1225 exit_code = xchg(&p->exit_code, 0);
1226 if (unlikely(p->exit_state)) {
1228 * The task resumed and then died. Let the next iteration
1229 * catch it in EXIT_ZOMBIE. Note that exit_code might
1230 * already be zero here if it resumed and did _exit(0).
1231 * The task itself is dead and won't touch exit_code again;
1232 * other processors in this function are locked out.
1234 p->exit_code = exit_code;
1237 if (unlikely(exit_code == 0)) {
1239 * Another thread in this function got to it first, or it
1240 * resumed, or it resumed and then died.
1242 write_unlock_irq(&tasklist_lock);
1246 * We are returning to the wait loop without having successfully
1247 * removed the process and having released the lock. We cannot
1248 * continue, since the "p" task pointer is potentially stale.
1250 * Return -EAGAIN, and do_wait() will restart the loop from the
1251 * beginning. Do _not_ re-acquire the lock.
1256 /* move to end of parent's list to avoid starvation */
1258 add_parent(p, p->parent);
1260 write_unlock_irq(&tasklist_lock);
1262 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1263 if (!retval && stat_addr)
1264 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1265 if (!retval && infop)
1266 retval = put_user(SIGCHLD, &infop->si_signo);
1267 if (!retval && infop)
1268 retval = put_user(0, &infop->si_errno);
1269 if (!retval && infop)
1270 retval = put_user((short)((p->ptrace & PT_PTRACED)
1271 ? CLD_TRAPPED : CLD_STOPPED),
1273 if (!retval && infop)
1274 retval = put_user(exit_code, &infop->si_status);
1275 if (!retval && infop)
1276 retval = put_user(p->pid, &infop->si_pid);
1277 if (!retval && infop)
1278 retval = put_user(p->uid, &infop->si_uid);
1288 * Handle do_wait work for one task in a live, non-stopped state.
1289 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1290 * the lock and this task is uninteresting. If we return nonzero, we have
1291 * released the lock and the system call should return.
1293 static int wait_task_continued(task_t *p, int noreap,
1294 struct siginfo __user *infop,
1295 int __user *stat_addr, struct rusage __user *ru)
1301 if (unlikely(!p->signal))
1304 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1307 spin_lock_irq(&p->sighand->siglock);
1308 /* Re-check with the lock held. */
1309 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1310 spin_unlock_irq(&p->sighand->siglock);
1314 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1315 spin_unlock_irq(&p->sighand->siglock);
1320 read_unlock(&tasklist_lock);
1323 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1325 if (!retval && stat_addr)
1326 retval = put_user(0xffff, stat_addr);
1330 retval = wait_noreap_copyout(p, pid, uid,
1331 CLD_CONTINUED, SIGCONT,
1333 BUG_ON(retval == 0);
1340 static inline int my_ptrace_child(struct task_struct *p)
1342 if (!(p->ptrace & PT_PTRACED))
1344 if (!(p->ptrace & PT_ATTACHED))
1347 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1348 * we are the attacher. If we are the real parent, this is a race
1349 * inside ptrace_attach. It is waiting for the tasklist_lock,
1350 * which we have to switch the parent links, but has already set
1351 * the flags in p->ptrace.
1353 return (p->parent != p->real_parent);
1356 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1357 int __user *stat_addr, struct rusage __user *ru)
1359 DECLARE_WAITQUEUE(wait, current);
1360 struct task_struct *tsk;
1363 add_wait_queue(¤t->signal->wait_chldexit,&wait);
1366 * We will set this flag if we see any child that might later
1367 * match our criteria, even if we are not able to reap it yet.
1370 current->state = TASK_INTERRUPTIBLE;
1371 read_lock(&tasklist_lock);
1374 struct task_struct *p;
1375 struct list_head *_p;
1378 list_for_each(_p,&tsk->children) {
1379 p = list_entry(_p,struct task_struct,sibling);
1381 ret = eligible_child(pid, options, p);
1388 * When we hit the race with PTRACE_ATTACH,
1389 * we will not report this child. But the
1390 * race means it has not yet been moved to
1391 * our ptrace_children list, so we need to
1392 * set the flag here to avoid a spurious ECHILD
1393 * when the race happens with the only child.
1396 if (!my_ptrace_child(p))
1401 * It's stopped now, so it might later
1402 * continue, exit, or stop again.
1405 if (!(options & WUNTRACED) &&
1406 !my_ptrace_child(p))
1408 retval = wait_task_stopped(p, ret == 2,
1409 (options & WNOWAIT),
1412 if (retval == -EAGAIN)
1414 if (retval != 0) /* He released the lock. */
1419 if (p->exit_state == EXIT_DEAD)
1421 // case EXIT_ZOMBIE:
1422 if (p->exit_state == EXIT_ZOMBIE) {
1424 * Eligible but we cannot release
1428 goto check_continued;
1429 if (!likely(options & WEXITED))
1431 retval = wait_task_zombie(
1432 p, (options & WNOWAIT),
1433 infop, stat_addr, ru);
1434 /* He released the lock. */
1441 * It's running now, so it might later
1442 * exit, stop, or stop and then continue.
1445 if (!unlikely(options & WCONTINUED))
1447 retval = wait_task_continued(
1448 p, (options & WNOWAIT),
1449 infop, stat_addr, ru);
1450 if (retval != 0) /* He released the lock. */
1456 list_for_each(_p, &tsk->ptrace_children) {
1457 p = list_entry(_p, struct task_struct,
1459 if (!eligible_child(pid, options, p))
1465 if (options & __WNOTHREAD)
1467 tsk = next_thread(tsk);
1468 if (tsk->signal != current->signal)
1470 } while (tsk != current);
1472 read_unlock(&tasklist_lock);
1475 if (options & WNOHANG)
1477 retval = -ERESTARTSYS;
1478 if (signal_pending(current))
1485 current->state = TASK_RUNNING;
1486 remove_wait_queue(¤t->signal->wait_chldexit,&wait);
1492 * For a WNOHANG return, clear out all the fields
1493 * we would set so the user can easily tell the
1497 retval = put_user(0, &infop->si_signo);
1499 retval = put_user(0, &infop->si_errno);
1501 retval = put_user(0, &infop->si_code);
1503 retval = put_user(0, &infop->si_pid);
1505 retval = put_user(0, &infop->si_uid);
1507 retval = put_user(0, &infop->si_status);
1513 asmlinkage long sys_waitid(int which, pid_t pid,
1514 struct siginfo __user *infop, int options,
1515 struct rusage __user *ru)
1519 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1521 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1541 ret = do_wait(pid, options, infop, NULL, ru);
1543 /* avoid REGPARM breakage on x86: */
1544 prevent_tail_call(ret);
1548 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1549 int options, struct rusage __user *ru)
1553 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1554 __WNOTHREAD|__WCLONE|__WALL))
1556 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1558 /* avoid REGPARM breakage on x86: */
1559 prevent_tail_call(ret);
1563 #ifdef __ARCH_WANT_SYS_WAITPID
1566 * sys_waitpid() remains for compatibility. waitpid() should be
1567 * implemented by calling sys_wait4() from libc.a.
1569 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1571 return sys_wait4(pid, stat_addr, options, NULL);