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/ckrm.h>
29 #include <linux/ckrm_tsk.h>
30 #include <linux/vs_limit.h>
31 #include <linux/ckrm_mem.h>
32 #include <linux/syscalls.h>
33 #include <linux/vs_limit.h>
35 #include <asm/uaccess.h>
36 #include <asm/unistd.h>
37 #include <asm/pgtable.h>
38 #include <asm/mmu_context.h>
40 extern void sem_exit (void);
41 extern struct task_struct *child_reaper;
43 int getrusage(struct task_struct *, int, struct rusage __user *);
45 static void __unhash_process(struct task_struct *p)
48 /* tasklist_lock is held, is this sufficient? */
50 atomic_dec(&p->vx_info->cvirt.nr_threads);
53 detach_pid(p, PIDTYPE_PID);
54 detach_pid(p, PIDTYPE_TGID);
55 if (thread_group_leader(p)) {
56 detach_pid(p, PIDTYPE_PGID);
57 detach_pid(p, PIDTYPE_SID);
59 __get_cpu_var(process_counts)--;
65 void release_task(struct task_struct * p)
69 struct dentry *proc_dentry;
72 atomic_dec(&p->user->processes);
73 spin_lock(&p->proc_lock);
74 proc_dentry = proc_pid_unhash(p);
75 write_lock_irq(&tasklist_lock);
76 if (unlikely(p->ptrace))
78 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
84 * If we are the last non-leader member of the thread
85 * group, and the leader is zombie, then notify the
86 * group leader's parent process. (if it wants notification.)
89 leader = p->group_leader;
90 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
91 BUG_ON(leader->exit_signal == -1);
92 do_notify_parent(leader, leader->exit_signal);
94 * If we were the last child thread and the leader has
95 * exited already, and the leader's parent ignores SIGCHLD,
96 * then we are the one who should release the leader.
98 * do_notify_parent() will have marked it self-reaping in
101 zap_leader = (leader->exit_signal == -1);
105 write_unlock_irq(&tasklist_lock);
106 spin_unlock(&p->proc_lock);
107 proc_pid_flush(proc_dentry);
112 if (unlikely(zap_leader))
116 /* we are using it only for SMP init */
118 void unhash_process(struct task_struct *p)
120 struct dentry *proc_dentry;
122 spin_lock(&p->proc_lock);
123 proc_dentry = proc_pid_unhash(p);
124 write_lock_irq(&tasklist_lock);
126 write_unlock_irq(&tasklist_lock);
127 spin_unlock(&p->proc_lock);
128 proc_pid_flush(proc_dentry);
132 * This checks not only the pgrp, but falls back on the pid if no
133 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
136 int session_of_pgrp(int pgrp)
138 struct task_struct *p;
141 read_lock(&tasklist_lock);
142 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
143 if (p->signal->session > 0) {
144 sid = p->signal->session;
147 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
148 p = find_task_by_pid(pgrp);
150 sid = p->signal->session;
152 read_unlock(&tasklist_lock);
158 * Determine if a process group is "orphaned", according to the POSIX
159 * definition in 2.2.2.52. Orphaned process groups are not to be affected
160 * by terminal-generated stop signals. Newly orphaned process groups are
161 * to receive a SIGHUP and a SIGCONT.
163 * "I ask you, have you ever known what it is to be an orphan?"
165 static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
167 struct task_struct *p;
170 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
171 if (p == ignored_task
172 || p->exit_state >= EXIT_ZOMBIE
173 || p->real_parent->pid == 1)
175 if (process_group(p->real_parent) != pgrp
176 && p->real_parent->signal->session == p->signal->session) {
180 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
181 return ret; /* (sighing) "Often!" */
184 int is_orphaned_pgrp(int pgrp)
188 read_lock(&tasklist_lock);
189 retval = will_become_orphaned_pgrp(pgrp, NULL);
190 read_unlock(&tasklist_lock);
195 static inline int has_stopped_jobs(int pgrp)
198 struct task_struct *p;
200 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
201 if (p->state != TASK_STOPPED)
204 /* If p is stopped by a debugger on a signal that won't
205 stop it, then don't count p as stopped. This isn't
206 perfect but it's a good approximation. */
207 if (unlikely (p->ptrace)
208 && p->exit_code != SIGSTOP
209 && p->exit_code != SIGTSTP
210 && p->exit_code != SIGTTOU
211 && p->exit_code != SIGTTIN)
216 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
221 * reparent_to_init() - Reparent the calling kernel thread to the init task.
223 * If a kernel thread is launched as a result of a system call, or if
224 * it ever exits, it should generally reparent itself to init so that
225 * it is correctly cleaned up on exit.
227 * The various task state such as scheduling policy and priority may have
228 * been inherited from a user process, so we reset them to sane values here.
230 * NOTE that reparent_to_init() gives the caller full capabilities.
232 void reparent_to_init(void)
234 write_lock_irq(&tasklist_lock);
236 ptrace_unlink(current);
237 /* Reparent to init */
238 REMOVE_LINKS(current);
239 /* FIXME handle vchild_reaper/initpid */
240 current->parent = child_reaper;
241 current->real_parent = child_reaper;
244 /* Set the exit signal to SIGCHLD so we signal init on exit */
245 current->exit_signal = SIGCHLD;
247 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
248 set_user_nice(current, 0);
252 security_task_reparent_to_init(current);
253 memcpy(current->signal->rlim, init_task.signal->rlim,
254 sizeof(current->signal->rlim));
255 atomic_inc(&(INIT_USER->__count));
256 write_unlock_irq(&tasklist_lock);
257 switch_uid(INIT_USER);
260 void __set_special_pids(pid_t session, pid_t pgrp)
262 struct task_struct *curr = current;
264 if (curr->signal->session != session) {
265 detach_pid(curr, PIDTYPE_SID);
266 curr->signal->session = session;
267 attach_pid(curr, PIDTYPE_SID, session);
269 if (process_group(curr) != pgrp) {
270 detach_pid(curr, PIDTYPE_PGID);
271 curr->signal->pgrp = pgrp;
272 attach_pid(curr, PIDTYPE_PGID, pgrp);
276 void set_special_pids(pid_t session, pid_t pgrp)
278 write_lock_irq(&tasklist_lock);
279 __set_special_pids(session, pgrp);
280 write_unlock_irq(&tasklist_lock);
284 * Let kernel threads use this to say that they
285 * allow a certain signal (since daemonize() will
286 * have disabled all of them by default).
288 int allow_signal(int sig)
290 if (sig < 1 || sig > _NSIG)
293 spin_lock_irq(¤t->sighand->siglock);
294 sigdelset(¤t->blocked, sig);
296 /* Kernel threads handle their own signals.
297 Let the signal code know it'll be handled, so
298 that they don't get converted to SIGKILL or
299 just silently dropped */
300 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
303 spin_unlock_irq(¤t->sighand->siglock);
307 EXPORT_SYMBOL(allow_signal);
309 int disallow_signal(int sig)
311 if (sig < 1 || sig > _NSIG)
314 spin_lock_irq(¤t->sighand->siglock);
315 sigaddset(¤t->blocked, sig);
317 spin_unlock_irq(¤t->sighand->siglock);
321 EXPORT_SYMBOL(disallow_signal);
324 * Put all the gunge required to become a kernel thread without
325 * attached user resources in one place where it belongs.
328 void daemonize(const char *name, ...)
331 struct fs_struct *fs;
334 va_start(args, name);
335 vsnprintf(current->comm, sizeof(current->comm), name, args);
339 * If we were started as result of loading a module, close all of the
340 * user space pages. We don't need them, and if we didn't close them
341 * they would be locked into memory.
345 set_special_pids(1, 1);
347 current->signal->tty = NULL;
350 /* Block and flush all signals */
351 sigfillset(&blocked);
352 sigprocmask(SIG_BLOCK, &blocked, NULL);
353 flush_signals(current);
355 /* Become as one with the init task */
357 exit_fs(current); /* current->fs->count--; */
360 atomic_inc(&fs->count);
362 current->files = init_task.files;
363 atomic_inc(¤t->files->count);
368 EXPORT_SYMBOL(daemonize);
370 static inline void close_files(struct files_struct * files)
378 if (i >= files->max_fdset || i >= files->max_fds)
380 set = files->open_fds->fds_bits[j++];
383 struct file * file = xchg(&files->fd[i], NULL);
385 filp_close(file, files);
394 struct files_struct *get_files_struct(struct task_struct *task)
396 struct files_struct *files;
401 atomic_inc(&files->count);
407 void fastcall put_files_struct(struct files_struct *files)
409 if (atomic_dec_and_test(&files->count)) {
412 * Free the fd and fdset arrays if we expanded them.
414 if (files->fd != &files->fd_array[0])
415 free_fd_array(files->fd, files->max_fds);
416 if (files->max_fdset > __FD_SETSIZE) {
417 free_fdset(files->open_fds, files->max_fdset);
418 free_fdset(files->close_on_exec, files->max_fdset);
420 kmem_cache_free(files_cachep, files);
424 EXPORT_SYMBOL(put_files_struct);
426 static inline void __exit_files(struct task_struct *tsk)
428 struct files_struct * files = tsk->files;
434 put_files_struct(files);
438 void exit_files(struct task_struct *tsk)
443 static inline void __put_fs_struct(struct fs_struct *fs)
445 /* No need to hold fs->lock if we are killing it */
446 if (atomic_dec_and_test(&fs->count)) {
453 mntput(fs->altrootmnt);
455 kmem_cache_free(fs_cachep, fs);
459 void put_fs_struct(struct fs_struct *fs)
464 static inline void __exit_fs(struct task_struct *tsk)
466 struct fs_struct * fs = tsk->fs;
476 void exit_fs(struct task_struct *tsk)
481 EXPORT_SYMBOL_GPL(exit_fs);
484 * Turn us into a lazy TLB process if we
487 static inline void __exit_mm(struct task_struct * tsk)
489 struct mm_struct *mm = tsk->mm;
495 * Serialize with any possible pending coredump.
496 * We must hold mmap_sem around checking core_waiters
497 * and clearing tsk->mm. The core-inducing thread
498 * will increment core_waiters for each thread in the
499 * group with ->mm != NULL.
501 down_read(&mm->mmap_sem);
502 if (mm->core_waiters) {
503 up_read(&mm->mmap_sem);
504 down_write(&mm->mmap_sem);
505 if (!--mm->core_waiters)
506 complete(mm->core_startup_done);
507 up_write(&mm->mmap_sem);
509 wait_for_completion(&mm->core_done);
510 down_read(&mm->mmap_sem);
512 atomic_inc(&mm->mm_count);
513 if (mm != tsk->active_mm) BUG();
514 /* more a memory barrier than a real lock */
517 up_read(&mm->mmap_sem);
518 #ifdef CONFIG_CKRM_RES_MEM
519 spin_lock(&mm->peertask_lock);
520 list_del_init(&tsk->mm_peers);
521 ckrm_mem_evaluate_mm(mm);
522 spin_unlock(&mm->peertask_lock);
524 enter_lazy_tlb(mm, current);
529 void exit_mm(struct task_struct *tsk)
534 static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
537 * Make sure we're not reparenting to ourselves and that
538 * the parent is not a zombie.
540 BUG_ON(p == reaper || reaper->state >= EXIT_ZOMBIE || reaper->exit_state >= EXIT_ZOMBIE);
541 p->real_parent = reaper;
542 if (p->parent == p->real_parent)
546 static inline void reparent_thread(task_t *p, task_t *father, int traced)
548 /* We don't want people slaying init. */
549 if (p->exit_signal != -1)
550 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->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
575 thread_group_empty(p))
576 do_notify_parent(p, p->exit_signal);
577 else if (p->state == TASK_TRACED) {
579 * If it was at a trace stop, turn it into
580 * a normal stop since it's no longer being
588 * process group orphan check
589 * Case ii: Our child is in a different pgrp
590 * than we are, and it was the only connection
591 * outside, so the child pgrp is now orphaned.
593 if ((process_group(p) != process_group(father)) &&
594 (p->signal->session == father->signal->session)) {
595 int pgrp = process_group(p);
597 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
598 __kill_pg_info(SIGHUP, (void *)1, pgrp);
599 __kill_pg_info(SIGCONT, (void *)1, pgrp);
605 * When we die, we re-parent all our children.
606 * Try to give them to another thread in our thread
607 * group, and if no such member exists, give it to
608 * the global child reaper process (ie "init")
610 static inline void forget_original_parent(struct task_struct * father,
611 struct list_head *to_release)
613 struct task_struct *p, *reaper = father;
614 struct list_head *_p, *_n;
616 /* FIXME handle vchild_reaper/initpid */
618 reaper = next_thread(reaper);
619 if (reaper == father) {
620 reaper = child_reaper;
623 } while (reaper->exit_state >= EXIT_ZOMBIE);
626 * There are only two places where our children can be:
628 * - in our child list
629 * - in our ptraced child list
631 * Search them and reparent children.
633 list_for_each_safe(_p, _n, &father->children) {
635 p = list_entry(_p,struct task_struct,sibling);
639 /* if father isn't the real parent, then ptrace must be enabled */
640 BUG_ON(father != p->real_parent && !ptrace);
642 if (father == p->real_parent) {
643 /* reparent with a reaper, real father it's us */
644 choose_new_parent(p, reaper, child_reaper);
645 reparent_thread(p, father, 0);
647 /* reparent ptraced task to its real parent */
649 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
650 thread_group_empty(p))
651 do_notify_parent(p, p->exit_signal);
655 * if the ptraced child is a zombie with exit_signal == -1
656 * we must collect it before we exit, or it will remain
657 * zombie forever since we prevented it from self-reap itself
658 * while it was being traced by us, to be able to see it in wait4.
660 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
661 list_add(&p->ptrace_list, to_release);
663 list_for_each_safe(_p, _n, &father->ptrace_children) {
664 p = list_entry(_p,struct task_struct,ptrace_list);
665 choose_new_parent(p, reaper, child_reaper);
666 reparent_thread(p, father, 1);
671 * Send signals to all our closest relatives so that they know
672 * to properly mourn us..
674 static void exit_notify(struct task_struct *tsk)
677 struct task_struct *t;
678 struct list_head ptrace_dead, *_p, *_n;
682 if (signal_pending(tsk) && !tsk->signal->group_exit
683 && !thread_group_empty(tsk)) {
685 * This occurs when there was a race between our exit
686 * syscall and a group signal choosing us as the one to
687 * wake up. It could be that we are the only thread
688 * alerted to check for pending signals, but another thread
689 * should be woken now to take the signal since we will not.
690 * Now we'll wake all the threads in the group just to make
691 * sure someone gets all the pending signals.
693 read_lock(&tasklist_lock);
694 spin_lock_irq(&tsk->sighand->siglock);
695 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
696 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
697 recalc_sigpending_tsk(t);
698 if (signal_pending(t))
699 signal_wake_up(t, 0);
701 spin_unlock_irq(&tsk->sighand->siglock);
702 read_unlock(&tasklist_lock);
705 write_lock_irq(&tasklist_lock);
708 * This does two things:
710 * A. Make init inherit all the child processes
711 * B. Check to see if any process groups have become orphaned
712 * as a result of our exiting, and if they have any stopped
713 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
716 INIT_LIST_HEAD(&ptrace_dead);
717 forget_original_parent(tsk, &ptrace_dead);
718 BUG_ON(!list_empty(&tsk->children));
719 BUG_ON(!list_empty(&tsk->ptrace_children));
722 * Check to see if any process groups have become orphaned
723 * as a result of our exiting, and if they have any stopped
724 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
726 * Case i: Our father is in a different pgrp than we are
727 * and we were the only connection outside, so our pgrp
728 * is about to become orphaned.
731 t = tsk->real_parent;
733 if ((process_group(t) != process_group(tsk)) &&
734 (t->signal->session == tsk->signal->session) &&
735 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
736 has_stopped_jobs(process_group(tsk))) {
737 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
738 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
741 /* Let father know we died
743 * Thread signals are configurable, but you aren't going to use
744 * that to send signals to arbitary processes.
745 * That stops right now.
747 * If the parent exec id doesn't match the exec id we saved
748 * when we started then we know the parent has changed security
751 * If our self_exec id doesn't match our parent_exec_id then
752 * we have changed execution domain as these two values started
753 * the same after a fork.
757 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
758 ( tsk->parent_exec_id != t->self_exec_id ||
759 tsk->self_exec_id != tsk->parent_exec_id)
760 && !capable(CAP_KILL))
761 tsk->exit_signal = SIGCHLD;
764 /* If something other than our normal parent is ptracing us, then
765 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
766 * only has special meaning to our real parent.
768 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
769 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
770 do_notify_parent(tsk, signal);
771 } else if (tsk->ptrace) {
772 do_notify_parent(tsk, SIGCHLD);
776 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
778 tsk->exit_state = state;
781 * Clear these here so that update_process_times() won't try to deliver
782 * itimer, profile or rlimit signals to this task while it is in late exit.
784 tsk->it_virt_value = 0;
785 tsk->it_prof_value = 0;
787 write_unlock_irq(&tasklist_lock);
789 list_for_each_safe(_p, _n, &ptrace_dead) {
791 t = list_entry(_p,struct task_struct,ptrace_list);
795 /* If the process is dead, release it - nobody will wait for it */
796 if (state == EXIT_DEAD)
799 /* PF_DEAD causes final put_task_struct after we schedule. */
801 tsk->flags |= PF_DEAD;
804 fastcall NORET_TYPE void do_exit(long code)
806 struct task_struct *tsk = current;
809 profile_task_exit(tsk);
811 if (unlikely(in_interrupt()))
812 panic("Aiee, killing interrupt handler!");
813 if (unlikely(!tsk->pid))
814 panic("Attempted to kill the idle task!");
815 if (unlikely(tsk->pid == 1))
816 panic("Attempted to kill init!");
819 tsk->flags |= PF_EXITING;
820 del_timer_sync(&tsk->real_timer);
822 if (unlikely(in_atomic()))
823 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
824 current->comm, current->pid,
827 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
828 current->ptrace_message = code;
829 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
832 group_dead = atomic_dec_and_test(&tsk->signal->live);
835 if (current->tux_info) {
836 #ifdef CONFIG_TUX_DEBUG
837 printk("Possibly unexpected TUX-thread exit(%ld) at %p?\n",
838 code, __builtin_return_address(0));
851 if (group_dead && tsk->signal->leader)
852 disassociate_ctty(1);
854 module_put(tsk->thread_info->exec_domain->module);
856 module_put(tsk->binfmt->module);
858 tsk->exit_code = code;
861 mpol_free(tsk->mempolicy);
862 tsk->mempolicy = NULL;
865 BUG_ON(!(current->flags & PF_DEAD));
868 /* Avoid "noreturn function does return". */
872 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
880 EXPORT_SYMBOL(complete_and_exit);
882 asmlinkage long sys_exit(int error_code)
884 do_exit((error_code&0xff)<<8);
887 task_t fastcall *next_thread(const task_t *p)
892 if (!spin_is_locked(&p->sighand->siglock) &&
893 !rwlock_is_locked(&tasklist_lock))
896 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
899 EXPORT_SYMBOL(next_thread);
902 * Take down every thread in the group. This is called by fatal signals
903 * as well as by sys_exit_group (below).
906 do_group_exit(int exit_code)
908 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
910 if (current->signal->group_exit)
911 exit_code = current->signal->group_exit_code;
912 else if (!thread_group_empty(current)) {
913 struct signal_struct *const sig = current->signal;
914 struct sighand_struct *const sighand = current->sighand;
915 read_lock(&tasklist_lock);
916 spin_lock_irq(&sighand->siglock);
918 /* Another thread got here before we took the lock. */
919 exit_code = sig->group_exit_code;
922 sig->group_exit_code = exit_code;
923 zap_other_threads(current);
925 spin_unlock_irq(&sighand->siglock);
926 read_unlock(&tasklist_lock);
934 * this kills every thread in the thread group. Note that any externally
935 * wait4()-ing process will get the correct exit code - even if this
936 * thread is not the thread group leader.
938 asmlinkage void sys_exit_group(int error_code)
940 do_group_exit((error_code & 0xff) << 8);
943 static int eligible_child(pid_t pid, int options, task_t *p)
949 if (process_group(p) != process_group(current))
951 } else if (pid != -1) {
952 if (process_group(p) != -pid)
957 * Do not consider detached threads that are
960 if (p->exit_signal == -1 && !p->ptrace)
963 /* Wait for all children (clone and not) if __WALL is set;
964 * otherwise, wait for clone children *only* if __WCLONE is
965 * set; otherwise, wait for non-clone children *only*. (Note:
966 * A "clone" child here is one that reports to its parent
967 * using a signal other than SIGCHLD.) */
968 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
969 && !(options & __WALL))
972 * Do not consider thread group leaders that are
973 * in a non-empty thread group:
975 if (current->tgid != p->tgid && delay_group_leader(p))
978 if (security_task_wait(p))
984 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
986 struct siginfo __user *infop,
987 struct rusage __user *rusagep)
989 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
992 retval = put_user(SIGCHLD, &infop->si_signo);
994 retval = put_user(0, &infop->si_errno);
996 retval = put_user((short)why, &infop->si_code);
998 retval = put_user(pid, &infop->si_pid);
1000 retval = put_user(uid, &infop->si_uid);
1002 retval = put_user(status, &infop->si_status);
1009 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. 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_zombie(task_t *p, int noreap,
1015 struct siginfo __user *infop,
1016 int __user *stat_addr, struct rusage __user *ru)
1018 unsigned long state;
1022 if (unlikely(noreap)) {
1025 int exit_code = p->exit_code;
1028 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1030 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1033 read_unlock(&tasklist_lock);
1034 if ((exit_code & 0x7f) == 0) {
1036 status = exit_code >> 8;
1038 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1039 status = exit_code & 0x7f;
1041 return wait_noreap_copyout(p, pid, uid, why,
1046 * Try to move the task's state to DEAD
1047 * only one thread is allowed to do this:
1049 state = xchg(&p->exit_state, EXIT_DEAD);
1050 if (state != EXIT_ZOMBIE) {
1051 BUG_ON(state != EXIT_DEAD);
1054 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1056 * This can only happen in a race with a ptraced thread
1057 * dying on another processor.
1062 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1064 * The resource counters for the group leader are in its
1065 * own task_struct. Those for dead threads in the group
1066 * are in its signal_struct, as are those for the child
1067 * processes it has previously reaped. All these
1068 * accumulate in the parent's signal_struct c* fields.
1070 * We don't bother to take a lock here to protect these
1071 * p->signal fields, because they are only touched by
1072 * __exit_signal, which runs with tasklist_lock
1073 * write-locked anyway, and so is excluded here. We do
1074 * need to protect the access to p->parent->signal fields,
1075 * as other threads in the parent group can be right
1076 * here reaping other children at the same time.
1078 spin_lock_irq(&p->parent->sighand->siglock);
1079 p->parent->signal->cutime +=
1080 p->utime + p->signal->utime + p->signal->cutime;
1081 p->parent->signal->cstime +=
1082 p->stime + p->signal->stime + p->signal->cstime;
1083 p->parent->signal->cmin_flt +=
1084 p->min_flt + p->signal->min_flt + p->signal->cmin_flt;
1085 p->parent->signal->cmaj_flt +=
1086 p->maj_flt + p->signal->maj_flt + p->signal->cmaj_flt;
1087 p->parent->signal->cnvcsw +=
1088 p->nvcsw + p->signal->nvcsw + p->signal->cnvcsw;
1089 p->parent->signal->cnivcsw +=
1090 p->nivcsw + p->signal->nivcsw + p->signal->cnivcsw;
1091 spin_unlock_irq(&p->parent->sighand->siglock);
1095 * Now we are sure this task is interesting, and no other
1096 * thread can reap it because we set its state to EXIT_DEAD.
1098 read_unlock(&tasklist_lock);
1100 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1101 status = p->signal->group_exit
1102 ? p->signal->group_exit_code : p->exit_code;
1103 if (!retval && stat_addr)
1104 retval = put_user(status, stat_addr);
1105 if (!retval && infop)
1106 retval = put_user(SIGCHLD, &infop->si_signo);
1107 if (!retval && infop)
1108 retval = put_user(0, &infop->si_errno);
1109 if (!retval && infop) {
1112 if ((status & 0x7f) == 0) {
1116 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1119 retval = put_user((short)why, &infop->si_code);
1121 retval = put_user(status, &infop->si_status);
1123 if (!retval && infop)
1124 retval = put_user(p->pid, &infop->si_pid);
1125 if (!retval && infop)
1126 retval = put_user(p->uid, &infop->si_uid);
1128 // TODO: is this safe?
1129 p->exit_state = EXIT_ZOMBIE;
1133 if (p->real_parent != p->parent) {
1134 write_lock_irq(&tasklist_lock);
1135 /* Double-check with lock held. */
1136 if (p->real_parent != p->parent) {
1138 // TODO: is this safe?
1139 p->exit_state = EXIT_ZOMBIE;
1141 * If this is not a detached task, notify the parent.
1142 * If it's still not detached after that, don't release
1145 if (p->exit_signal != -1) {
1146 do_notify_parent(p, p->exit_signal);
1147 if (p->exit_signal != -1)
1151 write_unlock_irq(&tasklist_lock);
1160 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1161 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1162 * the lock and this task is uninteresting. If we return nonzero, we have
1163 * released the lock and the system call should return.
1165 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1166 struct siginfo __user *infop,
1167 int __user *stat_addr, struct rusage __user *ru)
1169 int retval, exit_code;
1173 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1174 p->signal && p->signal->group_stop_count > 0)
1176 * A group stop is in progress and this is the group leader.
1177 * We won't report until all threads have stopped.
1182 * Now we are pretty sure this task is interesting.
1183 * Make sure it doesn't get reaped out from under us while we
1184 * give up the lock and then examine it below. We don't want to
1185 * keep holding onto the tasklist_lock while we call getrusage and
1186 * possibly take page faults for user memory.
1189 read_unlock(&tasklist_lock);
1191 if (unlikely(noreap)) {
1194 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1196 exit_code = p->exit_code;
1197 if (unlikely(!exit_code) ||
1198 unlikely(p->state > TASK_STOPPED))
1200 return wait_noreap_copyout(p, pid, uid,
1201 why, (exit_code << 8) | 0x7f,
1205 write_lock_irq(&tasklist_lock);
1208 * This uses xchg to be atomic with the thread resuming and setting
1209 * it. It must also be done with the write lock held to prevent a
1210 * race with the EXIT_ZOMBIE case.
1212 exit_code = xchg(&p->exit_code, 0);
1213 if (unlikely(p->exit_state >= EXIT_ZOMBIE)) {
1215 * The task resumed and then died. Let the next iteration
1216 * catch it in EXIT_ZOMBIE. Note that exit_code might
1217 * already be zero here if it resumed and did _exit(0).
1218 * The task itself is dead and won't touch exit_code again;
1219 * other processors in this function are locked out.
1221 p->exit_code = exit_code;
1224 if (unlikely(exit_code == 0)) {
1226 * Another thread in this function got to it first, or it
1227 * resumed, or it resumed and then died.
1229 write_unlock_irq(&tasklist_lock);
1233 * We are returning to the wait loop without having successfully
1234 * removed the process and having released the lock. We cannot
1235 * continue, since the "p" task pointer is potentially stale.
1237 * Return -EAGAIN, and do_wait() will restart the loop from the
1238 * beginning. Do _not_ re-acquire the lock.
1243 /* move to end of parent's list to avoid starvation */
1245 add_parent(p, p->parent);
1247 write_unlock_irq(&tasklist_lock);
1249 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1250 if (!retval && stat_addr)
1251 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1252 if (!retval && infop)
1253 retval = put_user(SIGCHLD, &infop->si_signo);
1254 if (!retval && infop)
1255 retval = put_user(0, &infop->si_errno);
1256 if (!retval && infop)
1257 retval = put_user((short)((p->ptrace & PT_PTRACED)
1258 ? CLD_TRAPPED : CLD_STOPPED),
1260 if (!retval && infop)
1261 retval = put_user(exit_code, &infop->si_status);
1262 if (!retval && infop)
1263 retval = put_user(p->pid, &infop->si_pid);
1264 if (!retval && infop)
1265 retval = put_user(p->uid, &infop->si_uid);
1275 * Handle do_wait work for one task in a live, non-stopped state.
1276 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1277 * the lock and this task is uninteresting. If we return nonzero, we have
1278 * released the lock and the system call should return.
1280 static int wait_task_continued(task_t *p, int noreap,
1281 struct siginfo __user *infop,
1282 int __user *stat_addr, struct rusage __user *ru)
1288 if (unlikely(!p->signal))
1291 if (p->signal->stop_state >= 0)
1294 spin_lock_irq(&p->sighand->siglock);
1295 if (p->signal->stop_state >= 0) { /* Re-check with the lock held. */
1296 spin_unlock_irq(&p->sighand->siglock);
1300 p->signal->stop_state = 0;
1301 spin_unlock_irq(&p->sighand->siglock);
1306 read_unlock(&tasklist_lock);
1309 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1311 if (!retval && stat_addr)
1312 retval = put_user(0xffff, stat_addr);
1316 retval = wait_noreap_copyout(p, pid, uid,
1317 CLD_CONTINUED, SIGCONT,
1319 BUG_ON(retval == 0);
1326 static inline int my_ptrace_child(struct task_struct *p)
1328 if (!(p->ptrace & PT_PTRACED))
1330 if (!(p->ptrace & PT_ATTACHED))
1333 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1334 * we are the attacher. If we are the real parent, this is a race
1335 * inside ptrace_attach. It is waiting for the tasklist_lock,
1336 * which we have to switch the parent links, but has already set
1337 * the flags in p->ptrace.
1339 return (p->parent != p->real_parent);
1342 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1343 int __user *stat_addr, struct rusage __user *ru)
1345 DECLARE_WAITQUEUE(wait, current);
1346 struct task_struct *tsk;
1349 add_wait_queue(¤t->wait_chldexit,&wait);
1352 * We will set this flag if we see any child that might later
1353 * match our criteria, even if we are not able to reap it yet.
1356 current->state = TASK_INTERRUPTIBLE;
1357 read_lock(&tasklist_lock);
1360 struct task_struct *p;
1361 struct list_head *_p;
1364 list_for_each(_p,&tsk->children) {
1365 p = list_entry(_p,struct task_struct,sibling);
1367 ret = eligible_child(pid, options, p);
1373 if (!my_ptrace_child(p))
1378 * It's stopped now, so it might later
1379 * continue, exit, or stop again.
1382 if (!(options & WUNTRACED) &&
1383 !my_ptrace_child(p))
1385 retval = wait_task_stopped(p, ret == 2,
1386 (options & WNOWAIT),
1389 if (retval == -EAGAIN)
1391 if (retval != 0) /* He released the lock. */
1396 if (p->exit_state == EXIT_DEAD)
1398 // case EXIT_ZOMBIE:
1399 if (p->exit_state == EXIT_ZOMBIE) {
1401 * Eligible but we cannot release
1405 goto check_continued;
1406 if (!likely(options & WEXITED))
1408 retval = wait_task_zombie(
1409 p, (options & WNOWAIT),
1410 infop, stat_addr, ru);
1411 /* He released the lock. */
1418 * It's running now, so it might later
1419 * exit, stop, or stop and then continue.
1422 if (!unlikely(options & WCONTINUED))
1425 retval = wait_task_continued(
1426 p, (options & WNOWAIT),
1427 infop, stat_addr, ru);
1428 if (retval != 0) /* He released the lock. */
1434 list_for_each(_p, &tsk->ptrace_children) {
1435 p = list_entry(_p, struct task_struct,
1437 if (!eligible_child(pid, options, p))
1443 if (options & __WNOTHREAD)
1445 tsk = next_thread(tsk);
1446 if (tsk->signal != current->signal)
1448 } while (tsk != current);
1450 read_unlock(&tasklist_lock);
1453 if (options & WNOHANG)
1455 retval = -ERESTARTSYS;
1456 if (signal_pending(current))
1463 current->state = TASK_RUNNING;
1464 remove_wait_queue(¤t->wait_chldexit,&wait);
1470 * For a WNOHANG return, clear out all the fields
1471 * we would set so the user can easily tell the
1475 retval = put_user(0, &infop->si_signo);
1477 retval = put_user(0, &infop->si_errno);
1479 retval = put_user(0, &infop->si_code);
1481 retval = put_user(0, &infop->si_pid);
1483 retval = put_user(0, &infop->si_uid);
1485 retval = put_user(0, &infop->si_status);
1491 asmlinkage long sys_waitid(int which, pid_t pid,
1492 struct siginfo __user *infop, int options,
1493 struct rusage __user *ru)
1497 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1499 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1519 ret = do_wait(pid, options, infop, NULL, ru);
1521 /* avoid REGPARM breakage on x86: */
1522 prevent_tail_call(ret);
1526 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1527 int options, struct rusage __user *ru)
1531 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1532 __WNOTHREAD|__WCLONE|__WALL))
1534 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1536 /* avoid REGPARM breakage on x86: */
1537 prevent_tail_call(ret);
1541 #ifdef __ARCH_WANT_SYS_WAITPID
1544 * sys_waitpid() remains for compatibility. waitpid() should be
1545 * implemented by calling sys_wait4() from libc.a.
1547 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1549 return sys_wait4(pid, stat_addr, options, NULL);