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>
34 #include <asm/uaccess.h>
35 #include <asm/unistd.h>
36 #include <asm/pgtable.h>
37 #include <asm/mmu_context.h>
39 extern void sem_exit (void);
40 extern struct task_struct *child_reaper;
42 int getrusage(struct task_struct *, int, struct rusage __user *);
44 static void __unhash_process(struct task_struct *p)
47 /* tasklist_lock is held, is this sufficient? */
49 atomic_dec(&p->vx_info->cvirt.nr_threads);
52 detach_pid(p, PIDTYPE_PID);
53 detach_pid(p, PIDTYPE_TGID);
54 if (thread_group_leader(p)) {
55 detach_pid(p, PIDTYPE_PGID);
56 detach_pid(p, PIDTYPE_SID);
58 __get_cpu_var(process_counts)--;
64 void release_task(struct task_struct * p)
68 struct dentry *proc_dentry;
71 atomic_dec(&p->user->processes);
72 spin_lock(&p->proc_lock);
73 proc_dentry = proc_pid_unhash(p);
74 write_lock_irq(&tasklist_lock);
75 if (unlikely(p->ptrace))
77 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
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
171 || p->exit_state >= EXIT_ZOMBIE
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 void reparent_to_init(void)
233 write_lock_irq(&tasklist_lock);
235 ptrace_unlink(current);
236 /* Reparent to init */
237 REMOVE_LINKS(current);
238 /* FIXME handle vchild_reaper/initpid */
239 current->parent = child_reaper;
240 current->real_parent = child_reaper;
243 /* Set the exit signal to SIGCHLD so we signal init on exit */
244 current->exit_signal = SIGCHLD;
246 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
247 set_user_nice(current, 0);
251 security_task_reparent_to_init(current);
252 memcpy(current->signal->rlim, init_task.signal->rlim,
253 sizeof(current->signal->rlim));
254 atomic_inc(&(INIT_USER->__count));
255 write_unlock_irq(&tasklist_lock);
256 switch_uid(INIT_USER);
259 void __set_special_pids(pid_t session, pid_t pgrp)
261 struct task_struct *curr = current;
263 if (curr->signal->session != session) {
264 detach_pid(curr, PIDTYPE_SID);
265 curr->signal->session = session;
266 attach_pid(curr, PIDTYPE_SID, session);
268 if (process_group(curr) != pgrp) {
269 detach_pid(curr, PIDTYPE_PGID);
270 curr->signal->pgrp = pgrp;
271 attach_pid(curr, PIDTYPE_PGID, pgrp);
275 void set_special_pids(pid_t session, pid_t pgrp)
277 write_lock_irq(&tasklist_lock);
278 __set_special_pids(session, pgrp);
279 write_unlock_irq(&tasklist_lock);
283 * Let kernel threads use this to say that they
284 * allow a certain signal (since daemonize() will
285 * have disabled all of them by default).
287 int allow_signal(int sig)
289 if (sig < 1 || sig > _NSIG)
292 spin_lock_irq(¤t->sighand->siglock);
293 sigdelset(¤t->blocked, sig);
295 /* Kernel threads handle their own signals.
296 Let the signal code know it'll be handled, so
297 that they don't get converted to SIGKILL or
298 just silently dropped */
299 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
302 spin_unlock_irq(¤t->sighand->siglock);
306 EXPORT_SYMBOL(allow_signal);
308 int disallow_signal(int sig)
310 if (sig < 1 || sig > _NSIG)
313 spin_lock_irq(¤t->sighand->siglock);
314 sigaddset(¤t->blocked, sig);
316 spin_unlock_irq(¤t->sighand->siglock);
320 EXPORT_SYMBOL(disallow_signal);
323 * Put all the gunge required to become a kernel thread without
324 * attached user resources in one place where it belongs.
327 void daemonize(const char *name, ...)
330 struct fs_struct *fs;
333 va_start(args, name);
334 vsnprintf(current->comm, sizeof(current->comm), name, args);
338 * If we were started as result of loading a module, close all of the
339 * user space pages. We don't need them, and if we didn't close them
340 * they would be locked into memory.
344 set_special_pids(1, 1);
346 current->signal->tty = NULL;
349 /* Block and flush all signals */
350 sigfillset(&blocked);
351 sigprocmask(SIG_BLOCK, &blocked, NULL);
352 flush_signals(current);
354 /* Become as one with the init task */
356 exit_fs(current); /* current->fs->count--; */
359 atomic_inc(&fs->count);
361 current->files = init_task.files;
362 atomic_inc(¤t->files->count);
367 EXPORT_SYMBOL(daemonize);
369 static inline void close_files(struct files_struct * files)
377 if (i >= files->max_fdset || i >= files->max_fds)
379 set = files->open_fds->fds_bits[j++];
382 struct file * file = xchg(&files->fd[i], NULL);
384 filp_close(file, files);
392 struct files_struct *get_files_struct(struct task_struct *task)
394 struct files_struct *files;
399 atomic_inc(&files->count);
405 void fastcall put_files_struct(struct files_struct *files)
407 if (atomic_dec_and_test(&files->count)) {
410 * Free the fd and fdset arrays if we expanded them.
412 if (files->fd != &files->fd_array[0])
413 free_fd_array(files->fd, files->max_fds);
414 if (files->max_fdset > __FD_SETSIZE) {
415 free_fdset(files->open_fds, files->max_fdset);
416 free_fdset(files->close_on_exec, files->max_fdset);
418 kmem_cache_free(files_cachep, files);
422 EXPORT_SYMBOL(put_files_struct);
424 static inline void __exit_files(struct task_struct *tsk)
426 struct files_struct * files = tsk->files;
432 put_files_struct(files);
436 void exit_files(struct task_struct *tsk)
441 static inline void __put_fs_struct(struct fs_struct *fs)
443 /* No need to hold fs->lock if we are killing it */
444 if (atomic_dec_and_test(&fs->count)) {
451 mntput(fs->altrootmnt);
453 kmem_cache_free(fs_cachep, fs);
457 void put_fs_struct(struct fs_struct *fs)
462 static inline void __exit_fs(struct task_struct *tsk)
464 struct fs_struct * fs = tsk->fs;
474 void exit_fs(struct task_struct *tsk)
479 EXPORT_SYMBOL_GPL(exit_fs);
482 * Turn us into a lazy TLB process if we
485 static inline void __exit_mm(struct task_struct * tsk)
487 struct mm_struct *mm = tsk->mm;
493 * Serialize with any possible pending coredump.
494 * We must hold mmap_sem around checking core_waiters
495 * and clearing tsk->mm. The core-inducing thread
496 * will increment core_waiters for each thread in the
497 * group with ->mm != NULL.
499 down_read(&mm->mmap_sem);
500 if (mm->core_waiters) {
501 up_read(&mm->mmap_sem);
502 down_write(&mm->mmap_sem);
503 if (!--mm->core_waiters)
504 complete(mm->core_startup_done);
505 up_write(&mm->mmap_sem);
507 wait_for_completion(&mm->core_done);
508 down_read(&mm->mmap_sem);
510 atomic_inc(&mm->mm_count);
511 if (mm != tsk->active_mm) BUG();
512 /* more a memory barrier than a real lock */
515 up_read(&mm->mmap_sem);
516 #ifdef CONFIG_CKRM_RES_MEM
517 spin_lock(&mm->peertask_lock);
518 list_del_init(&tsk->mm_peers);
519 ckrm_mem_evaluate_mm(mm);
520 spin_unlock(&mm->peertask_lock);
522 enter_lazy_tlb(mm, current);
527 void exit_mm(struct task_struct *tsk)
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->state >= EXIT_ZOMBIE || reaper->exit_state >= EXIT_ZOMBIE);
539 p->real_parent = reaper;
540 if (p->parent == p->real_parent)
544 static inline void reparent_thread(task_t *p, task_t *father, int traced)
546 /* We don't want people slaying init. */
547 if (p->exit_signal != -1)
548 p->exit_signal = SIGCHLD;
550 if (p->pdeath_signal)
551 /* We already hold the tasklist_lock here. */
552 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
554 /* Move the child from its dying parent to the new one. */
555 if (unlikely(traced)) {
556 /* Preserve ptrace links if someone else is tracing this child. */
557 list_del_init(&p->ptrace_list);
558 if (p->parent != p->real_parent)
559 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
561 /* If this child is being traced, then we're the one tracing it
562 * anyway, so let go of it.
565 list_del_init(&p->sibling);
566 p->parent = p->real_parent;
567 list_add_tail(&p->sibling, &p->parent->children);
569 /* If we'd notified the old parent about this child's death,
570 * also notify the new parent.
572 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
573 thread_group_empty(p))
574 do_notify_parent(p, p->exit_signal);
575 else if (p->state == TASK_TRACED) {
577 * If it was at a trace stop, turn it into
578 * a normal stop since it's no longer being
586 * process group orphan check
587 * Case ii: Our child is in a different pgrp
588 * than we are, and it was the only connection
589 * outside, so the child pgrp is now orphaned.
591 if ((process_group(p) != process_group(father)) &&
592 (p->signal->session == father->signal->session)) {
593 int pgrp = process_group(p);
595 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
596 __kill_pg_info(SIGHUP, (void *)1, pgrp);
597 __kill_pg_info(SIGCONT, (void *)1, pgrp);
603 * When we die, we re-parent all our children.
604 * Try to give them to another thread in our thread
605 * group, and if no such member exists, give it to
606 * the global child reaper process (ie "init")
608 static inline void forget_original_parent(struct task_struct * father,
609 struct list_head *to_release)
611 struct task_struct *p, *reaper = father;
612 struct list_head *_p, *_n;
615 reaper = next_thread(reaper);
616 if (reaper == father) {
617 reaper = child_reaper;
620 } while (reaper->exit_state >= EXIT_ZOMBIE);
623 * There are only two places where our children can be:
625 * - in our child list
626 * - in our ptraced child list
628 * Search them and reparent children.
630 list_for_each_safe(_p, _n, &father->children) {
632 p = list_entry(_p,struct task_struct,sibling);
636 /* if father isn't the real parent, then ptrace must be enabled */
637 BUG_ON(father != p->real_parent && !ptrace);
639 if (father == p->real_parent) {
640 /* reparent with a reaper, real father it's us */
641 choose_new_parent(p, reaper, child_reaper);
642 reparent_thread(p, father, 0);
644 /* reparent ptraced task to its real parent */
646 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
647 thread_group_empty(p))
648 do_notify_parent(p, p->exit_signal);
652 * if the ptraced child is a zombie with exit_signal == -1
653 * we must collect it before we exit, or it will remain
654 * zombie forever since we prevented it from self-reap itself
655 * while it was being traced by us, to be able to see it in wait4.
657 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
658 list_add(&p->ptrace_list, to_release);
660 list_for_each_safe(_p, _n, &father->ptrace_children) {
661 p = list_entry(_p,struct task_struct,ptrace_list);
662 choose_new_parent(p, reaper, child_reaper);
663 reparent_thread(p, father, 1);
668 * Send signals to all our closest relatives so that they know
669 * to properly mourn us..
671 static void exit_notify(struct task_struct *tsk)
674 struct task_struct *t;
675 struct list_head ptrace_dead, *_p, *_n;
679 if (signal_pending(tsk) && !tsk->signal->group_exit
680 && !thread_group_empty(tsk)) {
682 * This occurs when there was a race between our exit
683 * syscall and a group signal choosing us as the one to
684 * wake up. It could be that we are the only thread
685 * alerted to check for pending signals, but another thread
686 * should be woken now to take the signal since we will not.
687 * Now we'll wake all the threads in the group just to make
688 * sure someone gets all the pending signals.
690 read_lock(&tasklist_lock);
691 spin_lock_irq(&tsk->sighand->siglock);
692 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
693 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
694 recalc_sigpending_tsk(t);
695 if (signal_pending(t))
696 signal_wake_up(t, 0);
698 spin_unlock_irq(&tsk->sighand->siglock);
699 read_unlock(&tasklist_lock);
702 write_lock_irq(&tasklist_lock);
705 * This does two things:
707 * A. Make init inherit all the child processes
708 * B. Check to see if any process groups have become orphaned
709 * as a result of our exiting, and if they have any stopped
710 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
713 INIT_LIST_HEAD(&ptrace_dead);
714 forget_original_parent(tsk, &ptrace_dead);
715 BUG_ON(!list_empty(&tsk->children));
716 BUG_ON(!list_empty(&tsk->ptrace_children));
719 * Check to see if any process groups have become orphaned
720 * as a result of our exiting, and if they have any stopped
721 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
723 * Case i: Our father is in a different pgrp than we are
724 * and we were the only connection outside, so our pgrp
725 * is about to become orphaned.
728 t = tsk->real_parent;
730 if ((process_group(t) != process_group(tsk)) &&
731 (t->signal->session == tsk->signal->session) &&
732 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
733 has_stopped_jobs(process_group(tsk))) {
734 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
735 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
738 /* Let father know we died
740 * Thread signals are configurable, but you aren't going to use
741 * that to send signals to arbitary processes.
742 * That stops right now.
744 * If the parent exec id doesn't match the exec id we saved
745 * when we started then we know the parent has changed security
748 * If our self_exec id doesn't match our parent_exec_id then
749 * we have changed execution domain as these two values started
750 * the same after a fork.
754 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
755 ( tsk->parent_exec_id != t->self_exec_id ||
756 tsk->self_exec_id != tsk->parent_exec_id)
757 && !capable(CAP_KILL))
758 tsk->exit_signal = SIGCHLD;
761 /* If something other than our normal parent is ptracing us, then
762 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
763 * only has special meaning to our real parent.
765 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
766 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
767 do_notify_parent(tsk, signal);
768 } else if (tsk->ptrace) {
769 do_notify_parent(tsk, SIGCHLD);
773 if (tsk->exit_signal == -1 && tsk->ptrace == 0)
775 tsk->exit_state = state;
778 * Clear these here so that update_process_times() won't try to deliver
779 * itimer, profile or rlimit signals to this task while it is in late exit.
781 tsk->it_virt_value = 0;
782 tsk->it_prof_value = 0;
784 write_unlock_irq(&tasklist_lock);
786 list_for_each_safe(_p, _n, &ptrace_dead) {
788 t = list_entry(_p,struct task_struct,ptrace_list);
792 /* If the process is dead, release it - nobody will wait for it */
793 if (state == EXIT_DEAD)
796 /* PF_DEAD causes final put_task_struct after we schedule. */
798 tsk->flags |= PF_DEAD;
801 fastcall NORET_TYPE void do_exit(long code)
803 struct task_struct *tsk = current;
806 profile_task_exit(tsk);
808 if (unlikely(in_interrupt()))
809 panic("Aiee, killing interrupt handler!");
810 if (unlikely(!tsk->pid))
811 panic("Attempted to kill the idle task!");
812 if (unlikely(tsk->pid == 1))
813 panic("Attempted to kill init!");
816 tsk->flags |= PF_EXITING;
817 del_timer_sync(&tsk->real_timer);
819 if (unlikely(in_atomic()))
820 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
821 current->comm, current->pid,
824 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
825 current->ptrace_message = code;
826 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
829 group_dead = atomic_dec_and_test(&tsk->signal->live);
832 if (current->tux_info) {
833 #ifdef CONFIG_TUX_DEBUG
834 printk("Possibly unexpected TUX-thread exit(%ld) at %p?\n",
835 code, __builtin_return_address(0));
848 if (group_dead && tsk->signal->leader)
849 disassociate_ctty(1);
851 module_put(tsk->thread_info->exec_domain->module);
853 module_put(tsk->binfmt->module);
855 tsk->exit_code = code;
858 mpol_free(tsk->mempolicy);
859 tsk->mempolicy = NULL;
862 BUG_ON(!(current->flags & PF_DEAD));
865 /* Avoid "noreturn function does return". */
869 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
877 EXPORT_SYMBOL(complete_and_exit);
879 asmlinkage long sys_exit(int error_code)
881 do_exit((error_code&0xff)<<8);
884 task_t fastcall *next_thread(const task_t *p)
889 if (!spin_is_locked(&p->sighand->siglock) &&
890 !rwlock_is_locked(&tasklist_lock))
893 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
896 EXPORT_SYMBOL(next_thread);
899 * Take down every thread in the group. This is called by fatal signals
900 * as well as by sys_exit_group (below).
903 do_group_exit(int exit_code)
905 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
907 if (current->signal->group_exit)
908 exit_code = current->signal->group_exit_code;
909 else if (!thread_group_empty(current)) {
910 struct signal_struct *const sig = current->signal;
911 struct sighand_struct *const sighand = current->sighand;
912 read_lock(&tasklist_lock);
913 spin_lock_irq(&sighand->siglock);
915 /* Another thread got here before we took the lock. */
916 exit_code = sig->group_exit_code;
919 sig->group_exit_code = exit_code;
920 zap_other_threads(current);
922 spin_unlock_irq(&sighand->siglock);
923 read_unlock(&tasklist_lock);
931 * this kills every thread in the thread group. Note that any externally
932 * wait4()-ing process will get the correct exit code - even if this
933 * thread is not the thread group leader.
935 asmlinkage void sys_exit_group(int error_code)
937 do_group_exit((error_code & 0xff) << 8);
940 static int eligible_child(pid_t pid, int options, task_t *p)
946 if (process_group(p) != process_group(current))
948 } else if (pid != -1) {
949 if (process_group(p) != -pid)
954 * Do not consider detached threads that are
957 if (p->exit_signal == -1 && !p->ptrace)
960 /* Wait for all children (clone and not) if __WALL is set;
961 * otherwise, wait for clone children *only* if __WCLONE is
962 * set; otherwise, wait for non-clone children *only*. (Note:
963 * A "clone" child here is one that reports to its parent
964 * using a signal other than SIGCHLD.) */
965 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
966 && !(options & __WALL))
969 * Do not consider thread group leaders that are
970 * in a non-empty thread group:
972 if (current->tgid != p->tgid && delay_group_leader(p))
975 if (security_task_wait(p))
981 static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
983 struct siginfo __user *infop,
984 struct rusage __user *rusagep)
986 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
989 retval = put_user(SIGCHLD, &infop->si_signo);
991 retval = put_user(0, &infop->si_errno);
993 retval = put_user((short)why, &infop->si_code);
995 retval = put_user(pid, &infop->si_pid);
997 retval = put_user(uid, &infop->si_uid);
999 retval = put_user(status, &infop->si_status);
1006 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1007 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1008 * the lock and this task is uninteresting. If we return nonzero, we have
1009 * released the lock and the system call should return.
1011 static int wait_task_zombie(task_t *p, int noreap,
1012 struct siginfo __user *infop,
1013 int __user *stat_addr, struct rusage __user *ru)
1015 unsigned long state;
1019 if (unlikely(noreap)) {
1022 int exit_code = p->exit_code;
1025 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1027 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1030 read_unlock(&tasklist_lock);
1031 if ((exit_code & 0x7f) == 0) {
1033 status = exit_code >> 8;
1035 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1036 status = exit_code & 0x7f;
1038 return wait_noreap_copyout(p, pid, uid, why,
1043 * Try to move the task's state to DEAD
1044 * only one thread is allowed to do this:
1046 state = xchg(&p->exit_state, EXIT_DEAD);
1047 if (state != EXIT_ZOMBIE) {
1048 BUG_ON(state != EXIT_DEAD);
1051 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1053 * This can only happen in a race with a ptraced thread
1054 * dying on another processor.
1059 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1061 * The resource counters for the group leader are in its
1062 * own task_struct. Those for dead threads in the group
1063 * are in its signal_struct, as are those for the child
1064 * processes it has previously reaped. All these
1065 * accumulate in the parent's signal_struct c* fields.
1067 * We don't bother to take a lock here to protect these
1068 * p->signal fields, because they are only touched by
1069 * __exit_signal, which runs with tasklist_lock
1070 * write-locked anyway, and so is excluded here. We do
1071 * need to protect the access to p->parent->signal fields,
1072 * as other threads in the parent group can be right
1073 * here reaping other children at the same time.
1075 spin_lock_irq(&p->parent->sighand->siglock);
1076 p->parent->signal->cutime +=
1077 p->utime + p->signal->utime + p->signal->cutime;
1078 p->parent->signal->cstime +=
1079 p->stime + p->signal->stime + p->signal->cstime;
1080 p->parent->signal->cmin_flt +=
1081 p->min_flt + p->signal->min_flt + p->signal->cmin_flt;
1082 p->parent->signal->cmaj_flt +=
1083 p->maj_flt + p->signal->maj_flt + p->signal->cmaj_flt;
1084 p->parent->signal->cnvcsw +=
1085 p->nvcsw + p->signal->nvcsw + p->signal->cnvcsw;
1086 p->parent->signal->cnivcsw +=
1087 p->nivcsw + p->signal->nivcsw + p->signal->cnivcsw;
1088 spin_unlock_irq(&p->parent->sighand->siglock);
1092 * Now we are sure this task is interesting, and no other
1093 * thread can reap it because we set its state to EXIT_DEAD.
1095 read_unlock(&tasklist_lock);
1097 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1098 status = p->signal->group_exit
1099 ? p->signal->group_exit_code : p->exit_code;
1100 if (!retval && stat_addr)
1101 retval = put_user(status, stat_addr);
1102 if (!retval && infop)
1103 retval = put_user(SIGCHLD, &infop->si_signo);
1104 if (!retval && infop)
1105 retval = put_user(0, &infop->si_errno);
1106 if (!retval && infop) {
1109 if ((status & 0x7f) == 0) {
1113 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1116 retval = put_user((short)why, &infop->si_code);
1118 retval = put_user(status, &infop->si_status);
1120 if (!retval && infop)
1121 retval = put_user(p->pid, &infop->si_pid);
1122 if (!retval && infop)
1123 retval = put_user(p->uid, &infop->si_uid);
1125 // TODO: is this safe?
1126 p->exit_state = EXIT_ZOMBIE;
1130 if (p->real_parent != p->parent) {
1131 write_lock_irq(&tasklist_lock);
1132 /* Double-check with lock held. */
1133 if (p->real_parent != p->parent) {
1135 // TODO: is this safe?
1136 p->exit_state = EXIT_ZOMBIE;
1138 * If this is not a detached task, notify the parent.
1139 * If it's still not detached after that, don't release
1142 if (p->exit_signal != -1) {
1143 do_notify_parent(p, p->exit_signal);
1144 if (p->exit_signal != -1)
1148 write_unlock_irq(&tasklist_lock);
1157 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1158 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1159 * the lock and this task is uninteresting. If we return nonzero, we have
1160 * released the lock and the system call should return.
1162 static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1163 struct siginfo __user *infop,
1164 int __user *stat_addr, struct rusage __user *ru)
1166 int retval, exit_code;
1170 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1171 p->signal && p->signal->group_stop_count > 0)
1173 * A group stop is in progress and this is the group leader.
1174 * We won't report until all threads have stopped.
1179 * Now we are pretty sure this task is interesting.
1180 * Make sure it doesn't get reaped out from under us while we
1181 * give up the lock and then examine it below. We don't want to
1182 * keep holding onto the tasklist_lock while we call getrusage and
1183 * possibly take page faults for user memory.
1186 read_unlock(&tasklist_lock);
1188 if (unlikely(noreap)) {
1191 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1193 exit_code = p->exit_code;
1194 if (unlikely(!exit_code) ||
1195 unlikely(p->state > TASK_STOPPED))
1197 return wait_noreap_copyout(p, pid, uid,
1198 why, (exit_code << 8) | 0x7f,
1202 write_lock_irq(&tasklist_lock);
1205 * This uses xchg to be atomic with the thread resuming and setting
1206 * it. It must also be done with the write lock held to prevent a
1207 * race with the EXIT_ZOMBIE case.
1209 exit_code = xchg(&p->exit_code, 0);
1210 if (unlikely(p->exit_state >= EXIT_ZOMBIE)) {
1212 * The task resumed and then died. Let the next iteration
1213 * catch it in EXIT_ZOMBIE. Note that exit_code might
1214 * already be zero here if it resumed and did _exit(0).
1215 * The task itself is dead and won't touch exit_code again;
1216 * other processors in this function are locked out.
1218 p->exit_code = exit_code;
1221 if (unlikely(exit_code == 0)) {
1223 * Another thread in this function got to it first, or it
1224 * resumed, or it resumed and then died.
1226 write_unlock_irq(&tasklist_lock);
1230 * We are returning to the wait loop without having successfully
1231 * removed the process and having released the lock. We cannot
1232 * continue, since the "p" task pointer is potentially stale.
1234 * Return -EAGAIN, and do_wait() will restart the loop from the
1235 * beginning. Do _not_ re-acquire the lock.
1240 /* move to end of parent's list to avoid starvation */
1242 add_parent(p, p->parent);
1244 write_unlock_irq(&tasklist_lock);
1246 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1247 if (!retval && stat_addr)
1248 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1249 if (!retval && infop)
1250 retval = put_user(SIGCHLD, &infop->si_signo);
1251 if (!retval && infop)
1252 retval = put_user(0, &infop->si_errno);
1253 if (!retval && infop)
1254 retval = put_user((short)((p->ptrace & PT_PTRACED)
1255 ? CLD_TRAPPED : CLD_STOPPED),
1257 if (!retval && infop)
1258 retval = put_user(exit_code, &infop->si_status);
1259 if (!retval && infop)
1260 retval = put_user(p->pid, &infop->si_pid);
1261 if (!retval && infop)
1262 retval = put_user(p->uid, &infop->si_uid);
1272 * Handle do_wait work for one task in a live, non-stopped state.
1273 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1274 * the lock and this task is uninteresting. If we return nonzero, we have
1275 * released the lock and the system call should return.
1277 static int wait_task_continued(task_t *p, int noreap,
1278 struct siginfo __user *infop,
1279 int __user *stat_addr, struct rusage __user *ru)
1285 if (unlikely(!p->signal))
1288 if (p->signal->stop_state >= 0)
1291 spin_lock_irq(&p->sighand->siglock);
1292 if (p->signal->stop_state >= 0) { /* Re-check with the lock held. */
1293 spin_unlock_irq(&p->sighand->siglock);
1297 p->signal->stop_state = 0;
1298 spin_unlock_irq(&p->sighand->siglock);
1303 read_unlock(&tasklist_lock);
1306 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1308 if (!retval && stat_addr)
1309 retval = put_user(0xffff, stat_addr);
1313 retval = wait_noreap_copyout(p, pid, uid,
1314 CLD_CONTINUED, SIGCONT,
1316 BUG_ON(retval == 0);
1323 static inline int my_ptrace_child(struct task_struct *p)
1325 if (!(p->ptrace & PT_PTRACED))
1327 if (!(p->ptrace & PT_ATTACHED))
1330 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1331 * we are the attacher. If we are the real parent, this is a race
1332 * inside ptrace_attach. It is waiting for the tasklist_lock,
1333 * which we have to switch the parent links, but has already set
1334 * the flags in p->ptrace.
1336 return (p->parent != p->real_parent);
1339 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1340 int __user *stat_addr, struct rusage __user *ru)
1342 DECLARE_WAITQUEUE(wait, current);
1343 struct task_struct *tsk;
1346 add_wait_queue(¤t->wait_chldexit,&wait);
1349 * We will set this flag if we see any child that might later
1350 * match our criteria, even if we are not able to reap it yet.
1353 current->state = TASK_INTERRUPTIBLE;
1354 read_lock(&tasklist_lock);
1357 struct task_struct *p;
1358 struct list_head *_p;
1361 list_for_each(_p,&tsk->children) {
1362 p = list_entry(_p,struct task_struct,sibling);
1364 ret = eligible_child(pid, options, p);
1370 if (!my_ptrace_child(p))
1375 * It's stopped now, so it might later
1376 * continue, exit, or stop again.
1379 if (!(options & WUNTRACED) &&
1380 !my_ptrace_child(p))
1382 retval = wait_task_stopped(p, ret == 2,
1383 (options & WNOWAIT),
1386 if (retval == -EAGAIN)
1388 if (retval != 0) /* He released the lock. */
1393 if (p->exit_state == EXIT_DEAD)
1395 // case EXIT_ZOMBIE:
1396 if (p->exit_state == EXIT_ZOMBIE) {
1398 * Eligible but we cannot release
1402 goto check_continued;
1403 if (!likely(options & WEXITED))
1405 retval = wait_task_zombie(
1406 p, (options & WNOWAIT),
1407 infop, stat_addr, ru);
1408 /* He released the lock. */
1415 * It's running now, so it might later
1416 * exit, stop, or stop and then continue.
1419 if (!unlikely(options & WCONTINUED))
1422 retval = wait_task_continued(
1423 p, (options & WNOWAIT),
1424 infop, stat_addr, ru);
1425 if (retval != 0) /* He released the lock. */
1431 list_for_each(_p, &tsk->ptrace_children) {
1432 p = list_entry(_p, struct task_struct,
1434 if (!eligible_child(pid, options, p))
1440 if (options & __WNOTHREAD)
1442 tsk = next_thread(tsk);
1443 if (tsk->signal != current->signal)
1445 } while (tsk != current);
1447 read_unlock(&tasklist_lock);
1450 if (options & WNOHANG)
1452 retval = -ERESTARTSYS;
1453 if (signal_pending(current))
1460 current->state = TASK_RUNNING;
1461 remove_wait_queue(¤t->wait_chldexit,&wait);
1467 * For a WNOHANG return, clear out all the fields
1468 * we would set so the user can easily tell the
1472 retval = put_user(0, &infop->si_signo);
1474 retval = put_user(0, &infop->si_errno);
1476 retval = put_user(0, &infop->si_code);
1478 retval = put_user(0, &infop->si_pid);
1480 retval = put_user(0, &infop->si_uid);
1482 retval = put_user(0, &infop->si_status);
1488 asmlinkage long sys_waitid(int which, pid_t pid,
1489 struct siginfo __user *infop, int options,
1490 struct rusage __user *ru)
1494 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1496 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1516 ret = do_wait(pid, options, infop, NULL, ru);
1518 /* avoid REGPARM breakage on x86: */
1519 prevent_tail_call(ret);
1523 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1524 int options, struct rusage __user *ru)
1528 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1529 __WNOTHREAD|__WCLONE|__WALL))
1531 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1533 /* avoid REGPARM breakage on x86: */
1534 prevent_tail_call(ret);
1538 #ifdef __ARCH_WANT_SYS_WAITPID
1541 * sys_waitpid() remains for compatibility. waitpid() should be
1542 * implemented by calling sys_wait4() from libc.a.
1544 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1546 return sys_wait4(pid, stat_addr, options, NULL);