fed up with those stupid warnings
[mmotm.git] / kernel / exit.c
blob140475a5ff1e622dae5deccfedf6ca9a59f3cecd
1 /*
2 * linux/kernel/exit.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
7 #include <linux/mm.h>
8 #include <linux/slab.h>
9 #include <linux/interrupt.h>
10 #include <linux/module.h>
11 #include <linux/capability.h>
12 #include <linux/completion.h>
13 #include <linux/personality.h>
14 #include <linux/tty.h>
15 #include <linux/iocontext.h>
16 #include <linux/key.h>
17 #include <linux/security.h>
18 #include <linux/cpu.h>
19 #include <linux/acct.h>
20 #include <linux/tsacct_kern.h>
21 #include <linux/file.h>
22 #include <linux/fdtable.h>
23 #include <linux/binfmts.h>
24 #include <linux/nsproxy.h>
25 #include <linux/pid_namespace.h>
26 #include <linux/ptrace.h>
27 #include <linux/profile.h>
28 #include <linux/mount.h>
29 #include <linux/proc_fs.h>
30 #include <linux/kthread.h>
31 #include <linux/mempolicy.h>
32 #include <linux/taskstats_kern.h>
33 #include <linux/delayacct.h>
34 #include <linux/freezer.h>
35 #include <linux/cgroup.h>
36 #include <linux/syscalls.h>
37 #include <linux/signal.h>
38 #include <linux/posix-timers.h>
39 #include <linux/cn_proc.h>
40 #include <linux/mutex.h>
41 #include <linux/futex.h>
42 #include <linux/pipe_fs_i.h>
43 #include <linux/audit.h> /* for audit_free() */
44 #include <linux/resource.h>
45 #include <linux/blkdev.h>
46 #include <linux/task_io_accounting_ops.h>
47 #include <linux/tracehook.h>
48 #include <linux/fs_struct.h>
49 #include <linux/init_task.h>
50 #include <linux/perf_event.h>
51 #include <trace/events/sched.h>
53 #include <asm/uaccess.h>
54 #include <asm/unistd.h>
55 #include <asm/pgtable.h>
56 #include <asm/mmu_context.h>
57 #include "cred-internals.h"
59 static void exit_mm(struct task_struct * tsk);
61 static void __unhash_process(struct task_struct *p)
63 nr_threads--;
64 detach_pid(p, PIDTYPE_PID);
65 if (thread_group_leader(p)) {
66 detach_pid(p, PIDTYPE_PGID);
67 detach_pid(p, PIDTYPE_SID);
69 list_del_rcu(&p->tasks);
70 list_del_init(&p->sibling);
71 __get_cpu_var(process_counts)--;
73 list_del_rcu(&p->thread_group);
77 * This function expects the tasklist_lock write-locked.
79 static void __exit_signal(struct task_struct *tsk)
81 struct signal_struct *sig = tsk->signal;
82 struct sighand_struct *sighand;
84 BUG_ON(!sig);
85 BUG_ON(!atomic_read(&sig->count));
87 sighand = rcu_dereference(tsk->sighand);
88 spin_lock(&sighand->siglock);
90 posix_cpu_timers_exit(tsk);
91 if (atomic_dec_and_test(&sig->count))
92 posix_cpu_timers_exit_group(tsk);
93 else {
95 * If there is any task waiting for the group exit
96 * then notify it:
98 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count)
99 wake_up_process(sig->group_exit_task);
101 if (tsk == sig->curr_target)
102 sig->curr_target = next_thread(tsk);
104 * Accumulate here the counters for all threads but the
105 * group leader as they die, so they can be added into
106 * the process-wide totals when those are taken.
107 * The group leader stays around as a zombie as long
108 * as there are other threads. When it gets reaped,
109 * the exit.c code will add its counts into these totals.
110 * We won't ever get here for the group leader, since it
111 * will have been the last reference on the signal_struct.
113 sig->utime = cputime_add(sig->utime, task_utime(tsk));
114 sig->stime = cputime_add(sig->stime, task_stime(tsk));
115 sig->gtime = cputime_add(sig->gtime, task_gtime(tsk));
116 sig->min_flt += tsk->min_flt;
117 sig->maj_flt += tsk->maj_flt;
118 sig->nvcsw += tsk->nvcsw;
119 sig->nivcsw += tsk->nivcsw;
120 sig->inblock += task_io_get_inblock(tsk);
121 sig->oublock += task_io_get_oublock(tsk);
122 task_io_accounting_add(&sig->ioac, &tsk->ioac);
123 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
124 sig = NULL; /* Marker for below. */
127 __unhash_process(tsk);
130 * Do this under ->siglock, we can race with another thread
131 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
133 flush_sigqueue(&tsk->pending);
135 tsk->signal = NULL;
136 tsk->sighand = NULL;
137 spin_unlock(&sighand->siglock);
139 __cleanup_sighand(sighand);
140 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
141 if (sig) {
142 flush_sigqueue(&sig->shared_pending);
143 taskstats_tgid_free(sig);
145 * Make sure ->signal can't go away under rq->lock,
146 * see account_group_exec_runtime().
148 task_rq_unlock_wait(tsk);
149 __cleanup_signal(sig);
153 static void delayed_put_task_struct(struct rcu_head *rhp)
155 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
157 #ifdef CONFIG_PERF_EVENTS
158 WARN_ON_ONCE(tsk->perf_event_ctxp);
159 #endif
160 trace_sched_process_free(tsk);
161 put_task_struct(tsk);
165 void release_task(struct task_struct * p)
167 struct task_struct *leader;
168 int zap_leader;
169 repeat:
170 tracehook_prepare_release_task(p);
171 /* don't need to get the RCU readlock here - the process is dead and
172 * can't be modifying its own credentials */
173 atomic_dec(&__task_cred(p)->user->processes);
175 proc_flush_task(p);
177 write_lock_irq(&tasklist_lock);
178 tracehook_finish_release_task(p);
179 __exit_signal(p);
182 * If we are the last non-leader member of the thread
183 * group, and the leader is zombie, then notify the
184 * group leader's parent process. (if it wants notification.)
186 zap_leader = 0;
187 leader = p->group_leader;
188 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
189 BUG_ON(task_detached(leader));
190 do_notify_parent(leader, leader->exit_signal);
192 * If we were the last child thread and the leader has
193 * exited already, and the leader's parent ignores SIGCHLD,
194 * then we are the one who should release the leader.
196 * do_notify_parent() will have marked it self-reaping in
197 * that case.
199 zap_leader = task_detached(leader);
202 * This maintains the invariant that release_task()
203 * only runs on a task in EXIT_DEAD, just for sanity.
205 if (zap_leader)
206 leader->exit_state = EXIT_DEAD;
209 write_unlock_irq(&tasklist_lock);
210 release_thread(p);
211 call_rcu(&p->rcu, delayed_put_task_struct);
213 p = leader;
214 if (unlikely(zap_leader))
215 goto repeat;
219 * This checks not only the pgrp, but falls back on the pid if no
220 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
221 * without this...
223 * The caller must hold rcu lock or the tasklist lock.
225 struct pid *session_of_pgrp(struct pid *pgrp)
227 struct task_struct *p;
228 struct pid *sid = NULL;
230 p = pid_task(pgrp, PIDTYPE_PGID);
231 if (p == NULL)
232 p = pid_task(pgrp, PIDTYPE_PID);
233 if (p != NULL)
234 sid = task_session(p);
236 return sid;
240 * Determine if a process group is "orphaned", according to the POSIX
241 * definition in 2.2.2.52. Orphaned process groups are not to be affected
242 * by terminal-generated stop signals. Newly orphaned process groups are
243 * to receive a SIGHUP and a SIGCONT.
245 * "I ask you, have you ever known what it is to be an orphan?"
247 static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
249 struct task_struct *p;
251 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
252 if ((p == ignored_task) ||
253 (p->exit_state && thread_group_empty(p)) ||
254 is_global_init(p->real_parent))
255 continue;
257 if (task_pgrp(p->real_parent) != pgrp &&
258 task_session(p->real_parent) == task_session(p))
259 return 0;
260 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
262 return 1;
265 int is_current_pgrp_orphaned(void)
267 int retval;
269 read_lock(&tasklist_lock);
270 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
271 read_unlock(&tasklist_lock);
273 return retval;
276 static int has_stopped_jobs(struct pid *pgrp)
278 int retval = 0;
279 struct task_struct *p;
281 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
282 if (!task_is_stopped(p))
283 continue;
284 retval = 1;
285 break;
286 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
287 return retval;
291 * Check to see if any process groups have become orphaned as
292 * a result of our exiting, and if they have any stopped jobs,
293 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
295 static void
296 kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
298 struct pid *pgrp = task_pgrp(tsk);
299 struct task_struct *ignored_task = tsk;
301 if (!parent)
302 /* exit: our father is in a different pgrp than
303 * we are and we were the only connection outside.
305 parent = tsk->real_parent;
306 else
307 /* reparent: our child is in a different pgrp than
308 * we are, and it was the only connection outside.
310 ignored_task = NULL;
312 if (task_pgrp(parent) != pgrp &&
313 task_session(parent) == task_session(tsk) &&
314 will_become_orphaned_pgrp(pgrp, ignored_task) &&
315 has_stopped_jobs(pgrp)) {
316 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
317 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
322 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
324 * If a kernel thread is launched as a result of a system call, or if
325 * it ever exits, it should generally reparent itself to kthreadd so it
326 * isn't in the way of other processes and is correctly cleaned up on exit.
328 * The various task state such as scheduling policy and priority may have
329 * been inherited from a user process, so we reset them to sane values here.
331 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
333 static void reparent_to_kthreadd(void)
335 write_lock_irq(&tasklist_lock);
337 ptrace_unlink(current);
338 /* Reparent to init */
339 current->real_parent = current->parent = kthreadd_task;
340 list_move_tail(&current->sibling, &current->real_parent->children);
342 /* Set the exit signal to SIGCHLD so we signal init on exit */
343 current->exit_signal = SIGCHLD;
345 if (task_nice(current) < 0)
346 set_user_nice(current, 0);
347 /* cpus_allowed? */
348 /* rt_priority? */
349 /* signals? */
350 memcpy(current->signal->rlim, init_task.signal->rlim,
351 sizeof(current->signal->rlim));
353 atomic_inc(&init_cred.usage);
354 commit_creds(&init_cred);
355 write_unlock_irq(&tasklist_lock);
358 void __set_special_pids(struct pid *pid)
360 struct task_struct *curr = current->group_leader;
362 if (task_session(curr) != pid)
363 change_pid(curr, PIDTYPE_SID, pid);
365 if (task_pgrp(curr) != pid)
366 change_pid(curr, PIDTYPE_PGID, pid);
369 static void set_special_pids(struct pid *pid)
371 write_lock_irq(&tasklist_lock);
372 __set_special_pids(pid);
373 write_unlock_irq(&tasklist_lock);
377 * Let kernel threads use this to say that they allow a certain signal.
378 * Must not be used if kthread was cloned with CLONE_SIGHAND.
380 int allow_signal(int sig)
382 if (!valid_signal(sig) || sig < 1)
383 return -EINVAL;
385 spin_lock_irq(&current->sighand->siglock);
386 /* This is only needed for daemonize()'ed kthreads */
387 sigdelset(&current->blocked, sig);
389 * Kernel threads handle their own signals. Let the signal code
390 * know it'll be handled, so that they don't get converted to
391 * SIGKILL or just silently dropped.
393 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
394 recalc_sigpending();
395 spin_unlock_irq(&current->sighand->siglock);
396 return 0;
399 EXPORT_SYMBOL(allow_signal);
401 int disallow_signal(int sig)
403 if (!valid_signal(sig) || sig < 1)
404 return -EINVAL;
406 spin_lock_irq(&current->sighand->siglock);
407 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
408 recalc_sigpending();
409 spin_unlock_irq(&current->sighand->siglock);
410 return 0;
413 EXPORT_SYMBOL(disallow_signal);
416 * Put all the gunge required to become a kernel thread without
417 * attached user resources in one place where it belongs.
420 void daemonize(const char *name, ...)
422 va_list args;
423 sigset_t blocked;
425 va_start(args, name);
426 vsnprintf(current->comm, sizeof(current->comm), name, args);
427 va_end(args);
430 * If we were started as result of loading a module, close all of the
431 * user space pages. We don't need them, and if we didn't close them
432 * they would be locked into memory.
434 exit_mm(current);
436 * We don't want to have TIF_FREEZE set if the system-wide hibernation
437 * or suspend transition begins right now.
439 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
441 if (current->nsproxy != &init_nsproxy) {
442 get_nsproxy(&init_nsproxy);
443 switch_task_namespaces(current, &init_nsproxy);
445 set_special_pids(&init_struct_pid);
446 proc_clear_tty(current);
448 /* Block and flush all signals */
449 sigfillset(&blocked);
450 sigprocmask(SIG_BLOCK, &blocked, NULL);
451 flush_signals(current);
453 /* Become as one with the init task */
455 daemonize_fs_struct();
456 exit_files(current);
457 current->files = init_task.files;
458 atomic_inc(&current->files->count);
460 reparent_to_kthreadd();
463 EXPORT_SYMBOL(daemonize);
465 static void close_files(struct files_struct * files)
467 int i, j;
468 struct fdtable *fdt;
470 j = 0;
473 * It is safe to dereference the fd table without RCU or
474 * ->file_lock because this is the last reference to the
475 * files structure.
477 fdt = files_fdtable(files);
478 for (;;) {
479 unsigned long set;
480 i = j * __NFDBITS;
481 if (i >= fdt->max_fds)
482 break;
483 set = fdt->open_fds->fds_bits[j++];
484 while (set) {
485 if (set & 1) {
486 struct file * file = xchg(&fdt->fd[i], NULL);
487 if (file) {
488 filp_close(file, files);
489 cond_resched();
492 i++;
493 set >>= 1;
498 struct files_struct *get_files_struct(struct task_struct *task)
500 struct files_struct *files;
502 task_lock(task);
503 files = task->files;
504 if (files)
505 atomic_inc(&files->count);
506 task_unlock(task);
508 return files;
511 void put_files_struct(struct files_struct *files)
513 struct fdtable *fdt;
515 if (atomic_dec_and_test(&files->count)) {
516 close_files(files);
518 * Free the fd and fdset arrays if we expanded them.
519 * If the fdtable was embedded, pass files for freeing
520 * at the end of the RCU grace period. Otherwise,
521 * you can free files immediately.
523 fdt = files_fdtable(files);
524 if (fdt != &files->fdtab)
525 kmem_cache_free(files_cachep, files);
526 free_fdtable(fdt);
530 void reset_files_struct(struct files_struct *files)
532 struct task_struct *tsk = current;
533 struct files_struct *old;
535 old = tsk->files;
536 task_lock(tsk);
537 tsk->files = files;
538 task_unlock(tsk);
539 put_files_struct(old);
542 void exit_files(struct task_struct *tsk)
544 struct files_struct * files = tsk->files;
546 if (files) {
547 task_lock(tsk);
548 tsk->files = NULL;
549 task_unlock(tsk);
550 put_files_struct(files);
554 #ifdef CONFIG_MM_OWNER
556 * Task p is exiting and it owned mm, lets find a new owner for it
558 static inline int
559 mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
562 * If there are other users of the mm and the owner (us) is exiting
563 * we need to find a new owner to take on the responsibility.
565 if (atomic_read(&mm->mm_users) <= 1)
566 return 0;
567 if (mm->owner != p)
568 return 0;
569 return 1;
572 void mm_update_next_owner(struct mm_struct *mm)
574 struct task_struct *c, *g, *p = current;
576 retry:
577 if (!mm_need_new_owner(mm, p))
578 return;
580 read_lock(&tasklist_lock);
582 * Search in the children
584 list_for_each_entry(c, &p->children, sibling) {
585 if (c->mm == mm)
586 goto assign_new_owner;
590 * Search in the siblings
592 list_for_each_entry(c, &p->real_parent->children, sibling) {
593 if (c->mm == mm)
594 goto assign_new_owner;
598 * Search through everything else. We should not get
599 * here often
601 do_each_thread(g, c) {
602 if (c->mm == mm)
603 goto assign_new_owner;
604 } while_each_thread(g, c);
606 read_unlock(&tasklist_lock);
608 * We found no owner yet mm_users > 1: this implies that we are
609 * most likely racing with swapoff (try_to_unuse()) or /proc or
610 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
612 mm->owner = NULL;
613 return;
615 assign_new_owner:
616 BUG_ON(c == p);
617 get_task_struct(c);
619 * The task_lock protects c->mm from changing.
620 * We always want mm->owner->mm == mm
622 task_lock(c);
624 * Delay read_unlock() till we have the task_lock()
625 * to ensure that c does not slip away underneath us
627 read_unlock(&tasklist_lock);
628 if (c->mm != mm) {
629 task_unlock(c);
630 put_task_struct(c);
631 goto retry;
633 mm->owner = c;
634 task_unlock(c);
635 put_task_struct(c);
637 #endif /* CONFIG_MM_OWNER */
640 * Turn us into a lazy TLB process if we
641 * aren't already..
643 static void exit_mm(struct task_struct * tsk)
645 struct mm_struct *mm = tsk->mm;
646 struct core_state *core_state;
648 mm_release(tsk, mm);
649 if (!mm)
650 return;
652 * Serialize with any possible pending coredump.
653 * We must hold mmap_sem around checking core_state
654 * and clearing tsk->mm. The core-inducing thread
655 * will increment ->nr_threads for each thread in the
656 * group with ->mm != NULL.
658 down_read(&mm->mmap_sem);
659 core_state = mm->core_state;
660 if (core_state) {
661 struct core_thread self;
662 up_read(&mm->mmap_sem);
664 self.task = tsk;
665 self.next = xchg(&core_state->dumper.next, &self);
667 * Implies mb(), the result of xchg() must be visible
668 * to core_state->dumper.
670 if (atomic_dec_and_test(&core_state->nr_threads))
671 complete(&core_state->startup);
673 for (;;) {
674 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
675 if (!self.task) /* see coredump_finish() */
676 break;
677 schedule();
679 __set_task_state(tsk, TASK_RUNNING);
680 down_read(&mm->mmap_sem);
682 atomic_inc(&mm->mm_count);
683 BUG_ON(mm != tsk->active_mm);
684 /* more a memory barrier than a real lock */
685 task_lock(tsk);
686 tsk->mm = NULL;
687 up_read(&mm->mmap_sem);
688 enter_lazy_tlb(mm, current);
689 /* We don't want this task to be frozen prematurely */
690 clear_freeze_flag(tsk);
691 task_unlock(tsk);
692 mm_update_next_owner(mm);
693 mmput(mm);
697 * When we die, we re-parent all our children.
698 * Try to give them to another thread in our thread
699 * group, and if no such member exists, give it to
700 * the child reaper process (ie "init") in our pid
701 * space.
703 static struct task_struct *find_new_reaper(struct task_struct *father)
705 struct pid_namespace *pid_ns = task_active_pid_ns(father);
706 struct task_struct *thread;
708 thread = father;
709 while_each_thread(father, thread) {
710 if (thread->flags & PF_EXITING)
711 continue;
712 if (unlikely(pid_ns->child_reaper == father))
713 pid_ns->child_reaper = thread;
714 return thread;
717 if (unlikely(pid_ns->child_reaper == father)) {
718 write_unlock_irq(&tasklist_lock);
719 if (unlikely(pid_ns == &init_pid_ns))
720 panic("Attempted to kill init!");
722 zap_pid_ns_processes(pid_ns);
723 write_lock_irq(&tasklist_lock);
725 * We can not clear ->child_reaper or leave it alone.
726 * There may by stealth EXIT_DEAD tasks on ->children,
727 * forget_original_parent() must move them somewhere.
729 pid_ns->child_reaper = init_pid_ns.child_reaper;
732 return pid_ns->child_reaper;
736 * Any that need to be release_task'd are put on the @dead list.
738 static void reparent_leader(struct task_struct *father, struct task_struct *p,
739 struct list_head *dead)
741 list_move_tail(&p->sibling, &p->real_parent->children);
743 if (task_detached(p))
744 return;
746 * If this is a threaded reparent there is no need to
747 * notify anyone anything has happened.
749 if (same_thread_group(p->real_parent, father))
750 return;
752 /* We don't want people slaying init. */
753 p->exit_signal = SIGCHLD;
755 /* If it has exited notify the new parent about this child's death. */
756 if (!task_ptrace(p) &&
757 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
758 do_notify_parent(p, p->exit_signal);
759 if (task_detached(p)) {
760 p->exit_state = EXIT_DEAD;
761 list_move_tail(&p->sibling, dead);
765 kill_orphaned_pgrp(p, father);
768 static void forget_original_parent(struct task_struct *father)
770 struct task_struct *p, *n, *reaper;
771 LIST_HEAD(dead_children);
773 exit_ptrace(father);
775 write_lock_irq(&tasklist_lock);
776 reaper = find_new_reaper(father);
778 list_for_each_entry_safe(p, n, &father->children, sibling) {
779 struct task_struct *t = p;
780 do {
781 t->real_parent = reaper;
782 if (t->parent == father) {
783 BUG_ON(task_ptrace(t));
784 t->parent = t->real_parent;
786 if (t->pdeath_signal)
787 group_send_sig_info(t->pdeath_signal,
788 SEND_SIG_NOINFO, t);
789 } while_each_thread(p, t);
790 reparent_leader(father, p, &dead_children);
792 write_unlock_irq(&tasklist_lock);
794 BUG_ON(!list_empty(&father->children));
796 list_for_each_entry_safe(p, n, &dead_children, sibling) {
797 list_del_init(&p->sibling);
798 release_task(p);
803 * Send signals to all our closest relatives so that they know
804 * to properly mourn us..
806 static void exit_notify(struct task_struct *tsk, int group_dead)
808 int signal;
809 void *cookie;
812 * This does two things:
814 * A. Make init inherit all the child processes
815 * B. Check to see if any process groups have become orphaned
816 * as a result of our exiting, and if they have any stopped
817 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
819 forget_original_parent(tsk);
820 exit_task_namespaces(tsk);
822 write_lock_irq(&tasklist_lock);
823 if (group_dead)
824 kill_orphaned_pgrp(tsk->group_leader, NULL);
826 /* Let father know we died
828 * Thread signals are configurable, but you aren't going to use
829 * that to send signals to arbitary processes.
830 * That stops right now.
832 * If the parent exec id doesn't match the exec id we saved
833 * when we started then we know the parent has changed security
834 * domain.
836 * If our self_exec id doesn't match our parent_exec_id then
837 * we have changed execution domain as these two values started
838 * the same after a fork.
840 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
841 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
842 tsk->self_exec_id != tsk->parent_exec_id))
843 tsk->exit_signal = SIGCHLD;
845 signal = tracehook_notify_death(tsk, &cookie, group_dead);
846 if (signal >= 0)
847 signal = do_notify_parent(tsk, signal);
849 tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
851 /* mt-exec, de_thread() is waiting for us */
852 if (thread_group_leader(tsk) &&
853 tsk->signal->group_exit_task &&
854 tsk->signal->notify_count < 0)
855 wake_up_process(tsk->signal->group_exit_task);
857 write_unlock_irq(&tasklist_lock);
859 tracehook_report_death(tsk, signal, cookie, group_dead);
861 /* If the process is dead, release it - nobody will wait for it */
862 if (signal == DEATH_REAP)
863 release_task(tsk);
866 #ifdef CONFIG_DEBUG_STACK_USAGE
867 static void check_stack_usage(void)
869 static DEFINE_SPINLOCK(low_water_lock);
870 static int lowest_to_date = THREAD_SIZE;
871 unsigned long free;
873 free = stack_not_used(current);
875 if (free >= lowest_to_date)
876 return;
878 spin_lock(&low_water_lock);
879 if (free < lowest_to_date) {
880 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
881 "left\n",
882 current->comm, free);
883 lowest_to_date = free;
885 spin_unlock(&low_water_lock);
887 #else
888 static inline void check_stack_usage(void) {}
889 #endif
891 NORET_TYPE void do_exit(long code)
893 struct task_struct *tsk = current;
894 int group_dead;
896 profile_task_exit(tsk);
898 WARN_ON(atomic_read(&tsk->fs_excl));
900 if (unlikely(in_interrupt()))
901 panic("Aiee, killing interrupt handler!");
902 if (unlikely(!tsk->pid))
903 panic("Attempted to kill the idle task!");
905 tracehook_report_exit(&code);
907 validate_creds_for_do_exit(tsk);
910 * We're taking recursive faults here in do_exit. Safest is to just
911 * leave this task alone and wait for reboot.
913 if (unlikely(tsk->flags & PF_EXITING)) {
914 printk(KERN_ALERT
915 "Fixing recursive fault but reboot is needed!\n");
917 * We can do this unlocked here. The futex code uses
918 * this flag just to verify whether the pi state
919 * cleanup has been done or not. In the worst case it
920 * loops once more. We pretend that the cleanup was
921 * done as there is no way to return. Either the
922 * OWNER_DIED bit is set by now or we push the blocked
923 * task into the wait for ever nirwana as well.
925 tsk->flags |= PF_EXITPIDONE;
926 set_current_state(TASK_UNINTERRUPTIBLE);
927 schedule();
930 exit_irq_thread();
932 exit_signals(tsk); /* sets PF_EXITING */
934 * tsk->flags are checked in the futex code to protect against
935 * an exiting task cleaning up the robust pi futexes.
937 smp_mb();
938 spin_unlock_wait(&tsk->pi_lock);
940 if (unlikely(in_atomic()))
941 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
942 current->comm, task_pid_nr(current),
943 preempt_count());
945 acct_update_integrals(tsk);
947 group_dead = atomic_dec_and_test(&tsk->signal->live);
948 if (group_dead) {
949 hrtimer_cancel(&tsk->signal->real_timer);
950 exit_itimers(tsk->signal);
951 if (tsk->mm)
952 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
954 acct_collect(code, group_dead);
955 if (group_dead)
956 tty_audit_exit();
957 if (unlikely(tsk->audit_context))
958 audit_free(tsk);
960 tsk->exit_code = code;
961 taskstats_exit(tsk, group_dead);
963 exit_mm(tsk);
965 if (group_dead)
966 acct_process();
967 trace_sched_process_exit(tsk);
969 exit_sem(tsk);
970 exit_files(tsk);
971 exit_fs(tsk);
972 check_stack_usage();
973 exit_thread();
974 cgroup_exit(tsk, 1);
976 if (group_dead && tsk->signal->leader)
977 disassociate_ctty(1);
979 module_put(task_thread_info(tsk)->exec_domain->module);
981 proc_exit_connector(tsk);
984 * Flush inherited counters to the parent - before the parent
985 * gets woken up by child-exit notifications.
987 perf_event_exit_task(tsk);
989 exit_notify(tsk, group_dead);
990 #ifdef CONFIG_NUMA
991 mpol_put(tsk->mempolicy);
992 tsk->mempolicy = NULL;
993 #endif
994 #ifdef CONFIG_FUTEX
995 if (unlikely(current->pi_state_cache))
996 kfree(current->pi_state_cache);
997 #endif
999 * Make sure we are holding no locks:
1001 debug_check_no_locks_held(tsk);
1003 * We can do this unlocked here. The futex code uses this flag
1004 * just to verify whether the pi state cleanup has been done
1005 * or not. In the worst case it loops once more.
1007 tsk->flags |= PF_EXITPIDONE;
1009 if (tsk->io_context)
1010 exit_io_context();
1012 if (tsk->splice_pipe)
1013 __free_pipe_info(tsk->splice_pipe);
1015 validate_creds_for_do_exit(tsk);
1017 preempt_disable();
1018 exit_rcu();
1019 /* causes final put_task_struct in finish_task_switch(). */
1020 tsk->state = TASK_DEAD;
1021 schedule();
1022 BUG();
1023 /* Avoid "noreturn function does return". */
1024 for (;;)
1025 cpu_relax(); /* For when BUG is null */
1028 EXPORT_SYMBOL_GPL(do_exit);
1030 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1032 if (comp)
1033 complete(comp);
1035 do_exit(code);
1038 EXPORT_SYMBOL(complete_and_exit);
1040 SYSCALL_DEFINE1(exit, int, error_code)
1042 do_exit((error_code&0xff)<<8);
1046 * Take down every thread in the group. This is called by fatal signals
1047 * as well as by sys_exit_group (below).
1049 NORET_TYPE void
1050 do_group_exit(int exit_code)
1052 struct signal_struct *sig = current->signal;
1054 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1056 if (signal_group_exit(sig))
1057 exit_code = sig->group_exit_code;
1058 else if (!thread_group_empty(current)) {
1059 struct sighand_struct *const sighand = current->sighand;
1060 spin_lock_irq(&sighand->siglock);
1061 if (signal_group_exit(sig))
1062 /* Another thread got here before we took the lock. */
1063 exit_code = sig->group_exit_code;
1064 else {
1065 sig->group_exit_code = exit_code;
1066 sig->flags = SIGNAL_GROUP_EXIT;
1067 zap_other_threads(current);
1069 spin_unlock_irq(&sighand->siglock);
1072 do_exit(exit_code);
1073 /* NOTREACHED */
1077 * this kills every thread in the thread group. Note that any externally
1078 * wait4()-ing process will get the correct exit code - even if this
1079 * thread is not the thread group leader.
1081 SYSCALL_DEFINE1(exit_group, int, error_code)
1083 do_group_exit((error_code & 0xff) << 8);
1084 /* NOTREACHED */
1085 return 0;
1088 struct wait_opts {
1089 enum pid_type wo_type;
1090 int wo_flags;
1091 struct pid *wo_pid;
1093 struct siginfo __user *wo_info;
1094 int __user *wo_stat;
1095 struct rusage __user *wo_rusage;
1097 wait_queue_t child_wait;
1098 int notask_error;
1101 static inline
1102 struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1104 if (type != PIDTYPE_PID)
1105 task = task->group_leader;
1106 return task->pids[type].pid;
1109 static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1111 return wo->wo_type == PIDTYPE_MAX ||
1112 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1115 static int eligible_child(struct wait_opts *wo, struct task_struct *p)
1117 if (!eligible_pid(wo, p))
1118 return 0;
1119 /* Wait for all children (clone and not) if __WALL is set;
1120 * otherwise, wait for clone children *only* if __WCLONE is
1121 * set; otherwise, wait for non-clone children *only*. (Note:
1122 * A "clone" child here is one that reports to its parent
1123 * using a signal other than SIGCHLD.) */
1124 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1125 && !(wo->wo_flags & __WALL))
1126 return 0;
1128 return 1;
1131 static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1132 pid_t pid, uid_t uid, int why, int status)
1134 struct siginfo __user *infop;
1135 int retval = wo->wo_rusage
1136 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1138 put_task_struct(p);
1139 infop = wo->wo_info;
1140 if (infop) {
1141 if (!retval)
1142 retval = put_user(SIGCHLD, &infop->si_signo);
1143 if (!retval)
1144 retval = put_user(0, &infop->si_errno);
1145 if (!retval)
1146 retval = put_user((short)why, &infop->si_code);
1147 if (!retval)
1148 retval = put_user(pid, &infop->si_pid);
1149 if (!retval)
1150 retval = put_user(uid, &infop->si_uid);
1151 if (!retval)
1152 retval = put_user(status, &infop->si_status);
1154 if (!retval)
1155 retval = pid;
1156 return retval;
1160 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. 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_zombie(struct wait_opts *wo, struct task_struct *p)
1167 unsigned long state;
1168 int retval, status, traced;
1169 pid_t pid = task_pid_vnr(p);
1170 uid_t uid = __task_cred(p)->uid;
1171 struct siginfo __user *infop;
1173 if (!likely(wo->wo_flags & WEXITED))
1174 return 0;
1176 if (unlikely(wo->wo_flags & WNOWAIT)) {
1177 int exit_code = p->exit_code;
1178 int why, status;
1180 get_task_struct(p);
1181 read_unlock(&tasklist_lock);
1182 if ((exit_code & 0x7f) == 0) {
1183 why = CLD_EXITED;
1184 status = exit_code >> 8;
1185 } else {
1186 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1187 status = exit_code & 0x7f;
1189 return wait_noreap_copyout(wo, p, pid, uid, why, status);
1193 * Try to move the task's state to DEAD
1194 * only one thread is allowed to do this:
1196 state = xchg(&p->exit_state, EXIT_DEAD);
1197 if (state != EXIT_ZOMBIE) {
1198 BUG_ON(state != EXIT_DEAD);
1199 return 0;
1202 traced = ptrace_reparented(p);
1204 * It can be ptraced but not reparented, check
1205 * !task_detached() to filter out sub-threads.
1207 if (likely(!traced) && likely(!task_detached(p))) {
1208 struct signal_struct *psig;
1209 struct signal_struct *sig;
1210 unsigned long maxrss;
1213 * The resource counters for the group leader are in its
1214 * own task_struct. Those for dead threads in the group
1215 * are in its signal_struct, as are those for the child
1216 * processes it has previously reaped. All these
1217 * accumulate in the parent's signal_struct c* fields.
1219 * We don't bother to take a lock here to protect these
1220 * p->signal fields, because they are only touched by
1221 * __exit_signal, which runs with tasklist_lock
1222 * write-locked anyway, and so is excluded here. We do
1223 * need to protect the access to parent->signal fields,
1224 * as other threads in the parent group can be right
1225 * here reaping other children at the same time.
1227 spin_lock_irq(&p->real_parent->sighand->siglock);
1228 psig = p->real_parent->signal;
1229 sig = p->signal;
1230 psig->cutime =
1231 cputime_add(psig->cutime,
1232 cputime_add(p->utime,
1233 cputime_add(sig->utime,
1234 sig->cutime)));
1235 psig->cstime =
1236 cputime_add(psig->cstime,
1237 cputime_add(p->stime,
1238 cputime_add(sig->stime,
1239 sig->cstime)));
1240 psig->cgtime =
1241 cputime_add(psig->cgtime,
1242 cputime_add(p->gtime,
1243 cputime_add(sig->gtime,
1244 sig->cgtime)));
1245 psig->cmin_flt +=
1246 p->min_flt + sig->min_flt + sig->cmin_flt;
1247 psig->cmaj_flt +=
1248 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1249 psig->cnvcsw +=
1250 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1251 psig->cnivcsw +=
1252 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1253 psig->cinblock +=
1254 task_io_get_inblock(p) +
1255 sig->inblock + sig->cinblock;
1256 psig->coublock +=
1257 task_io_get_oublock(p) +
1258 sig->oublock + sig->coublock;
1259 maxrss = max(sig->maxrss, sig->cmaxrss);
1260 if (psig->cmaxrss < maxrss)
1261 psig->cmaxrss = maxrss;
1262 task_io_accounting_add(&psig->ioac, &p->ioac);
1263 task_io_accounting_add(&psig->ioac, &sig->ioac);
1264 spin_unlock_irq(&p->real_parent->sighand->siglock);
1268 * Now we are sure this task is interesting, and no other
1269 * thread can reap it because we set its state to EXIT_DEAD.
1271 read_unlock(&tasklist_lock);
1273 retval = wo->wo_rusage
1274 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1275 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1276 ? p->signal->group_exit_code : p->exit_code;
1277 if (!retval && wo->wo_stat)
1278 retval = put_user(status, wo->wo_stat);
1280 infop = wo->wo_info;
1281 if (!retval && infop)
1282 retval = put_user(SIGCHLD, &infop->si_signo);
1283 if (!retval && infop)
1284 retval = put_user(0, &infop->si_errno);
1285 if (!retval && infop) {
1286 int why;
1288 if ((status & 0x7f) == 0) {
1289 why = CLD_EXITED;
1290 status >>= 8;
1291 } else {
1292 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1293 status &= 0x7f;
1295 retval = put_user((short)why, &infop->si_code);
1296 if (!retval)
1297 retval = put_user(status, &infop->si_status);
1299 if (!retval && infop)
1300 retval = put_user(pid, &infop->si_pid);
1301 if (!retval && infop)
1302 retval = put_user(uid, &infop->si_uid);
1303 if (!retval)
1304 retval = pid;
1306 if (traced) {
1307 write_lock_irq(&tasklist_lock);
1308 /* We dropped tasklist, ptracer could die and untrace */
1309 ptrace_unlink(p);
1311 * If this is not a detached task, notify the parent.
1312 * If it's still not detached after that, don't release
1313 * it now.
1315 if (!task_detached(p)) {
1316 do_notify_parent(p, p->exit_signal);
1317 if (!task_detached(p)) {
1318 p->exit_state = EXIT_ZOMBIE;
1319 p = NULL;
1322 write_unlock_irq(&tasklist_lock);
1324 if (p != NULL)
1325 release_task(p);
1327 return retval;
1330 static int *task_stopped_code(struct task_struct *p, bool ptrace)
1332 if (ptrace) {
1333 if (task_is_stopped_or_traced(p))
1334 return &p->exit_code;
1335 } else {
1336 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1337 return &p->signal->group_exit_code;
1339 return NULL;
1343 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1344 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1345 * the lock and this task is uninteresting. If we return nonzero, we have
1346 * released the lock and the system call should return.
1348 static int wait_task_stopped(struct wait_opts *wo,
1349 int ptrace, struct task_struct *p)
1351 struct siginfo __user *infop;
1352 int retval, exit_code, *p_code, why;
1353 uid_t uid = 0; /* unneeded, required by compiler */
1354 pid_t pid;
1357 * Traditionally we see ptrace'd stopped tasks regardless of options.
1359 if (!ptrace && !(wo->wo_flags & WUNTRACED))
1360 return 0;
1362 exit_code = 0;
1363 spin_lock_irq(&p->sighand->siglock);
1365 p_code = task_stopped_code(p, ptrace);
1366 if (unlikely(!p_code))
1367 goto unlock_sig;
1369 exit_code = *p_code;
1370 if (!exit_code)
1371 goto unlock_sig;
1373 if (!unlikely(wo->wo_flags & WNOWAIT))
1374 *p_code = 0;
1376 /* don't need the RCU readlock here as we're holding a spinlock */
1377 uid = __task_cred(p)->uid;
1378 unlock_sig:
1379 spin_unlock_irq(&p->sighand->siglock);
1380 if (!exit_code)
1381 return 0;
1384 * Now we are pretty sure this task is interesting.
1385 * Make sure it doesn't get reaped out from under us while we
1386 * give up the lock and then examine it below. We don't want to
1387 * keep holding onto the tasklist_lock while we call getrusage and
1388 * possibly take page faults for user memory.
1390 get_task_struct(p);
1391 pid = task_pid_vnr(p);
1392 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1393 read_unlock(&tasklist_lock);
1395 if (unlikely(wo->wo_flags & WNOWAIT))
1396 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
1398 retval = wo->wo_rusage
1399 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1400 if (!retval && wo->wo_stat)
1401 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1403 infop = wo->wo_info;
1404 if (!retval && infop)
1405 retval = put_user(SIGCHLD, &infop->si_signo);
1406 if (!retval && infop)
1407 retval = put_user(0, &infop->si_errno);
1408 if (!retval && infop)
1409 retval = put_user((short)why, &infop->si_code);
1410 if (!retval && infop)
1411 retval = put_user(exit_code, &infop->si_status);
1412 if (!retval && infop)
1413 retval = put_user(pid, &infop->si_pid);
1414 if (!retval && infop)
1415 retval = put_user(uid, &infop->si_uid);
1416 if (!retval)
1417 retval = pid;
1418 put_task_struct(p);
1420 BUG_ON(!retval);
1421 return retval;
1425 * Handle do_wait work for one task in a live, non-stopped state.
1426 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1427 * the lock and this task is uninteresting. If we return nonzero, we have
1428 * released the lock and the system call should return.
1430 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1432 int retval;
1433 pid_t pid;
1434 uid_t uid;
1436 if (!unlikely(wo->wo_flags & WCONTINUED))
1437 return 0;
1439 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1440 return 0;
1442 spin_lock_irq(&p->sighand->siglock);
1443 /* Re-check with the lock held. */
1444 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1445 spin_unlock_irq(&p->sighand->siglock);
1446 return 0;
1448 if (!unlikely(wo->wo_flags & WNOWAIT))
1449 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1450 uid = __task_cred(p)->uid;
1451 spin_unlock_irq(&p->sighand->siglock);
1453 pid = task_pid_vnr(p);
1454 get_task_struct(p);
1455 read_unlock(&tasklist_lock);
1457 if (!wo->wo_info) {
1458 retval = wo->wo_rusage
1459 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1460 put_task_struct(p);
1461 if (!retval && wo->wo_stat)
1462 retval = put_user(0xffff, wo->wo_stat);
1463 if (!retval)
1464 retval = pid;
1465 } else {
1466 retval = wait_noreap_copyout(wo, p, pid, uid,
1467 CLD_CONTINUED, SIGCONT);
1468 BUG_ON(retval == 0);
1471 return retval;
1475 * Consider @p for a wait by @parent.
1477 * -ECHILD should be in ->notask_error before the first call.
1478 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1479 * Returns zero if the search for a child should continue;
1480 * then ->notask_error is 0 if @p is an eligible child,
1481 * or another error from security_task_wait(), or still -ECHILD.
1483 static int wait_consider_task(struct wait_opts *wo, int ptrace,
1484 struct task_struct *p)
1486 int ret = eligible_child(wo, p);
1487 if (!ret)
1488 return ret;
1490 ret = security_task_wait(p);
1491 if (unlikely(ret < 0)) {
1493 * If we have not yet seen any eligible child,
1494 * then let this error code replace -ECHILD.
1495 * A permission error will give the user a clue
1496 * to look for security policy problems, rather
1497 * than for mysterious wait bugs.
1499 if (wo->notask_error)
1500 wo->notask_error = ret;
1501 return 0;
1504 if (likely(!ptrace) && unlikely(task_ptrace(p))) {
1506 * This child is hidden by ptrace.
1507 * We aren't allowed to see it now, but eventually we will.
1509 wo->notask_error = 0;
1510 return 0;
1513 if (p->exit_state == EXIT_DEAD)
1514 return 0;
1517 * We don't reap group leaders with subthreads.
1519 if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
1520 return wait_task_zombie(wo, p);
1523 * It's stopped or running now, so it might
1524 * later continue, exit, or stop again.
1526 wo->notask_error = 0;
1528 if (task_stopped_code(p, ptrace))
1529 return wait_task_stopped(wo, ptrace, p);
1531 return wait_task_continued(wo, p);
1535 * Do the work of do_wait() for one thread in the group, @tsk.
1537 * -ECHILD should be in ->notask_error before the first call.
1538 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1539 * Returns zero if the search for a child should continue; then
1540 * ->notask_error is 0 if there were any eligible children,
1541 * or another error from security_task_wait(), or still -ECHILD.
1543 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1545 struct task_struct *p;
1547 list_for_each_entry(p, &tsk->children, sibling) {
1548 int ret = wait_consider_task(wo, 0, p);
1549 if (ret)
1550 return ret;
1553 return 0;
1556 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1558 struct task_struct *p;
1560 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1561 int ret = wait_consider_task(wo, 1, p);
1562 if (ret)
1563 return ret;
1566 return 0;
1569 static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1570 int sync, void *key)
1572 struct wait_opts *wo = container_of(wait, struct wait_opts,
1573 child_wait);
1574 struct task_struct *p = key;
1576 if (!eligible_pid(wo, p))
1577 return 0;
1579 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1580 return 0;
1582 return default_wake_function(wait, mode, sync, key);
1585 void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1587 __wake_up_sync_key(&parent->signal->wait_chldexit,
1588 TASK_INTERRUPTIBLE, 1, p);
1591 static long do_wait(struct wait_opts *wo)
1593 struct task_struct *tsk;
1594 int retval;
1596 trace_sched_process_wait(wo->wo_pid);
1598 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1599 wo->child_wait.private = current;
1600 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1601 repeat:
1603 * If there is nothing that can match our critiera just get out.
1604 * We will clear ->notask_error to zero if we see any child that
1605 * might later match our criteria, even if we are not able to reap
1606 * it yet.
1608 wo->notask_error = -ECHILD;
1609 if ((wo->wo_type < PIDTYPE_MAX) &&
1610 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
1611 goto notask;
1613 set_current_state(TASK_INTERRUPTIBLE);
1614 read_lock(&tasklist_lock);
1615 tsk = current;
1616 do {
1617 retval = do_wait_thread(wo, tsk);
1618 if (retval)
1619 goto end;
1621 retval = ptrace_do_wait(wo, tsk);
1622 if (retval)
1623 goto end;
1625 if (wo->wo_flags & __WNOTHREAD)
1626 break;
1627 } while_each_thread(current, tsk);
1628 read_unlock(&tasklist_lock);
1630 notask:
1631 retval = wo->notask_error;
1632 if (!retval && !(wo->wo_flags & WNOHANG)) {
1633 retval = -ERESTARTSYS;
1634 if (!signal_pending(current)) {
1635 schedule();
1636 goto repeat;
1639 end:
1640 __set_current_state(TASK_RUNNING);
1641 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1642 return retval;
1645 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1646 infop, int, options, struct rusage __user *, ru)
1648 struct wait_opts wo;
1649 struct pid *pid = NULL;
1650 enum pid_type type;
1651 long ret;
1653 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1654 return -EINVAL;
1655 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1656 return -EINVAL;
1658 switch (which) {
1659 case P_ALL:
1660 type = PIDTYPE_MAX;
1661 break;
1662 case P_PID:
1663 type = PIDTYPE_PID;
1664 if (upid <= 0)
1665 return -EINVAL;
1666 break;
1667 case P_PGID:
1668 type = PIDTYPE_PGID;
1669 if (upid <= 0)
1670 return -EINVAL;
1671 break;
1672 default:
1673 return -EINVAL;
1676 if (type < PIDTYPE_MAX)
1677 pid = find_get_pid(upid);
1679 wo.wo_type = type;
1680 wo.wo_pid = pid;
1681 wo.wo_flags = options;
1682 wo.wo_info = infop;
1683 wo.wo_stat = NULL;
1684 wo.wo_rusage = ru;
1685 ret = do_wait(&wo);
1687 if (ret > 0) {
1688 ret = 0;
1689 } else if (infop) {
1691 * For a WNOHANG return, clear out all the fields
1692 * we would set so the user can easily tell the
1693 * difference.
1695 if (!ret)
1696 ret = put_user(0, &infop->si_signo);
1697 if (!ret)
1698 ret = put_user(0, &infop->si_errno);
1699 if (!ret)
1700 ret = put_user(0, &infop->si_code);
1701 if (!ret)
1702 ret = put_user(0, &infop->si_pid);
1703 if (!ret)
1704 ret = put_user(0, &infop->si_uid);
1705 if (!ret)
1706 ret = put_user(0, &infop->si_status);
1709 put_pid(pid);
1711 /* avoid REGPARM breakage on x86: */
1712 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
1713 return ret;
1716 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1717 int, options, struct rusage __user *, ru)
1719 struct wait_opts wo;
1720 struct pid *pid = NULL;
1721 enum pid_type type;
1722 long ret;
1724 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1725 __WNOTHREAD|__WCLONE|__WALL))
1726 return -EINVAL;
1728 if (upid == -1)
1729 type = PIDTYPE_MAX;
1730 else if (upid < 0) {
1731 type = PIDTYPE_PGID;
1732 pid = find_get_pid(-upid);
1733 } else if (upid == 0) {
1734 type = PIDTYPE_PGID;
1735 pid = get_task_pid(current, PIDTYPE_PGID);
1736 } else /* upid > 0 */ {
1737 type = PIDTYPE_PID;
1738 pid = find_get_pid(upid);
1741 wo.wo_type = type;
1742 wo.wo_pid = pid;
1743 wo.wo_flags = options | WEXITED;
1744 wo.wo_info = NULL;
1745 wo.wo_stat = stat_addr;
1746 wo.wo_rusage = ru;
1747 ret = do_wait(&wo);
1748 put_pid(pid);
1750 /* avoid REGPARM breakage on x86: */
1751 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
1752 return ret;
1755 #ifdef __ARCH_WANT_SYS_WAITPID
1758 * sys_waitpid() remains for compatibility. waitpid() should be
1759 * implemented by calling sys_wait4() from libc.a.
1761 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1763 return sys_wait4(pid, stat_addr, options, NULL);
1766 #endif