USB: ipaq: fix oops when device is plugged in
[linux-2.6/mini2440.git] / kernel / exit.c
blobe61891f801238f3a386e9162f7bed8b456a4e268
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 __get_cpu_var(process_counts)--;
72 list_del_rcu(&p->thread_group);
73 list_del_init(&p->sibling);
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);
364 proc_sid_connector(curr);
367 if (task_pgrp(curr) != pid)
368 change_pid(curr, PIDTYPE_PGID, pid);
371 static void set_special_pids(struct pid *pid)
373 write_lock_irq(&tasklist_lock);
374 __set_special_pids(pid);
375 write_unlock_irq(&tasklist_lock);
379 * Let kernel threads use this to say that they allow a certain signal.
380 * Must not be used if kthread was cloned with CLONE_SIGHAND.
382 int allow_signal(int sig)
384 if (!valid_signal(sig) || sig < 1)
385 return -EINVAL;
387 spin_lock_irq(&current->sighand->siglock);
388 /* This is only needed for daemonize()'ed kthreads */
389 sigdelset(&current->blocked, sig);
391 * Kernel threads handle their own signals. Let the signal code
392 * know it'll be handled, so that they don't get converted to
393 * SIGKILL or just silently dropped.
395 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
396 recalc_sigpending();
397 spin_unlock_irq(&current->sighand->siglock);
398 return 0;
401 EXPORT_SYMBOL(allow_signal);
403 int disallow_signal(int sig)
405 if (!valid_signal(sig) || sig < 1)
406 return -EINVAL;
408 spin_lock_irq(&current->sighand->siglock);
409 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
410 recalc_sigpending();
411 spin_unlock_irq(&current->sighand->siglock);
412 return 0;
415 EXPORT_SYMBOL(disallow_signal);
418 * Put all the gunge required to become a kernel thread without
419 * attached user resources in one place where it belongs.
422 void daemonize(const char *name, ...)
424 va_list args;
425 sigset_t blocked;
427 va_start(args, name);
428 vsnprintf(current->comm, sizeof(current->comm), name, args);
429 va_end(args);
432 * If we were started as result of loading a module, close all of the
433 * user space pages. We don't need them, and if we didn't close them
434 * they would be locked into memory.
436 exit_mm(current);
438 * We don't want to have TIF_FREEZE set if the system-wide hibernation
439 * or suspend transition begins right now.
441 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
443 if (current->nsproxy != &init_nsproxy) {
444 get_nsproxy(&init_nsproxy);
445 switch_task_namespaces(current, &init_nsproxy);
447 set_special_pids(&init_struct_pid);
448 proc_clear_tty(current);
450 /* Block and flush all signals */
451 sigfillset(&blocked);
452 sigprocmask(SIG_BLOCK, &blocked, NULL);
453 flush_signals(current);
455 /* Become as one with the init task */
457 daemonize_fs_struct();
458 exit_files(current);
459 current->files = init_task.files;
460 atomic_inc(&current->files->count);
462 reparent_to_kthreadd();
465 EXPORT_SYMBOL(daemonize);
467 static void close_files(struct files_struct * files)
469 int i, j;
470 struct fdtable *fdt;
472 j = 0;
475 * It is safe to dereference the fd table without RCU or
476 * ->file_lock because this is the last reference to the
477 * files structure.
479 fdt = files_fdtable(files);
480 for (;;) {
481 unsigned long set;
482 i = j * __NFDBITS;
483 if (i >= fdt->max_fds)
484 break;
485 set = fdt->open_fds->fds_bits[j++];
486 while (set) {
487 if (set & 1) {
488 struct file * file = xchg(&fdt->fd[i], NULL);
489 if (file) {
490 filp_close(file, files);
491 cond_resched();
494 i++;
495 set >>= 1;
500 struct files_struct *get_files_struct(struct task_struct *task)
502 struct files_struct *files;
504 task_lock(task);
505 files = task->files;
506 if (files)
507 atomic_inc(&files->count);
508 task_unlock(task);
510 return files;
513 void put_files_struct(struct files_struct *files)
515 struct fdtable *fdt;
517 if (atomic_dec_and_test(&files->count)) {
518 close_files(files);
520 * Free the fd and fdset arrays if we expanded them.
521 * If the fdtable was embedded, pass files for freeing
522 * at the end of the RCU grace period. Otherwise,
523 * you can free files immediately.
525 fdt = files_fdtable(files);
526 if (fdt != &files->fdtab)
527 kmem_cache_free(files_cachep, files);
528 free_fdtable(fdt);
532 void reset_files_struct(struct files_struct *files)
534 struct task_struct *tsk = current;
535 struct files_struct *old;
537 old = tsk->files;
538 task_lock(tsk);
539 tsk->files = files;
540 task_unlock(tsk);
541 put_files_struct(old);
544 void exit_files(struct task_struct *tsk)
546 struct files_struct * files = tsk->files;
548 if (files) {
549 task_lock(tsk);
550 tsk->files = NULL;
551 task_unlock(tsk);
552 put_files_struct(files);
556 #ifdef CONFIG_MM_OWNER
558 * Task p is exiting and it owned mm, lets find a new owner for it
560 static inline int
561 mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
564 * If there are other users of the mm and the owner (us) is exiting
565 * we need to find a new owner to take on the responsibility.
567 if (atomic_read(&mm->mm_users) <= 1)
568 return 0;
569 if (mm->owner != p)
570 return 0;
571 return 1;
574 void mm_update_next_owner(struct mm_struct *mm)
576 struct task_struct *c, *g, *p = current;
578 retry:
579 if (!mm_need_new_owner(mm, p))
580 return;
582 read_lock(&tasklist_lock);
584 * Search in the children
586 list_for_each_entry(c, &p->children, sibling) {
587 if (c->mm == mm)
588 goto assign_new_owner;
592 * Search in the siblings
594 list_for_each_entry(c, &p->real_parent->children, sibling) {
595 if (c->mm == mm)
596 goto assign_new_owner;
600 * Search through everything else. We should not get
601 * here often
603 do_each_thread(g, c) {
604 if (c->mm == mm)
605 goto assign_new_owner;
606 } while_each_thread(g, c);
608 read_unlock(&tasklist_lock);
610 * We found no owner yet mm_users > 1: this implies that we are
611 * most likely racing with swapoff (try_to_unuse()) or /proc or
612 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
614 mm->owner = NULL;
615 return;
617 assign_new_owner:
618 BUG_ON(c == p);
619 get_task_struct(c);
621 * The task_lock protects c->mm from changing.
622 * We always want mm->owner->mm == mm
624 task_lock(c);
626 * Delay read_unlock() till we have the task_lock()
627 * to ensure that c does not slip away underneath us
629 read_unlock(&tasklist_lock);
630 if (c->mm != mm) {
631 task_unlock(c);
632 put_task_struct(c);
633 goto retry;
635 mm->owner = c;
636 task_unlock(c);
637 put_task_struct(c);
639 #endif /* CONFIG_MM_OWNER */
642 * Turn us into a lazy TLB process if we
643 * aren't already..
645 static void exit_mm(struct task_struct * tsk)
647 struct mm_struct *mm = tsk->mm;
648 struct core_state *core_state;
650 mm_release(tsk, mm);
651 if (!mm)
652 return;
654 * Serialize with any possible pending coredump.
655 * We must hold mmap_sem around checking core_state
656 * and clearing tsk->mm. The core-inducing thread
657 * will increment ->nr_threads for each thread in the
658 * group with ->mm != NULL.
660 down_read(&mm->mmap_sem);
661 core_state = mm->core_state;
662 if (core_state) {
663 struct core_thread self;
664 up_read(&mm->mmap_sem);
666 self.task = tsk;
667 self.next = xchg(&core_state->dumper.next, &self);
669 * Implies mb(), the result of xchg() must be visible
670 * to core_state->dumper.
672 if (atomic_dec_and_test(&core_state->nr_threads))
673 complete(&core_state->startup);
675 for (;;) {
676 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
677 if (!self.task) /* see coredump_finish() */
678 break;
679 schedule();
681 __set_task_state(tsk, TASK_RUNNING);
682 down_read(&mm->mmap_sem);
684 atomic_inc(&mm->mm_count);
685 BUG_ON(mm != tsk->active_mm);
686 /* more a memory barrier than a real lock */
687 task_lock(tsk);
688 tsk->mm = NULL;
689 up_read(&mm->mmap_sem);
690 enter_lazy_tlb(mm, current);
691 /* We don't want this task to be frozen prematurely */
692 clear_freeze_flag(tsk);
693 task_unlock(tsk);
694 mm_update_next_owner(mm);
695 mmput(mm);
699 * When we die, we re-parent all our children.
700 * Try to give them to another thread in our thread
701 * group, and if no such member exists, give it to
702 * the child reaper process (ie "init") in our pid
703 * space.
705 static struct task_struct *find_new_reaper(struct task_struct *father)
707 struct pid_namespace *pid_ns = task_active_pid_ns(father);
708 struct task_struct *thread;
710 thread = father;
711 while_each_thread(father, thread) {
712 if (thread->flags & PF_EXITING)
713 continue;
714 if (unlikely(pid_ns->child_reaper == father))
715 pid_ns->child_reaper = thread;
716 return thread;
719 if (unlikely(pid_ns->child_reaper == father)) {
720 write_unlock_irq(&tasklist_lock);
721 if (unlikely(pid_ns == &init_pid_ns))
722 panic("Attempted to kill init!");
724 zap_pid_ns_processes(pid_ns);
725 write_lock_irq(&tasklist_lock);
727 * We can not clear ->child_reaper or leave it alone.
728 * There may by stealth EXIT_DEAD tasks on ->children,
729 * forget_original_parent() must move them somewhere.
731 pid_ns->child_reaper = init_pid_ns.child_reaper;
734 return pid_ns->child_reaper;
738 * Any that need to be release_task'd are put on the @dead list.
740 static void reparent_thread(struct task_struct *father, struct task_struct *p,
741 struct list_head *dead)
743 if (p->pdeath_signal)
744 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
746 list_move_tail(&p->sibling, &p->real_parent->children);
748 if (task_detached(p))
749 return;
751 * If this is a threaded reparent there is no need to
752 * notify anyone anything has happened.
754 if (same_thread_group(p->real_parent, father))
755 return;
757 /* We don't want people slaying init. */
758 p->exit_signal = SIGCHLD;
760 /* If it has exited notify the new parent about this child's death. */
761 if (!task_ptrace(p) &&
762 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
763 do_notify_parent(p, p->exit_signal);
764 if (task_detached(p)) {
765 p->exit_state = EXIT_DEAD;
766 list_move_tail(&p->sibling, dead);
770 kill_orphaned_pgrp(p, father);
773 static void forget_original_parent(struct task_struct *father)
775 struct task_struct *p, *n, *reaper;
776 LIST_HEAD(dead_children);
778 exit_ptrace(father);
780 write_lock_irq(&tasklist_lock);
781 reaper = find_new_reaper(father);
783 list_for_each_entry_safe(p, n, &father->children, sibling) {
784 p->real_parent = reaper;
785 if (p->parent == father) {
786 BUG_ON(task_ptrace(p));
787 p->parent = p->real_parent;
789 reparent_thread(father, p, &dead_children);
791 write_unlock_irq(&tasklist_lock);
793 BUG_ON(!list_empty(&father->children));
795 list_for_each_entry_safe(p, n, &dead_children, sibling) {
796 list_del_init(&p->sibling);
797 release_task(p);
802 * Send signals to all our closest relatives so that they know
803 * to properly mourn us..
805 static void exit_notify(struct task_struct *tsk, int group_dead)
807 int signal;
808 void *cookie;
811 * This does two things:
813 * A. Make init inherit all the child processes
814 * B. Check to see if any process groups have become orphaned
815 * as a result of our exiting, and if they have any stopped
816 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
818 forget_original_parent(tsk);
819 exit_task_namespaces(tsk);
821 write_lock_irq(&tasklist_lock);
822 if (group_dead)
823 kill_orphaned_pgrp(tsk->group_leader, NULL);
825 /* Let father know we died
827 * Thread signals are configurable, but you aren't going to use
828 * that to send signals to arbitary processes.
829 * That stops right now.
831 * If the parent exec id doesn't match the exec id we saved
832 * when we started then we know the parent has changed security
833 * domain.
835 * If our self_exec id doesn't match our parent_exec_id then
836 * we have changed execution domain as these two values started
837 * the same after a fork.
839 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
840 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
841 tsk->self_exec_id != tsk->parent_exec_id))
842 tsk->exit_signal = SIGCHLD;
844 signal = tracehook_notify_death(tsk, &cookie, group_dead);
845 if (signal >= 0)
846 signal = do_notify_parent(tsk, signal);
848 tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
850 /* mt-exec, de_thread() is waiting for us */
851 if (thread_group_leader(tsk) &&
852 tsk->signal->group_exit_task &&
853 tsk->signal->notify_count < 0)
854 wake_up_process(tsk->signal->group_exit_task);
856 write_unlock_irq(&tasklist_lock);
858 tracehook_report_death(tsk, signal, cookie, group_dead);
860 /* If the process is dead, release it - nobody will wait for it */
861 if (signal == DEATH_REAP)
862 release_task(tsk);
865 #ifdef CONFIG_DEBUG_STACK_USAGE
866 static void check_stack_usage(void)
868 static DEFINE_SPINLOCK(low_water_lock);
869 static int lowest_to_date = THREAD_SIZE;
870 unsigned long free;
872 free = stack_not_used(current);
874 if (free >= lowest_to_date)
875 return;
877 spin_lock(&low_water_lock);
878 if (free < lowest_to_date) {
879 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
880 "left\n",
881 current->comm, free);
882 lowest_to_date = free;
884 spin_unlock(&low_water_lock);
886 #else
887 static inline void check_stack_usage(void) {}
888 #endif
890 NORET_TYPE void do_exit(long code)
892 struct task_struct *tsk = current;
893 int group_dead;
895 profile_task_exit(tsk);
897 WARN_ON(atomic_read(&tsk->fs_excl));
899 if (unlikely(in_interrupt()))
900 panic("Aiee, killing interrupt handler!");
901 if (unlikely(!tsk->pid))
902 panic("Attempted to kill the idle task!");
904 tracehook_report_exit(&code);
906 validate_creds_for_do_exit(tsk);
909 * We're taking recursive faults here in do_exit. Safest is to just
910 * leave this task alone and wait for reboot.
912 if (unlikely(tsk->flags & PF_EXITING)) {
913 printk(KERN_ALERT
914 "Fixing recursive fault but reboot is needed!\n");
916 * We can do this unlocked here. The futex code uses
917 * this flag just to verify whether the pi state
918 * cleanup has been done or not. In the worst case it
919 * loops once more. We pretend that the cleanup was
920 * done as there is no way to return. Either the
921 * OWNER_DIED bit is set by now or we push the blocked
922 * task into the wait for ever nirwana as well.
924 tsk->flags |= PF_EXITPIDONE;
925 set_current_state(TASK_UNINTERRUPTIBLE);
926 schedule();
929 exit_irq_thread();
931 exit_signals(tsk); /* sets PF_EXITING */
933 * tsk->flags are checked in the futex code to protect against
934 * an exiting task cleaning up the robust pi futexes.
936 smp_mb();
937 spin_unlock_wait(&tsk->pi_lock);
939 if (unlikely(in_atomic()))
940 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
941 current->comm, task_pid_nr(current),
942 preempt_count());
944 acct_update_integrals(tsk);
946 group_dead = atomic_dec_and_test(&tsk->signal->live);
947 if (group_dead) {
948 hrtimer_cancel(&tsk->signal->real_timer);
949 exit_itimers(tsk->signal);
950 if (tsk->mm)
951 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
953 acct_collect(code, group_dead);
954 if (group_dead)
955 tty_audit_exit();
956 if (unlikely(tsk->audit_context))
957 audit_free(tsk);
959 tsk->exit_code = code;
960 taskstats_exit(tsk, group_dead);
962 exit_mm(tsk);
964 if (group_dead)
965 acct_process();
966 trace_sched_process_exit(tsk);
968 exit_sem(tsk);
969 exit_files(tsk);
970 exit_fs(tsk);
971 check_stack_usage();
972 exit_thread();
973 cgroup_exit(tsk, 1);
975 if (group_dead && tsk->signal->leader)
976 disassociate_ctty(1);
978 module_put(task_thread_info(tsk)->exec_domain->module);
980 proc_exit_connector(tsk);
983 * Flush inherited counters to the parent - before the parent
984 * gets woken up by child-exit notifications.
986 perf_event_exit_task(tsk);
988 exit_notify(tsk, group_dead);
989 #ifdef CONFIG_NUMA
990 mpol_put(tsk->mempolicy);
991 tsk->mempolicy = NULL;
992 #endif
993 #ifdef CONFIG_FUTEX
994 if (unlikely(current->pi_state_cache))
995 kfree(current->pi_state_cache);
996 #endif
998 * Make sure we are holding no locks:
1000 debug_check_no_locks_held(tsk);
1002 * We can do this unlocked here. The futex code uses this flag
1003 * just to verify whether the pi state cleanup has been done
1004 * or not. In the worst case it loops once more.
1006 tsk->flags |= PF_EXITPIDONE;
1008 if (tsk->io_context)
1009 exit_io_context();
1011 if (tsk->splice_pipe)
1012 __free_pipe_info(tsk->splice_pipe);
1014 validate_creds_for_do_exit(tsk);
1016 preempt_disable();
1017 exit_rcu();
1018 /* causes final put_task_struct in finish_task_switch(). */
1019 tsk->state = TASK_DEAD;
1020 schedule();
1021 BUG();
1022 /* Avoid "noreturn function does return". */
1023 for (;;)
1024 cpu_relax(); /* For when BUG is null */
1027 EXPORT_SYMBOL_GPL(do_exit);
1029 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1031 if (comp)
1032 complete(comp);
1034 do_exit(code);
1037 EXPORT_SYMBOL(complete_and_exit);
1039 SYSCALL_DEFINE1(exit, int, error_code)
1041 do_exit((error_code&0xff)<<8);
1045 * Take down every thread in the group. This is called by fatal signals
1046 * as well as by sys_exit_group (below).
1048 NORET_TYPE void
1049 do_group_exit(int exit_code)
1051 struct signal_struct *sig = current->signal;
1053 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1055 if (signal_group_exit(sig))
1056 exit_code = sig->group_exit_code;
1057 else if (!thread_group_empty(current)) {
1058 struct sighand_struct *const sighand = current->sighand;
1059 spin_lock_irq(&sighand->siglock);
1060 if (signal_group_exit(sig))
1061 /* Another thread got here before we took the lock. */
1062 exit_code = sig->group_exit_code;
1063 else {
1064 sig->group_exit_code = exit_code;
1065 sig->flags = SIGNAL_GROUP_EXIT;
1066 zap_other_threads(current);
1068 spin_unlock_irq(&sighand->siglock);
1071 do_exit(exit_code);
1072 /* NOTREACHED */
1076 * this kills every thread in the thread group. Note that any externally
1077 * wait4()-ing process will get the correct exit code - even if this
1078 * thread is not the thread group leader.
1080 SYSCALL_DEFINE1(exit_group, int, error_code)
1082 do_group_exit((error_code & 0xff) << 8);
1083 /* NOTREACHED */
1084 return 0;
1087 struct wait_opts {
1088 enum pid_type wo_type;
1089 int wo_flags;
1090 struct pid *wo_pid;
1092 struct siginfo __user *wo_info;
1093 int __user *wo_stat;
1094 struct rusage __user *wo_rusage;
1096 wait_queue_t child_wait;
1097 int notask_error;
1100 static inline
1101 struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1103 if (type != PIDTYPE_PID)
1104 task = task->group_leader;
1105 return task->pids[type].pid;
1108 static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1110 return wo->wo_type == PIDTYPE_MAX ||
1111 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1114 static int eligible_child(struct wait_opts *wo, struct task_struct *p)
1116 if (!eligible_pid(wo, p))
1117 return 0;
1118 /* Wait for all children (clone and not) if __WALL is set;
1119 * otherwise, wait for clone children *only* if __WCLONE is
1120 * set; otherwise, wait for non-clone children *only*. (Note:
1121 * A "clone" child here is one that reports to its parent
1122 * using a signal other than SIGCHLD.) */
1123 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1124 && !(wo->wo_flags & __WALL))
1125 return 0;
1127 return 1;
1130 static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1131 pid_t pid, uid_t uid, int why, int status)
1133 struct siginfo __user *infop;
1134 int retval = wo->wo_rusage
1135 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1137 put_task_struct(p);
1138 infop = wo->wo_info;
1139 if (infop) {
1140 if (!retval)
1141 retval = put_user(SIGCHLD, &infop->si_signo);
1142 if (!retval)
1143 retval = put_user(0, &infop->si_errno);
1144 if (!retval)
1145 retval = put_user((short)why, &infop->si_code);
1146 if (!retval)
1147 retval = put_user(pid, &infop->si_pid);
1148 if (!retval)
1149 retval = put_user(uid, &infop->si_uid);
1150 if (!retval)
1151 retval = put_user(status, &infop->si_status);
1153 if (!retval)
1154 retval = pid;
1155 return retval;
1159 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1160 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1161 * the lock and this task is uninteresting. If we return nonzero, we have
1162 * released the lock and the system call should return.
1164 static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1166 unsigned long state;
1167 int retval, status, traced;
1168 pid_t pid = task_pid_vnr(p);
1169 uid_t uid = __task_cred(p)->uid;
1170 struct siginfo __user *infop;
1172 if (!likely(wo->wo_flags & WEXITED))
1173 return 0;
1175 if (unlikely(wo->wo_flags & WNOWAIT)) {
1176 int exit_code = p->exit_code;
1177 int why, status;
1179 get_task_struct(p);
1180 read_unlock(&tasklist_lock);
1181 if ((exit_code & 0x7f) == 0) {
1182 why = CLD_EXITED;
1183 status = exit_code >> 8;
1184 } else {
1185 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1186 status = exit_code & 0x7f;
1188 return wait_noreap_copyout(wo, p, pid, uid, why, status);
1192 * Try to move the task's state to DEAD
1193 * only one thread is allowed to do this:
1195 state = xchg(&p->exit_state, EXIT_DEAD);
1196 if (state != EXIT_ZOMBIE) {
1197 BUG_ON(state != EXIT_DEAD);
1198 return 0;
1201 traced = ptrace_reparented(p);
1203 * It can be ptraced but not reparented, check
1204 * !task_detached() to filter out sub-threads.
1206 if (likely(!traced) && likely(!task_detached(p))) {
1207 struct signal_struct *psig;
1208 struct signal_struct *sig;
1209 unsigned long maxrss;
1212 * The resource counters for the group leader are in its
1213 * own task_struct. Those for dead threads in the group
1214 * are in its signal_struct, as are those for the child
1215 * processes it has previously reaped. All these
1216 * accumulate in the parent's signal_struct c* fields.
1218 * We don't bother to take a lock here to protect these
1219 * p->signal fields, because they are only touched by
1220 * __exit_signal, which runs with tasklist_lock
1221 * write-locked anyway, and so is excluded here. We do
1222 * need to protect the access to parent->signal fields,
1223 * as other threads in the parent group can be right
1224 * here reaping other children at the same time.
1226 spin_lock_irq(&p->real_parent->sighand->siglock);
1227 psig = p->real_parent->signal;
1228 sig = p->signal;
1229 psig->cutime =
1230 cputime_add(psig->cutime,
1231 cputime_add(p->utime,
1232 cputime_add(sig->utime,
1233 sig->cutime)));
1234 psig->cstime =
1235 cputime_add(psig->cstime,
1236 cputime_add(p->stime,
1237 cputime_add(sig->stime,
1238 sig->cstime)));
1239 psig->cgtime =
1240 cputime_add(psig->cgtime,
1241 cputime_add(p->gtime,
1242 cputime_add(sig->gtime,
1243 sig->cgtime)));
1244 psig->cmin_flt +=
1245 p->min_flt + sig->min_flt + sig->cmin_flt;
1246 psig->cmaj_flt +=
1247 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1248 psig->cnvcsw +=
1249 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1250 psig->cnivcsw +=
1251 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1252 psig->cinblock +=
1253 task_io_get_inblock(p) +
1254 sig->inblock + sig->cinblock;
1255 psig->coublock +=
1256 task_io_get_oublock(p) +
1257 sig->oublock + sig->coublock;
1258 maxrss = max(sig->maxrss, sig->cmaxrss);
1259 if (psig->cmaxrss < maxrss)
1260 psig->cmaxrss = maxrss;
1261 task_io_accounting_add(&psig->ioac, &p->ioac);
1262 task_io_accounting_add(&psig->ioac, &sig->ioac);
1263 spin_unlock_irq(&p->real_parent->sighand->siglock);
1267 * Now we are sure this task is interesting, and no other
1268 * thread can reap it because we set its state to EXIT_DEAD.
1270 read_unlock(&tasklist_lock);
1272 retval = wo->wo_rusage
1273 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1274 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1275 ? p->signal->group_exit_code : p->exit_code;
1276 if (!retval && wo->wo_stat)
1277 retval = put_user(status, wo->wo_stat);
1279 infop = wo->wo_info;
1280 if (!retval && infop)
1281 retval = put_user(SIGCHLD, &infop->si_signo);
1282 if (!retval && infop)
1283 retval = put_user(0, &infop->si_errno);
1284 if (!retval && infop) {
1285 int why;
1287 if ((status & 0x7f) == 0) {
1288 why = CLD_EXITED;
1289 status >>= 8;
1290 } else {
1291 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1292 status &= 0x7f;
1294 retval = put_user((short)why, &infop->si_code);
1295 if (!retval)
1296 retval = put_user(status, &infop->si_status);
1298 if (!retval && infop)
1299 retval = put_user(pid, &infop->si_pid);
1300 if (!retval && infop)
1301 retval = put_user(uid, &infop->si_uid);
1302 if (!retval)
1303 retval = pid;
1305 if (traced) {
1306 write_lock_irq(&tasklist_lock);
1307 /* We dropped tasklist, ptracer could die and untrace */
1308 ptrace_unlink(p);
1310 * If this is not a detached task, notify the parent.
1311 * If it's still not detached after that, don't release
1312 * it now.
1314 if (!task_detached(p)) {
1315 do_notify_parent(p, p->exit_signal);
1316 if (!task_detached(p)) {
1317 p->exit_state = EXIT_ZOMBIE;
1318 p = NULL;
1321 write_unlock_irq(&tasklist_lock);
1323 if (p != NULL)
1324 release_task(p);
1326 return retval;
1329 static int *task_stopped_code(struct task_struct *p, bool ptrace)
1331 if (ptrace) {
1332 if (task_is_stopped_or_traced(p))
1333 return &p->exit_code;
1334 } else {
1335 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1336 return &p->signal->group_exit_code;
1338 return NULL;
1342 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1343 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1344 * the lock and this task is uninteresting. If we return nonzero, we have
1345 * released the lock and the system call should return.
1347 static int wait_task_stopped(struct wait_opts *wo,
1348 int ptrace, struct task_struct *p)
1350 struct siginfo __user *infop;
1351 int retval, exit_code, *p_code, why;
1352 uid_t uid = 0; /* unneeded, required by compiler */
1353 pid_t pid;
1356 * Traditionally we see ptrace'd stopped tasks regardless of options.
1358 if (!ptrace && !(wo->wo_flags & WUNTRACED))
1359 return 0;
1361 exit_code = 0;
1362 spin_lock_irq(&p->sighand->siglock);
1364 p_code = task_stopped_code(p, ptrace);
1365 if (unlikely(!p_code))
1366 goto unlock_sig;
1368 exit_code = *p_code;
1369 if (!exit_code)
1370 goto unlock_sig;
1372 if (!unlikely(wo->wo_flags & WNOWAIT))
1373 *p_code = 0;
1375 /* don't need the RCU readlock here as we're holding a spinlock */
1376 uid = __task_cred(p)->uid;
1377 unlock_sig:
1378 spin_unlock_irq(&p->sighand->siglock);
1379 if (!exit_code)
1380 return 0;
1383 * Now we are pretty sure this task is interesting.
1384 * Make sure it doesn't get reaped out from under us while we
1385 * give up the lock and then examine it below. We don't want to
1386 * keep holding onto the tasklist_lock while we call getrusage and
1387 * possibly take page faults for user memory.
1389 get_task_struct(p);
1390 pid = task_pid_vnr(p);
1391 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1392 read_unlock(&tasklist_lock);
1394 if (unlikely(wo->wo_flags & WNOWAIT))
1395 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
1397 retval = wo->wo_rusage
1398 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1399 if (!retval && wo->wo_stat)
1400 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1402 infop = wo->wo_info;
1403 if (!retval && infop)
1404 retval = put_user(SIGCHLD, &infop->si_signo);
1405 if (!retval && infop)
1406 retval = put_user(0, &infop->si_errno);
1407 if (!retval && infop)
1408 retval = put_user((short)why, &infop->si_code);
1409 if (!retval && infop)
1410 retval = put_user(exit_code, &infop->si_status);
1411 if (!retval && infop)
1412 retval = put_user(pid, &infop->si_pid);
1413 if (!retval && infop)
1414 retval = put_user(uid, &infop->si_uid);
1415 if (!retval)
1416 retval = pid;
1417 put_task_struct(p);
1419 BUG_ON(!retval);
1420 return retval;
1424 * Handle do_wait work for one task in a live, non-stopped state.
1425 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1426 * the lock and this task is uninteresting. If we return nonzero, we have
1427 * released the lock and the system call should return.
1429 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1431 int retval;
1432 pid_t pid;
1433 uid_t uid;
1435 if (!unlikely(wo->wo_flags & WCONTINUED))
1436 return 0;
1438 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1439 return 0;
1441 spin_lock_irq(&p->sighand->siglock);
1442 /* Re-check with the lock held. */
1443 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1444 spin_unlock_irq(&p->sighand->siglock);
1445 return 0;
1447 if (!unlikely(wo->wo_flags & WNOWAIT))
1448 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1449 uid = __task_cred(p)->uid;
1450 spin_unlock_irq(&p->sighand->siglock);
1452 pid = task_pid_vnr(p);
1453 get_task_struct(p);
1454 read_unlock(&tasklist_lock);
1456 if (!wo->wo_info) {
1457 retval = wo->wo_rusage
1458 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1459 put_task_struct(p);
1460 if (!retval && wo->wo_stat)
1461 retval = put_user(0xffff, wo->wo_stat);
1462 if (!retval)
1463 retval = pid;
1464 } else {
1465 retval = wait_noreap_copyout(wo, p, pid, uid,
1466 CLD_CONTINUED, SIGCONT);
1467 BUG_ON(retval == 0);
1470 return retval;
1474 * Consider @p for a wait by @parent.
1476 * -ECHILD should be in ->notask_error before the first call.
1477 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1478 * Returns zero if the search for a child should continue;
1479 * then ->notask_error is 0 if @p is an eligible child,
1480 * or another error from security_task_wait(), or still -ECHILD.
1482 static int wait_consider_task(struct wait_opts *wo, int ptrace,
1483 struct task_struct *p)
1485 int ret = eligible_child(wo, p);
1486 if (!ret)
1487 return ret;
1489 ret = security_task_wait(p);
1490 if (unlikely(ret < 0)) {
1492 * If we have not yet seen any eligible child,
1493 * then let this error code replace -ECHILD.
1494 * A permission error will give the user a clue
1495 * to look for security policy problems, rather
1496 * than for mysterious wait bugs.
1498 if (wo->notask_error)
1499 wo->notask_error = ret;
1500 return 0;
1503 if (likely(!ptrace) && unlikely(task_ptrace(p))) {
1505 * This child is hidden by ptrace.
1506 * We aren't allowed to see it now, but eventually we will.
1508 wo->notask_error = 0;
1509 return 0;
1512 if (p->exit_state == EXIT_DEAD)
1513 return 0;
1516 * We don't reap group leaders with subthreads.
1518 if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
1519 return wait_task_zombie(wo, p);
1522 * It's stopped or running now, so it might
1523 * later continue, exit, or stop again.
1525 wo->notask_error = 0;
1527 if (task_stopped_code(p, ptrace))
1528 return wait_task_stopped(wo, ptrace, p);
1530 return wait_task_continued(wo, p);
1534 * Do the work of do_wait() for one thread in the group, @tsk.
1536 * -ECHILD should be in ->notask_error before the first call.
1537 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1538 * Returns zero if the search for a child should continue; then
1539 * ->notask_error is 0 if there were any eligible children,
1540 * or another error from security_task_wait(), or still -ECHILD.
1542 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1544 struct task_struct *p;
1546 list_for_each_entry(p, &tsk->children, sibling) {
1548 * Do not consider detached threads.
1550 if (!task_detached(p)) {
1551 int ret = wait_consider_task(wo, 0, p);
1552 if (ret)
1553 return ret;
1557 return 0;
1560 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1562 struct task_struct *p;
1564 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1565 int ret = wait_consider_task(wo, 1, p);
1566 if (ret)
1567 return ret;
1570 return 0;
1573 static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1574 int sync, void *key)
1576 struct wait_opts *wo = container_of(wait, struct wait_opts,
1577 child_wait);
1578 struct task_struct *p = key;
1580 if (!eligible_pid(wo, p))
1581 return 0;
1583 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1584 return 0;
1586 return default_wake_function(wait, mode, sync, key);
1589 void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1591 __wake_up_sync_key(&parent->signal->wait_chldexit,
1592 TASK_INTERRUPTIBLE, 1, p);
1595 static long do_wait(struct wait_opts *wo)
1597 struct task_struct *tsk;
1598 int retval;
1600 trace_sched_process_wait(wo->wo_pid);
1602 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1603 wo->child_wait.private = current;
1604 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1605 repeat:
1607 * If there is nothing that can match our critiera just get out.
1608 * We will clear ->notask_error to zero if we see any child that
1609 * might later match our criteria, even if we are not able to reap
1610 * it yet.
1612 wo->notask_error = -ECHILD;
1613 if ((wo->wo_type < PIDTYPE_MAX) &&
1614 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
1615 goto notask;
1617 set_current_state(TASK_INTERRUPTIBLE);
1618 read_lock(&tasklist_lock);
1619 tsk = current;
1620 do {
1621 retval = do_wait_thread(wo, tsk);
1622 if (retval)
1623 goto end;
1625 retval = ptrace_do_wait(wo, tsk);
1626 if (retval)
1627 goto end;
1629 if (wo->wo_flags & __WNOTHREAD)
1630 break;
1631 } while_each_thread(current, tsk);
1632 read_unlock(&tasklist_lock);
1634 notask:
1635 retval = wo->notask_error;
1636 if (!retval && !(wo->wo_flags & WNOHANG)) {
1637 retval = -ERESTARTSYS;
1638 if (!signal_pending(current)) {
1639 schedule();
1640 goto repeat;
1643 end:
1644 __set_current_state(TASK_RUNNING);
1645 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1646 return retval;
1649 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1650 infop, int, options, struct rusage __user *, ru)
1652 struct wait_opts wo;
1653 struct pid *pid = NULL;
1654 enum pid_type type;
1655 long ret;
1657 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1658 return -EINVAL;
1659 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1660 return -EINVAL;
1662 switch (which) {
1663 case P_ALL:
1664 type = PIDTYPE_MAX;
1665 break;
1666 case P_PID:
1667 type = PIDTYPE_PID;
1668 if (upid <= 0)
1669 return -EINVAL;
1670 break;
1671 case P_PGID:
1672 type = PIDTYPE_PGID;
1673 if (upid <= 0)
1674 return -EINVAL;
1675 break;
1676 default:
1677 return -EINVAL;
1680 if (type < PIDTYPE_MAX)
1681 pid = find_get_pid(upid);
1683 wo.wo_type = type;
1684 wo.wo_pid = pid;
1685 wo.wo_flags = options;
1686 wo.wo_info = infop;
1687 wo.wo_stat = NULL;
1688 wo.wo_rusage = ru;
1689 ret = do_wait(&wo);
1691 if (ret > 0) {
1692 ret = 0;
1693 } else if (infop) {
1695 * For a WNOHANG return, clear out all the fields
1696 * we would set so the user can easily tell the
1697 * difference.
1699 if (!ret)
1700 ret = put_user(0, &infop->si_signo);
1701 if (!ret)
1702 ret = put_user(0, &infop->si_errno);
1703 if (!ret)
1704 ret = put_user(0, &infop->si_code);
1705 if (!ret)
1706 ret = put_user(0, &infop->si_pid);
1707 if (!ret)
1708 ret = put_user(0, &infop->si_uid);
1709 if (!ret)
1710 ret = put_user(0, &infop->si_status);
1713 put_pid(pid);
1715 /* avoid REGPARM breakage on x86: */
1716 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
1717 return ret;
1720 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1721 int, options, struct rusage __user *, ru)
1723 struct wait_opts wo;
1724 struct pid *pid = NULL;
1725 enum pid_type type;
1726 long ret;
1728 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1729 __WNOTHREAD|__WCLONE|__WALL))
1730 return -EINVAL;
1732 if (upid == -1)
1733 type = PIDTYPE_MAX;
1734 else if (upid < 0) {
1735 type = PIDTYPE_PGID;
1736 pid = find_get_pid(-upid);
1737 } else if (upid == 0) {
1738 type = PIDTYPE_PGID;
1739 pid = get_task_pid(current, PIDTYPE_PGID);
1740 } else /* upid > 0 */ {
1741 type = PIDTYPE_PID;
1742 pid = find_get_pid(upid);
1745 wo.wo_type = type;
1746 wo.wo_pid = pid;
1747 wo.wo_flags = options | WEXITED;
1748 wo.wo_info = NULL;
1749 wo.wo_stat = stat_addr;
1750 wo.wo_rusage = ru;
1751 ret = do_wait(&wo);
1752 put_pid(pid);
1754 /* avoid REGPARM breakage on x86: */
1755 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
1756 return ret;
1759 #ifdef __ARCH_WANT_SYS_WAITPID
1762 * sys_waitpid() remains for compatibility. waitpid() should be
1763 * implemented by calling sys_wait4() from libc.a.
1765 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1767 return sys_wait4(pid, stat_addr, options, NULL);
1770 #endif