4 * Copyright (C) 1991, 1992 Linus Torvalds
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/mnt_namespace.h>
16 #include <linux/iocontext.h>
17 #include <linux/key.h>
18 #include <linux/security.h>
19 #include <linux/cpu.h>
20 #include <linux/acct.h>
21 #include <linux/tsacct_kern.h>
22 #include <linux/file.h>
23 #include <linux/fdtable.h>
24 #include <linux/binfmts.h>
25 #include <linux/nsproxy.h>
26 #include <linux/pid_namespace.h>
27 #include <linux/ptrace.h>
28 #include <linux/profile.h>
29 #include <linux/mount.h>
30 #include <linux/proc_fs.h>
31 #include <linux/kthread.h>
32 #include <linux/mempolicy.h>
33 #include <linux/taskstats_kern.h>
34 #include <linux/delayacct.h>
35 #include <linux/freezer.h>
36 #include <linux/cgroup.h>
37 #include <linux/syscalls.h>
38 #include <linux/signal.h>
39 #include <linux/posix-timers.h>
40 #include <linux/cn_proc.h>
41 #include <linux/mutex.h>
42 #include <linux/futex.h>
43 #include <linux/compat.h>
44 #include <linux/pipe_fs_i.h>
45 #include <linux/audit.h> /* for audit_free() */
46 #include <linux/resource.h>
47 #include <linux/blkdev.h>
48 #include <linux/task_io_accounting_ops.h>
49 #include <linux/tracehook.h>
51 #include <asm/uaccess.h>
52 #include <asm/unistd.h>
53 #include <asm/pgtable.h>
54 #include <asm/mmu_context.h>
56 static void exit_mm(struct task_struct
* tsk
);
58 static inline int task_detached(struct task_struct
*p
)
60 return p
->exit_signal
== -1;
63 static void __unhash_process(struct task_struct
*p
)
66 detach_pid(p
, PIDTYPE_PID
);
67 if (thread_group_leader(p
)) {
68 detach_pid(p
, PIDTYPE_PGID
);
69 detach_pid(p
, PIDTYPE_SID
);
71 list_del_rcu(&p
->tasks
);
72 __get_cpu_var(process_counts
)--;
74 list_del_rcu(&p
->thread_group
);
75 list_del_init(&p
->sibling
);
79 * This function expects the tasklist_lock write-locked.
81 static void __exit_signal(struct task_struct
*tsk
)
83 struct signal_struct
*sig
= tsk
->signal
;
84 struct sighand_struct
*sighand
;
87 BUG_ON(!atomic_read(&sig
->count
));
89 sighand
= rcu_dereference(tsk
->sighand
);
90 spin_lock(&sighand
->siglock
);
92 posix_cpu_timers_exit(tsk
);
93 if (atomic_dec_and_test(&sig
->count
))
94 posix_cpu_timers_exit_group(tsk
);
97 * If there is any task waiting for the group exit
100 if (sig
->group_exit_task
&& atomic_read(&sig
->count
) == sig
->notify_count
)
101 wake_up_process(sig
->group_exit_task
);
103 if (tsk
== sig
->curr_target
)
104 sig
->curr_target
= next_thread(tsk
);
106 * Accumulate here the counters for all threads but the
107 * group leader as they die, so they can be added into
108 * the process-wide totals when those are taken.
109 * The group leader stays around as a zombie as long
110 * as there are other threads. When it gets reaped,
111 * the exit.c code will add its counts into these totals.
112 * We won't ever get here for the group leader, since it
113 * will have been the last reference on the signal_struct.
115 sig
->utime
= cputime_add(sig
->utime
, tsk
->utime
);
116 sig
->stime
= cputime_add(sig
->stime
, tsk
->stime
);
117 sig
->gtime
= cputime_add(sig
->gtime
, tsk
->gtime
);
118 sig
->min_flt
+= tsk
->min_flt
;
119 sig
->maj_flt
+= tsk
->maj_flt
;
120 sig
->nvcsw
+= tsk
->nvcsw
;
121 sig
->nivcsw
+= tsk
->nivcsw
;
122 sig
->inblock
+= task_io_get_inblock(tsk
);
123 sig
->oublock
+= task_io_get_oublock(tsk
);
124 #ifdef CONFIG_TASK_XACCT
125 sig
->rchar
+= tsk
->rchar
;
126 sig
->wchar
+= tsk
->wchar
;
127 sig
->syscr
+= tsk
->syscr
;
128 sig
->syscw
+= tsk
->syscw
;
129 #endif /* CONFIG_TASK_XACCT */
130 #ifdef CONFIG_TASK_IO_ACCOUNTING
131 sig
->ioac
.read_bytes
+= tsk
->ioac
.read_bytes
;
132 sig
->ioac
.write_bytes
+= tsk
->ioac
.write_bytes
;
133 sig
->ioac
.cancelled_write_bytes
+=
134 tsk
->ioac
.cancelled_write_bytes
;
135 #endif /* CONFIG_TASK_IO_ACCOUNTING */
136 sig
->sum_sched_runtime
+= tsk
->se
.sum_exec_runtime
;
137 sig
= NULL
; /* Marker for below. */
140 __unhash_process(tsk
);
143 * Do this under ->siglock, we can race with another thread
144 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
146 flush_sigqueue(&tsk
->pending
);
150 spin_unlock(&sighand
->siglock
);
152 __cleanup_sighand(sighand
);
153 clear_tsk_thread_flag(tsk
,TIF_SIGPENDING
);
155 flush_sigqueue(&sig
->shared_pending
);
156 taskstats_tgid_free(sig
);
157 __cleanup_signal(sig
);
161 static void delayed_put_task_struct(struct rcu_head
*rhp
)
163 put_task_struct(container_of(rhp
, struct task_struct
, rcu
));
167 * Do final ptrace-related cleanup of a zombie being reaped.
169 * Called with write_lock(&tasklist_lock) held.
171 static void ptrace_release_task(struct task_struct
*p
)
173 BUG_ON(!list_empty(&p
->ptraced
));
175 BUG_ON(!list_empty(&p
->ptrace_entry
));
178 void release_task(struct task_struct
* p
)
180 struct task_struct
*leader
;
183 atomic_dec(&p
->user
->processes
);
185 write_lock_irq(&tasklist_lock
);
186 ptrace_release_task(p
);
190 * If we are the last non-leader member of the thread
191 * group, and the leader is zombie, then notify the
192 * group leader's parent process. (if it wants notification.)
195 leader
= p
->group_leader
;
196 if (leader
!= p
&& thread_group_empty(leader
) && leader
->exit_state
== EXIT_ZOMBIE
) {
197 BUG_ON(task_detached(leader
));
198 do_notify_parent(leader
, leader
->exit_signal
);
200 * If we were the last child thread and the leader has
201 * exited already, and the leader's parent ignores SIGCHLD,
202 * then we are the one who should release the leader.
204 * do_notify_parent() will have marked it self-reaping in
207 zap_leader
= task_detached(leader
);
210 write_unlock_irq(&tasklist_lock
);
212 call_rcu(&p
->rcu
, delayed_put_task_struct
);
215 if (unlikely(zap_leader
))
220 * This checks not only the pgrp, but falls back on the pid if no
221 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
224 * The caller must hold rcu lock or the tasklist lock.
226 struct pid
*session_of_pgrp(struct pid
*pgrp
)
228 struct task_struct
*p
;
229 struct pid
*sid
= NULL
;
231 p
= pid_task(pgrp
, PIDTYPE_PGID
);
233 p
= pid_task(pgrp
, PIDTYPE_PID
);
235 sid
= task_session(p
);
241 * Determine if a process group is "orphaned", according to the POSIX
242 * definition in 2.2.2.52. Orphaned process groups are not to be affected
243 * by terminal-generated stop signals. Newly orphaned process groups are
244 * to receive a SIGHUP and a SIGCONT.
246 * "I ask you, have you ever known what it is to be an orphan?"
248 static int will_become_orphaned_pgrp(struct pid
*pgrp
, struct task_struct
*ignored_task
)
250 struct task_struct
*p
;
252 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
253 if ((p
== ignored_task
) ||
254 (p
->exit_state
&& thread_group_empty(p
)) ||
255 is_global_init(p
->real_parent
))
258 if (task_pgrp(p
->real_parent
) != pgrp
&&
259 task_session(p
->real_parent
) == task_session(p
))
261 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
266 int is_current_pgrp_orphaned(void)
270 read_lock(&tasklist_lock
);
271 retval
= will_become_orphaned_pgrp(task_pgrp(current
), NULL
);
272 read_unlock(&tasklist_lock
);
277 static int has_stopped_jobs(struct pid
*pgrp
)
280 struct task_struct
*p
;
282 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
283 if (!task_is_stopped(p
))
287 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
292 * Check to see if any process groups have become orphaned as
293 * a result of our exiting, and if they have any stopped jobs,
294 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
297 kill_orphaned_pgrp(struct task_struct
*tsk
, struct task_struct
*parent
)
299 struct pid
*pgrp
= task_pgrp(tsk
);
300 struct task_struct
*ignored_task
= tsk
;
303 /* exit: our father is in a different pgrp than
304 * we are and we were the only connection outside.
306 parent
= tsk
->real_parent
;
308 /* reparent: our child is in a different pgrp than
309 * we are, and it was the only connection outside.
313 if (task_pgrp(parent
) != pgrp
&&
314 task_session(parent
) == task_session(tsk
) &&
315 will_become_orphaned_pgrp(pgrp
, ignored_task
) &&
316 has_stopped_jobs(pgrp
)) {
317 __kill_pgrp_info(SIGHUP
, SEND_SIG_PRIV
, pgrp
);
318 __kill_pgrp_info(SIGCONT
, SEND_SIG_PRIV
, pgrp
);
323 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
325 * If a kernel thread is launched as a result of a system call, or if
326 * it ever exits, it should generally reparent itself to kthreadd so it
327 * isn't in the way of other processes and is correctly cleaned up on exit.
329 * The various task state such as scheduling policy and priority may have
330 * been inherited from a user process, so we reset them to sane values here.
332 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
334 static void reparent_to_kthreadd(void)
336 write_lock_irq(&tasklist_lock
);
338 ptrace_unlink(current
);
339 /* Reparent to init */
340 current
->real_parent
= current
->parent
= kthreadd_task
;
341 list_move_tail(¤t
->sibling
, ¤t
->real_parent
->children
);
343 /* Set the exit signal to SIGCHLD so we signal init on exit */
344 current
->exit_signal
= SIGCHLD
;
346 if (task_nice(current
) < 0)
347 set_user_nice(current
, 0);
351 security_task_reparent_to_init(current
);
352 memcpy(current
->signal
->rlim
, init_task
.signal
->rlim
,
353 sizeof(current
->signal
->rlim
));
354 atomic_inc(&(INIT_USER
->__count
));
355 write_unlock_irq(&tasklist_lock
);
356 switch_uid(INIT_USER
);
359 void __set_special_pids(struct pid
*pid
)
361 struct task_struct
*curr
= current
->group_leader
;
362 pid_t nr
= pid_nr(pid
);
364 if (task_session(curr
) != pid
) {
365 change_pid(curr
, PIDTYPE_SID
, pid
);
366 set_task_session(curr
, nr
);
368 if (task_pgrp(curr
) != pid
) {
369 change_pid(curr
, PIDTYPE_PGID
, pid
);
370 set_task_pgrp(curr
, nr
);
374 static void set_special_pids(struct pid
*pid
)
376 write_lock_irq(&tasklist_lock
);
377 __set_special_pids(pid
);
378 write_unlock_irq(&tasklist_lock
);
382 * Let kernel threads use this to say that they
383 * allow a certain signal (since daemonize() will
384 * have disabled all of them by default).
386 int allow_signal(int sig
)
388 if (!valid_signal(sig
) || sig
< 1)
391 spin_lock_irq(¤t
->sighand
->siglock
);
392 sigdelset(¤t
->blocked
, sig
);
394 /* Kernel threads handle their own signals.
395 Let the signal code know it'll be handled, so
396 that they don't get converted to SIGKILL or
397 just silently dropped */
398 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= (void __user
*)2;
401 spin_unlock_irq(¤t
->sighand
->siglock
);
405 EXPORT_SYMBOL(allow_signal
);
407 int disallow_signal(int sig
)
409 if (!valid_signal(sig
) || sig
< 1)
412 spin_lock_irq(¤t
->sighand
->siglock
);
413 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= SIG_IGN
;
415 spin_unlock_irq(¤t
->sighand
->siglock
);
419 EXPORT_SYMBOL(disallow_signal
);
422 * Put all the gunge required to become a kernel thread without
423 * attached user resources in one place where it belongs.
426 void daemonize(const char *name
, ...)
429 struct fs_struct
*fs
;
432 va_start(args
, name
);
433 vsnprintf(current
->comm
, sizeof(current
->comm
), name
, args
);
437 * If we were started as result of loading a module, close all of the
438 * user space pages. We don't need them, and if we didn't close them
439 * they would be locked into memory.
443 * We don't want to have TIF_FREEZE set if the system-wide hibernation
444 * or suspend transition begins right now.
446 current
->flags
|= (PF_NOFREEZE
| PF_KTHREAD
);
448 if (current
->nsproxy
!= &init_nsproxy
) {
449 get_nsproxy(&init_nsproxy
);
450 switch_task_namespaces(current
, &init_nsproxy
);
452 set_special_pids(&init_struct_pid
);
453 proc_clear_tty(current
);
455 /* Block and flush all signals */
456 sigfillset(&blocked
);
457 sigprocmask(SIG_BLOCK
, &blocked
, NULL
);
458 flush_signals(current
);
460 /* Become as one with the init task */
462 exit_fs(current
); /* current->fs->count--; */
465 atomic_inc(&fs
->count
);
468 current
->files
= init_task
.files
;
469 atomic_inc(¤t
->files
->count
);
471 reparent_to_kthreadd();
474 EXPORT_SYMBOL(daemonize
);
476 static void close_files(struct files_struct
* files
)
484 * It is safe to dereference the fd table without RCU or
485 * ->file_lock because this is the last reference to the
488 fdt
= files_fdtable(files
);
492 if (i
>= fdt
->max_fds
)
494 set
= fdt
->open_fds
->fds_bits
[j
++];
497 struct file
* file
= xchg(&fdt
->fd
[i
], NULL
);
499 filp_close(file
, files
);
509 struct files_struct
*get_files_struct(struct task_struct
*task
)
511 struct files_struct
*files
;
516 atomic_inc(&files
->count
);
522 void put_files_struct(struct files_struct
*files
)
526 if (atomic_dec_and_test(&files
->count
)) {
529 * Free the fd and fdset arrays if we expanded them.
530 * If the fdtable was embedded, pass files for freeing
531 * at the end of the RCU grace period. Otherwise,
532 * you can free files immediately.
534 fdt
= files_fdtable(files
);
535 if (fdt
!= &files
->fdtab
)
536 kmem_cache_free(files_cachep
, files
);
541 void reset_files_struct(struct files_struct
*files
)
543 struct task_struct
*tsk
= current
;
544 struct files_struct
*old
;
550 put_files_struct(old
);
553 void exit_files(struct task_struct
*tsk
)
555 struct files_struct
* files
= tsk
->files
;
561 put_files_struct(files
);
565 void put_fs_struct(struct fs_struct
*fs
)
567 /* No need to hold fs->lock if we are killing it */
568 if (atomic_dec_and_test(&fs
->count
)) {
571 if (fs
->altroot
.dentry
)
572 path_put(&fs
->altroot
);
573 kmem_cache_free(fs_cachep
, fs
);
577 void exit_fs(struct task_struct
*tsk
)
579 struct fs_struct
* fs
= tsk
->fs
;
589 EXPORT_SYMBOL_GPL(exit_fs
);
591 #ifdef CONFIG_MM_OWNER
593 * Task p is exiting and it owned mm, lets find a new owner for it
596 mm_need_new_owner(struct mm_struct
*mm
, struct task_struct
*p
)
599 * If there are other users of the mm and the owner (us) is exiting
600 * we need to find a new owner to take on the responsibility.
604 if (atomic_read(&mm
->mm_users
) <= 1)
611 void mm_update_next_owner(struct mm_struct
*mm
)
613 struct task_struct
*c
, *g
, *p
= current
;
616 if (!mm_need_new_owner(mm
, p
))
619 read_lock(&tasklist_lock
);
621 * Search in the children
623 list_for_each_entry(c
, &p
->children
, sibling
) {
625 goto assign_new_owner
;
629 * Search in the siblings
631 list_for_each_entry(c
, &p
->parent
->children
, sibling
) {
633 goto assign_new_owner
;
637 * Search through everything else. We should not get
640 do_each_thread(g
, c
) {
642 goto assign_new_owner
;
643 } while_each_thread(g
, c
);
645 read_unlock(&tasklist_lock
);
652 * The task_lock protects c->mm from changing.
653 * We always want mm->owner->mm == mm
657 * Delay read_unlock() till we have the task_lock()
658 * to ensure that c does not slip away underneath us
660 read_unlock(&tasklist_lock
);
666 cgroup_mm_owner_callbacks(mm
->owner
, c
);
671 #endif /* CONFIG_MM_OWNER */
674 * Turn us into a lazy TLB process if we
677 static void exit_mm(struct task_struct
* tsk
)
679 struct mm_struct
*mm
= tsk
->mm
;
680 struct core_state
*core_state
;
686 * Serialize with any possible pending coredump.
687 * We must hold mmap_sem around checking core_state
688 * and clearing tsk->mm. The core-inducing thread
689 * will increment ->nr_threads for each thread in the
690 * group with ->mm != NULL.
692 down_read(&mm
->mmap_sem
);
693 core_state
= mm
->core_state
;
695 struct core_thread self
;
696 up_read(&mm
->mmap_sem
);
699 self
.next
= xchg(&core_state
->dumper
.next
, &self
);
701 * Implies mb(), the result of xchg() must be visible
702 * to core_state->dumper.
704 if (atomic_dec_and_test(&core_state
->nr_threads
))
705 complete(&core_state
->startup
);
708 set_task_state(tsk
, TASK_UNINTERRUPTIBLE
);
709 if (!self
.task
) /* see coredump_finish() */
713 __set_task_state(tsk
, TASK_RUNNING
);
714 down_read(&mm
->mmap_sem
);
716 atomic_inc(&mm
->mm_count
);
717 BUG_ON(mm
!= tsk
->active_mm
);
718 /* more a memory barrier than a real lock */
721 up_read(&mm
->mmap_sem
);
722 enter_lazy_tlb(mm
, current
);
723 /* We don't want this task to be frozen prematurely */
724 clear_freeze_flag(tsk
);
726 mm_update_next_owner(mm
);
731 * Return nonzero if @parent's children should reap themselves.
733 * Called with write_lock_irq(&tasklist_lock) held.
735 static int ignoring_children(struct task_struct
*parent
)
738 struct sighand_struct
*psig
= parent
->sighand
;
740 spin_lock_irqsave(&psig
->siglock
, flags
);
741 ret
= (psig
->action
[SIGCHLD
-1].sa
.sa_handler
== SIG_IGN
||
742 (psig
->action
[SIGCHLD
-1].sa
.sa_flags
& SA_NOCLDWAIT
));
743 spin_unlock_irqrestore(&psig
->siglock
, flags
);
748 * Detach all tasks we were using ptrace on.
749 * Any that need to be release_task'd are put on the @dead list.
751 * Called with write_lock(&tasklist_lock) held.
753 static void ptrace_exit(struct task_struct
*parent
, struct list_head
*dead
)
755 struct task_struct
*p
, *n
;
758 list_for_each_entry_safe(p
, n
, &parent
->ptraced
, ptrace_entry
) {
761 if (p
->exit_state
!= EXIT_ZOMBIE
)
765 * If it's a zombie, our attachedness prevented normal
766 * parent notification or self-reaping. Do notification
767 * now if it would have happened earlier. If it should
768 * reap itself, add it to the @dead list. We can't call
769 * release_task() here because we already hold tasklist_lock.
771 * If it's our own child, there is no notification to do.
772 * But if our normal children self-reap, then this child
773 * was prevented by ptrace and we must reap it now.
775 if (!task_detached(p
) && thread_group_empty(p
)) {
776 if (!same_thread_group(p
->real_parent
, parent
))
777 do_notify_parent(p
, p
->exit_signal
);
780 ign
= ignoring_children(parent
);
786 if (task_detached(p
)) {
788 * Mark it as in the process of being reaped.
790 p
->exit_state
= EXIT_DEAD
;
791 list_add(&p
->ptrace_entry
, dead
);
797 * Finish up exit-time ptrace cleanup.
799 * Called without locks.
801 static void ptrace_exit_finish(struct task_struct
*parent
,
802 struct list_head
*dead
)
804 struct task_struct
*p
, *n
;
806 BUG_ON(!list_empty(&parent
->ptraced
));
808 list_for_each_entry_safe(p
, n
, dead
, ptrace_entry
) {
809 list_del_init(&p
->ptrace_entry
);
814 static void reparent_thread(struct task_struct
*p
, struct task_struct
*father
)
816 if (p
->pdeath_signal
)
817 /* We already hold the tasklist_lock here. */
818 group_send_sig_info(p
->pdeath_signal
, SEND_SIG_NOINFO
, p
);
820 list_move_tail(&p
->sibling
, &p
->real_parent
->children
);
822 /* If this is a threaded reparent there is no need to
823 * notify anyone anything has happened.
825 if (same_thread_group(p
->real_parent
, father
))
828 /* We don't want people slaying init. */
829 if (!task_detached(p
))
830 p
->exit_signal
= SIGCHLD
;
832 /* If we'd notified the old parent about this child's death,
833 * also notify the new parent.
835 if (!ptrace_reparented(p
) &&
836 p
->exit_state
== EXIT_ZOMBIE
&&
837 !task_detached(p
) && thread_group_empty(p
))
838 do_notify_parent(p
, p
->exit_signal
);
840 kill_orphaned_pgrp(p
, father
);
844 * When we die, we re-parent all our children.
845 * Try to give them to another thread in our thread
846 * group, and if no such member exists, give it to
847 * the child reaper process (ie "init") in our pid
850 static void forget_original_parent(struct task_struct
*father
)
852 struct task_struct
*p
, *n
, *reaper
= father
;
853 LIST_HEAD(ptrace_dead
);
855 write_lock_irq(&tasklist_lock
);
858 * First clean up ptrace if we were using it.
860 ptrace_exit(father
, &ptrace_dead
);
863 reaper
= next_thread(reaper
);
864 if (reaper
== father
) {
865 reaper
= task_child_reaper(father
);
868 } while (reaper
->flags
& PF_EXITING
);
870 list_for_each_entry_safe(p
, n
, &father
->children
, sibling
) {
871 p
->real_parent
= reaper
;
872 if (p
->parent
== father
) {
874 p
->parent
= p
->real_parent
;
876 reparent_thread(p
, father
);
879 write_unlock_irq(&tasklist_lock
);
880 BUG_ON(!list_empty(&father
->children
));
882 ptrace_exit_finish(father
, &ptrace_dead
);
886 * Send signals to all our closest relatives so that they know
887 * to properly mourn us..
889 static void exit_notify(struct task_struct
*tsk
, int group_dead
)
894 * This does two things:
896 * A. Make init inherit all the child processes
897 * B. Check to see if any process groups have become orphaned
898 * as a result of our exiting, and if they have any stopped
899 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
901 forget_original_parent(tsk
);
902 exit_task_namespaces(tsk
);
904 write_lock_irq(&tasklist_lock
);
906 kill_orphaned_pgrp(tsk
->group_leader
, NULL
);
908 /* Let father know we died
910 * Thread signals are configurable, but you aren't going to use
911 * that to send signals to arbitary processes.
912 * That stops right now.
914 * If the parent exec id doesn't match the exec id we saved
915 * when we started then we know the parent has changed security
918 * If our self_exec id doesn't match our parent_exec_id then
919 * we have changed execution domain as these two values started
920 * the same after a fork.
922 if (tsk
->exit_signal
!= SIGCHLD
&& !task_detached(tsk
) &&
923 (tsk
->parent_exec_id
!= tsk
->real_parent
->self_exec_id
||
924 tsk
->self_exec_id
!= tsk
->parent_exec_id
) &&
926 tsk
->exit_signal
= SIGCHLD
;
928 /* If something other than our normal parent is ptracing us, then
929 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
930 * only has special meaning to our real parent.
932 if (!task_detached(tsk
) && thread_group_empty(tsk
)) {
933 int signal
= ptrace_reparented(tsk
) ?
934 SIGCHLD
: tsk
->exit_signal
;
935 do_notify_parent(tsk
, signal
);
936 } else if (tsk
->ptrace
) {
937 do_notify_parent(tsk
, SIGCHLD
);
941 if (task_detached(tsk
) && likely(!tsk
->ptrace
))
943 tsk
->exit_state
= state
;
945 /* mt-exec, de_thread() is waiting for us */
946 if (thread_group_leader(tsk
) &&
947 tsk
->signal
->notify_count
< 0 &&
948 tsk
->signal
->group_exit_task
)
949 wake_up_process(tsk
->signal
->group_exit_task
);
951 write_unlock_irq(&tasklist_lock
);
953 /* If the process is dead, release it - nobody will wait for it */
954 if (state
== EXIT_DEAD
)
958 #ifdef CONFIG_DEBUG_STACK_USAGE
959 static void check_stack_usage(void)
961 static DEFINE_SPINLOCK(low_water_lock
);
962 static int lowest_to_date
= THREAD_SIZE
;
963 unsigned long *n
= end_of_stack(current
);
968 free
= (unsigned long)n
- (unsigned long)end_of_stack(current
);
970 if (free
>= lowest_to_date
)
973 spin_lock(&low_water_lock
);
974 if (free
< lowest_to_date
) {
975 printk(KERN_WARNING
"%s used greatest stack depth: %lu bytes "
977 current
->comm
, free
);
978 lowest_to_date
= free
;
980 spin_unlock(&low_water_lock
);
983 static inline void check_stack_usage(void) {}
986 static inline void exit_child_reaper(struct task_struct
*tsk
)
988 if (likely(tsk
->group_leader
!= task_child_reaper(tsk
)))
991 if (tsk
->nsproxy
->pid_ns
== &init_pid_ns
)
992 panic("Attempted to kill init!");
995 * @tsk is the last thread in the 'cgroup-init' and is exiting.
996 * Terminate all remaining processes in the namespace and reap them
997 * before exiting @tsk.
999 * Note that @tsk (last thread of cgroup-init) may not necessarily
1000 * be the child-reaper (i.e main thread of cgroup-init) of the
1001 * namespace i.e the child_reaper may have already exited.
1003 * Even after a child_reaper exits, we let it inherit orphaned children,
1004 * because, pid_ns->child_reaper remains valid as long as there is
1005 * at least one living sub-thread in the cgroup init.
1007 * This living sub-thread of the cgroup-init will be notified when
1008 * a child inherited by the 'child-reaper' exits (do_notify_parent()
1009 * uses __group_send_sig_info()). Further, when reaping child processes,
1010 * do_wait() iterates over children of all living sub threads.
1012 * i.e even though 'child_reaper' thread is listed as the parent of the
1013 * orphaned children, any living sub-thread in the cgroup-init can
1014 * perform the role of the child_reaper.
1016 zap_pid_ns_processes(tsk
->nsproxy
->pid_ns
);
1019 NORET_TYPE
void do_exit(long code
)
1021 struct task_struct
*tsk
= current
;
1024 profile_task_exit(tsk
);
1026 WARN_ON(atomic_read(&tsk
->fs_excl
));
1028 if (unlikely(in_interrupt()))
1029 panic("Aiee, killing interrupt handler!");
1030 if (unlikely(!tsk
->pid
))
1031 panic("Attempted to kill the idle task!");
1033 tracehook_report_exit(&code
);
1036 * We're taking recursive faults here in do_exit. Safest is to just
1037 * leave this task alone and wait for reboot.
1039 if (unlikely(tsk
->flags
& PF_EXITING
)) {
1041 "Fixing recursive fault but reboot is needed!\n");
1043 * We can do this unlocked here. The futex code uses
1044 * this flag just to verify whether the pi state
1045 * cleanup has been done or not. In the worst case it
1046 * loops once more. We pretend that the cleanup was
1047 * done as there is no way to return. Either the
1048 * OWNER_DIED bit is set by now or we push the blocked
1049 * task into the wait for ever nirwana as well.
1051 tsk
->flags
|= PF_EXITPIDONE
;
1052 if (tsk
->io_context
)
1054 set_current_state(TASK_UNINTERRUPTIBLE
);
1058 exit_signals(tsk
); /* sets PF_EXITING */
1060 * tsk->flags are checked in the futex code to protect against
1061 * an exiting task cleaning up the robust pi futexes.
1064 spin_unlock_wait(&tsk
->pi_lock
);
1066 if (unlikely(in_atomic()))
1067 printk(KERN_INFO
"note: %s[%d] exited with preempt_count %d\n",
1068 current
->comm
, task_pid_nr(current
),
1071 acct_update_integrals(tsk
);
1073 update_hiwater_rss(tsk
->mm
);
1074 update_hiwater_vm(tsk
->mm
);
1076 group_dead
= atomic_dec_and_test(&tsk
->signal
->live
);
1078 exit_child_reaper(tsk
);
1079 hrtimer_cancel(&tsk
->signal
->real_timer
);
1080 exit_itimers(tsk
->signal
);
1082 acct_collect(code
, group_dead
);
1084 if (unlikely(tsk
->robust_list
))
1085 exit_robust_list(tsk
);
1086 #ifdef CONFIG_COMPAT
1087 if (unlikely(tsk
->compat_robust_list
))
1088 compat_exit_robust_list(tsk
);
1093 if (unlikely(tsk
->audit_context
))
1096 tsk
->exit_code
= code
;
1097 taskstats_exit(tsk
, group_dead
);
1106 check_stack_usage();
1108 cgroup_exit(tsk
, 1);
1111 if (group_dead
&& tsk
->signal
->leader
)
1112 disassociate_ctty(1);
1114 module_put(task_thread_info(tsk
)->exec_domain
->module
);
1116 module_put(tsk
->binfmt
->module
);
1118 proc_exit_connector(tsk
);
1119 exit_notify(tsk
, group_dead
);
1121 mpol_put(tsk
->mempolicy
);
1122 tsk
->mempolicy
= NULL
;
1126 * This must happen late, after the PID is not
1129 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
1130 exit_pi_state_list(tsk
);
1131 if (unlikely(current
->pi_state_cache
))
1132 kfree(current
->pi_state_cache
);
1135 * Make sure we are holding no locks:
1137 debug_check_no_locks_held(tsk
);
1139 * We can do this unlocked here. The futex code uses this flag
1140 * just to verify whether the pi state cleanup has been done
1141 * or not. In the worst case it loops once more.
1143 tsk
->flags
|= PF_EXITPIDONE
;
1145 if (tsk
->io_context
)
1148 if (tsk
->splice_pipe
)
1149 __free_pipe_info(tsk
->splice_pipe
);
1152 /* causes final put_task_struct in finish_task_switch(). */
1153 tsk
->state
= TASK_DEAD
;
1157 /* Avoid "noreturn function does return". */
1159 cpu_relax(); /* For when BUG is null */
1162 EXPORT_SYMBOL_GPL(do_exit
);
1164 NORET_TYPE
void complete_and_exit(struct completion
*comp
, long code
)
1172 EXPORT_SYMBOL(complete_and_exit
);
1174 asmlinkage
long sys_exit(int error_code
)
1176 do_exit((error_code
&0xff)<<8);
1180 * Take down every thread in the group. This is called by fatal signals
1181 * as well as by sys_exit_group (below).
1184 do_group_exit(int exit_code
)
1186 struct signal_struct
*sig
= current
->signal
;
1188 BUG_ON(exit_code
& 0x80); /* core dumps don't get here */
1190 if (signal_group_exit(sig
))
1191 exit_code
= sig
->group_exit_code
;
1192 else if (!thread_group_empty(current
)) {
1193 struct sighand_struct
*const sighand
= current
->sighand
;
1194 spin_lock_irq(&sighand
->siglock
);
1195 if (signal_group_exit(sig
))
1196 /* Another thread got here before we took the lock. */
1197 exit_code
= sig
->group_exit_code
;
1199 sig
->group_exit_code
= exit_code
;
1200 sig
->flags
= SIGNAL_GROUP_EXIT
;
1201 zap_other_threads(current
);
1203 spin_unlock_irq(&sighand
->siglock
);
1211 * this kills every thread in the thread group. Note that any externally
1212 * wait4()-ing process will get the correct exit code - even if this
1213 * thread is not the thread group leader.
1215 asmlinkage
void sys_exit_group(int error_code
)
1217 do_group_exit((error_code
& 0xff) << 8);
1220 static struct pid
*task_pid_type(struct task_struct
*task
, enum pid_type type
)
1222 struct pid
*pid
= NULL
;
1223 if (type
== PIDTYPE_PID
)
1224 pid
= task
->pids
[type
].pid
;
1225 else if (type
< PIDTYPE_MAX
)
1226 pid
= task
->group_leader
->pids
[type
].pid
;
1230 static int eligible_child(enum pid_type type
, struct pid
*pid
, int options
,
1231 struct task_struct
*p
)
1235 if (type
< PIDTYPE_MAX
) {
1236 if (task_pid_type(p
, type
) != pid
)
1240 /* Wait for all children (clone and not) if __WALL is set;
1241 * otherwise, wait for clone children *only* if __WCLONE is
1242 * set; otherwise, wait for non-clone children *only*. (Note:
1243 * A "clone" child here is one that reports to its parent
1244 * using a signal other than SIGCHLD.) */
1245 if (((p
->exit_signal
!= SIGCHLD
) ^ ((options
& __WCLONE
) != 0))
1246 && !(options
& __WALL
))
1249 err
= security_task_wait(p
);
1256 static int wait_noreap_copyout(struct task_struct
*p
, pid_t pid
, uid_t uid
,
1257 int why
, int status
,
1258 struct siginfo __user
*infop
,
1259 struct rusage __user
*rusagep
)
1261 int retval
= rusagep
? getrusage(p
, RUSAGE_BOTH
, rusagep
) : 0;
1265 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1267 retval
= put_user(0, &infop
->si_errno
);
1269 retval
= put_user((short)why
, &infop
->si_code
);
1271 retval
= put_user(pid
, &infop
->si_pid
);
1273 retval
= put_user(uid
, &infop
->si_uid
);
1275 retval
= put_user(status
, &infop
->si_status
);
1282 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1283 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1284 * the lock and this task is uninteresting. If we return nonzero, we have
1285 * released the lock and the system call should return.
1287 static int wait_task_zombie(struct task_struct
*p
, int options
,
1288 struct siginfo __user
*infop
,
1289 int __user
*stat_addr
, struct rusage __user
*ru
)
1291 unsigned long state
;
1292 int retval
, status
, traced
;
1293 pid_t pid
= task_pid_vnr(p
);
1295 if (!likely(options
& WEXITED
))
1298 if (unlikely(options
& WNOWAIT
)) {
1300 int exit_code
= p
->exit_code
;
1304 read_unlock(&tasklist_lock
);
1305 if ((exit_code
& 0x7f) == 0) {
1307 status
= exit_code
>> 8;
1309 why
= (exit_code
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1310 status
= exit_code
& 0x7f;
1312 return wait_noreap_copyout(p
, pid
, uid
, why
,
1317 * Try to move the task's state to DEAD
1318 * only one thread is allowed to do this:
1320 state
= xchg(&p
->exit_state
, EXIT_DEAD
);
1321 if (state
!= EXIT_ZOMBIE
) {
1322 BUG_ON(state
!= EXIT_DEAD
);
1326 traced
= ptrace_reparented(p
);
1328 if (likely(!traced
)) {
1329 struct signal_struct
*psig
;
1330 struct signal_struct
*sig
;
1333 * The resource counters for the group leader are in its
1334 * own task_struct. Those for dead threads in the group
1335 * are in its signal_struct, as are those for the child
1336 * processes it has previously reaped. All these
1337 * accumulate in the parent's signal_struct c* fields.
1339 * We don't bother to take a lock here to protect these
1340 * p->signal fields, because they are only touched by
1341 * __exit_signal, which runs with tasklist_lock
1342 * write-locked anyway, and so is excluded here. We do
1343 * need to protect the access to p->parent->signal fields,
1344 * as other threads in the parent group can be right
1345 * here reaping other children at the same time.
1347 spin_lock_irq(&p
->parent
->sighand
->siglock
);
1348 psig
= p
->parent
->signal
;
1351 cputime_add(psig
->cutime
,
1352 cputime_add(p
->utime
,
1353 cputime_add(sig
->utime
,
1356 cputime_add(psig
->cstime
,
1357 cputime_add(p
->stime
,
1358 cputime_add(sig
->stime
,
1361 cputime_add(psig
->cgtime
,
1362 cputime_add(p
->gtime
,
1363 cputime_add(sig
->gtime
,
1366 p
->min_flt
+ sig
->min_flt
+ sig
->cmin_flt
;
1368 p
->maj_flt
+ sig
->maj_flt
+ sig
->cmaj_flt
;
1370 p
->nvcsw
+ sig
->nvcsw
+ sig
->cnvcsw
;
1372 p
->nivcsw
+ sig
->nivcsw
+ sig
->cnivcsw
;
1374 task_io_get_inblock(p
) +
1375 sig
->inblock
+ sig
->cinblock
;
1377 task_io_get_oublock(p
) +
1378 sig
->oublock
+ sig
->coublock
;
1379 #ifdef CONFIG_TASK_XACCT
1380 psig
->rchar
+= p
->rchar
+ sig
->rchar
;
1381 psig
->wchar
+= p
->wchar
+ sig
->wchar
;
1382 psig
->syscr
+= p
->syscr
+ sig
->syscr
;
1383 psig
->syscw
+= p
->syscw
+ sig
->syscw
;
1384 #endif /* CONFIG_TASK_XACCT */
1385 #ifdef CONFIG_TASK_IO_ACCOUNTING
1386 psig
->ioac
.read_bytes
+=
1387 p
->ioac
.read_bytes
+ sig
->ioac
.read_bytes
;
1388 psig
->ioac
.write_bytes
+=
1389 p
->ioac
.write_bytes
+ sig
->ioac
.write_bytes
;
1390 psig
->ioac
.cancelled_write_bytes
+=
1391 p
->ioac
.cancelled_write_bytes
+
1392 sig
->ioac
.cancelled_write_bytes
;
1393 #endif /* CONFIG_TASK_IO_ACCOUNTING */
1394 spin_unlock_irq(&p
->parent
->sighand
->siglock
);
1398 * Now we are sure this task is interesting, and no other
1399 * thread can reap it because we set its state to EXIT_DEAD.
1401 read_unlock(&tasklist_lock
);
1403 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1404 status
= (p
->signal
->flags
& SIGNAL_GROUP_EXIT
)
1405 ? p
->signal
->group_exit_code
: p
->exit_code
;
1406 if (!retval
&& stat_addr
)
1407 retval
= put_user(status
, stat_addr
);
1408 if (!retval
&& infop
)
1409 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1410 if (!retval
&& infop
)
1411 retval
= put_user(0, &infop
->si_errno
);
1412 if (!retval
&& infop
) {
1415 if ((status
& 0x7f) == 0) {
1419 why
= (status
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1422 retval
= put_user((short)why
, &infop
->si_code
);
1424 retval
= put_user(status
, &infop
->si_status
);
1426 if (!retval
&& infop
)
1427 retval
= put_user(pid
, &infop
->si_pid
);
1428 if (!retval
&& infop
)
1429 retval
= put_user(p
->uid
, &infop
->si_uid
);
1434 write_lock_irq(&tasklist_lock
);
1435 /* We dropped tasklist, ptracer could die and untrace */
1438 * If this is not a detached task, notify the parent.
1439 * If it's still not detached after that, don't release
1442 if (!task_detached(p
)) {
1443 do_notify_parent(p
, p
->exit_signal
);
1444 if (!task_detached(p
)) {
1445 p
->exit_state
= EXIT_ZOMBIE
;
1449 write_unlock_irq(&tasklist_lock
);
1458 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1459 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1460 * the lock and this task is uninteresting. If we return nonzero, we have
1461 * released the lock and the system call should return.
1463 static int wait_task_stopped(int ptrace
, struct task_struct
*p
,
1464 int options
, struct siginfo __user
*infop
,
1465 int __user
*stat_addr
, struct rusage __user
*ru
)
1467 int retval
, exit_code
, why
;
1468 uid_t uid
= 0; /* unneeded, required by compiler */
1471 if (!(options
& WUNTRACED
))
1475 spin_lock_irq(&p
->sighand
->siglock
);
1477 if (unlikely(!task_is_stopped_or_traced(p
)))
1480 if (!ptrace
&& p
->signal
->group_stop_count
> 0)
1482 * A group stop is in progress and this is the group leader.
1483 * We won't report until all threads have stopped.
1487 exit_code
= p
->exit_code
;
1491 if (!unlikely(options
& WNOWAIT
))
1496 spin_unlock_irq(&p
->sighand
->siglock
);
1501 * Now we are pretty sure this task is interesting.
1502 * Make sure it doesn't get reaped out from under us while we
1503 * give up the lock and then examine it below. We don't want to
1504 * keep holding onto the tasklist_lock while we call getrusage and
1505 * possibly take page faults for user memory.
1508 pid
= task_pid_vnr(p
);
1509 why
= ptrace
? CLD_TRAPPED
: CLD_STOPPED
;
1510 read_unlock(&tasklist_lock
);
1512 if (unlikely(options
& WNOWAIT
))
1513 return wait_noreap_copyout(p
, pid
, uid
,
1517 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1518 if (!retval
&& stat_addr
)
1519 retval
= put_user((exit_code
<< 8) | 0x7f, stat_addr
);
1520 if (!retval
&& infop
)
1521 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1522 if (!retval
&& infop
)
1523 retval
= put_user(0, &infop
->si_errno
);
1524 if (!retval
&& infop
)
1525 retval
= put_user((short)why
, &infop
->si_code
);
1526 if (!retval
&& infop
)
1527 retval
= put_user(exit_code
, &infop
->si_status
);
1528 if (!retval
&& infop
)
1529 retval
= put_user(pid
, &infop
->si_pid
);
1530 if (!retval
&& infop
)
1531 retval
= put_user(uid
, &infop
->si_uid
);
1541 * Handle do_wait work for one task in a live, non-stopped state.
1542 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1543 * the lock and this task is uninteresting. If we return nonzero, we have
1544 * released the lock and the system call should return.
1546 static int wait_task_continued(struct task_struct
*p
, int options
,
1547 struct siginfo __user
*infop
,
1548 int __user
*stat_addr
, struct rusage __user
*ru
)
1554 if (!unlikely(options
& WCONTINUED
))
1557 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
))
1560 spin_lock_irq(&p
->sighand
->siglock
);
1561 /* Re-check with the lock held. */
1562 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
)) {
1563 spin_unlock_irq(&p
->sighand
->siglock
);
1566 if (!unlikely(options
& WNOWAIT
))
1567 p
->signal
->flags
&= ~SIGNAL_STOP_CONTINUED
;
1568 spin_unlock_irq(&p
->sighand
->siglock
);
1570 pid
= task_pid_vnr(p
);
1573 read_unlock(&tasklist_lock
);
1576 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1578 if (!retval
&& stat_addr
)
1579 retval
= put_user(0xffff, stat_addr
);
1583 retval
= wait_noreap_copyout(p
, pid
, uid
,
1584 CLD_CONTINUED
, SIGCONT
,
1586 BUG_ON(retval
== 0);
1593 * Consider @p for a wait by @parent.
1595 * -ECHILD should be in *@notask_error before the first call.
1596 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1597 * Returns zero if the search for a child should continue;
1598 * then *@notask_error is 0 if @p is an eligible child,
1599 * or another error from security_task_wait(), or still -ECHILD.
1601 static int wait_consider_task(struct task_struct
*parent
, int ptrace
,
1602 struct task_struct
*p
, int *notask_error
,
1603 enum pid_type type
, struct pid
*pid
, int options
,
1604 struct siginfo __user
*infop
,
1605 int __user
*stat_addr
, struct rusage __user
*ru
)
1607 int ret
= eligible_child(type
, pid
, options
, p
);
1611 if (unlikely(ret
< 0)) {
1613 * If we have not yet seen any eligible child,
1614 * then let this error code replace -ECHILD.
1615 * A permission error will give the user a clue
1616 * to look for security policy problems, rather
1617 * than for mysterious wait bugs.
1620 *notask_error
= ret
;
1623 if (likely(!ptrace
) && unlikely(p
->ptrace
)) {
1625 * This child is hidden by ptrace.
1626 * We aren't allowed to see it now, but eventually we will.
1632 if (p
->exit_state
== EXIT_DEAD
)
1636 * We don't reap group leaders with subthreads.
1638 if (p
->exit_state
== EXIT_ZOMBIE
&& !delay_group_leader(p
))
1639 return wait_task_zombie(p
, options
, infop
, stat_addr
, ru
);
1642 * It's stopped or running now, so it might
1643 * later continue, exit, or stop again.
1647 if (task_is_stopped_or_traced(p
))
1648 return wait_task_stopped(ptrace
, p
, options
,
1649 infop
, stat_addr
, ru
);
1651 return wait_task_continued(p
, options
, infop
, stat_addr
, ru
);
1655 * Do the work of do_wait() for one thread in the group, @tsk.
1657 * -ECHILD should be in *@notask_error before the first call.
1658 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1659 * Returns zero if the search for a child should continue; then
1660 * *@notask_error is 0 if there were any eligible children,
1661 * or another error from security_task_wait(), or still -ECHILD.
1663 static int do_wait_thread(struct task_struct
*tsk
, int *notask_error
,
1664 enum pid_type type
, struct pid
*pid
, int options
,
1665 struct siginfo __user
*infop
, int __user
*stat_addr
,
1666 struct rusage __user
*ru
)
1668 struct task_struct
*p
;
1670 list_for_each_entry(p
, &tsk
->children
, sibling
) {
1672 * Do not consider detached threads.
1674 if (!task_detached(p
)) {
1675 int ret
= wait_consider_task(tsk
, 0, p
, notask_error
,
1677 infop
, stat_addr
, ru
);
1686 static int ptrace_do_wait(struct task_struct
*tsk
, int *notask_error
,
1687 enum pid_type type
, struct pid
*pid
, int options
,
1688 struct siginfo __user
*infop
, int __user
*stat_addr
,
1689 struct rusage __user
*ru
)
1691 struct task_struct
*p
;
1694 * Traditionally we see ptrace'd stopped tasks regardless of options.
1696 options
|= WUNTRACED
;
1698 list_for_each_entry(p
, &tsk
->ptraced
, ptrace_entry
) {
1699 int ret
= wait_consider_task(tsk
, 1, p
, notask_error
,
1701 infop
, stat_addr
, ru
);
1709 static long do_wait(enum pid_type type
, struct pid
*pid
, int options
,
1710 struct siginfo __user
*infop
, int __user
*stat_addr
,
1711 struct rusage __user
*ru
)
1713 DECLARE_WAITQUEUE(wait
, current
);
1714 struct task_struct
*tsk
;
1717 add_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1720 * If there is nothing that can match our critiera just get out.
1721 * We will clear @retval to zero if we see any child that might later
1722 * match our criteria, even if we are not able to reap it yet.
1725 if ((type
< PIDTYPE_MAX
) && (!pid
|| hlist_empty(&pid
->tasks
[type
])))
1728 current
->state
= TASK_INTERRUPTIBLE
;
1729 read_lock(&tasklist_lock
);
1732 int tsk_result
= do_wait_thread(tsk
, &retval
,
1734 infop
, stat_addr
, ru
);
1736 tsk_result
= ptrace_do_wait(tsk
, &retval
,
1738 infop
, stat_addr
, ru
);
1741 * tasklist_lock is unlocked and we have a final result.
1743 retval
= tsk_result
;
1747 if (options
& __WNOTHREAD
)
1749 tsk
= next_thread(tsk
);
1750 BUG_ON(tsk
->signal
!= current
->signal
);
1751 } while (tsk
!= current
);
1752 read_unlock(&tasklist_lock
);
1754 if (!retval
&& !(options
& WNOHANG
)) {
1755 retval
= -ERESTARTSYS
;
1756 if (!signal_pending(current
)) {
1763 current
->state
= TASK_RUNNING
;
1764 remove_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1770 * For a WNOHANG return, clear out all the fields
1771 * we would set so the user can easily tell the
1775 retval
= put_user(0, &infop
->si_signo
);
1777 retval
= put_user(0, &infop
->si_errno
);
1779 retval
= put_user(0, &infop
->si_code
);
1781 retval
= put_user(0, &infop
->si_pid
);
1783 retval
= put_user(0, &infop
->si_uid
);
1785 retval
= put_user(0, &infop
->si_status
);
1791 asmlinkage
long sys_waitid(int which
, pid_t upid
,
1792 struct siginfo __user
*infop
, int options
,
1793 struct rusage __user
*ru
)
1795 struct pid
*pid
= NULL
;
1799 if (options
& ~(WNOHANG
|WNOWAIT
|WEXITED
|WSTOPPED
|WCONTINUED
))
1801 if (!(options
& (WEXITED
|WSTOPPED
|WCONTINUED
)))
1814 type
= PIDTYPE_PGID
;
1822 if (type
< PIDTYPE_MAX
)
1823 pid
= find_get_pid(upid
);
1824 ret
= do_wait(type
, pid
, options
, infop
, NULL
, ru
);
1827 /* avoid REGPARM breakage on x86: */
1828 asmlinkage_protect(5, ret
, which
, upid
, infop
, options
, ru
);
1832 asmlinkage
long sys_wait4(pid_t upid
, int __user
*stat_addr
,
1833 int options
, struct rusage __user
*ru
)
1835 struct pid
*pid
= NULL
;
1839 if (options
& ~(WNOHANG
|WUNTRACED
|WCONTINUED
|
1840 __WNOTHREAD
|__WCLONE
|__WALL
))
1845 else if (upid
< 0) {
1846 type
= PIDTYPE_PGID
;
1847 pid
= find_get_pid(-upid
);
1848 } else if (upid
== 0) {
1849 type
= PIDTYPE_PGID
;
1850 pid
= get_pid(task_pgrp(current
));
1851 } else /* upid > 0 */ {
1853 pid
= find_get_pid(upid
);
1856 ret
= do_wait(type
, pid
, options
| WEXITED
, NULL
, stat_addr
, ru
);
1859 /* avoid REGPARM breakage on x86: */
1860 asmlinkage_protect(4, ret
, upid
, stat_addr
, options
, ru
);
1864 #ifdef __ARCH_WANT_SYS_WAITPID
1867 * sys_waitpid() remains for compatibility. waitpid() should be
1868 * implemented by calling sys_wait4() from libc.a.
1870 asmlinkage
long sys_waitpid(pid_t pid
, int __user
*stat_addr
, int options
)
1872 return sys_wait4(pid
, stat_addr
, options
, NULL
);