4 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/slab.h>
9 #include <linux/interrupt.h>
10 #include <linux/smp_lock.h>
11 #include <linux/module.h>
12 #include <linux/capability.h>
13 #include <linux/completion.h>
14 #include <linux/personality.h>
15 #include <linux/tty.h>
16 #include <linux/namespace.h>
17 #include <linux/key.h>
18 #include <linux/security.h>
19 #include <linux/cpu.h>
20 #include <linux/acct.h>
21 #include <linux/file.h>
22 #include <linux/binfmts.h>
23 #include <linux/ptrace.h>
24 #include <linux/profile.h>
25 #include <linux/mount.h>
26 #include <linux/proc_fs.h>
27 #include <linux/mempolicy.h>
28 #include <linux/taskstats_kern.h>
29 #include <linux/delayacct.h>
30 #include <linux/cpuset.h>
31 #include <linux/syscalls.h>
32 #include <linux/signal.h>
33 #include <linux/posix-timers.h>
34 #include <linux/cn_proc.h>
35 #include <linux/mutex.h>
36 #include <linux/futex.h>
37 #include <linux/compat.h>
38 #include <linux/pipe_fs_i.h>
39 #include <linux/audit.h> /* for audit_free() */
40 #include <linux/resource.h>
42 #include <asm/uaccess.h>
43 #include <asm/unistd.h>
44 #include <asm/pgtable.h>
45 #include <asm/mmu_context.h>
47 extern void sem_exit (void);
48 extern struct task_struct
*child_reaper
;
50 static void exit_mm(struct task_struct
* tsk
);
52 static void __unhash_process(struct task_struct
*p
)
55 detach_pid(p
, PIDTYPE_PID
);
56 if (thread_group_leader(p
)) {
57 detach_pid(p
, PIDTYPE_PGID
);
58 detach_pid(p
, PIDTYPE_SID
);
60 list_del_rcu(&p
->tasks
);
61 __get_cpu_var(process_counts
)--;
63 list_del_rcu(&p
->thread_group
);
68 * This function expects the tasklist_lock write-locked.
70 static void __exit_signal(struct task_struct
*tsk
)
72 struct signal_struct
*sig
= tsk
->signal
;
73 struct sighand_struct
*sighand
;
76 BUG_ON(!atomic_read(&sig
->count
));
79 sighand
= rcu_dereference(tsk
->sighand
);
80 spin_lock(&sighand
->siglock
);
82 posix_cpu_timers_exit(tsk
);
83 if (atomic_dec_and_test(&sig
->count
))
84 posix_cpu_timers_exit_group(tsk
);
87 * If there is any task waiting for the group exit
90 if (sig
->group_exit_task
&& atomic_read(&sig
->count
) == sig
->notify_count
) {
91 wake_up_process(sig
->group_exit_task
);
92 sig
->group_exit_task
= NULL
;
94 if (tsk
== sig
->curr_target
)
95 sig
->curr_target
= next_thread(tsk
);
97 * Accumulate here the counters for all threads but the
98 * group leader as they die, so they can be added into
99 * the process-wide totals when those are taken.
100 * The group leader stays around as a zombie as long
101 * as there are other threads. When it gets reaped,
102 * the exit.c code will add its counts into these totals.
103 * We won't ever get here for the group leader, since it
104 * will have been the last reference on the signal_struct.
106 sig
->utime
= cputime_add(sig
->utime
, tsk
->utime
);
107 sig
->stime
= cputime_add(sig
->stime
, tsk
->stime
);
108 sig
->min_flt
+= tsk
->min_flt
;
109 sig
->maj_flt
+= tsk
->maj_flt
;
110 sig
->nvcsw
+= tsk
->nvcsw
;
111 sig
->nivcsw
+= tsk
->nivcsw
;
112 sig
->sched_time
+= tsk
->sched_time
;
113 sig
= NULL
; /* Marker for below. */
116 __unhash_process(tsk
);
120 spin_unlock(&sighand
->siglock
);
123 __cleanup_sighand(sighand
);
124 clear_tsk_thread_flag(tsk
,TIF_SIGPENDING
);
125 flush_sigqueue(&tsk
->pending
);
127 flush_sigqueue(&sig
->shared_pending
);
128 __cleanup_signal(sig
);
132 static void delayed_put_task_struct(struct rcu_head
*rhp
)
134 put_task_struct(container_of(rhp
, struct task_struct
, rcu
));
137 void release_task(struct task_struct
* p
)
139 struct task_struct
*leader
;
142 atomic_dec(&p
->user
->processes
);
143 write_lock_irq(&tasklist_lock
);
145 BUG_ON(!list_empty(&p
->ptrace_list
) || !list_empty(&p
->ptrace_children
));
149 * If we are the last non-leader member of the thread
150 * group, and the leader is zombie, then notify the
151 * group leader's parent process. (if it wants notification.)
154 leader
= p
->group_leader
;
155 if (leader
!= p
&& thread_group_empty(leader
) && leader
->exit_state
== EXIT_ZOMBIE
) {
156 BUG_ON(leader
->exit_signal
== -1);
157 do_notify_parent(leader
, leader
->exit_signal
);
159 * If we were the last child thread and the leader has
160 * exited already, and the leader's parent ignores SIGCHLD,
161 * then we are the one who should release the leader.
163 * do_notify_parent() will have marked it self-reaping in
166 zap_leader
= (leader
->exit_signal
== -1);
170 write_unlock_irq(&tasklist_lock
);
173 call_rcu(&p
->rcu
, delayed_put_task_struct
);
176 if (unlikely(zap_leader
))
181 * This checks not only the pgrp, but falls back on the pid if no
182 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
185 int session_of_pgrp(int pgrp
)
187 struct task_struct
*p
;
190 read_lock(&tasklist_lock
);
191 do_each_task_pid(pgrp
, PIDTYPE_PGID
, p
) {
192 if (p
->signal
->session
> 0) {
193 sid
= p
->signal
->session
;
196 } while_each_task_pid(pgrp
, PIDTYPE_PGID
, p
);
197 p
= find_task_by_pid(pgrp
);
199 sid
= p
->signal
->session
;
201 read_unlock(&tasklist_lock
);
207 * Determine if a process group is "orphaned", according to the POSIX
208 * definition in 2.2.2.52. Orphaned process groups are not to be affected
209 * by terminal-generated stop signals. Newly orphaned process groups are
210 * to receive a SIGHUP and a SIGCONT.
212 * "I ask you, have you ever known what it is to be an orphan?"
214 static int will_become_orphaned_pgrp(int pgrp
, struct task_struct
*ignored_task
)
216 struct task_struct
*p
;
219 do_each_task_pid(pgrp
, PIDTYPE_PGID
, p
) {
220 if (p
== ignored_task
222 || is_init(p
->real_parent
))
224 if (process_group(p
->real_parent
) != pgrp
225 && p
->real_parent
->signal
->session
== p
->signal
->session
) {
229 } while_each_task_pid(pgrp
, PIDTYPE_PGID
, p
);
230 return ret
; /* (sighing) "Often!" */
233 int is_orphaned_pgrp(int pgrp
)
237 read_lock(&tasklist_lock
);
238 retval
= will_become_orphaned_pgrp(pgrp
, NULL
);
239 read_unlock(&tasklist_lock
);
244 static int has_stopped_jobs(int pgrp
)
247 struct task_struct
*p
;
249 do_each_task_pid(pgrp
, PIDTYPE_PGID
, p
) {
250 if (p
->state
!= TASK_STOPPED
)
254 } while_each_task_pid(pgrp
, PIDTYPE_PGID
, p
);
259 * reparent_to_init - Reparent the calling kernel thread to the init task.
261 * If a kernel thread is launched as a result of a system call, or if
262 * it ever exits, it should generally reparent itself to init so that
263 * it is correctly cleaned up on exit.
265 * The various task state such as scheduling policy and priority may have
266 * been inherited from a user process, so we reset them to sane values here.
268 * NOTE that reparent_to_init() gives the caller full capabilities.
270 static void reparent_to_init(void)
272 write_lock_irq(&tasklist_lock
);
274 ptrace_unlink(current
);
275 /* Reparent to init */
276 remove_parent(current
);
277 current
->parent
= child_reaper
;
278 current
->real_parent
= child_reaper
;
281 /* Set the exit signal to SIGCHLD so we signal init on exit */
282 current
->exit_signal
= SIGCHLD
;
284 if ((current
->policy
== SCHED_NORMAL
||
285 current
->policy
== SCHED_BATCH
)
286 && (task_nice(current
) < 0))
287 set_user_nice(current
, 0);
291 security_task_reparent_to_init(current
);
292 memcpy(current
->signal
->rlim
, init_task
.signal
->rlim
,
293 sizeof(current
->signal
->rlim
));
294 atomic_inc(&(INIT_USER
->__count
));
295 write_unlock_irq(&tasklist_lock
);
296 switch_uid(INIT_USER
);
299 void __set_special_pids(pid_t session
, pid_t pgrp
)
301 struct task_struct
*curr
= current
->group_leader
;
303 if (curr
->signal
->session
!= session
) {
304 detach_pid(curr
, PIDTYPE_SID
);
305 curr
->signal
->session
= session
;
306 attach_pid(curr
, PIDTYPE_SID
, session
);
308 if (process_group(curr
) != pgrp
) {
309 detach_pid(curr
, PIDTYPE_PGID
);
310 curr
->signal
->pgrp
= pgrp
;
311 attach_pid(curr
, PIDTYPE_PGID
, pgrp
);
315 void set_special_pids(pid_t session
, pid_t pgrp
)
317 write_lock_irq(&tasklist_lock
);
318 __set_special_pids(session
, pgrp
);
319 write_unlock_irq(&tasklist_lock
);
323 * Let kernel threads use this to say that they
324 * allow a certain signal (since daemonize() will
325 * have disabled all of them by default).
327 int allow_signal(int sig
)
329 if (!valid_signal(sig
) || sig
< 1)
332 spin_lock_irq(¤t
->sighand
->siglock
);
333 sigdelset(¤t
->blocked
, sig
);
335 /* Kernel threads handle their own signals.
336 Let the signal code know it'll be handled, so
337 that they don't get converted to SIGKILL or
338 just silently dropped */
339 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= (void __user
*)2;
342 spin_unlock_irq(¤t
->sighand
->siglock
);
346 EXPORT_SYMBOL(allow_signal
);
348 int disallow_signal(int sig
)
350 if (!valid_signal(sig
) || sig
< 1)
353 spin_lock_irq(¤t
->sighand
->siglock
);
354 sigaddset(¤t
->blocked
, sig
);
356 spin_unlock_irq(¤t
->sighand
->siglock
);
360 EXPORT_SYMBOL(disallow_signal
);
363 * Put all the gunge required to become a kernel thread without
364 * attached user resources in one place where it belongs.
367 void daemonize(const char *name
, ...)
370 struct fs_struct
*fs
;
373 va_start(args
, name
);
374 vsnprintf(current
->comm
, sizeof(current
->comm
), name
, args
);
378 * If we were started as result of loading a module, close all of the
379 * user space pages. We don't need them, and if we didn't close them
380 * they would be locked into memory.
384 set_special_pids(1, 1);
385 mutex_lock(&tty_mutex
);
386 current
->signal
->tty
= NULL
;
387 mutex_unlock(&tty_mutex
);
389 /* Block and flush all signals */
390 sigfillset(&blocked
);
391 sigprocmask(SIG_BLOCK
, &blocked
, NULL
);
392 flush_signals(current
);
394 /* Become as one with the init task */
396 exit_fs(current
); /* current->fs->count--; */
399 atomic_inc(&fs
->count
);
400 exit_namespace(current
);
401 current
->namespace = init_task
.namespace;
402 get_namespace(current
->namespace);
404 current
->files
= init_task
.files
;
405 atomic_inc(¤t
->files
->count
);
410 EXPORT_SYMBOL(daemonize
);
412 static void close_files(struct files_struct
* files
)
420 * It is safe to dereference the fd table without RCU or
421 * ->file_lock because this is the last reference to the
424 fdt
= files_fdtable(files
);
428 if (i
>= fdt
->max_fdset
|| i
>= fdt
->max_fds
)
430 set
= fdt
->open_fds
->fds_bits
[j
++];
433 struct file
* file
= xchg(&fdt
->fd
[i
], NULL
);
435 filp_close(file
, files
);
443 struct files_struct
*get_files_struct(struct task_struct
*task
)
445 struct files_struct
*files
;
450 atomic_inc(&files
->count
);
456 void fastcall
put_files_struct(struct files_struct
*files
)
460 if (atomic_dec_and_test(&files
->count
)) {
463 * Free the fd and fdset arrays if we expanded them.
464 * If the fdtable was embedded, pass files for freeing
465 * at the end of the RCU grace period. Otherwise,
466 * you can free files immediately.
468 fdt
= files_fdtable(files
);
469 if (fdt
== &files
->fdtab
)
470 fdt
->free_files
= files
;
472 kmem_cache_free(files_cachep
, files
);
477 EXPORT_SYMBOL(put_files_struct
);
479 void reset_files_struct(struct task_struct
*tsk
, struct files_struct
*files
)
481 struct files_struct
*old
;
487 put_files_struct(old
);
489 EXPORT_SYMBOL(reset_files_struct
);
491 static inline void __exit_files(struct task_struct
*tsk
)
493 struct files_struct
* files
= tsk
->files
;
499 put_files_struct(files
);
503 void exit_files(struct task_struct
*tsk
)
508 static inline void __put_fs_struct(struct fs_struct
*fs
)
510 /* No need to hold fs->lock if we are killing it */
511 if (atomic_dec_and_test(&fs
->count
)) {
518 mntput(fs
->altrootmnt
);
520 kmem_cache_free(fs_cachep
, fs
);
524 void put_fs_struct(struct fs_struct
*fs
)
529 static inline void __exit_fs(struct task_struct
*tsk
)
531 struct fs_struct
* fs
= tsk
->fs
;
541 void exit_fs(struct task_struct
*tsk
)
546 EXPORT_SYMBOL_GPL(exit_fs
);
549 * Turn us into a lazy TLB process if we
552 static void exit_mm(struct task_struct
* tsk
)
554 struct mm_struct
*mm
= tsk
->mm
;
560 * Serialize with any possible pending coredump.
561 * We must hold mmap_sem around checking core_waiters
562 * and clearing tsk->mm. The core-inducing thread
563 * will increment core_waiters for each thread in the
564 * group with ->mm != NULL.
566 down_read(&mm
->mmap_sem
);
567 if (mm
->core_waiters
) {
568 up_read(&mm
->mmap_sem
);
569 down_write(&mm
->mmap_sem
);
570 if (!--mm
->core_waiters
)
571 complete(mm
->core_startup_done
);
572 up_write(&mm
->mmap_sem
);
574 wait_for_completion(&mm
->core_done
);
575 down_read(&mm
->mmap_sem
);
577 atomic_inc(&mm
->mm_count
);
578 BUG_ON(mm
!= tsk
->active_mm
);
579 /* more a memory barrier than a real lock */
582 up_read(&mm
->mmap_sem
);
583 enter_lazy_tlb(mm
, current
);
589 choose_new_parent(struct task_struct
*p
, struct task_struct
*reaper
)
592 * Make sure we're not reparenting to ourselves and that
593 * the parent is not a zombie.
595 BUG_ON(p
== reaper
|| reaper
->exit_state
);
596 p
->real_parent
= reaper
;
600 reparent_thread(struct task_struct
*p
, struct task_struct
*father
, int traced
)
602 /* We don't want people slaying init. */
603 if (p
->exit_signal
!= -1)
604 p
->exit_signal
= SIGCHLD
;
606 if (p
->pdeath_signal
)
607 /* We already hold the tasklist_lock here. */
608 group_send_sig_info(p
->pdeath_signal
, SEND_SIG_NOINFO
, p
);
610 /* Move the child from its dying parent to the new one. */
611 if (unlikely(traced
)) {
612 /* Preserve ptrace links if someone else is tracing this child. */
613 list_del_init(&p
->ptrace_list
);
614 if (p
->parent
!= p
->real_parent
)
615 list_add(&p
->ptrace_list
, &p
->real_parent
->ptrace_children
);
617 /* If this child is being traced, then we're the one tracing it
618 * anyway, so let go of it.
622 p
->parent
= p
->real_parent
;
625 /* If we'd notified the old parent about this child's death,
626 * also notify the new parent.
628 if (p
->exit_state
== EXIT_ZOMBIE
&& p
->exit_signal
!= -1 &&
629 thread_group_empty(p
))
630 do_notify_parent(p
, p
->exit_signal
);
631 else if (p
->state
== TASK_TRACED
) {
633 * If it was at a trace stop, turn it into
634 * a normal stop since it's no longer being
642 * process group orphan check
643 * Case ii: Our child is in a different pgrp
644 * than we are, and it was the only connection
645 * outside, so the child pgrp is now orphaned.
647 if ((process_group(p
) != process_group(father
)) &&
648 (p
->signal
->session
== father
->signal
->session
)) {
649 int pgrp
= process_group(p
);
651 if (will_become_orphaned_pgrp(pgrp
, NULL
) && has_stopped_jobs(pgrp
)) {
652 __kill_pg_info(SIGHUP
, SEND_SIG_PRIV
, pgrp
);
653 __kill_pg_info(SIGCONT
, SEND_SIG_PRIV
, pgrp
);
659 * When we die, we re-parent all our children.
660 * Try to give them to another thread in our thread
661 * group, and if no such member exists, give it to
662 * the global child reaper process (ie "init")
665 forget_original_parent(struct task_struct
*father
, struct list_head
*to_release
)
667 struct task_struct
*p
, *reaper
= father
;
668 struct list_head
*_p
, *_n
;
671 reaper
= next_thread(reaper
);
672 if (reaper
== father
) {
673 reaper
= child_reaper
;
676 } while (reaper
->exit_state
);
679 * There are only two places where our children can be:
681 * - in our child list
682 * - in our ptraced child list
684 * Search them and reparent children.
686 list_for_each_safe(_p
, _n
, &father
->children
) {
688 p
= list_entry(_p
, struct task_struct
, sibling
);
692 /* if father isn't the real parent, then ptrace must be enabled */
693 BUG_ON(father
!= p
->real_parent
&& !ptrace
);
695 if (father
== p
->real_parent
) {
696 /* reparent with a reaper, real father it's us */
697 choose_new_parent(p
, reaper
);
698 reparent_thread(p
, father
, 0);
700 /* reparent ptraced task to its real parent */
702 if (p
->exit_state
== EXIT_ZOMBIE
&& p
->exit_signal
!= -1 &&
703 thread_group_empty(p
))
704 do_notify_parent(p
, p
->exit_signal
);
708 * if the ptraced child is a zombie with exit_signal == -1
709 * we must collect it before we exit, or it will remain
710 * zombie forever since we prevented it from self-reap itself
711 * while it was being traced by us, to be able to see it in wait4.
713 if (unlikely(ptrace
&& p
->exit_state
== EXIT_ZOMBIE
&& p
->exit_signal
== -1))
714 list_add(&p
->ptrace_list
, to_release
);
716 list_for_each_safe(_p
, _n
, &father
->ptrace_children
) {
717 p
= list_entry(_p
, struct task_struct
, ptrace_list
);
718 choose_new_parent(p
, reaper
);
719 reparent_thread(p
, father
, 1);
724 * Send signals to all our closest relatives so that they know
725 * to properly mourn us..
727 static void exit_notify(struct task_struct
*tsk
)
730 struct task_struct
*t
;
731 struct list_head ptrace_dead
, *_p
, *_n
;
733 if (signal_pending(tsk
) && !(tsk
->signal
->flags
& SIGNAL_GROUP_EXIT
)
734 && !thread_group_empty(tsk
)) {
736 * This occurs when there was a race between our exit
737 * syscall and a group signal choosing us as the one to
738 * wake up. It could be that we are the only thread
739 * alerted to check for pending signals, but another thread
740 * should be woken now to take the signal since we will not.
741 * Now we'll wake all the threads in the group just to make
742 * sure someone gets all the pending signals.
744 read_lock(&tasklist_lock
);
745 spin_lock_irq(&tsk
->sighand
->siglock
);
746 for (t
= next_thread(tsk
); t
!= tsk
; t
= next_thread(t
))
747 if (!signal_pending(t
) && !(t
->flags
& PF_EXITING
)) {
748 recalc_sigpending_tsk(t
);
749 if (signal_pending(t
))
750 signal_wake_up(t
, 0);
752 spin_unlock_irq(&tsk
->sighand
->siglock
);
753 read_unlock(&tasklist_lock
);
756 write_lock_irq(&tasklist_lock
);
759 * This does two things:
761 * A. Make init inherit all the child processes
762 * B. Check to see if any process groups have become orphaned
763 * as a result of our exiting, and if they have any stopped
764 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
767 INIT_LIST_HEAD(&ptrace_dead
);
768 forget_original_parent(tsk
, &ptrace_dead
);
769 BUG_ON(!list_empty(&tsk
->children
));
770 BUG_ON(!list_empty(&tsk
->ptrace_children
));
773 * Check to see if any process groups have become orphaned
774 * as a result of our exiting, and if they have any stopped
775 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
777 * Case i: Our father is in a different pgrp than we are
778 * and we were the only connection outside, so our pgrp
779 * is about to become orphaned.
782 t
= tsk
->real_parent
;
784 if ((process_group(t
) != process_group(tsk
)) &&
785 (t
->signal
->session
== tsk
->signal
->session
) &&
786 will_become_orphaned_pgrp(process_group(tsk
), tsk
) &&
787 has_stopped_jobs(process_group(tsk
))) {
788 __kill_pg_info(SIGHUP
, SEND_SIG_PRIV
, process_group(tsk
));
789 __kill_pg_info(SIGCONT
, SEND_SIG_PRIV
, process_group(tsk
));
792 /* Let father know we died
794 * Thread signals are configurable, but you aren't going to use
795 * that to send signals to arbitary processes.
796 * That stops right now.
798 * If the parent exec id doesn't match the exec id we saved
799 * when we started then we know the parent has changed security
802 * If our self_exec id doesn't match our parent_exec_id then
803 * we have changed execution domain as these two values started
804 * the same after a fork.
808 if (tsk
->exit_signal
!= SIGCHLD
&& tsk
->exit_signal
!= -1 &&
809 ( tsk
->parent_exec_id
!= t
->self_exec_id
||
810 tsk
->self_exec_id
!= tsk
->parent_exec_id
)
811 && !capable(CAP_KILL
))
812 tsk
->exit_signal
= SIGCHLD
;
815 /* If something other than our normal parent is ptracing us, then
816 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
817 * only has special meaning to our real parent.
819 if (tsk
->exit_signal
!= -1 && thread_group_empty(tsk
)) {
820 int signal
= tsk
->parent
== tsk
->real_parent
? tsk
->exit_signal
: SIGCHLD
;
821 do_notify_parent(tsk
, signal
);
822 } else if (tsk
->ptrace
) {
823 do_notify_parent(tsk
, SIGCHLD
);
827 if (tsk
->exit_signal
== -1 &&
828 (likely(tsk
->ptrace
== 0) ||
829 unlikely(tsk
->parent
->signal
->flags
& SIGNAL_GROUP_EXIT
)))
831 tsk
->exit_state
= state
;
833 write_unlock_irq(&tasklist_lock
);
835 list_for_each_safe(_p
, _n
, &ptrace_dead
) {
837 t
= list_entry(_p
, struct task_struct
, ptrace_list
);
841 /* If the process is dead, release it - nobody will wait for it */
842 if (state
== EXIT_DEAD
)
846 fastcall NORET_TYPE
void do_exit(long code
)
848 struct task_struct
*tsk
= current
;
849 struct taskstats
*tidstats
;
853 profile_task_exit(tsk
);
855 WARN_ON(atomic_read(&tsk
->fs_excl
));
857 if (unlikely(in_interrupt()))
858 panic("Aiee, killing interrupt handler!");
859 if (unlikely(!tsk
->pid
))
860 panic("Attempted to kill the idle task!");
861 if (unlikely(tsk
== child_reaper
))
862 panic("Attempted to kill init!");
864 if (unlikely(current
->ptrace
& PT_TRACE_EXIT
)) {
865 current
->ptrace_message
= code
;
866 ptrace_notify((PTRACE_EVENT_EXIT
<< 8) | SIGTRAP
);
870 * We're taking recursive faults here in do_exit. Safest is to just
871 * leave this task alone and wait for reboot.
873 if (unlikely(tsk
->flags
& PF_EXITING
)) {
875 "Fixing recursive fault but reboot is needed!\n");
878 set_current_state(TASK_UNINTERRUPTIBLE
);
882 tsk
->flags
|= PF_EXITING
;
884 if (unlikely(in_atomic()))
885 printk(KERN_INFO
"note: %s[%d] exited with preempt_count %d\n",
886 current
->comm
, current
->pid
,
889 taskstats_exit_alloc(&tidstats
, &mycpu
);
891 acct_update_integrals(tsk
);
893 update_hiwater_rss(tsk
->mm
);
894 update_hiwater_vm(tsk
->mm
);
896 group_dead
= atomic_dec_and_test(&tsk
->signal
->live
);
898 hrtimer_cancel(&tsk
->signal
->real_timer
);
899 exit_itimers(tsk
->signal
);
901 acct_collect(code
, group_dead
);
902 if (unlikely(tsk
->robust_list
))
903 exit_robust_list(tsk
);
904 #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
905 if (unlikely(tsk
->compat_robust_list
))
906 compat_exit_robust_list(tsk
);
908 if (unlikely(tsk
->audit_context
))
910 taskstats_exit_send(tsk
, tidstats
, group_dead
, mycpu
);
911 taskstats_exit_free(tidstats
);
925 if (group_dead
&& tsk
->signal
->leader
)
926 disassociate_ctty(1);
928 module_put(task_thread_info(tsk
)->exec_domain
->module
);
930 module_put(tsk
->binfmt
->module
);
932 tsk
->exit_code
= code
;
933 proc_exit_connector(tsk
);
936 mpol_free(tsk
->mempolicy
);
937 tsk
->mempolicy
= NULL
;
940 * This must happen late, after the PID is not
943 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
944 exit_pi_state_list(tsk
);
945 if (unlikely(current
->pi_state_cache
))
946 kfree(current
->pi_state_cache
);
948 * Make sure we are holding no locks:
950 debug_check_no_locks_held(tsk
);
955 if (tsk
->splice_pipe
)
956 __free_pipe_info(tsk
->splice_pipe
);
958 /* PF_DEAD causes final put_task_struct after we schedule. */
960 BUG_ON(tsk
->flags
& PF_DEAD
);
961 tsk
->flags
|= PF_DEAD
;
965 /* Avoid "noreturn function does return". */
969 EXPORT_SYMBOL_GPL(do_exit
);
971 NORET_TYPE
void complete_and_exit(struct completion
*comp
, long code
)
979 EXPORT_SYMBOL(complete_and_exit
);
981 asmlinkage
long sys_exit(int error_code
)
983 do_exit((error_code
&0xff)<<8);
987 * Take down every thread in the group. This is called by fatal signals
988 * as well as by sys_exit_group (below).
991 do_group_exit(int exit_code
)
993 BUG_ON(exit_code
& 0x80); /* core dumps don't get here */
995 if (current
->signal
->flags
& SIGNAL_GROUP_EXIT
)
996 exit_code
= current
->signal
->group_exit_code
;
997 else if (!thread_group_empty(current
)) {
998 struct signal_struct
*const sig
= current
->signal
;
999 struct sighand_struct
*const sighand
= current
->sighand
;
1000 spin_lock_irq(&sighand
->siglock
);
1001 if (sig
->flags
& SIGNAL_GROUP_EXIT
)
1002 /* Another thread got here before we took the lock. */
1003 exit_code
= sig
->group_exit_code
;
1005 sig
->group_exit_code
= exit_code
;
1006 zap_other_threads(current
);
1008 spin_unlock_irq(&sighand
->siglock
);
1016 * this kills every thread in the thread group. Note that any externally
1017 * wait4()-ing process will get the correct exit code - even if this
1018 * thread is not the thread group leader.
1020 asmlinkage
void sys_exit_group(int error_code
)
1022 do_group_exit((error_code
& 0xff) << 8);
1025 static int eligible_child(pid_t pid
, int options
, struct task_struct
*p
)
1031 if (process_group(p
) != process_group(current
))
1033 } else if (pid
!= -1) {
1034 if (process_group(p
) != -pid
)
1039 * Do not consider detached threads that are
1042 if (p
->exit_signal
== -1 && !p
->ptrace
)
1045 /* Wait for all children (clone and not) if __WALL is set;
1046 * otherwise, wait for clone children *only* if __WCLONE is
1047 * set; otherwise, wait for non-clone children *only*. (Note:
1048 * A "clone" child here is one that reports to its parent
1049 * using a signal other than SIGCHLD.) */
1050 if (((p
->exit_signal
!= SIGCHLD
) ^ ((options
& __WCLONE
) != 0))
1051 && !(options
& __WALL
))
1054 * Do not consider thread group leaders that are
1055 * in a non-empty thread group:
1057 if (delay_group_leader(p
))
1060 if (security_task_wait(p
))
1066 static int wait_noreap_copyout(struct task_struct
*p
, pid_t pid
, uid_t uid
,
1067 int why
, int status
,
1068 struct siginfo __user
*infop
,
1069 struct rusage __user
*rusagep
)
1071 int retval
= rusagep
? getrusage(p
, RUSAGE_BOTH
, rusagep
) : 0;
1075 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1077 retval
= put_user(0, &infop
->si_errno
);
1079 retval
= put_user((short)why
, &infop
->si_code
);
1081 retval
= put_user(pid
, &infop
->si_pid
);
1083 retval
= put_user(uid
, &infop
->si_uid
);
1085 retval
= put_user(status
, &infop
->si_status
);
1092 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1093 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1094 * the lock and this task is uninteresting. If we return nonzero, we have
1095 * released the lock and the system call should return.
1097 static int wait_task_zombie(struct task_struct
*p
, int noreap
,
1098 struct siginfo __user
*infop
,
1099 int __user
*stat_addr
, struct rusage __user
*ru
)
1101 unsigned long state
;
1105 if (unlikely(noreap
)) {
1108 int exit_code
= p
->exit_code
;
1111 if (unlikely(p
->exit_state
!= EXIT_ZOMBIE
))
1113 if (unlikely(p
->exit_signal
== -1 && p
->ptrace
== 0))
1116 read_unlock(&tasklist_lock
);
1117 if ((exit_code
& 0x7f) == 0) {
1119 status
= exit_code
>> 8;
1121 why
= (exit_code
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1122 status
= exit_code
& 0x7f;
1124 return wait_noreap_copyout(p
, pid
, uid
, why
,
1129 * Try to move the task's state to DEAD
1130 * only one thread is allowed to do this:
1132 state
= xchg(&p
->exit_state
, EXIT_DEAD
);
1133 if (state
!= EXIT_ZOMBIE
) {
1134 BUG_ON(state
!= EXIT_DEAD
);
1137 if (unlikely(p
->exit_signal
== -1 && p
->ptrace
== 0)) {
1139 * This can only happen in a race with a ptraced thread
1140 * dying on another processor.
1145 if (likely(p
->real_parent
== p
->parent
) && likely(p
->signal
)) {
1146 struct signal_struct
*psig
;
1147 struct signal_struct
*sig
;
1150 * The resource counters for the group leader are in its
1151 * own task_struct. Those for dead threads in the group
1152 * are in its signal_struct, as are those for the child
1153 * processes it has previously reaped. All these
1154 * accumulate in the parent's signal_struct c* fields.
1156 * We don't bother to take a lock here to protect these
1157 * p->signal fields, because they are only touched by
1158 * __exit_signal, which runs with tasklist_lock
1159 * write-locked anyway, and so is excluded here. We do
1160 * need to protect the access to p->parent->signal fields,
1161 * as other threads in the parent group can be right
1162 * here reaping other children at the same time.
1164 spin_lock_irq(&p
->parent
->sighand
->siglock
);
1165 psig
= p
->parent
->signal
;
1168 cputime_add(psig
->cutime
,
1169 cputime_add(p
->utime
,
1170 cputime_add(sig
->utime
,
1173 cputime_add(psig
->cstime
,
1174 cputime_add(p
->stime
,
1175 cputime_add(sig
->stime
,
1178 p
->min_flt
+ sig
->min_flt
+ sig
->cmin_flt
;
1180 p
->maj_flt
+ sig
->maj_flt
+ sig
->cmaj_flt
;
1182 p
->nvcsw
+ sig
->nvcsw
+ sig
->cnvcsw
;
1184 p
->nivcsw
+ sig
->nivcsw
+ sig
->cnivcsw
;
1185 spin_unlock_irq(&p
->parent
->sighand
->siglock
);
1189 * Now we are sure this task is interesting, and no other
1190 * thread can reap it because we set its state to EXIT_DEAD.
1192 read_unlock(&tasklist_lock
);
1194 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1195 status
= (p
->signal
->flags
& SIGNAL_GROUP_EXIT
)
1196 ? p
->signal
->group_exit_code
: p
->exit_code
;
1197 if (!retval
&& stat_addr
)
1198 retval
= put_user(status
, stat_addr
);
1199 if (!retval
&& infop
)
1200 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1201 if (!retval
&& infop
)
1202 retval
= put_user(0, &infop
->si_errno
);
1203 if (!retval
&& infop
) {
1206 if ((status
& 0x7f) == 0) {
1210 why
= (status
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1213 retval
= put_user((short)why
, &infop
->si_code
);
1215 retval
= put_user(status
, &infop
->si_status
);
1217 if (!retval
&& infop
)
1218 retval
= put_user(p
->pid
, &infop
->si_pid
);
1219 if (!retval
&& infop
)
1220 retval
= put_user(p
->uid
, &infop
->si_uid
);
1222 // TODO: is this safe?
1223 p
->exit_state
= EXIT_ZOMBIE
;
1227 if (p
->real_parent
!= p
->parent
) {
1228 write_lock_irq(&tasklist_lock
);
1229 /* Double-check with lock held. */
1230 if (p
->real_parent
!= p
->parent
) {
1232 // TODO: is this safe?
1233 p
->exit_state
= EXIT_ZOMBIE
;
1235 * If this is not a detached task, notify the parent.
1236 * If it's still not detached after that, don't release
1239 if (p
->exit_signal
!= -1) {
1240 do_notify_parent(p
, p
->exit_signal
);
1241 if (p
->exit_signal
!= -1)
1245 write_unlock_irq(&tasklist_lock
);
1254 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1255 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1256 * the lock and this task is uninteresting. If we return nonzero, we have
1257 * released the lock and the system call should return.
1259 static int wait_task_stopped(struct task_struct
*p
, int delayed_group_leader
,
1260 int noreap
, struct siginfo __user
*infop
,
1261 int __user
*stat_addr
, struct rusage __user
*ru
)
1263 int retval
, exit_code
;
1267 if (delayed_group_leader
&& !(p
->ptrace
& PT_PTRACED
) &&
1268 p
->signal
&& p
->signal
->group_stop_count
> 0)
1270 * A group stop is in progress and this is the group leader.
1271 * We won't report until all threads have stopped.
1276 * Now we are pretty sure this task is interesting.
1277 * Make sure it doesn't get reaped out from under us while we
1278 * give up the lock and then examine it below. We don't want to
1279 * keep holding onto the tasklist_lock while we call getrusage and
1280 * possibly take page faults for user memory.
1283 read_unlock(&tasklist_lock
);
1285 if (unlikely(noreap
)) {
1288 int why
= (p
->ptrace
& PT_PTRACED
) ? CLD_TRAPPED
: CLD_STOPPED
;
1290 exit_code
= p
->exit_code
;
1291 if (unlikely(!exit_code
) ||
1292 unlikely(p
->state
& TASK_TRACED
))
1294 return wait_noreap_copyout(p
, pid
, uid
,
1295 why
, (exit_code
<< 8) | 0x7f,
1299 write_lock_irq(&tasklist_lock
);
1302 * This uses xchg to be atomic with the thread resuming and setting
1303 * it. It must also be done with the write lock held to prevent a
1304 * race with the EXIT_ZOMBIE case.
1306 exit_code
= xchg(&p
->exit_code
, 0);
1307 if (unlikely(p
->exit_state
)) {
1309 * The task resumed and then died. Let the next iteration
1310 * catch it in EXIT_ZOMBIE. Note that exit_code might
1311 * already be zero here if it resumed and did _exit(0).
1312 * The task itself is dead and won't touch exit_code again;
1313 * other processors in this function are locked out.
1315 p
->exit_code
= exit_code
;
1318 if (unlikely(exit_code
== 0)) {
1320 * Another thread in this function got to it first, or it
1321 * resumed, or it resumed and then died.
1323 write_unlock_irq(&tasklist_lock
);
1327 * We are returning to the wait loop without having successfully
1328 * removed the process and having released the lock. We cannot
1329 * continue, since the "p" task pointer is potentially stale.
1331 * Return -EAGAIN, and do_wait() will restart the loop from the
1332 * beginning. Do _not_ re-acquire the lock.
1337 /* move to end of parent's list to avoid starvation */
1341 write_unlock_irq(&tasklist_lock
);
1343 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1344 if (!retval
&& stat_addr
)
1345 retval
= put_user((exit_code
<< 8) | 0x7f, stat_addr
);
1346 if (!retval
&& infop
)
1347 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1348 if (!retval
&& infop
)
1349 retval
= put_user(0, &infop
->si_errno
);
1350 if (!retval
&& infop
)
1351 retval
= put_user((short)((p
->ptrace
& PT_PTRACED
)
1352 ? CLD_TRAPPED
: CLD_STOPPED
),
1354 if (!retval
&& infop
)
1355 retval
= put_user(exit_code
, &infop
->si_status
);
1356 if (!retval
&& infop
)
1357 retval
= put_user(p
->pid
, &infop
->si_pid
);
1358 if (!retval
&& infop
)
1359 retval
= put_user(p
->uid
, &infop
->si_uid
);
1369 * Handle do_wait work for one task in a live, non-stopped state.
1370 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1371 * the lock and this task is uninteresting. If we return nonzero, we have
1372 * released the lock and the system call should return.
1374 static int wait_task_continued(struct task_struct
*p
, int noreap
,
1375 struct siginfo __user
*infop
,
1376 int __user
*stat_addr
, struct rusage __user
*ru
)
1382 if (unlikely(!p
->signal
))
1385 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
))
1388 spin_lock_irq(&p
->sighand
->siglock
);
1389 /* Re-check with the lock held. */
1390 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
)) {
1391 spin_unlock_irq(&p
->sighand
->siglock
);
1395 p
->signal
->flags
&= ~SIGNAL_STOP_CONTINUED
;
1396 spin_unlock_irq(&p
->sighand
->siglock
);
1401 read_unlock(&tasklist_lock
);
1404 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1406 if (!retval
&& stat_addr
)
1407 retval
= put_user(0xffff, stat_addr
);
1411 retval
= wait_noreap_copyout(p
, pid
, uid
,
1412 CLD_CONTINUED
, SIGCONT
,
1414 BUG_ON(retval
== 0);
1421 static inline int my_ptrace_child(struct task_struct
*p
)
1423 if (!(p
->ptrace
& PT_PTRACED
))
1425 if (!(p
->ptrace
& PT_ATTACHED
))
1428 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1429 * we are the attacher. If we are the real parent, this is a race
1430 * inside ptrace_attach. It is waiting for the tasklist_lock,
1431 * which we have to switch the parent links, but has already set
1432 * the flags in p->ptrace.
1434 return (p
->parent
!= p
->real_parent
);
1437 static long do_wait(pid_t pid
, int options
, struct siginfo __user
*infop
,
1438 int __user
*stat_addr
, struct rusage __user
*ru
)
1440 DECLARE_WAITQUEUE(wait
, current
);
1441 struct task_struct
*tsk
;
1444 add_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1447 * We will set this flag if we see any child that might later
1448 * match our criteria, even if we are not able to reap it yet.
1451 current
->state
= TASK_INTERRUPTIBLE
;
1452 read_lock(&tasklist_lock
);
1455 struct task_struct
*p
;
1456 struct list_head
*_p
;
1459 list_for_each(_p
,&tsk
->children
) {
1460 p
= list_entry(_p
, struct task_struct
, sibling
);
1462 ret
= eligible_child(pid
, options
, p
);
1469 * When we hit the race with PTRACE_ATTACH,
1470 * we will not report this child. But the
1471 * race means it has not yet been moved to
1472 * our ptrace_children list, so we need to
1473 * set the flag here to avoid a spurious ECHILD
1474 * when the race happens with the only child.
1477 if (!my_ptrace_child(p
))
1482 * It's stopped now, so it might later
1483 * continue, exit, or stop again.
1486 if (!(options
& WUNTRACED
) &&
1487 !my_ptrace_child(p
))
1489 retval
= wait_task_stopped(p
, ret
== 2,
1490 (options
& WNOWAIT
),
1493 if (retval
== -EAGAIN
)
1495 if (retval
!= 0) /* He released the lock. */
1500 if (p
->exit_state
== EXIT_DEAD
)
1502 // case EXIT_ZOMBIE:
1503 if (p
->exit_state
== EXIT_ZOMBIE
) {
1505 * Eligible but we cannot release
1509 goto check_continued
;
1510 if (!likely(options
& WEXITED
))
1512 retval
= wait_task_zombie(
1513 p
, (options
& WNOWAIT
),
1514 infop
, stat_addr
, ru
);
1515 /* He released the lock. */
1522 * It's running now, so it might later
1523 * exit, stop, or stop and then continue.
1526 if (!unlikely(options
& WCONTINUED
))
1528 retval
= wait_task_continued(
1529 p
, (options
& WNOWAIT
),
1530 infop
, stat_addr
, ru
);
1531 if (retval
!= 0) /* He released the lock. */
1537 list_for_each(_p
, &tsk
->ptrace_children
) {
1538 p
= list_entry(_p
, struct task_struct
,
1540 if (!eligible_child(pid
, options
, p
))
1546 if (options
& __WNOTHREAD
)
1548 tsk
= next_thread(tsk
);
1549 BUG_ON(tsk
->signal
!= current
->signal
);
1550 } while (tsk
!= current
);
1552 read_unlock(&tasklist_lock
);
1555 if (options
& WNOHANG
)
1557 retval
= -ERESTARTSYS
;
1558 if (signal_pending(current
))
1565 current
->state
= TASK_RUNNING
;
1566 remove_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1572 * For a WNOHANG return, clear out all the fields
1573 * we would set so the user can easily tell the
1577 retval
= put_user(0, &infop
->si_signo
);
1579 retval
= put_user(0, &infop
->si_errno
);
1581 retval
= put_user(0, &infop
->si_code
);
1583 retval
= put_user(0, &infop
->si_pid
);
1585 retval
= put_user(0, &infop
->si_uid
);
1587 retval
= put_user(0, &infop
->si_status
);
1593 asmlinkage
long sys_waitid(int which
, pid_t pid
,
1594 struct siginfo __user
*infop
, int options
,
1595 struct rusage __user
*ru
)
1599 if (options
& ~(WNOHANG
|WNOWAIT
|WEXITED
|WSTOPPED
|WCONTINUED
))
1601 if (!(options
& (WEXITED
|WSTOPPED
|WCONTINUED
)))
1621 ret
= do_wait(pid
, options
, infop
, NULL
, ru
);
1623 /* avoid REGPARM breakage on x86: */
1624 prevent_tail_call(ret
);
1628 asmlinkage
long sys_wait4(pid_t pid
, int __user
*stat_addr
,
1629 int options
, struct rusage __user
*ru
)
1633 if (options
& ~(WNOHANG
|WUNTRACED
|WCONTINUED
|
1634 __WNOTHREAD
|__WCLONE
|__WALL
))
1636 ret
= do_wait(pid
, options
| WEXITED
, NULL
, stat_addr
, ru
);
1638 /* avoid REGPARM breakage on x86: */
1639 prevent_tail_call(ret
);
1643 #ifdef __ARCH_WANT_SYS_WAITPID
1646 * sys_waitpid() remains for compatibility. waitpid() should be
1647 * implemented by calling sys_wait4() from libc.a.
1649 asmlinkage
long sys_waitpid(pid_t pid
, int __user
*stat_addr
, int options
)
1651 return sys_wait4(pid
, stat_addr
, options
, NULL
);