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/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/binfmts.h>
23 #include <linux/nsproxy.h>
24 #include <linux/pid_namespace.h>
25 #include <linux/ptrace.h>
26 #include <linux/profile.h>
27 #include <linux/mount.h>
28 #include <linux/proc_fs.h>
29 #include <linux/kthread.h>
30 #include <linux/mempolicy.h>
31 #include <linux/taskstats_kern.h>
32 #include <linux/delayacct.h>
33 #include <linux/freezer.h>
34 #include <linux/cgroup.h>
35 #include <linux/syscalls.h>
36 #include <linux/signal.h>
37 #include <linux/posix-timers.h>
38 #include <linux/cn_proc.h>
39 #include <linux/mutex.h>
40 #include <linux/futex.h>
41 #include <linux/compat.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>
48 #include <asm/uaccess.h>
49 #include <asm/unistd.h>
50 #include <asm/pgtable.h>
51 #include <asm/mmu_context.h>
53 static void exit_mm(struct task_struct
* tsk
);
55 static void __unhash_process(struct task_struct
*p
)
58 detach_pid(p
, PIDTYPE_PID
);
59 if (thread_group_leader(p
)) {
60 detach_pid(p
, PIDTYPE_PGID
);
61 detach_pid(p
, PIDTYPE_SID
);
63 list_del_rcu(&p
->tasks
);
64 __get_cpu_var(process_counts
)--;
66 list_del_rcu(&p
->thread_group
);
71 * This function expects the tasklist_lock write-locked.
73 static void __exit_signal(struct task_struct
*tsk
)
75 struct signal_struct
*sig
= tsk
->signal
;
76 struct sighand_struct
*sighand
;
79 BUG_ON(!atomic_read(&sig
->count
));
82 sighand
= rcu_dereference(tsk
->sighand
);
83 spin_lock(&sighand
->siglock
);
85 posix_cpu_timers_exit(tsk
);
86 if (atomic_dec_and_test(&sig
->count
))
87 posix_cpu_timers_exit_group(tsk
);
90 * If there is any task waiting for the group exit
93 if (sig
->group_exit_task
&& atomic_read(&sig
->count
) == sig
->notify_count
)
94 wake_up_process(sig
->group_exit_task
);
96 if (tsk
== sig
->curr_target
)
97 sig
->curr_target
= next_thread(tsk
);
99 * Accumulate here the counters for all threads but the
100 * group leader as they die, so they can be added into
101 * the process-wide totals when those are taken.
102 * The group leader stays around as a zombie as long
103 * as there are other threads. When it gets reaped,
104 * the exit.c code will add its counts into these totals.
105 * We won't ever get here for the group leader, since it
106 * will have been the last reference on the signal_struct.
108 sig
->utime
= cputime_add(sig
->utime
, tsk
->utime
);
109 sig
->stime
= cputime_add(sig
->stime
, tsk
->stime
);
110 sig
->gtime
= cputime_add(sig
->gtime
, tsk
->gtime
);
111 sig
->min_flt
+= tsk
->min_flt
;
112 sig
->maj_flt
+= tsk
->maj_flt
;
113 sig
->nvcsw
+= tsk
->nvcsw
;
114 sig
->nivcsw
+= tsk
->nivcsw
;
115 sig
->inblock
+= task_io_get_inblock(tsk
);
116 sig
->oublock
+= task_io_get_oublock(tsk
);
117 sig
->sum_sched_runtime
+= tsk
->se
.sum_exec_runtime
;
118 sig
= NULL
; /* Marker for below. */
121 __unhash_process(tsk
);
125 spin_unlock(&sighand
->siglock
);
128 __cleanup_sighand(sighand
);
129 clear_tsk_thread_flag(tsk
,TIF_SIGPENDING
);
130 flush_sigqueue(&tsk
->pending
);
132 flush_sigqueue(&sig
->shared_pending
);
133 taskstats_tgid_free(sig
);
134 __cleanup_signal(sig
);
138 static void delayed_put_task_struct(struct rcu_head
*rhp
)
140 put_task_struct(container_of(rhp
, struct task_struct
, rcu
));
143 void release_task(struct task_struct
* p
)
145 struct task_struct
*leader
;
148 atomic_dec(&p
->user
->processes
);
150 write_lock_irq(&tasklist_lock
);
152 BUG_ON(!list_empty(&p
->ptrace_list
) || !list_empty(&p
->ptrace_children
));
156 * If we are the last non-leader member of the thread
157 * group, and the leader is zombie, then notify the
158 * group leader's parent process. (if it wants notification.)
161 leader
= p
->group_leader
;
162 if (leader
!= p
&& thread_group_empty(leader
) && leader
->exit_state
== EXIT_ZOMBIE
) {
163 BUG_ON(leader
->exit_signal
== -1);
164 do_notify_parent(leader
, leader
->exit_signal
);
166 * If we were the last child thread and the leader has
167 * exited already, and the leader's parent ignores SIGCHLD,
168 * then we are the one who should release the leader.
170 * do_notify_parent() will have marked it self-reaping in
173 zap_leader
= (leader
->exit_signal
== -1);
176 write_unlock_irq(&tasklist_lock
);
178 call_rcu(&p
->rcu
, delayed_put_task_struct
);
181 if (unlikely(zap_leader
))
186 * This checks not only the pgrp, but falls back on the pid if no
187 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
190 * The caller must hold rcu lock or the tasklist lock.
192 struct pid
*session_of_pgrp(struct pid
*pgrp
)
194 struct task_struct
*p
;
195 struct pid
*sid
= NULL
;
197 p
= pid_task(pgrp
, PIDTYPE_PGID
);
199 p
= pid_task(pgrp
, PIDTYPE_PID
);
201 sid
= task_session(p
);
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(struct pid
*pgrp
, struct task_struct
*ignored_task
)
216 struct task_struct
*p
;
218 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
219 if ((p
== ignored_task
) ||
220 (p
->exit_state
&& thread_group_empty(p
)) ||
221 is_global_init(p
->real_parent
))
224 if (task_pgrp(p
->real_parent
) != pgrp
&&
225 task_session(p
->real_parent
) == task_session(p
))
227 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
232 int is_current_pgrp_orphaned(void)
236 read_lock(&tasklist_lock
);
237 retval
= will_become_orphaned_pgrp(task_pgrp(current
), NULL
);
238 read_unlock(&tasklist_lock
);
243 static int has_stopped_jobs(struct pid
*pgrp
)
246 struct task_struct
*p
;
248 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
249 if (!task_is_stopped(p
))
253 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
258 * Check to see if any process groups have become orphaned as
259 * a result of our exiting, and if they have any stopped jobs,
260 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
263 kill_orphaned_pgrp(struct task_struct
*tsk
, struct task_struct
*parent
)
265 struct pid
*pgrp
= task_pgrp(tsk
);
266 struct task_struct
*ignored_task
= tsk
;
269 /* exit: our father is in a different pgrp than
270 * we are and we were the only connection outside.
272 parent
= tsk
->real_parent
;
274 /* reparent: our child is in a different pgrp than
275 * we are, and it was the only connection outside.
279 if (task_pgrp(parent
) != pgrp
&&
280 task_session(parent
) == task_session(tsk
) &&
281 will_become_orphaned_pgrp(pgrp
, ignored_task
) &&
282 has_stopped_jobs(pgrp
)) {
283 __kill_pgrp_info(SIGHUP
, SEND_SIG_PRIV
, pgrp
);
284 __kill_pgrp_info(SIGCONT
, SEND_SIG_PRIV
, pgrp
);
289 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
291 * If a kernel thread is launched as a result of a system call, or if
292 * it ever exits, it should generally reparent itself to kthreadd so it
293 * isn't in the way of other processes and is correctly cleaned up on exit.
295 * The various task state such as scheduling policy and priority may have
296 * been inherited from a user process, so we reset them to sane values here.
298 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
300 static void reparent_to_kthreadd(void)
302 write_lock_irq(&tasklist_lock
);
304 ptrace_unlink(current
);
305 /* Reparent to init */
306 remove_parent(current
);
307 current
->real_parent
= current
->parent
= kthreadd_task
;
310 /* Set the exit signal to SIGCHLD so we signal init on exit */
311 current
->exit_signal
= SIGCHLD
;
313 if (task_nice(current
) < 0)
314 set_user_nice(current
, 0);
318 security_task_reparent_to_init(current
);
319 memcpy(current
->signal
->rlim
, init_task
.signal
->rlim
,
320 sizeof(current
->signal
->rlim
));
321 atomic_inc(&(INIT_USER
->__count
));
322 write_unlock_irq(&tasklist_lock
);
323 switch_uid(INIT_USER
);
326 void __set_special_pids(struct pid
*pid
)
328 struct task_struct
*curr
= current
->group_leader
;
329 pid_t nr
= pid_nr(pid
);
331 if (task_session(curr
) != pid
) {
332 detach_pid(curr
, PIDTYPE_SID
);
333 attach_pid(curr
, PIDTYPE_SID
, pid
);
334 set_task_session(curr
, nr
);
336 if (task_pgrp(curr
) != pid
) {
337 detach_pid(curr
, PIDTYPE_PGID
);
338 attach_pid(curr
, PIDTYPE_PGID
, pid
);
339 set_task_pgrp(curr
, nr
);
343 static void set_special_pids(struct pid
*pid
)
345 write_lock_irq(&tasklist_lock
);
346 __set_special_pids(pid
);
347 write_unlock_irq(&tasklist_lock
);
351 * Let kernel threads use this to say that they
352 * allow a certain signal (since daemonize() will
353 * have disabled all of them by default).
355 int allow_signal(int sig
)
357 if (!valid_signal(sig
) || sig
< 1)
360 spin_lock_irq(¤t
->sighand
->siglock
);
361 sigdelset(¤t
->blocked
, sig
);
363 /* Kernel threads handle their own signals.
364 Let the signal code know it'll be handled, so
365 that they don't get converted to SIGKILL or
366 just silently dropped */
367 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= (void __user
*)2;
370 spin_unlock_irq(¤t
->sighand
->siglock
);
374 EXPORT_SYMBOL(allow_signal
);
376 int disallow_signal(int sig
)
378 if (!valid_signal(sig
) || sig
< 1)
381 spin_lock_irq(¤t
->sighand
->siglock
);
382 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= SIG_IGN
;
384 spin_unlock_irq(¤t
->sighand
->siglock
);
388 EXPORT_SYMBOL(disallow_signal
);
391 * Put all the gunge required to become a kernel thread without
392 * attached user resources in one place where it belongs.
395 void daemonize(const char *name
, ...)
398 struct fs_struct
*fs
;
401 va_start(args
, name
);
402 vsnprintf(current
->comm
, sizeof(current
->comm
), name
, args
);
406 * If we were started as result of loading a module, close all of the
407 * user space pages. We don't need them, and if we didn't close them
408 * they would be locked into memory.
412 * We don't want to have TIF_FREEZE set if the system-wide hibernation
413 * or suspend transition begins right now.
415 current
->flags
|= PF_NOFREEZE
;
417 if (current
->nsproxy
!= &init_nsproxy
) {
418 get_nsproxy(&init_nsproxy
);
419 switch_task_namespaces(current
, &init_nsproxy
);
421 set_special_pids(&init_struct_pid
);
422 proc_clear_tty(current
);
424 /* Block and flush all signals */
425 sigfillset(&blocked
);
426 sigprocmask(SIG_BLOCK
, &blocked
, NULL
);
427 flush_signals(current
);
429 /* Become as one with the init task */
431 exit_fs(current
); /* current->fs->count--; */
434 atomic_inc(&fs
->count
);
437 current
->files
= init_task
.files
;
438 atomic_inc(¤t
->files
->count
);
440 reparent_to_kthreadd();
443 EXPORT_SYMBOL(daemonize
);
445 static void close_files(struct files_struct
* files
)
453 * It is safe to dereference the fd table without RCU or
454 * ->file_lock because this is the last reference to the
457 fdt
= files_fdtable(files
);
461 if (i
>= fdt
->max_fds
)
463 set
= fdt
->open_fds
->fds_bits
[j
++];
466 struct file
* file
= xchg(&fdt
->fd
[i
], NULL
);
468 filp_close(file
, files
);
478 struct files_struct
*get_files_struct(struct task_struct
*task
)
480 struct files_struct
*files
;
485 atomic_inc(&files
->count
);
491 void put_files_struct(struct files_struct
*files
)
495 if (atomic_dec_and_test(&files
->count
)) {
498 * Free the fd and fdset arrays if we expanded them.
499 * If the fdtable was embedded, pass files for freeing
500 * at the end of the RCU grace period. Otherwise,
501 * you can free files immediately.
503 fdt
= files_fdtable(files
);
504 if (fdt
!= &files
->fdtab
)
505 kmem_cache_free(files_cachep
, files
);
510 EXPORT_SYMBOL(put_files_struct
);
512 void reset_files_struct(struct task_struct
*tsk
, struct files_struct
*files
)
514 struct files_struct
*old
;
520 put_files_struct(old
);
522 EXPORT_SYMBOL(reset_files_struct
);
524 void exit_files(struct task_struct
*tsk
)
526 struct files_struct
* files
= tsk
->files
;
532 put_files_struct(files
);
536 void put_fs_struct(struct fs_struct
*fs
)
538 /* No need to hold fs->lock if we are killing it */
539 if (atomic_dec_and_test(&fs
->count
)) {
542 if (fs
->altroot
.dentry
)
543 path_put(&fs
->altroot
);
544 kmem_cache_free(fs_cachep
, fs
);
548 void exit_fs(struct task_struct
*tsk
)
550 struct fs_struct
* fs
= tsk
->fs
;
560 EXPORT_SYMBOL_GPL(exit_fs
);
563 * Turn us into a lazy TLB process if we
566 static void exit_mm(struct task_struct
* tsk
)
568 struct mm_struct
*mm
= tsk
->mm
;
574 * Serialize with any possible pending coredump.
575 * We must hold mmap_sem around checking core_waiters
576 * and clearing tsk->mm. The core-inducing thread
577 * will increment core_waiters for each thread in the
578 * group with ->mm != NULL.
580 down_read(&mm
->mmap_sem
);
581 if (mm
->core_waiters
) {
582 up_read(&mm
->mmap_sem
);
583 down_write(&mm
->mmap_sem
);
584 if (!--mm
->core_waiters
)
585 complete(mm
->core_startup_done
);
586 up_write(&mm
->mmap_sem
);
588 wait_for_completion(&mm
->core_done
);
589 down_read(&mm
->mmap_sem
);
591 atomic_inc(&mm
->mm_count
);
592 BUG_ON(mm
!= tsk
->active_mm
);
593 /* more a memory barrier than a real lock */
596 up_read(&mm
->mmap_sem
);
597 enter_lazy_tlb(mm
, current
);
598 /* We don't want this task to be frozen prematurely */
599 clear_freeze_flag(tsk
);
605 reparent_thread(struct task_struct
*p
, struct task_struct
*father
, int traced
)
607 if (p
->pdeath_signal
)
608 /* We already hold the tasklist_lock here. */
609 group_send_sig_info(p
->pdeath_signal
, SEND_SIG_NOINFO
, p
);
611 /* Move the child from its dying parent to the new one. */
612 if (unlikely(traced
)) {
613 /* Preserve ptrace links if someone else is tracing this child. */
614 list_del_init(&p
->ptrace_list
);
615 if (p
->parent
!= p
->real_parent
)
616 list_add(&p
->ptrace_list
, &p
->real_parent
->ptrace_children
);
618 /* If this child is being traced, then we're the one tracing it
619 * anyway, so let go of it.
623 p
->parent
= p
->real_parent
;
626 if (task_is_traced(p
)) {
628 * If it was at a trace stop, turn it into
629 * a normal stop since it's no longer being
636 /* If this is a threaded reparent there is no need to
637 * notify anyone anything has happened.
639 if (p
->real_parent
->group_leader
== father
->group_leader
)
642 /* We don't want people slaying init. */
643 if (p
->exit_signal
!= -1)
644 p
->exit_signal
= SIGCHLD
;
646 /* If we'd notified the old parent about this child's death,
647 * also notify the new parent.
649 if (!traced
&& p
->exit_state
== EXIT_ZOMBIE
&&
650 p
->exit_signal
!= -1 && thread_group_empty(p
))
651 do_notify_parent(p
, p
->exit_signal
);
653 kill_orphaned_pgrp(p
, father
);
657 * When we die, we re-parent all our children.
658 * Try to give them to another thread in our thread
659 * group, and if no such member exists, give it to
660 * the child reaper process (ie "init") in our pid
663 static void forget_original_parent(struct task_struct
*father
)
665 struct task_struct
*p
, *n
, *reaper
= father
;
666 struct list_head ptrace_dead
;
668 INIT_LIST_HEAD(&ptrace_dead
);
670 write_lock_irq(&tasklist_lock
);
673 reaper
= next_thread(reaper
);
674 if (reaper
== father
) {
675 reaper
= task_child_reaper(father
);
678 } while (reaper
->flags
& PF_EXITING
);
681 * There are only two places where our children can be:
683 * - in our child list
684 * - in our ptraced child list
686 * Search them and reparent children.
688 list_for_each_entry_safe(p
, n
, &father
->children
, sibling
) {
693 /* if father isn't the real parent, then ptrace must be enabled */
694 BUG_ON(father
!= p
->real_parent
&& !ptrace
);
696 if (father
== p
->real_parent
) {
697 /* reparent with a reaper, real father it's us */
698 p
->real_parent
= reaper
;
699 reparent_thread(p
, father
, 0);
701 /* reparent ptraced task to its real parent */
703 if (p
->exit_state
== EXIT_ZOMBIE
&& p
->exit_signal
!= -1 &&
704 thread_group_empty(p
))
705 do_notify_parent(p
, p
->exit_signal
);
709 * if the ptraced child is a zombie with exit_signal == -1
710 * we must collect it before we exit, or it will remain
711 * zombie forever since we prevented it from self-reap itself
712 * while it was being traced by us, to be able to see it in wait4.
714 if (unlikely(ptrace
&& p
->exit_state
== EXIT_ZOMBIE
&& p
->exit_signal
== -1))
715 list_add(&p
->ptrace_list
, &ptrace_dead
);
718 list_for_each_entry_safe(p
, n
, &father
->ptrace_children
, ptrace_list
) {
719 p
->real_parent
= reaper
;
720 reparent_thread(p
, father
, 1);
723 write_unlock_irq(&tasklist_lock
);
724 BUG_ON(!list_empty(&father
->children
));
725 BUG_ON(!list_empty(&father
->ptrace_children
));
727 list_for_each_entry_safe(p
, n
, &ptrace_dead
, ptrace_list
) {
728 list_del_init(&p
->ptrace_list
);
735 * Send signals to all our closest relatives so that they know
736 * to properly mourn us..
738 static void exit_notify(struct task_struct
*tsk
, int group_dead
)
743 * This does two things:
745 * A. Make init inherit all the child processes
746 * B. Check to see if any process groups have become orphaned
747 * as a result of our exiting, and if they have any stopped
748 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
750 forget_original_parent(tsk
);
751 exit_task_namespaces(tsk
);
753 write_lock_irq(&tasklist_lock
);
755 kill_orphaned_pgrp(tsk
->group_leader
, NULL
);
757 /* Let father know we died
759 * Thread signals are configurable, but you aren't going to use
760 * that to send signals to arbitary processes.
761 * That stops right now.
763 * If the parent exec id doesn't match the exec id we saved
764 * when we started then we know the parent has changed security
767 * If our self_exec id doesn't match our parent_exec_id then
768 * we have changed execution domain as these two values started
769 * the same after a fork.
771 if (tsk
->exit_signal
!= SIGCHLD
&& tsk
->exit_signal
!= -1 &&
772 (tsk
->parent_exec_id
!= tsk
->real_parent
->self_exec_id
||
773 tsk
->self_exec_id
!= tsk
->parent_exec_id
)
774 && !capable(CAP_KILL
))
775 tsk
->exit_signal
= SIGCHLD
;
778 /* If something other than our normal parent is ptracing us, then
779 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
780 * only has special meaning to our real parent.
782 if (tsk
->exit_signal
!= -1 && thread_group_empty(tsk
)) {
783 int signal
= tsk
->parent
== tsk
->real_parent
? tsk
->exit_signal
: SIGCHLD
;
784 do_notify_parent(tsk
, signal
);
785 } else if (tsk
->ptrace
) {
786 do_notify_parent(tsk
, SIGCHLD
);
790 if (tsk
->exit_signal
== -1 && likely(!tsk
->ptrace
))
792 tsk
->exit_state
= state
;
794 if (thread_group_leader(tsk
) &&
795 tsk
->signal
->notify_count
< 0 &&
796 tsk
->signal
->group_exit_task
)
797 wake_up_process(tsk
->signal
->group_exit_task
);
799 write_unlock_irq(&tasklist_lock
);
801 /* If the process is dead, release it - nobody will wait for it */
802 if (state
== EXIT_DEAD
)
806 #ifdef CONFIG_DEBUG_STACK_USAGE
807 static void check_stack_usage(void)
809 static DEFINE_SPINLOCK(low_water_lock
);
810 static int lowest_to_date
= THREAD_SIZE
;
811 unsigned long *n
= end_of_stack(current
);
816 free
= (unsigned long)n
- (unsigned long)end_of_stack(current
);
818 if (free
>= lowest_to_date
)
821 spin_lock(&low_water_lock
);
822 if (free
< lowest_to_date
) {
823 printk(KERN_WARNING
"%s used greatest stack depth: %lu bytes "
825 current
->comm
, free
);
826 lowest_to_date
= free
;
828 spin_unlock(&low_water_lock
);
831 static inline void check_stack_usage(void) {}
834 static inline void exit_child_reaper(struct task_struct
*tsk
)
836 if (likely(tsk
->group_leader
!= task_child_reaper(tsk
)))
839 if (tsk
->nsproxy
->pid_ns
== &init_pid_ns
)
840 panic("Attempted to kill init!");
843 * @tsk is the last thread in the 'cgroup-init' and is exiting.
844 * Terminate all remaining processes in the namespace and reap them
845 * before exiting @tsk.
847 * Note that @tsk (last thread of cgroup-init) may not necessarily
848 * be the child-reaper (i.e main thread of cgroup-init) of the
849 * namespace i.e the child_reaper may have already exited.
851 * Even after a child_reaper exits, we let it inherit orphaned children,
852 * because, pid_ns->child_reaper remains valid as long as there is
853 * at least one living sub-thread in the cgroup init.
855 * This living sub-thread of the cgroup-init will be notified when
856 * a child inherited by the 'child-reaper' exits (do_notify_parent()
857 * uses __group_send_sig_info()). Further, when reaping child processes,
858 * do_wait() iterates over children of all living sub threads.
860 * i.e even though 'child_reaper' thread is listed as the parent of the
861 * orphaned children, any living sub-thread in the cgroup-init can
862 * perform the role of the child_reaper.
864 zap_pid_ns_processes(tsk
->nsproxy
->pid_ns
);
867 NORET_TYPE
void do_exit(long code
)
869 struct task_struct
*tsk
= current
;
872 profile_task_exit(tsk
);
874 WARN_ON(atomic_read(&tsk
->fs_excl
));
876 if (unlikely(in_interrupt()))
877 panic("Aiee, killing interrupt handler!");
878 if (unlikely(!tsk
->pid
))
879 panic("Attempted to kill the idle task!");
881 if (unlikely(current
->ptrace
& PT_TRACE_EXIT
)) {
882 current
->ptrace_message
= code
;
883 ptrace_notify((PTRACE_EVENT_EXIT
<< 8) | SIGTRAP
);
887 * We're taking recursive faults here in do_exit. Safest is to just
888 * leave this task alone and wait for reboot.
890 if (unlikely(tsk
->flags
& PF_EXITING
)) {
892 "Fixing recursive fault but reboot is needed!\n");
894 * We can do this unlocked here. The futex code uses
895 * this flag just to verify whether the pi state
896 * cleanup has been done or not. In the worst case it
897 * loops once more. We pretend that the cleanup was
898 * done as there is no way to return. Either the
899 * OWNER_DIED bit is set by now or we push the blocked
900 * task into the wait for ever nirwana as well.
902 tsk
->flags
|= PF_EXITPIDONE
;
905 set_current_state(TASK_UNINTERRUPTIBLE
);
909 exit_signals(tsk
); /* sets PF_EXITING */
911 * tsk->flags are checked in the futex code to protect against
912 * an exiting task cleaning up the robust pi futexes.
915 spin_unlock_wait(&tsk
->pi_lock
);
917 if (unlikely(in_atomic()))
918 printk(KERN_INFO
"note: %s[%d] exited with preempt_count %d\n",
919 current
->comm
, task_pid_nr(current
),
922 acct_update_integrals(tsk
);
924 update_hiwater_rss(tsk
->mm
);
925 update_hiwater_vm(tsk
->mm
);
927 group_dead
= atomic_dec_and_test(&tsk
->signal
->live
);
929 exit_child_reaper(tsk
);
930 hrtimer_cancel(&tsk
->signal
->real_timer
);
931 exit_itimers(tsk
->signal
);
933 acct_collect(code
, group_dead
);
935 if (unlikely(tsk
->robust_list
))
936 exit_robust_list(tsk
);
938 if (unlikely(tsk
->compat_robust_list
))
939 compat_exit_robust_list(tsk
);
944 if (unlikely(tsk
->audit_context
))
947 tsk
->exit_code
= code
;
948 taskstats_exit(tsk
, group_dead
);
962 if (group_dead
&& tsk
->signal
->leader
)
963 disassociate_ctty(1);
965 module_put(task_thread_info(tsk
)->exec_domain
->module
);
967 module_put(tsk
->binfmt
->module
);
969 proc_exit_connector(tsk
);
970 exit_notify(tsk
, group_dead
);
972 mpol_free(tsk
->mempolicy
);
973 tsk
->mempolicy
= NULL
;
977 * This must happen late, after the PID is not
980 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
981 exit_pi_state_list(tsk
);
982 if (unlikely(current
->pi_state_cache
))
983 kfree(current
->pi_state_cache
);
986 * Make sure we are holding no locks:
988 debug_check_no_locks_held(tsk
);
990 * We can do this unlocked here. The futex code uses this flag
991 * just to verify whether the pi state cleanup has been done
992 * or not. In the worst case it loops once more.
994 tsk
->flags
|= PF_EXITPIDONE
;
999 if (tsk
->splice_pipe
)
1000 __free_pipe_info(tsk
->splice_pipe
);
1003 /* causes final put_task_struct in finish_task_switch(). */
1004 tsk
->state
= TASK_DEAD
;
1008 /* Avoid "noreturn function does return". */
1010 cpu_relax(); /* For when BUG is null */
1013 EXPORT_SYMBOL_GPL(do_exit
);
1015 NORET_TYPE
void complete_and_exit(struct completion
*comp
, long code
)
1023 EXPORT_SYMBOL(complete_and_exit
);
1025 asmlinkage
long sys_exit(int error_code
)
1027 do_exit((error_code
&0xff)<<8);
1031 * Take down every thread in the group. This is called by fatal signals
1032 * as well as by sys_exit_group (below).
1035 do_group_exit(int exit_code
)
1037 BUG_ON(exit_code
& 0x80); /* core dumps don't get here */
1039 if (current
->signal
->flags
& SIGNAL_GROUP_EXIT
)
1040 exit_code
= current
->signal
->group_exit_code
;
1041 else if (!thread_group_empty(current
)) {
1042 struct signal_struct
*const sig
= current
->signal
;
1043 struct sighand_struct
*const sighand
= current
->sighand
;
1044 spin_lock_irq(&sighand
->siglock
);
1045 if (signal_group_exit(sig
))
1046 /* Another thread got here before we took the lock. */
1047 exit_code
= sig
->group_exit_code
;
1049 sig
->group_exit_code
= exit_code
;
1050 sig
->flags
= SIGNAL_GROUP_EXIT
;
1051 zap_other_threads(current
);
1053 spin_unlock_irq(&sighand
->siglock
);
1061 * this kills every thread in the thread group. Note that any externally
1062 * wait4()-ing process will get the correct exit code - even if this
1063 * thread is not the thread group leader.
1065 asmlinkage
void sys_exit_group(int error_code
)
1067 do_group_exit((error_code
& 0xff) << 8);
1070 static struct pid
*task_pid_type(struct task_struct
*task
, enum pid_type type
)
1072 struct pid
*pid
= NULL
;
1073 if (type
== PIDTYPE_PID
)
1074 pid
= task
->pids
[type
].pid
;
1075 else if (type
< PIDTYPE_MAX
)
1076 pid
= task
->group_leader
->pids
[type
].pid
;
1080 static int eligible_child(enum pid_type type
, struct pid
*pid
, int options
,
1081 struct task_struct
*p
)
1085 if (type
< PIDTYPE_MAX
) {
1086 if (task_pid_type(p
, type
) != pid
)
1091 * Do not consider detached threads that are
1094 if (p
->exit_signal
== -1 && !p
->ptrace
)
1097 /* Wait for all children (clone and not) if __WALL is set;
1098 * otherwise, wait for clone children *only* if __WCLONE is
1099 * set; otherwise, wait for non-clone children *only*. (Note:
1100 * A "clone" child here is one that reports to its parent
1101 * using a signal other than SIGCHLD.) */
1102 if (((p
->exit_signal
!= SIGCHLD
) ^ ((options
& __WCLONE
) != 0))
1103 && !(options
& __WALL
))
1106 err
= security_task_wait(p
);
1110 if (type
!= PIDTYPE_PID
)
1112 /* This child was explicitly requested, abort */
1113 read_unlock(&tasklist_lock
);
1117 static int wait_noreap_copyout(struct task_struct
*p
, pid_t pid
, uid_t uid
,
1118 int why
, int status
,
1119 struct siginfo __user
*infop
,
1120 struct rusage __user
*rusagep
)
1122 int retval
= rusagep
? getrusage(p
, RUSAGE_BOTH
, rusagep
) : 0;
1126 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1128 retval
= put_user(0, &infop
->si_errno
);
1130 retval
= put_user((short)why
, &infop
->si_code
);
1132 retval
= put_user(pid
, &infop
->si_pid
);
1134 retval
= put_user(uid
, &infop
->si_uid
);
1136 retval
= put_user(status
, &infop
->si_status
);
1143 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1144 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1145 * the lock and this task is uninteresting. If we return nonzero, we have
1146 * released the lock and the system call should return.
1148 static int wait_task_zombie(struct task_struct
*p
, int noreap
,
1149 struct siginfo __user
*infop
,
1150 int __user
*stat_addr
, struct rusage __user
*ru
)
1152 unsigned long state
;
1153 int retval
, status
, traced
;
1154 pid_t pid
= task_pid_vnr(p
);
1156 if (unlikely(noreap
)) {
1158 int exit_code
= p
->exit_code
;
1162 read_unlock(&tasklist_lock
);
1163 if ((exit_code
& 0x7f) == 0) {
1165 status
= exit_code
>> 8;
1167 why
= (exit_code
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1168 status
= exit_code
& 0x7f;
1170 return wait_noreap_copyout(p
, pid
, uid
, why
,
1175 * Try to move the task's state to DEAD
1176 * only one thread is allowed to do this:
1178 state
= xchg(&p
->exit_state
, EXIT_DEAD
);
1179 if (state
!= EXIT_ZOMBIE
) {
1180 BUG_ON(state
!= EXIT_DEAD
);
1184 /* traced means p->ptrace, but not vice versa */
1185 traced
= (p
->real_parent
!= p
->parent
);
1187 if (likely(!traced
)) {
1188 struct signal_struct
*psig
;
1189 struct signal_struct
*sig
;
1192 * The resource counters for the group leader are in its
1193 * own task_struct. Those for dead threads in the group
1194 * are in its signal_struct, as are those for the child
1195 * processes it has previously reaped. All these
1196 * accumulate in the parent's signal_struct c* fields.
1198 * We don't bother to take a lock here to protect these
1199 * p->signal fields, because they are only touched by
1200 * __exit_signal, which runs with tasklist_lock
1201 * write-locked anyway, and so is excluded here. We do
1202 * need to protect the access to p->parent->signal fields,
1203 * as other threads in the parent group can be right
1204 * here reaping other children at the same time.
1206 spin_lock_irq(&p
->parent
->sighand
->siglock
);
1207 psig
= p
->parent
->signal
;
1210 cputime_add(psig
->cutime
,
1211 cputime_add(p
->utime
,
1212 cputime_add(sig
->utime
,
1215 cputime_add(psig
->cstime
,
1216 cputime_add(p
->stime
,
1217 cputime_add(sig
->stime
,
1220 cputime_add(psig
->cgtime
,
1221 cputime_add(p
->gtime
,
1222 cputime_add(sig
->gtime
,
1225 p
->min_flt
+ sig
->min_flt
+ sig
->cmin_flt
;
1227 p
->maj_flt
+ sig
->maj_flt
+ sig
->cmaj_flt
;
1229 p
->nvcsw
+ sig
->nvcsw
+ sig
->cnvcsw
;
1231 p
->nivcsw
+ sig
->nivcsw
+ sig
->cnivcsw
;
1233 task_io_get_inblock(p
) +
1234 sig
->inblock
+ sig
->cinblock
;
1236 task_io_get_oublock(p
) +
1237 sig
->oublock
+ sig
->coublock
;
1238 spin_unlock_irq(&p
->parent
->sighand
->siglock
);
1242 * Now we are sure this task is interesting, and no other
1243 * thread can reap it because we set its state to EXIT_DEAD.
1245 read_unlock(&tasklist_lock
);
1247 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1248 status
= (p
->signal
->flags
& SIGNAL_GROUP_EXIT
)
1249 ? p
->signal
->group_exit_code
: p
->exit_code
;
1250 if (!retval
&& stat_addr
)
1251 retval
= put_user(status
, stat_addr
);
1252 if (!retval
&& infop
)
1253 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1254 if (!retval
&& infop
)
1255 retval
= put_user(0, &infop
->si_errno
);
1256 if (!retval
&& infop
) {
1259 if ((status
& 0x7f) == 0) {
1263 why
= (status
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1266 retval
= put_user((short)why
, &infop
->si_code
);
1268 retval
= put_user(status
, &infop
->si_status
);
1270 if (!retval
&& infop
)
1271 retval
= put_user(pid
, &infop
->si_pid
);
1272 if (!retval
&& infop
)
1273 retval
= put_user(p
->uid
, &infop
->si_uid
);
1278 write_lock_irq(&tasklist_lock
);
1279 /* We dropped tasklist, ptracer could die and untrace */
1282 * If this is not a detached task, notify the parent.
1283 * If it's still not detached after that, don't release
1286 if (p
->exit_signal
!= -1) {
1287 do_notify_parent(p
, p
->exit_signal
);
1288 if (p
->exit_signal
!= -1) {
1289 p
->exit_state
= EXIT_ZOMBIE
;
1293 write_unlock_irq(&tasklist_lock
);
1302 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1303 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1304 * the lock and this task is uninteresting. If we return nonzero, we have
1305 * released the lock and the system call should return.
1307 static int wait_task_stopped(struct task_struct
*p
,
1308 int noreap
, struct siginfo __user
*infop
,
1309 int __user
*stat_addr
, struct rusage __user
*ru
)
1311 int retval
, exit_code
, why
;
1312 uid_t uid
= 0; /* unneeded, required by compiler */
1316 spin_lock_irq(&p
->sighand
->siglock
);
1318 if (unlikely(!task_is_stopped_or_traced(p
)))
1321 if (!(p
->ptrace
& PT_PTRACED
) && p
->signal
->group_stop_count
> 0)
1323 * A group stop is in progress and this is the group leader.
1324 * We won't report until all threads have stopped.
1328 exit_code
= p
->exit_code
;
1337 spin_unlock_irq(&p
->sighand
->siglock
);
1342 * Now we are pretty sure this task is interesting.
1343 * Make sure it doesn't get reaped out from under us while we
1344 * give up the lock and then examine it below. We don't want to
1345 * keep holding onto the tasklist_lock while we call getrusage and
1346 * possibly take page faults for user memory.
1349 pid
= task_pid_vnr(p
);
1350 why
= (p
->ptrace
& PT_PTRACED
) ? CLD_TRAPPED
: CLD_STOPPED
;
1351 read_unlock(&tasklist_lock
);
1353 if (unlikely(noreap
))
1354 return wait_noreap_copyout(p
, pid
, uid
,
1358 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1359 if (!retval
&& stat_addr
)
1360 retval
= put_user((exit_code
<< 8) | 0x7f, stat_addr
);
1361 if (!retval
&& infop
)
1362 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1363 if (!retval
&& infop
)
1364 retval
= put_user(0, &infop
->si_errno
);
1365 if (!retval
&& infop
)
1366 retval
= put_user((short)why
, &infop
->si_code
);
1367 if (!retval
&& infop
)
1368 retval
= put_user(exit_code
, &infop
->si_status
);
1369 if (!retval
&& infop
)
1370 retval
= put_user(pid
, &infop
->si_pid
);
1371 if (!retval
&& infop
)
1372 retval
= put_user(uid
, &infop
->si_uid
);
1382 * Handle do_wait work for one task in a live, non-stopped state.
1383 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1384 * the lock and this task is uninteresting. If we return nonzero, we have
1385 * released the lock and the system call should return.
1387 static int wait_task_continued(struct task_struct
*p
, int noreap
,
1388 struct siginfo __user
*infop
,
1389 int __user
*stat_addr
, struct rusage __user
*ru
)
1395 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
))
1398 spin_lock_irq(&p
->sighand
->siglock
);
1399 /* Re-check with the lock held. */
1400 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
)) {
1401 spin_unlock_irq(&p
->sighand
->siglock
);
1405 p
->signal
->flags
&= ~SIGNAL_STOP_CONTINUED
;
1406 spin_unlock_irq(&p
->sighand
->siglock
);
1408 pid
= task_pid_vnr(p
);
1411 read_unlock(&tasklist_lock
);
1414 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1416 if (!retval
&& stat_addr
)
1417 retval
= put_user(0xffff, stat_addr
);
1421 retval
= wait_noreap_copyout(p
, pid
, uid
,
1422 CLD_CONTINUED
, SIGCONT
,
1424 BUG_ON(retval
== 0);
1430 static long do_wait(enum pid_type type
, struct pid
*pid
, int options
,
1431 struct siginfo __user
*infop
, int __user
*stat_addr
,
1432 struct rusage __user
*ru
)
1434 DECLARE_WAITQUEUE(wait
, current
);
1435 struct task_struct
*tsk
;
1438 add_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1440 /* If there is nothing that can match our critier just get out */
1442 if ((type
< PIDTYPE_MAX
) && (!pid
|| hlist_empty(&pid
->tasks
[type
])))
1446 * We will set this flag if we see any child that might later
1447 * match our criteria, even if we are not able to reap it yet.
1450 current
->state
= TASK_INTERRUPTIBLE
;
1451 read_lock(&tasklist_lock
);
1454 struct task_struct
*p
;
1456 list_for_each_entry(p
, &tsk
->children
, sibling
) {
1457 int ret
= eligible_child(type
, pid
, options
, p
);
1461 if (unlikely(ret
< 0)) {
1463 } else if (task_is_stopped_or_traced(p
)) {
1465 * It's stopped now, so it might later
1466 * continue, exit, or stop again.
1469 if (!(p
->ptrace
& PT_PTRACED
) &&
1470 !(options
& WUNTRACED
))
1473 retval
= wait_task_stopped(p
,
1474 (options
& WNOWAIT
), infop
,
1476 } else if (p
->exit_state
== EXIT_ZOMBIE
&&
1477 !delay_group_leader(p
)) {
1479 * We don't reap group leaders with subthreads.
1481 if (!likely(options
& WEXITED
))
1483 retval
= wait_task_zombie(p
,
1484 (options
& WNOWAIT
), infop
,
1486 } else if (p
->exit_state
!= EXIT_DEAD
) {
1488 * It's running now, so it might later
1489 * exit, stop, or stop and then continue.
1492 if (!unlikely(options
& WCONTINUED
))
1494 retval
= wait_task_continued(p
,
1495 (options
& WNOWAIT
), infop
,
1498 if (retval
!= 0) /* tasklist_lock released */
1502 list_for_each_entry(p
, &tsk
->ptrace_children
,
1504 flag
= eligible_child(type
, pid
, options
, p
);
1507 if (likely(flag
> 0))
1513 if (options
& __WNOTHREAD
)
1515 tsk
= next_thread(tsk
);
1516 BUG_ON(tsk
->signal
!= current
->signal
);
1517 } while (tsk
!= current
);
1518 read_unlock(&tasklist_lock
);
1521 if (options
& WNOHANG
)
1523 retval
= -ERESTARTSYS
;
1524 if (signal_pending(current
))
1531 current
->state
= TASK_RUNNING
;
1532 remove_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1538 * For a WNOHANG return, clear out all the fields
1539 * we would set so the user can easily tell the
1543 retval
= put_user(0, &infop
->si_signo
);
1545 retval
= put_user(0, &infop
->si_errno
);
1547 retval
= put_user(0, &infop
->si_code
);
1549 retval
= put_user(0, &infop
->si_pid
);
1551 retval
= put_user(0, &infop
->si_uid
);
1553 retval
= put_user(0, &infop
->si_status
);
1559 asmlinkage
long sys_waitid(int which
, pid_t upid
,
1560 struct siginfo __user
*infop
, int options
,
1561 struct rusage __user
*ru
)
1563 struct pid
*pid
= NULL
;
1567 if (options
& ~(WNOHANG
|WNOWAIT
|WEXITED
|WSTOPPED
|WCONTINUED
))
1569 if (!(options
& (WEXITED
|WSTOPPED
|WCONTINUED
)))
1582 type
= PIDTYPE_PGID
;
1590 if (type
< PIDTYPE_MAX
)
1591 pid
= find_get_pid(upid
);
1592 ret
= do_wait(type
, pid
, options
, infop
, NULL
, ru
);
1595 /* avoid REGPARM breakage on x86: */
1596 asmlinkage_protect(5, ret
, which
, upid
, infop
, options
, ru
);
1600 asmlinkage
long sys_wait4(pid_t upid
, int __user
*stat_addr
,
1601 int options
, struct rusage __user
*ru
)
1603 struct pid
*pid
= NULL
;
1607 if (options
& ~(WNOHANG
|WUNTRACED
|WCONTINUED
|
1608 __WNOTHREAD
|__WCLONE
|__WALL
))
1613 else if (upid
< 0) {
1614 type
= PIDTYPE_PGID
;
1615 pid
= find_get_pid(-upid
);
1616 } else if (upid
== 0) {
1617 type
= PIDTYPE_PGID
;
1618 pid
= get_pid(task_pgrp(current
));
1619 } else /* upid > 0 */ {
1621 pid
= find_get_pid(upid
);
1624 ret
= do_wait(type
, pid
, options
| WEXITED
, NULL
, stat_addr
, ru
);
1627 /* avoid REGPARM breakage on x86: */
1628 asmlinkage_protect(4, ret
, upid
, stat_addr
, options
, ru
);
1632 #ifdef __ARCH_WANT_SYS_WAITPID
1635 * sys_waitpid() remains for compatibility. waitpid() should be
1636 * implemented by calling sys_wait4() from libc.a.
1638 asmlinkage
long sys_waitpid(pid_t pid
, int __user
*stat_addr
, int options
)
1640 return sys_wait4(pid
, stat_addr
, options
, NULL
);