]>
git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - kernel/exit.c
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/pipe_fs_i.h>
44 #include <linux/audit.h> /* for audit_free() */
45 #include <linux/resource.h>
46 #include <linux/blkdev.h>
47 #include <linux/task_io_accounting_ops.h>
48 #include <linux/tracehook.h>
49 #include <linux/fs_struct.h>
50 #include <linux/init_task.h>
51 #include <trace/events/sched.h>
53 #include <asm/uaccess.h>
54 #include <asm/unistd.h>
55 #include <asm/pgtable.h>
56 #include <asm/mmu_context.h>
57 #include "cred-internals.h"
59 static void exit_mm(struct task_struct
* tsk
);
61 static void __unhash_process(struct task_struct
*p
)
64 detach_pid(p
, PIDTYPE_PID
);
65 if (thread_group_leader(p
)) {
66 detach_pid(p
, PIDTYPE_PGID
);
67 detach_pid(p
, PIDTYPE_SID
);
69 list_del_rcu(&p
->tasks
);
70 __get_cpu_var(process_counts
)--;
72 list_del_rcu(&p
->thread_group
);
73 list_del_init(&p
->sibling
);
77 * This function expects the tasklist_lock write-locked.
79 static void __exit_signal(struct task_struct
*tsk
)
81 struct signal_struct
*sig
= tsk
->signal
;
82 struct sighand_struct
*sighand
;
85 BUG_ON(!atomic_read(&sig
->count
));
87 sighand
= rcu_dereference(tsk
->sighand
);
88 spin_lock(&sighand
->siglock
);
90 posix_cpu_timers_exit(tsk
);
91 if (atomic_dec_and_test(&sig
->count
))
92 posix_cpu_timers_exit_group(tsk
);
95 * If there is any task waiting for the group exit
98 if (sig
->group_exit_task
&& atomic_read(&sig
->count
) == sig
->notify_count
)
99 wake_up_process(sig
->group_exit_task
);
101 if (tsk
== sig
->curr_target
)
102 sig
->curr_target
= next_thread(tsk
);
104 * Accumulate here the counters for all threads but the
105 * group leader as they die, so they can be added into
106 * the process-wide totals when those are taken.
107 * The group leader stays around as a zombie as long
108 * as there are other threads. When it gets reaped,
109 * the exit.c code will add its counts into these totals.
110 * We won't ever get here for the group leader, since it
111 * will have been the last reference on the signal_struct.
113 sig
->utime
= cputime_add(sig
->utime
, task_utime(tsk
));
114 sig
->stime
= cputime_add(sig
->stime
, task_stime(tsk
));
115 sig
->gtime
= cputime_add(sig
->gtime
, task_gtime(tsk
));
116 sig
->min_flt
+= tsk
->min_flt
;
117 sig
->maj_flt
+= tsk
->maj_flt
;
118 sig
->nvcsw
+= tsk
->nvcsw
;
119 sig
->nivcsw
+= tsk
->nivcsw
;
120 sig
->inblock
+= task_io_get_inblock(tsk
);
121 sig
->oublock
+= task_io_get_oublock(tsk
);
122 task_io_accounting_add(&sig
->ioac
, &tsk
->ioac
);
123 sig
->sum_sched_runtime
+= tsk
->se
.sum_exec_runtime
;
124 sig
= NULL
; /* Marker for below. */
127 __unhash_process(tsk
);
130 * Do this under ->siglock, we can race with another thread
131 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
133 flush_sigqueue(&tsk
->pending
);
137 spin_unlock(&sighand
->siglock
);
139 __cleanup_sighand(sighand
);
140 clear_tsk_thread_flag(tsk
,TIF_SIGPENDING
);
142 flush_sigqueue(&sig
->shared_pending
);
143 taskstats_tgid_free(sig
);
145 * Make sure ->signal can't go away under rq->lock,
146 * see account_group_exec_runtime().
148 task_rq_unlock_wait(tsk
);
149 __cleanup_signal(sig
);
153 static void delayed_put_task_struct(struct rcu_head
*rhp
)
155 struct task_struct
*tsk
= container_of(rhp
, struct task_struct
, rcu
);
157 trace_sched_process_free(tsk
);
158 put_task_struct(tsk
);
162 void release_task(struct task_struct
* p
)
164 struct task_struct
*leader
;
167 tracehook_prepare_release_task(p
);
168 /* don't need to get the RCU readlock here - the process is dead and
169 * can't be modifying its own credentials */
170 atomic_dec(&__task_cred(p
)->user
->processes
);
173 write_lock_irq(&tasklist_lock
);
174 tracehook_finish_release_task(p
);
178 * If we are the last non-leader member of the thread
179 * group, and the leader is zombie, then notify the
180 * group leader's parent process. (if it wants notification.)
183 leader
= p
->group_leader
;
184 if (leader
!= p
&& thread_group_empty(leader
) && leader
->exit_state
== EXIT_ZOMBIE
) {
185 BUG_ON(task_detached(leader
));
186 do_notify_parent(leader
, leader
->exit_signal
);
188 * If we were the last child thread and the leader has
189 * exited already, and the leader's parent ignores SIGCHLD,
190 * then we are the one who should release the leader.
192 * do_notify_parent() will have marked it self-reaping in
195 zap_leader
= task_detached(leader
);
198 * This maintains the invariant that release_task()
199 * only runs on a task in EXIT_DEAD, just for sanity.
202 leader
->exit_state
= EXIT_DEAD
;
205 write_unlock_irq(&tasklist_lock
);
207 call_rcu(&p
->rcu
, delayed_put_task_struct
);
210 if (unlikely(zap_leader
))
215 * This checks not only the pgrp, but falls back on the pid if no
216 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
219 * The caller must hold rcu lock or the tasklist lock.
221 struct pid
*session_of_pgrp(struct pid
*pgrp
)
223 struct task_struct
*p
;
224 struct pid
*sid
= NULL
;
226 p
= pid_task(pgrp
, PIDTYPE_PGID
);
228 p
= pid_task(pgrp
, PIDTYPE_PID
);
230 sid
= task_session(p
);
236 * Determine if a process group is "orphaned", according to the POSIX
237 * definition in 2.2.2.52. Orphaned process groups are not to be affected
238 * by terminal-generated stop signals. Newly orphaned process groups are
239 * to receive a SIGHUP and a SIGCONT.
241 * "I ask you, have you ever known what it is to be an orphan?"
243 static int will_become_orphaned_pgrp(struct pid
*pgrp
, struct task_struct
*ignored_task
)
245 struct task_struct
*p
;
247 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
248 if ((p
== ignored_task
) ||
249 (p
->exit_state
&& thread_group_empty(p
)) ||
250 is_global_init(p
->real_parent
))
253 if (task_pgrp(p
->real_parent
) != pgrp
&&
254 task_session(p
->real_parent
) == task_session(p
))
256 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
261 int is_current_pgrp_orphaned(void)
265 read_lock(&tasklist_lock
);
266 retval
= will_become_orphaned_pgrp(task_pgrp(current
), NULL
);
267 read_unlock(&tasklist_lock
);
272 static int has_stopped_jobs(struct pid
*pgrp
)
275 struct task_struct
*p
;
277 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
278 if (!task_is_stopped(p
))
282 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
287 * Check to see if any process groups have become orphaned as
288 * a result of our exiting, and if they have any stopped jobs,
289 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
292 kill_orphaned_pgrp(struct task_struct
*tsk
, struct task_struct
*parent
)
294 struct pid
*pgrp
= task_pgrp(tsk
);
295 struct task_struct
*ignored_task
= tsk
;
298 /* exit: our father is in a different pgrp than
299 * we are and we were the only connection outside.
301 parent
= tsk
->real_parent
;
303 /* reparent: our child is in a different pgrp than
304 * we are, and it was the only connection outside.
308 if (task_pgrp(parent
) != pgrp
&&
309 task_session(parent
) == task_session(tsk
) &&
310 will_become_orphaned_pgrp(pgrp
, ignored_task
) &&
311 has_stopped_jobs(pgrp
)) {
312 __kill_pgrp_info(SIGHUP
, SEND_SIG_PRIV
, pgrp
);
313 __kill_pgrp_info(SIGCONT
, SEND_SIG_PRIV
, pgrp
);
318 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
320 * If a kernel thread is launched as a result of a system call, or if
321 * it ever exits, it should generally reparent itself to kthreadd so it
322 * isn't in the way of other processes and is correctly cleaned up on exit.
324 * The various task state such as scheduling policy and priority may have
325 * been inherited from a user process, so we reset them to sane values here.
327 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
329 static void reparent_to_kthreadd(void)
331 write_lock_irq(&tasklist_lock
);
333 ptrace_unlink(current
);
334 /* Reparent to init */
335 current
->real_parent
= current
->parent
= kthreadd_task
;
336 list_move_tail(¤t
->sibling
, ¤t
->real_parent
->children
);
338 /* Set the exit signal to SIGCHLD so we signal init on exit */
339 current
->exit_signal
= SIGCHLD
;
341 if (task_nice(current
) < 0)
342 set_user_nice(current
, 0);
346 memcpy(current
->signal
->rlim
, init_task
.signal
->rlim
,
347 sizeof(current
->signal
->rlim
));
349 atomic_inc(&init_cred
.usage
);
350 commit_creds(&init_cred
);
351 write_unlock_irq(&tasklist_lock
);
354 void __set_special_pids(struct pid
*pid
)
356 struct task_struct
*curr
= current
->group_leader
;
358 if (task_session(curr
) != pid
)
359 change_pid(curr
, PIDTYPE_SID
, pid
);
361 if (task_pgrp(curr
) != pid
)
362 change_pid(curr
, PIDTYPE_PGID
, pid
);
365 static void set_special_pids(struct pid
*pid
)
367 write_lock_irq(&tasklist_lock
);
368 __set_special_pids(pid
);
369 write_unlock_irq(&tasklist_lock
);
373 * Let kernel threads use this to say that they
374 * allow a certain signal (since daemonize() will
375 * have disabled all of them by default).
377 int allow_signal(int sig
)
379 if (!valid_signal(sig
) || sig
< 1)
382 spin_lock_irq(¤t
->sighand
->siglock
);
383 sigdelset(¤t
->blocked
, sig
);
385 /* Kernel threads handle their own signals.
386 Let the signal code know it'll be handled, so
387 that they don't get converted to SIGKILL or
388 just silently dropped */
389 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= (void __user
*)2;
392 spin_unlock_irq(¤t
->sighand
->siglock
);
396 EXPORT_SYMBOL(allow_signal
);
398 int disallow_signal(int sig
)
400 if (!valid_signal(sig
) || sig
< 1)
403 spin_lock_irq(¤t
->sighand
->siglock
);
404 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= SIG_IGN
;
406 spin_unlock_irq(¤t
->sighand
->siglock
);
410 EXPORT_SYMBOL(disallow_signal
);
413 * Put all the gunge required to become a kernel thread without
414 * attached user resources in one place where it belongs.
417 void daemonize(const char *name
, ...)
422 va_start(args
, name
);
423 vsnprintf(current
->comm
, sizeof(current
->comm
), name
, args
);
427 * If we were started as result of loading a module, close all of the
428 * user space pages. We don't need them, and if we didn't close them
429 * they would be locked into memory.
433 * We don't want to have TIF_FREEZE set if the system-wide hibernation
434 * or suspend transition begins right now.
436 current
->flags
|= (PF_NOFREEZE
| PF_KTHREAD
);
438 if (current
->nsproxy
!= &init_nsproxy
) {
439 get_nsproxy(&init_nsproxy
);
440 switch_task_namespaces(current
, &init_nsproxy
);
442 set_special_pids(&init_struct_pid
);
443 proc_clear_tty(current
);
445 /* Block and flush all signals */
446 sigfillset(&blocked
);
447 sigprocmask(SIG_BLOCK
, &blocked
, NULL
);
448 flush_signals(current
);
450 /* Become as one with the init task */
452 daemonize_fs_struct();
454 current
->files
= init_task
.files
;
455 atomic_inc(¤t
->files
->count
);
457 reparent_to_kthreadd();
460 EXPORT_SYMBOL(daemonize
);
462 static void close_files(struct files_struct
* files
)
470 * It is safe to dereference the fd table without RCU or
471 * ->file_lock because this is the last reference to the
474 fdt
= files_fdtable(files
);
478 if (i
>= fdt
->max_fds
)
480 set
= fdt
->open_fds
->fds_bits
[j
++];
483 struct file
* file
= xchg(&fdt
->fd
[i
], NULL
);
485 filp_close(file
, files
);
495 struct files_struct
*get_files_struct(struct task_struct
*task
)
497 struct files_struct
*files
;
502 atomic_inc(&files
->count
);
508 void put_files_struct(struct files_struct
*files
)
512 if (atomic_dec_and_test(&files
->count
)) {
515 * Free the fd and fdset arrays if we expanded them.
516 * If the fdtable was embedded, pass files for freeing
517 * at the end of the RCU grace period. Otherwise,
518 * you can free files immediately.
520 fdt
= files_fdtable(files
);
521 if (fdt
!= &files
->fdtab
)
522 kmem_cache_free(files_cachep
, files
);
527 void reset_files_struct(struct files_struct
*files
)
529 struct task_struct
*tsk
= current
;
530 struct files_struct
*old
;
536 put_files_struct(old
);
539 void exit_files(struct task_struct
*tsk
)
541 struct files_struct
* files
= tsk
->files
;
547 put_files_struct(files
);
551 #ifdef CONFIG_MM_OWNER
553 * Task p is exiting and it owned mm, lets find a new owner for it
556 mm_need_new_owner(struct mm_struct
*mm
, struct task_struct
*p
)
559 * If there are other users of the mm and the owner (us) is exiting
560 * we need to find a new owner to take on the responsibility.
562 if (atomic_read(&mm
->mm_users
) <= 1)
569 void mm_update_next_owner(struct mm_struct
*mm
)
571 struct task_struct
*c
, *g
, *p
= current
;
574 if (!mm_need_new_owner(mm
, p
))
577 read_lock(&tasklist_lock
);
579 * Search in the children
581 list_for_each_entry(c
, &p
->children
, sibling
) {
583 goto assign_new_owner
;
587 * Search in the siblings
589 list_for_each_entry(c
, &p
->parent
->children
, sibling
) {
591 goto assign_new_owner
;
595 * Search through everything else. We should not get
598 do_each_thread(g
, c
) {
600 goto assign_new_owner
;
601 } while_each_thread(g
, c
);
603 read_unlock(&tasklist_lock
);
605 * We found no owner yet mm_users > 1: this implies that we are
606 * most likely racing with swapoff (try_to_unuse()) or /proc or
607 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
616 * The task_lock protects c->mm from changing.
617 * We always want mm->owner->mm == mm
621 * Delay read_unlock() till we have the task_lock()
622 * to ensure that c does not slip away underneath us
624 read_unlock(&tasklist_lock
);
634 #endif /* CONFIG_MM_OWNER */
637 * Turn us into a lazy TLB process if we
640 static void exit_mm(struct task_struct
* tsk
)
642 struct mm_struct
*mm
= tsk
->mm
;
643 struct core_state
*core_state
;
649 * Serialize with any possible pending coredump.
650 * We must hold mmap_sem around checking core_state
651 * and clearing tsk->mm. The core-inducing thread
652 * will increment ->nr_threads for each thread in the
653 * group with ->mm != NULL.
655 down_read(&mm
->mmap_sem
);
656 core_state
= mm
->core_state
;
658 struct core_thread self
;
659 up_read(&mm
->mmap_sem
);
662 self
.next
= xchg(&core_state
->dumper
.next
, &self
);
664 * Implies mb(), the result of xchg() must be visible
665 * to core_state->dumper.
667 if (atomic_dec_and_test(&core_state
->nr_threads
))
668 complete(&core_state
->startup
);
671 set_task_state(tsk
, TASK_UNINTERRUPTIBLE
);
672 if (!self
.task
) /* see coredump_finish() */
676 __set_task_state(tsk
, TASK_RUNNING
);
677 down_read(&mm
->mmap_sem
);
679 atomic_inc(&mm
->mm_count
);
680 BUG_ON(mm
!= tsk
->active_mm
);
681 /* more a memory barrier than a real lock */
684 up_read(&mm
->mmap_sem
);
685 enter_lazy_tlb(mm
, current
);
686 /* We don't want this task to be frozen prematurely */
687 clear_freeze_flag(tsk
);
689 mm_update_next_owner(mm
);
694 * When we die, we re-parent all our children.
695 * Try to give them to another thread in our thread
696 * group, and if no such member exists, give it to
697 * the child reaper process (ie "init") in our pid
700 static struct task_struct
*find_new_reaper(struct task_struct
*father
)
702 struct pid_namespace
*pid_ns
= task_active_pid_ns(father
);
703 struct task_struct
*thread
;
706 while_each_thread(father
, thread
) {
707 if (thread
->flags
& PF_EXITING
)
709 if (unlikely(pid_ns
->child_reaper
== father
))
710 pid_ns
->child_reaper
= thread
;
714 if (unlikely(pid_ns
->child_reaper
== father
)) {
715 write_unlock_irq(&tasklist_lock
);
716 if (unlikely(pid_ns
== &init_pid_ns
))
717 panic("Attempted to kill init!");
719 zap_pid_ns_processes(pid_ns
);
720 write_lock_irq(&tasklist_lock
);
722 * We can not clear ->child_reaper or leave it alone.
723 * There may by stealth EXIT_DEAD tasks on ->children,
724 * forget_original_parent() must move them somewhere.
726 pid_ns
->child_reaper
= init_pid_ns
.child_reaper
;
729 return pid_ns
->child_reaper
;
733 * Any that need to be release_task'd are put on the @dead list.
735 static void reparent_thread(struct task_struct
*father
, struct task_struct
*p
,
736 struct list_head
*dead
)
738 if (p
->pdeath_signal
)
739 group_send_sig_info(p
->pdeath_signal
, SEND_SIG_NOINFO
, p
);
741 list_move_tail(&p
->sibling
, &p
->real_parent
->children
);
743 if (task_detached(p
))
746 * If this is a threaded reparent there is no need to
747 * notify anyone anything has happened.
749 if (same_thread_group(p
->real_parent
, father
))
752 /* We don't want people slaying init. */
753 p
->exit_signal
= SIGCHLD
;
755 /* If it has exited notify the new parent about this child's death. */
757 p
->exit_state
== EXIT_ZOMBIE
&& thread_group_empty(p
)) {
758 do_notify_parent(p
, p
->exit_signal
);
759 if (task_detached(p
)) {
760 p
->exit_state
= EXIT_DEAD
;
761 list_move_tail(&p
->sibling
, dead
);
765 kill_orphaned_pgrp(p
, father
);
768 static void forget_original_parent(struct task_struct
*father
)
770 struct task_struct
*p
, *n
, *reaper
;
771 LIST_HEAD(dead_children
);
775 write_lock_irq(&tasklist_lock
);
776 reaper
= find_new_reaper(father
);
778 list_for_each_entry_safe(p
, n
, &father
->children
, sibling
) {
779 p
->real_parent
= reaper
;
780 if (p
->parent
== father
) {
782 p
->parent
= p
->real_parent
;
784 reparent_thread(father
, p
, &dead_children
);
786 write_unlock_irq(&tasklist_lock
);
788 BUG_ON(!list_empty(&father
->children
));
790 list_for_each_entry_safe(p
, n
, &dead_children
, sibling
) {
791 list_del_init(&p
->sibling
);
797 * Send signals to all our closest relatives so that they know
798 * to properly mourn us..
800 static void exit_notify(struct task_struct
*tsk
, int group_dead
)
806 * This does two things:
808 * A. Make init inherit all the child processes
809 * B. Check to see if any process groups have become orphaned
810 * as a result of our exiting, and if they have any stopped
811 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
813 forget_original_parent(tsk
);
814 exit_task_namespaces(tsk
);
816 write_lock_irq(&tasklist_lock
);
818 kill_orphaned_pgrp(tsk
->group_leader
, NULL
);
820 /* Let father know we died
822 * Thread signals are configurable, but you aren't going to use
823 * that to send signals to arbitary processes.
824 * That stops right now.
826 * If the parent exec id doesn't match the exec id we saved
827 * when we started then we know the parent has changed security
830 * If our self_exec id doesn't match our parent_exec_id then
831 * we have changed execution domain as these two values started
832 * the same after a fork.
834 if (tsk
->exit_signal
!= SIGCHLD
&& !task_detached(tsk
) &&
835 (tsk
->parent_exec_id
!= tsk
->real_parent
->self_exec_id
||
836 tsk
->self_exec_id
!= tsk
->parent_exec_id
))
837 tsk
->exit_signal
= SIGCHLD
;
839 signal
= tracehook_notify_death(tsk
, &cookie
, group_dead
);
841 signal
= do_notify_parent(tsk
, signal
);
843 tsk
->exit_state
= signal
== DEATH_REAP
? EXIT_DEAD
: EXIT_ZOMBIE
;
845 /* mt-exec, de_thread() is waiting for us */
846 if (thread_group_leader(tsk
) &&
847 tsk
->signal
->group_exit_task
&&
848 tsk
->signal
->notify_count
< 0)
849 wake_up_process(tsk
->signal
->group_exit_task
);
851 write_unlock_irq(&tasklist_lock
);
853 tracehook_report_death(tsk
, signal
, cookie
, group_dead
);
855 /* If the process is dead, release it - nobody will wait for it */
856 if (signal
== DEATH_REAP
)
860 #ifdef CONFIG_DEBUG_STACK_USAGE
861 static void check_stack_usage(void)
863 static DEFINE_SPINLOCK(low_water_lock
);
864 static int lowest_to_date
= THREAD_SIZE
;
867 free
= stack_not_used(current
);
869 if (free
>= lowest_to_date
)
872 spin_lock(&low_water_lock
);
873 if (free
< lowest_to_date
) {
874 printk(KERN_WARNING
"%s used greatest stack depth: %lu bytes "
876 current
->comm
, free
);
877 lowest_to_date
= free
;
879 spin_unlock(&low_water_lock
);
882 static inline void check_stack_usage(void) {}
885 NORET_TYPE
void do_exit(long code
)
887 struct task_struct
*tsk
= current
;
890 profile_task_exit(tsk
);
892 WARN_ON(atomic_read(&tsk
->fs_excl
));
894 if (unlikely(in_interrupt()))
895 panic("Aiee, killing interrupt handler!");
896 if (unlikely(!tsk
->pid
))
897 panic("Attempted to kill the idle task!");
899 tracehook_report_exit(&code
);
902 * We're taking recursive faults here in do_exit. Safest is to just
903 * leave this task alone and wait for reboot.
905 if (unlikely(tsk
->flags
& PF_EXITING
)) {
907 "Fixing recursive fault but reboot is needed!\n");
909 * We can do this unlocked here. The futex code uses
910 * this flag just to verify whether the pi state
911 * cleanup has been done or not. In the worst case it
912 * loops once more. We pretend that the cleanup was
913 * done as there is no way to return. Either the
914 * OWNER_DIED bit is set by now or we push the blocked
915 * task into the wait for ever nirwana as well.
917 tsk
->flags
|= PF_EXITPIDONE
;
918 set_current_state(TASK_UNINTERRUPTIBLE
);
924 exit_signals(tsk
); /* sets PF_EXITING */
926 * tsk->flags are checked in the futex code to protect against
927 * an exiting task cleaning up the robust pi futexes.
930 spin_unlock_wait(&tsk
->pi_lock
);
932 if (unlikely(in_atomic()))
933 printk(KERN_INFO
"note: %s[%d] exited with preempt_count %d\n",
934 current
->comm
, task_pid_nr(current
),
937 acct_update_integrals(tsk
);
939 group_dead
= atomic_dec_and_test(&tsk
->signal
->live
);
941 hrtimer_cancel(&tsk
->signal
->real_timer
);
942 exit_itimers(tsk
->signal
);
944 acct_collect(code
, group_dead
);
947 if (unlikely(tsk
->audit_context
))
950 tsk
->exit_code
= code
;
951 taskstats_exit(tsk
, group_dead
);
957 trace_sched_process_exit(tsk
);
966 if (group_dead
&& tsk
->signal
->leader
)
967 disassociate_ctty(1);
969 module_put(task_thread_info(tsk
)->exec_domain
->module
);
971 module_put(tsk
->binfmt
->module
);
973 proc_exit_connector(tsk
);
974 exit_notify(tsk
, group_dead
);
976 mpol_put(tsk
->mempolicy
);
977 tsk
->mempolicy
= NULL
;
981 * This must happen late, after the PID is not
984 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
985 exit_pi_state_list(tsk
);
986 if (unlikely(current
->pi_state_cache
))
987 kfree(current
->pi_state_cache
);
990 * Make sure we are holding no locks:
992 debug_check_no_locks_held(tsk
);
994 * We can do this unlocked here. The futex code uses this flag
995 * just to verify whether the pi state cleanup has been done
996 * or not. In the worst case it loops once more.
998 tsk
->flags
|= PF_EXITPIDONE
;
1000 if (tsk
->io_context
)
1003 if (tsk
->splice_pipe
)
1004 __free_pipe_info(tsk
->splice_pipe
);
1007 /* causes final put_task_struct in finish_task_switch(). */
1008 tsk
->state
= TASK_DEAD
;
1011 /* Avoid "noreturn function does return". */
1013 cpu_relax(); /* For when BUG is null */
1016 EXPORT_SYMBOL_GPL(do_exit
);
1018 NORET_TYPE
void complete_and_exit(struct completion
*comp
, long code
)
1026 EXPORT_SYMBOL(complete_and_exit
);
1028 SYSCALL_DEFINE1(exit
, int, error_code
)
1030 do_exit((error_code
&0xff)<<8);
1034 * Take down every thread in the group. This is called by fatal signals
1035 * as well as by sys_exit_group (below).
1038 do_group_exit(int exit_code
)
1040 struct signal_struct
*sig
= current
->signal
;
1042 BUG_ON(exit_code
& 0x80); /* core dumps don't get here */
1044 if (signal_group_exit(sig
))
1045 exit_code
= sig
->group_exit_code
;
1046 else if (!thread_group_empty(current
)) {
1047 struct sighand_struct
*const sighand
= current
->sighand
;
1048 spin_lock_irq(&sighand
->siglock
);
1049 if (signal_group_exit(sig
))
1050 /* Another thread got here before we took the lock. */
1051 exit_code
= sig
->group_exit_code
;
1053 sig
->group_exit_code
= exit_code
;
1054 sig
->flags
= SIGNAL_GROUP_EXIT
;
1055 zap_other_threads(current
);
1057 spin_unlock_irq(&sighand
->siglock
);
1065 * this kills every thread in the thread group. Note that any externally
1066 * wait4()-ing process will get the correct exit code - even if this
1067 * thread is not the thread group leader.
1069 SYSCALL_DEFINE1(exit_group
, int, error_code
)
1071 do_group_exit((error_code
& 0xff) << 8);
1076 static struct pid
*task_pid_type(struct task_struct
*task
, enum pid_type type
)
1078 struct pid
*pid
= NULL
;
1079 if (type
== PIDTYPE_PID
)
1080 pid
= task
->pids
[type
].pid
;
1081 else if (type
< PIDTYPE_MAX
)
1082 pid
= task
->group_leader
->pids
[type
].pid
;
1086 static int eligible_child(enum pid_type type
, struct pid
*pid
, int options
,
1087 struct task_struct
*p
)
1091 if (type
< PIDTYPE_MAX
) {
1092 if (task_pid_type(p
, type
) != pid
)
1096 /* Wait for all children (clone and not) if __WALL is set;
1097 * otherwise, wait for clone children *only* if __WCLONE is
1098 * set; otherwise, wait for non-clone children *only*. (Note:
1099 * A "clone" child here is one that reports to its parent
1100 * using a signal other than SIGCHLD.) */
1101 if (((p
->exit_signal
!= SIGCHLD
) ^ ((options
& __WCLONE
) != 0))
1102 && !(options
& __WALL
))
1105 err
= security_task_wait(p
);
1112 static int wait_noreap_copyout(struct task_struct
*p
, pid_t pid
, uid_t uid
,
1113 int why
, int status
,
1114 struct siginfo __user
*infop
,
1115 struct rusage __user
*rusagep
)
1117 int retval
= rusagep
? getrusage(p
, RUSAGE_BOTH
, rusagep
) : 0;
1121 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1123 retval
= put_user(0, &infop
->si_errno
);
1125 retval
= put_user((short)why
, &infop
->si_code
);
1127 retval
= put_user(pid
, &infop
->si_pid
);
1129 retval
= put_user(uid
, &infop
->si_uid
);
1131 retval
= put_user(status
, &infop
->si_status
);
1138 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1139 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1140 * the lock and this task is uninteresting. If we return nonzero, we have
1141 * released the lock and the system call should return.
1143 static int wait_task_zombie(struct task_struct
*p
, int options
,
1144 struct siginfo __user
*infop
,
1145 int __user
*stat_addr
, struct rusage __user
*ru
)
1147 unsigned long state
;
1148 int retval
, status
, traced
;
1149 pid_t pid
= task_pid_vnr(p
);
1150 uid_t uid
= __task_cred(p
)->uid
;
1152 if (!likely(options
& WEXITED
))
1155 if (unlikely(options
& WNOWAIT
)) {
1156 int exit_code
= p
->exit_code
;
1160 read_unlock(&tasklist_lock
);
1161 if ((exit_code
& 0x7f) == 0) {
1163 status
= exit_code
>> 8;
1165 why
= (exit_code
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1166 status
= exit_code
& 0x7f;
1168 return wait_noreap_copyout(p
, pid
, uid
, why
,
1173 * Try to move the task's state to DEAD
1174 * only one thread is allowed to do this:
1176 state
= xchg(&p
->exit_state
, EXIT_DEAD
);
1177 if (state
!= EXIT_ZOMBIE
) {
1178 BUG_ON(state
!= EXIT_DEAD
);
1182 traced
= ptrace_reparented(p
);
1184 if (likely(!traced
)) {
1185 struct signal_struct
*psig
;
1186 struct signal_struct
*sig
;
1187 struct task_cputime cputime
;
1190 * The resource counters for the group leader are in its
1191 * own task_struct. Those for dead threads in the group
1192 * are in its signal_struct, as are those for the child
1193 * processes it has previously reaped. All these
1194 * accumulate in the parent's signal_struct c* fields.
1196 * We don't bother to take a lock here to protect these
1197 * p->signal fields, because they are only touched by
1198 * __exit_signal, which runs with tasklist_lock
1199 * write-locked anyway, and so is excluded here. We do
1200 * need to protect the access to p->parent->signal fields,
1201 * as other threads in the parent group can be right
1202 * here reaping other children at the same time.
1204 * We use thread_group_cputime() to get times for the thread
1205 * group, which consolidates times for all threads in the
1206 * group including the group leader.
1208 thread_group_cputime(p
, &cputime
);
1209 spin_lock_irq(&p
->parent
->sighand
->siglock
);
1210 psig
= p
->parent
->signal
;
1213 cputime_add(psig
->cutime
,
1214 cputime_add(cputime
.utime
,
1217 cputime_add(psig
->cstime
,
1218 cputime_add(cputime
.stime
,
1221 cputime_add(psig
->cgtime
,
1222 cputime_add(p
->gtime
,
1223 cputime_add(sig
->gtime
,
1226 p
->min_flt
+ sig
->min_flt
+ sig
->cmin_flt
;
1228 p
->maj_flt
+ sig
->maj_flt
+ sig
->cmaj_flt
;
1230 p
->nvcsw
+ sig
->nvcsw
+ sig
->cnvcsw
;
1232 p
->nivcsw
+ sig
->nivcsw
+ sig
->cnivcsw
;
1234 task_io_get_inblock(p
) +
1235 sig
->inblock
+ sig
->cinblock
;
1237 task_io_get_oublock(p
) +
1238 sig
->oublock
+ sig
->coublock
;
1239 task_io_accounting_add(&psig
->ioac
, &p
->ioac
);
1240 task_io_accounting_add(&psig
->ioac
, &sig
->ioac
);
1241 spin_unlock_irq(&p
->parent
->sighand
->siglock
);
1245 * Now we are sure this task is interesting, and no other
1246 * thread can reap it because we set its state to EXIT_DEAD.
1248 read_unlock(&tasklist_lock
);
1250 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1251 status
= (p
->signal
->flags
& SIGNAL_GROUP_EXIT
)
1252 ? p
->signal
->group_exit_code
: p
->exit_code
;
1253 if (!retval
&& stat_addr
)
1254 retval
= put_user(status
, stat_addr
);
1255 if (!retval
&& infop
)
1256 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1257 if (!retval
&& infop
)
1258 retval
= put_user(0, &infop
->si_errno
);
1259 if (!retval
&& infop
) {
1262 if ((status
& 0x7f) == 0) {
1266 why
= (status
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1269 retval
= put_user((short)why
, &infop
->si_code
);
1271 retval
= put_user(status
, &infop
->si_status
);
1273 if (!retval
&& infop
)
1274 retval
= put_user(pid
, &infop
->si_pid
);
1275 if (!retval
&& infop
)
1276 retval
= put_user(uid
, &infop
->si_uid
);
1281 write_lock_irq(&tasklist_lock
);
1282 /* We dropped tasklist, ptracer could die and untrace */
1285 * If this is not a detached task, notify the parent.
1286 * If it's still not detached after that, don't release
1289 if (!task_detached(p
)) {
1290 do_notify_parent(p
, p
->exit_signal
);
1291 if (!task_detached(p
)) {
1292 p
->exit_state
= EXIT_ZOMBIE
;
1296 write_unlock_irq(&tasklist_lock
);
1304 static int *task_stopped_code(struct task_struct
*p
, bool ptrace
)
1307 if (task_is_stopped_or_traced(p
))
1308 return &p
->exit_code
;
1310 if (p
->signal
->flags
& SIGNAL_STOP_STOPPED
)
1311 return &p
->signal
->group_exit_code
;
1317 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1318 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1319 * the lock and this task is uninteresting. If we return nonzero, we have
1320 * released the lock and the system call should return.
1322 static int wait_task_stopped(int ptrace
, struct task_struct
*p
,
1323 int options
, struct siginfo __user
*infop
,
1324 int __user
*stat_addr
, struct rusage __user
*ru
)
1326 int retval
, exit_code
, *p_code
, why
;
1327 uid_t uid
= 0; /* unneeded, required by compiler */
1330 if (!(options
& WUNTRACED
))
1334 spin_lock_irq(&p
->sighand
->siglock
);
1336 p_code
= task_stopped_code(p
, ptrace
);
1337 if (unlikely(!p_code
))
1340 exit_code
= *p_code
;
1344 if (!unlikely(options
& WNOWAIT
))
1347 /* don't need the RCU readlock here as we're holding a spinlock */
1348 uid
= __task_cred(p
)->uid
;
1350 spin_unlock_irq(&p
->sighand
->siglock
);
1355 * Now we are pretty sure this task is interesting.
1356 * Make sure it doesn't get reaped out from under us while we
1357 * give up the lock and then examine it below. We don't want to
1358 * keep holding onto the tasklist_lock while we call getrusage and
1359 * possibly take page faults for user memory.
1362 pid
= task_pid_vnr(p
);
1363 why
= ptrace
? CLD_TRAPPED
: CLD_STOPPED
;
1364 read_unlock(&tasklist_lock
);
1366 if (unlikely(options
& WNOWAIT
))
1367 return wait_noreap_copyout(p
, pid
, uid
,
1371 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1372 if (!retval
&& stat_addr
)
1373 retval
= put_user((exit_code
<< 8) | 0x7f, stat_addr
);
1374 if (!retval
&& infop
)
1375 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1376 if (!retval
&& infop
)
1377 retval
= put_user(0, &infop
->si_errno
);
1378 if (!retval
&& infop
)
1379 retval
= put_user((short)why
, &infop
->si_code
);
1380 if (!retval
&& infop
)
1381 retval
= put_user(exit_code
, &infop
->si_status
);
1382 if (!retval
&& infop
)
1383 retval
= put_user(pid
, &infop
->si_pid
);
1384 if (!retval
&& infop
)
1385 retval
= put_user(uid
, &infop
->si_uid
);
1395 * Handle do_wait work for one task in a live, non-stopped state.
1396 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1397 * the lock and this task is uninteresting. If we return nonzero, we have
1398 * released the lock and the system call should return.
1400 static int wait_task_continued(struct task_struct
*p
, int options
,
1401 struct siginfo __user
*infop
,
1402 int __user
*stat_addr
, struct rusage __user
*ru
)
1408 if (!unlikely(options
& WCONTINUED
))
1411 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
))
1414 spin_lock_irq(&p
->sighand
->siglock
);
1415 /* Re-check with the lock held. */
1416 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
)) {
1417 spin_unlock_irq(&p
->sighand
->siglock
);
1420 if (!unlikely(options
& WNOWAIT
))
1421 p
->signal
->flags
&= ~SIGNAL_STOP_CONTINUED
;
1422 uid
= __task_cred(p
)->uid
;
1423 spin_unlock_irq(&p
->sighand
->siglock
);
1425 pid
= task_pid_vnr(p
);
1427 read_unlock(&tasklist_lock
);
1430 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1432 if (!retval
&& stat_addr
)
1433 retval
= put_user(0xffff, stat_addr
);
1437 retval
= wait_noreap_copyout(p
, pid
, uid
,
1438 CLD_CONTINUED
, SIGCONT
,
1440 BUG_ON(retval
== 0);
1447 * Consider @p for a wait by @parent.
1449 * -ECHILD should be in *@notask_error before the first call.
1450 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1451 * Returns zero if the search for a child should continue;
1452 * then *@notask_error is 0 if @p is an eligible child,
1453 * or another error from security_task_wait(), or still -ECHILD.
1455 static int wait_consider_task(struct task_struct
*parent
, int ptrace
,
1456 struct task_struct
*p
, int *notask_error
,
1457 enum pid_type type
, struct pid
*pid
, int options
,
1458 struct siginfo __user
*infop
,
1459 int __user
*stat_addr
, struct rusage __user
*ru
)
1461 int ret
= eligible_child(type
, pid
, options
, p
);
1465 if (unlikely(ret
< 0)) {
1467 * If we have not yet seen any eligible child,
1468 * then let this error code replace -ECHILD.
1469 * A permission error will give the user a clue
1470 * to look for security policy problems, rather
1471 * than for mysterious wait bugs.
1474 *notask_error
= ret
;
1478 if (likely(!ptrace
) && unlikely(p
->ptrace
)) {
1480 * This child is hidden by ptrace.
1481 * We aren't allowed to see it now, but eventually we will.
1487 if (p
->exit_state
== EXIT_DEAD
)
1491 * We don't reap group leaders with subthreads.
1493 if (p
->exit_state
== EXIT_ZOMBIE
&& !delay_group_leader(p
))
1494 return wait_task_zombie(p
, options
, infop
, stat_addr
, ru
);
1497 * It's stopped or running now, so it might
1498 * later continue, exit, or stop again.
1502 if (task_stopped_code(p
, ptrace
))
1503 return wait_task_stopped(ptrace
, p
, options
,
1504 infop
, stat_addr
, ru
);
1506 return wait_task_continued(p
, options
, infop
, stat_addr
, ru
);
1510 * Do the work of do_wait() for one thread in the group, @tsk.
1512 * -ECHILD should be in *@notask_error before the first call.
1513 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1514 * Returns zero if the search for a child should continue; then
1515 * *@notask_error is 0 if there were any eligible children,
1516 * or another error from security_task_wait(), or still -ECHILD.
1518 static int do_wait_thread(struct task_struct
*tsk
, int *notask_error
,
1519 enum pid_type type
, struct pid
*pid
, int options
,
1520 struct siginfo __user
*infop
, int __user
*stat_addr
,
1521 struct rusage __user
*ru
)
1523 struct task_struct
*p
;
1525 list_for_each_entry(p
, &tsk
->children
, sibling
) {
1527 * Do not consider detached threads.
1529 if (!task_detached(p
)) {
1530 int ret
= wait_consider_task(tsk
, 0, p
, notask_error
,
1532 infop
, stat_addr
, ru
);
1541 static int ptrace_do_wait(struct task_struct
*tsk
, int *notask_error
,
1542 enum pid_type type
, struct pid
*pid
, int options
,
1543 struct siginfo __user
*infop
, int __user
*stat_addr
,
1544 struct rusage __user
*ru
)
1546 struct task_struct
*p
;
1549 * Traditionally we see ptrace'd stopped tasks regardless of options.
1551 options
|= WUNTRACED
;
1553 list_for_each_entry(p
, &tsk
->ptraced
, ptrace_entry
) {
1554 int ret
= wait_consider_task(tsk
, 1, p
, notask_error
,
1556 infop
, stat_addr
, ru
);
1564 static long do_wait(enum pid_type type
, struct pid
*pid
, int options
,
1565 struct siginfo __user
*infop
, int __user
*stat_addr
,
1566 struct rusage __user
*ru
)
1568 DECLARE_WAITQUEUE(wait
, current
);
1569 struct task_struct
*tsk
;
1572 trace_sched_process_wait(pid
);
1574 add_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1577 * If there is nothing that can match our critiera just get out.
1578 * We will clear @retval to zero if we see any child that might later
1579 * match our criteria, even if we are not able to reap it yet.
1582 if ((type
< PIDTYPE_MAX
) && (!pid
|| hlist_empty(&pid
->tasks
[type
])))
1585 current
->state
= TASK_INTERRUPTIBLE
;
1586 read_lock(&tasklist_lock
);
1589 int tsk_result
= do_wait_thread(tsk
, &retval
,
1591 infop
, stat_addr
, ru
);
1593 tsk_result
= ptrace_do_wait(tsk
, &retval
,
1595 infop
, stat_addr
, ru
);
1598 * tasklist_lock is unlocked and we have a final result.
1600 retval
= tsk_result
;
1604 if (options
& __WNOTHREAD
)
1606 tsk
= next_thread(tsk
);
1607 BUG_ON(tsk
->signal
!= current
->signal
);
1608 } while (tsk
!= current
);
1609 read_unlock(&tasklist_lock
);
1611 if (!retval
&& !(options
& WNOHANG
)) {
1612 retval
= -ERESTARTSYS
;
1613 if (!signal_pending(current
)) {
1620 current
->state
= TASK_RUNNING
;
1621 remove_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1627 * For a WNOHANG return, clear out all the fields
1628 * we would set so the user can easily tell the
1632 retval
= put_user(0, &infop
->si_signo
);
1634 retval
= put_user(0, &infop
->si_errno
);
1636 retval
= put_user(0, &infop
->si_code
);
1638 retval
= put_user(0, &infop
->si_pid
);
1640 retval
= put_user(0, &infop
->si_uid
);
1642 retval
= put_user(0, &infop
->si_status
);
1648 SYSCALL_DEFINE5(waitid
, int, which
, pid_t
, upid
, struct siginfo __user
*,
1649 infop
, int, options
, struct rusage __user
*, ru
)
1651 struct pid
*pid
= NULL
;
1655 if (options
& ~(WNOHANG
|WNOWAIT
|WEXITED
|WSTOPPED
|WCONTINUED
))
1657 if (!(options
& (WEXITED
|WSTOPPED
|WCONTINUED
)))
1670 type
= PIDTYPE_PGID
;
1678 if (type
< PIDTYPE_MAX
)
1679 pid
= find_get_pid(upid
);
1680 ret
= do_wait(type
, pid
, options
, infop
, NULL
, ru
);
1683 /* avoid REGPARM breakage on x86: */
1684 asmlinkage_protect(5, ret
, which
, upid
, infop
, options
, ru
);
1688 SYSCALL_DEFINE4(wait4
, pid_t
, upid
, int __user
*, stat_addr
,
1689 int, options
, struct rusage __user
*, ru
)
1691 struct pid
*pid
= NULL
;
1695 if (options
& ~(WNOHANG
|WUNTRACED
|WCONTINUED
|
1696 __WNOTHREAD
|__WCLONE
|__WALL
))
1701 else if (upid
< 0) {
1702 type
= PIDTYPE_PGID
;
1703 pid
= find_get_pid(-upid
);
1704 } else if (upid
== 0) {
1705 type
= PIDTYPE_PGID
;
1706 pid
= get_task_pid(current
, PIDTYPE_PGID
);
1707 } else /* upid > 0 */ {
1709 pid
= find_get_pid(upid
);
1712 ret
= do_wait(type
, pid
, options
| WEXITED
, NULL
, stat_addr
, ru
);
1715 /* avoid REGPARM breakage on x86: */
1716 asmlinkage_protect(4, ret
, upid
, stat_addr
, options
, ru
);
1720 #ifdef __ARCH_WANT_SYS_WAITPID
1723 * sys_waitpid() remains for compatibility. waitpid() should be
1724 * implemented by calling sys_wait4() from libc.a.
1726 SYSCALL_DEFINE3(waitpid
, pid_t
, pid
, int __user
*, stat_addr
, int, options
)
1728 return sys_wait4(pid
, stat_addr
, options
, NULL
);