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/fdtable.h>
23 #include <linux/binfmts.h>
24 #include <linux/nsproxy.h>
25 #include <linux/pid_namespace.h>
26 #include <linux/ptrace.h>
27 #include <linux/profile.h>
28 #include <linux/mount.h>
29 #include <linux/proc_fs.h>
30 #include <linux/kthread.h>
31 #include <linux/mempolicy.h>
32 #include <linux/taskstats_kern.h>
33 #include <linux/delayacct.h>
34 #include <linux/freezer.h>
35 #include <linux/cgroup.h>
36 #include <linux/syscalls.h>
37 #include <linux/signal.h>
38 #include <linux/posix-timers.h>
39 #include <linux/cn_proc.h>
40 #include <linux/mutex.h>
41 #include <linux/futex.h>
42 #include <linux/compat.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>
49 #include <asm/uaccess.h>
50 #include <asm/unistd.h>
51 #include <asm/pgtable.h>
52 #include <asm/mmu_context.h>
54 static void exit_mm(struct task_struct
* tsk
);
56 static inline int task_detached(struct task_struct
*p
)
58 return p
->exit_signal
== -1;
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
);
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
));
88 sighand
= rcu_dereference(tsk
->sighand
);
89 spin_lock(&sighand
->siglock
);
91 posix_cpu_timers_exit(tsk
);
92 if (atomic_dec_and_test(&sig
->count
))
93 posix_cpu_timers_exit_group(tsk
);
96 * If there is any task waiting for the group exit
99 if (sig
->group_exit_task
&& atomic_read(&sig
->count
) == sig
->notify_count
)
100 wake_up_process(sig
->group_exit_task
);
102 if (tsk
== sig
->curr_target
)
103 sig
->curr_target
= next_thread(tsk
);
105 * Accumulate here the counters for all threads but the
106 * group leader as they die, so they can be added into
107 * the process-wide totals when those are taken.
108 * The group leader stays around as a zombie as long
109 * as there are other threads. When it gets reaped,
110 * the exit.c code will add its counts into these totals.
111 * We won't ever get here for the group leader, since it
112 * will have been the last reference on the signal_struct.
114 sig
->utime
= cputime_add(sig
->utime
, task_utime(tsk
));
115 sig
->stime
= cputime_add(sig
->stime
, task_stime(tsk
));
116 sig
->gtime
= cputime_add(sig
->gtime
, task_gtime(tsk
));
117 sig
->min_flt
+= tsk
->min_flt
;
118 sig
->maj_flt
+= tsk
->maj_flt
;
119 sig
->nvcsw
+= tsk
->nvcsw
;
120 sig
->nivcsw
+= tsk
->nivcsw
;
121 sig
->inblock
+= task_io_get_inblock(tsk
);
122 sig
->oublock
+= task_io_get_oublock(tsk
);
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
);
140 __cleanup_sighand(sighand
);
141 clear_tsk_thread_flag(tsk
,TIF_SIGPENDING
);
143 flush_sigqueue(&sig
->shared_pending
);
144 taskstats_tgid_free(sig
);
145 __cleanup_signal(sig
);
149 static void delayed_put_task_struct(struct rcu_head
*rhp
)
151 put_task_struct(container_of(rhp
, struct task_struct
, rcu
));
154 void release_task(struct task_struct
* p
)
156 struct task_struct
*leader
;
159 atomic_dec(&p
->user
->processes
);
161 write_lock_irq(&tasklist_lock
);
163 BUG_ON(!list_empty(&p
->ptrace_list
) || !list_empty(&p
->ptrace_children
));
167 * If we are the last non-leader member of the thread
168 * group, and the leader is zombie, then notify the
169 * group leader's parent process. (if it wants notification.)
172 leader
= p
->group_leader
;
173 if (leader
!= p
&& thread_group_empty(leader
) && leader
->exit_state
== EXIT_ZOMBIE
) {
174 BUG_ON(task_detached(leader
));
175 do_notify_parent(leader
, leader
->exit_signal
);
177 * If we were the last child thread and the leader has
178 * exited already, and the leader's parent ignores SIGCHLD,
179 * then we are the one who should release the leader.
181 * do_notify_parent() will have marked it self-reaping in
184 zap_leader
= task_detached(leader
);
187 write_unlock_irq(&tasklist_lock
);
189 call_rcu(&p
->rcu
, delayed_put_task_struct
);
192 if (unlikely(zap_leader
))
197 * This checks not only the pgrp, but falls back on the pid if no
198 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
201 * The caller must hold rcu lock or the tasklist lock.
203 struct pid
*session_of_pgrp(struct pid
*pgrp
)
205 struct task_struct
*p
;
206 struct pid
*sid
= NULL
;
208 p
= pid_task(pgrp
, PIDTYPE_PGID
);
210 p
= pid_task(pgrp
, PIDTYPE_PID
);
212 sid
= task_session(p
);
218 * Determine if a process group is "orphaned", according to the POSIX
219 * definition in 2.2.2.52. Orphaned process groups are not to be affected
220 * by terminal-generated stop signals. Newly orphaned process groups are
221 * to receive a SIGHUP and a SIGCONT.
223 * "I ask you, have you ever known what it is to be an orphan?"
225 static int will_become_orphaned_pgrp(struct pid
*pgrp
, struct task_struct
*ignored_task
)
227 struct task_struct
*p
;
229 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
230 if ((p
== ignored_task
) ||
231 (p
->exit_state
&& thread_group_empty(p
)) ||
232 is_global_init(p
->real_parent
))
235 if (task_pgrp(p
->real_parent
) != pgrp
&&
236 task_session(p
->real_parent
) == task_session(p
))
238 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
243 int is_current_pgrp_orphaned(void)
247 read_lock(&tasklist_lock
);
248 retval
= will_become_orphaned_pgrp(task_pgrp(current
), NULL
);
249 read_unlock(&tasklist_lock
);
254 static int has_stopped_jobs(struct pid
*pgrp
)
257 struct task_struct
*p
;
259 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
260 if (!task_is_stopped(p
))
264 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
269 * Check to see if any process groups have become orphaned as
270 * a result of our exiting, and if they have any stopped jobs,
271 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
274 kill_orphaned_pgrp(struct task_struct
*tsk
, struct task_struct
*parent
)
276 struct pid
*pgrp
= task_pgrp(tsk
);
277 struct task_struct
*ignored_task
= tsk
;
280 /* exit: our father is in a different pgrp than
281 * we are and we were the only connection outside.
283 parent
= tsk
->real_parent
;
285 /* reparent: our child is in a different pgrp than
286 * we are, and it was the only connection outside.
290 if (task_pgrp(parent
) != pgrp
&&
291 task_session(parent
) == task_session(tsk
) &&
292 will_become_orphaned_pgrp(pgrp
, ignored_task
) &&
293 has_stopped_jobs(pgrp
)) {
294 __kill_pgrp_info(SIGHUP
, SEND_SIG_PRIV
, pgrp
);
295 __kill_pgrp_info(SIGCONT
, SEND_SIG_PRIV
, pgrp
);
300 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
302 * If a kernel thread is launched as a result of a system call, or if
303 * it ever exits, it should generally reparent itself to kthreadd so it
304 * isn't in the way of other processes and is correctly cleaned up on exit.
306 * The various task state such as scheduling policy and priority may have
307 * been inherited from a user process, so we reset them to sane values here.
309 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
311 static void reparent_to_kthreadd(void)
313 write_lock_irq(&tasklist_lock
);
315 ptrace_unlink(current
);
316 /* Reparent to init */
317 remove_parent(current
);
318 current
->real_parent
= current
->parent
= kthreadd_task
;
321 /* Set the exit signal to SIGCHLD so we signal init on exit */
322 current
->exit_signal
= SIGCHLD
;
324 if (task_nice(current
) < 0)
325 set_user_nice(current
, 0);
329 security_task_reparent_to_init(current
);
330 memcpy(current
->signal
->rlim
, init_task
.signal
->rlim
,
331 sizeof(current
->signal
->rlim
));
332 atomic_inc(&(INIT_USER
->__count
));
333 write_unlock_irq(&tasklist_lock
);
334 switch_uid(INIT_USER
);
337 void __set_special_pids(struct pid
*pid
)
339 struct task_struct
*curr
= current
->group_leader
;
340 pid_t nr
= pid_nr(pid
);
342 if (task_session(curr
) != pid
) {
343 change_pid(curr
, PIDTYPE_SID
, pid
);
344 set_task_session(curr
, nr
);
346 if (task_pgrp(curr
) != pid
) {
347 change_pid(curr
, PIDTYPE_PGID
, pid
);
348 set_task_pgrp(curr
, nr
);
352 static void set_special_pids(struct pid
*pid
)
354 write_lock_irq(&tasklist_lock
);
355 __set_special_pids(pid
);
356 write_unlock_irq(&tasklist_lock
);
360 * Let kernel threads use this to say that they
361 * allow a certain signal (since daemonize() will
362 * have disabled all of them by default).
364 int allow_signal(int sig
)
366 if (!valid_signal(sig
) || sig
< 1)
369 spin_lock_irq(¤t
->sighand
->siglock
);
370 sigdelset(¤t
->blocked
, sig
);
372 /* Kernel threads handle their own signals.
373 Let the signal code know it'll be handled, so
374 that they don't get converted to SIGKILL or
375 just silently dropped */
376 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= (void __user
*)2;
379 spin_unlock_irq(¤t
->sighand
->siglock
);
383 EXPORT_SYMBOL(allow_signal
);
385 int disallow_signal(int sig
)
387 if (!valid_signal(sig
) || sig
< 1)
390 spin_lock_irq(¤t
->sighand
->siglock
);
391 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= SIG_IGN
;
393 spin_unlock_irq(¤t
->sighand
->siglock
);
397 EXPORT_SYMBOL(disallow_signal
);
400 * Put all the gunge required to become a kernel thread without
401 * attached user resources in one place where it belongs.
404 void daemonize(const char *name
, ...)
407 struct fs_struct
*fs
;
410 va_start(args
, name
);
411 vsnprintf(current
->comm
, sizeof(current
->comm
), name
, args
);
415 * If we were started as result of loading a module, close all of the
416 * user space pages. We don't need them, and if we didn't close them
417 * they would be locked into memory.
421 * We don't want to have TIF_FREEZE set if the system-wide hibernation
422 * or suspend transition begins right now.
424 current
->flags
|= PF_NOFREEZE
;
426 if (current
->nsproxy
!= &init_nsproxy
) {
427 get_nsproxy(&init_nsproxy
);
428 switch_task_namespaces(current
, &init_nsproxy
);
430 set_special_pids(&init_struct_pid
);
431 proc_clear_tty(current
);
433 /* Block and flush all signals */
434 sigfillset(&blocked
);
435 sigprocmask(SIG_BLOCK
, &blocked
, NULL
);
436 flush_signals(current
);
438 /* Become as one with the init task */
440 exit_fs(current
); /* current->fs->count--; */
443 atomic_inc(&fs
->count
);
446 current
->files
= init_task
.files
;
447 atomic_inc(¤t
->files
->count
);
449 reparent_to_kthreadd();
452 EXPORT_SYMBOL(daemonize
);
454 static void close_files(struct files_struct
* files
)
462 * It is safe to dereference the fd table without RCU or
463 * ->file_lock because this is the last reference to the
466 fdt
= files_fdtable(files
);
470 if (i
>= fdt
->max_fds
)
472 set
= fdt
->open_fds
->fds_bits
[j
++];
475 struct file
* file
= xchg(&fdt
->fd
[i
], NULL
);
477 filp_close(file
, files
);
487 struct files_struct
*get_files_struct(struct task_struct
*task
)
489 struct files_struct
*files
;
494 atomic_inc(&files
->count
);
500 void put_files_struct(struct files_struct
*files
)
504 if (atomic_dec_and_test(&files
->count
)) {
507 * Free the fd and fdset arrays if we expanded them.
508 * If the fdtable was embedded, pass files for freeing
509 * at the end of the RCU grace period. Otherwise,
510 * you can free files immediately.
512 fdt
= files_fdtable(files
);
513 if (fdt
!= &files
->fdtab
)
514 kmem_cache_free(files_cachep
, files
);
519 void reset_files_struct(struct files_struct
*files
)
521 struct task_struct
*tsk
= current
;
522 struct files_struct
*old
;
528 put_files_struct(old
);
531 void exit_files(struct task_struct
*tsk
)
533 struct files_struct
* files
= tsk
->files
;
539 put_files_struct(files
);
543 void put_fs_struct(struct fs_struct
*fs
)
545 /* No need to hold fs->lock if we are killing it */
546 if (atomic_dec_and_test(&fs
->count
)) {
549 if (fs
->altroot
.dentry
)
550 path_put(&fs
->altroot
);
551 kmem_cache_free(fs_cachep
, fs
);
555 void exit_fs(struct task_struct
*tsk
)
557 struct fs_struct
* fs
= tsk
->fs
;
567 EXPORT_SYMBOL_GPL(exit_fs
);
569 #ifdef CONFIG_MM_OWNER
571 * Task p is exiting and it owned mm, lets find a new owner for it
574 mm_need_new_owner(struct mm_struct
*mm
, struct task_struct
*p
)
577 * If there are other users of the mm and the owner (us) is exiting
578 * we need to find a new owner to take on the responsibility.
580 if (atomic_read(&mm
->mm_users
) <= 1)
587 void mm_update_next_owner(struct mm_struct
*mm
)
589 struct task_struct
*c
, *g
, *p
= current
;
592 if (!mm_need_new_owner(mm
, p
))
595 read_lock(&tasklist_lock
);
597 * Search in the children
599 list_for_each_entry(c
, &p
->children
, sibling
) {
601 goto assign_new_owner
;
605 * Search in the siblings
607 list_for_each_entry(c
, &p
->parent
->children
, sibling
) {
609 goto assign_new_owner
;
613 * Search through everything else. We should not get
616 do_each_thread(g
, c
) {
618 goto assign_new_owner
;
619 } while_each_thread(g
, c
);
621 read_unlock(&tasklist_lock
);
623 * We found no owner yet mm_users > 1: this implies that we are
624 * most likely racing with swapoff (try_to_unuse()) or /proc or
625 * ptrace or page migration (get_task_mm()). Mark owner as NULL,
626 * so that subsystems can understand the callback and take action.
628 down_write(&mm
->mmap_sem
);
629 cgroup_mm_owner_callbacks(mm
->owner
, NULL
);
631 up_write(&mm
->mmap_sem
);
638 * The task_lock protects c->mm from changing.
639 * We always want mm->owner->mm == mm
643 * Delay read_unlock() till we have the task_lock()
644 * to ensure that c does not slip away underneath us
646 read_unlock(&tasklist_lock
);
652 cgroup_mm_owner_callbacks(mm
->owner
, c
);
657 #endif /* CONFIG_MM_OWNER */
660 * Turn us into a lazy TLB process if we
663 static void exit_mm(struct task_struct
* tsk
)
665 struct mm_struct
*mm
= tsk
->mm
;
671 * Serialize with any possible pending coredump.
672 * We must hold mmap_sem around checking core_waiters
673 * and clearing tsk->mm. The core-inducing thread
674 * will increment core_waiters for each thread in the
675 * group with ->mm != NULL.
677 down_read(&mm
->mmap_sem
);
678 if (mm
->core_waiters
) {
679 up_read(&mm
->mmap_sem
);
680 down_write(&mm
->mmap_sem
);
681 if (!--mm
->core_waiters
)
682 complete(mm
->core_startup_done
);
683 up_write(&mm
->mmap_sem
);
685 wait_for_completion(&mm
->core_done
);
686 down_read(&mm
->mmap_sem
);
688 atomic_inc(&mm
->mm_count
);
689 BUG_ON(mm
!= tsk
->active_mm
);
690 /* more a memory barrier than a real lock */
693 up_read(&mm
->mmap_sem
);
694 enter_lazy_tlb(mm
, current
);
695 /* We don't want this task to be frozen prematurely */
696 clear_freeze_flag(tsk
);
698 mm_update_next_owner(mm
);
703 reparent_thread(struct task_struct
*p
, struct task_struct
*father
, int traced
)
705 if (p
->pdeath_signal
)
706 /* We already hold the tasklist_lock here. */
707 group_send_sig_info(p
->pdeath_signal
, SEND_SIG_NOINFO
, p
);
709 /* Move the child from its dying parent to the new one. */
710 if (unlikely(traced
)) {
711 /* Preserve ptrace links if someone else is tracing this child. */
712 list_del_init(&p
->ptrace_list
);
713 if (ptrace_reparented(p
))
714 list_add(&p
->ptrace_list
, &p
->real_parent
->ptrace_children
);
716 /* If this child is being traced, then we're the one tracing it
717 * anyway, so let go of it.
721 p
->parent
= p
->real_parent
;
724 if (task_is_traced(p
)) {
726 * If it was at a trace stop, turn it into
727 * a normal stop since it's no longer being
734 /* If this is a threaded reparent there is no need to
735 * notify anyone anything has happened.
737 if (same_thread_group(p
->real_parent
, father
))
740 /* We don't want people slaying init. */
741 if (!task_detached(p
))
742 p
->exit_signal
= SIGCHLD
;
744 /* If we'd notified the old parent about this child's death,
745 * also notify the new parent.
747 if (!traced
&& p
->exit_state
== EXIT_ZOMBIE
&&
748 !task_detached(p
) && thread_group_empty(p
))
749 do_notify_parent(p
, p
->exit_signal
);
751 kill_orphaned_pgrp(p
, father
);
755 * When we die, we re-parent all our children.
756 * Try to give them to another thread in our thread
757 * group, and if no such member exists, give it to
758 * the child reaper process (ie "init") in our pid
761 static void forget_original_parent(struct task_struct
*father
)
763 struct task_struct
*p
, *n
, *reaper
= father
;
764 struct list_head ptrace_dead
;
766 INIT_LIST_HEAD(&ptrace_dead
);
768 write_lock_irq(&tasklist_lock
);
771 reaper
= next_thread(reaper
);
772 if (reaper
== father
) {
773 reaper
= task_child_reaper(father
);
776 } while (reaper
->flags
& PF_EXITING
);
779 * There are only two places where our children can be:
781 * - in our child list
782 * - in our ptraced child list
784 * Search them and reparent children.
786 list_for_each_entry_safe(p
, n
, &father
->children
, sibling
) {
791 /* if father isn't the real parent, then ptrace must be enabled */
792 BUG_ON(father
!= p
->real_parent
&& !ptrace
);
794 if (father
== p
->real_parent
) {
795 /* reparent with a reaper, real father it's us */
796 p
->real_parent
= reaper
;
797 reparent_thread(p
, father
, 0);
799 /* reparent ptraced task to its real parent */
801 if (p
->exit_state
== EXIT_ZOMBIE
&& !task_detached(p
) &&
802 thread_group_empty(p
))
803 do_notify_parent(p
, p
->exit_signal
);
807 * if the ptraced child is a detached zombie we must collect
808 * it before we exit, or it will remain zombie forever since
809 * we prevented it from self-reap itself while it was being
810 * traced by us, to be able to see it in wait4.
812 if (unlikely(ptrace
&& p
->exit_state
== EXIT_ZOMBIE
&& task_detached(p
)))
813 list_add(&p
->ptrace_list
, &ptrace_dead
);
816 list_for_each_entry_safe(p
, n
, &father
->ptrace_children
, ptrace_list
) {
817 p
->real_parent
= reaper
;
818 reparent_thread(p
, father
, 1);
821 write_unlock_irq(&tasklist_lock
);
822 BUG_ON(!list_empty(&father
->children
));
823 BUG_ON(!list_empty(&father
->ptrace_children
));
825 list_for_each_entry_safe(p
, n
, &ptrace_dead
, ptrace_list
) {
826 list_del_init(&p
->ptrace_list
);
833 * Send signals to all our closest relatives so that they know
834 * to properly mourn us..
836 static void exit_notify(struct task_struct
*tsk
, int group_dead
)
841 * This does two things:
843 * A. Make init inherit all the child processes
844 * B. Check to see if any process groups have become orphaned
845 * as a result of our exiting, and if they have any stopped
846 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
848 forget_original_parent(tsk
);
849 exit_task_namespaces(tsk
);
851 write_lock_irq(&tasklist_lock
);
853 kill_orphaned_pgrp(tsk
->group_leader
, NULL
);
855 /* Let father know we died
857 * Thread signals are configurable, but you aren't going to use
858 * that to send signals to arbitary processes.
859 * That stops right now.
861 * If the parent exec id doesn't match the exec id we saved
862 * when we started then we know the parent has changed security
865 * If our self_exec id doesn't match our parent_exec_id then
866 * we have changed execution domain as these two values started
867 * the same after a fork.
869 if (tsk
->exit_signal
!= SIGCHLD
&& !task_detached(tsk
) &&
870 (tsk
->parent_exec_id
!= tsk
->real_parent
->self_exec_id
||
871 tsk
->self_exec_id
!= tsk
->parent_exec_id
) &&
873 tsk
->exit_signal
= SIGCHLD
;
875 /* If something other than our normal parent is ptracing us, then
876 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
877 * only has special meaning to our real parent.
879 if (!task_detached(tsk
) && thread_group_empty(tsk
)) {
880 int signal
= ptrace_reparented(tsk
) ?
881 SIGCHLD
: tsk
->exit_signal
;
882 do_notify_parent(tsk
, signal
);
883 } else if (tsk
->ptrace
) {
884 do_notify_parent(tsk
, SIGCHLD
);
888 if (task_detached(tsk
) && likely(!tsk
->ptrace
))
890 tsk
->exit_state
= state
;
892 /* mt-exec, de_thread() is waiting for us */
893 if (thread_group_leader(tsk
) &&
894 tsk
->signal
->notify_count
< 0 &&
895 tsk
->signal
->group_exit_task
)
896 wake_up_process(tsk
->signal
->group_exit_task
);
898 write_unlock_irq(&tasklist_lock
);
900 /* If the process is dead, release it - nobody will wait for it */
901 if (state
== EXIT_DEAD
)
905 #ifdef CONFIG_DEBUG_STACK_USAGE
906 static void check_stack_usage(void)
908 static DEFINE_SPINLOCK(low_water_lock
);
909 static int lowest_to_date
= THREAD_SIZE
;
910 unsigned long *n
= end_of_stack(current
);
915 free
= (unsigned long)n
- (unsigned long)end_of_stack(current
);
917 if (free
>= lowest_to_date
)
920 spin_lock(&low_water_lock
);
921 if (free
< lowest_to_date
) {
922 printk(KERN_WARNING
"%s used greatest stack depth: %lu bytes "
924 current
->comm
, free
);
925 lowest_to_date
= free
;
927 spin_unlock(&low_water_lock
);
930 static inline void check_stack_usage(void) {}
933 static inline void exit_child_reaper(struct task_struct
*tsk
)
935 if (likely(tsk
->group_leader
!= task_child_reaper(tsk
)))
938 if (tsk
->nsproxy
->pid_ns
== &init_pid_ns
)
939 panic("Attempted to kill init!");
942 * @tsk is the last thread in the 'cgroup-init' and is exiting.
943 * Terminate all remaining processes in the namespace and reap them
944 * before exiting @tsk.
946 * Note that @tsk (last thread of cgroup-init) may not necessarily
947 * be the child-reaper (i.e main thread of cgroup-init) of the
948 * namespace i.e the child_reaper may have already exited.
950 * Even after a child_reaper exits, we let it inherit orphaned children,
951 * because, pid_ns->child_reaper remains valid as long as there is
952 * at least one living sub-thread in the cgroup init.
954 * This living sub-thread of the cgroup-init will be notified when
955 * a child inherited by the 'child-reaper' exits (do_notify_parent()
956 * uses __group_send_sig_info()). Further, when reaping child processes,
957 * do_wait() iterates over children of all living sub threads.
959 * i.e even though 'child_reaper' thread is listed as the parent of the
960 * orphaned children, any living sub-thread in the cgroup-init can
961 * perform the role of the child_reaper.
963 zap_pid_ns_processes(tsk
->nsproxy
->pid_ns
);
966 NORET_TYPE
void do_exit(long code
)
968 struct task_struct
*tsk
= current
;
971 profile_task_exit(tsk
);
973 WARN_ON(atomic_read(&tsk
->fs_excl
));
975 if (unlikely(in_interrupt()))
976 panic("Aiee, killing interrupt handler!");
977 if (unlikely(!tsk
->pid
))
978 panic("Attempted to kill the idle task!");
980 if (unlikely(current
->ptrace
& PT_TRACE_EXIT
)) {
981 current
->ptrace_message
= code
;
982 ptrace_notify((PTRACE_EVENT_EXIT
<< 8) | SIGTRAP
);
986 * We're taking recursive faults here in do_exit. Safest is to just
987 * leave this task alone and wait for reboot.
989 if (unlikely(tsk
->flags
& PF_EXITING
)) {
991 "Fixing recursive fault but reboot is needed!\n");
993 * We can do this unlocked here. The futex code uses
994 * this flag just to verify whether the pi state
995 * cleanup has been done or not. In the worst case it
996 * loops once more. We pretend that the cleanup was
997 * done as there is no way to return. Either the
998 * OWNER_DIED bit is set by now or we push the blocked
999 * task into the wait for ever nirwana as well.
1001 tsk
->flags
|= PF_EXITPIDONE
;
1002 if (tsk
->io_context
)
1004 set_current_state(TASK_UNINTERRUPTIBLE
);
1008 exit_signals(tsk
); /* sets PF_EXITING */
1010 * tsk->flags are checked in the futex code to protect against
1011 * an exiting task cleaning up the robust pi futexes.
1014 spin_unlock_wait(&tsk
->pi_lock
);
1016 if (unlikely(in_atomic()))
1017 printk(KERN_INFO
"note: %s[%d] exited with preempt_count %d\n",
1018 current
->comm
, task_pid_nr(current
),
1021 acct_update_integrals(tsk
);
1023 update_hiwater_rss(tsk
->mm
);
1024 update_hiwater_vm(tsk
->mm
);
1026 group_dead
= atomic_dec_and_test(&tsk
->signal
->live
);
1028 exit_child_reaper(tsk
);
1029 hrtimer_cancel(&tsk
->signal
->real_timer
);
1030 exit_itimers(tsk
->signal
);
1032 acct_collect(code
, group_dead
);
1034 if (unlikely(tsk
->robust_list
))
1035 exit_robust_list(tsk
);
1036 #ifdef CONFIG_COMPAT
1037 if (unlikely(tsk
->compat_robust_list
))
1038 compat_exit_robust_list(tsk
);
1043 if (unlikely(tsk
->audit_context
))
1046 tsk
->exit_code
= code
;
1047 taskstats_exit(tsk
, group_dead
);
1056 check_stack_usage();
1058 cgroup_exit(tsk
, 1);
1061 if (group_dead
&& tsk
->signal
->leader
)
1062 disassociate_ctty(1);
1064 module_put(task_thread_info(tsk
)->exec_domain
->module
);
1066 module_put(tsk
->binfmt
->module
);
1068 proc_exit_connector(tsk
);
1069 exit_notify(tsk
, group_dead
);
1071 mpol_put(tsk
->mempolicy
);
1072 tsk
->mempolicy
= NULL
;
1076 * This must happen late, after the PID is not
1079 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
1080 exit_pi_state_list(tsk
);
1081 if (unlikely(current
->pi_state_cache
))
1082 kfree(current
->pi_state_cache
);
1085 * Make sure we are holding no locks:
1087 debug_check_no_locks_held(tsk
);
1089 * We can do this unlocked here. The futex code uses this flag
1090 * just to verify whether the pi state cleanup has been done
1091 * or not. In the worst case it loops once more.
1093 tsk
->flags
|= PF_EXITPIDONE
;
1095 if (tsk
->io_context
)
1098 if (tsk
->splice_pipe
)
1099 __free_pipe_info(tsk
->splice_pipe
);
1102 /* causes final put_task_struct in finish_task_switch(). */
1103 tsk
->state
= TASK_DEAD
;
1107 /* Avoid "noreturn function does return". */
1109 cpu_relax(); /* For when BUG is null */
1112 EXPORT_SYMBOL_GPL(do_exit
);
1114 NORET_TYPE
void complete_and_exit(struct completion
*comp
, long code
)
1122 EXPORT_SYMBOL(complete_and_exit
);
1124 asmlinkage
long sys_exit(int error_code
)
1126 do_exit((error_code
&0xff)<<8);
1130 * Take down every thread in the group. This is called by fatal signals
1131 * as well as by sys_exit_group (below).
1134 do_group_exit(int exit_code
)
1136 struct signal_struct
*sig
= current
->signal
;
1138 BUG_ON(exit_code
& 0x80); /* core dumps don't get here */
1140 if (signal_group_exit(sig
))
1141 exit_code
= sig
->group_exit_code
;
1142 else if (!thread_group_empty(current
)) {
1143 struct sighand_struct
*const sighand
= current
->sighand
;
1144 spin_lock_irq(&sighand
->siglock
);
1145 if (signal_group_exit(sig
))
1146 /* Another thread got here before we took the lock. */
1147 exit_code
= sig
->group_exit_code
;
1149 sig
->group_exit_code
= exit_code
;
1150 sig
->flags
= SIGNAL_GROUP_EXIT
;
1151 zap_other_threads(current
);
1153 spin_unlock_irq(&sighand
->siglock
);
1161 * this kills every thread in the thread group. Note that any externally
1162 * wait4()-ing process will get the correct exit code - even if this
1163 * thread is not the thread group leader.
1165 asmlinkage
void sys_exit_group(int error_code
)
1167 do_group_exit((error_code
& 0xff) << 8);
1170 static struct pid
*task_pid_type(struct task_struct
*task
, enum pid_type type
)
1172 struct pid
*pid
= NULL
;
1173 if (type
== PIDTYPE_PID
)
1174 pid
= task
->pids
[type
].pid
;
1175 else if (type
< PIDTYPE_MAX
)
1176 pid
= task
->group_leader
->pids
[type
].pid
;
1180 static int eligible_child(enum pid_type type
, struct pid
*pid
, int options
,
1181 struct task_struct
*p
)
1185 if (type
< PIDTYPE_MAX
) {
1186 if (task_pid_type(p
, type
) != pid
)
1191 * Do not consider detached threads that are
1194 if (task_detached(p
) && !p
->ptrace
)
1197 /* Wait for all children (clone and not) if __WALL is set;
1198 * otherwise, wait for clone children *only* if __WCLONE is
1199 * set; otherwise, wait for non-clone children *only*. (Note:
1200 * A "clone" child here is one that reports to its parent
1201 * using a signal other than SIGCHLD.) */
1202 if (((p
->exit_signal
!= SIGCHLD
) ^ ((options
& __WCLONE
) != 0))
1203 && !(options
& __WALL
))
1206 err
= security_task_wait(p
);
1210 if (type
!= PIDTYPE_PID
)
1212 /* This child was explicitly requested, abort */
1213 read_unlock(&tasklist_lock
);
1217 static int wait_noreap_copyout(struct task_struct
*p
, pid_t pid
, uid_t uid
,
1218 int why
, int status
,
1219 struct siginfo __user
*infop
,
1220 struct rusage __user
*rusagep
)
1222 int retval
= rusagep
? getrusage(p
, RUSAGE_BOTH
, rusagep
) : 0;
1226 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1228 retval
= put_user(0, &infop
->si_errno
);
1230 retval
= put_user((short)why
, &infop
->si_code
);
1232 retval
= put_user(pid
, &infop
->si_pid
);
1234 retval
= put_user(uid
, &infop
->si_uid
);
1236 retval
= put_user(status
, &infop
->si_status
);
1243 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1244 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1245 * the lock and this task is uninteresting. If we return nonzero, we have
1246 * released the lock and the system call should return.
1248 static int wait_task_zombie(struct task_struct
*p
, int noreap
,
1249 struct siginfo __user
*infop
,
1250 int __user
*stat_addr
, struct rusage __user
*ru
)
1252 unsigned long state
;
1253 int retval
, status
, traced
;
1254 pid_t pid
= task_pid_vnr(p
);
1256 if (unlikely(noreap
)) {
1258 int exit_code
= p
->exit_code
;
1262 read_unlock(&tasklist_lock
);
1263 if ((exit_code
& 0x7f) == 0) {
1265 status
= exit_code
>> 8;
1267 why
= (exit_code
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1268 status
= exit_code
& 0x7f;
1270 return wait_noreap_copyout(p
, pid
, uid
, why
,
1275 * Try to move the task's state to DEAD
1276 * only one thread is allowed to do this:
1278 state
= xchg(&p
->exit_state
, EXIT_DEAD
);
1279 if (state
!= EXIT_ZOMBIE
) {
1280 BUG_ON(state
!= EXIT_DEAD
);
1284 traced
= ptrace_reparented(p
);
1286 if (likely(!traced
)) {
1287 struct signal_struct
*psig
;
1288 struct signal_struct
*sig
;
1291 * The resource counters for the group leader are in its
1292 * own task_struct. Those for dead threads in the group
1293 * are in its signal_struct, as are those for the child
1294 * processes it has previously reaped. All these
1295 * accumulate in the parent's signal_struct c* fields.
1297 * We don't bother to take a lock here to protect these
1298 * p->signal fields, because they are only touched by
1299 * __exit_signal, which runs with tasklist_lock
1300 * write-locked anyway, and so is excluded here. We do
1301 * need to protect the access to p->parent->signal fields,
1302 * as other threads in the parent group can be right
1303 * here reaping other children at the same time.
1305 spin_lock_irq(&p
->parent
->sighand
->siglock
);
1306 psig
= p
->parent
->signal
;
1309 cputime_add(psig
->cutime
,
1310 cputime_add(p
->utime
,
1311 cputime_add(sig
->utime
,
1314 cputime_add(psig
->cstime
,
1315 cputime_add(p
->stime
,
1316 cputime_add(sig
->stime
,
1319 cputime_add(psig
->cgtime
,
1320 cputime_add(p
->gtime
,
1321 cputime_add(sig
->gtime
,
1324 p
->min_flt
+ sig
->min_flt
+ sig
->cmin_flt
;
1326 p
->maj_flt
+ sig
->maj_flt
+ sig
->cmaj_flt
;
1328 p
->nvcsw
+ sig
->nvcsw
+ sig
->cnvcsw
;
1330 p
->nivcsw
+ sig
->nivcsw
+ sig
->cnivcsw
;
1332 task_io_get_inblock(p
) +
1333 sig
->inblock
+ sig
->cinblock
;
1335 task_io_get_oublock(p
) +
1336 sig
->oublock
+ sig
->coublock
;
1337 spin_unlock_irq(&p
->parent
->sighand
->siglock
);
1341 * Now we are sure this task is interesting, and no other
1342 * thread can reap it because we set its state to EXIT_DEAD.
1344 read_unlock(&tasklist_lock
);
1346 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1347 status
= (p
->signal
->flags
& SIGNAL_GROUP_EXIT
)
1348 ? p
->signal
->group_exit_code
: p
->exit_code
;
1349 if (!retval
&& stat_addr
)
1350 retval
= put_user(status
, stat_addr
);
1351 if (!retval
&& infop
)
1352 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1353 if (!retval
&& infop
)
1354 retval
= put_user(0, &infop
->si_errno
);
1355 if (!retval
&& infop
) {
1358 if ((status
& 0x7f) == 0) {
1362 why
= (status
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1365 retval
= put_user((short)why
, &infop
->si_code
);
1367 retval
= put_user(status
, &infop
->si_status
);
1369 if (!retval
&& infop
)
1370 retval
= put_user(pid
, &infop
->si_pid
);
1371 if (!retval
&& infop
)
1372 retval
= put_user(p
->uid
, &infop
->si_uid
);
1377 write_lock_irq(&tasklist_lock
);
1378 /* We dropped tasklist, ptracer could die and untrace */
1381 * If this is not a detached task, notify the parent.
1382 * If it's still not detached after that, don't release
1385 if (!task_detached(p
)) {
1386 do_notify_parent(p
, p
->exit_signal
);
1387 if (!task_detached(p
)) {
1388 p
->exit_state
= EXIT_ZOMBIE
;
1392 write_unlock_irq(&tasklist_lock
);
1401 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1402 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1403 * the lock and this task is uninteresting. If we return nonzero, we have
1404 * released the lock and the system call should return.
1406 static int wait_task_stopped(struct task_struct
*p
,
1407 int noreap
, struct siginfo __user
*infop
,
1408 int __user
*stat_addr
, struct rusage __user
*ru
)
1410 int retval
, exit_code
, why
;
1411 uid_t uid
= 0; /* unneeded, required by compiler */
1415 spin_lock_irq(&p
->sighand
->siglock
);
1417 if (unlikely(!task_is_stopped_or_traced(p
)))
1420 if (!(p
->ptrace
& PT_PTRACED
) && p
->signal
->group_stop_count
> 0)
1422 * A group stop is in progress and this is the group leader.
1423 * We won't report until all threads have stopped.
1427 exit_code
= p
->exit_code
;
1436 spin_unlock_irq(&p
->sighand
->siglock
);
1441 * Now we are pretty sure this task is interesting.
1442 * Make sure it doesn't get reaped out from under us while we
1443 * give up the lock and then examine it below. We don't want to
1444 * keep holding onto the tasklist_lock while we call getrusage and
1445 * possibly take page faults for user memory.
1448 pid
= task_pid_vnr(p
);
1449 why
= (p
->ptrace
& PT_PTRACED
) ? CLD_TRAPPED
: CLD_STOPPED
;
1450 read_unlock(&tasklist_lock
);
1452 if (unlikely(noreap
))
1453 return wait_noreap_copyout(p
, pid
, uid
,
1457 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1458 if (!retval
&& stat_addr
)
1459 retval
= put_user((exit_code
<< 8) | 0x7f, stat_addr
);
1460 if (!retval
&& infop
)
1461 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1462 if (!retval
&& infop
)
1463 retval
= put_user(0, &infop
->si_errno
);
1464 if (!retval
&& infop
)
1465 retval
= put_user((short)why
, &infop
->si_code
);
1466 if (!retval
&& infop
)
1467 retval
= put_user(exit_code
, &infop
->si_status
);
1468 if (!retval
&& infop
)
1469 retval
= put_user(pid
, &infop
->si_pid
);
1470 if (!retval
&& infop
)
1471 retval
= put_user(uid
, &infop
->si_uid
);
1481 * Handle do_wait work for one task in a live, non-stopped state.
1482 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1483 * the lock and this task is uninteresting. If we return nonzero, we have
1484 * released the lock and the system call should return.
1486 static int wait_task_continued(struct task_struct
*p
, int noreap
,
1487 struct siginfo __user
*infop
,
1488 int __user
*stat_addr
, struct rusage __user
*ru
)
1494 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
))
1497 spin_lock_irq(&p
->sighand
->siglock
);
1498 /* Re-check with the lock held. */
1499 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
)) {
1500 spin_unlock_irq(&p
->sighand
->siglock
);
1504 p
->signal
->flags
&= ~SIGNAL_STOP_CONTINUED
;
1505 spin_unlock_irq(&p
->sighand
->siglock
);
1507 pid
= task_pid_vnr(p
);
1510 read_unlock(&tasklist_lock
);
1513 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1515 if (!retval
&& stat_addr
)
1516 retval
= put_user(0xffff, stat_addr
);
1520 retval
= wait_noreap_copyout(p
, pid
, uid
,
1521 CLD_CONTINUED
, SIGCONT
,
1523 BUG_ON(retval
== 0);
1529 static long do_wait(enum pid_type type
, struct pid
*pid
, int options
,
1530 struct siginfo __user
*infop
, int __user
*stat_addr
,
1531 struct rusage __user
*ru
)
1533 DECLARE_WAITQUEUE(wait
, current
);
1534 struct task_struct
*tsk
;
1537 add_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1539 /* If there is nothing that can match our critier just get out */
1541 if ((type
< PIDTYPE_MAX
) && (!pid
|| hlist_empty(&pid
->tasks
[type
])))
1545 * We will set this flag if we see any child that might later
1546 * match our criteria, even if we are not able to reap it yet.
1549 current
->state
= TASK_INTERRUPTIBLE
;
1550 read_lock(&tasklist_lock
);
1553 struct task_struct
*p
;
1555 list_for_each_entry(p
, &tsk
->children
, sibling
) {
1556 int ret
= eligible_child(type
, pid
, options
, p
);
1560 if (unlikely(ret
< 0)) {
1562 } else if (task_is_stopped_or_traced(p
)) {
1564 * It's stopped now, so it might later
1565 * continue, exit, or stop again.
1568 if (!(p
->ptrace
& PT_PTRACED
) &&
1569 !(options
& WUNTRACED
))
1572 retval
= wait_task_stopped(p
,
1573 (options
& WNOWAIT
), infop
,
1575 } else if (p
->exit_state
== EXIT_ZOMBIE
&&
1576 !delay_group_leader(p
)) {
1578 * We don't reap group leaders with subthreads.
1580 if (!likely(options
& WEXITED
))
1582 retval
= wait_task_zombie(p
,
1583 (options
& WNOWAIT
), infop
,
1585 } else if (p
->exit_state
!= EXIT_DEAD
) {
1587 * It's running now, so it might later
1588 * exit, stop, or stop and then continue.
1591 if (!unlikely(options
& WCONTINUED
))
1593 retval
= wait_task_continued(p
,
1594 (options
& WNOWAIT
), infop
,
1597 if (retval
!= 0) /* tasklist_lock released */
1601 list_for_each_entry(p
, &tsk
->ptrace_children
,
1603 flag
= eligible_child(type
, pid
, options
, p
);
1606 if (likely(flag
> 0))
1612 if (options
& __WNOTHREAD
)
1614 tsk
= next_thread(tsk
);
1615 BUG_ON(tsk
->signal
!= current
->signal
);
1616 } while (tsk
!= current
);
1617 read_unlock(&tasklist_lock
);
1620 if (options
& WNOHANG
)
1622 retval
= -ERESTARTSYS
;
1623 if (signal_pending(current
))
1630 current
->state
= TASK_RUNNING
;
1631 remove_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1637 * For a WNOHANG return, clear out all the fields
1638 * we would set so the user can easily tell the
1642 retval
= put_user(0, &infop
->si_signo
);
1644 retval
= put_user(0, &infop
->si_errno
);
1646 retval
= put_user(0, &infop
->si_code
);
1648 retval
= put_user(0, &infop
->si_pid
);
1650 retval
= put_user(0, &infop
->si_uid
);
1652 retval
= put_user(0, &infop
->si_status
);
1658 asmlinkage
long sys_waitid(int which
, pid_t upid
,
1659 struct siginfo __user
*infop
, int options
,
1660 struct rusage __user
*ru
)
1662 struct pid
*pid
= NULL
;
1666 if (options
& ~(WNOHANG
|WNOWAIT
|WEXITED
|WSTOPPED
|WCONTINUED
))
1668 if (!(options
& (WEXITED
|WSTOPPED
|WCONTINUED
)))
1681 type
= PIDTYPE_PGID
;
1689 if (type
< PIDTYPE_MAX
)
1690 pid
= find_get_pid(upid
);
1691 ret
= do_wait(type
, pid
, options
, infop
, NULL
, ru
);
1694 /* avoid REGPARM breakage on x86: */
1695 asmlinkage_protect(5, ret
, which
, upid
, infop
, options
, ru
);
1699 asmlinkage
long sys_wait4(pid_t upid
, int __user
*stat_addr
,
1700 int options
, struct rusage __user
*ru
)
1702 struct pid
*pid
= NULL
;
1706 if (options
& ~(WNOHANG
|WUNTRACED
|WCONTINUED
|
1707 __WNOTHREAD
|__WCLONE
|__WALL
))
1712 else if (upid
< 0) {
1713 type
= PIDTYPE_PGID
;
1714 pid
= find_get_pid(-upid
);
1715 } else if (upid
== 0) {
1716 type
= PIDTYPE_PGID
;
1717 pid
= get_pid(task_pgrp(current
));
1718 } else /* upid > 0 */ {
1720 pid
= find_get_pid(upid
);
1723 ret
= do_wait(type
, pid
, options
| WEXITED
, NULL
, stat_addr
, ru
);
1726 /* avoid REGPARM breakage on x86: */
1727 asmlinkage_protect(4, ret
, upid
, stat_addr
, options
, ru
);
1731 #ifdef __ARCH_WANT_SYS_WAITPID
1734 * sys_waitpid() remains for compatibility. waitpid() should be
1735 * implemented by calling sys_wait4() from libc.a.
1737 asmlinkage
long sys_waitpid(pid_t pid
, int __user
*stat_addr
, int options
)
1739 return sys_wait4(pid
, stat_addr
, options
, NULL
);