da28745f7c3869854b9486ecc9d4b893f7571b19
4 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/slab.h>
9 #include <linux/interrupt.h>
10 #include <linux/module.h>
11 #include <linux/capability.h>
12 #include <linux/completion.h>
13 #include <linux/personality.h>
14 #include <linux/tty.h>
15 #include <linux/mnt_namespace.h>
16 #include <linux/iocontext.h>
17 #include <linux/key.h>
18 #include <linux/security.h>
19 #include <linux/cpu.h>
20 #include <linux/acct.h>
21 #include <linux/tsacct_kern.h>
22 #include <linux/file.h>
23 #include <linux/fdtable.h>
24 #include <linux/binfmts.h>
25 #include <linux/nsproxy.h>
26 #include <linux/pid_namespace.h>
27 #include <linux/ptrace.h>
28 #include <linux/profile.h>
29 #include <linux/mount.h>
30 #include <linux/proc_fs.h>
31 #include <linux/kthread.h>
32 #include <linux/mempolicy.h>
33 #include <linux/taskstats_kern.h>
34 #include <linux/delayacct.h>
35 #include <linux/freezer.h>
36 #include <linux/cgroup.h>
37 #include <linux/syscalls.h>
38 #include <linux/signal.h>
39 #include <linux/posix-timers.h>
40 #include <linux/cn_proc.h>
41 #include <linux/mutex.h>
42 #include <linux/futex.h>
43 #include <linux/compat.h>
44 #include <linux/pipe_fs_i.h>
45 #include <linux/audit.h> /* for audit_free() */
46 #include <linux/resource.h>
47 #include <linux/blkdev.h>
48 #include <linux/task_io_accounting_ops.h>
49 #include <linux/tracehook.h>
51 #include <asm/uaccess.h>
52 #include <asm/unistd.h>
53 #include <asm/pgtable.h>
54 #include <asm/mmu_context.h>
56 static void exit_mm(struct task_struct
* tsk
);
58 static inline int task_detached(struct task_struct
*p
)
60 return p
->exit_signal
== -1;
63 static void __unhash_process(struct task_struct
*p
)
66 detach_pid(p
, PIDTYPE_PID
);
67 if (thread_group_leader(p
)) {
68 detach_pid(p
, PIDTYPE_PGID
);
69 detach_pid(p
, PIDTYPE_SID
);
71 list_del_rcu(&p
->tasks
);
72 __get_cpu_var(process_counts
)--;
74 list_del_rcu(&p
->thread_group
);
75 list_del_init(&p
->sibling
);
79 * This function expects the tasklist_lock write-locked.
81 static void __exit_signal(struct task_struct
*tsk
)
83 struct signal_struct
*sig
= tsk
->signal
;
84 struct sighand_struct
*sighand
;
87 BUG_ON(!atomic_read(&sig
->count
));
89 sighand
= rcu_dereference(tsk
->sighand
);
90 spin_lock(&sighand
->siglock
);
92 posix_cpu_timers_exit(tsk
);
93 if (atomic_dec_and_test(&sig
->count
))
94 posix_cpu_timers_exit_group(tsk
);
97 * If there is any task waiting for the group exit
100 if (sig
->group_exit_task
&& atomic_read(&sig
->count
) == sig
->notify_count
)
101 wake_up_process(sig
->group_exit_task
);
103 if (tsk
== sig
->curr_target
)
104 sig
->curr_target
= next_thread(tsk
);
106 * Accumulate here the counters for all threads but the
107 * group leader as they die, so they can be added into
108 * the process-wide totals when those are taken.
109 * The group leader stays around as a zombie as long
110 * as there are other threads. When it gets reaped,
111 * the exit.c code will add its counts into these totals.
112 * We won't ever get here for the group leader, since it
113 * will have been the last reference on the signal_struct.
115 sig
->utime
= cputime_add(sig
->utime
, tsk
->utime
);
116 sig
->stime
= cputime_add(sig
->stime
, tsk
->stime
);
117 sig
->gtime
= cputime_add(sig
->gtime
, tsk
->gtime
);
118 sig
->min_flt
+= tsk
->min_flt
;
119 sig
->maj_flt
+= tsk
->maj_flt
;
120 sig
->nvcsw
+= tsk
->nvcsw
;
121 sig
->nivcsw
+= tsk
->nivcsw
;
122 sig
->inblock
+= task_io_get_inblock(tsk
);
123 sig
->oublock
+= task_io_get_oublock(tsk
);
124 #ifdef CONFIG_TASK_XACCT
125 sig
->rchar
+= tsk
->rchar
;
126 sig
->wchar
+= tsk
->wchar
;
127 sig
->syscr
+= tsk
->syscr
;
128 sig
->syscw
+= tsk
->syscw
;
129 #endif /* CONFIG_TASK_XACCT */
130 #ifdef CONFIG_TASK_IO_ACCOUNTING
131 sig
->ioac
.read_bytes
+= tsk
->ioac
.read_bytes
;
132 sig
->ioac
.write_bytes
+= tsk
->ioac
.write_bytes
;
133 sig
->ioac
.cancelled_write_bytes
+=
134 tsk
->ioac
.cancelled_write_bytes
;
135 #endif /* CONFIG_TASK_IO_ACCOUNTING */
136 sig
->sum_sched_runtime
+= tsk
->se
.sum_exec_runtime
;
137 sig
= NULL
; /* Marker for below. */
140 __unhash_process(tsk
);
143 * Do this under ->siglock, we can race with another thread
144 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
146 flush_sigqueue(&tsk
->pending
);
150 spin_unlock(&sighand
->siglock
);
152 __cleanup_sighand(sighand
);
153 clear_tsk_thread_flag(tsk
,TIF_SIGPENDING
);
155 flush_sigqueue(&sig
->shared_pending
);
156 taskstats_tgid_free(sig
);
157 __cleanup_signal(sig
);
161 static void delayed_put_task_struct(struct rcu_head
*rhp
)
163 put_task_struct(container_of(rhp
, struct task_struct
, rcu
));
167 void release_task(struct task_struct
* p
)
169 struct task_struct
*leader
;
172 tracehook_prepare_release_task(p
);
173 atomic_dec(&p
->user
->processes
);
175 write_lock_irq(&tasklist_lock
);
176 tracehook_finish_release_task(p
);
180 * If we are the last non-leader member of the thread
181 * group, and the leader is zombie, then notify the
182 * group leader's parent process. (if it wants notification.)
185 leader
= p
->group_leader
;
186 if (leader
!= p
&& thread_group_empty(leader
) && leader
->exit_state
== EXIT_ZOMBIE
) {
187 BUG_ON(task_detached(leader
));
188 do_notify_parent(leader
, leader
->exit_signal
);
190 * If we were the last child thread and the leader has
191 * exited already, and the leader's parent ignores SIGCHLD,
192 * then we are the one who should release the leader.
194 * do_notify_parent() will have marked it self-reaping in
197 zap_leader
= task_detached(leader
);
200 * This maintains the invariant that release_task()
201 * only runs on a task in EXIT_DEAD, just for sanity.
204 leader
->exit_state
= EXIT_DEAD
;
207 write_unlock_irq(&tasklist_lock
);
209 call_rcu(&p
->rcu
, delayed_put_task_struct
);
212 if (unlikely(zap_leader
))
217 * This checks not only the pgrp, but falls back on the pid if no
218 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
221 * The caller must hold rcu lock or the tasklist lock.
223 struct pid
*session_of_pgrp(struct pid
*pgrp
)
225 struct task_struct
*p
;
226 struct pid
*sid
= NULL
;
228 p
= pid_task(pgrp
, PIDTYPE_PGID
);
230 p
= pid_task(pgrp
, PIDTYPE_PID
);
232 sid
= task_session(p
);
238 * Determine if a process group is "orphaned", according to the POSIX
239 * definition in 2.2.2.52. Orphaned process groups are not to be affected
240 * by terminal-generated stop signals. Newly orphaned process groups are
241 * to receive a SIGHUP and a SIGCONT.
243 * "I ask you, have you ever known what it is to be an orphan?"
245 static int will_become_orphaned_pgrp(struct pid
*pgrp
, struct task_struct
*ignored_task
)
247 struct task_struct
*p
;
249 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
250 if ((p
== ignored_task
) ||
251 (p
->exit_state
&& thread_group_empty(p
)) ||
252 is_global_init(p
->real_parent
))
255 if (task_pgrp(p
->real_parent
) != pgrp
&&
256 task_session(p
->real_parent
) == task_session(p
))
258 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
263 int is_current_pgrp_orphaned(void)
267 read_lock(&tasklist_lock
);
268 retval
= will_become_orphaned_pgrp(task_pgrp(current
), NULL
);
269 read_unlock(&tasklist_lock
);
274 static int has_stopped_jobs(struct pid
*pgrp
)
277 struct task_struct
*p
;
279 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
280 if (!task_is_stopped(p
))
284 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
289 * Check to see if any process groups have become orphaned as
290 * a result of our exiting, and if they have any stopped jobs,
291 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
294 kill_orphaned_pgrp(struct task_struct
*tsk
, struct task_struct
*parent
)
296 struct pid
*pgrp
= task_pgrp(tsk
);
297 struct task_struct
*ignored_task
= tsk
;
300 /* exit: our father is in a different pgrp than
301 * we are and we were the only connection outside.
303 parent
= tsk
->real_parent
;
305 /* reparent: our child is in a different pgrp than
306 * we are, and it was the only connection outside.
310 if (task_pgrp(parent
) != pgrp
&&
311 task_session(parent
) == task_session(tsk
) &&
312 will_become_orphaned_pgrp(pgrp
, ignored_task
) &&
313 has_stopped_jobs(pgrp
)) {
314 __kill_pgrp_info(SIGHUP
, SEND_SIG_PRIV
, pgrp
);
315 __kill_pgrp_info(SIGCONT
, SEND_SIG_PRIV
, pgrp
);
320 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
322 * If a kernel thread is launched as a result of a system call, or if
323 * it ever exits, it should generally reparent itself to kthreadd so it
324 * isn't in the way of other processes and is correctly cleaned up on exit.
326 * The various task state such as scheduling policy and priority may have
327 * been inherited from a user process, so we reset them to sane values here.
329 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
331 static void reparent_to_kthreadd(void)
333 write_lock_irq(&tasklist_lock
);
335 ptrace_unlink(current
);
336 /* Reparent to init */
337 current
->real_parent
= current
->parent
= kthreadd_task
;
338 list_move_tail(¤t
->sibling
, ¤t
->real_parent
->children
);
340 /* Set the exit signal to SIGCHLD so we signal init on exit */
341 current
->exit_signal
= SIGCHLD
;
343 if (task_nice(current
) < 0)
344 set_user_nice(current
, 0);
348 security_task_reparent_to_init(current
);
349 memcpy(current
->signal
->rlim
, init_task
.signal
->rlim
,
350 sizeof(current
->signal
->rlim
));
351 atomic_inc(&(INIT_USER
->__count
));
352 write_unlock_irq(&tasklist_lock
);
353 switch_uid(INIT_USER
);
356 void __set_special_pids(struct pid
*pid
)
358 struct task_struct
*curr
= current
->group_leader
;
359 pid_t nr
= pid_nr(pid
);
361 if (task_session(curr
) != pid
) {
362 change_pid(curr
, PIDTYPE_SID
, pid
);
363 set_task_session(curr
, nr
);
365 if (task_pgrp(curr
) != pid
) {
366 change_pid(curr
, PIDTYPE_PGID
, pid
);
367 set_task_pgrp(curr
, nr
);
371 static void set_special_pids(struct pid
*pid
)
373 write_lock_irq(&tasklist_lock
);
374 __set_special_pids(pid
);
375 write_unlock_irq(&tasklist_lock
);
379 * Let kernel threads use this to say that they
380 * allow a certain signal (since daemonize() will
381 * have disabled all of them by default).
383 int allow_signal(int sig
)
385 if (!valid_signal(sig
) || sig
< 1)
388 spin_lock_irq(¤t
->sighand
->siglock
);
389 sigdelset(¤t
->blocked
, sig
);
391 /* Kernel threads handle their own signals.
392 Let the signal code know it'll be handled, so
393 that they don't get converted to SIGKILL or
394 just silently dropped */
395 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= (void __user
*)2;
398 spin_unlock_irq(¤t
->sighand
->siglock
);
402 EXPORT_SYMBOL(allow_signal
);
404 int disallow_signal(int sig
)
406 if (!valid_signal(sig
) || sig
< 1)
409 spin_lock_irq(¤t
->sighand
->siglock
);
410 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= SIG_IGN
;
412 spin_unlock_irq(¤t
->sighand
->siglock
);
416 EXPORT_SYMBOL(disallow_signal
);
419 * Put all the gunge required to become a kernel thread without
420 * attached user resources in one place where it belongs.
423 void daemonize(const char *name
, ...)
426 struct fs_struct
*fs
;
429 va_start(args
, name
);
430 vsnprintf(current
->comm
, sizeof(current
->comm
), name
, args
);
434 * If we were started as result of loading a module, close all of the
435 * user space pages. We don't need them, and if we didn't close them
436 * they would be locked into memory.
440 * We don't want to have TIF_FREEZE set if the system-wide hibernation
441 * or suspend transition begins right now.
443 current
->flags
|= (PF_NOFREEZE
| PF_KTHREAD
);
445 if (current
->nsproxy
!= &init_nsproxy
) {
446 get_nsproxy(&init_nsproxy
);
447 switch_task_namespaces(current
, &init_nsproxy
);
449 set_special_pids(&init_struct_pid
);
450 proc_clear_tty(current
);
452 /* Block and flush all signals */
453 sigfillset(&blocked
);
454 sigprocmask(SIG_BLOCK
, &blocked
, NULL
);
455 flush_signals(current
);
457 /* Become as one with the init task */
459 exit_fs(current
); /* current->fs->count--; */
462 atomic_inc(&fs
->count
);
465 current
->files
= init_task
.files
;
466 atomic_inc(¤t
->files
->count
);
468 reparent_to_kthreadd();
471 EXPORT_SYMBOL(daemonize
);
473 static void close_files(struct files_struct
* files
)
481 * It is safe to dereference the fd table without RCU or
482 * ->file_lock because this is the last reference to the
485 fdt
= files_fdtable(files
);
489 if (i
>= fdt
->max_fds
)
491 set
= fdt
->open_fds
->fds_bits
[j
++];
494 struct file
* file
= xchg(&fdt
->fd
[i
], NULL
);
496 filp_close(file
, files
);
506 struct files_struct
*get_files_struct(struct task_struct
*task
)
508 struct files_struct
*files
;
513 atomic_inc(&files
->count
);
519 void put_files_struct(struct files_struct
*files
)
523 if (atomic_dec_and_test(&files
->count
)) {
526 * Free the fd and fdset arrays if we expanded them.
527 * If the fdtable was embedded, pass files for freeing
528 * at the end of the RCU grace period. Otherwise,
529 * you can free files immediately.
531 fdt
= files_fdtable(files
);
532 if (fdt
!= &files
->fdtab
)
533 kmem_cache_free(files_cachep
, files
);
538 void reset_files_struct(struct files_struct
*files
)
540 struct task_struct
*tsk
= current
;
541 struct files_struct
*old
;
547 put_files_struct(old
);
550 void exit_files(struct task_struct
*tsk
)
552 struct files_struct
* files
= tsk
->files
;
558 put_files_struct(files
);
562 void put_fs_struct(struct fs_struct
*fs
)
564 /* No need to hold fs->lock if we are killing it */
565 if (atomic_dec_and_test(&fs
->count
)) {
568 if (fs
->altroot
.dentry
)
569 path_put(&fs
->altroot
);
570 kmem_cache_free(fs_cachep
, fs
);
574 void exit_fs(struct task_struct
*tsk
)
576 struct fs_struct
* fs
= tsk
->fs
;
586 EXPORT_SYMBOL_GPL(exit_fs
);
588 #ifdef CONFIG_MM_OWNER
590 * Task p is exiting and it owned mm, lets find a new owner for it
593 mm_need_new_owner(struct mm_struct
*mm
, struct task_struct
*p
)
596 * If there are other users of the mm and the owner (us) is exiting
597 * we need to find a new owner to take on the responsibility.
601 if (atomic_read(&mm
->mm_users
) <= 1)
608 void mm_update_next_owner(struct mm_struct
*mm
)
610 struct task_struct
*c
, *g
, *p
= current
;
613 if (!mm_need_new_owner(mm
, p
))
616 read_lock(&tasklist_lock
);
618 * Search in the children
620 list_for_each_entry(c
, &p
->children
, sibling
) {
622 goto assign_new_owner
;
626 * Search in the siblings
628 list_for_each_entry(c
, &p
->parent
->children
, sibling
) {
630 goto assign_new_owner
;
634 * Search through everything else. We should not get
637 do_each_thread(g
, c
) {
639 goto assign_new_owner
;
640 } while_each_thread(g
, c
);
642 read_unlock(&tasklist_lock
);
649 * The task_lock protects c->mm from changing.
650 * We always want mm->owner->mm == mm
654 * Delay read_unlock() till we have the task_lock()
655 * to ensure that c does not slip away underneath us
657 read_unlock(&tasklist_lock
);
663 cgroup_mm_owner_callbacks(mm
->owner
, c
);
668 #endif /* CONFIG_MM_OWNER */
671 * Turn us into a lazy TLB process if we
674 static void exit_mm(struct task_struct
* tsk
)
676 struct mm_struct
*mm
= tsk
->mm
;
677 struct core_state
*core_state
;
683 * Serialize with any possible pending coredump.
684 * We must hold mmap_sem around checking core_state
685 * and clearing tsk->mm. The core-inducing thread
686 * will increment ->nr_threads for each thread in the
687 * group with ->mm != NULL.
689 down_read(&mm
->mmap_sem
);
690 core_state
= mm
->core_state
;
692 struct core_thread self
;
693 up_read(&mm
->mmap_sem
);
696 self
.next
= xchg(&core_state
->dumper
.next
, &self
);
698 * Implies mb(), the result of xchg() must be visible
699 * to core_state->dumper.
701 if (atomic_dec_and_test(&core_state
->nr_threads
))
702 complete(&core_state
->startup
);
705 set_task_state(tsk
, TASK_UNINTERRUPTIBLE
);
706 if (!self
.task
) /* see coredump_finish() */
710 __set_task_state(tsk
, TASK_RUNNING
);
711 down_read(&mm
->mmap_sem
);
713 atomic_inc(&mm
->mm_count
);
714 BUG_ON(mm
!= tsk
->active_mm
);
715 /* more a memory barrier than a real lock */
718 up_read(&mm
->mmap_sem
);
719 enter_lazy_tlb(mm
, current
);
720 /* We don't want this task to be frozen prematurely */
721 clear_freeze_flag(tsk
);
723 mm_update_next_owner(mm
);
728 * Return nonzero if @parent's children should reap themselves.
730 * Called with write_lock_irq(&tasklist_lock) held.
732 static int ignoring_children(struct task_struct
*parent
)
735 struct sighand_struct
*psig
= parent
->sighand
;
737 spin_lock_irqsave(&psig
->siglock
, flags
);
738 ret
= (psig
->action
[SIGCHLD
-1].sa
.sa_handler
== SIG_IGN
||
739 (psig
->action
[SIGCHLD
-1].sa
.sa_flags
& SA_NOCLDWAIT
));
740 spin_unlock_irqrestore(&psig
->siglock
, flags
);
745 * Detach all tasks we were using ptrace on.
746 * Any that need to be release_task'd are put on the @dead list.
748 * Called with write_lock(&tasklist_lock) held.
750 static void ptrace_exit(struct task_struct
*parent
, struct list_head
*dead
)
752 struct task_struct
*p
, *n
;
755 list_for_each_entry_safe(p
, n
, &parent
->ptraced
, ptrace_entry
) {
758 if (p
->exit_state
!= EXIT_ZOMBIE
)
762 * If it's a zombie, our attachedness prevented normal
763 * parent notification or self-reaping. Do notification
764 * now if it would have happened earlier. If it should
765 * reap itself, add it to the @dead list. We can't call
766 * release_task() here because we already hold tasklist_lock.
768 * If it's our own child, there is no notification to do.
769 * But if our normal children self-reap, then this child
770 * was prevented by ptrace and we must reap it now.
772 if (!task_detached(p
) && thread_group_empty(p
)) {
773 if (!same_thread_group(p
->real_parent
, parent
))
774 do_notify_parent(p
, p
->exit_signal
);
777 ign
= ignoring_children(parent
);
783 if (task_detached(p
)) {
785 * Mark it as in the process of being reaped.
787 p
->exit_state
= EXIT_DEAD
;
788 list_add(&p
->ptrace_entry
, dead
);
794 * Finish up exit-time ptrace cleanup.
796 * Called without locks.
798 static void ptrace_exit_finish(struct task_struct
*parent
,
799 struct list_head
*dead
)
801 struct task_struct
*p
, *n
;
803 BUG_ON(!list_empty(&parent
->ptraced
));
805 list_for_each_entry_safe(p
, n
, dead
, ptrace_entry
) {
806 list_del_init(&p
->ptrace_entry
);
811 static void reparent_thread(struct task_struct
*p
, struct task_struct
*father
)
813 if (p
->pdeath_signal
)
814 /* We already hold the tasklist_lock here. */
815 group_send_sig_info(p
->pdeath_signal
, SEND_SIG_NOINFO
, p
);
817 list_move_tail(&p
->sibling
, &p
->real_parent
->children
);
819 /* If this is a threaded reparent there is no need to
820 * notify anyone anything has happened.
822 if (same_thread_group(p
->real_parent
, father
))
825 /* We don't want people slaying init. */
826 if (!task_detached(p
))
827 p
->exit_signal
= SIGCHLD
;
829 /* If we'd notified the old parent about this child's death,
830 * also notify the new parent.
832 if (!ptrace_reparented(p
) &&
833 p
->exit_state
== EXIT_ZOMBIE
&&
834 !task_detached(p
) && thread_group_empty(p
))
835 do_notify_parent(p
, p
->exit_signal
);
837 kill_orphaned_pgrp(p
, father
);
841 * When we die, we re-parent all our children.
842 * Try to give them to another thread in our thread
843 * group, and if no such member exists, give it to
844 * the child reaper process (ie "init") in our pid
847 static void forget_original_parent(struct task_struct
*father
)
849 struct task_struct
*p
, *n
, *reaper
= father
;
850 LIST_HEAD(ptrace_dead
);
852 write_lock_irq(&tasklist_lock
);
855 * First clean up ptrace if we were using it.
857 ptrace_exit(father
, &ptrace_dead
);
860 reaper
= next_thread(reaper
);
861 if (reaper
== father
) {
862 reaper
= task_child_reaper(father
);
865 } while (reaper
->flags
& PF_EXITING
);
867 list_for_each_entry_safe(p
, n
, &father
->children
, sibling
) {
868 p
->real_parent
= reaper
;
869 if (p
->parent
== father
) {
871 p
->parent
= p
->real_parent
;
873 reparent_thread(p
, father
);
876 write_unlock_irq(&tasklist_lock
);
877 BUG_ON(!list_empty(&father
->children
));
879 ptrace_exit_finish(father
, &ptrace_dead
);
883 * Send signals to all our closest relatives so that they know
884 * to properly mourn us..
886 static void exit_notify(struct task_struct
*tsk
, int group_dead
)
891 * This does two things:
893 * A. Make init inherit all the child processes
894 * B. Check to see if any process groups have become orphaned
895 * as a result of our exiting, and if they have any stopped
896 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
898 forget_original_parent(tsk
);
899 exit_task_namespaces(tsk
);
901 write_lock_irq(&tasklist_lock
);
903 kill_orphaned_pgrp(tsk
->group_leader
, NULL
);
905 /* Let father know we died
907 * Thread signals are configurable, but you aren't going to use
908 * that to send signals to arbitary processes.
909 * That stops right now.
911 * If the parent exec id doesn't match the exec id we saved
912 * when we started then we know the parent has changed security
915 * If our self_exec id doesn't match our parent_exec_id then
916 * we have changed execution domain as these two values started
917 * the same after a fork.
919 if (tsk
->exit_signal
!= SIGCHLD
&& !task_detached(tsk
) &&
920 (tsk
->parent_exec_id
!= tsk
->real_parent
->self_exec_id
||
921 tsk
->self_exec_id
!= tsk
->parent_exec_id
) &&
923 tsk
->exit_signal
= SIGCHLD
;
925 /* If something other than our normal parent is ptracing us, then
926 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
927 * only has special meaning to our real parent.
929 if (!task_detached(tsk
) && thread_group_empty(tsk
)) {
930 int signal
= ptrace_reparented(tsk
) ?
931 SIGCHLD
: tsk
->exit_signal
;
932 do_notify_parent(tsk
, signal
);
933 } else if (tsk
->ptrace
) {
934 do_notify_parent(tsk
, SIGCHLD
);
938 if (task_detached(tsk
) && likely(!tsk
->ptrace
))
940 tsk
->exit_state
= state
;
942 /* mt-exec, de_thread() is waiting for us */
943 if (thread_group_leader(tsk
) &&
944 tsk
->signal
->notify_count
< 0 &&
945 tsk
->signal
->group_exit_task
)
946 wake_up_process(tsk
->signal
->group_exit_task
);
948 write_unlock_irq(&tasklist_lock
);
950 /* If the process is dead, release it - nobody will wait for it */
951 if (state
== EXIT_DEAD
)
955 #ifdef CONFIG_DEBUG_STACK_USAGE
956 static void check_stack_usage(void)
958 static DEFINE_SPINLOCK(low_water_lock
);
959 static int lowest_to_date
= THREAD_SIZE
;
960 unsigned long *n
= end_of_stack(current
);
965 free
= (unsigned long)n
- (unsigned long)end_of_stack(current
);
967 if (free
>= lowest_to_date
)
970 spin_lock(&low_water_lock
);
971 if (free
< lowest_to_date
) {
972 printk(KERN_WARNING
"%s used greatest stack depth: %lu bytes "
974 current
->comm
, free
);
975 lowest_to_date
= free
;
977 spin_unlock(&low_water_lock
);
980 static inline void check_stack_usage(void) {}
983 static inline void exit_child_reaper(struct task_struct
*tsk
)
985 if (likely(tsk
->group_leader
!= task_child_reaper(tsk
)))
988 if (tsk
->nsproxy
->pid_ns
== &init_pid_ns
)
989 panic("Attempted to kill init!");
992 * @tsk is the last thread in the 'cgroup-init' and is exiting.
993 * Terminate all remaining processes in the namespace and reap them
994 * before exiting @tsk.
996 * Note that @tsk (last thread of cgroup-init) may not necessarily
997 * be the child-reaper (i.e main thread of cgroup-init) of the
998 * namespace i.e the child_reaper may have already exited.
1000 * Even after a child_reaper exits, we let it inherit orphaned children,
1001 * because, pid_ns->child_reaper remains valid as long as there is
1002 * at least one living sub-thread in the cgroup init.
1004 * This living sub-thread of the cgroup-init will be notified when
1005 * a child inherited by the 'child-reaper' exits (do_notify_parent()
1006 * uses __group_send_sig_info()). Further, when reaping child processes,
1007 * do_wait() iterates over children of all living sub threads.
1009 * i.e even though 'child_reaper' thread is listed as the parent of the
1010 * orphaned children, any living sub-thread in the cgroup-init can
1011 * perform the role of the child_reaper.
1013 zap_pid_ns_processes(tsk
->nsproxy
->pid_ns
);
1016 NORET_TYPE
void do_exit(long code
)
1018 struct task_struct
*tsk
= current
;
1021 profile_task_exit(tsk
);
1023 WARN_ON(atomic_read(&tsk
->fs_excl
));
1025 if (unlikely(in_interrupt()))
1026 panic("Aiee, killing interrupt handler!");
1027 if (unlikely(!tsk
->pid
))
1028 panic("Attempted to kill the idle task!");
1030 tracehook_report_exit(&code
);
1033 * We're taking recursive faults here in do_exit. Safest is to just
1034 * leave this task alone and wait for reboot.
1036 if (unlikely(tsk
->flags
& PF_EXITING
)) {
1038 "Fixing recursive fault but reboot is needed!\n");
1040 * We can do this unlocked here. The futex code uses
1041 * this flag just to verify whether the pi state
1042 * cleanup has been done or not. In the worst case it
1043 * loops once more. We pretend that the cleanup was
1044 * done as there is no way to return. Either the
1045 * OWNER_DIED bit is set by now or we push the blocked
1046 * task into the wait for ever nirwana as well.
1048 tsk
->flags
|= PF_EXITPIDONE
;
1049 if (tsk
->io_context
)
1051 set_current_state(TASK_UNINTERRUPTIBLE
);
1055 exit_signals(tsk
); /* sets PF_EXITING */
1057 * tsk->flags are checked in the futex code to protect against
1058 * an exiting task cleaning up the robust pi futexes.
1061 spin_unlock_wait(&tsk
->pi_lock
);
1063 if (unlikely(in_atomic()))
1064 printk(KERN_INFO
"note: %s[%d] exited with preempt_count %d\n",
1065 current
->comm
, task_pid_nr(current
),
1068 acct_update_integrals(tsk
);
1070 update_hiwater_rss(tsk
->mm
);
1071 update_hiwater_vm(tsk
->mm
);
1073 group_dead
= atomic_dec_and_test(&tsk
->signal
->live
);
1075 exit_child_reaper(tsk
);
1076 hrtimer_cancel(&tsk
->signal
->real_timer
);
1077 exit_itimers(tsk
->signal
);
1079 acct_collect(code
, group_dead
);
1081 if (unlikely(tsk
->robust_list
))
1082 exit_robust_list(tsk
);
1083 #ifdef CONFIG_COMPAT
1084 if (unlikely(tsk
->compat_robust_list
))
1085 compat_exit_robust_list(tsk
);
1090 if (unlikely(tsk
->audit_context
))
1093 tsk
->exit_code
= code
;
1094 taskstats_exit(tsk
, group_dead
);
1103 check_stack_usage();
1105 cgroup_exit(tsk
, 1);
1108 if (group_dead
&& tsk
->signal
->leader
)
1109 disassociate_ctty(1);
1111 module_put(task_thread_info(tsk
)->exec_domain
->module
);
1113 module_put(tsk
->binfmt
->module
);
1115 proc_exit_connector(tsk
);
1116 exit_notify(tsk
, group_dead
);
1118 mpol_put(tsk
->mempolicy
);
1119 tsk
->mempolicy
= NULL
;
1123 * This must happen late, after the PID is not
1126 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
1127 exit_pi_state_list(tsk
);
1128 if (unlikely(current
->pi_state_cache
))
1129 kfree(current
->pi_state_cache
);
1132 * Make sure we are holding no locks:
1134 debug_check_no_locks_held(tsk
);
1136 * We can do this unlocked here. The futex code uses this flag
1137 * just to verify whether the pi state cleanup has been done
1138 * or not. In the worst case it loops once more.
1140 tsk
->flags
|= PF_EXITPIDONE
;
1142 if (tsk
->io_context
)
1145 if (tsk
->splice_pipe
)
1146 __free_pipe_info(tsk
->splice_pipe
);
1149 /* causes final put_task_struct in finish_task_switch(). */
1150 tsk
->state
= TASK_DEAD
;
1154 /* Avoid "noreturn function does return". */
1156 cpu_relax(); /* For when BUG is null */
1159 EXPORT_SYMBOL_GPL(do_exit
);
1161 NORET_TYPE
void complete_and_exit(struct completion
*comp
, long code
)
1169 EXPORT_SYMBOL(complete_and_exit
);
1171 asmlinkage
long sys_exit(int error_code
)
1173 do_exit((error_code
&0xff)<<8);
1177 * Take down every thread in the group. This is called by fatal signals
1178 * as well as by sys_exit_group (below).
1181 do_group_exit(int exit_code
)
1183 struct signal_struct
*sig
= current
->signal
;
1185 BUG_ON(exit_code
& 0x80); /* core dumps don't get here */
1187 if (signal_group_exit(sig
))
1188 exit_code
= sig
->group_exit_code
;
1189 else if (!thread_group_empty(current
)) {
1190 struct sighand_struct
*const sighand
= current
->sighand
;
1191 spin_lock_irq(&sighand
->siglock
);
1192 if (signal_group_exit(sig
))
1193 /* Another thread got here before we took the lock. */
1194 exit_code
= sig
->group_exit_code
;
1196 sig
->group_exit_code
= exit_code
;
1197 sig
->flags
= SIGNAL_GROUP_EXIT
;
1198 zap_other_threads(current
);
1200 spin_unlock_irq(&sighand
->siglock
);
1208 * this kills every thread in the thread group. Note that any externally
1209 * wait4()-ing process will get the correct exit code - even if this
1210 * thread is not the thread group leader.
1212 asmlinkage
void sys_exit_group(int error_code
)
1214 do_group_exit((error_code
& 0xff) << 8);
1217 static struct pid
*task_pid_type(struct task_struct
*task
, enum pid_type type
)
1219 struct pid
*pid
= NULL
;
1220 if (type
== PIDTYPE_PID
)
1221 pid
= task
->pids
[type
].pid
;
1222 else if (type
< PIDTYPE_MAX
)
1223 pid
= task
->group_leader
->pids
[type
].pid
;
1227 static int eligible_child(enum pid_type type
, struct pid
*pid
, int options
,
1228 struct task_struct
*p
)
1232 if (type
< PIDTYPE_MAX
) {
1233 if (task_pid_type(p
, type
) != pid
)
1237 /* Wait for all children (clone and not) if __WALL is set;
1238 * otherwise, wait for clone children *only* if __WCLONE is
1239 * set; otherwise, wait for non-clone children *only*. (Note:
1240 * A "clone" child here is one that reports to its parent
1241 * using a signal other than SIGCHLD.) */
1242 if (((p
->exit_signal
!= SIGCHLD
) ^ ((options
& __WCLONE
) != 0))
1243 && !(options
& __WALL
))
1246 err
= security_task_wait(p
);
1253 static int wait_noreap_copyout(struct task_struct
*p
, pid_t pid
, uid_t uid
,
1254 int why
, int status
,
1255 struct siginfo __user
*infop
,
1256 struct rusage __user
*rusagep
)
1258 int retval
= rusagep
? getrusage(p
, RUSAGE_BOTH
, rusagep
) : 0;
1262 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1264 retval
= put_user(0, &infop
->si_errno
);
1266 retval
= put_user((short)why
, &infop
->si_code
);
1268 retval
= put_user(pid
, &infop
->si_pid
);
1270 retval
= put_user(uid
, &infop
->si_uid
);
1272 retval
= put_user(status
, &infop
->si_status
);
1279 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1280 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1281 * the lock and this task is uninteresting. If we return nonzero, we have
1282 * released the lock and the system call should return.
1284 static int wait_task_zombie(struct task_struct
*p
, int options
,
1285 struct siginfo __user
*infop
,
1286 int __user
*stat_addr
, struct rusage __user
*ru
)
1288 unsigned long state
;
1289 int retval
, status
, traced
;
1290 pid_t pid
= task_pid_vnr(p
);
1292 if (!likely(options
& WEXITED
))
1295 if (unlikely(options
& WNOWAIT
)) {
1297 int exit_code
= p
->exit_code
;
1301 read_unlock(&tasklist_lock
);
1302 if ((exit_code
& 0x7f) == 0) {
1304 status
= exit_code
>> 8;
1306 why
= (exit_code
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1307 status
= exit_code
& 0x7f;
1309 return wait_noreap_copyout(p
, pid
, uid
, why
,
1314 * Try to move the task's state to DEAD
1315 * only one thread is allowed to do this:
1317 state
= xchg(&p
->exit_state
, EXIT_DEAD
);
1318 if (state
!= EXIT_ZOMBIE
) {
1319 BUG_ON(state
!= EXIT_DEAD
);
1323 traced
= ptrace_reparented(p
);
1325 if (likely(!traced
)) {
1326 struct signal_struct
*psig
;
1327 struct signal_struct
*sig
;
1330 * The resource counters for the group leader are in its
1331 * own task_struct. Those for dead threads in the group
1332 * are in its signal_struct, as are those for the child
1333 * processes it has previously reaped. All these
1334 * accumulate in the parent's signal_struct c* fields.
1336 * We don't bother to take a lock here to protect these
1337 * p->signal fields, because they are only touched by
1338 * __exit_signal, which runs with tasklist_lock
1339 * write-locked anyway, and so is excluded here. We do
1340 * need to protect the access to p->parent->signal fields,
1341 * as other threads in the parent group can be right
1342 * here reaping other children at the same time.
1344 spin_lock_irq(&p
->parent
->sighand
->siglock
);
1345 psig
= p
->parent
->signal
;
1348 cputime_add(psig
->cutime
,
1349 cputime_add(p
->utime
,
1350 cputime_add(sig
->utime
,
1353 cputime_add(psig
->cstime
,
1354 cputime_add(p
->stime
,
1355 cputime_add(sig
->stime
,
1358 cputime_add(psig
->cgtime
,
1359 cputime_add(p
->gtime
,
1360 cputime_add(sig
->gtime
,
1363 p
->min_flt
+ sig
->min_flt
+ sig
->cmin_flt
;
1365 p
->maj_flt
+ sig
->maj_flt
+ sig
->cmaj_flt
;
1367 p
->nvcsw
+ sig
->nvcsw
+ sig
->cnvcsw
;
1369 p
->nivcsw
+ sig
->nivcsw
+ sig
->cnivcsw
;
1371 task_io_get_inblock(p
) +
1372 sig
->inblock
+ sig
->cinblock
;
1374 task_io_get_oublock(p
) +
1375 sig
->oublock
+ sig
->coublock
;
1376 #ifdef CONFIG_TASK_XACCT
1377 psig
->rchar
+= p
->rchar
+ sig
->rchar
;
1378 psig
->wchar
+= p
->wchar
+ sig
->wchar
;
1379 psig
->syscr
+= p
->syscr
+ sig
->syscr
;
1380 psig
->syscw
+= p
->syscw
+ sig
->syscw
;
1381 #endif /* CONFIG_TASK_XACCT */
1382 #ifdef CONFIG_TASK_IO_ACCOUNTING
1383 psig
->ioac
.read_bytes
+=
1384 p
->ioac
.read_bytes
+ sig
->ioac
.read_bytes
;
1385 psig
->ioac
.write_bytes
+=
1386 p
->ioac
.write_bytes
+ sig
->ioac
.write_bytes
;
1387 psig
->ioac
.cancelled_write_bytes
+=
1388 p
->ioac
.cancelled_write_bytes
+
1389 sig
->ioac
.cancelled_write_bytes
;
1390 #endif /* CONFIG_TASK_IO_ACCOUNTING */
1391 spin_unlock_irq(&p
->parent
->sighand
->siglock
);
1395 * Now we are sure this task is interesting, and no other
1396 * thread can reap it because we set its state to EXIT_DEAD.
1398 read_unlock(&tasklist_lock
);
1400 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1401 status
= (p
->signal
->flags
& SIGNAL_GROUP_EXIT
)
1402 ? p
->signal
->group_exit_code
: p
->exit_code
;
1403 if (!retval
&& stat_addr
)
1404 retval
= put_user(status
, stat_addr
);
1405 if (!retval
&& infop
)
1406 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1407 if (!retval
&& infop
)
1408 retval
= put_user(0, &infop
->si_errno
);
1409 if (!retval
&& infop
) {
1412 if ((status
& 0x7f) == 0) {
1416 why
= (status
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1419 retval
= put_user((short)why
, &infop
->si_code
);
1421 retval
= put_user(status
, &infop
->si_status
);
1423 if (!retval
&& infop
)
1424 retval
= put_user(pid
, &infop
->si_pid
);
1425 if (!retval
&& infop
)
1426 retval
= put_user(p
->uid
, &infop
->si_uid
);
1431 write_lock_irq(&tasklist_lock
);
1432 /* We dropped tasklist, ptracer could die and untrace */
1435 * If this is not a detached task, notify the parent.
1436 * If it's still not detached after that, don't release
1439 if (!task_detached(p
)) {
1440 do_notify_parent(p
, p
->exit_signal
);
1441 if (!task_detached(p
)) {
1442 p
->exit_state
= EXIT_ZOMBIE
;
1446 write_unlock_irq(&tasklist_lock
);
1455 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1456 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1457 * the lock and this task is uninteresting. If we return nonzero, we have
1458 * released the lock and the system call should return.
1460 static int wait_task_stopped(int ptrace
, struct task_struct
*p
,
1461 int options
, struct siginfo __user
*infop
,
1462 int __user
*stat_addr
, struct rusage __user
*ru
)
1464 int retval
, exit_code
, why
;
1465 uid_t uid
= 0; /* unneeded, required by compiler */
1468 if (!(options
& WUNTRACED
))
1472 spin_lock_irq(&p
->sighand
->siglock
);
1474 if (unlikely(!task_is_stopped_or_traced(p
)))
1477 if (!ptrace
&& p
->signal
->group_stop_count
> 0)
1479 * A group stop is in progress and this is the group leader.
1480 * We won't report until all threads have stopped.
1484 exit_code
= p
->exit_code
;
1488 if (!unlikely(options
& WNOWAIT
))
1493 spin_unlock_irq(&p
->sighand
->siglock
);
1498 * Now we are pretty sure this task is interesting.
1499 * Make sure it doesn't get reaped out from under us while we
1500 * give up the lock and then examine it below. We don't want to
1501 * keep holding onto the tasklist_lock while we call getrusage and
1502 * possibly take page faults for user memory.
1505 pid
= task_pid_vnr(p
);
1506 why
= ptrace
? CLD_TRAPPED
: CLD_STOPPED
;
1507 read_unlock(&tasklist_lock
);
1509 if (unlikely(options
& WNOWAIT
))
1510 return wait_noreap_copyout(p
, pid
, uid
,
1514 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1515 if (!retval
&& stat_addr
)
1516 retval
= put_user((exit_code
<< 8) | 0x7f, stat_addr
);
1517 if (!retval
&& infop
)
1518 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1519 if (!retval
&& infop
)
1520 retval
= put_user(0, &infop
->si_errno
);
1521 if (!retval
&& infop
)
1522 retval
= put_user((short)why
, &infop
->si_code
);
1523 if (!retval
&& infop
)
1524 retval
= put_user(exit_code
, &infop
->si_status
);
1525 if (!retval
&& infop
)
1526 retval
= put_user(pid
, &infop
->si_pid
);
1527 if (!retval
&& infop
)
1528 retval
= put_user(uid
, &infop
->si_uid
);
1538 * Handle do_wait work for one task in a live, non-stopped state.
1539 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1540 * the lock and this task is uninteresting. If we return nonzero, we have
1541 * released the lock and the system call should return.
1543 static int wait_task_continued(struct task_struct
*p
, int options
,
1544 struct siginfo __user
*infop
,
1545 int __user
*stat_addr
, struct rusage __user
*ru
)
1551 if (!unlikely(options
& WCONTINUED
))
1554 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
))
1557 spin_lock_irq(&p
->sighand
->siglock
);
1558 /* Re-check with the lock held. */
1559 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
)) {
1560 spin_unlock_irq(&p
->sighand
->siglock
);
1563 if (!unlikely(options
& WNOWAIT
))
1564 p
->signal
->flags
&= ~SIGNAL_STOP_CONTINUED
;
1565 spin_unlock_irq(&p
->sighand
->siglock
);
1567 pid
= task_pid_vnr(p
);
1570 read_unlock(&tasklist_lock
);
1573 retval
= ru
? getrusage(p
, RUSAGE_BOTH
, ru
) : 0;
1575 if (!retval
&& stat_addr
)
1576 retval
= put_user(0xffff, stat_addr
);
1580 retval
= wait_noreap_copyout(p
, pid
, uid
,
1581 CLD_CONTINUED
, SIGCONT
,
1583 BUG_ON(retval
== 0);
1590 * Consider @p for a wait by @parent.
1592 * -ECHILD should be in *@notask_error before the first call.
1593 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1594 * Returns zero if the search for a child should continue;
1595 * then *@notask_error is 0 if @p is an eligible child,
1596 * or another error from security_task_wait(), or still -ECHILD.
1598 static int wait_consider_task(struct task_struct
*parent
, int ptrace
,
1599 struct task_struct
*p
, int *notask_error
,
1600 enum pid_type type
, struct pid
*pid
, int options
,
1601 struct siginfo __user
*infop
,
1602 int __user
*stat_addr
, struct rusage __user
*ru
)
1604 int ret
= eligible_child(type
, pid
, options
, p
);
1608 if (unlikely(ret
< 0)) {
1610 * If we have not yet seen any eligible child,
1611 * then let this error code replace -ECHILD.
1612 * A permission error will give the user a clue
1613 * to look for security policy problems, rather
1614 * than for mysterious wait bugs.
1617 *notask_error
= ret
;
1620 if (likely(!ptrace
) && unlikely(p
->ptrace
)) {
1622 * This child is hidden by ptrace.
1623 * We aren't allowed to see it now, but eventually we will.
1629 if (p
->exit_state
== EXIT_DEAD
)
1633 * We don't reap group leaders with subthreads.
1635 if (p
->exit_state
== EXIT_ZOMBIE
&& !delay_group_leader(p
))
1636 return wait_task_zombie(p
, options
, infop
, stat_addr
, ru
);
1639 * It's stopped or running now, so it might
1640 * later continue, exit, or stop again.
1644 if (task_is_stopped_or_traced(p
))
1645 return wait_task_stopped(ptrace
, p
, options
,
1646 infop
, stat_addr
, ru
);
1648 return wait_task_continued(p
, options
, infop
, stat_addr
, ru
);
1652 * Do the work of do_wait() for one thread in the group, @tsk.
1654 * -ECHILD should be in *@notask_error before the first call.
1655 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1656 * Returns zero if the search for a child should continue; then
1657 * *@notask_error is 0 if there were any eligible children,
1658 * or another error from security_task_wait(), or still -ECHILD.
1660 static int do_wait_thread(struct task_struct
*tsk
, int *notask_error
,
1661 enum pid_type type
, struct pid
*pid
, int options
,
1662 struct siginfo __user
*infop
, int __user
*stat_addr
,
1663 struct rusage __user
*ru
)
1665 struct task_struct
*p
;
1667 list_for_each_entry(p
, &tsk
->children
, sibling
) {
1669 * Do not consider detached threads.
1671 if (!task_detached(p
)) {
1672 int ret
= wait_consider_task(tsk
, 0, p
, notask_error
,
1674 infop
, stat_addr
, ru
);
1683 static int ptrace_do_wait(struct task_struct
*tsk
, int *notask_error
,
1684 enum pid_type type
, struct pid
*pid
, int options
,
1685 struct siginfo __user
*infop
, int __user
*stat_addr
,
1686 struct rusage __user
*ru
)
1688 struct task_struct
*p
;
1691 * Traditionally we see ptrace'd stopped tasks regardless of options.
1693 options
|= WUNTRACED
;
1695 list_for_each_entry(p
, &tsk
->ptraced
, ptrace_entry
) {
1696 int ret
= wait_consider_task(tsk
, 1, p
, notask_error
,
1698 infop
, stat_addr
, ru
);
1706 static long do_wait(enum pid_type type
, struct pid
*pid
, int options
,
1707 struct siginfo __user
*infop
, int __user
*stat_addr
,
1708 struct rusage __user
*ru
)
1710 DECLARE_WAITQUEUE(wait
, current
);
1711 struct task_struct
*tsk
;
1714 add_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1717 * If there is nothing that can match our critiera just get out.
1718 * We will clear @retval to zero if we see any child that might later
1719 * match our criteria, even if we are not able to reap it yet.
1722 if ((type
< PIDTYPE_MAX
) && (!pid
|| hlist_empty(&pid
->tasks
[type
])))
1725 current
->state
= TASK_INTERRUPTIBLE
;
1726 read_lock(&tasklist_lock
);
1729 int tsk_result
= do_wait_thread(tsk
, &retval
,
1731 infop
, stat_addr
, ru
);
1733 tsk_result
= ptrace_do_wait(tsk
, &retval
,
1735 infop
, stat_addr
, ru
);
1738 * tasklist_lock is unlocked and we have a final result.
1740 retval
= tsk_result
;
1744 if (options
& __WNOTHREAD
)
1746 tsk
= next_thread(tsk
);
1747 BUG_ON(tsk
->signal
!= current
->signal
);
1748 } while (tsk
!= current
);
1749 read_unlock(&tasklist_lock
);
1751 if (!retval
&& !(options
& WNOHANG
)) {
1752 retval
= -ERESTARTSYS
;
1753 if (!signal_pending(current
)) {
1760 current
->state
= TASK_RUNNING
;
1761 remove_wait_queue(¤t
->signal
->wait_chldexit
,&wait
);
1767 * For a WNOHANG return, clear out all the fields
1768 * we would set so the user can easily tell the
1772 retval
= put_user(0, &infop
->si_signo
);
1774 retval
= put_user(0, &infop
->si_errno
);
1776 retval
= put_user(0, &infop
->si_code
);
1778 retval
= put_user(0, &infop
->si_pid
);
1780 retval
= put_user(0, &infop
->si_uid
);
1782 retval
= put_user(0, &infop
->si_status
);
1788 asmlinkage
long sys_waitid(int which
, pid_t upid
,
1789 struct siginfo __user
*infop
, int options
,
1790 struct rusage __user
*ru
)
1792 struct pid
*pid
= NULL
;
1796 if (options
& ~(WNOHANG
|WNOWAIT
|WEXITED
|WSTOPPED
|WCONTINUED
))
1798 if (!(options
& (WEXITED
|WSTOPPED
|WCONTINUED
)))
1811 type
= PIDTYPE_PGID
;
1819 if (type
< PIDTYPE_MAX
)
1820 pid
= find_get_pid(upid
);
1821 ret
= do_wait(type
, pid
, options
, infop
, NULL
, ru
);
1824 /* avoid REGPARM breakage on x86: */
1825 asmlinkage_protect(5, ret
, which
, upid
, infop
, options
, ru
);
1829 asmlinkage
long sys_wait4(pid_t upid
, int __user
*stat_addr
,
1830 int options
, struct rusage __user
*ru
)
1832 struct pid
*pid
= NULL
;
1836 if (options
& ~(WNOHANG
|WUNTRACED
|WCONTINUED
|
1837 __WNOTHREAD
|__WCLONE
|__WALL
))
1842 else if (upid
< 0) {
1843 type
= PIDTYPE_PGID
;
1844 pid
= find_get_pid(-upid
);
1845 } else if (upid
== 0) {
1846 type
= PIDTYPE_PGID
;
1847 pid
= get_pid(task_pgrp(current
));
1848 } else /* upid > 0 */ {
1850 pid
= find_get_pid(upid
);
1853 ret
= do_wait(type
, pid
, options
| WEXITED
, NULL
, stat_addr
, ru
);
1856 /* avoid REGPARM breakage on x86: */
1857 asmlinkage_protect(4, ret
, upid
, stat_addr
, options
, ru
);
1861 #ifdef __ARCH_WANT_SYS_WAITPID
1864 * sys_waitpid() remains for compatibility. waitpid() should be
1865 * implemented by calling sys_wait4() from libc.a.
1867 asmlinkage
long sys_waitpid(pid_t pid
, int __user
*stat_addr
, int options
)
1869 return sys_wait4(pid
, stat_addr
, options
, NULL
);
This page took 0.068632 seconds and 5 git commands to generate.