]>
git.proxmox.com Git - mirror_ubuntu-artful-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/iocontext.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/pipe_fs_i.h>
43 #include <linux/audit.h> /* for audit_free() */
44 #include <linux/resource.h>
45 #include <linux/blkdev.h>
46 #include <linux/task_io_accounting_ops.h>
47 #include <linux/tracehook.h>
48 #include <linux/fs_struct.h>
49 #include <linux/init_task.h>
50 #include <linux/perf_event.h>
51 #include <trace/events/sched.h>
52 #include <linux/hw_breakpoint.h>
53 #include <linux/oom.h>
55 #include <asm/uaccess.h>
56 #include <asm/unistd.h>
57 #include <asm/pgtable.h>
58 #include <asm/mmu_context.h>
60 static void exit_mm(struct task_struct
* tsk
);
62 static void __unhash_process(struct task_struct
*p
, bool group_dead
)
65 detach_pid(p
, PIDTYPE_PID
);
67 detach_pid(p
, PIDTYPE_PGID
);
68 detach_pid(p
, PIDTYPE_SID
);
70 list_del_rcu(&p
->tasks
);
71 list_del_init(&p
->sibling
);
72 __this_cpu_dec(process_counts
);
74 list_del_rcu(&p
->thread_group
);
78 * This function expects the tasklist_lock write-locked.
80 static void __exit_signal(struct task_struct
*tsk
)
82 struct signal_struct
*sig
= tsk
->signal
;
83 bool group_dead
= thread_group_leader(tsk
);
84 struct sighand_struct
*sighand
;
85 struct tty_struct
*uninitialized_var(tty
);
87 sighand
= rcu_dereference_check(tsk
->sighand
,
88 lockdep_tasklist_lock_is_held());
89 spin_lock(&sighand
->siglock
);
91 posix_cpu_timers_exit(tsk
);
93 posix_cpu_timers_exit_group(tsk
);
98 * This can only happen if the caller is de_thread().
99 * FIXME: this is the temporary hack, we should teach
100 * posix-cpu-timers to handle this case correctly.
102 if (unlikely(has_group_leader_pid(tsk
)))
103 posix_cpu_timers_exit_group(tsk
);
106 * If there is any task waiting for the group exit
109 if (sig
->notify_count
> 0 && !--sig
->notify_count
)
110 wake_up_process(sig
->group_exit_task
);
112 if (tsk
== sig
->curr_target
)
113 sig
->curr_target
= next_thread(tsk
);
115 * Accumulate here the counters for all threads but the
116 * group leader as they die, so they can be added into
117 * the process-wide totals when those are taken.
118 * The group leader stays around as a zombie as long
119 * as there are other threads. When it gets reaped,
120 * the exit.c code will add its counts into these totals.
121 * We won't ever get here for the group leader, since it
122 * will have been the last reference on the signal_struct.
124 sig
->utime
= cputime_add(sig
->utime
, tsk
->utime
);
125 sig
->stime
= cputime_add(sig
->stime
, tsk
->stime
);
126 sig
->gtime
= cputime_add(sig
->gtime
, tsk
->gtime
);
127 sig
->min_flt
+= tsk
->min_flt
;
128 sig
->maj_flt
+= tsk
->maj_flt
;
129 sig
->nvcsw
+= tsk
->nvcsw
;
130 sig
->nivcsw
+= tsk
->nivcsw
;
131 sig
->inblock
+= task_io_get_inblock(tsk
);
132 sig
->oublock
+= task_io_get_oublock(tsk
);
133 task_io_accounting_add(&sig
->ioac
, &tsk
->ioac
);
134 sig
->sum_sched_runtime
+= tsk
->se
.sum_exec_runtime
;
138 __unhash_process(tsk
, group_dead
);
141 * Do this under ->siglock, we can race with another thread
142 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
144 flush_sigqueue(&tsk
->pending
);
146 spin_unlock(&sighand
->siglock
);
148 __cleanup_sighand(sighand
);
149 clear_tsk_thread_flag(tsk
,TIF_SIGPENDING
);
151 flush_sigqueue(&sig
->shared_pending
);
156 static void delayed_put_task_struct(struct rcu_head
*rhp
)
158 struct task_struct
*tsk
= container_of(rhp
, struct task_struct
, rcu
);
160 perf_event_delayed_put(tsk
);
161 trace_sched_process_free(tsk
);
162 put_task_struct(tsk
);
166 void release_task(struct task_struct
* p
)
168 struct task_struct
*leader
;
171 /* don't need to get the RCU readlock here - the process is dead and
172 * can't be modifying its own credentials. But shut RCU-lockdep up */
174 atomic_dec(&__task_cred(p
)->user
->processes
);
179 write_lock_irq(&tasklist_lock
);
180 ptrace_release_task(p
);
184 * If we are the last non-leader member of the thread
185 * group, and the leader is zombie, then notify the
186 * group leader's parent process. (if it wants notification.)
189 leader
= p
->group_leader
;
190 if (leader
!= p
&& thread_group_empty(leader
) && leader
->exit_state
== EXIT_ZOMBIE
) {
192 * If we were the last child thread and the leader has
193 * exited already, and the leader's parent ignores SIGCHLD,
194 * then we are the one who should release the leader.
196 zap_leader
= do_notify_parent(leader
, leader
->exit_signal
);
198 leader
->exit_state
= EXIT_DEAD
;
201 write_unlock_irq(&tasklist_lock
);
203 call_rcu(&p
->rcu
, delayed_put_task_struct
);
206 if (unlikely(zap_leader
))
211 * This checks not only the pgrp, but falls back on the pid if no
212 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
215 * The caller must hold rcu lock or the tasklist lock.
217 struct pid
*session_of_pgrp(struct pid
*pgrp
)
219 struct task_struct
*p
;
220 struct pid
*sid
= NULL
;
222 p
= pid_task(pgrp
, PIDTYPE_PGID
);
224 p
= pid_task(pgrp
, PIDTYPE_PID
);
226 sid
= task_session(p
);
232 * Determine if a process group is "orphaned", according to the POSIX
233 * definition in 2.2.2.52. Orphaned process groups are not to be affected
234 * by terminal-generated stop signals. Newly orphaned process groups are
235 * to receive a SIGHUP and a SIGCONT.
237 * "I ask you, have you ever known what it is to be an orphan?"
239 static int will_become_orphaned_pgrp(struct pid
*pgrp
, struct task_struct
*ignored_task
)
241 struct task_struct
*p
;
243 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
244 if ((p
== ignored_task
) ||
245 (p
->exit_state
&& thread_group_empty(p
)) ||
246 is_global_init(p
->real_parent
))
249 if (task_pgrp(p
->real_parent
) != pgrp
&&
250 task_session(p
->real_parent
) == task_session(p
))
252 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
257 int is_current_pgrp_orphaned(void)
261 read_lock(&tasklist_lock
);
262 retval
= will_become_orphaned_pgrp(task_pgrp(current
), NULL
);
263 read_unlock(&tasklist_lock
);
268 static bool has_stopped_jobs(struct pid
*pgrp
)
270 struct task_struct
*p
;
272 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
273 if (p
->signal
->flags
& SIGNAL_STOP_STOPPED
)
275 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
281 * Check to see if any process groups have become orphaned as
282 * a result of our exiting, and if they have any stopped jobs,
283 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
286 kill_orphaned_pgrp(struct task_struct
*tsk
, struct task_struct
*parent
)
288 struct pid
*pgrp
= task_pgrp(tsk
);
289 struct task_struct
*ignored_task
= tsk
;
292 /* exit: our father is in a different pgrp than
293 * we are and we were the only connection outside.
295 parent
= tsk
->real_parent
;
297 /* reparent: our child is in a different pgrp than
298 * we are, and it was the only connection outside.
302 if (task_pgrp(parent
) != pgrp
&&
303 task_session(parent
) == task_session(tsk
) &&
304 will_become_orphaned_pgrp(pgrp
, ignored_task
) &&
305 has_stopped_jobs(pgrp
)) {
306 __kill_pgrp_info(SIGHUP
, SEND_SIG_PRIV
, pgrp
);
307 __kill_pgrp_info(SIGCONT
, SEND_SIG_PRIV
, pgrp
);
312 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
314 * If a kernel thread is launched as a result of a system call, or if
315 * it ever exits, it should generally reparent itself to kthreadd so it
316 * isn't in the way of other processes and is correctly cleaned up on exit.
318 * The various task state such as scheduling policy and priority may have
319 * been inherited from a user process, so we reset them to sane values here.
321 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
323 static void reparent_to_kthreadd(void)
325 write_lock_irq(&tasklist_lock
);
327 ptrace_unlink(current
);
328 /* Reparent to init */
329 current
->real_parent
= current
->parent
= kthreadd_task
;
330 list_move_tail(¤t
->sibling
, ¤t
->real_parent
->children
);
332 /* Set the exit signal to SIGCHLD so we signal init on exit */
333 current
->exit_signal
= SIGCHLD
;
335 if (task_nice(current
) < 0)
336 set_user_nice(current
, 0);
340 memcpy(current
->signal
->rlim
, init_task
.signal
->rlim
,
341 sizeof(current
->signal
->rlim
));
343 atomic_inc(&init_cred
.usage
);
344 commit_creds(&init_cred
);
345 write_unlock_irq(&tasklist_lock
);
348 void __set_special_pids(struct pid
*pid
)
350 struct task_struct
*curr
= current
->group_leader
;
352 if (task_session(curr
) != pid
)
353 change_pid(curr
, PIDTYPE_SID
, pid
);
355 if (task_pgrp(curr
) != pid
)
356 change_pid(curr
, PIDTYPE_PGID
, pid
);
359 static void set_special_pids(struct pid
*pid
)
361 write_lock_irq(&tasklist_lock
);
362 __set_special_pids(pid
);
363 write_unlock_irq(&tasklist_lock
);
367 * Let kernel threads use this to say that they allow a certain signal.
368 * Must not be used if kthread was cloned with CLONE_SIGHAND.
370 int allow_signal(int sig
)
372 if (!valid_signal(sig
) || sig
< 1)
375 spin_lock_irq(¤t
->sighand
->siglock
);
376 /* This is only needed for daemonize()'ed kthreads */
377 sigdelset(¤t
->blocked
, sig
);
379 * Kernel threads handle their own signals. Let the signal code
380 * know it'll be handled, so that they don't get converted to
381 * SIGKILL or just silently dropped.
383 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= (void __user
*)2;
385 spin_unlock_irq(¤t
->sighand
->siglock
);
389 EXPORT_SYMBOL(allow_signal
);
391 int disallow_signal(int sig
)
393 if (!valid_signal(sig
) || sig
< 1)
396 spin_lock_irq(¤t
->sighand
->siglock
);
397 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= SIG_IGN
;
399 spin_unlock_irq(¤t
->sighand
->siglock
);
403 EXPORT_SYMBOL(disallow_signal
);
406 * Put all the gunge required to become a kernel thread without
407 * attached user resources in one place where it belongs.
410 void daemonize(const char *name
, ...)
415 va_start(args
, name
);
416 vsnprintf(current
->comm
, sizeof(current
->comm
), name
, args
);
420 * If we were started as result of loading a module, close all of the
421 * user space pages. We don't need them, and if we didn't close them
422 * they would be locked into memory.
426 * We don't want to have TIF_FREEZE set if the system-wide hibernation
427 * or suspend transition begins right now.
429 current
->flags
|= (PF_NOFREEZE
| PF_KTHREAD
);
431 if (current
->nsproxy
!= &init_nsproxy
) {
432 get_nsproxy(&init_nsproxy
);
433 switch_task_namespaces(current
, &init_nsproxy
);
435 set_special_pids(&init_struct_pid
);
436 proc_clear_tty(current
);
438 /* Block and flush all signals */
439 sigfillset(&blocked
);
440 sigprocmask(SIG_BLOCK
, &blocked
, NULL
);
441 flush_signals(current
);
443 /* Become as one with the init task */
445 daemonize_fs_struct();
447 current
->files
= init_task
.files
;
448 atomic_inc(¤t
->files
->count
);
450 reparent_to_kthreadd();
453 EXPORT_SYMBOL(daemonize
);
455 static void close_files(struct files_struct
* files
)
463 * It is safe to dereference the fd table without RCU or
464 * ->file_lock because this is the last reference to the
465 * files structure. But use RCU to shut RCU-lockdep up.
468 fdt
= files_fdtable(files
);
473 if (i
>= fdt
->max_fds
)
475 set
= fdt
->open_fds
->fds_bits
[j
++];
478 struct file
* file
= xchg(&fdt
->fd
[i
], NULL
);
480 filp_close(file
, files
);
490 struct files_struct
*get_files_struct(struct task_struct
*task
)
492 struct files_struct
*files
;
497 atomic_inc(&files
->count
);
503 void put_files_struct(struct files_struct
*files
)
507 if (atomic_dec_and_test(&files
->count
)) {
510 * Free the fd and fdset arrays if we expanded them.
511 * If the fdtable was embedded, pass files for freeing
512 * at the end of the RCU grace period. Otherwise,
513 * you can free files immediately.
516 fdt
= files_fdtable(files
);
517 if (fdt
!= &files
->fdtab
)
518 kmem_cache_free(files_cachep
, files
);
524 void reset_files_struct(struct files_struct
*files
)
526 struct task_struct
*tsk
= current
;
527 struct files_struct
*old
;
533 put_files_struct(old
);
536 void exit_files(struct task_struct
*tsk
)
538 struct files_struct
* files
= tsk
->files
;
544 put_files_struct(files
);
548 #ifdef CONFIG_MM_OWNER
550 * A task is exiting. If it owned this mm, find a new owner for the mm.
552 void mm_update_next_owner(struct mm_struct
*mm
)
554 struct task_struct
*c
, *g
, *p
= current
;
558 * If the exiting or execing task is not the owner, it's
559 * someone else's problem.
564 * The current owner is exiting/execing and there are no other
565 * candidates. Do not leave the mm pointing to a possibly
566 * freed task structure.
568 if (atomic_read(&mm
->mm_users
) <= 1) {
573 read_lock(&tasklist_lock
);
575 * Search in the children
577 list_for_each_entry(c
, &p
->children
, sibling
) {
579 goto assign_new_owner
;
583 * Search in the siblings
585 list_for_each_entry(c
, &p
->real_parent
->children
, sibling
) {
587 goto assign_new_owner
;
591 * Search through everything else. We should not get
594 do_each_thread(g
, c
) {
596 goto assign_new_owner
;
597 } while_each_thread(g
, c
);
599 read_unlock(&tasklist_lock
);
601 * We found no owner yet mm_users > 1: this implies that we are
602 * most likely racing with swapoff (try_to_unuse()) or /proc or
603 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
612 * The task_lock protects c->mm from changing.
613 * We always want mm->owner->mm == mm
617 * Delay read_unlock() till we have the task_lock()
618 * to ensure that c does not slip away underneath us
620 read_unlock(&tasklist_lock
);
630 #endif /* CONFIG_MM_OWNER */
633 * Turn us into a lazy TLB process if we
636 static void exit_mm(struct task_struct
* tsk
)
638 struct mm_struct
*mm
= tsk
->mm
;
639 struct core_state
*core_state
;
645 * Serialize with any possible pending coredump.
646 * We must hold mmap_sem around checking core_state
647 * and clearing tsk->mm. The core-inducing thread
648 * will increment ->nr_threads for each thread in the
649 * group with ->mm != NULL.
651 down_read(&mm
->mmap_sem
);
652 core_state
= mm
->core_state
;
654 struct core_thread self
;
655 up_read(&mm
->mmap_sem
);
658 self
.next
= xchg(&core_state
->dumper
.next
, &self
);
660 * Implies mb(), the result of xchg() must be visible
661 * to core_state->dumper.
663 if (atomic_dec_and_test(&core_state
->nr_threads
))
664 complete(&core_state
->startup
);
667 set_task_state(tsk
, TASK_UNINTERRUPTIBLE
);
668 if (!self
.task
) /* see coredump_finish() */
672 __set_task_state(tsk
, TASK_RUNNING
);
673 down_read(&mm
->mmap_sem
);
675 atomic_inc(&mm
->mm_count
);
676 BUG_ON(mm
!= tsk
->active_mm
);
677 /* more a memory barrier than a real lock */
680 up_read(&mm
->mmap_sem
);
681 enter_lazy_tlb(mm
, current
);
682 /* We don't want this task to be frozen prematurely */
683 clear_freeze_flag(tsk
);
684 if (tsk
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MIN
)
685 atomic_dec(&mm
->oom_disable_count
);
687 mm_update_next_owner(mm
);
692 * When we die, we re-parent all our children.
693 * Try to give them to another thread in our thread
694 * group, and if no such member exists, give it to
695 * the child reaper process (ie "init") in our pid
698 static struct task_struct
*find_new_reaper(struct task_struct
*father
)
699 __releases(&tasklist_lock
)
700 __acquires(&tasklist_lock
)
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_leader(struct task_struct
*father
, struct task_struct
*p
,
736 struct list_head
*dead
)
738 list_move_tail(&p
->sibling
, &p
->real_parent
->children
);
740 if (p
->exit_state
== EXIT_DEAD
)
743 * If this is a threaded reparent there is no need to
744 * notify anyone anything has happened.
746 if (same_thread_group(p
->real_parent
, father
))
749 /* We don't want people slaying init. */
750 p
->exit_signal
= SIGCHLD
;
752 /* If it has exited notify the new parent about this child's death. */
754 p
->exit_state
== EXIT_ZOMBIE
&& thread_group_empty(p
)) {
755 if (do_notify_parent(p
, p
->exit_signal
)) {
756 p
->exit_state
= EXIT_DEAD
;
757 list_move_tail(&p
->sibling
, dead
);
761 kill_orphaned_pgrp(p
, father
);
764 static void forget_original_parent(struct task_struct
*father
)
766 struct task_struct
*p
, *n
, *reaper
;
767 LIST_HEAD(dead_children
);
769 write_lock_irq(&tasklist_lock
);
771 * Note that exit_ptrace() and find_new_reaper() might
772 * drop tasklist_lock and reacquire it.
775 reaper
= find_new_reaper(father
);
777 list_for_each_entry_safe(p
, n
, &father
->children
, sibling
) {
778 struct task_struct
*t
= p
;
780 t
->real_parent
= reaper
;
781 if (t
->parent
== father
) {
783 t
->parent
= t
->real_parent
;
785 if (t
->pdeath_signal
)
786 group_send_sig_info(t
->pdeath_signal
,
788 } while_each_thread(p
, t
);
789 reparent_leader(father
, p
, &dead_children
);
791 write_unlock_irq(&tasklist_lock
);
793 BUG_ON(!list_empty(&father
->children
));
795 list_for_each_entry_safe(p
, n
, &dead_children
, sibling
) {
796 list_del_init(&p
->sibling
);
802 * Send signals to all our closest relatives so that they know
803 * to properly mourn us..
805 static void exit_notify(struct task_struct
*tsk
, int group_dead
)
810 * This does two things:
812 * A. Make init inherit all the child processes
813 * B. Check to see if any process groups have become orphaned
814 * as a result of our exiting, and if they have any stopped
815 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
817 forget_original_parent(tsk
);
818 exit_task_namespaces(tsk
);
820 write_lock_irq(&tasklist_lock
);
822 kill_orphaned_pgrp(tsk
->group_leader
, NULL
);
824 /* Let father know we died
826 * Thread signals are configurable, but you aren't going to use
827 * that to send signals to arbitrary processes.
828 * That stops right now.
830 * If the parent exec id doesn't match the exec id we saved
831 * when we started then we know the parent has changed security
834 * If our self_exec id doesn't match our parent_exec_id then
835 * we have changed execution domain as these two values started
836 * the same after a fork.
838 if (thread_group_leader(tsk
) && tsk
->exit_signal
!= SIGCHLD
&&
839 (tsk
->parent_exec_id
!= tsk
->real_parent
->self_exec_id
||
840 tsk
->self_exec_id
!= tsk
->parent_exec_id
))
841 tsk
->exit_signal
= SIGCHLD
;
843 if (unlikely(tsk
->ptrace
)) {
844 int sig
= thread_group_leader(tsk
) &&
845 thread_group_empty(tsk
) &&
846 !ptrace_reparented(tsk
) ?
847 tsk
->exit_signal
: SIGCHLD
;
848 autoreap
= do_notify_parent(tsk
, sig
);
849 } else if (thread_group_leader(tsk
)) {
850 autoreap
= thread_group_empty(tsk
) &&
851 do_notify_parent(tsk
, tsk
->exit_signal
);
856 tsk
->exit_state
= autoreap
? EXIT_DEAD
: EXIT_ZOMBIE
;
858 /* mt-exec, de_thread() is waiting for group leader */
859 if (unlikely(tsk
->signal
->notify_count
< 0))
860 wake_up_process(tsk
->signal
->group_exit_task
);
861 write_unlock_irq(&tasklist_lock
);
863 /* If the process is dead, release it - nobody will wait for it */
868 #ifdef CONFIG_DEBUG_STACK_USAGE
869 static void check_stack_usage(void)
871 static DEFINE_SPINLOCK(low_water_lock
);
872 static int lowest_to_date
= THREAD_SIZE
;
875 free
= stack_not_used(current
);
877 if (free
>= lowest_to_date
)
880 spin_lock(&low_water_lock
);
881 if (free
< lowest_to_date
) {
882 printk(KERN_WARNING
"%s used greatest stack depth: %lu bytes "
884 current
->comm
, free
);
885 lowest_to_date
= free
;
887 spin_unlock(&low_water_lock
);
890 static inline void check_stack_usage(void) {}
893 NORET_TYPE
void do_exit(long code
)
895 struct task_struct
*tsk
= current
;
898 profile_task_exit(tsk
);
900 WARN_ON(blk_needs_flush_plug(tsk
));
902 if (unlikely(in_interrupt()))
903 panic("Aiee, killing interrupt handler!");
904 if (unlikely(!tsk
->pid
))
905 panic("Attempted to kill the idle task!");
908 * If do_exit is called because this processes oopsed, it's possible
909 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
910 * continuing. Amongst other possible reasons, this is to prevent
911 * mm_release()->clear_child_tid() from writing to a user-controlled
916 ptrace_event(PTRACE_EVENT_EXIT
, code
);
918 validate_creds_for_do_exit(tsk
);
921 * We're taking recursive faults here in do_exit. Safest is to just
922 * leave this task alone and wait for reboot.
924 if (unlikely(tsk
->flags
& PF_EXITING
)) {
926 "Fixing recursive fault but reboot is needed!\n");
928 * We can do this unlocked here. The futex code uses
929 * this flag just to verify whether the pi state
930 * cleanup has been done or not. In the worst case it
931 * loops once more. We pretend that the cleanup was
932 * done as there is no way to return. Either the
933 * OWNER_DIED bit is set by now or we push the blocked
934 * task into the wait for ever nirwana as well.
936 tsk
->flags
|= PF_EXITPIDONE
;
937 set_current_state(TASK_UNINTERRUPTIBLE
);
943 exit_signals(tsk
); /* sets PF_EXITING */
945 * tsk->flags are checked in the futex code to protect against
946 * an exiting task cleaning up the robust pi futexes.
949 raw_spin_unlock_wait(&tsk
->pi_lock
);
951 if (unlikely(in_atomic()))
952 printk(KERN_INFO
"note: %s[%d] exited with preempt_count %d\n",
953 current
->comm
, task_pid_nr(current
),
956 acct_update_integrals(tsk
);
957 /* sync mm's RSS info before statistics gathering */
959 sync_mm_rss(tsk
, tsk
->mm
);
960 group_dead
= atomic_dec_and_test(&tsk
->signal
->live
);
962 hrtimer_cancel(&tsk
->signal
->real_timer
);
963 exit_itimers(tsk
->signal
);
965 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, tsk
->mm
);
967 acct_collect(code
, group_dead
);
970 if (unlikely(tsk
->audit_context
))
973 tsk
->exit_code
= code
;
974 taskstats_exit(tsk
, group_dead
);
980 trace_sched_process_exit(tsk
);
989 * Flush inherited counters to the parent - before the parent
990 * gets woken up by child-exit notifications.
992 * because of cgroup mode, must be called before cgroup_exit()
994 perf_event_exit_task(tsk
);
999 disassociate_ctty(1);
1001 module_put(task_thread_info(tsk
)->exec_domain
->module
);
1003 proc_exit_connector(tsk
);
1006 * FIXME: do that only when needed, using sched_exit tracepoint
1008 ptrace_put_breakpoints(tsk
);
1010 exit_notify(tsk
, group_dead
);
1013 mpol_put(tsk
->mempolicy
);
1014 tsk
->mempolicy
= NULL
;
1018 if (unlikely(current
->pi_state_cache
))
1019 kfree(current
->pi_state_cache
);
1022 * Make sure we are holding no locks:
1024 debug_check_no_locks_held(tsk
);
1026 * We can do this unlocked here. The futex code uses this flag
1027 * just to verify whether the pi state cleanup has been done
1028 * or not. In the worst case it loops once more.
1030 tsk
->flags
|= PF_EXITPIDONE
;
1032 if (tsk
->io_context
)
1033 exit_io_context(tsk
);
1035 if (tsk
->splice_pipe
)
1036 __free_pipe_info(tsk
->splice_pipe
);
1038 validate_creds_for_do_exit(tsk
);
1042 /* causes final put_task_struct in finish_task_switch(). */
1043 tsk
->state
= TASK_DEAD
;
1046 /* Avoid "noreturn function does return". */
1048 cpu_relax(); /* For when BUG is null */
1051 EXPORT_SYMBOL_GPL(do_exit
);
1053 NORET_TYPE
void complete_and_exit(struct completion
*comp
, long code
)
1061 EXPORT_SYMBOL(complete_and_exit
);
1063 SYSCALL_DEFINE1(exit
, int, error_code
)
1065 do_exit((error_code
&0xff)<<8);
1069 * Take down every thread in the group. This is called by fatal signals
1070 * as well as by sys_exit_group (below).
1073 do_group_exit(int exit_code
)
1075 struct signal_struct
*sig
= current
->signal
;
1077 BUG_ON(exit_code
& 0x80); /* core dumps don't get here */
1079 if (signal_group_exit(sig
))
1080 exit_code
= sig
->group_exit_code
;
1081 else if (!thread_group_empty(current
)) {
1082 struct sighand_struct
*const sighand
= current
->sighand
;
1083 spin_lock_irq(&sighand
->siglock
);
1084 if (signal_group_exit(sig
))
1085 /* Another thread got here before we took the lock. */
1086 exit_code
= sig
->group_exit_code
;
1088 sig
->group_exit_code
= exit_code
;
1089 sig
->flags
= SIGNAL_GROUP_EXIT
;
1090 zap_other_threads(current
);
1092 spin_unlock_irq(&sighand
->siglock
);
1100 * this kills every thread in the thread group. Note that any externally
1101 * wait4()-ing process will get the correct exit code - even if this
1102 * thread is not the thread group leader.
1104 SYSCALL_DEFINE1(exit_group
, int, error_code
)
1106 do_group_exit((error_code
& 0xff) << 8);
1112 enum pid_type wo_type
;
1116 struct siginfo __user
*wo_info
;
1117 int __user
*wo_stat
;
1118 struct rusage __user
*wo_rusage
;
1120 wait_queue_t child_wait
;
1125 struct pid
*task_pid_type(struct task_struct
*task
, enum pid_type type
)
1127 if (type
!= PIDTYPE_PID
)
1128 task
= task
->group_leader
;
1129 return task
->pids
[type
].pid
;
1132 static int eligible_pid(struct wait_opts
*wo
, struct task_struct
*p
)
1134 return wo
->wo_type
== PIDTYPE_MAX
||
1135 task_pid_type(p
, wo
->wo_type
) == wo
->wo_pid
;
1138 static int eligible_child(struct wait_opts
*wo
, struct task_struct
*p
)
1140 if (!eligible_pid(wo
, p
))
1142 /* Wait for all children (clone and not) if __WALL is set;
1143 * otherwise, wait for clone children *only* if __WCLONE is
1144 * set; otherwise, wait for non-clone children *only*. (Note:
1145 * A "clone" child here is one that reports to its parent
1146 * using a signal other than SIGCHLD.) */
1147 if (((p
->exit_signal
!= SIGCHLD
) ^ !!(wo
->wo_flags
& __WCLONE
))
1148 && !(wo
->wo_flags
& __WALL
))
1154 static int wait_noreap_copyout(struct wait_opts
*wo
, struct task_struct
*p
,
1155 pid_t pid
, uid_t uid
, int why
, int status
)
1157 struct siginfo __user
*infop
;
1158 int retval
= wo
->wo_rusage
1159 ? getrusage(p
, RUSAGE_BOTH
, wo
->wo_rusage
) : 0;
1162 infop
= wo
->wo_info
;
1165 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1167 retval
= put_user(0, &infop
->si_errno
);
1169 retval
= put_user((short)why
, &infop
->si_code
);
1171 retval
= put_user(pid
, &infop
->si_pid
);
1173 retval
= put_user(uid
, &infop
->si_uid
);
1175 retval
= put_user(status
, &infop
->si_status
);
1183 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1184 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1185 * the lock and this task is uninteresting. If we return nonzero, we have
1186 * released the lock and the system call should return.
1188 static int wait_task_zombie(struct wait_opts
*wo
, struct task_struct
*p
)
1190 unsigned long state
;
1191 int retval
, status
, traced
;
1192 pid_t pid
= task_pid_vnr(p
);
1193 uid_t uid
= __task_cred(p
)->uid
;
1194 struct siginfo __user
*infop
;
1196 if (!likely(wo
->wo_flags
& WEXITED
))
1199 if (unlikely(wo
->wo_flags
& WNOWAIT
)) {
1200 int exit_code
= p
->exit_code
;
1204 read_unlock(&tasklist_lock
);
1205 if ((exit_code
& 0x7f) == 0) {
1207 status
= exit_code
>> 8;
1209 why
= (exit_code
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1210 status
= exit_code
& 0x7f;
1212 return wait_noreap_copyout(wo
, p
, pid
, uid
, why
, status
);
1216 * Try to move the task's state to DEAD
1217 * only one thread is allowed to do this:
1219 state
= xchg(&p
->exit_state
, EXIT_DEAD
);
1220 if (state
!= EXIT_ZOMBIE
) {
1221 BUG_ON(state
!= EXIT_DEAD
);
1225 traced
= ptrace_reparented(p
);
1227 * It can be ptraced but not reparented, check
1228 * thread_group_leader() to filter out sub-threads.
1230 if (likely(!traced
) && thread_group_leader(p
)) {
1231 struct signal_struct
*psig
;
1232 struct signal_struct
*sig
;
1233 unsigned long maxrss
;
1234 cputime_t tgutime
, tgstime
;
1237 * The resource counters for the group leader are in its
1238 * own task_struct. Those for dead threads in the group
1239 * are in its signal_struct, as are those for the child
1240 * processes it has previously reaped. All these
1241 * accumulate in the parent's signal_struct c* fields.
1243 * We don't bother to take a lock here to protect these
1244 * p->signal fields, because they are only touched by
1245 * __exit_signal, which runs with tasklist_lock
1246 * write-locked anyway, and so is excluded here. We do
1247 * need to protect the access to parent->signal fields,
1248 * as other threads in the parent group can be right
1249 * here reaping other children at the same time.
1251 * We use thread_group_times() to get times for the thread
1252 * group, which consolidates times for all threads in the
1253 * group including the group leader.
1255 thread_group_times(p
, &tgutime
, &tgstime
);
1256 spin_lock_irq(&p
->real_parent
->sighand
->siglock
);
1257 psig
= p
->real_parent
->signal
;
1260 cputime_add(psig
->cutime
,
1261 cputime_add(tgutime
,
1264 cputime_add(psig
->cstime
,
1265 cputime_add(tgstime
,
1268 cputime_add(psig
->cgtime
,
1269 cputime_add(p
->gtime
,
1270 cputime_add(sig
->gtime
,
1273 p
->min_flt
+ sig
->min_flt
+ sig
->cmin_flt
;
1275 p
->maj_flt
+ sig
->maj_flt
+ sig
->cmaj_flt
;
1277 p
->nvcsw
+ sig
->nvcsw
+ sig
->cnvcsw
;
1279 p
->nivcsw
+ sig
->nivcsw
+ sig
->cnivcsw
;
1281 task_io_get_inblock(p
) +
1282 sig
->inblock
+ sig
->cinblock
;
1284 task_io_get_oublock(p
) +
1285 sig
->oublock
+ sig
->coublock
;
1286 maxrss
= max(sig
->maxrss
, sig
->cmaxrss
);
1287 if (psig
->cmaxrss
< maxrss
)
1288 psig
->cmaxrss
= maxrss
;
1289 task_io_accounting_add(&psig
->ioac
, &p
->ioac
);
1290 task_io_accounting_add(&psig
->ioac
, &sig
->ioac
);
1291 spin_unlock_irq(&p
->real_parent
->sighand
->siglock
);
1295 * Now we are sure this task is interesting, and no other
1296 * thread can reap it because we set its state to EXIT_DEAD.
1298 read_unlock(&tasklist_lock
);
1300 retval
= wo
->wo_rusage
1301 ? getrusage(p
, RUSAGE_BOTH
, wo
->wo_rusage
) : 0;
1302 status
= (p
->signal
->flags
& SIGNAL_GROUP_EXIT
)
1303 ? p
->signal
->group_exit_code
: p
->exit_code
;
1304 if (!retval
&& wo
->wo_stat
)
1305 retval
= put_user(status
, wo
->wo_stat
);
1307 infop
= wo
->wo_info
;
1308 if (!retval
&& infop
)
1309 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1310 if (!retval
&& infop
)
1311 retval
= put_user(0, &infop
->si_errno
);
1312 if (!retval
&& infop
) {
1315 if ((status
& 0x7f) == 0) {
1319 why
= (status
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1322 retval
= put_user((short)why
, &infop
->si_code
);
1324 retval
= put_user(status
, &infop
->si_status
);
1326 if (!retval
&& infop
)
1327 retval
= put_user(pid
, &infop
->si_pid
);
1328 if (!retval
&& infop
)
1329 retval
= put_user(uid
, &infop
->si_uid
);
1334 write_lock_irq(&tasklist_lock
);
1335 /* We dropped tasklist, ptracer could die and untrace */
1338 * If this is not a sub-thread, notify the parent.
1339 * If parent wants a zombie, don't release it now.
1341 if (thread_group_leader(p
) &&
1342 !do_notify_parent(p
, p
->exit_signal
)) {
1343 p
->exit_state
= EXIT_ZOMBIE
;
1346 write_unlock_irq(&tasklist_lock
);
1354 static int *task_stopped_code(struct task_struct
*p
, bool ptrace
)
1357 if (task_is_stopped_or_traced(p
) &&
1358 !(p
->jobctl
& JOBCTL_LISTENING
))
1359 return &p
->exit_code
;
1361 if (p
->signal
->flags
& SIGNAL_STOP_STOPPED
)
1362 return &p
->signal
->group_exit_code
;
1368 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1370 * @ptrace: is the wait for ptrace
1371 * @p: task to wait for
1373 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1376 * read_lock(&tasklist_lock), which is released if return value is
1377 * non-zero. Also, grabs and releases @p->sighand->siglock.
1380 * 0 if wait condition didn't exist and search for other wait conditions
1381 * should continue. Non-zero return, -errno on failure and @p's pid on
1382 * success, implies that tasklist_lock is released and wait condition
1383 * search should terminate.
1385 static int wait_task_stopped(struct wait_opts
*wo
,
1386 int ptrace
, struct task_struct
*p
)
1388 struct siginfo __user
*infop
;
1389 int retval
, exit_code
, *p_code
, why
;
1390 uid_t uid
= 0; /* unneeded, required by compiler */
1394 * Traditionally we see ptrace'd stopped tasks regardless of options.
1396 if (!ptrace
&& !(wo
->wo_flags
& WUNTRACED
))
1399 if (!task_stopped_code(p
, ptrace
))
1403 spin_lock_irq(&p
->sighand
->siglock
);
1405 p_code
= task_stopped_code(p
, ptrace
);
1406 if (unlikely(!p_code
))
1409 exit_code
= *p_code
;
1413 if (!unlikely(wo
->wo_flags
& WNOWAIT
))
1418 spin_unlock_irq(&p
->sighand
->siglock
);
1423 * Now we are pretty sure this task is interesting.
1424 * Make sure it doesn't get reaped out from under us while we
1425 * give up the lock and then examine it below. We don't want to
1426 * keep holding onto the tasklist_lock while we call getrusage and
1427 * possibly take page faults for user memory.
1430 pid
= task_pid_vnr(p
);
1431 why
= ptrace
? CLD_TRAPPED
: CLD_STOPPED
;
1432 read_unlock(&tasklist_lock
);
1434 if (unlikely(wo
->wo_flags
& WNOWAIT
))
1435 return wait_noreap_copyout(wo
, p
, pid
, uid
, why
, exit_code
);
1437 retval
= wo
->wo_rusage
1438 ? getrusage(p
, RUSAGE_BOTH
, wo
->wo_rusage
) : 0;
1439 if (!retval
&& wo
->wo_stat
)
1440 retval
= put_user((exit_code
<< 8) | 0x7f, wo
->wo_stat
);
1442 infop
= wo
->wo_info
;
1443 if (!retval
&& infop
)
1444 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1445 if (!retval
&& infop
)
1446 retval
= put_user(0, &infop
->si_errno
);
1447 if (!retval
&& infop
)
1448 retval
= put_user((short)why
, &infop
->si_code
);
1449 if (!retval
&& infop
)
1450 retval
= put_user(exit_code
, &infop
->si_status
);
1451 if (!retval
&& infop
)
1452 retval
= put_user(pid
, &infop
->si_pid
);
1453 if (!retval
&& infop
)
1454 retval
= put_user(uid
, &infop
->si_uid
);
1464 * Handle do_wait work for one task in a live, non-stopped state.
1465 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1466 * the lock and this task is uninteresting. If we return nonzero, we have
1467 * released the lock and the system call should return.
1469 static int wait_task_continued(struct wait_opts
*wo
, struct task_struct
*p
)
1475 if (!unlikely(wo
->wo_flags
& WCONTINUED
))
1478 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
))
1481 spin_lock_irq(&p
->sighand
->siglock
);
1482 /* Re-check with the lock held. */
1483 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
)) {
1484 spin_unlock_irq(&p
->sighand
->siglock
);
1487 if (!unlikely(wo
->wo_flags
& WNOWAIT
))
1488 p
->signal
->flags
&= ~SIGNAL_STOP_CONTINUED
;
1490 spin_unlock_irq(&p
->sighand
->siglock
);
1492 pid
= task_pid_vnr(p
);
1494 read_unlock(&tasklist_lock
);
1497 retval
= wo
->wo_rusage
1498 ? getrusage(p
, RUSAGE_BOTH
, wo
->wo_rusage
) : 0;
1500 if (!retval
&& wo
->wo_stat
)
1501 retval
= put_user(0xffff, wo
->wo_stat
);
1505 retval
= wait_noreap_copyout(wo
, p
, pid
, uid
,
1506 CLD_CONTINUED
, SIGCONT
);
1507 BUG_ON(retval
== 0);
1514 * Consider @p for a wait by @parent.
1516 * -ECHILD should be in ->notask_error before the first call.
1517 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1518 * Returns zero if the search for a child should continue;
1519 * then ->notask_error is 0 if @p is an eligible child,
1520 * or another error from security_task_wait(), or still -ECHILD.
1522 static int wait_consider_task(struct wait_opts
*wo
, int ptrace
,
1523 struct task_struct
*p
)
1525 int ret
= eligible_child(wo
, p
);
1529 ret
= security_task_wait(p
);
1530 if (unlikely(ret
< 0)) {
1532 * If we have not yet seen any eligible child,
1533 * then let this error code replace -ECHILD.
1534 * A permission error will give the user a clue
1535 * to look for security policy problems, rather
1536 * than for mysterious wait bugs.
1538 if (wo
->notask_error
)
1539 wo
->notask_error
= ret
;
1543 /* dead body doesn't have much to contribute */
1544 if (p
->exit_state
== EXIT_DEAD
)
1548 if (p
->exit_state
== EXIT_ZOMBIE
) {
1550 * A zombie ptracee is only visible to its ptracer.
1551 * Notification and reaping will be cascaded to the real
1552 * parent when the ptracer detaches.
1554 if (likely(!ptrace
) && unlikely(p
->ptrace
)) {
1555 /* it will become visible, clear notask_error */
1556 wo
->notask_error
= 0;
1560 /* we don't reap group leaders with subthreads */
1561 if (!delay_group_leader(p
))
1562 return wait_task_zombie(wo
, p
);
1565 * Allow access to stopped/continued state via zombie by
1566 * falling through. Clearing of notask_error is complex.
1570 * If WEXITED is set, notask_error should naturally be
1571 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1572 * so, if there are live subthreads, there are events to
1573 * wait for. If all subthreads are dead, it's still safe
1574 * to clear - this function will be called again in finite
1575 * amount time once all the subthreads are released and
1576 * will then return without clearing.
1580 * Stopped state is per-task and thus can't change once the
1581 * target task dies. Only continued and exited can happen.
1582 * Clear notask_error if WCONTINUED | WEXITED.
1584 if (likely(!ptrace
) || (wo
->wo_flags
& (WCONTINUED
| WEXITED
)))
1585 wo
->notask_error
= 0;
1588 * If @p is ptraced by a task in its real parent's group,
1589 * hide group stop/continued state when looking at @p as
1590 * the real parent; otherwise, a single stop can be
1591 * reported twice as group and ptrace stops.
1593 * If a ptracer wants to distinguish the two events for its
1594 * own children, it should create a separate process which
1595 * takes the role of real parent.
1597 if (likely(!ptrace
) && p
->ptrace
&& !ptrace_reparented(p
))
1601 * @p is alive and it's gonna stop, continue or exit, so
1602 * there always is something to wait for.
1604 wo
->notask_error
= 0;
1608 * Wait for stopped. Depending on @ptrace, different stopped state
1609 * is used and the two don't interact with each other.
1611 ret
= wait_task_stopped(wo
, ptrace
, p
);
1616 * Wait for continued. There's only one continued state and the
1617 * ptracer can consume it which can confuse the real parent. Don't
1618 * use WCONTINUED from ptracer. You don't need or want it.
1620 return wait_task_continued(wo
, p
);
1624 * Do the work of do_wait() for one thread in the group, @tsk.
1626 * -ECHILD should be in ->notask_error before the first call.
1627 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1628 * Returns zero if the search for a child should continue; then
1629 * ->notask_error is 0 if there were any eligible children,
1630 * or another error from security_task_wait(), or still -ECHILD.
1632 static int do_wait_thread(struct wait_opts
*wo
, struct task_struct
*tsk
)
1634 struct task_struct
*p
;
1636 list_for_each_entry(p
, &tsk
->children
, sibling
) {
1637 int ret
= wait_consider_task(wo
, 0, p
);
1645 static int ptrace_do_wait(struct wait_opts
*wo
, struct task_struct
*tsk
)
1647 struct task_struct
*p
;
1649 list_for_each_entry(p
, &tsk
->ptraced
, ptrace_entry
) {
1650 int ret
= wait_consider_task(wo
, 1, p
);
1658 static int child_wait_callback(wait_queue_t
*wait
, unsigned mode
,
1659 int sync
, void *key
)
1661 struct wait_opts
*wo
= container_of(wait
, struct wait_opts
,
1663 struct task_struct
*p
= key
;
1665 if (!eligible_pid(wo
, p
))
1668 if ((wo
->wo_flags
& __WNOTHREAD
) && wait
->private != p
->parent
)
1671 return default_wake_function(wait
, mode
, sync
, key
);
1674 void __wake_up_parent(struct task_struct
*p
, struct task_struct
*parent
)
1676 __wake_up_sync_key(&parent
->signal
->wait_chldexit
,
1677 TASK_INTERRUPTIBLE
, 1, p
);
1680 static long do_wait(struct wait_opts
*wo
)
1682 struct task_struct
*tsk
;
1685 trace_sched_process_wait(wo
->wo_pid
);
1687 init_waitqueue_func_entry(&wo
->child_wait
, child_wait_callback
);
1688 wo
->child_wait
.private = current
;
1689 add_wait_queue(¤t
->signal
->wait_chldexit
, &wo
->child_wait
);
1692 * If there is nothing that can match our critiera just get out.
1693 * We will clear ->notask_error to zero if we see any child that
1694 * might later match our criteria, even if we are not able to reap
1697 wo
->notask_error
= -ECHILD
;
1698 if ((wo
->wo_type
< PIDTYPE_MAX
) &&
1699 (!wo
->wo_pid
|| hlist_empty(&wo
->wo_pid
->tasks
[wo
->wo_type
])))
1702 set_current_state(TASK_INTERRUPTIBLE
);
1703 read_lock(&tasklist_lock
);
1706 retval
= do_wait_thread(wo
, tsk
);
1710 retval
= ptrace_do_wait(wo
, tsk
);
1714 if (wo
->wo_flags
& __WNOTHREAD
)
1716 } while_each_thread(current
, tsk
);
1717 read_unlock(&tasklist_lock
);
1720 retval
= wo
->notask_error
;
1721 if (!retval
&& !(wo
->wo_flags
& WNOHANG
)) {
1722 retval
= -ERESTARTSYS
;
1723 if (!signal_pending(current
)) {
1729 __set_current_state(TASK_RUNNING
);
1730 remove_wait_queue(¤t
->signal
->wait_chldexit
, &wo
->child_wait
);
1734 SYSCALL_DEFINE5(waitid
, int, which
, pid_t
, upid
, struct siginfo __user
*,
1735 infop
, int, options
, struct rusage __user
*, ru
)
1737 struct wait_opts wo
;
1738 struct pid
*pid
= NULL
;
1742 if (options
& ~(WNOHANG
|WNOWAIT
|WEXITED
|WSTOPPED
|WCONTINUED
))
1744 if (!(options
& (WEXITED
|WSTOPPED
|WCONTINUED
)))
1757 type
= PIDTYPE_PGID
;
1765 if (type
< PIDTYPE_MAX
)
1766 pid
= find_get_pid(upid
);
1770 wo
.wo_flags
= options
;
1780 * For a WNOHANG return, clear out all the fields
1781 * we would set so the user can easily tell the
1785 ret
= put_user(0, &infop
->si_signo
);
1787 ret
= put_user(0, &infop
->si_errno
);
1789 ret
= put_user(0, &infop
->si_code
);
1791 ret
= put_user(0, &infop
->si_pid
);
1793 ret
= put_user(0, &infop
->si_uid
);
1795 ret
= put_user(0, &infop
->si_status
);
1800 /* avoid REGPARM breakage on x86: */
1801 asmlinkage_protect(5, ret
, which
, upid
, infop
, options
, ru
);
1805 SYSCALL_DEFINE4(wait4
, pid_t
, upid
, int __user
*, stat_addr
,
1806 int, options
, struct rusage __user
*, ru
)
1808 struct wait_opts wo
;
1809 struct pid
*pid
= NULL
;
1813 if (options
& ~(WNOHANG
|WUNTRACED
|WCONTINUED
|
1814 __WNOTHREAD
|__WCLONE
|__WALL
))
1819 else if (upid
< 0) {
1820 type
= PIDTYPE_PGID
;
1821 pid
= find_get_pid(-upid
);
1822 } else if (upid
== 0) {
1823 type
= PIDTYPE_PGID
;
1824 pid
= get_task_pid(current
, PIDTYPE_PGID
);
1825 } else /* upid > 0 */ {
1827 pid
= find_get_pid(upid
);
1832 wo
.wo_flags
= options
| WEXITED
;
1834 wo
.wo_stat
= stat_addr
;
1839 /* avoid REGPARM breakage on x86: */
1840 asmlinkage_protect(4, ret
, upid
, stat_addr
, options
, ru
);
1844 #ifdef __ARCH_WANT_SYS_WAITPID
1847 * sys_waitpid() remains for compatibility. waitpid() should be
1848 * implemented by calling sys_wait4() from libc.a.
1850 SYSCALL_DEFINE3(waitpid
, pid_t
, pid
, int __user
*, stat_addr
, int, options
)
1852 return sys_wait4(pid
, stat_addr
, options
, NULL
);