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>
54 #include <asm/uaccess.h>
55 #include <asm/unistd.h>
56 #include <asm/pgtable.h>
57 #include <asm/mmu_context.h>
59 static void exit_mm(struct task_struct
* tsk
);
61 static void __unhash_process(struct task_struct
*p
)
64 detach_pid(p
, PIDTYPE_PID
);
65 if (thread_group_leader(p
)) {
66 detach_pid(p
, PIDTYPE_PGID
);
67 detach_pid(p
, PIDTYPE_SID
);
69 list_del_rcu(&p
->tasks
);
70 list_del_init(&p
->sibling
);
71 __get_cpu_var(process_counts
)--;
73 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
));
87 sighand
= rcu_dereference_check(tsk
->sighand
,
88 rcu_read_lock_held() ||
89 lockdep_tasklist_lock_is_held());
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 * This can only happen if the caller is de_thread().
98 * FIXME: this is the temporary hack, we should teach
99 * posix-cpu-timers to handle this case correctly.
101 if (unlikely(has_group_leader_pid(tsk
)))
102 posix_cpu_timers_exit_group(tsk
);
105 * If there is any task waiting for the group exit
108 if (sig
->group_exit_task
&& atomic_read(&sig
->count
) == sig
->notify_count
)
109 wake_up_process(sig
->group_exit_task
);
111 if (tsk
== sig
->curr_target
)
112 sig
->curr_target
= next_thread(tsk
);
114 * Accumulate here the counters for all threads but the
115 * group leader as they die, so they can be added into
116 * the process-wide totals when those are taken.
117 * The group leader stays around as a zombie as long
118 * as there are other threads. When it gets reaped,
119 * the exit.c code will add its counts into these totals.
120 * We won't ever get here for the group leader, since it
121 * will have been the last reference on the signal_struct.
123 sig
->utime
= cputime_add(sig
->utime
, tsk
->utime
);
124 sig
->stime
= cputime_add(sig
->stime
, tsk
->stime
);
125 sig
->gtime
= cputime_add(sig
->gtime
, tsk
->gtime
);
126 sig
->min_flt
+= tsk
->min_flt
;
127 sig
->maj_flt
+= tsk
->maj_flt
;
128 sig
->nvcsw
+= tsk
->nvcsw
;
129 sig
->nivcsw
+= tsk
->nivcsw
;
130 sig
->inblock
+= task_io_get_inblock(tsk
);
131 sig
->oublock
+= task_io_get_oublock(tsk
);
132 task_io_accounting_add(&sig
->ioac
, &tsk
->ioac
);
133 sig
->sum_sched_runtime
+= tsk
->se
.sum_exec_runtime
;
134 sig
= NULL
; /* Marker for below. */
137 __unhash_process(tsk
);
140 * Do this under ->siglock, we can race with another thread
141 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
143 flush_sigqueue(&tsk
->pending
);
147 spin_unlock(&sighand
->siglock
);
149 __cleanup_sighand(sighand
);
150 clear_tsk_thread_flag(tsk
,TIF_SIGPENDING
);
152 flush_sigqueue(&sig
->shared_pending
);
153 taskstats_tgid_free(sig
);
155 * Make sure ->signal can't go away under rq->lock,
156 * see account_group_exec_runtime().
158 task_rq_unlock_wait(tsk
);
159 __cleanup_signal(sig
);
163 static void delayed_put_task_struct(struct rcu_head
*rhp
)
165 struct task_struct
*tsk
= container_of(rhp
, struct task_struct
, rcu
);
167 #ifdef CONFIG_PERF_EVENTS
168 WARN_ON_ONCE(tsk
->perf_event_ctxp
);
170 trace_sched_process_free(tsk
);
171 put_task_struct(tsk
);
175 void release_task(struct task_struct
* p
)
177 struct task_struct
*leader
;
180 tracehook_prepare_release_task(p
);
181 /* don't need to get the RCU readlock here - the process is dead and
182 * can't be modifying its own credentials. But shut RCU-lockdep up */
184 atomic_dec(&__task_cred(p
)->user
->processes
);
189 write_lock_irq(&tasklist_lock
);
190 tracehook_finish_release_task(p
);
194 * If we are the last non-leader member of the thread
195 * group, and the leader is zombie, then notify the
196 * group leader's parent process. (if it wants notification.)
199 leader
= p
->group_leader
;
200 if (leader
!= p
&& thread_group_empty(leader
) && leader
->exit_state
== EXIT_ZOMBIE
) {
201 BUG_ON(task_detached(leader
));
202 do_notify_parent(leader
, leader
->exit_signal
);
204 * If we were the last child thread and the leader has
205 * exited already, and the leader's parent ignores SIGCHLD,
206 * then we are the one who should release the leader.
208 * do_notify_parent() will have marked it self-reaping in
211 zap_leader
= task_detached(leader
);
214 * This maintains the invariant that release_task()
215 * only runs on a task in EXIT_DEAD, just for sanity.
218 leader
->exit_state
= EXIT_DEAD
;
221 write_unlock_irq(&tasklist_lock
);
223 call_rcu(&p
->rcu
, delayed_put_task_struct
);
226 if (unlikely(zap_leader
))
231 * This checks not only the pgrp, but falls back on the pid if no
232 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
235 * The caller must hold rcu lock or the tasklist lock.
237 struct pid
*session_of_pgrp(struct pid
*pgrp
)
239 struct task_struct
*p
;
240 struct pid
*sid
= NULL
;
242 p
= pid_task(pgrp
, PIDTYPE_PGID
);
244 p
= pid_task(pgrp
, PIDTYPE_PID
);
246 sid
= task_session(p
);
252 * Determine if a process group is "orphaned", according to the POSIX
253 * definition in 2.2.2.52. Orphaned process groups are not to be affected
254 * by terminal-generated stop signals. Newly orphaned process groups are
255 * to receive a SIGHUP and a SIGCONT.
257 * "I ask you, have you ever known what it is to be an orphan?"
259 static int will_become_orphaned_pgrp(struct pid
*pgrp
, struct task_struct
*ignored_task
)
261 struct task_struct
*p
;
263 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
264 if ((p
== ignored_task
) ||
265 (p
->exit_state
&& thread_group_empty(p
)) ||
266 is_global_init(p
->real_parent
))
269 if (task_pgrp(p
->real_parent
) != pgrp
&&
270 task_session(p
->real_parent
) == task_session(p
))
272 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
277 int is_current_pgrp_orphaned(void)
281 read_lock(&tasklist_lock
);
282 retval
= will_become_orphaned_pgrp(task_pgrp(current
), NULL
);
283 read_unlock(&tasklist_lock
);
288 static int has_stopped_jobs(struct pid
*pgrp
)
291 struct task_struct
*p
;
293 do_each_pid_task(pgrp
, PIDTYPE_PGID
, p
) {
294 if (!task_is_stopped(p
))
298 } while_each_pid_task(pgrp
, PIDTYPE_PGID
, p
);
303 * Check to see if any process groups have become orphaned as
304 * a result of our exiting, and if they have any stopped jobs,
305 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
308 kill_orphaned_pgrp(struct task_struct
*tsk
, struct task_struct
*parent
)
310 struct pid
*pgrp
= task_pgrp(tsk
);
311 struct task_struct
*ignored_task
= tsk
;
314 /* exit: our father is in a different pgrp than
315 * we are and we were the only connection outside.
317 parent
= tsk
->real_parent
;
319 /* reparent: our child is in a different pgrp than
320 * we are, and it was the only connection outside.
324 if (task_pgrp(parent
) != pgrp
&&
325 task_session(parent
) == task_session(tsk
) &&
326 will_become_orphaned_pgrp(pgrp
, ignored_task
) &&
327 has_stopped_jobs(pgrp
)) {
328 __kill_pgrp_info(SIGHUP
, SEND_SIG_PRIV
, pgrp
);
329 __kill_pgrp_info(SIGCONT
, SEND_SIG_PRIV
, pgrp
);
334 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
336 * If a kernel thread is launched as a result of a system call, or if
337 * it ever exits, it should generally reparent itself to kthreadd so it
338 * isn't in the way of other processes and is correctly cleaned up on exit.
340 * The various task state such as scheduling policy and priority may have
341 * been inherited from a user process, so we reset them to sane values here.
343 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
345 static void reparent_to_kthreadd(void)
347 write_lock_irq(&tasklist_lock
);
349 ptrace_unlink(current
);
350 /* Reparent to init */
351 current
->real_parent
= current
->parent
= kthreadd_task
;
352 list_move_tail(¤t
->sibling
, ¤t
->real_parent
->children
);
354 /* Set the exit signal to SIGCHLD so we signal init on exit */
355 current
->exit_signal
= SIGCHLD
;
357 if (task_nice(current
) < 0)
358 set_user_nice(current
, 0);
362 memcpy(current
->signal
->rlim
, init_task
.signal
->rlim
,
363 sizeof(current
->signal
->rlim
));
365 atomic_inc(&init_cred
.usage
);
366 commit_creds(&init_cred
);
367 write_unlock_irq(&tasklist_lock
);
370 void __set_special_pids(struct pid
*pid
)
372 struct task_struct
*curr
= current
->group_leader
;
374 if (task_session(curr
) != pid
)
375 change_pid(curr
, PIDTYPE_SID
, pid
);
377 if (task_pgrp(curr
) != pid
)
378 change_pid(curr
, PIDTYPE_PGID
, pid
);
381 static void set_special_pids(struct pid
*pid
)
383 write_lock_irq(&tasklist_lock
);
384 __set_special_pids(pid
);
385 write_unlock_irq(&tasklist_lock
);
389 * Let kernel threads use this to say that they allow a certain signal.
390 * Must not be used if kthread was cloned with CLONE_SIGHAND.
392 int allow_signal(int sig
)
394 if (!valid_signal(sig
) || sig
< 1)
397 spin_lock_irq(¤t
->sighand
->siglock
);
398 /* This is only needed for daemonize()'ed kthreads */
399 sigdelset(¤t
->blocked
, sig
);
401 * Kernel threads handle their own signals. Let the signal code
402 * know it'll be handled, so that they don't get converted to
403 * SIGKILL or just silently dropped.
405 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= (void __user
*)2;
407 spin_unlock_irq(¤t
->sighand
->siglock
);
411 EXPORT_SYMBOL(allow_signal
);
413 int disallow_signal(int sig
)
415 if (!valid_signal(sig
) || sig
< 1)
418 spin_lock_irq(¤t
->sighand
->siglock
);
419 current
->sighand
->action
[(sig
)-1].sa
.sa_handler
= SIG_IGN
;
421 spin_unlock_irq(¤t
->sighand
->siglock
);
425 EXPORT_SYMBOL(disallow_signal
);
428 * Put all the gunge required to become a kernel thread without
429 * attached user resources in one place where it belongs.
432 void daemonize(const char *name
, ...)
437 va_start(args
, name
);
438 vsnprintf(current
->comm
, sizeof(current
->comm
), name
, args
);
442 * If we were started as result of loading a module, close all of the
443 * user space pages. We don't need them, and if we didn't close them
444 * they would be locked into memory.
448 * We don't want to have TIF_FREEZE set if the system-wide hibernation
449 * or suspend transition begins right now.
451 current
->flags
|= (PF_NOFREEZE
| PF_KTHREAD
);
453 if (current
->nsproxy
!= &init_nsproxy
) {
454 get_nsproxy(&init_nsproxy
);
455 switch_task_namespaces(current
, &init_nsproxy
);
457 set_special_pids(&init_struct_pid
);
458 proc_clear_tty(current
);
460 /* Block and flush all signals */
461 sigfillset(&blocked
);
462 sigprocmask(SIG_BLOCK
, &blocked
, NULL
);
463 flush_signals(current
);
465 /* Become as one with the init task */
467 daemonize_fs_struct();
469 current
->files
= init_task
.files
;
470 atomic_inc(¤t
->files
->count
);
472 reparent_to_kthreadd();
475 EXPORT_SYMBOL(daemonize
);
477 static void close_files(struct files_struct
* files
)
485 * It is safe to dereference the fd table without RCU or
486 * ->file_lock because this is the last reference to the
487 * files structure. But use RCU to shut RCU-lockdep up.
490 fdt
= files_fdtable(files
);
495 if (i
>= fdt
->max_fds
)
497 set
= fdt
->open_fds
->fds_bits
[j
++];
500 struct file
* file
= xchg(&fdt
->fd
[i
], NULL
);
502 filp_close(file
, files
);
512 struct files_struct
*get_files_struct(struct task_struct
*task
)
514 struct files_struct
*files
;
519 atomic_inc(&files
->count
);
525 void put_files_struct(struct files_struct
*files
)
529 if (atomic_dec_and_test(&files
->count
)) {
532 * Free the fd and fdset arrays if we expanded them.
533 * If the fdtable was embedded, pass files for freeing
534 * at the end of the RCU grace period. Otherwise,
535 * you can free files immediately.
538 fdt
= files_fdtable(files
);
539 if (fdt
!= &files
->fdtab
)
540 kmem_cache_free(files_cachep
, files
);
546 void reset_files_struct(struct files_struct
*files
)
548 struct task_struct
*tsk
= current
;
549 struct files_struct
*old
;
555 put_files_struct(old
);
558 void exit_files(struct task_struct
*tsk
)
560 struct files_struct
* files
= tsk
->files
;
566 put_files_struct(files
);
570 #ifdef CONFIG_MM_OWNER
572 * Task p is exiting and it owned mm, lets find a new owner for it
575 mm_need_new_owner(struct mm_struct
*mm
, struct task_struct
*p
)
578 * If there are other users of the mm and the owner (us) is exiting
579 * we need to find a new owner to take on the responsibility.
581 if (atomic_read(&mm
->mm_users
) <= 1)
588 void mm_update_next_owner(struct mm_struct
*mm
)
590 struct task_struct
*c
, *g
, *p
= current
;
593 if (!mm_need_new_owner(mm
, p
))
596 read_lock(&tasklist_lock
);
598 * Search in the children
600 list_for_each_entry(c
, &p
->children
, sibling
) {
602 goto assign_new_owner
;
606 * Search in the siblings
608 list_for_each_entry(c
, &p
->real_parent
->children
, sibling
) {
610 goto assign_new_owner
;
614 * Search through everything else. We should not get
617 do_each_thread(g
, c
) {
619 goto assign_new_owner
;
620 } while_each_thread(g
, c
);
622 read_unlock(&tasklist_lock
);
624 * We found no owner yet mm_users > 1: this implies that we are
625 * most likely racing with swapoff (try_to_unuse()) or /proc or
626 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
635 * The task_lock protects c->mm from changing.
636 * We always want mm->owner->mm == mm
640 * Delay read_unlock() till we have the task_lock()
641 * to ensure that c does not slip away underneath us
643 read_unlock(&tasklist_lock
);
653 #endif /* CONFIG_MM_OWNER */
656 * Turn us into a lazy TLB process if we
659 static void exit_mm(struct task_struct
* tsk
)
661 struct mm_struct
*mm
= tsk
->mm
;
662 struct core_state
*core_state
;
668 * Serialize with any possible pending coredump.
669 * We must hold mmap_sem around checking core_state
670 * and clearing tsk->mm. The core-inducing thread
671 * will increment ->nr_threads for each thread in the
672 * group with ->mm != NULL.
674 down_read(&mm
->mmap_sem
);
675 core_state
= mm
->core_state
;
677 struct core_thread self
;
678 up_read(&mm
->mmap_sem
);
681 self
.next
= xchg(&core_state
->dumper
.next
, &self
);
683 * Implies mb(), the result of xchg() must be visible
684 * to core_state->dumper.
686 if (atomic_dec_and_test(&core_state
->nr_threads
))
687 complete(&core_state
->startup
);
690 set_task_state(tsk
, TASK_UNINTERRUPTIBLE
);
691 if (!self
.task
) /* see coredump_finish() */
695 __set_task_state(tsk
, TASK_RUNNING
);
696 down_read(&mm
->mmap_sem
);
698 atomic_inc(&mm
->mm_count
);
699 BUG_ON(mm
!= tsk
->active_mm
);
700 /* more a memory barrier than a real lock */
703 up_read(&mm
->mmap_sem
);
704 enter_lazy_tlb(mm
, current
);
705 /* We don't want this task to be frozen prematurely */
706 clear_freeze_flag(tsk
);
708 mm_update_next_owner(mm
);
713 * When we die, we re-parent all our children.
714 * Try to give them to another thread in our thread
715 * group, and if no such member exists, give it to
716 * the child reaper process (ie "init") in our pid
719 static struct task_struct
*find_new_reaper(struct task_struct
*father
)
721 struct pid_namespace
*pid_ns
= task_active_pid_ns(father
);
722 struct task_struct
*thread
;
725 while_each_thread(father
, thread
) {
726 if (thread
->flags
& PF_EXITING
)
728 if (unlikely(pid_ns
->child_reaper
== father
))
729 pid_ns
->child_reaper
= thread
;
733 if (unlikely(pid_ns
->child_reaper
== father
)) {
734 write_unlock_irq(&tasklist_lock
);
735 if (unlikely(pid_ns
== &init_pid_ns
))
736 panic("Attempted to kill init!");
738 zap_pid_ns_processes(pid_ns
);
739 write_lock_irq(&tasklist_lock
);
741 * We can not clear ->child_reaper or leave it alone.
742 * There may by stealth EXIT_DEAD tasks on ->children,
743 * forget_original_parent() must move them somewhere.
745 pid_ns
->child_reaper
= init_pid_ns
.child_reaper
;
748 return pid_ns
->child_reaper
;
752 * Any that need to be release_task'd are put on the @dead list.
754 static void reparent_leader(struct task_struct
*father
, struct task_struct
*p
,
755 struct list_head
*dead
)
757 list_move_tail(&p
->sibling
, &p
->real_parent
->children
);
759 if (task_detached(p
))
762 * If this is a threaded reparent there is no need to
763 * notify anyone anything has happened.
765 if (same_thread_group(p
->real_parent
, father
))
768 /* We don't want people slaying init. */
769 p
->exit_signal
= SIGCHLD
;
771 /* If it has exited notify the new parent about this child's death. */
772 if (!task_ptrace(p
) &&
773 p
->exit_state
== EXIT_ZOMBIE
&& thread_group_empty(p
)) {
774 do_notify_parent(p
, p
->exit_signal
);
775 if (task_detached(p
)) {
776 p
->exit_state
= EXIT_DEAD
;
777 list_move_tail(&p
->sibling
, dead
);
781 kill_orphaned_pgrp(p
, father
);
784 static void forget_original_parent(struct task_struct
*father
)
786 struct task_struct
*p
, *n
, *reaper
;
787 LIST_HEAD(dead_children
);
791 write_lock_irq(&tasklist_lock
);
792 reaper
= find_new_reaper(father
);
794 list_for_each_entry_safe(p
, n
, &father
->children
, sibling
) {
795 struct task_struct
*t
= p
;
797 t
->real_parent
= reaper
;
798 if (t
->parent
== father
) {
799 BUG_ON(task_ptrace(t
));
800 t
->parent
= t
->real_parent
;
802 if (t
->pdeath_signal
)
803 group_send_sig_info(t
->pdeath_signal
,
805 } while_each_thread(p
, t
);
806 reparent_leader(father
, p
, &dead_children
);
808 write_unlock_irq(&tasklist_lock
);
810 BUG_ON(!list_empty(&father
->children
));
812 list_for_each_entry_safe(p
, n
, &dead_children
, sibling
) {
813 list_del_init(&p
->sibling
);
819 * Send signals to all our closest relatives so that they know
820 * to properly mourn us..
822 static void exit_notify(struct task_struct
*tsk
, int group_dead
)
828 * This does two things:
830 * A. Make init inherit all the child processes
831 * B. Check to see if any process groups have become orphaned
832 * as a result of our exiting, and if they have any stopped
833 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
835 forget_original_parent(tsk
);
836 exit_task_namespaces(tsk
);
838 write_lock_irq(&tasklist_lock
);
840 kill_orphaned_pgrp(tsk
->group_leader
, NULL
);
842 /* Let father know we died
844 * Thread signals are configurable, but you aren't going to use
845 * that to send signals to arbitary processes.
846 * That stops right now.
848 * If the parent exec id doesn't match the exec id we saved
849 * when we started then we know the parent has changed security
852 * If our self_exec id doesn't match our parent_exec_id then
853 * we have changed execution domain as these two values started
854 * the same after a fork.
856 if (tsk
->exit_signal
!= SIGCHLD
&& !task_detached(tsk
) &&
857 (tsk
->parent_exec_id
!= tsk
->real_parent
->self_exec_id
||
858 tsk
->self_exec_id
!= tsk
->parent_exec_id
))
859 tsk
->exit_signal
= SIGCHLD
;
861 signal
= tracehook_notify_death(tsk
, &cookie
, group_dead
);
863 signal
= do_notify_parent(tsk
, signal
);
865 tsk
->exit_state
= signal
== DEATH_REAP
? EXIT_DEAD
: EXIT_ZOMBIE
;
867 /* mt-exec, de_thread() is waiting for us */
868 if (thread_group_leader(tsk
) &&
869 tsk
->signal
->group_exit_task
&&
870 tsk
->signal
->notify_count
< 0)
871 wake_up_process(tsk
->signal
->group_exit_task
);
873 write_unlock_irq(&tasklist_lock
);
875 tracehook_report_death(tsk
, signal
, cookie
, group_dead
);
877 /* If the process is dead, release it - nobody will wait for it */
878 if (signal
== DEATH_REAP
)
882 #ifdef CONFIG_DEBUG_STACK_USAGE
883 static void check_stack_usage(void)
885 static DEFINE_SPINLOCK(low_water_lock
);
886 static int lowest_to_date
= THREAD_SIZE
;
889 free
= stack_not_used(current
);
891 if (free
>= lowest_to_date
)
894 spin_lock(&low_water_lock
);
895 if (free
< lowest_to_date
) {
896 printk(KERN_WARNING
"%s used greatest stack depth: %lu bytes "
898 current
->comm
, free
);
899 lowest_to_date
= free
;
901 spin_unlock(&low_water_lock
);
904 static inline void check_stack_usage(void) {}
907 NORET_TYPE
void do_exit(long code
)
909 struct task_struct
*tsk
= current
;
912 profile_task_exit(tsk
);
914 WARN_ON(atomic_read(&tsk
->fs_excl
));
916 if (unlikely(in_interrupt()))
917 panic("Aiee, killing interrupt handler!");
918 if (unlikely(!tsk
->pid
))
919 panic("Attempted to kill the idle task!");
922 * If do_exit is called because this processes oopsed, it's possible
923 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
924 * continuing. Amongst other possible reasons, this is to prevent
925 * mm_release()->clear_child_tid() from writing to a user-controlled
930 tracehook_report_exit(&code
);
932 validate_creds_for_do_exit(tsk
);
935 * We're taking recursive faults here in do_exit. Safest is to just
936 * leave this task alone and wait for reboot.
938 if (unlikely(tsk
->flags
& PF_EXITING
)) {
940 "Fixing recursive fault but reboot is needed!\n");
942 * We can do this unlocked here. The futex code uses
943 * this flag just to verify whether the pi state
944 * cleanup has been done or not. In the worst case it
945 * loops once more. We pretend that the cleanup was
946 * done as there is no way to return. Either the
947 * OWNER_DIED bit is set by now or we push the blocked
948 * task into the wait for ever nirwana as well.
950 tsk
->flags
|= PF_EXITPIDONE
;
951 set_current_state(TASK_UNINTERRUPTIBLE
);
957 exit_signals(tsk
); /* sets PF_EXITING */
959 * tsk->flags are checked in the futex code to protect against
960 * an exiting task cleaning up the robust pi futexes.
963 raw_spin_unlock_wait(&tsk
->pi_lock
);
965 if (unlikely(in_atomic()))
966 printk(KERN_INFO
"note: %s[%d] exited with preempt_count %d\n",
967 current
->comm
, task_pid_nr(current
),
970 acct_update_integrals(tsk
);
971 /* sync mm's RSS info before statistics gathering */
973 sync_mm_rss(tsk
, tsk
->mm
);
974 group_dead
= atomic_dec_and_test(&tsk
->signal
->live
);
976 hrtimer_cancel(&tsk
->signal
->real_timer
);
977 exit_itimers(tsk
->signal
);
979 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, tsk
->mm
);
981 acct_collect(code
, group_dead
);
984 if (unlikely(tsk
->audit_context
))
987 tsk
->exit_code
= code
;
988 taskstats_exit(tsk
, group_dead
);
994 trace_sched_process_exit(tsk
);
1001 cgroup_exit(tsk
, 1);
1004 disassociate_ctty(1);
1006 module_put(task_thread_info(tsk
)->exec_domain
->module
);
1008 proc_exit_connector(tsk
);
1011 * FIXME: do that only when needed, using sched_exit tracepoint
1013 flush_ptrace_hw_breakpoint(tsk
);
1015 * Flush inherited counters to the parent - before the parent
1016 * gets woken up by child-exit notifications.
1018 perf_event_exit_task(tsk
);
1020 exit_notify(tsk
, group_dead
);
1022 mpol_put(tsk
->mempolicy
);
1023 tsk
->mempolicy
= NULL
;
1026 if (unlikely(current
->pi_state_cache
))
1027 kfree(current
->pi_state_cache
);
1030 * Make sure we are holding no locks:
1032 debug_check_no_locks_held(tsk
);
1034 * We can do this unlocked here. The futex code uses this flag
1035 * just to verify whether the pi state cleanup has been done
1036 * or not. In the worst case it loops once more.
1038 tsk
->flags
|= PF_EXITPIDONE
;
1040 if (tsk
->io_context
)
1041 exit_io_context(tsk
);
1043 if (tsk
->splice_pipe
)
1044 __free_pipe_info(tsk
->splice_pipe
);
1046 validate_creds_for_do_exit(tsk
);
1050 /* causes final put_task_struct in finish_task_switch(). */
1051 tsk
->state
= TASK_DEAD
;
1054 /* Avoid "noreturn function does return". */
1056 cpu_relax(); /* For when BUG is null */
1059 EXPORT_SYMBOL_GPL(do_exit
);
1061 NORET_TYPE
void complete_and_exit(struct completion
*comp
, long code
)
1069 EXPORT_SYMBOL(complete_and_exit
);
1071 SYSCALL_DEFINE1(exit
, int, error_code
)
1073 do_exit((error_code
&0xff)<<8);
1077 * Take down every thread in the group. This is called by fatal signals
1078 * as well as by sys_exit_group (below).
1081 do_group_exit(int exit_code
)
1083 struct signal_struct
*sig
= current
->signal
;
1085 BUG_ON(exit_code
& 0x80); /* core dumps don't get here */
1087 if (signal_group_exit(sig
))
1088 exit_code
= sig
->group_exit_code
;
1089 else if (!thread_group_empty(current
)) {
1090 struct sighand_struct
*const sighand
= current
->sighand
;
1091 spin_lock_irq(&sighand
->siglock
);
1092 if (signal_group_exit(sig
))
1093 /* Another thread got here before we took the lock. */
1094 exit_code
= sig
->group_exit_code
;
1096 sig
->group_exit_code
= exit_code
;
1097 sig
->flags
= SIGNAL_GROUP_EXIT
;
1098 zap_other_threads(current
);
1100 spin_unlock_irq(&sighand
->siglock
);
1108 * this kills every thread in the thread group. Note that any externally
1109 * wait4()-ing process will get the correct exit code - even if this
1110 * thread is not the thread group leader.
1112 SYSCALL_DEFINE1(exit_group
, int, error_code
)
1114 do_group_exit((error_code
& 0xff) << 8);
1120 enum pid_type wo_type
;
1124 struct siginfo __user
*wo_info
;
1125 int __user
*wo_stat
;
1126 struct rusage __user
*wo_rusage
;
1128 wait_queue_t child_wait
;
1133 struct pid
*task_pid_type(struct task_struct
*task
, enum pid_type type
)
1135 if (type
!= PIDTYPE_PID
)
1136 task
= task
->group_leader
;
1137 return task
->pids
[type
].pid
;
1140 static int eligible_pid(struct wait_opts
*wo
, struct task_struct
*p
)
1142 return wo
->wo_type
== PIDTYPE_MAX
||
1143 task_pid_type(p
, wo
->wo_type
) == wo
->wo_pid
;
1146 static int eligible_child(struct wait_opts
*wo
, struct task_struct
*p
)
1148 if (!eligible_pid(wo
, p
))
1150 /* Wait for all children (clone and not) if __WALL is set;
1151 * otherwise, wait for clone children *only* if __WCLONE is
1152 * set; otherwise, wait for non-clone children *only*. (Note:
1153 * A "clone" child here is one that reports to its parent
1154 * using a signal other than SIGCHLD.) */
1155 if (((p
->exit_signal
!= SIGCHLD
) ^ !!(wo
->wo_flags
& __WCLONE
))
1156 && !(wo
->wo_flags
& __WALL
))
1162 static int wait_noreap_copyout(struct wait_opts
*wo
, struct task_struct
*p
,
1163 pid_t pid
, uid_t uid
, int why
, int status
)
1165 struct siginfo __user
*infop
;
1166 int retval
= wo
->wo_rusage
1167 ? getrusage(p
, RUSAGE_BOTH
, wo
->wo_rusage
) : 0;
1170 infop
= wo
->wo_info
;
1173 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1175 retval
= put_user(0, &infop
->si_errno
);
1177 retval
= put_user((short)why
, &infop
->si_code
);
1179 retval
= put_user(pid
, &infop
->si_pid
);
1181 retval
= put_user(uid
, &infop
->si_uid
);
1183 retval
= put_user(status
, &infop
->si_status
);
1191 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1192 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1193 * the lock and this task is uninteresting. If we return nonzero, we have
1194 * released the lock and the system call should return.
1196 static int wait_task_zombie(struct wait_opts
*wo
, struct task_struct
*p
)
1198 unsigned long state
;
1199 int retval
, status
, traced
;
1200 pid_t pid
= task_pid_vnr(p
);
1201 uid_t uid
= __task_cred(p
)->uid
;
1202 struct siginfo __user
*infop
;
1204 if (!likely(wo
->wo_flags
& WEXITED
))
1207 if (unlikely(wo
->wo_flags
& WNOWAIT
)) {
1208 int exit_code
= p
->exit_code
;
1212 read_unlock(&tasklist_lock
);
1213 if ((exit_code
& 0x7f) == 0) {
1215 status
= exit_code
>> 8;
1217 why
= (exit_code
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1218 status
= exit_code
& 0x7f;
1220 return wait_noreap_copyout(wo
, p
, pid
, uid
, why
, status
);
1224 * Try to move the task's state to DEAD
1225 * only one thread is allowed to do this:
1227 state
= xchg(&p
->exit_state
, EXIT_DEAD
);
1228 if (state
!= EXIT_ZOMBIE
) {
1229 BUG_ON(state
!= EXIT_DEAD
);
1233 traced
= ptrace_reparented(p
);
1235 * It can be ptraced but not reparented, check
1236 * !task_detached() to filter out sub-threads.
1238 if (likely(!traced
) && likely(!task_detached(p
))) {
1239 struct signal_struct
*psig
;
1240 struct signal_struct
*sig
;
1241 unsigned long maxrss
;
1242 cputime_t tgutime
, tgstime
;
1245 * The resource counters for the group leader are in its
1246 * own task_struct. Those for dead threads in the group
1247 * are in its signal_struct, as are those for the child
1248 * processes it has previously reaped. All these
1249 * accumulate in the parent's signal_struct c* fields.
1251 * We don't bother to take a lock here to protect these
1252 * p->signal fields, because they are only touched by
1253 * __exit_signal, which runs with tasklist_lock
1254 * write-locked anyway, and so is excluded here. We do
1255 * need to protect the access to parent->signal fields,
1256 * as other threads in the parent group can be right
1257 * here reaping other children at the same time.
1259 * We use thread_group_times() to get times for the thread
1260 * group, which consolidates times for all threads in the
1261 * group including the group leader.
1263 thread_group_times(p
, &tgutime
, &tgstime
);
1264 spin_lock_irq(&p
->real_parent
->sighand
->siglock
);
1265 psig
= p
->real_parent
->signal
;
1268 cputime_add(psig
->cutime
,
1269 cputime_add(tgutime
,
1272 cputime_add(psig
->cstime
,
1273 cputime_add(tgstime
,
1276 cputime_add(psig
->cgtime
,
1277 cputime_add(p
->gtime
,
1278 cputime_add(sig
->gtime
,
1281 p
->min_flt
+ sig
->min_flt
+ sig
->cmin_flt
;
1283 p
->maj_flt
+ sig
->maj_flt
+ sig
->cmaj_flt
;
1285 p
->nvcsw
+ sig
->nvcsw
+ sig
->cnvcsw
;
1287 p
->nivcsw
+ sig
->nivcsw
+ sig
->cnivcsw
;
1289 task_io_get_inblock(p
) +
1290 sig
->inblock
+ sig
->cinblock
;
1292 task_io_get_oublock(p
) +
1293 sig
->oublock
+ sig
->coublock
;
1294 maxrss
= max(sig
->maxrss
, sig
->cmaxrss
);
1295 if (psig
->cmaxrss
< maxrss
)
1296 psig
->cmaxrss
= maxrss
;
1297 task_io_accounting_add(&psig
->ioac
, &p
->ioac
);
1298 task_io_accounting_add(&psig
->ioac
, &sig
->ioac
);
1299 spin_unlock_irq(&p
->real_parent
->sighand
->siglock
);
1303 * Now we are sure this task is interesting, and no other
1304 * thread can reap it because we set its state to EXIT_DEAD.
1306 read_unlock(&tasklist_lock
);
1308 retval
= wo
->wo_rusage
1309 ? getrusage(p
, RUSAGE_BOTH
, wo
->wo_rusage
) : 0;
1310 status
= (p
->signal
->flags
& SIGNAL_GROUP_EXIT
)
1311 ? p
->signal
->group_exit_code
: p
->exit_code
;
1312 if (!retval
&& wo
->wo_stat
)
1313 retval
= put_user(status
, wo
->wo_stat
);
1315 infop
= wo
->wo_info
;
1316 if (!retval
&& infop
)
1317 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1318 if (!retval
&& infop
)
1319 retval
= put_user(0, &infop
->si_errno
);
1320 if (!retval
&& infop
) {
1323 if ((status
& 0x7f) == 0) {
1327 why
= (status
& 0x80) ? CLD_DUMPED
: CLD_KILLED
;
1330 retval
= put_user((short)why
, &infop
->si_code
);
1332 retval
= put_user(status
, &infop
->si_status
);
1334 if (!retval
&& infop
)
1335 retval
= put_user(pid
, &infop
->si_pid
);
1336 if (!retval
&& infop
)
1337 retval
= put_user(uid
, &infop
->si_uid
);
1342 write_lock_irq(&tasklist_lock
);
1343 /* We dropped tasklist, ptracer could die and untrace */
1346 * If this is not a detached task, notify the parent.
1347 * If it's still not detached after that, don't release
1350 if (!task_detached(p
)) {
1351 do_notify_parent(p
, p
->exit_signal
);
1352 if (!task_detached(p
)) {
1353 p
->exit_state
= EXIT_ZOMBIE
;
1357 write_unlock_irq(&tasklist_lock
);
1365 static int *task_stopped_code(struct task_struct
*p
, bool ptrace
)
1368 if (task_is_stopped_or_traced(p
))
1369 return &p
->exit_code
;
1371 if (p
->signal
->flags
& SIGNAL_STOP_STOPPED
)
1372 return &p
->signal
->group_exit_code
;
1378 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1379 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1380 * the lock and this task is uninteresting. If we return nonzero, we have
1381 * released the lock and the system call should return.
1383 static int wait_task_stopped(struct wait_opts
*wo
,
1384 int ptrace
, struct task_struct
*p
)
1386 struct siginfo __user
*infop
;
1387 int retval
, exit_code
, *p_code
, why
;
1388 uid_t uid
= 0; /* unneeded, required by compiler */
1392 * Traditionally we see ptrace'd stopped tasks regardless of options.
1394 if (!ptrace
&& !(wo
->wo_flags
& WUNTRACED
))
1398 spin_lock_irq(&p
->sighand
->siglock
);
1400 p_code
= task_stopped_code(p
, ptrace
);
1401 if (unlikely(!p_code
))
1404 exit_code
= *p_code
;
1408 if (!unlikely(wo
->wo_flags
& WNOWAIT
))
1413 spin_unlock_irq(&p
->sighand
->siglock
);
1418 * Now we are pretty sure this task is interesting.
1419 * Make sure it doesn't get reaped out from under us while we
1420 * give up the lock and then examine it below. We don't want to
1421 * keep holding onto the tasklist_lock while we call getrusage and
1422 * possibly take page faults for user memory.
1425 pid
= task_pid_vnr(p
);
1426 why
= ptrace
? CLD_TRAPPED
: CLD_STOPPED
;
1427 read_unlock(&tasklist_lock
);
1429 if (unlikely(wo
->wo_flags
& WNOWAIT
))
1430 return wait_noreap_copyout(wo
, p
, pid
, uid
, why
, exit_code
);
1432 retval
= wo
->wo_rusage
1433 ? getrusage(p
, RUSAGE_BOTH
, wo
->wo_rusage
) : 0;
1434 if (!retval
&& wo
->wo_stat
)
1435 retval
= put_user((exit_code
<< 8) | 0x7f, wo
->wo_stat
);
1437 infop
= wo
->wo_info
;
1438 if (!retval
&& infop
)
1439 retval
= put_user(SIGCHLD
, &infop
->si_signo
);
1440 if (!retval
&& infop
)
1441 retval
= put_user(0, &infop
->si_errno
);
1442 if (!retval
&& infop
)
1443 retval
= put_user((short)why
, &infop
->si_code
);
1444 if (!retval
&& infop
)
1445 retval
= put_user(exit_code
, &infop
->si_status
);
1446 if (!retval
&& infop
)
1447 retval
= put_user(pid
, &infop
->si_pid
);
1448 if (!retval
&& infop
)
1449 retval
= put_user(uid
, &infop
->si_uid
);
1459 * Handle do_wait work for one task in a live, non-stopped state.
1460 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1461 * the lock and this task is uninteresting. If we return nonzero, we have
1462 * released the lock and the system call should return.
1464 static int wait_task_continued(struct wait_opts
*wo
, struct task_struct
*p
)
1470 if (!unlikely(wo
->wo_flags
& WCONTINUED
))
1473 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
))
1476 spin_lock_irq(&p
->sighand
->siglock
);
1477 /* Re-check with the lock held. */
1478 if (!(p
->signal
->flags
& SIGNAL_STOP_CONTINUED
)) {
1479 spin_unlock_irq(&p
->sighand
->siglock
);
1482 if (!unlikely(wo
->wo_flags
& WNOWAIT
))
1483 p
->signal
->flags
&= ~SIGNAL_STOP_CONTINUED
;
1485 spin_unlock_irq(&p
->sighand
->siglock
);
1487 pid
= task_pid_vnr(p
);
1489 read_unlock(&tasklist_lock
);
1492 retval
= wo
->wo_rusage
1493 ? getrusage(p
, RUSAGE_BOTH
, wo
->wo_rusage
) : 0;
1495 if (!retval
&& wo
->wo_stat
)
1496 retval
= put_user(0xffff, wo
->wo_stat
);
1500 retval
= wait_noreap_copyout(wo
, p
, pid
, uid
,
1501 CLD_CONTINUED
, SIGCONT
);
1502 BUG_ON(retval
== 0);
1509 * Consider @p for a wait by @parent.
1511 * -ECHILD should be in ->notask_error before the first call.
1512 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1513 * Returns zero if the search for a child should continue;
1514 * then ->notask_error is 0 if @p is an eligible child,
1515 * or another error from security_task_wait(), or still -ECHILD.
1517 static int wait_consider_task(struct wait_opts
*wo
, int ptrace
,
1518 struct task_struct
*p
)
1520 int ret
= eligible_child(wo
, p
);
1524 ret
= security_task_wait(p
);
1525 if (unlikely(ret
< 0)) {
1527 * If we have not yet seen any eligible child,
1528 * then let this error code replace -ECHILD.
1529 * A permission error will give the user a clue
1530 * to look for security policy problems, rather
1531 * than for mysterious wait bugs.
1533 if (wo
->notask_error
)
1534 wo
->notask_error
= ret
;
1538 if (likely(!ptrace
) && unlikely(task_ptrace(p
))) {
1540 * This child is hidden by ptrace.
1541 * We aren't allowed to see it now, but eventually we will.
1543 wo
->notask_error
= 0;
1547 if (p
->exit_state
== EXIT_DEAD
)
1551 * We don't reap group leaders with subthreads.
1553 if (p
->exit_state
== EXIT_ZOMBIE
&& !delay_group_leader(p
))
1554 return wait_task_zombie(wo
, p
);
1557 * It's stopped or running now, so it might
1558 * later continue, exit, or stop again.
1560 wo
->notask_error
= 0;
1562 if (task_stopped_code(p
, ptrace
))
1563 return wait_task_stopped(wo
, ptrace
, p
);
1565 return wait_task_continued(wo
, p
);
1569 * Do the work of do_wait() for one thread in the group, @tsk.
1571 * -ECHILD should be in ->notask_error before the first call.
1572 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1573 * Returns zero if the search for a child should continue; then
1574 * ->notask_error is 0 if there were any eligible children,
1575 * or another error from security_task_wait(), or still -ECHILD.
1577 static int do_wait_thread(struct wait_opts
*wo
, struct task_struct
*tsk
)
1579 struct task_struct
*p
;
1581 list_for_each_entry(p
, &tsk
->children
, sibling
) {
1582 int ret
= wait_consider_task(wo
, 0, p
);
1590 static int ptrace_do_wait(struct wait_opts
*wo
, struct task_struct
*tsk
)
1592 struct task_struct
*p
;
1594 list_for_each_entry(p
, &tsk
->ptraced
, ptrace_entry
) {
1595 int ret
= wait_consider_task(wo
, 1, p
);
1603 static int child_wait_callback(wait_queue_t
*wait
, unsigned mode
,
1604 int sync
, void *key
)
1606 struct wait_opts
*wo
= container_of(wait
, struct wait_opts
,
1608 struct task_struct
*p
= key
;
1610 if (!eligible_pid(wo
, p
))
1613 if ((wo
->wo_flags
& __WNOTHREAD
) && wait
->private != p
->parent
)
1616 return default_wake_function(wait
, mode
, sync
, key
);
1619 void __wake_up_parent(struct task_struct
*p
, struct task_struct
*parent
)
1621 __wake_up_sync_key(&parent
->signal
->wait_chldexit
,
1622 TASK_INTERRUPTIBLE
, 1, p
);
1625 static long do_wait(struct wait_opts
*wo
)
1627 struct task_struct
*tsk
;
1630 trace_sched_process_wait(wo
->wo_pid
);
1632 init_waitqueue_func_entry(&wo
->child_wait
, child_wait_callback
);
1633 wo
->child_wait
.private = current
;
1634 add_wait_queue(¤t
->signal
->wait_chldexit
, &wo
->child_wait
);
1637 * If there is nothing that can match our critiera just get out.
1638 * We will clear ->notask_error to zero if we see any child that
1639 * might later match our criteria, even if we are not able to reap
1642 wo
->notask_error
= -ECHILD
;
1643 if ((wo
->wo_type
< PIDTYPE_MAX
) &&
1644 (!wo
->wo_pid
|| hlist_empty(&wo
->wo_pid
->tasks
[wo
->wo_type
])))
1647 set_current_state(TASK_INTERRUPTIBLE
);
1648 read_lock(&tasklist_lock
);
1651 retval
= do_wait_thread(wo
, tsk
);
1655 retval
= ptrace_do_wait(wo
, tsk
);
1659 if (wo
->wo_flags
& __WNOTHREAD
)
1661 } while_each_thread(current
, tsk
);
1662 read_unlock(&tasklist_lock
);
1665 retval
= wo
->notask_error
;
1666 if (!retval
&& !(wo
->wo_flags
& WNOHANG
)) {
1667 retval
= -ERESTARTSYS
;
1668 if (!signal_pending(current
)) {
1674 __set_current_state(TASK_RUNNING
);
1675 remove_wait_queue(¤t
->signal
->wait_chldexit
, &wo
->child_wait
);
1679 SYSCALL_DEFINE5(waitid
, int, which
, pid_t
, upid
, struct siginfo __user
*,
1680 infop
, int, options
, struct rusage __user
*, ru
)
1682 struct wait_opts wo
;
1683 struct pid
*pid
= NULL
;
1687 if (options
& ~(WNOHANG
|WNOWAIT
|WEXITED
|WSTOPPED
|WCONTINUED
))
1689 if (!(options
& (WEXITED
|WSTOPPED
|WCONTINUED
)))
1702 type
= PIDTYPE_PGID
;
1710 if (type
< PIDTYPE_MAX
)
1711 pid
= find_get_pid(upid
);
1715 wo
.wo_flags
= options
;
1725 * For a WNOHANG return, clear out all the fields
1726 * we would set so the user can easily tell the
1730 ret
= put_user(0, &infop
->si_signo
);
1732 ret
= put_user(0, &infop
->si_errno
);
1734 ret
= put_user(0, &infop
->si_code
);
1736 ret
= put_user(0, &infop
->si_pid
);
1738 ret
= put_user(0, &infop
->si_uid
);
1740 ret
= put_user(0, &infop
->si_status
);
1745 /* avoid REGPARM breakage on x86: */
1746 asmlinkage_protect(5, ret
, which
, upid
, infop
, options
, ru
);
1750 SYSCALL_DEFINE4(wait4
, pid_t
, upid
, int __user
*, stat_addr
,
1751 int, options
, struct rusage __user
*, ru
)
1753 struct wait_opts wo
;
1754 struct pid
*pid
= NULL
;
1758 if (options
& ~(WNOHANG
|WUNTRACED
|WCONTINUED
|
1759 __WNOTHREAD
|__WCLONE
|__WALL
))
1764 else if (upid
< 0) {
1765 type
= PIDTYPE_PGID
;
1766 pid
= find_get_pid(-upid
);
1767 } else if (upid
== 0) {
1768 type
= PIDTYPE_PGID
;
1769 pid
= get_task_pid(current
, PIDTYPE_PGID
);
1770 } else /* upid > 0 */ {
1772 pid
= find_get_pid(upid
);
1777 wo
.wo_flags
= options
| WEXITED
;
1779 wo
.wo_stat
= stat_addr
;
1784 /* avoid REGPARM breakage on x86: */
1785 asmlinkage_protect(4, ret
, upid
, stat_addr
, options
, ru
);
1789 #ifdef __ARCH_WANT_SYS_WAITPID
1792 * sys_waitpid() remains for compatibility. waitpid() should be
1793 * implemented by calling sys_wait4() from libc.a.
1795 SYSCALL_DEFINE3(waitpid
, pid_t
, pid
, int __user
*, stat_addr
, int, options
)
1797 return sys_wait4(pid
, stat_addr
, options
, NULL
);