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>
54 #include <linux/writeback.h>
55 #include <linux/shm.h>
57 #include <asm/uaccess.h>
58 #include <asm/unistd.h>
59 #include <asm/pgtable.h>
60 #include <asm/mmu_context.h>
62 static void exit_mm(struct task_struct * tsk);
64 static void __unhash_process(struct task_struct *p, bool group_dead)
67 detach_pid(p, PIDTYPE_PID);
69 detach_pid(p, PIDTYPE_PGID);
70 detach_pid(p, PIDTYPE_SID);
72 list_del_rcu(&p->tasks);
73 list_del_init(&p->sibling);
74 __this_cpu_dec(process_counts);
76 list_del_rcu(&p->thread_group);
80 * This function expects the tasklist_lock write-locked.
82 static void __exit_signal(struct task_struct *tsk)
84 struct signal_struct *sig = tsk->signal;
85 bool group_dead = thread_group_leader(tsk);
86 struct sighand_struct *sighand;
87 struct tty_struct *uninitialized_var(tty);
89 sighand = rcu_dereference_check(tsk->sighand,
90 lockdep_tasklist_lock_is_held());
91 spin_lock(&sighand->siglock);
93 posix_cpu_timers_exit(tsk);
95 posix_cpu_timers_exit_group(tsk);
100 * This can only happen if the caller is de_thread().
101 * FIXME: this is the temporary hack, we should teach
102 * posix-cpu-timers to handle this case correctly.
104 if (unlikely(has_group_leader_pid(tsk)))
105 posix_cpu_timers_exit_group(tsk);
108 * If there is any task waiting for the group exit
111 if (sig->notify_count > 0 && !--sig->notify_count)
112 wake_up_process(sig->group_exit_task);
114 if (tsk == sig->curr_target)
115 sig->curr_target = next_thread(tsk);
117 * Accumulate here the counters for all threads but the
118 * group leader as they die, so they can be added into
119 * the process-wide totals when those are taken.
120 * The group leader stays around as a zombie as long
121 * as there are other threads. When it gets reaped,
122 * the exit.c code will add its counts into these totals.
123 * We won't ever get here for the group leader, since it
124 * will have been the last reference on the signal_struct.
126 sig->utime += tsk->utime;
127 sig->stime += tsk->stime;
128 sig->gtime += tsk->gtime;
129 sig->min_flt += tsk->min_flt;
130 sig->maj_flt += tsk->maj_flt;
131 sig->nvcsw += tsk->nvcsw;
132 sig->nivcsw += tsk->nivcsw;
133 sig->inblock += task_io_get_inblock(tsk);
134 sig->oublock += task_io_get_oublock(tsk);
135 task_io_accounting_add(&sig->ioac, &tsk->ioac);
136 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
140 __unhash_process(tsk, group_dead);
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);
148 spin_unlock(&sighand->siglock);
150 __cleanup_sighand(sighand);
151 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
153 flush_sigqueue(&sig->shared_pending);
158 static void delayed_put_task_struct(struct rcu_head *rhp)
160 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
162 perf_event_delayed_put(tsk);
163 trace_sched_process_free(tsk);
164 put_task_struct(tsk);
168 void release_task(struct task_struct * p)
170 struct task_struct *leader;
173 /* don't need to get the RCU readlock here - the process is dead and
174 * can't be modifying its own credentials. But shut RCU-lockdep up */
176 atomic_dec(&__task_cred(p)->user->processes);
181 write_lock_irq(&tasklist_lock);
182 ptrace_release_task(p);
186 * If we are the last non-leader member of the thread
187 * group, and the leader is zombie, then notify the
188 * group leader's parent process. (if it wants notification.)
191 leader = p->group_leader;
192 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
194 * If we were the last child thread and the leader has
195 * exited already, and the leader's parent ignores SIGCHLD,
196 * then we are the one who should release the leader.
198 zap_leader = do_notify_parent(leader, leader->exit_signal);
200 leader->exit_state = EXIT_DEAD;
203 write_unlock_irq(&tasklist_lock);
205 call_rcu(&p->rcu, delayed_put_task_struct);
208 if (unlikely(zap_leader))
213 * This checks not only the pgrp, but falls back on the pid if no
214 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
217 * The caller must hold rcu lock or the tasklist lock.
219 struct pid *session_of_pgrp(struct pid *pgrp)
221 struct task_struct *p;
222 struct pid *sid = NULL;
224 p = pid_task(pgrp, PIDTYPE_PGID);
226 p = pid_task(pgrp, PIDTYPE_PID);
228 sid = task_session(p);
234 * Determine if a process group is "orphaned", according to the POSIX
235 * definition in 2.2.2.52. Orphaned process groups are not to be affected
236 * by terminal-generated stop signals. Newly orphaned process groups are
237 * to receive a SIGHUP and a SIGCONT.
239 * "I ask you, have you ever known what it is to be an orphan?"
241 static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
243 struct task_struct *p;
245 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
246 if ((p == ignored_task) ||
247 (p->exit_state && thread_group_empty(p)) ||
248 is_global_init(p->real_parent))
251 if (task_pgrp(p->real_parent) != pgrp &&
252 task_session(p->real_parent) == task_session(p))
254 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
259 int is_current_pgrp_orphaned(void)
263 read_lock(&tasklist_lock);
264 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
265 read_unlock(&tasklist_lock);
270 static bool has_stopped_jobs(struct pid *pgrp)
272 struct task_struct *p;
274 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
275 if (p->signal->flags & SIGNAL_STOP_STOPPED)
277 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
283 * Check to see if any process groups have become orphaned as
284 * a result of our exiting, and if they have any stopped jobs,
285 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
288 kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
290 struct pid *pgrp = task_pgrp(tsk);
291 struct task_struct *ignored_task = tsk;
294 /* exit: our father is in a different pgrp than
295 * we are and we were the only connection outside.
297 parent = tsk->real_parent;
299 /* reparent: our child is in a different pgrp than
300 * we are, and it was the only connection outside.
304 if (task_pgrp(parent) != pgrp &&
305 task_session(parent) == task_session(tsk) &&
306 will_become_orphaned_pgrp(pgrp, ignored_task) &&
307 has_stopped_jobs(pgrp)) {
308 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
309 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
314 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
316 * If a kernel thread is launched as a result of a system call, or if
317 * it ever exits, it should generally reparent itself to kthreadd so it
318 * isn't in the way of other processes and is correctly cleaned up on exit.
320 * The various task state such as scheduling policy and priority may have
321 * been inherited from a user process, so we reset them to sane values here.
323 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
325 static void reparent_to_kthreadd(void)
327 write_lock_irq(&tasklist_lock);
329 ptrace_unlink(current);
330 /* Reparent to init */
331 current->real_parent = current->parent = kthreadd_task;
332 list_move_tail(¤t->sibling, ¤t->real_parent->children);
334 /* Set the exit signal to SIGCHLD so we signal init on exit */
335 current->exit_signal = SIGCHLD;
337 if (task_nice(current) < 0)
338 set_user_nice(current, 0);
342 memcpy(current->signal->rlim, init_task.signal->rlim,
343 sizeof(current->signal->rlim));
345 atomic_inc(&init_cred.usage);
346 commit_creds(&init_cred);
347 write_unlock_irq(&tasklist_lock);
350 void __set_special_pids(struct pid *pid)
352 struct task_struct *curr = current->group_leader;
354 if (task_session(curr) != pid)
355 change_pid(curr, PIDTYPE_SID, pid);
357 if (task_pgrp(curr) != pid)
358 change_pid(curr, PIDTYPE_PGID, pid);
361 static void set_special_pids(struct pid *pid)
363 write_lock_irq(&tasklist_lock);
364 __set_special_pids(pid);
365 write_unlock_irq(&tasklist_lock);
369 * Let kernel threads use this to say that they allow a certain signal.
370 * Must not be used if kthread was cloned with CLONE_SIGHAND.
372 int allow_signal(int sig)
374 if (!valid_signal(sig) || sig < 1)
377 spin_lock_irq(¤t->sighand->siglock);
378 /* This is only needed for daemonize()'ed kthreads */
379 sigdelset(¤t->blocked, sig);
381 * Kernel threads handle their own signals. Let the signal code
382 * know it'll be handled, so that they don't get converted to
383 * SIGKILL or just silently dropped.
385 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
387 spin_unlock_irq(¤t->sighand->siglock);
391 EXPORT_SYMBOL(allow_signal);
393 int disallow_signal(int sig)
395 if (!valid_signal(sig) || sig < 1)
398 spin_lock_irq(¤t->sighand->siglock);
399 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
401 spin_unlock_irq(¤t->sighand->siglock);
405 EXPORT_SYMBOL(disallow_signal);
408 * Put all the gunge required to become a kernel thread without
409 * attached user resources in one place where it belongs.
412 void daemonize(const char *name, ...)
417 va_start(args, name);
418 vsnprintf(current->comm, sizeof(current->comm), name, args);
422 * If we were started as result of loading a module, close all of the
423 * user space pages. We don't need them, and if we didn't close them
424 * they would be locked into memory.
428 * We don't want to get frozen, in case system-wide hibernation
429 * or suspend transition begins right now.
431 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
433 if (current->nsproxy != &init_nsproxy) {
434 get_nsproxy(&init_nsproxy);
435 switch_task_namespaces(current, &init_nsproxy);
437 set_special_pids(&init_struct_pid);
438 proc_clear_tty(current);
440 /* Block and flush all signals */
441 sigfillset(&blocked);
442 sigprocmask(SIG_BLOCK, &blocked, NULL);
443 flush_signals(current);
445 /* Become as one with the init task */
447 daemonize_fs_struct();
449 current->files = init_task.files;
450 atomic_inc(¤t->files->count);
452 reparent_to_kthreadd();
455 EXPORT_SYMBOL(daemonize);
457 static void close_files(struct files_struct * files)
465 * It is safe to dereference the fd table without RCU or
466 * ->file_lock because this is the last reference to the
467 * files structure. But use RCU to shut RCU-lockdep up.
470 fdt = files_fdtable(files);
475 if (i >= fdt->max_fds)
477 set = fdt->open_fds->fds_bits[j++];
480 struct file * file = xchg(&fdt->fd[i], NULL);
482 filp_close(file, files);
492 struct files_struct *get_files_struct(struct task_struct *task)
494 struct files_struct *files;
499 atomic_inc(&files->count);
505 void put_files_struct(struct files_struct *files)
509 if (atomic_dec_and_test(&files->count)) {
512 * Free the fd and fdset arrays if we expanded them.
513 * If the fdtable was embedded, pass files for freeing
514 * at the end of the RCU grace period. Otherwise,
515 * you can free files immediately.
518 fdt = files_fdtable(files);
519 if (fdt != &files->fdtab)
520 kmem_cache_free(files_cachep, files);
526 void reset_files_struct(struct files_struct *files)
528 struct task_struct *tsk = current;
529 struct files_struct *old;
535 put_files_struct(old);
538 void exit_files(struct task_struct *tsk)
540 struct files_struct * files = tsk->files;
546 put_files_struct(files);
550 #ifdef CONFIG_MM_OWNER
552 * A task is exiting. If it owned this mm, find a new owner for the mm.
554 void mm_update_next_owner(struct mm_struct *mm)
556 struct task_struct *c, *g, *p = current;
560 * If the exiting or execing task is not the owner, it's
561 * someone else's problem.
566 * The current owner is exiting/execing and there are no other
567 * candidates. Do not leave the mm pointing to a possibly
568 * freed task structure.
570 if (atomic_read(&mm->mm_users) <= 1) {
575 read_lock(&tasklist_lock);
577 * Search in the children
579 list_for_each_entry(c, &p->children, sibling) {
581 goto assign_new_owner;
585 * Search in the siblings
587 list_for_each_entry(c, &p->real_parent->children, sibling) {
589 goto assign_new_owner;
593 * Search through everything else. We should not get
596 do_each_thread(g, c) {
598 goto assign_new_owner;
599 } while_each_thread(g, c);
601 read_unlock(&tasklist_lock);
603 * We found no owner yet mm_users > 1: this implies that we are
604 * most likely racing with swapoff (try_to_unuse()) or /proc or
605 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
614 * The task_lock protects c->mm from changing.
615 * We always want mm->owner->mm == mm
619 * Delay read_unlock() till we have the task_lock()
620 * to ensure that c does not slip away underneath us
622 read_unlock(&tasklist_lock);
632 #endif /* CONFIG_MM_OWNER */
635 * Turn us into a lazy TLB process if we
638 static void exit_mm(struct task_struct * tsk)
640 struct mm_struct *mm = tsk->mm;
641 struct core_state *core_state;
647 * Serialize with any possible pending coredump.
648 * We must hold mmap_sem around checking core_state
649 * and clearing tsk->mm. The core-inducing thread
650 * will increment ->nr_threads for each thread in the
651 * group with ->mm != NULL.
653 down_read(&mm->mmap_sem);
654 core_state = mm->core_state;
656 struct core_thread self;
657 up_read(&mm->mmap_sem);
660 self.next = xchg(&core_state->dumper.next, &self);
662 * Implies mb(), the result of xchg() must be visible
663 * to core_state->dumper.
665 if (atomic_dec_and_test(&core_state->nr_threads))
666 complete(&core_state->startup);
669 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
670 if (!self.task) /* see coredump_finish() */
674 __set_task_state(tsk, TASK_RUNNING);
675 down_read(&mm->mmap_sem);
677 atomic_inc(&mm->mm_count);
678 BUG_ON(mm != tsk->active_mm);
679 /* more a memory barrier than a real lock */
682 up_read(&mm->mmap_sem);
683 enter_lazy_tlb(mm, current);
685 mm_update_next_owner(mm);
690 * When we die, we re-parent all our children, and try to:
691 * 1. give them to another thread in our thread group, if such a member exists
692 * 2. give it to the first ancestor process which prctl'd itself as a
693 * child_subreaper for its children (like a service manager)
694 * 3. give it to the init process (PID 1) in our pid namespace
696 static struct task_struct *find_new_reaper(struct task_struct *father)
697 __releases(&tasklist_lock)
698 __acquires(&tasklist_lock)
700 struct pid_namespace *pid_ns = task_active_pid_ns(father);
701 struct task_struct *thread;
704 while_each_thread(father, thread) {
705 if (thread->flags & PF_EXITING)
707 if (unlikely(pid_ns->child_reaper == father))
708 pid_ns->child_reaper = thread;
712 if (unlikely(pid_ns->child_reaper == father)) {
713 write_unlock_irq(&tasklist_lock);
714 if (unlikely(pid_ns == &init_pid_ns))
715 panic("Attempted to kill init!");
717 zap_pid_ns_processes(pid_ns);
718 write_lock_irq(&tasklist_lock);
720 * We can not clear ->child_reaper or leave it alone.
721 * There may by stealth EXIT_DEAD tasks on ->children,
722 * forget_original_parent() must move them somewhere.
724 pid_ns->child_reaper = init_pid_ns.child_reaper;
725 } else if (father->signal->has_child_subreaper) {
726 struct task_struct *reaper;
729 * Find the first ancestor marked as child_subreaper.
730 * Note that the code below checks same_thread_group(reaper,
731 * pid_ns->child_reaper). This is what we need to DTRT in a
732 * PID namespace. However we still need the check above, see
733 * http://marc.info/?l=linux-kernel&m=131385460420380
735 for (reaper = father->real_parent;
736 reaper != &init_task;
737 reaper = reaper->real_parent) {
738 if (same_thread_group(reaper, pid_ns->child_reaper))
740 if (!reaper->signal->is_child_subreaper)
744 if (!(thread->flags & PF_EXITING))
746 } while_each_thread(reaper, thread);
750 return pid_ns->child_reaper;
754 * Any that need to be release_task'd are put on the @dead list.
756 static void reparent_leader(struct task_struct *father, struct task_struct *p,
757 struct list_head *dead)
759 list_move_tail(&p->sibling, &p->real_parent->children);
761 if (p->exit_state == EXIT_DEAD)
764 * If this is a threaded reparent there is no need to
765 * notify anyone anything has happened.
767 if (same_thread_group(p->real_parent, father))
770 /* We don't want people slaying init. */
771 p->exit_signal = SIGCHLD;
773 /* If it has exited notify the new parent about this child's death. */
775 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
776 if (do_notify_parent(p, p->exit_signal)) {
777 p->exit_state = EXIT_DEAD;
778 list_move_tail(&p->sibling, dead);
782 kill_orphaned_pgrp(p, father);
785 static void forget_original_parent(struct task_struct *father)
787 struct task_struct *p, *n, *reaper;
788 LIST_HEAD(dead_children);
790 write_lock_irq(&tasklist_lock);
792 * Note that exit_ptrace() and find_new_reaper() might
793 * drop tasklist_lock and reacquire it.
796 reaper = find_new_reaper(father);
798 list_for_each_entry_safe(p, n, &father->children, sibling) {
799 struct task_struct *t = p;
801 t->real_parent = reaper;
802 if (t->parent == father) {
804 t->parent = t->real_parent;
806 if (t->pdeath_signal)
807 group_send_sig_info(t->pdeath_signal,
809 } while_each_thread(p, t);
810 reparent_leader(father, p, &dead_children);
812 write_unlock_irq(&tasklist_lock);
814 BUG_ON(!list_empty(&father->children));
816 list_for_each_entry_safe(p, n, &dead_children, sibling) {
817 list_del_init(&p->sibling);
823 * Send signals to all our closest relatives so that they know
824 * to properly mourn us..
826 static void exit_notify(struct task_struct *tsk, int group_dead)
831 * This does two things:
833 * A. Make init inherit all the child processes
834 * B. Check to see if any process groups have become orphaned
835 * as a result of our exiting, and if they have any stopped
836 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
838 forget_original_parent(tsk);
839 exit_task_namespaces(tsk);
841 write_lock_irq(&tasklist_lock);
843 kill_orphaned_pgrp(tsk->group_leader, NULL);
845 if (unlikely(tsk->ptrace)) {
846 int sig = thread_group_leader(tsk) &&
847 thread_group_empty(tsk) &&
848 !ptrace_reparented(tsk) ?
849 tsk->exit_signal : SIGCHLD;
850 autoreap = do_notify_parent(tsk, sig);
851 } else if (thread_group_leader(tsk)) {
852 autoreap = thread_group_empty(tsk) &&
853 do_notify_parent(tsk, tsk->exit_signal);
858 tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE;
860 /* mt-exec, de_thread() is waiting for group leader */
861 if (unlikely(tsk->signal->notify_count < 0))
862 wake_up_process(tsk->signal->group_exit_task);
863 write_unlock_irq(&tasklist_lock);
865 /* If the process is dead, release it - nobody will wait for it */
870 #ifdef CONFIG_DEBUG_STACK_USAGE
871 static void check_stack_usage(void)
873 static DEFINE_SPINLOCK(low_water_lock);
874 static int lowest_to_date = THREAD_SIZE;
877 free = stack_not_used(current);
879 if (free >= lowest_to_date)
882 spin_lock(&low_water_lock);
883 if (free < lowest_to_date) {
884 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
886 current->comm, free);
887 lowest_to_date = free;
889 spin_unlock(&low_water_lock);
892 static inline void check_stack_usage(void) {}
895 void do_exit(long code)
897 struct task_struct *tsk = current;
900 profile_task_exit(tsk);
902 WARN_ON(blk_needs_flush_plug(tsk));
904 if (unlikely(in_interrupt()))
905 panic("Aiee, killing interrupt handler!");
906 if (unlikely(!tsk->pid))
907 panic("Attempted to kill the idle task!");
910 * If do_exit is called because this processes oopsed, it's possible
911 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
912 * continuing. Amongst other possible reasons, this is to prevent
913 * mm_release()->clear_child_tid() from writing to a user-controlled
918 ptrace_event(PTRACE_EVENT_EXIT, code);
920 validate_creds_for_do_exit(tsk);
923 * We're taking recursive faults here in do_exit. Safest is to just
924 * leave this task alone and wait for reboot.
926 if (unlikely(tsk->flags & PF_EXITING)) {
928 "Fixing recursive fault but reboot is needed!\n");
930 * We can do this unlocked here. The futex code uses
931 * this flag just to verify whether the pi state
932 * cleanup has been done or not. In the worst case it
933 * loops once more. We pretend that the cleanup was
934 * done as there is no way to return. Either the
935 * OWNER_DIED bit is set by now or we push the blocked
936 * task into the wait for ever nirwana as well.
938 tsk->flags |= PF_EXITPIDONE;
939 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);
953 if (unlikely(in_atomic()))
954 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
955 current->comm, task_pid_nr(current),
958 acct_update_integrals(tsk);
959 /* sync mm's RSS info before statistics gathering */
961 sync_mm_rss(tsk->mm);
962 group_dead = atomic_dec_and_test(&tsk->signal->live);
964 hrtimer_cancel(&tsk->signal->real_timer);
965 exit_itimers(tsk->signal);
967 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
969 acct_collect(code, group_dead);
974 tsk->exit_code = code;
975 taskstats_exit(tsk, group_dead);
981 trace_sched_process_exit(tsk);
991 * Flush inherited counters to the parent - before the parent
992 * gets woken up by child-exit notifications.
994 * because of cgroup mode, must be called before cgroup_exit()
996 perf_event_exit_task(tsk);
1001 disassociate_ctty(1);
1003 module_put(task_thread_info(tsk)->exec_domain->module);
1005 proc_exit_connector(tsk);
1008 * FIXME: do that only when needed, using sched_exit tracepoint
1010 ptrace_put_breakpoints(tsk);
1012 exit_notify(tsk, group_dead);
1015 mpol_put(tsk->mempolicy);
1016 tsk->mempolicy = NULL;
1020 if (unlikely(current->pi_state_cache))
1021 kfree(current->pi_state_cache);
1024 * Make sure we are holding no locks:
1026 debug_check_no_locks_held(tsk);
1028 * We can do this unlocked here. The futex code uses this flag
1029 * just to verify whether the pi state cleanup has been done
1030 * or not. In the worst case it loops once more.
1032 tsk->flags |= PF_EXITPIDONE;
1034 if (tsk->io_context)
1035 exit_io_context(tsk);
1037 if (tsk->splice_pipe)
1038 __free_pipe_info(tsk->splice_pipe);
1040 validate_creds_for_do_exit(tsk);
1043 if (tsk->nr_dirtied)
1044 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
1048 * The setting of TASK_RUNNING by try_to_wake_up() may be delayed
1049 * when the following two conditions become true.
1050 * - There is race condition of mmap_sem (It is acquired by
1052 * - SMI occurs before setting TASK_RUNINNG.
1053 * (or hypervisor of virtual machine switches to other guest)
1054 * As a result, we may become TASK_RUNNING after becoming TASK_DEAD
1056 * To avoid it, we have to wait for releasing tsk->pi_lock which
1057 * is held by try_to_wake_up()
1060 raw_spin_unlock_wait(&tsk->pi_lock);
1062 /* causes final put_task_struct in finish_task_switch(). */
1063 tsk->state = TASK_DEAD;
1064 tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */
1067 /* Avoid "noreturn function does return". */
1069 cpu_relax(); /* For when BUG is null */
1072 EXPORT_SYMBOL_GPL(do_exit);
1074 void complete_and_exit(struct completion *comp, long code)
1082 EXPORT_SYMBOL(complete_and_exit);
1084 SYSCALL_DEFINE1(exit, int, error_code)
1086 do_exit((error_code&0xff)<<8);
1090 * Take down every thread in the group. This is called by fatal signals
1091 * as well as by sys_exit_group (below).
1094 do_group_exit(int exit_code)
1096 struct signal_struct *sig = current->signal;
1098 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1100 if (signal_group_exit(sig))
1101 exit_code = sig->group_exit_code;
1102 else if (!thread_group_empty(current)) {
1103 struct sighand_struct *const sighand = current->sighand;
1104 spin_lock_irq(&sighand->siglock);
1105 if (signal_group_exit(sig))
1106 /* Another thread got here before we took the lock. */
1107 exit_code = sig->group_exit_code;
1109 sig->group_exit_code = exit_code;
1110 sig->flags = SIGNAL_GROUP_EXIT;
1111 zap_other_threads(current);
1113 spin_unlock_irq(&sighand->siglock);
1121 * this kills every thread in the thread group. Note that any externally
1122 * wait4()-ing process will get the correct exit code - even if this
1123 * thread is not the thread group leader.
1125 SYSCALL_DEFINE1(exit_group, int, error_code)
1127 do_group_exit((error_code & 0xff) << 8);
1133 enum pid_type wo_type;
1137 struct siginfo __user *wo_info;
1138 int __user *wo_stat;
1139 struct rusage __user *wo_rusage;
1141 wait_queue_t child_wait;
1146 struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1148 if (type != PIDTYPE_PID)
1149 task = task->group_leader;
1150 return task->pids[type].pid;
1153 static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1155 return wo->wo_type == PIDTYPE_MAX ||
1156 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1159 static int eligible_child(struct wait_opts *wo, struct task_struct *p)
1161 if (!eligible_pid(wo, p))
1163 /* Wait for all children (clone and not) if __WALL is set;
1164 * otherwise, wait for clone children *only* if __WCLONE is
1165 * set; otherwise, wait for non-clone children *only*. (Note:
1166 * A "clone" child here is one that reports to its parent
1167 * using a signal other than SIGCHLD.) */
1168 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1169 && !(wo->wo_flags & __WALL))
1175 static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1176 pid_t pid, uid_t uid, int why, int status)
1178 struct siginfo __user *infop;
1179 int retval = wo->wo_rusage
1180 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1183 infop = wo->wo_info;
1186 retval = put_user(SIGCHLD, &infop->si_signo);
1188 retval = put_user(0, &infop->si_errno);
1190 retval = put_user((short)why, &infop->si_code);
1192 retval = put_user(pid, &infop->si_pid);
1194 retval = put_user(uid, &infop->si_uid);
1196 retval = put_user(status, &infop->si_status);
1204 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1205 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1206 * the lock and this task is uninteresting. If we return nonzero, we have
1207 * released the lock and the system call should return.
1209 static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1211 unsigned long state;
1212 int retval, status, traced;
1213 pid_t pid = task_pid_vnr(p);
1214 uid_t uid = __task_cred(p)->uid;
1215 struct siginfo __user *infop;
1217 if (!likely(wo->wo_flags & WEXITED))
1220 if (unlikely(wo->wo_flags & WNOWAIT)) {
1221 int exit_code = p->exit_code;
1225 read_unlock(&tasklist_lock);
1226 if ((exit_code & 0x7f) == 0) {
1228 status = exit_code >> 8;
1230 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1231 status = exit_code & 0x7f;
1233 return wait_noreap_copyout(wo, p, pid, uid, why, status);
1237 * Try to move the task's state to DEAD
1238 * only one thread is allowed to do this:
1240 state = xchg(&p->exit_state, EXIT_DEAD);
1241 if (state != EXIT_ZOMBIE) {
1242 BUG_ON(state != EXIT_DEAD);
1246 traced = ptrace_reparented(p);
1248 * It can be ptraced but not reparented, check
1249 * thread_group_leader() to filter out sub-threads.
1251 if (likely(!traced) && thread_group_leader(p)) {
1252 struct signal_struct *psig;
1253 struct signal_struct *sig;
1254 unsigned long maxrss;
1255 cputime_t tgutime, tgstime;
1258 * The resource counters for the group leader are in its
1259 * own task_struct. Those for dead threads in the group
1260 * are in its signal_struct, as are those for the child
1261 * processes it has previously reaped. All these
1262 * accumulate in the parent's signal_struct c* fields.
1264 * We don't bother to take a lock here to protect these
1265 * p->signal fields, because they are only touched by
1266 * __exit_signal, which runs with tasklist_lock
1267 * write-locked anyway, and so is excluded here. We do
1268 * need to protect the access to parent->signal fields,
1269 * as other threads in the parent group can be right
1270 * here reaping other children at the same time.
1272 * We use thread_group_times() to get times for the thread
1273 * group, which consolidates times for all threads in the
1274 * group including the group leader.
1276 thread_group_times(p, &tgutime, &tgstime);
1277 spin_lock_irq(&p->real_parent->sighand->siglock);
1278 psig = p->real_parent->signal;
1280 psig->cutime += tgutime + sig->cutime;
1281 psig->cstime += tgstime + sig->cstime;
1282 psig->cgtime += p->gtime + sig->gtime + sig->cgtime;
1284 p->min_flt + sig->min_flt + sig->cmin_flt;
1286 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1288 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1290 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1292 task_io_get_inblock(p) +
1293 sig->inblock + sig->cinblock;
1295 task_io_get_oublock(p) +
1296 sig->oublock + sig->coublock;
1297 maxrss = max(sig->maxrss, sig->cmaxrss);
1298 if (psig->cmaxrss < maxrss)
1299 psig->cmaxrss = maxrss;
1300 task_io_accounting_add(&psig->ioac, &p->ioac);
1301 task_io_accounting_add(&psig->ioac, &sig->ioac);
1302 spin_unlock_irq(&p->real_parent->sighand->siglock);
1306 * Now we are sure this task is interesting, and no other
1307 * thread can reap it because we set its state to EXIT_DEAD.
1309 read_unlock(&tasklist_lock);
1311 retval = wo->wo_rusage
1312 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1313 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1314 ? p->signal->group_exit_code : p->exit_code;
1315 if (!retval && wo->wo_stat)
1316 retval = put_user(status, wo->wo_stat);
1318 infop = wo->wo_info;
1319 if (!retval && infop)
1320 retval = put_user(SIGCHLD, &infop->si_signo);
1321 if (!retval && infop)
1322 retval = put_user(0, &infop->si_errno);
1323 if (!retval && infop) {
1326 if ((status & 0x7f) == 0) {
1330 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1333 retval = put_user((short)why, &infop->si_code);
1335 retval = put_user(status, &infop->si_status);
1337 if (!retval && infop)
1338 retval = put_user(pid, &infop->si_pid);
1339 if (!retval && infop)
1340 retval = put_user(uid, &infop->si_uid);
1345 write_lock_irq(&tasklist_lock);
1346 /* We dropped tasklist, ptracer could die and untrace */
1349 * If this is not a sub-thread, notify the parent.
1350 * If parent wants a zombie, don't release it now.
1352 if (thread_group_leader(p) &&
1353 !do_notify_parent(p, p->exit_signal)) {
1354 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 !(p->jobctl & JOBCTL_LISTENING))
1370 return &p->exit_code;
1372 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1373 return &p->signal->group_exit_code;
1379 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1381 * @ptrace: is the wait for ptrace
1382 * @p: task to wait for
1384 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1387 * read_lock(&tasklist_lock), which is released if return value is
1388 * non-zero. Also, grabs and releases @p->sighand->siglock.
1391 * 0 if wait condition didn't exist and search for other wait conditions
1392 * should continue. Non-zero return, -errno on failure and @p's pid on
1393 * success, implies that tasklist_lock is released and wait condition
1394 * search should terminate.
1396 static int wait_task_stopped(struct wait_opts *wo,
1397 int ptrace, struct task_struct *p)
1399 struct siginfo __user *infop;
1400 int retval, exit_code, *p_code, why;
1401 uid_t uid = 0; /* unneeded, required by compiler */
1405 * Traditionally we see ptrace'd stopped tasks regardless of options.
1407 if (!ptrace && !(wo->wo_flags & WUNTRACED))
1410 if (!task_stopped_code(p, ptrace))
1414 spin_lock_irq(&p->sighand->siglock);
1416 p_code = task_stopped_code(p, ptrace);
1417 if (unlikely(!p_code))
1420 exit_code = *p_code;
1424 if (!unlikely(wo->wo_flags & WNOWAIT))
1429 spin_unlock_irq(&p->sighand->siglock);
1434 * Now we are pretty sure this task is interesting.
1435 * Make sure it doesn't get reaped out from under us while we
1436 * give up the lock and then examine it below. We don't want to
1437 * keep holding onto the tasklist_lock while we call getrusage and
1438 * possibly take page faults for user memory.
1441 pid = task_pid_vnr(p);
1442 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1443 read_unlock(&tasklist_lock);
1445 if (unlikely(wo->wo_flags & WNOWAIT))
1446 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
1448 retval = wo->wo_rusage
1449 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1450 if (!retval && wo->wo_stat)
1451 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1453 infop = wo->wo_info;
1454 if (!retval && infop)
1455 retval = put_user(SIGCHLD, &infop->si_signo);
1456 if (!retval && infop)
1457 retval = put_user(0, &infop->si_errno);
1458 if (!retval && infop)
1459 retval = put_user((short)why, &infop->si_code);
1460 if (!retval && infop)
1461 retval = put_user(exit_code, &infop->si_status);
1462 if (!retval && infop)
1463 retval = put_user(pid, &infop->si_pid);
1464 if (!retval && infop)
1465 retval = put_user(uid, &infop->si_uid);
1475 * Handle do_wait work for one task in a live, non-stopped state.
1476 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1477 * the lock and this task is uninteresting. If we return nonzero, we have
1478 * released the lock and the system call should return.
1480 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1486 if (!unlikely(wo->wo_flags & WCONTINUED))
1489 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1492 spin_lock_irq(&p->sighand->siglock);
1493 /* Re-check with the lock held. */
1494 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1495 spin_unlock_irq(&p->sighand->siglock);
1498 if (!unlikely(wo->wo_flags & WNOWAIT))
1499 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1501 spin_unlock_irq(&p->sighand->siglock);
1503 pid = task_pid_vnr(p);
1505 read_unlock(&tasklist_lock);
1508 retval = wo->wo_rusage
1509 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1511 if (!retval && wo->wo_stat)
1512 retval = put_user(0xffff, wo->wo_stat);
1516 retval = wait_noreap_copyout(wo, p, pid, uid,
1517 CLD_CONTINUED, SIGCONT);
1518 BUG_ON(retval == 0);
1525 * Consider @p for a wait by @parent.
1527 * -ECHILD should be in ->notask_error before the first call.
1528 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1529 * Returns zero if the search for a child should continue;
1530 * then ->notask_error is 0 if @p is an eligible child,
1531 * or another error from security_task_wait(), or still -ECHILD.
1533 static int wait_consider_task(struct wait_opts *wo, int ptrace,
1534 struct task_struct *p)
1536 int ret = eligible_child(wo, p);
1540 ret = security_task_wait(p);
1541 if (unlikely(ret < 0)) {
1543 * If we have not yet seen any eligible child,
1544 * then let this error code replace -ECHILD.
1545 * A permission error will give the user a clue
1546 * to look for security policy problems, rather
1547 * than for mysterious wait bugs.
1549 if (wo->notask_error)
1550 wo->notask_error = ret;
1554 /* dead body doesn't have much to contribute */
1555 if (unlikely(p->exit_state == EXIT_DEAD)) {
1557 * But do not ignore this task until the tracer does
1558 * wait_task_zombie()->do_notify_parent().
1560 if (likely(!ptrace) && unlikely(ptrace_reparented(p)))
1561 wo->notask_error = 0;
1566 if (p->exit_state == EXIT_ZOMBIE) {
1568 * A zombie ptracee is only visible to its ptracer.
1569 * Notification and reaping will be cascaded to the real
1570 * parent when the ptracer detaches.
1572 if (likely(!ptrace) && unlikely(p->ptrace)) {
1573 /* it will become visible, clear notask_error */
1574 wo->notask_error = 0;
1578 /* we don't reap group leaders with subthreads */
1579 if (!delay_group_leader(p))
1580 return wait_task_zombie(wo, p);
1583 * Allow access to stopped/continued state via zombie by
1584 * falling through. Clearing of notask_error is complex.
1588 * If WEXITED is set, notask_error should naturally be
1589 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1590 * so, if there are live subthreads, there are events to
1591 * wait for. If all subthreads are dead, it's still safe
1592 * to clear - this function will be called again in finite
1593 * amount time once all the subthreads are released and
1594 * will then return without clearing.
1598 * Stopped state is per-task and thus can't change once the
1599 * target task dies. Only continued and exited can happen.
1600 * Clear notask_error if WCONTINUED | WEXITED.
1602 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1603 wo->notask_error = 0;
1606 * If @p is ptraced by a task in its real parent's group,
1607 * hide group stop/continued state when looking at @p as
1608 * the real parent; otherwise, a single stop can be
1609 * reported twice as group and ptrace stops.
1611 * If a ptracer wants to distinguish the two events for its
1612 * own children, it should create a separate process which
1613 * takes the role of real parent.
1615 if (likely(!ptrace) && p->ptrace && !ptrace_reparented(p))
1619 * @p is alive and it's gonna stop, continue or exit, so
1620 * there always is something to wait for.
1622 wo->notask_error = 0;
1626 * Wait for stopped. Depending on @ptrace, different stopped state
1627 * is used and the two don't interact with each other.
1629 ret = wait_task_stopped(wo, ptrace, p);
1634 * Wait for continued. There's only one continued state and the
1635 * ptracer can consume it which can confuse the real parent. Don't
1636 * use WCONTINUED from ptracer. You don't need or want it.
1638 return wait_task_continued(wo, p);
1642 * Do the work of do_wait() for one thread in the group, @tsk.
1644 * -ECHILD should be in ->notask_error before the first call.
1645 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1646 * Returns zero if the search for a child should continue; then
1647 * ->notask_error is 0 if there were any eligible children,
1648 * or another error from security_task_wait(), or still -ECHILD.
1650 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1652 struct task_struct *p;
1654 list_for_each_entry(p, &tsk->children, sibling) {
1655 int ret = wait_consider_task(wo, 0, p);
1663 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1665 struct task_struct *p;
1667 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1668 int ret = wait_consider_task(wo, 1, p);
1676 static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1677 int sync, void *key)
1679 struct wait_opts *wo = container_of(wait, struct wait_opts,
1681 struct task_struct *p = key;
1683 if (!eligible_pid(wo, p))
1686 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1689 return default_wake_function(wait, mode, sync, key);
1692 void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1694 __wake_up_sync_key(&parent->signal->wait_chldexit,
1695 TASK_INTERRUPTIBLE, 1, p);
1698 static long do_wait(struct wait_opts *wo)
1700 struct task_struct *tsk;
1703 trace_sched_process_wait(wo->wo_pid);
1705 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1706 wo->child_wait.private = current;
1707 add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
1710 * If there is nothing that can match our critiera just get out.
1711 * We will clear ->notask_error to zero if we see any child that
1712 * might later match our criteria, even if we are not able to reap
1715 wo->notask_error = -ECHILD;
1716 if ((wo->wo_type < PIDTYPE_MAX) &&
1717 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
1720 set_current_state(TASK_INTERRUPTIBLE);
1721 read_lock(&tasklist_lock);
1724 retval = do_wait_thread(wo, tsk);
1728 retval = ptrace_do_wait(wo, tsk);
1732 if (wo->wo_flags & __WNOTHREAD)
1734 } while_each_thread(current, tsk);
1735 read_unlock(&tasklist_lock);
1738 retval = wo->notask_error;
1739 if (!retval && !(wo->wo_flags & WNOHANG)) {
1740 retval = -ERESTARTSYS;
1741 if (!signal_pending(current)) {
1747 __set_current_state(TASK_RUNNING);
1748 remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
1752 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1753 infop, int, options, struct rusage __user *, ru)
1755 struct wait_opts wo;
1756 struct pid *pid = NULL;
1760 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1762 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1775 type = PIDTYPE_PGID;
1783 if (type < PIDTYPE_MAX)
1784 pid = find_get_pid(upid);
1788 wo.wo_flags = options;
1798 * For a WNOHANG return, clear out all the fields
1799 * we would set so the user can easily tell the
1803 ret = put_user(0, &infop->si_signo);
1805 ret = put_user(0, &infop->si_errno);
1807 ret = put_user(0, &infop->si_code);
1809 ret = put_user(0, &infop->si_pid);
1811 ret = put_user(0, &infop->si_uid);
1813 ret = put_user(0, &infop->si_status);
1818 /* avoid REGPARM breakage on x86: */
1819 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
1823 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1824 int, options, struct rusage __user *, ru)
1826 struct wait_opts wo;
1827 struct pid *pid = NULL;
1831 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1832 __WNOTHREAD|__WCLONE|__WALL))
1837 else if (upid < 0) {
1838 type = PIDTYPE_PGID;
1839 pid = find_get_pid(-upid);
1840 } else if (upid == 0) {
1841 type = PIDTYPE_PGID;
1842 pid = get_task_pid(current, PIDTYPE_PGID);
1843 } else /* upid > 0 */ {
1845 pid = find_get_pid(upid);
1850 wo.wo_flags = options | WEXITED;
1852 wo.wo_stat = stat_addr;
1857 /* avoid REGPARM breakage on x86: */
1858 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
1862 #ifdef __ARCH_WANT_SYS_WAITPID
1865 * sys_waitpid() remains for compatibility. waitpid() should be
1866 * implemented by calling sys_wait4() from libc.a.
1868 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1870 return sys_wait4(pid, stat_addr, options, NULL);