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.
426 mm_release(current, current->mm);
429 * We don't want to get frozen, in case system-wide hibernation
430 * or suspend transition begins right now.
432 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
434 if (current->nsproxy != &init_nsproxy) {
435 get_nsproxy(&init_nsproxy);
436 switch_task_namespaces(current, &init_nsproxy);
438 set_special_pids(&init_struct_pid);
439 proc_clear_tty(current);
441 /* Block and flush all signals */
442 sigfillset(&blocked);
443 sigprocmask(SIG_BLOCK, &blocked, NULL);
444 flush_signals(current);
446 /* Become as one with the init task */
448 daemonize_fs_struct();
450 current->files = init_task.files;
451 atomic_inc(¤t->files->count);
453 reparent_to_kthreadd();
456 EXPORT_SYMBOL(daemonize);
458 static void close_files(struct files_struct * files)
466 * It is safe to dereference the fd table without RCU or
467 * ->file_lock because this is the last reference to the
468 * files structure. But use RCU to shut RCU-lockdep up.
471 fdt = files_fdtable(files);
476 if (i >= fdt->max_fds)
478 set = fdt->open_fds[j++];
481 struct file * file = xchg(&fdt->fd[i], NULL);
483 filp_close(file, files);
493 struct files_struct *get_files_struct(struct task_struct *task)
495 struct files_struct *files;
500 atomic_inc(&files->count);
506 void put_files_struct(struct files_struct *files)
510 if (atomic_dec_and_test(&files->count)) {
513 * Free the fd and fdset arrays if we expanded them.
514 * If the fdtable was embedded, pass files for freeing
515 * at the end of the RCU grace period. Otherwise,
516 * you can free files immediately.
519 fdt = files_fdtable(files);
520 if (fdt != &files->fdtab)
521 kmem_cache_free(files_cachep, files);
527 void reset_files_struct(struct files_struct *files)
529 struct task_struct *tsk = current;
530 struct files_struct *old;
536 put_files_struct(old);
539 void exit_files(struct task_struct *tsk)
541 struct files_struct * files = tsk->files;
547 put_files_struct(files);
551 #ifdef CONFIG_MM_OWNER
553 * A task is exiting. If it owned this mm, find a new owner for the mm.
555 void mm_update_next_owner(struct mm_struct *mm)
557 struct task_struct *c, *g, *p = current;
561 * If the exiting or execing task is not the owner, it's
562 * someone else's problem.
567 * The current owner is exiting/execing and there are no other
568 * candidates. Do not leave the mm pointing to a possibly
569 * freed task structure.
571 if (atomic_read(&mm->mm_users) <= 1) {
576 read_lock(&tasklist_lock);
578 * Search in the children
580 list_for_each_entry(c, &p->children, sibling) {
582 goto assign_new_owner;
586 * Search in the siblings
588 list_for_each_entry(c, &p->real_parent->children, sibling) {
590 goto assign_new_owner;
594 * Search through everything else. We should not get
597 do_each_thread(g, c) {
599 goto assign_new_owner;
600 } while_each_thread(g, c);
602 read_unlock(&tasklist_lock);
604 * We found no owner yet mm_users > 1: this implies that we are
605 * most likely racing with swapoff (try_to_unuse()) or /proc or
606 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
615 * The task_lock protects c->mm from changing.
616 * We always want mm->owner->mm == mm
620 * Delay read_unlock() till we have the task_lock()
621 * to ensure that c does not slip away underneath us
623 read_unlock(&tasklist_lock);
633 #endif /* CONFIG_MM_OWNER */
636 * Turn us into a lazy TLB process if we
639 static void exit_mm(struct task_struct * tsk)
641 struct mm_struct *mm = tsk->mm;
642 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! exitcode=0x%08x\n",
716 father->signal->group_exit_code ?:
720 zap_pid_ns_processes(pid_ns);
721 write_lock_irq(&tasklist_lock);
723 * We can not clear ->child_reaper or leave it alone.
724 * There may by stealth EXIT_DEAD tasks on ->children,
725 * forget_original_parent() must move them somewhere.
727 pid_ns->child_reaper = init_pid_ns.child_reaper;
728 } else if (father->signal->has_child_subreaper) {
729 struct task_struct *reaper;
732 * Find the first ancestor marked as child_subreaper.
733 * Note that the code below checks same_thread_group(reaper,
734 * pid_ns->child_reaper). This is what we need to DTRT in a
735 * PID namespace. However we still need the check above, see
736 * http://marc.info/?l=linux-kernel&m=131385460420380
738 for (reaper = father->real_parent;
739 reaper != &init_task;
740 reaper = reaper->real_parent) {
741 if (same_thread_group(reaper, pid_ns->child_reaper))
743 if (!reaper->signal->is_child_subreaper)
747 if (!(thread->flags & PF_EXITING))
749 } while_each_thread(reaper, thread);
753 return pid_ns->child_reaper;
757 * Any that need to be release_task'd are put on the @dead list.
759 static void reparent_leader(struct task_struct *father, struct task_struct *p,
760 struct list_head *dead)
762 list_move_tail(&p->sibling, &p->real_parent->children);
764 if (p->exit_state == EXIT_DEAD)
767 * If this is a threaded reparent there is no need to
768 * notify anyone anything has happened.
770 if (same_thread_group(p->real_parent, father))
773 /* We don't want people slaying init. */
774 p->exit_signal = SIGCHLD;
776 /* If it has exited notify the new parent about this child's death. */
778 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
779 if (do_notify_parent(p, p->exit_signal)) {
780 p->exit_state = EXIT_DEAD;
781 list_move_tail(&p->sibling, dead);
785 kill_orphaned_pgrp(p, father);
788 static void forget_original_parent(struct task_struct *father)
790 struct task_struct *p, *n, *reaper;
791 LIST_HEAD(dead_children);
793 write_lock_irq(&tasklist_lock);
795 * Note that exit_ptrace() and find_new_reaper() might
796 * drop tasklist_lock and reacquire it.
799 reaper = find_new_reaper(father);
801 list_for_each_entry_safe(p, n, &father->children, sibling) {
802 struct task_struct *t = p;
804 t->real_parent = reaper;
805 if (t->parent == father) {
807 t->parent = t->real_parent;
809 if (t->pdeath_signal)
810 group_send_sig_info(t->pdeath_signal,
812 } while_each_thread(p, t);
813 reparent_leader(father, p, &dead_children);
815 write_unlock_irq(&tasklist_lock);
817 BUG_ON(!list_empty(&father->children));
819 list_for_each_entry_safe(p, n, &dead_children, sibling) {
820 list_del_init(&p->sibling);
826 * Send signals to all our closest relatives so that they know
827 * to properly mourn us..
829 static void exit_notify(struct task_struct *tsk, int group_dead)
834 * This does two things:
836 * A. Make init inherit all the child processes
837 * B. Check to see if any process groups have become orphaned
838 * as a result of our exiting, and if they have any stopped
839 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
841 forget_original_parent(tsk);
842 exit_task_namespaces(tsk);
844 write_lock_irq(&tasklist_lock);
846 kill_orphaned_pgrp(tsk->group_leader, NULL);
848 if (unlikely(tsk->ptrace)) {
849 int sig = thread_group_leader(tsk) &&
850 thread_group_empty(tsk) &&
851 !ptrace_reparented(tsk) ?
852 tsk->exit_signal : SIGCHLD;
853 autoreap = do_notify_parent(tsk, sig);
854 } else if (thread_group_leader(tsk)) {
855 autoreap = thread_group_empty(tsk) &&
856 do_notify_parent(tsk, tsk->exit_signal);
861 tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE;
863 /* mt-exec, de_thread() is waiting for group leader */
864 if (unlikely(tsk->signal->notify_count < 0))
865 wake_up_process(tsk->signal->group_exit_task);
866 write_unlock_irq(&tasklist_lock);
868 /* If the process is dead, release it - nobody will wait for it */
873 #ifdef CONFIG_DEBUG_STACK_USAGE
874 static void check_stack_usage(void)
876 static DEFINE_SPINLOCK(low_water_lock);
877 static int lowest_to_date = THREAD_SIZE;
880 free = stack_not_used(current);
882 if (free >= lowest_to_date)
885 spin_lock(&low_water_lock);
886 if (free < lowest_to_date) {
887 printk(KERN_WARNING "%s (%d) used greatest stack depth: "
889 current->comm, task_pid_nr(current), free);
890 lowest_to_date = free;
892 spin_unlock(&low_water_lock);
895 static inline void check_stack_usage(void) {}
898 void do_exit(long code)
900 struct task_struct *tsk = current;
903 profile_task_exit(tsk);
905 WARN_ON(blk_needs_flush_plug(tsk));
907 if (unlikely(in_interrupt()))
908 panic("Aiee, killing interrupt handler!");
909 if (unlikely(!tsk->pid))
910 panic("Attempted to kill the idle task!");
913 * If do_exit is called because this processes oopsed, it's possible
914 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
915 * continuing. Amongst other possible reasons, this is to prevent
916 * mm_release()->clear_child_tid() from writing to a user-controlled
921 ptrace_event(PTRACE_EVENT_EXIT, code);
923 validate_creds_for_do_exit(tsk);
926 * We're taking recursive faults here in do_exit. Safest is to just
927 * leave this task alone and wait for reboot.
929 if (unlikely(tsk->flags & PF_EXITING)) {
931 "Fixing recursive fault but reboot is needed!\n");
933 * We can do this unlocked here. The futex code uses
934 * this flag just to verify whether the pi state
935 * cleanup has been done or not. In the worst case it
936 * loops once more. We pretend that the cleanup was
937 * done as there is no way to return. Either the
938 * OWNER_DIED bit is set by now or we push the blocked
939 * task into the wait for ever nirwana as well.
941 tsk->flags |= PF_EXITPIDONE;
942 set_current_state(TASK_UNINTERRUPTIBLE);
946 exit_signals(tsk); /* sets PF_EXITING */
948 * tsk->flags are checked in the futex code to protect against
949 * an exiting task cleaning up the robust pi futexes, and in
950 * task_work_add() to avoid the race with exit_task_work().
953 raw_spin_unlock_wait(&tsk->pi_lock);
957 if (unlikely(in_atomic()))
958 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
959 current->comm, task_pid_nr(current),
962 acct_update_integrals(tsk);
964 /* Set exit_code before complete_vfork_done() in mm_release() */
965 tsk->exit_code = code;
967 /* Release mm and sync mm's RSS info before statistics gathering */
968 mm_release(tsk, tsk->mm);
970 group_dead = atomic_dec_and_test(&tsk->signal->live);
972 hrtimer_cancel(&tsk->signal->real_timer);
973 exit_itimers(tsk->signal);
975 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
977 acct_collect(code, group_dead);
982 taskstats_exit(tsk, group_dead);
988 trace_sched_process_exit(tsk);
998 * Flush inherited counters to the parent - before the parent
999 * gets woken up by child-exit notifications.
1001 * because of cgroup mode, must be called before cgroup_exit()
1003 perf_event_exit_task(tsk);
1005 cgroup_exit(tsk, 1);
1008 disassociate_ctty(1);
1010 module_put(task_thread_info(tsk)->exec_domain->module);
1012 proc_exit_connector(tsk);
1015 * FIXME: do that only when needed, using sched_exit tracepoint
1017 ptrace_put_breakpoints(tsk);
1019 exit_notify(tsk, group_dead);
1022 mpol_put(tsk->mempolicy);
1023 tsk->mempolicy = NULL;
1027 if (unlikely(current->pi_state_cache))
1028 kfree(current->pi_state_cache);
1031 * Make sure we are holding no locks:
1033 debug_check_no_locks_held(tsk);
1035 * We can do this unlocked here. The futex code uses this flag
1036 * just to verify whether the pi state cleanup has been done
1037 * or not. In the worst case it loops once more.
1039 tsk->flags |= PF_EXITPIDONE;
1041 if (tsk->io_context)
1042 exit_io_context(tsk);
1044 if (tsk->splice_pipe)
1045 __free_pipe_info(tsk->splice_pipe);
1047 validate_creds_for_do_exit(tsk);
1050 if (tsk->nr_dirtied)
1051 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
1055 * The setting of TASK_RUNNING by try_to_wake_up() may be delayed
1056 * when the following two conditions become true.
1057 * - There is race condition of mmap_sem (It is acquired by
1059 * - SMI occurs before setting TASK_RUNINNG.
1060 * (or hypervisor of virtual machine switches to other guest)
1061 * As a result, we may become TASK_RUNNING after becoming TASK_DEAD
1063 * To avoid it, we have to wait for releasing tsk->pi_lock which
1064 * is held by try_to_wake_up()
1067 raw_spin_unlock_wait(&tsk->pi_lock);
1069 /* causes final put_task_struct in finish_task_switch(). */
1070 tsk->state = TASK_DEAD;
1071 tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */
1074 /* Avoid "noreturn function does return". */
1076 cpu_relax(); /* For when BUG is null */
1079 EXPORT_SYMBOL_GPL(do_exit);
1081 void complete_and_exit(struct completion *comp, long code)
1089 EXPORT_SYMBOL(complete_and_exit);
1091 SYSCALL_DEFINE1(exit, int, error_code)
1093 do_exit((error_code&0xff)<<8);
1097 * Take down every thread in the group. This is called by fatal signals
1098 * as well as by sys_exit_group (below).
1101 do_group_exit(int exit_code)
1103 struct signal_struct *sig = current->signal;
1105 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1107 if (signal_group_exit(sig))
1108 exit_code = sig->group_exit_code;
1109 else if (!thread_group_empty(current)) {
1110 struct sighand_struct *const sighand = current->sighand;
1111 spin_lock_irq(&sighand->siglock);
1112 if (signal_group_exit(sig))
1113 /* Another thread got here before we took the lock. */
1114 exit_code = sig->group_exit_code;
1116 sig->group_exit_code = exit_code;
1117 sig->flags = SIGNAL_GROUP_EXIT;
1118 zap_other_threads(current);
1120 spin_unlock_irq(&sighand->siglock);
1128 * this kills every thread in the thread group. Note that any externally
1129 * wait4()-ing process will get the correct exit code - even if this
1130 * thread is not the thread group leader.
1132 SYSCALL_DEFINE1(exit_group, int, error_code)
1134 do_group_exit((error_code & 0xff) << 8);
1140 enum pid_type wo_type;
1144 struct siginfo __user *wo_info;
1145 int __user *wo_stat;
1146 struct rusage __user *wo_rusage;
1148 wait_queue_t child_wait;
1153 struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1155 if (type != PIDTYPE_PID)
1156 task = task->group_leader;
1157 return task->pids[type].pid;
1160 static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1162 return wo->wo_type == PIDTYPE_MAX ||
1163 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1166 static int eligible_child(struct wait_opts *wo, struct task_struct *p)
1168 if (!eligible_pid(wo, p))
1170 /* Wait for all children (clone and not) if __WALL is set;
1171 * otherwise, wait for clone children *only* if __WCLONE is
1172 * set; otherwise, wait for non-clone children *only*. (Note:
1173 * A "clone" child here is one that reports to its parent
1174 * using a signal other than SIGCHLD.) */
1175 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1176 && !(wo->wo_flags & __WALL))
1182 static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1183 pid_t pid, uid_t uid, int why, int status)
1185 struct siginfo __user *infop;
1186 int retval = wo->wo_rusage
1187 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1190 infop = wo->wo_info;
1193 retval = put_user(SIGCHLD, &infop->si_signo);
1195 retval = put_user(0, &infop->si_errno);
1197 retval = put_user((short)why, &infop->si_code);
1199 retval = put_user(pid, &infop->si_pid);
1201 retval = put_user(uid, &infop->si_uid);
1203 retval = put_user(status, &infop->si_status);
1211 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1212 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1213 * the lock and this task is uninteresting. If we return nonzero, we have
1214 * released the lock and the system call should return.
1216 static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1218 unsigned long state;
1219 int retval, status, traced;
1220 pid_t pid = task_pid_vnr(p);
1221 uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p));
1222 struct siginfo __user *infop;
1224 if (!likely(wo->wo_flags & WEXITED))
1227 if (unlikely(wo->wo_flags & WNOWAIT)) {
1228 int exit_code = p->exit_code;
1232 read_unlock(&tasklist_lock);
1233 if ((exit_code & 0x7f) == 0) {
1235 status = exit_code >> 8;
1237 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1238 status = exit_code & 0x7f;
1240 return wait_noreap_copyout(wo, p, pid, uid, why, status);
1244 * Try to move the task's state to DEAD
1245 * only one thread is allowed to do this:
1247 state = xchg(&p->exit_state, EXIT_DEAD);
1248 if (state != EXIT_ZOMBIE) {
1249 BUG_ON(state != EXIT_DEAD);
1253 traced = ptrace_reparented(p);
1255 * It can be ptraced but not reparented, check
1256 * thread_group_leader() to filter out sub-threads.
1258 if (likely(!traced) && thread_group_leader(p)) {
1259 struct signal_struct *psig;
1260 struct signal_struct *sig;
1261 unsigned long maxrss;
1262 cputime_t tgutime, tgstime;
1265 * The resource counters for the group leader are in its
1266 * own task_struct. Those for dead threads in the group
1267 * are in its signal_struct, as are those for the child
1268 * processes it has previously reaped. All these
1269 * accumulate in the parent's signal_struct c* fields.
1271 * We don't bother to take a lock here to protect these
1272 * p->signal fields, because they are only touched by
1273 * __exit_signal, which runs with tasklist_lock
1274 * write-locked anyway, and so is excluded here. We do
1275 * need to protect the access to parent->signal fields,
1276 * as other threads in the parent group can be right
1277 * here reaping other children at the same time.
1279 * We use thread_group_times() to get times for the thread
1280 * group, which consolidates times for all threads in the
1281 * group including the group leader.
1283 thread_group_times(p, &tgutime, &tgstime);
1284 spin_lock_irq(&p->real_parent->sighand->siglock);
1285 psig = p->real_parent->signal;
1287 psig->cutime += tgutime + sig->cutime;
1288 psig->cstime += tgstime + sig->cstime;
1289 psig->cgtime += p->gtime + sig->gtime + sig->cgtime;
1291 p->min_flt + sig->min_flt + sig->cmin_flt;
1293 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1295 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1297 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1299 task_io_get_inblock(p) +
1300 sig->inblock + sig->cinblock;
1302 task_io_get_oublock(p) +
1303 sig->oublock + sig->coublock;
1304 maxrss = max(sig->maxrss, sig->cmaxrss);
1305 if (psig->cmaxrss < maxrss)
1306 psig->cmaxrss = maxrss;
1307 task_io_accounting_add(&psig->ioac, &p->ioac);
1308 task_io_accounting_add(&psig->ioac, &sig->ioac);
1309 spin_unlock_irq(&p->real_parent->sighand->siglock);
1313 * Now we are sure this task is interesting, and no other
1314 * thread can reap it because we set its state to EXIT_DEAD.
1316 read_unlock(&tasklist_lock);
1318 retval = wo->wo_rusage
1319 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1320 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1321 ? p->signal->group_exit_code : p->exit_code;
1322 if (!retval && wo->wo_stat)
1323 retval = put_user(status, wo->wo_stat);
1325 infop = wo->wo_info;
1326 if (!retval && infop)
1327 retval = put_user(SIGCHLD, &infop->si_signo);
1328 if (!retval && infop)
1329 retval = put_user(0, &infop->si_errno);
1330 if (!retval && infop) {
1333 if ((status & 0x7f) == 0) {
1337 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1340 retval = put_user((short)why, &infop->si_code);
1342 retval = put_user(status, &infop->si_status);
1344 if (!retval && infop)
1345 retval = put_user(pid, &infop->si_pid);
1346 if (!retval && infop)
1347 retval = put_user(uid, &infop->si_uid);
1352 write_lock_irq(&tasklist_lock);
1353 /* We dropped tasklist, ptracer could die and untrace */
1356 * If this is not a sub-thread, notify the parent.
1357 * If parent wants a zombie, don't release it now.
1359 if (thread_group_leader(p) &&
1360 !do_notify_parent(p, p->exit_signal)) {
1361 p->exit_state = EXIT_ZOMBIE;
1364 write_unlock_irq(&tasklist_lock);
1372 static int *task_stopped_code(struct task_struct *p, bool ptrace)
1375 if (task_is_stopped_or_traced(p) &&
1376 !(p->jobctl & JOBCTL_LISTENING))
1377 return &p->exit_code;
1379 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1380 return &p->signal->group_exit_code;
1386 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1388 * @ptrace: is the wait for ptrace
1389 * @p: task to wait for
1391 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1394 * read_lock(&tasklist_lock), which is released if return value is
1395 * non-zero. Also, grabs and releases @p->sighand->siglock.
1398 * 0 if wait condition didn't exist and search for other wait conditions
1399 * should continue. Non-zero return, -errno on failure and @p's pid on
1400 * success, implies that tasklist_lock is released and wait condition
1401 * search should terminate.
1403 static int wait_task_stopped(struct wait_opts *wo,
1404 int ptrace, struct task_struct *p)
1406 struct siginfo __user *infop;
1407 int retval, exit_code, *p_code, why;
1408 uid_t uid = 0; /* unneeded, required by compiler */
1412 * Traditionally we see ptrace'd stopped tasks regardless of options.
1414 if (!ptrace && !(wo->wo_flags & WUNTRACED))
1417 if (!task_stopped_code(p, ptrace))
1421 spin_lock_irq(&p->sighand->siglock);
1423 p_code = task_stopped_code(p, ptrace);
1424 if (unlikely(!p_code))
1427 exit_code = *p_code;
1431 if (!unlikely(wo->wo_flags & WNOWAIT))
1434 uid = from_kuid_munged(current_user_ns(), task_uid(p));
1436 spin_unlock_irq(&p->sighand->siglock);
1441 * Now we are pretty sure this task is interesting.
1442 * Make sure it doesn't get reaped out from under us while we
1443 * give up the lock and then examine it below. We don't want to
1444 * keep holding onto the tasklist_lock while we call getrusage and
1445 * possibly take page faults for user memory.
1448 pid = task_pid_vnr(p);
1449 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1450 read_unlock(&tasklist_lock);
1452 if (unlikely(wo->wo_flags & WNOWAIT))
1453 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
1455 retval = wo->wo_rusage
1456 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1457 if (!retval && wo->wo_stat)
1458 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1460 infop = wo->wo_info;
1461 if (!retval && infop)
1462 retval = put_user(SIGCHLD, &infop->si_signo);
1463 if (!retval && infop)
1464 retval = put_user(0, &infop->si_errno);
1465 if (!retval && infop)
1466 retval = put_user((short)why, &infop->si_code);
1467 if (!retval && infop)
1468 retval = put_user(exit_code, &infop->si_status);
1469 if (!retval && infop)
1470 retval = put_user(pid, &infop->si_pid);
1471 if (!retval && infop)
1472 retval = put_user(uid, &infop->si_uid);
1482 * Handle do_wait work for one task in a live, non-stopped state.
1483 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1484 * the lock and this task is uninteresting. If we return nonzero, we have
1485 * released the lock and the system call should return.
1487 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1493 if (!unlikely(wo->wo_flags & WCONTINUED))
1496 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1499 spin_lock_irq(&p->sighand->siglock);
1500 /* Re-check with the lock held. */
1501 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1502 spin_unlock_irq(&p->sighand->siglock);
1505 if (!unlikely(wo->wo_flags & WNOWAIT))
1506 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1507 uid = from_kuid_munged(current_user_ns(), task_uid(p));
1508 spin_unlock_irq(&p->sighand->siglock);
1510 pid = task_pid_vnr(p);
1512 read_unlock(&tasklist_lock);
1515 retval = wo->wo_rusage
1516 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1518 if (!retval && wo->wo_stat)
1519 retval = put_user(0xffff, wo->wo_stat);
1523 retval = wait_noreap_copyout(wo, p, pid, uid,
1524 CLD_CONTINUED, SIGCONT);
1525 BUG_ON(retval == 0);
1532 * Consider @p for a wait by @parent.
1534 * -ECHILD should be in ->notask_error before the first call.
1535 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1536 * Returns zero if the search for a child should continue;
1537 * then ->notask_error is 0 if @p is an eligible child,
1538 * or another error from security_task_wait(), or still -ECHILD.
1540 static int wait_consider_task(struct wait_opts *wo, int ptrace,
1541 struct task_struct *p)
1543 int ret = eligible_child(wo, p);
1547 ret = security_task_wait(p);
1548 if (unlikely(ret < 0)) {
1550 * If we have not yet seen any eligible child,
1551 * then let this error code replace -ECHILD.
1552 * A permission error will give the user a clue
1553 * to look for security policy problems, rather
1554 * than for mysterious wait bugs.
1556 if (wo->notask_error)
1557 wo->notask_error = ret;
1561 /* dead body doesn't have much to contribute */
1562 if (unlikely(p->exit_state == EXIT_DEAD)) {
1564 * But do not ignore this task until the tracer does
1565 * wait_task_zombie()->do_notify_parent().
1567 if (likely(!ptrace) && unlikely(ptrace_reparented(p)))
1568 wo->notask_error = 0;
1573 if (p->exit_state == EXIT_ZOMBIE) {
1575 * A zombie ptracee is only visible to its ptracer.
1576 * Notification and reaping will be cascaded to the real
1577 * parent when the ptracer detaches.
1579 if (likely(!ptrace) && unlikely(p->ptrace)) {
1580 /* it will become visible, clear notask_error */
1581 wo->notask_error = 0;
1585 /* we don't reap group leaders with subthreads */
1586 if (!delay_group_leader(p))
1587 return wait_task_zombie(wo, p);
1590 * Allow access to stopped/continued state via zombie by
1591 * falling through. Clearing of notask_error is complex.
1595 * If WEXITED is set, notask_error should naturally be
1596 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1597 * so, if there are live subthreads, there are events to
1598 * wait for. If all subthreads are dead, it's still safe
1599 * to clear - this function will be called again in finite
1600 * amount time once all the subthreads are released and
1601 * will then return without clearing.
1605 * Stopped state is per-task and thus can't change once the
1606 * target task dies. Only continued and exited can happen.
1607 * Clear notask_error if WCONTINUED | WEXITED.
1609 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1610 wo->notask_error = 0;
1613 * If @p is ptraced by a task in its real parent's group,
1614 * hide group stop/continued state when looking at @p as
1615 * the real parent; otherwise, a single stop can be
1616 * reported twice as group and ptrace stops.
1618 * If a ptracer wants to distinguish the two events for its
1619 * own children, it should create a separate process which
1620 * takes the role of real parent.
1622 if (likely(!ptrace) && p->ptrace && !ptrace_reparented(p))
1626 * @p is alive and it's gonna stop, continue or exit, so
1627 * there always is something to wait for.
1629 wo->notask_error = 0;
1633 * Wait for stopped. Depending on @ptrace, different stopped state
1634 * is used and the two don't interact with each other.
1636 ret = wait_task_stopped(wo, ptrace, p);
1641 * Wait for continued. There's only one continued state and the
1642 * ptracer can consume it which can confuse the real parent. Don't
1643 * use WCONTINUED from ptracer. You don't need or want it.
1645 return wait_task_continued(wo, p);
1649 * Do the work of do_wait() for one thread in the group, @tsk.
1651 * -ECHILD should be in ->notask_error before the first call.
1652 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1653 * Returns zero if the search for a child should continue; then
1654 * ->notask_error is 0 if there were any eligible children,
1655 * or another error from security_task_wait(), or still -ECHILD.
1657 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1659 struct task_struct *p;
1661 list_for_each_entry(p, &tsk->children, sibling) {
1662 int ret = wait_consider_task(wo, 0, p);
1670 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1672 struct task_struct *p;
1674 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1675 int ret = wait_consider_task(wo, 1, p);
1683 static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1684 int sync, void *key)
1686 struct wait_opts *wo = container_of(wait, struct wait_opts,
1688 struct task_struct *p = key;
1690 if (!eligible_pid(wo, p))
1693 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1696 return default_wake_function(wait, mode, sync, key);
1699 void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1701 __wake_up_sync_key(&parent->signal->wait_chldexit,
1702 TASK_INTERRUPTIBLE, 1, p);
1705 static long do_wait(struct wait_opts *wo)
1707 struct task_struct *tsk;
1710 trace_sched_process_wait(wo->wo_pid);
1712 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1713 wo->child_wait.private = current;
1714 add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
1717 * If there is nothing that can match our critiera just get out.
1718 * We will clear ->notask_error to zero if we see any child that
1719 * might later match our criteria, even if we are not able to reap
1722 wo->notask_error = -ECHILD;
1723 if ((wo->wo_type < PIDTYPE_MAX) &&
1724 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
1727 set_current_state(TASK_INTERRUPTIBLE);
1728 read_lock(&tasklist_lock);
1731 retval = do_wait_thread(wo, tsk);
1735 retval = ptrace_do_wait(wo, tsk);
1739 if (wo->wo_flags & __WNOTHREAD)
1741 } while_each_thread(current, tsk);
1742 read_unlock(&tasklist_lock);
1745 retval = wo->notask_error;
1746 if (!retval && !(wo->wo_flags & WNOHANG)) {
1747 retval = -ERESTARTSYS;
1748 if (!signal_pending(current)) {
1754 __set_current_state(TASK_RUNNING);
1755 remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
1759 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1760 infop, int, options, struct rusage __user *, ru)
1762 struct wait_opts wo;
1763 struct pid *pid = NULL;
1767 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1769 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1782 type = PIDTYPE_PGID;
1790 if (type < PIDTYPE_MAX)
1791 pid = find_get_pid(upid);
1795 wo.wo_flags = options;
1805 * For a WNOHANG return, clear out all the fields
1806 * we would set so the user can easily tell the
1810 ret = put_user(0, &infop->si_signo);
1812 ret = put_user(0, &infop->si_errno);
1814 ret = put_user(0, &infop->si_code);
1816 ret = put_user(0, &infop->si_pid);
1818 ret = put_user(0, &infop->si_uid);
1820 ret = put_user(0, &infop->si_status);
1825 /* avoid REGPARM breakage on x86: */
1826 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
1830 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1831 int, options, struct rusage __user *, ru)
1833 struct wait_opts wo;
1834 struct pid *pid = NULL;
1838 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1839 __WNOTHREAD|__WCLONE|__WALL))
1844 else if (upid < 0) {
1845 type = PIDTYPE_PGID;
1846 pid = find_get_pid(-upid);
1847 } else if (upid == 0) {
1848 type = PIDTYPE_PGID;
1849 pid = get_task_pid(current, PIDTYPE_PGID);
1850 } else /* upid > 0 */ {
1852 pid = find_get_pid(upid);
1857 wo.wo_flags = options | WEXITED;
1859 wo.wo_stat = stat_addr;
1864 /* avoid REGPARM breakage on x86: */
1865 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
1869 #ifdef __ARCH_WANT_SYS_WAITPID
1872 * sys_waitpid() remains for compatibility. waitpid() should be
1873 * implemented by calling sys_wait4() from libc.a.
1875 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1877 return sys_wait4(pid, stat_addr, options, NULL);