xref: /openbmc/linux/kernel/exit.c (revision e868d61272caa648214046a096e5a6bfc068dc8c)
1 /*
2  *  linux/kernel/exit.c
3  *
4  *  Copyright (C) 1991, 1992  Linus Torvalds
5  */
6 
7 #include <linux/mm.h>
8 #include <linux/slab.h>
9 #include <linux/interrupt.h>
10 #include <linux/module.h>
11 #include <linux/capability.h>
12 #include <linux/completion.h>
13 #include <linux/personality.h>
14 #include <linux/tty.h>
15 #include <linux/mnt_namespace.h>
16 #include <linux/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/binfmts.h>
23 #include <linux/nsproxy.h>
24 #include <linux/pid_namespace.h>
25 #include <linux/ptrace.h>
26 #include <linux/profile.h>
27 #include <linux/signalfd.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/cpuset.h>
35 #include <linux/syscalls.h>
36 #include <linux/signal.h>
37 #include <linux/posix-timers.h>
38 #include <linux/cn_proc.h>
39 #include <linux/mutex.h>
40 #include <linux/futex.h>
41 #include <linux/compat.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 
48 #include <asm/uaccess.h>
49 #include <asm/unistd.h>
50 #include <asm/pgtable.h>
51 #include <asm/mmu_context.h>
52 
53 extern void sem_exit (void);
54 
55 static void exit_mm(struct task_struct * tsk);
56 
57 static void __unhash_process(struct task_struct *p)
58 {
59 	nr_threads--;
60 	detach_pid(p, PIDTYPE_PID);
61 	if (thread_group_leader(p)) {
62 		detach_pid(p, PIDTYPE_PGID);
63 		detach_pid(p, PIDTYPE_SID);
64 
65 		list_del_rcu(&p->tasks);
66 		__get_cpu_var(process_counts)--;
67 	}
68 	list_del_rcu(&p->thread_group);
69 	remove_parent(p);
70 }
71 
72 /*
73  * This function expects the tasklist_lock write-locked.
74  */
75 static void __exit_signal(struct task_struct *tsk)
76 {
77 	struct signal_struct *sig = tsk->signal;
78 	struct sighand_struct *sighand;
79 
80 	BUG_ON(!sig);
81 	BUG_ON(!atomic_read(&sig->count));
82 
83 	rcu_read_lock();
84 	sighand = rcu_dereference(tsk->sighand);
85 	spin_lock(&sighand->siglock);
86 
87 	/*
88 	 * Notify that this sighand has been detached. This must
89 	 * be called with the tsk->sighand lock held. Also, this
90 	 * access tsk->sighand internally, so it must be called
91 	 * before tsk->sighand is reset.
92 	 */
93 	signalfd_detach_locked(tsk);
94 
95 	posix_cpu_timers_exit(tsk);
96 	if (atomic_dec_and_test(&sig->count))
97 		posix_cpu_timers_exit_group(tsk);
98 	else {
99 		/*
100 		 * If there is any task waiting for the group exit
101 		 * then notify it:
102 		 */
103 		if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
104 			wake_up_process(sig->group_exit_task);
105 			sig->group_exit_task = NULL;
106 		}
107 		if (tsk == sig->curr_target)
108 			sig->curr_target = next_thread(tsk);
109 		/*
110 		 * Accumulate here the counters for all threads but the
111 		 * group leader as they die, so they can be added into
112 		 * the process-wide totals when those are taken.
113 		 * The group leader stays around as a zombie as long
114 		 * as there are other threads.  When it gets reaped,
115 		 * the exit.c code will add its counts into these totals.
116 		 * We won't ever get here for the group leader, since it
117 		 * will have been the last reference on the signal_struct.
118 		 */
119 		sig->utime = cputime_add(sig->utime, tsk->utime);
120 		sig->stime = cputime_add(sig->stime, tsk->stime);
121 		sig->min_flt += tsk->min_flt;
122 		sig->maj_flt += tsk->maj_flt;
123 		sig->nvcsw += tsk->nvcsw;
124 		sig->nivcsw += tsk->nivcsw;
125 		sig->sched_time += tsk->sched_time;
126 		sig->inblock += task_io_get_inblock(tsk);
127 		sig->oublock += task_io_get_oublock(tsk);
128 		sig = NULL; /* Marker for below. */
129 	}
130 
131 	__unhash_process(tsk);
132 
133 	tsk->signal = NULL;
134 	tsk->sighand = NULL;
135 	spin_unlock(&sighand->siglock);
136 	rcu_read_unlock();
137 
138 	__cleanup_sighand(sighand);
139 	clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
140 	flush_sigqueue(&tsk->pending);
141 	if (sig) {
142 		flush_sigqueue(&sig->shared_pending);
143 		taskstats_tgid_free(sig);
144 		__cleanup_signal(sig);
145 	}
146 }
147 
148 static void delayed_put_task_struct(struct rcu_head *rhp)
149 {
150 	put_task_struct(container_of(rhp, struct task_struct, rcu));
151 }
152 
153 void release_task(struct task_struct * p)
154 {
155 	struct task_struct *leader;
156 	int zap_leader;
157 repeat:
158 	atomic_dec(&p->user->processes);
159 	write_lock_irq(&tasklist_lock);
160 	ptrace_unlink(p);
161 	BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
162 	__exit_signal(p);
163 
164 	/*
165 	 * If we are the last non-leader member of the thread
166 	 * group, and the leader is zombie, then notify the
167 	 * group leader's parent process. (if it wants notification.)
168 	 */
169 	zap_leader = 0;
170 	leader = p->group_leader;
171 	if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
172 		BUG_ON(leader->exit_signal == -1);
173 		do_notify_parent(leader, leader->exit_signal);
174 		/*
175 		 * If we were the last child thread and the leader has
176 		 * exited already, and the leader's parent ignores SIGCHLD,
177 		 * then we are the one who should release the leader.
178 		 *
179 		 * do_notify_parent() will have marked it self-reaping in
180 		 * that case.
181 		 */
182 		zap_leader = (leader->exit_signal == -1);
183 	}
184 
185 	sched_exit(p);
186 	write_unlock_irq(&tasklist_lock);
187 	proc_flush_task(p);
188 	release_thread(p);
189 	call_rcu(&p->rcu, delayed_put_task_struct);
190 
191 	p = leader;
192 	if (unlikely(zap_leader))
193 		goto repeat;
194 }
195 
196 /*
197  * This checks not only the pgrp, but falls back on the pid if no
198  * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
199  * without this...
200  *
201  * The caller must hold rcu lock or the tasklist lock.
202  */
203 struct pid *session_of_pgrp(struct pid *pgrp)
204 {
205 	struct task_struct *p;
206 	struct pid *sid = NULL;
207 
208 	p = pid_task(pgrp, PIDTYPE_PGID);
209 	if (p == NULL)
210 		p = pid_task(pgrp, PIDTYPE_PID);
211 	if (p != NULL)
212 		sid = task_session(p);
213 
214 	return sid;
215 }
216 
217 /*
218  * Determine if a process group is "orphaned", according to the POSIX
219  * definition in 2.2.2.52.  Orphaned process groups are not to be affected
220  * by terminal-generated stop signals.  Newly orphaned process groups are
221  * to receive a SIGHUP and a SIGCONT.
222  *
223  * "I ask you, have you ever known what it is to be an orphan?"
224  */
225 static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
226 {
227 	struct task_struct *p;
228 	int ret = 1;
229 
230 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
231 		if (p == ignored_task
232 				|| p->exit_state
233 				|| is_init(p->real_parent))
234 			continue;
235 		if (task_pgrp(p->real_parent) != pgrp &&
236 		    task_session(p->real_parent) == task_session(p)) {
237 			ret = 0;
238 			break;
239 		}
240 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
241 	return ret;	/* (sighing) "Often!" */
242 }
243 
244 int is_current_pgrp_orphaned(void)
245 {
246 	int retval;
247 
248 	read_lock(&tasklist_lock);
249 	retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
250 	read_unlock(&tasklist_lock);
251 
252 	return retval;
253 }
254 
255 static int has_stopped_jobs(struct pid *pgrp)
256 {
257 	int retval = 0;
258 	struct task_struct *p;
259 
260 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
261 		if (p->state != TASK_STOPPED)
262 			continue;
263 		retval = 1;
264 		break;
265 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
266 	return retval;
267 }
268 
269 /**
270  * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
271  *
272  * If a kernel thread is launched as a result of a system call, or if
273  * it ever exits, it should generally reparent itself to kthreadd so it
274  * isn't in the way of other processes and is correctly cleaned up on exit.
275  *
276  * The various task state such as scheduling policy and priority may have
277  * been inherited from a user process, so we reset them to sane values here.
278  *
279  * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
280  */
281 static void reparent_to_kthreadd(void)
282 {
283 	write_lock_irq(&tasklist_lock);
284 
285 	ptrace_unlink(current);
286 	/* Reparent to init */
287 	remove_parent(current);
288 	current->real_parent = current->parent = kthreadd_task;
289 	add_parent(current);
290 
291 	/* Set the exit signal to SIGCHLD so we signal init on exit */
292 	current->exit_signal = SIGCHLD;
293 
294 	if (!has_rt_policy(current) && (task_nice(current) < 0))
295 		set_user_nice(current, 0);
296 	/* cpus_allowed? */
297 	/* rt_priority? */
298 	/* signals? */
299 	security_task_reparent_to_init(current);
300 	memcpy(current->signal->rlim, init_task.signal->rlim,
301 	       sizeof(current->signal->rlim));
302 	atomic_inc(&(INIT_USER->__count));
303 	write_unlock_irq(&tasklist_lock);
304 	switch_uid(INIT_USER);
305 }
306 
307 void __set_special_pids(pid_t session, pid_t pgrp)
308 {
309 	struct task_struct *curr = current->group_leader;
310 
311 	if (process_session(curr) != session) {
312 		detach_pid(curr, PIDTYPE_SID);
313 		set_signal_session(curr->signal, session);
314 		attach_pid(curr, PIDTYPE_SID, find_pid(session));
315 	}
316 	if (process_group(curr) != pgrp) {
317 		detach_pid(curr, PIDTYPE_PGID);
318 		curr->signal->pgrp = pgrp;
319 		attach_pid(curr, PIDTYPE_PGID, find_pid(pgrp));
320 	}
321 }
322 
323 static void set_special_pids(pid_t session, pid_t pgrp)
324 {
325 	write_lock_irq(&tasklist_lock);
326 	__set_special_pids(session, pgrp);
327 	write_unlock_irq(&tasklist_lock);
328 }
329 
330 /*
331  * Let kernel threads use this to say that they
332  * allow a certain signal (since daemonize() will
333  * have disabled all of them by default).
334  */
335 int allow_signal(int sig)
336 {
337 	if (!valid_signal(sig) || sig < 1)
338 		return -EINVAL;
339 
340 	spin_lock_irq(&current->sighand->siglock);
341 	sigdelset(&current->blocked, sig);
342 	if (!current->mm) {
343 		/* Kernel threads handle their own signals.
344 		   Let the signal code know it'll be handled, so
345 		   that they don't get converted to SIGKILL or
346 		   just silently dropped */
347 		current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
348 	}
349 	recalc_sigpending();
350 	spin_unlock_irq(&current->sighand->siglock);
351 	return 0;
352 }
353 
354 EXPORT_SYMBOL(allow_signal);
355 
356 int disallow_signal(int sig)
357 {
358 	if (!valid_signal(sig) || sig < 1)
359 		return -EINVAL;
360 
361 	spin_lock_irq(&current->sighand->siglock);
362 	current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
363 	recalc_sigpending();
364 	spin_unlock_irq(&current->sighand->siglock);
365 	return 0;
366 }
367 
368 EXPORT_SYMBOL(disallow_signal);
369 
370 /*
371  *	Put all the gunge required to become a kernel thread without
372  *	attached user resources in one place where it belongs.
373  */
374 
375 void daemonize(const char *name, ...)
376 {
377 	va_list args;
378 	struct fs_struct *fs;
379 	sigset_t blocked;
380 
381 	va_start(args, name);
382 	vsnprintf(current->comm, sizeof(current->comm), name, args);
383 	va_end(args);
384 
385 	/*
386 	 * If we were started as result of loading a module, close all of the
387 	 * user space pages.  We don't need them, and if we didn't close them
388 	 * they would be locked into memory.
389 	 */
390 	exit_mm(current);
391 
392 	set_special_pids(1, 1);
393 	proc_clear_tty(current);
394 
395 	/* Block and flush all signals */
396 	sigfillset(&blocked);
397 	sigprocmask(SIG_BLOCK, &blocked, NULL);
398 	flush_signals(current);
399 
400 	/* Become as one with the init task */
401 
402 	exit_fs(current);	/* current->fs->count--; */
403 	fs = init_task.fs;
404 	current->fs = fs;
405 	atomic_inc(&fs->count);
406 
407 	exit_task_namespaces(current);
408 	current->nsproxy = init_task.nsproxy;
409 	get_task_namespaces(current);
410 
411  	exit_files(current);
412 	current->files = init_task.files;
413 	atomic_inc(&current->files->count);
414 
415 	reparent_to_kthreadd();
416 }
417 
418 EXPORT_SYMBOL(daemonize);
419 
420 static void close_files(struct files_struct * files)
421 {
422 	int i, j;
423 	struct fdtable *fdt;
424 
425 	j = 0;
426 
427 	/*
428 	 * It is safe to dereference the fd table without RCU or
429 	 * ->file_lock because this is the last reference to the
430 	 * files structure.
431 	 */
432 	fdt = files_fdtable(files);
433 	for (;;) {
434 		unsigned long set;
435 		i = j * __NFDBITS;
436 		if (i >= fdt->max_fds)
437 			break;
438 		set = fdt->open_fds->fds_bits[j++];
439 		while (set) {
440 			if (set & 1) {
441 				struct file * file = xchg(&fdt->fd[i], NULL);
442 				if (file) {
443 					filp_close(file, files);
444 					cond_resched();
445 				}
446 			}
447 			i++;
448 			set >>= 1;
449 		}
450 	}
451 }
452 
453 struct files_struct *get_files_struct(struct task_struct *task)
454 {
455 	struct files_struct *files;
456 
457 	task_lock(task);
458 	files = task->files;
459 	if (files)
460 		atomic_inc(&files->count);
461 	task_unlock(task);
462 
463 	return files;
464 }
465 
466 void fastcall put_files_struct(struct files_struct *files)
467 {
468 	struct fdtable *fdt;
469 
470 	if (atomic_dec_and_test(&files->count)) {
471 		close_files(files);
472 		/*
473 		 * Free the fd and fdset arrays if we expanded them.
474 		 * If the fdtable was embedded, pass files for freeing
475 		 * at the end of the RCU grace period. Otherwise,
476 		 * you can free files immediately.
477 		 */
478 		fdt = files_fdtable(files);
479 		if (fdt != &files->fdtab)
480 			kmem_cache_free(files_cachep, files);
481 		free_fdtable(fdt);
482 	}
483 }
484 
485 EXPORT_SYMBOL(put_files_struct);
486 
487 void reset_files_struct(struct task_struct *tsk, struct files_struct *files)
488 {
489 	struct files_struct *old;
490 
491 	old = tsk->files;
492 	task_lock(tsk);
493 	tsk->files = files;
494 	task_unlock(tsk);
495 	put_files_struct(old);
496 }
497 EXPORT_SYMBOL(reset_files_struct);
498 
499 static inline void __exit_files(struct task_struct *tsk)
500 {
501 	struct files_struct * files = tsk->files;
502 
503 	if (files) {
504 		task_lock(tsk);
505 		tsk->files = NULL;
506 		task_unlock(tsk);
507 		put_files_struct(files);
508 	}
509 }
510 
511 void exit_files(struct task_struct *tsk)
512 {
513 	__exit_files(tsk);
514 }
515 
516 static inline void __put_fs_struct(struct fs_struct *fs)
517 {
518 	/* No need to hold fs->lock if we are killing it */
519 	if (atomic_dec_and_test(&fs->count)) {
520 		dput(fs->root);
521 		mntput(fs->rootmnt);
522 		dput(fs->pwd);
523 		mntput(fs->pwdmnt);
524 		if (fs->altroot) {
525 			dput(fs->altroot);
526 			mntput(fs->altrootmnt);
527 		}
528 		kmem_cache_free(fs_cachep, fs);
529 	}
530 }
531 
532 void put_fs_struct(struct fs_struct *fs)
533 {
534 	__put_fs_struct(fs);
535 }
536 
537 static inline void __exit_fs(struct task_struct *tsk)
538 {
539 	struct fs_struct * fs = tsk->fs;
540 
541 	if (fs) {
542 		task_lock(tsk);
543 		tsk->fs = NULL;
544 		task_unlock(tsk);
545 		__put_fs_struct(fs);
546 	}
547 }
548 
549 void exit_fs(struct task_struct *tsk)
550 {
551 	__exit_fs(tsk);
552 }
553 
554 EXPORT_SYMBOL_GPL(exit_fs);
555 
556 /*
557  * Turn us into a lazy TLB process if we
558  * aren't already..
559  */
560 static void exit_mm(struct task_struct * tsk)
561 {
562 	struct mm_struct *mm = tsk->mm;
563 
564 	mm_release(tsk, mm);
565 	if (!mm)
566 		return;
567 	/*
568 	 * Serialize with any possible pending coredump.
569 	 * We must hold mmap_sem around checking core_waiters
570 	 * and clearing tsk->mm.  The core-inducing thread
571 	 * will increment core_waiters for each thread in the
572 	 * group with ->mm != NULL.
573 	 */
574 	down_read(&mm->mmap_sem);
575 	if (mm->core_waiters) {
576 		up_read(&mm->mmap_sem);
577 		down_write(&mm->mmap_sem);
578 		if (!--mm->core_waiters)
579 			complete(mm->core_startup_done);
580 		up_write(&mm->mmap_sem);
581 
582 		wait_for_completion(&mm->core_done);
583 		down_read(&mm->mmap_sem);
584 	}
585 	atomic_inc(&mm->mm_count);
586 	BUG_ON(mm != tsk->active_mm);
587 	/* more a memory barrier than a real lock */
588 	task_lock(tsk);
589 	tsk->mm = NULL;
590 	up_read(&mm->mmap_sem);
591 	enter_lazy_tlb(mm, current);
592 	task_unlock(tsk);
593 	mmput(mm);
594 }
595 
596 static inline void
597 choose_new_parent(struct task_struct *p, struct task_struct *reaper)
598 {
599 	/*
600 	 * Make sure we're not reparenting to ourselves and that
601 	 * the parent is not a zombie.
602 	 */
603 	BUG_ON(p == reaper || reaper->exit_state);
604 	p->real_parent = reaper;
605 }
606 
607 static void
608 reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
609 {
610 	if (p->pdeath_signal)
611 		/* We already hold the tasklist_lock here.  */
612 		group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
613 
614 	/* Move the child from its dying parent to the new one.  */
615 	if (unlikely(traced)) {
616 		/* Preserve ptrace links if someone else is tracing this child.  */
617 		list_del_init(&p->ptrace_list);
618 		if (p->parent != p->real_parent)
619 			list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
620 	} else {
621 		/* If this child is being traced, then we're the one tracing it
622 		 * anyway, so let go of it.
623 		 */
624 		p->ptrace = 0;
625 		remove_parent(p);
626 		p->parent = p->real_parent;
627 		add_parent(p);
628 
629 		if (p->state == TASK_TRACED) {
630 			/*
631 			 * If it was at a trace stop, turn it into
632 			 * a normal stop since it's no longer being
633 			 * traced.
634 			 */
635 			ptrace_untrace(p);
636 		}
637 	}
638 
639 	/* If this is a threaded reparent there is no need to
640 	 * notify anyone anything has happened.
641 	 */
642 	if (p->real_parent->group_leader == father->group_leader)
643 		return;
644 
645 	/* We don't want people slaying init.  */
646 	if (p->exit_signal != -1)
647 		p->exit_signal = SIGCHLD;
648 
649 	/* If we'd notified the old parent about this child's death,
650 	 * also notify the new parent.
651 	 */
652 	if (!traced && p->exit_state == EXIT_ZOMBIE &&
653 	    p->exit_signal != -1 && thread_group_empty(p))
654 		do_notify_parent(p, p->exit_signal);
655 
656 	/*
657 	 * process group orphan check
658 	 * Case ii: Our child is in a different pgrp
659 	 * than we are, and it was the only connection
660 	 * outside, so the child pgrp is now orphaned.
661 	 */
662 	if ((task_pgrp(p) != task_pgrp(father)) &&
663 	    (task_session(p) == task_session(father))) {
664 		struct pid *pgrp = task_pgrp(p);
665 
666 		if (will_become_orphaned_pgrp(pgrp, NULL) &&
667 		    has_stopped_jobs(pgrp)) {
668 			__kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
669 			__kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
670 		}
671 	}
672 }
673 
674 /*
675  * When we die, we re-parent all our children.
676  * Try to give them to another thread in our thread
677  * group, and if no such member exists, give it to
678  * the child reaper process (ie "init") in our pid
679  * space.
680  */
681 static void
682 forget_original_parent(struct task_struct *father, struct list_head *to_release)
683 {
684 	struct task_struct *p, *reaper = father;
685 	struct list_head *_p, *_n;
686 
687 	do {
688 		reaper = next_thread(reaper);
689 		if (reaper == father) {
690 			reaper = child_reaper(father);
691 			break;
692 		}
693 	} while (reaper->exit_state);
694 
695 	/*
696 	 * There are only two places where our children can be:
697 	 *
698 	 * - in our child list
699 	 * - in our ptraced child list
700 	 *
701 	 * Search them and reparent children.
702 	 */
703 	list_for_each_safe(_p, _n, &father->children) {
704 		int ptrace;
705 		p = list_entry(_p, struct task_struct, sibling);
706 
707 		ptrace = p->ptrace;
708 
709 		/* if father isn't the real parent, then ptrace must be enabled */
710 		BUG_ON(father != p->real_parent && !ptrace);
711 
712 		if (father == p->real_parent) {
713 			/* reparent with a reaper, real father it's us */
714 			choose_new_parent(p, reaper);
715 			reparent_thread(p, father, 0);
716 		} else {
717 			/* reparent ptraced task to its real parent */
718 			__ptrace_unlink (p);
719 			if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
720 			    thread_group_empty(p))
721 				do_notify_parent(p, p->exit_signal);
722 		}
723 
724 		/*
725 		 * if the ptraced child is a zombie with exit_signal == -1
726 		 * we must collect it before we exit, or it will remain
727 		 * zombie forever since we prevented it from self-reap itself
728 		 * while it was being traced by us, to be able to see it in wait4.
729 		 */
730 		if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
731 			list_add(&p->ptrace_list, to_release);
732 	}
733 	list_for_each_safe(_p, _n, &father->ptrace_children) {
734 		p = list_entry(_p, struct task_struct, ptrace_list);
735 		choose_new_parent(p, reaper);
736 		reparent_thread(p, father, 1);
737 	}
738 }
739 
740 /*
741  * Send signals to all our closest relatives so that they know
742  * to properly mourn us..
743  */
744 static void exit_notify(struct task_struct *tsk)
745 {
746 	int state;
747 	struct task_struct *t;
748 	struct list_head ptrace_dead, *_p, *_n;
749 	struct pid *pgrp;
750 
751 	if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
752 	    && !thread_group_empty(tsk)) {
753 		/*
754 		 * This occurs when there was a race between our exit
755 		 * syscall and a group signal choosing us as the one to
756 		 * wake up.  It could be that we are the only thread
757 		 * alerted to check for pending signals, but another thread
758 		 * should be woken now to take the signal since we will not.
759 		 * Now we'll wake all the threads in the group just to make
760 		 * sure someone gets all the pending signals.
761 		 */
762 		read_lock(&tasklist_lock);
763 		spin_lock_irq(&tsk->sighand->siglock);
764 		for (t = next_thread(tsk); t != tsk; t = next_thread(t))
765 			if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
766 				recalc_sigpending_tsk(t);
767 				if (signal_pending(t))
768 					signal_wake_up(t, 0);
769 			}
770 		spin_unlock_irq(&tsk->sighand->siglock);
771 		read_unlock(&tasklist_lock);
772 	}
773 
774 	write_lock_irq(&tasklist_lock);
775 
776 	/*
777 	 * This does two things:
778 	 *
779   	 * A.  Make init inherit all the child processes
780 	 * B.  Check to see if any process groups have become orphaned
781 	 *	as a result of our exiting, and if they have any stopped
782 	 *	jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
783 	 */
784 
785 	INIT_LIST_HEAD(&ptrace_dead);
786 	forget_original_parent(tsk, &ptrace_dead);
787 	BUG_ON(!list_empty(&tsk->children));
788 	BUG_ON(!list_empty(&tsk->ptrace_children));
789 
790 	/*
791 	 * Check to see if any process groups have become orphaned
792 	 * as a result of our exiting, and if they have any stopped
793 	 * jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
794 	 *
795 	 * Case i: Our father is in a different pgrp than we are
796 	 * and we were the only connection outside, so our pgrp
797 	 * is about to become orphaned.
798 	 */
799 
800 	t = tsk->real_parent;
801 
802 	pgrp = task_pgrp(tsk);
803 	if ((task_pgrp(t) != pgrp) &&
804 	    (task_session(t) == task_session(tsk)) &&
805 	    will_become_orphaned_pgrp(pgrp, tsk) &&
806 	    has_stopped_jobs(pgrp)) {
807 		__kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
808 		__kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
809 	}
810 
811 	/* Let father know we died
812 	 *
813 	 * Thread signals are configurable, but you aren't going to use
814 	 * that to send signals to arbitary processes.
815 	 * That stops right now.
816 	 *
817 	 * If the parent exec id doesn't match the exec id we saved
818 	 * when we started then we know the parent has changed security
819 	 * domain.
820 	 *
821 	 * If our self_exec id doesn't match our parent_exec_id then
822 	 * we have changed execution domain as these two values started
823 	 * the same after a fork.
824 	 *
825 	 */
826 
827 	if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
828 	    ( tsk->parent_exec_id != t->self_exec_id  ||
829 	      tsk->self_exec_id != tsk->parent_exec_id)
830 	    && !capable(CAP_KILL))
831 		tsk->exit_signal = SIGCHLD;
832 
833 
834 	/* If something other than our normal parent is ptracing us, then
835 	 * send it a SIGCHLD instead of honoring exit_signal.  exit_signal
836 	 * only has special meaning to our real parent.
837 	 */
838 	if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
839 		int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
840 		do_notify_parent(tsk, signal);
841 	} else if (tsk->ptrace) {
842 		do_notify_parent(tsk, SIGCHLD);
843 	}
844 
845 	state = EXIT_ZOMBIE;
846 	if (tsk->exit_signal == -1 &&
847 	    (likely(tsk->ptrace == 0) ||
848 	     unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
849 		state = EXIT_DEAD;
850 	tsk->exit_state = state;
851 
852 	write_unlock_irq(&tasklist_lock);
853 
854 	list_for_each_safe(_p, _n, &ptrace_dead) {
855 		list_del_init(_p);
856 		t = list_entry(_p, struct task_struct, ptrace_list);
857 		release_task(t);
858 	}
859 
860 	/* If the process is dead, release it - nobody will wait for it */
861 	if (state == EXIT_DEAD)
862 		release_task(tsk);
863 }
864 
865 fastcall NORET_TYPE void do_exit(long code)
866 {
867 	struct task_struct *tsk = current;
868 	int group_dead;
869 
870 	profile_task_exit(tsk);
871 
872 	WARN_ON(atomic_read(&tsk->fs_excl));
873 
874 	if (unlikely(in_interrupt()))
875 		panic("Aiee, killing interrupt handler!");
876 	if (unlikely(!tsk->pid))
877 		panic("Attempted to kill the idle task!");
878 	if (unlikely(tsk == child_reaper(tsk))) {
879 		if (tsk->nsproxy->pid_ns != &init_pid_ns)
880 			tsk->nsproxy->pid_ns->child_reaper = init_pid_ns.child_reaper;
881 		else
882 			panic("Attempted to kill init!");
883 	}
884 
885 
886 	if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
887 		current->ptrace_message = code;
888 		ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
889 	}
890 
891 	/*
892 	 * We're taking recursive faults here in do_exit. Safest is to just
893 	 * leave this task alone and wait for reboot.
894 	 */
895 	if (unlikely(tsk->flags & PF_EXITING)) {
896 		printk(KERN_ALERT
897 			"Fixing recursive fault but reboot is needed!\n");
898 		if (tsk->io_context)
899 			exit_io_context();
900 		set_current_state(TASK_UNINTERRUPTIBLE);
901 		schedule();
902 	}
903 
904 	tsk->flags |= PF_EXITING;
905 
906 	if (unlikely(in_atomic()))
907 		printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
908 				current->comm, current->pid,
909 				preempt_count());
910 
911 	acct_update_integrals(tsk);
912 	if (tsk->mm) {
913 		update_hiwater_rss(tsk->mm);
914 		update_hiwater_vm(tsk->mm);
915 	}
916 	group_dead = atomic_dec_and_test(&tsk->signal->live);
917 	if (group_dead) {
918  		hrtimer_cancel(&tsk->signal->real_timer);
919 		exit_itimers(tsk->signal);
920 	}
921 	acct_collect(code, group_dead);
922 	if (unlikely(tsk->robust_list))
923 		exit_robust_list(tsk);
924 #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
925 	if (unlikely(tsk->compat_robust_list))
926 		compat_exit_robust_list(tsk);
927 #endif
928 	if (unlikely(tsk->audit_context))
929 		audit_free(tsk);
930 
931 	taskstats_exit(tsk, group_dead);
932 
933 	exit_mm(tsk);
934 
935 	if (group_dead)
936 		acct_process();
937 	exit_sem(tsk);
938 	__exit_files(tsk);
939 	__exit_fs(tsk);
940 	exit_thread();
941 	cpuset_exit(tsk);
942 	exit_keys(tsk);
943 
944 	if (group_dead && tsk->signal->leader)
945 		disassociate_ctty(1);
946 
947 	module_put(task_thread_info(tsk)->exec_domain->module);
948 	if (tsk->binfmt)
949 		module_put(tsk->binfmt->module);
950 
951 	tsk->exit_code = code;
952 	proc_exit_connector(tsk);
953 	exit_task_namespaces(tsk);
954 	exit_notify(tsk);
955 #ifdef CONFIG_NUMA
956 	mpol_free(tsk->mempolicy);
957 	tsk->mempolicy = NULL;
958 #endif
959 	/*
960 	 * This must happen late, after the PID is not
961 	 * hashed anymore:
962 	 */
963 	if (unlikely(!list_empty(&tsk->pi_state_list)))
964 		exit_pi_state_list(tsk);
965 	if (unlikely(current->pi_state_cache))
966 		kfree(current->pi_state_cache);
967 	/*
968 	 * Make sure we are holding no locks:
969 	 */
970 	debug_check_no_locks_held(tsk);
971 
972 	if (tsk->io_context)
973 		exit_io_context();
974 
975 	if (tsk->splice_pipe)
976 		__free_pipe_info(tsk->splice_pipe);
977 
978 	preempt_disable();
979 	/* causes final put_task_struct in finish_task_switch(). */
980 	tsk->state = TASK_DEAD;
981 
982 	schedule();
983 	BUG();
984 	/* Avoid "noreturn function does return".  */
985 	for (;;)
986 		cpu_relax();	/* For when BUG is null */
987 }
988 
989 EXPORT_SYMBOL_GPL(do_exit);
990 
991 NORET_TYPE void complete_and_exit(struct completion *comp, long code)
992 {
993 	if (comp)
994 		complete(comp);
995 
996 	do_exit(code);
997 }
998 
999 EXPORT_SYMBOL(complete_and_exit);
1000 
1001 asmlinkage long sys_exit(int error_code)
1002 {
1003 	do_exit((error_code&0xff)<<8);
1004 }
1005 
1006 /*
1007  * Take down every thread in the group.  This is called by fatal signals
1008  * as well as by sys_exit_group (below).
1009  */
1010 NORET_TYPE void
1011 do_group_exit(int exit_code)
1012 {
1013 	BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1014 
1015 	if (current->signal->flags & SIGNAL_GROUP_EXIT)
1016 		exit_code = current->signal->group_exit_code;
1017 	else if (!thread_group_empty(current)) {
1018 		struct signal_struct *const sig = current->signal;
1019 		struct sighand_struct *const sighand = current->sighand;
1020 		spin_lock_irq(&sighand->siglock);
1021 		if (sig->flags & SIGNAL_GROUP_EXIT)
1022 			/* Another thread got here before we took the lock.  */
1023 			exit_code = sig->group_exit_code;
1024 		else {
1025 			sig->group_exit_code = exit_code;
1026 			zap_other_threads(current);
1027 		}
1028 		spin_unlock_irq(&sighand->siglock);
1029 	}
1030 
1031 	do_exit(exit_code);
1032 	/* NOTREACHED */
1033 }
1034 
1035 /*
1036  * this kills every thread in the thread group. Note that any externally
1037  * wait4()-ing process will get the correct exit code - even if this
1038  * thread is not the thread group leader.
1039  */
1040 asmlinkage void sys_exit_group(int error_code)
1041 {
1042 	do_group_exit((error_code & 0xff) << 8);
1043 }
1044 
1045 static int eligible_child(pid_t pid, int options, struct task_struct *p)
1046 {
1047 	int err;
1048 
1049 	if (pid > 0) {
1050 		if (p->pid != pid)
1051 			return 0;
1052 	} else if (!pid) {
1053 		if (process_group(p) != process_group(current))
1054 			return 0;
1055 	} else if (pid != -1) {
1056 		if (process_group(p) != -pid)
1057 			return 0;
1058 	}
1059 
1060 	/*
1061 	 * Do not consider detached threads that are
1062 	 * not ptraced:
1063 	 */
1064 	if (p->exit_signal == -1 && !p->ptrace)
1065 		return 0;
1066 
1067 	/* Wait for all children (clone and not) if __WALL is set;
1068 	 * otherwise, wait for clone children *only* if __WCLONE is
1069 	 * set; otherwise, wait for non-clone children *only*.  (Note:
1070 	 * A "clone" child here is one that reports to its parent
1071 	 * using a signal other than SIGCHLD.) */
1072 	if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1073 	    && !(options & __WALL))
1074 		return 0;
1075 	/*
1076 	 * Do not consider thread group leaders that are
1077 	 * in a non-empty thread group:
1078 	 */
1079 	if (delay_group_leader(p))
1080 		return 2;
1081 
1082 	err = security_task_wait(p);
1083 	if (err)
1084 		return err;
1085 
1086 	return 1;
1087 }
1088 
1089 static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
1090 			       int why, int status,
1091 			       struct siginfo __user *infop,
1092 			       struct rusage __user *rusagep)
1093 {
1094 	int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
1095 
1096 	put_task_struct(p);
1097 	if (!retval)
1098 		retval = put_user(SIGCHLD, &infop->si_signo);
1099 	if (!retval)
1100 		retval = put_user(0, &infop->si_errno);
1101 	if (!retval)
1102 		retval = put_user((short)why, &infop->si_code);
1103 	if (!retval)
1104 		retval = put_user(pid, &infop->si_pid);
1105 	if (!retval)
1106 		retval = put_user(uid, &infop->si_uid);
1107 	if (!retval)
1108 		retval = put_user(status, &infop->si_status);
1109 	if (!retval)
1110 		retval = pid;
1111 	return retval;
1112 }
1113 
1114 /*
1115  * Handle sys_wait4 work for one task in state EXIT_ZOMBIE.  We hold
1116  * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
1117  * the lock and this task is uninteresting.  If we return nonzero, we have
1118  * released the lock and the system call should return.
1119  */
1120 static int wait_task_zombie(struct task_struct *p, int noreap,
1121 			    struct siginfo __user *infop,
1122 			    int __user *stat_addr, struct rusage __user *ru)
1123 {
1124 	unsigned long state;
1125 	int retval;
1126 	int status;
1127 
1128 	if (unlikely(noreap)) {
1129 		pid_t pid = p->pid;
1130 		uid_t uid = p->uid;
1131 		int exit_code = p->exit_code;
1132 		int why, status;
1133 
1134 		if (unlikely(p->exit_state != EXIT_ZOMBIE))
1135 			return 0;
1136 		if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1137 			return 0;
1138 		get_task_struct(p);
1139 		read_unlock(&tasklist_lock);
1140 		if ((exit_code & 0x7f) == 0) {
1141 			why = CLD_EXITED;
1142 			status = exit_code >> 8;
1143 		} else {
1144 			why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1145 			status = exit_code & 0x7f;
1146 		}
1147 		return wait_noreap_copyout(p, pid, uid, why,
1148 					   status, infop, ru);
1149 	}
1150 
1151 	/*
1152 	 * Try to move the task's state to DEAD
1153 	 * only one thread is allowed to do this:
1154 	 */
1155 	state = xchg(&p->exit_state, EXIT_DEAD);
1156 	if (state != EXIT_ZOMBIE) {
1157 		BUG_ON(state != EXIT_DEAD);
1158 		return 0;
1159 	}
1160 	if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1161 		/*
1162 		 * This can only happen in a race with a ptraced thread
1163 		 * dying on another processor.
1164 		 */
1165 		return 0;
1166 	}
1167 
1168 	if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1169 		struct signal_struct *psig;
1170 		struct signal_struct *sig;
1171 
1172 		/*
1173 		 * The resource counters for the group leader are in its
1174 		 * own task_struct.  Those for dead threads in the group
1175 		 * are in its signal_struct, as are those for the child
1176 		 * processes it has previously reaped.  All these
1177 		 * accumulate in the parent's signal_struct c* fields.
1178 		 *
1179 		 * We don't bother to take a lock here to protect these
1180 		 * p->signal fields, because they are only touched by
1181 		 * __exit_signal, which runs with tasklist_lock
1182 		 * write-locked anyway, and so is excluded here.  We do
1183 		 * need to protect the access to p->parent->signal fields,
1184 		 * as other threads in the parent group can be right
1185 		 * here reaping other children at the same time.
1186 		 */
1187 		spin_lock_irq(&p->parent->sighand->siglock);
1188 		psig = p->parent->signal;
1189 		sig = p->signal;
1190 		psig->cutime =
1191 			cputime_add(psig->cutime,
1192 			cputime_add(p->utime,
1193 			cputime_add(sig->utime,
1194 				    sig->cutime)));
1195 		psig->cstime =
1196 			cputime_add(psig->cstime,
1197 			cputime_add(p->stime,
1198 			cputime_add(sig->stime,
1199 				    sig->cstime)));
1200 		psig->cmin_flt +=
1201 			p->min_flt + sig->min_flt + sig->cmin_flt;
1202 		psig->cmaj_flt +=
1203 			p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1204 		psig->cnvcsw +=
1205 			p->nvcsw + sig->nvcsw + sig->cnvcsw;
1206 		psig->cnivcsw +=
1207 			p->nivcsw + sig->nivcsw + sig->cnivcsw;
1208 		psig->cinblock +=
1209 			task_io_get_inblock(p) +
1210 			sig->inblock + sig->cinblock;
1211 		psig->coublock +=
1212 			task_io_get_oublock(p) +
1213 			sig->oublock + sig->coublock;
1214 		spin_unlock_irq(&p->parent->sighand->siglock);
1215 	}
1216 
1217 	/*
1218 	 * Now we are sure this task is interesting, and no other
1219 	 * thread can reap it because we set its state to EXIT_DEAD.
1220 	 */
1221 	read_unlock(&tasklist_lock);
1222 
1223 	retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1224 	status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1225 		? p->signal->group_exit_code : p->exit_code;
1226 	if (!retval && stat_addr)
1227 		retval = put_user(status, stat_addr);
1228 	if (!retval && infop)
1229 		retval = put_user(SIGCHLD, &infop->si_signo);
1230 	if (!retval && infop)
1231 		retval = put_user(0, &infop->si_errno);
1232 	if (!retval && infop) {
1233 		int why;
1234 
1235 		if ((status & 0x7f) == 0) {
1236 			why = CLD_EXITED;
1237 			status >>= 8;
1238 		} else {
1239 			why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1240 			status &= 0x7f;
1241 		}
1242 		retval = put_user((short)why, &infop->si_code);
1243 		if (!retval)
1244 			retval = put_user(status, &infop->si_status);
1245 	}
1246 	if (!retval && infop)
1247 		retval = put_user(p->pid, &infop->si_pid);
1248 	if (!retval && infop)
1249 		retval = put_user(p->uid, &infop->si_uid);
1250 	if (retval) {
1251 		// TODO: is this safe?
1252 		p->exit_state = EXIT_ZOMBIE;
1253 		return retval;
1254 	}
1255 	retval = p->pid;
1256 	if (p->real_parent != p->parent) {
1257 		write_lock_irq(&tasklist_lock);
1258 		/* Double-check with lock held.  */
1259 		if (p->real_parent != p->parent) {
1260 			__ptrace_unlink(p);
1261 			// TODO: is this safe?
1262 			p->exit_state = EXIT_ZOMBIE;
1263 			/*
1264 			 * If this is not a detached task, notify the parent.
1265 			 * If it's still not detached after that, don't release
1266 			 * it now.
1267 			 */
1268 			if (p->exit_signal != -1) {
1269 				do_notify_parent(p, p->exit_signal);
1270 				if (p->exit_signal != -1)
1271 					p = NULL;
1272 			}
1273 		}
1274 		write_unlock_irq(&tasklist_lock);
1275 	}
1276 	if (p != NULL)
1277 		release_task(p);
1278 	BUG_ON(!retval);
1279 	return retval;
1280 }
1281 
1282 /*
1283  * Handle sys_wait4 work for one task in state TASK_STOPPED.  We hold
1284  * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
1285  * the lock and this task is uninteresting.  If we return nonzero, we have
1286  * released the lock and the system call should return.
1287  */
1288 static int wait_task_stopped(struct task_struct *p, int delayed_group_leader,
1289 			     int noreap, struct siginfo __user *infop,
1290 			     int __user *stat_addr, struct rusage __user *ru)
1291 {
1292 	int retval, exit_code;
1293 
1294 	if (!p->exit_code)
1295 		return 0;
1296 	if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1297 	    p->signal && p->signal->group_stop_count > 0)
1298 		/*
1299 		 * A group stop is in progress and this is the group leader.
1300 		 * We won't report until all threads have stopped.
1301 		 */
1302 		return 0;
1303 
1304 	/*
1305 	 * Now we are pretty sure this task is interesting.
1306 	 * Make sure it doesn't get reaped out from under us while we
1307 	 * give up the lock and then examine it below.  We don't want to
1308 	 * keep holding onto the tasklist_lock while we call getrusage and
1309 	 * possibly take page faults for user memory.
1310 	 */
1311 	get_task_struct(p);
1312 	read_unlock(&tasklist_lock);
1313 
1314 	if (unlikely(noreap)) {
1315 		pid_t pid = p->pid;
1316 		uid_t uid = p->uid;
1317 		int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1318 
1319 		exit_code = p->exit_code;
1320 		if (unlikely(!exit_code) ||
1321 		    unlikely(p->state & TASK_TRACED))
1322 			goto bail_ref;
1323 		return wait_noreap_copyout(p, pid, uid,
1324 					   why, (exit_code << 8) | 0x7f,
1325 					   infop, ru);
1326 	}
1327 
1328 	write_lock_irq(&tasklist_lock);
1329 
1330 	/*
1331 	 * This uses xchg to be atomic with the thread resuming and setting
1332 	 * it.  It must also be done with the write lock held to prevent a
1333 	 * race with the EXIT_ZOMBIE case.
1334 	 */
1335 	exit_code = xchg(&p->exit_code, 0);
1336 	if (unlikely(p->exit_state)) {
1337 		/*
1338 		 * The task resumed and then died.  Let the next iteration
1339 		 * catch it in EXIT_ZOMBIE.  Note that exit_code might
1340 		 * already be zero here if it resumed and did _exit(0).
1341 		 * The task itself is dead and won't touch exit_code again;
1342 		 * other processors in this function are locked out.
1343 		 */
1344 		p->exit_code = exit_code;
1345 		exit_code = 0;
1346 	}
1347 	if (unlikely(exit_code == 0)) {
1348 		/*
1349 		 * Another thread in this function got to it first, or it
1350 		 * resumed, or it resumed and then died.
1351 		 */
1352 		write_unlock_irq(&tasklist_lock);
1353 bail_ref:
1354 		put_task_struct(p);
1355 		/*
1356 		 * We are returning to the wait loop without having successfully
1357 		 * removed the process and having released the lock. We cannot
1358 		 * continue, since the "p" task pointer is potentially stale.
1359 		 *
1360 		 * Return -EAGAIN, and do_wait() will restart the loop from the
1361 		 * beginning. Do _not_ re-acquire the lock.
1362 		 */
1363 		return -EAGAIN;
1364 	}
1365 
1366 	/* move to end of parent's list to avoid starvation */
1367 	remove_parent(p);
1368 	add_parent(p);
1369 
1370 	write_unlock_irq(&tasklist_lock);
1371 
1372 	retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1373 	if (!retval && stat_addr)
1374 		retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1375 	if (!retval && infop)
1376 		retval = put_user(SIGCHLD, &infop->si_signo);
1377 	if (!retval && infop)
1378 		retval = put_user(0, &infop->si_errno);
1379 	if (!retval && infop)
1380 		retval = put_user((short)((p->ptrace & PT_PTRACED)
1381 					  ? CLD_TRAPPED : CLD_STOPPED),
1382 				  &infop->si_code);
1383 	if (!retval && infop)
1384 		retval = put_user(exit_code, &infop->si_status);
1385 	if (!retval && infop)
1386 		retval = put_user(p->pid, &infop->si_pid);
1387 	if (!retval && infop)
1388 		retval = put_user(p->uid, &infop->si_uid);
1389 	if (!retval)
1390 		retval = p->pid;
1391 	put_task_struct(p);
1392 
1393 	BUG_ON(!retval);
1394 	return retval;
1395 }
1396 
1397 /*
1398  * Handle do_wait work for one task in a live, non-stopped state.
1399  * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
1400  * the lock and this task is uninteresting.  If we return nonzero, we have
1401  * released the lock and the system call should return.
1402  */
1403 static int wait_task_continued(struct task_struct *p, int noreap,
1404 			       struct siginfo __user *infop,
1405 			       int __user *stat_addr, struct rusage __user *ru)
1406 {
1407 	int retval;
1408 	pid_t pid;
1409 	uid_t uid;
1410 
1411 	if (unlikely(!p->signal))
1412 		return 0;
1413 
1414 	if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1415 		return 0;
1416 
1417 	spin_lock_irq(&p->sighand->siglock);
1418 	/* Re-check with the lock held.  */
1419 	if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1420 		spin_unlock_irq(&p->sighand->siglock);
1421 		return 0;
1422 	}
1423 	if (!noreap)
1424 		p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1425 	spin_unlock_irq(&p->sighand->siglock);
1426 
1427 	pid = p->pid;
1428 	uid = p->uid;
1429 	get_task_struct(p);
1430 	read_unlock(&tasklist_lock);
1431 
1432 	if (!infop) {
1433 		retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1434 		put_task_struct(p);
1435 		if (!retval && stat_addr)
1436 			retval = put_user(0xffff, stat_addr);
1437 		if (!retval)
1438 			retval = p->pid;
1439 	} else {
1440 		retval = wait_noreap_copyout(p, pid, uid,
1441 					     CLD_CONTINUED, SIGCONT,
1442 					     infop, ru);
1443 		BUG_ON(retval == 0);
1444 	}
1445 
1446 	return retval;
1447 }
1448 
1449 
1450 static inline int my_ptrace_child(struct task_struct *p)
1451 {
1452 	if (!(p->ptrace & PT_PTRACED))
1453 		return 0;
1454 	if (!(p->ptrace & PT_ATTACHED))
1455 		return 1;
1456 	/*
1457 	 * This child was PTRACE_ATTACH'd.  We should be seeing it only if
1458 	 * we are the attacher.  If we are the real parent, this is a race
1459 	 * inside ptrace_attach.  It is waiting for the tasklist_lock,
1460 	 * which we have to switch the parent links, but has already set
1461 	 * the flags in p->ptrace.
1462 	 */
1463 	return (p->parent != p->real_parent);
1464 }
1465 
1466 static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1467 		    int __user *stat_addr, struct rusage __user *ru)
1468 {
1469 	DECLARE_WAITQUEUE(wait, current);
1470 	struct task_struct *tsk;
1471 	int flag, retval;
1472 	int allowed, denied;
1473 
1474 	add_wait_queue(&current->signal->wait_chldexit,&wait);
1475 repeat:
1476 	/*
1477 	 * We will set this flag if we see any child that might later
1478 	 * match our criteria, even if we are not able to reap it yet.
1479 	 */
1480 	flag = 0;
1481 	allowed = denied = 0;
1482 	current->state = TASK_INTERRUPTIBLE;
1483 	read_lock(&tasklist_lock);
1484 	tsk = current;
1485 	do {
1486 		struct task_struct *p;
1487 		struct list_head *_p;
1488 		int ret;
1489 
1490 		list_for_each(_p,&tsk->children) {
1491 			p = list_entry(_p, struct task_struct, sibling);
1492 
1493 			ret = eligible_child(pid, options, p);
1494 			if (!ret)
1495 				continue;
1496 
1497 			if (unlikely(ret < 0)) {
1498 				denied = ret;
1499 				continue;
1500 			}
1501 			allowed = 1;
1502 
1503 			switch (p->state) {
1504 			case TASK_TRACED:
1505 				/*
1506 				 * When we hit the race with PTRACE_ATTACH,
1507 				 * we will not report this child.  But the
1508 				 * race means it has not yet been moved to
1509 				 * our ptrace_children list, so we need to
1510 				 * set the flag here to avoid a spurious ECHILD
1511 				 * when the race happens with the only child.
1512 				 */
1513 				flag = 1;
1514 				if (!my_ptrace_child(p))
1515 					continue;
1516 				/*FALLTHROUGH*/
1517 			case TASK_STOPPED:
1518 				/*
1519 				 * It's stopped now, so it might later
1520 				 * continue, exit, or stop again.
1521 				 */
1522 				flag = 1;
1523 				if (!(options & WUNTRACED) &&
1524 				    !my_ptrace_child(p))
1525 					continue;
1526 				retval = wait_task_stopped(p, ret == 2,
1527 							   (options & WNOWAIT),
1528 							   infop,
1529 							   stat_addr, ru);
1530 				if (retval == -EAGAIN)
1531 					goto repeat;
1532 				if (retval != 0) /* He released the lock.  */
1533 					goto end;
1534 				break;
1535 			default:
1536 			// case EXIT_DEAD:
1537 				if (p->exit_state == EXIT_DEAD)
1538 					continue;
1539 			// case EXIT_ZOMBIE:
1540 				if (p->exit_state == EXIT_ZOMBIE) {
1541 					/*
1542 					 * Eligible but we cannot release
1543 					 * it yet:
1544 					 */
1545 					if (ret == 2)
1546 						goto check_continued;
1547 					if (!likely(options & WEXITED))
1548 						continue;
1549 					retval = wait_task_zombie(
1550 						p, (options & WNOWAIT),
1551 						infop, stat_addr, ru);
1552 					/* He released the lock.  */
1553 					if (retval != 0)
1554 						goto end;
1555 					break;
1556 				}
1557 check_continued:
1558 				/*
1559 				 * It's running now, so it might later
1560 				 * exit, stop, or stop and then continue.
1561 				 */
1562 				flag = 1;
1563 				if (!unlikely(options & WCONTINUED))
1564 					continue;
1565 				retval = wait_task_continued(
1566 					p, (options & WNOWAIT),
1567 					infop, stat_addr, ru);
1568 				if (retval != 0) /* He released the lock.  */
1569 					goto end;
1570 				break;
1571 			}
1572 		}
1573 		if (!flag) {
1574 			list_for_each(_p, &tsk->ptrace_children) {
1575 				p = list_entry(_p, struct task_struct,
1576 						ptrace_list);
1577 				if (!eligible_child(pid, options, p))
1578 					continue;
1579 				flag = 1;
1580 				break;
1581 			}
1582 		}
1583 		if (options & __WNOTHREAD)
1584 			break;
1585 		tsk = next_thread(tsk);
1586 		BUG_ON(tsk->signal != current->signal);
1587 	} while (tsk != current);
1588 
1589 	read_unlock(&tasklist_lock);
1590 	if (flag) {
1591 		retval = 0;
1592 		if (options & WNOHANG)
1593 			goto end;
1594 		retval = -ERESTARTSYS;
1595 		if (signal_pending(current))
1596 			goto end;
1597 		schedule();
1598 		goto repeat;
1599 	}
1600 	retval = -ECHILD;
1601 	if (unlikely(denied) && !allowed)
1602 		retval = denied;
1603 end:
1604 	current->state = TASK_RUNNING;
1605 	remove_wait_queue(&current->signal->wait_chldexit,&wait);
1606 	if (infop) {
1607 		if (retval > 0)
1608 		retval = 0;
1609 		else {
1610 			/*
1611 			 * For a WNOHANG return, clear out all the fields
1612 			 * we would set so the user can easily tell the
1613 			 * difference.
1614 			 */
1615 			if (!retval)
1616 				retval = put_user(0, &infop->si_signo);
1617 			if (!retval)
1618 				retval = put_user(0, &infop->si_errno);
1619 			if (!retval)
1620 				retval = put_user(0, &infop->si_code);
1621 			if (!retval)
1622 				retval = put_user(0, &infop->si_pid);
1623 			if (!retval)
1624 				retval = put_user(0, &infop->si_uid);
1625 			if (!retval)
1626 				retval = put_user(0, &infop->si_status);
1627 		}
1628 	}
1629 	return retval;
1630 }
1631 
1632 asmlinkage long sys_waitid(int which, pid_t pid,
1633 			   struct siginfo __user *infop, int options,
1634 			   struct rusage __user *ru)
1635 {
1636 	long ret;
1637 
1638 	if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1639 		return -EINVAL;
1640 	if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1641 		return -EINVAL;
1642 
1643 	switch (which) {
1644 	case P_ALL:
1645 		pid = -1;
1646 		break;
1647 	case P_PID:
1648 		if (pid <= 0)
1649 			return -EINVAL;
1650 		break;
1651 	case P_PGID:
1652 		if (pid <= 0)
1653 			return -EINVAL;
1654 		pid = -pid;
1655 		break;
1656 	default:
1657 		return -EINVAL;
1658 	}
1659 
1660 	ret = do_wait(pid, options, infop, NULL, ru);
1661 
1662 	/* avoid REGPARM breakage on x86: */
1663 	prevent_tail_call(ret);
1664 	return ret;
1665 }
1666 
1667 asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1668 			  int options, struct rusage __user *ru)
1669 {
1670 	long ret;
1671 
1672 	if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1673 			__WNOTHREAD|__WCLONE|__WALL))
1674 		return -EINVAL;
1675 	ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1676 
1677 	/* avoid REGPARM breakage on x86: */
1678 	prevent_tail_call(ret);
1679 	return ret;
1680 }
1681 
1682 #ifdef __ARCH_WANT_SYS_WAITPID
1683 
1684 /*
1685  * sys_waitpid() remains for compatibility. waitpid() should be
1686  * implemented by calling sys_wait4() from libc.a.
1687  */
1688 asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1689 {
1690 	return sys_wait4(pid, stat_addr, options, NULL);
1691 }
1692 
1693 #endif
1694