xref: /freebsd/sys/kern/kern_exit.c (revision 43d1e6ee299ad4e143d90d3ad374d1c24bd3306f)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)kern_exit.c	8.7 (Berkeley) 2/12/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_ktrace.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/sysproto.h>
47 #include <sys/capsicum.h>
48 #include <sys/eventhandler.h>
49 #include <sys/kernel.h>
50 #include <sys/malloc.h>
51 #include <sys/lock.h>
52 #include <sys/mutex.h>
53 #include <sys/proc.h>
54 #include <sys/procdesc.h>
55 #include <sys/pioctl.h>
56 #include <sys/jail.h>
57 #include <sys/tty.h>
58 #include <sys/wait.h>
59 #include <sys/vmmeter.h>
60 #include <sys/vnode.h>
61 #include <sys/racct.h>
62 #include <sys/resourcevar.h>
63 #include <sys/sbuf.h>
64 #include <sys/signalvar.h>
65 #include <sys/sched.h>
66 #include <sys/sx.h>
67 #include <sys/syscallsubr.h>
68 #include <sys/syslog.h>
69 #include <sys/ptrace.h>
70 #include <sys/acct.h>		/* for acct_process() function prototype */
71 #include <sys/filedesc.h>
72 #include <sys/sdt.h>
73 #include <sys/shm.h>
74 #include <sys/sem.h>
75 #include <sys/umtx.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <security/audit/audit.h>
81 #include <security/mac/mac_framework.h>
82 
83 #include <vm/vm.h>
84 #include <vm/vm_extern.h>
85 #include <vm/vm_param.h>
86 #include <vm/pmap.h>
87 #include <vm/vm_map.h>
88 #include <vm/vm_page.h>
89 #include <vm/uma.h>
90 
91 #ifdef KDTRACE_HOOKS
92 #include <sys/dtrace_bsd.h>
93 dtrace_execexit_func_t	dtrace_fasttrap_exit;
94 #endif
95 
96 SDT_PROVIDER_DECLARE(proc);
97 SDT_PROBE_DEFINE1(proc, , , exit, "int");
98 
99 /* Hook for NFS teardown procedure. */
100 void (*nlminfo_release_p)(struct proc *p);
101 
102 EVENTHANDLER_LIST_DECLARE(process_exit);
103 
104 struct proc *
105 proc_realparent(struct proc *child)
106 {
107 	struct proc *p, *parent;
108 
109 	sx_assert(&proctree_lock, SX_LOCKED);
110 	if ((child->p_treeflag & P_TREE_ORPHANED) == 0)
111 		return (child->p_pptr->p_pid == child->p_oppid ?
112 			    child->p_pptr : initproc);
113 	for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
114 		/* Cannot use LIST_PREV(), since the list head is not known. */
115 		p = __containerof(p->p_orphan.le_prev, struct proc,
116 		    p_orphan.le_next);
117 		KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
118 		    ("missing P_ORPHAN %p", p));
119 	}
120 	parent = __containerof(p->p_orphan.le_prev, struct proc,
121 	    p_orphans.lh_first);
122 	return (parent);
123 }
124 
125 void
126 reaper_abandon_children(struct proc *p, bool exiting)
127 {
128 	struct proc *p1, *p2, *ptmp;
129 
130 	sx_assert(&proctree_lock, SX_LOCKED);
131 	KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
132 	if ((p->p_treeflag & P_TREE_REAPER) == 0)
133 		return;
134 	p1 = p->p_reaper;
135 	LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
136 		LIST_REMOVE(p2, p_reapsibling);
137 		p2->p_reaper = p1;
138 		p2->p_reapsubtree = p->p_reapsubtree;
139 		LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
140 		if (exiting && p2->p_pptr == p) {
141 			PROC_LOCK(p2);
142 			proc_reparent(p2, p1, true);
143 			PROC_UNLOCK(p2);
144 		}
145 	}
146 	KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
147 	p->p_treeflag &= ~P_TREE_REAPER;
148 }
149 
150 static void
151 reaper_clear(struct proc *p)
152 {
153 	struct proc *p1;
154 	bool clear;
155 
156 	sx_assert(&proctree_lock, SX_LOCKED);
157 	LIST_REMOVE(p, p_reapsibling);
158 	if (p->p_reapsubtree == 1)
159 		return;
160 	clear = true;
161 	LIST_FOREACH(p1, &p->p_reaper->p_reaplist, p_reapsibling) {
162 		if (p1->p_reapsubtree == p->p_reapsubtree) {
163 			clear = false;
164 			break;
165 		}
166 	}
167 	if (clear)
168 		proc_id_clear(PROC_ID_REAP, p->p_reapsubtree);
169 }
170 
171 static void
172 clear_orphan(struct proc *p)
173 {
174 	struct proc *p1;
175 
176 	sx_assert(&proctree_lock, SA_XLOCKED);
177 	if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
178 		return;
179 	if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
180 		p1 = LIST_NEXT(p, p_orphan);
181 		if (p1 != NULL)
182 			p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
183 		p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
184 	}
185 	LIST_REMOVE(p, p_orphan);
186 	p->p_treeflag &= ~P_TREE_ORPHANED;
187 }
188 
189 /*
190  * exit -- death of process.
191  */
192 void
193 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
194 {
195 
196 	exit1(td, uap->rval, 0);
197 	/* NOTREACHED */
198 }
199 
200 /*
201  * Exit: deallocate address space and other resources, change proc state to
202  * zombie, and unlink proc from allproc and parent's lists.  Save exit status
203  * and rusage for wait().  Check for child processes and orphan them.
204  */
205 void
206 exit1(struct thread *td, int rval, int signo)
207 {
208 	struct proc *p, *nq, *q, *t;
209 	struct thread *tdt;
210 	ksiginfo_t *ksi, *ksi1;
211 	int signal_parent;
212 
213 	mtx_assert(&Giant, MA_NOTOWNED);
214 	KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
215 
216 	p = td->td_proc;
217 	/*
218 	 * XXX in case we're rebooting we just let init die in order to
219 	 * work around an unsolved stack overflow seen very late during
220 	 * shutdown on sparc64 when the gmirror worker process exists.
221 	 */
222 	if (p == initproc && rebooting == 0) {
223 		printf("init died (signal %d, exit %d)\n", signo, rval);
224 		panic("Going nowhere without my init!");
225 	}
226 
227 	/*
228 	 * Deref SU mp, since the thread does not return to userspace.
229 	 */
230 	td_softdep_cleanup(td);
231 
232 	/*
233 	 * MUST abort all other threads before proceeding past here.
234 	 */
235 	PROC_LOCK(p);
236 	/*
237 	 * First check if some other thread or external request got
238 	 * here before us.  If so, act appropriately: exit or suspend.
239 	 * We must ensure that stop requests are handled before we set
240 	 * P_WEXIT.
241 	 */
242 	thread_suspend_check(0);
243 	while (p->p_flag & P_HADTHREADS) {
244 		/*
245 		 * Kill off the other threads. This requires
246 		 * some co-operation from other parts of the kernel
247 		 * so it may not be instantaneous.  With this state set
248 		 * any thread entering the kernel from userspace will
249 		 * thread_exit() in trap().  Any thread attempting to
250 		 * sleep will return immediately with EINTR or EWOULDBLOCK
251 		 * which will hopefully force them to back out to userland
252 		 * freeing resources as they go.  Any thread attempting
253 		 * to return to userland will thread_exit() from userret().
254 		 * thread_exit() will unsuspend us when the last of the
255 		 * other threads exits.
256 		 * If there is already a thread singler after resumption,
257 		 * calling thread_single will fail; in that case, we just
258 		 * re-check all suspension request, the thread should
259 		 * either be suspended there or exit.
260 		 */
261 		if (!thread_single(p, SINGLE_EXIT))
262 			/*
263 			 * All other activity in this process is now
264 			 * stopped.  Threading support has been turned
265 			 * off.
266 			 */
267 			break;
268 		/*
269 		 * Recheck for new stop or suspend requests which
270 		 * might appear while process lock was dropped in
271 		 * thread_single().
272 		 */
273 		thread_suspend_check(0);
274 	}
275 	KASSERT(p->p_numthreads == 1,
276 	    ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
277 	racct_sub(p, RACCT_NTHR, 1);
278 
279 	/* Let event handler change exit status */
280 	p->p_xexit = rval;
281 	p->p_xsig = signo;
282 
283 	/*
284 	 * Wakeup anyone in procfs' PIOCWAIT.  They should have a hold
285 	 * on our vmspace, so we should block below until they have
286 	 * released their reference to us.  Note that if they have
287 	 * requested S_EXIT stops we will block here until they ack
288 	 * via PIOCCONT.
289 	 */
290 	_STOPEVENT(p, S_EXIT, 0);
291 
292 	/*
293 	 * Ignore any pending request to stop due to a stop signal.
294 	 * Once P_WEXIT is set, future requests will be ignored as
295 	 * well.
296 	 */
297 	p->p_flag &= ~P_STOPPED_SIG;
298 	KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
299 
300 	/*
301 	 * Note that we are exiting and do another wakeup of anyone in
302 	 * PIOCWAIT in case they aren't listening for S_EXIT stops or
303 	 * decided to wait again after we told them we are exiting.
304 	 */
305 	p->p_flag |= P_WEXIT;
306 	wakeup(&p->p_stype);
307 
308 	/*
309 	 * Wait for any processes that have a hold on our vmspace to
310 	 * release their reference.
311 	 */
312 	while (p->p_lock > 0)
313 		msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
314 
315 	PROC_UNLOCK(p);
316 	/* Drain the limit callout while we don't have the proc locked */
317 	callout_drain(&p->p_limco);
318 
319 #ifdef AUDIT
320 	/*
321 	 * The Sun BSM exit token contains two components: an exit status as
322 	 * passed to exit(), and a return value to indicate what sort of exit
323 	 * it was.  The exit status is WEXITSTATUS(rv), but it's not clear
324 	 * what the return value is.
325 	 */
326 	AUDIT_ARG_EXIT(rval, 0);
327 	AUDIT_SYSCALL_EXIT(0, td);
328 #endif
329 
330 	/* Are we a task leader with peers? */
331 	if (p->p_peers != NULL && p == p->p_leader) {
332 		mtx_lock(&ppeers_lock);
333 		q = p->p_peers;
334 		while (q != NULL) {
335 			PROC_LOCK(q);
336 			kern_psignal(q, SIGKILL);
337 			PROC_UNLOCK(q);
338 			q = q->p_peers;
339 		}
340 		while (p->p_peers != NULL)
341 			msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
342 		mtx_unlock(&ppeers_lock);
343 	}
344 
345 	/*
346 	 * Check if any loadable modules need anything done at process exit.
347 	 * E.g. SYSV IPC stuff.
348 	 * Event handler could change exit status.
349 	 * XXX what if one of these generates an error?
350 	 */
351 	EVENTHANDLER_DIRECT_INVOKE(process_exit, p);
352 
353 	/*
354 	 * If parent is waiting for us to exit or exec,
355 	 * P_PPWAIT is set; we will wakeup the parent below.
356 	 */
357 	PROC_LOCK(p);
358 	stopprofclock(p);
359 	p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
360 	p->p_ptevents = 0;
361 
362 	/*
363 	 * Stop the real interval timer.  If the handler is currently
364 	 * executing, prevent it from rearming itself and let it finish.
365 	 */
366 	if (timevalisset(&p->p_realtimer.it_value) &&
367 	    _callout_stop_safe(&p->p_itcallout, CS_EXECUTING, NULL) == 0) {
368 		timevalclear(&p->p_realtimer.it_interval);
369 		msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
370 		KASSERT(!timevalisset(&p->p_realtimer.it_value),
371 		    ("realtime timer is still armed"));
372 	}
373 
374 	PROC_UNLOCK(p);
375 
376 	umtx_thread_exit(td);
377 
378 	/*
379 	 * Reset any sigio structures pointing to us as a result of
380 	 * F_SETOWN with our pid.
381 	 */
382 	funsetownlst(&p->p_sigiolst);
383 
384 	/*
385 	 * If this process has an nlminfo data area (for lockd), release it
386 	 */
387 	if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
388 		(*nlminfo_release_p)(p);
389 
390 	/*
391 	 * Close open files and release open-file table.
392 	 * This may block!
393 	 */
394 	fdescfree(td);
395 
396 	/*
397 	 * If this thread tickled GEOM, we need to wait for the giggling to
398 	 * stop before we return to userland
399 	 */
400 	if (td->td_pflags & TDP_GEOM)
401 		g_waitidle();
402 
403 	/*
404 	 * Remove ourself from our leader's peer list and wake our leader.
405 	 */
406 	if (p->p_leader->p_peers != NULL) {
407 		mtx_lock(&ppeers_lock);
408 		if (p->p_leader->p_peers != NULL) {
409 			q = p->p_leader;
410 			while (q->p_peers != p)
411 				q = q->p_peers;
412 			q->p_peers = p->p_peers;
413 			wakeup(p->p_leader);
414 		}
415 		mtx_unlock(&ppeers_lock);
416 	}
417 
418 	vmspace_exit(td);
419 	killjobc();
420 	(void)acct_process(td);
421 
422 #ifdef KTRACE
423 	ktrprocexit(td);
424 #endif
425 	/*
426 	 * Release reference to text vnode
427 	 */
428 	if (p->p_textvp != NULL) {
429 		vrele(p->p_textvp);
430 		p->p_textvp = NULL;
431 	}
432 
433 	/*
434 	 * Release our limits structure.
435 	 */
436 	lim_free(p->p_limit);
437 	p->p_limit = NULL;
438 
439 	tidhash_remove(td);
440 
441 	/*
442 	 * Call machine-dependent code to release any
443 	 * machine-dependent resources other than the address space.
444 	 * The address space is released by "vmspace_exitfree(p)" in
445 	 * vm_waitproc().
446 	 */
447 	cpu_exit(td);
448 
449 	WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
450 
451 	/*
452 	 * Move proc from allproc queue to zombproc.
453 	 */
454 	sx_xlock(&allproc_lock);
455 	sx_xlock(&zombproc_lock);
456 	LIST_REMOVE(p, p_list);
457 	LIST_INSERT_HEAD(&zombproc, p, p_list);
458 	sx_xunlock(&zombproc_lock);
459 	sx_xunlock(&allproc_lock);
460 
461 	sx_xlock(&proctree_lock);
462 
463 	/*
464 	 * Reparent all children processes:
465 	 * - traced ones to the original parent (or init if we are that parent)
466 	 * - the rest to init
467 	 */
468 	q = LIST_FIRST(&p->p_children);
469 	if (q != NULL)		/* only need this if any child is S_ZOMB */
470 		wakeup(q->p_reaper);
471 	for (; q != NULL; q = nq) {
472 		nq = LIST_NEXT(q, p_sibling);
473 		ksi = ksiginfo_alloc(TRUE);
474 		PROC_LOCK(q);
475 		q->p_sigparent = SIGCHLD;
476 
477 		if (!(q->p_flag & P_TRACED)) {
478 			proc_reparent(q, q->p_reaper, true);
479 			if (q->p_state == PRS_ZOMBIE) {
480 				/*
481 				 * Inform reaper about the reparented
482 				 * zombie, since wait(2) has something
483 				 * new to report.  Guarantee queueing
484 				 * of the SIGCHLD signal, similar to
485 				 * the _exit() behaviour, by providing
486 				 * our ksiginfo.  Ksi is freed by the
487 				 * signal delivery.
488 				 */
489 				if (q->p_ksi == NULL) {
490 					ksi1 = NULL;
491 				} else {
492 					ksiginfo_copy(q->p_ksi, ksi);
493 					ksi->ksi_flags |= KSI_INS;
494 					ksi1 = ksi;
495 					ksi = NULL;
496 				}
497 				PROC_LOCK(q->p_reaper);
498 				pksignal(q->p_reaper, SIGCHLD, ksi1);
499 				PROC_UNLOCK(q->p_reaper);
500 			} else if (q->p_pdeathsig > 0) {
501 				/*
502 				 * The child asked to received a signal
503 				 * when we exit.
504 				 */
505 				kern_psignal(q, q->p_pdeathsig);
506 			}
507 		} else {
508 			/*
509 			 * Traced processes are killed since their existence
510 			 * means someone is screwing up.
511 			 */
512 			t = proc_realparent(q);
513 			if (t == p) {
514 				proc_reparent(q, q->p_reaper, true);
515 			} else {
516 				PROC_LOCK(t);
517 				proc_reparent(q, t, true);
518 				PROC_UNLOCK(t);
519 			}
520 			/*
521 			 * Since q was found on our children list, the
522 			 * proc_reparent() call moved q to the orphan
523 			 * list due to present P_TRACED flag. Clear
524 			 * orphan link for q now while q is locked.
525 			 */
526 			clear_orphan(q);
527 			q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
528 			q->p_flag2 &= ~P2_PTRACE_FSTP;
529 			q->p_ptevents = 0;
530 			FOREACH_THREAD_IN_PROC(q, tdt) {
531 				tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
532 				    TDB_FSTP);
533 			}
534 			kern_psignal(q, SIGKILL);
535 		}
536 		PROC_UNLOCK(q);
537 		if (ksi != NULL)
538 			ksiginfo_free(ksi);
539 	}
540 
541 	/*
542 	 * Also get rid of our orphans.
543 	 */
544 	while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
545 		PROC_LOCK(q);
546 		KASSERT(q->p_oppid == p->p_pid,
547 		    ("orphan %p of %p has unexpected oppid %d", q, p,
548 		    q->p_oppid));
549 		q->p_oppid = q->p_reaper->p_pid;
550 
551 		/*
552 		 * If we are the real parent of this process
553 		 * but it has been reparented to a debugger, then
554 		 * check if it asked for a signal when we exit.
555 		 */
556 		if (q->p_pdeathsig > 0)
557 			kern_psignal(q, q->p_pdeathsig);
558 		CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
559 		    q->p_pid);
560 		clear_orphan(q);
561 		PROC_UNLOCK(q);
562 	}
563 
564 #ifdef KDTRACE_HOOKS
565 	if (SDT_PROBES_ENABLED()) {
566 		int reason = CLD_EXITED;
567 		if (WCOREDUMP(signo))
568 			reason = CLD_DUMPED;
569 		else if (WIFSIGNALED(signo))
570 			reason = CLD_KILLED;
571 		SDT_PROBE1(proc, , , exit, reason);
572 	}
573 #endif
574 
575 	/* Save exit status. */
576 	PROC_LOCK(p);
577 	p->p_xthread = td;
578 
579 #ifdef KDTRACE_HOOKS
580 	/*
581 	 * Tell the DTrace fasttrap provider about the exit if it
582 	 * has declared an interest.
583 	 */
584 	if (dtrace_fasttrap_exit)
585 		dtrace_fasttrap_exit(p);
586 #endif
587 
588 	/*
589 	 * Notify interested parties of our demise.
590 	 */
591 	KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
592 
593 	/*
594 	 * If this is a process with a descriptor, we may not need to deliver
595 	 * a signal to the parent.  proctree_lock is held over
596 	 * procdesc_exit() to serialize concurrent calls to close() and
597 	 * exit().
598 	 */
599 	signal_parent = 0;
600 	if (p->p_procdesc == NULL || procdesc_exit(p)) {
601 		/*
602 		 * Notify parent that we're gone.  If parent has the
603 		 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
604 		 * notify process 1 instead (and hope it will handle this
605 		 * situation).
606 		 */
607 		PROC_LOCK(p->p_pptr);
608 		mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
609 		if (p->p_pptr->p_sigacts->ps_flag &
610 		    (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
611 			struct proc *pp;
612 
613 			mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
614 			pp = p->p_pptr;
615 			PROC_UNLOCK(pp);
616 			proc_reparent(p, p->p_reaper, true);
617 			p->p_sigparent = SIGCHLD;
618 			PROC_LOCK(p->p_pptr);
619 
620 			/*
621 			 * Notify parent, so in case he was wait(2)ing or
622 			 * executing waitpid(2) with our pid, he will
623 			 * continue.
624 			 */
625 			wakeup(pp);
626 		} else
627 			mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
628 
629 		if (p->p_pptr == p->p_reaper || p->p_pptr == initproc) {
630 			signal_parent = 1;
631 		} else if (p->p_sigparent != 0) {
632 			if (p->p_sigparent == SIGCHLD) {
633 				signal_parent = 1;
634 			} else { /* LINUX thread */
635 				signal_parent = 2;
636 			}
637 		}
638 	} else
639 		PROC_LOCK(p->p_pptr);
640 	sx_xunlock(&proctree_lock);
641 
642 	if (signal_parent == 1) {
643 		childproc_exited(p);
644 	} else if (signal_parent == 2) {
645 		kern_psignal(p->p_pptr, p->p_sigparent);
646 	}
647 
648 	/* Tell the prison that we are gone. */
649 	prison_proc_free(p->p_ucred->cr_prison);
650 
651 	/*
652 	 * The state PRS_ZOMBIE prevents other proesses from sending
653 	 * signal to the process, to avoid memory leak, we free memory
654 	 * for signal queue at the time when the state is set.
655 	 */
656 	sigqueue_flush(&p->p_sigqueue);
657 	sigqueue_flush(&td->td_sigqueue);
658 
659 	/*
660 	 * We have to wait until after acquiring all locks before
661 	 * changing p_state.  We need to avoid all possible context
662 	 * switches (including ones from blocking on a mutex) while
663 	 * marked as a zombie.  We also have to set the zombie state
664 	 * before we release the parent process' proc lock to avoid
665 	 * a lost wakeup.  So, we first call wakeup, then we grab the
666 	 * sched lock, update the state, and release the parent process'
667 	 * proc lock.
668 	 */
669 	wakeup(p->p_pptr);
670 	cv_broadcast(&p->p_pwait);
671 	sched_exit(p->p_pptr, td);
672 	PROC_SLOCK(p);
673 	p->p_state = PRS_ZOMBIE;
674 	PROC_UNLOCK(p->p_pptr);
675 
676 	/*
677 	 * Save our children's rusage information in our exit rusage.
678 	 */
679 	PROC_STATLOCK(p);
680 	ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
681 	PROC_STATUNLOCK(p);
682 
683 	/*
684 	 * Make sure the scheduler takes this thread out of its tables etc.
685 	 * This will also release this thread's reference to the ucred.
686 	 * Other thread parts to release include pcb bits and such.
687 	 */
688 	thread_exit();
689 }
690 
691 
692 #ifndef _SYS_SYSPROTO_H_
693 struct abort2_args {
694 	char *why;
695 	int nargs;
696 	void **args;
697 };
698 #endif
699 
700 int
701 sys_abort2(struct thread *td, struct abort2_args *uap)
702 {
703 	struct proc *p = td->td_proc;
704 	struct sbuf *sb;
705 	void *uargs[16];
706 	int error, i, sig;
707 
708 	/*
709 	 * Do it right now so we can log either proper call of abort2(), or
710 	 * note, that invalid argument was passed. 512 is big enough to
711 	 * handle 16 arguments' descriptions with additional comments.
712 	 */
713 	sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
714 	sbuf_clear(sb);
715 	sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
716 	    p->p_comm, p->p_pid, td->td_ucred->cr_uid);
717 	/*
718 	 * Since we can't return from abort2(), send SIGKILL in cases, where
719 	 * abort2() was called improperly
720 	 */
721 	sig = SIGKILL;
722 	/* Prevent from DoSes from user-space. */
723 	if (uap->nargs < 0 || uap->nargs > 16)
724 		goto out;
725 	if (uap->nargs > 0) {
726 		if (uap->args == NULL)
727 			goto out;
728 		error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
729 		if (error != 0)
730 			goto out;
731 	}
732 	/*
733 	 * Limit size of 'reason' string to 128. Will fit even when
734 	 * maximal number of arguments was chosen to be logged.
735 	 */
736 	if (uap->why != NULL) {
737 		error = sbuf_copyin(sb, uap->why, 128);
738 		if (error < 0)
739 			goto out;
740 	} else {
741 		sbuf_printf(sb, "(null)");
742 	}
743 	if (uap->nargs > 0) {
744 		sbuf_printf(sb, "(");
745 		for (i = 0;i < uap->nargs; i++)
746 			sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
747 		sbuf_printf(sb, ")");
748 	}
749 	/*
750 	 * Final stage: arguments were proper, string has been
751 	 * successfully copied from userspace, and copying pointers
752 	 * from user-space succeed.
753 	 */
754 	sig = SIGABRT;
755 out:
756 	if (sig == SIGKILL) {
757 		sbuf_trim(sb);
758 		sbuf_printf(sb, " (Reason text inaccessible)");
759 	}
760 	sbuf_cat(sb, "\n");
761 	sbuf_finish(sb);
762 	log(LOG_INFO, "%s", sbuf_data(sb));
763 	sbuf_delete(sb);
764 	exit1(td, 0, sig);
765 	return (0);
766 }
767 
768 
769 #ifdef COMPAT_43
770 /*
771  * The dirty work is handled by kern_wait().
772  */
773 int
774 owait(struct thread *td, struct owait_args *uap __unused)
775 {
776 	int error, status;
777 
778 	error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
779 	if (error == 0)
780 		td->td_retval[1] = status;
781 	return (error);
782 }
783 #endif /* COMPAT_43 */
784 
785 /*
786  * The dirty work is handled by kern_wait().
787  */
788 int
789 sys_wait4(struct thread *td, struct wait4_args *uap)
790 {
791 	struct rusage ru, *rup;
792 	int error, status;
793 
794 	if (uap->rusage != NULL)
795 		rup = &ru;
796 	else
797 		rup = NULL;
798 	error = kern_wait(td, uap->pid, &status, uap->options, rup);
799 	if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
800 		error = copyout(&status, uap->status, sizeof(status));
801 	if (uap->rusage != NULL && error == 0 && td->td_retval[0] != 0)
802 		error = copyout(&ru, uap->rusage, sizeof(struct rusage));
803 	return (error);
804 }
805 
806 int
807 sys_wait6(struct thread *td, struct wait6_args *uap)
808 {
809 	struct __wrusage wru, *wrup;
810 	siginfo_t si, *sip;
811 	idtype_t idtype;
812 	id_t id;
813 	int error, status;
814 
815 	idtype = uap->idtype;
816 	id = uap->id;
817 
818 	if (uap->wrusage != NULL)
819 		wrup = &wru;
820 	else
821 		wrup = NULL;
822 
823 	if (uap->info != NULL) {
824 		sip = &si;
825 		bzero(sip, sizeof(*sip));
826 	} else
827 		sip = NULL;
828 
829 	/*
830 	 *  We expect all callers of wait6() to know about WEXITED and
831 	 *  WTRAPPED.
832 	 */
833 	error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
834 
835 	if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
836 		error = copyout(&status, uap->status, sizeof(status));
837 	if (uap->wrusage != NULL && error == 0 && td->td_retval[0] != 0)
838 		error = copyout(&wru, uap->wrusage, sizeof(wru));
839 	if (uap->info != NULL && error == 0)
840 		error = copyout(&si, uap->info, sizeof(si));
841 	return (error);
842 }
843 
844 /*
845  * Reap the remains of a zombie process and optionally return status and
846  * rusage.  Asserts and will release both the proctree_lock and the process
847  * lock as part of its work.
848  */
849 void
850 proc_reap(struct thread *td, struct proc *p, int *status, int options)
851 {
852 	struct proc *q, *t;
853 
854 	sx_assert(&proctree_lock, SA_XLOCKED);
855 	PROC_LOCK_ASSERT(p, MA_OWNED);
856 	KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
857 
858 	mtx_spin_wait_unlocked(&p->p_slock);
859 
860 	q = td->td_proc;
861 
862 	if (status)
863 		*status = KW_EXITCODE(p->p_xexit, p->p_xsig);
864 	if (options & WNOWAIT) {
865 		/*
866 		 *  Only poll, returning the status.  Caller does not wish to
867 		 * release the proc struct just yet.
868 		 */
869 		PROC_UNLOCK(p);
870 		sx_xunlock(&proctree_lock);
871 		return;
872 	}
873 
874 	PROC_LOCK(q);
875 	sigqueue_take(p->p_ksi);
876 	PROC_UNLOCK(q);
877 
878 	/*
879 	 * If we got the child via a ptrace 'attach', we need to give it back
880 	 * to the old parent.
881 	 */
882 	if (p->p_oppid != p->p_pptr->p_pid) {
883 		PROC_UNLOCK(p);
884 		t = proc_realparent(p);
885 		PROC_LOCK(t);
886 		PROC_LOCK(p);
887 		CTR2(KTR_PTRACE,
888 		    "wait: traced child %d moved back to parent %d", p->p_pid,
889 		    t->p_pid);
890 		proc_reparent(p, t, false);
891 		PROC_UNLOCK(p);
892 		pksignal(t, SIGCHLD, p->p_ksi);
893 		wakeup(t);
894 		cv_broadcast(&p->p_pwait);
895 		PROC_UNLOCK(t);
896 		sx_xunlock(&proctree_lock);
897 		return;
898 	}
899 	PROC_UNLOCK(p);
900 
901 	/*
902 	 * Remove other references to this process to ensure we have an
903 	 * exclusive reference.
904 	 */
905 	sx_xlock(&zombproc_lock);
906 	LIST_REMOVE(p, p_list);	/* off zombproc */
907 	sx_xunlock(&zombproc_lock);
908 	sx_xlock(PIDHASHLOCK(p->p_pid));
909 	LIST_REMOVE(p, p_hash);
910 	sx_xunlock(PIDHASHLOCK(p->p_pid));
911 	LIST_REMOVE(p, p_sibling);
912 	reaper_abandon_children(p, true);
913 	reaper_clear(p);
914 	proc_id_clear(PROC_ID_PID, p->p_pid);
915 	PROC_LOCK(p);
916 	clear_orphan(p);
917 	PROC_UNLOCK(p);
918 	leavepgrp(p);
919 	if (p->p_procdesc != NULL)
920 		procdesc_reap(p);
921 	sx_xunlock(&proctree_lock);
922 
923 	PROC_LOCK(p);
924 	knlist_detach(p->p_klist);
925 	p->p_klist = NULL;
926 	PROC_UNLOCK(p);
927 
928 	/*
929 	 * Removal from allproc list and process group list paired with
930 	 * PROC_LOCK which was executed during that time should guarantee
931 	 * nothing can reach this process anymore. As such further locking
932 	 * is unnecessary.
933 	 */
934 	p->p_xexit = p->p_xsig = 0;		/* XXX: why? */
935 
936 	PROC_LOCK(q);
937 	ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
938 	PROC_UNLOCK(q);
939 
940 	/*
941 	 * Decrement the count of procs running with this uid.
942 	 */
943 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
944 
945 	/*
946 	 * Destroy resource accounting information associated with the process.
947 	 */
948 #ifdef RACCT
949 	if (racct_enable) {
950 		PROC_LOCK(p);
951 		racct_sub(p, RACCT_NPROC, 1);
952 		PROC_UNLOCK(p);
953 	}
954 #endif
955 	racct_proc_exit(p);
956 
957 	/*
958 	 * Free credentials, arguments, and sigacts.
959 	 */
960 	crfree(p->p_ucred);
961 	proc_set_cred(p, NULL);
962 	pargs_drop(p->p_args);
963 	p->p_args = NULL;
964 	sigacts_free(p->p_sigacts);
965 	p->p_sigacts = NULL;
966 
967 	/*
968 	 * Do any thread-system specific cleanups.
969 	 */
970 	thread_wait(p);
971 
972 	/*
973 	 * Give vm and machine-dependent layer a chance to free anything that
974 	 * cpu_exit couldn't release while still running in process context.
975 	 */
976 	vm_waitproc(p);
977 #ifdef MAC
978 	mac_proc_destroy(p);
979 #endif
980 
981 	KASSERT(FIRST_THREAD_IN_PROC(p),
982 	    ("proc_reap: no residual thread!"));
983 	uma_zfree(proc_zone, p);
984 	atomic_add_int(&nprocs, -1);
985 }
986 
987 static int
988 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
989     int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
990     int check_only)
991 {
992 	struct rusage *rup;
993 
994 	sx_assert(&proctree_lock, SA_XLOCKED);
995 
996 	PROC_LOCK(p);
997 
998 	switch (idtype) {
999 	case P_ALL:
1000 		if (p->p_procdesc != NULL) {
1001 			PROC_UNLOCK(p);
1002 			return (0);
1003 		}
1004 		break;
1005 	case P_PID:
1006 		if (p->p_pid != (pid_t)id) {
1007 			PROC_UNLOCK(p);
1008 			return (0);
1009 		}
1010 		break;
1011 	case P_PGID:
1012 		if (p->p_pgid != (pid_t)id) {
1013 			PROC_UNLOCK(p);
1014 			return (0);
1015 		}
1016 		break;
1017 	case P_SID:
1018 		if (p->p_session->s_sid != (pid_t)id) {
1019 			PROC_UNLOCK(p);
1020 			return (0);
1021 		}
1022 		break;
1023 	case P_UID:
1024 		if (p->p_ucred->cr_uid != (uid_t)id) {
1025 			PROC_UNLOCK(p);
1026 			return (0);
1027 		}
1028 		break;
1029 	case P_GID:
1030 		if (p->p_ucred->cr_gid != (gid_t)id) {
1031 			PROC_UNLOCK(p);
1032 			return (0);
1033 		}
1034 		break;
1035 	case P_JAILID:
1036 		if (p->p_ucred->cr_prison->pr_id != (int)id) {
1037 			PROC_UNLOCK(p);
1038 			return (0);
1039 		}
1040 		break;
1041 	/*
1042 	 * It seems that the thread structures get zeroed out
1043 	 * at process exit.  This makes it impossible to
1044 	 * support P_SETID, P_CID or P_CPUID.
1045 	 */
1046 	default:
1047 		PROC_UNLOCK(p);
1048 		return (0);
1049 	}
1050 
1051 	if (p_canwait(td, p)) {
1052 		PROC_UNLOCK(p);
1053 		return (0);
1054 	}
1055 
1056 	if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1057 		PROC_UNLOCK(p);
1058 		return (0);
1059 	}
1060 
1061 	/*
1062 	 * This special case handles a kthread spawned by linux_clone
1063 	 * (see linux_misc.c).  The linux_wait4 and linux_waitpid
1064 	 * functions need to be able to distinguish between waiting
1065 	 * on a process and waiting on a thread.  It is a thread if
1066 	 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1067 	 * signifies we want to wait for threads and not processes.
1068 	 */
1069 	if ((p->p_sigparent != SIGCHLD) ^
1070 	    ((options & WLINUXCLONE) != 0)) {
1071 		PROC_UNLOCK(p);
1072 		return (0);
1073 	}
1074 
1075 	if (siginfo != NULL) {
1076 		bzero(siginfo, sizeof(*siginfo));
1077 		siginfo->si_errno = 0;
1078 
1079 		/*
1080 		 * SUSv4 requires that the si_signo value is always
1081 		 * SIGCHLD. Obey it despite the rfork(2) interface
1082 		 * allows to request other signal for child exit
1083 		 * notification.
1084 		 */
1085 		siginfo->si_signo = SIGCHLD;
1086 
1087 		/*
1088 		 *  This is still a rough estimate.  We will fix the
1089 		 *  cases TRAPPED, STOPPED, and CONTINUED later.
1090 		 */
1091 		if (WCOREDUMP(p->p_xsig)) {
1092 			siginfo->si_code = CLD_DUMPED;
1093 			siginfo->si_status = WTERMSIG(p->p_xsig);
1094 		} else if (WIFSIGNALED(p->p_xsig)) {
1095 			siginfo->si_code = CLD_KILLED;
1096 			siginfo->si_status = WTERMSIG(p->p_xsig);
1097 		} else {
1098 			siginfo->si_code = CLD_EXITED;
1099 			siginfo->si_status = p->p_xexit;
1100 		}
1101 
1102 		siginfo->si_pid = p->p_pid;
1103 		siginfo->si_uid = p->p_ucred->cr_uid;
1104 
1105 		/*
1106 		 * The si_addr field would be useful additional
1107 		 * detail, but apparently the PC value may be lost
1108 		 * when we reach this point.  bzero() above sets
1109 		 * siginfo->si_addr to NULL.
1110 		 */
1111 	}
1112 
1113 	/*
1114 	 * There should be no reason to limit resources usage info to
1115 	 * exited processes only.  A snapshot about any resources used
1116 	 * by a stopped process may be exactly what is needed.
1117 	 */
1118 	if (wrusage != NULL) {
1119 		rup = &wrusage->wru_self;
1120 		*rup = p->p_ru;
1121 		PROC_STATLOCK(p);
1122 		calcru(p, &rup->ru_utime, &rup->ru_stime);
1123 		PROC_STATUNLOCK(p);
1124 
1125 		rup = &wrusage->wru_children;
1126 		*rup = p->p_stats->p_cru;
1127 		calccru(p, &rup->ru_utime, &rup->ru_stime);
1128 	}
1129 
1130 	if (p->p_state == PRS_ZOMBIE && !check_only) {
1131 		proc_reap(td, p, status, options);
1132 		return (-1);
1133 	}
1134 	return (1);
1135 }
1136 
1137 int
1138 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1139     struct rusage *rusage)
1140 {
1141 	struct __wrusage wru, *wrup;
1142 	idtype_t idtype;
1143 	id_t id;
1144 	int ret;
1145 
1146 	/*
1147 	 * Translate the special pid values into the (idtype, pid)
1148 	 * pair for kern_wait6.  The WAIT_MYPGRP case is handled by
1149 	 * kern_wait6() on its own.
1150 	 */
1151 	if (pid == WAIT_ANY) {
1152 		idtype = P_ALL;
1153 		id = 0;
1154 	} else if (pid < 0) {
1155 		idtype = P_PGID;
1156 		id = (id_t)-pid;
1157 	} else {
1158 		idtype = P_PID;
1159 		id = (id_t)pid;
1160 	}
1161 
1162 	if (rusage != NULL)
1163 		wrup = &wru;
1164 	else
1165 		wrup = NULL;
1166 
1167 	/*
1168 	 * For backward compatibility we implicitly add flags WEXITED
1169 	 * and WTRAPPED here.
1170 	 */
1171 	options |= WEXITED | WTRAPPED;
1172 	ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1173 	if (rusage != NULL)
1174 		*rusage = wru.wru_self;
1175 	return (ret);
1176 }
1177 
1178 static void
1179 report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo,
1180     int *status, int options, int si_code)
1181 {
1182 	bool cont;
1183 
1184 	PROC_LOCK_ASSERT(p, MA_OWNED);
1185 	sx_assert(&proctree_lock, SA_XLOCKED);
1186 	MPASS(si_code == CLD_TRAPPED || si_code == CLD_STOPPED ||
1187 	    si_code == CLD_CONTINUED);
1188 
1189 	cont = si_code == CLD_CONTINUED;
1190 	if ((options & WNOWAIT) == 0) {
1191 		if (cont)
1192 			p->p_flag &= ~P_CONTINUED;
1193 		else
1194 			p->p_flag |= P_WAITED;
1195 		PROC_LOCK(td->td_proc);
1196 		sigqueue_take(p->p_ksi);
1197 		PROC_UNLOCK(td->td_proc);
1198 	}
1199 	sx_xunlock(&proctree_lock);
1200 	if (siginfo != NULL) {
1201 		siginfo->si_code = si_code;
1202 		siginfo->si_status = cont ? SIGCONT : p->p_xsig;
1203 	}
1204 	if (status != NULL)
1205 		*status = cont ? SIGCONT : W_STOPCODE(p->p_xsig);
1206 	PROC_UNLOCK(p);
1207 	td->td_retval[0] = p->p_pid;
1208 }
1209 
1210 int
1211 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1212     int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1213 {
1214 	struct proc *p, *q;
1215 	pid_t pid;
1216 	int error, nfound, ret;
1217 	bool report;
1218 
1219 	AUDIT_ARG_VALUE((int)idtype);	/* XXX - This is likely wrong! */
1220 	AUDIT_ARG_PID((pid_t)id);	/* XXX - This may be wrong! */
1221 	AUDIT_ARG_VALUE(options);
1222 
1223 	q = td->td_proc;
1224 
1225 	if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1226 		PROC_LOCK(q);
1227 		id = (id_t)q->p_pgid;
1228 		PROC_UNLOCK(q);
1229 		idtype = P_PGID;
1230 	}
1231 
1232 	/* If we don't know the option, just return. */
1233 	if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1234 	    WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1235 		return (EINVAL);
1236 	if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1237 		/*
1238 		 * We will be unable to find any matching processes,
1239 		 * because there are no known events to look for.
1240 		 * Prefer to return error instead of blocking
1241 		 * indefinitely.
1242 		 */
1243 		return (EINVAL);
1244 	}
1245 
1246 loop:
1247 	if (q->p_flag & P_STATCHILD) {
1248 		PROC_LOCK(q);
1249 		q->p_flag &= ~P_STATCHILD;
1250 		PROC_UNLOCK(q);
1251 	}
1252 	sx_xlock(&proctree_lock);
1253 loop_locked:
1254 	nfound = 0;
1255 	LIST_FOREACH(p, &q->p_children, p_sibling) {
1256 		pid = p->p_pid;
1257 		ret = proc_to_reap(td, p, idtype, id, status, options,
1258 		    wrusage, siginfo, 0);
1259 		if (ret == 0)
1260 			continue;
1261 		else if (ret != 1) {
1262 			td->td_retval[0] = pid;
1263 			return (0);
1264 		}
1265 
1266 		nfound++;
1267 		PROC_LOCK_ASSERT(p, MA_OWNED);
1268 
1269 		if ((options & WTRAPPED) != 0 &&
1270 		    (p->p_flag & P_TRACED) != 0) {
1271 			PROC_SLOCK(p);
1272 			report =
1273 			    ((p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) &&
1274 			    p->p_suspcount == p->p_numthreads &&
1275 			    (p->p_flag & P_WAITED) == 0);
1276 			PROC_SUNLOCK(p);
1277 			if (report) {
1278 			CTR4(KTR_PTRACE,
1279 			    "wait: returning trapped pid %d status %#x "
1280 			    "(xstat %d) xthread %d",
1281 			    p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1282 			    p->p_xthread != NULL ?
1283 			    p->p_xthread->td_tid : -1);
1284 				report_alive_proc(td, p, siginfo, status,
1285 				    options, CLD_TRAPPED);
1286 				return (0);
1287 			}
1288 		}
1289 		if ((options & WUNTRACED) != 0 &&
1290 		    (p->p_flag & P_STOPPED_SIG) != 0) {
1291 			PROC_SLOCK(p);
1292 			report = (p->p_suspcount == p->p_numthreads &&
1293 			    ((p->p_flag & P_WAITED) == 0));
1294 			PROC_SUNLOCK(p);
1295 			if (report) {
1296 				report_alive_proc(td, p, siginfo, status,
1297 				    options, CLD_STOPPED);
1298 				return (0);
1299 			}
1300 		}
1301 		if ((options & WCONTINUED) != 0 &&
1302 		    (p->p_flag & P_CONTINUED) != 0) {
1303 			report_alive_proc(td, p, siginfo, status, options,
1304 			    CLD_CONTINUED);
1305 			return (0);
1306 		}
1307 		PROC_UNLOCK(p);
1308 	}
1309 
1310 	/*
1311 	 * Look in the orphans list too, to allow the parent to
1312 	 * collect it's child exit status even if child is being
1313 	 * debugged.
1314 	 *
1315 	 * Debugger detaches from the parent upon successful
1316 	 * switch-over from parent to child.  At this point due to
1317 	 * re-parenting the parent loses the child to debugger and a
1318 	 * wait4(2) call would report that it has no children to wait
1319 	 * for.  By maintaining a list of orphans we allow the parent
1320 	 * to successfully wait until the child becomes a zombie.
1321 	 */
1322 	if (nfound == 0) {
1323 		LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1324 			ret = proc_to_reap(td, p, idtype, id, NULL, options,
1325 			    NULL, NULL, 1);
1326 			if (ret != 0) {
1327 				KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1328 				    (int)td->td_retval[0]));
1329 				PROC_UNLOCK(p);
1330 				nfound++;
1331 				break;
1332 			}
1333 		}
1334 	}
1335 	if (nfound == 0) {
1336 		sx_xunlock(&proctree_lock);
1337 		return (ECHILD);
1338 	}
1339 	if (options & WNOHANG) {
1340 		sx_xunlock(&proctree_lock);
1341 		td->td_retval[0] = 0;
1342 		return (0);
1343 	}
1344 	PROC_LOCK(q);
1345 	if (q->p_flag & P_STATCHILD) {
1346 		q->p_flag &= ~P_STATCHILD;
1347 		PROC_UNLOCK(q);
1348 		goto loop_locked;
1349 	}
1350 	sx_xunlock(&proctree_lock);
1351 	error = msleep(q, &q->p_mtx, PWAIT | PCATCH | PDROP, "wait", 0);
1352 	if (error)
1353 		return (error);
1354 	goto loop;
1355 }
1356 
1357 /*
1358  * Make process 'parent' the new parent of process 'child'.
1359  * Must be called with an exclusive hold of proctree lock.
1360  */
1361 void
1362 proc_reparent(struct proc *child, struct proc *parent, bool set_oppid)
1363 {
1364 
1365 	sx_assert(&proctree_lock, SX_XLOCKED);
1366 	PROC_LOCK_ASSERT(child, MA_OWNED);
1367 	if (child->p_pptr == parent)
1368 		return;
1369 
1370 	PROC_LOCK(child->p_pptr);
1371 	sigqueue_take(child->p_ksi);
1372 	PROC_UNLOCK(child->p_pptr);
1373 	LIST_REMOVE(child, p_sibling);
1374 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1375 
1376 	clear_orphan(child);
1377 	if (child->p_flag & P_TRACED) {
1378 		if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1379 			child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1380 			LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1381 			    p_orphan);
1382 		} else {
1383 			LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1384 			    child, p_orphan);
1385 		}
1386 		child->p_treeflag |= P_TREE_ORPHANED;
1387 	}
1388 
1389 	child->p_pptr = parent;
1390 	if (set_oppid)
1391 		child->p_oppid = parent->p_pid;
1392 }
1393