xref: /freebsd/sys/kern/kern_exit.c (revision 63f9a4cb2684a303e3eb2ffed39c03a2e2b28ae0)
1 /*
2  * Copyright (c) 1982, 1986, 1989, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)kern_exit.c	8.7 (Berkeley) 2/12/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_compat.h"
41 #include "opt_ktrace.h"
42 #include "opt_mac.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/sysproto.h>
47 #include <sys/eventhandler.h>
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
50 #include <sys/lock.h>
51 #include <sys/mutex.h>
52 #include <sys/proc.h>
53 #include <sys/pioctl.h>
54 #include <sys/tty.h>
55 #include <sys/wait.h>
56 #include <sys/vmmeter.h>
57 #include <sys/vnode.h>
58 #include <sys/resourcevar.h>
59 #include <sys/signalvar.h>
60 #include <sys/sched.h>
61 #include <sys/sx.h>
62 #include <sys/ptrace.h>
63 #include <sys/acct.h>		/* for acct_process() function prototype */
64 #include <sys/filedesc.h>
65 #include <sys/mac.h>
66 #include <sys/shm.h>
67 #include <sys/sem.h>
68 #ifdef KTRACE
69 #include <sys/ktrace.h>
70 #endif
71 
72 #include <vm/vm.h>
73 #include <vm/vm_extern.h>
74 #include <vm/vm_param.h>
75 #include <vm/pmap.h>
76 #include <vm/vm_map.h>
77 #include <vm/vm_page.h>
78 #include <vm/uma.h>
79 
80 /* Required to be non-static for SysVR4 emulator */
81 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status");
82 
83 /*
84  * exit --
85  *	Death of process.
86  *
87  * MPSAFE
88  */
89 void
90 sys_exit(struct thread *td, struct sys_exit_args *uap)
91 {
92 
93 	exit1(td, W_EXITCODE(uap->rval, 0));
94 	/* NOTREACHED */
95 }
96 
97 /*
98  * Exit: deallocate address space and other resources, change proc state
99  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
100  * status and rusage for wait().  Check for child processes and orphan them.
101  */
102 void
103 exit1(struct thread *td, int rv)
104 {
105 	struct bintime new_switchtime;
106 	struct proc *p, *nq, *q;
107 	struct tty *tp;
108 	struct vnode *ttyvp;
109 	struct vmspace *vm;
110 	struct vnode *vtmp;
111 #ifdef KTRACE
112 	struct vnode *tracevp;
113 	struct ucred *tracecred;
114 #endif
115 	struct plimit *plim;
116 	int refcnt;
117 
118 	/*
119 	 * Drop Giant if caller has it.  Eventually we should warn about
120 	 * being called with Giant held.
121 	 */
122 	while (mtx_owned(&Giant))
123 		mtx_unlock(&Giant);
124 
125 	p = td->td_proc;
126 	if (p == initproc) {
127 		printf("init died (signal %d, exit %d)\n",
128 		    WTERMSIG(rv), WEXITSTATUS(rv));
129 		panic("Going nowhere without my init!");
130 	}
131 
132 	/*
133 	 * MUST abort all other threads before proceeding past here.
134 	 */
135 	PROC_LOCK(p);
136 	if (p->p_flag & P_HADTHREADS) {
137 retry:
138 		/*
139 		 * First check if some other thread got here before us..
140 		 * if so, act apropriatly, (exit or suspend);
141 		 */
142 		thread_suspend_check(0);
143 
144 		/*
145 		 * Kill off the other threads. This requires
146 		 * some co-operation from other parts of the kernel
147 		 * so it may not be instantaneous.  With this state set
148 		 * any thread entering the kernel from userspace will
149 		 * thread_exit() in trap().  Any thread attempting to
150 		 * sleep will return immediately with EINTR or EWOULDBLOCK
151 		 * which will hopefully force them to back out to userland
152 		 * freeing resources as they go.  Any thread attempting
153 		 * to return to userland will thread_exit() from userret().
154 		 * thread_exit() will unsuspend us when the last of the
155 		 * other threads exits.
156 		 * If there is already a thread singler after resumption,
157 		 * calling thread_single will fail; in that case, we just
158 		 * re-check all suspension request, the thread should
159 		 * either be suspended there or exit.
160 		 */
161 		if (thread_single(SINGLE_EXIT))
162 			goto retry;
163 
164 		/*
165 		 * All other activity in this process is now stopped.
166 		 * Threading support has been turned off.
167 		 */
168 	}
169 
170 	p->p_flag |= P_WEXIT;
171 	PROC_UNLOCK(p);
172 
173 	/* Are we a task leader? */
174 	if (p == p->p_leader) {
175 		mtx_lock(&ppeers_lock);
176 		q = p->p_peers;
177 		while (q != NULL) {
178 			PROC_LOCK(q);
179 			psignal(q, SIGKILL);
180 			PROC_UNLOCK(q);
181 			q = q->p_peers;
182 		}
183 		while (p->p_peers != NULL)
184 			msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
185 		mtx_unlock(&ppeers_lock);
186 	}
187 
188 	PROC_LOCK(p);
189 	_STOPEVENT(p, S_EXIT, rv);
190 	wakeup(&p->p_stype);	/* Wakeup anyone in procfs' PIOCWAIT */
191 	PROC_UNLOCK(p);
192 
193 	/*
194 	 * Check if any loadable modules need anything done at process exit.
195 	 * E.g. SYSV IPC stuff
196 	 * XXX what if one of these generates an error?
197 	 */
198 	EVENTHANDLER_INVOKE(process_exit, p);
199 
200 	MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage),
201 		M_ZOMBIE, M_WAITOK);
202 	/*
203 	 * If parent is waiting for us to exit or exec,
204 	 * P_PPWAIT is set; we will wakeup the parent below.
205 	 */
206 	PROC_LOCK(p);
207 	stopprofclock(p);
208 	p->p_flag &= ~(P_TRACED | P_PPWAIT);
209 	SIGEMPTYSET(p->p_siglist);
210 	SIGEMPTYSET(td->td_siglist);
211 
212 	/*
213 	 * Stop the real interval timer.  If the handler is currently
214 	 * executing, prevent it from rearming itself and let it finish.
215 	 */
216 	if (timevalisset(&p->p_realtimer.it_value) &&
217 	    callout_stop(&p->p_itcallout) == 0) {
218 		timevalclear(&p->p_realtimer.it_interval);
219 		msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
220 		KASSERT(!timevalisset(&p->p_realtimer.it_value),
221 		    ("realtime timer is still armed"));
222 	}
223 	PROC_UNLOCK(p);
224 
225 	/*
226 	 * Reset any sigio structures pointing to us as a result of
227 	 * F_SETOWN with our pid.
228 	 */
229 	mtx_lock(&Giant);	/* XXX: not sure if needed */
230 	funsetownlst(&p->p_sigiolst);
231 
232 	/*
233 	 * Close open files and release open-file table.
234 	 * This may block!
235 	 */
236 	fdfree(td);
237 	mtx_unlock(&Giant);
238 
239 	/*
240 	 * Remove ourself from our leader's peer list and wake our leader.
241 	 */
242 	mtx_lock(&ppeers_lock);
243 	if (p->p_leader->p_peers) {
244 		q = p->p_leader;
245 		while (q->p_peers != p)
246 			q = q->p_peers;
247 		q->p_peers = p->p_peers;
248 		wakeup(p->p_leader);
249 	}
250 	mtx_unlock(&ppeers_lock);
251 
252 	/* The next two chunks should probably be moved to vmspace_exit. */
253 	vm = p->p_vmspace;
254 	/*
255 	 * Release user portion of address space.
256 	 * This releases references to vnodes,
257 	 * which could cause I/O if the file has been unlinked.
258 	 * Need to do this early enough that we can still sleep.
259 	 * Can't free the entire vmspace as the kernel stack
260 	 * may be mapped within that space also.
261 	 *
262 	 * Processes sharing the same vmspace may exit in one order, and
263 	 * get cleaned up by vmspace_exit() in a different order.  The
264 	 * last exiting process to reach this point releases as much of
265 	 * the environment as it can, and the last process cleaned up
266 	 * by vmspace_exit() (which decrements exitingcnt) cleans up the
267 	 * remainder.
268 	 */
269 	atomic_add_int(&vm->vm_exitingcnt, 1);
270 	do
271 		refcnt = vm->vm_refcnt;
272 	while (!atomic_cmpset_int(&vm->vm_refcnt, refcnt, refcnt - 1));
273 	if (refcnt == 1) {
274 		shmexit(vm);
275 		pmap_remove_pages(vmspace_pmap(vm), vm_map_min(&vm->vm_map),
276 		    vm_map_max(&vm->vm_map));
277 		(void) vm_map_remove(&vm->vm_map, vm_map_min(&vm->vm_map),
278 		    vm_map_max(&vm->vm_map));
279 	}
280 
281 	mtx_lock(&Giant);
282 	sx_xlock(&proctree_lock);
283 	if (SESS_LEADER(p)) {
284 		struct session *sp;
285 
286 		sp = p->p_session;
287 		if (sp->s_ttyvp) {
288 			/*
289 			 * Controlling process.
290 			 * Signal foreground pgrp,
291 			 * drain controlling terminal
292 			 * and revoke access to controlling terminal.
293 			 */
294 			if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
295 				tp = sp->s_ttyp;
296 				if (sp->s_ttyp->t_pgrp) {
297 					PGRP_LOCK(sp->s_ttyp->t_pgrp);
298 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
299 					PGRP_UNLOCK(sp->s_ttyp->t_pgrp);
300 				}
301 				/* XXX tp should be locked. */
302 				sx_xunlock(&proctree_lock);
303 				(void) ttywait(tp);
304 				sx_xlock(&proctree_lock);
305 				/*
306 				 * The tty could have been revoked
307 				 * if we blocked.
308 				 */
309 				if (sp->s_ttyvp) {
310 					ttyvp = sp->s_ttyvp;
311 					SESS_LOCK(p->p_session);
312 					sp->s_ttyvp = NULL;
313 					SESS_UNLOCK(p->p_session);
314 					sx_xunlock(&proctree_lock);
315 					VOP_REVOKE(ttyvp, REVOKEALL);
316 					vrele(ttyvp);
317 					sx_xlock(&proctree_lock);
318 				}
319 			}
320 			if (sp->s_ttyvp) {
321 				ttyvp = sp->s_ttyvp;
322 				SESS_LOCK(p->p_session);
323 				sp->s_ttyvp = NULL;
324 				SESS_UNLOCK(p->p_session);
325 				vrele(ttyvp);
326 			}
327 			/*
328 			 * s_ttyp is not zero'd; we use this to indicate
329 			 * that the session once had a controlling terminal.
330 			 * (for logging and informational purposes)
331 			 */
332 		}
333 		SESS_LOCK(p->p_session);
334 		sp->s_leader = NULL;
335 		SESS_UNLOCK(p->p_session);
336 	}
337 	fixjobc(p, p->p_pgrp, 0);
338 	sx_xunlock(&proctree_lock);
339 	(void)acct_process(td);
340 	mtx_unlock(&Giant);
341 #ifdef KTRACE
342 	/*
343 	 * release trace file
344 	 */
345 	PROC_LOCK(p);
346 	mtx_lock(&ktrace_mtx);
347 	p->p_traceflag = 0;	/* don't trace the vrele() */
348 	tracevp = p->p_tracevp;
349 	p->p_tracevp = NULL;
350 	tracecred = p->p_tracecred;
351 	p->p_tracecred = NULL;
352 	mtx_unlock(&ktrace_mtx);
353 	PROC_UNLOCK(p);
354 	if (tracevp != NULL) {
355 		mtx_lock(&Giant);
356 		vrele(tracevp);
357 		mtx_unlock(&Giant);
358 	}
359 	if (tracecred != NULL)
360 		crfree(tracecred);
361 #endif
362 	/*
363 	 * Release reference to text vnode
364 	 */
365 	if ((vtmp = p->p_textvp) != NULL) {
366 		p->p_textvp = NULL;
367 		mtx_lock(&Giant);
368 		vrele(vtmp);
369 		mtx_unlock(&Giant);
370 	}
371 
372 	/*
373 	 * Release our limits structure.
374 	 */
375 	PROC_LOCK(p);
376 	plim = p->p_limit;
377 	p->p_limit = NULL;
378 	PROC_UNLOCK(p);
379 	lim_free(plim);
380 
381 	/*
382 	 * Remove proc from allproc queue and pidhash chain.
383 	 * Place onto zombproc.  Unlink from parent's child list.
384 	 */
385 	sx_xlock(&allproc_lock);
386 	LIST_REMOVE(p, p_list);
387 	LIST_INSERT_HEAD(&zombproc, p, p_list);
388 	LIST_REMOVE(p, p_hash);
389 	sx_xunlock(&allproc_lock);
390 
391 	sx_xlock(&proctree_lock);
392 	q = LIST_FIRST(&p->p_children);
393 	if (q != NULL)		/* only need this if any child is S_ZOMB */
394 		wakeup(initproc);
395 	for (; q != NULL; q = nq) {
396 		nq = LIST_NEXT(q, p_sibling);
397 		PROC_LOCK(q);
398 		proc_reparent(q, initproc);
399 		q->p_sigparent = SIGCHLD;
400 		/*
401 		 * Traced processes are killed
402 		 * since their existence means someone is screwing up.
403 		 */
404 		if (q->p_flag & P_TRACED) {
405 			q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
406 			psignal(q, SIGKILL);
407 		}
408 		PROC_UNLOCK(q);
409 	}
410 
411 	/*
412 	 * Save exit status and finalize rusage info except for times,
413 	 * adding in child rusage info.
414 	 */
415 	PROC_LOCK(p);
416 	p->p_xstat = rv;
417 	p->p_xthread = td;
418 	p->p_stats->p_ru.ru_nvcsw++;
419 	*p->p_ru = p->p_stats->p_ru;
420 	ruadd(p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
421 
422 	/*
423 	 * Notify interested parties of our demise.
424 	 */
425 	KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
426 
427 	/*
428 	 * Just delete all entries in the p_klist. At this point we won't
429 	 * report any more events, and there are nasty race conditions that
430 	 * can beat us if we don't.
431 	 */
432 	knlist_clear(&p->p_klist, 1);
433 
434 	/*
435 	 * Notify parent that we're gone.  If parent has the PS_NOCLDWAIT
436 	 * flag set, or if the handler is set to SIG_IGN, notify process
437 	 * 1 instead (and hope it will handle this situation).
438 	 */
439 	PROC_LOCK(p->p_pptr);
440 	mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
441 	if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
442 		struct proc *pp;
443 
444 		mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
445 		pp = p->p_pptr;
446 		PROC_UNLOCK(pp);
447 		proc_reparent(p, initproc);
448 		p->p_sigparent = SIGCHLD;
449 		PROC_LOCK(p->p_pptr);
450 		/*
451 		 * If this was the last child of our parent, notify
452 		 * parent, so in case he was wait(2)ing, he will
453 		 * continue.
454 		 */
455 		if (LIST_EMPTY(&pp->p_children))
456 			wakeup(pp);
457 	} else
458 		mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
459 
460 	if (p->p_pptr == initproc)
461 		psignal(p->p_pptr, SIGCHLD);
462 	else if (p->p_sigparent != 0)
463 		psignal(p->p_pptr, p->p_sigparent);
464 	PROC_UNLOCK(p->p_pptr);
465 
466 	/*
467 	 * If this is a kthread, then wakeup anyone waiting for it to exit.
468 	 */
469 	if (p->p_flag & P_KTHREAD)
470 		wakeup(p);
471 	PROC_UNLOCK(p);
472 
473 	/*
474 	 * Finally, call machine-dependent code to release the remaining
475 	 * resources including address space.
476 	 * The address space is released by "vmspace_exitfree(p)" in
477 	 * vm_waitproc().
478 	 */
479 	cpu_exit(td);
480 
481 	PROC_LOCK(p);
482 	PROC_LOCK(p->p_pptr);
483 	sx_xunlock(&proctree_lock);
484 
485 	/*
486 	 * We have to wait until after acquiring all locks before
487 	 * changing p_state.  We need to avoid all possible context
488 	 * switches (including ones from blocking on a mutex) while
489 	 * marked as a zombie.
490 	 */
491 	mtx_lock_spin(&sched_lock);
492 	p->p_state = PRS_ZOMBIE;
493 
494 	critical_enter();
495 	mtx_unlock_spin(&sched_lock);
496 	wakeup(p->p_pptr);
497 	PROC_UNLOCK(p->p_pptr);
498 	mtx_lock_spin(&sched_lock);
499 	critical_exit();
500 
501 	/* Do the same timestamp bookkeeping that mi_switch() would do. */
502 	binuptime(&new_switchtime);
503 	bintime_add(&p->p_rux.rux_runtime, &new_switchtime);
504 	bintime_sub(&p->p_rux.rux_runtime, PCPU_PTR(switchtime));
505 	PCPU_SET(switchtime, new_switchtime);
506 	PCPU_SET(switchticks, ticks);
507 	cnt.v_swtch++;
508 
509 	sched_exit(p->p_pptr, td);
510 
511 	/*
512 	 * Hopefully no one will try to deliver a signal to the process this
513 	 * late in the game.
514 	 */
515 	knlist_destroy(&p->p_klist);
516 
517 	/*
518 	 * Make sure the scheduler takes this thread out of its tables etc.
519 	 * This will also release this thread's reference to the ucred.
520 	 * Other thread parts to release include pcb bits and such.
521 	 */
522 	thread_exit();
523 }
524 
525 #ifdef COMPAT_43
526 /*
527  * The dirty work is handled by kern_wait().
528  *
529  * MPSAFE.
530  */
531 int
532 owait(struct thread *td, struct owait_args *uap __unused)
533 {
534 	int error, status;
535 
536 	error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
537 	if (error == 0)
538 		td->td_retval[1] = status;
539 	return (error);
540 }
541 #endif /* COMPAT_43 */
542 
543 /*
544  * The dirty work is handled by kern_wait().
545  *
546  * MPSAFE.
547  */
548 int
549 wait4(struct thread *td, struct wait_args *uap)
550 {
551 	struct rusage ru, *rup;
552 	int error, status;
553 
554 	if (uap->rusage != NULL)
555 		rup = &ru;
556 	else
557 		rup = NULL;
558 	error = kern_wait(td, uap->pid, &status, uap->options, rup);
559 	if (uap->status != NULL && error == 0)
560 		error = copyout(&status, uap->status, sizeof(status));
561 	if (uap->rusage != NULL && error == 0)
562 		error = copyout(&ru, uap->rusage, sizeof(struct rusage));
563 	return (error);
564 }
565 
566 int
567 kern_wait(struct thread *td, pid_t pid, int *status, int options,
568     struct rusage *rusage)
569 {
570 	struct proc *p, *q, *t;
571 	int error, nfound;
572 
573 	q = td->td_proc;
574 	if (pid == 0) {
575 		PROC_LOCK(q);
576 		pid = -q->p_pgid;
577 		PROC_UNLOCK(q);
578 	}
579 	if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WLINUXCLONE))
580 		return (EINVAL);
581 loop:
582 	nfound = 0;
583 	sx_xlock(&proctree_lock);
584 	LIST_FOREACH(p, &q->p_children, p_sibling) {
585 		PROC_LOCK(p);
586 		if (pid != WAIT_ANY &&
587 		    p->p_pid != pid && p->p_pgid != -pid) {
588 			PROC_UNLOCK(p);
589 			continue;
590 		}
591 
592 		/*
593 		 * This special case handles a kthread spawned by linux_clone
594 		 * (see linux_misc.c).  The linux_wait4 and linux_waitpid
595 		 * functions need to be able to distinguish between waiting
596 		 * on a process and waiting on a thread.  It is a thread if
597 		 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
598 		 * signifies we want to wait for threads and not processes.
599 		 */
600 		if ((p->p_sigparent != SIGCHLD) ^
601 		    ((options & WLINUXCLONE) != 0)) {
602 			PROC_UNLOCK(p);
603 			continue;
604 		}
605 
606 		nfound++;
607 		if (p->p_state == PRS_ZOMBIE) {
608 			td->td_retval[0] = p->p_pid;
609 			if (status)
610 				*status = p->p_xstat;	/* convert to int */
611 			if (rusage) {
612 				bcopy(p->p_ru, rusage, sizeof(struct rusage));
613 				calcru(p, &rusage->ru_utime, &rusage->ru_stime);
614 			}
615 
616 			/*
617 			 * If we got the child via a ptrace 'attach',
618 			 * we need to give it back to the old parent.
619 			 */
620 			PROC_UNLOCK(p);
621 			if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) {
622 				PROC_LOCK(p);
623 				p->p_oppid = 0;
624 				proc_reparent(p, t);
625 				PROC_UNLOCK(p);
626 				psignal(t, SIGCHLD);
627 				wakeup(t);
628 				PROC_UNLOCK(t);
629 				sx_xunlock(&proctree_lock);
630 				return (0);
631 			}
632 
633 			/*
634 			 * Remove other references to this process to ensure
635 			 * we have an exclusive reference.
636 			 */
637 			sx_xlock(&allproc_lock);
638 			LIST_REMOVE(p, p_list);	/* off zombproc */
639 			sx_xunlock(&allproc_lock);
640 			LIST_REMOVE(p, p_sibling);
641 			leavepgrp(p);
642 			sx_xunlock(&proctree_lock);
643 
644 			/*
645 			 * As a side effect of this lock, we know that
646 			 * all other writes to this proc are visible now, so
647 			 * no more locking is needed for p.
648 			 */
649 			PROC_LOCK(p);
650 			p->p_xstat = 0;		/* XXX: why? */
651 			PROC_UNLOCK(p);
652 			PROC_LOCK(q);
653 			ruadd(&q->p_stats->p_cru, &q->p_crux, p->p_ru,
654 			    &p->p_rux);
655 			PROC_UNLOCK(q);
656 			FREE(p->p_ru, M_ZOMBIE);
657 			p->p_ru = NULL;
658 
659 			/*
660 			 * Decrement the count of procs running with this uid.
661 			 */
662 			(void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
663 
664 			/*
665 			 * Free credentials, arguments, and sigacts.
666 			 */
667 			crfree(p->p_ucred);
668 			p->p_ucred = NULL;
669 			pargs_drop(p->p_args);
670 			p->p_args = NULL;
671 			sigacts_free(p->p_sigacts);
672 			p->p_sigacts = NULL;
673 
674 			/*
675 			 * Do any thread-system specific cleanups.
676 			 */
677 			thread_wait(p);
678 
679 			/*
680 			 * Give vm and machine-dependent layer a chance
681 			 * to free anything that cpu_exit couldn't
682 			 * release while still running in process context.
683 			 */
684 			vm_waitproc(p);
685 #ifdef MAC
686 			mac_destroy_proc(p);
687 #endif
688 			KASSERT(FIRST_THREAD_IN_PROC(p),
689 			    ("kern_wait: no residual thread!"));
690 			uma_zfree(proc_zone, p);
691 			sx_xlock(&allproc_lock);
692 			nprocs--;
693 			sx_xunlock(&allproc_lock);
694 			return (0);
695 		}
696 		mtx_lock_spin(&sched_lock);
697 		if (P_SHOULDSTOP(p) && p->p_suspcount == p->p_numthreads &&
698 		    (p->p_flag & P_WAITED) == 0 &&
699 		    (p->p_flag & P_TRACED || options & WUNTRACED)) {
700 			mtx_unlock_spin(&sched_lock);
701 			p->p_flag |= P_WAITED;
702 			sx_xunlock(&proctree_lock);
703 			td->td_retval[0] = p->p_pid;
704 			if (status)
705 				*status = W_STOPCODE(p->p_xstat);
706 			PROC_UNLOCK(p);
707 			return (0);
708 		}
709 		mtx_unlock_spin(&sched_lock);
710 		if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) {
711 			sx_xunlock(&proctree_lock);
712 			td->td_retval[0] = p->p_pid;
713 			p->p_flag &= ~P_CONTINUED;
714 			PROC_UNLOCK(p);
715 
716 			if (status)
717 				*status = SIGCONT;
718 			return (0);
719 		}
720 		PROC_UNLOCK(p);
721 	}
722 	if (nfound == 0) {
723 		sx_xunlock(&proctree_lock);
724 		return (ECHILD);
725 	}
726 	if (options & WNOHANG) {
727 		sx_xunlock(&proctree_lock);
728 		td->td_retval[0] = 0;
729 		return (0);
730 	}
731 	PROC_LOCK(q);
732 	sx_xunlock(&proctree_lock);
733 	error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
734 	PROC_UNLOCK(q);
735 	if (error)
736 		return (error);
737 	goto loop;
738 }
739 
740 /*
741  * Make process 'parent' the new parent of process 'child'.
742  * Must be called with an exclusive hold of proctree lock.
743  */
744 void
745 proc_reparent(struct proc *child, struct proc *parent)
746 {
747 
748 	sx_assert(&proctree_lock, SX_XLOCKED);
749 	PROC_LOCK_ASSERT(child, MA_OWNED);
750 	if (child->p_pptr == parent)
751 		return;
752 
753 	LIST_REMOVE(child, p_sibling);
754 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
755 	child->p_pptr = parent;
756 }
757