xref: /freebsd/sys/kern/kern_exit.c (revision 2172a5cbc5f6999263fbcf5a367032154db963b9)
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  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the University of
21  *	California, Berkeley and its contributors.
22  * 4. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  *
38  *	@(#)kern_exit.c	8.7 (Berkeley) 2/12/94
39  */
40 
41 #include <sys/cdefs.h>
42 __FBSDID("$FreeBSD$");
43 
44 #include "opt_compat.h"
45 #include "opt_ktrace.h"
46 #include "opt_mac.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/sysproto.h>
51 #include <sys/eventhandler.h>
52 #include <sys/kernel.h>
53 #include <sys/malloc.h>
54 #include <sys/lock.h>
55 #include <sys/mutex.h>
56 #include <sys/proc.h>
57 #include <sys/pioctl.h>
58 #include <sys/tty.h>
59 #include <sys/wait.h>
60 #include <sys/vmmeter.h>
61 #include <sys/vnode.h>
62 #include <sys/resourcevar.h>
63 #include <sys/signalvar.h>
64 #include <sys/sched.h>
65 #include <sys/sx.h>
66 #include <sys/ptrace.h>
67 #include <sys/acct.h>		/* for acct_process() function prototype */
68 #include <sys/filedesc.h>
69 #include <sys/mac.h>
70 #include <sys/shm.h>
71 #include <sys/sem.h>
72 #ifdef KTRACE
73 #include <sys/ktrace.h>
74 #endif
75 
76 #include <vm/vm.h>
77 #include <vm/vm_extern.h>
78 #include <vm/vm_param.h>
79 #include <vm/pmap.h>
80 #include <vm/vm_map.h>
81 #include <vm/vm_page.h>
82 #include <vm/uma.h>
83 #include <sys/user.h>
84 
85 /* Required to be non-static for SysVR4 emulator */
86 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status");
87 
88 static int wait1(struct thread *, struct wait_args *, int);
89 
90 /*
91  * exit --
92  *	Death of process.
93  *
94  * MPSAFE
95  */
96 void
97 sys_exit(struct thread *td, struct sys_exit_args *uap)
98 {
99 
100 	mtx_lock(&Giant);
101 	exit1(td, W_EXITCODE(uap->rval, 0));
102 	/* NOTREACHED */
103 }
104 
105 /*
106  * Exit: deallocate address space and other resources, change proc state
107  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
108  * status and rusage for wait().  Check for child processes and orphan them.
109  */
110 void
111 exit1(struct thread *td, int rv)
112 {
113 	struct proc *p, *nq, *q;
114 	struct tty *tp;
115 	struct vnode *ttyvp;
116 	struct vmspace *vm;
117 	struct vnode *vtmp;
118 #ifdef KTRACE
119 	struct vnode *tracevp;
120 	struct ucred *tracecred;
121 #endif
122 
123 	GIANT_REQUIRED;
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_SA || p->p_numthreads > 1) {
137 		/*
138 		 * First check if some other thread got here before us..
139 		 * if so, act apropriatly, (exit or suspend);
140 		 */
141 		thread_suspend_check(0);
142 
143 		/*
144 		 * Kill off the other threads. This requires
145 		 * Some co-operation from other parts of the kernel
146 		 * so it may not be instant.
147 		 * 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 immediatly
151 		 * with EINTR or EWOULDBLOCK, which will hopefully force them
152 		 * to back out to userland, freeing resources as they go, and
153 		 * anything attempting to return to userland will thread_exit()
154 		 * from userret().  thread_exit() will unsuspend us
155 		 * when the last other thread exits.
156 		 */
157 		if (thread_single(SINGLE_EXIT)) {
158 			panic ("Exit: Single threading fouled up");
159 		}
160 		/*
161 		 * All other activity in this process is now stopped.
162 		 * Remove excess KSEs and KSEGRPS. XXXKSE (when we have them)
163 		 * ...
164 		 * Turn off threading support.
165 		 */
166 		p->p_flag &= ~P_SA;
167 		thread_single_end();	/* Don't need this any more. */
168 	}
169 	/*
170 	 * With this state set:
171 	 * Any thread entering the kernel from userspace will thread_exit()
172 	 * in trap().  Any thread attempting to sleep will return immediatly
173 	 * with EINTR or EWOULDBLOCK, which will hopefully force them
174 	 * to back out to userland, freeing resources as they go, and
175 	 * anything attempting to return to userland will thread_exit()
176 	 * from userret().  thread_exit() will do a wakeup on p->p_numthreads
177 	 * if it transitions to 1.
178 	 */
179 
180 	p->p_flag |= P_WEXIT;
181 	PROC_UNLOCK(p);
182 
183 	/* Are we a task leader? */
184 	if (p == p->p_leader) {
185 		mtx_lock(&ppeers_lock);
186 		q = p->p_peers;
187 		while (q != NULL) {
188 			PROC_LOCK(q);
189 			psignal(q, SIGKILL);
190 			PROC_UNLOCK(q);
191 			q = q->p_peers;
192 		}
193 		while (p->p_peers != NULL)
194 			msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
195 		mtx_unlock(&ppeers_lock);
196 	}
197 
198 #ifdef PGINPROF
199 	vmsizmon();
200 #endif
201 	STOPEVENT(p, S_EXIT, rv);
202 	wakeup(&p->p_stype);	/* Wakeup anyone in procfs' PIOCWAIT */
203 
204 	/*
205 	 * Check if any loadable modules need anything done at process exit.
206 	 * e.g. SYSV IPC stuff
207 	 * XXX what if one of these generates an error?
208 	 */
209 	EVENTHANDLER_INVOKE(process_exit, p);
210 
211 	MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage),
212 		M_ZOMBIE, M_WAITOK);
213 	/*
214 	 * If parent is waiting for us to exit or exec,
215 	 * P_PPWAIT is set; we will wakeup the parent below.
216 	 */
217 	PROC_LOCK(p);
218 	stopprofclock(p);
219 	p->p_flag &= ~(P_TRACED | P_PPWAIT);
220 	SIGEMPTYSET(p->p_siglist);
221 	SIGEMPTYSET(td->td_siglist);
222 
223 	/*
224 	 * Stop the real interval timer.  If the handler is currently
225 	 * executing, prevent it from rearming itself and let it finish.
226 	 */
227 	if (timevalisset(&p->p_realtimer.it_value) &&
228 	    callout_stop(&p->p_itcallout) == 0) {
229 		timevalclear(&p->p_realtimer.it_interval);
230 		msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
231 		KASSERT(!timevalisset(&p->p_realtimer.it_value),
232 		    ("realtime timer is still armed"));
233 	}
234 	PROC_UNLOCK(p);
235 
236 	/*
237 	 * Reset any sigio structures pointing to us as a result of
238 	 * F_SETOWN with our pid.
239 	 */
240 	funsetownlst(&p->p_sigiolst);
241 
242 	/*
243 	 * Close open files and release open-file table.
244 	 * This may block!
245 	 */
246 	fdfree(td);
247 
248 	/*
249 	 * Remove ourself from our leader's peer list and wake our leader.
250 	 */
251 	mtx_lock(&ppeers_lock);
252 	if (p->p_leader->p_peers) {
253 		q = p->p_leader;
254 		while (q->p_peers != p)
255 			q = q->p_peers;
256 		q->p_peers = p->p_peers;
257 		wakeup(p->p_leader);
258 	}
259 	mtx_unlock(&ppeers_lock);
260 
261 	/* The next two chunks should probably be moved to vmspace_exit. */
262 	vm = p->p_vmspace;
263 	/*
264 	 * Release user portion of address space.
265 	 * This releases references to vnodes,
266 	 * which could cause I/O if the file has been unlinked.
267 	 * Need to do this early enough that we can still sleep.
268 	 * Can't free the entire vmspace as the kernel stack
269 	 * may be mapped within that space also.
270 	 *
271 	 * Processes sharing the same vmspace may exit in one order, and
272 	 * get cleaned up by vmspace_exit() in a different order.  The
273 	 * last exiting process to reach this point releases as much of
274 	 * the environment as it can, and the last process cleaned up
275 	 * by vmspace_exit() (which decrements exitingcnt) cleans up the
276 	 * remainder.
277 	 */
278 	++vm->vm_exitingcnt;
279 	if (--vm->vm_refcnt == 0) {
280 		shmexit(vm);
281 		vm_page_lock_queues();
282 		pmap_remove_pages(vmspace_pmap(vm), vm_map_min(&vm->vm_map),
283 		    vm_map_max(&vm->vm_map));
284 		vm_page_unlock_queues();
285 		(void) vm_map_remove(&vm->vm_map, vm_map_min(&vm->vm_map),
286 		    vm_map_max(&vm->vm_map));
287 	}
288 
289 	sx_xlock(&proctree_lock);
290 	if (SESS_LEADER(p)) {
291 		struct session *sp;
292 
293 		sp = p->p_session;
294 		if (sp->s_ttyvp) {
295 			/*
296 			 * Controlling process.
297 			 * Signal foreground pgrp,
298 			 * drain controlling terminal
299 			 * and revoke access to controlling terminal.
300 			 */
301 			if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
302 				tp = sp->s_ttyp;
303 				if (sp->s_ttyp->t_pgrp) {
304 					PGRP_LOCK(sp->s_ttyp->t_pgrp);
305 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
306 					PGRP_UNLOCK(sp->s_ttyp->t_pgrp);
307 				}
308 				/* XXX tp should be locked. */
309 				sx_xunlock(&proctree_lock);
310 				(void) ttywait(tp);
311 				sx_xlock(&proctree_lock);
312 				/*
313 				 * The tty could have been revoked
314 				 * if we blocked.
315 				 */
316 				if (sp->s_ttyvp) {
317 					ttyvp = sp->s_ttyvp;
318 					SESS_LOCK(p->p_session);
319 					sp->s_ttyvp = NULL;
320 					SESS_UNLOCK(p->p_session);
321 					sx_xunlock(&proctree_lock);
322 					VOP_REVOKE(ttyvp, REVOKEALL);
323 					vrele(ttyvp);
324 					sx_xlock(&proctree_lock);
325 				}
326 			}
327 			if (sp->s_ttyvp) {
328 				ttyvp = sp->s_ttyvp;
329 				SESS_LOCK(p->p_session);
330 				sp->s_ttyvp = NULL;
331 				SESS_UNLOCK(p->p_session);
332 				vrele(ttyvp);
333 			}
334 			/*
335 			 * s_ttyp is not zero'd; we use this to indicate
336 			 * that the session once had a controlling terminal.
337 			 * (for logging and informational purposes)
338 			 */
339 		}
340 		SESS_LOCK(p->p_session);
341 		sp->s_leader = NULL;
342 		SESS_UNLOCK(p->p_session);
343 	}
344 	fixjobc(p, p->p_pgrp, 0);
345 	sx_xunlock(&proctree_lock);
346 	(void)acct_process(td);
347 #ifdef KTRACE
348 	/*
349 	 * release trace file
350 	 */
351 	PROC_LOCK(p);
352 	mtx_lock(&ktrace_mtx);
353 	p->p_traceflag = 0;	/* don't trace the vrele() */
354 	tracevp = p->p_tracevp;
355 	p->p_tracevp = NULL;
356 	tracecred = p->p_tracecred;
357 	p->p_tracecred = NULL;
358 	mtx_unlock(&ktrace_mtx);
359 	PROC_UNLOCK(p);
360 	if (tracevp != NULL)
361 		vrele(tracevp);
362 	if (tracecred != NULL)
363 		crfree(tracecred);
364 #endif
365 	/*
366 	 * Release reference to text vnode
367 	 */
368 	if ((vtmp = p->p_textvp) != NULL) {
369 		p->p_textvp = NULL;
370 		vrele(vtmp);
371 	}
372 
373 	/*
374 	 * Release our limits structure.
375 	 */
376 	mtx_assert(&Giant, MA_OWNED);
377 	if (--p->p_limit->p_refcnt == 0) {
378 		FREE(p->p_limit, M_SUBPROC);
379 		p->p_limit = NULL;
380 	}
381 
382 	/*
383 	 * Release this thread's reference to the ucred.  The actual proc
384 	 * reference will stay around until the proc is harvested by
385 	 * wait().  At this point the ucred is immutable (no other threads
386 	 * from this proc are around that can change it) so we leave the
387 	 * per-thread ucred pointer intact in case it is needed although
388 	 * in theory nothing should be using it at this point.
389 	 */
390 	crfree(td->td_ucred);
391 
392 	/*
393 	 * Remove proc from allproc queue and pidhash chain.
394 	 * Place onto zombproc.  Unlink from parent's child list.
395 	 */
396 	sx_xlock(&allproc_lock);
397 	LIST_REMOVE(p, p_list);
398 	LIST_INSERT_HEAD(&zombproc, p, p_list);
399 	LIST_REMOVE(p, p_hash);
400 	sx_xunlock(&allproc_lock);
401 
402 	sx_xlock(&proctree_lock);
403 	q = LIST_FIRST(&p->p_children);
404 	if (q != NULL)		/* only need this if any child is S_ZOMB */
405 		wakeup(initproc);
406 	for (; q != NULL; q = nq) {
407 		nq = LIST_NEXT(q, p_sibling);
408 		PROC_LOCK(q);
409 		proc_reparent(q, initproc);
410 		q->p_sigparent = SIGCHLD;
411 		/*
412 		 * Traced processes are killed
413 		 * since their existence means someone is screwing up.
414 		 */
415 		if (q->p_flag & P_TRACED) {
416 			q->p_flag &= ~P_TRACED;
417 			psignal(q, SIGKILL);
418 		}
419 		PROC_UNLOCK(q);
420 	}
421 
422 	/*
423 	 * Save exit status and final rusage info, adding in child rusage
424 	 * info and self times.
425 	 */
426 	PROC_LOCK(p);
427 	p->p_xstat = rv;
428 	*p->p_ru = p->p_stats->p_ru;
429 	mtx_lock_spin(&sched_lock);
430 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
431 	mtx_unlock_spin(&sched_lock);
432 	ruadd(p->p_ru, &p->p_stats->p_cru);
433 
434 	/*
435 	 * Notify interested parties of our demise.
436 	 */
437 	KNOTE(&p->p_klist, NOTE_EXIT);
438 	/*
439 	 * Just delete all entries in the p_klist. At this point we won't
440 	 * report any more events, and there are nasty race conditions that
441 	 * can beat us if we don't.
442 	 */
443 	while (SLIST_FIRST(&p->p_klist))
444 		SLIST_REMOVE_HEAD(&p->p_klist, kn_selnext);
445 
446 	/*
447 	 * Notify parent that we're gone.  If parent has the PS_NOCLDWAIT
448 	 * flag set, or if the handler is set to SIG_IGN, notify process
449 	 * 1 instead (and hope it will handle this situation).
450 	 */
451 	PROC_LOCK(p->p_pptr);
452 	mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
453 	if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
454 		struct proc *pp;
455 
456 		mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
457 		pp = p->p_pptr;
458 		PROC_UNLOCK(pp);
459 		proc_reparent(p, initproc);
460 		PROC_LOCK(p->p_pptr);
461 		/*
462 		 * If this was the last child of our parent, notify
463 		 * parent, so in case he was wait(2)ing, he will
464 		 * continue.
465 		 */
466 		if (LIST_EMPTY(&pp->p_children))
467 			wakeup(pp);
468 	} else
469 		mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
470 
471 	if (p->p_sigparent && p->p_pptr != initproc)
472 		psignal(p->p_pptr, p->p_sigparent);
473 	else
474 		psignal(p->p_pptr, SIGCHLD);
475 	PROC_UNLOCK(p->p_pptr);
476 
477 	/*
478 	 * If this is a kthread, then wakeup anyone waiting for it to exit.
479 	 */
480 	if (p->p_flag & P_KTHREAD)
481 		wakeup(p);
482 	PROC_UNLOCK(p);
483 
484 	/*
485 	 * Finally, call machine-dependent code to release the remaining
486 	 * resources including address space.
487 	 * The address space is released by "vmspace_exitfree(p)" in
488 	 * vm_waitproc().
489 	 */
490 	cpu_exit(td);
491 
492 	PROC_LOCK(p);
493 	PROC_LOCK(p->p_pptr);
494 	sx_xunlock(&proctree_lock);
495 	mtx_lock_spin(&sched_lock);
496 
497 	while (mtx_owned(&Giant))
498 		mtx_unlock(&Giant);
499 
500 	/*
501 	 * We have to wait until after acquiring all locks before
502 	 * changing p_state.  If we block on a mutex then we will be
503 	 * back at SRUN when we resume and our parent will never
504 	 * harvest us.
505 	 */
506 	p->p_state = PRS_ZOMBIE;
507 
508 	wakeup(p->p_pptr);
509 	PROC_UNLOCK(p->p_pptr);
510 	cnt.v_swtch++;
511 	binuptime(PCPU_PTR(switchtime));
512 	PCPU_SET(switchticks, ticks);
513 
514 	cpu_sched_exit(td); /* XXXKSE check if this should be in thread_exit */
515 	/*
516 	 * Allow the scheduler to adjust the priority of the
517 	 * parent when a kseg is exiting.
518 	 */
519 	if (p->p_pid != 1)
520 		sched_exit(p->p_pptr, p);
521 
522 	/*
523 	 * Make sure the scheduler takes this thread out of its tables etc.
524 	 * This will also release this thread's reference to the ucred.
525 	 * Other thread parts to release include pcb bits and such.
526 	 */
527 	thread_exit();
528 }
529 
530 #ifdef COMPAT_43
531 /*
532  * MPSAFE.  The dirty work is handled by wait1().
533  */
534 int
535 owait(struct thread *td, struct owait_args *uap __unused)
536 {
537 	struct wait_args w;
538 
539 	w.options = 0;
540 	w.rusage = NULL;
541 	w.pid = WAIT_ANY;
542 	w.status = NULL;
543 	return (wait1(td, &w, 1));
544 }
545 #endif /* COMPAT_43 */
546 
547 /*
548  * MPSAFE.  The dirty work is handled by wait1().
549  */
550 int
551 wait4(struct thread *td, struct wait_args *uap)
552 {
553 
554 	return (wait1(td, uap, 0));
555 }
556 
557 /*
558  * MPSAFE
559  */
560 static int
561 wait1(struct thread *td, struct wait_args *uap, int compat)
562 {
563 	struct rusage ru;
564 	int nfound;
565 	struct proc *p, *q, *t;
566 	int status, error;
567 
568 	q = td->td_proc;
569 	if (uap->pid == 0) {
570 		PROC_LOCK(q);
571 		uap->pid = -q->p_pgid;
572 		PROC_UNLOCK(q);
573 	}
574 	if (uap->options &~ (WUNTRACED|WNOHANG|WCONTINUED|WLINUXCLONE))
575 		return (EINVAL);
576 	mtx_lock(&Giant);
577 loop:
578 	nfound = 0;
579 	sx_xlock(&proctree_lock);
580 	LIST_FOREACH(p, &q->p_children, p_sibling) {
581 		PROC_LOCK(p);
582 		if (uap->pid != WAIT_ANY &&
583 		    p->p_pid != uap->pid && p->p_pgid != -uap->pid) {
584 			PROC_UNLOCK(p);
585 			continue;
586 		}
587 
588 		/*
589 		 * This special case handles a kthread spawned by linux_clone
590 		 * (see linux_misc.c).  The linux_wait4 and linux_waitpid
591 		 * functions need to be able to distinguish between waiting
592 		 * on a process and waiting on a thread.  It is a thread if
593 		 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
594 		 * signifies we want to wait for threads and not processes.
595 		 */
596 		if ((p->p_sigparent != SIGCHLD) ^
597 		    ((uap->options & WLINUXCLONE) != 0)) {
598 			PROC_UNLOCK(p);
599 			continue;
600 		}
601 
602 		nfound++;
603 		if (p->p_state == PRS_ZOMBIE) {
604 			td->td_retval[0] = p->p_pid;
605 #ifdef COMPAT_43
606 			if (compat)
607 				td->td_retval[1] = p->p_xstat;
608 			else
609 #endif
610 			if (uap->status) {
611 				status = p->p_xstat;	/* convert to int */
612 				PROC_UNLOCK(p);
613 				if ((error = copyout(&status,
614 				    uap->status, sizeof(status)))) {
615 					sx_xunlock(&proctree_lock);
616 					mtx_unlock(&Giant);
617 					return (error);
618 				}
619 				PROC_LOCK(p);
620 			}
621 			if (uap->rusage) {
622 				bcopy(p->p_ru, &ru, sizeof(ru));
623 				PROC_UNLOCK(p);
624 				if ((error = copyout(&ru,
625 				    uap->rusage, sizeof (struct rusage)))) {
626 					sx_xunlock(&proctree_lock);
627 					mtx_unlock(&Giant);
628 					return (error);
629 				}
630 			} else
631 				PROC_UNLOCK(p);
632 			/*
633 			 * If we got the child via a ptrace 'attach',
634 			 * we need to give it back to the old parent.
635 			 */
636 			if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) {
637 				PROC_LOCK(p);
638 				p->p_oppid = 0;
639 				proc_reparent(p, t);
640 				PROC_UNLOCK(p);
641 				psignal(t, SIGCHLD);
642 				wakeup(t);
643 				PROC_UNLOCK(t);
644 				sx_xunlock(&proctree_lock);
645 				mtx_unlock(&Giant);
646 				return (0);
647 			}
648 
649 			/*
650 			 * Remove other references to this process to ensure
651 			 * we have an exclusive reference.
652 			 */
653 			sx_xlock(&allproc_lock);
654 			LIST_REMOVE(p, p_list);	/* off zombproc */
655 			sx_xunlock(&allproc_lock);
656 			LIST_REMOVE(p, p_sibling);
657 			leavepgrp(p);
658 			sx_xunlock(&proctree_lock);
659 
660 			/*
661 			 * As a side effect of this lock, we know that
662 			 * all other writes to this proc are visible now, so
663 			 * no more locking is needed for p.
664 			 */
665 			PROC_LOCK(p);
666 			p->p_xstat = 0;		/* XXX: why? */
667 			PROC_UNLOCK(p);
668 			PROC_LOCK(q);
669 			ruadd(&q->p_stats->p_cru, p->p_ru);
670 			PROC_UNLOCK(q);
671 			FREE(p->p_ru, M_ZOMBIE);
672 			p->p_ru = NULL;
673 
674 			/*
675 			 * Decrement the count of procs running with this uid.
676 			 */
677 			(void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
678 
679 			/*
680 			 * Free credentials, arguments, and sigacts
681 			 */
682 			crfree(p->p_ucred);
683 			p->p_ucred = NULL;
684 			pargs_drop(p->p_args);
685 			p->p_args = NULL;
686 			sigacts_free(p->p_sigacts);
687 			p->p_sigacts = NULL;
688 
689 			/*
690 			 * do any thread-system specific cleanups
691 			 */
692 			thread_wait(p);
693 
694 			/*
695 			 * Give vm and machine-dependent layer a chance
696 			 * to free anything that cpu_exit couldn't
697 			 * release while still running in process context.
698 			 */
699 			vm_waitproc(p);
700 #ifdef MAC
701 			mac_destroy_proc(p);
702 #endif
703 			KASSERT(FIRST_THREAD_IN_PROC(p),
704 			    ("wait1: no residual thread!"));
705 			uma_zfree(proc_zone, p);
706 			sx_xlock(&allproc_lock);
707 			nprocs--;
708 			sx_xunlock(&allproc_lock);
709 			mtx_unlock(&Giant);
710 			return (0);
711 		}
712 		mtx_lock_spin(&sched_lock);
713 		if (P_SHOULDSTOP(p) && (p->p_suspcount == p->p_numthreads) &&
714 		    ((p->p_flag & P_WAITED) == 0) &&
715 		    (p->p_flag & P_TRACED || uap->options & WUNTRACED)) {
716 			mtx_unlock_spin(&sched_lock);
717 			p->p_flag |= P_WAITED;
718 			sx_xunlock(&proctree_lock);
719 			td->td_retval[0] = p->p_pid;
720 #ifdef COMPAT_43
721 			if (compat) {
722 				td->td_retval[1] = W_STOPCODE(p->p_xstat);
723 				PROC_UNLOCK(p);
724 				error = 0;
725 			} else
726 #endif
727 			if (uap->status) {
728 				status = W_STOPCODE(p->p_xstat);
729 				PROC_UNLOCK(p);
730 				error = copyout(&status,
731 					uap->status, sizeof(status));
732 			} else {
733 				PROC_UNLOCK(p);
734 				error = 0;
735 			}
736 			mtx_unlock(&Giant);
737 			return (error);
738 		}
739 		mtx_unlock_spin(&sched_lock);
740 		if (uap->options & WCONTINUED && (p->p_flag & P_CONTINUED)) {
741 			sx_xunlock(&proctree_lock);
742 			td->td_retval[0] = p->p_pid;
743 			p->p_flag &= ~P_CONTINUED;
744 			PROC_UNLOCK(p);
745 
746 			if (uap->status) {
747 				status = SIGCONT;
748 				error = copyout(&status,
749 				    uap->status, sizeof(status));
750 			} else
751 				error = 0;
752 
753 			mtx_unlock(&Giant);
754 			return (error);
755 		}
756 		PROC_UNLOCK(p);
757 	}
758 	if (nfound == 0) {
759 		sx_xunlock(&proctree_lock);
760 		mtx_unlock(&Giant);
761 		return (ECHILD);
762 	}
763 	if (uap->options & WNOHANG) {
764 		sx_xunlock(&proctree_lock);
765 		td->td_retval[0] = 0;
766 		mtx_unlock(&Giant);
767 		return (0);
768 	}
769 	PROC_LOCK(q);
770 	sx_xunlock(&proctree_lock);
771 	error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
772 	PROC_UNLOCK(q);
773 	if (error) {
774 		mtx_unlock(&Giant);
775 		return (error);
776 	}
777 	goto loop;
778 }
779 
780 /*
781  * Make process 'parent' the new parent of process 'child'.
782  * Must be called with an exclusive hold of proctree lock.
783  */
784 void
785 proc_reparent(struct proc *child, struct proc *parent)
786 {
787 
788 	sx_assert(&proctree_lock, SX_XLOCKED);
789 	PROC_LOCK_ASSERT(child, MA_OWNED);
790 	if (child->p_pptr == parent)
791 		return;
792 
793 	LIST_REMOVE(child, p_sibling);
794 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
795 	child->p_pptr = parent;
796 }
797