xref: /freebsd/sys/kern/kern_exit.c (revision 69c9999d0ca45b210e75706ab4952ad5a33ce6ec)
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  * $FreeBSD$
40  */
41 
42 #include "opt_compat.h"
43 #include "opt_ktrace.h"
44 #include "opt_mac.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/sysproto.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/pioctl.h>
55 #include <sys/tty.h>
56 #include <sys/wait.h>
57 #include <sys/vmmeter.h>
58 #include <sys/vnode.h>
59 #include <sys/resourcevar.h>
60 #include <sys/signalvar.h>
61 #include <sys/sched.h>
62 #include <sys/sx.h>
63 #include <sys/ptrace.h>
64 #include <sys/acct.h>		/* for acct_process() function prototype */
65 #include <sys/filedesc.h>
66 #include <sys/mac.h>
67 #include <sys/shm.h>
68 #include <sys/sem.h>
69 #include <sys/jail.h>
70 #ifdef KTRACE
71 #include <sys/ktrace.h>
72 #endif
73 
74 #include <vm/vm.h>
75 #include <vm/vm_extern.h>
76 #include <vm/vm_param.h>
77 #include <vm/pmap.h>
78 #include <vm/vm_map.h>
79 #include <vm/vm_page.h>
80 #include <vm/uma.h>
81 #include <sys/user.h>
82 
83 /* Required to be non-static for SysVR4 emulator */
84 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status");
85 
86 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback");
87 
88 static int wait1(struct thread *, struct wait_args *, int);
89 
90 /*
91  * callout list for things to do at exit time
92  */
93 struct exitlist {
94 	exitlist_fn function;
95 	TAILQ_ENTRY(exitlist) next;
96 };
97 
98 TAILQ_HEAD(exit_list_head, exitlist);
99 static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list);
100 
101 /*
102  * exit --
103  *	Death of process.
104  *
105  * MPSAFE
106  */
107 void
108 sys_exit(td, uap)
109 	struct thread *td;
110 	struct sys_exit_args /* {
111 		int	rval;
112 	} */ *uap;
113 {
114 
115 	mtx_lock(&Giant);
116 	exit1(td, W_EXITCODE(uap->rval, 0));
117 	/* NOTREACHED */
118 }
119 
120 /*
121  * Exit: deallocate address space and other resources, change proc state
122  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
123  * status and rusage for wait().  Check for child processes and orphan them.
124  */
125 void
126 exit1(td, rv)
127 	register struct thread *td;
128 	int rv;
129 {
130 	struct exitlist *ep;
131 	struct proc *p, *nq, *q;
132 	struct tty *tp;
133 	struct vnode *ttyvp;
134 	register struct vmspace *vm;
135 	struct vnode *vtmp;
136 #ifdef KTRACE
137 	struct vnode *tracevp;
138 #endif
139 
140 	GIANT_REQUIRED;
141 
142 	p = td->td_proc;
143 	if (p == initproc) {
144 		printf("init died (signal %d, exit %d)\n",
145 		    WTERMSIG(rv), WEXITSTATUS(rv));
146 		panic("Going nowhere without my init!");
147 	}
148 
149 	/*
150 	 * XXXXKSE: MUST abort all other threads before proceeding past here.
151 	 */
152 	PROC_LOCK(p);
153 	if (p->p_flag & P_KSES) {
154 		/*
155 		 * First check if some other thread got here before us..
156 		 * if so, act apropriatly, (exit or suspend);
157 		 */
158 		thread_suspend_check(0);
159 		/*
160 		 * Here is a trick..
161 		 * We need to free up our KSE to process other threads
162 		 * so that we can safely set the UNBOUND flag
163 		 * (whether or not we have a mailbox) as we are NEVER
164 		 * going to return to the user.
165 		 * The flag will not be set yet if we are exiting
166 		 * because of a signal, pagefault, or similar
167 		 * (or even an exit(2) from the UTS).
168 		 */
169 		td->td_flags |= TDF_UNBOUND;
170 
171 		/*
172 		 * Kill off the other threads. This requires
173 		 * Some co-operation from other parts of the kernel
174 		 * so it may not be instant.
175 		 * With this state set:
176 		 * Any thread entering the kernel from userspace will
177 		 * thread_exit() in trap().  Any thread attempting to
178 		 * sleep will return immediatly
179 		 * with EINTR or EWOULDBLOCK, which will hopefully force them
180 		 * to back out to userland, freeing resources as they go, and
181 		 * anything attempting to return to userland will thread_exit()
182 		 * from userret().  thread_exit() will unsuspend us
183 		 * when the last other thread exits.
184 		 */
185 		if (thread_single(SINGLE_EXIT)) {
186 			panic ("Exit: Single threading fouled up");
187 		}
188 		/*
189 		 * All other activity in this process is now stopped.
190 		 * Remove excess KSEs and KSEGRPS. XXXKSE (when we have them)
191 		 * ...
192 		 * Turn off threading support.
193 		 */
194 		p->p_flag &= ~P_KSES;
195 		td->td_flags &= ~TDF_UNBOUND;
196 		thread_single_end(); 	/* Don't need this any more. */
197 	}
198 	/*
199 	 * With this state set:
200 	 * Any thread entering the kernel from userspace will thread_exit()
201 	 * in trap().  Any thread attempting to sleep will return immediatly
202 	 * with EINTR or EWOULDBLOCK, which will hopefully force them
203 	 * to back out to userland, freeing resources as they go, and
204 	 * anything attempting to return to userland will thread_exit()
205 	 * from userret().  thread_exit() will do a wakeup on p->p_numthreads
206 	 * if it transitions to 1.
207 	 */
208 
209 	p->p_flag |= P_WEXIT;
210 	PROC_UNLOCK(p);
211 
212 	/* Are we a task leader? */
213 	if (p == p->p_leader) {
214 		mtx_lock(&ppeers_lock);
215 		q = p->p_peers;
216 		while (q != NULL) {
217 			PROC_LOCK(q);
218 			psignal(q, SIGKILL);
219 			PROC_UNLOCK(q);
220 			q = q->p_peers;
221 		}
222 		while (p->p_peers != NULL)
223 			msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
224 		mtx_unlock(&ppeers_lock);
225 	}
226 
227 #ifdef PGINPROF
228 	vmsizmon();
229 #endif
230 	STOPEVENT(p, S_EXIT, rv);
231 	wakeup(&p->p_stype);	/* Wakeup anyone in procfs' PIOCWAIT */
232 
233 	/*
234 	 * Check if any loadable modules need anything done at process exit.
235 	 * e.g. SYSV IPC stuff
236 	 * XXX what if one of these generates an error?
237 	 */
238 	TAILQ_FOREACH(ep, &exit_list, next)
239 		(*ep->function)(p);
240 
241 	stopprofclock(p);
242 
243 	MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage),
244 		M_ZOMBIE, M_WAITOK);
245 	/*
246 	 * If parent is waiting for us to exit or exec,
247 	 * P_PPWAIT is set; we will wakeup the parent below.
248 	 */
249 	PROC_LOCK(p);
250 	p->p_flag &= ~(P_TRACED | P_PPWAIT);
251 	SIGEMPTYSET(p->p_siglist);
252 	PROC_UNLOCK(p);
253 	if (timevalisset(&p->p_realtimer.it_value))
254 		callout_stop(&p->p_itcallout);
255 
256 	/*
257 	 * Reset any sigio structures pointing to us as a result of
258 	 * F_SETOWN with our pid.
259 	 */
260 	funsetownlst(&p->p_sigiolst);
261 
262 	/*
263 	 * Close open files and release open-file table.
264 	 * This may block!
265 	 */
266 	fdfree(td); /* XXXKSE *//* may not be the one in proc */
267 
268 	/*
269 	 * Remove ourself from our leader's peer list and wake our leader.
270 	 */
271 	mtx_lock(&ppeers_lock);
272 	if (p->p_leader->p_peers) {
273 		q = p->p_leader;
274 		while (q->p_peers != p)
275 			q = q->p_peers;
276 		q->p_peers = p->p_peers;
277 		wakeup(p->p_leader);
278 	}
279 	mtx_unlock(&ppeers_lock);
280 
281 	/* The next two chunks should probably be moved to vmspace_exit. */
282 	vm = p->p_vmspace;
283 	/*
284 	 * Release user portion of address space.
285 	 * This releases references to vnodes,
286 	 * which could cause I/O if the file has been unlinked.
287 	 * Need to do this early enough that we can still sleep.
288 	 * Can't free the entire vmspace as the kernel stack
289 	 * may be mapped within that space also.
290 	 *
291 	 * Processes sharing the same vmspace may exit in one order, and
292 	 * get cleaned up by vmspace_exit() in a different order.  The
293 	 * last exiting process to reach this point releases as much of
294 	 * the environment as it can, and the last process cleaned up
295 	 * by vmspace_exit() (which decrements exitingcnt) cleans up the
296 	 * remainder.
297 	 */
298 	++vm->vm_exitingcnt;
299 	if (--vm->vm_refcnt == 0) {
300 		if (vm->vm_shm)
301 			shmexit(p);
302 		vm_page_lock_queues();
303 		pmap_remove_pages(vmspace_pmap(vm), vm_map_min(&vm->vm_map),
304 		    vm_map_max(&vm->vm_map));
305 		vm_page_unlock_queues();
306 		(void) vm_map_remove(&vm->vm_map, vm_map_min(&vm->vm_map),
307 		    vm_map_max(&vm->vm_map));
308 	}
309 
310 	sx_xlock(&proctree_lock);
311 	if (SESS_LEADER(p)) {
312 		register struct session *sp;
313 
314 		sp = p->p_session;
315 		if (sp->s_ttyvp) {
316 			/*
317 			 * Controlling process.
318 			 * Signal foreground pgrp,
319 			 * drain controlling terminal
320 			 * and revoke access to controlling terminal.
321 			 */
322 			if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
323 				tp = sp->s_ttyp;
324 				if (sp->s_ttyp->t_pgrp) {
325 					PGRP_LOCK(sp->s_ttyp->t_pgrp);
326 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
327 					PGRP_UNLOCK(sp->s_ttyp->t_pgrp);
328 				}
329 				/* XXX tp should be locked. */
330 				sx_xunlock(&proctree_lock);
331 				(void) ttywait(tp);
332 				sx_xlock(&proctree_lock);
333 				/*
334 				 * The tty could have been revoked
335 				 * if we blocked.
336 				 */
337 				if (sp->s_ttyvp) {
338 					ttyvp = sp->s_ttyvp;
339 					SESS_LOCK(p->p_session);
340 					sp->s_ttyvp = NULL;
341 					SESS_UNLOCK(p->p_session);
342 					sx_xunlock(&proctree_lock);
343 					VOP_REVOKE(ttyvp, REVOKEALL);
344 					vrele(ttyvp);
345 					sx_xlock(&proctree_lock);
346 				}
347 			}
348 			if (sp->s_ttyvp) {
349 				ttyvp = sp->s_ttyvp;
350 				SESS_LOCK(p->p_session);
351 				sp->s_ttyvp = NULL;
352 				SESS_UNLOCK(p->p_session);
353 				vrele(ttyvp);
354 			}
355 			/*
356 			 * s_ttyp is not zero'd; we use this to indicate
357 			 * that the session once had a controlling terminal.
358 			 * (for logging and informational purposes)
359 			 */
360 		}
361 		SESS_LOCK(p->p_session);
362 		sp->s_leader = NULL;
363 		SESS_UNLOCK(p->p_session);
364 	}
365 	fixjobc(p, p->p_pgrp, 0);
366 	sx_xunlock(&proctree_lock);
367 	(void)acct_process(td);
368 #ifdef KTRACE
369 	/*
370 	 * release trace file
371 	 */
372 	PROC_LOCK(p);
373 	mtx_lock(&ktrace_mtx);
374 	p->p_traceflag = 0;	/* don't trace the vrele() */
375 	tracevp = p->p_tracep;
376 	p->p_tracep = NULL;
377 	mtx_unlock(&ktrace_mtx);
378 	PROC_UNLOCK(p);
379 	if (tracevp != NULL)
380 		vrele(tracevp);
381 #endif
382 	/*
383 	 * Release reference to text vnode
384 	 */
385 	if ((vtmp = p->p_textvp) != NULL) {
386 		p->p_textvp = NULL;
387 		vrele(vtmp);
388 	}
389 
390 	/*
391 	 * Release our limits structure.
392 	 */
393 	mtx_assert(&Giant, MA_OWNED);
394 	if (--p->p_limit->p_refcnt == 0) {
395 		FREE(p->p_limit, M_SUBPROC);
396 		p->p_limit = NULL;
397 	}
398 
399 	/*
400 	 * Release this thread's reference to the ucred.  The actual proc
401 	 * reference will stay around until the proc is harvested by
402 	 * wait().  At this point the ucred is immutable (no other threads
403 	 * from this proc are around that can change it) so we leave the
404 	 * per-thread ucred pointer intact in case it is needed although
405 	 * in theory nothing should be using it at this point.
406 	 */
407 	crfree(td->td_ucred);
408 
409 	/*
410 	 * Remove proc from allproc queue and pidhash chain.
411 	 * Place onto zombproc.  Unlink from parent's child list.
412 	 */
413 	sx_xlock(&allproc_lock);
414 	LIST_REMOVE(p, p_list);
415 	LIST_INSERT_HEAD(&zombproc, p, p_list);
416 	LIST_REMOVE(p, p_hash);
417 	sx_xunlock(&allproc_lock);
418 
419 	sx_xlock(&proctree_lock);
420 	q = LIST_FIRST(&p->p_children);
421 	if (q != NULL)		/* only need this if any child is S_ZOMB */
422 		wakeup(initproc);
423 	for (; q != NULL; q = nq) {
424 		nq = LIST_NEXT(q, p_sibling);
425 		PROC_LOCK(q);
426 		proc_reparent(q, initproc);
427 		q->p_sigparent = SIGCHLD;
428 		/*
429 		 * Traced processes are killed
430 		 * since their existence means someone is screwing up.
431 		 */
432 		if (q->p_flag & P_TRACED) {
433 			q->p_flag &= ~P_TRACED;
434 			psignal(q, SIGKILL);
435 		}
436 		PROC_UNLOCK(q);
437 	}
438 
439 	/*
440 	 * Save exit status and final rusage info, adding in child rusage
441 	 * info and self times.
442 	 */
443 	PROC_LOCK(p);
444 	p->p_xstat = rv;
445 	*p->p_ru = p->p_stats->p_ru;
446 	mtx_lock_spin(&sched_lock);
447 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
448 	mtx_unlock_spin(&sched_lock);
449 	ruadd(p->p_ru, &p->p_stats->p_cru);
450 
451 	/*
452 	 * Notify interested parties of our demise.
453 	 */
454 	KNOTE(&p->p_klist, NOTE_EXIT);
455 
456 	/*
457 	 * Notify parent that we're gone.  If parent has the PS_NOCLDWAIT
458 	 * flag set, or if the handler is set to SIG_IGN, notify process
459 	 * 1 instead (and hope it will handle this situation).
460 	 */
461 	PROC_LOCK(p->p_pptr);
462 	if (p->p_pptr->p_procsig->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
463 		struct proc *pp;
464 
465 		pp = p->p_pptr;
466 		PROC_UNLOCK(pp);
467 		proc_reparent(p, initproc);
468 		PROC_LOCK(p->p_pptr);
469 		/*
470 		 * If this was the last child of our parent, notify
471 		 * parent, so in case he was wait(2)ing, he will
472 		 * continue.
473 		 */
474 		if (LIST_EMPTY(&pp->p_children))
475 			wakeup(pp);
476 	}
477 
478 	if (p->p_sigparent && p->p_pptr != initproc)
479 		psignal(p->p_pptr, p->p_sigparent);
480 	else
481 		psignal(p->p_pptr, SIGCHLD);
482 	PROC_UNLOCK(p->p_pptr);
483 
484 	/*
485 	 * If this is a kthread, then wakeup anyone waiting for it to exit.
486 	 */
487 	if (p->p_flag & P_KTHREAD)
488 		wakeup(p);
489 	PROC_UNLOCK(p);
490 
491 	/*
492 	 * Finally, call machine-dependent code to release the remaining
493 	 * resources including address space.
494 	 * The address space is released by "vmspace_exitfree(p)" in
495 	 * vm_waitproc().
496 	 */
497 	cpu_exit(td);
498 
499 	PROC_LOCK(p);
500 	PROC_LOCK(p->p_pptr);
501 	sx_xunlock(&proctree_lock);
502 	mtx_lock_spin(&sched_lock);
503 
504 	while (mtx_owned(&Giant))
505 		mtx_unlock(&Giant);
506 
507 	/*
508 	 * We have to wait until after releasing all locks before
509 	 * changing p_state.  If we block on a mutex then we will be
510 	 * back at SRUN when we resume and our parent will never
511 	 * harvest us.
512 	 */
513 	p->p_state = PRS_ZOMBIE;
514 
515 	wakeup(p->p_pptr);
516 	PROC_UNLOCK(p->p_pptr);
517 	cnt.v_swtch++;
518 	binuptime(PCPU_PTR(switchtime));
519 	PCPU_SET(switchticks, ticks);
520 
521 	cpu_sched_exit(td); /* XXXKSE check if this should be in thread_exit */
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(td, uap)
536 	struct thread *td;
537 	register struct owait_args /* {
538 		int     dummy;
539 	} */ *uap;
540 {
541 	struct wait_args w;
542 
543 	w.options = 0;
544 	w.rusage = NULL;
545 	w.pid = WAIT_ANY;
546 	w.status = NULL;
547 	return (wait1(td, &w, 1));
548 }
549 #endif /* COMPAT_43 */
550 
551 /*
552  * MPSAFE.  The dirty work is handled by wait1().
553  */
554 int
555 wait4(td, uap)
556 	struct thread *td;
557 	struct wait_args *uap;
558 {
559 
560 	return (wait1(td, uap, 0));
561 }
562 
563 /*
564  * MPSAFE
565  */
566 static int
567 wait1(td, uap, compat)
568 	register struct thread *td;
569 	register struct wait_args /* {
570 		int pid;
571 		int *status;
572 		int options;
573 		struct rusage *rusage;
574 	} */ *uap;
575 	int compat;
576 {
577 	struct rusage ru;
578 	int nfound;
579 	struct proc *p, *q, *t;
580 	int status, error;
581 
582 	q = td->td_proc;
583 	if (uap->pid == 0) {
584 		PROC_LOCK(q);
585 		uap->pid = -q->p_pgid;
586 		PROC_UNLOCK(q);
587 	}
588 	if (uap->options &~ (WUNTRACED|WNOHANG|WCONTINUED|WLINUXCLONE))
589 		return (EINVAL);
590 	mtx_lock(&Giant);
591 loop:
592 	nfound = 0;
593 	sx_xlock(&proctree_lock);
594 	LIST_FOREACH(p, &q->p_children, p_sibling) {
595 		PROC_LOCK(p);
596 		if (uap->pid != WAIT_ANY &&
597 		    p->p_pid != uap->pid && p->p_pgid != -uap->pid) {
598 			PROC_UNLOCK(p);
599 			continue;
600 		}
601 
602 		/*
603 		 * This special case handles a kthread spawned by linux_clone
604 		 * (see linux_misc.c).  The linux_wait4 and linux_waitpid
605 		 * functions need to be able to distinguish between waiting
606 		 * on a process and waiting on a thread.  It is a thread if
607 		 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
608 		 * signifies we want to wait for threads and not processes.
609 		 */
610 		if ((p->p_sigparent != SIGCHLD) ^
611 		    ((uap->options & WLINUXCLONE) != 0)) {
612 			PROC_UNLOCK(p);
613 			continue;
614 		}
615 
616 		nfound++;
617 		if (p->p_state == PRS_ZOMBIE) {
618 			/*
619 			 * Allow the scheduler to adjust the priority of the
620 			 * parent when a kseg is exiting.
621 			 */
622 			if (curthread->td_proc->p_pid != 1) {
623 				mtx_lock_spin(&sched_lock);
624 				sched_exit(curthread->td_ksegrp,
625 				    FIRST_KSEGRP_IN_PROC(p));
626 				mtx_unlock_spin(&sched_lock);
627 			}
628 
629 			td->td_retval[0] = p->p_pid;
630 #ifdef COMPAT_43
631 			if (compat)
632 				td->td_retval[1] = p->p_xstat;
633 			else
634 #endif
635 			if (uap->status) {
636 				status = p->p_xstat;	/* convert to int */
637 				PROC_UNLOCK(p);
638 				if ((error = copyout(&status,
639 				    uap->status, sizeof(status)))) {
640 					sx_xunlock(&proctree_lock);
641 					mtx_unlock(&Giant);
642 					return (error);
643 				}
644 				PROC_LOCK(p);
645 			}
646 			if (uap->rusage) {
647 				bcopy(p->p_ru, &ru, sizeof(ru));
648 				PROC_UNLOCK(p);
649 				if ((error = copyout(&ru,
650 				    uap->rusage, sizeof (struct rusage)))) {
651 					sx_xunlock(&proctree_lock);
652 					mtx_unlock(&Giant);
653 					return (error);
654 				}
655 			} else
656 				PROC_UNLOCK(p);
657 			/*
658 			 * If we got the child via a ptrace 'attach',
659 			 * we need to give it back to the old parent.
660 			 */
661 			if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) {
662 				PROC_LOCK(p);
663 				p->p_oppid = 0;
664 				proc_reparent(p, t);
665 				PROC_UNLOCK(p);
666 				psignal(t, SIGCHLD);
667 				wakeup(t);
668 				PROC_UNLOCK(t);
669 				sx_xunlock(&proctree_lock);
670 				mtx_unlock(&Giant);
671 				return (0);
672 			}
673 			/*
674 			 * Remove other references to this process to ensure
675 			 * we have an exclusive reference.
676 			 */
677 			leavepgrp(p);
678 
679 			sx_xlock(&allproc_lock);
680 			LIST_REMOVE(p, p_list);	/* off zombproc */
681 			sx_xunlock(&allproc_lock);
682 
683 			LIST_REMOVE(p, p_sibling);
684 			sx_xunlock(&proctree_lock);
685 
686 			/*
687 			 * As a side effect of this lock, we know that
688 			 * all other writes to this proc are visible now, so
689 			 * no more locking is needed for p.
690 			 */
691 			PROC_LOCK(p);
692 			p->p_xstat = 0;		/* XXX: why? */
693 			PROC_UNLOCK(p);
694 			PROC_LOCK(q);
695 			ruadd(&q->p_stats->p_cru, p->p_ru);
696 			PROC_UNLOCK(q);
697 			FREE(p->p_ru, M_ZOMBIE);
698 			p->p_ru = NULL;
699 
700 			/*
701 			 * Decrement the count of procs running with this uid.
702 			 */
703 			(void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
704 
705 			/*
706 			 * Free up credentials.
707 			 */
708 			crfree(p->p_ucred);
709 			p->p_ucred = NULL;	/* XXX: why? */
710 
711 			/*
712 			 * Remove unused arguments
713 			 */
714 			pargs_drop(p->p_args);
715 			p->p_args = NULL;
716 
717 			if (--p->p_procsig->ps_refcnt == 0) {
718 				if (p->p_sigacts != &p->p_uarea->u_sigacts)
719 					FREE(p->p_sigacts, M_SUBPROC);
720 				FREE(p->p_procsig, M_SUBPROC);
721 				p->p_procsig = NULL;
722 			}
723 
724 			/*
725 			 * do any thread-system specific cleanups
726 			 */
727 			thread_wait(p);
728 
729 			/*
730 			 * Give vm and machine-dependent layer a chance
731 			 * to free anything that cpu_exit couldn't
732 			 * release while still running in process context.
733 			 */
734 			vm_waitproc(p);
735 			mtx_destroy(&p->p_mtx);
736 #ifdef MAC
737 			mac_destroy_proc(p);
738 #endif
739 			KASSERT(FIRST_THREAD_IN_PROC(p),
740 			    ("wait1: no residual thread!"));
741 			uma_zfree(proc_zone, p);
742 			sx_xlock(&allproc_lock);
743 			nprocs--;
744 			sx_xunlock(&allproc_lock);
745 			mtx_unlock(&Giant);
746 			return (0);
747 		}
748 		if (P_SHOULDSTOP(p) && ((p->p_flag & P_WAITED) == 0) &&
749 		    (p->p_flag & P_TRACED || uap->options & WUNTRACED)) {
750 			p->p_flag |= P_WAITED;
751 			sx_xunlock(&proctree_lock);
752 			td->td_retval[0] = p->p_pid;
753 #ifdef COMPAT_43
754 			if (compat) {
755 				td->td_retval[1] = W_STOPCODE(p->p_xstat);
756 				PROC_UNLOCK(p);
757 				error = 0;
758 			} else
759 #endif
760 			if (uap->status) {
761 				status = W_STOPCODE(p->p_xstat);
762 				PROC_UNLOCK(p);
763 				error = copyout(&status,
764 					uap->status, sizeof(status));
765 			} else {
766 				PROC_UNLOCK(p);
767 				error = 0;
768 			}
769 			mtx_unlock(&Giant);
770 			return (error);
771 		}
772 		if (uap->options & WCONTINUED && (p->p_flag & P_CONTINUED)) {
773 			sx_xunlock(&proctree_lock);
774 			td->td_retval[0] = p->p_pid;
775 			p->p_flag &= ~P_CONTINUED;
776 			PROC_UNLOCK(p);
777 
778 			if (uap->status) {
779 				status = SIGCONT;
780 				error = copyout(&status,
781 				    uap->status, sizeof(status));
782 			} else
783 				error = 0;
784 
785 			mtx_unlock(&Giant);
786 			return (error);
787 		}
788 		PROC_UNLOCK(p);
789 	}
790 	if (nfound == 0) {
791 		sx_xunlock(&proctree_lock);
792 		mtx_unlock(&Giant);
793 		return (ECHILD);
794 	}
795 	if (uap->options & WNOHANG) {
796 		sx_xunlock(&proctree_lock);
797 		td->td_retval[0] = 0;
798 		mtx_unlock(&Giant);
799 		return (0);
800 	}
801 	PROC_LOCK(q);
802 	sx_xunlock(&proctree_lock);
803 	error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
804 	PROC_UNLOCK(q);
805 	if (error) {
806 		mtx_unlock(&Giant);
807 		return (error);
808 	}
809 	goto loop;
810 }
811 
812 /*
813  * Make process 'parent' the new parent of process 'child'.
814  * Must be called with an exclusive hold of proctree lock.
815  */
816 void
817 proc_reparent(child, parent)
818 	register struct proc *child;
819 	register struct proc *parent;
820 {
821 
822 	sx_assert(&proctree_lock, SX_XLOCKED);
823 	PROC_LOCK_ASSERT(child, MA_OWNED);
824 	if (child->p_pptr == parent)
825 		return;
826 
827 	LIST_REMOVE(child, p_sibling);
828 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
829 	child->p_pptr = parent;
830 }
831 
832 /*
833  * The next two functions are to handle adding/deleting items on the
834  * exit callout list
835  *
836  * at_exit():
837  * Take the arguments given and put them onto the exit callout list,
838  * However first make sure that it's not already there.
839  * returns 0 on success.
840  */
841 
842 int
843 at_exit(function)
844 	exitlist_fn function;
845 {
846 	struct exitlist *ep;
847 
848 #ifdef INVARIANTS
849 	/* Be noisy if the programmer has lost track of things */
850 	if (rm_at_exit(function))
851 		printf("WARNING: exit callout entry (%p) already present\n",
852 		    function);
853 #endif
854 	ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT);
855 	if (ep == NULL)
856 		return (ENOMEM);
857 	ep->function = function;
858 	TAILQ_INSERT_TAIL(&exit_list, ep, next);
859 	return (0);
860 }
861 
862 /*
863  * Scan the exit callout list for the given item and remove it.
864  * Returns the number of items removed (0 or 1)
865  */
866 int
867 rm_at_exit(function)
868 	exitlist_fn function;
869 {
870 	struct exitlist *ep;
871 
872 	TAILQ_FOREACH(ep, &exit_list, next) {
873 		if (ep->function == function) {
874 			TAILQ_REMOVE(&exit_list, ep, next);
875 			free(ep, M_ATEXIT);
876 			return (1);
877 		}
878 	}
879 	return (0);
880 }
881