xref: /freebsd/sys/kern/kern_exit.c (revision 721351876cd4d3a8a700f62d2061331fa951a488)
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_kdtrace.h"
42 #include "opt_ktrace.h"
43 #include "opt_mac.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/sysproto.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/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/sbuf.h>
61 #include <sys/signalvar.h>
62 #include <sys/sched.h>
63 #include <sys/sx.h>
64 #include <sys/syscallsubr.h>
65 #include <sys/syslog.h>
66 #include <sys/ptrace.h>
67 #include <sys/acct.h>		/* for acct_process() function prototype */
68 #include <sys/filedesc.h>
69 #include <sys/sdt.h>
70 #include <sys/shm.h>
71 #include <sys/sem.h>
72 #ifdef KTRACE
73 #include <sys/ktrace.h>
74 #endif
75 
76 #include <security/audit/audit.h>
77 #include <security/mac/mac_framework.h>
78 
79 #include <vm/vm.h>
80 #include <vm/vm_extern.h>
81 #include <vm/vm_param.h>
82 #include <vm/pmap.h>
83 #include <vm/vm_map.h>
84 #include <vm/vm_page.h>
85 #include <vm/uma.h>
86 
87 #ifdef KDTRACE_HOOKS
88 #include <sys/dtrace_bsd.h>
89 dtrace_execexit_func_t	dtrace_fasttrap_exit;
90 #endif
91 
92 SDT_PROVIDER_DECLARE(proc);
93 SDT_PROBE_DEFINE(proc, kernel, , exit);
94 SDT_PROBE_ARGTYPE(proc, kernel, , exit, 0, "int");
95 
96 /* Required to be non-static for SysVR4 emulator */
97 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status");
98 
99 /* Hook for NFS teardown procedure. */
100 void (*nlminfo_release_p)(struct proc *p);
101 
102 /*
103  * exit -- death of process.
104  */
105 void
106 sys_exit(struct thread *td, struct sys_exit_args *uap)
107 {
108 
109 	exit1(td, W_EXITCODE(uap->rval, 0));
110 	/* NOTREACHED */
111 }
112 
113 /*
114  * Exit: deallocate address space and other resources, change proc state to
115  * zombie, and unlink proc from allproc and parent's lists.  Save exit status
116  * and rusage for wait().  Check for child processes and orphan them.
117  */
118 void
119 exit1(struct thread *td, int rv)
120 {
121 	struct proc *p, *nq, *q;
122 	struct tty *tp;
123 	struct vnode *ttyvp;
124 	struct vnode *vtmp;
125 #ifdef KTRACE
126 	struct vnode *tracevp;
127 	struct ucred *tracecred;
128 #endif
129 	struct plimit *plim;
130 	int locked;
131 
132 	mtx_assert(&Giant, MA_NOTOWNED);
133 
134 	p = td->td_proc;
135 	if (p == initproc) {
136 		printf("init died (signal %d, exit %d)\n",
137 		    WTERMSIG(rv), WEXITSTATUS(rv));
138 		panic("Going nowhere without my init!");
139 	}
140 
141 	/*
142 	 * MUST abort all other threads before proceeding past here.
143 	 */
144 	PROC_LOCK(p);
145 	while (p->p_flag & P_HADTHREADS) {
146 		/*
147 		 * First check if some other thread got here before us..
148 		 * if so, act apropriatly, (exit or suspend);
149 		 */
150 		thread_suspend_check(0);
151 
152 		/*
153 		 * Kill off the other threads. This requires
154 		 * some co-operation from other parts of the kernel
155 		 * so it may not be instantaneous.  With this state set
156 		 * any thread entering the kernel from userspace will
157 		 * thread_exit() in trap().  Any thread attempting to
158 		 * sleep will return immediately with EINTR or EWOULDBLOCK
159 		 * which will hopefully force them to back out to userland
160 		 * freeing resources as they go.  Any thread attempting
161 		 * to return to userland will thread_exit() from userret().
162 		 * thread_exit() will unsuspend us when the last of the
163 		 * other threads exits.
164 		 * If there is already a thread singler after resumption,
165 		 * calling thread_single will fail; in that case, we just
166 		 * re-check all suspension request, the thread should
167 		 * either be suspended there or exit.
168 		 */
169 		if (! thread_single(SINGLE_EXIT))
170 			break;
171 
172 		/*
173 		 * All other activity in this process is now stopped.
174 		 * Threading support has been turned off.
175 		 */
176 	}
177 	KASSERT(p->p_numthreads == 1,
178 	    ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
179 	/*
180 	 * Wakeup anyone in procfs' PIOCWAIT.  They should have a hold
181 	 * on our vmspace, so we should block below until they have
182 	 * released their reference to us.  Note that if they have
183 	 * requested S_EXIT stops we will block here until they ack
184 	 * via PIOCCONT.
185 	 */
186 	_STOPEVENT(p, S_EXIT, rv);
187 
188 	/*
189 	 * Note that we are exiting and do another wakeup of anyone in
190 	 * PIOCWAIT in case they aren't listening for S_EXIT stops or
191 	 * decided to wait again after we told them we are exiting.
192 	 */
193 	p->p_flag |= P_WEXIT;
194 	wakeup(&p->p_stype);
195 
196 	/*
197 	 * Wait for any processes that have a hold on our vmspace to
198 	 * release their reference.
199 	 */
200 	while (p->p_lock > 0)
201 		msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
202 
203 	PROC_UNLOCK(p);
204 	/* Drain the limit callout while we don't have the proc locked */
205 	callout_drain(&p->p_limco);
206 
207 #ifdef AUDIT
208 	/*
209 	 * The Sun BSM exit token contains two components: an exit status as
210 	 * passed to exit(), and a return value to indicate what sort of exit
211 	 * it was.  The exit status is WEXITSTATUS(rv), but it's not clear
212 	 * what the return value is.
213 	 */
214 	AUDIT_ARG(exit, WEXITSTATUS(rv), 0);
215 	AUDIT_SYSCALL_EXIT(0, td);
216 #endif
217 
218 	/* Are we a task leader? */
219 	if (p == p->p_leader) {
220 		mtx_lock(&ppeers_lock);
221 		q = p->p_peers;
222 		while (q != NULL) {
223 			PROC_LOCK(q);
224 			psignal(q, SIGKILL);
225 			PROC_UNLOCK(q);
226 			q = q->p_peers;
227 		}
228 		while (p->p_peers != NULL)
229 			msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
230 		mtx_unlock(&ppeers_lock);
231 	}
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 	EVENTHANDLER_INVOKE(process_exit, p);
239 
240 	/*
241 	 * If parent is waiting for us to exit or exec,
242 	 * P_PPWAIT is set; we will wakeup the parent below.
243 	 */
244 	PROC_LOCK(p);
245 	stopprofclock(p);
246 	p->p_flag &= ~(P_TRACED | P_PPWAIT);
247 
248 	/*
249 	 * Stop the real interval timer.  If the handler is currently
250 	 * executing, prevent it from rearming itself and let it finish.
251 	 */
252 	if (timevalisset(&p->p_realtimer.it_value) &&
253 	    callout_stop(&p->p_itcallout) == 0) {
254 		timevalclear(&p->p_realtimer.it_interval);
255 		msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
256 		KASSERT(!timevalisset(&p->p_realtimer.it_value),
257 		    ("realtime timer is still armed"));
258 	}
259 	PROC_UNLOCK(p);
260 
261 	/*
262 	 * Reset any sigio structures pointing to us as a result of
263 	 * F_SETOWN with our pid.
264 	 */
265 	funsetownlst(&p->p_sigiolst);
266 
267 	/*
268 	 * If this process has an nlminfo data area (for lockd), release it
269 	 */
270 	if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
271 		(*nlminfo_release_p)(p);
272 
273 	/*
274 	 * Close open files and release open-file table.
275 	 * This may block!
276 	 */
277 	fdfree(td);
278 
279 	/*
280 	 * If this thread tickled GEOM, we need to wait for the giggling to
281 	 * stop before we return to userland
282 	 */
283 	if (td->td_pflags & TDP_GEOM)
284 		g_waitidle();
285 
286 	/*
287 	 * Remove ourself from our leader's peer list and wake our leader.
288 	 */
289 	mtx_lock(&ppeers_lock);
290 	if (p->p_leader->p_peers) {
291 		q = p->p_leader;
292 		while (q->p_peers != p)
293 			q = q->p_peers;
294 		q->p_peers = p->p_peers;
295 		wakeup(p->p_leader);
296 	}
297 	mtx_unlock(&ppeers_lock);
298 
299 	vmspace_exit(td);
300 
301 	mtx_lock(&Giant);	/* XXX TTY */
302 	sx_xlock(&proctree_lock);
303 	if (SESS_LEADER(p)) {
304 		struct session *sp;
305 
306 		sp = p->p_session;
307 		if (sp->s_ttyvp) {
308 			/*
309 			 * Controlling process.
310 			 * Signal foreground pgrp,
311 			 * drain controlling terminal
312 			 * and revoke access to controlling terminal.
313 			 */
314 			if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
315 				tp = sp->s_ttyp;
316 				if (sp->s_ttyp->t_pgrp) {
317 					PGRP_LOCK(sp->s_ttyp->t_pgrp);
318 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
319 					PGRP_UNLOCK(sp->s_ttyp->t_pgrp);
320 				}
321 				/* XXX tp should be locked. */
322 				sx_xunlock(&proctree_lock);
323 				(void) ttywait(tp);
324 				sx_xlock(&proctree_lock);
325 				/*
326 				 * The tty could have been revoked
327 				 * if we blocked.
328 				 */
329 				if (sp->s_ttyvp) {
330 					ttyvp = sp->s_ttyvp;
331 					SESS_LOCK(p->p_session);
332 					sp->s_ttyvp = NULL;
333 					SESS_UNLOCK(p->p_session);
334 					sx_xunlock(&proctree_lock);
335 					VOP_LOCK(ttyvp, LK_EXCLUSIVE);
336 					VOP_REVOKE(ttyvp, REVOKEALL);
337 					vput(ttyvp);
338 					sx_xlock(&proctree_lock);
339 				}
340 			}
341 			if (sp->s_ttyvp) {
342 				ttyvp = sp->s_ttyvp;
343 				SESS_LOCK(p->p_session);
344 				sp->s_ttyvp = NULL;
345 				SESS_UNLOCK(p->p_session);
346 				vrele(ttyvp);
347 			}
348 			/*
349 			 * s_ttyp is not zero'd; we use this to indicate
350 			 * that the session once had a controlling terminal.
351 			 * (for logging and informational purposes)
352 			 */
353 		}
354 		SESS_LOCK(p->p_session);
355 		sp->s_leader = NULL;
356 		SESS_UNLOCK(p->p_session);
357 	}
358 	fixjobc(p, p->p_pgrp, 0);
359 	sx_xunlock(&proctree_lock);
360 	(void)acct_process(td);
361 	mtx_unlock(&Giant);
362 #ifdef KTRACE
363 	/*
364 	 * Disable tracing, then drain any pending records and release
365 	 * the trace file.
366 	 */
367 	if (p->p_traceflag != 0) {
368 		PROC_LOCK(p);
369 		mtx_lock(&ktrace_mtx);
370 		p->p_traceflag = 0;
371 		mtx_unlock(&ktrace_mtx);
372 		PROC_UNLOCK(p);
373 		ktrprocexit(td);
374 		PROC_LOCK(p);
375 		mtx_lock(&ktrace_mtx);
376 		tracevp = p->p_tracevp;
377 		p->p_tracevp = NULL;
378 		tracecred = p->p_tracecred;
379 		p->p_tracecred = NULL;
380 		mtx_unlock(&ktrace_mtx);
381 		PROC_UNLOCK(p);
382 		if (tracevp != NULL) {
383 			locked = VFS_LOCK_GIANT(tracevp->v_mount);
384 			vrele(tracevp);
385 			VFS_UNLOCK_GIANT(locked);
386 		}
387 		if (tracecred != NULL)
388 			crfree(tracecred);
389 	}
390 #endif
391 	/*
392 	 * Release reference to text vnode
393 	 */
394 	if ((vtmp = p->p_textvp) != NULL) {
395 		p->p_textvp = NULL;
396 		locked = VFS_LOCK_GIANT(vtmp->v_mount);
397 		vrele(vtmp);
398 		VFS_UNLOCK_GIANT(locked);
399 	}
400 
401 	/*
402 	 * Release our limits structure.
403 	 */
404 	PROC_LOCK(p);
405 	plim = p->p_limit;
406 	p->p_limit = NULL;
407 	PROC_UNLOCK(p);
408 	lim_free(plim);
409 
410 	/*
411 	 * Remove proc from allproc queue and pidhash chain.
412 	 * Place onto zombproc.  Unlink from parent's child list.
413 	 */
414 	sx_xlock(&allproc_lock);
415 	LIST_REMOVE(p, p_list);
416 	LIST_INSERT_HEAD(&zombproc, p, p_list);
417 	LIST_REMOVE(p, p_hash);
418 	sx_xunlock(&allproc_lock);
419 
420 	/*
421 	 * Call machine-dependent code to release any
422 	 * machine-dependent resources other than the address space.
423 	 * The address space is released by "vmspace_exitfree(p)" in
424 	 * vm_waitproc().
425 	 */
426 	cpu_exit(td);
427 
428 	WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
429 
430 	/*
431 	 * Reparent all of our children to init.
432 	 */
433 	sx_xlock(&proctree_lock);
434 	q = LIST_FIRST(&p->p_children);
435 	if (q != NULL)		/* only need this if any child is S_ZOMB */
436 		wakeup(initproc);
437 	for (; q != NULL; q = nq) {
438 		nq = LIST_NEXT(q, p_sibling);
439 		PROC_LOCK(q);
440 		proc_reparent(q, initproc);
441 		q->p_sigparent = SIGCHLD;
442 		/*
443 		 * Traced processes are killed
444 		 * since their existence means someone is screwing up.
445 		 */
446 		if (q->p_flag & P_TRACED) {
447 			q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
448 			psignal(q, SIGKILL);
449 		}
450 		PROC_UNLOCK(q);
451 	}
452 
453 	/* Save exit status. */
454 	PROC_LOCK(p);
455 	p->p_xstat = rv;
456 	p->p_xthread = td;
457 
458 #ifdef KDTRACE_HOOKS
459 	/*
460 	 * Tell the DTrace fasttrap provider about the exit if it
461 	 * has declared an interest.
462 	 */
463 	if (dtrace_fasttrap_exit)
464 		dtrace_fasttrap_exit(p);
465 #endif
466 
467 	/*
468 	 * Notify interested parties of our demise.
469 	 */
470 	KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
471 
472 #ifdef KDTRACE_HOOKS
473 	int reason = CLD_EXITED;
474 	if (WCOREDUMP(rv))
475 		reason = CLD_DUMPED;
476 	else if (WIFSIGNALED(rv))
477 		reason = CLD_KILLED;
478 	SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
479 #endif
480 
481 	/*
482 	 * Just delete all entries in the p_klist. At this point we won't
483 	 * report any more events, and there are nasty race conditions that
484 	 * can beat us if we don't.
485 	 */
486 	knlist_clear(&p->p_klist, 1);
487 
488 	/*
489 	 * Notify parent that we're gone.  If parent has the PS_NOCLDWAIT
490 	 * flag set, or if the handler is set to SIG_IGN, notify process
491 	 * 1 instead (and hope it will handle this situation).
492 	 */
493 	PROC_LOCK(p->p_pptr);
494 	mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
495 	if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
496 		struct proc *pp;
497 
498 		mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
499 		pp = p->p_pptr;
500 		PROC_UNLOCK(pp);
501 		proc_reparent(p, initproc);
502 		p->p_sigparent = SIGCHLD;
503 		PROC_LOCK(p->p_pptr);
504 		/*
505 		 * If this was the last child of our parent, notify
506 		 * parent, so in case he was wait(2)ing, he will
507 		 * continue.
508 		 */
509 		if (LIST_EMPTY(&pp->p_children))
510 			wakeup(pp);
511 	} else
512 		mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
513 
514 	if (p->p_pptr == initproc)
515 		psignal(p->p_pptr, SIGCHLD);
516 	else if (p->p_sigparent != 0) {
517 		if (p->p_sigparent == SIGCHLD)
518 			childproc_exited(p);
519 		else	/* LINUX thread */
520 			psignal(p->p_pptr, p->p_sigparent);
521 	}
522 	sx_xunlock(&proctree_lock);
523 
524 	/*
525 	 * The state PRS_ZOMBIE prevents other proesses from sending
526 	 * signal to the process, to avoid memory leak, we free memory
527 	 * for signal queue at the time when the state is set.
528 	 */
529 	sigqueue_flush(&p->p_sigqueue);
530 	sigqueue_flush(&td->td_sigqueue);
531 
532 	/*
533 	 * We have to wait until after acquiring all locks before
534 	 * changing p_state.  We need to avoid all possible context
535 	 * switches (including ones from blocking on a mutex) while
536 	 * marked as a zombie.  We also have to set the zombie state
537 	 * before we release the parent process' proc lock to avoid
538 	 * a lost wakeup.  So, we first call wakeup, then we grab the
539 	 * sched lock, update the state, and release the parent process'
540 	 * proc lock.
541 	 */
542 	wakeup(p->p_pptr);
543 	sched_exit(p->p_pptr, td);
544 	PROC_SLOCK(p);
545 	p->p_state = PRS_ZOMBIE;
546 	PROC_UNLOCK(p->p_pptr);
547 
548 	/*
549 	 * Hopefully no one will try to deliver a signal to the process this
550 	 * late in the game.
551 	 */
552 	knlist_destroy(&p->p_klist);
553 
554 	/*
555 	 * Save our children's rusage information in our exit rusage.
556 	 */
557 	ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
558 
559 	/*
560 	 * Make sure the scheduler takes this thread out of its tables etc.
561 	 * This will also release this thread's reference to the ucred.
562 	 * Other thread parts to release include pcb bits and such.
563 	 */
564 	thread_exit();
565 }
566 
567 
568 #ifndef _SYS_SYSPROTO_H_
569 struct abort2_args {
570 	char *why;
571 	int nargs;
572 	void **args;
573 };
574 #endif
575 
576 int
577 abort2(struct thread *td, struct abort2_args *uap)
578 {
579 	struct proc *p = td->td_proc;
580 	struct sbuf *sb;
581 	void *uargs[16];
582 	int error, i, sig;
583 
584 	error = 0;	/* satisfy compiler */
585 
586 	/*
587 	 * Do it right now so we can log either proper call of abort2(), or
588 	 * note, that invalid argument was passed. 512 is big enough to
589 	 * handle 16 arguments' descriptions with additional comments.
590 	 */
591 	sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
592 	sbuf_clear(sb);
593 	sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
594 	    p->p_comm, p->p_pid, td->td_ucred->cr_uid);
595 	/*
596 	 * Since we can't return from abort2(), send SIGKILL in cases, where
597 	 * abort2() was called improperly
598 	 */
599 	sig = SIGKILL;
600 	/* Prevent from DoSes from user-space. */
601 	if (uap->nargs < 0 || uap->nargs > 16)
602 		goto out;
603 	if (uap->nargs > 0) {
604 		if (uap->args == NULL)
605 			goto out;
606 		error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
607 		if (error != 0)
608 			goto out;
609 	}
610 	/*
611 	 * Limit size of 'reason' string to 128. Will fit even when
612 	 * maximal number of arguments was chosen to be logged.
613 	 */
614 	if (uap->why != NULL) {
615 		error = sbuf_copyin(sb, uap->why, 128);
616 		if (error < 0)
617 			goto out;
618 	} else {
619 		sbuf_printf(sb, "(null)");
620 	}
621 	if (uap->nargs > 0) {
622 		sbuf_printf(sb, "(");
623 		for (i = 0;i < uap->nargs; i++)
624 			sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
625 		sbuf_printf(sb, ")");
626 	}
627 	/*
628 	 * Final stage: arguments were proper, string has been
629 	 * successfully copied from userspace, and copying pointers
630 	 * from user-space succeed.
631 	 */
632 	sig = SIGABRT;
633 out:
634 	if (sig == SIGKILL) {
635 		sbuf_trim(sb);
636 		sbuf_printf(sb, " (Reason text inaccessible)");
637 	}
638 	sbuf_cat(sb, "\n");
639 	sbuf_finish(sb);
640 	log(LOG_INFO, "%s", sbuf_data(sb));
641 	sbuf_delete(sb);
642 	exit1(td, W_EXITCODE(0, sig));
643 	return (0);
644 }
645 
646 
647 #ifdef COMPAT_43
648 /*
649  * The dirty work is handled by kern_wait().
650  */
651 int
652 owait(struct thread *td, struct owait_args *uap __unused)
653 {
654 	int error, status;
655 
656 	error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
657 	if (error == 0)
658 		td->td_retval[1] = status;
659 	return (error);
660 }
661 #endif /* COMPAT_43 */
662 
663 /*
664  * The dirty work is handled by kern_wait().
665  */
666 int
667 wait4(struct thread *td, struct wait_args *uap)
668 {
669 	struct rusage ru, *rup;
670 	int error, status;
671 
672 	if (uap->rusage != NULL)
673 		rup = &ru;
674 	else
675 		rup = NULL;
676 	error = kern_wait(td, uap->pid, &status, uap->options, rup);
677 	if (uap->status != NULL && error == 0)
678 		error = copyout(&status, uap->status, sizeof(status));
679 	if (uap->rusage != NULL && error == 0)
680 		error = copyout(&ru, uap->rusage, sizeof(struct rusage));
681 	return (error);
682 }
683 
684 int
685 kern_wait(struct thread *td, pid_t pid, int *status, int options,
686     struct rusage *rusage)
687 {
688 	struct proc *p, *q, *t;
689 	int error, nfound;
690 
691 	AUDIT_ARG(pid, pid);
692 
693 	q = td->td_proc;
694 	if (pid == 0) {
695 		PROC_LOCK(q);
696 		pid = -q->p_pgid;
697 		PROC_UNLOCK(q);
698 	}
699 	if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WLINUXCLONE))
700 		return (EINVAL);
701 loop:
702 	if (q->p_flag & P_STATCHILD) {
703 		PROC_LOCK(q);
704 		q->p_flag &= ~P_STATCHILD;
705 		PROC_UNLOCK(q);
706 	}
707 	nfound = 0;
708 	sx_xlock(&proctree_lock);
709 	LIST_FOREACH(p, &q->p_children, p_sibling) {
710 		PROC_LOCK(p);
711 		if (pid != WAIT_ANY &&
712 		    p->p_pid != pid && p->p_pgid != -pid) {
713 			PROC_UNLOCK(p);
714 			continue;
715 		}
716 		if (p_canwait(td, p)) {
717 			PROC_UNLOCK(p);
718 			continue;
719 		}
720 
721 		/*
722 		 * This special case handles a kthread spawned by linux_clone
723 		 * (see linux_misc.c).  The linux_wait4 and linux_waitpid
724 		 * functions need to be able to distinguish between waiting
725 		 * on a process and waiting on a thread.  It is a thread if
726 		 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
727 		 * signifies we want to wait for threads and not processes.
728 		 */
729 		if ((p->p_sigparent != SIGCHLD) ^
730 		    ((options & WLINUXCLONE) != 0)) {
731 			PROC_UNLOCK(p);
732 			continue;
733 		}
734 
735 		nfound++;
736 		PROC_SLOCK(p);
737 		if (p->p_state == PRS_ZOMBIE) {
738 			if (rusage) {
739 				*rusage = p->p_ru;
740 				calcru(p, &rusage->ru_utime, &rusage->ru_stime);
741 			}
742 			PROC_SUNLOCK(p);
743 			td->td_retval[0] = p->p_pid;
744 			if (status)
745 				*status = p->p_xstat;	/* convert to int */
746 			PROC_LOCK(q);
747 			sigqueue_take(p->p_ksi);
748 			PROC_UNLOCK(q);
749 
750 			/*
751 			 * If we got the child via a ptrace 'attach',
752 			 * we need to give it back to the old parent.
753 			 */
754 			PROC_UNLOCK(p);
755 			if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) {
756 				PROC_LOCK(p);
757 				p->p_oppid = 0;
758 				proc_reparent(p, t);
759 				PROC_UNLOCK(p);
760 				tdsignal(t, NULL, SIGCHLD, p->p_ksi);
761 				wakeup(t);
762 				PROC_UNLOCK(t);
763 				sx_xunlock(&proctree_lock);
764 				return (0);
765 			}
766 
767 			/*
768 			 * Remove other references to this process to ensure
769 			 * we have an exclusive reference.
770 			 */
771 			sx_xlock(&allproc_lock);
772 			LIST_REMOVE(p, p_list);	/* off zombproc */
773 			sx_xunlock(&allproc_lock);
774 			LIST_REMOVE(p, p_sibling);
775 			leavepgrp(p);
776 			sx_xunlock(&proctree_lock);
777 
778 			/*
779 			 * As a side effect of this lock, we know that
780 			 * all other writes to this proc are visible now, so
781 			 * no more locking is needed for p.
782 			 */
783 			PROC_LOCK(p);
784 			p->p_xstat = 0;		/* XXX: why? */
785 			PROC_UNLOCK(p);
786 			PROC_LOCK(q);
787 			ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru,
788 			    &p->p_rux);
789 			PROC_UNLOCK(q);
790 
791 			/*
792 			 * Decrement the count of procs running with this uid.
793 			 */
794 			(void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
795 
796 			/*
797 			 * Free credentials, arguments, and sigacts.
798 			 */
799 			crfree(p->p_ucred);
800 			p->p_ucred = NULL;
801 			pargs_drop(p->p_args);
802 			p->p_args = NULL;
803 			sigacts_free(p->p_sigacts);
804 			p->p_sigacts = NULL;
805 
806 			/*
807 			 * Do any thread-system specific cleanups.
808 			 */
809 			thread_wait(p);
810 
811 			/*
812 			 * Give vm and machine-dependent layer a chance
813 			 * to free anything that cpu_exit couldn't
814 			 * release while still running in process context.
815 			 */
816 			vm_waitproc(p);
817 #ifdef MAC
818 			mac_proc_destroy(p);
819 #endif
820 			KASSERT(FIRST_THREAD_IN_PROC(p),
821 			    ("kern_wait: no residual thread!"));
822 			uma_zfree(proc_zone, p);
823 			sx_xlock(&allproc_lock);
824 			nprocs--;
825 			sx_xunlock(&allproc_lock);
826 			return (0);
827 		}
828 		if ((p->p_flag & P_STOPPED_SIG) &&
829 		    (p->p_suspcount == p->p_numthreads) &&
830 		    (p->p_flag & P_WAITED) == 0 &&
831 		    (p->p_flag & P_TRACED || options & WUNTRACED)) {
832 			PROC_SUNLOCK(p);
833 			p->p_flag |= P_WAITED;
834 			sx_xunlock(&proctree_lock);
835 			td->td_retval[0] = p->p_pid;
836 			if (status)
837 				*status = W_STOPCODE(p->p_xstat);
838 
839 			PROC_LOCK(q);
840 			sigqueue_take(p->p_ksi);
841 			PROC_UNLOCK(q);
842 			PROC_UNLOCK(p);
843 
844 			return (0);
845 		}
846 		PROC_SUNLOCK(p);
847 		if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) {
848 			sx_xunlock(&proctree_lock);
849 			td->td_retval[0] = p->p_pid;
850 			p->p_flag &= ~P_CONTINUED;
851 
852 			PROC_LOCK(q);
853 			sigqueue_take(p->p_ksi);
854 			PROC_UNLOCK(q);
855 			PROC_UNLOCK(p);
856 
857 			if (status)
858 				*status = SIGCONT;
859 			return (0);
860 		}
861 		PROC_UNLOCK(p);
862 	}
863 	if (nfound == 0) {
864 		sx_xunlock(&proctree_lock);
865 		return (ECHILD);
866 	}
867 	if (options & WNOHANG) {
868 		sx_xunlock(&proctree_lock);
869 		td->td_retval[0] = 0;
870 		return (0);
871 	}
872 	PROC_LOCK(q);
873 	sx_xunlock(&proctree_lock);
874 	if (q->p_flag & P_STATCHILD) {
875 		q->p_flag &= ~P_STATCHILD;
876 		error = 0;
877 	} else
878 		error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
879 	PROC_UNLOCK(q);
880 	if (error)
881 		return (error);
882 	goto loop;
883 }
884 
885 /*
886  * Make process 'parent' the new parent of process 'child'.
887  * Must be called with an exclusive hold of proctree lock.
888  */
889 void
890 proc_reparent(struct proc *child, struct proc *parent)
891 {
892 
893 	sx_assert(&proctree_lock, SX_XLOCKED);
894 	PROC_LOCK_ASSERT(child, MA_OWNED);
895 	if (child->p_pptr == parent)
896 		return;
897 
898 	PROC_LOCK(child->p_pptr);
899 	sigqueue_take(child->p_ksi);
900 	PROC_UNLOCK(child->p_pptr);
901 	LIST_REMOVE(child, p_sibling);
902 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
903 	child->p_pptr = parent;
904 }
905