xref: /netbsd-src/sys/kern/kern_exit.c (revision 4472dbe5e3bd91ef2540bada7a7ca7384627ff9b)
1 /*	$NetBSD: kern_exit.c,v 1.79 2000/05/27 00:11:12 sommerfeld Exp $	*/
2 
3 /*-
4  * Copyright (c) 1998, 1999 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1982, 1986, 1989, 1991, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  * (c) UNIX System Laboratories, Inc.
44  * All or some portions of this file are derived from material licensed
45  * to the University of California by American Telephone and Telegraph
46  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
47  * the permission of UNIX System Laboratories, Inc.
48  *
49  * Redistribution and use in source and binary forms, with or without
50  * modification, are permitted provided that the following conditions
51  * are met:
52  * 1. Redistributions of source code must retain the above copyright
53  *    notice, this list of conditions and the following disclaimer.
54  * 2. Redistributions in binary form must reproduce the above copyright
55  *    notice, this list of conditions and the following disclaimer in the
56  *    documentation and/or other materials provided with the distribution.
57  * 3. All advertising materials mentioning features or use of this software
58  *    must display the following acknowledgement:
59  *	This product includes software developed by the University of
60  *	California, Berkeley and its contributors.
61  * 4. Neither the name of the University nor the names of its contributors
62  *    may be used to endorse or promote products derived from this software
63  *    without specific prior written permission.
64  *
65  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
66  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
67  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
68  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
69  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
70  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
71  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
72  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
73  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
74  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
75  * SUCH DAMAGE.
76  *
77  *	@(#)kern_exit.c	8.10 (Berkeley) 2/23/95
78  */
79 
80 #include "opt_ktrace.h"
81 #include "opt_sysv.h"
82 
83 #include <sys/param.h>
84 #include <sys/systm.h>
85 #include <sys/map.h>
86 #include <sys/ioctl.h>
87 #include <sys/proc.h>
88 #include <sys/tty.h>
89 #include <sys/time.h>
90 #include <sys/resource.h>
91 #include <sys/kernel.h>
92 #include <sys/ktrace.h>
93 #include <sys/proc.h>
94 #include <sys/buf.h>
95 #include <sys/wait.h>
96 #include <sys/file.h>
97 #include <sys/vnode.h>
98 #include <sys/syslog.h>
99 #include <sys/malloc.h>
100 #include <sys/pool.h>
101 #include <sys/resourcevar.h>
102 #include <sys/ptrace.h>
103 #include <sys/acct.h>
104 #include <sys/filedesc.h>
105 #include <sys/signalvar.h>
106 #include <sys/sched.h>
107 #ifdef SYSVSHM
108 #include <sys/shm.h>
109 #endif
110 #ifdef SYSVSEM
111 #include <sys/sem.h>
112 #endif
113 
114 #include <sys/mount.h>
115 #include <sys/syscallargs.h>
116 
117 #include <machine/cpu.h>
118 
119 #include <vm/vm.h>
120 #include <vm/vm_kern.h>
121 
122 #include <uvm/uvm_extern.h>
123 
124 /*
125  * exit --
126  *	Death of process.
127  */
128 int
129 sys_exit(p, v, retval)
130 	struct proc *p;
131 	void *v;
132 	register_t *retval;
133 {
134 	struct sys_exit_args /* {
135 		syscallarg(int) rval;
136 	} */ *uap = v;
137 
138 	exit1(p, W_EXITCODE(SCARG(uap, rval), 0));
139 	/* NOTREACHED */
140 	return (0);
141 }
142 
143 /*
144  * Exit: deallocate address space and other resources, change proc state
145  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
146  * status and rusage for wait().  Check for child processes and orphan them.
147  */
148 void
149 exit1(p, rv)
150 	struct proc *p;
151 	int rv;
152 {
153 	struct proc *q, *nq;
154 	struct vmspace *vm;
155 	int s;
156 
157 	if (__predict_false(p == initproc))
158 		panic("init died (signal %d, exit %d)",
159 		    WTERMSIG(rv), WEXITSTATUS(rv));
160 
161 #ifdef PGINPROF
162 	vmsizmon();
163 #endif
164 	if (p->p_flag & P_PROFIL)
165 		stopprofclock(p);
166 	p->p_ru = pool_get(&rusage_pool, PR_WAITOK);
167 	/*
168 	 * If parent is waiting for us to exit or exec, P_PPWAIT is set; we
169 	 * wake up the parent early to avoid deadlock.
170 	 */
171 	p->p_flag |= P_WEXIT;
172 	if (p->p_flag & P_PPWAIT) {
173 		p->p_flag &= ~P_PPWAIT;
174 		wakeup((caddr_t)p->p_pptr);
175 	}
176 	sigfillset(&p->p_sigignore);
177 	sigemptyset(&p->p_siglist);
178 	p->p_sigcheck = 0;
179 	callout_stop(&p->p_realit_ch);
180 
181 	/*
182 	 * Close open files and release open-file table.
183 	 * This may block!
184 	 */
185 	fdfree(p);
186 	cwdfree(p);
187 
188 	/* The next three chunks should probably be moved to vmspace_exit. */
189 	vm = p->p_vmspace;
190 #ifdef SYSVSHM
191 	if (vm->vm_shm && vm->vm_refcnt == 1)
192 		shmexit(vm);
193 #endif
194 #ifdef SYSVSEM
195 	semexit(p);
196 #endif
197 	/*
198 	 * Release user portion of address space.
199 	 * This releases references to vnodes,
200 	 * which could cause I/O if the file has been unlinked.
201 	 * Need to do this early enough that we can still sleep.
202 	 * Can't free the entire vmspace as the kernel stack
203 	 * may be mapped within that space also.
204 	 */
205 	if (vm->vm_refcnt == 1)
206 		(void) uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
207 		    VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
208 
209 	if (SESS_LEADER(p)) {
210 		struct session *sp = p->p_session;
211 
212 		if (sp->s_ttyvp) {
213 			/*
214 			 * Controlling process.
215 			 * Signal foreground pgrp,
216 			 * drain controlling terminal
217 			 * and revoke access to controlling terminal.
218 			 */
219 			if (sp->s_ttyp->t_session == sp) {
220 				if (sp->s_ttyp->t_pgrp)
221 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
222 				(void) ttywait(sp->s_ttyp);
223 				/*
224 				 * The tty could have been revoked
225 				 * if we blocked.
226 				 */
227 				if (sp->s_ttyvp)
228 					VOP_REVOKE(sp->s_ttyvp, REVOKEALL);
229 			}
230 			if (sp->s_ttyvp)
231 				vrele(sp->s_ttyvp);
232 			sp->s_ttyvp = NULL;
233 			/*
234 			 * s_ttyp is not zero'd; we use this to indicate
235 			 * that the session once had a controlling terminal.
236 			 * (for logging and informational purposes)
237 			 */
238 		}
239 		sp->s_leader = NULL;
240 	}
241 	fixjobc(p, p->p_pgrp, 0);
242 	(void)acct_process(p);
243 #ifdef KTRACE
244 	/*
245 	 * release trace file
246 	 */
247 	ktrderef(p);
248 #endif
249 	/*
250 	 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP!
251 	 */
252 	p->p_stat = SDEAD;
253 
254 	/*
255 	 * Remove proc from pidhash chain so looking it up won't
256 	 * work.  Move it from allproc to zombproc, but do not yet
257 	 * wake up the reaper.  We will put the proc on the
258 	 * deadproc list later (using the p_hash member), and
259 	 * wake up the reaper when we do.
260 	 */
261 	s = proclist_lock_write();
262 	LIST_REMOVE(p, p_hash);
263 	LIST_REMOVE(p, p_list);
264 	LIST_INSERT_HEAD(&zombproc, p, p_list);
265 	proclist_unlock_write(s);
266 
267 	/*
268 	 * Give orphaned children to init(8).
269 	 */
270 	q = p->p_children.lh_first;
271 	if (q)		/* only need this if any child is S_ZOMB */
272 		wakeup((caddr_t)initproc);
273 	for (; q != 0; q = nq) {
274 		nq = q->p_sibling.le_next;
275 		proc_reparent(q, initproc);
276 		/*
277 		 * Traced processes are killed
278 		 * since their existence means someone is screwing up.
279 		 */
280 		if (q->p_flag & P_TRACED) {
281 			q->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
282 			psignal(q, SIGKILL);
283 		}
284 	}
285 
286 	/*
287 	 * Save exit status and final rusage info, adding in child rusage
288 	 * info and self times.
289 	 */
290 	p->p_xstat = rv;
291 	*p->p_ru = p->p_stats->p_ru;
292 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
293 	ruadd(p->p_ru, &p->p_stats->p_cru);
294 
295 	/*
296 	 * Notify parent that we're gone.  If parent has the P_NOCLDWAIT
297 	 * flag set, notify init instead (and hope it will handle
298 	 * this situation).
299 	 */
300 	if (p->p_pptr->p_flag & P_NOCLDWAIT) {
301 		struct proc *pp = p->p_pptr;
302 		proc_reparent(p, initproc);
303 		/*
304 		 * If this was the last child of our parent, notify
305 		 * parent, so in case he was wait(2)ing, he will
306 		 * continue.
307 		 */
308 		if (pp->p_children.lh_first == NULL)
309 			wakeup((caddr_t)pp);
310 	}
311 
312 	/*
313 	 * Release the process's signal state.
314 	 */
315 	sigactsfree(p);
316 
317 	/*
318 	 * Clear curproc after we've done all operations
319 	 * that could block, and before tearing down the rest
320 	 * of the process state that might be used from clock, etc.
321 	 * Also, can't clear curproc while we're still runnable,
322 	 * as we're not on a run queue (we are current, just not
323 	 * a proper proc any longer!).
324 	 *
325 	 * Other substructures are freed from wait().
326 	 */
327 	curproc = NULL;
328 	limfree(p->p_limit);
329 	p->p_limit = NULL;
330 
331 	/*
332 	 * Finally, call machine-dependent code to switch to a new
333 	 * context (possibly the idle context).  Once we are no longer
334 	 * using the dead process's vmspace and stack, exit2() will be
335 	 * called to schedule those resources to be released by the
336 	 * reaper thread.
337 	 *
338 	 * Note that cpu_exit() will end with a call equivalent to
339 	 * cpu_switch(), finishing our execution (pun intended).
340 	 */
341 	cpu_exit(p);
342 }
343 
344 /*
345  * We are called from cpu_exit() once it is safe to schedule the
346  * dead process's resources to be freed.
347  *
348  * NOTE: One must be careful with locking in this routine.  It's
349  * called from a critical section in machine-dependent code, so
350  * we should refrain from changing any interrupt state.
351  *
352  * We lock the deadproc list (a spin lock), place the proc on that
353  * list (using the p_hash member), and wake up the reaper.
354  */
355 void
356 exit2(p)
357 	struct proc *p;
358 {
359 
360 	simple_lock(&deadproc_slock);
361 	LIST_INSERT_HEAD(&deadproc, p, p_hash);
362 	simple_unlock(&deadproc_slock);
363 
364 	wakeup(&deadproc);
365 }
366 
367 /*
368  * Process reaper.  This is run by a kernel thread to free the resources
369  * of a dead process.  Once the resources are free, the process becomes
370  * a zombie, and the parent is allowed to read the undead's status.
371  */
372 void
373 reaper()
374 {
375 	struct proc *p;
376 
377 	for (;;) {
378 		simple_lock(&deadproc_slock);
379 		p = LIST_FIRST(&deadproc);
380 		if (p == NULL) {
381 			/* No work for us; go to sleep until someone exits. */
382 			simple_unlock(&deadproc_slock);
383 			(void) tsleep(&deadproc, PVM, "reaper", 0);
384 			continue;
385 		}
386 
387 		/* Remove us from the deadproc list. */
388 		LIST_REMOVE(p, p_hash);
389 		simple_unlock(&deadproc_slock);
390 
391 		/*
392 		 * Give machine-dependent code a chance to free any
393 		 * resources it couldn't free while still running on
394 		 * that process's context.  This must be done before
395 		 * uvm_exit(), in case these resources are in the PCB.
396 		 */
397 		cpu_wait(p);
398 
399 		/*
400 		 * Free the VM resources we're still holding on to.
401 		 * We must do this from a valid thread because doing
402 		 * so may block.
403 		 */
404 		uvm_exit(p);
405 
406 		/* Process is now a true zombie. */
407 		p->p_stat = SZOMB;
408 
409 		/* Wake up the parent so it can get exit status. */
410 		if ((p->p_flag & P_FSTRACE) == 0 && p->p_exitsig != 0)
411 			psignal(p->p_pptr, P_EXITSIG(p));
412 		wakeup((caddr_t)p->p_pptr);
413 	}
414 }
415 
416 int
417 sys_wait4(q, v, retval)
418 	struct proc *q;
419 	void *v;
420 	register_t *retval;
421 {
422 	struct sys_wait4_args /* {
423 		syscallarg(int) pid;
424 		syscallarg(int *) status;
425 		syscallarg(int) options;
426 		syscallarg(struct rusage *) rusage;
427 	} */ *uap = v;
428 	int nfound;
429 	struct proc *p, *t;
430 	int status, error, s;
431 
432 	if (SCARG(uap, pid) == 0)
433 		SCARG(uap, pid) = -q->p_pgid;
434 	if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG|WALTSIG))
435 		return (EINVAL);
436 
437 loop:
438 	nfound = 0;
439 	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
440 		if (SCARG(uap, pid) != WAIT_ANY &&
441 		    p->p_pid != SCARG(uap, pid) &&
442 		    p->p_pgid != -SCARG(uap, pid))
443 			continue;
444 		/*
445 		 * Wait for processes with p_exitsig != SIGCHLD processes only
446 		 * if WALTSIG is set; wait for processes with p_exitsig ==
447 		 * SIGCHLD only if WALTSIG is clear.
448 		 */
449 		if ((SCARG(uap, options) & WALTSIG) ?
450 		    (p->p_exitsig == SIGCHLD) : (P_EXITSIG(p) != SIGCHLD))
451 			continue;
452 
453 		nfound++;
454 		if (p->p_stat == SZOMB) {
455 			retval[0] = p->p_pid;
456 
457 			if (SCARG(uap, status)) {
458 				status = p->p_xstat;	/* convert to int */
459 				error = copyout((caddr_t)&status,
460 						(caddr_t)SCARG(uap, status),
461 						sizeof(status));
462 				if (error)
463 					return (error);
464 			}
465 			if (SCARG(uap, rusage) &&
466 			    (error = copyout((caddr_t)p->p_ru,
467 			    (caddr_t)SCARG(uap, rusage),
468 			    sizeof(struct rusage))))
469 				return (error);
470 			/*
471 			 * If we got the child via ptrace(2) or procfs, and
472 			 * the parent is different (meaning the process was
473 			 * attached, rather than run as a child), then we need
474 			 * to give it back to the old parent, and send the
475 			 * parent the exit signal.  The rest of the cleanup
476 			 * will be done when the old parent waits on the child.
477 			 */
478 			if ((p->p_flag & P_TRACED) &&
479 			    p->p_oppid != p->p_pptr->p_pid) {
480 				t = pfind(p->p_oppid);
481 				proc_reparent(p, t ? t : initproc);
482 				p->p_oppid = 0;
483 				p->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
484 				if (p->p_exitsig != 0)
485 					psignal(p->p_pptr, P_EXITSIG(p));
486 				wakeup((caddr_t)p->p_pptr);
487 				return (0);
488 			}
489 			scheduler_wait_hook(q, p);
490 			p->p_xstat = 0;
491 			ruadd(&q->p_stats->p_cru, p->p_ru);
492 			pool_put(&rusage_pool, p->p_ru);
493 
494 			/*
495 			 * Finally finished with old proc entry.
496 			 * Unlink it from its process group and free it.
497 			 */
498 			leavepgrp(p);
499 
500 			s = proclist_lock_write();
501 			LIST_REMOVE(p, p_list);	/* off zombproc */
502 			proclist_unlock_write(s);
503 
504 			LIST_REMOVE(p, p_sibling);
505 
506 			/*
507 			 * Decrement the count of procs running with this uid.
508 			 */
509 			(void)chgproccnt(p->p_cred->p_ruid, -1);
510 
511 			/*
512 			 * Free up credentials.
513 			 */
514 			if (--p->p_cred->p_refcnt == 0) {
515 				crfree(p->p_cred->pc_ucred);
516 				pool_put(&pcred_pool, p->p_cred);
517 			}
518 
519 			/*
520 			 * Release reference to text vnode
521 			 */
522 			if (p->p_textvp)
523 				vrele(p->p_textvp);
524 
525 			pool_put(&proc_pool, p);
526 			nprocs--;
527 			return (0);
528 		}
529 		if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
530 		    (p->p_flag & P_TRACED || SCARG(uap, options) & WUNTRACED)) {
531 			p->p_flag |= P_WAITED;
532 			retval[0] = p->p_pid;
533 
534 			if (SCARG(uap, status)) {
535 				status = W_STOPCODE(p->p_xstat);
536 				error = copyout((caddr_t)&status,
537 				    (caddr_t)SCARG(uap, status),
538 				    sizeof(status));
539 			} else
540 				error = 0;
541 			return (error);
542 		}
543 	}
544 	if (nfound == 0)
545 		return (ECHILD);
546 	if (SCARG(uap, options) & WNOHANG) {
547 		retval[0] = 0;
548 		return (0);
549 	}
550 	if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)) != 0)
551 		return (error);
552 	goto loop;
553 }
554 
555 /*
556  * make process 'parent' the new parent of process 'child'.
557  */
558 void
559 proc_reparent(child, parent)
560 	struct proc *child;
561 	struct proc *parent;
562 {
563 
564 	if (child->p_pptr == parent)
565 		return;
566 
567 	if (parent == initproc)
568 		child->p_exitsig = SIGCHLD;
569 
570 	LIST_REMOVE(child, p_sibling);
571 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
572 	child->p_pptr = parent;
573 }
574