xref: /netbsd-src/sys/kern/kern_exit.c (revision 3c618b8f0665ca5acf2910b7200a6f7d6be3506e)
1 /*	$NetBSD: kern_exit.c,v 1.65 1999/03/24 05:51:22 mrg Exp $	*/
2 
3 /*-
4  * Copyright (c) 1998 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 	register struct proc *p;
151 	int rv;
152 {
153 	register struct proc *q, *nq;
154 	register struct vmspace *vm;
155 
156 	if (p->p_pid == 1)
157 		panic("init died (signal %d, exit %d)",
158 		    WTERMSIG(rv), WEXITSTATUS(rv));
159 #ifdef PGINPROF
160 	vmsizmon();
161 #endif
162 	if (p->p_flag & P_PROFIL)
163 		stopprofclock(p);
164 	p->p_ru = pool_get(&rusage_pool, PR_WAITOK);
165 	/*
166 	 * If parent is waiting for us to exit or exec, P_PPWAIT is set; we
167 	 * wake up the parent early to avoid deadlock.
168 	 */
169 	p->p_flag |= P_WEXIT;
170 	if (p->p_flag & P_PPWAIT) {
171 		p->p_flag &= ~P_PPWAIT;
172 		wakeup((caddr_t)p->p_pptr);
173 	}
174 	sigfillset(&p->p_sigignore);
175 	sigemptyset(&p->p_siglist);
176 	p->p_sigcheck = 0;
177 	untimeout(realitexpire, (caddr_t)p);
178 
179 	/*
180 	 * Close open files and release open-file table.
181 	 * This may block!
182 	 */
183 	fdfree(p);
184 
185 	/* The next three chunks should probably be moved to vmspace_exit. */
186 	vm = p->p_vmspace;
187 #ifdef SYSVSHM
188 	if (vm->vm_shm && vm->vm_refcnt == 1)
189 		shmexit(vm);
190 #endif
191 #ifdef SYSVSEM
192 	semexit(p);
193 #endif
194 	/*
195 	 * Release user portion of address space.
196 	 * This releases references to vnodes,
197 	 * which could cause I/O if the file has been unlinked.
198 	 * Need to do this early enough that we can still sleep.
199 	 * Can't free the entire vmspace as the kernel stack
200 	 * may be mapped within that space also.
201 	 */
202 	if (vm->vm_refcnt == 1)
203 		(void) uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
204 		    VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
205 
206 	if (SESS_LEADER(p)) {
207 		register struct session *sp = p->p_session;
208 
209 		if (sp->s_ttyvp) {
210 			/*
211 			 * Controlling process.
212 			 * Signal foreground pgrp,
213 			 * drain controlling terminal
214 			 * and revoke access to controlling terminal.
215 			 */
216 			if (sp->s_ttyp->t_session == sp) {
217 				if (sp->s_ttyp->t_pgrp)
218 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
219 				(void) ttywait(sp->s_ttyp);
220 				/*
221 				 * The tty could have been revoked
222 				 * if we blocked.
223 				 */
224 				if (sp->s_ttyvp)
225 					VOP_REVOKE(sp->s_ttyvp, REVOKEALL);
226 			}
227 			if (sp->s_ttyvp)
228 				vrele(sp->s_ttyvp);
229 			sp->s_ttyvp = NULL;
230 			/*
231 			 * s_ttyp is not zero'd; we use this to indicate
232 			 * that the session once had a controlling terminal.
233 			 * (for logging and informational purposes)
234 			 */
235 		}
236 		sp->s_leader = NULL;
237 	}
238 	fixjobc(p, p->p_pgrp, 0);
239 	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
240 	(void)acct_process(p);
241 #ifdef KTRACE
242 	/*
243 	 * release trace file
244 	 */
245 	ktrderef(p);
246 #endif
247 	/*
248 	 * Remove proc from allproc queue and pidhash chain.
249 	 * Unlink from parent's child list.
250 	 */
251 	LIST_REMOVE(p, p_list);
252 	p->p_stat = SZOMB;
253 
254 	/*
255 	 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP!
256 	 */
257 
258 	LIST_REMOVE(p, p_hash);
259 
260 	q = p->p_children.lh_first;
261 	if (q)		/* only need this if any child is S_ZOMB */
262 		wakeup((caddr_t)initproc);
263 	for (; q != 0; q = nq) {
264 		nq = q->p_sibling.le_next;
265 		proc_reparent(q, initproc);
266 		/*
267 		 * Traced processes are killed
268 		 * since their existence means someone is screwing up.
269 		 */
270 		if (q->p_flag & P_TRACED) {
271 			q->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
272 			psignal(q, SIGKILL);
273 		}
274 	}
275 
276 	/*
277 	 * Save exit status and final rusage info, adding in child rusage
278 	 * info and self times.
279 	 */
280 	p->p_xstat = rv;
281 	*p->p_ru = p->p_stats->p_ru;
282 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
283 	ruadd(p->p_ru, &p->p_stats->p_cru);
284 
285 	/*
286 	 * Notify parent that we're gone.  If parent has the P_NOCLDWAIT
287 	 * flag set, notify init instead (and hope it will handle
288 	 * this situation).
289 	 */
290 	if (p->p_pptr->p_flag & P_NOCLDWAIT) {
291 		struct proc *pp = p->p_pptr;
292 		proc_reparent(p, initproc);
293 		/*
294 		 * If this was the last child of our parent, notify
295 		 * parent, so in case he was wait(2)ing, he will
296 		 * continue.
297 		 */
298 		if (pp->p_children.lh_first == NULL)
299 			wakeup((caddr_t)pp);
300 	}
301 
302 	/*
303 	 * Clear curproc after we've done all operations
304 	 * that could block, and before tearing down the rest
305 	 * of the process state that might be used from clock, etc.
306 	 * Also, can't clear curproc while we're still runnable,
307 	 * as we're not on a run queue (we are current, just not
308 	 * a proper proc any longer!).
309 	 *
310 	 * Other substructures are freed from wait().
311 	 */
312 	curproc = NULL;
313 	if (--p->p_limit->p_refcnt == 0)
314 		pool_put(&plimit_pool, p->p_limit);
315 
316 	/*
317 	 * Finally, call machine-dependent code to switch to a new
318 	 * context (possibly the idle context).  Once we are no longer
319 	 * using the dead process's vmspace and stack, exit2() will be
320 	 * called to schedule those resources to be released by the
321 	 * reaper thread.
322 	 *
323 	 * Note that cpu_exit() will end with a call equivalent to
324 	 * cpu_switch(), finishing our execution (pun intended).
325 	 */
326 	cpu_exit(p);
327 }
328 
329 /*
330  * We are called from cpu_exit() once it is safe to schedule the
331  * dead process's resources to be freed.
332  */
333 void
334 exit2(p)
335 	struct proc *p;
336 {
337 
338 	LIST_INSERT_HEAD(&deadproc, p, p_list);
339 	wakeup(&deadproc);
340 }
341 
342 /*
343  * Process reaper.  This is run by a kernel thread to free the resources
344  * of a dead process.  Once the resources are free, the process becomes
345  * a zombie, and the parent is allowed to read the undead's status.
346  */
347 void
348 reaper()
349 {
350 	struct proc *p;
351 
352 	for (;;) {
353 		p = LIST_FIRST(&deadproc);
354 		if (p == NULL) {
355 			/* No work for us; go to sleep until someone exits. */
356 			(void) tsleep(&deadproc, PVM, "reaper", 0);
357 			continue;
358 		}
359 
360 		/* Remove us from the deadproc list. */
361 		LIST_REMOVE(p, p_list);
362 
363 		/*
364 		 * Free the VM resources we're still holding on to.
365 		 * We must do this from a valid thread because doing
366 		 * so may block.
367 		 */
368 		uvm_exit(p);
369 
370 		/* Process is now a true zombie. */
371 		LIST_INSERT_HEAD(&zombproc, p, p_list);
372 
373 		/* Wake up the parent so it can get exit satus. */
374 		if ((p->p_flag & P_FSTRACE) == 0)
375 			psignal(p->p_pptr, SIGCHLD);
376 		wakeup((caddr_t)p->p_pptr);
377 	}
378 }
379 
380 int
381 sys_wait4(q, v, retval)
382 	register struct proc *q;
383 	void *v;
384 	register_t *retval;
385 {
386 	register struct sys_wait4_args /* {
387 		syscallarg(int) pid;
388 		syscallarg(int *) status;
389 		syscallarg(int) options;
390 		syscallarg(struct rusage *) rusage;
391 	} */ *uap = v;
392 	register int nfound;
393 	register struct proc *p, *t;
394 	int status, error;
395 
396 	if (SCARG(uap, pid) == 0)
397 		SCARG(uap, pid) = -q->p_pgid;
398 	if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG))
399 		return (EINVAL);
400 
401 loop:
402 	nfound = 0;
403 	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
404 		if (SCARG(uap, pid) != WAIT_ANY &&
405 		    p->p_pid != SCARG(uap, pid) &&
406 		    p->p_pgid != -SCARG(uap, pid))
407 			continue;
408 		nfound++;
409 		if (p->p_stat == SZOMB) {
410 			retval[0] = p->p_pid;
411 
412 			if (SCARG(uap, status)) {
413 				status = p->p_xstat;	/* convert to int */
414 				error = copyout((caddr_t)&status,
415 						(caddr_t)SCARG(uap, status),
416 						sizeof(status));
417 				if (error)
418 					return (error);
419 			}
420 			if (SCARG(uap, rusage) &&
421 			    (error = copyout((caddr_t)p->p_ru,
422 			    (caddr_t)SCARG(uap, rusage),
423 			    sizeof(struct rusage))))
424 				return (error);
425 			/*
426 			 * If we got the child via ptrace(2) or procfs, and
427 			 * the parent is different (meaning the process was
428 			 * attached, rather than run as a child), then we need
429 			 * to give it back to the old parent, and send the
430 			 * parent a SIGCHLD.  The rest of the cleanup will be
431 			 * done when the old parent waits on the child.
432 			 */
433 			if ((p->p_flag & P_TRACED) &&
434 			    p->p_oppid != p->p_pptr->p_pid) {
435 				t = pfind(p->p_oppid);
436 				proc_reparent(p, t ? t : initproc);
437 				p->p_oppid = 0;
438 				p->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
439 				psignal(p->p_pptr, SIGCHLD);
440 				wakeup((caddr_t)p->p_pptr);
441 				return (0);
442 			}
443 			scheduler_wait_hook(curproc, p);
444 			p->p_xstat = 0;
445 			ruadd(&q->p_stats->p_cru, p->p_ru);
446 			pool_put(&rusage_pool, p->p_ru);
447 
448 			/*
449 			 * Finally finished with old proc entry.
450 			 * Unlink it from its process group and free it.
451 			 */
452 			leavepgrp(p);
453 
454 			LIST_REMOVE(p, p_list);	/* off zombproc */
455 
456 			LIST_REMOVE(p, p_sibling);
457 
458 			/*
459 			 * Decrement the count of procs running with this uid.
460 			 */
461 			(void)chgproccnt(p->p_cred->p_ruid, -1);
462 
463 			/*
464 			 * Free up credentials.
465 			 */
466 			if (--p->p_cred->p_refcnt == 0) {
467 				crfree(p->p_cred->pc_ucred);
468 				pool_put(&pcred_pool, p->p_cred);
469 			}
470 
471 			/*
472 			 * Release reference to text vnode
473 			 */
474 			if (p->p_textvp)
475 				vrele(p->p_textvp);
476 
477 			/*
478 			 * Give machine-dependent layer a chance
479 			 * to free anything that cpu_exit couldn't
480 			 * release while still running in process context.
481 			 */
482 			cpu_wait(p);
483 			pool_put(&proc_pool, p);
484 			nprocs--;
485 			return (0);
486 		}
487 		if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
488 		    (p->p_flag & P_TRACED || SCARG(uap, options) & WUNTRACED)) {
489 			p->p_flag |= P_WAITED;
490 			retval[0] = p->p_pid;
491 
492 			if (SCARG(uap, status)) {
493 				status = W_STOPCODE(p->p_xstat);
494 				error = copyout((caddr_t)&status,
495 				    (caddr_t)SCARG(uap, status),
496 				    sizeof(status));
497 			} else
498 				error = 0;
499 			return (error);
500 		}
501 	}
502 	if (nfound == 0)
503 		return (ECHILD);
504 	if (SCARG(uap, options) & WNOHANG) {
505 		retval[0] = 0;
506 		return (0);
507 	}
508 	if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)) != 0)
509 		return (error);
510 	goto loop;
511 }
512 
513 /*
514  * make process 'parent' the new parent of process 'child'.
515  */
516 void
517 proc_reparent(child, parent)
518 	register struct proc *child;
519 	register struct proc *parent;
520 {
521 
522 	if (child->p_pptr == parent)
523 		return;
524 
525 	LIST_REMOVE(child, p_sibling);
526 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
527 	child->p_pptr = parent;
528 }
529