xref: /netbsd-src/sys/kern/kern_exit.c (revision dc306354b0b29af51801a7632f1e95265a68cd81)
1 /*	$NetBSD: kern_exit.c,v 1.61 1998/11/11 22:37:13 thorpej 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_uvm.h"
82 #include "opt_sysv.h"
83 
84 #include <sys/param.h>
85 #include <sys/systm.h>
86 #include <sys/map.h>
87 #include <sys/ioctl.h>
88 #include <sys/proc.h>
89 #include <sys/tty.h>
90 #include <sys/time.h>
91 #include <sys/resource.h>
92 #include <sys/kernel.h>
93 #include <sys/ktrace.h>
94 #include <sys/proc.h>
95 #include <sys/buf.h>
96 #include <sys/wait.h>
97 #include <sys/file.h>
98 #include <sys/vnode.h>
99 #include <sys/syslog.h>
100 #include <sys/malloc.h>
101 #include <sys/pool.h>
102 #include <sys/resourcevar.h>
103 #include <sys/ptrace.h>
104 #include <sys/acct.h>
105 #include <sys/filedesc.h>
106 #include <sys/signalvar.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 #if defined(UVM)
123 #include <uvm/uvm_extern.h>
124 #endif
125 
126 /*
127  * exit --
128  *	Death of process.
129  */
130 int
131 sys_exit(p, v, retval)
132 	struct proc *p;
133 	void *v;
134 	register_t *retval;
135 {
136 	struct sys_exit_args /* {
137 		syscallarg(int) rval;
138 	} */ *uap = v;
139 
140 	exit1(p, W_EXITCODE(SCARG(uap, rval), 0));
141 	/* NOTREACHED */
142 	return (0);
143 }
144 
145 /*
146  * Exit: deallocate address space and other resources, change proc state
147  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
148  * status and rusage for wait().  Check for child processes and orphan them.
149  */
150 void
151 exit1(p, rv)
152 	register struct proc *p;
153 	int rv;
154 {
155 	register struct proc *q, *nq;
156 	register struct vmspace *vm;
157 
158 	if (p->p_pid == 1)
159 		panic("init died (signal %d, exit %d)",
160 		    WTERMSIG(rv), WEXITSTATUS(rv));
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 	untimeout(realitexpire, (caddr_t)p);
180 
181 	/*
182 	 * Close open files and release open-file table.
183 	 * This may block!
184 	 */
185 	fdfree(p);
186 
187 	/* The next three chunks should probably be moved to vmspace_exit. */
188 	vm = p->p_vmspace;
189 #ifdef SYSVSHM
190 	if (vm->vm_shm && vm->vm_refcnt == 1)
191 		shmexit(vm);
192 #endif
193 #ifdef SYSVSEM
194 	semexit(p);
195 #endif
196 	/*
197 	 * Release user portion of address space.
198 	 * This releases references to vnodes,
199 	 * which could cause I/O if the file has been unlinked.
200 	 * Need to do this early enough that we can still sleep.
201 	 * Can't free the entire vmspace as the kernel stack
202 	 * may be mapped within that space also.
203 	 */
204 #if defined(UVM)
205 	if (vm->vm_refcnt == 1)
206 		(void) uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
207 		    VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
208 #else
209 	if (vm->vm_refcnt == 1)
210 		(void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS,
211 		    VM_MAXUSER_ADDRESS);
212 #endif
213 
214 	if (SESS_LEADER(p)) {
215 		register struct session *sp = p->p_session;
216 
217 		if (sp->s_ttyvp) {
218 			/*
219 			 * Controlling process.
220 			 * Signal foreground pgrp,
221 			 * drain controlling terminal
222 			 * and revoke access to controlling terminal.
223 			 */
224 			if (sp->s_ttyp->t_session == sp) {
225 				if (sp->s_ttyp->t_pgrp)
226 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
227 				(void) ttywait(sp->s_ttyp);
228 				/*
229 				 * The tty could have been revoked
230 				 * if we blocked.
231 				 */
232 				if (sp->s_ttyvp)
233 					VOP_REVOKE(sp->s_ttyvp, REVOKEALL);
234 			}
235 			if (sp->s_ttyvp)
236 				vrele(sp->s_ttyvp);
237 			sp->s_ttyvp = NULL;
238 			/*
239 			 * s_ttyp is not zero'd; we use this to indicate
240 			 * that the session once had a controlling terminal.
241 			 * (for logging and informational purposes)
242 			 */
243 		}
244 		sp->s_leader = NULL;
245 	}
246 	fixjobc(p, p->p_pgrp, 0);
247 	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
248 	(void)acct_process(p);
249 #ifdef KTRACE
250 	/*
251 	 * release trace file
252 	 */
253 	ktrderef(p);
254 #endif
255 	/*
256 	 * Remove proc from allproc queue and pidhash chain.
257 	 * Unlink from parent's child list.
258 	 */
259 	LIST_REMOVE(p, p_list);
260 	p->p_stat = SZOMB;
261 
262 	/*
263 	 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP!
264 	 */
265 
266 	LIST_REMOVE(p, p_hash);
267 
268 	q = p->p_children.lh_first;
269 	if (q)		/* only need this if any child is S_ZOMB */
270 		wakeup((caddr_t)initproc);
271 	for (; q != 0; q = nq) {
272 		nq = q->p_sibling.le_next;
273 		proc_reparent(q, initproc);
274 		/*
275 		 * Traced processes are killed
276 		 * since their existence means someone is screwing up.
277 		 */
278 		if (q->p_flag & P_TRACED) {
279 			q->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
280 			psignal(q, SIGKILL);
281 		}
282 	}
283 
284 	/*
285 	 * Save exit status and final rusage info, adding in child rusage
286 	 * info and self times.
287 	 */
288 	p->p_xstat = rv;
289 	*p->p_ru = p->p_stats->p_ru;
290 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
291 	ruadd(p->p_ru, &p->p_stats->p_cru);
292 
293 	/*
294 	 * Notify parent that we're gone.  If parent has the P_NOCLDWAIT
295 	 * flag set, notify init instead (and hope it will handle
296 	 * this situation).
297 	 */
298 	if (p->p_pptr->p_flag & P_NOCLDWAIT) {
299 		struct proc *pp = p->p_pptr;
300 		proc_reparent(p, initproc);
301 		/*
302 		 * If this was the last child of our parent, notify
303 		 * parent, so in case he was wait(2)ing, he will
304 		 * continue.
305 		 */
306 		if (pp->p_children.lh_first == NULL)
307 			wakeup((caddr_t)pp);
308 	}
309 
310 	/*
311 	 * Clear curproc after we've done all operations
312 	 * that could block, and before tearing down the rest
313 	 * of the process state that might be used from clock, etc.
314 	 * Also, can't clear curproc while we're still runnable,
315 	 * as we're not on a run queue (we are current, just not
316 	 * a proper proc any longer!).
317 	 *
318 	 * Other substructures are freed from wait().
319 	 */
320 	curproc = NULL;
321 	if (--p->p_limit->p_refcnt == 0)
322 		pool_put(&plimit_pool, p->p_limit);
323 
324 	/*
325 	 * Finally, call machine-dependent code to switch to a new
326 	 * context (possibly the idle context).  Once we are no longer
327 	 * using the dead process's vmspace and stack, exit2() will be
328 	 * called to schedule those resources to be released by the
329 	 * reaper thread.
330 	 *
331 	 * Note that cpu_exit() will end with a call equivalent to
332 	 * cpu_switch(), finishing our execution (pun intended).
333 	 */
334 	cpu_exit(p);
335 }
336 
337 /*
338  * We are called from cpu_exit() once it is safe to schedule the
339  * dead process's resources to be freed.
340  */
341 void
342 exit2(p)
343 	struct proc *p;
344 {
345 
346 	LIST_INSERT_HEAD(&deadproc, p, p_list);
347 	wakeup(&deadproc);
348 }
349 
350 /*
351  * Process reaper.  This is run by a kernel thread to free the resources
352  * of a dead process.  Once the resources are free, the process becomes
353  * a zombie, and the parent is allowed to read the undead's status.
354  */
355 void
356 reaper()
357 {
358 	struct proc *p;
359 
360 	for (;;) {
361 		p = LIST_FIRST(&deadproc);
362 		if (p == NULL) {
363 			/* No work for us; go to sleep until someone exits. */
364 			(void) tsleep(&deadproc, PVM, "reaper", 0);
365 			continue;
366 		}
367 
368 		/* Remove us from the deadproc list. */
369 		LIST_REMOVE(p, p_list);
370 
371 		/*
372 		 * Free the VM resources we're still holding on to.
373 		 * We must do this from a valid thread because doing
374 		 * so may block.
375 		 */
376 #if defined(UVM)
377 		uvm_exit(p);
378 #else
379 		vm_exit(p);
380 #endif
381 
382 		/* Process is now a true zombie. */
383 		LIST_INSERT_HEAD(&zombproc, p, p_list);
384 
385 		/* Wake up the parent so it can get exit satus. */
386 		if ((p->p_flag & P_FSTRACE) == 0)
387 			psignal(p->p_pptr, SIGCHLD);
388 		wakeup((caddr_t)p->p_pptr);
389 	}
390 }
391 
392 int
393 sys_wait4(q, v, retval)
394 	register struct proc *q;
395 	void *v;
396 	register_t *retval;
397 {
398 	register struct sys_wait4_args /* {
399 		syscallarg(int) pid;
400 		syscallarg(int *) status;
401 		syscallarg(int) options;
402 		syscallarg(struct rusage *) rusage;
403 	} */ *uap = v;
404 	register int nfound;
405 	register struct proc *p, *t;
406 	int status, error;
407 
408 	if (SCARG(uap, pid) == 0)
409 		SCARG(uap, pid) = -q->p_pgid;
410 	if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG))
411 		return (EINVAL);
412 
413 loop:
414 	nfound = 0;
415 	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
416 		if (SCARG(uap, pid) != WAIT_ANY &&
417 		    p->p_pid != SCARG(uap, pid) &&
418 		    p->p_pgid != -SCARG(uap, pid))
419 			continue;
420 		nfound++;
421 		if (p->p_stat == SZOMB) {
422 			retval[0] = p->p_pid;
423 
424 			if (SCARG(uap, status)) {
425 				status = p->p_xstat;	/* convert to int */
426 				error = copyout((caddr_t)&status,
427 						(caddr_t)SCARG(uap, status),
428 						sizeof(status));
429 				if (error)
430 					return (error);
431 			}
432 			if (SCARG(uap, rusage) &&
433 			    (error = copyout((caddr_t)p->p_ru,
434 			    (caddr_t)SCARG(uap, rusage),
435 			    sizeof(struct rusage))))
436 				return (error);
437 			/*
438 			 * If we got the child via ptrace(2) or procfs, and
439 			 * the parent is different (meaning the process was
440 			 * attached, rather than run as a child), then we need
441 			 * to give it back to the old parent, and send the
442 			 * parent a SIGCHLD.  The rest of the cleanup will be
443 			 * done when the old parent waits on the child.
444 			 */
445 			if ((p->p_flag & P_TRACED) &&
446 			    p->p_oppid != p->p_pptr->p_pid) {
447 				t = pfind(p->p_oppid);
448 				proc_reparent(p, t ? t : initproc);
449 				p->p_oppid = 0;
450 				p->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
451 				psignal(p->p_pptr, SIGCHLD);
452 				wakeup((caddr_t)p->p_pptr);
453 				return (0);
454 			}
455 			p->p_xstat = 0;
456 			ruadd(&q->p_stats->p_cru, p->p_ru);
457 			pool_put(&rusage_pool, p->p_ru);
458 
459 			/*
460 			 * Finally finished with old proc entry.
461 			 * Unlink it from its process group and free it.
462 			 */
463 			leavepgrp(p);
464 
465 			LIST_REMOVE(p, p_list);	/* off zombproc */
466 
467 			LIST_REMOVE(p, p_sibling);
468 
469 			/*
470 			 * Decrement the count of procs running with this uid.
471 			 */
472 			(void)chgproccnt(p->p_cred->p_ruid, -1);
473 
474 			/*
475 			 * Free up credentials.
476 			 */
477 			if (--p->p_cred->p_refcnt == 0) {
478 				crfree(p->p_cred->pc_ucred);
479 				pool_put(&pcred_pool, p->p_cred);
480 			}
481 
482 			/*
483 			 * Release reference to text vnode
484 			 */
485 			if (p->p_textvp)
486 				vrele(p->p_textvp);
487 
488 			/*
489 			 * Give machine-dependent layer a chance
490 			 * to free anything that cpu_exit couldn't
491 			 * release while still running in process context.
492 			 */
493 			cpu_wait(p);
494 			pool_put(&proc_pool, p);
495 			nprocs--;
496 			return (0);
497 		}
498 		if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
499 		    (p->p_flag & P_TRACED || SCARG(uap, options) & WUNTRACED)) {
500 			p->p_flag |= P_WAITED;
501 			retval[0] = p->p_pid;
502 
503 			if (SCARG(uap, status)) {
504 				status = W_STOPCODE(p->p_xstat);
505 				error = copyout((caddr_t)&status,
506 				    (caddr_t)SCARG(uap, status),
507 				    sizeof(status));
508 			} else
509 				error = 0;
510 			return (error);
511 		}
512 	}
513 	if (nfound == 0)
514 		return (ECHILD);
515 	if (SCARG(uap, options) & WNOHANG) {
516 		retval[0] = 0;
517 		return (0);
518 	}
519 	if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)) != 0)
520 		return (error);
521 	goto loop;
522 }
523 
524 /*
525  * make process 'parent' the new parent of process 'child'.
526  */
527 void
528 proc_reparent(child, parent)
529 	register struct proc *child;
530 	register struct proc *parent;
531 {
532 
533 	if (child->p_pptr == parent)
534 		return;
535 
536 	LIST_REMOVE(child, p_sibling);
537 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
538 	child->p_pptr = parent;
539 }
540