xref: /netbsd-src/sys/kern/kern_exit.c (revision 481fca6e59249d8ffcf24fef7cfbe7b131bfb080)
1 /*	$NetBSD: kern_exit.c,v 1.81 2000/06/27 17:41:18 mrg 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 <uvm/uvm_extern.h>
120 
121 /*
122  * exit --
123  *	Death of process.
124  */
125 int
126 sys_exit(p, v, retval)
127 	struct proc *p;
128 	void *v;
129 	register_t *retval;
130 {
131 	struct sys_exit_args /* {
132 		syscallarg(int) rval;
133 	} */ *uap = v;
134 
135 	exit1(p, W_EXITCODE(SCARG(uap, rval), 0));
136 	/* NOTREACHED */
137 	return (0);
138 }
139 
140 /*
141  * Exit: deallocate address space and other resources, change proc state
142  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
143  * status and rusage for wait().  Check for child processes and orphan them.
144  */
145 void
146 exit1(p, rv)
147 	struct proc *p;
148 	int rv;
149 {
150 	struct proc *q, *nq;
151 	struct vmspace *vm;
152 	int s;
153 
154 	if (__predict_false(p == initproc))
155 		panic("init died (signal %d, exit %d)",
156 		    WTERMSIG(rv), WEXITSTATUS(rv));
157 
158 #ifdef PGINPROF
159 	vmsizmon();
160 #endif
161 	if (p->p_flag & P_PROFIL)
162 		stopprofclock(p);
163 	p->p_ru = pool_get(&rusage_pool, PR_WAITOK);
164 	/*
165 	 * If parent is waiting for us to exit or exec, P_PPWAIT is set; we
166 	 * wake up the parent early to avoid deadlock.
167 	 */
168 	p->p_flag |= P_WEXIT;
169 	if (p->p_flag & P_PPWAIT) {
170 		p->p_flag &= ~P_PPWAIT;
171 		wakeup((caddr_t)p->p_pptr);
172 	}
173 	sigfillset(&p->p_sigignore);
174 	sigemptyset(&p->p_siglist);
175 	p->p_sigcheck = 0;
176 	callout_stop(&p->p_realit_ch);
177 
178 	/*
179 	 * Close open files and release open-file table.
180 	 * This may block!
181 	 */
182 	fdfree(p);
183 	cwdfree(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 		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 	(void)acct_process(p);
240 #ifdef KTRACE
241 	/*
242 	 * release trace file
243 	 */
244 	ktrderef(p);
245 #endif
246 	/*
247 	 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP!
248 	 */
249 	p->p_stat = SDEAD;
250 
251 	/*
252 	 * Remove proc from pidhash chain so looking it up won't
253 	 * work.  Move it from allproc to zombproc, but do not yet
254 	 * wake up the reaper.  We will put the proc on the
255 	 * deadproc list later (using the p_hash member), and
256 	 * wake up the reaper when we do.
257 	 */
258 	s = proclist_lock_write();
259 	LIST_REMOVE(p, p_hash);
260 	LIST_REMOVE(p, p_list);
261 	LIST_INSERT_HEAD(&zombproc, p, p_list);
262 	proclist_unlock_write(s);
263 
264 	/*
265 	 * Give orphaned children to init(8).
266 	 */
267 	q = p->p_children.lh_first;
268 	if (q)		/* only need this if any child is S_ZOMB */
269 		wakeup((caddr_t)initproc);
270 	for (; q != 0; q = nq) {
271 		nq = q->p_sibling.le_next;
272 		proc_reparent(q, initproc);
273 		/*
274 		 * Traced processes are killed
275 		 * since their existence means someone is screwing up.
276 		 */
277 		if (q->p_flag & P_TRACED) {
278 			q->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
279 			psignal(q, SIGKILL);
280 		}
281 	}
282 
283 	/*
284 	 * Save exit status and final rusage info, adding in child rusage
285 	 * info and self times.
286 	 */
287 	p->p_xstat = rv;
288 	*p->p_ru = p->p_stats->p_ru;
289 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
290 	ruadd(p->p_ru, &p->p_stats->p_cru);
291 
292 	/*
293 	 * Notify parent that we're gone.  If parent has the P_NOCLDWAIT
294 	 * flag set, notify init instead (and hope it will handle
295 	 * this situation).
296 	 */
297 	if (p->p_pptr->p_flag & P_NOCLDWAIT) {
298 		struct proc *pp = p->p_pptr;
299 		proc_reparent(p, initproc);
300 		/*
301 		 * If this was the last child of our parent, notify
302 		 * parent, so in case he was wait(2)ing, he will
303 		 * continue.
304 		 */
305 		if (pp->p_children.lh_first == NULL)
306 			wakeup((caddr_t)pp);
307 	}
308 
309 	/*
310 	 * Release the process's signal state.
311 	 */
312 	sigactsfree(p);
313 
314 	/*
315 	 * Clear curproc after we've done all operations
316 	 * that could block, and before tearing down the rest
317 	 * of the process state that might be used from clock, etc.
318 	 * Also, can't clear curproc while we're still runnable,
319 	 * as we're not on a run queue (we are current, just not
320 	 * a proper proc any longer!).
321 	 *
322 	 * Other substructures are freed from wait().
323 	 */
324 	curproc = NULL;
325 	limfree(p->p_limit);
326 	p->p_limit = NULL;
327 
328 	/*
329 	 * Finally, call machine-dependent code to switch to a new
330 	 * context (possibly the idle context).  Once we are no longer
331 	 * using the dead process's vmspace and stack, exit2() will be
332 	 * called to schedule those resources to be released by the
333 	 * reaper thread.
334 	 *
335 	 * Note that cpu_exit() will end with a call equivalent to
336 	 * cpu_switch(), finishing our execution (pun intended).
337 	 */
338 	cpu_exit(p);
339 }
340 
341 /*
342  * We are called from cpu_exit() once it is safe to schedule the
343  * dead process's resources to be freed.
344  *
345  * NOTE: One must be careful with locking in this routine.  It's
346  * called from a critical section in machine-dependent code, so
347  * we should refrain from changing any interrupt state.
348  *
349  * We lock the deadproc list (a spin lock), place the proc on that
350  * list (using the p_hash member), and wake up the reaper.
351  */
352 void
353 exit2(p)
354 	struct proc *p;
355 {
356 
357 	simple_lock(&deadproc_slock);
358 	LIST_INSERT_HEAD(&deadproc, p, p_hash);
359 	simple_unlock(&deadproc_slock);
360 
361 	wakeup(&deadproc);
362 }
363 
364 /*
365  * Process reaper.  This is run by a kernel thread to free the resources
366  * of a dead process.  Once the resources are free, the process becomes
367  * a zombie, and the parent is allowed to read the undead's status.
368  */
369 void
370 reaper()
371 {
372 	struct proc *p;
373 
374 	for (;;) {
375 		simple_lock(&deadproc_slock);
376 		p = LIST_FIRST(&deadproc);
377 		if (p == NULL) {
378 			/* No work for us; go to sleep until someone exits. */
379 			simple_unlock(&deadproc_slock);
380 			(void) tsleep(&deadproc, PVM, "reaper", 0);
381 			continue;
382 		}
383 
384 		/* Remove us from the deadproc list. */
385 		LIST_REMOVE(p, p_hash);
386 		simple_unlock(&deadproc_slock);
387 
388 		/*
389 		 * Give machine-dependent code a chance to free any
390 		 * resources it couldn't free while still running on
391 		 * that process's context.  This must be done before
392 		 * uvm_exit(), in case these resources are in the PCB.
393 		 */
394 		cpu_wait(p);
395 
396 		/*
397 		 * Free the VM resources we're still holding on to.
398 		 * We must do this from a valid thread because doing
399 		 * so may block.
400 		 */
401 		uvm_exit(p);
402 
403 		/* Process is now a true zombie. */
404 		p->p_stat = SZOMB;
405 
406 		/* Wake up the parent so it can get exit status. */
407 		if ((p->p_flag & P_FSTRACE) == 0 && p->p_exitsig != 0)
408 			psignal(p->p_pptr, P_EXITSIG(p));
409 		wakeup((caddr_t)p->p_pptr);
410 	}
411 }
412 
413 int
414 sys_wait4(q, v, retval)
415 	struct proc *q;
416 	void *v;
417 	register_t *retval;
418 {
419 	struct sys_wait4_args /* {
420 		syscallarg(int) pid;
421 		syscallarg(int *) status;
422 		syscallarg(int) options;
423 		syscallarg(struct rusage *) rusage;
424 	} */ *uap = v;
425 	int nfound;
426 	struct proc *p, *t;
427 	int status, error, s;
428 
429 	if (SCARG(uap, pid) == 0)
430 		SCARG(uap, pid) = -q->p_pgid;
431 	if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG|WALTSIG))
432 		return (EINVAL);
433 
434 loop:
435 	nfound = 0;
436 	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
437 		if (SCARG(uap, pid) != WAIT_ANY &&
438 		    p->p_pid != SCARG(uap, pid) &&
439 		    p->p_pgid != -SCARG(uap, pid))
440 			continue;
441 		/*
442 		 * Wait for processes with p_exitsig != SIGCHLD processes only
443 		 * if WALTSIG is set; wait for processes with p_exitsig ==
444 		 * SIGCHLD only if WALTSIG is clear.
445 		 */
446 		if ((SCARG(uap, options) & WALTSIG) ?
447 		    (p->p_exitsig == SIGCHLD) : (P_EXITSIG(p) != SIGCHLD))
448 			continue;
449 
450 		nfound++;
451 		if (p->p_stat == SZOMB) {
452 			retval[0] = p->p_pid;
453 
454 			if (SCARG(uap, status)) {
455 				status = p->p_xstat;	/* convert to int */
456 				error = copyout((caddr_t)&status,
457 						(caddr_t)SCARG(uap, status),
458 						sizeof(status));
459 				if (error)
460 					return (error);
461 			}
462 			if (SCARG(uap, rusage) &&
463 			    (error = copyout((caddr_t)p->p_ru,
464 			    (caddr_t)SCARG(uap, rusage),
465 			    sizeof(struct rusage))))
466 				return (error);
467 			/*
468 			 * If we got the child via ptrace(2) or procfs, and
469 			 * the parent is different (meaning the process was
470 			 * attached, rather than run as a child), then we need
471 			 * to give it back to the old parent, and send the
472 			 * parent the exit signal.  The rest of the cleanup
473 			 * will be done when the old parent waits on the child.
474 			 */
475 			if ((p->p_flag & P_TRACED) &&
476 			    p->p_oppid != p->p_pptr->p_pid) {
477 				t = pfind(p->p_oppid);
478 				proc_reparent(p, t ? t : initproc);
479 				p->p_oppid = 0;
480 				p->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
481 				if (p->p_exitsig != 0)
482 					psignal(p->p_pptr, P_EXITSIG(p));
483 				wakeup((caddr_t)p->p_pptr);
484 				return (0);
485 			}
486 			scheduler_wait_hook(q, p);
487 			p->p_xstat = 0;
488 			ruadd(&q->p_stats->p_cru, p->p_ru);
489 			pool_put(&rusage_pool, p->p_ru);
490 
491 			/*
492 			 * Finally finished with old proc entry.
493 			 * Unlink it from its process group and free it.
494 			 */
495 			leavepgrp(p);
496 
497 			s = proclist_lock_write();
498 			LIST_REMOVE(p, p_list);	/* off zombproc */
499 			proclist_unlock_write(s);
500 
501 			LIST_REMOVE(p, p_sibling);
502 
503 			/*
504 			 * Decrement the count of procs running with this uid.
505 			 */
506 			(void)chgproccnt(p->p_cred->p_ruid, -1);
507 
508 			/*
509 			 * Free up credentials.
510 			 */
511 			if (--p->p_cred->p_refcnt == 0) {
512 				crfree(p->p_cred->pc_ucred);
513 				pool_put(&pcred_pool, p->p_cred);
514 			}
515 
516 			/*
517 			 * Release reference to text vnode
518 			 */
519 			if (p->p_textvp)
520 				vrele(p->p_textvp);
521 
522 			pool_put(&proc_pool, p);
523 			nprocs--;
524 			return (0);
525 		}
526 		if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
527 		    (p->p_flag & P_TRACED || SCARG(uap, options) & WUNTRACED)) {
528 			p->p_flag |= P_WAITED;
529 			retval[0] = p->p_pid;
530 
531 			if (SCARG(uap, status)) {
532 				status = W_STOPCODE(p->p_xstat);
533 				error = copyout((caddr_t)&status,
534 				    (caddr_t)SCARG(uap, status),
535 				    sizeof(status));
536 			} else
537 				error = 0;
538 			return (error);
539 		}
540 	}
541 	if (nfound == 0)
542 		return (ECHILD);
543 	if (SCARG(uap, options) & WNOHANG) {
544 		retval[0] = 0;
545 		return (0);
546 	}
547 	if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)) != 0)
548 		return (error);
549 	goto loop;
550 }
551 
552 /*
553  * make process 'parent' the new parent of process 'child'.
554  */
555 void
556 proc_reparent(child, parent)
557 	struct proc *child;
558 	struct proc *parent;
559 {
560 
561 	if (child->p_pptr == parent)
562 		return;
563 
564 	if (parent == initproc)
565 		child->p_exitsig = SIGCHLD;
566 
567 	LIST_REMOVE(child, p_sibling);
568 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
569 	child->p_pptr = parent;
570 }
571