xref: /netbsd-src/sys/kern/kern_exit.c (revision 27578b9aac214cc7796ead81dcc5427e79d5f2a0)
1 /*	$NetBSD: kern_exit.c,v 1.91 2001/07/18 19:10:27 thorpej 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(struct proc *p, void *v, register_t *retval)
127 {
128 	struct sys_exit_args /* {
129 		syscallarg(int)	rval;
130 	} */ *uap = v;
131 
132 	exit1(p, W_EXITCODE(SCARG(uap, rval), 0));
133 	/* NOTREACHED */
134 	return (0);
135 }
136 
137 /*
138  * Exit: deallocate address space and other resources, change proc state
139  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
140  * status and rusage for wait().  Check for child processes and orphan them.
141  */
142 void
143 exit1(struct proc *p, int rv)
144 {
145 	struct proc	*q, *nq;
146 	int		s;
147 
148 	if (__predict_false(p == initproc))
149 		panic("init died (signal %d, exit %d)",
150 		    WTERMSIG(rv), WEXITSTATUS(rv));
151 
152 #ifdef PGINPROF
153 	vmsizmon();
154 #endif
155 	if (p->p_flag & P_PROFIL)
156 		stopprofclock(p);
157 	p->p_ru = pool_get(&rusage_pool, PR_WAITOK);
158 	/*
159 	 * If parent is waiting for us to exit or exec, P_PPWAIT is set; we
160 	 * wake up the parent early to avoid deadlock.
161 	 */
162 	p->p_flag |= P_WEXIT;
163 	if (p->p_flag & P_PPWAIT) {
164 		p->p_flag &= ~P_PPWAIT;
165 		wakeup((caddr_t)p->p_pptr);
166 	}
167 	sigfillset(&p->p_sigctx.ps_sigignore);
168 	sigemptyset(&p->p_sigctx.ps_siglist);
169 	p->p_sigctx.ps_sigcheck = 0;
170 	callout_stop(&p->p_realit_ch);
171 
172 	/*
173 	 * Close open files and release open-file table.
174 	 * This may block!
175 	 */
176 	fdfree(p);
177 	cwdfree(p);
178 
179 #ifdef SYSVSEM
180 	semexit(p);
181 #endif
182 	if (SESS_LEADER(p)) {
183 		struct session *sp = p->p_session;
184 
185 		if (sp->s_ttyvp) {
186 			/*
187 			 * Controlling process.
188 			 * Signal foreground pgrp,
189 			 * drain controlling terminal
190 			 * and revoke access to controlling terminal.
191 			 */
192 			if (sp->s_ttyp->t_session == sp) {
193 				if (sp->s_ttyp->t_pgrp)
194 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
195 				(void) ttywait(sp->s_ttyp);
196 				/*
197 				 * The tty could have been revoked
198 				 * if we blocked.
199 				 */
200 				if (sp->s_ttyvp)
201 					VOP_REVOKE(sp->s_ttyvp, REVOKEALL);
202 			}
203 			if (sp->s_ttyvp)
204 				vrele(sp->s_ttyvp);
205 			sp->s_ttyvp = NULL;
206 			/*
207 			 * s_ttyp is not zero'd; we use this to indicate
208 			 * that the session once had a controlling terminal.
209 			 * (for logging and informational purposes)
210 			 */
211 		}
212 		sp->s_leader = NULL;
213 	}
214 	fixjobc(p, p->p_pgrp, 0);
215 	(void)acct_process(p);
216 #ifdef KTRACE
217 	/*
218 	 * release trace file
219 	 */
220 	ktrderef(p);
221 #endif
222 	/*
223 	 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP!
224 	 */
225 	p->p_stat = SDEAD;
226 
227 	/*
228 	 * Remove proc from pidhash chain so looking it up won't
229 	 * work.  Move it from allproc to zombproc, but do not yet
230 	 * wake up the reaper.  We will put the proc on the
231 	 * deadproc list later (using the p_hash member), and
232 	 * wake up the reaper when we do.
233 	 */
234 	s = proclist_lock_write();
235 	LIST_REMOVE(p, p_hash);
236 	LIST_REMOVE(p, p_list);
237 	LIST_INSERT_HEAD(&zombproc, p, p_list);
238 	proclist_unlock_write(s);
239 
240 	/*
241 	 * Give orphaned children to init(8).
242 	 */
243 	q = p->p_children.lh_first;
244 	if (q)		/* only need this if any child is S_ZOMB */
245 		wakeup((caddr_t)initproc);
246 	for (; q != 0; q = nq) {
247 		nq = q->p_sibling.le_next;
248 		proc_reparent(q, initproc);
249 		/*
250 		 * Traced processes are killed
251 		 * since their existence means someone is screwing up.
252 		 */
253 		if (q->p_flag & P_TRACED) {
254 			q->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
255 			psignal(q, SIGKILL);
256 		}
257 	}
258 
259 	/*
260 	 * Save exit status and final rusage info, adding in child rusage
261 	 * info and self times.
262 	 */
263 	p->p_xstat = rv;
264 	*p->p_ru = p->p_stats->p_ru;
265 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
266 	ruadd(p->p_ru, &p->p_stats->p_cru);
267 
268 	/*
269 	 * Notify parent that we're gone.  If parent has the P_NOCLDWAIT
270 	 * flag set, notify init instead (and hope it will handle
271 	 * this situation).
272 	 */
273 	if (p->p_pptr->p_flag & P_NOCLDWAIT) {
274 		struct proc *pp = p->p_pptr;
275 		proc_reparent(p, initproc);
276 		/*
277 		 * If this was the last child of our parent, notify
278 		 * parent, so in case he was wait(2)ing, he will
279 		 * continue.
280 		 */
281 		if (pp->p_children.lh_first == NULL)
282 			wakeup((caddr_t)pp);
283 	}
284 
285 	/*
286 	 * Release the process's signal state.
287 	 */
288 	sigactsfree(p);
289 
290 	/*
291 	 * Clear curproc after we've done all operations
292 	 * that could block, and before tearing down the rest
293 	 * of the process state that might be used from clock, etc.
294 	 * Also, can't clear curproc while we're still runnable,
295 	 * as we're not on a run queue (we are current, just not
296 	 * a proper proc any longer!).
297 	 *
298 	 * Other substructures are freed from wait().
299 	 */
300 	curproc = NULL;
301 	limfree(p->p_limit);
302 	p->p_limit = NULL;
303 
304 	/*
305 	 * If emulation has process exit hook, call it now.
306 	 */
307 	if (p->p_emul->e_proc_exit)
308 		(*p->p_emul->e_proc_exit)(p);
309 
310 	/* This process no longer needs to hold the kernel lock. */
311 	KERNEL_PROC_UNLOCK(p);
312 
313 	/*
314 	 * Finally, call machine-dependent code to switch to a new
315 	 * context (possibly the idle context).  Once we are no longer
316 	 * using the dead process's vmspace and stack, exit2() will be
317 	 * called to schedule those resources to be released by the
318 	 * reaper thread.
319 	 *
320 	 * Note that cpu_exit() will end with a call equivalent to
321 	 * cpu_switch(), finishing our execution (pun intended).
322 	 */
323 	cpu_exit(p);
324 }
325 
326 /*
327  * We are called from cpu_exit() once it is safe to schedule the
328  * dead process's resources to be freed (i.e., once we've switched to
329  * the idle PCB for the current CPU).
330  *
331  * NOTE: One must be careful with locking in this routine.  It's
332  * called from a critical section in machine-dependent code, so
333  * we should refrain from changing any interrupt state.
334  *
335  * We lock the deadproc list (a spin lock), place the proc on that
336  * list (using the p_hash member), and wake up the reaper.
337  */
338 void
339 exit2(struct proc *p)
340 {
341 
342 	simple_lock(&deadproc_slock);
343 	LIST_INSERT_HEAD(&deadproc, p, p_hash);
344 	simple_unlock(&deadproc_slock);
345 
346 	wakeup(&deadproc);
347 }
348 
349 /*
350  * Process reaper.  This is run by a kernel thread to free the resources
351  * of a dead process.  Once the resources are free, the process becomes
352  * a zombie, and the parent is allowed to read the undead's status.
353  */
354 void
355 reaper(void *arg)
356 {
357 	struct proc *p;
358 
359 	KERNEL_PROC_UNLOCK(curproc);
360 
361 	for (;;) {
362 		simple_lock(&deadproc_slock);
363 		p = LIST_FIRST(&deadproc);
364 		if (p == NULL) {
365 			/* No work for us; go to sleep until someone exits. */
366 			(void) ltsleep(&deadproc, PVM|PNORELOCK,
367 			    "reaper", 0, &deadproc_slock);
368 			continue;
369 		}
370 
371 		/* Remove us from the deadproc list. */
372 		LIST_REMOVE(p, p_hash);
373 		simple_unlock(&deadproc_slock);
374 		KERNEL_PROC_LOCK(curproc);
375 
376 		/*
377 		 * Give machine-dependent code a chance to free any
378 		 * resources it couldn't free while still running on
379 		 * that process's context.  This must be done before
380 		 * uvm_exit(), in case these resources are in the PCB.
381 		 */
382 		cpu_wait(p);
383 
384 		/*
385 		 * Free the VM resources we're still holding on to.
386 		 * We must do this from a valid thread because doing
387 		 * so may block.
388 		 */
389 		uvm_exit(p);
390 
391 		/* Process is now a true zombie. */
392 		p->p_stat = SZOMB;
393 
394 		/* Wake up the parent so it can get exit status. */
395 		if ((p->p_flag & P_FSTRACE) == 0 && p->p_exitsig != 0)
396 			psignal(p->p_pptr, P_EXITSIG(p));
397 		KERNEL_PROC_UNLOCK(curproc);
398 		wakeup((caddr_t)p->p_pptr);
399 	}
400 }
401 
402 int
403 sys_wait4(struct proc *q, void *v, register_t *retval)
404 {
405 	struct sys_wait4_args /* {
406 		syscallarg(int)			pid;
407 		syscallarg(int *)		status;
408 		syscallarg(int)			options;
409 		syscallarg(struct rusage *)	rusage;
410 	} */ *uap = v;
411 	struct proc	*p, *t;
412 	int		nfound, status, error, s;
413 
414 	if (SCARG(uap, pid) == 0)
415 		SCARG(uap, pid) = -q->p_pgid;
416 	if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG|WALTSIG))
417 		return (EINVAL);
418 
419  loop:
420 	nfound = 0;
421 	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
422 		if (SCARG(uap, pid) != WAIT_ANY &&
423 		    p->p_pid != SCARG(uap, pid) &&
424 		    p->p_pgid != -SCARG(uap, pid))
425 			continue;
426 		/*
427 		 * Wait for processes with p_exitsig != SIGCHLD processes only
428 		 * if WALTSIG is set; wait for processes with p_exitsig ==
429 		 * SIGCHLD only if WALTSIG is clear.
430 		 */
431 		if (((SCARG(uap, options) & WALLSIG) == 0) &&
432 		    ((SCARG(uap, options) & WALTSIG) ?
433 		     (p->p_exitsig == SIGCHLD) : (P_EXITSIG(p) != SIGCHLD)))
434 			continue;
435 
436 		nfound++;
437 		if (p->p_stat == SZOMB) {
438 			retval[0] = p->p_pid;
439 
440 			if (SCARG(uap, status)) {
441 				status = p->p_xstat;	/* convert to int */
442 				error = copyout((caddr_t)&status,
443 						(caddr_t)SCARG(uap, status),
444 						sizeof(status));
445 				if (error)
446 					return (error);
447 			}
448 			if (SCARG(uap, rusage) &&
449 			    (error = copyout((caddr_t)p->p_ru,
450 			    (caddr_t)SCARG(uap, rusage),
451 			    sizeof(struct rusage))))
452 				return (error);
453 			/*
454 			 * If we got the child via ptrace(2) or procfs, and
455 			 * the parent is different (meaning the process was
456 			 * attached, rather than run as a child), then we need
457 			 * to give it back to the old parent, and send the
458 			 * parent the exit signal.  The rest of the cleanup
459 			 * will be done when the old parent waits on the child.
460 			 */
461 			if ((p->p_flag & P_TRACED) &&
462 			    p->p_oppid != p->p_pptr->p_pid) {
463 				t = pfind(p->p_oppid);
464 				proc_reparent(p, t ? t : initproc);
465 				p->p_oppid = 0;
466 				p->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
467 				if (p->p_exitsig != 0)
468 					psignal(p->p_pptr, P_EXITSIG(p));
469 				wakeup((caddr_t)p->p_pptr);
470 				return (0);
471 			}
472 			scheduler_wait_hook(q, p);
473 			p->p_xstat = 0;
474 			ruadd(&q->p_stats->p_cru, p->p_ru);
475 			pool_put(&rusage_pool, p->p_ru);
476 
477 			/*
478 			 * Finally finished with old proc entry.
479 			 * Unlink it from its process group and free it.
480 			 */
481 			leavepgrp(p);
482 
483 			s = proclist_lock_write();
484 			LIST_REMOVE(p, p_list);	/* off zombproc */
485 			proclist_unlock_write(s);
486 
487 			LIST_REMOVE(p, p_sibling);
488 
489 			/*
490 			 * Decrement the count of procs running with this uid.
491 			 */
492 			(void)chgproccnt(p->p_cred->p_ruid, -1);
493 
494 			/*
495 			 * Free up credentials.
496 			 */
497 			if (--p->p_cred->p_refcnt == 0) {
498 				crfree(p->p_cred->pc_ucred);
499 				pool_put(&pcred_pool, p->p_cred);
500 			}
501 
502 			/*
503 			 * Release reference to text vnode
504 			 */
505 			if (p->p_textvp)
506 				vrele(p->p_textvp);
507 
508 			pool_put(&proc_pool, p);
509 			nprocs--;
510 			return (0);
511 		}
512 		if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
513 		    (p->p_flag & P_TRACED || SCARG(uap, options) & WUNTRACED)) {
514 			p->p_flag |= P_WAITED;
515 			retval[0] = p->p_pid;
516 
517 			if (SCARG(uap, status)) {
518 				status = W_STOPCODE(p->p_xstat);
519 				error = copyout((caddr_t)&status,
520 				    (caddr_t)SCARG(uap, status),
521 				    sizeof(status));
522 			} else
523 				error = 0;
524 			return (error);
525 		}
526 	}
527 	if (nfound == 0)
528 		return (ECHILD);
529 	if (SCARG(uap, options) & WNOHANG) {
530 		retval[0] = 0;
531 		return (0);
532 	}
533 	if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)) != 0)
534 		return (error);
535 	goto loop;
536 }
537 
538 /*
539  * make process 'parent' the new parent of process 'child'.
540  */
541 void
542 proc_reparent(struct proc *child, struct proc *parent)
543 {
544 
545 	if (child->p_pptr == parent)
546 		return;
547 
548 	if (parent == initproc)
549 		child->p_exitsig = SIGCHLD;
550 
551 	LIST_REMOVE(child, p_sibling);
552 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
553 	child->p_pptr = parent;
554 }
555