xref: /netbsd-src/sys/kern/kern_exit.c (revision 5aefcfdc06931dd97e76246d2fe0302f7b3fe094)
1 /*	$NetBSD: kern_exit.c,v 1.87 2000/12/22 22:59:00 jdolecek 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 	struct vmspace *vm;
147 	int s;
148 
149 	if (__predict_false(p == initproc))
150 		panic("init died (signal %d, exit %d)",
151 		    WTERMSIG(rv), WEXITSTATUS(rv));
152 
153 #ifdef PGINPROF
154 	vmsizmon();
155 #endif
156 	if (p->p_flag & P_PROFIL)
157 		stopprofclock(p);
158 	p->p_ru = pool_get(&rusage_pool, PR_WAITOK);
159 	/*
160 	 * If parent is waiting for us to exit or exec, P_PPWAIT is set; we
161 	 * wake up the parent early to avoid deadlock.
162 	 */
163 	p->p_flag |= P_WEXIT;
164 	if (p->p_flag & P_PPWAIT) {
165 		p->p_flag &= ~P_PPWAIT;
166 		wakeup((caddr_t)p->p_pptr);
167 	}
168 	sigfillset(&p->p_sigctx.ps_sigignore);
169 	sigemptyset(&p->p_sigctx.ps_siglist);
170 	p->p_sigctx.ps_sigcheck = 0;
171 	callout_stop(&p->p_realit_ch);
172 
173 	/*
174 	 * Close open files and release open-file table.
175 	 * This may block!
176 	 */
177 	fdfree(p);
178 	cwdfree(p);
179 
180 	/* The next three chunks should probably be moved to vmspace_exit. */
181 	vm = p->p_vmspace;
182 #ifdef SYSVSHM
183 	if (vm->vm_shm && vm->vm_refcnt == 1)
184 		shmexit(vm);
185 #endif
186 #ifdef SYSVSEM
187 	semexit(p);
188 #endif
189 	/*
190 	 * Release user portion of address space.
191 	 * This releases references to vnodes,
192 	 * which could cause I/O if the file has been unlinked.
193 	 * Need to do this early enough that we can still sleep.
194 	 * Can't free the entire vmspace as the kernel stack
195 	 * may be mapped within that space also.
196 	 */
197 	if (vm->vm_refcnt == 1)
198 		(void) uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
199 		    VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
200 
201 	if (SESS_LEADER(p)) {
202 		struct session *sp = p->p_session;
203 
204 		if (sp->s_ttyvp) {
205 			/*
206 			 * Controlling process.
207 			 * Signal foreground pgrp,
208 			 * drain controlling terminal
209 			 * and revoke access to controlling terminal.
210 			 */
211 			if (sp->s_ttyp->t_session == sp) {
212 				if (sp->s_ttyp->t_pgrp)
213 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
214 				(void) ttywait(sp->s_ttyp);
215 				/*
216 				 * The tty could have been revoked
217 				 * if we blocked.
218 				 */
219 				if (sp->s_ttyvp)
220 					VOP_REVOKE(sp->s_ttyvp, REVOKEALL);
221 			}
222 			if (sp->s_ttyvp)
223 				vrele(sp->s_ttyvp);
224 			sp->s_ttyvp = NULL;
225 			/*
226 			 * s_ttyp is not zero'd; we use this to indicate
227 			 * that the session once had a controlling terminal.
228 			 * (for logging and informational purposes)
229 			 */
230 		}
231 		sp->s_leader = NULL;
232 	}
233 	fixjobc(p, p->p_pgrp, 0);
234 	(void)acct_process(p);
235 #ifdef KTRACE
236 	/*
237 	 * release trace file
238 	 */
239 	ktrderef(p);
240 #endif
241 	/*
242 	 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP!
243 	 */
244 	p->p_stat = SDEAD;
245 
246 	/*
247 	 * Remove proc from pidhash chain so looking it up won't
248 	 * work.  Move it from allproc to zombproc, but do not yet
249 	 * wake up the reaper.  We will put the proc on the
250 	 * deadproc list later (using the p_hash member), and
251 	 * wake up the reaper when we do.
252 	 */
253 	s = proclist_lock_write();
254 	LIST_REMOVE(p, p_hash);
255 	LIST_REMOVE(p, p_list);
256 	LIST_INSERT_HEAD(&zombproc, p, p_list);
257 	proclist_unlock_write(s);
258 
259 	/*
260 	 * Give orphaned children to init(8).
261 	 */
262 	q = p->p_children.lh_first;
263 	if (q)		/* only need this if any child is S_ZOMB */
264 		wakeup((caddr_t)initproc);
265 	for (; q != 0; q = nq) {
266 		nq = q->p_sibling.le_next;
267 		proc_reparent(q, initproc);
268 		/*
269 		 * Traced processes are killed
270 		 * since their existence means someone is screwing up.
271 		 */
272 		if (q->p_flag & P_TRACED) {
273 			q->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
274 			psignal(q, SIGKILL);
275 		}
276 	}
277 
278 	/*
279 	 * Save exit status and final rusage info, adding in child rusage
280 	 * info and self times.
281 	 */
282 	p->p_xstat = rv;
283 	*p->p_ru = p->p_stats->p_ru;
284 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
285 	ruadd(p->p_ru, &p->p_stats->p_cru);
286 
287 	/*
288 	 * Notify parent that we're gone.  If parent has the P_NOCLDWAIT
289 	 * flag set, notify init instead (and hope it will handle
290 	 * this situation).
291 	 */
292 	if (p->p_pptr->p_flag & P_NOCLDWAIT) {
293 		struct proc *pp = p->p_pptr;
294 		proc_reparent(p, initproc);
295 		/*
296 		 * If this was the last child of our parent, notify
297 		 * parent, so in case he was wait(2)ing, he will
298 		 * continue.
299 		 */
300 		if (pp->p_children.lh_first == NULL)
301 			wakeup((caddr_t)pp);
302 	}
303 
304 	/*
305 	 * Release the process's signal state.
306 	 */
307 	sigactsfree(p);
308 
309 	/*
310 	 * Clear curproc after we've done all operations
311 	 * that could block, and before tearing down the rest
312 	 * of the process state that might be used from clock, etc.
313 	 * Also, can't clear curproc while we're still runnable,
314 	 * as we're not on a run queue (we are current, just not
315 	 * a proper proc any longer!).
316 	 *
317 	 * Other substructures are freed from wait().
318 	 */
319 	curproc = NULL;
320 	limfree(p->p_limit);
321 	p->p_limit = NULL;
322 
323 	/*
324 	 * If emulation has process exit hook, call it now.
325 	 */
326 	if (p->p_emul->e_proc_exit)
327 		(*p->p_emul->e_proc_exit)(p);
328 
329 	/* This process no longer needs to hold the kernel lock. */
330 	KERNEL_PROC_UNLOCK(p);
331 
332 	/*
333 	 * Finally, call machine-dependent code to switch to a new
334 	 * context (possibly the idle context).  Once we are no longer
335 	 * using the dead process's vmspace and stack, exit2() will be
336 	 * called to schedule those resources to be released by the
337 	 * reaper thread.
338 	 *
339 	 * Note that cpu_exit() will end with a call equivalent to
340 	 * cpu_switch(), finishing our execution (pun intended).
341 	 */
342 	cpu_exit(p);
343 }
344 
345 /*
346  * We are called from cpu_exit() once it is safe to schedule the
347  * dead process's resources to be freed (i.e., once we've switched to
348  * the idle PCB for the current CPU).
349  *
350  * NOTE: One must be careful with locking in this routine.  It's
351  * called from a critical section in machine-dependent code, so
352  * we should refrain from changing any interrupt state.
353  *
354  * We lock the deadproc list (a spin lock), place the proc on that
355  * list (using the p_hash member), and wake up the reaper.
356  */
357 void
358 exit2(struct proc *p)
359 {
360 
361 	simple_lock(&deadproc_slock);
362 	LIST_INSERT_HEAD(&deadproc, p, p_hash);
363 	simple_unlock(&deadproc_slock);
364 
365 	wakeup(&deadproc);
366 }
367 
368 /*
369  * Process reaper.  This is run by a kernel thread to free the resources
370  * of a dead process.  Once the resources are free, the process becomes
371  * a zombie, and the parent is allowed to read the undead's status.
372  */
373 void
374 reaper(void *arg)
375 {
376 	struct proc *p;
377 
378 	for (;;) {
379 		simple_lock(&deadproc_slock);
380 		p = LIST_FIRST(&deadproc);
381 		if (p == NULL) {
382 			/* No work for us; go to sleep until someone exits. */
383 			(void) ltsleep(&deadproc, PVM|PNORELOCK,
384 			    "reaper", 0, &deadproc_slock);
385 			continue;
386 		}
387 
388 		/* Remove us from the deadproc list. */
389 		LIST_REMOVE(p, p_hash);
390 		simple_unlock(&deadproc_slock);
391 
392 		/*
393 		 * Give machine-dependent code a chance to free any
394 		 * resources it couldn't free while still running on
395 		 * that process's context.  This must be done before
396 		 * uvm_exit(), in case these resources are in the PCB.
397 		 */
398 		cpu_wait(p);
399 
400 		/*
401 		 * Free the VM resources we're still holding on to.
402 		 * We must do this from a valid thread because doing
403 		 * so may block.
404 		 */
405 		uvm_exit(p);
406 
407 		/* Process is now a true zombie. */
408 		p->p_stat = SZOMB;
409 
410 		/* Wake up the parent so it can get exit status. */
411 		if ((p->p_flag & P_FSTRACE) == 0 && p->p_exitsig != 0)
412 			psignal(p->p_pptr, P_EXITSIG(p));
413 		wakeup((caddr_t)p->p_pptr);
414 	}
415 }
416 
417 int
418 sys_wait4(struct proc *q, void *v, register_t *retval)
419 {
420 	struct sys_wait4_args /* {
421 		syscallarg(int) pid;
422 		syscallarg(int *) status;
423 		syscallarg(int) options;
424 		syscallarg(struct rusage *) rusage;
425 	} */ *uap = v;
426 	int nfound;
427 	struct proc *p, *t;
428 	int status, error, s;
429 
430 	if (SCARG(uap, pid) == 0)
431 		SCARG(uap, pid) = -q->p_pgid;
432 	if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG|WALTSIG))
433 		return (EINVAL);
434 
435 loop:
436 	nfound = 0;
437 	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
438 		if (SCARG(uap, pid) != WAIT_ANY &&
439 		    p->p_pid != SCARG(uap, pid) &&
440 		    p->p_pgid != -SCARG(uap, pid))
441 			continue;
442 		/*
443 		 * Wait for processes with p_exitsig != SIGCHLD processes only
444 		 * if WALTSIG is set; wait for processes with p_exitsig ==
445 		 * SIGCHLD only if WALTSIG is clear.
446 		 */
447 		if ((SCARG(uap, options) & WALTSIG) ?
448 		    (p->p_exitsig == SIGCHLD) : (P_EXITSIG(p) != SIGCHLD))
449 			continue;
450 
451 		nfound++;
452 		if (p->p_stat == SZOMB) {
453 			retval[0] = p->p_pid;
454 
455 			if (SCARG(uap, status)) {
456 				status = p->p_xstat;	/* convert to int */
457 				error = copyout((caddr_t)&status,
458 						(caddr_t)SCARG(uap, status),
459 						sizeof(status));
460 				if (error)
461 					return (error);
462 			}
463 			if (SCARG(uap, rusage) &&
464 			    (error = copyout((caddr_t)p->p_ru,
465 			    (caddr_t)SCARG(uap, rusage),
466 			    sizeof(struct rusage))))
467 				return (error);
468 			/*
469 			 * If we got the child via ptrace(2) or procfs, and
470 			 * the parent is different (meaning the process was
471 			 * attached, rather than run as a child), then we need
472 			 * to give it back to the old parent, and send the
473 			 * parent the exit signal.  The rest of the cleanup
474 			 * will be done when the old parent waits on the child.
475 			 */
476 			if ((p->p_flag & P_TRACED) &&
477 			    p->p_oppid != p->p_pptr->p_pid) {
478 				t = pfind(p->p_oppid);
479 				proc_reparent(p, t ? t : initproc);
480 				p->p_oppid = 0;
481 				p->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE);
482 				if (p->p_exitsig != 0)
483 					psignal(p->p_pptr, P_EXITSIG(p));
484 				wakeup((caddr_t)p->p_pptr);
485 				return (0);
486 			}
487 			scheduler_wait_hook(q, p);
488 			p->p_xstat = 0;
489 			ruadd(&q->p_stats->p_cru, p->p_ru);
490 			pool_put(&rusage_pool, p->p_ru);
491 
492 			/*
493 			 * Finally finished with old proc entry.
494 			 * Unlink it from its process group and free it.
495 			 */
496 			leavepgrp(p);
497 
498 			s = proclist_lock_write();
499 			LIST_REMOVE(p, p_list);	/* off zombproc */
500 			proclist_unlock_write(s);
501 
502 			LIST_REMOVE(p, p_sibling);
503 
504 			/*
505 			 * Decrement the count of procs running with this uid.
506 			 */
507 			(void)chgproccnt(p->p_cred->p_ruid, -1);
508 
509 			/*
510 			 * Free up credentials.
511 			 */
512 			if (--p->p_cred->p_refcnt == 0) {
513 				crfree(p->p_cred->pc_ucred);
514 				pool_put(&pcred_pool, p->p_cred);
515 			}
516 
517 			/*
518 			 * Release reference to text vnode
519 			 */
520 			if (p->p_textvp)
521 				vrele(p->p_textvp);
522 
523 			pool_put(&proc_pool, p);
524 			nprocs--;
525 			return (0);
526 		}
527 		if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
528 		    (p->p_flag & P_TRACED || SCARG(uap, options) & WUNTRACED)) {
529 			p->p_flag |= P_WAITED;
530 			retval[0] = p->p_pid;
531 
532 			if (SCARG(uap, status)) {
533 				status = W_STOPCODE(p->p_xstat);
534 				error = copyout((caddr_t)&status,
535 				    (caddr_t)SCARG(uap, status),
536 				    sizeof(status));
537 			} else
538 				error = 0;
539 			return (error);
540 		}
541 	}
542 	if (nfound == 0)
543 		return (ECHILD);
544 	if (SCARG(uap, options) & WNOHANG) {
545 		retval[0] = 0;
546 		return (0);
547 	}
548 	if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)) != 0)
549 		return (error);
550 	goto loop;
551 }
552 
553 /*
554  * make process 'parent' the new parent of process 'child'.
555  */
556 void
557 proc_reparent(struct proc *child, struct proc *parent)
558 {
559 
560 	if (child->p_pptr == parent)
561 		return;
562 
563 	if (parent == initproc)
564 		child->p_exitsig = SIGCHLD;
565 
566 	LIST_REMOVE(child, p_sibling);
567 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
568 	child->p_pptr = parent;
569 }
570