xref: /freebsd/sys/kern/kern_exit.c (revision ce834215a70ff69e7e222827437116eee2f9ac6f)
1 /*
2  * Copyright (c) 1982, 1986, 1989, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the University of
21  *	California, Berkeley and its contributors.
22  * 4. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  *
38  *	@(#)kern_exit.c	8.7 (Berkeley) 2/12/94
39  * $Id: kern_exit.c,v 1.49 1997/06/16 00:29:30 dyson Exp $
40  */
41 
42 #include "opt_ktrace.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/sysproto.h>
47 #include <sys/sysent.h>
48 #include <sys/proc.h>
49 #include <sys/tty.h>
50 #include <sys/time.h>
51 #include <sys/resource.h>
52 #include <sys/kernel.h>
53 #include <sys/buf.h>
54 #include <sys/wait.h>
55 #include <sys/file.h>
56 #include <sys/vnode.h>
57 #include <sys/syslog.h>
58 #include <sys/malloc.h>
59 #include <sys/resourcevar.h>
60 #include <sys/signalvar.h>
61 #include <sys/ptrace.h>
62 #include <sys/acct.h>		/* for acct_process() function prototype */
63 #include <sys/filedesc.h>
64 #include <sys/shm.h>
65 #include <sys/sem.h>
66 
67 #ifdef COMPAT_43
68 #include <machine/reg.h>
69 #include <machine/psl.h>
70 #endif
71 
72 #include <vm/vm.h>
73 #include <vm/vm_param.h>
74 #include <vm/vm_prot.h>
75 #include <sys/lock.h>
76 #include <vm/pmap.h>
77 #include <vm/vm_map.h>
78 #include <vm/vm_kern.h>
79 
80 static int wait1 __P((struct proc *, struct wait_args *, int [], int));
81 
82 /*
83  * callout list for things to do at exit time
84  */
85 typedef struct exit_list_element {
86 	struct exit_list_element *next;
87 	exitlist_fn function;
88 } *ele_p;
89 
90 static ele_p exit_list;
91 
92 /*
93  * exit --
94  *	Death of process.
95  */
96 void
97 exit(p, uap, retval)
98 	struct proc *p;
99 	struct rexit_args /* {
100 		int	rval;
101 	} */ *uap;
102 	int *retval;
103 {
104 
105 	exit1(p, W_EXITCODE(uap->rval, 0));
106 	/* NOTREACHED */
107 }
108 
109 /*
110  * Exit: deallocate address space and other resources, change proc state
111  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
112  * status and rusage for wait().  Check for child processes and orphan them.
113  */
114 void
115 exit1(p, rv)
116 	register struct proc *p;
117 	int rv;
118 {
119 	register struct proc *q, *nq;
120 	register struct vmspace *vm;
121 	ele_p ep = exit_list;
122 
123 	if (p->p_pid == 1) {
124 		printf("init died (signal %d, exit %d)\n",
125 		    WTERMSIG(rv), WEXITSTATUS(rv));
126 		panic("Going nowhere without my init!");
127 	}
128 
129 	aio_proc_rundown(p);
130 
131 	/* are we a task leader? */
132 	if(p == p->p_leader) {
133         	struct kill_args killArgs;
134 		killArgs.signum = SIGKILL;
135 		q = p->p_peers;
136 		while(q) {
137 			killArgs.pid = q->p_pid;
138 			/*
139 		         * The interface for kill is better
140 			 * than the internal signal
141 			 */
142 			kill(p, &killArgs, &rv);
143 			nq = q;
144 			q = q->p_peers;
145 			/*
146 			 * orphan the threads so we don't mess up
147 			 * when they call exit
148 			 */
149 			nq->p_peers = 0;
150 			nq->p_leader = nq;
151 		}
152 
153 	/* otherwise are we a peer? */
154 	} else if(p->p_peers) {
155 		q = p->p_leader;
156 		while(q->p_peers != p)
157 			q = q->p_peers;
158 		q->p_peers = p->p_peers;
159 	}
160 
161 #ifdef PGINPROF
162 	vmsizmon();
163 #endif
164 	/*
165 	 * Check if any LKMs need anything done at process exit.
166 	 * e.g. SYSV IPC stuff
167 	 * XXX what if one of these generates an error?
168 	 */
169 	while (ep) {
170 		(*ep->function)(p);
171 		ep = ep->next;
172 	}
173 
174 	if (p->p_flag & P_PROFIL)
175 		stopprofclock(p);
176 	MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage),
177 		M_ZOMBIE, M_WAITOK);
178 	/*
179 	 * If parent is waiting for us to exit or exec,
180 	 * P_PPWAIT is set; we will wakeup the parent below.
181 	 */
182 	p->p_flag &= ~(P_TRACED | P_PPWAIT);
183 	p->p_flag |= P_WEXIT;
184 	p->p_sigignore = ~0;
185 	p->p_siglist = 0;
186 	untimeout(realitexpire, (caddr_t)p);
187 
188 	/*
189 	 * Close open files and release open-file table.
190 	 * This may block!
191 	 */
192 	fdfree(p);
193 
194 	/*
195 	 * XXX Shutdown SYSV semaphores
196 	 */
197 	semexit(p);
198 
199 	/* The next two chunks should probably be moved to vmspace_exit. */
200 	vm = p->p_vmspace;
201 	/*
202 	 * Release user portion of address space.
203 	 * This releases references to vnodes,
204 	 * which could cause I/O if the file has been unlinked.
205 	 * Need to do this early enough that we can still sleep.
206 	 * Can't free the entire vmspace as the kernel stack
207 	 * may be mapped within that space also.
208 	 */
209 	if (vm->vm_refcnt == 1) {
210 		if (vm->vm_shm)
211 			shmexit(p);
212 		pmap_remove_pages(&vm->vm_pmap, VM_MIN_ADDRESS,
213 		    VM_MAXUSER_ADDRESS);
214 		(void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS,
215 		    VM_MAXUSER_ADDRESS);
216 	}
217 
218 	if (SESS_LEADER(p)) {
219 		register struct session *sp = p->p_session;
220 
221 		if (sp->s_ttyvp) {
222 			/*
223 			 * Controlling process.
224 			 * Signal foreground pgrp,
225 			 * drain controlling terminal
226 			 * and revoke access to controlling terminal.
227 			 */
228 			if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
229 				if (sp->s_ttyp->t_pgrp)
230 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
231 				(void) ttywait(sp->s_ttyp);
232 				/*
233 				 * The tty could have been revoked
234 				 * if we blocked.
235 				 */
236 				if (sp->s_ttyvp)
237 					VOP_REVOKE(sp->s_ttyvp, REVOKEALL);
238 			}
239 			if (sp->s_ttyvp)
240 				vrele(sp->s_ttyvp);
241 			sp->s_ttyvp = NULL;
242 			/*
243 			 * s_ttyp is not zero'd; we use this to indicate
244 			 * that the session once had a controlling terminal.
245 			 * (for logging and informational purposes)
246 			 */
247 		}
248 		sp->s_leader = NULL;
249 	}
250 	fixjobc(p, p->p_pgrp, 0);
251 	if (p->p_limit->p_refcnt > 1 &&
252 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
253 		p->p_limit->p_refcnt--;
254 		p->p_limit = limcopy(p->p_limit);
255 	}
256 	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
257 	(void)acct_process(p);
258 #ifdef KTRACE
259 	/*
260 	 * release trace file
261 	 */
262 	p->p_traceflag = 0;	/* don't trace the vrele() */
263 	if (p->p_tracep)
264 		vrele(p->p_tracep);
265 #endif
266 	/*
267 	 * Remove proc from allproc queue and pidhash chain.
268 	 * Place onto zombproc.  Unlink from parent's child list.
269 	 */
270 	LIST_REMOVE(p, p_list);
271 	LIST_INSERT_HEAD(&zombproc, p, p_list);
272 	p->p_stat = SZOMB;
273 
274 	LIST_REMOVE(p, p_hash);
275 
276 	q = p->p_children.lh_first;
277 	if (q)		/* only need this if any child is S_ZOMB */
278 		wakeup((caddr_t) initproc);
279 	for (; q != 0; q = nq) {
280 		nq = q->p_sibling.le_next;
281 		LIST_REMOVE(q, p_sibling);
282 		LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling);
283 		q->p_pptr = initproc;
284 		/*
285 		 * Traced processes are killed
286 		 * since their existence means someone is screwing up.
287 		 */
288 		if (q->p_flag & P_TRACED) {
289 			q->p_flag &= ~P_TRACED;
290 			psignal(q, SIGKILL);
291 		}
292 	}
293 
294 	/*
295 	 * Save exit status and final rusage info, adding in child rusage
296 	 * info and self times.
297 	 */
298 	p->p_xstat = rv;
299 	*p->p_ru = p->p_stats->p_ru;
300 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
301 	ruadd(p->p_ru, &p->p_stats->p_cru);
302 
303 	/*
304 	 * Notify parent that we're gone.
305 	 */
306 	psignal(p->p_pptr, SIGCHLD);
307 	wakeup((caddr_t)p->p_pptr);
308 #if defined(tahoe)
309 	/* move this to cpu_exit */
310 	p->p_addr->u_pcb.pcb_savacc.faddr = (float *)NULL;
311 #endif
312 	/*
313 	 * Clear curproc after we've done all operations
314 	 * that could block, and before tearing down the rest
315 	 * of the process state that might be used from clock, etc.
316 	 * Also, can't clear curproc while we're still runnable,
317 	 * as we're not on a run queue (we are current, just not
318 	 * a proper proc any longer!).
319 	 *
320 	 * Other substructures are freed from wait().
321 	 */
322 	curproc = NULL;
323 	if (--p->p_limit->p_refcnt == 0) {
324 		FREE(p->p_limit, M_SUBPROC);
325 		p->p_limit = NULL;
326 	}
327 
328 	/*
329 	 * Finally, call machine-dependent code to release the remaining
330 	 * resources including address space, the kernel stack and pcb.
331 	 * The address space is released by "vmspace_free(p->p_vmspace)";
332 	 * This is machine-dependent, as we may have to change stacks
333 	 * or ensure that the current one isn't reallocated before we
334 	 * finish.  cpu_exit will end with a call to cpu_switch(), finishing
335 	 * our execution (pun intended).
336 	 */
337 	cpu_exit(p);
338 }
339 
340 #ifdef COMPAT_43
341 #if defined(hp300) || defined(luna68k)
342 #include <machine/frame.h>
343 #define GETPS(rp)	((struct frame *)(rp))->f_sr
344 #else
345 #define GETPS(rp)	(rp)[PS]
346 #endif
347 
348 int
349 owait(p, uap, retval)
350 	struct proc *p;
351 	register struct owait_args /* {
352 		int     dummy;
353 	} */ *uap;
354 	int *retval;
355 {
356 	struct wait_args w;
357 
358 #ifdef PSL_ALLCC
359 	if ((GETPS(p->p_md.md_regs) & PSL_ALLCC) != PSL_ALLCC) {
360 		w.options = 0;
361 		w.rusage = NULL;
362 	} else {
363 		w.options = p->p_md.md_regs[R0];
364 		w.rusage = (struct rusage *)p->p_md.md_regs[R1];
365 	}
366 #else
367 	w.options = 0;
368 	w.rusage = NULL;
369 #endif
370 	w.pid = WAIT_ANY;
371 	w.status = NULL;
372 	return (wait1(p, &w, retval, 1));
373 }
374 #endif /* COMPAT_43 */
375 
376 int
377 wait4(p, uap, retval)
378 	struct proc *p;
379 	struct wait_args *uap;
380 	int *retval;
381 {
382 
383 	return (wait1(p, uap, retval, 0));
384 }
385 
386 static int
387 wait1(q, uap, retval, compat)
388 	register struct proc *q;
389 	register struct wait_args /* {
390 		int pid;
391 		int *status;
392 		int options;
393 		struct rusage *rusage;
394 	} */ *uap;
395 	int retval[];
396 	int compat;
397 {
398 	register int nfound;
399 	register struct proc *p, *t;
400 	int status, error;
401 
402 	if (uap->pid == 0)
403 		uap->pid = -q->p_pgid;
404 #ifdef notyet
405 	if (uap->options &~ (WUNTRACED|WNOHANG))
406 		return (EINVAL);
407 #endif
408 loop:
409 	nfound = 0;
410 	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
411 		if (uap->pid != WAIT_ANY &&
412 		    p->p_pid != uap->pid && p->p_pgid != -uap->pid)
413 			continue;
414 		nfound++;
415 		if (p->p_stat == SZOMB) {
416 			/* charge childs scheduling cpu usage to parent */
417 			if (curproc->p_pid != 1) {
418 				curproc->p_estcpu = min(curproc->p_estcpu +
419 				    p->p_estcpu, UCHAR_MAX);
420 			}
421 
422 			retval[0] = p->p_pid;
423 #ifdef COMPAT_43
424 			if (compat)
425 				retval[1] = p->p_xstat;
426 			else
427 #endif
428 			if (uap->status) {
429 				status = p->p_xstat;	/* convert to int */
430 				if ((error = copyout((caddr_t)&status,
431 				    (caddr_t)uap->status, sizeof(status))))
432 					return (error);
433 			}
434 			if (uap->rusage && (error = copyout((caddr_t)p->p_ru,
435 			    (caddr_t)uap->rusage, sizeof (struct rusage))))
436 				return (error);
437 			/*
438 			 * If we got the child via a ptrace 'attach',
439 			 * we need to give it back to the old parent.
440 			 */
441 			if (p->p_oppid && (t = pfind(p->p_oppid))) {
442 				p->p_oppid = 0;
443 				proc_reparent(p, t);
444 				psignal(t, SIGCHLD);
445 				wakeup((caddr_t)t);
446 				return (0);
447 			}
448 			p->p_xstat = 0;
449 			ruadd(&q->p_stats->p_cru, p->p_ru);
450 			FREE(p->p_ru, M_ZOMBIE);
451 			p->p_ru = NULL;
452 
453 			/*
454 			 * Decrement the count of procs running with this uid.
455 			 */
456 			(void)chgproccnt(p->p_cred->p_ruid, -1);
457 
458 			/*
459 			 * Release reference to text vnode
460 			 */
461 			if (p->p_textvp)
462 				vrele(p->p_textvp);
463 
464 			/*
465 			 * Free up credentials.
466 			 */
467 			if (--p->p_cred->p_refcnt == 0) {
468 				crfree(p->p_cred->pc_ucred);
469 				FREE(p->p_cred, M_SUBPROC);
470 				p->p_cred = NULL;
471 			}
472 
473 			/*
474 			 * Finally finished with old proc entry.
475 			 * Unlink it from its process group and free it.
476 			 */
477 			leavepgrp(p);
478 			LIST_REMOVE(p, p_list);	/* off zombproc */
479 			LIST_REMOVE(p, p_sibling);
480 
481 			/*
482 			 * Give machine-dependent layer a chance
483 			 * to free anything that cpu_exit couldn't
484 			 * release while still running in process context.
485 			 */
486 			cpu_wait(p);
487 			FREE(p, M_PROC);
488 			nprocs--;
489 			return (0);
490 		}
491 		if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
492 		    (p->p_flag & P_TRACED || uap->options & WUNTRACED)) {
493 			p->p_flag |= P_WAITED;
494 			retval[0] = p->p_pid;
495 #ifdef COMPAT_43
496 			if (compat) {
497 				retval[1] = W_STOPCODE(p->p_xstat);
498 				error = 0;
499 			} else
500 #endif
501 			if (uap->status) {
502 				status = W_STOPCODE(p->p_xstat);
503 				error = copyout((caddr_t)&status,
504 					(caddr_t)uap->status, sizeof(status));
505 			} else
506 				error = 0;
507 			return (error);
508 		}
509 	}
510 	if (nfound == 0)
511 		return (ECHILD);
512 	if (uap->options & WNOHANG) {
513 		retval[0] = 0;
514 		return (0);
515 	}
516 	if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)))
517 		return (error);
518 	goto loop;
519 }
520 
521 /*
522  * make process 'parent' the new parent of process 'child'.
523  */
524 void
525 proc_reparent(child, parent)
526 	register struct proc *child;
527 	register struct proc *parent;
528 {
529 
530 	if (child->p_pptr == parent)
531 		return;
532 
533 	LIST_REMOVE(child, p_sibling);
534 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
535 	child->p_pptr = parent;
536 }
537 
538 /*
539  * The next two functions are to handle adding/deleting items on the
540  * exit callout list
541  *
542  * at_exit():
543  * Take the arguments given and put them onto the exit callout list,
544  * However first make sure that it's not already there.
545  * returns 0 on success.
546  */
547 int
548 at_exit(exitlist_fn function)
549 {
550 	ele_p ep;
551 
552 	/* Be noisy if the programmer has lost track of things */
553 	if (rm_at_exit(function))
554 		printf("exit callout entry already present\n");
555 	ep = malloc(sizeof(*ep), M_TEMP, M_NOWAIT);
556 	if (ep == NULL)
557 		return (ENOMEM);
558 	ep->next = exit_list;
559 	ep->function = function;
560 	exit_list = ep;
561 	return (0);
562 }
563 /*
564  * Scan the exit callout list for the given items and remove them.
565  * Returns the number of items removed.
566  * Logically this can only be 0 or 1.
567  */
568 int
569 rm_at_exit(exitlist_fn function)
570 {
571 	ele_p *epp, ep;
572 	int count;
573 
574 	count = 0;
575 	epp = &exit_list;
576 	ep = *epp;
577 	while (ep) {
578 		if (ep->function == function) {
579 			*epp = ep->next;
580 			free(ep, M_TEMP);
581 			count++;
582 		} else {
583 			epp = &ep->next;
584 		}
585 		ep = *epp;
586 	}
587 	return (count);
588 }
589 
590 
591