xref: /openbsd-src/sys/kern/kern_exit.c (revision 62a742911104f98b9185b2c6b6007d9b1c36396c)
1 /*	$OpenBSD: kern_exit.c,v 1.17 1999/03/12 17:49:37 deraadt Exp $	*/
2 /*	$NetBSD: kern_exit.c,v 1.39 1996/04/22 01:38:25 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1989, 1991, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  * (c) UNIX System Laboratories, Inc.
8  * All or some portions of this file are derived from material licensed
9  * to the University of California by American Telephone and Telegraph
10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11  * the permission of UNIX System Laboratories, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. All advertising materials mentioning features or use of this software
22  *    must display the following acknowledgement:
23  *	This product includes software developed by the University of
24  *	California, Berkeley and its contributors.
25  * 4. Neither the name of the University nor the names of its contributors
26  *    may be used to endorse or promote products derived from this software
27  *    without specific prior written permission.
28  *
29  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39  * SUCH DAMAGE.
40  *
41  *	@(#)kern_exit.c	8.7 (Berkeley) 2/12/94
42  */
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/map.h>
47 #include <sys/ioctl.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/proc.h>
54 #include <sys/buf.h>
55 #include <sys/wait.h>
56 #include <sys/file.h>
57 #include <sys/vnode.h>
58 #include <sys/syslog.h>
59 #include <sys/malloc.h>
60 #include <sys/resourcevar.h>
61 #include <sys/ptrace.h>
62 #include <sys/acct.h>
63 #include <sys/filedesc.h>
64 #include <sys/signalvar.h>
65 #ifdef SYSVSHM
66 #include <sys/shm.h>
67 #endif
68 #ifdef SYSVSEM
69 #include <sys/sem.h>
70 #endif
71 
72 #include <sys/mount.h>
73 #include <sys/syscallargs.h>
74 
75 #include <machine/cpu.h>
76 
77 #include <vm/vm.h>
78 #include <vm/vm_kern.h>
79 
80 #if defined(UVM)
81 #include <uvm/uvm_extern.h>
82 #endif
83 
84 /*
85  * exit --
86  *	Death of process.
87  */
88 int
89 sys_exit(p, v, retval)
90 	struct proc *p;
91 	void *v;
92 	register_t *retval;
93 {
94 	struct sys_exit_args /* {
95 		syscallarg(int) rval;
96 	} */ *uap = v;
97 
98 	exit1(p, W_EXITCODE(SCARG(uap, rval), 0));
99 	/* NOTREACHED */
100 	return (0);
101 }
102 
103 /*
104  * Exit: deallocate address space and other resources, change proc state
105  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
106  * status and rusage for wait().  Check for child processes and orphan them.
107  */
108 void
109 exit1(p, rv)
110 	register struct proc *p;
111 	int rv;
112 {
113 	register struct proc *q, *nq;
114 	register struct vmspace *vm;
115 
116 	if (p->p_pid == 1)
117 		panic("init died (signal %d, exit %d)",
118 		    WTERMSIG(rv), WEXITSTATUS(rv));
119 #ifdef PGINPROF
120 	vmsizmon();
121 #endif
122 	if (p->p_flag & P_PROFIL)
123 		stopprofclock(p);
124 	MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage),
125 		M_ZOMBIE, M_WAITOK);
126 	/*
127 	 * If parent is waiting for us to exit or exec, P_PPWAIT is set; we
128 	 * wake up the parent early to avoid deadlock.
129 	 */
130 	p->p_flag |= P_WEXIT;
131 	p->p_flag &= ~P_TRACED;
132 	if (p->p_flag & P_PPWAIT) {
133 		p->p_flag &= ~P_PPWAIT;
134 		wakeup((caddr_t)p->p_pptr);
135 	}
136 	p->p_sigignore = ~0;
137 	p->p_siglist = 0;
138 	untimeout(realitexpire, (caddr_t)p);
139 
140 	/*
141 	 * Close open files and release open-file table.
142 	 * This may block!
143 	 */
144 	fdfree(p);
145 
146 	/* The next three chunks should probably be moved to vmspace_exit. */
147 	vm = p->p_vmspace;
148 #ifdef SYSVSHM
149 	if (vm->vm_shm)
150 		shmexit(p);
151 #endif
152 #ifdef SYSVSEM
153 	semexit(p);
154 #endif
155 	/*
156 	 * Release user portion of address space.
157 	 * This releases references to vnodes,
158 	 * which could cause I/O if the file has been unlinked.
159 	 * Need to do this early enough that we can still sleep.
160 	 * Can't free the entire vmspace as the kernel stack
161 	 * may be mapped within that space also.
162 	 */
163 #if defined(UVM)
164 	if (vm->vm_refcnt == 1)
165 		(void) uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
166 		    VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
167 #else
168 	if (vm->vm_refcnt == 1)
169 		(void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS,
170 		    VM_MAXUSER_ADDRESS);
171 #endif
172 
173 	if (SESS_LEADER(p)) {
174 		register struct session *sp = p->p_session;
175 
176 		if (sp->s_ttyvp) {
177 			/*
178 			 * Controlling process.
179 			 * Signal foreground pgrp,
180 			 * drain controlling terminal
181 			 * and revoke access to controlling terminal.
182 			 */
183 			if (sp->s_ttyp->t_session == sp) {
184 				if (sp->s_ttyp->t_pgrp)
185 					pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
186 				(void) ttywait(sp->s_ttyp);
187 				/*
188 				 * The tty could have been revoked
189 				 * if we blocked.
190 				 */
191 				if (sp->s_ttyvp)
192 					VOP_REVOKE(sp->s_ttyvp, REVOKEALL);
193 			}
194 			if (sp->s_ttyvp)
195 				vrele(sp->s_ttyvp);
196 			sp->s_ttyvp = NULL;
197 			/*
198 			 * s_ttyp is not zero'd; we use this to indicate
199 			 * that the session once had a controlling terminal.
200 			 * (for logging and informational purposes)
201 			 */
202 		}
203 		sp->s_leader = NULL;
204 	}
205 	fixjobc(p, p->p_pgrp, 0);
206 	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
207 	(void)acct_process(p);
208 #ifdef KTRACE
209 	/*
210 	 * release trace file
211 	 */
212 	p->p_traceflag = 0;	/* don't trace the vrele() */
213 	if (p->p_tracep)
214 		vrele(p->p_tracep);
215 #endif
216 	/*
217 	 * Remove proc from allproc queue and pidhash chain.
218 	 * Place onto zombproc.  Unlink from parent's child list.
219 	 */
220 	LIST_REMOVE(p, p_list);
221 	LIST_INSERT_HEAD(&zombproc, p, p_list);
222 	p->p_stat = SZOMB;
223 
224 	LIST_REMOVE(p, p_hash);
225 
226 	q = p->p_children.lh_first;
227 	if (q)		/* only need this if any child is S_ZOMB */
228 		wakeup((caddr_t)initproc);
229 	for (; q != 0; q = nq) {
230 		nq = q->p_sibling.le_next;
231 		proc_reparent(q, initproc);
232 		/*
233 		 * Traced processes are killed
234 		 * since their existence means someone is screwing up.
235 		 */
236 		if (q->p_flag & P_TRACED) {
237 			q->p_flag &= ~P_TRACED;
238 			psignal(q, SIGKILL);
239 		}
240 	}
241 
242 	/*
243 	 * Save exit status and final rusage info, adding in child rusage
244 	 * info and self times.
245 	 */
246 	p->p_xstat = rv;
247 	*p->p_ru = p->p_stats->p_ru;
248 	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
249 	ruadd(p->p_ru, &p->p_stats->p_cru);
250 
251 	/*
252 	 * clear %cpu usage during swap
253 	 */
254 	p->p_pctcpu = 0;
255 
256 	/*
257 	 * Notify parent that we're gone.  If we have P_NOZOMBIE or parent has
258 	 * the P_NOCLDWAIT flag set, notify process 1 instead (and hope it
259 	 * will handle this situation).
260 	 */
261 	if ((p->p_flag & P_NOZOMBIE) || (p->p_pptr->p_flag & P_NOCLDWAIT)) {
262 		struct proc *pp = p->p_pptr;
263 		proc_reparent(p, initproc);
264 		/*
265 		 * If this was the last child of our parent, notify
266 		 * parent, so in case he was wait(2)ing, he will
267 		 * continue.
268 		 */
269 		if (pp->p_children.lh_first == NULL)
270 			wakeup((caddr_t)pp);
271 	}
272 	psignal(p->p_pptr, SIGCHLD);
273 	wakeup((caddr_t)p->p_pptr);
274 
275 	/*
276 	 * Notify procfs debugger
277 	 */
278 	if (p->p_flag & P_FSTRACE)
279 		wakeup((caddr_t)p);
280 #if defined(tahoe)
281 	/* move this to cpu_exit */
282 	p->p_addr->u_pcb.pcb_savacc.faddr = (float *)NULL;
283 #endif
284 	/*
285 	 * Clear curproc after we've done all operations
286 	 * that could block, and before tearing down the rest
287 	 * of the process state that might be used from clock, etc.
288 	 * Also, can't clear curproc while we're still runnable,
289 	 * as we're not on a run queue (we are current, just not
290 	 * a proper proc any longer!).
291 	 *
292 	 * Other substructures are freed from wait().
293 	 */
294 	curproc = NULL;
295 	if (--p->p_limit->p_refcnt == 0)
296 		FREE(p->p_limit, M_SUBPROC);
297 
298 	/*
299 	 * Finally, call machine-dependent code to release the remaining
300 	 * resources including address space, the kernel stack and pcb.
301 	 * The address space is released by "vmspace_free(p->p_vmspace)";
302 	 * This is machine-dependent, as we may have to change stacks
303 	 * or ensure that the current one isn't reallocated before we
304 	 * finish.  cpu_exit will end with a call to cpu_swtch(), finishing
305 	 * our execution (pun intended).
306 	 */
307 	cpu_exit(p);
308 }
309 
310 int
311 sys_wait4(q, v, retval)
312 	register struct proc *q;
313 	void *v;
314 	register_t *retval;
315 {
316 	register struct sys_wait4_args /* {
317 		syscallarg(int) pid;
318 		syscallarg(int *) status;
319 		syscallarg(int) options;
320 		syscallarg(struct rusage *) rusage;
321 	} */ *uap = v;
322 	register int nfound;
323 	register struct proc *p, *t;
324 	int status, error;
325 
326 #ifdef COMPAT_09
327 	SCARG(uap, pid) = (short)SCARG(uap, pid);
328 #endif
329 
330 	if (SCARG(uap, pid) == 0)
331 		SCARG(uap, pid) = -q->p_pgid;
332 	if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG))
333 		return (EINVAL);
334 
335 loop:
336 	nfound = 0;
337 	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
338 		if ((p->p_flag & P_NOZOMBIE) ||
339 		    (SCARG(uap, pid) != WAIT_ANY &&
340 		    p->p_pid != SCARG(uap, pid) &&
341 		    p->p_pgid != -SCARG(uap, pid)))
342 			continue;
343 		nfound++;
344 		if (p->p_stat == SZOMB) {
345 			retval[0] = p->p_pid;
346 
347 			if (SCARG(uap, status)) {
348 				status = p->p_xstat;	/* convert to int */
349 				error = copyout((caddr_t)&status,
350 						(caddr_t)SCARG(uap, status),
351 						sizeof(status));
352 				if (error)
353 					return (error);
354 			}
355 			if (SCARG(uap, rusage) &&
356 			    (error = copyout((caddr_t)p->p_ru,
357 			    (caddr_t)SCARG(uap, rusage),
358 			    sizeof (struct rusage))))
359 				return (error);
360 			/*
361 			 * If we got the child via a ptrace 'attach',
362 			 * we need to give it back to the old parent.
363 			 */
364 			if (p->p_oppid && (t = pfind(p->p_oppid))) {
365 				p->p_oppid = 0;
366 				proc_reparent(p, t);
367 				psignal(t, SIGCHLD);
368 				wakeup((caddr_t)t);
369 				return (0);
370 			}
371 
372 			/* Charge us for our child's sins */
373 			curproc->p_estcpu = min(curproc->p_estcpu +
374 			    p->p_estcpu, UCHAR_MAX);
375 			p->p_xstat = 0;
376 			ruadd(&q->p_stats->p_cru, p->p_ru);
377 			FREE(p->p_ru, M_ZOMBIE);
378 
379 			/*
380 			 * Finally finished with old proc entry.
381 			 * Unlink it from its process group and free it.
382 			 */
383 			leavepgrp(p);
384 			LIST_REMOVE(p, p_list);	/* off zombproc */
385 			LIST_REMOVE(p, p_sibling);
386 
387 			/*
388 			 * Decrement the count of procs running with this uid.
389 			 */
390 			(void)chgproccnt(p->p_cred->p_ruid, -1);
391 
392 			/*
393 			 * Free up credentials.
394 			 */
395 			if (--p->p_cred->p_refcnt == 0) {
396 				crfree(p->p_cred->pc_ucred);
397 				FREE(p->p_cred, M_SUBPROC);
398 			}
399 
400 			/*
401 			 * Release reference to text vnode
402 			 */
403 			if (p->p_textvp)
404 				vrele(p->p_textvp);
405 
406 			/*
407 			 * Give machine-dependent layer a chance
408 			 * to free anything that cpu_exit couldn't
409 			 * release while still running in process context.
410 			 */
411 			cpu_wait(p);
412 			FREE(p, M_PROC);
413 			nprocs--;
414 			return (0);
415 		}
416 		if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
417 		    (p->p_flag & P_TRACED || SCARG(uap, options) & WUNTRACED)) {
418 			p->p_flag |= P_WAITED;
419 			retval[0] = p->p_pid;
420 
421 			if (SCARG(uap, status)) {
422 				status = W_STOPCODE(p->p_xstat);
423 				error = copyout((caddr_t)&status,
424 				    (caddr_t)SCARG(uap, status),
425 				    sizeof(status));
426 			} else
427 				error = 0;
428 			return (error);
429 		}
430 	}
431 	if (nfound == 0)
432 		return (ECHILD);
433 	if (SCARG(uap, options) & WNOHANG) {
434 		retval[0] = 0;
435 		return (0);
436 	}
437 	if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)) != 0)
438 		return (error);
439 	goto loop;
440 }
441 
442 /*
443  * make process 'parent' the new parent of process 'child'.
444  */
445 void
446 proc_reparent(child, parent)
447 	register struct proc *child;
448 	register struct proc *parent;
449 {
450 
451 	if (child->p_pptr == parent)
452 		return;
453 
454 	LIST_REMOVE(child, p_sibling);
455 	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
456 	child->p_pptr = parent;
457 }
458