xref: /openbsd-src/sys/kern/kern_fork.c (revision 3388b1b4ab64cffc1d1bf306ad1f48c006aa17ed)
1 /*	$NetBSD: kern_fork.c,v 1.27 1995/12/10 08:26:02 mycroft Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *	This product includes software developed by the University of
23  *	California, Berkeley and its contributors.
24  * 4. Neither the name of the University nor the names of its contributors
25  *    may be used to endorse or promote products derived from this software
26  *    without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38  * SUCH DAMAGE.
39  *
40  *	@(#)kern_fork.c	8.6 (Berkeley) 4/8/94
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/map.h>
46 #include <sys/filedesc.h>
47 #include <sys/kernel.h>
48 #include <sys/malloc.h>
49 #include <sys/proc.h>
50 #include <sys/resourcevar.h>
51 #include <sys/vnode.h>
52 #include <sys/file.h>
53 #include <sys/acct.h>
54 #include <sys/ktrace.h>
55 
56 #include <vm/vm.h>
57 
58 #include <sys/mount.h>
59 #include <sys/syscallargs.h>
60 
61 int	nprocs = 1;		/* process 0 */
62 
63 #define	ISFORK	0
64 #define	ISVFORK	1
65 #define	ISRFORK	2
66 
67 int
68 sys_fork(p, v, retval)
69 	struct proc *p;
70 	void *v;
71 	register_t *retval;
72 {
73 
74 	return (fork1(p, ISFORK, 0, retval));
75 }
76 
77 int
78 sys_vfork(p, v, retval)
79 	struct proc *p;
80 	void *v;
81 	register_t *retval;
82 {
83 
84 	return (fork1(p, ISVFORK, 0, retval));
85 }
86 
87 int
88 sys_rfork(p, v, retval)
89 	struct proc *p;
90 	void *v;
91 	register_t *retval;
92 {
93 	struct sys_rfork_args /* {
94 		syscallarg(int) flags;
95 	} */ *uap = v;
96 
97 	return (fork1(p, ISRFORK, SCARG(uap, flags), retval));
98 }
99 
100 int
101 fork1(p1, forktype, rforkflags, retval)
102 	register struct proc *p1;
103 	int forktype;
104 	int rforkflags;
105 	register_t *retval;
106 {
107 	register struct proc *p2;
108 	register uid_t uid;
109 	struct proc *newproc;
110 	struct proc **hash;
111 	int count;
112 	static int nextpid, pidchecked = 0;
113 	int dupfd = 1, cleanfd = 0;
114 
115 	if (forktype == ISRFORK) {
116 		dupfd = 0;
117 		if ((rforkflags & RFPROC) == 0)
118 			return (EINVAL);
119 		if ((rforkflags & (RFFDG|RFCFDG)) == (RFFDG|RFCFDG))
120 			return (EINVAL);
121 		if (rforkflags & RFFDG)
122 			dupfd = 1;
123 		if (rforkflags & RFNOWAIT)
124 			return (EINVAL);	/* XXX unimplimented */
125 		if (rforkflags & RFCFDG)
126 			cleanfd = 1;
127 	}
128 
129 	/*
130 	 * Although process entries are dynamically created, we still keep
131 	 * a global limit on the maximum number we will create.  Don't allow
132 	 * a nonprivileged user to use the last process; don't let root
133 	 * exceed the limit. The variable nprocs is the current number of
134 	 * processes, maxproc is the limit.
135 	 */
136 	uid = p1->p_cred->p_ruid;
137 	if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) {
138 		tablefull("proc");
139 		return (EAGAIN);
140 	}
141 
142 	/*
143 	 * Increment the count of procs running with this uid. Don't allow
144 	 * a nonprivileged user to exceed their current limit.
145 	 */
146 	count = chgproccnt(uid, 1);
147 	if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) {
148 		(void)chgproccnt(uid, -1);
149 		return (EAGAIN);
150 	}
151 
152 	/* Allocate new proc. */
153 	MALLOC(newproc, struct proc *, sizeof(struct proc), M_PROC, M_WAITOK);
154 
155 	/*
156 	 * Find an unused process ID.  We remember a range of unused IDs
157 	 * ready to use (from nextpid+1 through pidchecked-1).
158 	 */
159 	nextpid++;
160 retry:
161 	/*
162 	 * If the process ID prototype has wrapped around,
163 	 * restart somewhat above 0, as the low-numbered procs
164 	 * tend to include daemons that don't exit.
165 	 */
166 	if (nextpid >= PID_MAX) {
167 		nextpid = 100;
168 		pidchecked = 0;
169 	}
170 	if (nextpid >= pidchecked) {
171 		int doingzomb = 0;
172 
173 		pidchecked = PID_MAX;
174 		/*
175 		 * Scan the active and zombie procs to check whether this pid
176 		 * is in use.  Remember the lowest pid that's greater
177 		 * than nextpid, so we can avoid checking for a while.
178 		 */
179 		p2 = allproc.lh_first;
180 again:
181 		for (; p2 != 0; p2 = p2->p_list.le_next) {
182 			while (p2->p_pid == nextpid ||
183 			    p2->p_pgrp->pg_id == nextpid) {
184 				nextpid++;
185 				if (nextpid >= pidchecked)
186 					goto retry;
187 			}
188 			if (p2->p_pid > nextpid && pidchecked > p2->p_pid)
189 				pidchecked = p2->p_pid;
190 			if (p2->p_pgrp->pg_id > nextpid &&
191 			    pidchecked > p2->p_pgrp->pg_id)
192 				pidchecked = p2->p_pgrp->pg_id;
193 		}
194 		if (!doingzomb) {
195 			doingzomb = 1;
196 			p2 = zombproc.lh_first;
197 			goto again;
198 		}
199 	}
200 
201 	nprocs++;
202 	p2 = newproc;
203 	p2->p_stat = SIDL;			/* protect against others */
204 	p2->p_pid = nextpid;
205 	LIST_INSERT_HEAD(&allproc, p2, p_list);
206 	p2->p_forw = p2->p_back = NULL;		/* shouldn't be necessary */
207 	LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
208 
209 	/*
210 	 * Make a proc table entry for the new process.
211 	 * Start by zeroing the section of proc that is zero-initialized,
212 	 * then copy the section that is copied directly from the parent.
213 	 */
214 	bzero(&p2->p_startzero,
215 	    (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
216 	bcopy(&p1->p_startcopy, &p2->p_startcopy,
217 	    (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
218 
219 	/*
220 	 * Duplicate sub-structures as needed.
221 	 * Increase reference counts on shared objects.
222 	 * The p_stats and p_sigacts substructs are set in vm_fork.
223 	 */
224 	p2->p_flag = P_INMEM;
225 	p2->p_emul = p1->p_emul;
226 	if (p1->p_flag & P_PROFIL)
227 		startprofclock(p2);
228 	MALLOC(p2->p_cred, struct pcred *, sizeof(struct pcred),
229 	    M_SUBPROC, M_WAITOK);
230 	bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred));
231 	p2->p_cred->p_refcnt = 1;
232 	crhold(p1->p_ucred);
233 
234 	/* bump references to the text vnode (for procfs) */
235 	p2->p_textvp = p1->p_textvp;
236 	if (p2->p_textvp)
237 		VREF(p2->p_textvp);
238 
239 	if (cleanfd)
240 		p2->p_fd = fdinit(p1);
241 	else if (dupfd)
242 		p2->p_fd = fdcopy(p1);
243 	else
244 		p2->p_fd = fdshare(p1);
245 
246 	/*
247 	 * If p_limit is still copy-on-write, bump refcnt,
248 	 * otherwise get a copy that won't be modified.
249 	 * (If PL_SHAREMOD is clear, the structure is shared
250 	 * copy-on-write.)
251 	 */
252 	if (p1->p_limit->p_lflags & PL_SHAREMOD)
253 		p2->p_limit = limcopy(p1->p_limit);
254 	else {
255 		p2->p_limit = p1->p_limit;
256 		p2->p_limit->p_refcnt++;
257 	}
258 
259 	if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
260 		p2->p_flag |= P_CONTROLT;
261 	if (forktype == ISVFORK)
262 		p2->p_flag |= P_PPWAIT;
263 	LIST_INSERT_AFTER(p1, p2, p_pglist);
264 	p2->p_pptr = p1;
265 	if (rforkflags & RFNOWAIT) {
266 		/* XXX should we do anything? */
267 	} else {
268 		LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling);
269 	}
270 	LIST_INIT(&p2->p_children);
271 
272 #ifdef KTRACE
273 	/*
274 	 * Copy traceflag and tracefile if enabled.
275 	 * If not inherited, these were zeroed above.
276 	 */
277 	if (p1->p_traceflag&KTRFAC_INHERIT) {
278 		p2->p_traceflag = p1->p_traceflag;
279 		if ((p2->p_tracep = p1->p_tracep) != NULL)
280 			VREF(p2->p_tracep);
281 	}
282 #endif
283 
284 	/*
285 	 * This begins the section where we must prevent the parent
286 	 * from being swapped.
287 	 */
288 	p1->p_holdcnt++;
289 
290 	if (forktype == ISRFORK && (rforkflags & RFMEM)) {
291 		/* share as much address space as possible */
292 		(void) vm_map_inherit(&p1->p_vmspace->vm_map,
293 		    VM_MIN_ADDRESS, VM_MAXUSER_ADDRESS - MAXSSIZ,
294 		    VM_INHERIT_SHARE);
295 	}
296 
297 #ifdef __FORK_BRAINDAMAGE
298 	/*
299 	 * Set return values for child before vm_fork,
300 	 * so they can be copied to child stack.
301 	 * We return 0, rather than the traditional behaviour of modifying the
302 	 * return value in the system call stub.
303 	 * NOTE: the kernel stack may be at a different location in the child
304 	 * process, and thus addresses of automatic variables (including retval)
305 	 * may be invalid after vm_fork returns in the child process.
306 	 */
307 	retval[0] = 0;
308 	retval[1] = 1;
309 	if (vm_fork(p1, p2))
310 		return (0);
311 #else
312 	/*
313 	 * Finish creating the child process.  It will return through a
314 	 * different path later.
315 	 */
316 	vm_fork(p1, p2);
317 #endif
318 
319 	/*
320 	 * Make child runnable, set start time, and add to run queue.
321 	 */
322 	(void) splstatclock();
323 	p2->p_stats->p_start = time;
324 	p2->p_acflag = AFORK;
325 	p2->p_stat = SRUN;
326 	setrunqueue(p2);
327 	(void) spl0();
328 
329 	/*
330 	 * Now can be swapped.
331 	 */
332 	p1->p_holdcnt--;
333 
334 	/*
335 	 * Preserve synchronization semantics of vfork.  If waiting for
336 	 * child to exec or exit, set P_PPWAIT on child, and sleep on our
337 	 * proc (in case of exit).
338 	 */
339 	if (forktype == ISVFORK)
340 		while (p2->p_flag & P_PPWAIT)
341 			tsleep(p1, PWAIT, "ppwait", 0);
342 
343 	/*
344 	 * Return child pid to parent process,
345 	 * marking us as parent via retval[1].
346 	 */
347 	retval[0] = p2->p_pid;
348 	retval[1] = 0;
349 	return (0);
350 }
351