xref: /freebsd/sys/kern/kern_fork.c (revision 8e6b01171e30297084bb0b4457c4183c2746aacc)
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_fork.c	8.6 (Berkeley) 4/8/94
39  * $Id: kern_fork.c,v 1.12 1995/05/30 08:05:27 rgrimes Exp $
40  */
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/filedesc.h>
45 #include <sys/kernel.h>
46 #include <sys/malloc.h>
47 #include <sys/proc.h>
48 #include <sys/resourcevar.h>
49 #include <sys/vnode.h>
50 #include <sys/file.h>
51 #include <sys/acct.h>
52 #include <sys/ktrace.h>
53 
54 #include <vm/vm.h>
55 
56 static int fork1(struct proc *, int, int *);
57 
58 struct fork_args {
59         int     dummy;
60 };
61 
62 /* ARGSUSED */
63 int
64 fork(p, uap, retval)
65 	struct proc *p;
66 	struct fork_args *uap;
67 	int retval[];
68 {
69 
70 	return (fork1(p, 0, retval));
71 }
72 
73 /* ARGSUSED */
74 int
75 vfork(p, uap, retval)
76 	struct proc *p;
77 	struct fork_args *uap;
78 	int retval[];
79 {
80 
81 	return (fork1(p, 1, retval));
82 }
83 
84 int	nprocs = 1;		/* process 0 */
85 
86 static int
87 fork1(p1, isvfork, retval)
88 	register struct proc *p1;
89 	int isvfork, retval[];
90 {
91 	register struct proc *p2;
92 	register uid_t uid;
93 	struct proc *newproc;
94 	struct proc **hash;
95 	int count;
96 	static int nextpid, pidchecked = 0;
97 
98 	/*
99 	 * Although process entries are dynamically created, we still keep
100 	 * a global limit on the maximum number we will create.  Don't allow
101 	 * a nonprivileged user to use the last process; don't let root
102 	 * exceed the limit. The variable nprocs is the current number of
103 	 * processes, maxproc is the limit.
104 	 */
105 	uid = p1->p_cred->p_ruid;
106 	if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) {
107 		tablefull("proc");
108 		return (EAGAIN);
109 	}
110 	/*
111 	 * Increment the count of procs running with this uid. Don't allow
112 	 * a nonprivileged user to exceed their current limit.
113 	 */
114 	count = chgproccnt(uid, 1);
115 	if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) {
116 		(void)chgproccnt(uid, -1);
117 		return (EAGAIN);
118 	}
119 
120 	/* Allocate new proc. */
121 	MALLOC(newproc, struct proc *, sizeof(struct proc), M_PROC, M_WAITOK);
122 
123 	/*
124 	 * Find an unused process ID.  We remember a range of unused IDs
125 	 * ready to use (from nextpid+1 through pidchecked-1).
126 	 */
127 	nextpid++;
128 retry:
129 	/*
130 	 * If the process ID prototype has wrapped around,
131 	 * restart somewhat above 0, as the low-numbered procs
132 	 * tend to include daemons that don't exit.
133 	 */
134 	if (nextpid >= PID_MAX) {
135 		nextpid = 100;
136 		pidchecked = 0;
137 	}
138 	if (nextpid >= pidchecked) {
139 		int doingzomb = 0;
140 
141 		pidchecked = PID_MAX;
142 		/*
143 		 * Scan the active and zombie procs to check whether this pid
144 		 * is in use.  Remember the lowest pid that's greater
145 		 * than nextpid, so we can avoid checking for a while.
146 		 */
147 		p2 = (struct proc *)allproc;
148 again:
149 		for (; p2 != NULL; p2 = p2->p_next) {
150 			while (p2->p_pid == nextpid ||
151 			    p2->p_pgrp->pg_id == nextpid) {
152 				nextpid++;
153 				if (nextpid >= pidchecked)
154 					goto retry;
155 			}
156 			if (p2->p_pid > nextpid && pidchecked > p2->p_pid)
157 				pidchecked = p2->p_pid;
158 			if (p2->p_pgrp->pg_id > nextpid &&
159 			    pidchecked > p2->p_pgrp->pg_id)
160 				pidchecked = p2->p_pgrp->pg_id;
161 		}
162 		if (!doingzomb) {
163 			doingzomb = 1;
164 			p2 = zombproc;
165 			goto again;
166 		}
167 	}
168 
169 
170 	/*
171 	 * Link onto allproc (this should probably be delayed).
172 	 * Heavy use of volatile here to prevent the compiler from
173 	 * rearranging code.  Yes, it *is* terribly ugly, but at least
174 	 * it works.
175 	 */
176 	nprocs++;
177 	p2 = newproc;
178 #define	Vp2 ((volatile struct proc *)p2)
179 	Vp2->p_stat = SIDL;			/* protect against others */
180 	Vp2->p_pid = nextpid;
181 	/*
182 	 * This is really:
183 	 *	p2->p_next = allproc;
184 	 *	allproc->p_prev = &p2->p_next;
185 	 *	p2->p_prev = &allproc;
186 	 *	allproc = p2;
187 	 * The assignment via allproc is legal since it is never NULL.
188 	 */
189 	*(volatile struct proc **)&Vp2->p_next = allproc;
190 	*(volatile struct proc ***)&allproc->p_prev =
191 	    (volatile struct proc **)&Vp2->p_next;
192 	*(volatile struct proc ***)&Vp2->p_prev = &allproc;
193 	allproc = Vp2;
194 #undef Vp2
195 	p2->p_forw = p2->p_back = NULL;		/* shouldn't be necessary */
196 
197 	/* Insert on the hash chain. */
198 	hash = &pidhash[PIDHASH(p2->p_pid)];
199 	p2->p_hash = *hash;
200 	*hash = p2;
201 
202 	/*
203 	 * Make a proc table entry for the new process.
204 	 * Start by zeroing the section of proc that is zero-initialized,
205 	 * then copy the section that is copied directly from the parent.
206 	 */
207 	bzero(&p2->p_startzero,
208 	    (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
209 	bcopy(&p1->p_startcopy, &p2->p_startcopy,
210 	    (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
211 
212 	/*
213 	 * Duplicate sub-structures as needed.
214 	 * Increase reference counts on shared objects.
215 	 * The p_stats and p_sigacts substructs are set in vm_fork.
216 	 */
217 	p2->p_flag = P_INMEM;
218 	if (p1->p_flag & P_PROFIL)
219 		startprofclock(p2);
220 	MALLOC(p2->p_cred, struct pcred *, sizeof(struct pcred),
221 	    M_SUBPROC, M_WAITOK);
222 	bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred));
223 	p2->p_cred->p_refcnt = 1;
224 	crhold(p1->p_ucred);
225 
226 	/* bump references to the text vnode (for procfs) */
227 	p2->p_textvp = p1->p_textvp;
228 	if (p2->p_textvp)
229 		VREF(p2->p_textvp);
230 
231 	p2->p_fd = fdcopy(p1);
232 	/*
233 	 * If p_limit is still copy-on-write, bump refcnt,
234 	 * otherwise get a copy that won't be modified.
235 	 * (If PL_SHAREMOD is clear, the structure is shared
236 	 * copy-on-write.)
237 	 */
238 	if (p1->p_limit->p_lflags & PL_SHAREMOD)
239 		p2->p_limit = limcopy(p1->p_limit);
240 	else {
241 		p2->p_limit = p1->p_limit;
242 		p2->p_limit->p_refcnt++;
243 	}
244 
245 	if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
246 		p2->p_flag |= P_CONTROLT;
247 	if (isvfork)
248 		p2->p_flag |= P_PPWAIT;
249 	p2->p_pgrpnxt = p1->p_pgrpnxt;
250 	p1->p_pgrpnxt = p2;
251 	p2->p_pptr = p1;
252 	p2->p_osptr = p1->p_cptr;
253 	if (p1->p_cptr)
254 		p1->p_cptr->p_ysptr = p2;
255 	p1->p_cptr = p2;
256 #ifdef KTRACE
257 	/*
258 	 * Copy traceflag and tracefile if enabled.
259 	 * If not inherited, these were zeroed above.
260 	 */
261 	if (p1->p_traceflag&KTRFAC_INHERIT) {
262 		p2->p_traceflag = p1->p_traceflag;
263 		if ((p2->p_tracep = p1->p_tracep) != NULL)
264 			VREF(p2->p_tracep);
265 	}
266 #endif
267 
268 	/*
269 	 * set priority of child to be that of parent
270 	 */
271 	p2->p_estcpu = p1->p_estcpu;
272 
273 	/*
274 	 * This begins the section where we must prevent the parent
275 	 * from being swapped.
276 	 */
277 	p1->p_flag |= P_NOSWAP;
278 
279 	/*
280 	 * Set return values for child before vm_fork,
281 	 * so they can be copied to child stack.
282 	 * We return parent pid, and mark as child in retval[1].
283 	 * NOTE: the kernel stack may be at a different location in the child
284 	 * process, and thus addresses of automatic variables (including retval)
285 	 * may be invalid after vm_fork returns in the child process.
286 	 */
287 	retval[0] = p1->p_pid;
288 	retval[1] = 1;
289 	if (vm_fork(p1, p2, isvfork)) {
290 		/*
291 		 * Child process.  Set start time and get to work.
292 		 */
293 		microtime(&runtime);
294 		p2->p_stats->p_start = runtime;
295 		p2->p_acflag = AFORK;
296 		return (0);
297 	}
298 
299 	/*
300 	 * Make child runnable and add to run queue.
301 	 */
302 	(void) splhigh();
303 	p2->p_stat = SRUN;
304 	setrunqueue(p2);
305 	(void) spl0();
306 
307 	/*
308 	 * Now can be swapped.
309 	 */
310 	p1->p_flag &= ~P_NOSWAP;
311 
312 	/*
313 	 * Preserve synchronization semantics of vfork.  If waiting for
314 	 * child to exec or exit, set P_PPWAIT on child, and sleep on our
315 	 * proc (in case of exit).
316 	 */
317 	if (isvfork)
318 		while (p2->p_flag & P_PPWAIT)
319 			tsleep(p1, PWAIT, "ppwait", 0);
320 
321 	/*
322 	 * Return child pid to parent process,
323 	 * marking us as parent via retval[1].
324 	 */
325 	retval[0] = p2->p_pid;
326 	retval[1] = 0;
327 	return (0);
328 }
329