xref: /freebsd/sys/kern/kern_fork.c (revision 5ebc7e6281887681c3a348a5a4c902e262ccd656)
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 1.10 1995/01/21 15:08:57 bde 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 /* ARGSUSED */
59 int
60 fork(p, uap, retval)
61 	struct proc *p;
62 	struct fork_args *uap;
63 	int retval[];
64 {
65 
66 	return (fork1(p, 0, retval));
67 }
68 
69 /* ARGSUSED */
70 int
71 vfork(p, uap, retval)
72 	struct proc *p;
73 	struct fork_args *uap;
74 	int retval[];
75 {
76 
77 	return (fork1(p, 1, retval));
78 }
79 
80 int	nprocs = 1;		/* process 0 */
81 
82 static int
83 fork1(p1, isvfork, retval)
84 	register struct proc *p1;
85 	int isvfork, retval[];
86 {
87 	register struct proc *p2;
88 	register uid_t uid;
89 	struct proc *newproc;
90 	struct proc **hash;
91 	int count;
92 	static int nextpid, pidchecked = 0;
93 
94 	/*
95 	 * Although process entries are dynamically created, we still keep
96 	 * a global limit on the maximum number we will create.  Don't allow
97 	 * a nonprivileged user to use the last process; don't let root
98 	 * exceed the limit. The variable nprocs is the current number of
99 	 * processes, maxproc is the limit.
100 	 */
101 	uid = p1->p_cred->p_ruid;
102 	if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) {
103 		tablefull("proc");
104 		return (EAGAIN);
105 	}
106 	/*
107 	 * Increment the count of procs running with this uid. Don't allow
108 	 * a nonprivileged user to exceed their current limit.
109 	 */
110 	count = chgproccnt(uid, 1);
111 	if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) {
112 		(void)chgproccnt(uid, -1);
113 		return (EAGAIN);
114 	}
115 
116 	/* Allocate new proc. */
117 	MALLOC(newproc, struct proc *, sizeof(struct proc), M_PROC, M_WAITOK);
118 
119 	/*
120 	 * Find an unused process ID.  We remember a range of unused IDs
121 	 * ready to use (from nextpid+1 through pidchecked-1).
122 	 */
123 	nextpid++;
124 retry:
125 	/*
126 	 * If the process ID prototype has wrapped around,
127 	 * restart somewhat above 0, as the low-numbered procs
128 	 * tend to include daemons that don't exit.
129 	 */
130 	if (nextpid >= PID_MAX) {
131 		nextpid = 100;
132 		pidchecked = 0;
133 	}
134 	if (nextpid >= pidchecked) {
135 		int doingzomb = 0;
136 
137 		pidchecked = PID_MAX;
138 		/*
139 		 * Scan the active and zombie procs to check whether this pid
140 		 * is in use.  Remember the lowest pid that's greater
141 		 * than nextpid, so we can avoid checking for a while.
142 		 */
143 		p2 = (struct proc *)allproc;
144 again:
145 		for (; p2 != NULL; p2 = p2->p_next) {
146 			while (p2->p_pid == nextpid ||
147 			    p2->p_pgrp->pg_id == nextpid) {
148 				nextpid++;
149 				if (nextpid >= pidchecked)
150 					goto retry;
151 			}
152 			if (p2->p_pid > nextpid && pidchecked > p2->p_pid)
153 				pidchecked = p2->p_pid;
154 			if (p2->p_pgrp->pg_id > nextpid &&
155 			    pidchecked > p2->p_pgrp->pg_id)
156 				pidchecked = p2->p_pgrp->pg_id;
157 		}
158 		if (!doingzomb) {
159 			doingzomb = 1;
160 			p2 = zombproc;
161 			goto again;
162 		}
163 	}
164 
165 
166 	/*
167 	 * Link onto allproc (this should probably be delayed).
168 	 * Heavy use of volatile here to prevent the compiler from
169 	 * rearranging code.  Yes, it *is* terribly ugly, but at least
170 	 * it works.
171 	 */
172 	nprocs++;
173 	p2 = newproc;
174 #define	Vp2 ((volatile struct proc *)p2)
175 	Vp2->p_stat = SIDL;			/* protect against others */
176 	Vp2->p_pid = nextpid;
177 	/*
178 	 * This is really:
179 	 *	p2->p_next = allproc;
180 	 *	allproc->p_prev = &p2->p_next;
181 	 *	p2->p_prev = &allproc;
182 	 *	allproc = p2;
183 	 * The assignment via allproc is legal since it is never NULL.
184 	 */
185 	*(volatile struct proc **)&Vp2->p_next = allproc;
186 	*(volatile struct proc ***)&allproc->p_prev =
187 	    (volatile struct proc **)&Vp2->p_next;
188 	*(volatile struct proc ***)&Vp2->p_prev = &allproc;
189 	allproc = Vp2;
190 #undef Vp2
191 	p2->p_forw = p2->p_back = NULL;		/* shouldn't be necessary */
192 
193 	/* Insert on the hash chain. */
194 	hash = &pidhash[PIDHASH(p2->p_pid)];
195 	p2->p_hash = *hash;
196 	*hash = p2;
197 
198 	/*
199 	 * Make a proc table entry for the new process.
200 	 * Start by zeroing the section of proc that is zero-initialized,
201 	 * then copy the section that is copied directly from the parent.
202 	 */
203 	bzero(&p2->p_startzero,
204 	    (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
205 	bcopy(&p1->p_startcopy, &p2->p_startcopy,
206 	    (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
207 
208 	/*
209 	 * Duplicate sub-structures as needed.
210 	 * Increase reference counts on shared objects.
211 	 * The p_stats and p_sigacts substructs are set in vm_fork.
212 	 */
213 	p2->p_flag = P_INMEM;
214 	if (p1->p_flag & P_PROFIL)
215 		startprofclock(p2);
216 	MALLOC(p2->p_cred, struct pcred *, sizeof(struct pcred),
217 	    M_SUBPROC, M_WAITOK);
218 	bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred));
219 	p2->p_cred->p_refcnt = 1;
220 	crhold(p1->p_ucred);
221 
222 	/* bump references to the text vnode (for procfs) */
223 	p2->p_textvp = p1->p_textvp;
224 	if (p2->p_textvp)
225 		VREF(p2->p_textvp);
226 
227 	p2->p_fd = fdcopy(p1);
228 	/*
229 	 * If p_limit is still copy-on-write, bump refcnt,
230 	 * otherwise get a copy that won't be modified.
231 	 * (If PL_SHAREMOD is clear, the structure is shared
232 	 * copy-on-write.)
233 	 */
234 	if (p1->p_limit->p_lflags & PL_SHAREMOD)
235 		p2->p_limit = limcopy(p1->p_limit);
236 	else {
237 		p2->p_limit = p1->p_limit;
238 		p2->p_limit->p_refcnt++;
239 	}
240 
241 	if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
242 		p2->p_flag |= P_CONTROLT;
243 	if (isvfork)
244 		p2->p_flag |= P_PPWAIT;
245 	p2->p_pgrpnxt = p1->p_pgrpnxt;
246 	p1->p_pgrpnxt = p2;
247 	p2->p_pptr = p1;
248 	p2->p_osptr = p1->p_cptr;
249 	if (p1->p_cptr)
250 		p1->p_cptr->p_ysptr = p2;
251 	p1->p_cptr = p2;
252 #ifdef KTRACE
253 	/*
254 	 * Copy traceflag and tracefile if enabled.
255 	 * If not inherited, these were zeroed above.
256 	 */
257 	if (p1->p_traceflag&KTRFAC_INHERIT) {
258 		p2->p_traceflag = p1->p_traceflag;
259 		if ((p2->p_tracep = p1->p_tracep) != NULL)
260 			VREF(p2->p_tracep);
261 	}
262 #endif
263 
264 	/*
265 	 * set priority of child to be that of parent
266 	 */
267 	p2->p_estcpu = p1->p_estcpu;
268 
269 	/*
270 	 * This begins the section where we must prevent the parent
271 	 * from being swapped.
272 	 */
273 	p1->p_flag |= P_NOSWAP;
274 
275 	/*
276 	 * Set return values for child before vm_fork,
277 	 * so they can be copied to child stack.
278 	 * We return parent pid, and mark as child in retval[1].
279 	 * NOTE: the kernel stack may be at a different location in the child
280 	 * process, and thus addresses of automatic variables (including retval)
281 	 * may be invalid after vm_fork returns in the child process.
282 	 */
283 	retval[0] = p1->p_pid;
284 	retval[1] = 1;
285 	if (vm_fork(p1, p2, isvfork)) {
286 		/*
287 		 * Child process.  Set start time and get to work.
288 		 */
289 		microtime(&runtime);
290 		p2->p_stats->p_start = runtime;
291 		p2->p_acflag = AFORK;
292 		return (0);
293 	}
294 
295 	/*
296 	 * Make child runnable and add to run queue.
297 	 */
298 	(void) splhigh();
299 	p2->p_stat = SRUN;
300 	setrunqueue(p2);
301 	(void) spl0();
302 
303 	/*
304 	 * Now can be swapped.
305 	 */
306 	p1->p_flag &= ~P_NOSWAP;
307 
308 	/*
309 	 * Preserve synchronization semantics of vfork.  If waiting for
310 	 * child to exec or exit, set P_PPWAIT on child, and sleep on our
311 	 * proc (in case of exit).
312 	 */
313 	if (isvfork)
314 		while (p2->p_flag & P_PPWAIT)
315 			tsleep(p1, PWAIT, "ppwait", 0);
316 
317 	/*
318 	 * Return child pid to parent process,
319 	 * marking us as parent via retval[1].
320 	 */
321 	retval[0] = p2->p_pid;
322 	retval[1] = 0;
323 	return (0);
324 }
325