Home | History | Annotate | Line # | Download | only in kern
kern_fork.c revision 1.32.6.1
      1 /*	$NetBSD: kern_fork.c,v 1.32.6.1 1997/09/08 23:09:38 thorpej 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/mount.h>
     50 #include <sys/proc.h>
     51 #include <sys/resourcevar.h>
     52 #include <sys/vnode.h>
     53 #include <sys/file.h>
     54 #include <sys/acct.h>
     55 #include <sys/ktrace.h>
     56 #include <sys/signalvar.h>
     57 
     58 #include <sys/syscallargs.h>
     59 
     60 #include <vm/vm.h>
     61 
     62 int	nprocs = 1;		/* process 0 */
     63 
     64 int fork1 __P((struct proc *, int, register_t *));
     65 
     66 /*ARGSUSED*/
     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, 0, retval));
     75 }
     76 
     77 /*ARGSUSED*/
     78 int
     79 sys_vfork(p, v, retval)
     80 	struct proc *p;
     81 	void *v;
     82 	register_t *retval;
     83 {
     84 
     85 	return (fork1(p, 1, retval));
     86 }
     87 
     88 int
     89 fork1(p1, isvfork, retval)
     90 	register struct proc *p1;
     91 	int isvfork;
     92 	register_t *retval;
     93 {
     94 	register struct proc *p2;
     95 	register uid_t uid;
     96 	struct proc *newproc;
     97 	int count;
     98 	static int nextpid, pidchecked = 0;
     99 
    100 	/*
    101 	 * Although process entries are dynamically created, we still keep
    102 	 * a global limit on the maximum number we will create.  Don't allow
    103 	 * a nonprivileged user to use the last process; don't let root
    104 	 * exceed the limit. The variable nprocs is the current number of
    105 	 * processes, maxproc is the limit.
    106 	 */
    107 	uid = p1->p_cred->p_ruid;
    108 	if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) {
    109 		tablefull("proc");
    110 		return (EAGAIN);
    111 	}
    112 
    113 	/*
    114 	 * Increment the count of procs running with this uid. Don't allow
    115 	 * a nonprivileged user to exceed their current limit.
    116 	 */
    117 	count = chgproccnt(uid, 1);
    118 	if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) {
    119 		(void)chgproccnt(uid, -1);
    120 		return (EAGAIN);
    121 	}
    122 
    123 	/* Allocate new proc. */
    124 	MALLOC(newproc, struct proc *, sizeof(struct proc), M_PROC, M_WAITOK);
    125 
    126 	/*
    127 	 * Find an unused process ID.  We remember a range of unused IDs
    128 	 * ready to use (from nextpid+1 through pidchecked-1).
    129 	 */
    130 	nextpid++;
    131 retry:
    132 	/*
    133 	 * If the process ID prototype has wrapped around,
    134 	 * restart somewhat above 0, as the low-numbered procs
    135 	 * tend to include daemons that don't exit.
    136 	 */
    137 	if (nextpid >= PID_MAX) {
    138 		nextpid = 100;
    139 		pidchecked = 0;
    140 	}
    141 	if (nextpid >= pidchecked) {
    142 		int doingzomb = 0;
    143 
    144 		pidchecked = PID_MAX;
    145 		/*
    146 		 * Scan the active and zombie procs to check whether this pid
    147 		 * is in use.  Remember the lowest pid that's greater
    148 		 * than nextpid, so we can avoid checking for a while.
    149 		 */
    150 		p2 = allproc.lh_first;
    151 again:
    152 		for (; p2 != 0; p2 = p2->p_list.le_next) {
    153 			while (p2->p_pid == nextpid ||
    154 			    p2->p_pgrp->pg_id == nextpid) {
    155 				nextpid++;
    156 				if (nextpid >= pidchecked)
    157 					goto retry;
    158 			}
    159 			if (p2->p_pid > nextpid && pidchecked > p2->p_pid)
    160 				pidchecked = p2->p_pid;
    161 			if (p2->p_pgrp->pg_id > nextpid &&
    162 			    pidchecked > p2->p_pgrp->pg_id)
    163 				pidchecked = p2->p_pgrp->pg_id;
    164 		}
    165 		if (!doingzomb) {
    166 			doingzomb = 1;
    167 			p2 = zombproc.lh_first;
    168 			goto again;
    169 		}
    170 	}
    171 
    172 	nprocs++;
    173 	p2 = newproc;
    174 	p2->p_stat = SIDL;			/* protect against others */
    175 	p2->p_pid = nextpid;
    176 	LIST_INSERT_HEAD(&allproc, p2, p_list);
    177 	p2->p_forw = p2->p_back = NULL;		/* shouldn't be necessary */
    178 	LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
    179 
    180 	/*
    181 	 * Make a proc table entry for the new process.
    182 	 * Start by zeroing the section of proc that is zero-initialized,
    183 	 * then copy the section that is copied directly from the parent.
    184 	 */
    185 	bzero(&p2->p_startzero,
    186 	    (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
    187 	bcopy(&p1->p_startcopy, &p2->p_startcopy,
    188 	    (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
    189 
    190 	/*
    191 	 * Duplicate sub-structures as needed.
    192 	 * Increase reference counts on shared objects.
    193 	 * The p_stats and p_sigacts substructs are set in vm_fork.
    194 	 */
    195 	p2->p_flag = P_INMEM | (p1->p_flag & P_SUGID);
    196 	p2->p_emul = p1->p_emul;
    197 	if (p1->p_flag & P_PROFIL)
    198 		startprofclock(p2);
    199 	MALLOC(p2->p_cred, struct pcred *, sizeof(struct pcred),
    200 	    M_SUBPROC, M_WAITOK);
    201 	bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred));
    202 	p2->p_cred->p_refcnt = 1;
    203 	crhold(p1->p_ucred);
    204 
    205 	/* bump references to the text vnode (for procfs) */
    206 	p2->p_textvp = p1->p_textvp;
    207 	if (p2->p_textvp)
    208 		VREF(p2->p_textvp);
    209 
    210 	p2->p_fd = fdcopy(p1);
    211 	/*
    212 	 * If p_limit is still copy-on-write, bump refcnt,
    213 	 * otherwise get a copy that won't be modified.
    214 	 * (If PL_SHAREMOD is clear, the structure is shared
    215 	 * copy-on-write.)
    216 	 */
    217 	if (p1->p_limit->p_lflags & PL_SHAREMOD)
    218 		p2->p_limit = limcopy(p1->p_limit);
    219 	else {
    220 		p2->p_limit = p1->p_limit;
    221 		p2->p_limit->p_refcnt++;
    222 	}
    223 
    224 	if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
    225 		p2->p_flag |= P_CONTROLT;
    226 	if (isvfork)
    227 		p2->p_flag |= P_PPWAIT;
    228 	LIST_INSERT_AFTER(p1, p2, p_pglist);
    229 	p2->p_pptr = p1;
    230 	LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling);
    231 	LIST_INIT(&p2->p_children);
    232 
    233 #ifdef KTRACE
    234 	/*
    235 	 * Copy traceflag and tracefile if enabled.
    236 	 * If not inherited, these were zeroed above.
    237 	 */
    238 	if (p1->p_traceflag&KTRFAC_INHERIT) {
    239 		p2->p_traceflag = p1->p_traceflag;
    240 		if ((p2->p_tracep = p1->p_tracep) != NULL)
    241 			VREF(p2->p_tracep);
    242 	}
    243 #endif
    244 
    245 	/*
    246 	 * Set up signal actions for the new process.
    247 	 */
    248 	p2->p_sigacts = sigacts_copy(p1);
    249 
    250 	/*
    251 	 * This begins the section where we must prevent the parent
    252 	 * from being swapped.
    253 	 */
    254 	PHOLD(p1);
    255 
    256 	/*
    257 	 * Finish creating the child process.  It will return through a
    258 	 * different path later.
    259 	 */
    260 	vm_fork(p1, p2);
    261 
    262 	/*
    263 	 * Make child runnable, set start time, and add to run queue.
    264 	 */
    265 	(void) splstatclock();
    266 	p2->p_stats->p_start = time;
    267 	p2->p_acflag = AFORK;
    268 	p2->p_stat = SRUN;
    269 	setrunqueue(p2);
    270 	(void) spl0();
    271 
    272 	/*
    273 	 * Now can be swapped.
    274 	 */
    275 	PRELE(p1);
    276 
    277 	/*
    278 	 * Preserve synchronization semantics of vfork.  If waiting for
    279 	 * child to exec or exit, set P_PPWAIT on child, and sleep on our
    280 	 * proc (in case of exit).
    281 	 */
    282 	if (isvfork)
    283 		while (p2->p_flag & P_PPWAIT)
    284 			tsleep(p1, PWAIT, "ppwait", 0);
    285 
    286 	/*
    287 	 * Return child pid to parent process,
    288 	 * marking us as parent via retval[1].
    289 	 */
    290 	retval[0] = p2->p_pid;
    291 	retval[1] = 0;
    292 	return (0);
    293 }
    294