Home | History | Annotate | Line # | Download | only in kern
kern_fork.c revision 1.39
      1 /*	$NetBSD: kern_fork.c,v 1.39 1998/02/14 00:37:31 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 "opt_uvm.h"
     44 
     45 #include <sys/param.h>
     46 #include <sys/systm.h>
     47 #include <sys/map.h>
     48 #include <sys/filedesc.h>
     49 #include <sys/kernel.h>
     50 #include <sys/malloc.h>
     51 #include <sys/mount.h>
     52 #include <sys/proc.h>
     53 #include <sys/resourcevar.h>
     54 #include <sys/vnode.h>
     55 #include <sys/file.h>
     56 #include <sys/acct.h>
     57 #include <sys/ktrace.h>
     58 #include <sys/vmmeter.h>
     59 
     60 #include <sys/syscallargs.h>
     61 
     62 #include <vm/vm.h>
     63 
     64 #if defined(UVM)
     65 #include <uvm/uvm_extern.h>
     66 #endif
     67 
     68 int	nprocs = 1;		/* process 0 */
     69 
     70 /*ARGSUSED*/
     71 int
     72 sys_fork(p, v, retval)
     73 	struct proc *p;
     74 	void *v;
     75 	register_t *retval;
     76 {
     77 
     78 	return (fork1(p, 0, retval, NULL));
     79 }
     80 
     81 /*
     82  * vfork(2) system call compatible with 4.4BSD (i.e. BSD with Mach VM).
     83  * Address space is not shared, but parent is blocked until child exit.
     84  */
     85 /*ARGSUSED*/
     86 int
     87 sys_vfork(p, v, retval)
     88 	struct proc *p;
     89 	void *v;
     90 	register_t *retval;
     91 {
     92 
     93 	return (fork1(p, FORK_PPWAIT, retval, NULL));
     94 }
     95 
     96 /*
     97  * New vfork(2) system call for NetBSD, which implements original 3BSD vfork(2)
     98  * semantics.  Address space is shared, and parent is blocked until child exit.
     99  */
    100 /*ARGSUSED*/
    101 int
    102 sys___vfork14(p, v, retval)
    103 	struct proc *p;
    104 	void *v;
    105 	register_t *retval;
    106 {
    107 
    108 	return (fork1(p, FORK_PPWAIT|FORK_SHAREVM, retval, NULL));
    109 }
    110 
    111 int
    112 fork1(p1, flags, retval, rnewprocp)
    113 	register struct proc *p1;
    114 	int flags;
    115 	register_t *retval;
    116 	struct proc **rnewprocp;
    117 {
    118 	register struct proc *p2;
    119 	register uid_t uid;
    120 	struct proc *newproc;
    121 	int count;
    122 	static int nextpid, pidchecked = 0;
    123 
    124 	/*
    125 	 * Although process entries are dynamically created, we still keep
    126 	 * a global limit on the maximum number we will create.  Don't allow
    127 	 * a nonprivileged user to use the last process; don't let root
    128 	 * exceed the limit. The variable nprocs is the current number of
    129 	 * processes, maxproc is the limit.
    130 	 */
    131 	uid = p1->p_cred->p_ruid;
    132 	if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) {
    133 		tablefull("proc");
    134 		return (EAGAIN);
    135 	}
    136 
    137 	/*
    138 	 * Increment the count of procs running with this uid. Don't allow
    139 	 * a nonprivileged user to exceed their current limit.
    140 	 */
    141 	count = chgproccnt(uid, 1);
    142 	if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) {
    143 		(void)chgproccnt(uid, -1);
    144 		return (EAGAIN);
    145 	}
    146 
    147 	/* Allocate new proc. */
    148 	MALLOC(newproc, struct proc *, sizeof(struct proc), M_PROC, M_WAITOK);
    149 
    150 	/*
    151 	 * Find an unused process ID.  We remember a range of unused IDs
    152 	 * ready to use (from nextpid+1 through pidchecked-1).
    153 	 */
    154 	nextpid++;
    155 retry:
    156 	/*
    157 	 * If the process ID prototype has wrapped around,
    158 	 * restart somewhat above 0, as the low-numbered procs
    159 	 * tend to include daemons that don't exit.
    160 	 */
    161 	if (nextpid >= PID_MAX) {
    162 		nextpid = 100;
    163 		pidchecked = 0;
    164 	}
    165 	if (nextpid >= pidchecked) {
    166 		int doingzomb = 0;
    167 
    168 		pidchecked = PID_MAX;
    169 		/*
    170 		 * Scan the active and zombie procs to check whether this pid
    171 		 * is in use.  Remember the lowest pid that's greater
    172 		 * than nextpid, so we can avoid checking for a while.
    173 		 */
    174 		p2 = allproc.lh_first;
    175 again:
    176 		for (; p2 != 0; p2 = p2->p_list.le_next) {
    177 			while (p2->p_pid == nextpid ||
    178 			    p2->p_pgrp->pg_id == nextpid ||
    179 			    p2->p_session->s_sid == nextpid) {
    180 				nextpid++;
    181 				if (nextpid >= pidchecked)
    182 					goto retry;
    183 			}
    184 			if (p2->p_pid > nextpid && pidchecked > p2->p_pid)
    185 				pidchecked = p2->p_pid;
    186 
    187 			if (p2->p_pgrp->pg_id > nextpid &&
    188 			    pidchecked > p2->p_pgrp->pg_id)
    189 				pidchecked = p2->p_pgrp->pg_id;
    190 
    191 			if (p2->p_session->s_sid > nextpid &&
    192 			    pidchecked > p2->p_session->s_sid)
    193 				pidchecked = p2->p_session->s_sid;
    194 		}
    195 		if (!doingzomb) {
    196 			doingzomb = 1;
    197 			p2 = zombproc.lh_first;
    198 			goto again;
    199 		}
    200 	}
    201 
    202 	nprocs++;
    203 	p2 = newproc;
    204 	p2->p_stat = SIDL;			/* protect against others */
    205 	p2->p_pid = nextpid;
    206 	LIST_INSERT_HEAD(&allproc, p2, p_list);
    207 	p2->p_forw = p2->p_back = NULL;		/* shouldn't be necessary */
    208 	LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
    209 
    210 	/*
    211 	 * Make a proc table entry for the new process.
    212 	 * Start by zeroing the section of proc that is zero-initialized,
    213 	 * then copy the section that is copied directly from the parent.
    214 	 */
    215 	bzero(&p2->p_startzero,
    216 	    (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
    217 	bcopy(&p1->p_startcopy, &p2->p_startcopy,
    218 	    (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
    219 
    220 	/*
    221 	 * Duplicate sub-structures as needed.
    222 	 * Increase reference counts on shared objects.
    223 	 * The p_stats and p_sigacts substructs are set in vm_fork.
    224 	 */
    225 	p2->p_flag = P_INMEM | (p1->p_flag & P_SUGID);
    226 	p2->p_emul = p1->p_emul;
    227 	if (p1->p_flag & P_PROFIL)
    228 		startprofclock(p2);
    229 	MALLOC(p2->p_cred, struct pcred *, sizeof(struct pcred),
    230 	    M_SUBPROC, M_WAITOK);
    231 	bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred));
    232 	p2->p_cred->p_refcnt = 1;
    233 	crhold(p1->p_ucred);
    234 
    235 	/* bump references to the text vnode (for procfs) */
    236 	p2->p_textvp = p1->p_textvp;
    237 	if (p2->p_textvp)
    238 		VREF(p2->p_textvp);
    239 
    240 	p2->p_fd = fdcopy(p1);
    241 	/*
    242 	 * If p_limit is still copy-on-write, bump refcnt,
    243 	 * otherwise get a copy that won't be modified.
    244 	 * (If PL_SHAREMOD is clear, the structure is shared
    245 	 * copy-on-write.)
    246 	 */
    247 	if (p1->p_limit->p_lflags & PL_SHAREMOD)
    248 		p2->p_limit = limcopy(p1->p_limit);
    249 	else {
    250 		p2->p_limit = p1->p_limit;
    251 		p2->p_limit->p_refcnt++;
    252 	}
    253 
    254 	if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
    255 		p2->p_flag |= P_CONTROLT;
    256 	if (flags & FORK_PPWAIT)
    257 		p2->p_flag |= P_PPWAIT;
    258 	LIST_INSERT_AFTER(p1, p2, p_pglist);
    259 	p2->p_pptr = p1;
    260 	LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling);
    261 	LIST_INIT(&p2->p_children);
    262 
    263 #ifdef KTRACE
    264 	/*
    265 	 * Copy traceflag and tracefile if enabled.
    266 	 * If not inherited, these were zeroed above.
    267 	 */
    268 	if (p1->p_traceflag&KTRFAC_INHERIT) {
    269 		p2->p_traceflag = p1->p_traceflag;
    270 		if ((p2->p_tracep = p1->p_tracep) != NULL)
    271 			VREF(p2->p_tracep);
    272 	}
    273 #endif
    274 
    275 	/*
    276 	 * This begins the section where we must prevent the parent
    277 	 * from being swapped.
    278 	 */
    279 	PHOLD(p1);
    280 
    281 	/*
    282 	 * Finish creating the child process.  It will return through a
    283 	 * different path later.
    284 	 */
    285 #if defined(UVM)
    286 	uvm_fork(p1, p2, (flags & FORK_SHAREVM) ? TRUE : FALSE);
    287 #else
    288 	vm_fork(p1, p2, (flags & FORK_SHAREVM) ? TRUE : FALSE);
    289 #endif
    290 
    291 	/*
    292 	 * Make child runnable, set start time, and add to run queue.
    293 	 */
    294 	(void) splstatclock();
    295 	p2->p_stats->p_start = time;
    296 	p2->p_acflag = AFORK;
    297 	p2->p_stat = SRUN;
    298 	setrunqueue(p2);
    299 	(void) spl0();
    300 
    301 	/*
    302 	 * Now can be swapped.
    303 	 */
    304 	PRELE(p1);
    305 
    306 	/*
    307 	 * Update stats now that we know the fork was successful.
    308 	 */
    309 #if defined(UVM)
    310 	uvmexp.forks++;
    311 	if (flags & FORK_PPWAIT)
    312 		uvmexp.forks_ppwait++;
    313 	if (flags & FORK_SHAREVM)
    314 		uvmexp.forks_sharevm++;
    315 #else
    316 	cnt.v_forks++;
    317 	if (flags & FORK_PPWAIT)
    318 		cnt.v_forks_ppwait++;
    319 	if (flags & FORK_SHAREVM)
    320 		cnt.v_forks_sharevm++;
    321 #endif
    322 
    323 	/*
    324 	 * Pass a pointer to the new process to the caller.
    325 	 */
    326 	if (rnewprocp != NULL)
    327 		*rnewprocp = p2;
    328 
    329 	/*
    330 	 * Preserve synchronization semantics of vfork.  If waiting for
    331 	 * child to exec or exit, set P_PPWAIT on child, and sleep on our
    332 	 * proc (in case of exit).
    333 	 */
    334 	if (flags & FORK_PPWAIT)
    335 		while (p2->p_flag & P_PPWAIT)
    336 			tsleep(p1, PWAIT, "ppwait", 0);
    337 
    338 	/*
    339 	 * Return child pid to parent process,
    340 	 * marking us as parent via retval[1].
    341 	 */
    342 	if (retval != NULL) {
    343 		retval[0] = p2->p_pid;
    344 		retval[1] = 0;
    345 	}
    346 	return (0);
    347 }
    348