[unix-history] / usr / src / sys / kern / kern_fork.c

/*
 * Copyright (c) 1982, 1986, 1989 Regents of the University of California.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms are permitted
 * provided that the above copyright notice and this paragraph are
 * duplicated in all such forms and that any documentation,
 * advertising materials, and other materials related to such
 * distribution and use acknowledge that the software was developed
 * by the University of California, Berkeley.  The name of the
 * University may not be used to endorse or promote products derived
 * from this software without specific prior written permission.
 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
 *
 *	@(#)kern_fork.c	7.11 (Berkeley) %G%
 */

#include "param.h"
#include "systm.h"
#include "map.h"
#include "user.h"
#include "kernel.h"
#include "proc.h"
#include "vnode.h"
#include "seg.h"
#include "vm.h"
#include "text.h"
#include "file.h"
#include "acct.h"
#include "ktrace.h"
#include "../ufs/quota.h"

#include "machine/reg.h"
#include "machine/pte.h"
#include "machine/psl.h"

/*
 * fork system call.
 */
fork()
{

	u.u_cdmap = zdmap;
	u.u_csmap = zdmap;
	if (swpexpand(u.u_dsize, u.u_ssize, &u.u_cdmap, &u.u_csmap) == 0) {
		u.u_r.r_val2 = 0;
		return;
	}
	fork1(0);
}

vfork()
{

	fork1(1);
}

fork1(isvfork)
	int isvfork;
{
	register struct proc *p1, *p2;
	register a;

	a = 0;
	if (u.u_uid != 0) {
		for (p1 = allproc; p1; p1 = p1->p_nxt)
			if (p1->p_uid == u.u_uid)
				a++;
		for (p1 = zombproc; p1; p1 = p1->p_nxt)
			if (p1->p_uid == u.u_uid)
				a++;
	}
	/*
	 * Disallow if
	 *  No processes at all;
	 *  not su and too many procs owned; or
	 *  not su and would take last slot.
	 */
	p2 = freeproc;
	if (p2==NULL)
		tablefull("proc");
	if (p2==NULL || (u.u_uid!=0 && (p2->p_nxt == NULL || a>MAXUPRC))) {
		u.u_error = EAGAIN;
		if (!isvfork) {
			(void) vsexpand((size_t)0, &u.u_cdmap, 1);
			(void) vsexpand((size_t)0, &u.u_csmap, 1);
		}
		goto out;
	}
	p1 = u.u_procp;
	if (newproc(isvfork)) {
		u.u_r.r_val1 = p1->p_pid;
		u.u_r.r_val2 = 1;  /* child */
		u.u_start = time;
		u.u_acflag = AFORK;
		return;
	}
	u.u_r.r_val1 = p2->p_pid;

out:
	u.u_r.r_val2 = 0;
}

/*
 * Create a new process-- the internal version of
 * sys fork.
 * It returns 1 in the new process, 0 in the old.
 */
newproc(isvfork)
	int isvfork;
{
	register struct proc *rpp, *rip;
	register int n;
	register struct file *fp;
	static int pidchecked = 0;

	/*
	 * First, just locate a slot for a process
	 * and copy the useful info from this process into it.
	 * The panic "cannot happen" because fork has already
	 * checked for the existence of a slot.
	 */
	mpid++;
retry:
	if (mpid >= 30000) {
		mpid = 100;
		pidchecked = 0;
	}
	if (mpid >= pidchecked) {
		int doingzomb = 0;

		pidchecked = 30000;
		/*
		 * Scan the proc table to check whether this pid
		 * is in use.  Remember the lowest pid that's greater
		 * than mpid, so we can avoid checking for a while.
		 */
		rpp = allproc;
again:
		for (; rpp != NULL; rpp = rpp->p_nxt) {
			if (rpp->p_pid == mpid || rpp->p_pgrp->pg_id == mpid) {
				mpid++;
				if (mpid >= pidchecked)
					goto retry;
			}
			if (rpp->p_pid > mpid && pidchecked > rpp->p_pid)
				pidchecked = rpp->p_pid;
			if (rpp->p_pgrp->pg_id > mpid && 
			    pidchecked > rpp->p_pgrp->pg_id)
				pidchecked = rpp->p_pgrp->pg_id;
		}
		if (!doingzomb) {
			doingzomb = 1;
			rpp = zombproc;
			goto again;
		}
	}
	if ((rpp = freeproc) == NULL)
		panic("no procs");

	freeproc = rpp->p_nxt;			/* off freeproc */
	rpp->p_nxt = allproc;			/* onto allproc */
	rpp->p_nxt->p_prev = &rpp->p_nxt;	/*   (allproc is never NULL) */
	rpp->p_prev = &allproc;
	allproc = rpp;

	/*
	 * Make a proc table entry for the new process.
	 */
	rip = u.u_procp;
#ifdef QUOTA
	rpp->p_quota = rip->p_quota;
	rpp->p_quota->q_cnt++;
#endif
#if defined(tahoe)
	rpp->p_ckey = rip->p_ckey;
	rpp->p_dkey = 0;
#endif
	rpp->p_stat = SIDL;
	timerclear(&rpp->p_realtimer.it_value);
	rpp->p_flag = SLOAD | (rip->p_flag & (SPAGV|SCTTY));
	if (isvfork) {
		rpp->p_flag |= SVFORK;
		rpp->p_ndx = rip->p_ndx;
	} else
		rpp->p_ndx = rpp - proc;
	bcopy(rip->p_comm, rpp->p_comm, MAXCOMLEN+1);
	bcopy(rip->p_logname, rpp->p_logname, MAXLOGNAME);
	rpp->p_uid = rip->p_uid;
	rpp->p_ruid = rip->p_ruid;
	rpp->p_rgid = rip->p_rgid;
	rpp->p_pgrp = rip->p_pgrp;
	rpp->p_pgrpnxt = rip->p_pgrpnxt;
	rip->p_pgrpnxt = rpp;
	rpp->p_nice = rip->p_nice;
	rpp->p_textp = isvfork ? 0 : rip->p_textp;
	rpp->p_pid = mpid;
	rpp->p_ppid = rip->p_pid;
	rpp->p_pptr = rip;
	rpp->p_osptr = rip->p_cptr;
	if (rip->p_cptr)
		rip->p_cptr->p_ysptr = rpp;
	rpp->p_ysptr = NULL;
	rpp->p_cptr = NULL;
	rip->p_cptr = rpp;
	rpp->p_time = 0;
	rpp->p_cpu = 0;
	rpp->p_sigmask = rip->p_sigmask;
	rpp->p_sigcatch = rip->p_sigcatch;
	rpp->p_sigignore = rip->p_sigignore;
	/* take along any pending signals like stops? */
	if (isvfork) {
		rpp->p_tsize = rpp->p_dsize = rpp->p_ssize = 0;
		rpp->p_szpt = clrnd(ctopt(UPAGES));
		forkstat.cntvfork++;
		forkstat.sizvfork += rip->p_dsize + rip->p_ssize;
	} else {
		rpp->p_tsize = rip->p_tsize;
		rpp->p_dsize = rip->p_dsize;
		rpp->p_ssize = rip->p_ssize;
		rpp->p_szpt = rip->p_szpt;
		forkstat.cntfork++;
		forkstat.sizfork += rip->p_dsize + rip->p_ssize;
	}
#ifdef KTRACE
	if (rip->p_traceflag&KTRFAC_INHERIT) {
		rpp->p_traceflag = rip->p_traceflag;
		if ((rpp->p_tracep = rip->p_tracep) != NULL)
			VREF(rpp->p_tracep);
	} else {
		rpp->p_tracep = NULL;
		rpp->p_traceflag = 0;
	}
#endif
	rpp->p_rssize = 0;
	rpp->p_maxrss = rip->p_maxrss;
	rpp->p_wchan = 0;
	rpp->p_slptime = 0;
	rpp->p_pctcpu = 0;
	rpp->p_cpticks = 0;
	{
	struct proc **hash = &pidhash[PIDHASH(rpp->p_pid)];

	rpp->p_hash = *hash;
	*hash = rpp;
	}
	multprog++;

	/*
	 * Increase reference counts on shared objects.
	 */
	for (n = 0; n <= u.u_lastfile; n++) {
		fp = u.u_ofile[n];
		if (fp == NULL)
			continue;
		fp->f_count++;
	}
	VREF(u.u_cdir);
	if (u.u_rdir)
		VREF(u.u_rdir);
	crhold(u.u_cred);

	/*
	 * This begins the section where we must prevent the parent
	 * from being swapped.
	 */
	rip->p_flag |= SKEEP;
	if (procdup(rpp, isvfork))
		return (1);

	/*
	 * Make child runnable and add to run queue.
	 */
	(void) splclock();
	rpp->p_stat = SRUN;
	setrq(rpp);
	(void) spl0();

	/*
	 * Cause child to take a non-local goto as soon as it runs.
	 * On older systems this was done with SSWAP bit in proc
	 * table; on VAX we use u.u_pcb.pcb_sswap so don't need
	 * to do rpp->p_flag |= SSWAP.  Actually do nothing here.
	 */
	/* rpp->p_flag |= SSWAP; */

	/*
	 * Now can be swapped.
	 */
	rip->p_flag &= ~SKEEP;

	/*
	 * If vfork make chain from parent process to child
	 * (where virtal memory is temporarily).  Wait for
	 * child to finish, steal virtual memory back,
	 * and wakeup child to let it die.
	 */
	if (isvfork) {
		u.u_procp->p_xlink = rpp;
		u.u_procp->p_flag |= SNOVM;
		while (rpp->p_flag & SVFORK)
			sleep((caddr_t)rpp, PZERO - 1);
		if ((rpp->p_flag & SLOAD) == 0)
			panic("newproc vfork");
		uaccess(rpp, Vfmap, &vfutl);
		u.u_procp->p_xlink = 0;
		vpassvm(rpp, u.u_procp, &vfutl, &u, Vfmap);
		u.u_procp->p_flag &= ~SNOVM;
		rpp->p_ndx = rpp - proc;
		rpp->p_flag |= SVFDONE;
		wakeup((caddr_t)rpp);
	}

	/*
	 * 0 return means parent.
	 */
	return (0);
}
Commit	Line	Data
da7c5cc6	1	/*
c4ec2128 KM	2	* Copyright (c) 1982, 1986, 1989 Regents of the University of California.
c4ec2128 KM	3	* All rights reserved.
da7c5cc6	4	*
c4ec2128 KM	5	* Redistribution and use in source and binary forms are permitted
	6	* provided that the above copyright notice and this paragraph are
	7	* duplicated in all such forms and that any documentation,
	8	* advertising materials, and other materials related to such
	9	* distribution and use acknowledge that the software was developed
	10	* by the University of California, Berkeley. The name of the
	11	* University may not be used to endorse or promote products derived
	12	* from this software without specific prior written permission.
	13	* THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
	14	* IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
	15	* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
	16	*
56a12c82	17	* @(#)kern_fork.c 7.11 (Berkeley) %G%
da7c5cc6	18	*/
50108d5c	19
94368568 JB	20	#include "param.h"
	21	#include "systm.h"
	22	#include "map.h"
94368568 JB	23	#include "user.h"
	24	#include "kernel.h"
	25	#include "proc.h"
c4ec2128	26	#include "vnode.h"
94368568 JB	27	#include "seg.h"
	28	#include "vm.h"
	29	#include "text.h"
	30	#include "file.h"
	31	#include "acct.h"
56a12c82	32	#include "ktrace.h"
c4ec2128	33	#include "../ufs/quota.h"
50108d5c	34
d301d150 KM	35	#include "machine/reg.h"
	36	#include "machine/pte.h"
	37	#include "machine/psl.h"
	38
50108d5c SL	39	/*
	40	* fork system call.
	41	*/
	42	fork()
	43	{
	44
	45	u.u_cdmap = zdmap;
	46	u.u_csmap = zdmap;
	47	if (swpexpand(u.u_dsize, u.u_ssize, &u.u_cdmap, &u.u_csmap) == 0) {
	48	u.u_r.r_val2 = 0;
	49	return;
	50	}
	51	fork1(0);
	52	}
	53
	54	vfork()
	55	{
	56
	57	fork1(1);
	58	}
	59
	60	fork1(isvfork)
	61	int isvfork;
	62	{
	63	register struct proc p1, p2;
	64	register a;
	65
	66	a = 0;
1d348849 MK	67	if (u.u_uid != 0) {
	68	for (p1 = allproc; p1; p1 = p1->p_nxt)
	69	if (p1->p_uid == u.u_uid)
	70	a++;
	71	for (p1 = zombproc; p1; p1 = p1->p_nxt)
	72	if (p1->p_uid == u.u_uid)
50108d5c	73	a++;
50108d5c SL	74	}
	75	/*
	76	* Disallow if
	77	* No processes at all;
	78	* not su and too many procs owned; or
	79	* not su and would take last slot.
	80	*/
1d348849	81	p2 = freeproc;
50108d5c SL	82	if (p2==NULL)
50108d5c SL	83	tablefull("proc");
1d348849	84	if (p2==NULL \|\| (u.u_uid!=0 && (p2->p_nxt == NULL \|\| a>MAXUPRC))) {
50108d5c SL	85	u.u_error = EAGAIN;
50108d5c SL	86	if (!isvfork) {
8011f5df MK	87	(void) vsexpand((size_t)0, &u.u_cdmap, 1);
8011f5df MK	88	(void) vsexpand((size_t)0, &u.u_csmap, 1);
50108d5c SL	89	}
	90	goto out;
	91	}
	92	p1 = u.u_procp;
	93	if (newproc(isvfork)) {
	94	u.u_r.r_val1 = p1->p_pid;
	95	u.u_r.r_val2 = 1; /* child */
e8109af3	96	u.u_start = time;
50108d5c SL	97	u.u_acflag = AFORK;
	98	return;
	99	}
	100	u.u_r.r_val1 = p2->p_pid;
	101
	102	out:
	103	u.u_r.r_val2 = 0;
	104	}
	105
	106	/*
	107	* Create a new process-- the internal version of
	108	* sys fork.
	109	* It returns 1 in the new process, 0 in the old.
	110	*/
	111	newproc(isvfork)
	112	int isvfork;
	113	{
50108d5c SL	114	register struct proc rpp, rip;
	115	register int n;
	116	register struct file *fp;
1d348849	117	static int pidchecked = 0;
50108d5c	118
50108d5c SL	119	/*
	120	* First, just locate a slot for a process
	121	* and copy the useful info from this process into it.
	122	* The panic "cannot happen" because fork has already
	123	* checked for the existence of a slot.
	124	*/
50108d5c	125	mpid++;
1d348849	126	retry:
50108d5c	127	if (mpid >= 30000) {
1d348849 MK	128	mpid = 100;
1d348849 MK	129	pidchecked = 0;
50108d5c	130	}
1d348849 MK	131	if (mpid >= pidchecked) {
1d348849 MK	132	int doingzomb = 0;
e9539440	133
1d348849 MK	134	pidchecked = 30000;
	135	/*
	136	* Scan the proc table to check whether this pid
	137	* is in use. Remember the lowest pid that's greater
	138	* than mpid, so we can avoid checking for a while.
	139	*/
	140	rpp = allproc;
	141	again:
	142	for (; rpp != NULL; rpp = rpp->p_nxt) {
8fe87cbb	143	if (rpp->p_pid == mpid \|\| rpp->p_pgrp->pg_id == mpid) {
1d348849 MK	144	mpid++;
	145	if (mpid >= pidchecked)
	146	goto retry;
	147	}
e9539440	148	if (rpp->p_pid > mpid && pidchecked > rpp->p_pid)
1d348849	149	pidchecked = rpp->p_pid;
8fe87cbb MT	150	if (rpp->p_pgrp->pg_id > mpid &&
	151	pidchecked > rpp->p_pgrp->pg_id)
	152	pidchecked = rpp->p_pgrp->pg_id;
1d348849 MK	153	}
	154	if (!doingzomb) {
	155	doingzomb = 1;
	156	rpp = zombproc;
	157	goto again;
	158	}
50108d5c	159	}
1d348849	160	if ((rpp = freeproc) == NULL)
50108d5c SL	161	panic("no procs");
50108d5c SL	162
1d348849 MK	163	freeproc = rpp->p_nxt; /* off freeproc */
	164	rpp->p_nxt = allproc; /* onto allproc */
	165	rpp->p_nxt->p_prev = &rpp->p_nxt; /* (allproc is never NULL) */
	166	rpp->p_prev = &allproc;
	167	allproc = rpp;
	168
50108d5c SL	169	/*
	170	* Make a proc table entry for the new process.
	171	*/
	172	rip = u.u_procp;
	173	#ifdef QUOTA
	174	rpp->p_quota = rip->p_quota;
	175	rpp->p_quota->q_cnt++;
fb1db32c MK	176	#endif
	177	#if defined(tahoe)
	178	rpp->p_ckey = rip->p_ckey;
	179	rpp->p_dkey = 0;
50108d5c SL	180	#endif
	181	rpp->p_stat = SIDL;
	182	timerclear(&rpp->p_realtimer.it_value);
e3fb2940	183	rpp->p_flag = SLOAD \| (rip->p_flag & (SPAGV\|SCTTY));
50108d5c SL	184	if (isvfork) {
	185	rpp->p_flag \|= SVFORK;
	186	rpp->p_ndx = rip->p_ndx;
	187	} else
	188	rpp->p_ndx = rpp - proc;
56a12c82 MT	189	bcopy(rip->p_comm, rpp->p_comm, MAXCOMLEN+1);
56a12c82 MT	190	bcopy(rip->p_logname, rpp->p_logname, MAXLOGNAME);
50108d5c	191	rpp->p_uid = rip->p_uid;
d9d1cf27 MK	192	rpp->p_ruid = rip->p_ruid;
d9d1cf27 MK	193	rpp->p_rgid = rip->p_rgid;
50108d5c	194	rpp->p_pgrp = rip->p_pgrp;
8fe87cbb MT	195	rpp->p_pgrpnxt = rip->p_pgrpnxt;
8fe87cbb MT	196	rip->p_pgrpnxt = rpp;
50108d5c SL	197	rpp->p_nice = rip->p_nice;
	198	rpp->p_textp = isvfork ? 0 : rip->p_textp;
	199	rpp->p_pid = mpid;
	200	rpp->p_ppid = rip->p_pid;
	201	rpp->p_pptr = rip;
	202	rpp->p_osptr = rip->p_cptr;
	203	if (rip->p_cptr)
	204	rip->p_cptr->p_ysptr = rpp;
	205	rpp->p_ysptr = NULL;
	206	rpp->p_cptr = NULL;
	207	rip->p_cptr = rpp;
	208	rpp->p_time = 0;
	209	rpp->p_cpu = 0;
dd012d1e SL	210	rpp->p_sigmask = rip->p_sigmask;
	211	rpp->p_sigcatch = rip->p_sigcatch;
	212	rpp->p_sigignore = rip->p_sigignore;
	213	/* take along any pending signals like stops? */
50108d5c SL	214	if (isvfork) {
	215	rpp->p_tsize = rpp->p_dsize = rpp->p_ssize = 0;
	216	rpp->p_szpt = clrnd(ctopt(UPAGES));
	217	forkstat.cntvfork++;
	218	forkstat.sizvfork += rip->p_dsize + rip->p_ssize;
	219	} else {
	220	rpp->p_tsize = rip->p_tsize;
	221	rpp->p_dsize = rip->p_dsize;
	222	rpp->p_ssize = rip->p_ssize;
	223	rpp->p_szpt = rip->p_szpt;
	224	forkstat.cntfork++;
	225	forkstat.sizfork += rip->p_dsize + rip->p_ssize;
	226	}
98111078	227	#ifdef KTRACE
56a12c82 MT	228	if (rip->p_traceflag&KTRFAC_INHERIT) {
56a12c82 MT	229	rpp->p_traceflag = rip->p_traceflag;
20b702d3 MT	230	if ((rpp->p_tracep = rip->p_tracep) != NULL)
20b702d3 MT	231	VREF(rpp->p_tracep);
98111078 MT	232	} else {
	233	rpp->p_tracep = NULL;
	234	rpp->p_traceflag = 0;
	235	}
	236	#endif
50108d5c SL	237	rpp->p_rssize = 0;
	238	rpp->p_maxrss = rip->p_maxrss;
	239	rpp->p_wchan = 0;
	240	rpp->p_slptime = 0;
	241	rpp->p_pctcpu = 0;
	242	rpp->p_cpticks = 0;
56a12c82 MT	243	{
	244	struct proc **hash = &pidhash[PIDHASH(rpp->p_pid)];
	245
	246	rpp->p_hash = *hash;
	247	*hash = rpp;
	248	}
50108d5c SL	249	multprog++;
	250
	251	/*
	252	* Increase reference counts on shared objects.
	253	*/
8694aaad	254	for (n = 0; n <= u.u_lastfile; n++) {
50108d5c SL	255	fp = u.u_ofile[n];
	256	if (fp == NULL)
	257	continue;
	258	fp->f_count++;
50108d5c	259	}
8fe1c702	260	VREF(u.u_cdir);
50108d5c	261	if (u.u_rdir)
8fe1c702	262	VREF(u.u_rdir);
c4ec2128	263	crhold(u.u_cred);
50108d5c SL	264
50108d5c SL	265	/*
50108d5c SL	266	* This begins the section where we must prevent the parent
	267	* from being swapped.
	268	*/
	269	rip->p_flag \|= SKEEP;
	270	if (procdup(rpp, isvfork))
	271	return (1);
	272
	273	/*
	274	* Make child runnable and add to run queue.
	275	*/
01b0e233	276	(void) splclock();
50108d5c SL	277	rpp->p_stat = SRUN;
	278	setrq(rpp);
	279	(void) spl0();
	280
	281	/*
	282	* Cause child to take a non-local goto as soon as it runs.
	283	* On older systems this was done with SSWAP bit in proc
	284	* table; on VAX we use u.u_pcb.pcb_sswap so don't need
	285	* to do rpp->p_flag \|= SSWAP. Actually do nothing here.
	286	*/
	287	/* rpp->p_flag \|= SSWAP; */
	288
	289	/*
	290	* Now can be swapped.
	291	*/
	292	rip->p_flag &= ~SKEEP;
	293
	294	/*
	295	* If vfork make chain from parent process to child
	296	* (where virtal memory is temporarily). Wait for
	297	* child to finish, steal virtual memory back,
	298	* and wakeup child to let it die.
	299	*/
	300	if (isvfork) {
	301	u.u_procp->p_xlink = rpp;
	302	u.u_procp->p_flag \|= SNOVM;
	303	while (rpp->p_flag & SVFORK)
	304	sleep((caddr_t)rpp, PZERO - 1);
	305	if ((rpp->p_flag & SLOAD) == 0)
	306	panic("newproc vfork");
	307	uaccess(rpp, Vfmap, &vfutl);
	308	u.u_procp->p_xlink = 0;
	309	vpassvm(rpp, u.u_procp, &vfutl, &u, Vfmap);
	310	u.u_procp->p_flag &= ~SNOVM;
	311	rpp->p_ndx = rpp - proc;
	312	rpp->p_flag \|= SVFDONE;
	313	wakeup((caddr_t)rpp);
	314	}
	315
	316	/*
	317	* 0 return means parent.
	318	*/
	319	return (0);
	320	}