[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	7.10 (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 "../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;
	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_flag&SKTR) {
		rpp->p_flag |= SKTR;
		if ((rpp->p_tracep = rip->p_tracep) != NULL)
			VREF(rpp->p_tracep);
		rpp->p_traceflag = rip->p_traceflag;
	} 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;
	n = PIDHASH(rpp->p_pid);
	rpp->p_idhash = pidhash[n];
	pidhash[n] = rpp - proc;
	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	*
e3fb2940	17	* @(#)kern_fork.c 7.10 (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"
c4ec2128	32	#include "../ufs/quota.h"
50108d5c	33
d301d150 KM	34	#include "machine/reg.h"
	35	#include "machine/pte.h"
	36	#include "machine/psl.h"
	37
50108d5c SL	38	/*
	39	* fork system call.
	40	*/
	41	fork()
	42	{
	43
	44	u.u_cdmap = zdmap;
	45	u.u_csmap = zdmap;
	46	if (swpexpand(u.u_dsize, u.u_ssize, &u.u_cdmap, &u.u_csmap) == 0) {
	47	u.u_r.r_val2 = 0;
	48	return;
	49	}
	50	fork1(0);
	51	}
	52
	53	vfork()
	54	{
	55
	56	fork1(1);
	57	}
	58
	59	fork1(isvfork)
	60	int isvfork;
	61	{
	62	register struct proc p1, p2;
	63	register a;
	64
	65	a = 0;
1d348849 MK	66	if (u.u_uid != 0) {
	67	for (p1 = allproc; p1; p1 = p1->p_nxt)
	68	if (p1->p_uid == u.u_uid)
	69	a++;
	70	for (p1 = zombproc; p1; p1 = p1->p_nxt)
	71	if (p1->p_uid == u.u_uid)
50108d5c	72	a++;
50108d5c SL	73	}
	74	/*
	75	* Disallow if
	76	* No processes at all;
	77	* not su and too many procs owned; or
	78	* not su and would take last slot.
	79	*/
1d348849	80	p2 = freeproc;
50108d5c SL	81	if (p2==NULL)
50108d5c SL	82	tablefull("proc");
1d348849	83	if (p2==NULL \|\| (u.u_uid!=0 && (p2->p_nxt == NULL \|\| a>MAXUPRC))) {
50108d5c SL	84	u.u_error = EAGAIN;
50108d5c SL	85	if (!isvfork) {
8011f5df MK	86	(void) vsexpand((size_t)0, &u.u_cdmap, 1);
8011f5df MK	87	(void) vsexpand((size_t)0, &u.u_csmap, 1);
50108d5c SL	88	}
	89	goto out;
	90	}
	91	p1 = u.u_procp;
	92	if (newproc(isvfork)) {
	93	u.u_r.r_val1 = p1->p_pid;
	94	u.u_r.r_val2 = 1; /* child */
e8109af3	95	u.u_start = time;
50108d5c SL	96	u.u_acflag = AFORK;
	97	return;
	98	}
	99	u.u_r.r_val1 = p2->p_pid;
	100
	101	out:
	102	u.u_r.r_val2 = 0;
	103	}
	104
	105	/*
	106	* Create a new process-- the internal version of
	107	* sys fork.
	108	* It returns 1 in the new process, 0 in the old.
	109	*/
	110	newproc(isvfork)
	111	int isvfork;
	112	{
50108d5c SL	113	register struct proc rpp, rip;
	114	register int n;
	115	register struct file *fp;
1d348849	116	static int pidchecked = 0;
50108d5c	117
50108d5c SL	118	/*
	119	* First, just locate a slot for a process
	120	* and copy the useful info from this process into it.
	121	* The panic "cannot happen" because fork has already
	122	* checked for the existence of a slot.
	123	*/
50108d5c	124	mpid++;
1d348849	125	retry:
50108d5c	126	if (mpid >= 30000) {
1d348849 MK	127	mpid = 100;
1d348849 MK	128	pidchecked = 0;
50108d5c	129	}
1d348849 MK	130	if (mpid >= pidchecked) {
1d348849 MK	131	int doingzomb = 0;
e9539440	132
1d348849 MK	133	pidchecked = 30000;
	134	/*
	135	* Scan the proc table to check whether this pid
	136	* is in use. Remember the lowest pid that's greater
	137	* than mpid, so we can avoid checking for a while.
	138	*/
	139	rpp = allproc;
	140	again:
	141	for (; rpp != NULL; rpp = rpp->p_nxt) {
8fe87cbb	142	if (rpp->p_pid == mpid \|\| rpp->p_pgrp->pg_id == mpid) {
1d348849 MK	143	mpid++;
	144	if (mpid >= pidchecked)
	145	goto retry;
	146	}
e9539440	147	if (rpp->p_pid > mpid && pidchecked > rpp->p_pid)
1d348849	148	pidchecked = rpp->p_pid;
8fe87cbb MT	149	if (rpp->p_pgrp->pg_id > mpid &&
	150	pidchecked > rpp->p_pgrp->pg_id)
	151	pidchecked = rpp->p_pgrp->pg_id;
1d348849 MK	152	}
	153	if (!doingzomb) {
	154	doingzomb = 1;
	155	rpp = zombproc;
	156	goto again;
	157	}
50108d5c	158	}
1d348849	159	if ((rpp = freeproc) == NULL)
50108d5c SL	160	panic("no procs");
50108d5c SL	161
1d348849 MK	162	freeproc = rpp->p_nxt; /* off freeproc */
	163	rpp->p_nxt = allproc; /* onto allproc */
	164	rpp->p_nxt->p_prev = &rpp->p_nxt; /* (allproc is never NULL) */
	165	rpp->p_prev = &allproc;
	166	allproc = rpp;
	167
50108d5c SL	168	/*
	169	* Make a proc table entry for the new process.
	170	*/
	171	rip = u.u_procp;
	172	#ifdef QUOTA
	173	rpp->p_quota = rip->p_quota;
	174	rpp->p_quota->q_cnt++;
fb1db32c MK	175	#endif
	176	#if defined(tahoe)
	177	rpp->p_ckey = rip->p_ckey;
	178	rpp->p_dkey = 0;
50108d5c SL	179	#endif
	180	rpp->p_stat = SIDL;
	181	timerclear(&rpp->p_realtimer.it_value);
e3fb2940	182	rpp->p_flag = SLOAD \| (rip->p_flag & (SPAGV\|SCTTY));
50108d5c SL	183	if (isvfork) {
	184	rpp->p_flag \|= SVFORK;
	185	rpp->p_ndx = rip->p_ndx;
	186	} else
	187	rpp->p_ndx = rpp - proc;
	188	rpp->p_uid = rip->p_uid;
d9d1cf27 MK	189	rpp->p_ruid = rip->p_ruid;
d9d1cf27 MK	190	rpp->p_rgid = rip->p_rgid;
50108d5c	191	rpp->p_pgrp = rip->p_pgrp;
8fe87cbb MT	192	rpp->p_pgrpnxt = rip->p_pgrpnxt;
8fe87cbb MT	193	rip->p_pgrpnxt = rpp;
50108d5c SL	194	rpp->p_nice = rip->p_nice;
	195	rpp->p_textp = isvfork ? 0 : rip->p_textp;
	196	rpp->p_pid = mpid;
	197	rpp->p_ppid = rip->p_pid;
	198	rpp->p_pptr = rip;
	199	rpp->p_osptr = rip->p_cptr;
	200	if (rip->p_cptr)
	201	rip->p_cptr->p_ysptr = rpp;
	202	rpp->p_ysptr = NULL;
	203	rpp->p_cptr = NULL;
	204	rip->p_cptr = rpp;
	205	rpp->p_time = 0;
	206	rpp->p_cpu = 0;
dd012d1e SL	207	rpp->p_sigmask = rip->p_sigmask;
	208	rpp->p_sigcatch = rip->p_sigcatch;
	209	rpp->p_sigignore = rip->p_sigignore;
	210	/* take along any pending signals like stops? */
50108d5c SL	211	if (isvfork) {
	212	rpp->p_tsize = rpp->p_dsize = rpp->p_ssize = 0;
	213	rpp->p_szpt = clrnd(ctopt(UPAGES));
	214	forkstat.cntvfork++;
	215	forkstat.sizvfork += rip->p_dsize + rip->p_ssize;
	216	} else {
	217	rpp->p_tsize = rip->p_tsize;
	218	rpp->p_dsize = rip->p_dsize;
	219	rpp->p_ssize = rip->p_ssize;
	220	rpp->p_szpt = rip->p_szpt;
	221	forkstat.cntfork++;
	222	forkstat.sizfork += rip->p_dsize + rip->p_ssize;
	223	}
98111078 MT	224	#ifdef KTRACE
	225	if (rip->p_flag&SKTR) {
	226	rpp->p_flag \|= SKTR;
20b702d3 MT	227	if ((rpp->p_tracep = rip->p_tracep) != NULL)
20b702d3 MT	228	VREF(rpp->p_tracep);
98111078 MT	229	rpp->p_traceflag = rip->p_traceflag;
	230	} else {
	231	rpp->p_tracep = NULL;
	232	rpp->p_traceflag = 0;
	233	}
	234	#endif
50108d5c SL	235	rpp->p_rssize = 0;
	236	rpp->p_maxrss = rip->p_maxrss;
	237	rpp->p_wchan = 0;
	238	rpp->p_slptime = 0;
	239	rpp->p_pctcpu = 0;
	240	rpp->p_cpticks = 0;
	241	n = PIDHASH(rpp->p_pid);
1d348849	242	rpp->p_idhash = pidhash[n];
50108d5c SL	243	pidhash[n] = rpp - proc;
	244	multprog++;
	245
	246	/*
	247	* Increase reference counts on shared objects.
	248	*/
8694aaad	249	for (n = 0; n <= u.u_lastfile; n++) {
50108d5c SL	250	fp = u.u_ofile[n];
	251	if (fp == NULL)
	252	continue;
	253	fp->f_count++;
50108d5c	254	}
8fe1c702	255	VREF(u.u_cdir);
50108d5c	256	if (u.u_rdir)
8fe1c702	257	VREF(u.u_rdir);
c4ec2128	258	crhold(u.u_cred);
50108d5c SL	259
50108d5c SL	260	/*
50108d5c SL	261	* This begins the section where we must prevent the parent
	262	* from being swapped.
	263	*/
	264	rip->p_flag \|= SKEEP;
	265	if (procdup(rpp, isvfork))
	266	return (1);
	267
	268	/*
	269	* Make child runnable and add to run queue.
	270	*/
01b0e233	271	(void) splclock();
50108d5c SL	272	rpp->p_stat = SRUN;
	273	setrq(rpp);
	274	(void) spl0();
	275
	276	/*
	277	* Cause child to take a non-local goto as soon as it runs.
	278	* On older systems this was done with SSWAP bit in proc
	279	* table; on VAX we use u.u_pcb.pcb_sswap so don't need
	280	* to do rpp->p_flag \|= SSWAP. Actually do nothing here.
	281	*/
	282	/* rpp->p_flag \|= SSWAP; */
	283
	284	/*
	285	* Now can be swapped.
	286	*/
	287	rip->p_flag &= ~SKEEP;
	288
	289	/*
	290	* If vfork make chain from parent process to child
	291	* (where virtal memory is temporarily). Wait for
	292	* child to finish, steal virtual memory back,
	293	* and wakeup child to let it die.
	294	*/
	295	if (isvfork) {
	296	u.u_procp->p_xlink = rpp;
	297	u.u_procp->p_flag \|= SNOVM;
	298	while (rpp->p_flag & SVFORK)
	299	sleep((caddr_t)rpp, PZERO - 1);
	300	if ((rpp->p_flag & SLOAD) == 0)
	301	panic("newproc vfork");
	302	uaccess(rpp, Vfmap, &vfutl);
	303	u.u_procp->p_xlink = 0;
	304	vpassvm(rpp, u.u_procp, &vfutl, &u, Vfmap);
	305	u.u_procp->p_flag &= ~SNOVM;
	306	rpp->p_ndx = rpp - proc;
	307	rpp->p_flag \|= SVFDONE;
	308	wakeup((caddr_t)rpp);
	309	}
	310
	311	/*
	312	* 0 return means parent.
	313	*/
	314	return (0);
	315	}