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

/*	kern_proc.c	3.5	%H%	*/

#include "../h/param.h"
#include "../h/systm.h"
#include "../h/map.h"
#include "../h/mtpr.h"
#include "../h/dir.h"
#include "../h/user.h"
#include "../h/proc.h"
#include "../h/buf.h"
#include "../h/reg.h"
#include "../h/inode.h"
#include "../h/seg.h"
#include "../h/acct.h"
#include <wait.h>
#include "../h/pte.h"
#include "../h/vm.h"
#include "../h/text.h"
#include "../h/psl.h"

/*
 * exec system call, with and without environments.
 */
struct execa {
	char	*fname;
	char	**argp;
	char	**envp;
};

exec()
{
	((struct execa *)u.u_ap)->envp = NULL;
	exece();
}

exece()
{
	register nc;
	register char *cp;
	register struct buf *bp;
	register struct execa *uap;
	int na, ne, ucp, ap, c;
	struct inode *ip;
	swblk_t bno;

	if ((ip = namei(uchar, 0)) == NULL)
		return;
	bno = 0;
	bp = 0;
	if(access(ip, IEXEC))
		goto bad;
	if((ip->i_mode & IFMT) != IFREG ||
	   (ip->i_mode & (IEXEC|(IEXEC>>3)|(IEXEC>>6))) == 0) {
		u.u_error = EACCES;
		goto bad;
	}
	/*
	 * Collect arguments on "file" in swap space.
	 */
	na = 0;
	ne = 0;
	nc = 0;
	uap = (struct execa *)u.u_ap;
	if ((bno = malloc(swapmap, ctod(clrnd((int) btoc(NCARGS))))) == 0) {
		swkill(u.u_procp, "exece");
		goto bad;
	}
	if (bno % CLSIZE)
		panic("execa malloc");
	if (uap->argp) for (;;) {
		ap = NULL;
		if (uap->argp) {
			ap = fuword((caddr_t)uap->argp);
			uap->argp++;
		}
		if (ap==NULL && uap->envp) {
			uap->argp = NULL;
			if ((ap = fuword((caddr_t)uap->envp)) == NULL)
				break;
			uap->envp++;
			ne++;
		}
		if (ap==NULL)
			break;
		na++;
		if(ap == -1)
			u.u_error = EFAULT;
		do {
			if (nc >= NCARGS-1)
				u.u_error = E2BIG;
			if ((c = fubyte((caddr_t)ap++)) < 0)
				u.u_error = EFAULT;
			if (u.u_error) {
				if (bp)
					brelse(bp);
				bp = 0;
				goto badarg;
			}
			if ((nc&BMASK) == 0) {
				if (bp)
					bdwrite(bp);
				bp = getblk(swapdev, (daddr_t)(dbtofsb(swplo+bno)+(nc>>BSHIFT)));
				cp = bp->b_un.b_addr;
			}
			nc++;
			*cp++ = c;
		} while (c>0);
	}
	if (bp)
		bdwrite(bp);
	bp = 0;
	nc = (nc + NBPW-1) & ~(NBPW-1);
	if (getxfile(ip, nc) || u.u_error) {
badarg:
		for (c = 0; c < nc; c += BSIZE)
			if (bp = baddr(swapdev, dbtofsb(swplo+bno)+(c>>BSHIFT))) {
				bp->b_flags |= B_AGE;		/* throw away */
				bp->b_flags &= ~B_DELWRI;	/* cancel io */
				brelse(bp);
				bp = 0;
			}
		goto bad;
	}

	/*
	 * copy back arglist
	 */

	ucp = USRSTACK - nc - NBPW;
	ap = ucp - na*NBPW - 3*NBPW;
	u.u_ar0[SP] = ap;
	(void) suword((caddr_t)ap, na-ne);
	nc = 0;
	for (;;) {
		ap += NBPW;
		if (na==ne) {
			(void) suword((caddr_t)ap, 0);
			ap += NBPW;
		}
		if (--na < 0)
			break;
		(void) suword((caddr_t)ap, ucp);
		do {
			if ((nc&BMASK) == 0) {
				if (bp)
					brelse(bp);
				bp = bread(swapdev, (daddr_t)(dbtofsb(swplo+bno)+(nc>>BSHIFT)));
				bp->b_flags |= B_AGE;		/* throw away */
				bp->b_flags &= ~B_DELWRI;	/* cancel io */
				cp = bp->b_un.b_addr;
			}
			(void) subyte((caddr_t)ucp++, (c = *cp++));
			nc++;
		} while(c&0377);
	}
	(void) suword((caddr_t)ap, 0);
	(void) suword((caddr_t)ucp, 0);
	setregs();
bad:
	if (bp)
		brelse(bp);
	if (bno)
		mfree(swapmap, ctod(clrnd((int) btoc(NCARGS))), bno);
	iput(ip);
}

/*
 * Read in and set up memory for executed file.
 * Zero return is normal;
 * non-zero means only the text is being replaced
 */
getxfile(ip, nargc)
register struct inode *ip;
{
	register sep;
	register size_t ts, ds, ss;
	register int overlay;
	int pagi = 0;

	/*
	 * read in first few bytes
	 * of file for segment
	 * sizes:
	 * ux_mag = 407/410/411/405
	 *  407 is plain executable
	 *  410 is RO text
	 *  411 is separated ID
	 *  405 is overlaid text
	 *  412 is demand paged plain executable (NOT IMPLEMENTED)
	 *  413 is demand paged RO text
	 */

	u.u_base = (caddr_t)&u.u_exdata;
	u.u_count = sizeof(u.u_exdata);
	u.u_offset = 0;
	u.u_segflg = 1;
	readi(ip);
	u.u_segflg = 0;
	if(u.u_error)
		goto bad;
	if (u.u_count!=0) {
		u.u_error = ENOEXEC;
		goto bad;
	}
	sep = 0;
	overlay = 0;
	switch (u.u_exdata.ux_mag) {

	case 0405:
		overlay++;
		break;

	case 0412:
		u.u_error = ENOEXEC;
		goto bad;

	case 0407:
		u.u_exdata.ux_dsize += u.u_exdata.ux_tsize;
		u.u_exdata.ux_tsize = 0;
		break;

	case 0413:
		pagi = SPAGI;
		/* fall into ... */

	case 0410:
		if (u.u_exdata.ux_tsize == 0) {
			u.u_error = ENOEXEC;
			goto bad;
		}
		break;

	case 0411:
		u.u_error = ENOEXEC;
		goto bad;

	default:
		u.u_error = ENOEXEC;
		goto bad;
	}
	if(u.u_exdata.ux_tsize!=0 && (ip->i_flag&ITEXT)==0 && ip->i_count!=1) {
		u.u_error = ETXTBSY;
		goto bad;
	}

	/*
	 * find text and data sizes
	 * try them out for possible
	 * exceed of max sizes
	 */

	ts = clrnd(btoc(u.u_exdata.ux_tsize));
	ds = clrnd(btoc((u.u_exdata.ux_dsize+u.u_exdata.ux_bsize)));
	ss = clrnd(SSIZE + btoc(nargc));
	if (overlay) {
		if ((u.u_procp->p_flag & SPAGI) || u.u_sep==0 && ctos(ts) != ctos(u.u_tsize) || nargc) {
			u.u_error = ENOMEM;
			goto bad;
		}
		ds = u.u_dsize;
		ss = u.u_ssize;
		sep = u.u_sep;
		xfree();
		xalloc(ip, pagi);
		u.u_ar0[PC] = u.u_exdata.ux_entloc + 2; /* skip over entry mask */
	} else {
		if (chksize(ts, ds, ss))
			goto bad;
		u.u_cdmap = zdmap;
		u.u_csmap = zdmap;
		if (swpexpand(ds, ss, &u.u_cdmap, &u.u_csmap) == NULL)
			goto bad;

		/*
		 * At this point, committed to the new image!
		 * Release virtual memory resources of old process, and
		 * initialize the virtual memory of the new process.
		 * If we resulted from vfork(), instead wakeup our
		 * parent who will set SVFDONE when he has taken back
		 * our resources.
		 */
		u.u_prof.pr_scale = 0;
		if ((u.u_procp->p_flag & SVFORK) == 0)
			vrelvm();
		else {
			u.u_procp->p_flag &= ~SVFORK;
			u.u_procp->p_flag |= SKEEP;
			wakeup((caddr_t)u.u_procp);
			while ((u.u_procp->p_flag & SVFDONE) == 0)
				sleep((caddr_t)u.u_procp, PZERO - 1);
			u.u_procp->p_flag &= ~(SVFDONE|SKEEP);
		}
		u.u_procp->p_flag &= ~(SPAGI|SANOM|SUANOM);
		u.u_procp->p_flag |= pagi;
		u.u_dmap = u.u_cdmap;
		u.u_smap = u.u_csmap;
		vgetvm(ts, ds, ss);

		if (pagi == 0) {
			/*
			 * Read in data segment.
			 */
			u.u_base = (char *)ctob(ts);
			u.u_offset = sizeof(u.u_exdata)+u.u_exdata.ux_tsize;
			u.u_count = u.u_exdata.ux_dsize;
			readi(ip);
		}
		xalloc(ip, pagi);
		if (pagi && u.u_procp->p_textp)
			vinifod((struct fpte *)dptopte(u.u_procp, 0),
			    PG_FTEXT, u.u_procp->p_textp->x_iptr,
			    1 + ts/CLSIZE, (int)btoc(u.u_exdata.ux_dsize));

		/* THIS SHOULD BE DONE AT A LOWER LEVEL, IF AT ALL */
		mtpr(TBIA,1);

		/*
		 * set SUID/SGID protections, if no tracing
		 */
		if ((u.u_procp->p_flag&STRC)==0) {
			if(ip->i_mode&ISUID)
				if(u.u_uid != 0) {
					u.u_uid = ip->i_uid;
					u.u_procp->p_uid = ip->i_uid;
				}
			if(ip->i_mode&ISGID)
				u.u_gid = ip->i_gid;
		} else
			psignal(u.u_procp, SIGTRAP);
	}
	u.u_tsize = ts;
	u.u_dsize = ds;
	u.u_ssize = ss;
	u.u_sep = sep;
bad:
	return(overlay);
}

/*
 * Clear registers on exec
 */
setregs()
{
	register int (**rp)();
	register i;
	long sigmask;

	for(rp = &u.u_signal[0], sigmask = 1L; rp < &u.u_signal[NSIG];
	    sigmask <<= 1, rp++) {
		switch (*rp) {

		case SIG_IGN:
		case SIG_DFL:
		case SIG_HOLD:
			continue;

		default:
			/*
			 * Normal or deferring catch.
			 * If deferred, now hold it, else
			 * revert to default.
			 */
			if (SIGISDEFER(*rp))
				*rp = SIG_HOLD;
			else
				*rp = SIG_DFL;
			if ((int)*rp & 1)
				u.u_procp->p_siga0 |= sigmask;
			else
				u.u_procp->p_siga1 &= ~sigmask;
			if ((int)*rp & 2)
				u.u_procp->p_siga1 |= sigmask;
			else
				u.u_procp->p_siga1 &= ~sigmask;
			continue;
		}
	}
/*
	for(rp = &u.u_ar0[0]; rp < &u.u_ar0[16];)
		*rp++ = 0;
*/
	u.u_ar0[PC] = u.u_exdata.ux_entloc + 2; /* skip over entry mask */
	for(i=0; i<NOFILE; i++) {
		if (u.u_pofile[i]&EXCLOSE) {
			closef(u.u_ofile[i]);
			u.u_ofile[i] = NULL;
			u.u_pofile[i] &= ~EXCLOSE;
		}
	}
	/*
	 * Remember file name for accounting.
	 */
	u.u_acflag &= ~AFORK;
	bcopy((caddr_t)u.u_dbuf, (caddr_t)u.u_comm, DIRSIZ);
}

/*
 * exit system call:
 * pass back caller's arg
 */
rexit()
{
	register struct a {
		int	rval;
	} *uap;

	uap = (struct a *)u.u_ap;
	exit((uap->rval & 0377) << 8);
}

/*
 * Release resources.
 * Save u. area for parent to look at.
 * Enter zombie state.
 * Wake up parent and init processes,
 * and dispose of children.
 */
exit(rv)
{
	register int i;
	register struct proc *p, *q;
	register struct file *f;
	register int x;

#ifdef PGINPROF
	vmsizmon();
#endif
	p = u.u_procp;
	p->p_flag &= ~(STRC|SULOCK);
	p->p_flag |= SWEXIT;
	p->p_clktim = 0;
	(void) spl6();
	if ((int)SIG_IGN & 1)
		p->p_siga0 = ~0;
	else
		p->p_siga0 = 0;
	if ((int)SIG_IGN & 2)
		p->p_siga1 = ~0;
	else
		p->p_siga0 = 0;
	(void) spl0();
	rate.v_pgin -= p->p_aveflt;
	p->p_aveflt = 0;
	for(i=0; i<NSIG; i++)
		u.u_signal[i] = SIG_IGN;
	/*
	 * Release virtual memory.  If we resulted from
	 * a vfork(), instead give the resources back to
	 * the parent.
	 */
	if ((p->p_flag & SVFORK) == 0)
		vrelvm();
	else {
		p->p_flag &= ~SVFORK;
		wakeup((caddr_t)p);
		while ((p->p_flag & SVFDONE) == 0)
			sleep((caddr_t)p, PZERO - 1);
		p->p_flag &= ~SVFDONE;
	}
	for(i=0; i<NOFILE; i++) {
		f = u.u_ofile[i];
		u.u_ofile[i] = NULL;
		closef(f);
	}
	plock(u.u_cdir);
	iput(u.u_cdir);
	if (u.u_rdir) {
		plock(u.u_rdir);
		iput(u.u_rdir);
	}
	acct();
	vrelpt(u.u_procp);
	vrelu(u.u_procp, 0);
	multprog--;
	spl7();			/* clock will get mad because of overlaying */
	p->p_stat = SZOMB;
	i = PIDHASH(p->p_pid);
	x = p - proc;
	if (pidhash[i] == x)
		pidhash[i] = p->p_idhash;
	else {
		for (i = pidhash[i]; i != 0; i = proc[i].p_idhash)
			if (proc[i].p_idhash == x) {
				proc[i].p_idhash = p->p_idhash;
				goto done;
			}
		panic("exit");
	}
done:
	((struct xproc *)p)->xp_xstat = rv;		/* overlay */
	((struct xproc *)p)->xp_vm = u.u_vm;		/* overlay */
	vmsadd(&((struct xproc *)p)->xp_vm, &u.u_cvm);
	for(q = &proc[0]; q < &proc[NPROC]; q++)
		if(q->p_pptr == p) {
			q->p_pptr = &proc[1];
			q->p_ppid = 1;
			wakeup((caddr_t)&proc[1]);
			/*
			 * Stopped or traced processes are killed
			 * since their existence means someone is screwing up.
			 */
			if (q->p_stat == SSTOP || q->p_flag&STRC) {
				q->p_flag &= ~STRC;
				psignal(q, SIGKILL);
			}
		}
	wakeup((caddr_t)p->p_pptr);
	psignal(p->p_pptr, SIGCHLD);
	swtch();
}

wait()
{
	struct vtimes vm;
	struct vtimes *vp;

	if ((u.u_ar0[PS] & PSL_ALLCC) != PSL_ALLCC) {
		wait1(0, (struct vtimes *)0);
		return;
	}
	vp = (struct vtimes *)u.u_ar0[R1];
	wait1(u.u_ar0[R0], &vm);
	if (u.u_error)
		return;
	(void) copyout((caddr_t)&vm, (caddr_t)vp, sizeof (struct vtimes));
}

/*
 * Wait system call.
 * Search for a terminated (zombie) child,
 * finally lay it to rest, and collect its status.
 * Look also for stopped (traced) children,
 * and pass back status from them.
 */
wait1(options, vp)
	register options;
	struct vtimes *vp;
{
	register f;
	register struct proc *p;

	f = 0;
loop:
	for(p = &proc[0]; p < &proc[NPROC]; p++)
	if(p->p_pptr == u.u_procp) {
		f++;
		if(p->p_stat == SZOMB) {
			u.u_r.r_val1 = p->p_pid;
			u.u_r.r_val2 = ((struct xproc *)p)->xp_xstat;
			((struct xproc *)p)->xp_xstat = 0;
			if (vp)
				*vp = ((struct xproc *)p)->xp_vm;
			vmsadd(&u.u_cvm, &((struct xproc *)p)->xp_vm);
			((struct xproc *)p)->xp_vm = zvms;
			p->p_stat = NULL;
			p->p_pid = 0;
			p->p_ppid = 0;
			p->p_pptr = 0;
			p->p_sig = 0;
			p->p_siga0 = 0;
			p->p_siga1 = 0;
			p->p_pgrp = 0;
			p->p_flag = 0;
			p->p_wchan = 0;
			p->p_cursig = 0;
			return;
		}
		if (p->p_stat == SSTOP && (p->p_flag&SWTED)==0 &&
		    (p->p_flag&STRC || options&WUNTRACED)) {
			p->p_flag |= SWTED;
			u.u_r.r_val1 = p->p_pid;
			u.u_r.r_val2 = (p->p_cursig<<8) | WSTOPPED;
			return;
		}
	}
	if (f==0) {
		u.u_error = ECHILD;
		return;
	}
	if (options&WNOHANG) {
		u.u_r.r_val1 = 0;
		return;
	}
	if (setjmp(u.u_qsav)) {
		u.u_eosys = RESTARTSYS;
		return;
	}
	sleep((caddr_t)u.u_procp, PWAIT);
	goto loop;
}

/*
 * 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);
}

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

	a = 0;
	p2 = NULL;
	for(p1 = &proc[0]; p1 < &proc[NPROC]; p1++) {
		if (p1->p_stat==NULL && p2==NULL)
			p2 = p1;
		else {
			if (p1->p_uid==u.u_uid && p1->p_stat!=NULL)
				a++;
		}
	}
	/*
	 * Disallow if
	 *  No processes at all;
	 *  not su and too many procs owned; or
	 *  not su and would take last slot.
	 */
	if (p2==NULL || (u.u_uid!=0 && (p2==&proc[NPROC-1] || a>MAXUPRC))) {
		u.u_error = EAGAIN;
		if (!isvfork) {
			(void) vsexpand(0, &u.u_cdmap, 1);
			(void) vsexpand(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;
}

/*
 * break system call.
 *  -- bad planning: "break" is a dirty word in C.
 */
sbreak()
{
	struct a {
		char	*nsiz;
	};
	register int n, d;

	/*
	 * set n to new data size
	 * set d to new-old
	 */

	n = btoc(((struct a *)u.u_ap)->nsiz);
	if (!u.u_sep)
		n -= ctos(u.u_tsize) * stoc(1);
	if (n < 0)
		n = 0;
	d = clrnd(n - u.u_dsize);
	if (chksize(u.u_tsize, u.u_dsize+d, u.u_ssize))
		return;
	if (swpexpand(u.u_dsize+d, u.u_ssize, &u.u_dmap, &u.u_smap)==0)
		return;
	expand(d, P0BR);
}
Commit	Line	Data
bdfe5b0f	1	/* kern_proc.c 3.5 %H% */
29dd101b BJ	2
	3	#include "../h/param.h"
	4	#include "../h/systm.h"
	5	#include "../h/map.h"
	6	#include "../h/mtpr.h"
	7	#include "../h/dir.h"
	8	#include "../h/user.h"
	9	#include "../h/proc.h"
	10	#include "../h/buf.h"
	11	#include "../h/reg.h"
	12	#include "../h/inode.h"
	13	#include "../h/seg.h"
	14	#include "../h/acct.h"
bdfe5b0f	15	#include <wait.h>
29dd101b BJ	16	#include "../h/pte.h"
	17	#include "../h/vm.h"
	18	#include "../h/text.h"
bdfe5b0f	19	#include "../h/psl.h"
29dd101b BJ	20
	21	/*
	22	* exec system call, with and without environments.
	23	*/
	24	struct execa {
	25	char *fname;
	26	char **argp;
	27	char **envp;
	28	};
	29
	30	exec()
	31	{
	32	((struct execa *)u.u_ap)->envp = NULL;
	33	exece();
	34	}
	35
	36	exece()
	37	{
	38	register nc;
	39	register char *cp;
	40	register struct buf *bp;
	41	register struct execa *uap;
	42	int na, ne, ucp, ap, c;
	43	struct inode *ip;
	44	swblk_t bno;
	45
	46	if ((ip = namei(uchar, 0)) == NULL)
	47	return;
	48	bno = 0;
	49	bp = 0;
	50	if(access(ip, IEXEC))
	51	goto bad;
	52	if((ip->i_mode & IFMT) != IFREG \|\|
	53	(ip->i_mode & (IEXEC\|(IEXEC>>3)\|(IEXEC>>6))) == 0) {
	54	u.u_error = EACCES;
	55	goto bad;
	56	}
	57	/*
	58	* Collect arguments on "file" in swap space.
	59	*/
	60	na = 0;
	61	ne = 0;
	62	nc = 0;
	63	uap = (struct execa *)u.u_ap;
	64	if ((bno = malloc(swapmap, ctod(clrnd((int) btoc(NCARGS))))) == 0) {
	65	swkill(u.u_procp, "exece");
	66	goto bad;
	67	}
	68	if (bno % CLSIZE)
	69	panic("execa malloc");
	70	if (uap->argp) for (;;) {
	71	ap = NULL;
	72	if (uap->argp) {
	73	ap = fuword((caddr_t)uap->argp);
	74	uap->argp++;
	75	}
	76	if (ap==NULL && uap->envp) {
	77	uap->argp = NULL;
	78	if ((ap = fuword((caddr_t)uap->envp)) == NULL)
	79	break;
	80	uap->envp++;
	81	ne++;
	82	}
	83	if (ap==NULL)
84	break;
85	na++;
86	if(ap == -1)
87	u.u_error = EFAULT;
88	do {
89	if (nc >= NCARGS-1)
90	u.u_error = E2BIG;
91	if ((c = fubyte((caddr_t)ap++)) < 0)
92	u.u_error = EFAULT;
64d6118b BJ	93	if (u.u_error) {
	94	if (bp)
	95	brelse(bp);
	96	bp = 0;
29dd101b	97	goto badarg;
64d6118b	98	}
29dd101b BJ	99	if ((nc&BMASK) == 0) {
	100	if (bp)
	101	bdwrite(bp);
	102	bp = getblk(swapdev, (daddr_t)(dbtofsb(swplo+bno)+(nc>>BSHIFT)));
	103	cp = bp->b_un.b_addr;
	104	}
	105	nc++;
	106	*cp++ = c;
	107	} while (c>0);
	108	}
	109	if (bp)
	110	bdwrite(bp);
	111	bp = 0;
	112	nc = (nc + NBPW-1) & ~(NBPW-1);
	113	if (getxfile(ip, nc) \|\| u.u_error) {
	114	badarg:
	115	for (c = 0; c < nc; c += BSIZE)
	116	if (bp = baddr(swapdev, dbtofsb(swplo+bno)+(c>>BSHIFT))) {
	117	bp->b_flags \|= B_AGE; /* throw away */
	118	bp->b_flags &= ~B_DELWRI; /* cancel io */
	119	brelse(bp);
	120	bp = 0;
	121	}
	122	goto bad;
	123	}
	124
	125	/*
	126	* copy back arglist
	127	*/
	128
	129	ucp = USRSTACK - nc - NBPW;
	130	ap = ucp - naNBPW - 3NBPW;
	131	u.u_ar0[SP] = ap;
81263dba	132	(void) suword((caddr_t)ap, na-ne);
29dd101b BJ	133	nc = 0;
	134	for (;;) {
	135	ap += NBPW;
	136	if (na==ne) {
81263dba	137	(void) suword((caddr_t)ap, 0);
29dd101b BJ	138	ap += NBPW;
	139	}
	140	if (--na < 0)
	141	break;
81263dba	142	(void) suword((caddr_t)ap, ucp);
29dd101b BJ	143	do {
	144	if ((nc&BMASK) == 0) {
	145	if (bp)
	146	brelse(bp);
	147	bp = bread(swapdev, (daddr_t)(dbtofsb(swplo+bno)+(nc>>BSHIFT)));
	148	bp->b_flags \|= B_AGE; /* throw away */
	149	bp->b_flags &= ~B_DELWRI; /* cancel io */
	150	cp = bp->b_un.b_addr;
	151	}
81263dba	152	(void) subyte((caddr_t)ucp++, (c = *cp++));
29dd101b BJ	153	nc++;
	154	} while(c&0377);
	155	}
81263dba BJ	156	(void) suword((caddr_t)ap, 0);
81263dba BJ	157	(void) suword((caddr_t)ucp, 0);
29dd101b BJ	158	setregs();
	159	bad:
	160	if (bp)
	161	brelse(bp);
	162	if (bno)
	163	mfree(swapmap, ctod(clrnd((int) btoc(NCARGS))), bno);
	164	iput(ip);
	165	}
	166
	167	/*
	168	* Read in and set up memory for executed file.
	169	* Zero return is normal;
	170	* non-zero means only the text is being replaced
	171	*/
	172	getxfile(ip, nargc)
	173	register struct inode *ip;
	174	{
	175	register sep;
	176	register size_t ts, ds, ss;
	177	register int overlay;
	178	int pagi = 0;
	179
	180	/*
	181	* read in first few bytes
	182	* of file for segment
	183	* sizes:
	184	* ux_mag = 407/410/411/405
	185	* 407 is plain executable
	186	* 410 is RO text
	187	* 411 is separated ID
	188	* 405 is overlaid text
	189	* 412 is demand paged plain executable (NOT IMPLEMENTED)
	190	* 413 is demand paged RO text
	191	*/
	192
	193	u.u_base = (caddr_t)&u.u_exdata;
	194	u.u_count = sizeof(u.u_exdata);
	195	u.u_offset = 0;
	196	u.u_segflg = 1;
	197	readi(ip);
	198	u.u_segflg = 0;
	199	if(u.u_error)
	200	goto bad;
	201	if (u.u_count!=0) {
	202	u.u_error = ENOEXEC;
	203	goto bad;
	204	}
	205	sep = 0;
	206	overlay = 0;
	207	switch (u.u_exdata.ux_mag) {
	208
	209	case 0405:
	210	overlay++;
	211	break;
	212
	213	case 0412:
	214	u.u_error = ENOEXEC;
	215	goto bad;
	216
	217	case 0407:
	218	u.u_exdata.ux_dsize += u.u_exdata.ux_tsize;
	219	u.u_exdata.ux_tsize = 0;
	220	break;
	221
222	case 0413:
223	pagi = SPAGI;
224	/* fall into ... */
225
226	case 0410:
227	if (u.u_exdata.ux_tsize == 0) {
228	u.u_error = ENOEXEC;
229	goto bad;
230	}
231	break;
232
233	case 0411:
234	u.u_error = ENOEXEC;
235	goto bad;
236
237	default:
238	u.u_error = ENOEXEC;
239	goto bad;
240	}
241	if(u.u_exdata.ux_tsize!=0 && (ip->i_flag&ITEXT)==0 && ip->i_count!=1) {
242	u.u_error = ETXTBSY;
243	goto bad;
244	}
245
246	/*
247	* find text and data sizes
248	* try them out for possible
249	* exceed of max sizes
250	*/
251
252	ts = clrnd(btoc(u.u_exdata.ux_tsize));
253	ds = clrnd(btoc((u.u_exdata.ux_dsize+u.u_exdata.ux_bsize)));
254	ss = clrnd(SSIZE + btoc(nargc));
255	if (overlay) {
256	if ((u.u_procp->p_flag & SPAGI) \|\| u.u_sep==0 && ctos(ts) != ctos(u.u_tsize) \|\| nargc) {
257	u.u_error = ENOMEM;
258	goto bad;
259	}
260	ds = u.u_dsize;
261	ss = u.u_ssize;
262	sep = u.u_sep;
263	xfree();
264	xalloc(ip, pagi);
265	u.u_ar0[PC] = u.u_exdata.ux_entloc + 2; /* skip over entry mask */
266	} else {
267	if (chksize(ts, ds, ss))
268	goto bad;
269	u.u_cdmap = zdmap;
270	u.u_csmap = zdmap;
271	if (swpexpand(ds, ss, &u.u_cdmap, &u.u_csmap) == NULL)
272	goto bad;
273
274	/*
275	* At this point, committed to the new image!
276	* Release virtual memory resources of old process, and
277	* initialize the virtual memory of the new process.
278	* If we resulted from vfork(), instead wakeup our
279	* parent who will set SVFDONE when he has taken back
280	* our resources.
281	*/
282	u.u_prof.pr_scale = 0;
283	if ((u.u_procp->p_flag & SVFORK) == 0)
284	vrelvm();
285	else {
286	u.u_procp->p_flag &= ~SVFORK;
287	u.u_procp->p_flag \|= SKEEP;
288	wakeup((caddr_t)u.u_procp);
289	while ((u.u_procp->p_flag & SVFDONE) == 0)
290	sleep((caddr_t)u.u_procp, PZERO - 1);
291	u.u_procp->p_flag &= ~(SVFDONE\|SKEEP);
292	}
293	u.u_procp->p_flag &= ~(SPAGI\|SANOM\|SUANOM);
294	u.u_procp->p_flag \|= pagi;
295	u.u_dmap = u.u_cdmap;
296	u.u_smap = u.u_csmap;
297	vgetvm(ts, ds, ss);
298
299	if (pagi == 0) {
300	/*
301	* Read in data segment.
302	*/
303	u.u_base = (char *)ctob(ts);
304	u.u_offset = sizeof(u.u_exdata)+u.u_exdata.ux_tsize;
305	u.u_count = u.u_exdata.ux_dsize;
306	readi(ip);
307	}
308	xalloc(ip, pagi);
309	if (pagi && u.u_procp->p_textp)
310	vinifod((struct fpte *)dptopte(u.u_procp, 0),
311	PG_FTEXT, u.u_procp->p_textp->x_iptr,
312	1 + ts/CLSIZE, (int)btoc(u.u_exdata.ux_dsize));
313
314	/* THIS SHOULD BE DONE AT A LOWER LEVEL, IF AT ALL */
315	mtpr(TBIA,1);
316
317	/*
318	* set SUID/SGID protections, if no tracing
319	*/
320	if ((u.u_procp->p_flag&STRC)==0) {
321	if(ip->i_mode&ISUID)
322	if(u.u_uid != 0) {
323	u.u_uid = ip->i_uid;
324	u.u_procp->p_uid = ip->i_uid;
325	}
326	if(ip->i_mode&ISGID)
327	u.u_gid = ip->i_gid;
328	} else
594ebedd	329	psignal(u.u_procp, SIGTRAP);
29dd101b BJ	330	}
	331	u.u_tsize = ts;
	332	u.u_dsize = ds;
	333	u.u_ssize = ss;
	334	u.u_sep = sep;
	335	bad:
	336	return(overlay);
	337	}
	338
	339	/*
	340	* Clear registers on exec
	341	*/
	342	setregs()
	343	{
594ebedd	344	register int (**rp)();
29dd101b	345	register i;
bdfe5b0f	346	long sigmask;
29dd101b	347
bdfe5b0f BJ	348	for(rp = &u.u_signal[0], sigmask = 1L; rp < &u.u_signal[NSIG];
	349	sigmask <<= 1, rp++) {
	350	switch (*rp) {
	351
	352	case SIG_IGN:
	353	case SIG_DFL:
	354	case SIG_HOLD:
	355	continue;
	356
	357	default:
	358	/*
	359	* Normal or deferring catch.
	360	* If deferred, now hold it, else
	361	* revert to default.
	362	*/
	363	if (SIGISDEFER(*rp))
	364	*rp = SIG_HOLD;
	365	else
	366	*rp = SIG_DFL;
	367	if ((int)*rp & 1)
	368	u.u_procp->p_siga0 \|= sigmask;
	369	else
	370	u.u_procp->p_siga1 &= ~sigmask;
	371	if ((int)*rp & 2)
	372	u.u_procp->p_siga1 \|= sigmask;
	373	else
	374	u.u_procp->p_siga1 &= ~sigmask;
	375	continue;
	376	}
	377	}
29dd101b BJ	378	/*
	379	for(rp = &u.u_ar0[0]; rp < &u.u_ar0[16];)
	380	*rp++ = 0;
	381	*/
	382	u.u_ar0[PC] = u.u_exdata.ux_entloc + 2; /* skip over entry mask */
	383	for(i=0; i<NOFILE; i++) {
	384	if (u.u_pofile[i]&EXCLOSE) {
	385	closef(u.u_ofile[i]);
	386	u.u_ofile[i] = NULL;
bdfe5b0f	387	u.u_pofile[i] &= ~EXCLOSE;
29dd101b	388	}
29dd101b BJ	389	}
	390	/*
	391	* Remember file name for accounting.
	392	*/
	393	u.u_acflag &= ~AFORK;
	394	bcopy((caddr_t)u.u_dbuf, (caddr_t)u.u_comm, DIRSIZ);
	395	}
	396
	397	/*
	398	* exit system call:
	399	* pass back caller's arg
	400	*/
	401	rexit()
	402	{
	403	register struct a {
	404	int rval;
	405	} *uap;
	406
	407	uap = (struct a *)u.u_ap;
	408	exit((uap->rval & 0377) << 8);
	409	}
	410
	411	/*
	412	* Release resources.
	413	* Save u. area for parent to look at.
	414	* Enter zombie state.
	415	* Wake up parent and init processes,
	416	* and dispose of children.
	417	*/
	418	exit(rv)
	419	{
	420	register int i;
	421	register struct proc p, q;
	422	register struct file *f;
	423	register int x;
	424
	425	#ifdef PGINPROF
	426	vmsizmon();
	427	#endif
	428	p = u.u_procp;
	429	p->p_flag &= ~(STRC\|SULOCK);
	430	p->p_flag \|= SWEXIT;
	431	p->p_clktim = 0;
bdfe5b0f BJ	432	(void) spl6();
	433	if ((int)SIG_IGN & 1)
	434	p->p_siga0 = ~0;
	435	else
	436	p->p_siga0 = 0;
	437	if ((int)SIG_IGN & 2)
	438	p->p_siga1 = ~0;
	439	else
	440	p->p_siga0 = 0;
	441	(void) spl0();
29dd101b BJ	442	rate.v_pgin -= p->p_aveflt;
	443	p->p_aveflt = 0;
	444	for(i=0; i<NSIG; i++)
594ebedd	445	u.u_signal[i] = SIG_IGN;
29dd101b BJ	446	/*
	447	* Release virtual memory. If we resulted from
	448	* a vfork(), instead give the resources back to
	449	* the parent.
	450	*/
	451	if ((p->p_flag & SVFORK) == 0)
	452	vrelvm();
	453	else {
	454	p->p_flag &= ~SVFORK;
	455	wakeup((caddr_t)p);
	456	while ((p->p_flag & SVFDONE) == 0)
	457	sleep((caddr_t)p, PZERO - 1);
	458	p->p_flag &= ~SVFDONE;
	459	}
	460	for(i=0; i<NOFILE; i++) {
	461	f = u.u_ofile[i];
	462	u.u_ofile[i] = NULL;
	463	closef(f);
	464	}
	465	plock(u.u_cdir);
	466	iput(u.u_cdir);
	467	if (u.u_rdir) {
	468	plock(u.u_rdir);
	469	iput(u.u_rdir);
	470	}
	471	acct();
	472	vrelpt(u.u_procp);
	473	vrelu(u.u_procp, 0);
	474	multprog--;
	475	spl7(); /* clock will get mad because of overlaying */
	476	p->p_stat = SZOMB;
	477	i = PIDHASH(p->p_pid);
	478	x = p - proc;
	479	if (pidhash[i] == x)
	480	pidhash[i] = p->p_idhash;
	481	else {
	482	for (i = pidhash[i]; i != 0; i = proc[i].p_idhash)
	483	if (proc[i].p_idhash == x) {
	484	proc[i].p_idhash = p->p_idhash;
	485	goto done;
	486	}
	487	panic("exit");
	488	}
	489	done:
	490	((struct xproc )p)->xp_xstat = rv; / overlay */
	491	((struct xproc )p)->xp_vm = u.u_vm; / overlay */
	492	vmsadd(&((struct xproc *)p)->xp_vm, &u.u_cvm);
	493	for(q = &proc[0]; q < &proc[NPROC]; q++)
bdfe5b0f BJ	494	if(q->p_pptr == p) {
bdfe5b0f BJ	495	q->p_pptr = &proc[1];
29dd101b	496	q->p_ppid = 1;
bdfe5b0f BJ	497	wakeup((caddr_t)&proc[1]);
	498	/*
	499	* Stopped or traced processes are killed
	500	* since their existence means someone is screwing up.
	501	*/
	502	if (q->p_stat == SSTOP \|\| q->p_flag&STRC) {
	503	q->p_flag &= ~STRC;
	504	psignal(q, SIGKILL);
	505	}
29dd101b	506	}
bdfe5b0f BJ	507	wakeup((caddr_t)p->p_pptr);
bdfe5b0f BJ	508	psignal(p->p_pptr, SIGCHLD);
29dd101b BJ	509	swtch();
	510	}
	511
	512	wait()
	513	{
bdfe5b0f BJ	514	struct vtimes vm;
bdfe5b0f BJ	515	struct vtimes *vp;
29dd101b	516
bdfe5b0f BJ	517	if ((u.u_ar0[PS] & PSL_ALLCC) != PSL_ALLCC) {
	518	wait1(0, (struct vtimes *)0);
	519	return;
	520	}
	521	vp = (struct vtimes *)u.u_ar0[R1];
	522	wait1(u.u_ar0[R0], &vm);
	523	if (u.u_error)
	524	return;
	525	(void) copyout((caddr_t)&vm, (caddr_t)vp, sizeof (struct vtimes));
29dd101b BJ	526	}
	527
	528	/*
	529	* Wait system call.
	530	* Search for a terminated (zombie) child,
	531	* finally lay it to rest, and collect its status.
	532	* Look also for stopped (traced) children,
	533	* and pass back status from them.
	534	*/
bdfe5b0f BJ	535	wait1(options, vp)
bdfe5b0f BJ	536	register options;
29dd101b BJ	537	struct vtimes *vp;
	538	{
	539	register f;
	540	register struct proc *p;
	541
	542	f = 0;
29dd101b BJ	543	loop:
29dd101b BJ	544	for(p = &proc[0]; p < &proc[NPROC]; p++)
bdfe5b0f	545	if(p->p_pptr == u.u_procp) {
29dd101b BJ	546	f++;
	547	if(p->p_stat == SZOMB) {
	548	u.u_r.r_val1 = p->p_pid;
	549	u.u_r.r_val2 = ((struct xproc *)p)->xp_xstat;
	550	((struct xproc *)p)->xp_xstat = 0;
	551	if (vp)
	552	vp = ((struct xproc )p)->xp_vm;
	553	vmsadd(&u.u_cvm, &((struct xproc *)p)->xp_vm);
	554	((struct xproc *)p)->xp_vm = zvms;
	555	p->p_stat = NULL;
	556	p->p_pid = 0;
	557	p->p_ppid = 0;
bdfe5b0f	558	p->p_pptr = 0;
29dd101b	559	p->p_sig = 0;
bdfe5b0f BJ	560	p->p_siga0 = 0;
bdfe5b0f BJ	561	p->p_siga1 = 0;
29dd101b BJ	562	p->p_pgrp = 0;
	563	p->p_flag = 0;
	564	p->p_wchan = 0;
bdfe5b0f	565	p->p_cursig = 0;
29dd101b BJ	566	return;
29dd101b BJ	567	}
bdfe5b0f BJ	568	if (p->p_stat == SSTOP && (p->p_flag&SWTED)==0 &&
	569	(p->p_flag&STRC \|\| options&WUNTRACED)) {
	570	p->p_flag \|= SWTED;
	571	u.u_r.r_val1 = p->p_pid;
	572	u.u_r.r_val2 = (p->p_cursig<<8) \| WSTOPPED;
	573	return;
29dd101b BJ	574	}
29dd101b BJ	575	}
bdfe5b0f BJ	576	if (f==0) {
	577	u.u_error = ECHILD;
	578	return;
	579	}
	580	if (options&WNOHANG) {
	581	u.u_r.r_val1 = 0;
	582	return;
	583	}
	584	if (setjmp(u.u_qsav)) {
	585	u.u_eosys = RESTARTSYS;
	586	return;
29dd101b	587	}
bdfe5b0f BJ	588	sleep((caddr_t)u.u_procp, PWAIT);
bdfe5b0f BJ	589	goto loop;
29dd101b BJ	590	}
	591
	592	/*
	593	* fork system call.
	594	*/
	595	fork()
	596	{
	597
	598	u.u_cdmap = zdmap;
	599	u.u_csmap = zdmap;
	600	if (swpexpand(u.u_dsize, u.u_ssize, &u.u_cdmap, &u.u_csmap) == 0) {
	601	u.u_r.r_val2 = 0;
	602	return;
	603	}
	604	fork1(0);
	605	}
	606
	607	fork1(isvfork)
	608	{
	609	register struct proc p1, p2;
	610	register a;
	611
	612	a = 0;
	613	p2 = NULL;
	614	for(p1 = &proc[0]; p1 < &proc[NPROC]; p1++) {
	615	if (p1->p_stat==NULL && p2==NULL)
	616	p2 = p1;
	617	else {
	618	if (p1->p_uid==u.u_uid && p1->p_stat!=NULL)
	619	a++;
	620	}
	621	}
	622	/*
	623	* Disallow if
	624	* No processes at all;
	625	* not su and too many procs owned; or
	626	* not su and would take last slot.
	627	*/
	628	if (p2==NULL \|\| (u.u_uid!=0 && (p2==&proc[NPROC-1] \|\| a>MAXUPRC))) {
	629	u.u_error = EAGAIN;
	630	if (!isvfork) {
81263dba BJ	631	(void) vsexpand(0, &u.u_cdmap, 1);
81263dba BJ	632	(void) vsexpand(0, &u.u_csmap, 1);
29dd101b BJ	633	}
	634	goto out;
	635	}
	636	p1 = u.u_procp;
	637	if(newproc(isvfork)) {
	638	u.u_r.r_val1 = p1->p_pid;
	639	u.u_r.r_val2 = 1; /* child */
	640	u.u_start = time;
	641	u.u_acflag = AFORK;
	642	return;
	643	}
	644	u.u_r.r_val1 = p2->p_pid;
	645
	646	out:
	647	u.u_r.r_val2 = 0;
	648	}
	649
	650	/*
	651	* break system call.
	652	* -- bad planning: "break" is a dirty word in C.
	653	*/
	654	sbreak()
	655	{
	656	struct a {
	657	char *nsiz;
	658	};
	659	register int n, d;
	660
	661	/*
	662	* set n to new data size
	663	* set d to new-old
	664	*/
	665
	666	n = btoc(((struct a *)u.u_ap)->nsiz);
	667	if (!u.u_sep)
	668	n -= ctos(u.u_tsize) * stoc(1);
	669	if (n < 0)
	670	n = 0;
	671	d = clrnd(n - u.u_dsize);
	672	if (chksize(u.u_tsize, u.u_dsize+d, u.u_ssize))
	673	return;
	674	if (swpexpand(u.u_dsize+d, u.u_ssize, &u.u_dmap, &u.u_smap)==0)
	675	return;
	676	expand(d, P0BR);
	677	}