Home | History | Annotate | Line # | Download | only in kern
kern_descrip.c revision 1.26
      1 /*	$NetBSD: kern_descrip.c,v 1.26 1994/10/30 21:47:38 cgd 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_descrip.c	8.6 (Berkeley) 4/19/94
     41  */
     42 
     43 #include <sys/param.h>
     44 #include <sys/systm.h>
     45 #include <sys/filedesc.h>
     46 #include <sys/kernel.h>
     47 #include <sys/vnode.h>
     48 #include <sys/proc.h>
     49 #include <sys/file.h>
     50 #include <sys/socket.h>
     51 #include <sys/socketvar.h>
     52 #include <sys/stat.h>
     53 #include <sys/ioctl.h>
     54 #include <sys/fcntl.h>
     55 #include <sys/malloc.h>
     56 #include <sys/syslog.h>
     57 #include <sys/unistd.h>
     58 #include <sys/resourcevar.h>
     59 
     60 #include <sys/mount.h>
     61 #include <sys/syscallargs.h>
     62 
     63 /*
     64  * Descriptor management.
     65  */
     66 struct filelist filehead;	/* head of list of open files */
     67 int nfiles;			/* actual number of open files */
     68 
     69 /*
     70  * System calls on descriptors.
     71  */
     72 
     73 #if defined(COMPAT_43) || defined(COMPAT_SUNOS) || defined(COMPAT_ULTRIX) || defined(COMPAT_HPUX)
     74 /* ARGSUSED */
     75 compat_43_getdtablesize(p, uap, retval)
     76 	struct proc *p;
     77 	void *uap;
     78 	register_t *retval;
     79 {
     80 
     81 	*retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
     82 	return (0);
     83 }
     84 #endif
     85 
     86 /*
     87  * Duplicate a file descriptor.
     88  */
     89 /* ARGSUSED */
     90 dup(p, uap, retval)
     91 	struct proc *p;
     92 	struct dup_args /* {
     93 		syscallarg(u_int) fd;
     94 	} */ *uap;
     95 	register_t *retval;
     96 {
     97 	register struct filedesc *fdp;
     98 	u_int old;
     99 	int new, error;
    100 
    101 	old = SCARG(uap, fd);
    102 	/*
    103 	 * XXX Compatibility
    104 	 */
    105 	if (old &~ 077) {
    106 		SCARG(uap, fd) &= 077;
    107 		return (dup2(p, uap, retval));
    108 	}
    109 
    110 	fdp = p->p_fd;
    111 	if (old >= fdp->fd_nfiles || fdp->fd_ofiles[old] == NULL)
    112 		return (EBADF);
    113 	if (error = fdalloc(p, 0, &new))
    114 		return (error);
    115 	return (finishdup(fdp, (int)old, new, retval));
    116 }
    117 
    118 /*
    119  * Duplicate a file descriptor to a particular value.
    120  */
    121 /* ARGSUSED */
    122 dup2(p, uap, retval)
    123 	struct proc *p;
    124 	struct dup2_args /* {
    125 		syscallarg(u_int) from;
    126 		syscallarg(u_int) to;
    127 	} */ *uap;
    128 	register_t *retval;
    129 {
    130 	register struct filedesc *fdp = p->p_fd;
    131 	register u_int old = SCARG(uap, from), new = SCARG(uap, to);
    132 	int i, error;
    133 
    134 	if (old >= fdp->fd_nfiles ||
    135 	    fdp->fd_ofiles[old] == NULL ||
    136 	    new >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
    137 	    new >= maxfiles)
    138 		return (EBADF);
    139 	if (old == new) {
    140 		*retval = new;
    141 		return (0);
    142 	}
    143 	if (new >= fdp->fd_nfiles) {
    144 		if (error = fdalloc(p, new, &i))
    145 			return (error);
    146 		if (new != i)
    147 			panic("dup2: fdalloc");
    148 	} else if (fdp->fd_ofiles[new]) {
    149 		if (fdp->fd_ofileflags[new] & UF_MAPPED)
    150 			(void) munmapfd(p, new);
    151 		/*
    152 		 * dup2() must succeed even if the close has an error.
    153 		 */
    154 		(void) closef(fdp->fd_ofiles[new], p);
    155 	}
    156 	return (finishdup(fdp, (int)old, (int)new, retval));
    157 }
    158 
    159 /*
    160  * The file control system call.
    161  */
    162 /* ARGSUSED */
    163 fcntl(p, uap, retval)
    164 	struct proc *p;
    165 	register struct fcntl_args /* {
    166 		syscallarg(int) fd;
    167 		syscallarg(int) cmd;
    168 		syscallarg(void *) arg;
    169 	} */ *uap;
    170 	register_t *retval;
    171 {
    172 	register struct filedesc *fdp = p->p_fd;
    173 	register struct file *fp;
    174 	register char *pop;
    175 	struct vnode *vp;
    176 	int i, tmp, error, flg = F_POSIX;
    177 	struct flock fl;
    178 	u_long newmin;
    179 
    180 	if ((unsigned)SCARG(uap, fd) >= fdp->fd_nfiles ||
    181 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
    182 		return (EBADF);
    183 	pop = &fdp->fd_ofileflags[SCARG(uap, fd)];
    184 	switch (SCARG(uap, cmd)) {
    185 
    186 	case F_DUPFD:
    187 		newmin = (u_long)SCARG(uap, arg);
    188 		if (newmin >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
    189 		    newmin >= maxfiles)
    190 			return (EINVAL);
    191 		if (error = fdalloc(p, newmin, &i))
    192 			return (error);
    193 		return (finishdup(fdp, SCARG(uap, fd), i, retval));
    194 
    195 	case F_GETFD:
    196 		*retval = *pop & 1;
    197 		return (0);
    198 
    199 	case F_SETFD:
    200 		*pop = (*pop &~ 1) | ((long)SCARG(uap, arg) & 1);
    201 		return (0);
    202 
    203 	case F_GETFL:
    204 		*retval = OFLAGS(fp->f_flag);
    205 		return (0);
    206 
    207 	case F_SETFL:
    208 		fp->f_flag &= ~FCNTLFLAGS;
    209 		fp->f_flag |= FFLAGS((long)SCARG(uap, arg)) & FCNTLFLAGS;
    210 		tmp = fp->f_flag & FNONBLOCK;
    211 		error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p);
    212 		if (error)
    213 			return (error);
    214 		tmp = fp->f_flag & FASYNC;
    215 		error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p);
    216 		if (!error)
    217 			return (0);
    218 		fp->f_flag &= ~FNONBLOCK;
    219 		tmp = 0;
    220 		(void) (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p);
    221 		return (error);
    222 
    223 	case F_GETOWN:
    224 		if (fp->f_type == DTYPE_SOCKET) {
    225 			*retval = ((struct socket *)fp->f_data)->so_pgid;
    226 			return (0);
    227 		}
    228 		error = (*fp->f_ops->fo_ioctl)
    229 			(fp, (int)TIOCGPGRP, (caddr_t)retval, p);
    230 		*retval = -*retval;
    231 		return (error);
    232 
    233 	case F_SETOWN:
    234 		if (fp->f_type == DTYPE_SOCKET) {
    235 			((struct socket *)fp->f_data)->so_pgid =
    236 			    (long)SCARG(uap, arg);
    237 			return (0);
    238 		}
    239 		if ((long)SCARG(uap, arg) <= 0) {
    240 			SCARG(uap, arg) = (void *)(-(long)SCARG(uap, arg));
    241 		} else {
    242 			struct proc *p1 = pfind((long)SCARG(uap, arg));
    243 			if (p1 == 0)
    244 				return (ESRCH);
    245 			SCARG(uap, arg) = (void *)(long)p1->p_pgrp->pg_id;
    246 		}
    247 		return ((*fp->f_ops->fo_ioctl)
    248 			(fp, (int)TIOCSPGRP, (caddr_t)&SCARG(uap, arg), p));
    249 
    250 	case F_SETLKW:
    251 		flg |= F_WAIT;
    252 		/* Fall into F_SETLK */
    253 
    254 	case F_SETLK:
    255 		if (fp->f_type != DTYPE_VNODE)
    256 			return (EBADF);
    257 		vp = (struct vnode *)fp->f_data;
    258 		/* Copy in the lock structure */
    259 		error = copyin((caddr_t)SCARG(uap, arg), (caddr_t)&fl,
    260 		    sizeof (fl));
    261 		if (error)
    262 			return (error);
    263 		if (fl.l_whence == SEEK_CUR)
    264 			fl.l_start += fp->f_offset;
    265 		switch (fl.l_type) {
    266 
    267 		case F_RDLCK:
    268 			if ((fp->f_flag & FREAD) == 0)
    269 				return (EBADF);
    270 			p->p_flag |= P_ADVLOCK;
    271 			return (VOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg));
    272 
    273 		case F_WRLCK:
    274 			if ((fp->f_flag & FWRITE) == 0)
    275 				return (EBADF);
    276 			p->p_flag |= P_ADVLOCK;
    277 			return (VOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg));
    278 
    279 		case F_UNLCK:
    280 			return (VOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &fl,
    281 				F_POSIX));
    282 
    283 		default:
    284 			return (EINVAL);
    285 		}
    286 
    287 	case F_GETLK:
    288 		if (fp->f_type != DTYPE_VNODE)
    289 			return (EBADF);
    290 		vp = (struct vnode *)fp->f_data;
    291 		/* Copy in the lock structure */
    292 		error = copyin((caddr_t)SCARG(uap, arg), (caddr_t)&fl,
    293 		    sizeof (fl));
    294 		if (error)
    295 			return (error);
    296 		if (fl.l_whence == SEEK_CUR)
    297 			fl.l_start += fp->f_offset;
    298 		if (error = VOP_ADVLOCK(vp, (caddr_t)p, F_GETLK, &fl, F_POSIX))
    299 			return (error);
    300 		return (copyout((caddr_t)&fl, (caddr_t)SCARG(uap, arg),
    301 		    sizeof (fl)));
    302 
    303 	default:
    304 		return (EINVAL);
    305 	}
    306 	/* NOTREACHED */
    307 }
    308 
    309 /*
    310  * Common code for dup, dup2, and fcntl(F_DUPFD).
    311  */
    312 int
    313 finishdup(fdp, old, new, retval)
    314 	register struct filedesc *fdp;
    315 	register int old, new;
    316 	register_t *retval;
    317 {
    318 	register struct file *fp;
    319 
    320 	fp = fdp->fd_ofiles[old];
    321 	fdp->fd_ofiles[new] = fp;
    322 	fdp->fd_ofileflags[new] = fdp->fd_ofileflags[old] &~ UF_EXCLOSE;
    323 	fp->f_count++;
    324 	if (new > fdp->fd_lastfile)
    325 		fdp->fd_lastfile = new;
    326 	*retval = new;
    327 	return (0);
    328 }
    329 
    330 /*
    331  * Close a file descriptor.
    332  */
    333 /* ARGSUSED */
    334 close(p, uap, retval)
    335 	struct proc *p;
    336 	struct close_args /* {
    337 		syscallarg(int) fd;
    338 	} */ *uap;
    339 	register_t *retval;
    340 {
    341 	register struct filedesc *fdp = p->p_fd;
    342 	register struct file *fp;
    343 	register int fd = SCARG(uap, fd);
    344 	register u_char *pf;
    345 
    346 	if ((unsigned)fd >= fdp->fd_nfiles ||
    347 	    (fp = fdp->fd_ofiles[fd]) == NULL)
    348 		return (EBADF);
    349 	pf = (u_char *)&fdp->fd_ofileflags[fd];
    350 	if (*pf & UF_MAPPED)
    351 		(void) munmapfd(p, fd);
    352 	fdp->fd_ofiles[fd] = NULL;
    353 	while (fdp->fd_lastfile > 0 && fdp->fd_ofiles[fdp->fd_lastfile] == NULL)
    354 		fdp->fd_lastfile--;
    355 	if (fd < fdp->fd_freefile)
    356 		fdp->fd_freefile = fd;
    357 	*pf = 0;
    358 	return (closef(fp, p));
    359 }
    360 
    361 #if defined(COMPAT_43) || defined(COMPAT_SUNOS) || defined(COMPAT_IBCS2)
    362 /*
    363  * Return status information about a file descriptor.
    364  */
    365 /* ARGSUSED */
    366 compat_43_fstat(p, uap, retval)
    367 	struct proc *p;
    368 	register struct compat_43_fstat_args /* {
    369 		syscallarg(int) fd;
    370 		syscallarg(struct ostat *) sb;
    371 	} */ *uap;
    372 	register_t *retval;
    373 {
    374 	register struct filedesc *fdp = p->p_fd;
    375 	register struct file *fp;
    376 	struct stat ub;
    377 	struct ostat oub;
    378 	int error;
    379 
    380 	if ((unsigned)SCARG(uap, fd) >= fdp->fd_nfiles ||
    381 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
    382 		return (EBADF);
    383 	switch (fp->f_type) {
    384 
    385 	case DTYPE_VNODE:
    386 		error = vn_stat((struct vnode *)fp->f_data, &ub, p);
    387 		break;
    388 
    389 	case DTYPE_SOCKET:
    390 		error = soo_stat((struct socket *)fp->f_data, &ub);
    391 		break;
    392 
    393 	default:
    394 		panic("ofstat");
    395 		/*NOTREACHED*/
    396 	}
    397 	cvtstat(&ub, &oub);
    398 	if (error == 0)
    399 		error = copyout((caddr_t)&oub, (caddr_t)SCARG(uap, sb),
    400 		    sizeof (oub));
    401 	return (error);
    402 }
    403 #endif /* COMPAT_43 || COMPAT_SUNOS || COMPAT_IBCS2 */
    404 
    405 /*
    406  * Return status information about a file descriptor.
    407  */
    408 /* ARGSUSED */
    409 fstat(p, uap, retval)
    410 	struct proc *p;
    411 	register struct fstat_args /* {
    412 		syscallarg(int) fd;
    413 		syscallarg(struct stat *) sb;
    414 	} */ *uap;
    415 	register_t *retval;
    416 {
    417 	register struct filedesc *fdp = p->p_fd;
    418 	register struct file *fp;
    419 	struct stat ub;
    420 	int error;
    421 
    422 	if ((unsigned)SCARG(uap, fd) >= fdp->fd_nfiles ||
    423 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
    424 		return (EBADF);
    425 	switch (fp->f_type) {
    426 
    427 	case DTYPE_VNODE:
    428 		error = vn_stat((struct vnode *)fp->f_data, &ub, p);
    429 		break;
    430 
    431 	case DTYPE_SOCKET:
    432 		error = soo_stat((struct socket *)fp->f_data, &ub);
    433 		break;
    434 
    435 	default:
    436 		panic("fstat");
    437 		/*NOTREACHED*/
    438 	}
    439 	if (error == 0)
    440 		error = copyout((caddr_t)&ub, (caddr_t)SCARG(uap, sb),
    441 		    sizeof (ub));
    442 	return (error);
    443 }
    444 
    445 /*
    446  * Return pathconf information about a file descriptor.
    447  */
    448 /* ARGSUSED */
    449 fpathconf(p, uap, retval)
    450 	struct proc *p;
    451 	register struct fpathconf_args /* {
    452 		syscallarg(int) fd;
    453 		syscallarg(int) name;
    454 	} */ *uap;
    455 	register_t *retval;
    456 {
    457 	struct filedesc *fdp = p->p_fd;
    458 	struct file *fp;
    459 	struct vnode *vp;
    460 
    461 	if ((unsigned)SCARG(uap, fd) >= fdp->fd_nfiles ||
    462 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
    463 		return (EBADF);
    464 	switch (fp->f_type) {
    465 
    466 	case DTYPE_SOCKET:
    467 		if (SCARG(uap, name) != _PC_PIPE_BUF)
    468 			return (EINVAL);
    469 		*retval = PIPE_BUF;
    470 		return (0);
    471 
    472 	case DTYPE_VNODE:
    473 		vp = (struct vnode *)fp->f_data;
    474 #ifdef notyet
    475 		return (VOP_PATHCONF(vp, SCARG(uap, name), retval));
    476 #else
    477 		return (ENOSYS);
    478 #endif
    479 
    480 	default:
    481 		panic("fpathconf");
    482 	}
    483 	/*NOTREACHED*/
    484 }
    485 
    486 /*
    487  * Allocate a file descriptor for the process.
    488  */
    489 int fdexpand;
    490 
    491 fdalloc(p, want, result)
    492 	struct proc *p;
    493 	int want;
    494 	int *result;
    495 {
    496 	register struct filedesc *fdp = p->p_fd;
    497 	register int i;
    498 	int lim, last, nfiles;
    499 	struct file **newofile;
    500 	char *newofileflags;
    501 
    502 	/*
    503 	 * Search for a free descriptor starting at the higher
    504 	 * of want or fd_freefile.  If that fails, consider
    505 	 * expanding the ofile array.
    506 	 */
    507 	lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
    508 	for (;;) {
    509 		last = min(fdp->fd_nfiles, lim);
    510 		if ((i = want) < fdp->fd_freefile)
    511 			i = fdp->fd_freefile;
    512 		for (; i < last; i++) {
    513 			if (fdp->fd_ofiles[i] == NULL) {
    514 				fdp->fd_ofileflags[i] = 0;
    515 				if (i > fdp->fd_lastfile)
    516 					fdp->fd_lastfile = i;
    517 				if (want <= fdp->fd_freefile)
    518 					fdp->fd_freefile = i;
    519 				*result = i;
    520 				return (0);
    521 			}
    522 		}
    523 
    524 		/*
    525 		 * No space in current array.  Expand?
    526 		 */
    527 		if (fdp->fd_nfiles >= lim)
    528 			return (EMFILE);
    529 		if (fdp->fd_nfiles < NDEXTENT)
    530 			nfiles = NDEXTENT;
    531 		else
    532 			nfiles = 2 * fdp->fd_nfiles;
    533 		MALLOC(newofile, struct file **, nfiles * OFILESIZE,
    534 		    M_FILEDESC, M_WAITOK);
    535 		newofileflags = (char *) &newofile[nfiles];
    536 		/*
    537 		 * Copy the existing ofile and ofileflags arrays
    538 		 * and zero the new portion of each array.
    539 		 */
    540 		bcopy(fdp->fd_ofiles, newofile,
    541 			(i = sizeof(struct file *) * fdp->fd_nfiles));
    542 		bzero((char *)newofile + i, nfiles * sizeof(struct file *) - i);
    543 		bcopy(fdp->fd_ofileflags, newofileflags,
    544 			(i = sizeof(char) * fdp->fd_nfiles));
    545 		bzero(newofileflags + i, nfiles * sizeof(char) - i);
    546 		if (fdp->fd_nfiles > NDFILE)
    547 			FREE(fdp->fd_ofiles, M_FILEDESC);
    548 		fdp->fd_ofiles = newofile;
    549 		fdp->fd_ofileflags = newofileflags;
    550 		fdp->fd_nfiles = nfiles;
    551 		fdexpand++;
    552 	}
    553 }
    554 
    555 /*
    556  * Check to see whether n user file descriptors
    557  * are available to the process p.
    558  */
    559 fdavail(p, n)
    560 	struct proc *p;
    561 	register int n;
    562 {
    563 	register struct filedesc *fdp = p->p_fd;
    564 	register struct file **fpp;
    565 	register int i, lim;
    566 
    567 	lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
    568 	if ((i = lim - fdp->fd_nfiles) > 0 && (n -= i) <= 0)
    569 		return (1);
    570 	fpp = &fdp->fd_ofiles[fdp->fd_freefile];
    571 	for (i = fdp->fd_nfiles - fdp->fd_freefile; --i >= 0; fpp++)
    572 		if (*fpp == NULL && --n <= 0)
    573 			return (1);
    574 	return (0);
    575 }
    576 
    577 /*
    578  * Create a new open file structure and allocate
    579  * a file decriptor for the process that refers to it.
    580  */
    581 falloc(p, resultfp, resultfd)
    582 	register struct proc *p;
    583 	struct file **resultfp;
    584 	int *resultfd;
    585 {
    586 	register struct file *fp, *fq;
    587 	int error, i;
    588 
    589 	if (error = fdalloc(p, 0, &i))
    590 		return (error);
    591 	if (nfiles >= maxfiles) {
    592 		tablefull("file");
    593 		return (ENFILE);
    594 	}
    595 	/*
    596 	 * Allocate a new file descriptor.
    597 	 * If the process has file descriptor zero open, add to the list
    598 	 * of open files at that point, otherwise put it at the front of
    599 	 * the list of open files.
    600 	 */
    601 	nfiles++;
    602 	MALLOC(fp, struct file *, sizeof(struct file), M_FILE, M_WAITOK);
    603 	bzero(fp, sizeof(struct file));
    604 	if (fq = p->p_fd->fd_ofiles[0]) {
    605 		LIST_INSERT_AFTER(fq, fp, f_list);
    606 	} else {
    607 		LIST_INSERT_HEAD(&filehead, fp, f_list);
    608 	}
    609 	p->p_fd->fd_ofiles[i] = fp;
    610 	fp->f_count = 1;
    611 	fp->f_cred = p->p_ucred;
    612 	crhold(fp->f_cred);
    613 	if (resultfp)
    614 		*resultfp = fp;
    615 	if (resultfd)
    616 		*resultfd = i;
    617 	return (0);
    618 }
    619 
    620 /*
    621  * Free a file descriptor.
    622  */
    623 ffree(fp)
    624 	register struct file *fp;
    625 {
    626 	register struct file *fq;
    627 
    628 	LIST_REMOVE(fp, f_list);
    629 	crfree(fp->f_cred);
    630 #ifdef DIAGNOSTIC
    631 	fp->f_count = 0;
    632 #endif
    633 	nfiles--;
    634 	FREE(fp, M_FILE);
    635 }
    636 
    637 /*
    638  * Copy a filedesc structure.
    639  */
    640 struct filedesc *
    641 fdcopy(p)
    642 	struct proc *p;
    643 {
    644 	register struct filedesc *newfdp, *fdp = p->p_fd;
    645 	register struct file **fpp;
    646 	register int i;
    647 
    648 	MALLOC(newfdp, struct filedesc *, sizeof(struct filedesc0),
    649 	    M_FILEDESC, M_WAITOK);
    650 	bcopy(fdp, newfdp, sizeof(struct filedesc));
    651 	VREF(newfdp->fd_cdir);
    652 	if (newfdp->fd_rdir)
    653 		VREF(newfdp->fd_rdir);
    654 	newfdp->fd_refcnt = 1;
    655 
    656 	/*
    657 	 * If the number of open files fits in the internal arrays
    658 	 * of the open file structure, use them, otherwise allocate
    659 	 * additional memory for the number of descriptors currently
    660 	 * in use.
    661 	 */
    662 	if (newfdp->fd_lastfile < NDFILE) {
    663 		newfdp->fd_ofiles = ((struct filedesc0 *) newfdp)->fd_dfiles;
    664 		newfdp->fd_ofileflags =
    665 		    ((struct filedesc0 *) newfdp)->fd_dfileflags;
    666 		i = NDFILE;
    667 	} else {
    668 		/*
    669 		 * Compute the smallest multiple of NDEXTENT needed
    670 		 * for the file descriptors currently in use,
    671 		 * allowing the table to shrink.
    672 		 */
    673 		i = newfdp->fd_nfiles;
    674 		while (i > 2 * NDEXTENT && i > newfdp->fd_lastfile * 2)
    675 			i /= 2;
    676 		MALLOC(newfdp->fd_ofiles, struct file **, i * OFILESIZE,
    677 		    M_FILEDESC, M_WAITOK);
    678 		newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
    679 	}
    680 	newfdp->fd_nfiles = i;
    681 	bcopy(fdp->fd_ofiles, newfdp->fd_ofiles, i * sizeof(struct file **));
    682 	bcopy(fdp->fd_ofileflags, newfdp->fd_ofileflags, i * sizeof(char));
    683 	fpp = newfdp->fd_ofiles;
    684 	for (i = newfdp->fd_lastfile; i-- >= 0; fpp++)
    685 		if (*fpp != NULL)
    686 			(*fpp)->f_count++;
    687 	return (newfdp);
    688 }
    689 
    690 /*
    691  * Release a filedesc structure.
    692  */
    693 void
    694 fdfree(p)
    695 	struct proc *p;
    696 {
    697 	register struct filedesc *fdp = p->p_fd;
    698 	struct file **fpp;
    699 	register int i;
    700 
    701 	if (--fdp->fd_refcnt > 0)
    702 		return;
    703 	fpp = fdp->fd_ofiles;
    704 	for (i = fdp->fd_lastfile; i-- >= 0; fpp++)
    705 		if (*fpp)
    706 			(void) closef(*fpp, p);
    707 	if (fdp->fd_nfiles > NDFILE)
    708 		FREE(fdp->fd_ofiles, M_FILEDESC);
    709 	vrele(fdp->fd_cdir);
    710 	if (fdp->fd_rdir)
    711 		vrele(fdp->fd_rdir);
    712 	FREE(fdp, M_FILEDESC);
    713 }
    714 
    715 /*
    716  * Close any files on exec?
    717  */
    718 void
    719 fdcloseexec(p)
    720 	struct proc *p;
    721 {
    722 	struct filedesc *fdp = p->p_fd;
    723 	struct file **fpp;
    724 	char *fdfp;
    725 	register int i;
    726 
    727 	fpp = fdp->fd_ofiles;
    728 	fdfp = fdp->fd_ofileflags;
    729 	for (i = 0; i <= fdp->fd_lastfile; i++, fpp++, fdfp++)
    730 		if (*fpp != NULL && (*fdfp & UF_EXCLOSE)) {
    731 			if (*fdfp & UF_MAPPED)
    732 				(void) munmapfd(p, i);
    733 			(void) closef(*fpp, p);
    734 			*fpp = NULL;
    735 			*fdfp = 0;
    736 			if (i < fdp->fd_freefile)
    737 				fdp->fd_freefile = i;
    738 		}
    739 	while (fdp->fd_lastfile > 0 && fdp->fd_ofiles[fdp->fd_lastfile] == NULL)
    740 		fdp->fd_lastfile--;
    741 }
    742 
    743 /*
    744  * Internal form of close.
    745  * Decrement reference count on file structure.
    746  * Note: p may be NULL when closing a file
    747  * that was being passed in a message.
    748  */
    749 closef(fp, p)
    750 	register struct file *fp;
    751 	register struct proc *p;
    752 {
    753 	struct vnode *vp;
    754 	struct flock lf;
    755 	int error;
    756 
    757 	if (fp == NULL)
    758 		return (0);
    759 	/*
    760 	 * POSIX record locking dictates that any close releases ALL
    761 	 * locks owned by this process.  This is handled by setting
    762 	 * a flag in the unlock to free ONLY locks obeying POSIX
    763 	 * semantics, and not to free BSD-style file locks.
    764 	 * If the descriptor was in a message, POSIX-style locks
    765 	 * aren't passed with the descriptor.
    766 	 */
    767 	if (p && (p->p_flag & P_ADVLOCK) && fp->f_type == DTYPE_VNODE) {
    768 		lf.l_whence = SEEK_SET;
    769 		lf.l_start = 0;
    770 		lf.l_len = 0;
    771 		lf.l_type = F_UNLCK;
    772 		vp = (struct vnode *)fp->f_data;
    773 		(void) VOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_POSIX);
    774 	}
    775 	if (--fp->f_count > 0)
    776 		return (0);
    777 	if (fp->f_count < 0)
    778 		panic("closef: count < 0");
    779 	if ((fp->f_flag & FHASLOCK) && fp->f_type == DTYPE_VNODE) {
    780 		lf.l_whence = SEEK_SET;
    781 		lf.l_start = 0;
    782 		lf.l_len = 0;
    783 		lf.l_type = F_UNLCK;
    784 		vp = (struct vnode *)fp->f_data;
    785 		(void) VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK);
    786 	}
    787 	if (fp->f_ops)
    788 		error = (*fp->f_ops->fo_close)(fp, p);
    789 	else
    790 		error = 0;
    791 	ffree(fp);
    792 	return (error);
    793 }
    794 
    795 /*
    796  * Apply an advisory lock on a file descriptor.
    797  *
    798  * Just attempt to get a record lock of the requested type on
    799  * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
    800  */
    801 /* ARGSUSED */
    802 flock(p, uap, retval)
    803 	struct proc *p;
    804 	register struct flock_args /* {
    805 		syscallarg(int) fd;
    806 		syscallarg(int) how;
    807 	} */ *uap;
    808 	register_t *retval;
    809 {
    810 	register struct filedesc *fdp = p->p_fd;
    811 	register struct file *fp;
    812 	struct vnode *vp;
    813 	struct flock lf;
    814 
    815 	if ((unsigned)SCARG(uap, fd) >= fdp->fd_nfiles ||
    816 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
    817 		return (EBADF);
    818 	if (fp->f_type != DTYPE_VNODE)
    819 		return (EOPNOTSUPP);
    820 	vp = (struct vnode *)fp->f_data;
    821 	lf.l_whence = SEEK_SET;
    822 	lf.l_start = 0;
    823 	lf.l_len = 0;
    824 	if (SCARG(uap, how) & LOCK_UN) {
    825 		lf.l_type = F_UNLCK;
    826 		fp->f_flag &= ~FHASLOCK;
    827 		return (VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK));
    828 	}
    829 	if (SCARG(uap, how) & LOCK_EX)
    830 		lf.l_type = F_WRLCK;
    831 	else if (SCARG(uap, how) & LOCK_SH)
    832 		lf.l_type = F_RDLCK;
    833 	else
    834 		return (EBADF);
    835 	fp->f_flag |= FHASLOCK;
    836 	if (SCARG(uap, how) & LOCK_NB)
    837 		return (VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK));
    838 	return (VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK|F_WAIT));
    839 }
    840 
    841 /*
    842  * File Descriptor pseudo-device driver (/dev/fd/).
    843  *
    844  * Opening minor device N dup()s the file (if any) connected to file
    845  * descriptor N belonging to the calling process.  Note that this driver
    846  * consists of only the ``open()'' routine, because all subsequent
    847  * references to this file will be direct to the other driver.
    848  */
    849 /* ARGSUSED */
    850 fdopen(dev, mode, type, p)
    851 	dev_t dev;
    852 	int mode, type;
    853 	struct proc *p;
    854 {
    855 
    856 	/*
    857 	 * XXX Kludge: set curproc->p_dupfd to contain the value of the
    858 	 * the file descriptor being sought for duplication. The error
    859 	 * return ensures that the vnode for this device will be released
    860 	 * by vn_open. Open will detect this special error and take the
    861 	 * actions in dupfdopen below. Other callers of vn_open or VOP_OPEN
    862 	 * will simply report the error.
    863 	 */
    864 	p->p_dupfd = minor(dev);
    865 	return (ENODEV);
    866 }
    867 
    868 /*
    869  * Duplicate the specified descriptor to a free descriptor.
    870  */
    871 dupfdopen(fdp, indx, dfd, mode, error)
    872 	register struct filedesc *fdp;
    873 	register int indx, dfd;
    874 	int mode;
    875 	int error;
    876 {
    877 	register struct file *wfp;
    878 	struct file *fp;
    879 
    880 	/*
    881 	 * If the to-be-dup'd fd number is greater than the allowed number
    882 	 * of file descriptors, or the fd to be dup'd has already been
    883 	 * closed, reject.  Note, check for new == old is necessary as
    884 	 * falloc could allocate an already closed to-be-dup'd descriptor
    885 	 * as the new descriptor.
    886 	 */
    887 	fp = fdp->fd_ofiles[indx];
    888 	if ((u_int)dfd >= fdp->fd_nfiles ||
    889 	    (wfp = fdp->fd_ofiles[dfd]) == NULL || fp == wfp)
    890 		return (EBADF);
    891 
    892 	/*
    893 	 * There are two cases of interest here.
    894 	 *
    895 	 * For ENODEV simply dup (dfd) to file descriptor
    896 	 * (indx) and return.
    897 	 *
    898 	 * For ENXIO steal away the file structure from (dfd) and
    899 	 * store it in (indx).  (dfd) is effectively closed by
    900 	 * this operation.
    901 	 *
    902 	 * Any other error code is just returned.
    903 	 */
    904 	switch (error) {
    905 	case ENODEV:
    906 		/*
    907 		 * Check that the mode the file is being opened for is a
    908 		 * subset of the mode of the existing descriptor.
    909 		 */
    910 		if (((mode & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag)
    911 			return (EACCES);
    912 		fdp->fd_ofiles[indx] = wfp;
    913 		fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd];
    914 		wfp->f_count++;
    915 		if (indx > fdp->fd_lastfile)
    916 			fdp->fd_lastfile = indx;
    917 		return (0);
    918 
    919 	case ENXIO:
    920 		/*
    921 		 * Steal away the file pointer from dfd, and stuff it into indx.
    922 		 */
    923 		fdp->fd_ofiles[indx] = fdp->fd_ofiles[dfd];
    924 		fdp->fd_ofiles[dfd] = NULL;
    925 		fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd];
    926 		fdp->fd_ofileflags[dfd] = 0;
    927 		/*
    928 		 * Complete the clean up of the filedesc structure by
    929 		 * recomputing the various hints.
    930 		 */
    931 		if (indx > fdp->fd_lastfile)
    932 			fdp->fd_lastfile = indx;
    933 		else
    934 			while (fdp->fd_lastfile > 0 &&
    935 			       fdp->fd_ofiles[fdp->fd_lastfile] == NULL)
    936 				fdp->fd_lastfile--;
    937 			if (dfd < fdp->fd_freefile)
    938 				fdp->fd_freefile = dfd;
    939 		return (0);
    940 
    941 	default:
    942 		return (error);
    943 	}
    944 	/* NOTREACHED */
    945 }
    946