Home | History | Annotate | Line # | Download | only in kern
kern_descrip.c revision 1.1.1.3
      1 /*
      2  * Copyright (c) 1982, 1986, 1989, 1991, 1993
      3  *	The Regents of the University of California.  All rights reserved.
      4  * (c) UNIX System Laboratories, Inc.
      5  * All or some portions of this file are derived from material licensed
      6  * to the University of California by American Telephone and Telegraph
      7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
      8  * the permission of UNIX System Laboratories, Inc.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  * 3. All advertising materials mentioning features or use of this software
     19  *    must display the following acknowledgement:
     20  *	This product includes software developed by the University of
     21  *	California, Berkeley and its contributors.
     22  * 4. Neither the name of the University nor the names of its contributors
     23  *    may be used to endorse or promote products derived from this software
     24  *    without specific prior written permission.
     25  *
     26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     36  * SUCH DAMAGE.
     37  *
     38  *	@(#)kern_descrip.c	8.8 (Berkeley) 2/14/95
     39  */
     40 
     41 #include <sys/param.h>
     42 #include <sys/systm.h>
     43 #include <sys/filedesc.h>
     44 #include <sys/kernel.h>
     45 #include <sys/vnode.h>
     46 #include <sys/proc.h>
     47 #include <sys/file.h>
     48 #include <sys/socket.h>
     49 #include <sys/socketvar.h>
     50 #include <sys/stat.h>
     51 #include <sys/ioctl.h>
     52 #include <sys/fcntl.h>
     53 #include <sys/malloc.h>
     54 #include <sys/syslog.h>
     55 #include <sys/unistd.h>
     56 #include <sys/resourcevar.h>
     57 
     58 #include <sys/mount.h>
     59 #include <sys/syscallargs.h>
     60 
     61 /*
     62  * Descriptor management.
     63  */
     64 struct filelist filehead;	/* head of list of open files */
     65 int nfiles;			/* actual number of open files */
     66 
     67 /*
     68  * System calls on descriptors.
     69  */
     70 /* ARGSUSED */
     71 int
     72 getdtablesize(p, uap, retval)
     73 	struct proc *p;
     74 	void *uap;
     75 	register_t *retval;
     76 {
     77 
     78 	*retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
     79 	return (0);
     80 }
     81 
     82 /*
     83  * Duplicate a file descriptor.
     84  */
     85 /* ARGSUSED */
     86 int
     87 dup(p, uap, retval)
     88 	struct proc *p;
     89 	struct dup_args /* {
     90 		syscallarg(u_int) fd;
     91 	} */ *uap;
     92 	register_t *retval;
     93 {
     94 	register struct filedesc *fdp;
     95 	u_int old;
     96 	int new, error;
     97 
     98 	old = SCARG(uap, fd);
     99 	/*
    100 	 * XXX Compatibility
    101 	 */
    102 	if (old &~ 077) {
    103 		SCARG(uap, fd) &= 077;
    104 		return (dup2(p, uap, retval));
    105 	}
    106 
    107 	fdp = p->p_fd;
    108 	if (old >= fdp->fd_nfiles || fdp->fd_ofiles[old] == NULL)
    109 		return (EBADF);
    110 	if (error = fdalloc(p, 0, &new))
    111 		return (error);
    112 	return (finishdup(fdp, (int)old, new, retval));
    113 }
    114 
    115 /*
    116  * Duplicate a file descriptor to a particular value.
    117  */
    118 /* ARGSUSED */
    119 int
    120 dup2(p, uap, retval)
    121 	struct proc *p;
    122 	struct dup2_args /* {
    123 		syscallarg(u_int) from;
    124 		syscallarg(u_int) to;
    125 	} */ *uap;
    126 	register_t *retval;
    127 {
    128 	register struct filedesc *fdp = p->p_fd;
    129 	register int old = SCARG(uap, from), new = SCARG(uap, to);
    130 	int i, error;
    131 
    132 	if (old >= fdp->fd_nfiles ||
    133 	    fdp->fd_ofiles[old] == NULL ||
    134 	    new >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
    135 	    new >= maxfiles)
    136 		return (EBADF);
    137 	if (old == new) {
    138 		*retval = new;
    139 		return (0);
    140 	}
    141 	if (new >= fdp->fd_nfiles) {
    142 		if (error = fdalloc(p, new, &i))
    143 			return (error);
    144 		if (new != i)
    145 			panic("dup2: fdalloc");
    146 	} else if (fdp->fd_ofiles[new]) {
    147 		if (fdp->fd_ofileflags[new] & UF_MAPPED)
    148 			(void) munmapfd(p, new);
    149 		/*
    150 		 * dup2() must succeed even if the close has an error.
    151 		 */
    152 		(void) closef(fdp->fd_ofiles[new], p);
    153 	}
    154 	return (finishdup(fdp, (int)old, (int)new, retval));
    155 }
    156 
    157 /*
    158  * The file control system call.
    159  */
    160 /* ARGSUSED */
    161 int
    162 fcntl(p, uap, retval)
    163 	struct proc *p;
    164 	register struct fcntl_args /* {
    165 		syscallarg(int) fd;
    166 		syscallarg(int) cmd;
    167 		syscallarg(void *) arg;
    168 	} */ *uap;
    169 	register_t *retval;
    170 {
    171 	int fd = SCARG(uap, fd);
    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_int newmin;
    179 
    180 	if ((u_int)fd >= fdp->fd_nfiles ||
    181 	    (fp = fdp->fd_ofiles[fd]) == NULL)
    182 		return (EBADF);
    183 	pop = &fdp->fd_ofileflags[fd];
    184 	switch (SCARG(uap, cmd)) {
    185 
    186 	case F_DUPFD:
    187 		newmin = (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, 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, 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, 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 int
    335 close(p, uap, retval)
    336 	struct proc *p;
    337 	struct close_args /* {
    338 		syscallarg(int) fd;
    339 	} */ *uap;
    340 	register_t *retval;
    341 {
    342 	int fd = SCARG(uap, fd);
    343 	register struct filedesc *fdp = p->p_fd;
    344 	register struct file *fp;
    345 	register u_char *pf;
    346 
    347 	if ((u_int)fd >= fdp->fd_nfiles ||
    348 	    (fp = fdp->fd_ofiles[fd]) == NULL)
    349 		return (EBADF);
    350 	pf = (u_char *)&fdp->fd_ofileflags[fd];
    351 	if (*pf & UF_MAPPED)
    352 		(void) munmapfd(p, fd);
    353 	fdp->fd_ofiles[fd] = NULL;
    354 	while (fdp->fd_lastfile > 0 && fdp->fd_ofiles[fdp->fd_lastfile] == NULL)
    355 		fdp->fd_lastfile--;
    356 	if (fd < fdp->fd_freefile)
    357 		fdp->fd_freefile = fd;
    358 	*pf = 0;
    359 	return (closef(fp, p));
    360 }
    361 
    362 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
    363 /*
    364  * Return status information about a file descriptor.
    365  */
    366 /* ARGSUSED */
    367 int
    368 compat_43_fstat(p, uap, retval)
    369 	struct proc *p;
    370 	register struct compat_43_fstat_args /* {
    371 		syscallarg(int) fd;
    372 		syscallarg(struct ostat *) sb;
    373 	} */ *uap;
    374 	register_t *retval;
    375 {
    376 	int fd = SCARG(uap, fd);
    377 	register struct filedesc *fdp = p->p_fd;
    378 	register struct file *fp;
    379 	struct stat ub;
    380 	struct ostat oub;
    381 	int error;
    382 
    383 	if ((u_int)fd >= fdp->fd_nfiles ||
    384 	    (fp = fdp->fd_ofiles[fd]) == NULL)
    385 		return (EBADF);
    386 	switch (fp->f_type) {
    387 
    388 	case DTYPE_VNODE:
    389 		error = vn_stat((struct vnode *)fp->f_data, &ub, p);
    390 		break;
    391 
    392 	case DTYPE_SOCKET:
    393 		error = soo_stat((struct socket *)fp->f_data, &ub);
    394 		break;
    395 
    396 	default:
    397 		panic("ofstat");
    398 		/*NOTREACHED*/
    399 	}
    400 	cvtstat(&ub, &oub);
    401 	if (error == 0)
    402 		error = copyout((caddr_t)&oub, (caddr_t)SCARG(uap, sb),
    403 		    sizeof (oub));
    404 	return (error);
    405 }
    406 #endif /* COMPAT_43 || COMPAT_SUNOS */
    407 
    408 /*
    409  * Return status information about a file descriptor.
    410  */
    411 /* ARGSUSED */
    412 int
    413 fstat(p, uap, retval)
    414 	struct proc *p;
    415 	register struct fstat_args /* {
    416 		syscallarg(int) fd;
    417 		syscallarg(struct stat *) sb;
    418 	} */ *uap;
    419 	register_t *retval;
    420 {
    421 	int fd = SCARG(uap, fd);
    422 	register struct filedesc *fdp = p->p_fd;
    423 	register struct file *fp;
    424 	struct stat ub;
    425 	int error;
    426 
    427 	if ((u_int)fd >= fdp->fd_nfiles ||
    428 	    (fp = fdp->fd_ofiles[fd]) == NULL)
    429 		return (EBADF);
    430 	switch (fp->f_type) {
    431 
    432 	case DTYPE_VNODE:
    433 		error = vn_stat((struct vnode *)fp->f_data, &ub, p);
    434 		break;
    435 
    436 	case DTYPE_SOCKET:
    437 		error = soo_stat((struct socket *)fp->f_data, &ub);
    438 		break;
    439 
    440 	default:
    441 		panic("fstat");
    442 		/*NOTREACHED*/
    443 	}
    444 	if (error == 0)
    445 		error = copyout((caddr_t)&ub, (caddr_t)SCARG(uap, sb),
    446 		    sizeof (ub));
    447 	return (error);
    448 }
    449 
    450 /*
    451  * Return pathconf information about a file descriptor.
    452  */
    453 /* ARGSUSED */
    454 int
    455 fpathconf(p, uap, retval)
    456 	struct proc *p;
    457 	register struct fpathconf_args /* {
    458 		syscallarg(int) fd;
    459 		syscallarg(int) name;
    460 	} */ *uap;
    461 	register_t *retval;
    462 {
    463 	int fd = SCARG(uap, fd);
    464 	struct filedesc *fdp = p->p_fd;
    465 	struct file *fp;
    466 	struct vnode *vp;
    467 
    468 	if ((u_int)fd >= fdp->fd_nfiles ||
    469 	    (fp = fdp->fd_ofiles[fd]) == NULL)
    470 		return (EBADF);
    471 	switch (fp->f_type) {
    472 
    473 	case DTYPE_SOCKET:
    474 		if (SCARG(uap, name) != _PC_PIPE_BUF)
    475 			return (EINVAL);
    476 		*retval = PIPE_BUF;
    477 		return (0);
    478 
    479 	case DTYPE_VNODE:
    480 		vp = (struct vnode *)fp->f_data;
    481 		return (VOP_PATHCONF(vp, SCARG(uap, name), retval));
    482 
    483 	default:
    484 		panic("fpathconf");
    485 	}
    486 	/*NOTREACHED*/
    487 }
    488 
    489 /*
    490  * Allocate a file descriptor for the process.
    491  */
    492 int fdexpand;
    493 
    494 int
    495 fdalloc(p, want, result)
    496 	struct proc *p;
    497 	int want;
    498 	int *result;
    499 {
    500 	register struct filedesc *fdp = p->p_fd;
    501 	register int i;
    502 	int lim, last, nfiles;
    503 	struct file **newofile;
    504 	char *newofileflags;
    505 
    506 	/*
    507 	 * Search for a free descriptor starting at the higher
    508 	 * of want or fd_freefile.  If that fails, consider
    509 	 * expanding the ofile array.
    510 	 */
    511 	lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
    512 	for (;;) {
    513 		last = min(fdp->fd_nfiles, lim);
    514 		if ((i = want) < fdp->fd_freefile)
    515 			i = fdp->fd_freefile;
    516 		for (; i < last; i++) {
    517 			if (fdp->fd_ofiles[i] == NULL) {
    518 				fdp->fd_ofileflags[i] = 0;
    519 				if (i > fdp->fd_lastfile)
    520 					fdp->fd_lastfile = i;
    521 				if (want <= fdp->fd_freefile)
    522 					fdp->fd_freefile = i;
    523 				*result = i;
    524 				return (0);
    525 			}
    526 		}
    527 
    528 		/*
    529 		 * No space in current array.  Expand?
    530 		 */
    531 		if (fdp->fd_nfiles >= lim)
    532 			return (EMFILE);
    533 		if (fdp->fd_nfiles < NDEXTENT)
    534 			nfiles = NDEXTENT;
    535 		else
    536 			nfiles = 2 * fdp->fd_nfiles;
    537 		MALLOC(newofile, struct file **, nfiles * OFILESIZE,
    538 		    M_FILEDESC, M_WAITOK);
    539 		newofileflags = (char *) &newofile[nfiles];
    540 		/*
    541 		 * Copy the existing ofile and ofileflags arrays
    542 		 * and zero the new portion of each array.
    543 		 */
    544 		bcopy(fdp->fd_ofiles, newofile,
    545 			(i = sizeof(struct file *) * fdp->fd_nfiles));
    546 		bzero((char *)newofile + i, nfiles * sizeof(struct file *) - i);
    547 		bcopy(fdp->fd_ofileflags, newofileflags,
    548 			(i = sizeof(char) * fdp->fd_nfiles));
    549 		bzero(newofileflags + i, nfiles * sizeof(char) - i);
    550 		if (fdp->fd_nfiles > NDFILE)
    551 			FREE(fdp->fd_ofiles, M_FILEDESC);
    552 		fdp->fd_ofiles = newofile;
    553 		fdp->fd_ofileflags = newofileflags;
    554 		fdp->fd_nfiles = nfiles;
    555 		fdexpand++;
    556 	}
    557 }
    558 
    559 /*
    560  * Check to see whether n user file descriptors
    561  * are available to the process p.
    562  */
    563 int
    564 fdavail(p, n)
    565 	struct proc *p;
    566 	register int n;
    567 {
    568 	register struct filedesc *fdp = p->p_fd;
    569 	register struct file **fpp;
    570 	register int i, lim;
    571 
    572 	lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
    573 	if ((i = lim - fdp->fd_nfiles) > 0 && (n -= i) <= 0)
    574 		return (1);
    575 	fpp = &fdp->fd_ofiles[fdp->fd_freefile];
    576 	for (i = fdp->fd_nfiles - fdp->fd_freefile; --i >= 0; fpp++)
    577 		if (*fpp == NULL && --n <= 0)
    578 			return (1);
    579 	return (0);
    580 }
    581 
    582 /*
    583  * Create a new open file structure and allocate
    584  * a file decriptor for the process that refers to it.
    585  */
    586 int
    587 falloc(p, resultfp, resultfd)
    588 	register struct proc *p;
    589 	struct file **resultfp;
    590 	int *resultfd;
    591 {
    592 	register struct file *fp, *fq;
    593 	int error, i;
    594 
    595 	if (error = fdalloc(p, 0, &i))
    596 		return (error);
    597 	if (nfiles >= maxfiles) {
    598 		tablefull("file");
    599 		return (ENFILE);
    600 	}
    601 	/*
    602 	 * Allocate a new file descriptor.
    603 	 * If the process has file descriptor zero open, add to the list
    604 	 * of open files at that point, otherwise put it at the front of
    605 	 * the list of open files.
    606 	 */
    607 	nfiles++;
    608 	MALLOC(fp, struct file *, sizeof(struct file), M_FILE, M_WAITOK);
    609 	bzero(fp, sizeof(struct file));
    610 	if (fq = p->p_fd->fd_ofiles[0]) {
    611 		LIST_INSERT_AFTER(fq, fp, f_list);
    612 	} else {
    613 		LIST_INSERT_HEAD(&filehead, fp, f_list);
    614 	}
    615 	p->p_fd->fd_ofiles[i] = fp;
    616 	fp->f_count = 1;
    617 	fp->f_cred = p->p_ucred;
    618 	crhold(fp->f_cred);
    619 	if (resultfp)
    620 		*resultfp = fp;
    621 	if (resultfd)
    622 		*resultfd = i;
    623 	return (0);
    624 }
    625 
    626 /*
    627  * Free a file descriptor.
    628  */
    629 void
    630 ffree(fp)
    631 	register struct file *fp;
    632 {
    633 	register struct file *fq;
    634 
    635 	LIST_REMOVE(fp, f_list);
    636 	crfree(fp->f_cred);
    637 #ifdef DIAGNOSTIC
    638 	fp->f_count = 0;
    639 #endif
    640 	nfiles--;
    641 	FREE(fp, M_FILE);
    642 }
    643 
    644 /*
    645  * Copy a filedesc structure.
    646  */
    647 struct filedesc *
    648 fdcopy(p)
    649 	struct proc *p;
    650 {
    651 	register struct filedesc *newfdp, *fdp = p->p_fd;
    652 	register struct file **fpp;
    653 	register int i;
    654 
    655 	MALLOC(newfdp, struct filedesc *, sizeof(struct filedesc0),
    656 	    M_FILEDESC, M_WAITOK);
    657 	bcopy(fdp, newfdp, sizeof(struct filedesc));
    658 	VREF(newfdp->fd_cdir);
    659 	if (newfdp->fd_rdir)
    660 		VREF(newfdp->fd_rdir);
    661 	newfdp->fd_refcnt = 1;
    662 
    663 	/*
    664 	 * If the number of open files fits in the internal arrays
    665 	 * of the open file structure, use them, otherwise allocate
    666 	 * additional memory for the number of descriptors currently
    667 	 * in use.
    668 	 */
    669 	if (newfdp->fd_lastfile < NDFILE) {
    670 		newfdp->fd_ofiles = ((struct filedesc0 *) newfdp)->fd_dfiles;
    671 		newfdp->fd_ofileflags =
    672 		    ((struct filedesc0 *) newfdp)->fd_dfileflags;
    673 		i = NDFILE;
    674 	} else {
    675 		/*
    676 		 * Compute the smallest multiple of NDEXTENT needed
    677 		 * for the file descriptors currently in use,
    678 		 * allowing the table to shrink.
    679 		 */
    680 		i = newfdp->fd_nfiles;
    681 		while (i > 2 * NDEXTENT && i > newfdp->fd_lastfile * 2)
    682 			i /= 2;
    683 		MALLOC(newfdp->fd_ofiles, struct file **, i * OFILESIZE,
    684 		    M_FILEDESC, M_WAITOK);
    685 		newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
    686 	}
    687 	newfdp->fd_nfiles = i;
    688 	bcopy(fdp->fd_ofiles, newfdp->fd_ofiles, i * sizeof(struct file **));
    689 	bcopy(fdp->fd_ofileflags, newfdp->fd_ofileflags, i * sizeof(char));
    690 	fpp = newfdp->fd_ofiles;
    691 	for (i = newfdp->fd_lastfile; i-- >= 0; fpp++)
    692 		if (*fpp != NULL)
    693 			(*fpp)->f_count++;
    694 	return (newfdp);
    695 }
    696 
    697 /*
    698  * Release a filedesc structure.
    699  */
    700 void
    701 fdfree(p)
    702 	struct proc *p;
    703 {
    704 	register struct filedesc *fdp = p->p_fd;
    705 	struct file **fpp;
    706 	register int i;
    707 
    708 	if (--fdp->fd_refcnt > 0)
    709 		return;
    710 	fpp = fdp->fd_ofiles;
    711 	for (i = fdp->fd_lastfile; i-- >= 0; fpp++)
    712 		if (*fpp)
    713 			(void) closef(*fpp, p);
    714 	if (fdp->fd_nfiles > NDFILE)
    715 		FREE(fdp->fd_ofiles, M_FILEDESC);
    716 	vrele(fdp->fd_cdir);
    717 	if (fdp->fd_rdir)
    718 		vrele(fdp->fd_rdir);
    719 	FREE(fdp, M_FILEDESC);
    720 }
    721 
    722 /*
    723  * Internal form of close.
    724  * Decrement reference count on file structure.
    725  * Note: p may be NULL when closing a file
    726  * that was being passed in a message.
    727  */
    728 int
    729 closef(fp, p)
    730 	register struct file *fp;
    731 	register struct proc *p;
    732 {
    733 	struct vnode *vp;
    734 	struct flock lf;
    735 	int error;
    736 
    737 	if (fp == NULL)
    738 		return (0);
    739 	/*
    740 	 * POSIX record locking dictates that any close releases ALL
    741 	 * locks owned by this process.  This is handled by setting
    742 	 * a flag in the unlock to free ONLY locks obeying POSIX
    743 	 * semantics, and not to free BSD-style file locks.
    744 	 * If the descriptor was in a message, POSIX-style locks
    745 	 * aren't passed with the descriptor.
    746 	 */
    747 	if (p && (p->p_flag & P_ADVLOCK) && fp->f_type == DTYPE_VNODE) {
    748 		lf.l_whence = SEEK_SET;
    749 		lf.l_start = 0;
    750 		lf.l_len = 0;
    751 		lf.l_type = F_UNLCK;
    752 		vp = (struct vnode *)fp->f_data;
    753 		(void) VOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_POSIX);
    754 	}
    755 	if (--fp->f_count > 0)
    756 		return (0);
    757 	if (fp->f_count < 0)
    758 		panic("closef: count < 0");
    759 	if ((fp->f_flag & FHASLOCK) && fp->f_type == DTYPE_VNODE) {
    760 		lf.l_whence = SEEK_SET;
    761 		lf.l_start = 0;
    762 		lf.l_len = 0;
    763 		lf.l_type = F_UNLCK;
    764 		vp = (struct vnode *)fp->f_data;
    765 		(void) VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK);
    766 	}
    767 	if (fp->f_ops)
    768 		error = (*fp->f_ops->fo_close)(fp, p);
    769 	else
    770 		error = 0;
    771 	ffree(fp);
    772 	return (error);
    773 }
    774 
    775 /*
    776  * Apply an advisory lock on a file descriptor.
    777  *
    778  * Just attempt to get a record lock of the requested type on
    779  * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
    780  */
    781 /* ARGSUSED */
    782 int
    783 flock(p, uap, retval)
    784 	struct proc *p;
    785 	register struct flock_args /* {
    786 		syscallarg(int) fd;
    787 		syscallarg(int) how;
    788 	} */ *uap;
    789 	register_t *retval;
    790 {
    791 	int fd = SCARG(uap, fd);
    792 	int how = SCARG(uap, how);
    793 	register struct filedesc *fdp = p->p_fd;
    794 	register struct file *fp;
    795 	struct vnode *vp;
    796 	struct flock lf;
    797 
    798 	if ((u_int)fd >= fdp->fd_nfiles ||
    799 	    (fp = fdp->fd_ofiles[fd]) == NULL)
    800 		return (EBADF);
    801 	if (fp->f_type != DTYPE_VNODE)
    802 		return (EOPNOTSUPP);
    803 	vp = (struct vnode *)fp->f_data;
    804 	lf.l_whence = SEEK_SET;
    805 	lf.l_start = 0;
    806 	lf.l_len = 0;
    807 	if (how & LOCK_UN) {
    808 		lf.l_type = F_UNLCK;
    809 		fp->f_flag &= ~FHASLOCK;
    810 		return (VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK));
    811 	}
    812 	if (how & LOCK_EX)
    813 		lf.l_type = F_WRLCK;
    814 	else if (how & LOCK_SH)
    815 		lf.l_type = F_RDLCK;
    816 	else
    817 		return (EBADF);
    818 	fp->f_flag |= FHASLOCK;
    819 	if (how & LOCK_NB)
    820 		return (VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK));
    821 	return (VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK|F_WAIT));
    822 }
    823 
    824 /*
    825  * File Descriptor pseudo-device driver (/dev/fd/).
    826  *
    827  * Opening minor device N dup()s the file (if any) connected to file
    828  * descriptor N belonging to the calling process.  Note that this driver
    829  * consists of only the ``open()'' routine, because all subsequent
    830  * references to this file will be direct to the other driver.
    831  */
    832 /* ARGSUSED */
    833 int
    834 fdopen(dev, mode, type, p)
    835 	dev_t dev;
    836 	int mode, type;
    837 	struct proc *p;
    838 {
    839 
    840 	/*
    841 	 * XXX Kludge: set curproc->p_dupfd to contain the value of the
    842 	 * the file descriptor being sought for duplication. The error
    843 	 * return ensures that the vnode for this device will be released
    844 	 * by vn_open. Open will detect this special error and take the
    845 	 * actions in dupfdopen below. Other callers of vn_open or VOP_OPEN
    846 	 * will simply report the error.
    847 	 */
    848 	p->p_dupfd = minor(dev);
    849 	return (ENODEV);
    850 }
    851 
    852 /*
    853  * Duplicate the specified descriptor to a free descriptor.
    854  */
    855 int
    856 dupfdopen(fdp, indx, dfd, mode, error)
    857 	register struct filedesc *fdp;
    858 	register int indx, dfd;
    859 	int mode;
    860 	int error;
    861 {
    862 	register struct file *wfp;
    863 	struct file *fp;
    864 
    865 	/*
    866 	 * If the to-be-dup'd fd number is greater than the allowed number
    867 	 * of file descriptors, or the fd to be dup'd has already been
    868 	 * closed, reject.  Note, check for new == old is necessary as
    869 	 * falloc could allocate an already closed to-be-dup'd descriptor
    870 	 * as the new descriptor.
    871 	 */
    872 	fp = fdp->fd_ofiles[indx];
    873 	if ((u_int)dfd >= fdp->fd_nfiles ||
    874 	    (wfp = fdp->fd_ofiles[dfd]) == NULL || fp == wfp)
    875 		return (EBADF);
    876 
    877 	/*
    878 	 * There are two cases of interest here.
    879 	 *
    880 	 * For ENODEV simply dup (dfd) to file descriptor
    881 	 * (indx) and return.
    882 	 *
    883 	 * For ENXIO steal away the file structure from (dfd) and
    884 	 * store it in (indx).  (dfd) is effectively closed by
    885 	 * this operation.
    886 	 *
    887 	 * Any other error code is just returned.
    888 	 */
    889 	switch (error) {
    890 	case ENODEV:
    891 		/*
    892 		 * Check that the mode the file is being opened for is a
    893 		 * subset of the mode of the existing descriptor.
    894 		 */
    895 		if (((mode & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag)
    896 			return (EACCES);
    897 		fdp->fd_ofiles[indx] = wfp;
    898 		fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd];
    899 		wfp->f_count++;
    900 		if (indx > fdp->fd_lastfile)
    901 			fdp->fd_lastfile = indx;
    902 		return (0);
    903 
    904 	case ENXIO:
    905 		/*
    906 		 * Steal away the file pointer from dfd, and stuff it into indx.
    907 		 */
    908 		fdp->fd_ofiles[indx] = fdp->fd_ofiles[dfd];
    909 		fdp->fd_ofiles[dfd] = NULL;
    910 		fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd];
    911 		fdp->fd_ofileflags[dfd] = 0;
    912 		/*
    913 		 * Complete the clean up of the filedesc structure by
    914 		 * recomputing the various hints.
    915 		 */
    916 		if (indx > fdp->fd_lastfile)
    917 			fdp->fd_lastfile = indx;
    918 		else
    919 			while (fdp->fd_lastfile > 0 &&
    920 			       fdp->fd_ofiles[fdp->fd_lastfile] == NULL)
    921 				fdp->fd_lastfile--;
    922 			if (dfd < fdp->fd_freefile)
    923 				fdp->fd_freefile = dfd;
    924 		return (0);
    925 
    926 	default:
    927 		return (error);
    928 	}
    929 	/* NOTREACHED */
    930 }
    931