Home | History | Annotate | Line # | Download | only in kern
vfs_vnops.c revision 1.27
      1 /*	$NetBSD: vfs_vnops.c,v 1.27 1998/02/19 00:53:46 thorpej Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1982, 1986, 1989, 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  *	@(#)vfs_vnops.c	8.5 (Berkeley) 12/8/94
     41  */
     42 
     43 #include "fs_union.h"
     44 #include "opt_uvm.h"
     45 
     46 #include <sys/param.h>
     47 #include <sys/systm.h>
     48 #include <sys/kernel.h>
     49 #include <sys/file.h>
     50 #include <sys/stat.h>
     51 #include <sys/buf.h>
     52 #include <sys/proc.h>
     53 #include <sys/mount.h>
     54 #include <sys/namei.h>
     55 #include <sys/vnode.h>
     56 #include <sys/ioctl.h>
     57 #include <sys/tty.h>
     58 #include <sys/poll.h>
     59 
     60 #include <vm/vm.h>
     61 
     62 #if defined(UVM)
     63 #include <uvm/uvm_extern.h>
     64 #endif
     65 
     66 struct 	fileops vnops =
     67 	{ vn_read, vn_write, vn_ioctl, vn_poll, vn_closefile };
     68 
     69 /*
     70  * Common code for vnode open operations.
     71  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
     72  */
     73 int
     74 vn_open(ndp, fmode, cmode)
     75 	register struct nameidata *ndp;
     76 	int fmode, cmode;
     77 {
     78 	register struct vnode *vp;
     79 	register struct proc *p = ndp->ni_cnd.cn_proc;
     80 	register struct ucred *cred = p->p_ucred;
     81 	struct vattr va;
     82 	int error;
     83 
     84 	if (fmode & O_CREAT) {
     85 		ndp->ni_cnd.cn_nameiop = CREATE;
     86 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
     87 		if ((fmode & O_EXCL) == 0)
     88 			ndp->ni_cnd.cn_flags |= FOLLOW;
     89 		if ((error = namei(ndp)) != 0)
     90 			return (error);
     91 		if (ndp->ni_vp == NULL) {
     92 			VATTR_NULL(&va);
     93 			va.va_type = VREG;
     94 			va.va_mode = cmode;
     95 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
     96 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
     97 					   &ndp->ni_cnd, &va);
     98 			if (error)
     99 				return (error);
    100 			fmode &= ~O_TRUNC;
    101 			vp = ndp->ni_vp;
    102 		} else {
    103 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    104 			if (ndp->ni_dvp == ndp->ni_vp)
    105 				vrele(ndp->ni_dvp);
    106 			else
    107 				vput(ndp->ni_dvp);
    108 			ndp->ni_dvp = NULL;
    109 			vp = ndp->ni_vp;
    110 			if (fmode & O_EXCL) {
    111 				error = EEXIST;
    112 				goto bad;
    113 			}
    114 			fmode &= ~O_CREAT;
    115 		}
    116 	} else {
    117 		ndp->ni_cnd.cn_nameiop = LOOKUP;
    118 		ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF;
    119 		if ((error = namei(ndp)) != 0)
    120 			return (error);
    121 		vp = ndp->ni_vp;
    122 	}
    123 	if (vp->v_type == VSOCK) {
    124 		error = EOPNOTSUPP;
    125 		goto bad;
    126 	}
    127 	if ((fmode & O_CREAT) == 0) {
    128 		if (fmode & FREAD) {
    129 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
    130 				goto bad;
    131 		}
    132 		if (fmode & (FWRITE | O_TRUNC)) {
    133 			if (vp->v_type == VDIR) {
    134 				error = EISDIR;
    135 				goto bad;
    136 			}
    137 			if ((error = vn_writechk(vp)) != 0 ||
    138 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
    139 				goto bad;
    140 		}
    141 	}
    142 	if (fmode & O_TRUNC) {
    143 		VOP_UNLOCK(vp);				/* XXX */
    144 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
    145 		VOP_LOCK(vp);				/* XXX */
    146 		VATTR_NULL(&va);
    147 		va.va_size = 0;
    148 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
    149 			goto bad;
    150 	}
    151 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
    152 		goto bad;
    153 	if (fmode & FWRITE)
    154 		vp->v_writecount++;
    155 	return (0);
    156 bad:
    157 	vput(vp);
    158 	return (error);
    159 }
    160 
    161 /*
    162  * Check for write permissions on the specified vnode.
    163  * The read-only status of the file system is checked.
    164  * Also, prototype text segments cannot be written.
    165  */
    166 int
    167 vn_writechk(vp)
    168 	register struct vnode *vp;
    169 {
    170 
    171 	/*
    172 	 * Disallow write attempts on read-only file systems;
    173 	 * unless the file is a socket or a block or character
    174 	 * device resident on the file system.
    175 	 */
    176 	if (vp->v_mount->mnt_flag & MNT_RDONLY) {
    177 		switch (vp->v_type) {
    178 		case VREG: case VDIR: case VLNK:
    179 			return (EROFS);
    180 		case VNON: case VCHR: case VSOCK:
    181 		case VFIFO: case VBAD: case VBLK:
    182 			break;
    183 		}
    184 	}
    185 	/*
    186 	 * If there's shared text associated with
    187 	 * the vnode, try to free it up once.  If
    188 	 * we fail, we can't allow writing.
    189 	 */
    190 #if defined(UVM)
    191 	if ((vp->v_flag & VTEXT) && !uvm_vnp_uncache(vp))
    192 		return (ETXTBSY);
    193 #else
    194 	if ((vp->v_flag & VTEXT) && !vnode_pager_uncache(vp))
    195 		return (ETXTBSY);
    196 #endif
    197 	return (0);
    198 }
    199 
    200 /*
    201  * Vnode close call
    202  */
    203 int
    204 vn_close(vp, flags, cred, p)
    205 	register struct vnode *vp;
    206 	int flags;
    207 	struct ucred *cred;
    208 	struct proc *p;
    209 {
    210 	int error;
    211 
    212 	if (flags & FWRITE)
    213 		vp->v_writecount--;
    214 	error = VOP_CLOSE(vp, flags, cred, p);
    215 	vrele(vp);
    216 	return (error);
    217 }
    218 
    219 /*
    220  * Package up an I/O request on a vnode into a uio and do it.
    221  */
    222 int
    223 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p)
    224 	enum uio_rw rw;
    225 	struct vnode *vp;
    226 	caddr_t base;
    227 	int len;
    228 	off_t offset;
    229 	enum uio_seg segflg;
    230 	int ioflg;
    231 	struct ucred *cred;
    232 	int *aresid;
    233 	struct proc *p;
    234 {
    235 	struct uio auio;
    236 	struct iovec aiov;
    237 	int error;
    238 
    239 	if ((ioflg & IO_NODELOCKED) == 0)
    240 		VOP_LOCK(vp);
    241 	auio.uio_iov = &aiov;
    242 	auio.uio_iovcnt = 1;
    243 	aiov.iov_base = base;
    244 	aiov.iov_len = len;
    245 	auio.uio_resid = len;
    246 	auio.uio_offset = offset;
    247 	auio.uio_segflg = segflg;
    248 	auio.uio_rw = rw;
    249 	auio.uio_procp = p;
    250 	if (rw == UIO_READ) {
    251 		error = VOP_READ(vp, &auio, ioflg, cred);
    252 	} else {
    253 		error = VOP_WRITE(vp, &auio, ioflg, cred);
    254 	}
    255 	if (aresid)
    256 		*aresid = auio.uio_resid;
    257 	else
    258 		if (auio.uio_resid && error == 0)
    259 			error = EIO;
    260 	if ((ioflg & IO_NODELOCKED) == 0)
    261 		VOP_UNLOCK(vp);
    262 	return (error);
    263 }
    264 
    265 int
    266 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies)
    267 	struct file *fp;
    268 	char *buf;
    269 	int segflg, *done, ncookies;
    270 	u_int count;
    271 	struct proc *p;
    272 	off_t *cookies;
    273 {
    274 	struct vnode *vp = (struct vnode *)fp->f_data;
    275 	struct iovec aiov;
    276 	struct uio auio;
    277 	int error, eofflag;
    278 
    279 unionread:
    280 	if (vp->v_type != VDIR)
    281 		return (EINVAL);
    282 	aiov.iov_base = buf;
    283 	aiov.iov_len = count;
    284 	auio.uio_iov = &aiov;
    285 	auio.uio_iovcnt = 1;
    286 	auio.uio_rw = UIO_READ;
    287 	auio.uio_segflg = segflg;
    288 	auio.uio_procp = p;
    289 	auio.uio_resid = count;
    290 	VOP_LOCK(vp);
    291 	auio.uio_offset = fp->f_offset;
    292 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, (off_t *)cookies,
    293 		    ncookies);
    294 	fp->f_offset = auio.uio_offset;
    295 	VOP_UNLOCK(vp);
    296 	if (error)
    297 		return (error);
    298 
    299 #ifdef UNION
    300 {
    301 	extern int (**union_vnodeop_p) __P((void *));
    302 	extern struct vnode *union_dircache __P((struct vnode *));
    303 
    304 	if (count == auio.uio_resid && (vp->v_op == union_vnodeop_p)) {
    305 		struct vnode *lvp;
    306 
    307 		lvp = union_dircache(vp);
    308 		if (lvp != NULLVP) {
    309 			struct vattr va;
    310 
    311 			/*
    312 			 * If the directory is opaque,
    313 			 * then don't show lower entries
    314 			 */
    315 			error = VOP_GETATTR(vp, &va, fp->f_cred, p);
    316 			if (va.va_flags & OPAQUE) {
    317 				vput(lvp);
    318 				lvp = NULL;
    319 			}
    320 		}
    321 
    322 		if (lvp != NULLVP) {
    323 			error = VOP_OPEN(lvp, FREAD, fp->f_cred, p);
    324 			VOP_UNLOCK(lvp);
    325 
    326 			if (error) {
    327 				vrele(lvp);
    328 				return (error);
    329 			}
    330 			fp->f_data = (caddr_t) lvp;
    331 			fp->f_offset = 0;
    332 			error = vn_close(vp, FREAD, fp->f_cred, p);
    333 			if (error)
    334 				return (error);
    335 			vp = lvp;
    336 			goto unionread;
    337 		}
    338 	}
    339 }
    340 #endif /* UNION */
    341 
    342 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
    343 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
    344 		struct vnode *tvp = vp;
    345 		vp = vp->v_mount->mnt_vnodecovered;
    346 		VREF(vp);
    347 		fp->f_data = (caddr_t) vp;
    348 		fp->f_offset = 0;
    349 		vrele(tvp);
    350 		goto unionread;
    351 	}
    352 	*done = count - auio.uio_resid;
    353 	return error;
    354 }
    355 
    356 /*
    357  * File table vnode read routine.
    358  */
    359 int
    360 vn_read(fp, uio, cred)
    361 	struct file *fp;
    362 	struct uio *uio;
    363 	struct ucred *cred;
    364 {
    365 	register struct vnode *vp = (struct vnode *)fp->f_data;
    366 	int count, error;
    367 
    368 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
    369 	VOP_LOCK(vp);
    370 	uio->uio_offset = fp->f_offset;
    371 	count = uio->uio_resid;
    372 	error = VOP_READ(vp, uio, (fp->f_flag & FNONBLOCK) ? IO_NDELAY : 0,
    373 		cred);
    374 	fp->f_offset += count - uio->uio_resid;
    375 	VOP_UNLOCK(vp);
    376 	return (error);
    377 }
    378 
    379 /*
    380  * File table vnode write routine.
    381  */
    382 int
    383 vn_write(fp, uio, cred)
    384 	struct file *fp;
    385 	struct uio *uio;
    386 	struct ucred *cred;
    387 {
    388 	register struct vnode *vp = (struct vnode *)fp->f_data;
    389 	int count, error, ioflag = IO_UNIT;
    390 
    391 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
    392 		ioflag |= IO_APPEND;
    393 	if (fp->f_flag & FNONBLOCK)
    394 		ioflag |= IO_NDELAY;
    395 	if (fp->f_flag & FFSYNC ||
    396 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
    397 		ioflag |= IO_SYNC;
    398 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
    399 	VOP_LOCK(vp);
    400 	uio->uio_offset = fp->f_offset;
    401 	count = uio->uio_resid;
    402 	error = VOP_WRITE(vp, uio, ioflag, cred);
    403 	if (ioflag & IO_APPEND)
    404 		fp->f_offset = uio->uio_offset;
    405 	else
    406 		fp->f_offset += count - uio->uio_resid;
    407 	VOP_UNLOCK(vp);
    408 	return (error);
    409 }
    410 
    411 /*
    412  * File table vnode stat routine.
    413  */
    414 int
    415 vn_stat(vp, sb, p)
    416 	struct vnode *vp;
    417 	register struct stat *sb;
    418 	struct proc *p;
    419 {
    420 	struct vattr va;
    421 	int error;
    422 	u_short mode;
    423 
    424 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
    425 	if (error)
    426 		return (error);
    427 	/*
    428 	 * Copy from vattr table
    429 	 */
    430 	sb->st_dev = va.va_fsid;
    431 	sb->st_ino = va.va_fileid;
    432 	mode = va.va_mode;
    433 	switch (vp->v_type) {
    434 	case VREG:
    435 		mode |= S_IFREG;
    436 		break;
    437 	case VDIR:
    438 		mode |= S_IFDIR;
    439 		break;
    440 	case VBLK:
    441 		mode |= S_IFBLK;
    442 		break;
    443 	case VCHR:
    444 		mode |= S_IFCHR;
    445 		break;
    446 	case VLNK:
    447 		mode |= S_IFLNK;
    448 		break;
    449 	case VSOCK:
    450 		mode |= S_IFSOCK;
    451 		break;
    452 	case VFIFO:
    453 		mode |= S_IFIFO;
    454 		break;
    455 	default:
    456 		return (EBADF);
    457 	};
    458 	sb->st_mode = mode;
    459 	sb->st_nlink = va.va_nlink;
    460 	sb->st_uid = va.va_uid;
    461 	sb->st_gid = va.va_gid;
    462 	sb->st_rdev = va.va_rdev;
    463 	sb->st_size = va.va_size;
    464 	sb->st_atimespec = va.va_atime;
    465 	sb->st_mtimespec = va.va_mtime;
    466 	sb->st_ctimespec = va.va_ctime;
    467 	sb->st_blksize = va.va_blocksize;
    468 	sb->st_flags = va.va_flags;
    469 	sb->st_gen = 0;
    470 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
    471 	return (0);
    472 }
    473 
    474 /*
    475  * File table vnode ioctl routine.
    476  */
    477 int
    478 vn_ioctl(fp, com, data, p)
    479 	struct file *fp;
    480 	u_long com;
    481 	caddr_t data;
    482 	struct proc *p;
    483 {
    484 	register struct vnode *vp = ((struct vnode *)fp->f_data);
    485 	struct vattr vattr;
    486 	int error;
    487 
    488 	switch (vp->v_type) {
    489 
    490 	case VREG:
    491 	case VDIR:
    492 		if (com == FIONREAD) {
    493 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
    494 			if (error)
    495 				return (error);
    496 			*(int *)data = vattr.va_size - fp->f_offset;
    497 			return (0);
    498 		}
    499 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
    500 			return (0);			/* XXX */
    501 		/* fall into ... */
    502 
    503 	default:
    504 		return (ENOTTY);
    505 
    506 	case VFIFO:
    507 	case VCHR:
    508 	case VBLK:
    509 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    510 		if (error == 0 && com == TIOCSCTTY) {
    511 			if (p->p_session->s_ttyvp)
    512 				vrele(p->p_session->s_ttyvp);
    513 			p->p_session->s_ttyvp = vp;
    514 			VREF(vp);
    515 		}
    516 		return (error);
    517 	}
    518 }
    519 
    520 /*
    521  * File table vnode poll routine.
    522  */
    523 int
    524 vn_poll(fp, events, p)
    525 	struct file *fp;
    526 	int events;
    527 	struct proc *p;
    528 {
    529 
    530 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
    531 }
    532 
    533 /*
    534  * File table vnode close routine.
    535  */
    536 int
    537 vn_closefile(fp, p)
    538 	struct file *fp;
    539 	struct proc *p;
    540 {
    541 
    542 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
    543 		fp->f_cred, p));
    544 }
    545