Home | History | Annotate | Line # | Download | only in kern
vfs_vnops.c revision 1.26
      1 /*	$NetBSD: vfs_vnops.c,v 1.26 1998/02/10 14:09:57 mrg 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 "opt_uvm.h"
     44 
     45 #include <sys/param.h>
     46 #include <sys/systm.h>
     47 #include <sys/kernel.h>
     48 #include <sys/file.h>
     49 #include <sys/stat.h>
     50 #include <sys/buf.h>
     51 #include <sys/proc.h>
     52 #include <sys/mount.h>
     53 #include <sys/namei.h>
     54 #include <sys/vnode.h>
     55 #include <sys/ioctl.h>
     56 #include <sys/tty.h>
     57 #include <sys/poll.h>
     58 
     59 #include <vm/vm.h>
     60 
     61 #if defined(UVM)
     62 #include <uvm/uvm_extern.h>
     63 #endif
     64 
     65 struct 	fileops vnops =
     66 	{ vn_read, vn_write, vn_ioctl, vn_poll, vn_closefile };
     67 
     68 /*
     69  * Common code for vnode open operations.
     70  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
     71  */
     72 int
     73 vn_open(ndp, fmode, cmode)
     74 	register struct nameidata *ndp;
     75 	int fmode, cmode;
     76 {
     77 	register struct vnode *vp;
     78 	register struct proc *p = ndp->ni_cnd.cn_proc;
     79 	register struct ucred *cred = p->p_ucred;
     80 	struct vattr va;
     81 	int error;
     82 
     83 	if (fmode & O_CREAT) {
     84 		ndp->ni_cnd.cn_nameiop = CREATE;
     85 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
     86 		if ((fmode & O_EXCL) == 0)
     87 			ndp->ni_cnd.cn_flags |= FOLLOW;
     88 		if ((error = namei(ndp)) != 0)
     89 			return (error);
     90 		if (ndp->ni_vp == NULL) {
     91 			VATTR_NULL(&va);
     92 			va.va_type = VREG;
     93 			va.va_mode = cmode;
     94 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
     95 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
     96 					   &ndp->ni_cnd, &va);
     97 			if (error)
     98 				return (error);
     99 			fmode &= ~O_TRUNC;
    100 			vp = ndp->ni_vp;
    101 		} else {
    102 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    103 			if (ndp->ni_dvp == ndp->ni_vp)
    104 				vrele(ndp->ni_dvp);
    105 			else
    106 				vput(ndp->ni_dvp);
    107 			ndp->ni_dvp = NULL;
    108 			vp = ndp->ni_vp;
    109 			if (fmode & O_EXCL) {
    110 				error = EEXIST;
    111 				goto bad;
    112 			}
    113 			fmode &= ~O_CREAT;
    114 		}
    115 	} else {
    116 		ndp->ni_cnd.cn_nameiop = LOOKUP;
    117 		ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF;
    118 		if ((error = namei(ndp)) != 0)
    119 			return (error);
    120 		vp = ndp->ni_vp;
    121 	}
    122 	if (vp->v_type == VSOCK) {
    123 		error = EOPNOTSUPP;
    124 		goto bad;
    125 	}
    126 	if ((fmode & O_CREAT) == 0) {
    127 		if (fmode & FREAD) {
    128 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
    129 				goto bad;
    130 		}
    131 		if (fmode & (FWRITE | O_TRUNC)) {
    132 			if (vp->v_type == VDIR) {
    133 				error = EISDIR;
    134 				goto bad;
    135 			}
    136 			if ((error = vn_writechk(vp)) != 0 ||
    137 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
    138 				goto bad;
    139 		}
    140 	}
    141 	if (fmode & O_TRUNC) {
    142 		VOP_UNLOCK(vp);				/* XXX */
    143 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
    144 		VOP_LOCK(vp);				/* XXX */
    145 		VATTR_NULL(&va);
    146 		va.va_size = 0;
    147 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
    148 			goto bad;
    149 	}
    150 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
    151 		goto bad;
    152 	if (fmode & FWRITE)
    153 		vp->v_writecount++;
    154 	return (0);
    155 bad:
    156 	vput(vp);
    157 	return (error);
    158 }
    159 
    160 /*
    161  * Check for write permissions on the specified vnode.
    162  * The read-only status of the file system is checked.
    163  * Also, prototype text segments cannot be written.
    164  */
    165 int
    166 vn_writechk(vp)
    167 	register struct vnode *vp;
    168 {
    169 
    170 	/*
    171 	 * Disallow write attempts on read-only file systems;
    172 	 * unless the file is a socket or a block or character
    173 	 * device resident on the file system.
    174 	 */
    175 	if (vp->v_mount->mnt_flag & MNT_RDONLY) {
    176 		switch (vp->v_type) {
    177 		case VREG: case VDIR: case VLNK:
    178 			return (EROFS);
    179 		case VNON: case VCHR: case VSOCK:
    180 		case VFIFO: case VBAD: case VBLK:
    181 			break;
    182 		}
    183 	}
    184 	/*
    185 	 * If there's shared text associated with
    186 	 * the vnode, try to free it up once.  If
    187 	 * we fail, we can't allow writing.
    188 	 */
    189 #if defined(UVM)
    190 	if ((vp->v_flag & VTEXT) && !uvm_vnp_uncache(vp))
    191 		return (ETXTBSY);
    192 #else
    193 	if ((vp->v_flag & VTEXT) && !vnode_pager_uncache(vp))
    194 		return (ETXTBSY);
    195 #endif
    196 	return (0);
    197 }
    198 
    199 /*
    200  * Vnode close call
    201  */
    202 int
    203 vn_close(vp, flags, cred, p)
    204 	register struct vnode *vp;
    205 	int flags;
    206 	struct ucred *cred;
    207 	struct proc *p;
    208 {
    209 	int error;
    210 
    211 	if (flags & FWRITE)
    212 		vp->v_writecount--;
    213 	error = VOP_CLOSE(vp, flags, cred, p);
    214 	vrele(vp);
    215 	return (error);
    216 }
    217 
    218 /*
    219  * Package up an I/O request on a vnode into a uio and do it.
    220  */
    221 int
    222 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p)
    223 	enum uio_rw rw;
    224 	struct vnode *vp;
    225 	caddr_t base;
    226 	int len;
    227 	off_t offset;
    228 	enum uio_seg segflg;
    229 	int ioflg;
    230 	struct ucred *cred;
    231 	int *aresid;
    232 	struct proc *p;
    233 {
    234 	struct uio auio;
    235 	struct iovec aiov;
    236 	int error;
    237 
    238 	if ((ioflg & IO_NODELOCKED) == 0)
    239 		VOP_LOCK(vp);
    240 	auio.uio_iov = &aiov;
    241 	auio.uio_iovcnt = 1;
    242 	aiov.iov_base = base;
    243 	aiov.iov_len = len;
    244 	auio.uio_resid = len;
    245 	auio.uio_offset = offset;
    246 	auio.uio_segflg = segflg;
    247 	auio.uio_rw = rw;
    248 	auio.uio_procp = p;
    249 	if (rw == UIO_READ) {
    250 		error = VOP_READ(vp, &auio, ioflg, cred);
    251 	} else {
    252 		error = VOP_WRITE(vp, &auio, ioflg, cred);
    253 	}
    254 	if (aresid)
    255 		*aresid = auio.uio_resid;
    256 	else
    257 		if (auio.uio_resid && error == 0)
    258 			error = EIO;
    259 	if ((ioflg & IO_NODELOCKED) == 0)
    260 		VOP_UNLOCK(vp);
    261 	return (error);
    262 }
    263 
    264 int
    265 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies)
    266 	struct file *fp;
    267 	char *buf;
    268 	int segflg, *done, ncookies;
    269 	u_int count;
    270 	struct proc *p;
    271 	off_t *cookies;
    272 {
    273 	struct vnode *vp = (struct vnode *)fp->f_data;
    274 	struct iovec aiov;
    275 	struct uio auio;
    276 	int error, eofflag;
    277 
    278 unionread:
    279 	if (vp->v_type != VDIR)
    280 		return (EINVAL);
    281 	aiov.iov_base = buf;
    282 	aiov.iov_len = count;
    283 	auio.uio_iov = &aiov;
    284 	auio.uio_iovcnt = 1;
    285 	auio.uio_rw = UIO_READ;
    286 	auio.uio_segflg = segflg;
    287 	auio.uio_procp = p;
    288 	auio.uio_resid = count;
    289 	VOP_LOCK(vp);
    290 	auio.uio_offset = fp->f_offset;
    291 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, (off_t *)cookies,
    292 		    ncookies);
    293 	fp->f_offset = auio.uio_offset;
    294 	VOP_UNLOCK(vp);
    295 	if (error)
    296 		return (error);
    297 
    298 #ifdef UNION
    299 {
    300 	extern int (**union_vnodeop_p) __P((void *));
    301 	extern struct vnode *union_dircache __P((struct vnode *));
    302 
    303 	if (count == auio.uio_resid && (vp->v_op == union_vnodeop_p)) {
    304 		struct vnode *lvp;
    305 
    306 		lvp = union_dircache(vp);
    307 		if (lvp != NULLVP) {
    308 			struct vattr va;
    309 
    310 			/*
    311 			 * If the directory is opaque,
    312 			 * then don't show lower entries
    313 			 */
    314 			error = VOP_GETATTR(vp, &va, fp->f_cred, p);
    315 			if (va.va_flags & OPAQUE) {
    316 				vput(lvp);
    317 				lvp = NULL;
    318 			}
    319 		}
    320 
    321 		if (lvp != NULLVP) {
    322 			error = VOP_OPEN(lvp, FREAD, fp->f_cred, p);
    323 			VOP_UNLOCK(lvp);
    324 
    325 			if (error) {
    326 				vrele(lvp);
    327 				return (error);
    328 			}
    329 			fp->f_data = (caddr_t) lvp;
    330 			fp->f_offset = 0;
    331 			error = vn_close(vp, FREAD, fp->f_cred, p);
    332 			if (error)
    333 				return (error);
    334 			vp = lvp;
    335 			goto unionread;
    336 		}
    337 	}
    338 }
    339 #endif /* UNION */
    340 
    341 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
    342 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
    343 		struct vnode *tvp = vp;
    344 		vp = vp->v_mount->mnt_vnodecovered;
    345 		VREF(vp);
    346 		fp->f_data = (caddr_t) vp;
    347 		fp->f_offset = 0;
    348 		vrele(tvp);
    349 		goto unionread;
    350 	}
    351 	*done = count - auio.uio_resid;
    352 	return error;
    353 }
    354 
    355 /*
    356  * File table vnode read routine.
    357  */
    358 int
    359 vn_read(fp, uio, cred)
    360 	struct file *fp;
    361 	struct uio *uio;
    362 	struct ucred *cred;
    363 {
    364 	register struct vnode *vp = (struct vnode *)fp->f_data;
    365 	int count, error;
    366 
    367 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
    368 	VOP_LOCK(vp);
    369 	uio->uio_offset = fp->f_offset;
    370 	count = uio->uio_resid;
    371 	error = VOP_READ(vp, uio, (fp->f_flag & FNONBLOCK) ? IO_NDELAY : 0,
    372 		cred);
    373 	fp->f_offset += count - uio->uio_resid;
    374 	VOP_UNLOCK(vp);
    375 	return (error);
    376 }
    377 
    378 /*
    379  * File table vnode write routine.
    380  */
    381 int
    382 vn_write(fp, uio, cred)
    383 	struct file *fp;
    384 	struct uio *uio;
    385 	struct ucred *cred;
    386 {
    387 	register struct vnode *vp = (struct vnode *)fp->f_data;
    388 	int count, error, ioflag = IO_UNIT;
    389 
    390 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
    391 		ioflag |= IO_APPEND;
    392 	if (fp->f_flag & FNONBLOCK)
    393 		ioflag |= IO_NDELAY;
    394 	if (fp->f_flag & FFSYNC ||
    395 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
    396 		ioflag |= IO_SYNC;
    397 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
    398 	VOP_LOCK(vp);
    399 	uio->uio_offset = fp->f_offset;
    400 	count = uio->uio_resid;
    401 	error = VOP_WRITE(vp, uio, ioflag, cred);
    402 	if (ioflag & IO_APPEND)
    403 		fp->f_offset = uio->uio_offset;
    404 	else
    405 		fp->f_offset += count - uio->uio_resid;
    406 	VOP_UNLOCK(vp);
    407 	return (error);
    408 }
    409 
    410 /*
    411  * File table vnode stat routine.
    412  */
    413 int
    414 vn_stat(vp, sb, p)
    415 	struct vnode *vp;
    416 	register struct stat *sb;
    417 	struct proc *p;
    418 {
    419 	struct vattr va;
    420 	int error;
    421 	u_short mode;
    422 
    423 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
    424 	if (error)
    425 		return (error);
    426 	/*
    427 	 * Copy from vattr table
    428 	 */
    429 	sb->st_dev = va.va_fsid;
    430 	sb->st_ino = va.va_fileid;
    431 	mode = va.va_mode;
    432 	switch (vp->v_type) {
    433 	case VREG:
    434 		mode |= S_IFREG;
    435 		break;
    436 	case VDIR:
    437 		mode |= S_IFDIR;
    438 		break;
    439 	case VBLK:
    440 		mode |= S_IFBLK;
    441 		break;
    442 	case VCHR:
    443 		mode |= S_IFCHR;
    444 		break;
    445 	case VLNK:
    446 		mode |= S_IFLNK;
    447 		break;
    448 	case VSOCK:
    449 		mode |= S_IFSOCK;
    450 		break;
    451 	case VFIFO:
    452 		mode |= S_IFIFO;
    453 		break;
    454 	default:
    455 		return (EBADF);
    456 	};
    457 	sb->st_mode = mode;
    458 	sb->st_nlink = va.va_nlink;
    459 	sb->st_uid = va.va_uid;
    460 	sb->st_gid = va.va_gid;
    461 	sb->st_rdev = va.va_rdev;
    462 	sb->st_size = va.va_size;
    463 	sb->st_atimespec = va.va_atime;
    464 	sb->st_mtimespec = va.va_mtime;
    465 	sb->st_ctimespec = va.va_ctime;
    466 	sb->st_blksize = va.va_blocksize;
    467 	sb->st_flags = va.va_flags;
    468 	sb->st_gen = 0;
    469 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
    470 	return (0);
    471 }
    472 
    473 /*
    474  * File table vnode ioctl routine.
    475  */
    476 int
    477 vn_ioctl(fp, com, data, p)
    478 	struct file *fp;
    479 	u_long com;
    480 	caddr_t data;
    481 	struct proc *p;
    482 {
    483 	register struct vnode *vp = ((struct vnode *)fp->f_data);
    484 	struct vattr vattr;
    485 	int error;
    486 
    487 	switch (vp->v_type) {
    488 
    489 	case VREG:
    490 	case VDIR:
    491 		if (com == FIONREAD) {
    492 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
    493 			if (error)
    494 				return (error);
    495 			*(int *)data = vattr.va_size - fp->f_offset;
    496 			return (0);
    497 		}
    498 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
    499 			return (0);			/* XXX */
    500 		/* fall into ... */
    501 
    502 	default:
    503 		return (ENOTTY);
    504 
    505 	case VFIFO:
    506 	case VCHR:
    507 	case VBLK:
    508 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    509 		if (error == 0 && com == TIOCSCTTY) {
    510 			if (p->p_session->s_ttyvp)
    511 				vrele(p->p_session->s_ttyvp);
    512 			p->p_session->s_ttyvp = vp;
    513 			VREF(vp);
    514 		}
    515 		return (error);
    516 	}
    517 }
    518 
    519 /*
    520  * File table vnode poll routine.
    521  */
    522 int
    523 vn_poll(fp, events, p)
    524 	struct file *fp;
    525 	int events;
    526 	struct proc *p;
    527 {
    528 
    529 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
    530 }
    531 
    532 /*
    533  * File table vnode close routine.
    534  */
    535 int
    536 vn_closefile(fp, p)
    537 	struct file *fp;
    538 	struct proc *p;
    539 {
    540 
    541 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
    542 		fp->f_cred, p));
    543 }
    544