Home | History | Annotate | Line # | Download | only in kern
vfs_vnops.c revision 1.37
      1 /*	$NetBSD: vfs_vnops.c,v 1.37 1999/08/03 20:19:17 wrstuden 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	8.14 (Berkeley) 6/15/95
     41  */
     42 
     43 #include "fs_union.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 #include <uvm/uvm_extern.h>
     62 
     63 #ifdef UNION
     64 #include <miscfs/union/union.h>
     65 #endif
     66 
     67 struct 	fileops vnops =
     68 	{ vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, vn_closefile };
     69 
     70 /*
     71  * Common code for vnode open operations.
     72  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
     73  */
     74 int
     75 vn_open(ndp, fmode, cmode)
     76 	register struct nameidata *ndp;
     77 	int fmode, cmode;
     78 {
     79 	register struct vnode *vp;
     80 	register struct proc *p = ndp->ni_cnd.cn_proc;
     81 	register struct ucred *cred = p->p_ucred;
     82 	struct vattr va;
     83 	int error;
     84 
     85 	if (fmode & O_CREAT) {
     86 		ndp->ni_cnd.cn_nameiop = CREATE;
     87 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
     88 		if ((fmode & O_EXCL) == 0)
     89 			ndp->ni_cnd.cn_flags |= FOLLOW;
     90 		if ((error = namei(ndp)) != 0)
     91 			return (error);
     92 		if (ndp->ni_vp == NULL) {
     93 			VATTR_NULL(&va);
     94 			va.va_type = VREG;
     95 			va.va_mode = cmode;
     96 			if (fmode & O_EXCL)
     97 				 va.va_vaflags |= VA_EXCLUSIVE;
     98 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
     99 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
    100 					   &ndp->ni_cnd, &va);
    101 			if (error)
    102 				return (error);
    103 			fmode &= ~O_TRUNC;
    104 			vp = ndp->ni_vp;
    105 		} else {
    106 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    107 			if (ndp->ni_dvp == ndp->ni_vp)
    108 				vrele(ndp->ni_dvp);
    109 			else
    110 				vput(ndp->ni_dvp);
    111 			ndp->ni_dvp = NULL;
    112 			vp = ndp->ni_vp;
    113 			if (fmode & O_EXCL) {
    114 				error = EEXIST;
    115 				goto bad;
    116 			}
    117 			fmode &= ~O_CREAT;
    118 		}
    119 	} else {
    120 		ndp->ni_cnd.cn_nameiop = LOOKUP;
    121 		ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF;
    122 		if ((error = namei(ndp)) != 0)
    123 			return (error);
    124 		vp = ndp->ni_vp;
    125 	}
    126 	if (vp->v_type == VSOCK) {
    127 		error = EOPNOTSUPP;
    128 		goto bad;
    129 	}
    130 	if ((fmode & O_CREAT) == 0) {
    131 		if (fmode & FREAD) {
    132 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
    133 				goto bad;
    134 		}
    135 		if (fmode & (FWRITE | O_TRUNC)) {
    136 			if (vp->v_type == VDIR) {
    137 				error = EISDIR;
    138 				goto bad;
    139 			}
    140 			if ((error = vn_writechk(vp)) != 0 ||
    141 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
    142 				goto bad;
    143 		}
    144 	}
    145 	if (fmode & O_TRUNC) {
    146 		VOP_UNLOCK(vp, 0);			/* XXX */
    147 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
    148 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
    149 		VATTR_NULL(&va);
    150 		va.va_size = 0;
    151 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
    152 			goto bad;
    153 	}
    154 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
    155 		goto bad;
    156 	if (fmode & FWRITE)
    157 		vp->v_writecount++;
    158 	return (0);
    159 bad:
    160 	vput(vp);
    161 	return (error);
    162 }
    163 
    164 /*
    165  * Check for write permissions on the specified vnode.
    166  * Prototype text segments cannot be written.
    167  */
    168 int
    169 vn_writechk(vp)
    170 	register struct vnode *vp;
    171 {
    172 
    173 	/*
    174 	 * If there's shared text associated with
    175 	 * the vnode, try to free it up once.  If
    176 	 * we fail, we can't allow writing.
    177 	 */
    178 	if ((vp->v_flag & VTEXT) && !uvm_vnp_uncache(vp))
    179 		return (ETXTBSY);
    180 	return (0);
    181 }
    182 
    183 /*
    184  * Vnode close call
    185  *
    186  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
    187  */
    188 int
    189 vn_close(vp, flags, cred, p)
    190 	register struct vnode *vp;
    191 	int flags;
    192 	struct ucred *cred;
    193 	struct proc *p;
    194 {
    195 	int error;
    196 
    197 	if (flags & FWRITE)
    198 		vp->v_writecount--;
    199 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    200 	error = VOP_CLOSE(vp, flags, cred, p);
    201 	vput(vp);
    202 	return (error);
    203 }
    204 
    205 /*
    206  * Package up an I/O request on a vnode into a uio and do it.
    207  */
    208 int
    209 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p)
    210 	enum uio_rw rw;
    211 	struct vnode *vp;
    212 	caddr_t base;
    213 	int len;
    214 	off_t offset;
    215 	enum uio_seg segflg;
    216 	int ioflg;
    217 	struct ucred *cred;
    218 	size_t *aresid;
    219 	struct proc *p;
    220 {
    221 	struct uio auio;
    222 	struct iovec aiov;
    223 	int error;
    224 
    225 	if ((ioflg & IO_NODELOCKED) == 0)
    226 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    227 	auio.uio_iov = &aiov;
    228 	auio.uio_iovcnt = 1;
    229 	aiov.iov_base = base;
    230 	aiov.iov_len = len;
    231 	auio.uio_resid = len;
    232 	auio.uio_offset = offset;
    233 	auio.uio_segflg = segflg;
    234 	auio.uio_rw = rw;
    235 	auio.uio_procp = p;
    236 	if (rw == UIO_READ) {
    237 		error = VOP_READ(vp, &auio, ioflg, cred);
    238 	} else {
    239 		error = VOP_WRITE(vp, &auio, ioflg, cred);
    240 	}
    241 	if (aresid)
    242 		*aresid = auio.uio_resid;
    243 	else
    244 		if (auio.uio_resid && error == 0)
    245 			error = EIO;
    246 	if ((ioflg & IO_NODELOCKED) == 0)
    247 		VOP_UNLOCK(vp, 0);
    248 	return (error);
    249 }
    250 
    251 int
    252 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies)
    253 	struct file *fp;
    254 	char *buf;
    255 	int segflg, *done, *ncookies;
    256 	u_int count;
    257 	struct proc *p;
    258 	off_t **cookies;
    259 {
    260 	struct vnode *vp = (struct vnode *)fp->f_data;
    261 	struct iovec aiov;
    262 	struct uio auio;
    263 	int error, eofflag;
    264 
    265 unionread:
    266 	if (vp->v_type != VDIR)
    267 		return (EINVAL);
    268 	aiov.iov_base = buf;
    269 	aiov.iov_len = count;
    270 	auio.uio_iov = &aiov;
    271 	auio.uio_iovcnt = 1;
    272 	auio.uio_rw = UIO_READ;
    273 	auio.uio_segflg = segflg;
    274 	auio.uio_procp = p;
    275 	auio.uio_resid = count;
    276 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    277 	auio.uio_offset = fp->f_offset;
    278 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
    279 		    ncookies);
    280 	fp->f_offset = auio.uio_offset;
    281 	VOP_UNLOCK(vp, 0);
    282 	if (error)
    283 		return (error);
    284 
    285 #ifdef UNION
    286 {
    287 	extern int (**union_vnodeop_p) __P((void *));
    288 	extern struct vnode *union_dircache __P((struct vnode *));
    289 
    290 	if (count == auio.uio_resid && (vp->v_op == union_vnodeop_p)) {
    291 		struct vnode *lvp;
    292 
    293 		lvp = union_dircache(vp);
    294 		if (lvp != NULLVP) {
    295 			struct vattr va;
    296 
    297 			/*
    298 			 * If the directory is opaque,
    299 			 * then don't show lower entries
    300 			 */
    301 			error = VOP_GETATTR(vp, &va, fp->f_cred, p);
    302 			if (va.va_flags & OPAQUE) {
    303 				vput(lvp);
    304 				lvp = NULL;
    305 			}
    306 		}
    307 
    308 		if (lvp != NULLVP) {
    309 			error = VOP_OPEN(lvp, FREAD, fp->f_cred, p);
    310 			if (error) {
    311 				vput(lvp);
    312 				return (error);
    313 			}
    314 			VOP_UNLOCK(lvp, 0);
    315 			fp->f_data = (caddr_t) lvp;
    316 			fp->f_offset = 0;
    317 			error = vn_close(vp, FREAD, fp->f_cred, p);
    318 			if (error)
    319 				return (error);
    320 			vp = lvp;
    321 			goto unionread;
    322 		}
    323 	}
    324 }
    325 #endif /* UNION */
    326 
    327 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
    328 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
    329 		struct vnode *tvp = vp;
    330 		vp = vp->v_mount->mnt_vnodecovered;
    331 		VREF(vp);
    332 		fp->f_data = (caddr_t) vp;
    333 		fp->f_offset = 0;
    334 		vrele(tvp);
    335 		goto unionread;
    336 	}
    337 	*done = count - auio.uio_resid;
    338 	return error;
    339 }
    340 
    341 /*
    342  * File table vnode read routine.
    343  */
    344 int
    345 vn_read(fp, offset, uio, cred, flags)
    346 	struct file *fp;
    347 	off_t *offset;
    348 	struct uio *uio;
    349 	struct ucred *cred;
    350 	int flags;
    351 {
    352 	struct vnode *vp = (struct vnode *)fp->f_data;
    353 	int count, error, ioflag = 0;
    354 
    355 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
    356 	if (fp->f_flag & FNONBLOCK)
    357 		ioflag |= IO_NDELAY;
    358 	if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
    359 		ioflag |= IO_SYNC;
    360 	if (fp->f_flag & FALTIO)
    361 		ioflag |= IO_ALTSEMANTICS;
    362 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    363 	uio->uio_offset = *offset;
    364 	count = uio->uio_resid;
    365 	error = VOP_READ(vp, uio, ioflag, cred);
    366 	if (flags & FOF_UPDATE_OFFSET)
    367 		*offset += count - uio->uio_resid;
    368 	VOP_UNLOCK(vp, 0);
    369 	return (error);
    370 }
    371 
    372 /*
    373  * File table vnode write routine.
    374  */
    375 int
    376 vn_write(fp, offset, uio, cred, flags)
    377 	struct file *fp;
    378 	off_t *offset;
    379 	struct uio *uio;
    380 	struct ucred *cred;
    381 	int flags;
    382 {
    383 	struct vnode *vp = (struct vnode *)fp->f_data;
    384 	int count, error, ioflag = IO_UNIT;
    385 
    386 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
    387 		ioflag |= IO_APPEND;
    388 	if (fp->f_flag & FNONBLOCK)
    389 		ioflag |= IO_NDELAY;
    390 	if (fp->f_flag & FFSYNC ||
    391 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
    392 		ioflag |= IO_SYNC;
    393 	else if (fp->f_flag & FDSYNC)
    394 		ioflag |= IO_DSYNC;
    395 	if (fp->f_flag & FALTIO)
    396 		ioflag |= IO_ALTSEMANTICS;
    397 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
    398 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    399 	uio->uio_offset = *offset;
    400 	count = uio->uio_resid;
    401 	error = VOP_WRITE(vp, uio, ioflag, cred);
    402 	if (flags & FOF_UPDATE_OFFSET) {
    403 		if (ioflag & IO_APPEND)
    404 			*offset = uio->uio_offset;
    405 		else
    406 			*offset += count - uio->uio_resid;
    407 	}
    408 	VOP_UNLOCK(vp, 0);
    409 	return (error);
    410 }
    411 
    412 /*
    413  * File table vnode stat routine.
    414  */
    415 int
    416 vn_stat(vp, sb, p)
    417 	struct vnode *vp;
    418 	register struct stat *sb;
    419 	struct proc *p;
    420 {
    421 	struct vattr va;
    422 	int error;
    423 	mode_t mode;
    424 
    425 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
    426 	if (error)
    427 		return (error);
    428 	/*
    429 	 * Copy from vattr table
    430 	 */
    431 	sb->st_dev = va.va_fsid;
    432 	sb->st_ino = va.va_fileid;
    433 	mode = va.va_mode;
    434 	switch (vp->v_type) {
    435 	case VREG:
    436 		mode |= S_IFREG;
    437 		break;
    438 	case VDIR:
    439 		mode |= S_IFDIR;
    440 		break;
    441 	case VBLK:
    442 		mode |= S_IFBLK;
    443 		break;
    444 	case VCHR:
    445 		mode |= S_IFCHR;
    446 		break;
    447 	case VLNK:
    448 		mode |= S_IFLNK;
    449 		break;
    450 	case VSOCK:
    451 		mode |= S_IFSOCK;
    452 		break;
    453 	case VFIFO:
    454 		mode |= S_IFIFO;
    455 		break;
    456 	default:
    457 		return (EBADF);
    458 	};
    459 	sb->st_mode = mode;
    460 	sb->st_nlink = va.va_nlink;
    461 	sb->st_uid = va.va_uid;
    462 	sb->st_gid = va.va_gid;
    463 	sb->st_rdev = va.va_rdev;
    464 	sb->st_size = va.va_size;
    465 	sb->st_atimespec = va.va_atime;
    466 	sb->st_mtimespec = va.va_mtime;
    467 	sb->st_ctimespec = va.va_ctime;
    468 	sb->st_blksize = va.va_blocksize;
    469 	sb->st_flags = va.va_flags;
    470 	sb->st_gen = 0;
    471 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
    472 	return (0);
    473 }
    474 
    475 /*
    476  * File table vnode fcntl routine.
    477  */
    478 int
    479 vn_fcntl(fp, com, data, p)
    480 	struct file *fp;
    481 	u_int com;
    482 	caddr_t data;
    483 	struct proc *p;
    484 {
    485 	register struct vnode *vp = ((struct vnode *)fp->f_data);
    486 	int error;
    487 
    488 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    489 	error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    490 	VOP_UNLOCK(vp, 0);
    491 	return (error);
    492 }
    493 
    494 /*
    495  * File table vnode ioctl routine.
    496  */
    497 int
    498 vn_ioctl(fp, com, data, p)
    499 	struct file *fp;
    500 	u_long com;
    501 	caddr_t data;
    502 	struct proc *p;
    503 {
    504 	register struct vnode *vp = ((struct vnode *)fp->f_data);
    505 	struct vattr vattr;
    506 	int error;
    507 
    508 	switch (vp->v_type) {
    509 
    510 	case VREG:
    511 	case VDIR:
    512 		if (com == FIONREAD) {
    513 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
    514 			if (error)
    515 				return (error);
    516 			*(int *)data = vattr.va_size - fp->f_offset;
    517 			return (0);
    518 		}
    519 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
    520 			return (0);			/* XXX */
    521 		/* fall into ... */
    522 
    523 	default:
    524 		return (ENOTTY);
    525 
    526 	case VFIFO:
    527 	case VCHR:
    528 	case VBLK:
    529 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    530 		if (error == 0 && com == TIOCSCTTY) {
    531 			if (p->p_session->s_ttyvp)
    532 				vrele(p->p_session->s_ttyvp);
    533 			p->p_session->s_ttyvp = vp;
    534 			VREF(vp);
    535 		}
    536 		return (error);
    537 	}
    538 }
    539 
    540 /*
    541  * File table vnode poll routine.
    542  */
    543 int
    544 vn_poll(fp, events, p)
    545 	struct file *fp;
    546 	int events;
    547 	struct proc *p;
    548 {
    549 
    550 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
    551 }
    552 
    553 /*
    554  * Check that the vnode is still valid, and if so
    555  * acquire requested lock.
    556  */
    557 int
    558 vn_lock(vp, flags)
    559 	struct vnode *vp;
    560 	int flags;
    561 {
    562 	int error;
    563 
    564 	do {
    565 		if ((flags & LK_INTERLOCK) == 0)
    566 			simple_lock(&vp->v_interlock);
    567 		if (vp->v_flag & VXLOCK) {
    568 			vp->v_flag |= VXWANT;
    569 			simple_unlock(&vp->v_interlock);
    570 			tsleep((caddr_t)vp, PINOD, "vn_lock", 0);
    571 			error = ENOENT;
    572 		} else {
    573 			error = VOP_LOCK(vp, flags | LK_INTERLOCK);
    574 			if (error == 0 || error == EDEADLK)
    575 				return (error);
    576 		}
    577 		flags &= ~LK_INTERLOCK;
    578 	} while (flags & LK_RETRY);
    579 	return (error);
    580 }
    581 
    582 /*
    583  * File table vnode close routine.
    584  */
    585 int
    586 vn_closefile(fp, p)
    587 	struct file *fp;
    588 	struct proc *p;
    589 {
    590 
    591 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
    592 		fp->f_cred, p));
    593 }
    594