Home | History | Annotate | Line # | Download | only in kern
vfs_vnops.c revision 1.31.2.1
      1 /*	$NetBSD: vfs_vnops.c,v 1.31.2.1 1998/11/09 06:06:33 chs 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 #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 #ifdef UNION
     67 #include <miscfs/union/union.h>
     68 #endif
     69 
     70 struct 	fileops vnops =
     71 	{ vn_read, vn_write, vn_ioctl, vn_poll, vn_closefile };
     72 
     73 /*
     74  * Common code for vnode open operations.
     75  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
     76  */
     77 int
     78 vn_open(ndp, fmode, cmode)
     79 	register struct nameidata *ndp;
     80 	int fmode, cmode;
     81 {
     82 	register struct vnode *vp;
     83 	register struct proc *p = ndp->ni_cnd.cn_proc;
     84 	register struct ucred *cred = p->p_ucred;
     85 	struct vattr va;
     86 	int error;
     87 
     88 	if (fmode & O_CREAT) {
     89 		ndp->ni_cnd.cn_nameiop = CREATE;
     90 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
     91 		if ((fmode & O_EXCL) == 0)
     92 			ndp->ni_cnd.cn_flags |= FOLLOW;
     93 		if ((error = namei(ndp)) != 0)
     94 			return (error);
     95 		if (ndp->ni_vp == NULL) {
     96 			VATTR_NULL(&va);
     97 			va.va_type = VREG;
     98 			va.va_mode = cmode;
     99 			if (fmode & O_EXCL)
    100 				 va.va_vaflags |= VA_EXCLUSIVE;
    101 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
    102 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
    103 					   &ndp->ni_cnd, &va);
    104 			if (error)
    105 				return (error);
    106 			fmode &= ~O_TRUNC;
    107 			vp = ndp->ni_vp;
    108 		} else {
    109 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    110 			if (ndp->ni_dvp == ndp->ni_vp)
    111 				vrele(ndp->ni_dvp);
    112 			else
    113 				vput(ndp->ni_dvp);
    114 			ndp->ni_dvp = NULL;
    115 			vp = ndp->ni_vp;
    116 			if (fmode & O_EXCL) {
    117 				error = EEXIST;
    118 				goto bad;
    119 			}
    120 			fmode &= ~O_CREAT;
    121 		}
    122 	} else {
    123 		ndp->ni_cnd.cn_nameiop = LOOKUP;
    124 		ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF;
    125 		if ((error = namei(ndp)) != 0)
    126 			return (error);
    127 		vp = ndp->ni_vp;
    128 	}
    129 	if (vp->v_type == VSOCK) {
    130 		error = EOPNOTSUPP;
    131 		goto bad;
    132 	}
    133 	if ((fmode & O_CREAT) == 0) {
    134 		if (fmode & FREAD) {
    135 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
    136 				goto bad;
    137 		}
    138 		if (fmode & (FWRITE | O_TRUNC)) {
    139 			if (vp->v_type == VDIR) {
    140 				error = EISDIR;
    141 				goto bad;
    142 			}
    143 			if ((error = vn_writechk(vp)) != 0 ||
    144 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
    145 				goto bad;
    146 		}
    147 	}
    148 	if (fmode & O_TRUNC) {
    149 		VOP_UNLOCK(vp, 0);			/* XXX */
    150 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
    151 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
    152 		VATTR_NULL(&va);
    153 		va.va_size = 0;
    154 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
    155 			goto bad;
    156 	}
    157 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
    158 		goto bad;
    159 	if (fmode & FWRITE)
    160 		vp->v_writecount++;
    161 
    162 #ifdef UBC
    163 	if (vp->v_type == VREG &&
    164 	    uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
    165 		error = EIO;
    166 		goto bad;
    167 	}
    168 #endif
    169 	return (0);
    170 bad:
    171 	vput(vp);
    172 	return (error);
    173 }
    174 
    175 /*
    176  * Check for write permissions on the specified vnode.
    177  * Prototype text segments cannot be written.
    178  */
    179 int
    180 vn_writechk(vp)
    181 	register struct vnode *vp;
    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 	size_t *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 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    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, 0);
    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 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    290 	auio.uio_offset = fp->f_offset;
    291 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
    292 		    ncookies);
    293 	fp->f_offset = auio.uio_offset;
    294 	VOP_UNLOCK(vp, 0);
    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 			if (error) {
    324 				vput(lvp);
    325 				return (error);
    326 			}
    327 			VOP_UNLOCK(lvp, 0);
    328 			fp->f_data = (caddr_t) lvp;
    329 			fp->f_offset = 0;
    330 			error = vn_close(vp, FREAD, fp->f_cred, p);
    331 			if (error)
    332 				return (error);
    333 			vp = lvp;
    334 			goto unionread;
    335 		}
    336 	}
    337 }
    338 #endif /* UNION */
    339 
    340 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
    341 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
    342 		struct vnode *tvp = vp;
    343 		vp = vp->v_mount->mnt_vnodecovered;
    344 		VREF(vp);
    345 		fp->f_data = (caddr_t) vp;
    346 		fp->f_offset = 0;
    347 		vrele(tvp);
    348 		goto unionread;
    349 	}
    350 	*done = count - auio.uio_resid;
    351 	return error;
    352 }
    353 
    354 /*
    355  * File table vnode read routine.
    356  */
    357 int
    358 vn_read(fp, offset, uio, cred, flags)
    359 	struct file *fp;
    360 	off_t *offset;
    361 	struct uio *uio;
    362 	struct ucred *cred;
    363 	int flags;
    364 {
    365 	struct vnode *vp = (struct vnode *)fp->f_data;
    366 	int count, error, ioflag = 0;
    367 
    368 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
    369 	if (fp->f_flag & FNONBLOCK)
    370 		ioflag |= IO_NDELAY;
    371 	if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
    372 		ioflag |= IO_SYNC;
    373 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    374 	uio->uio_offset = *offset;
    375 	count = uio->uio_resid;
    376 	error = VOP_READ(vp, uio, ioflag, cred);
    377 	if (flags & FOF_UPDATE_OFFSET)
    378 		*offset += count - uio->uio_resid;
    379 	VOP_UNLOCK(vp, 0);
    380 	return (error);
    381 }
    382 
    383 /*
    384  * File table vnode write routine.
    385  */
    386 int
    387 vn_write(fp, offset, uio, cred, flags)
    388 	struct file *fp;
    389 	off_t *offset;
    390 	struct uio *uio;
    391 	struct ucred *cred;
    392 	int flags;
    393 {
    394 	struct vnode *vp = (struct vnode *)fp->f_data;
    395 	int count, error, ioflag = IO_UNIT;
    396 
    397 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
    398 		ioflag |= IO_APPEND;
    399 	if (fp->f_flag & FNONBLOCK)
    400 		ioflag |= IO_NDELAY;
    401 	if (fp->f_flag & FFSYNC ||
    402 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
    403 		ioflag |= IO_SYNC;
    404 	else if (fp->f_flag & FDSYNC)
    405 		ioflag |= IO_DSYNC;
    406 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
    407 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    408 	uio->uio_offset = *offset;
    409 	count = uio->uio_resid;
    410 	error = VOP_WRITE(vp, uio, ioflag, cred);
    411 	if (flags & FOF_UPDATE_OFFSET) {
    412 		if (ioflag & IO_APPEND)
    413 			*offset = uio->uio_offset;
    414 		else
    415 			*offset += count - uio->uio_resid;
    416 	}
    417 	VOP_UNLOCK(vp, 0);
    418 	return (error);
    419 }
    420 
    421 /*
    422  * File table vnode stat routine.
    423  */
    424 int
    425 vn_stat(vp, sb, p)
    426 	struct vnode *vp;
    427 	register struct stat *sb;
    428 	struct proc *p;
    429 {
    430 	struct vattr va;
    431 	int error;
    432 	u_short mode;
    433 
    434 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
    435 	if (error)
    436 		return (error);
    437 	/*
    438 	 * Copy from vattr table
    439 	 */
    440 	sb->st_dev = va.va_fsid;
    441 	sb->st_ino = va.va_fileid;
    442 	mode = va.va_mode;
    443 	switch (vp->v_type) {
    444 	case VREG:
    445 		mode |= S_IFREG;
    446 		break;
    447 	case VDIR:
    448 		mode |= S_IFDIR;
    449 		break;
    450 	case VBLK:
    451 		mode |= S_IFBLK;
    452 		break;
    453 	case VCHR:
    454 		mode |= S_IFCHR;
    455 		break;
    456 	case VLNK:
    457 		mode |= S_IFLNK;
    458 		break;
    459 	case VSOCK:
    460 		mode |= S_IFSOCK;
    461 		break;
    462 	case VFIFO:
    463 		mode |= S_IFIFO;
    464 		break;
    465 	default:
    466 		return (EBADF);
    467 	};
    468 	sb->st_mode = mode;
    469 	sb->st_nlink = va.va_nlink;
    470 	sb->st_uid = va.va_uid;
    471 	sb->st_gid = va.va_gid;
    472 	sb->st_rdev = va.va_rdev;
    473 	sb->st_size = va.va_size;
    474 	sb->st_atimespec = va.va_atime;
    475 	sb->st_mtimespec = va.va_mtime;
    476 	sb->st_ctimespec = va.va_ctime;
    477 	sb->st_blksize = va.va_blocksize;
    478 	sb->st_flags = va.va_flags;
    479 	sb->st_gen = 0;
    480 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
    481 	return (0);
    482 }
    483 
    484 /*
    485  * File table vnode ioctl routine.
    486  */
    487 int
    488 vn_ioctl(fp, com, data, p)
    489 	struct file *fp;
    490 	u_long com;
    491 	caddr_t data;
    492 	struct proc *p;
    493 {
    494 	register struct vnode *vp = ((struct vnode *)fp->f_data);
    495 	struct vattr vattr;
    496 	int error;
    497 
    498 	switch (vp->v_type) {
    499 
    500 	case VREG:
    501 	case VDIR:
    502 		if (com == FIONREAD) {
    503 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
    504 			if (error)
    505 				return (error);
    506 			*(int *)data = vattr.va_size - fp->f_offset;
    507 			return (0);
    508 		}
    509 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
    510 			return (0);			/* XXX */
    511 		/* fall into ... */
    512 
    513 	default:
    514 		return (ENOTTY);
    515 
    516 	case VFIFO:
    517 	case VCHR:
    518 	case VBLK:
    519 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    520 		if (error == 0 && com == TIOCSCTTY) {
    521 			if (p->p_session->s_ttyvp)
    522 				vrele(p->p_session->s_ttyvp);
    523 			p->p_session->s_ttyvp = vp;
    524 			VREF(vp);
    525 		}
    526 		return (error);
    527 	}
    528 }
    529 
    530 /*
    531  * File table vnode poll routine.
    532  */
    533 int
    534 vn_poll(fp, events, p)
    535 	struct file *fp;
    536 	int events;
    537 	struct proc *p;
    538 {
    539 
    540 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
    541 }
    542 
    543 /*
    544  * Check that the vnode is still valid, and if so
    545  * acquire requested lock.
    546  */
    547 int
    548 vn_lock(vp, flags)
    549 	struct vnode *vp;
    550 	int flags;
    551 {
    552 	int error;
    553 
    554 	do {
    555 		if ((flags & LK_INTERLOCK) == 0)
    556 			simple_lock(&vp->v_interlock);
    557 		if (vp->v_flag & VXLOCK) {
    558 			vp->v_flag |= VXWANT;
    559 			simple_unlock(&vp->v_interlock);
    560 			tsleep((caddr_t)vp, PINOD, "vn_lock", 0);
    561 			error = ENOENT;
    562 		} else {
    563 			error = VOP_LOCK(vp, flags | LK_INTERLOCK);
    564 			if (error == 0)
    565 				return (error);
    566 		}
    567 		flags &= ~LK_INTERLOCK;
    568 	} while (flags & LK_RETRY);
    569 	return (error);
    570 }
    571 
    572 /*
    573  * File table vnode close routine.
    574  */
    575 int
    576 vn_closefile(fp, p)
    577 	struct file *fp;
    578 	struct proc *p;
    579 {
    580 
    581 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
    582 		fp->f_cred, p));
    583 }
    584