Home | History | Annotate | Line # | Download | only in kern
vfs_vnops.c revision 1.72
      1 /*	$NetBSD: vfs_vnops.c,v 1.72 2003/06/29 22:31:35 fvdl 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 <sys/cdefs.h>
     44 __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.72 2003/06/29 22:31:35 fvdl Exp $");
     45 
     46 #include "fs_union.h"
     47 
     48 #include <sys/param.h>
     49 #include <sys/systm.h>
     50 #include <sys/kernel.h>
     51 #include <sys/file.h>
     52 #include <sys/stat.h>
     53 #include <sys/buf.h>
     54 #include <sys/proc.h>
     55 #include <sys/mount.h>
     56 #include <sys/namei.h>
     57 #include <sys/vnode.h>
     58 #include <sys/ioctl.h>
     59 #include <sys/tty.h>
     60 #include <sys/poll.h>
     61 
     62 #include <uvm/uvm_extern.h>
     63 
     64 #ifdef UNION
     65 #include <fs/union/union.h>
     66 #endif
     67 
     68 #if defined(LKM) || defined(UNION)
     69 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct proc *);
     70 #endif
     71 
     72 #ifdef VERIFIED_EXEC
     73 #include <sys/verified_exec.h>
     74 
     75 extern LIST_HEAD(veriexec_devhead, veriexec_dev_list) veriexec_dev_head;
     76 extern struct veriexec_devhead veriexec_file_dev_head;
     77 #endif
     78 
     79 static int vn_read(struct file *fp, off_t *offset, struct uio *uio,
     80 	    struct ucred *cred, int flags);
     81 static int vn_write(struct file *fp, off_t *offset, struct uio *uio,
     82 	    struct ucred *cred, int flags);
     83 static int vn_closefile(struct file *fp, struct proc *p);
     84 static int vn_poll(struct file *fp, int events, struct proc *p);
     85 static int vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p);
     86 static int vn_statfile(struct file *fp, struct stat *sb, struct proc *p);
     87 static int vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p);
     88 
     89 struct 	fileops vnops = {
     90 	vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll,
     91 	vn_statfile, vn_closefile, vn_kqfilter
     92 };
     93 
     94 /*
     95  * Common code for vnode open operations.
     96  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
     97  */
     98 int
     99 vn_open(ndp, fmode, cmode)
    100 	struct nameidata *ndp;
    101 	int fmode, cmode;
    102 {
    103 	struct vnode *vp;
    104 	struct proc *p = ndp->ni_cnd.cn_proc;
    105 	struct ucred *cred = p->p_ucred;
    106 	struct vattr va;
    107 	int error;
    108 #ifdef VERIFIED_EXEC
    109 	char got_dev;
    110 	struct veriexec_inode_list *veriexec_node;
    111 	char fingerprint[MAXFINGERPRINTLEN];
    112 #endif
    113 
    114 	if (fmode & O_CREAT) {
    115 		ndp->ni_cnd.cn_nameiop = CREATE;
    116 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
    117 		if ((fmode & O_EXCL) == 0 &&
    118 		    ((fmode & O_NOFOLLOW) == 0))
    119 			ndp->ni_cnd.cn_flags |= FOLLOW;
    120 		if ((error = namei(ndp)) != 0)
    121 			return (error);
    122 		if (ndp->ni_vp == NULL) {
    123 			VATTR_NULL(&va);
    124 			va.va_type = VREG;
    125 			va.va_mode = cmode;
    126 			if (fmode & O_EXCL)
    127 				 va.va_vaflags |= VA_EXCLUSIVE;
    128 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
    129 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
    130 					   &ndp->ni_cnd, &va);
    131 			if (error)
    132 				return (error);
    133 			fmode &= ~O_TRUNC;
    134 			vp = ndp->ni_vp;
    135 		} else {
    136 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    137 			if (ndp->ni_dvp == ndp->ni_vp)
    138 				vrele(ndp->ni_dvp);
    139 			else
    140 				vput(ndp->ni_dvp);
    141 			ndp->ni_dvp = NULL;
    142 			vp = ndp->ni_vp;
    143 			if (fmode & O_EXCL) {
    144 				error = EEXIST;
    145 				goto bad;
    146 			}
    147 			if (ndp->ni_vp->v_type == VLNK) {
    148 				error = EFTYPE;
    149 				goto bad;
    150 			}
    151 			fmode &= ~O_CREAT;
    152 		}
    153 	} else {
    154 		ndp->ni_cnd.cn_nameiop = LOOKUP;
    155 		ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF;
    156 		if ((error = namei(ndp)) != 0)
    157 			return (error);
    158 		vp = ndp->ni_vp;
    159 	}
    160 	if (vp->v_type == VSOCK) {
    161 		error = EOPNOTSUPP;
    162 		goto bad;
    163 	}
    164 
    165 #ifdef VERIFIED_EXEC
    166 	veriexec_node = NULL;
    167 
    168 	if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0)
    169 		goto bad;
    170 #endif
    171 
    172 	if ((fmode & O_CREAT) == 0) {
    173 #ifdef VERIFIED_EXEC
    174 		  /*
    175 		   * Look for the file on the fingerprint lists iff
    176 		   * it has not been seen before.
    177 		   */
    178 		if ((vp->fp_status == FINGERPRINT_INVALID) ||
    179 		    (vp->fp_status == FINGERPRINT_NODEV)) {
    180 			  /* check the file list for the finger print */
    181 			veriexec_node = get_veriexec_inode(&veriexec_file_dev_head,
    182 						     va.va_fsid,
    183 						     va.va_fileid,
    184 						     &got_dev);
    185 			if (veriexec_node == NULL) {
    186 				/* failing that, check the exec list */
    187 				veriexec_node = get_veriexec_inode(
    188 					&veriexec_dev_head, va.va_fsid,
    189 					va.va_fileid, &got_dev);
    190 			}
    191 
    192 			if ((veriexec_node == NULL) && (got_dev == 1))
    193 				vp->fp_status = FINGERPRINT_NOENTRY;
    194 
    195 			if (veriexec_node != NULL) {
    196 				if ((error = evaluate_fingerprint(vp,
    197 						veriexec_node, p, va.va_size,
    198 						fingerprint)) != 0)
    199 					goto bad;
    200 
    201 				if (fingerprintcmp(veriexec_node,
    202 						   fingerprint) == 0) {
    203 					  /* fingerprint ok */
    204 					vp->fp_status =	FINGERPRINT_VALID;
    205 #ifdef VERIFIED_EXEC_DEBUG
    206 					printf(
    207 			"file fingerprint matches for dev %lu, file %lu\n",
    208 						va.va_fsid, va.va_fileid);
    209 #endif
    210 				} else {
    211 					vp->fp_status =	FINGERPRINT_NOMATCH;
    212 				}
    213 			}
    214 		}
    215 #endif
    216 
    217 		if (fmode & FREAD) {
    218 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
    219 				goto bad;
    220 
    221 #ifdef VERIFIED_EXEC
    222 				/* file is on finger print list */
    223 			if (vp->fp_status == FINGERPRINT_NOMATCH) {
    224 				  /* fingerprint bad */
    225 				printf(
    226 		"file fingerprint does not match on dev %lu, file %lu\n",
    227 					va.va_fsid, va.va_fileid);
    228 				if (securelevel > 2) {
    229 					error = EPERM;
    230 					goto bad;
    231 				}
    232 			}
    233 #endif
    234 		}
    235 		if (fmode & (FWRITE | O_TRUNC)) {
    236 			if (vp->v_type == VDIR) {
    237 				error = EISDIR;
    238 				goto bad;
    239 			}
    240 			if ((error = vn_writechk(vp)) != 0 ||
    241 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
    242 				goto bad;
    243 #ifdef VERIFIED_EXEC
    244 			  /*
    245 			   * If file has a fingerprint then
    246 			   * deny the write request, otherwise
    247 			   * invalidate the status so we don't
    248 			   * keep checking for the file having
    249 			   * a fingerprint.
    250 			   */
    251 			if (vp->fp_status == FINGERPRINT_VALID) {
    252 				printf(
    253 		      "writing to fingerprinted file for dev %lu, file %lu\n",
    254 		      va.va_fsid, va.va_fileid);
    255 				if (securelevel > 2) {
    256 					error = EPERM;
    257 					goto bad;
    258 				} else {
    259 					vp->fp_status =	FINGERPRINT_INVALID;
    260 				}
    261 			}
    262 #endif
    263 		}
    264 	}
    265 	if (fmode & O_TRUNC) {
    266 		VOP_UNLOCK(vp, 0);			/* XXX */
    267 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
    268 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
    269 		VATTR_NULL(&va);
    270 		va.va_size = 0;
    271 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
    272 			goto bad;
    273 	}
    274 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
    275 		goto bad;
    276 	if (vp->v_type == VREG &&
    277 	    uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
    278 		error = EIO;
    279 		goto bad;
    280 	}
    281 	if (fmode & FWRITE)
    282 		vp->v_writecount++;
    283 
    284 	return (0);
    285 bad:
    286 	vput(vp);
    287 	return (error);
    288 }
    289 
    290 /*
    291  * Check for write permissions on the specified vnode.
    292  * Prototype text segments cannot be written.
    293  */
    294 int
    295 vn_writechk(vp)
    296 	struct vnode *vp;
    297 {
    298 
    299 	/*
    300 	 * If the vnode is in use as a process's text,
    301 	 * we can't allow writing.
    302 	 */
    303 	if (vp->v_flag & VTEXT)
    304 		return (ETXTBSY);
    305 	return (0);
    306 }
    307 
    308 /*
    309  * Mark a vnode as having executable mappings.
    310  */
    311 void
    312 vn_markexec(vp)
    313 	struct vnode *vp;
    314 {
    315 	if ((vp->v_flag & VEXECMAP) == 0) {
    316 		uvmexp.filepages -= vp->v_uobj.uo_npages;
    317 		uvmexp.execpages += vp->v_uobj.uo_npages;
    318 	}
    319 	vp->v_flag |= VEXECMAP;
    320 }
    321 
    322 /*
    323  * Mark a vnode as being the text of a process.
    324  * Fail if the vnode is currently writable.
    325  */
    326 int
    327 vn_marktext(vp)
    328 	struct vnode *vp;
    329 {
    330 
    331 	if (vp->v_writecount != 0) {
    332 		KASSERT((vp->v_flag & VTEXT) == 0);
    333 		return (ETXTBSY);
    334 	}
    335 	vp->v_flag |= VTEXT;
    336 	vn_markexec(vp);
    337 	return (0);
    338 }
    339 
    340 /*
    341  * Vnode close call
    342  *
    343  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
    344  */
    345 int
    346 vn_close(vp, flags, cred, p)
    347 	struct vnode *vp;
    348 	int flags;
    349 	struct ucred *cred;
    350 	struct proc *p;
    351 {
    352 	int error;
    353 
    354 	if (flags & FWRITE)
    355 		vp->v_writecount--;
    356 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    357 	error = VOP_CLOSE(vp, flags, cred, p);
    358 	vput(vp);
    359 	return (error);
    360 }
    361 
    362 /*
    363  * Package up an I/O request on a vnode into a uio and do it.
    364  */
    365 int
    366 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p)
    367 	enum uio_rw rw;
    368 	struct vnode *vp;
    369 	caddr_t base;
    370 	int len;
    371 	off_t offset;
    372 	enum uio_seg segflg;
    373 	int ioflg;
    374 	struct ucred *cred;
    375 	size_t *aresid;
    376 	struct proc *p;
    377 {
    378 	struct uio auio;
    379 	struct iovec aiov;
    380 	int error;
    381 
    382 	if ((ioflg & IO_NODELOCKED) == 0) {
    383 		if (rw == UIO_READ) {
    384 			vn_lock(vp, LK_SHARED | LK_RETRY);
    385 		} else {
    386 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    387 		}
    388 	}
    389 	auio.uio_iov = &aiov;
    390 	auio.uio_iovcnt = 1;
    391 	aiov.iov_base = base;
    392 	aiov.iov_len = len;
    393 	auio.uio_resid = len;
    394 	auio.uio_offset = offset;
    395 	auio.uio_segflg = segflg;
    396 	auio.uio_rw = rw;
    397 	auio.uio_procp = p;
    398 	if (rw == UIO_READ) {
    399 		error = VOP_READ(vp, &auio, ioflg, cred);
    400 	} else {
    401 		error = VOP_WRITE(vp, &auio, ioflg, cred);
    402 	}
    403 	if (aresid)
    404 		*aresid = auio.uio_resid;
    405 	else
    406 		if (auio.uio_resid && error == 0)
    407 			error = EIO;
    408 	if ((ioflg & IO_NODELOCKED) == 0)
    409 		VOP_UNLOCK(vp, 0);
    410 	return (error);
    411 }
    412 
    413 int
    414 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies)
    415 	struct file *fp;
    416 	char *buf;
    417 	int segflg, *done, *ncookies;
    418 	u_int count;
    419 	struct proc *p;
    420 	off_t **cookies;
    421 {
    422 	struct vnode *vp = (struct vnode *)fp->f_data;
    423 	struct iovec aiov;
    424 	struct uio auio;
    425 	int error, eofflag;
    426 
    427 unionread:
    428 	if (vp->v_type != VDIR)
    429 		return (EINVAL);
    430 	aiov.iov_base = buf;
    431 	aiov.iov_len = count;
    432 	auio.uio_iov = &aiov;
    433 	auio.uio_iovcnt = 1;
    434 	auio.uio_rw = UIO_READ;
    435 	auio.uio_segflg = segflg;
    436 	auio.uio_procp = p;
    437 	auio.uio_resid = count;
    438 	vn_lock(vp, LK_SHARED | LK_RETRY);
    439 	auio.uio_offset = fp->f_offset;
    440 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
    441 		    ncookies);
    442 	fp->f_offset = auio.uio_offset;
    443 	VOP_UNLOCK(vp, 0);
    444 	if (error)
    445 		return (error);
    446 
    447 #if defined(UNION) || defined(LKM)
    448 	if (count == auio.uio_resid && vn_union_readdir_hook) {
    449 		struct vnode *ovp = vp;
    450 
    451 		error = (*vn_union_readdir_hook)(&vp, fp, p);
    452 		if (error)
    453 			return (error);
    454 		if (vp != ovp)
    455 			goto unionread;
    456 	}
    457 #endif /* UNION || LKM */
    458 
    459 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
    460 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
    461 		struct vnode *tvp = vp;
    462 		vp = vp->v_mount->mnt_vnodecovered;
    463 		VREF(vp);
    464 		fp->f_data = vp;
    465 		fp->f_offset = 0;
    466 		vrele(tvp);
    467 		goto unionread;
    468 	}
    469 	*done = count - auio.uio_resid;
    470 	return error;
    471 }
    472 
    473 /*
    474  * File table vnode read routine.
    475  */
    476 static int
    477 vn_read(fp, offset, uio, cred, flags)
    478 	struct file *fp;
    479 	off_t *offset;
    480 	struct uio *uio;
    481 	struct ucred *cred;
    482 	int flags;
    483 {
    484 	struct vnode *vp = (struct vnode *)fp->f_data;
    485 	int count, error, ioflag = 0;
    486 
    487 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
    488 	if (fp->f_flag & FNONBLOCK)
    489 		ioflag |= IO_NDELAY;
    490 	if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
    491 		ioflag |= IO_SYNC;
    492 	if (fp->f_flag & FALTIO)
    493 		ioflag |= IO_ALTSEMANTICS;
    494 	vn_lock(vp, LK_SHARED | LK_RETRY);
    495 	uio->uio_offset = *offset;
    496 	count = uio->uio_resid;
    497 	error = VOP_READ(vp, uio, ioflag, cred);
    498 	if (flags & FOF_UPDATE_OFFSET)
    499 		*offset += count - uio->uio_resid;
    500 	VOP_UNLOCK(vp, 0);
    501 	return (error);
    502 }
    503 
    504 /*
    505  * File table vnode write routine.
    506  */
    507 static int
    508 vn_write(fp, offset, uio, cred, flags)
    509 	struct file *fp;
    510 	off_t *offset;
    511 	struct uio *uio;
    512 	struct ucred *cred;
    513 	int flags;
    514 {
    515 	struct vnode *vp = (struct vnode *)fp->f_data;
    516 	int count, error, ioflag = IO_UNIT;
    517 
    518 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
    519 		ioflag |= IO_APPEND;
    520 	if (fp->f_flag & FNONBLOCK)
    521 		ioflag |= IO_NDELAY;
    522 	if (fp->f_flag & FFSYNC ||
    523 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
    524 		ioflag |= IO_SYNC;
    525 	else if (fp->f_flag & FDSYNC)
    526 		ioflag |= IO_DSYNC;
    527 	if (fp->f_flag & FALTIO)
    528 		ioflag |= IO_ALTSEMANTICS;
    529 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
    530 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    531 	uio->uio_offset = *offset;
    532 	count = uio->uio_resid;
    533 	error = VOP_WRITE(vp, uio, ioflag, cred);
    534 	if (flags & FOF_UPDATE_OFFSET) {
    535 		if (ioflag & IO_APPEND)
    536 			*offset = uio->uio_offset;
    537 		else
    538 			*offset += count - uio->uio_resid;
    539 	}
    540 	VOP_UNLOCK(vp, 0);
    541 	return (error);
    542 }
    543 
    544 /*
    545  * File table vnode stat routine.
    546  */
    547 static int
    548 vn_statfile(fp, sb, p)
    549 	struct file *fp;
    550 	struct stat *sb;
    551 	struct proc *p;
    552 {
    553 	struct vnode *vp = (struct vnode *)fp->f_data;
    554 
    555 	return vn_stat(vp, sb, p);
    556 }
    557 
    558 int
    559 vn_stat(vp, sb, p)
    560 	struct vnode *vp;
    561 	struct stat *sb;
    562 	struct proc *p;
    563 {
    564 	struct vattr va;
    565 	int error;
    566 	mode_t mode;
    567 
    568 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
    569 	if (error)
    570 		return (error);
    571 	/*
    572 	 * Copy from vattr table
    573 	 */
    574 	sb->st_dev = va.va_fsid;
    575 	sb->st_ino = va.va_fileid;
    576 	mode = va.va_mode;
    577 	switch (vp->v_type) {
    578 	case VREG:
    579 		mode |= S_IFREG;
    580 		break;
    581 	case VDIR:
    582 		mode |= S_IFDIR;
    583 		break;
    584 	case VBLK:
    585 		mode |= S_IFBLK;
    586 		break;
    587 	case VCHR:
    588 		mode |= S_IFCHR;
    589 		break;
    590 	case VLNK:
    591 		mode |= S_IFLNK;
    592 		break;
    593 	case VSOCK:
    594 		mode |= S_IFSOCK;
    595 		break;
    596 	case VFIFO:
    597 		mode |= S_IFIFO;
    598 		break;
    599 	default:
    600 		return (EBADF);
    601 	};
    602 	sb->st_mode = mode;
    603 	sb->st_nlink = va.va_nlink;
    604 	sb->st_uid = va.va_uid;
    605 	sb->st_gid = va.va_gid;
    606 	sb->st_rdev = va.va_rdev;
    607 	sb->st_size = va.va_size;
    608 	sb->st_atimespec = va.va_atime;
    609 	sb->st_mtimespec = va.va_mtime;
    610 	sb->st_ctimespec = va.va_ctime;
    611 	sb->st_birthtimespec = va.va_birthtime;
    612 	sb->st_blksize = va.va_blocksize;
    613 	sb->st_flags = va.va_flags;
    614 	sb->st_gen = 0;
    615 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
    616 	return (0);
    617 }
    618 
    619 /*
    620  * File table vnode fcntl routine.
    621  */
    622 static int
    623 vn_fcntl(fp, com, data, p)
    624 	struct file *fp;
    625 	u_int com;
    626 	void *data;
    627 	struct proc *p;
    628 {
    629 	struct vnode *vp = ((struct vnode *)fp->f_data);
    630 	int error;
    631 
    632 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    633 	error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    634 	VOP_UNLOCK(vp, 0);
    635 	return (error);
    636 }
    637 
    638 /*
    639  * File table vnode ioctl routine.
    640  */
    641 static int
    642 vn_ioctl(fp, com, data, p)
    643 	struct file *fp;
    644 	u_long com;
    645 	void *data;
    646 	struct proc *p;
    647 {
    648 	struct vnode *vp = ((struct vnode *)fp->f_data);
    649 	struct vattr vattr;
    650 	int error;
    651 
    652 	switch (vp->v_type) {
    653 
    654 	case VREG:
    655 	case VDIR:
    656 		if (com == FIONREAD) {
    657 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
    658 			if (error)
    659 				return (error);
    660 			*(int *)data = vattr.va_size - fp->f_offset;
    661 			return (0);
    662 		}
    663 		if (com == FIOGETBMAP) {
    664 			daddr_t *block;
    665 
    666 			if (*(daddr_t *)data < 0)
    667 				return (EINVAL);
    668 			block = (daddr_t *)data;
    669 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
    670 		}
    671 		if (com == OFIOGETBMAP) {
    672 			daddr_t ibn, obn;
    673 
    674 			if (*(int32_t *)data < 0)
    675 				return (EINVAL);
    676 			ibn = (daddr_t)*(int32_t *)data;
    677 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
    678 			*(int32_t *)data = (int32_t)obn;
    679 			return error;
    680 		}
    681 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
    682 			return (0);			/* XXX */
    683 		/* fall into ... */
    684 
    685 	case VFIFO:
    686 	case VCHR:
    687 	case VBLK:
    688 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    689 		if (error == 0 && com == TIOCSCTTY) {
    690 			if (p->p_session->s_ttyvp)
    691 				vrele(p->p_session->s_ttyvp);
    692 			p->p_session->s_ttyvp = vp;
    693 			VREF(vp);
    694 		}
    695 		return (error);
    696 
    697 	default:
    698 		return (EPASSTHROUGH);
    699 	}
    700 }
    701 
    702 /*
    703  * File table vnode poll routine.
    704  */
    705 static int
    706 vn_poll(fp, events, p)
    707 	struct file *fp;
    708 	int events;
    709 	struct proc *p;
    710 {
    711 
    712 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
    713 }
    714 
    715 /*
    716  * File table vnode kqfilter routine.
    717  */
    718 int
    719 vn_kqfilter(fp, kn)
    720 	struct file *fp;
    721 	struct knote *kn;
    722 {
    723 
    724 	return (VOP_KQFILTER((struct vnode *)fp->f_data, kn));
    725 }
    726 
    727 /*
    728  * Check that the vnode is still valid, and if so
    729  * acquire requested lock.
    730  */
    731 int
    732 vn_lock(vp, flags)
    733 	struct vnode *vp;
    734 	int flags;
    735 {
    736 	int error;
    737 
    738 	do {
    739 		if ((flags & LK_INTERLOCK) == 0)
    740 			simple_lock(&vp->v_interlock);
    741 		if (vp->v_flag & VXLOCK) {
    742 			if (flags & LK_NOWAIT) {
    743 				simple_unlock(&vp->v_interlock);
    744 				return EBUSY;
    745 			}
    746 			vp->v_flag |= VXWANT;
    747 			ltsleep(vp, PINOD | PNORELOCK,
    748 			    "vn_lock", 0, &vp->v_interlock);
    749 			error = ENOENT;
    750 		} else {
    751 			error = VOP_LOCK(vp, flags | LK_INTERLOCK);
    752 			if (error == 0 || error == EDEADLK || error == EBUSY)
    753 				return (error);
    754 		}
    755 		flags &= ~LK_INTERLOCK;
    756 	} while (flags & LK_RETRY);
    757 	return (error);
    758 }
    759 
    760 /*
    761  * File table vnode close routine.
    762  */
    763 static int
    764 vn_closefile(fp, p)
    765 	struct file *fp;
    766 	struct proc *p;
    767 {
    768 
    769 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
    770 		fp->f_cred, p));
    771 }
    772 
    773 /*
    774  * Enable LK_CANRECURSE on lock. Return prior status.
    775  */
    776 u_int
    777 vn_setrecurse(vp)
    778 	struct vnode *vp;
    779 {
    780 	struct lock *lkp = &vp->v_lock;
    781 	u_int retval = lkp->lk_flags & LK_CANRECURSE;
    782 
    783 	lkp->lk_flags |= LK_CANRECURSE;
    784 	return retval;
    785 }
    786 
    787 /*
    788  * Called when done with locksetrecurse.
    789  */
    790 void
    791 vn_restorerecurse(vp, flags)
    792 	struct vnode *vp;
    793 	u_int flags;
    794 {
    795 	struct lock *lkp = &vp->v_lock;
    796 
    797 	lkp->lk_flags &= ~LK_CANRECURSE;
    798 	lkp->lk_flags |= flags;
    799 }
    800