Home | History | Annotate | Line # | Download | only in kern
vfs_vnops.c revision 1.75
      1 /*	$NetBSD: vfs_vnops.c,v 1.75 2003/10/15 11:29:01 hannken 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. Neither the name of the University nor the names of its contributors
     21  *    may be used to endorse or promote products derived from this software
     22  *    without specific prior written permission.
     23  *
     24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     34  * SUCH DAMAGE.
     35  *
     36  *	@(#)vfs_vnops.c	8.14 (Berkeley) 6/15/95
     37  */
     38 
     39 #include <sys/cdefs.h>
     40 __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.75 2003/10/15 11:29:01 hannken Exp $");
     41 
     42 #include "fs_union.h"
     43 
     44 #include <sys/param.h>
     45 #include <sys/systm.h>
     46 #include <sys/kernel.h>
     47 #include <sys/file.h>
     48 #include <sys/stat.h>
     49 #include <sys/buf.h>
     50 #include <sys/proc.h>
     51 #include <sys/mount.h>
     52 #include <sys/namei.h>
     53 #include <sys/vnode.h>
     54 #include <sys/ioctl.h>
     55 #include <sys/tty.h>
     56 #include <sys/poll.h>
     57 
     58 #include <uvm/uvm_extern.h>
     59 
     60 #ifdef UNION
     61 #include <fs/union/union.h>
     62 #endif
     63 
     64 #if defined(LKM) || defined(UNION)
     65 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct proc *);
     66 #endif
     67 
     68 #ifdef VERIFIED_EXEC
     69 #include <sys/verified_exec.h>
     70 
     71 extern LIST_HEAD(veriexec_devhead, veriexec_dev_list) veriexec_dev_head;
     72 extern struct veriexec_devhead veriexec_file_dev_head;
     73 #endif
     74 
     75 static int vn_read(struct file *fp, off_t *offset, struct uio *uio,
     76 	    struct ucred *cred, int flags);
     77 static int vn_write(struct file *fp, off_t *offset, struct uio *uio,
     78 	    struct ucred *cred, int flags);
     79 static int vn_closefile(struct file *fp, struct proc *p);
     80 static int vn_poll(struct file *fp, int events, struct proc *p);
     81 static int vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p);
     82 static int vn_statfile(struct file *fp, struct stat *sb, struct proc *p);
     83 static int vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p);
     84 
     85 struct 	fileops vnops = {
     86 	vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll,
     87 	vn_statfile, vn_closefile, vn_kqfilter
     88 };
     89 
     90 /*
     91  * Common code for vnode open operations.
     92  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
     93  */
     94 int
     95 vn_open(ndp, fmode, cmode)
     96 	struct nameidata *ndp;
     97 	int fmode, cmode;
     98 {
     99 	struct vnode *vp;
    100 	struct mount *mp;
    101 	struct proc *p = ndp->ni_cnd.cn_proc;
    102 	struct ucred *cred = p->p_ucred;
    103 	struct vattr va;
    104 	int error;
    105 #ifdef VERIFIED_EXEC
    106 	char got_dev;
    107 	struct veriexec_inode_list *veriexec_node;
    108 	char fingerprint[MAXFINGERPRINTLEN];
    109 #endif
    110 
    111 restart:
    112 	if (fmode & O_CREAT) {
    113 		ndp->ni_cnd.cn_nameiop = CREATE;
    114 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
    115 		if ((fmode & O_EXCL) == 0 &&
    116 		    ((fmode & O_NOFOLLOW) == 0))
    117 			ndp->ni_cnd.cn_flags |= FOLLOW;
    118 		if ((error = namei(ndp)) != 0)
    119 			return (error);
    120 		if (ndp->ni_vp == NULL) {
    121 			VATTR_NULL(&va);
    122 			va.va_type = VREG;
    123 			va.va_mode = cmode;
    124 			if (fmode & O_EXCL)
    125 				 va.va_vaflags |= VA_EXCLUSIVE;
    126 			if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) {
    127 				VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    128 				vput(ndp->ni_dvp);
    129 				if ((error = vn_start_write(NULL, &mp,
    130 				    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
    131 					return (error);
    132 				goto restart;
    133 			}
    134 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
    135 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
    136 					   &ndp->ni_cnd, &va);
    137 			vn_finished_write(mp, 0);
    138 			if (error)
    139 				return (error);
    140 			fmode &= ~O_TRUNC;
    141 			vp = ndp->ni_vp;
    142 		} else {
    143 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    144 			if (ndp->ni_dvp == ndp->ni_vp)
    145 				vrele(ndp->ni_dvp);
    146 			else
    147 				vput(ndp->ni_dvp);
    148 			ndp->ni_dvp = NULL;
    149 			vp = ndp->ni_vp;
    150 			if (fmode & O_EXCL) {
    151 				error = EEXIST;
    152 				goto bad;
    153 			}
    154 			fmode &= ~O_CREAT;
    155 		}
    156 	} else {
    157 		ndp->ni_cnd.cn_nameiop = LOOKUP;
    158 		ndp->ni_cnd.cn_flags = LOCKLEAF;
    159 		if ((fmode & O_NOFOLLOW) == 0)
    160 			ndp->ni_cnd.cn_flags |= FOLLOW;
    161 		if ((error = namei(ndp)) != 0)
    162 			return (error);
    163 		vp = ndp->ni_vp;
    164 	}
    165 	if (vp->v_type == VSOCK) {
    166 		error = EOPNOTSUPP;
    167 		goto bad;
    168 	}
    169 	if (ndp->ni_vp->v_type == VLNK) {
    170 		error = EFTYPE;
    171 		goto bad;
    172 	}
    173 
    174 #ifdef VERIFIED_EXEC
    175 	veriexec_node = NULL;
    176 
    177 	if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0)
    178 		goto bad;
    179 #endif
    180 
    181 	if ((fmode & O_CREAT) == 0) {
    182 #ifdef VERIFIED_EXEC
    183 		  /*
    184 		   * Look for the file on the fingerprint lists iff
    185 		   * it has not been seen before.
    186 		   */
    187 		if ((vp->fp_status == FINGERPRINT_INVALID) ||
    188 		    (vp->fp_status == FINGERPRINT_NODEV)) {
    189 			  /* check the file list for the finger print */
    190 			veriexec_node = get_veriexec_inode(&veriexec_file_dev_head,
    191 						     va.va_fsid,
    192 						     va.va_fileid,
    193 						     &got_dev);
    194 			if (veriexec_node == NULL) {
    195 				/* failing that, check the exec list */
    196 				veriexec_node = get_veriexec_inode(
    197 					&veriexec_dev_head, va.va_fsid,
    198 					va.va_fileid, &got_dev);
    199 			}
    200 
    201 			if ((veriexec_node == NULL) && (got_dev == 1))
    202 				vp->fp_status = FINGERPRINT_NOENTRY;
    203 
    204 			if (veriexec_node != NULL) {
    205 				if ((error = evaluate_fingerprint(vp,
    206 						veriexec_node, p, va.va_size,
    207 						fingerprint)) != 0)
    208 					goto bad;
    209 
    210 				if (fingerprintcmp(veriexec_node,
    211 						   fingerprint) == 0) {
    212 					  /* fingerprint ok */
    213 					vp->fp_status =	FINGERPRINT_VALID;
    214 #ifdef VERIFIED_EXEC_DEBUG
    215 					printf(
    216 			"file fingerprint matches for dev %lu, file %lu\n",
    217 						va.va_fsid, va.va_fileid);
    218 #endif
    219 				} else {
    220 					vp->fp_status =	FINGERPRINT_NOMATCH;
    221 				}
    222 			}
    223 		}
    224 #endif
    225 
    226 		if (fmode & FREAD) {
    227 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
    228 				goto bad;
    229 
    230 #ifdef VERIFIED_EXEC
    231 				/* file is on finger print list */
    232 			if (vp->fp_status == FINGERPRINT_NOMATCH) {
    233 				  /* fingerprint bad */
    234 				printf(
    235 		"file fingerprint does not match on dev %lu, file %lu\n",
    236 					va.va_fsid, va.va_fileid);
    237 				if (securelevel > 2) {
    238 					error = EPERM;
    239 					goto bad;
    240 				}
    241 			}
    242 #endif
    243 		}
    244 		if (fmode & (FWRITE | O_TRUNC)) {
    245 			if (vp->v_type == VDIR) {
    246 				error = EISDIR;
    247 				goto bad;
    248 			}
    249 			if ((error = vn_writechk(vp)) != 0 ||
    250 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
    251 				goto bad;
    252 #ifdef VERIFIED_EXEC
    253 			  /*
    254 			   * If file has a fingerprint then
    255 			   * deny the write request, otherwise
    256 			   * invalidate the status so we don't
    257 			   * keep checking for the file having
    258 			   * a fingerprint.
    259 			   */
    260 			if (vp->fp_status == FINGERPRINT_VALID) {
    261 				printf(
    262 		      "writing to fingerprinted file for dev %lu, file %lu\n",
    263 		      va.va_fsid, va.va_fileid);
    264 				if (securelevel > 2) {
    265 					error = EPERM;
    266 					goto bad;
    267 				} else {
    268 					vp->fp_status =	FINGERPRINT_INVALID;
    269 				}
    270 			}
    271 #endif
    272 		}
    273 	}
    274 	if (fmode & O_TRUNC) {
    275 		VOP_UNLOCK(vp, 0);			/* XXX */
    276 		if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
    277 			vput(vp);
    278 			return (error);
    279 		}
    280 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
    281 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
    282 		VATTR_NULL(&va);
    283 		va.va_size = 0;
    284 		error = VOP_SETATTR(vp, &va, cred, p);
    285 		vn_finished_write(mp, 0);
    286 		if (error != 0)
    287 			goto bad;
    288 	}
    289 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
    290 		goto bad;
    291 	if (vp->v_type == VREG &&
    292 	    uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
    293 		error = EIO;
    294 		goto bad;
    295 	}
    296 	if (fmode & FWRITE)
    297 		vp->v_writecount++;
    298 
    299 	return (0);
    300 bad:
    301 	vput(vp);
    302 	return (error);
    303 }
    304 
    305 /*
    306  * Check for write permissions on the specified vnode.
    307  * Prototype text segments cannot be written.
    308  */
    309 int
    310 vn_writechk(vp)
    311 	struct vnode *vp;
    312 {
    313 
    314 	/*
    315 	 * If the vnode is in use as a process's text,
    316 	 * we can't allow writing.
    317 	 */
    318 	if (vp->v_flag & VTEXT)
    319 		return (ETXTBSY);
    320 	return (0);
    321 }
    322 
    323 /*
    324  * Mark a vnode as having executable mappings.
    325  */
    326 void
    327 vn_markexec(vp)
    328 	struct vnode *vp;
    329 {
    330 	if ((vp->v_flag & VEXECMAP) == 0) {
    331 		uvmexp.filepages -= vp->v_uobj.uo_npages;
    332 		uvmexp.execpages += vp->v_uobj.uo_npages;
    333 	}
    334 	vp->v_flag |= VEXECMAP;
    335 }
    336 
    337 /*
    338  * Mark a vnode as being the text of a process.
    339  * Fail if the vnode is currently writable.
    340  */
    341 int
    342 vn_marktext(vp)
    343 	struct vnode *vp;
    344 {
    345 
    346 	if (vp->v_writecount != 0) {
    347 		KASSERT((vp->v_flag & VTEXT) == 0);
    348 		return (ETXTBSY);
    349 	}
    350 	vp->v_flag |= VTEXT;
    351 	vn_markexec(vp);
    352 	return (0);
    353 }
    354 
    355 /*
    356  * Vnode close call
    357  *
    358  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
    359  */
    360 int
    361 vn_close(vp, flags, cred, p)
    362 	struct vnode *vp;
    363 	int flags;
    364 	struct ucred *cred;
    365 	struct proc *p;
    366 {
    367 	int error;
    368 
    369 	if (flags & FWRITE)
    370 		vp->v_writecount--;
    371 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    372 	error = VOP_CLOSE(vp, flags, cred, p);
    373 	vput(vp);
    374 	return (error);
    375 }
    376 
    377 /*
    378  * Package up an I/O request on a vnode into a uio and do it.
    379  */
    380 int
    381 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p)
    382 	enum uio_rw rw;
    383 	struct vnode *vp;
    384 	caddr_t base;
    385 	int len;
    386 	off_t offset;
    387 	enum uio_seg segflg;
    388 	int ioflg;
    389 	struct ucred *cred;
    390 	size_t *aresid;
    391 	struct proc *p;
    392 {
    393 	struct uio auio;
    394 	struct iovec aiov;
    395 	struct mount *mp;
    396 	int error;
    397 
    398 	if ((ioflg & IO_NODELOCKED) == 0) {
    399 		if (rw == UIO_READ) {
    400 			vn_lock(vp, LK_SHARED | LK_RETRY);
    401 		} else /* UIO_WRITE */ {
    402 			if (vp->v_type != VCHR &&
    403 			    (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH))
    404 			    != 0)
    405 				return (error);
    406 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    407 		}
    408 	}
    409 	auio.uio_iov = &aiov;
    410 	auio.uio_iovcnt = 1;
    411 	aiov.iov_base = base;
    412 	aiov.iov_len = len;
    413 	auio.uio_resid = len;
    414 	auio.uio_offset = offset;
    415 	auio.uio_segflg = segflg;
    416 	auio.uio_rw = rw;
    417 	auio.uio_procp = p;
    418 	if (rw == UIO_READ) {
    419 		error = VOP_READ(vp, &auio, ioflg, cred);
    420 	} else {
    421 		error = VOP_WRITE(vp, &auio, ioflg, cred);
    422 	}
    423 	if (aresid)
    424 		*aresid = auio.uio_resid;
    425 	else
    426 		if (auio.uio_resid && error == 0)
    427 			error = EIO;
    428 	if ((ioflg & IO_NODELOCKED) == 0) {
    429 		if (rw == UIO_WRITE)
    430 			vn_finished_write(mp, 0);
    431 		VOP_UNLOCK(vp, 0);
    432 	}
    433 	return (error);
    434 }
    435 
    436 int
    437 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies)
    438 	struct file *fp;
    439 	char *buf;
    440 	int segflg, *done, *ncookies;
    441 	u_int count;
    442 	struct proc *p;
    443 	off_t **cookies;
    444 {
    445 	struct vnode *vp = (struct vnode *)fp->f_data;
    446 	struct iovec aiov;
    447 	struct uio auio;
    448 	int error, eofflag;
    449 
    450 unionread:
    451 	if (vp->v_type != VDIR)
    452 		return (EINVAL);
    453 	aiov.iov_base = buf;
    454 	aiov.iov_len = count;
    455 	auio.uio_iov = &aiov;
    456 	auio.uio_iovcnt = 1;
    457 	auio.uio_rw = UIO_READ;
    458 	auio.uio_segflg = segflg;
    459 	auio.uio_procp = p;
    460 	auio.uio_resid = count;
    461 	vn_lock(vp, LK_SHARED | LK_RETRY);
    462 	auio.uio_offset = fp->f_offset;
    463 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
    464 		    ncookies);
    465 	fp->f_offset = auio.uio_offset;
    466 	VOP_UNLOCK(vp, 0);
    467 	if (error)
    468 		return (error);
    469 
    470 #if defined(UNION) || defined(LKM)
    471 	if (count == auio.uio_resid && vn_union_readdir_hook) {
    472 		struct vnode *ovp = vp;
    473 
    474 		error = (*vn_union_readdir_hook)(&vp, fp, p);
    475 		if (error)
    476 			return (error);
    477 		if (vp != ovp)
    478 			goto unionread;
    479 	}
    480 #endif /* UNION || LKM */
    481 
    482 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
    483 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
    484 		struct vnode *tvp = vp;
    485 		vp = vp->v_mount->mnt_vnodecovered;
    486 		VREF(vp);
    487 		fp->f_data = vp;
    488 		fp->f_offset = 0;
    489 		vrele(tvp);
    490 		goto unionread;
    491 	}
    492 	*done = count - auio.uio_resid;
    493 	return error;
    494 }
    495 
    496 /*
    497  * File table vnode read routine.
    498  */
    499 static int
    500 vn_read(fp, offset, uio, cred, flags)
    501 	struct file *fp;
    502 	off_t *offset;
    503 	struct uio *uio;
    504 	struct ucred *cred;
    505 	int flags;
    506 {
    507 	struct vnode *vp = (struct vnode *)fp->f_data;
    508 	int count, error, ioflag = 0;
    509 
    510 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
    511 	if (fp->f_flag & FNONBLOCK)
    512 		ioflag |= IO_NDELAY;
    513 	if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
    514 		ioflag |= IO_SYNC;
    515 	if (fp->f_flag & FALTIO)
    516 		ioflag |= IO_ALTSEMANTICS;
    517 	vn_lock(vp, LK_SHARED | LK_RETRY);
    518 	uio->uio_offset = *offset;
    519 	count = uio->uio_resid;
    520 	error = VOP_READ(vp, uio, ioflag, cred);
    521 	if (flags & FOF_UPDATE_OFFSET)
    522 		*offset += count - uio->uio_resid;
    523 	VOP_UNLOCK(vp, 0);
    524 	return (error);
    525 }
    526 
    527 /*
    528  * File table vnode write routine.
    529  */
    530 static int
    531 vn_write(fp, offset, uio, cred, flags)
    532 	struct file *fp;
    533 	off_t *offset;
    534 	struct uio *uio;
    535 	struct ucred *cred;
    536 	int flags;
    537 {
    538 	struct vnode *vp = (struct vnode *)fp->f_data;
    539 	struct mount *mp;
    540 	int count, error, ioflag = IO_UNIT;
    541 
    542 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
    543 		ioflag |= IO_APPEND;
    544 	if (fp->f_flag & FNONBLOCK)
    545 		ioflag |= IO_NDELAY;
    546 	if (fp->f_flag & FFSYNC ||
    547 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
    548 		ioflag |= IO_SYNC;
    549 	else if (fp->f_flag & FDSYNC)
    550 		ioflag |= IO_DSYNC;
    551 	if (fp->f_flag & FALTIO)
    552 		ioflag |= IO_ALTSEMANTICS;
    553 	mp = NULL;
    554 	if (vp->v_type != VCHR &&
    555 	    (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
    556 		return (error);
    557 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
    558 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    559 	uio->uio_offset = *offset;
    560 	count = uio->uio_resid;
    561 	error = VOP_WRITE(vp, uio, ioflag, cred);
    562 	if (flags & FOF_UPDATE_OFFSET) {
    563 		if (ioflag & IO_APPEND)
    564 			*offset = uio->uio_offset;
    565 		else
    566 			*offset += count - uio->uio_resid;
    567 	}
    568 	VOP_UNLOCK(vp, 0);
    569 	vn_finished_write(mp, 0);
    570 	return (error);
    571 }
    572 
    573 /*
    574  * File table vnode stat routine.
    575  */
    576 static int
    577 vn_statfile(fp, sb, p)
    578 	struct file *fp;
    579 	struct stat *sb;
    580 	struct proc *p;
    581 {
    582 	struct vnode *vp = (struct vnode *)fp->f_data;
    583 
    584 	return vn_stat(vp, sb, p);
    585 }
    586 
    587 int
    588 vn_stat(vp, sb, p)
    589 	struct vnode *vp;
    590 	struct stat *sb;
    591 	struct proc *p;
    592 {
    593 	struct vattr va;
    594 	int error;
    595 	mode_t mode;
    596 
    597 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
    598 	if (error)
    599 		return (error);
    600 	/*
    601 	 * Copy from vattr table
    602 	 */
    603 	sb->st_dev = va.va_fsid;
    604 	sb->st_ino = va.va_fileid;
    605 	mode = va.va_mode;
    606 	switch (vp->v_type) {
    607 	case VREG:
    608 		mode |= S_IFREG;
    609 		break;
    610 	case VDIR:
    611 		mode |= S_IFDIR;
    612 		break;
    613 	case VBLK:
    614 		mode |= S_IFBLK;
    615 		break;
    616 	case VCHR:
    617 		mode |= S_IFCHR;
    618 		break;
    619 	case VLNK:
    620 		mode |= S_IFLNK;
    621 		break;
    622 	case VSOCK:
    623 		mode |= S_IFSOCK;
    624 		break;
    625 	case VFIFO:
    626 		mode |= S_IFIFO;
    627 		break;
    628 	default:
    629 		return (EBADF);
    630 	};
    631 	sb->st_mode = mode;
    632 	sb->st_nlink = va.va_nlink;
    633 	sb->st_uid = va.va_uid;
    634 	sb->st_gid = va.va_gid;
    635 	sb->st_rdev = va.va_rdev;
    636 	sb->st_size = va.va_size;
    637 	sb->st_atimespec = va.va_atime;
    638 	sb->st_mtimespec = va.va_mtime;
    639 	sb->st_ctimespec = va.va_ctime;
    640 	sb->st_birthtimespec = va.va_birthtime;
    641 	sb->st_blksize = va.va_blocksize;
    642 	sb->st_flags = va.va_flags;
    643 	sb->st_gen = 0;
    644 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
    645 	return (0);
    646 }
    647 
    648 /*
    649  * File table vnode fcntl routine.
    650  */
    651 static int
    652 vn_fcntl(fp, com, data, p)
    653 	struct file *fp;
    654 	u_int com;
    655 	void *data;
    656 	struct proc *p;
    657 {
    658 	struct vnode *vp = ((struct vnode *)fp->f_data);
    659 	int error;
    660 
    661 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    662 	error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    663 	VOP_UNLOCK(vp, 0);
    664 	return (error);
    665 }
    666 
    667 /*
    668  * File table vnode ioctl routine.
    669  */
    670 static int
    671 vn_ioctl(fp, com, data, p)
    672 	struct file *fp;
    673 	u_long com;
    674 	void *data;
    675 	struct proc *p;
    676 {
    677 	struct vnode *vp = ((struct vnode *)fp->f_data);
    678 	struct vattr vattr;
    679 	int error;
    680 
    681 	switch (vp->v_type) {
    682 
    683 	case VREG:
    684 	case VDIR:
    685 		if (com == FIONREAD) {
    686 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
    687 			if (error)
    688 				return (error);
    689 			*(int *)data = vattr.va_size - fp->f_offset;
    690 			return (0);
    691 		}
    692 		if (com == FIOGETBMAP) {
    693 			daddr_t *block;
    694 
    695 			if (*(daddr_t *)data < 0)
    696 				return (EINVAL);
    697 			block = (daddr_t *)data;
    698 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
    699 		}
    700 		if (com == OFIOGETBMAP) {
    701 			daddr_t ibn, obn;
    702 
    703 			if (*(int32_t *)data < 0)
    704 				return (EINVAL);
    705 			ibn = (daddr_t)*(int32_t *)data;
    706 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
    707 			*(int32_t *)data = (int32_t)obn;
    708 			return error;
    709 		}
    710 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
    711 			return (0);			/* XXX */
    712 		/* fall into ... */
    713 	case VFIFO:
    714 	case VCHR:
    715 	case VBLK:
    716 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    717 		if (error == 0 && com == TIOCSCTTY) {
    718 			if (p->p_session->s_ttyvp)
    719 				vrele(p->p_session->s_ttyvp);
    720 			p->p_session->s_ttyvp = vp;
    721 			VREF(vp);
    722 		}
    723 		return (error);
    724 
    725 	default:
    726 		return (EPASSTHROUGH);
    727 	}
    728 }
    729 
    730 /*
    731  * File table vnode poll routine.
    732  */
    733 static int
    734 vn_poll(fp, events, p)
    735 	struct file *fp;
    736 	int events;
    737 	struct proc *p;
    738 {
    739 
    740 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
    741 }
    742 
    743 /*
    744  * File table vnode kqfilter routine.
    745  */
    746 int
    747 vn_kqfilter(fp, kn)
    748 	struct file *fp;
    749 	struct knote *kn;
    750 {
    751 
    752 	return (VOP_KQFILTER((struct vnode *)fp->f_data, kn));
    753 }
    754 
    755 /*
    756  * Check that the vnode is still valid, and if so
    757  * acquire requested lock.
    758  */
    759 int
    760 vn_lock(vp, flags)
    761 	struct vnode *vp;
    762 	int flags;
    763 {
    764 	int error;
    765 
    766 	do {
    767 		if ((flags & LK_INTERLOCK) == 0)
    768 			simple_lock(&vp->v_interlock);
    769 		if (vp->v_flag & VXLOCK) {
    770 			if (flags & LK_NOWAIT) {
    771 				simple_unlock(&vp->v_interlock);
    772 				return EBUSY;
    773 			}
    774 			vp->v_flag |= VXWANT;
    775 			ltsleep(vp, PINOD | PNORELOCK,
    776 			    "vn_lock", 0, &vp->v_interlock);
    777 			error = ENOENT;
    778 		} else {
    779 			error = VOP_LOCK(vp, flags | LK_INTERLOCK);
    780 			if (error == 0 || error == EDEADLK || error == EBUSY)
    781 				return (error);
    782 		}
    783 		flags &= ~LK_INTERLOCK;
    784 	} while (flags & LK_RETRY);
    785 	return (error);
    786 }
    787 
    788 /*
    789  * File table vnode close routine.
    790  */
    791 static int
    792 vn_closefile(fp, p)
    793 	struct file *fp;
    794 	struct proc *p;
    795 {
    796 
    797 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
    798 		fp->f_cred, p));
    799 }
    800 
    801 /*
    802  * Enable LK_CANRECURSE on lock. Return prior status.
    803  */
    804 u_int
    805 vn_setrecurse(vp)
    806 	struct vnode *vp;
    807 {
    808 	struct lock *lkp = &vp->v_lock;
    809 	u_int retval = lkp->lk_flags & LK_CANRECURSE;
    810 
    811 	lkp->lk_flags |= LK_CANRECURSE;
    812 	return retval;
    813 }
    814 
    815 /*
    816  * Called when done with locksetrecurse.
    817  */
    818 void
    819 vn_restorerecurse(vp, flags)
    820 	struct vnode *vp;
    821 	u_int flags;
    822 {
    823 	struct lock *lkp = &vp->v_lock;
    824 
    825 	lkp->lk_flags &= ~LK_CANRECURSE;
    826 	lkp->lk_flags |= flags;
    827 }
    828 
    829 /*
    830  * Request a filesystem to suspend write operations.
    831  */
    832 int
    833 vfs_write_suspend(struct mount *mp)
    834 {
    835 	struct proc *p = curproc;	/* XXX */
    836 	int error;
    837 
    838 	if (mp->mnt_iflag & IMNT_SUSPEND)
    839 		return (0);
    840 	mp->mnt_iflag |= IMNT_SUSPEND;
    841 	if (mp->mnt_writeopcountupper > 0)
    842 		tsleep(&mp->mnt_writeopcountupper, PUSER - 1, "suspwt", 0);
    843 	if ((error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) != 0) {
    844 		vfs_write_resume(mp);
    845 		return (error);
    846 	}
    847 	mp->mnt_iflag |= IMNT_SUSPENDLOW;
    848 	if (mp->mnt_writeopcountlower > 0)
    849 		tsleep(&mp->mnt_writeopcountlower, PUSER - 1, "suspwt", 0);
    850 	mp->mnt_iflag |= IMNT_SUSPENDED;
    851 	return (0);
    852 }
    853 
    854 /*
    855  * Request a filesystem to resume write operations.
    856  */
    857 void
    858 vfs_write_resume(struct mount *mp)
    859 {
    860 
    861 	if ((mp->mnt_iflag & IMNT_SUSPEND) == 0)
    862 		return;
    863 	mp->mnt_iflag &= ~(IMNT_SUSPEND | IMNT_SUSPENDLOW | IMNT_SUSPENDED);
    864 	wakeup(&mp->mnt_flag);
    865 }
    866