Home | History | Annotate | Line # | Download | only in kern
vfs_vnops.c revision 1.63
      1 /*	$NetBSD: vfs_vnops.c,v 1.63 2003/02/24 08:34:30 perseant 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 1.63 2003/02/24 08:34:30 perseant 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 <miscfs/union/union.h>
     66 #endif
     67 #include <sys/verified_exec.h>
     68 
     69 extern LIST_HEAD(veriexec_devhead, veriexec_dev_list) veriexec_dev_head;
     70 extern struct veriexec_devhead veriexec_file_dev_head;
     71 
     72 static int vn_read(struct file *fp, off_t *offset, struct uio *uio,
     73 	    struct ucred *cred, int flags);
     74 static int vn_write(struct file *fp, off_t *offset, struct uio *uio,
     75 	    struct ucred *cred, int flags);
     76 static int vn_closefile(struct file *fp, struct proc *p);
     77 static int vn_poll(struct file *fp, int events, struct proc *p);
     78 static int vn_fcntl(struct file *fp, u_int com, caddr_t data, struct proc *p);
     79 static int vn_statfile(struct file *fp, struct stat *sb, struct proc *p);
     80 static int vn_ioctl(struct file *fp, u_long com, caddr_t data, struct proc *p);
     81 
     82 struct 	fileops vnops = {
     83 	vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll,
     84 	vn_statfile, vn_closefile, vn_kqfilter
     85 };
     86 
     87 /*
     88  * Common code for vnode open operations.
     89  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
     90  */
     91 int
     92 vn_open(ndp, fmode, cmode)
     93 	struct nameidata *ndp;
     94 	int fmode, cmode;
     95 {
     96 	struct vnode *vp;
     97 	struct proc *p = ndp->ni_cnd.cn_proc;
     98 	struct ucred *cred = p->p_ucred;
     99 	struct vattr va;
    100 	int error;
    101 #ifdef VERIFIED_EXEC
    102 	char got_dev;
    103 	struct veriexec_inode_list *veriexec_node;
    104 	char fingerprint[MAXFINGERPRINTLEN];
    105 #endif
    106 
    107 	if (fmode & O_CREAT) {
    108 		ndp->ni_cnd.cn_nameiop = CREATE;
    109 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
    110 		if ((fmode & O_EXCL) == 0 &&
    111 		    ((fmode & O_NOFOLLOW) == 0))
    112 			ndp->ni_cnd.cn_flags |= FOLLOW;
    113 		if ((error = namei(ndp)) != 0)
    114 			return (error);
    115 		if (ndp->ni_vp == NULL) {
    116 			VATTR_NULL(&va);
    117 			va.va_type = VREG;
    118 			va.va_mode = cmode;
    119 			if (fmode & O_EXCL)
    120 				 va.va_vaflags |= VA_EXCLUSIVE;
    121 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
    122 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
    123 					   &ndp->ni_cnd, &va);
    124 			if (error)
    125 				return (error);
    126 			fmode &= ~O_TRUNC;
    127 			vp = ndp->ni_vp;
    128 		} else {
    129 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    130 			if (ndp->ni_dvp == ndp->ni_vp)
    131 				vrele(ndp->ni_dvp);
    132 			else
    133 				vput(ndp->ni_dvp);
    134 			ndp->ni_dvp = NULL;
    135 			vp = ndp->ni_vp;
    136 			if (fmode & O_EXCL) {
    137 				error = EEXIST;
    138 				goto bad;
    139 			}
    140 			if (ndp->ni_vp->v_type == VLNK) {
    141 				error = EFTYPE;
    142 				goto bad;
    143 			}
    144 			fmode &= ~O_CREAT;
    145 		}
    146 	} else {
    147 		ndp->ni_cnd.cn_nameiop = LOOKUP;
    148 		ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF;
    149 		if ((error = namei(ndp)) != 0)
    150 			return (error);
    151 		vp = ndp->ni_vp;
    152 	}
    153 	if (vp->v_type == VSOCK) {
    154 		error = EOPNOTSUPP;
    155 		goto bad;
    156 	}
    157 
    158 #ifdef VERIFIED_EXEC
    159 	veriexec_node = NULL;
    160 
    161 	if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0)
    162 		goto bad;
    163 #endif
    164 
    165 	if ((fmode & O_CREAT) == 0) {
    166 #ifdef VERIFIED_EXEC
    167 		  /*
    168 		   * Look for the file on the fingerprint lists iff
    169 		   * it has not been seen before.
    170 		   */
    171 		if ((vp->fp_status == FINGERPRINT_INVALID) ||
    172 		    (vp->fp_status == FINGERPRINT_NODEV)) {
    173 			  /* check the file list for the finger print */
    174 			veriexec_node = get_veriexec_inode(&veriexec_file_dev_head,
    175 						     va.va_fsid,
    176 						     va.va_fileid,
    177 						     &got_dev);
    178 			if (veriexec_node == NULL) {
    179 				/* failing that, check the exec list */
    180 				veriexec_node = get_veriexec_inode(
    181 					&veriexec_dev_head, va.va_fsid,
    182 					va.va_fileid, &got_dev);
    183 			}
    184 
    185 			if ((veriexec_node == NULL) && (got_dev == 1))
    186 				vp->fp_status = FINGERPRINT_NOENTRY;
    187 
    188 			if (veriexec_node != NULL) {
    189 				if ((error = evaluate_fingerprint(vp,
    190 						veriexec_node, p, va.va_size,
    191 						fingerprint)) != 0)
    192 					goto bad;
    193 
    194 				if (fingerprintcmp(veriexec_node,
    195 						   fingerprint) == 0) {
    196 					  /* fingerprint ok */
    197 					vp->fp_status =	FINGERPRINT_VALID;
    198 #ifdef VERIFIED_EXEC_DEBUG
    199 					printf(
    200 			"file fingerprint matches for dev %lu, file %lu\n",
    201 						va.va_fsid, va.va_fileid);
    202 #endif
    203 				} else {
    204 					vp->fp_status =	FINGERPRINT_NOMATCH;
    205 				}
    206 			}
    207 		}
    208 #endif
    209 
    210 		if (fmode & FREAD) {
    211 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
    212 				goto bad;
    213 
    214 #ifdef VERIFIED_EXEC
    215 				/* file is on finger print list */
    216 			if (vp->fp_status == FINGERPRINT_NOMATCH) {
    217 				  /* fingerprint bad */
    218 				printf(
    219 		"file fingerprint does not match on dev %lu, file %lu\n",
    220 					va.va_fsid, va.va_fileid);
    221 				if (securelevel > 2) {
    222 					error = EPERM;
    223 					goto bad;
    224 				}
    225 			}
    226 #endif
    227 		}
    228 		if (fmode & (FWRITE | O_TRUNC)) {
    229 			if (vp->v_type == VDIR) {
    230 				error = EISDIR;
    231 				goto bad;
    232 			}
    233 			if ((error = vn_writechk(vp)) != 0 ||
    234 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
    235 				goto bad;
    236 #ifdef VERIFIED_EXEC
    237 			  /*
    238 			   * If file has a fingerprint then
    239 			   * deny the write request, otherwise
    240 			   * invalidate the status so we don't
    241 			   * keep checking for the file having
    242 			   * a fingerprint.
    243 			   */
    244 			if (vp->fp_status == FINGERPRINT_VALID) {
    245 				printf(
    246 		      "writing to fingerprinted file for dev %lu, file %lu\n",
    247 		      va.va_fsid, va.va_fileid);
    248 				if (securelevel > 2) {
    249 					error = EPERM;
    250 					goto bad;
    251 				} else {
    252 					vp->fp_status =	FINGERPRINT_INVALID;
    253 				}
    254 			}
    255 #endif
    256 		}
    257 	}
    258 	if (fmode & O_TRUNC) {
    259 		VOP_UNLOCK(vp, 0);			/* XXX */
    260 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
    261 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
    262 		VATTR_NULL(&va);
    263 		va.va_size = 0;
    264 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
    265 			goto bad;
    266 	}
    267 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
    268 		goto bad;
    269 	if (vp->v_type == VREG &&
    270 	    uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
    271 		error = EIO;
    272 		goto bad;
    273 	}
    274 	if (fmode & FWRITE)
    275 		vp->v_writecount++;
    276 
    277 	return (0);
    278 bad:
    279 	vput(vp);
    280 	return (error);
    281 }
    282 
    283 /*
    284  * Check for write permissions on the specified vnode.
    285  * Prototype text segments cannot be written.
    286  */
    287 int
    288 vn_writechk(vp)
    289 	struct vnode *vp;
    290 {
    291 
    292 	/*
    293 	 * If the vnode is in use as a process's text,
    294 	 * we can't allow writing.
    295 	 */
    296 	if (vp->v_flag & VTEXT)
    297 		return (ETXTBSY);
    298 	return (0);
    299 }
    300 
    301 /*
    302  * Mark a vnode as having executable mappings.
    303  */
    304 void
    305 vn_markexec(vp)
    306 	struct vnode *vp;
    307 {
    308 	if ((vp->v_flag & VEXECMAP) == 0) {
    309 		uvmexp.filepages -= vp->v_uobj.uo_npages;
    310 		uvmexp.execpages += vp->v_uobj.uo_npages;
    311 	}
    312 	vp->v_flag |= VEXECMAP;
    313 }
    314 
    315 /*
    316  * Mark a vnode as being the text of a process.
    317  * Fail if the vnode is currently writable.
    318  */
    319 int
    320 vn_marktext(vp)
    321 	struct vnode *vp;
    322 {
    323 
    324 	if (vp->v_writecount != 0) {
    325 		KASSERT((vp->v_flag & VTEXT) == 0);
    326 		return (ETXTBSY);
    327 	}
    328 	vp->v_flag |= VTEXT;
    329 	vn_markexec(vp);
    330 	return (0);
    331 }
    332 
    333 /*
    334  * Vnode close call
    335  *
    336  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
    337  */
    338 int
    339 vn_close(vp, flags, cred, p)
    340 	struct vnode *vp;
    341 	int flags;
    342 	struct ucred *cred;
    343 	struct proc *p;
    344 {
    345 	int error;
    346 
    347 	if (flags & FWRITE)
    348 		vp->v_writecount--;
    349 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    350 	error = VOP_CLOSE(vp, flags, cred, p);
    351 	vput(vp);
    352 	return (error);
    353 }
    354 
    355 /*
    356  * Package up an I/O request on a vnode into a uio and do it.
    357  */
    358 int
    359 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p)
    360 	enum uio_rw rw;
    361 	struct vnode *vp;
    362 	caddr_t base;
    363 	int len;
    364 	off_t offset;
    365 	enum uio_seg segflg;
    366 	int ioflg;
    367 	struct ucred *cred;
    368 	size_t *aresid;
    369 	struct proc *p;
    370 {
    371 	struct uio auio;
    372 	struct iovec aiov;
    373 	int error;
    374 
    375 	if ((ioflg & IO_NODELOCKED) == 0) {
    376 		if (rw == UIO_READ) {
    377 			vn_lock(vp, LK_SHARED | LK_RETRY);
    378 		} else {
    379 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    380 		}
    381 	}
    382 	auio.uio_iov = &aiov;
    383 	auio.uio_iovcnt = 1;
    384 	aiov.iov_base = base;
    385 	aiov.iov_len = len;
    386 	auio.uio_resid = len;
    387 	auio.uio_offset = offset;
    388 	auio.uio_segflg = segflg;
    389 	auio.uio_rw = rw;
    390 	auio.uio_procp = p;
    391 	if (rw == UIO_READ) {
    392 		error = VOP_READ(vp, &auio, ioflg, cred);
    393 	} else {
    394 		error = VOP_WRITE(vp, &auio, ioflg, cred);
    395 	}
    396 	if (aresid)
    397 		*aresid = auio.uio_resid;
    398 	else
    399 		if (auio.uio_resid && error == 0)
    400 			error = EIO;
    401 	if ((ioflg & IO_NODELOCKED) == 0)
    402 		VOP_UNLOCK(vp, 0);
    403 	return (error);
    404 }
    405 
    406 int
    407 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies)
    408 	struct file *fp;
    409 	char *buf;
    410 	int segflg, *done, *ncookies;
    411 	u_int count;
    412 	struct proc *p;
    413 	off_t **cookies;
    414 {
    415 	struct vnode *vp = (struct vnode *)fp->f_data;
    416 	struct iovec aiov;
    417 	struct uio auio;
    418 	int error, eofflag;
    419 
    420 unionread:
    421 	if (vp->v_type != VDIR)
    422 		return (EINVAL);
    423 	aiov.iov_base = buf;
    424 	aiov.iov_len = count;
    425 	auio.uio_iov = &aiov;
    426 	auio.uio_iovcnt = 1;
    427 	auio.uio_rw = UIO_READ;
    428 	auio.uio_segflg = segflg;
    429 	auio.uio_procp = p;
    430 	auio.uio_resid = count;
    431 	vn_lock(vp, LK_SHARED | LK_RETRY);
    432 	auio.uio_offset = fp->f_offset;
    433 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
    434 		    ncookies);
    435 	fp->f_offset = auio.uio_offset;
    436 	VOP_UNLOCK(vp, 0);
    437 	if (error)
    438 		return (error);
    439 
    440 #ifdef UNION
    441 {
    442 	extern struct vnode *union_dircache __P((struct vnode *));
    443 
    444 	if (count == auio.uio_resid && (vp->v_op == union_vnodeop_p)) {
    445 		struct vnode *lvp;
    446 
    447 		lvp = union_dircache(vp);
    448 		if (lvp != NULLVP) {
    449 			struct vattr va;
    450 
    451 			/*
    452 			 * If the directory is opaque,
    453 			 * then don't show lower entries
    454 			 */
    455 			error = VOP_GETATTR(vp, &va, fp->f_cred, p);
    456 			if (va.va_flags & OPAQUE) {
    457 				vput(lvp);
    458 				lvp = NULL;
    459 			}
    460 		}
    461 
    462 		if (lvp != NULLVP) {
    463 			error = VOP_OPEN(lvp, FREAD, fp->f_cred, p);
    464 			if (error) {
    465 				vput(lvp);
    466 				return (error);
    467 			}
    468 			VOP_UNLOCK(lvp, 0);
    469 			fp->f_data = (caddr_t) lvp;
    470 			fp->f_offset = 0;
    471 			error = vn_close(vp, FREAD, fp->f_cred, p);
    472 			if (error)
    473 				return (error);
    474 			vp = lvp;
    475 			goto unionread;
    476 		}
    477 	}
    478 }
    479 #endif /* UNION */
    480 
    481 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
    482 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
    483 		struct vnode *tvp = vp;
    484 		vp = vp->v_mount->mnt_vnodecovered;
    485 		VREF(vp);
    486 		fp->f_data = (caddr_t) vp;
    487 		fp->f_offset = 0;
    488 		vrele(tvp);
    489 		goto unionread;
    490 	}
    491 	*done = count - auio.uio_resid;
    492 	return error;
    493 }
    494 
    495 /*
    496  * File table vnode read routine.
    497  */
    498 static int
    499 vn_read(fp, offset, uio, cred, flags)
    500 	struct file *fp;
    501 	off_t *offset;
    502 	struct uio *uio;
    503 	struct ucred *cred;
    504 	int flags;
    505 {
    506 	struct vnode *vp = (struct vnode *)fp->f_data;
    507 	int count, error, ioflag = 0;
    508 
    509 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
    510 	if (fp->f_flag & FNONBLOCK)
    511 		ioflag |= IO_NDELAY;
    512 	if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
    513 		ioflag |= IO_SYNC;
    514 	if (fp->f_flag & FALTIO)
    515 		ioflag |= IO_ALTSEMANTICS;
    516 	vn_lock(vp, LK_SHARED | LK_RETRY);
    517 	uio->uio_offset = *offset;
    518 	count = uio->uio_resid;
    519 	error = VOP_READ(vp, uio, ioflag, cred);
    520 	if (flags & FOF_UPDATE_OFFSET)
    521 		*offset += count - uio->uio_resid;
    522 	VOP_UNLOCK(vp, 0);
    523 	return (error);
    524 }
    525 
    526 /*
    527  * File table vnode write routine.
    528  */
    529 static int
    530 vn_write(fp, offset, uio, cred, flags)
    531 	struct file *fp;
    532 	off_t *offset;
    533 	struct uio *uio;
    534 	struct ucred *cred;
    535 	int flags;
    536 {
    537 	struct vnode *vp = (struct vnode *)fp->f_data;
    538 	int count, error, ioflag = IO_UNIT;
    539 
    540 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
    541 		ioflag |= IO_APPEND;
    542 	if (fp->f_flag & FNONBLOCK)
    543 		ioflag |= IO_NDELAY;
    544 	if (fp->f_flag & FFSYNC ||
    545 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
    546 		ioflag |= IO_SYNC;
    547 	else if (fp->f_flag & FDSYNC)
    548 		ioflag |= IO_DSYNC;
    549 	if (fp->f_flag & FALTIO)
    550 		ioflag |= IO_ALTSEMANTICS;
    551 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
    552 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    553 	uio->uio_offset = *offset;
    554 	count = uio->uio_resid;
    555 	error = VOP_WRITE(vp, uio, ioflag, cred);
    556 	if (flags & FOF_UPDATE_OFFSET) {
    557 		if (ioflag & IO_APPEND)
    558 			*offset = uio->uio_offset;
    559 		else
    560 			*offset += count - uio->uio_resid;
    561 	}
    562 	VOP_UNLOCK(vp, 0);
    563 	return (error);
    564 }
    565 
    566 /*
    567  * File table vnode stat routine.
    568  */
    569 static int
    570 vn_statfile(fp, sb, p)
    571 	struct file *fp;
    572 	struct stat *sb;
    573 	struct proc *p;
    574 {
    575 	struct vnode *vp = (struct vnode *)fp->f_data;
    576 
    577 	return vn_stat(vp, sb, p);
    578 }
    579 
    580 int
    581 vn_stat(vp, sb, p)
    582 	struct vnode *vp;
    583 	struct stat *sb;
    584 	struct proc *p;
    585 {
    586 	struct vattr va;
    587 	int error;
    588 	mode_t mode;
    589 
    590 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
    591 	if (error)
    592 		return (error);
    593 	/*
    594 	 * Copy from vattr table
    595 	 */
    596 	sb->st_dev = va.va_fsid;
    597 	sb->st_ino = va.va_fileid;
    598 	mode = va.va_mode;
    599 	switch (vp->v_type) {
    600 	case VREG:
    601 		mode |= S_IFREG;
    602 		break;
    603 	case VDIR:
    604 		mode |= S_IFDIR;
    605 		break;
    606 	case VBLK:
    607 		mode |= S_IFBLK;
    608 		break;
    609 	case VCHR:
    610 		mode |= S_IFCHR;
    611 		break;
    612 	case VLNK:
    613 		mode |= S_IFLNK;
    614 		break;
    615 	case VSOCK:
    616 		mode |= S_IFSOCK;
    617 		break;
    618 	case VFIFO:
    619 		mode |= S_IFIFO;
    620 		break;
    621 	default:
    622 		return (EBADF);
    623 	};
    624 	sb->st_mode = mode;
    625 	sb->st_nlink = va.va_nlink;
    626 	sb->st_uid = va.va_uid;
    627 	sb->st_gid = va.va_gid;
    628 	sb->st_rdev = va.va_rdev;
    629 	sb->st_size = va.va_size;
    630 	sb->st_atimespec = va.va_atime;
    631 	sb->st_mtimespec = va.va_mtime;
    632 	sb->st_ctimespec = va.va_ctime;
    633 	sb->st_blksize = va.va_blocksize;
    634 	sb->st_flags = va.va_flags;
    635 	sb->st_gen = 0;
    636 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
    637 	return (0);
    638 }
    639 
    640 /*
    641  * File table vnode fcntl routine.
    642  */
    643 static int
    644 vn_fcntl(fp, com, data, p)
    645 	struct file *fp;
    646 	u_int com;
    647 	caddr_t data;
    648 	struct proc *p;
    649 {
    650 	struct vnode *vp = ((struct vnode *)fp->f_data);
    651 	int error;
    652 
    653 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    654 	error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    655 	VOP_UNLOCK(vp, 0);
    656 	return (error);
    657 }
    658 
    659 /*
    660  * File table vnode ioctl routine.
    661  */
    662 static int
    663 vn_ioctl(fp, com, data, p)
    664 	struct file *fp;
    665 	u_long com;
    666 	caddr_t data;
    667 	struct proc *p;
    668 {
    669 	struct vnode *vp = ((struct vnode *)fp->f_data);
    670 	struct vattr vattr;
    671 	int error;
    672 
    673 	switch (vp->v_type) {
    674 
    675 	case VREG:
    676 	case VDIR:
    677 		if (com == FIONREAD) {
    678 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
    679 			if (error)
    680 				return (error);
    681 			*(int *)data = vattr.va_size - fp->f_offset;
    682 			return (0);
    683 		}
    684 		if (com == FIOGETBMAP) {
    685 			daddr_t *block;
    686 
    687 			if (*(daddr_t *)data < 0)
    688 				return (EINVAL);
    689 			block = (daddr_t *)data;
    690 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
    691 		}
    692 		if (com == OFIOGETBMAP) {
    693 			daddr_t ibn, obn;
    694 
    695 			if (*(int32_t *)data < 0)
    696 				return (EINVAL);
    697 			ibn = (daddr_t)*(int32_t *)data;
    698 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
    699 			*(int32_t *)data = (int32_t)obn;
    700 			return error;
    701 		}
    702 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
    703 			return (0);			/* XXX */
    704 		/* fall into ... */
    705 
    706 	case VFIFO:
    707 	case VCHR:
    708 	case VBLK:
    709 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
    710 		if (error == 0 && com == TIOCSCTTY) {
    711 			if (p->p_session->s_ttyvp)
    712 				vrele(p->p_session->s_ttyvp);
    713 			p->p_session->s_ttyvp = vp;
    714 			VREF(vp);
    715 		}
    716 		return (error);
    717 
    718 	default:
    719 		return (EPASSTHROUGH);
    720 	}
    721 }
    722 
    723 /*
    724  * File table vnode poll routine.
    725  */
    726 static int
    727 vn_poll(fp, events, p)
    728 	struct file *fp;
    729 	int events;
    730 	struct proc *p;
    731 {
    732 
    733 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
    734 }
    735 
    736 /*
    737  * File table vnode kqfilter routine.
    738  */
    739 int
    740 vn_kqfilter(fp, kn)
    741 	struct file *fp;
    742 	struct knote *kn;
    743 {
    744 
    745 	return (VOP_KQFILTER((struct vnode *)fp->f_data, kn));
    746 }
    747 
    748 /*
    749  * Check that the vnode is still valid, and if so
    750  * acquire requested lock.
    751  */
    752 int
    753 vn_lock(vp, flags)
    754 	struct vnode *vp;
    755 	int flags;
    756 {
    757 	int error;
    758 
    759 	do {
    760 		if ((flags & LK_INTERLOCK) == 0)
    761 			simple_lock(&vp->v_interlock);
    762 		if (vp->v_flag & VXLOCK) {
    763 			if (flags & LK_NOWAIT) {
    764 				simple_unlock(&vp->v_interlock);
    765 				return EBUSY;
    766 			}
    767 			vp->v_flag |= VXWANT;
    768 			ltsleep(vp, PINOD | PNORELOCK,
    769 			    "vn_lock", 0, &vp->v_interlock);
    770 			error = ENOENT;
    771 		} else {
    772 			error = VOP_LOCK(vp, flags | LK_INTERLOCK);
    773 			if (error == 0 || error == EDEADLK || error == EBUSY)
    774 				return (error);
    775 		}
    776 		flags &= ~LK_INTERLOCK;
    777 	} while (flags & LK_RETRY);
    778 	return (error);
    779 }
    780 
    781 /*
    782  * File table vnode close routine.
    783  */
    784 static int
    785 vn_closefile(fp, p)
    786 	struct file *fp;
    787 	struct proc *p;
    788 {
    789 
    790 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
    791 		fp->f_cred, p));
    792 }
    793 
    794 /*
    795  * Enable LK_CANRECURSE on lock. Return prior status.
    796  */
    797 u_int
    798 vn_setrecurse(vp)
    799 	struct vnode *vp;
    800 {
    801 	struct lock *lkp = &vp->v_lock;
    802 	u_int retval = lkp->lk_flags & LK_CANRECURSE;
    803 
    804 	lkp->lk_flags |= LK_CANRECURSE;
    805 	return retval;
    806 }
    807 
    808 /*
    809  * Called when done with locksetrecurse.
    810  */
    811 void
    812 vn_restorerecurse(vp, flags)
    813 	struct vnode *vp;
    814 	u_int flags;
    815 {
    816 	struct lock *lkp = &vp->v_lock;
    817 
    818 	lkp->lk_flags &= ~LK_CANRECURSE;
    819 	lkp->lk_flags |= flags;
    820 }
    821