Home | History | Annotate | Line # | Download | only in specfs
spec_vnops.c revision 1.75
      1 /*	$NetBSD: spec_vnops.c,v 1.75 2003/12/10 11:40:12 hannken Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1989, 1993
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. Neither the name of the University nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  *
     31  *	@(#)spec_vnops.c	8.15 (Berkeley) 7/14/95
     32  */
     33 
     34 #include <sys/cdefs.h>
     35 __KERNEL_RCSID(0, "$NetBSD: spec_vnops.c,v 1.75 2003/12/10 11:40:12 hannken Exp $");
     36 
     37 #include "fss.h"
     38 
     39 #include <sys/param.h>
     40 #include <sys/proc.h>
     41 #include <sys/systm.h>
     42 #include <sys/kernel.h>
     43 #include <sys/conf.h>
     44 #include <sys/buf.h>
     45 #include <sys/mount.h>
     46 #include <sys/namei.h>
     47 #include <sys/vnode.h>
     48 #include <sys/stat.h>
     49 #include <sys/errno.h>
     50 #include <sys/ioctl.h>
     51 #include <sys/file.h>
     52 #include <sys/disklabel.h>
     53 #include <sys/lockf.h>
     54 #include <sys/tty.h>
     55 
     56 #include <miscfs/genfs/genfs.h>
     57 #include <miscfs/specfs/specdev.h>
     58 
     59 #if NFSS > 0
     60 #include <dev/fssvar.h>
     61 #endif
     62 
     63 /* symbolic sleep message strings for devices */
     64 const char	devopn[] = "devopn";
     65 const char	devio[] = "devio";
     66 const char	devwait[] = "devwait";
     67 const char	devin[] = "devin";
     68 const char	devout[] = "devout";
     69 const char	devioc[] = "devioc";
     70 const char	devcls[] = "devcls";
     71 
     72 struct vnode	*speclisth[SPECHSZ];
     73 
     74 /*
     75  * This vnode operations vector is used for two things only:
     76  * - special device nodes created from whole cloth by the kernel.
     77  * - as a temporary vnodeops replacement for vnodes which were found to
     78  *	be aliased by callers of checkalias().
     79  * For the ops vector for vnodes built from special devices found in a
     80  * filesystem, see (e.g) ffs_specop_entries[] in ffs_vnops.c or the
     81  * equivalent for other filesystems.
     82  */
     83 
     84 int (**spec_vnodeop_p) __P((void *));
     85 const struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
     86 	{ &vop_default_desc, vn_default_error },
     87 	{ &vop_lookup_desc, spec_lookup },		/* lookup */
     88 	{ &vop_create_desc, spec_create },		/* create */
     89 	{ &vop_mknod_desc, spec_mknod },		/* mknod */
     90 	{ &vop_open_desc, spec_open },			/* open */
     91 	{ &vop_close_desc, spec_close },		/* close */
     92 	{ &vop_access_desc, spec_access },		/* access */
     93 	{ &vop_getattr_desc, spec_getattr },		/* getattr */
     94 	{ &vop_setattr_desc, spec_setattr },		/* setattr */
     95 	{ &vop_read_desc, spec_read },			/* read */
     96 	{ &vop_write_desc, spec_write },		/* write */
     97 	{ &vop_lease_desc, spec_lease_check },		/* lease */
     98 	{ &vop_fcntl_desc, spec_fcntl },		/* fcntl */
     99 	{ &vop_ioctl_desc, spec_ioctl },		/* ioctl */
    100 	{ &vop_poll_desc, spec_poll },			/* poll */
    101 	{ &vop_kqfilter_desc, spec_kqfilter },		/* kqfilter */
    102 	{ &vop_revoke_desc, spec_revoke },		/* revoke */
    103 	{ &vop_mmap_desc, spec_mmap },			/* mmap */
    104 	{ &vop_fsync_desc, spec_fsync },		/* fsync */
    105 	{ &vop_seek_desc, spec_seek },			/* seek */
    106 	{ &vop_remove_desc, spec_remove },		/* remove */
    107 	{ &vop_link_desc, spec_link },			/* link */
    108 	{ &vop_rename_desc, spec_rename },		/* rename */
    109 	{ &vop_mkdir_desc, spec_mkdir },		/* mkdir */
    110 	{ &vop_rmdir_desc, spec_rmdir },		/* rmdir */
    111 	{ &vop_symlink_desc, spec_symlink },		/* symlink */
    112 	{ &vop_readdir_desc, spec_readdir },		/* readdir */
    113 	{ &vop_readlink_desc, spec_readlink },		/* readlink */
    114 	{ &vop_abortop_desc, spec_abortop },		/* abortop */
    115 	{ &vop_inactive_desc, spec_inactive },		/* inactive */
    116 	{ &vop_reclaim_desc, spec_reclaim },		/* reclaim */
    117 	{ &vop_lock_desc, spec_lock },			/* lock */
    118 	{ &vop_unlock_desc, spec_unlock },		/* unlock */
    119 	{ &vop_bmap_desc, spec_bmap },			/* bmap */
    120 	{ &vop_strategy_desc, spec_strategy },		/* strategy */
    121 	{ &vop_print_desc, spec_print },		/* print */
    122 	{ &vop_islocked_desc, spec_islocked },		/* islocked */
    123 	{ &vop_pathconf_desc, spec_pathconf },		/* pathconf */
    124 	{ &vop_advlock_desc, spec_advlock },		/* advlock */
    125 	{ &vop_blkatoff_desc, spec_blkatoff },		/* blkatoff */
    126 	{ &vop_valloc_desc, spec_valloc },		/* valloc */
    127 	{ &vop_vfree_desc, spec_vfree },		/* vfree */
    128 	{ &vop_truncate_desc, spec_truncate },		/* truncate */
    129 	{ &vop_update_desc, spec_update },		/* update */
    130 	{ &vop_bwrite_desc, spec_bwrite },		/* bwrite */
    131 	{ &vop_getpages_desc, spec_getpages },		/* getpages */
    132 	{ &vop_putpages_desc, spec_putpages },		/* putpages */
    133 	{ NULL, NULL }
    134 };
    135 const struct vnodeopv_desc spec_vnodeop_opv_desc =
    136 	{ &spec_vnodeop_p, spec_vnodeop_entries };
    137 
    138 /*
    139  * Trivial lookup routine that always fails.
    140  */
    141 int
    142 spec_lookup(v)
    143 	void *v;
    144 {
    145 	struct vop_lookup_args /* {
    146 		struct vnode *a_dvp;
    147 		struct vnode **a_vpp;
    148 		struct componentname *a_cnp;
    149 	} */ *ap = v;
    150 
    151 	*ap->a_vpp = NULL;
    152 	return (ENOTDIR);
    153 }
    154 
    155 /*
    156  * Returns true if dev is /dev/mem or /dev/kmem.
    157  */
    158 static int
    159 iskmemdev(dev_t dev)
    160 {
    161 	/* mem_no is emitted by config(8) to generated devsw.c */
    162 	extern const int mem_no;
    163 
    164 	/* minor 14 is /dev/io on i386 with COMPAT_10 */
    165 	return (major(dev) == mem_no && (minor(dev) < 2 || minor(dev) == 14));
    166 }
    167 
    168 /*
    169  * Open a special file.
    170  */
    171 /* ARGSUSED */
    172 int
    173 spec_open(v)
    174 	void *v;
    175 {
    176 	struct vop_open_args /* {
    177 		struct vnode *a_vp;
    178 		int  a_mode;
    179 		struct ucred *a_cred;
    180 		struct proc *a_p;
    181 	} */ *ap = v;
    182 	struct proc *p = ap->a_p;
    183 	struct vnode *bvp, *vp = ap->a_vp;
    184 	const struct bdevsw *bdev;
    185 	const struct cdevsw *cdev;
    186 	dev_t blkdev, dev = (dev_t)vp->v_rdev;
    187 	int error;
    188 	struct partinfo pi;
    189 	int (*d_ioctl)(dev_t, u_long, caddr_t, int, struct proc *);
    190 
    191 	/*
    192 	 * Don't allow open if fs is mounted -nodev.
    193 	 */
    194 	if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
    195 		return (ENXIO);
    196 
    197 	switch (vp->v_type) {
    198 
    199 	case VCHR:
    200 		cdev = cdevsw_lookup(dev);
    201 		if (cdev == NULL)
    202 			return (ENXIO);
    203 		if (ap->a_cred != FSCRED && (ap->a_mode & FWRITE)) {
    204 			/*
    205 			 * When running in very secure mode, do not allow
    206 			 * opens for writing of any disk character devices.
    207 			 */
    208 			if (securelevel >= 2 && cdev->d_type == D_DISK)
    209 				return (EPERM);
    210 			/*
    211 			 * When running in secure mode, do not allow opens
    212 			 * for writing of /dev/mem, /dev/kmem, or character
    213 			 * devices whose corresponding block devices are
    214 			 * currently mounted.
    215 			 */
    216 			if (securelevel >= 1) {
    217 				blkdev = devsw_chr2blk(dev);
    218 				if (blkdev != (dev_t)NODEV &&
    219 				    vfinddev(blkdev, VBLK, &bvp) &&
    220 				    (error = vfs_mountedon(bvp)))
    221 					return (error);
    222 				if (iskmemdev(dev))
    223 					return (EPERM);
    224 			}
    225 		}
    226 		if (cdev->d_type == D_TTY)
    227 			vp->v_flag |= VISTTY;
    228 		VOP_UNLOCK(vp, 0);
    229 		error = (*cdev->d_open)(dev, ap->a_mode, S_IFCHR, p);
    230 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    231 		if (cdev->d_type != D_DISK)
    232 			return error;
    233 		d_ioctl = cdev->d_ioctl;
    234 		break;
    235 
    236 	case VBLK:
    237 		bdev = bdevsw_lookup(dev);
    238 		if (bdev == NULL)
    239 			return (ENXIO);
    240 		/*
    241 		 * When running in very secure mode, do not allow
    242 		 * opens for writing of any disk block devices.
    243 		 */
    244 		if (securelevel >= 2 && ap->a_cred != FSCRED &&
    245 		    (ap->a_mode & FWRITE) && bdev->d_type == D_DISK)
    246 			return (EPERM);
    247 		/*
    248 		 * Do not allow opens of block devices that are
    249 		 * currently mounted.
    250 		 */
    251 		if ((error = vfs_mountedon(vp)) != 0)
    252 			return (error);
    253 		error = (*bdev->d_open)(dev, ap->a_mode, S_IFBLK, p);
    254 		d_ioctl = bdev->d_ioctl;
    255 		break;
    256 
    257 	case VNON:
    258 	case VLNK:
    259 	case VDIR:
    260 	case VREG:
    261 	case VBAD:
    262 	case VFIFO:
    263 	case VSOCK:
    264 	default:
    265 		return 0;
    266 	}
    267 
    268 	if (error)
    269 		return error;
    270 	if (!(*d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&pi, FREAD, curproc))
    271 		vp->v_size = (voff_t)pi.disklab->d_secsize * pi.part->p_size;
    272 	return 0;
    273 }
    274 
    275 /*
    276  * Vnode op for read
    277  */
    278 /* ARGSUSED */
    279 int
    280 spec_read(v)
    281 	void *v;
    282 {
    283 	struct vop_read_args /* {
    284 		struct vnode *a_vp;
    285 		struct uio *a_uio;
    286 		int  a_ioflag;
    287 		struct ucred *a_cred;
    288 	} */ *ap = v;
    289 	struct vnode *vp = ap->a_vp;
    290 	struct uio *uio = ap->a_uio;
    291  	struct proc *p = uio->uio_procp;
    292 	struct buf *bp;
    293 	const struct bdevsw *bdev;
    294 	const struct cdevsw *cdev;
    295 	daddr_t bn;
    296 	int bsize, bscale;
    297 	struct partinfo dpart;
    298 	int n, on;
    299 	int error = 0;
    300 
    301 #ifdef DIAGNOSTIC
    302 	if (uio->uio_rw != UIO_READ)
    303 		panic("spec_read mode");
    304 	if (uio->uio_segflg == UIO_USERSPACE && uio->uio_procp != curproc)
    305 		panic("spec_read proc");
    306 #endif
    307 	if (uio->uio_resid == 0)
    308 		return (0);
    309 
    310 	switch (vp->v_type) {
    311 
    312 	case VCHR:
    313 		VOP_UNLOCK(vp, 0);
    314 		cdev = cdevsw_lookup(vp->v_rdev);
    315 		if (cdev != NULL)
    316 			error = (*cdev->d_read)(vp->v_rdev, uio, ap->a_ioflag);
    317 		else
    318 			error = ENXIO;
    319 		vn_lock(vp, LK_SHARED | LK_RETRY);
    320 		return (error);
    321 
    322 	case VBLK:
    323 		if (uio->uio_offset < 0)
    324 			return (EINVAL);
    325 		bsize = BLKDEV_IOSIZE;
    326 		bdev = bdevsw_lookup(vp->v_rdev);
    327 		if (bdev != NULL &&
    328 		    (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart,
    329 				     FREAD, p) == 0) {
    330 			if (dpart.part->p_fstype == FS_BSDFFS &&
    331 			    dpart.part->p_frag != 0 && dpart.part->p_fsize != 0)
    332 				bsize = dpart.part->p_frag *
    333 				    dpart.part->p_fsize;
    334 		}
    335 		bscale = bsize >> DEV_BSHIFT;
    336 		do {
    337 			bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1);
    338 			on = uio->uio_offset % bsize;
    339 			n = min((unsigned)(bsize - on), uio->uio_resid);
    340 			error = bread(vp, bn, bsize, NOCRED, &bp);
    341 			n = min(n, bsize - bp->b_resid);
    342 			if (error) {
    343 				brelse(bp);
    344 				return (error);
    345 			}
    346 			error = uiomove((char *)bp->b_data + on, n, uio);
    347 			brelse(bp);
    348 		} while (error == 0 && uio->uio_resid > 0 && n != 0);
    349 		return (error);
    350 
    351 	default:
    352 		panic("spec_read type");
    353 	}
    354 	/* NOTREACHED */
    355 }
    356 
    357 /*
    358  * Vnode op for write
    359  */
    360 /* ARGSUSED */
    361 int
    362 spec_write(v)
    363 	void *v;
    364 {
    365 	struct vop_write_args /* {
    366 		struct vnode *a_vp;
    367 		struct uio *a_uio;
    368 		int  a_ioflag;
    369 		struct ucred *a_cred;
    370 	} */ *ap = v;
    371 	struct vnode *vp = ap->a_vp;
    372 	struct uio *uio = ap->a_uio;
    373 	struct proc *p = uio->uio_procp;
    374 	struct buf *bp;
    375 	const struct bdevsw *bdev;
    376 	const struct cdevsw *cdev;
    377 	daddr_t bn;
    378 	int bsize, bscale;
    379 	struct partinfo dpart;
    380 	int n, on;
    381 	int error = 0;
    382 
    383 #ifdef DIAGNOSTIC
    384 	if (uio->uio_rw != UIO_WRITE)
    385 		panic("spec_write mode");
    386 	if (uio->uio_segflg == UIO_USERSPACE && uio->uio_procp != curproc)
    387 		panic("spec_write proc");
    388 #endif
    389 
    390 	switch (vp->v_type) {
    391 
    392 	case VCHR:
    393 		VOP_UNLOCK(vp, 0);
    394 		cdev = cdevsw_lookup(vp->v_rdev);
    395 		if (cdev != NULL)
    396 			error = (*cdev->d_write)(vp->v_rdev, uio, ap->a_ioflag);
    397 		else
    398 			error = ENXIO;
    399 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    400 		return (error);
    401 
    402 	case VBLK:
    403 		if (uio->uio_resid == 0)
    404 			return (0);
    405 		if (uio->uio_offset < 0)
    406 			return (EINVAL);
    407 		bsize = BLKDEV_IOSIZE;
    408 		bdev = bdevsw_lookup(vp->v_rdev);
    409 		if (bdev != NULL &&
    410 		    (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart,
    411 				    FREAD, p) == 0) {
    412 			if (dpart.part->p_fstype == FS_BSDFFS &&
    413 			    dpart.part->p_frag != 0 && dpart.part->p_fsize != 0)
    414 				bsize = dpart.part->p_frag *
    415 				    dpart.part->p_fsize;
    416 		}
    417 		bscale = bsize >> DEV_BSHIFT;
    418 		do {
    419 			bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1);
    420 			on = uio->uio_offset % bsize;
    421 			n = min((unsigned)(bsize - on), uio->uio_resid);
    422 			if (n == bsize)
    423 				bp = getblk(vp, bn, bsize, 0, 0);
    424 			else
    425 				error = bread(vp, bn, bsize, NOCRED, &bp);
    426 			if (error) {
    427 				brelse(bp);
    428 				return (error);
    429 			}
    430 			n = min(n, bsize - bp->b_resid);
    431 			error = uiomove((char *)bp->b_data + on, n, uio);
    432 			if (error)
    433 				brelse(bp);
    434 			else {
    435 				if (n + on == bsize)
    436 					bawrite(bp);
    437 				else
    438 					bdwrite(bp);
    439 				if (bp->b_flags & B_ERROR)
    440 					error = bp->b_error;
    441 			}
    442 		} while (error == 0 && uio->uio_resid > 0 && n != 0);
    443 		return (error);
    444 
    445 	default:
    446 		panic("spec_write type");
    447 	}
    448 	/* NOTREACHED */
    449 }
    450 
    451 /*
    452  * Device ioctl operation.
    453  */
    454 /* ARGSUSED */
    455 int
    456 spec_ioctl(v)
    457 	void *v;
    458 {
    459 	struct vop_ioctl_args /* {
    460 		struct vnode *a_vp;
    461 		u_long a_command;
    462 		caddr_t  a_data;
    463 		int  a_fflag;
    464 		struct ucred *a_cred;
    465 		struct proc *a_p;
    466 	} */ *ap = v;
    467 	const struct bdevsw *bdev;
    468 	const struct cdevsw *cdev;
    469 	dev_t dev = ap->a_vp->v_rdev;
    470 
    471 	switch (ap->a_vp->v_type) {
    472 
    473 	case VCHR:
    474 		cdev = cdevsw_lookup(dev);
    475 		if (cdev == NULL)
    476 			return (ENXIO);
    477 		return ((*cdev->d_ioctl)(dev, ap->a_command, ap->a_data,
    478 		    ap->a_fflag, ap->a_p));
    479 
    480 	case VBLK:
    481 		bdev = bdevsw_lookup(dev);
    482 		if (bdev == NULL)
    483 			return (ENXIO);
    484 		if (ap->a_command == 0 && (long)ap->a_data == B_TAPE) {
    485 			if (bdev->d_type == D_TAPE)
    486 				return (0);
    487 			else
    488 				return (1);
    489 		}
    490 		return ((*bdev->d_ioctl)(dev, ap->a_command, ap->a_data,
    491 		   ap->a_fflag, ap->a_p));
    492 
    493 	default:
    494 		panic("spec_ioctl");
    495 		/* NOTREACHED */
    496 	}
    497 }
    498 
    499 /* ARGSUSED */
    500 int
    501 spec_poll(v)
    502 	void *v;
    503 {
    504 	struct vop_poll_args /* {
    505 		struct vnode *a_vp;
    506 		int a_events;
    507 		struct proc *a_p;
    508 	} */ *ap = v;
    509 	const struct cdevsw *cdev;
    510 	dev_t dev;
    511 
    512 	switch (ap->a_vp->v_type) {
    513 
    514 	case VCHR:
    515 		dev = ap->a_vp->v_rdev;
    516 		cdev = cdevsw_lookup(dev);
    517 		if (cdev == NULL)
    518 			return (ENXIO);
    519 		return (*cdev->d_poll)(dev, ap->a_events, ap->a_p);
    520 
    521 	default:
    522 		return (genfs_poll(v));
    523 	}
    524 }
    525 
    526 /* ARGSUSED */
    527 int
    528 spec_kqfilter(v)
    529 	void *v;
    530 {
    531 	struct vop_kqfilter_args /* {
    532 		struct vnode	*a_vp;
    533 		struct proc	*a_kn;
    534 	} */ *ap = v;
    535 	const struct cdevsw *cdev;
    536 	dev_t dev;
    537 
    538 	switch (ap->a_vp->v_type) {
    539 
    540 	case VCHR:
    541 		dev = ap->a_vp->v_rdev;
    542 		cdev = cdevsw_lookup(dev);
    543 		if (cdev == NULL)
    544 			return (ENXIO);
    545 		return (*cdev->d_kqfilter)(dev, ap->a_kn);
    546 	default:
    547 		/*
    548 		 * Block devices don't support kqfilter, and refuse it
    549 		 * for any other files (like those vflush()ed) too.
    550 		 */
    551 		return (EOPNOTSUPP);
    552 	}
    553 }
    554 
    555 /*
    556  * Synch buffers associated with a block device
    557  */
    558 /* ARGSUSED */
    559 int
    560 spec_fsync(v)
    561 	void *v;
    562 {
    563 	struct vop_fsync_args /* {
    564 		struct vnode *a_vp;
    565 		struct ucred *a_cred;
    566 		int  a_flags;
    567 		off_t offlo;
    568 		off_t offhi;
    569 		struct proc *a_p;
    570 	} */ *ap = v;
    571 	struct vnode *vp = ap->a_vp;
    572 
    573 	if (vp->v_type == VBLK)
    574 		vflushbuf(vp, (ap->a_flags & FSYNC_WAIT) != 0);
    575 	return (0);
    576 }
    577 
    578 /*
    579  * Just call the device strategy routine
    580  */
    581 int
    582 spec_strategy(v)
    583 	void *v;
    584 {
    585 	struct vop_strategy_args /* {
    586 		struct buf *a_bp;
    587 	} */ *ap = v;
    588 	struct buf *bp;
    589 	const struct bdevsw *bdev;
    590 
    591 	bp = ap->a_bp;
    592 	if (!(bp->b_flags & B_READ) &&
    593 	    (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
    594 		(*bioops.io_start)(bp);
    595 	bdev = bdevsw_lookup(bp->b_dev);
    596 	if (bdev != NULL) {
    597 #if NFSS > 0
    598 		fss_cow_hook(bp);
    599 #endif
    600 		(*bdev->d_strategy)(bp);
    601 	}
    602 	return (0);
    603 }
    604 
    605 int
    606 spec_inactive(v)
    607 	void *v;
    608 {
    609 	struct vop_inactive_args /* {
    610 		struct vnode *a_vp;
    611 		struct proc *a_p;
    612 	} */ *ap = v;
    613 
    614 	VOP_UNLOCK(ap->a_vp, 0);
    615 	return (0);
    616 }
    617 
    618 /*
    619  * This is a noop, simply returning what one has been given.
    620  */
    621 int
    622 spec_bmap(v)
    623 	void *v;
    624 {
    625 	struct vop_bmap_args /* {
    626 		struct vnode *a_vp;
    627 		daddr_t  a_bn;
    628 		struct vnode **a_vpp;
    629 		daddr_t *a_bnp;
    630 		int *a_runp;
    631 	} */ *ap = v;
    632 
    633 	if (ap->a_vpp != NULL)
    634 		*ap->a_vpp = ap->a_vp;
    635 	if (ap->a_bnp != NULL)
    636 		*ap->a_bnp = ap->a_bn;
    637 	if (ap->a_runp != NULL)
    638 		*ap->a_runp = (MAXBSIZE >> DEV_BSHIFT) - 1;
    639 	return (0);
    640 }
    641 
    642 /*
    643  * Device close routine
    644  */
    645 /* ARGSUSED */
    646 int
    647 spec_close(v)
    648 	void *v;
    649 {
    650 	struct vop_close_args /* {
    651 		struct vnode *a_vp;
    652 		int  a_fflag;
    653 		struct ucred *a_cred;
    654 		struct proc *a_p;
    655 	} */ *ap = v;
    656 	struct vnode *vp = ap->a_vp;
    657 	const struct bdevsw *bdev;
    658 	const struct cdevsw *cdev;
    659 	struct session *sess;
    660 	dev_t dev = vp->v_rdev;
    661 	int (*devclose) __P((dev_t, int, int, struct proc *));
    662 	int mode, error, count, flags, flags1;
    663 
    664 	count = vcount(vp);
    665 	flags = vp->v_flag;
    666 
    667 	switch (vp->v_type) {
    668 
    669 	case VCHR:
    670 		/*
    671 		 * Hack: a tty device that is a controlling terminal
    672 		 * has a reference from the session structure.
    673 		 * We cannot easily tell that a character device is
    674 		 * a controlling terminal, unless it is the closing
    675 		 * process' controlling terminal.  In that case,
    676 		 * if the reference count is 2 (this last descriptor
    677 		 * plus the session), release the reference from the session.
    678 		 * Also remove the link from the tty back to the session
    679 		 * and pgrp - due to the way consoles are handled we cannot
    680 		 * guarantee that the vrele() will do the final close on the
    681 		 * actual tty device.
    682 		 */
    683 		if (count == 2 && ap->a_p &&
    684 		    vp == (sess = ap->a_p->p_session)->s_ttyvp) {
    685 			sess->s_ttyvp = NULL;
    686 			if (sess->s_ttyp->t_session != NULL) {
    687 				sess->s_ttyp->t_pgrp = NULL;
    688 				sess->s_ttyp->t_session = NULL;
    689 				SESSRELE(sess);
    690 			} else if (sess->s_ttyp->t_pgrp != NULL)
    691 				panic("spec_close: spurious pgrp ref");
    692 			vrele(vp);
    693 			count--;
    694 		}
    695 		/*
    696 		 * If the vnode is locked, then we are in the midst
    697 		 * of forcably closing the device, otherwise we only
    698 		 * close on last reference.
    699 		 */
    700 		if (count > 1 && (flags & VXLOCK) == 0)
    701 			return (0);
    702 		cdev = cdevsw_lookup(dev);
    703 		if (cdev != NULL)
    704 			devclose = cdev->d_close;
    705 		else
    706 			devclose = NULL;
    707 		mode = S_IFCHR;
    708 		break;
    709 
    710 	case VBLK:
    711 		/*
    712 		 * On last close of a block device (that isn't mounted)
    713 		 * we must invalidate any in core blocks, so that
    714 		 * we can, for instance, change floppy disks.
    715 		 */
    716 		error = vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 0, 0);
    717 		if (error)
    718 			return (error);
    719 		/*
    720 		 * We do not want to really close the device if it
    721 		 * is still in use unless we are trying to close it
    722 		 * forcibly. Since every use (buffer, vnode, swap, cmap)
    723 		 * holds a reference to the vnode, and because we mark
    724 		 * any other vnodes that alias this device, when the
    725 		 * sum of the reference counts on all the aliased
    726 		 * vnodes descends to one, we are on last close.
    727 		 */
    728 		if (count > 1 && (flags & VXLOCK) == 0)
    729 			return (0);
    730 		bdev = bdevsw_lookup(dev);
    731 		if (bdev != NULL)
    732 			devclose = bdev->d_close;
    733 		else
    734 			devclose = NULL;
    735 		mode = S_IFBLK;
    736 		break;
    737 
    738 	default:
    739 		panic("spec_close: not special");
    740 	}
    741 
    742 	flags1 = ap->a_fflag;
    743 
    744 	/*
    745 	 * if VXLOCK is set, then we're going away soon, so make this
    746 	 * non-blocking. Also ensures that we won't wedge in vn_lock below.
    747 	 */
    748 	if (flags & VXLOCK)
    749 		flags1 |= FNONBLOCK;
    750 
    751 	/*
    752 	 * If we're able to block, release the vnode lock & reacquire. We
    753 	 * might end up sleeping for someone else who wants our queues. They
    754 	 * won't get them if we hold the vnode locked. Also, if VXLOCK is set,
    755 	 * don't release the lock as we won't be able to regain it.
    756 	 */
    757 	if (!(flags1 & FNONBLOCK))
    758 		VOP_UNLOCK(vp, 0);
    759 
    760 	if (devclose != NULL)
    761 		error = (*devclose)(dev, flags1, mode, ap->a_p);
    762 	else
    763 		error = ENXIO;
    764 
    765 	if (!(flags1 & FNONBLOCK))
    766 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    767 
    768 	return (error);
    769 }
    770 
    771 /*
    772  * Print out the contents of a special device vnode.
    773  */
    774 int
    775 spec_print(v)
    776 	void *v;
    777 {
    778 	struct vop_print_args /* {
    779 		struct vnode *a_vp;
    780 	} */ *ap = v;
    781 
    782 	printf("tag VT_NON, dev %d, %d\n", major(ap->a_vp->v_rdev),
    783 	    minor(ap->a_vp->v_rdev));
    784 	return 0;
    785 }
    786 
    787 /*
    788  * Return POSIX pathconf information applicable to special devices.
    789  */
    790 int
    791 spec_pathconf(v)
    792 	void *v;
    793 {
    794 	struct vop_pathconf_args /* {
    795 		struct vnode *a_vp;
    796 		int a_name;
    797 		register_t *a_retval;
    798 	} */ *ap = v;
    799 
    800 	switch (ap->a_name) {
    801 	case _PC_LINK_MAX:
    802 		*ap->a_retval = LINK_MAX;
    803 		return (0);
    804 	case _PC_MAX_CANON:
    805 		*ap->a_retval = MAX_CANON;
    806 		return (0);
    807 	case _PC_MAX_INPUT:
    808 		*ap->a_retval = MAX_INPUT;
    809 		return (0);
    810 	case _PC_PIPE_BUF:
    811 		*ap->a_retval = PIPE_BUF;
    812 		return (0);
    813 	case _PC_CHOWN_RESTRICTED:
    814 		*ap->a_retval = 1;
    815 		return (0);
    816 	case _PC_VDISABLE:
    817 		*ap->a_retval = _POSIX_VDISABLE;
    818 		return (0);
    819 	case _PC_SYNC_IO:
    820 		*ap->a_retval = 1;
    821 		return (0);
    822 	default:
    823 		return (EINVAL);
    824 	}
    825 	/* NOTREACHED */
    826 }
    827 
    828 /*
    829  * Advisory record locking support.
    830  */
    831 int
    832 spec_advlock(v)
    833 	void *v;
    834 {
    835 	struct vop_advlock_args /* {
    836 		struct vnode *a_vp;
    837 		caddr_t a_id;
    838 		int a_op;
    839 		struct flock *a_fl;
    840 		int a_flags;
    841 	} */ *ap = v;
    842 	struct vnode *vp = ap->a_vp;
    843 
    844 	return lf_advlock(ap, &vp->v_speclockf, (off_t)0);
    845 }
    846