Home | History | Annotate | Line # | Download | only in specfs
spec_vnops.c revision 1.104
      1 /*	$NetBSD: spec_vnops.c,v 1.104 2007/08/03 08:45:36 pooka 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 1.104 2007/08/03 08:45:36 pooka Exp $");
     36 
     37 #include <sys/param.h>
     38 #include <sys/proc.h>
     39 #include <sys/systm.h>
     40 #include <sys/kernel.h>
     41 #include <sys/conf.h>
     42 #include <sys/buf.h>
     43 #include <sys/mount.h>
     44 #include <sys/namei.h>
     45 #include <sys/vnode.h>
     46 #include <sys/stat.h>
     47 #include <sys/errno.h>
     48 #include <sys/ioctl.h>
     49 #include <sys/poll.h>
     50 #include <sys/file.h>
     51 #include <sys/disklabel.h>
     52 #include <sys/lockf.h>
     53 #include <sys/tty.h>
     54 #include <sys/kauth.h>
     55 
     56 #include <miscfs/genfs/genfs.h>
     57 #include <miscfs/specfs/specdev.h>
     58 
     59 /* symbolic sleep message strings for devices */
     60 const char	devopn[] = "devopn";
     61 const char	devio[] = "devio";
     62 const char	devwait[] = "devwait";
     63 const char	devin[] = "devin";
     64 const char	devout[] = "devout";
     65 const char	devioc[] = "devioc";
     66 const char	devcls[] = "devcls";
     67 
     68 struct vnode	*speclisth[SPECHSZ];
     69 
     70 /*
     71  * This vnode operations vector is used for two things only:
     72  * - special device nodes created from whole cloth by the kernel.
     73  * - as a temporary vnodeops replacement for vnodes which were found to
     74  *	be aliased by callers of checkalias().
     75  * For the ops vector for vnodes built from special devices found in a
     76  * filesystem, see (e.g) ffs_specop_entries[] in ffs_vnops.c or the
     77  * equivalent for other filesystems.
     78  */
     79 
     80 int (**spec_vnodeop_p)(void *);
     81 const struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
     82 	{ &vop_default_desc, vn_default_error },
     83 	{ &vop_lookup_desc, spec_lookup },		/* lookup */
     84 	{ &vop_create_desc, spec_create },		/* create */
     85 	{ &vop_mknod_desc, spec_mknod },		/* mknod */
     86 	{ &vop_open_desc, spec_open },			/* open */
     87 	{ &vop_close_desc, spec_close },		/* close */
     88 	{ &vop_access_desc, spec_access },		/* access */
     89 	{ &vop_getattr_desc, spec_getattr },		/* getattr */
     90 	{ &vop_setattr_desc, spec_setattr },		/* setattr */
     91 	{ &vop_read_desc, spec_read },			/* read */
     92 	{ &vop_write_desc, spec_write },		/* write */
     93 	{ &vop_lease_desc, spec_lease_check },		/* lease */
     94 	{ &vop_fcntl_desc, spec_fcntl },		/* fcntl */
     95 	{ &vop_ioctl_desc, spec_ioctl },		/* ioctl */
     96 	{ &vop_poll_desc, spec_poll },			/* poll */
     97 	{ &vop_kqfilter_desc, spec_kqfilter },		/* kqfilter */
     98 	{ &vop_revoke_desc, spec_revoke },		/* revoke */
     99 	{ &vop_mmap_desc, spec_mmap },			/* mmap */
    100 	{ &vop_fsync_desc, spec_fsync },		/* fsync */
    101 	{ &vop_seek_desc, spec_seek },			/* seek */
    102 	{ &vop_remove_desc, spec_remove },		/* remove */
    103 	{ &vop_link_desc, spec_link },			/* link */
    104 	{ &vop_rename_desc, spec_rename },		/* rename */
    105 	{ &vop_mkdir_desc, spec_mkdir },		/* mkdir */
    106 	{ &vop_rmdir_desc, spec_rmdir },		/* rmdir */
    107 	{ &vop_symlink_desc, spec_symlink },		/* symlink */
    108 	{ &vop_readdir_desc, spec_readdir },		/* readdir */
    109 	{ &vop_readlink_desc, spec_readlink },		/* readlink */
    110 	{ &vop_abortop_desc, spec_abortop },		/* abortop */
    111 	{ &vop_inactive_desc, spec_inactive },		/* inactive */
    112 	{ &vop_reclaim_desc, spec_reclaim },		/* reclaim */
    113 	{ &vop_lock_desc, spec_lock },			/* lock */
    114 	{ &vop_unlock_desc, spec_unlock },		/* unlock */
    115 	{ &vop_bmap_desc, spec_bmap },			/* bmap */
    116 	{ &vop_strategy_desc, spec_strategy },		/* strategy */
    117 	{ &vop_print_desc, spec_print },		/* print */
    118 	{ &vop_islocked_desc, spec_islocked },		/* islocked */
    119 	{ &vop_pathconf_desc, spec_pathconf },		/* pathconf */
    120 	{ &vop_advlock_desc, spec_advlock },		/* advlock */
    121 	{ &vop_bwrite_desc, spec_bwrite },		/* bwrite */
    122 	{ &vop_getpages_desc, spec_getpages },		/* getpages */
    123 	{ &vop_putpages_desc, spec_putpages },		/* putpages */
    124 	{ NULL, NULL }
    125 };
    126 const struct vnodeopv_desc spec_vnodeop_opv_desc =
    127 	{ &spec_vnodeop_p, spec_vnodeop_entries };
    128 
    129 /*
    130  * Trivial lookup routine that always fails.
    131  */
    132 int
    133 spec_lookup(void *v)
    134 {
    135 	struct vop_lookup_args /* {
    136 		struct vnode *a_dvp;
    137 		struct vnode **a_vpp;
    138 		struct componentname *a_cnp;
    139 	} */ *ap = v;
    140 
    141 	*ap->a_vpp = NULL;
    142 	return (ENOTDIR);
    143 }
    144 
    145 /*
    146  * Returns true if dev is /dev/mem or /dev/kmem.
    147  */
    148 int
    149 iskmemdev(dev_t dev)
    150 {
    151 	/* mem_no is emitted by config(8) to generated devsw.c */
    152 	extern const int mem_no;
    153 
    154 	/* minor 14 is /dev/io on i386 with COMPAT_10 */
    155 	return (major(dev) == mem_no && (minor(dev) < 2 || minor(dev) == 14));
    156 }
    157 
    158 /*
    159  * Open a special file.
    160  */
    161 /* ARGSUSED */
    162 int
    163 spec_open(void *v)
    164 {
    165 	struct vop_open_args /* {
    166 		struct vnode *a_vp;
    167 		int  a_mode;
    168 		kauth_cred_t a_cred;
    169 		struct lwp *a_l;
    170 	} */ *ap = v;
    171 	struct lwp *l = ap->a_l;
    172 	struct vnode *vp = ap->a_vp;
    173 	dev_t dev = (dev_t)vp->v_rdev;
    174 	int error;
    175 	struct partinfo pi;
    176 	enum kauth_device_req req;
    177 
    178 	/*
    179 	 * Don't allow open if fs is mounted -nodev.
    180 	 */
    181 	if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
    182 		return (ENXIO);
    183 
    184 #define M2K(m)	(((m) & FREAD) && ((m) & FWRITE) ? \
    185 		 KAUTH_REQ_DEVICE_RAWIO_SPEC_RW : \
    186 		 (m) & FWRITE ? KAUTH_REQ_DEVICE_RAWIO_SPEC_WRITE : \
    187 		 KAUTH_REQ_DEVICE_RAWIO_SPEC_READ)
    188 
    189 	switch (vp->v_type) {
    190 
    191 	case VCHR:
    192 		req = M2K(ap->a_mode);
    193 		error = kauth_authorize_device_spec(ap->a_cred, req, vp);
    194 		if (error)
    195 			return (error);
    196 
    197 		if (cdev_type(dev) == D_TTY)
    198 			vp->v_flag |= VISTTY;
    199 		VOP_UNLOCK(vp, 0);
    200 		error = cdev_open(dev, ap->a_mode, S_IFCHR, l);
    201 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    202 		if (cdev_type(dev) != D_DISK)
    203 			return error;
    204 		break;
    205 
    206 	case VBLK:
    207 		req = M2K(ap->a_mode);
    208 		error = kauth_authorize_device_spec(ap->a_cred, req, vp);
    209 		if (error)
    210 			return (error);
    211 		error = bdev_open(dev, ap->a_mode, S_IFBLK, l);
    212 		break;
    213 
    214 	case VNON:
    215 	case VLNK:
    216 	case VDIR:
    217 	case VREG:
    218 	case VBAD:
    219 	case VFIFO:
    220 	case VSOCK:
    221 	default:
    222 		return 0;
    223 	}
    224 
    225 #undef M2K
    226 
    227 	if (error)
    228 		return error;
    229 	if (vp->v_type == VCHR)
    230 		error = cdev_ioctl(vp->v_rdev, DIOCGPART, &pi, FREAD, curlwp);
    231 	else
    232 		error = bdev_ioctl(vp->v_rdev, DIOCGPART, &pi, FREAD, curlwp);
    233 	if (error == 0)
    234 		uvm_vnp_setsize(vp,
    235 		    (voff_t)pi.disklab->d_secsize * pi.part->p_size);
    236 	return 0;
    237 }
    238 
    239 /*
    240  * Vnode op for read
    241  */
    242 /* ARGSUSED */
    243 int
    244 spec_read(void *v)
    245 {
    246 	struct vop_read_args /* {
    247 		struct vnode *a_vp;
    248 		struct uio *a_uio;
    249 		int  a_ioflag;
    250 		kauth_cred_t a_cred;
    251 	} */ *ap = v;
    252 	struct vnode *vp = ap->a_vp;
    253 	struct uio *uio = ap->a_uio;
    254  	struct lwp *l = curlwp;
    255 	struct buf *bp;
    256 	daddr_t bn;
    257 	int bsize, bscale;
    258 	struct partinfo dpart;
    259 	int n, on;
    260 	int error = 0;
    261 
    262 #ifdef DIAGNOSTIC
    263 	if (uio->uio_rw != UIO_READ)
    264 		panic("spec_read mode");
    265 	if (&uio->uio_vmspace->vm_map != kernel_map &&
    266 	    uio->uio_vmspace != curproc->p_vmspace)
    267 		panic("spec_read proc");
    268 #endif
    269 	if (uio->uio_resid == 0)
    270 		return (0);
    271 
    272 	switch (vp->v_type) {
    273 
    274 	case VCHR:
    275 		VOP_UNLOCK(vp, 0);
    276 		error = cdev_read(vp->v_rdev, uio, ap->a_ioflag);
    277 		vn_lock(vp, LK_SHARED | LK_RETRY);
    278 		return (error);
    279 
    280 	case VBLK:
    281 		if (uio->uio_offset < 0)
    282 			return (EINVAL);
    283 		bsize = BLKDEV_IOSIZE;
    284 		if (bdev_ioctl(vp->v_rdev, DIOCGPART, &dpart, FREAD, l) == 0) {
    285 			if (dpart.part->p_fstype == FS_BSDFFS &&
    286 			    dpart.part->p_frag != 0 && dpart.part->p_fsize != 0)
    287 				bsize = dpart.part->p_frag *
    288 				    dpart.part->p_fsize;
    289 		}
    290 		bscale = bsize >> DEV_BSHIFT;
    291 		do {
    292 			bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1);
    293 			on = uio->uio_offset % bsize;
    294 			n = min((unsigned)(bsize - on), uio->uio_resid);
    295 			error = bread(vp, bn, bsize, NOCRED, &bp);
    296 			n = min(n, bsize - bp->b_resid);
    297 			if (error) {
    298 				brelse(bp);
    299 				return (error);
    300 			}
    301 			error = uiomove((char *)bp->b_data + on, n, uio);
    302 			brelse(bp);
    303 		} while (error == 0 && uio->uio_resid > 0 && n != 0);
    304 		return (error);
    305 
    306 	default:
    307 		panic("spec_read type");
    308 	}
    309 	/* NOTREACHED */
    310 }
    311 
    312 /*
    313  * Vnode op for write
    314  */
    315 /* ARGSUSED */
    316 int
    317 spec_write(void *v)
    318 {
    319 	struct vop_write_args /* {
    320 		struct vnode *a_vp;
    321 		struct uio *a_uio;
    322 		int  a_ioflag;
    323 		kauth_cred_t a_cred;
    324 	} */ *ap = v;
    325 	struct vnode *vp = ap->a_vp;
    326 	struct uio *uio = ap->a_uio;
    327 	struct lwp *l = curlwp;
    328 	struct buf *bp;
    329 	daddr_t bn;
    330 	int bsize, bscale;
    331 	struct partinfo dpart;
    332 	int n, on;
    333 	int error = 0;
    334 
    335 #ifdef DIAGNOSTIC
    336 	if (uio->uio_rw != UIO_WRITE)
    337 		panic("spec_write mode");
    338 	if (&uio->uio_vmspace->vm_map != kernel_map &&
    339 	    uio->uio_vmspace != curproc->p_vmspace)
    340 		panic("spec_write proc");
    341 #endif
    342 
    343 	switch (vp->v_type) {
    344 
    345 	case VCHR:
    346 		VOP_UNLOCK(vp, 0);
    347 		error = cdev_write(vp->v_rdev, uio, ap->a_ioflag);
    348 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    349 		return (error);
    350 
    351 	case VBLK:
    352 		if (uio->uio_resid == 0)
    353 			return (0);
    354 		if (uio->uio_offset < 0)
    355 			return (EINVAL);
    356 		bsize = BLKDEV_IOSIZE;
    357 		if (bdev_ioctl(vp->v_rdev, DIOCGPART, &dpart, FREAD, l) == 0) {
    358 			if (dpart.part->p_fstype == FS_BSDFFS &&
    359 			    dpart.part->p_frag != 0 && dpart.part->p_fsize != 0)
    360 				bsize = dpart.part->p_frag *
    361 				    dpart.part->p_fsize;
    362 		}
    363 		bscale = bsize >> DEV_BSHIFT;
    364 		do {
    365 			bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1);
    366 			on = uio->uio_offset % bsize;
    367 			n = min((unsigned)(bsize - on), uio->uio_resid);
    368 			if (n == bsize)
    369 				bp = getblk(vp, bn, bsize, 0, 0);
    370 			else
    371 				error = bread(vp, bn, bsize, NOCRED, &bp);
    372 			if (error) {
    373 				brelse(bp);
    374 				return (error);
    375 			}
    376 			n = min(n, bsize - bp->b_resid);
    377 			error = uiomove((char *)bp->b_data + on, n, uio);
    378 			if (error)
    379 				brelse(bp);
    380 			else {
    381 				if (n + on == bsize)
    382 					bawrite(bp);
    383 				else
    384 					bdwrite(bp);
    385 				if (bp->b_error != 0)
    386 					error = bp->b_error;
    387 			}
    388 		} while (error == 0 && uio->uio_resid > 0 && n != 0);
    389 		return (error);
    390 
    391 	default:
    392 		panic("spec_write type");
    393 	}
    394 	/* NOTREACHED */
    395 }
    396 
    397 /*
    398  * Device ioctl operation.
    399  */
    400 /* ARGSUSED */
    401 int
    402 spec_ioctl(void *v)
    403 {
    404 	struct vop_ioctl_args /* {
    405 		struct vnode *a_vp;
    406 		u_long a_command;
    407 		void  *a_data;
    408 		int  a_fflag;
    409 		kauth_cred_t a_cred;
    410 		struct lwp *a_l;
    411 	} */ *ap = v;
    412 	struct vnode *vp;
    413 	dev_t dev;
    414 
    415 	/*
    416 	 * Extract all the info we need from the vnode, taking care to
    417 	 * avoid a race with VOP_REVOKE().
    418 	 */
    419 
    420 	vp = ap->a_vp;
    421 	dev = NODEV;
    422 	simple_lock(&vp->v_interlock);
    423 	if ((vp->v_flag & VXLOCK) == 0 && vp->v_specinfo) {
    424 		dev = vp->v_rdev;
    425 	}
    426 	simple_unlock(&vp->v_interlock);
    427 	if (dev == NODEV) {
    428 		return ENXIO;
    429 	}
    430 
    431 	switch (vp->v_type) {
    432 
    433 	case VCHR:
    434 		return cdev_ioctl(dev, ap->a_command, ap->a_data,
    435 		    ap->a_fflag, ap->a_l);
    436 
    437 	case VBLK:
    438 		if (ap->a_command == 0 && (long)ap->a_data == B_TAPE) {
    439 			if (bdev_type(dev) == D_TAPE)
    440 				return (0);
    441 			else
    442 				return (1);
    443 		}
    444 		return bdev_ioctl(dev, ap->a_command, ap->a_data,
    445 		   ap->a_fflag, ap->a_l);
    446 
    447 	default:
    448 		panic("spec_ioctl");
    449 		/* NOTREACHED */
    450 	}
    451 }
    452 
    453 /* ARGSUSED */
    454 int
    455 spec_poll(void *v)
    456 {
    457 	struct vop_poll_args /* {
    458 		struct vnode *a_vp;
    459 		int a_events;
    460 		struct lwp *a_l;
    461 	} */ *ap = v;
    462 	struct vnode *vp;
    463 	dev_t dev;
    464 
    465 	/*
    466 	 * Extract all the info we need from the vnode, taking care to
    467 	 * avoid a race with VOP_REVOKE().
    468 	 */
    469 
    470 	vp = ap->a_vp;
    471 	dev = NODEV;
    472 	simple_lock(&vp->v_interlock);
    473 	if ((vp->v_flag & VXLOCK) == 0 && vp->v_specinfo) {
    474 		dev = vp->v_rdev;
    475 	}
    476 	simple_unlock(&vp->v_interlock);
    477 	if (dev == NODEV) {
    478 		return POLLERR;
    479 	}
    480 
    481 	switch (vp->v_type) {
    482 
    483 	case VCHR:
    484 		return cdev_poll(dev, ap->a_events, ap->a_l);
    485 
    486 	default:
    487 		return (genfs_poll(v));
    488 	}
    489 }
    490 
    491 /* ARGSUSED */
    492 int
    493 spec_kqfilter(void *v)
    494 {
    495 	struct vop_kqfilter_args /* {
    496 		struct vnode	*a_vp;
    497 		struct proc	*a_kn;
    498 	} */ *ap = v;
    499 	dev_t dev;
    500 
    501 	switch (ap->a_vp->v_type) {
    502 
    503 	case VCHR:
    504 		dev = ap->a_vp->v_rdev;
    505 		return cdev_kqfilter(dev, ap->a_kn);
    506 	default:
    507 		/*
    508 		 * Block devices don't support kqfilter, and refuse it
    509 		 * for any other files (like those vflush()ed) too.
    510 		 */
    511 		return (EOPNOTSUPP);
    512 	}
    513 }
    514 
    515 /*
    516  * Allow mapping of only D_DISK.  This is called only for VBLK.
    517  */
    518 int
    519 spec_mmap(void *v)
    520 {
    521 	struct vop_mmap_args /* {
    522 		struct vnode *a_vp;
    523 		vm_prot_t a_prot;
    524 		kauth_cred_t a_cred;
    525 		struct lwp *a_l;
    526 	} */ *ap = v;
    527 	struct vnode *vp = ap->a_vp;
    528 
    529 	KASSERT(vp->v_type == VBLK);
    530 	if (bdev_type(vp->v_rdev) != D_DISK)
    531 		return EINVAL;
    532 
    533 	return 0;
    534 }
    535 
    536 /*
    537  * Synch buffers associated with a block device
    538  */
    539 /* ARGSUSED */
    540 int
    541 spec_fsync(void *v)
    542 {
    543 	struct vop_fsync_args /* {
    544 		struct vnode *a_vp;
    545 		kauth_cred_t a_cred;
    546 		int  a_flags;
    547 		off_t offlo;
    548 		off_t offhi;
    549 		struct lwp *a_l;
    550 	} */ *ap = v;
    551 	struct vnode *vp = ap->a_vp;
    552 
    553 	if (vp->v_type == VBLK)
    554 		vflushbuf(vp, (ap->a_flags & FSYNC_WAIT) != 0);
    555 	return (0);
    556 }
    557 
    558 /*
    559  * Just call the device strategy routine
    560  */
    561 int
    562 spec_strategy(void *v)
    563 {
    564 	struct vop_strategy_args /* {
    565 		struct vnode *a_vp;
    566 		struct buf *a_bp;
    567 	} */ *ap = v;
    568 	struct vnode *vp = ap->a_vp;
    569 	struct buf *bp = ap->a_bp;
    570 	int error, s;
    571 	struct spec_cow_entry *e;
    572 
    573 	error = 0;
    574 	bp->b_dev = vp->v_rdev;
    575 	if (!(bp->b_flags & B_READ) &&
    576 	    (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
    577 		(*bioops.io_start)(bp);
    578 
    579 	if (!(bp->b_flags & B_READ) && !SLIST_EMPTY(&vp->v_spec_cow_head)) {
    580 		SPEC_COW_LOCK(vp->v_specinfo, s);
    581 		while (vp->v_spec_cow_req > 0)
    582 			ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0,
    583 			    &vp->v_spec_cow_slock);
    584 		vp->v_spec_cow_count++;
    585 		SPEC_COW_UNLOCK(vp->v_specinfo, s);
    586 
    587 		SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) {
    588 			if ((error = (*e->ce_func)(e->ce_cookie, bp)) != 0)
    589 				break;
    590 		}
    591 
    592 		SPEC_COW_LOCK(vp->v_specinfo, s);
    593 		vp->v_spec_cow_count--;
    594 		if (vp->v_spec_cow_req && vp->v_spec_cow_count == 0)
    595 			wakeup(&vp->v_spec_cow_req);
    596 		SPEC_COW_UNLOCK(vp->v_specinfo, s);
    597 	}
    598 
    599 	if (error) {
    600 		bp->b_error = error;
    601 		biodone(bp);
    602 		return (error);
    603 	}
    604 
    605 	bdev_strategy(bp);
    606 
    607 	return (0);
    608 }
    609 
    610 int
    611 spec_inactive(void *v)
    612 {
    613 	struct vop_inactive_args /* {
    614 		struct vnode *a_vp;
    615 		struct proc *a_l;
    616 	} */ *ap = v;
    617 
    618 	VOP_UNLOCK(ap->a_vp, 0);
    619 	return (0);
    620 }
    621 
    622 /*
    623  * This is a noop, simply returning what one has been given.
    624  */
    625 int
    626 spec_bmap(void *v)
    627 {
    628 	struct vop_bmap_args /* {
    629 		struct vnode *a_vp;
    630 		daddr_t  a_bn;
    631 		struct vnode **a_vpp;
    632 		daddr_t *a_bnp;
    633 		int *a_runp;
    634 	} */ *ap = v;
    635 
    636 	if (ap->a_vpp != NULL)
    637 		*ap->a_vpp = ap->a_vp;
    638 	if (ap->a_bnp != NULL)
    639 		*ap->a_bnp = ap->a_bn;
    640 	if (ap->a_runp != NULL)
    641 		*ap->a_runp = (MAXBSIZE >> DEV_BSHIFT) - 1;
    642 	return (0);
    643 }
    644 
    645 /*
    646  * Device close routine
    647  */
    648 /* ARGSUSED */
    649 int
    650 spec_close(void *v)
    651 {
    652 	struct vop_close_args /* {
    653 		struct vnode *a_vp;
    654 		int  a_fflag;
    655 		kauth_cred_t a_cred;
    656 		struct lwp *a_l;
    657 	} */ *ap = v;
    658 	struct vnode *vp = ap->a_vp;
    659 	struct session *sess;
    660 	dev_t dev = vp->v_rdev;
    661 	int mode, error, count, flags, flags1;
    662 
    663 	count = vcount(vp);
    664 	flags = vp->v_flag;
    665 
    666 	switch (vp->v_type) {
    667 
    668 	case VCHR:
    669 		/*
    670 		 * Hack: a tty device that is a controlling terminal
    671 		 * has a reference from the session structure.
    672 		 * We cannot easily tell that a character device is
    673 		 * a controlling terminal, unless it is the closing
    674 		 * process' controlling terminal.  In that case,
    675 		 * if the reference count is 2 (this last descriptor
    676 		 * plus the session), release the reference from the session.
    677 		 * Also remove the link from the tty back to the session
    678 		 * and pgrp - due to the way consoles are handled we cannot
    679 		 * guarantee that the vrele() will do the final close on the
    680 		 * actual tty device.
    681 		 */
    682 		mutex_enter(&proclist_lock);
    683 		if (count == 2 && ap->a_l &&
    684 		    vp == (sess = ap->a_l->l_proc->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 				mutex_exit(&proclist_lock);
    690 				SESSRELE(sess);
    691 			} else {
    692 				if (sess->s_ttyp->t_pgrp != NULL)
    693 					panic("spec_close: spurious pgrp ref");
    694 				mutex_exit(&proclist_lock);
    695 			}
    696 			vrele(vp);
    697 			count--;
    698 		} else
    699 			mutex_exit(&proclist_lock);
    700 
    701 		/*
    702 		 * If the vnode is locked, then we are in the midst
    703 		 * of forcably closing the device, otherwise we only
    704 		 * close on last reference.
    705 		 */
    706 		if (count > 1 && (flags & VXLOCK) == 0)
    707 			return (0);
    708 		mode = S_IFCHR;
    709 		break;
    710 
    711 	case VBLK:
    712 		/*
    713 		 * On last close of a block device (that isn't mounted)
    714 		 * we must invalidate any in core blocks, so that
    715 		 * we can, for instance, change floppy disks.
    716 		 */
    717 		error = vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_l, 0, 0);
    718 		if (error)
    719 			return (error);
    720 		/*
    721 		 * We do not want to really close the device if it
    722 		 * is still in use unless we are trying to close it
    723 		 * forcibly. Since every use (buffer, vnode, swap, cmap)
    724 		 * holds a reference to the vnode, and because we mark
    725 		 * any other vnodes that alias this device, when the
    726 		 * sum of the reference counts on all the aliased
    727 		 * vnodes descends to one, we are on last close.
    728 		 */
    729 		if (count > 1 && (flags & VXLOCK) == 0)
    730 			return (0);
    731 		mode = S_IFBLK;
    732 		break;
    733 
    734 	default:
    735 		panic("spec_close: not special");
    736 	}
    737 
    738 	flags1 = ap->a_fflag;
    739 
    740 	/*
    741 	 * if VXLOCK is set, then we're going away soon, so make this
    742 	 * non-blocking. Also ensures that we won't wedge in vn_lock below.
    743 	 */
    744 	if (flags & VXLOCK)
    745 		flags1 |= FNONBLOCK;
    746 
    747 	/*
    748 	 * If we're able to block, release the vnode lock & reacquire. We
    749 	 * might end up sleeping for someone else who wants our queues. They
    750 	 * won't get them if we hold the vnode locked. Also, if VXLOCK is
    751 	 * set, don't release the lock as we won't be able to regain it.
    752 	 */
    753 	if (!(flags1 & FNONBLOCK))
    754 		VOP_UNLOCK(vp, 0);
    755 
    756 	if (vp->v_type == VBLK)
    757 		error = bdev_close(dev, flags1, mode, ap->a_l);
    758 	else
    759 		error = cdev_close(dev, flags1, mode, ap->a_l);
    760 
    761 	if (!(flags1 & FNONBLOCK))
    762 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    763 
    764 	return (error);
    765 }
    766 
    767 /*
    768  * Print out the contents of a special device vnode.
    769  */
    770 int
    771 spec_print(void *v)
    772 {
    773 	struct vop_print_args /* {
    774 		struct vnode *a_vp;
    775 	} */ *ap = v;
    776 
    777 	printf("tag VT_NON, dev %d, %d\n", major(ap->a_vp->v_rdev),
    778 	    minor(ap->a_vp->v_rdev));
    779 	return 0;
    780 }
    781 
    782 /*
    783  * Return POSIX pathconf information applicable to special devices.
    784  */
    785 int
    786 spec_pathconf(void *v)
    787 {
    788 	struct vop_pathconf_args /* {
    789 		struct vnode *a_vp;
    790 		int a_name;
    791 		register_t *a_retval;
    792 	} */ *ap = v;
    793 
    794 	switch (ap->a_name) {
    795 	case _PC_LINK_MAX:
    796 		*ap->a_retval = LINK_MAX;
    797 		return (0);
    798 	case _PC_MAX_CANON:
    799 		*ap->a_retval = MAX_CANON;
    800 		return (0);
    801 	case _PC_MAX_INPUT:
    802 		*ap->a_retval = MAX_INPUT;
    803 		return (0);
    804 	case _PC_PIPE_BUF:
    805 		*ap->a_retval = PIPE_BUF;
    806 		return (0);
    807 	case _PC_CHOWN_RESTRICTED:
    808 		*ap->a_retval = 1;
    809 		return (0);
    810 	case _PC_VDISABLE:
    811 		*ap->a_retval = _POSIX_VDISABLE;
    812 		return (0);
    813 	case _PC_SYNC_IO:
    814 		*ap->a_retval = 1;
    815 		return (0);
    816 	default:
    817 		return (EINVAL);
    818 	}
    819 	/* NOTREACHED */
    820 }
    821 
    822 /*
    823  * Advisory record locking support.
    824  */
    825 int
    826 spec_advlock(void *v)
    827 {
    828 	struct vop_advlock_args /* {
    829 		struct vnode *a_vp;
    830 		void *a_id;
    831 		int a_op;
    832 		struct flock *a_fl;
    833 		int a_flags;
    834 	} */ *ap = v;
    835 	struct vnode *vp = ap->a_vp;
    836 
    837 	return lf_advlock(ap, &vp->v_speclockf, (off_t)0);
    838 }
    839