xref: /netbsd/sys/miscfs/specfs/spec_vnops.c (revision c4a72b64)
1 /*	$NetBSD: spec_vnops.c,v 1.66 2002/10/26 13:50:17 jdolecek 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. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)spec_vnops.c	8.15 (Berkeley) 7/14/95
36  */
37 
38 #include <sys/cdefs.h>
39 __KERNEL_RCSID(0, "$NetBSD: spec_vnops.c,v 1.66 2002/10/26 13:50:17 jdolecek Exp $");
40 
41 #include <sys/param.h>
42 #include <sys/proc.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/conf.h>
46 #include <sys/buf.h>
47 #include <sys/mount.h>
48 #include <sys/namei.h>
49 #include <sys/vnode.h>
50 #include <sys/stat.h>
51 #include <sys/errno.h>
52 #include <sys/ioctl.h>
53 #include <sys/file.h>
54 #include <sys/disklabel.h>
55 #include <sys/lockf.h>
56 
57 #include <miscfs/genfs/genfs.h>
58 #include <miscfs/specfs/specdev.h>
59 
60 /* symbolic sleep message strings for devices */
61 const char	devopn[] = "devopn";
62 const char	devio[] = "devio";
63 const char	devwait[] = "devwait";
64 const char	devin[] = "devin";
65 const char	devout[] = "devout";
66 const char	devioc[] = "devioc";
67 const char	devcls[] = "devcls";
68 
69 struct vnode	*speclisth[SPECHSZ];
70 
71 /*
72  * This vnode operations vector is used for two things only:
73  * - special device nodes created from whole cloth by the kernel.
74  * - as a temporary vnodeops replacement for vnodes which were found to
75  *	be aliased by callers of checkalias().
76  * For the ops vector for vnodes built from special devices found in a
77  * filesystem, see (e.g) ffs_specop_entries[] in ffs_vnops.c or the
78  * equivalent for other filesystems.
79  */
80 
81 int (**spec_vnodeop_p) __P((void *));
82 const struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
83 	{ &vop_default_desc, vn_default_error },
84 	{ &vop_lookup_desc, spec_lookup },		/* lookup */
85 	{ &vop_create_desc, spec_create },		/* create */
86 	{ &vop_mknod_desc, spec_mknod },		/* mknod */
87 	{ &vop_open_desc, spec_open },			/* open */
88 	{ &vop_close_desc, spec_close },		/* close */
89 	{ &vop_access_desc, spec_access },		/* access */
90 	{ &vop_getattr_desc, spec_getattr },		/* getattr */
91 	{ &vop_setattr_desc, spec_setattr },		/* setattr */
92 	{ &vop_read_desc, spec_read },			/* read */
93 	{ &vop_write_desc, spec_write },		/* write */
94 	{ &vop_lease_desc, spec_lease_check },		/* lease */
95 	{ &vop_fcntl_desc, spec_fcntl },		/* fcntl */
96 	{ &vop_ioctl_desc, spec_ioctl },		/* ioctl */
97 	{ &vop_poll_desc, spec_poll },			/* poll */
98 	{ &vop_kqfilter_desc, spec_kqfilter },		/* kqfilter */
99 	{ &vop_revoke_desc, spec_revoke },		/* revoke */
100 	{ &vop_mmap_desc, spec_mmap },			/* mmap */
101 	{ &vop_fsync_desc, spec_fsync },		/* fsync */
102 	{ &vop_seek_desc, spec_seek },			/* seek */
103 	{ &vop_remove_desc, spec_remove },		/* remove */
104 	{ &vop_link_desc, spec_link },			/* link */
105 	{ &vop_rename_desc, spec_rename },		/* rename */
106 	{ &vop_mkdir_desc, spec_mkdir },		/* mkdir */
107 	{ &vop_rmdir_desc, spec_rmdir },		/* rmdir */
108 	{ &vop_symlink_desc, spec_symlink },		/* symlink */
109 	{ &vop_readdir_desc, spec_readdir },		/* readdir */
110 	{ &vop_readlink_desc, spec_readlink },		/* readlink */
111 	{ &vop_abortop_desc, spec_abortop },		/* abortop */
112 	{ &vop_inactive_desc, spec_inactive },		/* inactive */
113 	{ &vop_reclaim_desc, spec_reclaim },		/* reclaim */
114 	{ &vop_lock_desc, spec_lock },			/* lock */
115 	{ &vop_unlock_desc, spec_unlock },		/* unlock */
116 	{ &vop_bmap_desc, spec_bmap },			/* bmap */
117 	{ &vop_strategy_desc, spec_strategy },		/* strategy */
118 	{ &vop_print_desc, spec_print },		/* print */
119 	{ &vop_islocked_desc, spec_islocked },		/* islocked */
120 	{ &vop_pathconf_desc, spec_pathconf },		/* pathconf */
121 	{ &vop_advlock_desc, spec_advlock },		/* advlock */
122 	{ &vop_blkatoff_desc, spec_blkatoff },		/* blkatoff */
123 	{ &vop_valloc_desc, spec_valloc },		/* valloc */
124 	{ &vop_vfree_desc, spec_vfree },		/* vfree */
125 	{ &vop_truncate_desc, spec_truncate },		/* truncate */
126 	{ &vop_update_desc, spec_update },		/* update */
127 	{ &vop_bwrite_desc, spec_bwrite },		/* bwrite */
128 	{ &vop_getpages_desc, spec_getpages },		/* getpages */
129 	{ &vop_putpages_desc, spec_putpages },		/* putpages */
130 	{ NULL, NULL }
131 };
132 const struct vnodeopv_desc spec_vnodeop_opv_desc =
133 	{ &spec_vnodeop_p, spec_vnodeop_entries };
134 
135 /*
136  * Trivial lookup routine that always fails.
137  */
138 int
139 spec_lookup(v)
140 	void *v;
141 {
142 	struct vop_lookup_args /* {
143 		struct vnode *a_dvp;
144 		struct vnode **a_vpp;
145 		struct componentname *a_cnp;
146 	} */ *ap = v;
147 
148 	*ap->a_vpp = NULL;
149 	return (ENOTDIR);
150 }
151 
152 /*
153  * Returns true if dev is /dev/mem or /dev/kmem.
154  */
155 static int
156 iskmemdev(dev_t dev)
157 {
158 	/* mem_no is emitted by config(8) to generated devsw.c */
159 	extern const int mem_no;
160 
161 	/* minor 14 is /dev/io on i386 with COMPAT_10 */
162 	return (major(dev) == mem_no && (minor(dev) < 2 || minor(dev) == 14));
163 }
164 
165 /*
166  * Open a special file.
167  */
168 /* ARGSUSED */
169 int
170 spec_open(v)
171 	void *v;
172 {
173 	struct vop_open_args /* {
174 		struct vnode *a_vp;
175 		int  a_mode;
176 		struct ucred *a_cred;
177 		struct proc *a_p;
178 	} */ *ap = v;
179 	struct proc *p = ap->a_p;
180 	struct vnode *bvp, *vp = ap->a_vp;
181 	const struct bdevsw *bdev;
182 	const struct cdevsw *cdev;
183 	dev_t blkdev, dev = (dev_t)vp->v_rdev;
184 	int error;
185 	struct partinfo pi;
186 
187 	/*
188 	 * Don't allow open if fs is mounted -nodev.
189 	 */
190 	if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
191 		return (ENXIO);
192 
193 	switch (vp->v_type) {
194 
195 	case VCHR:
196 		cdev = cdevsw_lookup(dev);
197 		if (cdev == NULL)
198 			return (ENXIO);
199 		if (ap->a_cred != FSCRED && (ap->a_mode & FWRITE)) {
200 			/*
201 			 * When running in very secure mode, do not allow
202 			 * opens for writing of any disk character devices.
203 			 */
204 			if (securelevel >= 2 && cdev->d_type == D_DISK)
205 				return (EPERM);
206 			/*
207 			 * When running in secure mode, do not allow opens
208 			 * for writing of /dev/mem, /dev/kmem, or character
209 			 * devices whose corresponding block devices are
210 			 * currently mounted.
211 			 */
212 			if (securelevel >= 1) {
213 				blkdev = devsw_chr2blk(dev);
214 				if (blkdev != (dev_t)NODEV &&
215 				    vfinddev(blkdev, VBLK, &bvp) &&
216 				    (error = vfs_mountedon(bvp)))
217 					return (error);
218 				if (iskmemdev(dev))
219 					return (EPERM);
220 			}
221 		}
222 		if (cdev->d_type == D_TTY)
223 			vp->v_flag |= VISTTY;
224 		VOP_UNLOCK(vp, 0);
225 		error = (*cdev->d_open)(dev, ap->a_mode, S_IFCHR, p);
226 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
227 		return (error);
228 
229 	case VBLK:
230 		bdev = bdevsw_lookup(dev);
231 		if (bdev == NULL)
232 			return (ENXIO);
233 		/*
234 		 * When running in very secure mode, do not allow
235 		 * opens for writing of any disk block devices.
236 		 */
237 		if (securelevel >= 2 && ap->a_cred != FSCRED &&
238 		    (ap->a_mode & FWRITE) && bdev->d_type == D_DISK)
239 			return (EPERM);
240 		/*
241 		 * Do not allow opens of block devices that are
242 		 * currently mounted.
243 		 */
244 		if ((error = vfs_mountedon(vp)) != 0)
245 			return (error);
246 		error = (*bdev->d_open)(dev, ap->a_mode, S_IFBLK, p);
247 		if (error) {
248 			return error;
249 		}
250 		error = (*bdev->d_ioctl)(vp->v_rdev,
251 		    DIOCGPART, (caddr_t)&pi, FREAD, curproc);
252 		if (error == 0) {
253 			vp->v_size = (voff_t)pi.disklab->d_secsize *
254 			    pi.part->p_size;
255 		}
256 		return 0;
257 
258 	case VNON:
259 	case VLNK:
260 	case VDIR:
261 	case VREG:
262 	case VBAD:
263 	case VFIFO:
264 	case VSOCK:
265 		break;
266 	}
267 	return (0);
268 }
269 
270 /*
271  * Vnode op for read
272  */
273 /* ARGSUSED */
274 int
275 spec_read(v)
276 	void *v;
277 {
278 	struct vop_read_args /* {
279 		struct vnode *a_vp;
280 		struct uio *a_uio;
281 		int  a_ioflag;
282 		struct ucred *a_cred;
283 	} */ *ap = v;
284 	struct vnode *vp = ap->a_vp;
285 	struct uio *uio = ap->a_uio;
286  	struct proc *p = uio->uio_procp;
287 	struct buf *bp;
288 	const struct bdevsw *bdev;
289 	const struct cdevsw *cdev;
290 	daddr_t bn;
291 	int bsize, bscale;
292 	struct partinfo dpart;
293 	int n, on;
294 	int error = 0;
295 
296 #ifdef DIAGNOSTIC
297 	if (uio->uio_rw != UIO_READ)
298 		panic("spec_read mode");
299 	if (uio->uio_segflg == UIO_USERSPACE && uio->uio_procp != curproc)
300 		panic("spec_read proc");
301 #endif
302 	if (uio->uio_resid == 0)
303 		return (0);
304 
305 	switch (vp->v_type) {
306 
307 	case VCHR:
308 		VOP_UNLOCK(vp, 0);
309 		cdev = cdevsw_lookup(vp->v_rdev);
310 		if (cdev != NULL)
311 			error = (*cdev->d_read)(vp->v_rdev, uio, ap->a_ioflag);
312 		else
313 			error = ENXIO;
314 		vn_lock(vp, LK_SHARED | LK_RETRY);
315 		return (error);
316 
317 	case VBLK:
318 		if (uio->uio_offset < 0)
319 			return (EINVAL);
320 		bsize = BLKDEV_IOSIZE;
321 		bdev = bdevsw_lookup(vp->v_rdev);
322 		if (bdev != NULL &&
323 		    (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart,
324 				     FREAD, p) == 0) {
325 			if (dpart.part->p_fstype == FS_BSDFFS &&
326 			    dpart.part->p_frag != 0 && dpart.part->p_fsize != 0)
327 				bsize = dpart.part->p_frag *
328 				    dpart.part->p_fsize;
329 		}
330 		bscale = bsize >> DEV_BSHIFT;
331 		do {
332 			bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1);
333 			on = uio->uio_offset % bsize;
334 			n = min((unsigned)(bsize - on), uio->uio_resid);
335 			error = bread(vp, bn, bsize, NOCRED, &bp);
336 			n = min(n, bsize - bp->b_resid);
337 			if (error) {
338 				brelse(bp);
339 				return (error);
340 			}
341 			error = uiomove((char *)bp->b_data + on, n, uio);
342 			brelse(bp);
343 		} while (error == 0 && uio->uio_resid > 0 && n != 0);
344 		return (error);
345 
346 	default:
347 		panic("spec_read type");
348 	}
349 	/* NOTREACHED */
350 }
351 
352 /*
353  * Vnode op for write
354  */
355 /* ARGSUSED */
356 int
357 spec_write(v)
358 	void *v;
359 {
360 	struct vop_write_args /* {
361 		struct vnode *a_vp;
362 		struct uio *a_uio;
363 		int  a_ioflag;
364 		struct ucred *a_cred;
365 	} */ *ap = v;
366 	struct vnode *vp = ap->a_vp;
367 	struct uio *uio = ap->a_uio;
368 	struct proc *p = uio->uio_procp;
369 	struct buf *bp;
370 	const struct bdevsw *bdev;
371 	const struct cdevsw *cdev;
372 	daddr_t bn;
373 	int bsize, bscale;
374 	struct partinfo dpart;
375 	int n, on;
376 	int error = 0;
377 
378 #ifdef DIAGNOSTIC
379 	if (uio->uio_rw != UIO_WRITE)
380 		panic("spec_write mode");
381 	if (uio->uio_segflg == UIO_USERSPACE && uio->uio_procp != curproc)
382 		panic("spec_write proc");
383 #endif
384 
385 	switch (vp->v_type) {
386 
387 	case VCHR:
388 		VOP_UNLOCK(vp, 0);
389 		cdev = cdevsw_lookup(vp->v_rdev);
390 		if (cdev != NULL)
391 			error = (*cdev->d_write)(vp->v_rdev, uio, ap->a_ioflag);
392 		else
393 			error = ENXIO;
394 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
395 		return (error);
396 
397 	case VBLK:
398 		if (uio->uio_resid == 0)
399 			return (0);
400 		if (uio->uio_offset < 0)
401 			return (EINVAL);
402 		bsize = BLKDEV_IOSIZE;
403 		bdev = bdevsw_lookup(vp->v_rdev);
404 		if (bdev != NULL &&
405 		    (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart,
406 				    FREAD, p) == 0) {
407 			if (dpart.part->p_fstype == FS_BSDFFS &&
408 			    dpart.part->p_frag != 0 && dpart.part->p_fsize != 0)
409 				bsize = dpart.part->p_frag *
410 				    dpart.part->p_fsize;
411 		}
412 		bscale = bsize >> DEV_BSHIFT;
413 		do {
414 			bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1);
415 			on = uio->uio_offset % bsize;
416 			n = min((unsigned)(bsize - on), uio->uio_resid);
417 			if (n == bsize)
418 				bp = getblk(vp, bn, bsize, 0, 0);
419 			else
420 				error = bread(vp, bn, bsize, NOCRED, &bp);
421 			if (error) {
422 				brelse(bp);
423 				return (error);
424 			}
425 			n = min(n, bsize - bp->b_resid);
426 			error = uiomove((char *)bp->b_data + on, n, uio);
427 			if (error)
428 				brelse(bp);
429 			else {
430 				if (n + on == bsize)
431 					bawrite(bp);
432 				else
433 					bdwrite(bp);
434 				if (bp->b_flags & B_ERROR)
435 					error = bp->b_error;
436 			}
437 		} while (error == 0 && uio->uio_resid > 0 && n != 0);
438 		return (error);
439 
440 	default:
441 		panic("spec_write type");
442 	}
443 	/* NOTREACHED */
444 }
445 
446 /*
447  * Device ioctl operation.
448  */
449 /* ARGSUSED */
450 int
451 spec_ioctl(v)
452 	void *v;
453 {
454 	struct vop_ioctl_args /* {
455 		struct vnode *a_vp;
456 		u_long a_command;
457 		caddr_t  a_data;
458 		int  a_fflag;
459 		struct ucred *a_cred;
460 		struct proc *a_p;
461 	} */ *ap = v;
462 	const struct bdevsw *bdev;
463 	const struct cdevsw *cdev;
464 	dev_t dev = ap->a_vp->v_rdev;
465 
466 	switch (ap->a_vp->v_type) {
467 
468 	case VCHR:
469 		cdev = cdevsw_lookup(dev);
470 		if (cdev == NULL)
471 			return (ENXIO);
472 		return ((*cdev->d_ioctl)(dev, ap->a_command, ap->a_data,
473 		    ap->a_fflag, ap->a_p));
474 
475 	case VBLK:
476 		bdev = bdevsw_lookup(dev);
477 		if (bdev == NULL)
478 			return (ENXIO);
479 		if (ap->a_command == 0 && (long)ap->a_data == B_TAPE) {
480 			if (bdev->d_type == D_TAPE)
481 				return (0);
482 			else
483 				return (1);
484 		}
485 		return ((*bdev->d_ioctl)(dev, ap->a_command, ap->a_data,
486 		   ap->a_fflag, ap->a_p));
487 
488 	default:
489 		panic("spec_ioctl");
490 		/* NOTREACHED */
491 	}
492 }
493 
494 /* ARGSUSED */
495 int
496 spec_poll(v)
497 	void *v;
498 {
499 	struct vop_poll_args /* {
500 		struct vnode *a_vp;
501 		int a_events;
502 		struct proc *a_p;
503 	} */ *ap = v;
504 	const struct cdevsw *cdev;
505 	dev_t dev;
506 
507 	switch (ap->a_vp->v_type) {
508 
509 	case VCHR:
510 		dev = ap->a_vp->v_rdev;
511 		cdev = cdevsw_lookup(dev);
512 		if (cdev == NULL)
513 			return (ENXIO);
514 		return (*cdev->d_poll)(dev, ap->a_events, ap->a_p);
515 
516 	default:
517 		return (genfs_poll(v));
518 	}
519 }
520 
521 /* ARGSUSED */
522 int
523 spec_kqfilter(v)
524 	void *v;
525 {
526 	struct vop_kqfilter_args /* {
527 		struct vnode	*a_vp;
528 		struct proc	*a_kn;
529 	} */ *ap = v;
530 	const struct cdevsw *cdev;
531 	dev_t dev;
532 
533 	switch (ap->a_vp->v_type) {
534 
535 	case VCHR:
536 		dev = ap->a_vp->v_rdev;
537 		cdev = cdevsw_lookup(dev);
538 		if (cdev == NULL)
539 			return (ENXIO);
540 		return (*cdev->d_kqfilter)(dev, ap->a_kn);
541 	default:
542 		/*
543 		 * Block devices don't support kqfilter, and refuse it
544 		 * for any other files (like those vflush()ed) too.
545 		 */
546 		return (EOPNOTSUPP);
547 	}
548 }
549 
550 /*
551  * Synch buffers associated with a block device
552  */
553 /* ARGSUSED */
554 int
555 spec_fsync(v)
556 	void *v;
557 {
558 	struct vop_fsync_args /* {
559 		struct vnode *a_vp;
560 		struct ucred *a_cred;
561 		int  a_flags;
562 		off_t offlo;
563 		off_t offhi;
564 		struct proc *a_p;
565 	} */ *ap = v;
566 	struct vnode *vp = ap->a_vp;
567 
568 	if (vp->v_type == VBLK)
569 		vflushbuf(vp, (ap->a_flags & FSYNC_WAIT) != 0);
570 	return (0);
571 }
572 
573 /*
574  * Just call the device strategy routine
575  */
576 int
577 spec_strategy(v)
578 	void *v;
579 {
580 	struct vop_strategy_args /* {
581 		struct buf *a_bp;
582 	} */ *ap = v;
583 	struct buf *bp;
584 	const struct bdevsw *bdev;
585 
586 	bp = ap->a_bp;
587 	if (!(bp->b_flags & B_READ) &&
588 	    (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
589 		(*bioops.io_start)(bp);
590 	bdev = bdevsw_lookup(bp->b_dev);
591 	if (bdev != NULL)
592 		(*bdev->d_strategy)(bp);
593 	return (0);
594 }
595 
596 int
597 spec_inactive(v)
598 	void *v;
599 {
600 	struct vop_inactive_args /* {
601 		struct vnode *a_vp;
602 		struct proc *a_p;
603 	} */ *ap = v;
604 
605 	VOP_UNLOCK(ap->a_vp, 0);
606 	return (0);
607 }
608 
609 /*
610  * This is a noop, simply returning what one has been given.
611  */
612 int
613 spec_bmap(v)
614 	void *v;
615 {
616 	struct vop_bmap_args /* {
617 		struct vnode *a_vp;
618 		daddr_t  a_bn;
619 		struct vnode **a_vpp;
620 		daddr_t *a_bnp;
621 		int *a_runp;
622 	} */ *ap = v;
623 
624 	if (ap->a_vpp != NULL)
625 		*ap->a_vpp = ap->a_vp;
626 	if (ap->a_bnp != NULL)
627 		*ap->a_bnp = ap->a_bn;
628 	if (ap->a_runp != NULL)
629 		*ap->a_runp = (MAXBSIZE >> DEV_BSHIFT) - 1;
630 	return (0);
631 }
632 
633 /*
634  * Device close routine
635  */
636 /* ARGSUSED */
637 int
638 spec_close(v)
639 	void *v;
640 {
641 	struct vop_close_args /* {
642 		struct vnode *a_vp;
643 		int  a_fflag;
644 		struct ucred *a_cred;
645 		struct proc *a_p;
646 	} */ *ap = v;
647 	struct vnode *vp = ap->a_vp;
648 	const struct bdevsw *bdev;
649 	const struct cdevsw *cdev;
650 	dev_t dev = vp->v_rdev;
651 	int (*devclose) __P((dev_t, int, int, struct proc *));
652 	int mode, error, count, flags, flags1;
653 
654 	count = vcount(vp);
655 	simple_lock(&vp->v_interlock);
656 	flags = vp->v_flag;
657 	simple_unlock(&vp->v_interlock);
658 
659 	switch (vp->v_type) {
660 
661 	case VCHR:
662 		/*
663 		 * Hack: a tty device that is a controlling terminal
664 		 * has a reference from the session structure.
665 		 * We cannot easily tell that a character device is
666 		 * a controlling terminal, unless it is the closing
667 		 * process' controlling terminal.  In that case,
668 		 * if the reference count is 2 (this last descriptor
669 		 * plus the session), release the reference from the session.
670 		 */
671 		if (count == 2 && ap->a_p &&
672 		    vp == ap->a_p->p_session->s_ttyvp) {
673 			vrele(vp);
674 			count--;
675 			ap->a_p->p_session->s_ttyvp = NULL;
676 		}
677 		/*
678 		 * If the vnode is locked, then we are in the midst
679 		 * of forcably closing the device, otherwise we only
680 		 * close on last reference.
681 		 */
682 		if (count > 1 && (flags & VXLOCK) == 0)
683 			return (0);
684 		cdev = cdevsw_lookup(dev);
685 		if (cdev != NULL)
686 			devclose = cdev->d_close;
687 		else
688 			devclose = NULL;
689 		mode = S_IFCHR;
690 		break;
691 
692 	case VBLK:
693 		/*
694 		 * On last close of a block device (that isn't mounted)
695 		 * we must invalidate any in core blocks, so that
696 		 * we can, for instance, change floppy disks.
697 		 */
698 		error = vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 0, 0);
699 		if (error)
700 			return (error);
701 		/*
702 		 * We do not want to really close the device if it
703 		 * is still in use unless we are trying to close it
704 		 * forcibly. Since every use (buffer, vnode, swap, cmap)
705 		 * holds a reference to the vnode, and because we mark
706 		 * any other vnodes that alias this device, when the
707 		 * sum of the reference counts on all the aliased
708 		 * vnodes descends to one, we are on last close.
709 		 */
710 		if (count > 1 && (flags & VXLOCK) == 0)
711 			return (0);
712 		bdev = bdevsw_lookup(dev);
713 		if (bdev != NULL)
714 			devclose = bdev->d_close;
715 		else
716 			devclose = NULL;
717 		mode = S_IFBLK;
718 		break;
719 
720 	default:
721 		panic("spec_close: not special");
722 	}
723 
724 	flags1 = ap->a_fflag;
725 
726 	/*
727 	 * if VXLOCK is set, then we're going away soon, so make this
728 	 * non-blocking. Also ensures that we won't wedge in vn_lock below.
729 	 */
730 	if (flags & VXLOCK)
731 		flags1 |= FNONBLOCK;
732 
733 	/*
734 	 * If we're able to block, release the vnode lock & reacquire. We
735 	 * might end up sleaping for someone else who wants our queues. They
736 	 * won't get them if we hold the vnode locked. Also, if VXLOCK is set,
737 	 * don't release the lock as we won't be able to regain it.
738 	 */
739 	if (!(flags1 & FNONBLOCK))
740 		VOP_UNLOCK(vp, 0);
741 
742 	if (devclose != NULL)
743 		error = (*devclose)(dev, flags1, mode, ap->a_p);
744 	else
745 		error = ENXIO;
746 
747 	if (!(flags1 & FNONBLOCK))
748 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
749 
750 	return (error);
751 }
752 
753 /*
754  * Print out the contents of a special device vnode.
755  */
756 int
757 spec_print(v)
758 	void *v;
759 {
760 	struct vop_print_args /* {
761 		struct vnode *a_vp;
762 	} */ *ap = v;
763 
764 	printf("tag VT_NON, dev %d, %d\n", major(ap->a_vp->v_rdev),
765 	    minor(ap->a_vp->v_rdev));
766 	return 0;
767 }
768 
769 /*
770  * Return POSIX pathconf information applicable to special devices.
771  */
772 int
773 spec_pathconf(v)
774 	void *v;
775 {
776 	struct vop_pathconf_args /* {
777 		struct vnode *a_vp;
778 		int a_name;
779 		register_t *a_retval;
780 	} */ *ap = v;
781 
782 	switch (ap->a_name) {
783 	case _PC_LINK_MAX:
784 		*ap->a_retval = LINK_MAX;
785 		return (0);
786 	case _PC_MAX_CANON:
787 		*ap->a_retval = MAX_CANON;
788 		return (0);
789 	case _PC_MAX_INPUT:
790 		*ap->a_retval = MAX_INPUT;
791 		return (0);
792 	case _PC_PIPE_BUF:
793 		*ap->a_retval = PIPE_BUF;
794 		return (0);
795 	case _PC_CHOWN_RESTRICTED:
796 		*ap->a_retval = 1;
797 		return (0);
798 	case _PC_VDISABLE:
799 		*ap->a_retval = _POSIX_VDISABLE;
800 		return (0);
801 	case _PC_SYNC_IO:
802 		*ap->a_retval = 1;
803 		return (0);
804 	default:
805 		return (EINVAL);
806 	}
807 	/* NOTREACHED */
808 }
809 
810 /*
811  * Advisory record locking support.
812  */
813 int
814 spec_advlock(v)
815 	void *v;
816 {
817 	struct vop_advlock_args /* {
818 		struct vnode *a_vp;
819 		caddr_t a_id;
820 		int a_op;
821 		struct flock *a_fl;
822 		int a_flags;
823 	} */ *ap = v;
824 	struct vnode *vp = ap->a_vp;
825 
826 	return lf_advlock(ap, &vp->v_speclockf, (off_t)0);
827 }
828