xref: /csrg-svn/sys/miscfs/specfs/spec_vnops.c (revision 40652)
1 /*
2  * Copyright (c) 1989 The Regents of the University of California.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms are permitted
6  * provided that the above copyright notice and this paragraph are
7  * duplicated in all such forms and that any documentation,
8  * advertising materials, and other materials related to such
9  * distribution and use acknowledge that the software was developed
10  * by the University of California, Berkeley.  The name of the
11  * University may not be used to endorse or promote products derived
12  * from this software without specific prior written permission.
13  * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
14  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
15  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
16  *
17  *	@(#)spec_vnops.c	7.24 (Berkeley) 03/27/90
18  */
19 
20 #include "param.h"
21 #include "systm.h"
22 #include "user.h"
23 #include "kernel.h"
24 #include "conf.h"
25 #include "buf.h"
26 #include "mount.h"
27 #include "vnode.h"
28 #include "specdev.h"
29 #include "stat.h"
30 #include "errno.h"
31 #include "ioctl.h"
32 #include "file.h"
33 #include "disklabel.h"
34 
35 int	spec_lookup(),
36 	spec_open(),
37 	spec_read(),
38 	spec_write(),
39 	spec_strategy(),
40 	spec_bmap(),
41 	spec_ioctl(),
42 	spec_select(),
43 	spec_lock(),
44 	spec_unlock(),
45 	spec_close(),
46 	spec_print(),
47 	spec_ebadf(),
48 	spec_badop(),
49 	spec_nullop();
50 
51 struct vnodeops spec_vnodeops = {
52 	spec_lookup,		/* lookup */
53 	spec_badop,		/* create */
54 	spec_badop,		/* mknod */
55 	spec_open,		/* open */
56 	spec_close,		/* close */
57 	spec_ebadf,		/* access */
58 	spec_ebadf,		/* getattr */
59 	spec_ebadf,		/* setattr */
60 	spec_read,		/* read */
61 	spec_write,		/* write */
62 	spec_ioctl,		/* ioctl */
63 	spec_select,		/* select */
64 	spec_badop,		/* mmap */
65 	spec_nullop,		/* fsync */
66 	spec_badop,		/* seek */
67 	spec_badop,		/* remove */
68 	spec_badop,		/* link */
69 	spec_badop,		/* rename */
70 	spec_badop,		/* mkdir */
71 	spec_badop,		/* rmdir */
72 	spec_badop,		/* symlink */
73 	spec_badop,		/* readdir */
74 	spec_badop,		/* readlink */
75 	spec_badop,		/* abortop */
76 	spec_nullop,		/* inactive */
77 	spec_nullop,		/* reclaim */
78 	spec_lock,		/* lock */
79 	spec_unlock,		/* unlock */
80 	spec_bmap,		/* bmap */
81 	spec_strategy,		/* strategy */
82 	spec_print,		/* print */
83 	spec_nullop,		/* islocked */
84 };
85 
86 /*
87  * Trivial lookup routine that always fails.
88  */
89 spec_lookup(vp, ndp)
90 	struct vnode *vp;
91 	struct nameidata *ndp;
92 {
93 
94 	ndp->ni_dvp = vp;
95 	ndp->ni_vp = NULL;
96 	return (ENOTDIR);
97 }
98 
99 /*
100  * Open called to allow handler
101  * of special files to initialize and
102  * validate before actual IO.
103  */
104 /* ARGSUSED */
105 spec_open(vp, mode, cred)
106 	register struct vnode *vp;
107 	int mode;
108 	struct ucred *cred;
109 {
110 	dev_t dev = (dev_t)vp->v_rdev;
111 	register int maj = major(dev);
112 	int error;
113 
114 	if (vp->v_mount && (vp->v_mount->m_flag & M_NODEV))
115 		return (ENXIO);
116 
117 	switch (vp->v_type) {
118 
119 	case VCHR:
120 		if ((u_int)maj >= nchrdev)
121 			return (ENXIO);
122 		return ((*cdevsw[maj].d_open)(dev, mode, S_IFCHR));
123 
124 	case VBLK:
125 		if ((u_int)maj >= nblkdev)
126 			return (ENXIO);
127 		if (error = mountedon(vp))
128 			return (error);
129 		return ((*bdevsw[maj].d_open)(dev, mode, S_IFBLK));
130 	}
131 	return (0);
132 }
133 
134 /*
135  * Vnode op for read
136  */
137 spec_read(vp, uio, ioflag, cred)
138 	register struct vnode *vp;
139 	register struct uio *uio;
140 	int ioflag;
141 	struct ucred *cred;
142 {
143 	struct buf *bp;
144 	daddr_t bn;
145 	long bsize, bscale;
146 	struct partinfo dpart;
147 	register int n, on;
148 	int error = 0;
149 	extern int mem_no;
150 
151 	if (uio->uio_rw != UIO_READ)
152 		panic("spec_read mode");
153 	if (uio->uio_resid == 0)
154 		return (0);
155 
156 	switch (vp->v_type) {
157 
158 	case VCHR:
159 		/*
160 		 * Negative offsets allowed only for /dev/kmem
161 		 */
162 		if (uio->uio_offset < 0 && major(vp->v_rdev) != mem_no)
163 			return (EINVAL);
164 		VOP_UNLOCK(vp);
165 		error = (*cdevsw[major(vp->v_rdev)].d_read)
166 			(vp->v_rdev, uio, ioflag);
167 		VOP_LOCK(vp);
168 		return (error);
169 
170 	case VBLK:
171 		if (uio->uio_offset < 0)
172 			return (EINVAL);
173 		bsize = BLKDEV_IOSIZE;
174 		if ((*bdevsw[major(vp->v_rdev)].d_ioctl)(vp->v_rdev, DIOCGPART,
175 		    (caddr_t)&dpart, FREAD) == 0) {
176 			if (dpart.part->p_fstype == FS_BSDFFS &&
177 			    dpart.part->p_frag != 0 && dpart.part->p_fsize != 0)
178 				bsize = dpart.part->p_frag *
179 				    dpart.part->p_fsize;
180 		}
181 		bscale = bsize / DEV_BSIZE;
182 		do {
183 			bn = (uio->uio_offset / DEV_BSIZE) &~ (bscale - 1);
184 			on = uio->uio_offset % bsize;
185 			n = MIN((unsigned)(bsize - on), uio->uio_resid);
186 			if (vp->v_lastr + bscale == bn)
187 				error = breada(vp, bn, (int)bsize, bn + bscale,
188 					(int)bsize, NOCRED, &bp);
189 			else
190 				error = bread(vp, bn, (int)bsize, NOCRED, &bp);
191 			vp->v_lastr = bn;
192 			n = MIN(n, bsize - bp->b_resid);
193 			if (error) {
194 				brelse(bp);
195 				return (error);
196 			}
197 			error = uiomove(bp->b_un.b_addr + on, n, uio);
198 			if (n + on == bsize)
199 				bp->b_flags |= B_AGE;
200 			brelse(bp);
201 		} while (error == 0 && uio->uio_resid > 0 && n != 0);
202 		return (error);
203 
204 	default:
205 		panic("spec_read type");
206 	}
207 	/* NOTREACHED */
208 }
209 
210 /*
211  * Vnode op for write
212  */
213 spec_write(vp, uio, ioflag, cred)
214 	register struct vnode *vp;
215 	register struct uio *uio;
216 	int ioflag;
217 	struct ucred *cred;
218 {
219 	struct buf *bp;
220 	daddr_t bn;
221 	int bsize, blkmask;
222 	struct partinfo dpart;
223 	register int n, on, i;
224 	int count, error = 0;
225 	extern int mem_no;
226 
227 	if (uio->uio_rw != UIO_WRITE)
228 		panic("spec_write mode");
229 
230 	switch (vp->v_type) {
231 
232 	case VCHR:
233 		/*
234 		 * Negative offsets allowed only for /dev/kmem
235 		 */
236 		if (uio->uio_offset < 0 && major(vp->v_rdev) != mem_no)
237 			return (EINVAL);
238 		VOP_UNLOCK(vp);
239 		error = (*cdevsw[major(vp->v_rdev)].d_write)
240 			(vp->v_rdev, uio, ioflag);
241 		VOP_LOCK(vp);
242 		return (error);
243 
244 	case VBLK:
245 		if (uio->uio_resid == 0)
246 			return (0);
247 		if (uio->uio_offset < 0)
248 			return (EINVAL);
249 		bsize = BLKDEV_IOSIZE;
250 		if ((*bdevsw[major(vp->v_rdev)].d_ioctl)(vp->v_rdev, DIOCGPART,
251 		    (caddr_t)&dpart, FREAD) == 0) {
252 			if (dpart.part->p_fstype == FS_BSDFFS &&
253 			    dpart.part->p_frag != 0 && dpart.part->p_fsize != 0)
254 				bsize = dpart.part->p_frag *
255 				    dpart.part->p_fsize;
256 		}
257 		blkmask = (bsize / DEV_BSIZE) - 1;
258 		do {
259 			bn = (uio->uio_offset / DEV_BSIZE) &~ blkmask;
260 			on = uio->uio_offset % bsize;
261 			n = MIN((unsigned)(bsize - on), uio->uio_resid);
262 			count = howmany(bsize, CLBYTES);
263 			for (i = 0; i < count; i++)
264 				munhash(vp, bn + i * (CLBYTES / DEV_BSIZE));
265 			if (n == bsize)
266 				bp = getblk(vp, bn, bsize);
267 			else
268 				error = bread(vp, bn, bsize, NOCRED, &bp);
269 			n = MIN(n, bsize - bp->b_resid);
270 			if (error) {
271 				brelse(bp);
272 				return (error);
273 			}
274 			error = uiomove(bp->b_un.b_addr + on, n, uio);
275 			if (n + on == bsize) {
276 				bp->b_flags |= B_AGE;
277 				bawrite(bp);
278 			} else
279 				bdwrite(bp);
280 		} while (error == 0 && uio->uio_resid > 0 && n != 0);
281 		return (error);
282 
283 	default:
284 		panic("spec_write type");
285 	}
286 	/* NOTREACHED */
287 }
288 
289 /*
290  * Device ioctl operation.
291  */
292 /* ARGSUSED */
293 spec_ioctl(vp, com, data, fflag, cred)
294 	struct vnode *vp;
295 	int com;
296 	caddr_t data;
297 	int fflag;
298 	struct ucred *cred;
299 {
300 	dev_t dev = vp->v_rdev;
301 
302 	switch (vp->v_type) {
303 
304 	case VCHR:
305 		return ((*cdevsw[major(dev)].d_ioctl)(dev, com, data, fflag));
306 
307 	case VBLK:
308 		if (com == 0 && (int)data == B_TAPE)
309 			if (bdevsw[major(dev)].d_flags & B_TAPE)
310 				return (0);
311 			else
312 				return (1);
313 		return ((*bdevsw[major(dev)].d_ioctl)(dev, com, data, fflag));
314 
315 	default:
316 		panic("spec_ioctl");
317 		/* NOTREACHED */
318 	}
319 }
320 
321 /* ARGSUSED */
322 spec_select(vp, which, fflags, cred)
323 	struct vnode *vp;
324 	int which, fflags;
325 	struct ucred *cred;
326 {
327 	register dev_t dev;
328 
329 	switch (vp->v_type) {
330 
331 	default:
332 		return (1);		/* XXX */
333 
334 	case VCHR:
335 		dev = vp->v_rdev;
336 		return (*cdevsw[major(dev)].d_select)(dev, which);
337 	}
338 }
339 
340 /*
341  * Just call the device strategy routine
342  */
343 spec_strategy(bp)
344 	register struct buf *bp;
345 {
346 
347 	(*bdevsw[major(bp->b_dev)].d_strategy)(bp);
348 	return (0);
349 }
350 
351 /*
352  * This is a noop, simply returning what one has been given.
353  */
354 spec_bmap(vp, bn, vpp, bnp)
355 	struct vnode *vp;
356 	daddr_t bn;
357 	struct vnode **vpp;
358 	daddr_t *bnp;
359 {
360 
361 	if (vpp != NULL)
362 		*vpp = vp;
363 	if (bnp != NULL)
364 		*bnp = bn;
365 	return (0);
366 }
367 
368 /*
369  * At the moment we do not do any locking.
370  */
371 /* ARGSUSED */
372 spec_lock(vp)
373 	struct vnode *vp;
374 {
375 
376 	return (0);
377 }
378 
379 /* ARGSUSED */
380 spec_unlock(vp)
381 	struct vnode *vp;
382 {
383 
384 	return (0);
385 }
386 
387 /*
388  * Device close routine
389  */
390 /* ARGSUSED */
391 spec_close(vp, flag, cred)
392 	register struct vnode *vp;
393 	int flag;
394 	struct ucred *cred;
395 {
396 	dev_t dev = vp->v_rdev;
397 	int (*cfunc)();
398 	int error, mode;
399 
400 	switch (vp->v_type) {
401 
402 	case VCHR:
403 		/*
404 		 * If the vnode is locked, then we are in the midst
405 		 * of forcably closing the device, otherwise we only
406 		 * close on last reference.
407 		 */
408 		if (vcount(vp) > 1 && (vp->v_flag & VXLOCK) == 0)
409 			return (0);
410 		cfunc = cdevsw[major(dev)].d_close;
411 		mode = S_IFCHR;
412 		break;
413 
414 	case VBLK:
415 		/*
416 		 * On last close of a block device (that isn't mounted)
417 		 * we must invalidate any in core blocks, so that
418 		 * we can, for instance, change floppy disks.
419 		 */
420 		vflushbuf(vp, 0);
421 		if (vinvalbuf(vp, 1))
422 			return (0);
423 		/*
424 		 * We do not want to really close the device if it
425 		 * is still in use unless we are trying to close it
426 		 * forcibly. Since every use (buffer, vnode, swap, cmap)
427 		 * holds a reference to the vnode, and because we mark
428 		 * any other vnodes that alias this device, when the
429 		 * sum of the reference counts on all the aliased
430 		 * vnodes descends to one, we are on last close.
431 		 */
432 		if (vcount(vp) > 1 && (vp->v_flag & VXLOCK) == 0)
433 			return (0);
434 		cfunc = bdevsw[major(dev)].d_close;
435 		mode = S_IFBLK;
436 		break;
437 
438 	default:
439 		panic("spec_close: not special");
440 	}
441 
442 	if (setjmp(&u.u_qsave)) {
443 		/*
444 		 * If device close routine is interrupted,
445 		 * must return so closef can clean up.
446 		 */
447 		error = EINTR;
448 	} else
449 		error = (*cfunc)(dev, flag, mode);
450 	return (error);
451 }
452 
453 /*
454  * Print out the contents of a special device vnode.
455  */
456 spec_print(vp)
457 	struct vnode *vp;
458 {
459 
460 	printf("tag VT_NON, dev %d, %d\n", major(vp->v_rdev),
461 		minor(vp->v_rdev));
462 }
463 
464 /*
465  * Special device failed operation
466  */
467 spec_ebadf()
468 {
469 
470 	return (EBADF);
471 }
472 
473 /*
474  * Special device bad operation
475  */
476 spec_badop()
477 {
478 
479 	panic("spec_badop called");
480 	/* NOTREACHED */
481 }
482 
483 /*
484  * Special device null operation
485  */
486 spec_nullop()
487 {
488 
489 	return (0);
490 }
491