xref: /netbsd-src/sys/kern/subr_disk.c (revision 946379e7b37692fc43f68eb0d1c10daa0a7f3b6c)
1 /*	$NetBSD: subr_disk.c,v 1.116 2016/01/06 00:22:30 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 1996, 1997, 1999, 2000, 2009 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  * (c) UNIX System Laboratories, Inc.
37  * All or some portions of this file are derived from material licensed
38  * to the University of California by American Telephone and Telegraph
39  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
40  * the permission of UNIX System Laboratories, Inc.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ufs_disksubr.c	8.5 (Berkeley) 1/21/94
67  */
68 
69 #include <sys/cdefs.h>
70 __KERNEL_RCSID(0, "$NetBSD: subr_disk.c,v 1.116 2016/01/06 00:22:30 christos Exp $");
71 
72 #include <sys/param.h>
73 #include <sys/kernel.h>
74 #include <sys/kmem.h>
75 #include <sys/buf.h>
76 #include <sys/fcntl.h>
77 #include <sys/syslog.h>
78 #include <sys/disklabel.h>
79 #include <sys/disk.h>
80 #include <sys/sysctl.h>
81 #include <lib/libkern/libkern.h>
82 
83 /*
84  * Compute checksum for disk label.
85  */
86 u_int
87 dkcksum(struct disklabel *lp)
88 {
89 
90 	return dkcksum_sized(lp, lp->d_npartitions);
91 }
92 
93 u_int
94 dkcksum_sized(struct disklabel *lp, size_t npartitions)
95 {
96 	uint16_t *start, *end;
97 	uint16_t sum = 0;
98 
99 	start = (uint16_t *)lp;
100 	end = (uint16_t *)&lp->d_partitions[npartitions];
101 	while (start < end)
102 		sum ^= *start++;
103 	return sum;
104 }
105 
106 /*
107  * Disk error is the preface to plaintive error messages
108  * about failing disk transfers.  It prints messages of the form
109 
110 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
111 
112  * if the offset of the error in the transfer and a disk label
113  * are both available.  blkdone should be -1 if the position of the error
114  * is unknown; the disklabel pointer may be null from drivers that have not
115  * been converted to use them.  The message is printed with printf
116  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
117  * The message should be completed (with at least a newline) with printf
118  * or addlog, respectively.  There is no trailing space.
119  */
120 #ifndef PRIdaddr
121 #define PRIdaddr PRId64
122 #endif
123 void
124 diskerr(const struct buf *bp, const char *dname, const char *what, int pri,
125     int blkdone, const struct disklabel *lp)
126 {
127 	int unit = DISKUNIT(bp->b_dev), part = DISKPART(bp->b_dev);
128 	void (*pr)(const char *, ...) __printflike(1, 2);
129 	char partname = 'a' + part;
130 	daddr_t sn;
131 
132 	if (/*CONSTCOND*/0)
133 		/* Compiler will error this is the format is wrong... */
134 		printf("%" PRIdaddr, bp->b_blkno);
135 
136 	if (pri != LOG_PRINTF) {
137 		static const char fmt[] = "";
138 		log(pri, fmt);
139 		pr = addlog;
140 	} else
141 		pr = printf;
142 	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
143 	    bp->b_flags & B_READ ? "read" : "writ");
144 	sn = bp->b_blkno;
145 	if (bp->b_bcount <= DEV_BSIZE)
146 		(*pr)("%" PRIdaddr, sn);
147 	else {
148 		if (blkdone >= 0) {
149 			sn += blkdone;
150 			(*pr)("%" PRIdaddr " of ", sn);
151 		}
152 		(*pr)("%" PRIdaddr "-%" PRIdaddr "", bp->b_blkno,
153 		    bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
154 	}
155 	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
156 		sn += lp->d_partitions[part].p_offset;
157 		(*pr)(" (%s%d bn %" PRIdaddr "; cn %" PRIdaddr "",
158 		    dname, unit, sn, sn / lp->d_secpercyl);
159 		sn %= lp->d_secpercyl;
160 		(*pr)(" tn %" PRIdaddr " sn %" PRIdaddr ")",
161 		    sn / lp->d_nsectors, sn % lp->d_nsectors);
162 	}
163 }
164 
165 /*
166  * Searches the iostatlist for the disk corresponding to the
167  * name provided.
168  */
169 struct disk *
170 disk_find(const char *name)
171 {
172 	struct io_stats *stat;
173 
174 	stat = iostat_find(name);
175 
176 	if ((stat != NULL) && (stat->io_type == IOSTAT_DISK))
177 		return stat->io_parent;
178 
179 	return (NULL);
180 }
181 
182 void
183 disk_init(struct disk *diskp, const char *name, const struct dkdriver *driver)
184 {
185 	u_int blocksize = DEV_BSIZE;
186 
187 	/*
188 	 * Initialize the wedge-related locks and other fields.
189 	 */
190 	mutex_init(&diskp->dk_rawlock, MUTEX_DEFAULT, IPL_NONE);
191 	mutex_init(&diskp->dk_openlock, MUTEX_DEFAULT, IPL_NONE);
192 	LIST_INIT(&diskp->dk_wedges);
193 	diskp->dk_nwedges = 0;
194 	diskp->dk_labelsector = LABELSECTOR;
195 	diskp->dk_blkshift = DK_BSIZE2BLKSHIFT(blocksize);
196 	diskp->dk_byteshift = DK_BSIZE2BYTESHIFT(blocksize);
197 	diskp->dk_name = name;
198 	diskp->dk_driver = driver;
199 }
200 
201 /*
202  * Attach a disk.
203  */
204 void
205 disk_attach(struct disk *diskp)
206 {
207 
208 	/*
209 	 * Allocate and initialize the disklabel structures.
210 	 */
211 	diskp->dk_label = kmem_zalloc(sizeof(struct disklabel), KM_SLEEP);
212 	diskp->dk_cpulabel = kmem_zalloc(sizeof(struct cpu_disklabel),
213 	    KM_SLEEP);
214 	if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL))
215 		panic("disk_attach: can't allocate storage for disklabel");
216 
217 	/*
218 	 * Set up the stats collection.
219 	 */
220 	diskp->dk_stats = iostat_alloc(IOSTAT_DISK, diskp, diskp->dk_name);
221 }
222 
223 int
224 disk_begindetach(struct disk *dk, int (*lastclose)(device_t),
225     device_t self, int flags)
226 {
227 	int rc;
228 
229 	rc = 0;
230 	mutex_enter(&dk->dk_openlock);
231 	if (dk->dk_openmask == 0)
232 		;	/* nothing to do */
233 	else if ((flags & DETACH_FORCE) == 0)
234 		rc = EBUSY;
235 	else if (lastclose != NULL)
236 		rc = (*lastclose)(self);
237 	mutex_exit(&dk->dk_openlock);
238 
239 	return rc;
240 }
241 
242 /*
243  * Detach a disk.
244  */
245 void
246 disk_detach(struct disk *diskp)
247 {
248 
249 	/*
250 	 * Remove from the drivelist.
251 	 */
252 	iostat_free(diskp->dk_stats);
253 
254 	/*
255 	 * Release the disk-info dictionary.
256 	 */
257 	if (diskp->dk_info) {
258 		prop_object_release(diskp->dk_info);
259 		diskp->dk_info = NULL;
260 	}
261 
262 	/*
263 	 * Free the space used by the disklabel structures.
264 	 */
265 	kmem_free(diskp->dk_label, sizeof(*diskp->dk_label));
266 	kmem_free(diskp->dk_cpulabel, sizeof(*diskp->dk_cpulabel));
267 }
268 
269 void
270 disk_destroy(struct disk *diskp)
271 {
272 
273 	mutex_destroy(&diskp->dk_openlock);
274 	mutex_destroy(&diskp->dk_rawlock);
275 }
276 
277 /*
278  * Mark the disk as busy for metrics collection.
279  */
280 void
281 disk_busy(struct disk *diskp)
282 {
283 
284 	iostat_busy(diskp->dk_stats);
285 }
286 
287 /*
288  * Finished disk operations, gather metrics.
289  */
290 void
291 disk_unbusy(struct disk *diskp, long bcount, int read)
292 {
293 
294 	iostat_unbusy(diskp->dk_stats, bcount, read);
295 }
296 
297 /*
298  * Return true if disk has an I/O operation in flight.
299  */
300 bool
301 disk_isbusy(struct disk *diskp)
302 {
303 
304 	return iostat_isbusy(diskp->dk_stats);
305 }
306 
307 /*
308  * Bounds checking against the media size, used for the raw partition.
309  * secsize, mediasize and b_blkno must all be the same units.
310  * Possibly this has to be DEV_BSIZE (512).
311  */
312 int
313 bounds_check_with_mediasize(struct buf *bp, int secsize, uint64_t mediasize)
314 {
315 	int64_t sz;
316 
317 	if (bp->b_blkno < 0) {
318 		/* Reject negative offsets immediately. */
319 		bp->b_error = EINVAL;
320 		return 0;
321 	}
322 
323 	sz = howmany((int64_t)bp->b_bcount, secsize);
324 
325 	/*
326 	 * bp->b_bcount is a 32-bit value, and we rejected a negative
327 	 * bp->b_blkno already, so "bp->b_blkno + sz" cannot overflow.
328 	 */
329 
330 	if (bp->b_blkno + sz > mediasize) {
331 		sz = mediasize - bp->b_blkno;
332 		if (sz == 0) {
333 			/* If exactly at end of disk, return EOF. */
334 			bp->b_resid = bp->b_bcount;
335 			return 0;
336 		}
337 		if (sz < 0) {
338 			/* If past end of disk, return EINVAL. */
339 			bp->b_error = EINVAL;
340 			return 0;
341 		}
342 		/* Otherwise, truncate request. */
343 		bp->b_bcount = sz * secsize;
344 	}
345 
346 	return 1;
347 }
348 
349 /*
350  * Determine the size of the transfer, and make sure it is
351  * within the boundaries of the partition. Adjust transfer
352  * if needed, and signal errors or early completion.
353  */
354 int
355 bounds_check_with_label(struct disk *dk, struct buf *bp, int wlabel)
356 {
357 	struct disklabel *lp = dk->dk_label;
358 	struct partition *p = lp->d_partitions + DISKPART(bp->b_dev);
359 	uint64_t p_size, p_offset, labelsector;
360 	int64_t sz;
361 
362 	if (bp->b_blkno < 0) {
363 		/* Reject negative offsets immediately. */
364 		bp->b_error = EINVAL;
365 		return -1;
366 	}
367 
368 	/* Protect against division by zero. XXX: Should never happen?!?! */
369 	if (lp->d_secpercyl == 0) {
370 		bp->b_error = EINVAL;
371 		return -1;
372 	}
373 
374 	p_size = (uint64_t)p->p_size << dk->dk_blkshift;
375 	p_offset = (uint64_t)p->p_offset << dk->dk_blkshift;
376 #if RAW_PART == 3
377 	labelsector = lp->d_partitions[2].p_offset;
378 #else
379 	labelsector = lp->d_partitions[RAW_PART].p_offset;
380 #endif
381 	labelsector = (labelsector + dk->dk_labelsector) << dk->dk_blkshift;
382 
383 	sz = howmany((int64_t)bp->b_bcount, DEV_BSIZE);
384 
385 	/*
386 	 * bp->b_bcount is a 32-bit value, and we rejected a negative
387 	 * bp->b_blkno already, so "bp->b_blkno + sz" cannot overflow.
388 	 */
389 
390 	if (bp->b_blkno + sz > p_size) {
391 		sz = p_size - bp->b_blkno;
392 		if (sz == 0) {
393 			/* If exactly at end of disk, return EOF. */
394 			bp->b_resid = bp->b_bcount;
395 			return 0;
396 		}
397 		if (sz < 0) {
398 			/* If past end of disk, return EINVAL. */
399 			bp->b_error = EINVAL;
400 			return -1;
401 		}
402 		/* Otherwise, truncate request. */
403 		bp->b_bcount = sz << DEV_BSHIFT;
404 	}
405 
406 	/* Overwriting disk label? */
407 	if (bp->b_blkno + p_offset <= labelsector &&
408 	    bp->b_blkno + p_offset + sz > labelsector &&
409 	    (bp->b_flags & B_READ) == 0 && !wlabel) {
410 		bp->b_error = EROFS;
411 		return -1;
412 	}
413 
414 	/* calculate cylinder for disksort to order transfers with */
415 	bp->b_cylinder = (bp->b_blkno + p->p_offset) /
416 	    (lp->d_secsize / DEV_BSIZE) / lp->d_secpercyl;
417 	return 1;
418 }
419 
420 int
421 disk_read_sectors(void (*strat)(struct buf *), const struct disklabel *lp,
422     struct buf *bp, unsigned int sector, int count)
423 {
424 	bp->b_blkno = btodb((off_t)sector * lp->d_secsize);
425 	bp->b_bcount = count * lp->d_secsize;
426 	bp->b_flags = (bp->b_flags & ~B_WRITE) | B_READ;
427 	bp->b_oflags &= ~BO_DONE;
428 	bp->b_cylinder = sector / lp->d_secpercyl;
429 	(*strat)(bp);
430 	return biowait(bp);
431 }
432 
433 const char *
434 convertdisklabel(struct disklabel *lp, void (*strat)(struct buf *),
435     struct buf *bp, uint32_t secperunit)
436 {
437 	struct partition rp, *altp, *p;
438 	int geom_ok;
439 	const char *str;
440 
441 	memset(&rp, 0, sizeof(rp));
442 	rp.p_size = secperunit;
443 	rp.p_fstype = FS_UNUSED;
444 
445 	/* If we can seek to d_secperunit - 1, believe the disk geometry. */
446 	if (secperunit != 0 &&
447 	    disk_read_sectors(strat, lp, bp, secperunit - 1, 1) == 0)
448 		geom_ok = 1;
449 	else
450 		geom_ok = 0;
451 
452 #if 0
453 	printf("%s: secperunit (%" PRIu32 ") %s\n", __func__,
454 	    secperunit, geom_ok ? "ok" : "not ok");
455 #endif
456 
457 	p = &lp->d_partitions[RAW_PART];
458 	if (RAW_PART == 'c' - 'a')
459 		altp = &lp->d_partitions['d' - 'a'];
460 	else
461 		altp = &lp->d_partitions['c' - 'a'];
462 
463 	if (lp->d_npartitions > RAW_PART && p->p_offset == 0 && p->p_size != 0)
464 		return NULL;	/* already a raw partition */
465 	else if (lp->d_npartitions > MAX('c', 'd') - 'a' &&
466 		 altp->p_offset == 0 && altp->p_size != 0) {
467 		/* alternate partition ('c' or 'd') is suitable for raw slot,
468 		 * swap with 'd' or 'c'.
469 		 */
470 		rp = *p;
471 		*p = *altp;
472 		*altp = rp;
473 		return NULL;
474 	} else if (lp->d_npartitions <= RAW_PART &&
475 	           lp->d_npartitions > 'c' - 'a') {
476 		/* No raw partition is present, but the alternate is present.
477 		 * Copy alternate to raw partition.
478 		 */
479 		lp->d_npartitions = RAW_PART + 1;
480 		*p = *altp;
481 		return NULL;
482 	} else if (!geom_ok)
483 		str = "no raw partition and disk reports bad geometry";
484 	else if (lp->d_npartitions <= RAW_PART) {
485 		memset(&lp->d_partitions[lp->d_npartitions], 0,
486 		    sizeof(struct partition) * (RAW_PART - lp->d_npartitions));
487 		*p = rp;
488 		lp->d_npartitions = RAW_PART + 1;
489 		return NULL;
490 	} else if (lp->d_npartitions < MAXPARTITIONS) {
491 		memmove(p + 1, p,
492 		    sizeof(struct partition) * (lp->d_npartitions - RAW_PART));
493 		*p = rp;
494 		lp->d_npartitions++;
495 		return NULL;
496 	} else
497 		str = "no raw partition and partition table is full";
498 #ifdef DIAGNOSTIC
499 	printf("Bad partition: %s\n", str);
500 	printf("type = %u, subtype = %u, typename = %s\n",
501 	    lp->d_type, lp->d_subtype, lp->d_typename);
502 	printf("secsize = %u, nsectors = %u, ntracks = %u\n",
503 	    lp->d_secsize, lp->d_nsectors, lp->d_ntracks);
504 	printf("ncylinders = %u, secpercyl = %u, secperunit = %u\n",
505 	    lp->d_ncylinders, lp->d_secpercyl, lp->d_secperunit);
506 	printf("npartitions = %u\n", lp->d_npartitions);
507 
508 	for (size_t i = 0; i < MIN(lp->d_npartitions, MAXPARTITIONS); i++) {
509 		p = &lp->d_partitions[i];
510 		printf("\t%c: offset = %u size = %u fstype = %u\n",
511 		    (char)(i + 'a'), p->p_offset, p->p_size, p->p_fstype);
512 	}
513 #endif
514 	return str;
515 }
516 
517 /*
518  * disk_ioctl --
519  *	Generic disk ioctl handling.
520  */
521 int
522 disk_ioctl(struct disk *dk, dev_t dev, u_long cmd, void *data, int flag,
523     struct lwp *l)
524 {
525 	struct dkwedge_info *dkw;
526 	struct partinfo *pi;
527 	struct partition *dp;
528 #ifdef __HAVE_OLD_DISKLABEL
529 	struct disklabel newlabel;
530 #endif
531 
532 	switch (cmd) {
533 	case DIOCGDISKINFO:
534 		if (dk->dk_info == NULL)
535 			return ENOTSUP;
536 		return prop_dictionary_copyout_ioctl(data, cmd, dk->dk_info);
537 
538 	case DIOCGSECTORSIZE:
539 		*(u_int *)data = dk->dk_geom.dg_secsize;
540 		return 0;
541 
542 	case DIOCGMEDIASIZE:
543 		*(off_t *)data = (off_t)dk->dk_geom.dg_secsize *
544 		    dk->dk_geom.dg_secperunit;
545 		return 0;
546 	default:
547 		break;
548 	}
549 
550 	if (dev == NODEV)
551 		return EPASSTHROUGH;
552 
553 	/* The following should be moved to dk_ioctl */
554 	switch (cmd) {
555 	case DIOCGDINFO:
556 		if (dk->dk_label == NULL)
557 			return EBUSY;
558 		memcpy(data, dk->dk_label, sizeof (*dk->dk_label));
559 		return 0;
560 
561 #ifdef __HAVE_OLD_DISKLABEL
562 	case ODIOCGDINFO:
563 		if (dk->dk_label == NULL)
564 			return EBUSY;
565 		memcpy(&newlabel, dk->dk_label, sizeof(newlabel));
566 		if (newlabel.d_npartitions > OLDMAXPARTITIONS)
567 			return ENOTTY;
568 		memcpy(data, &newlabel, sizeof(struct olddisklabel));
569 		return 0;
570 #endif
571 
572 	case DIOCGPARTINFO:
573 		pi = data;
574 		memset(pi, 0, sizeof(*pi));
575 		pi->pi_secsize = dk->dk_geom.dg_secsize;
576 		pi->pi_bsize = BLKDEV_IOSIZE;
577 
578 		if (DISKPART(dev) == RAW_PART) {
579 			pi->pi_size = dk->dk_geom.dg_secperunit;
580 			return 0;
581 		}
582 
583 		if (dk->dk_label == NULL)
584 			return EBUSY;
585 
586 		dp = &dk->dk_label->d_partitions[DISKPART(dev)];
587 		pi->pi_offset = dp->p_offset;
588 		pi->pi_size = dp->p_size;
589 
590 		pi->pi_fstype = dp->p_fstype;
591 		pi->pi_frag = dp->p_frag;
592 		pi->pi_fsize = dp->p_fsize;
593 		pi->pi_cpg = dp->p_cpg;
594 
595 		/*
596 		 * dholland 20130616: XXX this logic should not be
597 		 * here. It is here because the old buffer cache
598 		 * demands that all accesses to the same blocks need
599 		 * to be the same size; but it only works for FFS and
600 		 * nowadays I think it'll fail silently if the size
601 		 * info in the disklabel is wrong. (Or missing.) The
602 		 * buffer cache needs to be smarter; or failing that
603 		 * we need a reliable way here to get the right block
604 		 * size; or a reliable way to guarantee that (a) the
605 		 * fs is not mounted when we get here and (b) any
606 		 * buffers generated here will get purged when the fs
607 		 * does get mounted.
608 		 */
609 		if (dp->p_fstype == FS_BSDFFS &&
610 		    dp->p_frag != 0 && dp->p_fsize != 0)
611 			pi->pi_bsize = dp->p_frag * dp->p_fsize;
612 		return 0;
613 
614 	case DIOCAWEDGE:
615 		if ((flag & FWRITE) == 0)
616 			return EBADF;
617 
618 		dkw = data;
619 		strlcpy(dkw->dkw_parent, dk->dk_name, sizeof(dkw->dkw_parent));
620 		return dkwedge_add(dkw);
621 
622 	case DIOCDWEDGE:
623 		if ((flag & FWRITE) == 0)
624 			return EBADF;
625 
626 		dkw = data;
627 		strlcpy(dkw->dkw_parent, dk->dk_name, sizeof(dkw->dkw_parent));
628 		return dkwedge_del(dkw);
629 
630 	case DIOCLWEDGES:
631 		return dkwedge_list(dk, data, l);
632 
633 	case DIOCMWEDGES:
634 		if ((flag & FWRITE) == 0)
635 			return EBADF;
636 
637 		dkwedge_discover(dk);
638 		return 0;
639 
640 	default:
641 		return EPASSTHROUGH;
642 	}
643 }
644 
645 void
646 disk_set_info(device_t dev, struct disk *dk, const char *type)
647 {
648 	struct disk_geom *dg = &dk->dk_geom;
649 
650 	if (dg->dg_secsize == 0) {
651 #ifdef DIAGNOSTIC
652 		printf("%s: fixing 0 sector size\n", dk->dk_name);
653 #endif
654 		dg->dg_secsize = DEV_BSIZE;
655 	}
656 
657 	dk->dk_blkshift = DK_BSIZE2BLKSHIFT(dg->dg_secsize);
658 	dk->dk_byteshift = DK_BSIZE2BYTESHIFT(dg->dg_secsize);
659 
660 	if (dg->dg_secperunit == 0 && dg->dg_ncylinders == 0) {
661 #ifdef DIAGNOSTIC
662 		printf("%s: secperunit and ncylinders are zero\n", dk->dk_name);
663 #endif
664 		return;
665 	}
666 
667 	if (dg->dg_secperunit == 0) {
668 		if (dg->dg_nsectors == 0 || dg->dg_ntracks == 0) {
669 #ifdef DIAGNOSTIC
670 			printf("%s: secperunit and (sectors or tracks) "
671 			    "are zero\n", dk->dk_name);
672 #endif
673 			return;
674 		}
675 		dg->dg_secperunit = (int64_t) dg->dg_nsectors *
676 		    dg->dg_ntracks * dg->dg_ncylinders;
677 	}
678 
679 	if (dg->dg_ncylinders == 0) {
680 		if (dg->dg_ntracks && dg->dg_nsectors)
681 			dg->dg_ncylinders = dg->dg_secperunit /
682 			    (dg->dg_ntracks * dg->dg_nsectors);
683 	}
684 
685 	prop_dictionary_t disk_info, odisk_info, geom;
686 
687 	disk_info = prop_dictionary_create();
688 	geom = prop_dictionary_create();
689 
690 	prop_dictionary_set_uint64(geom, "sectors-per-unit",
691 	    dg->dg_secperunit);
692 
693 	prop_dictionary_set_uint32(geom, "sector-size", dg->dg_secsize);
694 
695 	if (dg->dg_nsectors)
696 		prop_dictionary_set_uint16(geom, "sectors-per-track",
697 		    dg->dg_nsectors);
698 
699 	if (dg->dg_ntracks)
700 		prop_dictionary_set_uint16(geom, "tracks-per-cylinder",
701 		    dg->dg_ntracks);
702 
703 	if (dg->dg_ncylinders)
704 		prop_dictionary_set_uint64(geom, "cylinders-per-unit",
705 		    dg->dg_ncylinders);
706 
707 	prop_dictionary_set(disk_info, "geometry", geom);
708 
709 	if (type)
710 		prop_dictionary_set_cstring_nocopy(disk_info, "type", type);
711 
712 	prop_object_release(geom);
713 
714 	odisk_info = dk->dk_info;
715 	dk->dk_info = disk_info;
716 
717 	if (dev)
718 		prop_dictionary_set(device_properties(dev), "disk-info",
719 		    disk_info);
720 
721 	/*
722 	 * Don't release disk_info here; we keep a reference to it.
723 	 * disk_detach() will release it when we go away.
724 	 */
725 	if (odisk_info)
726 		prop_object_release(odisk_info);
727 }
728