xref: /netbsd-src/sys/kern/subr_disk.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: subr_disk.c,v 1.89 2007/10/08 16:41:15 ad Exp $	*/
2 
3 /*-
4  * Copyright (c) 1996, 1997, 1999, 2000 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  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1982, 1986, 1988, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  * (c) UNIX System Laboratories, Inc.
44  * All or some portions of this file are derived from material licensed
45  * to the University of California by American Telephone and Telegraph
46  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
47  * the permission of UNIX System Laboratories, Inc.
48  *
49  * Redistribution and use in source and binary forms, with or without
50  * modification, are permitted provided that the following conditions
51  * are met:
52  * 1. Redistributions of source code must retain the above copyright
53  *    notice, this list of conditions and the following disclaimer.
54  * 2. Redistributions in binary form must reproduce the above copyright
55  *    notice, this list of conditions and the following disclaimer in the
56  *    documentation and/or other materials provided with the distribution.
57  * 3. Neither the name of the University nor the names of its contributors
58  *    may be used to endorse or promote products derived from this software
59  *    without specific prior written permission.
60  *
61  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
62  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
63  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
64  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
65  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
66  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
67  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
68  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
69  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
70  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
71  * SUCH DAMAGE.
72  *
73  *	@(#)ufs_disksubr.c	8.5 (Berkeley) 1/21/94
74  */
75 
76 #include <sys/cdefs.h>
77 __KERNEL_RCSID(0, "$NetBSD: subr_disk.c,v 1.89 2007/10/08 16:41:15 ad Exp $");
78 
79 #include <sys/param.h>
80 #include <sys/kernel.h>
81 #include <sys/malloc.h>
82 #include <sys/buf.h>
83 #include <sys/syslog.h>
84 #include <sys/disklabel.h>
85 #include <sys/disk.h>
86 #include <sys/sysctl.h>
87 #include <lib/libkern/libkern.h>
88 
89 /*
90  * Compute checksum for disk label.
91  */
92 u_int
93 dkcksum(struct disklabel *lp)
94 {
95 	return dkcksum_sized(lp, lp->d_npartitions);
96 }
97 
98 u_int
99 dkcksum_sized(struct disklabel *lp, size_t npartitions)
100 {
101 	u_short *start, *end;
102 	u_short sum = 0;
103 
104 	start = (u_short *)lp;
105 	end = (u_short *)&lp->d_partitions[npartitions];
106 	while (start < end)
107 		sum ^= *start++;
108 	return (sum);
109 }
110 
111 /*
112  * Disk error is the preface to plaintive error messages
113  * about failing disk transfers.  It prints messages of the form
114 
115 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
116 
117  * if the offset of the error in the transfer and a disk label
118  * are both available.  blkdone should be -1 if the position of the error
119  * is unknown; the disklabel pointer may be null from drivers that have not
120  * been converted to use them.  The message is printed with printf
121  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
122  * The message should be completed (with at least a newline) with printf
123  * or addlog, respectively.  There is no trailing space.
124  */
125 #ifndef PRIdaddr
126 #define PRIdaddr PRId64
127 #endif
128 void
129 diskerr(const struct buf *bp, const char *dname, const char *what, int pri,
130     int blkdone, const struct disklabel *lp)
131 {
132 	int unit = DISKUNIT(bp->b_dev), part = DISKPART(bp->b_dev);
133 	void (*pr)(const char *, ...);
134 	char partname = 'a' + part;
135 	daddr_t sn;
136 
137 	if (/*CONSTCOND*/0)
138 		/* Compiler will error this is the format is wrong... */
139 		printf("%" PRIdaddr, bp->b_blkno);
140 
141 	if (pri != LOG_PRINTF) {
142 		static const char fmt[] = "";
143 		log(pri, fmt);
144 		pr = addlog;
145 	} else
146 		pr = printf;
147 	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
148 	    bp->b_flags & B_READ ? "read" : "writ");
149 	sn = bp->b_blkno;
150 	if (bp->b_bcount <= DEV_BSIZE)
151 		(*pr)("%" PRIdaddr, sn);
152 	else {
153 		if (blkdone >= 0) {
154 			sn += blkdone;
155 			(*pr)("%" PRIdaddr " of ", sn);
156 		}
157 		(*pr)("%" PRIdaddr "-%" PRIdaddr "", bp->b_blkno,
158 		    bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
159 	}
160 	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
161 		sn += lp->d_partitions[part].p_offset;
162 		(*pr)(" (%s%d bn %" PRIdaddr "; cn %" PRIdaddr "",
163 		    dname, unit, sn, sn / lp->d_secpercyl);
164 		sn %= lp->d_secpercyl;
165 		(*pr)(" tn %" PRIdaddr " sn %" PRIdaddr ")",
166 		    sn / lp->d_nsectors, sn % lp->d_nsectors);
167 	}
168 }
169 
170 /*
171  * Searches the iostatlist for the disk corresponding to the
172  * name provided.
173  */
174 struct disk *
175 disk_find(const char *name)
176 {
177 	struct io_stats *stat;
178 
179 	stat = iostat_find(name);
180 
181 	if ((stat != NULL) && (stat->io_type == IOSTAT_DISK))
182 		return stat->io_parent;
183 
184 	return (NULL);
185 }
186 
187 void
188 disk_init(struct disk *diskp, char *name, struct dkdriver *driver)
189 {
190 
191 	/*
192 	 * Initialize the wedge-related locks and other fields.
193 	 */
194 	mutex_init(&diskp->dk_rawlock, MUTEX_DEFAULT, IPL_NONE);
195 	mutex_init(&diskp->dk_openlock, MUTEX_DEFAULT, IPL_NONE);
196 	LIST_INIT(&diskp->dk_wedges);
197 	diskp->dk_nwedges = 0;
198 	diskp->dk_labelsector = LABELSECTOR;
199 	disk_blocksize(diskp, DEV_BSIZE);
200 	diskp->dk_name = name;
201 	diskp->dk_driver = driver;
202 }
203 
204 /*
205  * Attach a disk.
206  */
207 void
208 disk_attach(struct disk *diskp)
209 {
210 
211 	/*
212 	 * Allocate and initialize the disklabel structures.  Note that
213 	 * it's not safe to sleep here, since we're probably going to be
214 	 * called during autoconfiguration.
215 	 */
216 	diskp->dk_label = malloc(sizeof(struct disklabel), M_DEVBUF, M_NOWAIT);
217 	diskp->dk_cpulabel = malloc(sizeof(struct cpu_disklabel), M_DEVBUF,
218 	    M_NOWAIT);
219 	if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL))
220 		panic("disk_attach: can't allocate storage for disklabel");
221 
222 	memset(diskp->dk_label, 0, sizeof(struct disklabel));
223 	memset(diskp->dk_cpulabel, 0, sizeof(struct cpu_disklabel));
224 
225 	/*
226 	 * Set up the stats collection.
227 	 */
228 	diskp->dk_stats = iostat_alloc(IOSTAT_DISK, diskp, diskp->dk_name);
229 }
230 
231 /*
232  * Detach a disk.
233  */
234 void
235 disk_detach(struct disk *diskp)
236 {
237 
238 	/*
239 	 * Remove from the drivelist.
240 	 */
241 	iostat_free(diskp->dk_stats);
242 
243 	/*
244 	 * Release the disk-info dictionary.
245 	 */
246 	if (diskp->dk_info) {
247 		prop_object_release(diskp->dk_info);
248 		diskp->dk_info = NULL;
249 	}
250 
251 	/*
252 	 * Free the space used by the disklabel structures.
253 	 */
254 	free(diskp->dk_label, M_DEVBUF);
255 	free(diskp->dk_cpulabel, M_DEVBUF);
256 }
257 
258 void
259 disk_destroy(struct disk *diskp)
260 {
261 
262 	mutex_destroy(&diskp->dk_openlock);
263 	mutex_destroy(&diskp->dk_rawlock);
264 }
265 
266 /*
267  * Mark the disk as busy for metrics collection.
268  */
269 void
270 disk_busy(struct disk *diskp)
271 {
272 
273 	iostat_busy(diskp->dk_stats);
274 }
275 
276 /*
277  * Finished disk operations, gather metrics.
278  */
279 void
280 disk_unbusy(struct disk *diskp, long bcount, int read)
281 {
282 
283 	iostat_unbusy(diskp->dk_stats, bcount, read);
284 }
285 
286 /*
287  * Set the physical blocksize of a disk, in bytes.
288  * Only necessary if blocksize != DEV_BSIZE.
289  */
290 void
291 disk_blocksize(struct disk *diskp, int blocksize)
292 {
293 
294 	diskp->dk_blkshift = DK_BSIZE2BLKSHIFT(blocksize);
295 	diskp->dk_byteshift = DK_BSIZE2BYTESHIFT(blocksize);
296 }
297 
298 /*
299  * Bounds checking against the media size, used for the raw partition.
300  * The sector size passed in should currently always be DEV_BSIZE,
301  * and the media size the size of the device in DEV_BSIZE sectors.
302  */
303 int
304 bounds_check_with_mediasize(struct buf *bp, int secsize, uint64_t mediasize)
305 {
306 	int64_t sz;
307 
308 	sz = howmany(bp->b_bcount, secsize);
309 
310 	if (bp->b_blkno + sz > mediasize) {
311 		sz = mediasize - bp->b_blkno;
312 		if (sz == 0) {
313 			/* If exactly at end of disk, return EOF. */
314 			bp->b_resid = bp->b_bcount;
315 			return 0;
316 		}
317 		if (sz < 0) {
318 			/* If past end of disk, return EINVAL. */
319 			bp->b_error = EINVAL;
320 			return 0;
321 		}
322 		/* Otherwise, truncate request. */
323 		bp->b_bcount = sz << DEV_BSHIFT;
324 	}
325 
326 	return 1;
327 }
328 
329 /*
330  * Determine the size of the transfer, and make sure it is
331  * within the boundaries of the partition. Adjust transfer
332  * if needed, and signal errors or early completion.
333  */
334 int
335 bounds_check_with_label(struct disk *dk, struct buf *bp, int wlabel)
336 {
337 	struct disklabel *lp = dk->dk_label;
338 	struct partition *p = lp->d_partitions + DISKPART(bp->b_dev);
339 	uint64_t p_size, p_offset, labelsector;
340 	int64_t sz;
341 
342 	/* Protect against division by zero. XXX: Should never happen?!?! */
343 	if (lp->d_secpercyl == 0) {
344 		bp->b_error = EINVAL;
345 		return -1;
346 	}
347 
348 	p_size = p->p_size << dk->dk_blkshift;
349 	p_offset = p->p_offset << dk->dk_blkshift;
350 #if RAW_PART == 3
351 	labelsector = lp->d_partitions[2].p_offset;
352 #else
353 	labelsector = lp->d_partitions[RAW_PART].p_offset;
354 #endif
355 	labelsector = (labelsector + dk->dk_labelsector) << dk->dk_blkshift;
356 
357 	sz = howmany(bp->b_bcount, DEV_BSIZE);
358 	if ((bp->b_blkno + sz) > p_size) {
359 		sz = p_size - bp->b_blkno;
360 		if (sz == 0) {
361 			/* If exactly at end of disk, return EOF. */
362 			bp->b_resid = bp->b_bcount;
363 			return 0;
364 		}
365 		if (sz < 0) {
366 			/* If past end of disk, return EINVAL. */
367 			bp->b_error = EINVAL;
368 			return -1;
369 		}
370 		/* Otherwise, truncate request. */
371 		bp->b_bcount = sz << DEV_BSHIFT;
372 	}
373 
374 	/* Overwriting disk label? */
375 	if (bp->b_blkno + p_offset <= labelsector &&
376 	    bp->b_blkno + p_offset + sz > labelsector &&
377 	    (bp->b_flags & B_READ) == 0 && !wlabel) {
378 		bp->b_error = EROFS;
379 		return -1;
380 	}
381 
382 	/* calculate cylinder for disksort to order transfers with */
383 	bp->b_cylinder = (bp->b_blkno + p->p_offset) /
384 	    (lp->d_secsize / DEV_BSIZE) / lp->d_secpercyl;
385 	return 1;
386 }
387 
388 int
389 disk_read_sectors(void (*strat)(struct buf *), const struct disklabel *lp,
390     struct buf *bp, unsigned int sector, int count)
391 {
392 	bp->b_blkno = sector;
393 	bp->b_bcount = count * lp->d_secsize;
394 	bp->b_flags = (bp->b_flags & ~(B_WRITE | B_DONE)) | B_READ;
395 	bp->b_cylinder = sector / lp->d_secpercyl;
396 	(*strat)(bp);
397 	return biowait(bp);
398 }
399 
400 const char *
401 convertdisklabel(struct disklabel *lp, void (*strat)(struct buf *),
402     struct buf *bp, uint32_t secperunit)
403 {
404 	struct partition rp, *altp, *p;
405 	int geom_ok;
406 
407 	memset(&rp, 0, sizeof(rp));
408 	rp.p_size = secperunit;
409 	rp.p_fstype = FS_UNUSED;
410 
411 	/* If we can seek to d_secperunit - 1, believe the disk geometry. */
412 	if (secperunit != 0 &&
413 	    disk_read_sectors(strat, lp, bp, secperunit - 1, 1) == 0)
414 		geom_ok = 1;
415 	else
416 		geom_ok = 0;
417 
418 #if 0
419 	printf("%s: secperunit (%" PRIu32 ") %s\n", __func__,
420 	    secperunit, geom_ok ? "ok" : "not ok");
421 #endif
422 
423 	p = &lp->d_partitions[RAW_PART];
424 	if (RAW_PART == 'c' - 'a')
425 		altp = &lp->d_partitions['d' - 'a'];
426 	else
427 		altp = &lp->d_partitions['c' - 'a'];
428 
429 	if (lp->d_npartitions > RAW_PART && p->p_offset == 0 && p->p_size != 0)
430 		;	/* already a raw partition */
431 	else if (lp->d_npartitions > MAX('c', 'd') - 'a' &&
432 		 altp->p_offset == 0 && altp->p_size != 0) {
433 		/* alternate partition ('c' or 'd') is suitable for raw slot,
434 		 * swap with 'd' or 'c'.
435 		 */
436 		rp = *p;
437 		*p = *altp;
438 		*altp = rp;
439 	} else if (lp->d_npartitions <= RAW_PART &&
440 	           lp->d_npartitions > 'c' - 'a') {
441 		/* No raw partition is present, but the alternate is present.
442 		 * Copy alternate to raw partition.
443 		 */
444 		lp->d_npartitions = RAW_PART + 1;
445 		*p = *altp;
446 	} else if (!geom_ok)
447 		return "no raw partition and disk reports bad geometry";
448 	else if (lp->d_npartitions <= RAW_PART) {
449 		memset(&lp->d_partitions[lp->d_npartitions], 0,
450 		    sizeof(struct partition) * (RAW_PART - lp->d_npartitions));
451 		*p = rp;
452 		lp->d_npartitions = RAW_PART + 1;
453 	} else if (lp->d_npartitions < MAXPARTITIONS) {
454 		memmove(p + 1, p,
455 		    sizeof(struct partition) * (lp->d_npartitions - RAW_PART));
456 		*p = rp;
457 		lp->d_npartitions++;
458 	} else
459 		return "no raw partition and partition table is full";
460 	return NULL;
461 }
462 
463 /*
464  * disk_ioctl --
465  *	Generic disk ioctl handling.
466  */
467 int
468 disk_ioctl(struct disk *diskp, u_long cmd, void *data, int flag,
469 	   struct lwp *l)
470 {
471 	int error;
472 
473 	switch (cmd) {
474 	case DIOCGDISKINFO:
475 	    {
476 		struct plistref *pref = (struct plistref *) data;
477 
478 		if (diskp->dk_info == NULL)
479 			error = ENOTSUP;
480 		else
481 			error = prop_dictionary_copyout_ioctl(pref, cmd,
482 							diskp->dk_info);
483 		break;
484 	    }
485 
486 	default:
487 		error = EPASSTHROUGH;
488 	}
489 
490 	return (error);
491 }
492