xref: /netbsd-src/sys/kern/subr_disk.c (revision b5677b36047b601b9addaaa494a58ceae82c2a6c)
1 /*	$NetBSD: subr_disk.c,v 1.95 2009/04/04 07:30:10 ad 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 1.95 2009/04/04 07:30:10 ad 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/syslog.h>
77 #include <sys/disklabel.h>
78 #include <sys/disk.h>
79 #include <sys/sysctl.h>
80 #include <lib/libkern/libkern.h>
81 
82 /*
83  * Compute checksum for disk label.
84  */
85 u_int
86 dkcksum(struct disklabel *lp)
87 {
88 	return dkcksum_sized(lp, lp->d_npartitions);
89 }
90 
91 u_int
92 dkcksum_sized(struct disklabel *lp, size_t npartitions)
93 {
94 	u_short *start, *end;
95 	u_short sum = 0;
96 
97 	start = (u_short *)lp;
98 	end = (u_short *)&lp->d_partitions[npartitions];
99 	while (start < end)
100 		sum ^= *start++;
101 	return (sum);
102 }
103 
104 /*
105  * Disk error is the preface to plaintive error messages
106  * about failing disk transfers.  It prints messages of the form
107 
108 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
109 
110  * if the offset of the error in the transfer and a disk label
111  * are both available.  blkdone should be -1 if the position of the error
112  * is unknown; the disklabel pointer may be null from drivers that have not
113  * been converted to use them.  The message is printed with printf
114  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
115  * The message should be completed (with at least a newline) with printf
116  * or addlog, respectively.  There is no trailing space.
117  */
118 #ifndef PRIdaddr
119 #define PRIdaddr PRId64
120 #endif
121 void
122 diskerr(const struct buf *bp, const char *dname, const char *what, int pri,
123     int blkdone, const struct disklabel *lp)
124 {
125 	int unit = DISKUNIT(bp->b_dev), part = DISKPART(bp->b_dev);
126 	void (*pr)(const char *, ...);
127 	char partname = 'a' + part;
128 	daddr_t sn;
129 
130 	if (/*CONSTCOND*/0)
131 		/* Compiler will error this is the format is wrong... */
132 		printf("%" PRIdaddr, bp->b_blkno);
133 
134 	if (pri != LOG_PRINTF) {
135 		static const char fmt[] = "";
136 		log(pri, fmt);
137 		pr = addlog;
138 	} else
139 		pr = printf;
140 	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
141 	    bp->b_flags & B_READ ? "read" : "writ");
142 	sn = bp->b_blkno;
143 	if (bp->b_bcount <= DEV_BSIZE)
144 		(*pr)("%" PRIdaddr, sn);
145 	else {
146 		if (blkdone >= 0) {
147 			sn += blkdone;
148 			(*pr)("%" PRIdaddr " of ", sn);
149 		}
150 		(*pr)("%" PRIdaddr "-%" PRIdaddr "", bp->b_blkno,
151 		    bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
152 	}
153 	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
154 		sn += lp->d_partitions[part].p_offset;
155 		(*pr)(" (%s%d bn %" PRIdaddr "; cn %" PRIdaddr "",
156 		    dname, unit, sn, sn / lp->d_secpercyl);
157 		sn %= lp->d_secpercyl;
158 		(*pr)(" tn %" PRIdaddr " sn %" PRIdaddr ")",
159 		    sn / lp->d_nsectors, sn % lp->d_nsectors);
160 	}
161 }
162 
163 /*
164  * Searches the iostatlist for the disk corresponding to the
165  * name provided.
166  */
167 struct disk *
168 disk_find(const char *name)
169 {
170 	struct io_stats *stat;
171 
172 	stat = iostat_find(name);
173 
174 	if ((stat != NULL) && (stat->io_type == IOSTAT_DISK))
175 		return stat->io_parent;
176 
177 	return (NULL);
178 }
179 
180 void
181 disk_init(struct disk *diskp, const char *name, const struct dkdriver *driver)
182 {
183 
184 	/*
185 	 * Initialize the wedge-related locks and other fields.
186 	 */
187 	mutex_init(&diskp->dk_rawlock, MUTEX_DEFAULT, IPL_NONE);
188 	mutex_init(&diskp->dk_openlock, MUTEX_DEFAULT, IPL_NONE);
189 	LIST_INIT(&diskp->dk_wedges);
190 	diskp->dk_nwedges = 0;
191 	diskp->dk_labelsector = LABELSECTOR;
192 	disk_blocksize(diskp, DEV_BSIZE);
193 	diskp->dk_name = name;
194 	diskp->dk_driver = driver;
195 }
196 
197 /*
198  * Attach a disk.
199  */
200 void
201 disk_attach(struct disk *diskp)
202 {
203 
204 	/*
205 	 * Allocate and initialize the disklabel structures.
206 	 */
207 	diskp->dk_label = kmem_zalloc(sizeof(struct disklabel), KM_SLEEP);
208 	diskp->dk_cpulabel = kmem_zalloc(sizeof(struct cpu_disklabel),
209 	    KM_SLEEP);
210 	if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL))
211 		panic("disk_attach: can't allocate storage for disklabel");
212 
213 	/*
214 	 * Set up the stats collection.
215 	 */
216 	diskp->dk_stats = iostat_alloc(IOSTAT_DISK, diskp, diskp->dk_name);
217 }
218 
219 /*
220  * Detach a disk.
221  */
222 void
223 disk_detach(struct disk *diskp)
224 {
225 
226 	/*
227 	 * Remove from the drivelist.
228 	 */
229 	iostat_free(diskp->dk_stats);
230 
231 	/*
232 	 * Release the disk-info dictionary.
233 	 */
234 	if (diskp->dk_info) {
235 		prop_object_release(diskp->dk_info);
236 		diskp->dk_info = NULL;
237 	}
238 
239 	/*
240 	 * Free the space used by the disklabel structures.
241 	 */
242 	kmem_free(diskp->dk_label, sizeof(*diskp->dk_label));
243 	kmem_free(diskp->dk_cpulabel, sizeof(*diskp->dk_cpulabel));
244 }
245 
246 void
247 disk_destroy(struct disk *diskp)
248 {
249 
250 	mutex_destroy(&diskp->dk_openlock);
251 	mutex_destroy(&diskp->dk_rawlock);
252 }
253 
254 /*
255  * Mark the disk as busy for metrics collection.
256  */
257 void
258 disk_busy(struct disk *diskp)
259 {
260 
261 	iostat_busy(diskp->dk_stats);
262 }
263 
264 /*
265  * Finished disk operations, gather metrics.
266  */
267 void
268 disk_unbusy(struct disk *diskp, long bcount, int read)
269 {
270 
271 	iostat_unbusy(diskp->dk_stats, bcount, read);
272 }
273 
274 /*
275  * Return true if disk has an I/O operation in flight.
276  */
277 bool
278 disk_isbusy(struct disk *diskp)
279 {
280 
281 	return iostat_isbusy(diskp->dk_stats);
282 }
283 
284 /*
285  * Set the physical blocksize of a disk, in bytes.
286  * Only necessary if blocksize != DEV_BSIZE.
287  */
288 void
289 disk_blocksize(struct disk *diskp, int blocksize)
290 {
291 
292 	diskp->dk_blkshift = DK_BSIZE2BLKSHIFT(blocksize);
293 	diskp->dk_byteshift = DK_BSIZE2BYTESHIFT(blocksize);
294 }
295 
296 /*
297  * Bounds checking against the media size, used for the raw partition.
298  * The sector size passed in should currently always be DEV_BSIZE,
299  * and the media size the size of the device in DEV_BSIZE sectors.
300  */
301 int
302 bounds_check_with_mediasize(struct buf *bp, int secsize, uint64_t mediasize)
303 {
304 	int64_t sz;
305 
306 	sz = howmany(bp->b_bcount, secsize);
307 
308 	if (bp->b_blkno + sz > mediasize) {
309 		sz = mediasize - bp->b_blkno;
310 		if (sz == 0) {
311 			/* If exactly at end of disk, return EOF. */
312 			bp->b_resid = bp->b_bcount;
313 			return 0;
314 		}
315 		if (sz < 0) {
316 			/* If past end of disk, return EINVAL. */
317 			bp->b_error = EINVAL;
318 			return 0;
319 		}
320 		/* Otherwise, truncate request. */
321 		bp->b_bcount = sz << DEV_BSHIFT;
322 	}
323 
324 	return 1;
325 }
326 
327 /*
328  * Determine the size of the transfer, and make sure it is
329  * within the boundaries of the partition. Adjust transfer
330  * if needed, and signal errors or early completion.
331  */
332 int
333 bounds_check_with_label(struct disk *dk, struct buf *bp, int wlabel)
334 {
335 	struct disklabel *lp = dk->dk_label;
336 	struct partition *p = lp->d_partitions + DISKPART(bp->b_dev);
337 	uint64_t p_size, p_offset, labelsector;
338 	int64_t sz;
339 
340 	/* Protect against division by zero. XXX: Should never happen?!?! */
341 	if (lp->d_secpercyl == 0) {
342 		bp->b_error = EINVAL;
343 		return -1;
344 	}
345 
346 	p_size = p->p_size << dk->dk_blkshift;
347 	p_offset = p->p_offset << dk->dk_blkshift;
348 #if RAW_PART == 3
349 	labelsector = lp->d_partitions[2].p_offset;
350 #else
351 	labelsector = lp->d_partitions[RAW_PART].p_offset;
352 #endif
353 	labelsector = (labelsector + dk->dk_labelsector) << dk->dk_blkshift;
354 
355 	sz = howmany(bp->b_bcount, DEV_BSIZE);
356 	if ((bp->b_blkno + sz) > p_size) {
357 		sz = p_size - bp->b_blkno;
358 		if (sz == 0) {
359 			/* If exactly at end of disk, return EOF. */
360 			bp->b_resid = bp->b_bcount;
361 			return 0;
362 		}
363 		if (sz < 0) {
364 			/* If past end of disk, return EINVAL. */
365 			bp->b_error = EINVAL;
366 			return -1;
367 		}
368 		/* Otherwise, truncate request. */
369 		bp->b_bcount = sz << DEV_BSHIFT;
370 	}
371 
372 	/* Overwriting disk label? */
373 	if (bp->b_blkno + p_offset <= labelsector &&
374 	    bp->b_blkno + p_offset + sz > labelsector &&
375 	    (bp->b_flags & B_READ) == 0 && !wlabel) {
376 		bp->b_error = EROFS;
377 		return -1;
378 	}
379 
380 	/* calculate cylinder for disksort to order transfers with */
381 	bp->b_cylinder = (bp->b_blkno + p->p_offset) /
382 	    (lp->d_secsize / DEV_BSIZE) / lp->d_secpercyl;
383 	return 1;
384 }
385 
386 int
387 disk_read_sectors(void (*strat)(struct buf *), const struct disklabel *lp,
388     struct buf *bp, unsigned int sector, int count)
389 {
390 	bp->b_blkno = sector;
391 	bp->b_bcount = count * lp->d_secsize;
392 	bp->b_flags = (bp->b_flags & ~B_WRITE) | B_READ;
393 	bp->b_oflags &= ~BO_DONE;
394 	bp->b_cylinder = sector / lp->d_secpercyl;
395 	(*strat)(bp);
396 	return biowait(bp);
397 }
398 
399 const char *
400 convertdisklabel(struct disklabel *lp, void (*strat)(struct buf *),
401     struct buf *bp, uint32_t secperunit)
402 {
403 	struct partition rp, *altp, *p;
404 	int geom_ok;
405 
406 	memset(&rp, 0, sizeof(rp));
407 	rp.p_size = secperunit;
408 	rp.p_fstype = FS_UNUSED;
409 
410 	/* If we can seek to d_secperunit - 1, believe the disk geometry. */
411 	if (secperunit != 0 &&
412 	    disk_read_sectors(strat, lp, bp, secperunit - 1, 1) == 0)
413 		geom_ok = 1;
414 	else
415 		geom_ok = 0;
416 
417 #if 0
418 	printf("%s: secperunit (%" PRIu32 ") %s\n", __func__,
419 	    secperunit, geom_ok ? "ok" : "not ok");
420 #endif
421 
422 	p = &lp->d_partitions[RAW_PART];
423 	if (RAW_PART == 'c' - 'a')
424 		altp = &lp->d_partitions['d' - 'a'];
425 	else
426 		altp = &lp->d_partitions['c' - 'a'];
427 
428 	if (lp->d_npartitions > RAW_PART && p->p_offset == 0 && p->p_size != 0)
429 		;	/* already a raw partition */
430 	else if (lp->d_npartitions > MAX('c', 'd') - 'a' &&
431 		 altp->p_offset == 0 && altp->p_size != 0) {
432 		/* alternate partition ('c' or 'd') is suitable for raw slot,
433 		 * swap with 'd' or 'c'.
434 		 */
435 		rp = *p;
436 		*p = *altp;
437 		*altp = rp;
438 	} else if (lp->d_npartitions <= RAW_PART &&
439 	           lp->d_npartitions > 'c' - 'a') {
440 		/* No raw partition is present, but the alternate is present.
441 		 * Copy alternate to raw partition.
442 		 */
443 		lp->d_npartitions = RAW_PART + 1;
444 		*p = *altp;
445 	} else if (!geom_ok)
446 		return "no raw partition and disk reports bad geometry";
447 	else if (lp->d_npartitions <= RAW_PART) {
448 		memset(&lp->d_partitions[lp->d_npartitions], 0,
449 		    sizeof(struct partition) * (RAW_PART - lp->d_npartitions));
450 		*p = rp;
451 		lp->d_npartitions = RAW_PART + 1;
452 	} else if (lp->d_npartitions < MAXPARTITIONS) {
453 		memmove(p + 1, p,
454 		    sizeof(struct partition) * (lp->d_npartitions - RAW_PART));
455 		*p = rp;
456 		lp->d_npartitions++;
457 	} else
458 		return "no raw partition and partition table is full";
459 	return NULL;
460 }
461 
462 /*
463  * disk_ioctl --
464  *	Generic disk ioctl handling.
465  */
466 int
467 disk_ioctl(struct disk *diskp, u_long cmd, void *data, int flag,
468 	   struct lwp *l)
469 {
470 	int error;
471 
472 	switch (cmd) {
473 	case DIOCGDISKINFO:
474 	    {
475 		struct plistref *pref = (struct plistref *) data;
476 
477 		if (diskp->dk_info == NULL)
478 			error = ENOTSUP;
479 		else
480 			error = prop_dictionary_copyout_ioctl(pref, cmd,
481 							diskp->dk_info);
482 		break;
483 	    }
484 
485 	default:
486 		error = EPASSTHROUGH;
487 	}
488 
489 	return (error);
490 }
491