xref: /netbsd-src/sys/kern/subr_disk.c (revision c0179c282a5968435315a82f4128c61372c68fc3)
1 /*	$NetBSD: subr_disk.c,v 1.83 2006/11/25 11:59:58 scw 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 1.83 2006/11/25 11:59:58 scw 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 	u_short *start, *end;
96 	u_short sum = 0;
97 
98 	start = (u_short *)lp;
99 	end = (u_short *)&lp->d_partitions[lp->d_npartitions];
100 	while (start < end)
101 		sum ^= *start++;
102 	return (sum);
103 }
104 
105 /*
106  * Disk error is the preface to plaintive error messages
107  * about failing disk transfers.  It prints messages of the form
108 
109 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
110 
111  * if the offset of the error in the transfer and a disk label
112  * are both available.  blkdone should be -1 if the position of the error
113  * is unknown; the disklabel pointer may be null from drivers that have not
114  * been converted to use them.  The message is printed with printf
115  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
116  * The message should be completed (with at least a newline) with printf
117  * or addlog, respectively.  There is no trailing space.
118  */
119 #ifndef PRIdaddr
120 #define PRIdaddr PRId64
121 #endif
122 void
123 diskerr(const struct buf *bp, const char *dname, const char *what, int pri,
124     int blkdone, const struct disklabel *lp)
125 {
126 	int unit = DISKUNIT(bp->b_dev), part = DISKPART(bp->b_dev);
127 	void (*pr)(const char *, ...);
128 	char partname = 'a' + part;
129 	daddr_t sn;
130 
131 	if (/*CONSTCOND*/0)
132 		/* Compiler will error this is the format is wrong... */
133 		printf("%" PRIdaddr, bp->b_blkno);
134 
135 	if (pri != LOG_PRINTF) {
136 		static const char fmt[] = "";
137 		log(pri, fmt);
138 		pr = addlog;
139 	} else
140 		pr = printf;
141 	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
142 	    bp->b_flags & B_READ ? "read" : "writ");
143 	sn = bp->b_blkno;
144 	if (bp->b_bcount <= DEV_BSIZE)
145 		(*pr)("%" PRIdaddr, sn);
146 	else {
147 		if (blkdone >= 0) {
148 			sn += blkdone;
149 			(*pr)("%" PRIdaddr " of ", sn);
150 		}
151 		(*pr)("%" PRIdaddr "-%" PRIdaddr "", bp->b_blkno,
152 		    bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
153 	}
154 	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
155 		sn += lp->d_partitions[part].p_offset;
156 		(*pr)(" (%s%d bn %" PRIdaddr "; cn %" PRIdaddr "",
157 		    dname, unit, sn, sn / lp->d_secpercyl);
158 		sn %= lp->d_secpercyl;
159 		(*pr)(" tn %" PRIdaddr " sn %" PRIdaddr ")",
160 		    sn / lp->d_nsectors, sn % lp->d_nsectors);
161 	}
162 }
163 
164 /*
165  * Searches the iostatlist for the disk corresponding to the
166  * name provided.
167  */
168 struct disk *
169 disk_find(const char *name)
170 {
171 	struct io_stats *stat;
172 
173 	stat = iostat_find(name);
174 
175 	if ((stat != NULL) && (stat->io_type == IOSTAT_DISK))
176 		return stat->io_parent;
177 
178 	return (NULL);
179 }
180 
181 static void
182 disk_init0(struct disk *diskp)
183 {
184 
185 	/*
186 	 * Initialize the wedge-related locks and other fields.
187 	 */
188 	lockinit(&diskp->dk_rawlock, PRIBIO, "dkrawlk", 0, 0);
189 	lockinit(&diskp->dk_openlock, PRIBIO, "dkoplk", 0, 0);
190 	LIST_INIT(&diskp->dk_wedges);
191 	diskp->dk_nwedges = 0;
192 	diskp->dk_labelsector = LABELSECTOR;
193 	disk_blocksize(diskp, DEV_BSIZE);
194 }
195 
196 static void
197 disk_attach0(struct disk *diskp)
198 {
199 
200 	/*
201 	 * Allocate and initialize the disklabel structures.  Note that
202 	 * it's not safe to sleep here, since we're probably going to be
203 	 * called during autoconfiguration.
204 	 */
205 	diskp->dk_label = malloc(sizeof(struct disklabel), M_DEVBUF, M_NOWAIT);
206 	diskp->dk_cpulabel = malloc(sizeof(struct cpu_disklabel), M_DEVBUF,
207 	    M_NOWAIT);
208 	if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL))
209 		panic("disk_attach: can't allocate storage for disklabel");
210 
211 	memset(diskp->dk_label, 0, sizeof(struct disklabel));
212 	memset(diskp->dk_cpulabel, 0, sizeof(struct cpu_disklabel));
213 
214 	/*
215 	 * Set up the stats collection.
216 	 */
217 	diskp->dk_stats = iostat_alloc(IOSTAT_DISK, diskp, diskp->dk_name);
218 }
219 
220 static void
221 disk_detach0(struct disk *diskp)
222 {
223 
224 	/*
225 	 * Remove from the drivelist.
226 	 */
227 	iostat_free(diskp->dk_stats);
228 
229 	/*
230 	 * Release the disk-info dictionary.
231 	 */
232 	if (diskp->dk_info) {
233 		prop_object_release(diskp->dk_info);
234 		diskp->dk_info = NULL;
235 	}
236 
237 	/*
238 	 * Free the space used by the disklabel structures.
239 	 */
240 	free(diskp->dk_label, M_DEVBUF);
241 	free(diskp->dk_cpulabel, M_DEVBUF);
242 }
243 
244 /*
245  * Attach a disk.
246  */
247 void
248 disk_attach(struct disk *diskp)
249 {
250 
251 	disk_init0(diskp);
252 	disk_attach0(diskp);
253 }
254 
255 /*
256  * Detach a disk.
257  */
258 void
259 disk_detach(struct disk *diskp)
260 {
261 
262 	(void) lockmgr(&diskp->dk_openlock, LK_DRAIN, NULL);
263 	disk_detach0(diskp);
264 }
265 
266 /*
267  * Initialize a pseudo disk.
268  */
269 void
270 pseudo_disk_init(struct disk *diskp)
271 {
272 
273 	disk_init0(diskp);
274 }
275 
276 /*
277  * Attach a pseudo disk.
278  */
279 void
280 pseudo_disk_attach(struct disk *diskp)
281 {
282 
283 	disk_attach0(diskp);
284 }
285 
286 /*
287  * Detach a pseudo disk.
288  */
289 void
290 pseudo_disk_detach(struct disk *diskp)
291 {
292 
293 	disk_detach0(diskp);
294 }
295 
296 /*
297  * Mark the disk as busy for metrics collection.
298  */
299 void
300 disk_busy(struct disk *diskp)
301 {
302 
303 	iostat_busy(diskp->dk_stats);
304 }
305 
306 /*
307  * Finished disk operations, gather metrics.
308  */
309 void
310 disk_unbusy(struct disk *diskp, long bcount, int read)
311 {
312 
313 	iostat_unbusy(diskp->dk_stats, bcount, read);
314 }
315 
316 /*
317  * Set the physical blocksize of a disk, in bytes.
318  * Only necessary if blocksize != DEV_BSIZE.
319  */
320 void
321 disk_blocksize(struct disk *diskp, int blocksize)
322 {
323 
324 	diskp->dk_blkshift = DK_BSIZE2BLKSHIFT(blocksize);
325 	diskp->dk_byteshift = DK_BSIZE2BYTESHIFT(blocksize);
326 }
327 
328 /*
329  * Bounds checking against the media size, used for the raw partition.
330  * The sector size passed in should currently always be DEV_BSIZE,
331  * and the media size the size of the device in DEV_BSIZE sectors.
332  */
333 int
334 bounds_check_with_mediasize(struct buf *bp, int secsize, uint64_t mediasize)
335 {
336 	int64_t sz;
337 
338 	sz = howmany(bp->b_bcount, secsize);
339 
340 	if (bp->b_blkno + sz > mediasize) {
341 		sz = mediasize - bp->b_blkno;
342 		if (sz == 0) {
343 			/* If exactly at end of disk, return EOF. */
344 			bp->b_resid = bp->b_bcount;
345 			goto done;
346 		}
347 		if (sz < 0) {
348 			/* If past end of disk, return EINVAL. */
349 			bp->b_error = EINVAL;
350 			goto bad;
351 		}
352 		/* Otherwise, truncate request. */
353 		bp->b_bcount = sz << DEV_BSHIFT;
354 	}
355 
356 	return 1;
357 
358 bad:
359 	bp->b_flags |= B_ERROR;
360 done:
361 	return 0;
362 }
363 
364 /*
365  * Determine the size of the transfer, and make sure it is
366  * within the boundaries of the partition. Adjust transfer
367  * if needed, and signal errors or early completion.
368  */
369 int
370 bounds_check_with_label(struct disk *dk, struct buf *bp, int wlabel)
371 {
372 	struct disklabel *lp = dk->dk_label;
373 	struct partition *p = lp->d_partitions + DISKPART(bp->b_dev);
374 	uint64_t p_size, p_offset, labelsector;
375 	int64_t sz;
376 
377 	/* Protect against division by zero. XXX: Should never happen?!?! */
378 	if (lp->d_secpercyl == 0) {
379 		bp->b_error = EINVAL;
380 		goto bad;
381 	}
382 
383 	p_size = p->p_size << dk->dk_blkshift;
384 	p_offset = p->p_offset << dk->dk_blkshift;
385 #if RAW_PART == 3
386 	labelsector = lp->d_partitions[2].p_offset;
387 #else
388 	labelsector = lp->d_partitions[RAW_PART].p_offset;
389 #endif
390 	labelsector = (labelsector + dk->dk_labelsector) << dk->dk_blkshift;
391 
392 	sz = howmany(bp->b_bcount, DEV_BSIZE);
393 	if ((bp->b_blkno + sz) > p_size) {
394 		sz = p_size - bp->b_blkno;
395 		if (sz == 0) {
396 			/* If exactly at end of disk, return EOF. */
397 			bp->b_resid = bp->b_bcount;
398 			return (0);
399 		}
400 		if (sz < 0) {
401 			/* If past end of disk, return EINVAL. */
402 			bp->b_error = EINVAL;
403 			goto bad;
404 		}
405 		/* Otherwise, truncate request. */
406 		bp->b_bcount = sz << DEV_BSHIFT;
407 	}
408 
409 	/* Overwriting disk label? */
410 	if (bp->b_blkno + p_offset <= labelsector &&
411 	    bp->b_blkno + p_offset + sz > labelsector &&
412 	    (bp->b_flags & B_READ) == 0 && !wlabel) {
413 		bp->b_error = EROFS;
414 		goto bad;
415 	}
416 
417 	/* calculate cylinder for disksort to order transfers with */
418 	bp->b_cylinder = (bp->b_blkno + p->p_offset) /
419 	    (lp->d_secsize / DEV_BSIZE) / lp->d_secpercyl;
420 	return (1);
421 
422 bad:
423 	bp->b_flags |= B_ERROR;
424 	return (-1);
425 }
426 
427 /*
428  * disk_ioctl --
429  *	Generic disk ioctl handling.
430  */
431 int
432 disk_ioctl(struct disk *diskp, u_long cmd, caddr_t data, int flag,
433 	   struct lwp *l)
434 {
435 	int error;
436 
437 	switch (cmd) {
438 	case DIOCGDISKINFO:
439 	    {
440 		struct plistref *pref = (struct plistref *) data;
441 
442 		if (diskp->dk_info == NULL)
443 			error = ENOTSUP;
444 		else
445 			error = prop_dictionary_copyout_ioctl(pref, cmd,
446 							diskp->dk_info);
447 		break;
448 	    }
449 
450 	default:
451 		error = EPASSTHROUGH;
452 	}
453 
454 	return (error);
455 }
456