xref: /netbsd-src/sys/kern/subr_disk.c (revision de2138164cd16145ee5fd4d0aef1e8f952c1a9fb)
1 /*	$NetBSD: subr_disk.c,v 1.84 2007/03/01 21:30:50 martin 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 1.84 2007/03/01 21:30:50 martin 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 static void
188 disk_init0(struct disk *diskp)
189 {
190 
191 	/*
192 	 * Initialize the wedge-related locks and other fields.
193 	 */
194 	lockinit(&diskp->dk_rawlock, PRIBIO, "dkrawlk", 0, 0);
195 	lockinit(&diskp->dk_openlock, PRIBIO, "dkoplk", 0, 0);
196 	LIST_INIT(&diskp->dk_wedges);
197 	diskp->dk_nwedges = 0;
198 	diskp->dk_labelsector = LABELSECTOR;
199 	disk_blocksize(diskp, DEV_BSIZE);
200 }
201 
202 static void
203 disk_attach0(struct disk *diskp)
204 {
205 
206 	/*
207 	 * Allocate and initialize the disklabel structures.  Note that
208 	 * it's not safe to sleep here, since we're probably going to be
209 	 * called during autoconfiguration.
210 	 */
211 	diskp->dk_label = malloc(sizeof(struct disklabel), M_DEVBUF, M_NOWAIT);
212 	diskp->dk_cpulabel = malloc(sizeof(struct cpu_disklabel), M_DEVBUF,
213 	    M_NOWAIT);
214 	if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL))
215 		panic("disk_attach: can't allocate storage for disklabel");
216 
217 	memset(diskp->dk_label, 0, sizeof(struct disklabel));
218 	memset(diskp->dk_cpulabel, 0, sizeof(struct cpu_disklabel));
219 
220 	/*
221 	 * Set up the stats collection.
222 	 */
223 	diskp->dk_stats = iostat_alloc(IOSTAT_DISK, diskp, diskp->dk_name);
224 }
225 
226 static void
227 disk_detach0(struct disk *diskp)
228 {
229 
230 	/*
231 	 * Remove from the drivelist.
232 	 */
233 	iostat_free(diskp->dk_stats);
234 
235 	/*
236 	 * Release the disk-info dictionary.
237 	 */
238 	if (diskp->dk_info) {
239 		prop_object_release(diskp->dk_info);
240 		diskp->dk_info = NULL;
241 	}
242 
243 	/*
244 	 * Free the space used by the disklabel structures.
245 	 */
246 	free(diskp->dk_label, M_DEVBUF);
247 	free(diskp->dk_cpulabel, M_DEVBUF);
248 }
249 
250 /*
251  * Attach a disk.
252  */
253 void
254 disk_attach(struct disk *diskp)
255 {
256 
257 	disk_init0(diskp);
258 	disk_attach0(diskp);
259 }
260 
261 /*
262  * Detach a disk.
263  */
264 void
265 disk_detach(struct disk *diskp)
266 {
267 
268 	(void) lockmgr(&diskp->dk_openlock, LK_DRAIN, NULL);
269 	disk_detach0(diskp);
270 }
271 
272 /*
273  * Initialize a pseudo disk.
274  */
275 void
276 pseudo_disk_init(struct disk *diskp)
277 {
278 
279 	disk_init0(diskp);
280 }
281 
282 /*
283  * Attach a pseudo disk.
284  */
285 void
286 pseudo_disk_attach(struct disk *diskp)
287 {
288 
289 	disk_attach0(diskp);
290 }
291 
292 /*
293  * Detach a pseudo disk.
294  */
295 void
296 pseudo_disk_detach(struct disk *diskp)
297 {
298 
299 	disk_detach0(diskp);
300 }
301 
302 /*
303  * Mark the disk as busy for metrics collection.
304  */
305 void
306 disk_busy(struct disk *diskp)
307 {
308 
309 	iostat_busy(diskp->dk_stats);
310 }
311 
312 /*
313  * Finished disk operations, gather metrics.
314  */
315 void
316 disk_unbusy(struct disk *diskp, long bcount, int read)
317 {
318 
319 	iostat_unbusy(diskp->dk_stats, bcount, read);
320 }
321 
322 /*
323  * Set the physical blocksize of a disk, in bytes.
324  * Only necessary if blocksize != DEV_BSIZE.
325  */
326 void
327 disk_blocksize(struct disk *diskp, int blocksize)
328 {
329 
330 	diskp->dk_blkshift = DK_BSIZE2BLKSHIFT(blocksize);
331 	diskp->dk_byteshift = DK_BSIZE2BYTESHIFT(blocksize);
332 }
333 
334 /*
335  * Bounds checking against the media size, used for the raw partition.
336  * The sector size passed in should currently always be DEV_BSIZE,
337  * and the media size the size of the device in DEV_BSIZE sectors.
338  */
339 int
340 bounds_check_with_mediasize(struct buf *bp, int secsize, uint64_t mediasize)
341 {
342 	int64_t sz;
343 
344 	sz = howmany(bp->b_bcount, secsize);
345 
346 	if (bp->b_blkno + sz > mediasize) {
347 		sz = mediasize - bp->b_blkno;
348 		if (sz == 0) {
349 			/* If exactly at end of disk, return EOF. */
350 			bp->b_resid = bp->b_bcount;
351 			goto done;
352 		}
353 		if (sz < 0) {
354 			/* If past end of disk, return EINVAL. */
355 			bp->b_error = EINVAL;
356 			goto bad;
357 		}
358 		/* Otherwise, truncate request. */
359 		bp->b_bcount = sz << DEV_BSHIFT;
360 	}
361 
362 	return 1;
363 
364 bad:
365 	bp->b_flags |= B_ERROR;
366 done:
367 	return 0;
368 }
369 
370 /*
371  * Determine the size of the transfer, and make sure it is
372  * within the boundaries of the partition. Adjust transfer
373  * if needed, and signal errors or early completion.
374  */
375 int
376 bounds_check_with_label(struct disk *dk, struct buf *bp, int wlabel)
377 {
378 	struct disklabel *lp = dk->dk_label;
379 	struct partition *p = lp->d_partitions + DISKPART(bp->b_dev);
380 	uint64_t p_size, p_offset, labelsector;
381 	int64_t sz;
382 
383 	/* Protect against division by zero. XXX: Should never happen?!?! */
384 	if (lp->d_secpercyl == 0) {
385 		bp->b_error = EINVAL;
386 		goto bad;
387 	}
388 
389 	p_size = p->p_size << dk->dk_blkshift;
390 	p_offset = p->p_offset << dk->dk_blkshift;
391 #if RAW_PART == 3
392 	labelsector = lp->d_partitions[2].p_offset;
393 #else
394 	labelsector = lp->d_partitions[RAW_PART].p_offset;
395 #endif
396 	labelsector = (labelsector + dk->dk_labelsector) << dk->dk_blkshift;
397 
398 	sz = howmany(bp->b_bcount, DEV_BSIZE);
399 	if ((bp->b_blkno + sz) > p_size) {
400 		sz = p_size - bp->b_blkno;
401 		if (sz == 0) {
402 			/* If exactly at end of disk, return EOF. */
403 			bp->b_resid = bp->b_bcount;
404 			return (0);
405 		}
406 		if (sz < 0) {
407 			/* If past end of disk, return EINVAL. */
408 			bp->b_error = EINVAL;
409 			goto bad;
410 		}
411 		/* Otherwise, truncate request. */
412 		bp->b_bcount = sz << DEV_BSHIFT;
413 	}
414 
415 	/* Overwriting disk label? */
416 	if (bp->b_blkno + p_offset <= labelsector &&
417 	    bp->b_blkno + p_offset + sz > labelsector &&
418 	    (bp->b_flags & B_READ) == 0 && !wlabel) {
419 		bp->b_error = EROFS;
420 		goto bad;
421 	}
422 
423 	/* calculate cylinder for disksort to order transfers with */
424 	bp->b_cylinder = (bp->b_blkno + p->p_offset) /
425 	    (lp->d_secsize / DEV_BSIZE) / lp->d_secpercyl;
426 	return (1);
427 
428 bad:
429 	bp->b_flags |= B_ERROR;
430 	return (-1);
431 }
432 
433 /*
434  * disk_ioctl --
435  *	Generic disk ioctl handling.
436  */
437 int
438 disk_ioctl(struct disk *diskp, u_long cmd, caddr_t data, int flag,
439 	   struct lwp *l)
440 {
441 	int error;
442 
443 	switch (cmd) {
444 	case DIOCGDISKINFO:
445 	    {
446 		struct plistref *pref = (struct plistref *) data;
447 
448 		if (diskp->dk_info == NULL)
449 			error = ENOTSUP;
450 		else
451 			error = prop_dictionary_copyout_ioctl(pref, cmd,
452 							diskp->dk_info);
453 		break;
454 	    }
455 
456 	default:
457 		error = EPASSTHROUGH;
458 	}
459 
460 	return (error);
461 }
462