xref: /netbsd-src/sys/kern/subr_disk.c (revision aaf4ece63a859a04e37cf3a7229b5fab0157cc06)
1 /*	$NetBSD: subr_disk.c,v 1.73 2005/12/26 18:45:27 perry 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 1.73 2005/12/26 18:45:27 perry Exp $");
78 
79 #include "opt_compat_netbsd.h"
80 
81 #include <sys/param.h>
82 #include <sys/kernel.h>
83 #include <sys/malloc.h>
84 #include <sys/buf.h>
85 #include <sys/syslog.h>
86 #include <sys/disklabel.h>
87 #include <sys/disk.h>
88 #include <sys/sysctl.h>
89 #include <lib/libkern/libkern.h>
90 
91 /*
92  * A global list of all disks attached to the system.  May grow or
93  * shrink over time.
94  */
95 struct	disklist_head disklist = TAILQ_HEAD_INITIALIZER(disklist);
96 int	disk_count;		/* number of drives in global disklist */
97 struct simplelock disklist_slock = SIMPLELOCK_INITIALIZER;
98 
99 /*
100  * Compute checksum for disk label.
101  */
102 u_int
103 dkcksum(struct disklabel *lp)
104 {
105 	u_short *start, *end;
106 	u_short sum = 0;
107 
108 	start = (u_short *)lp;
109 	end = (u_short *)&lp->d_partitions[lp->d_npartitions];
110 	while (start < end)
111 		sum ^= *start++;
112 	return (sum);
113 }
114 
115 /*
116  * Disk error is the preface to plaintive error messages
117  * about failing disk transfers.  It prints messages of the form
118 
119 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
120 
121  * if the offset of the error in the transfer and a disk label
122  * are both available.  blkdone should be -1 if the position of the error
123  * is unknown; the disklabel pointer may be null from drivers that have not
124  * been converted to use them.  The message is printed with printf
125  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
126  * The message should be completed (with at least a newline) with printf
127  * or addlog, respectively.  There is no trailing space.
128  */
129 #ifndef PRIdaddr
130 #define PRIdaddr PRId64
131 #endif
132 void
133 diskerr(const struct buf *bp, const char *dname, const char *what, int pri,
134     int blkdone, const struct disklabel *lp)
135 {
136 	int unit = DISKUNIT(bp->b_dev), part = DISKPART(bp->b_dev);
137 	void (*pr)(const char *, ...);
138 	char partname = 'a' + part;
139 	daddr_t sn;
140 
141 	if (/*CONSTCOND*/0)
142 		/* Compiler will error this is the format is wrong... */
143 		printf("%" PRIdaddr, bp->b_blkno);
144 
145 	if (pri != LOG_PRINTF) {
146 		static const char fmt[] = "";
147 		log(pri, fmt);
148 		pr = addlog;
149 	} else
150 		pr = printf;
151 	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
152 	    bp->b_flags & B_READ ? "read" : "writ");
153 	sn = bp->b_blkno;
154 	if (bp->b_bcount <= DEV_BSIZE)
155 		(*pr)("%" PRIdaddr, sn);
156 	else {
157 		if (blkdone >= 0) {
158 			sn += blkdone;
159 			(*pr)("%" PRIdaddr " of ", sn);
160 		}
161 		(*pr)("%" PRIdaddr "-%" PRIdaddr "", bp->b_blkno,
162 		    bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
163 	}
164 	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
165 		sn += lp->d_partitions[part].p_offset;
166 		(*pr)(" (%s%d bn %" PRIdaddr "; cn %" PRIdaddr "",
167 		    dname, unit, sn, sn / lp->d_secpercyl);
168 		sn %= lp->d_secpercyl;
169 		(*pr)(" tn %" PRIdaddr " sn %" PRIdaddr ")",
170 		    sn / lp->d_nsectors, sn % lp->d_nsectors);
171 	}
172 }
173 
174 /*
175  * Searches the disklist for the disk corresponding to the
176  * name provided.
177  */
178 struct disk *
179 disk_find(char *name)
180 {
181 	struct disk *diskp;
182 
183 	if ((name == NULL) || (disk_count <= 0))
184 		return (NULL);
185 
186 	simple_lock(&disklist_slock);
187 	for (diskp = TAILQ_FIRST(&disklist); diskp != NULL;
188 	    diskp = TAILQ_NEXT(diskp, dk_link))
189 		if (strcmp(diskp->dk_name, name) == 0) {
190 			simple_unlock(&disklist_slock);
191 			return (diskp);
192 		}
193 	simple_unlock(&disklist_slock);
194 
195 	return (NULL);
196 }
197 
198 static void
199 disk_init0(struct disk *diskp)
200 {
201 
202 	/*
203 	 * Initialize the wedge-related locks and other fields.
204 	 */
205 	lockinit(&diskp->dk_rawlock, PRIBIO, "dkrawlk", 0, 0);
206 	lockinit(&diskp->dk_openlock, PRIBIO, "dkoplk", 0, 0);
207 	LIST_INIT(&diskp->dk_wedges);
208 	diskp->dk_nwedges = 0;
209 }
210 
211 static void
212 disk_attach0(struct disk *diskp)
213 {
214 	int s;
215 
216 	/*
217 	 * Allocate and initialize the disklabel structures.  Note that
218 	 * it's not safe to sleep here, since we're probably going to be
219 	 * called during autoconfiguration.
220 	 */
221 	diskp->dk_label = malloc(sizeof(struct disklabel), M_DEVBUF, M_NOWAIT);
222 	diskp->dk_cpulabel = malloc(sizeof(struct cpu_disklabel), M_DEVBUF,
223 	    M_NOWAIT);
224 	if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL))
225 		panic("disk_attach: can't allocate storage for disklabel");
226 
227 	memset(diskp->dk_label, 0, sizeof(struct disklabel));
228 	memset(diskp->dk_cpulabel, 0, sizeof(struct cpu_disklabel));
229 
230 	/*
231 	 * Set the attached timestamp.
232 	 */
233 	s = splclock();
234 	diskp->dk_attachtime = mono_time;
235 	splx(s);
236 
237 	/*
238 	 * Link into the disklist.
239 	 */
240 	simple_lock(&disklist_slock);
241 	TAILQ_INSERT_TAIL(&disklist, diskp, dk_link);
242 	disk_count++;
243 	simple_unlock(&disklist_slock);
244 }
245 
246 static void
247 disk_detach0(struct disk *diskp)
248 {
249 
250 	/*
251 	 * Remove from the disklist.
252 	 */
253 	if (disk_count == 0)
254 		panic("disk_detach: disk_count == 0");
255 	simple_lock(&disklist_slock);
256 	TAILQ_REMOVE(&disklist, diskp, dk_link);
257 	disk_count--;
258 	simple_unlock(&disklist_slock);
259 
260 	/*
261 	 * Free the space used by the disklabel structures.
262 	 */
263 	free(diskp->dk_label, M_DEVBUF);
264 	free(diskp->dk_cpulabel, M_DEVBUF);
265 }
266 
267 /*
268  * Attach a disk.
269  */
270 void
271 disk_attach(struct disk *diskp)
272 {
273 
274 	disk_init0(diskp);
275 	disk_attach0(diskp);
276 }
277 
278 /*
279  * Detach a disk.
280  */
281 void
282 disk_detach(struct disk *diskp)
283 {
284 
285 	(void) lockmgr(&diskp->dk_openlock, LK_DRAIN, NULL);
286 	disk_detach0(diskp);
287 }
288 
289 /*
290  * Initialize a pseudo disk.
291  */
292 void
293 pseudo_disk_init(struct disk *diskp)
294 {
295 
296 	disk_init0(diskp);
297 }
298 
299 /*
300  * Attach a pseudo disk.
301  */
302 void
303 pseudo_disk_attach(struct disk *diskp)
304 {
305 
306 	disk_attach0(diskp);
307 }
308 
309 /*
310  * Detach a pseudo disk.
311  */
312 void
313 pseudo_disk_detach(struct disk *diskp)
314 {
315 
316 	disk_detach0(diskp);
317 }
318 
319 
320 /*
321  * Increment a disk's busy counter.  If the counter is going from
322  * 0 to 1, set the timestamp.
323  */
324 void
325 disk_busy(struct disk *diskp)
326 {
327 	int s;
328 
329 	/*
330 	 * XXX We'd like to use something as accurate as microtime(),
331 	 * but that doesn't depend on the system TOD clock.
332 	 */
333 	if (diskp->dk_busy++ == 0) {
334 		s = splclock();
335 		diskp->dk_timestamp = mono_time;
336 		splx(s);
337 	}
338 }
339 
340 /*
341  * Decrement a disk's busy counter, increment the byte count, total busy
342  * time, and reset the timestamp.
343  */
344 void
345 disk_unbusy(struct disk *diskp, long bcount, int read)
346 {
347 	int s;
348 	struct timeval dv_time, diff_time;
349 
350 	if (diskp->dk_busy-- == 0) {
351 		printf("%s: dk_busy < 0\n", diskp->dk_name);
352 		panic("disk_unbusy");
353 	}
354 
355 	s = splclock();
356 	dv_time = mono_time;
357 	splx(s);
358 
359 	timersub(&dv_time, &diskp->dk_timestamp, &diff_time);
360 	timeradd(&diskp->dk_time, &diff_time, &diskp->dk_time);
361 
362 	diskp->dk_timestamp = dv_time;
363 	if (bcount > 0) {
364 		if (read) {
365 			diskp->dk_rbytes += bcount;
366 			diskp->dk_rxfer++;
367 		} else {
368 			diskp->dk_wbytes += bcount;
369 			diskp->dk_wxfer++;
370 		}
371 	}
372 }
373 
374 /*
375  * Reset the metrics counters on the given disk.  Note that we cannot
376  * reset the busy counter, as it may case a panic in disk_unbusy().
377  * We also must avoid playing with the timestamp information, as it
378  * may skew any pending transfer results.
379  */
380 void
381 disk_resetstat(struct disk *diskp)
382 {
383 	int s = splbio(), t;
384 
385 	diskp->dk_rxfer = 0;
386 	diskp->dk_rbytes = 0;
387 	diskp->dk_wxfer = 0;
388 	diskp->dk_wbytes = 0;
389 
390 	t = splclock();
391 	diskp->dk_attachtime = mono_time;
392 	splx(t);
393 
394 	timerclear(&diskp->dk_time);
395 
396 	splx(s);
397 }
398 
399 int
400 sysctl_hw_disknames(SYSCTLFN_ARGS)
401 {
402 	char bf[DK_DISKNAMELEN + 1];
403 	char *where = oldp;
404 	struct disk *diskp;
405 	size_t needed, left, slen;
406 	int error, first;
407 
408 	if (newp != NULL)
409 		return (EPERM);
410 	if (namelen != 0)
411 		return (EINVAL);
412 
413 	first = 1;
414 	error = 0;
415 	needed = 0;
416 	left = *oldlenp;
417 
418 	simple_lock(&disklist_slock);
419 	for (diskp = TAILQ_FIRST(&disklist); diskp != NULL;
420 	    diskp = TAILQ_NEXT(diskp, dk_link)) {
421 		if (where == NULL)
422 			needed += strlen(diskp->dk_name) + 1;
423 		else {
424 			memset(bf, 0, sizeof(bf));
425 			if (first) {
426 				strncpy(bf, diskp->dk_name, sizeof(bf));
427 				first = 0;
428 			} else {
429 				bf[0] = ' ';
430 				strncpy(bf + 1, diskp->dk_name,
431 				    sizeof(bf) - 1);
432 			}
433 			bf[DK_DISKNAMELEN] = '\0';
434 			slen = strlen(bf);
435 			if (left < slen + 1)
436 				break;
437 			/* +1 to copy out the trailing NUL byte */
438 			error = copyout(bf, where, slen + 1);
439 			if (error)
440 				break;
441 			where += slen;
442 			needed += slen;
443 			left -= slen;
444 		}
445 	}
446 	simple_unlock(&disklist_slock);
447 	*oldlenp = needed;
448 	return (error);
449 }
450 
451 int
452 sysctl_hw_diskstats(SYSCTLFN_ARGS)
453 {
454 	struct disk_sysctl sdisk;
455 	struct disk *diskp;
456 	char *where = oldp;
457 	size_t tocopy, left;
458 	int error;
459 
460 	if (newp != NULL)
461 		return (EPERM);
462 
463 	/*
464 	 * The original hw.diskstats call was broken and did not require
465 	 * the userland to pass in it's size of struct disk_sysctl.  This
466 	 * was fixed after NetBSD 1.6 was released, and any applications
467 	 * that do not pass in the size are given an error only, unless
468 	 * we care about 1.6 compatibility.
469 	 */
470 	if (namelen == 0)
471 #ifdef COMPAT_16
472 		tocopy = offsetof(struct disk_sysctl, dk_rxfer);
473 #else
474 		return (EINVAL);
475 #endif
476 	else
477 		tocopy = name[0];
478 
479 	if (where == NULL) {
480 		*oldlenp = disk_count * tocopy;
481 		return (0);
482 	}
483 
484 	error = 0;
485 	left = *oldlenp;
486 	memset(&sdisk, 0, sizeof(sdisk));
487 	*oldlenp = 0;
488 
489 	simple_lock(&disklist_slock);
490 	TAILQ_FOREACH(diskp, &disklist, dk_link) {
491 		if (left < tocopy)
492 			break;
493 		strncpy(sdisk.dk_name, diskp->dk_name, sizeof(sdisk.dk_name));
494 		sdisk.dk_xfer = diskp->dk_rxfer + diskp->dk_wxfer;
495 		sdisk.dk_rxfer = diskp->dk_rxfer;
496 		sdisk.dk_wxfer = diskp->dk_wxfer;
497 		sdisk.dk_seek = diskp->dk_seek;
498 		sdisk.dk_bytes = diskp->dk_rbytes + diskp->dk_wbytes;
499 		sdisk.dk_rbytes = diskp->dk_rbytes;
500 		sdisk.dk_wbytes = diskp->dk_wbytes;
501 		sdisk.dk_attachtime_sec = diskp->dk_attachtime.tv_sec;
502 		sdisk.dk_attachtime_usec = diskp->dk_attachtime.tv_usec;
503 		sdisk.dk_timestamp_sec = diskp->dk_timestamp.tv_sec;
504 		sdisk.dk_timestamp_usec = diskp->dk_timestamp.tv_usec;
505 		sdisk.dk_time_sec = diskp->dk_time.tv_sec;
506 		sdisk.dk_time_usec = diskp->dk_time.tv_usec;
507 		sdisk.dk_busy = diskp->dk_busy;
508 
509 		error = copyout(&sdisk, where, min(tocopy, sizeof(sdisk)));
510 		if (error)
511 			break;
512 		where += tocopy;
513 		*oldlenp += tocopy;
514 		left -= tocopy;
515 	}
516 	simple_unlock(&disklist_slock);
517 	return (error);
518 }
519 
520 /*
521  * Bounds checking against the media size, used for the raw partition.
522  * The sector size passed in should currently always be DEV_BSIZE,
523  * and the media size the size of the device in DEV_BSIZE sectors.
524  */
525 int
526 bounds_check_with_mediasize(struct buf *bp, int secsize, uint64_t mediasize)
527 {
528 	int64_t sz;
529 
530 	sz = howmany(bp->b_bcount, secsize);
531 
532 	if (bp->b_blkno + sz > mediasize) {
533 		sz = mediasize - bp->b_blkno;
534 		if (sz == 0) {
535 			/* If exactly at end of disk, return EOF. */
536 			bp->b_resid = bp->b_bcount;
537 			goto done;
538 		}
539 		if (sz < 0) {
540 			/* If past end of disk, return EINVAL. */
541 			bp->b_error = EINVAL;
542 			goto bad;
543 		}
544 		/* Otherwise, truncate request. */
545 		bp->b_bcount = sz << DEV_BSHIFT;
546 	}
547 
548 	return 1;
549 
550 bad:
551 	bp->b_flags |= B_ERROR;
552 done:
553 	return 0;
554 }
555