Restore queue depth settings across tag disable events. The system often
[dragonfly.git] / sys / kern / subr_disklabel32.c
blob47d19198750ebe1a206036f63f15881c16227a54
1 /*
2 * Copyright (c) 2003-2007 The DragonFly Project. All rights reserved.
3 *
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
16 * distribution.
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
34 * ----------------------------------------------------------------------------
35 * "THE BEER-WARE LICENSE" (Revision 42):
36 * <phk@FreeBSD.ORG> wrote this file. As long as you retain this notice you
37 * can do whatever you want with this stuff. If we meet some day, and you think
38 * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp
39 * ----------------------------------------------------------------------------
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.
49 * Copyright (c) 1994 Bruce D. Evans.
50 * All rights reserved.
52 * Copyright (c) 1990 The Regents of the University of California.
53 * All rights reserved.
55 * This code is derived from software contributed to Berkeley by
56 * William Jolitz.
58 * Copyright (c) 1982, 1986, 1988 Regents of the University of California.
59 * All rights reserved.
61 * Redistribution and use in source and binary forms, with or without
62 * modification, are permitted provided that the following conditions
63 * are met:
64 * 1. Redistributions of source code must retain the above copyright
65 * notice, this list of conditions and the following disclaimer.
66 * 2. Redistributions in binary form must reproduce the above copyright
67 * notice, this list of conditions and the following disclaimer in the
68 * documentation and/or other materials provided with the distribution.
69 * 3. All advertising materials mentioning features or use of this software
70 * must display the following acknowledgement:
71 * This product includes software developed by the University of
72 * California, Berkeley and its contributors.
73 * 4. Neither the name of the University nor the names of its contributors
74 * may be used to endorse or promote products derived from this software
75 * without specific prior written permission.
77 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
78 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
79 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
80 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
81 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
82 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
83 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
84 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
85 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
86 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
87 * SUCH DAMAGE.
89 * @(#)ufs_disksubr.c 8.5 (Berkeley) 1/21/94
90 * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $
91 * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $
92 * $DragonFly: src/sys/kern/subr_disklabel32.c,v 1.4 2007/06/19 06:07:57 dillon Exp $
95 #include <sys/param.h>
96 #include <sys/systm.h>
97 #include <sys/kernel.h>
98 #include <sys/proc.h>
99 #include <sys/sysctl.h>
100 #include <sys/buf.h>
101 #include <sys/conf.h>
102 #include <sys/disklabel.h>
103 #include <sys/disklabel32.h>
104 #include <sys/diskslice.h>
105 #include <sys/disk.h>
106 #include <sys/dtype.h> /* DTYPE_* constants */
107 #include <sys/sysctl.h>
108 #include <machine/md_var.h>
109 #include <sys/ctype.h>
110 #include <sys/syslog.h>
111 #include <sys/device.h>
112 #include <sys/msgport.h>
113 #include <sys/msgport2.h>
114 #include <sys/buf2.h>
116 #include <vfs/ufs/dinode.h> /* XXX used only for fs.h */
117 #include <vfs/ufs/fs.h> /* XXX used only to get BBSIZE/SBSIZE */
119 static void partition_info(const char *sname, int part, struct partition32 *pp);
120 static void slice_info(const char *sname, struct diskslice *sp);
121 static const char *l32_fixlabel(const char *sname, struct diskslice *sp,
122 disklabel_t lpx, int writeflag);
125 * Retrieve the partition start and extent, in blocks. Return 0 on success,
126 * EINVAL on error.
128 static int
129 l32_getpartbounds(struct diskslices *ssp, disklabel_t lp, u_int32_t part,
130 u_int64_t *start, u_int64_t *blocks)
132 struct partition32 *pp;
134 if (part >= lp.lab32->d_npartitions)
135 return (EINVAL);
136 pp = &lp.lab32->d_partitions[part];
137 *start = pp->p_offset;
138 *blocks = pp->p_size;
139 return(0);
142 static void
143 l32_loadpartinfo(disklabel_t lp, u_int32_t part, struct partinfo *dpart)
145 struct partition32 *pp;
146 const size_t uuid_size = sizeof(struct uuid);
148 bzero(&dpart->fstype_uuid, uuid_size);
149 bzero(&dpart->storage_uuid, uuid_size);
150 if (part < lp.lab32->d_npartitions) {
151 pp = &lp.lab32->d_partitions[part];
152 dpart->fstype = pp->p_fstype;
153 } else {
154 dpart->fstype = 0;
158 static u_int32_t
159 l32_getnumparts(disklabel_t lp)
161 return(lp.lab32->d_npartitions);
165 * Attempt to read a disk label from a device.
167 * Returns NULL on sucess, and an error string on failure
169 static const char *
170 l32_readdisklabel(cdev_t dev, struct diskslice *sp, disklabel_t *lpp,
171 struct disk_info *info)
173 disklabel_t lpx;
174 struct buf *bp;
175 struct disklabel32 *dlp;
176 const char *msg = NULL;
177 int secsize = info->d_media_blksize;
179 bp = geteblk(secsize);
180 bp->b_bio1.bio_offset = (off_t)LABELSECTOR32 * secsize;
181 bp->b_bcount = secsize;
182 bp->b_flags &= ~B_INVAL;
183 bp->b_cmd = BUF_CMD_READ;
184 dev_dstrategy(dev, &bp->b_bio1);
185 if (biowait(bp))
186 msg = "I/O error";
187 else for (dlp = (struct disklabel32 *)bp->b_data;
188 dlp <= (struct disklabel32 *)((char *)bp->b_data +
189 secsize - sizeof(*dlp));
190 dlp = (struct disklabel32 *)((char *)dlp + sizeof(long))) {
191 if (dlp->d_magic != DISKMAGIC32 ||
192 dlp->d_magic2 != DISKMAGIC32) {
194 * NOTE! dsreadandsetlabel() does a strcmp() on
195 * this string.
197 if (msg == NULL)
198 msg = "no disk label";
199 } else if (dlp->d_npartitions > MAXPARTITIONS32 ||
200 dkcksum32(dlp) != 0) {
201 msg = "disk label corrupted";
202 } else {
203 lpx.lab32 = dlp;
204 msg = l32_fixlabel(NULL, sp, lpx, FALSE);
205 if (msg == NULL) {
206 (*lpp).lab32 = kmalloc(sizeof(*dlp),
207 M_DEVBUF, M_WAITOK|M_ZERO);
208 *(*lpp).lab32 = *dlp;
210 break;
213 bp->b_flags |= B_INVAL | B_AGE;
214 brelse(bp);
215 return (msg);
219 * Check new disk label for sensibility before setting it.
221 static int
222 l32_setdisklabel(disklabel_t olpx, disklabel_t nlpx, struct diskslices *ssp,
223 struct diskslice *sp, u_int32_t *openmask)
225 struct disklabel32 *olp, *nlp;
226 struct partition32 *opp, *npp;
227 int part;
228 int i;
230 olp = olpx.lab32;
231 nlp = nlpx.lab32;
234 * Check it is actually a disklabel we are looking at.
236 if (nlp->d_magic != DISKMAGIC32 || nlp->d_magic2 != DISKMAGIC32 ||
237 dkcksum32(nlp) != 0)
238 return (EINVAL);
241 * For each partition that we think is open, check the new disklabel
242 * for compatibility. Ignore special partitions (>= 128).
244 i = 0;
245 while (i < 128) {
246 if (openmask[i >> 5] == 0) {
247 i += 32;
248 continue;
250 if ((openmask[i >> 5] & (1 << (i & 31))) == 0) {
251 ++i;
252 continue;
254 if (nlp->d_npartitions <= i)
255 return (EBUSY);
256 opp = &olp->d_partitions[i];
257 npp = &nlp->d_partitions[i];
258 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size)
259 return (EBUSY);
261 * Copy internally-set partition information
262 * if new label doesn't include it. XXX
263 * (If we are using it then we had better stay the same type)
264 * This is possibly dubious, as someone else noted (XXX)
266 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
267 npp->p_fstype = opp->p_fstype;
268 npp->p_fsize = opp->p_fsize;
269 npp->p_frag = opp->p_frag;
270 npp->p_cpg = opp->p_cpg;
272 ++i;
274 nlp->d_checksum = 0;
275 nlp->d_checksum = dkcksum32(nlp);
276 *olp = *nlp;
278 if (olp->d_partitions[RAW_PART].p_offset)
279 return (EXDEV);
280 if (olp->d_secperunit > sp->ds_size)
281 return (ENOSPC);
282 for (part = 0; part < olp->d_npartitions; ++part) {
283 if (olp->d_partitions[part].p_size > sp->ds_size)
284 return(ENOSPC);
286 return (0);
290 * Write disk label back to device after modification.
292 static int
293 l32_writedisklabel(cdev_t dev, struct diskslices *ssp, struct diskslice *sp,
294 disklabel_t lpx)
296 struct disklabel32 *lp;
297 struct disklabel32 *dlp;
298 struct buf *bp;
299 const char *msg;
300 int error = 0;
302 lp = lpx.lab32;
304 if (lp->d_partitions[RAW_PART].p_offset != 0)
305 return (EXDEV); /* not quite right */
306 bp = geteblk((int)lp->d_secsize);
307 bp->b_bio1.bio_offset = (off_t)LABELSECTOR32 * lp->d_secsize;
308 bp->b_bcount = lp->d_secsize;
309 #if 1
311 * We read the label first to see if it's there,
312 * in which case we will put ours at the same offset into the block..
313 * (I think this is stupid [Julian])
314 * Note that you can't write a label out over a corrupted label!
315 * (also stupid.. how do you write the first one? by raw writes?)
317 bp->b_flags &= ~B_INVAL;
318 bp->b_cmd = BUF_CMD_READ;
319 dev_dstrategy(dkmodpart(dev, WHOLE_SLICE_PART), &bp->b_bio1);
320 error = biowait(bp);
321 if (error)
322 goto done;
323 for (dlp = (struct disklabel32 *)bp->b_data;
324 dlp <= (struct disklabel32 *)
325 ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp));
326 dlp = (struct disklabel32 *)((char *)dlp + sizeof(long))) {
327 if (dlp->d_magic == DISKMAGIC32 &&
328 dlp->d_magic2 == DISKMAGIC32 && dkcksum32(dlp) == 0) {
329 *dlp = *lp;
330 lpx.lab32 = dlp;
331 msg = l32_fixlabel(NULL, sp, lpx, TRUE);
332 if (msg) {
333 error = EINVAL;
334 } else {
335 bp->b_cmd = BUF_CMD_WRITE;
336 dev_dstrategy(dkmodpart(dev, WHOLE_SLICE_PART),
337 &bp->b_bio1);
338 error = biowait(bp);
340 goto done;
343 error = ESRCH;
344 done:
345 #else
346 bzero(bp->b_data, lp->d_secsize);
347 dlp = (struct disklabel32 *)bp->b_data;
348 *dlp = *lp;
349 bp->b_flags &= ~B_INVAL;
350 bp->b_cmd = BUF_CMD_WRITE;
351 BUF_STRATEGY(bp, 1);
352 error = biowait(bp);
353 #endif
354 bp->b_flags |= B_INVAL | B_AGE;
355 brelse(bp);
356 return (error);
360 * Create a disklabel based on a disk_info structure, initializing
361 * the appropriate fields and creating a raw partition that covers the
362 * whole disk.
364 * If a diskslice is passed, the label is truncated to the slice
366 static disklabel_t
367 l32_clone_label(struct disk_info *info, struct diskslice *sp)
369 struct disklabel32 *lp;
370 disklabel_t res;
372 lp = kmalloc(sizeof *lp, M_DEVBUF, M_WAITOK | M_ZERO);
373 lp->d_nsectors = info->d_secpertrack;
374 lp->d_ntracks = info->d_nheads;
375 lp->d_secpercyl = info->d_secpercyl;
376 lp->d_secsize = info->d_media_blksize;
378 if (sp)
379 lp->d_secperunit = (u_int)sp->ds_size;
380 else
381 lp->d_secperunit = (u_int)info->d_media_blocks;
383 if (lp->d_typename[0] == '\0')
384 strncpy(lp->d_typename, "amnesiac", sizeof(lp->d_typename));
385 if (lp->d_packname[0] == '\0')
386 strncpy(lp->d_packname, "fictitious", sizeof(lp->d_packname));
387 if (lp->d_nsectors == 0)
388 lp->d_nsectors = 32;
389 if (lp->d_ntracks == 0)
390 lp->d_ntracks = 64;
391 lp->d_secpercyl = lp->d_nsectors * lp->d_ntracks;
392 lp->d_ncylinders = lp->d_secperunit / lp->d_secpercyl;
393 if (lp->d_rpm == 0)
394 lp->d_rpm = 3600;
395 if (lp->d_interleave == 0)
396 lp->d_interleave = 1;
397 if (lp->d_npartitions < RAW_PART + 1)
398 lp->d_npartitions = MAXPARTITIONS32;
399 if (lp->d_bbsize == 0)
400 lp->d_bbsize = BBSIZE;
401 if (lp->d_sbsize == 0)
402 lp->d_sbsize = SBSIZE;
405 * Used by various devices to create a compatibility slice which
406 * allows us to mount root from devices which do not have a
407 * disklabel. Particularly: CDs.
409 lp->d_partitions[RAW_PART].p_size = lp->d_secperunit;
410 if (info->d_dsflags & DSO_COMPATPARTA) {
411 lp->d_partitions[0].p_size = lp->d_secperunit;
412 lp->d_partitions[0].p_fstype = FS_OTHER;
414 lp->d_magic = DISKMAGIC32;
415 lp->d_magic2 = DISKMAGIC32;
416 lp->d_checksum = dkcksum32(lp);
417 res.lab32 = lp;
418 return (res);
421 static void
422 l32_makevirginlabel(disklabel_t lpx, struct diskslices *ssp,
423 struct diskslice *sp, struct disk_info *info)
425 struct disklabel32 *lp = lpx.lab32;
426 struct partition32 *pp;
427 disklabel_t template;
429 template = l32_clone_label(info, NULL);
430 bcopy(template.opaque, lp, sizeof(struct disklabel32));
432 lp->d_magic = DISKMAGIC32;
433 lp->d_magic2 = DISKMAGIC32;
435 lp->d_npartitions = MAXPARTITIONS32;
436 if (lp->d_interleave == 0)
437 lp->d_interleave = 1;
438 if (lp->d_rpm == 0)
439 lp->d_rpm = 3600;
440 if (lp->d_nsectors == 0) /* sectors per track */
441 lp->d_nsectors = 32;
442 if (lp->d_ntracks == 0) /* heads */
443 lp->d_ntracks = 64;
444 lp->d_ncylinders = 0;
445 lp->d_bbsize = BBSIZE;
446 lp->d_sbsize = SBSIZE;
449 * If the slice or GPT partition is really small we could
450 * wind up with an absurd calculation for ncylinders.
452 while (lp->d_ncylinders < 4) {
453 if (lp->d_ntracks > 1)
454 lp->d_ntracks >>= 1;
455 else if (lp->d_nsectors > 1)
456 lp->d_nsectors >>= 1;
457 else
458 break;
459 lp->d_secpercyl = lp->d_nsectors * lp->d_ntracks;
460 lp->d_ncylinders = sp->ds_size / lp->d_secpercyl;
464 * Set or Modify the partition sizes to accomodate the slice,
465 * since we started with a copy of the virgin label stored
466 * in the whole-disk-slice and we are probably not a
467 * whole-disk slice.
469 lp->d_secperunit = sp->ds_size;
470 pp = &lp->d_partitions[RAW_PART];
471 pp->p_offset = 0;
472 pp->p_size = lp->d_secperunit;
473 if (info->d_dsflags & DSO_COMPATPARTA) {
474 pp = &lp->d_partitions[0];
475 pp->p_offset = 0;
476 pp->p_size = lp->d_secperunit;
477 pp->p_fstype = FS_OTHER;
479 lp->d_checksum = 0;
480 lp->d_checksum = dkcksum32(lp);
482 kfree(template.opaque, M_DEVBUF);
485 static const char *
486 l32_fixlabel(const char *sname, struct diskslice *sp,
487 disklabel_t lpx, int writeflag)
489 struct disklabel32 *lp;
490 struct partition32 *pp;
491 u_int64_t start;
492 u_int64_t end;
493 u_int64_t offset;
494 int part;
495 int warned;
497 lp = lpx.lab32;
499 /* These errors "can't happen" so don't bother reporting details. */
500 if (lp->d_magic != DISKMAGIC32 || lp->d_magic2 != DISKMAGIC32)
501 return ("fixlabel: invalid magic");
502 if (dkcksum32(lp) != 0)
503 return ("fixlabel: invalid checksum");
505 pp = &lp->d_partitions[RAW_PART];
508 * What a mess. For ages old backwards compatibility the disklabel
509 * on-disk stores absolute offsets instead of slice-relative offsets.
510 * So fix it up when reading, writing, or snooping.
512 * The in-core label is always slice-relative.
514 if (writeflag) {
515 start = 0;
516 offset = sp->ds_offset;
517 } else {
518 start = sp->ds_offset;
519 offset = -sp->ds_offset;
521 if (pp->p_offset != start) {
522 if (sname != NULL) {
523 kprintf(
524 "%s: rejecting BSD label: raw partition offset != slice offset\n",
525 sname);
526 slice_info(sname, sp);
527 partition_info(sname, RAW_PART, pp);
529 return ("fixlabel: raw partition offset != slice offset");
531 if (pp->p_size != sp->ds_size) {
532 if (sname != NULL) {
533 kprintf("%s: raw partition size != slice size\n", sname);
534 slice_info(sname, sp);
535 partition_info(sname, RAW_PART, pp);
537 if (pp->p_size > sp->ds_size) {
538 if (sname == NULL)
539 return ("fixlabel: raw partition size > slice size");
540 kprintf("%s: truncating raw partition\n", sname);
541 pp->p_size = sp->ds_size;
544 end = start + sp->ds_size;
545 if (start > end)
546 return ("fixlabel: slice wraps");
547 if (lp->d_secpercyl <= 0)
548 return ("fixlabel: d_secpercyl <= 0");
549 pp -= RAW_PART;
550 warned = FALSE;
551 for (part = 0; part < lp->d_npartitions; part++, pp++) {
552 if (pp->p_offset != 0 || pp->p_size != 0) {
553 if (pp->p_offset < start
554 || pp->p_offset + pp->p_size > end
555 || pp->p_offset + pp->p_size < pp->p_offset) {
556 if (sname != NULL) {
557 kprintf(
558 "%s: rejecting partition in BSD label: it isn't entirely within the slice\n",
559 sname);
560 if (!warned) {
561 slice_info(sname, sp);
562 warned = TRUE;
564 partition_info(sname, part, pp);
566 /* XXX else silently discard junk. */
567 bzero(pp, sizeof *pp);
568 } else {
569 pp->p_offset += offset;
573 lp->d_ncylinders = sp->ds_size / lp->d_secpercyl;
574 lp->d_secperunit = sp->ds_size;
575 lp->d_checksum = 0;
576 lp->d_checksum = dkcksum32(lp);
577 return (NULL);
581 * Set the number of blocks at the beginning of the slice which have
582 * been reserved for label operations. This area will be write-protected
583 * when accessed via the slice.
585 static void
586 l32_adjust_label_reserved(struct diskslices *ssp, int slice,
587 struct diskslice *sp)
589 /*struct disklabel32 *lp = sp->ds_label.lab32;*/
590 sp->ds_reserved = SBSIZE / ssp->dss_secsize;
593 static void
594 partition_info(const char *sname, int part, struct partition32 *pp)
596 kprintf("%s%c: start %lu, end %lu, size %lu\n", sname, 'a' + part,
597 (u_long)pp->p_offset, (u_long)(pp->p_offset + pp->p_size - 1),
598 (u_long)pp->p_size);
601 static void
602 slice_info(const char *sname, struct diskslice *sp)
604 kprintf("%s: start %llu, end %llu, size %llu\n", sname,
605 sp->ds_offset, sp->ds_offset + sp->ds_size - 1, sp->ds_size);
608 struct disklabel_ops disklabel32_ops = {
609 .labelsize = sizeof(struct disklabel32),
610 .op_readdisklabel = l32_readdisklabel,
611 .op_setdisklabel = l32_setdisklabel,
612 .op_writedisklabel = l32_writedisklabel,
613 .op_clone_label = l32_clone_label,
614 .op_adjust_label_reserved = l32_adjust_label_reserved,
615 .op_getpartbounds = l32_getpartbounds,
616 .op_loadpartinfo = l32_loadpartinfo,
617 .op_getnumparts = l32_getnumparts,
618 .op_makevirginlabel = l32_makevirginlabel