2 * Copyright (c) 2007 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
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
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/kernel.h>
39 #include <sys/disklabel.h>
40 #include <sys/disklabel64.h>
41 #include <sys/diskslice.h>
43 #include <sys/kern_syscall.h>
47 * Alignment against physical start (verses slice start). We use a megabyte
48 * here. Why do we use a megabyte? Because SSDs already use large 128K
49 * blocks internally (for MLC) and who the hell knows in the future.
51 * This way if the sysop picks sane values for partition sizes everything
52 * will be nicely aligned, particularly swap for e.g. swapcache, and
53 * clustered operations against larger physical sector sizes for newer HDs,
56 #define PALIGN_SIZE (1024 * 1024)
57 #define PALIGN_MASK (PALIGN_SIZE - 1)
60 * Retrieve the partition start and extent, in blocks. Return 0 on success,
64 l64_getpartbounds(struct diskslices *ssp, disklabel_t lp, u_int32_t part,
65 u_int64_t *start, u_int64_t *blocks)
67 struct partition64 *pp;
69 if (part >= lp.lab64->d_npartitions)
72 pp = &lp.lab64->d_partitions[part];
74 if ((pp->p_boffset & (ssp->dss_secsize - 1)) ||
75 (pp->p_bsize & (ssp->dss_secsize - 1))) {
78 *start = pp->p_boffset / ssp->dss_secsize;
79 *blocks = pp->p_bsize / ssp->dss_secsize;
84 * Get the filesystem type XXX - diskslices code needs to use uuids
87 l64_loadpartinfo(disklabel_t lp, u_int32_t part, struct partinfo *dpart)
89 struct partition64 *pp;
90 const size_t uuid_size = sizeof(struct uuid);
92 if (part < lp.lab64->d_npartitions) {
93 pp = &lp.lab64->d_partitions[part];
94 dpart->fstype_uuid = pp->p_type_uuid;
95 dpart->storage_uuid = pp->p_stor_uuid;
96 dpart->fstype = pp->p_fstype;
98 bzero(&dpart->fstype_uuid, uuid_size);
99 bzero(&dpart->storage_uuid, uuid_size);
105 * Get the number of partitions
108 l64_getnumparts(disklabel_t lp)
110 return(lp.lab64->d_npartitions);
114 l64_getpackname(disklabel_t lp, char *buf, size_t bytes)
118 if (lp.lab64->d_packname[0] == 0) {
122 slen = strnlen(lp.lab64->d_packname, sizeof(lp.lab64->d_packname));
125 bcopy(lp.lab64->d_packname, buf, slen);
132 l64_freedisklabel(disklabel_t *lpp)
134 kfree((*lpp).lab64, M_DEVBUF);
139 * Attempt to read a disk label from a device. 64 bit disklabels are
140 * sector-agnostic and begin at offset 0 on the device.
142 * Returns NULL on sucess, and an error string on failure.
145 l64_readdisklabel(cdev_t dev, struct diskslice *sp, disklabel_t *lpp,
146 struct disk_info *info)
149 struct disklabel64 *dlp;
157 * XXX I/O size is subject to device DMA limitations
159 secsize = info->d_media_blksize;
160 bpsize = roundup2(sizeof(*dlp), secsize);
162 bp = getpbuf_mem(NULL);
163 KKASSERT(bpsize <= bp->b_bufsize);
164 bp->b_bio1.bio_offset = 0;
165 bp->b_bio1.bio_done = biodone_sync;
166 bp->b_bio1.bio_flags |= BIO_SYNC;
167 bp->b_bcount = bpsize;
168 bp->b_flags &= ~B_INVAL;
169 bp->b_flags |= B_FAILONDIS;
170 bp->b_cmd = BUF_CMD_READ;
171 dev_dstrategy(dev, &bp->b_bio1);
173 if (biowait(&bp->b_bio1, "labrd")) {
176 dlp = (struct disklabel64 *)bp->b_data;
177 dlpcrcsize = offsetof(struct disklabel64,
178 d_partitions[dlp->d_npartitions]) -
179 offsetof(struct disklabel64, d_magic);
180 savecrc = dlp->d_crc;
182 if (dlp->d_magic != DISKMAGIC64) {
183 msg = "no disk label";
184 } else if (dlp->d_npartitions > MAXPARTITIONS64) {
185 msg = "disklabel64 corrupted, too many partitions";
186 } else if (savecrc != crc32(&dlp->d_magic, dlpcrcsize)) {
187 msg = "disklabel64 corrupted, bad CRC";
189 dlp->d_crc = savecrc;
190 (*lpp).lab64 = kmalloc(sizeof(*dlp),
191 M_DEVBUF, M_WAITOK|M_ZERO);
192 *(*lpp).lab64 = *dlp;
196 bp->b_flags |= B_INVAL | B_AGE;
203 * If everything is good, copy olpx to nlpx. Check to see if any
204 * open partitions would change.
207 l64_setdisklabel(disklabel_t olpx, disklabel_t nlpx, struct diskslices *ssp,
208 struct diskslice *sp, u_int32_t *openmask)
210 struct disklabel64 *olp, *nlp;
211 struct partition64 *opp, *npp;
220 slicebsize = (uint64_t)sp->ds_size * ssp->dss_secsize;
222 if (nlp->d_magic != DISKMAGIC64)
224 if (nlp->d_npartitions > MAXPARTITIONS64)
226 savecrc = nlp->d_crc;
228 nlpcrcsize = offsetof(struct disklabel64,
229 d_partitions[nlp->d_npartitions]) -
230 offsetof(struct disklabel64, d_magic);
231 if (crc32(&nlp->d_magic, nlpcrcsize) != savecrc) {
232 nlp->d_crc = savecrc;
235 nlp->d_crc = savecrc;
238 * Check if open partitions have changed
241 while (i < MAXPARTITIONS64) {
242 if (openmask[i >> 5] == 0) {
246 if ((openmask[i >> 5] & (1 << (i & 31))) == 0) {
250 if (nlp->d_npartitions <= i)
252 opp = &olp->d_partitions[i];
253 npp = &nlp->d_partitions[i];
254 if (npp->p_boffset != opp->p_boffset ||
255 npp->p_bsize < opp->p_bsize) {
260 * Do not allow p_type_uuid or p_stor_uuid to change if
261 * the partition is currently open.
263 if (bcmp(&npp->p_type_uuid, &opp->p_type_uuid,
264 sizeof(npp->p_type_uuid)) != 0) {
267 if (bcmp(&npp->p_stor_uuid, &opp->p_stor_uuid,
268 sizeof(npp->p_stor_uuid)) != 0) {
275 * Make sure the label and partition offsets and sizes are sane.
277 if (nlp->d_total_size > slicebsize)
279 if (nlp->d_total_size & (ssp->dss_secsize - 1))
281 if (nlp->d_bbase & (ssp->dss_secsize - 1))
283 if (nlp->d_pbase & (ssp->dss_secsize - 1))
285 if (nlp->d_pstop & (ssp->dss_secsize - 1))
287 if (nlp->d_abase & (ssp->dss_secsize - 1))
290 for (i = 0; i < nlp->d_npartitions; ++i) {
291 npp = &nlp->d_partitions[i];
292 if (npp->p_bsize == 0) {
293 if (npp->p_boffset != 0)
297 if (npp->p_boffset & (ssp->dss_secsize - 1))
299 if (npp->p_bsize & (ssp->dss_secsize - 1))
301 if (npp->p_boffset < nlp->d_pbase)
303 if (npp->p_boffset + npp->p_bsize > nlp->d_total_size)
308 * Structurally we may add code to make modifications above in the
309 * future, so regenerate the crc anyway.
312 nlp->d_crc = crc32(&nlp->d_magic, nlpcrcsize);
319 * Write disk label back to device after modification.
322 l64_writedisklabel(cdev_t dev, struct diskslices *ssp,
323 struct diskslice *sp, disklabel_t lpx)
325 struct disklabel64 *lp;
326 struct disklabel64 *dlp;
335 * XXX I/O size is subject to device DMA limitations
337 secsize = ssp->dss_secsize;
338 bpsize = roundup2(sizeof(*lp), secsize);
340 bp = getpbuf_mem(NULL);
341 KKASSERT(bpsize <= bp->b_bufsize);
342 bp->b_bio1.bio_offset = 0;
343 bp->b_bio1.bio_done = biodone_sync;
344 bp->b_bio1.bio_flags |= BIO_SYNC;
345 bp->b_bcount = bpsize;
346 bp->b_flags |= B_FAILONDIS;
349 * Because our I/O is larger then the label, and because we do not
350 * write the d_reserved0[] area, do a read-modify-write.
352 bp->b_flags &= ~B_INVAL;
353 bp->b_cmd = BUF_CMD_READ;
354 KKASSERT(dkpart(dev) == WHOLE_SLICE_PART);
355 dev_dstrategy(dev, &bp->b_bio1);
356 error = biowait(&bp->b_bio1, "labrd");
360 dlp = (void *)bp->b_data;
361 bcopy(&lp->d_magic, &dlp->d_magic,
362 sizeof(*lp) - offsetof(struct disklabel64, d_magic));
363 bp->b_cmd = BUF_CMD_WRITE;
364 bp->b_bio1.bio_done = biodone_sync;
365 bp->b_bio1.bio_flags |= BIO_SYNC;
366 KKASSERT(dkpart(dev) == WHOLE_SLICE_PART);
367 dev_dstrategy(dev, &bp->b_bio1);
368 error = biowait(&bp->b_bio1, "labwr");
370 bp->b_flags |= B_INVAL | B_AGE;
377 * Create a disklabel based on a disk_info structure for the purposes of
378 * DSO_COMPATLABEL - cases where no real label exists on the storage medium.
380 * If a diskslice is passed, the label is truncated to the slice.
382 * NOTE! This is not a legal label because d_bbase and d_pbase are both
386 l64_clone_label(struct disk_info *info, struct diskslice *sp)
388 struct disklabel64 *lp;
390 uint32_t blksize = info->d_media_blksize;
393 lp = kmalloc(sizeof *lp, M_DEVBUF, M_WAITOK | M_ZERO);
396 lp->d_total_size = (uint64_t)sp->ds_size * blksize;
398 lp->d_total_size = info->d_media_blocks * blksize;
400 lp->d_magic = DISKMAGIC64;
401 lp->d_align = blksize;
402 lp->d_npartitions = MAXPARTITIONS64;
403 lp->d_pstop = lp->d_total_size;
406 * Create a dummy 'c' part and a dummy 'a' part (if requested).
407 * Note that the 'c' part is really a hack. 64 bit disklabels
408 * do not use 'c' to mean the raw partition.
411 lp->d_partitions[2].p_boffset = 0;
412 lp->d_partitions[2].p_bsize = lp->d_total_size;
413 /* XXX SET FS TYPE */
415 if (info->d_dsflags & DSO_COMPATPARTA) {
416 lp->d_partitions[0].p_boffset = 0;
417 lp->d_partitions[0].p_bsize = lp->d_total_size;
418 /* XXX SET FS TYPE */
421 lpcrcsize = offsetof(struct disklabel64,
422 d_partitions[lp->d_npartitions]) -
423 offsetof(struct disklabel64, d_magic);
425 lp->d_crc = crc32(&lp->d_magic, lpcrcsize);
431 * Create a virgin disklabel64 suitable for writing to the media.
433 * disklabel64 always reserves 32KB for a boot area and leaves room
434 * for up to RESPARTITIONS64 partitions.
437 l64_makevirginlabel(disklabel_t lpx, struct diskslices *ssp,
438 struct diskslice *sp, struct disk_info *info)
440 struct disklabel64 *lp = lpx.lab64;
441 struct partition64 *pp;
444 uint64_t blkmask; /* 64 bits so we can ~ */
448 doffset = sp->ds_offset * info->d_media_blksize;
451 * Setup the initial label. Use of a block size of at least 4KB
452 * for calculating the initial reserved areas to allow some degree
453 * of portability between media with different sector sizes.
455 * Note that the modified blksize is stored in d_align as a hint
456 * to the disklabeling program.
458 bzero(lp, sizeof(*lp));
459 if ((blksize = info->d_media_blksize) < 4096)
461 blkmask = blksize - 1;
464 lp->d_total_size = (uint64_t)sp->ds_size * ssp->dss_secsize;
466 lp->d_total_size = info->d_media_blocks * info->d_media_blksize;
468 lp->d_magic = DISKMAGIC64;
469 lp->d_align = blksize;
470 lp->d_npartitions = MAXPARTITIONS64;
471 kern_uuidgen(&lp->d_stor_uuid, 1);
473 ressize = offsetof(struct disklabel64, d_partitions[RESPARTITIONS64]);
474 ressize = (ressize + (uint32_t)blkmask) & ~blkmask;
476 /* Reserve space for the stage2 boot code */
477 lp->d_bbase = ressize;
478 lp->d_pbase = lp->d_bbase + ((BOOT2SIZE64 + blkmask) & ~blkmask);
480 /* Reserve space for the backup label at the slice end */
481 lp->d_abase = lp->d_total_size - ressize;
484 * NOTE: The pbase and pstop are calculated to align to PALIGN_SIZE
485 * and adjusted with the slice offset, so the partitions are
486 * aligned relative to the start of the physical disk.
488 lp->d_pbase = ((doffset + lp->d_pbase + PALIGN_MASK) &
489 ~(uint64_t)PALIGN_MASK) - doffset;
490 lp->d_pstop = ((lp->d_abase - lp->d_pbase) &
491 ~(uint64_t)PALIGN_MASK) + lp->d_pbase;
494 * All partitions are left empty unless DSO_COMPATPARTA is set
497 if (info->d_dsflags & DSO_COMPATPARTA) {
498 pp = &lp->d_partitions[0];
499 pp->p_boffset = lp->d_pbase;
500 pp->p_bsize = lp->d_pstop - lp->d_pbase;
501 /* XXX SET FS TYPE */
504 lpcrcsize = offsetof(struct disklabel64,
505 d_partitions[lp->d_npartitions]) -
506 offsetof(struct disklabel64, d_magic);
507 lp->d_crc = crc32(&lp->d_magic, lpcrcsize);
511 * Set the number of blocks at the beginning of the slice which have
512 * been reserved for label operations. This area will be write-protected
513 * when accessed via the slice.
515 * For now just protect the label area proper. Do not protect the
516 * boot area. Note partitions in 64 bit disklabels do not overlap
517 * the disklabel or boot area.
520 l64_adjust_label_reserved(struct diskslices *ssp, int slice,
521 struct diskslice *sp)
523 struct disklabel64 *lp = sp->ds_label.lab64;
525 sp->ds_reserved = lp->d_bbase / ssp->dss_secsize;
528 struct disklabel_ops disklabel64_ops = {
529 .labelsize = sizeof(struct disklabel64),
530 .op_readdisklabel = l64_readdisklabel,
531 .op_setdisklabel = l64_setdisklabel,
532 .op_writedisklabel = l64_writedisklabel,
533 .op_clone_label = l64_clone_label,
534 .op_adjust_label_reserved = l64_adjust_label_reserved,
535 .op_getpartbounds = l64_getpartbounds,
536 .op_loadpartinfo = l64_loadpartinfo,
537 .op_getnumparts = l64_getnumparts,
538 .op_getpackname = l64_getpackname,
539 .op_makevirginlabel = l64_makevirginlabel,
540 .op_freedisklabel = l64_freedisklabel