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
34 * $DragonFly: src/sys/kern/subr_disklabel64.c,v 1.5 2007/07/20 17:21:51 dillon Exp $
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/kernel.h>
41 #include <sys/disklabel.h>
42 #include <sys/disklabel64.h>
43 #include <sys/diskslice.h>
45 #include <sys/kern_syscall.h>
49 * Alignment against physical start (verses slice start). We use a megabyte
50 * here. Why do we use a megabyte? Because SSDs already use large 128K
51 * blocks internally (for MLC) and who the hell knows in the future.
53 * This way if the sysop picks sane values for partition sizes everything
54 * will be nicely aligned, particularly swap for e.g. swapcache, and
55 * clustered operations against larger physical sector sizes for newer HDs,
58 #define PALIGN_SIZE (1024 * 1024)
59 #define PALIGN_MASK (PALIGN_SIZE - 1)
62 * Retrieve the partition start and extent, in blocks. Return 0 on success,
66 l64_getpartbounds(struct diskslices
*ssp
, disklabel_t lp
, u_int32_t part
,
67 u_int64_t
*start
, u_int64_t
*blocks
)
69 struct partition64
*pp
;
71 if (part
>= lp
.lab64
->d_npartitions
)
74 pp
= &lp
.lab64
->d_partitions
[part
];
76 if ((pp
->p_boffset
& (ssp
->dss_secsize
- 1)) ||
77 (pp
->p_bsize
& (ssp
->dss_secsize
- 1))) {
80 *start
= pp
->p_boffset
/ ssp
->dss_secsize
;
81 *blocks
= pp
->p_bsize
/ ssp
->dss_secsize
;
86 * Get the filesystem type XXX - diskslices code needs to use uuids
89 l64_loadpartinfo(disklabel_t lp
, u_int32_t part
, struct partinfo
*dpart
)
91 struct partition64
*pp
;
92 const size_t uuid_size
= sizeof(struct uuid
);
94 if (part
< lp
.lab64
->d_npartitions
) {
95 pp
= &lp
.lab64
->d_partitions
[part
];
96 dpart
->fstype_uuid
= pp
->p_type_uuid
;
97 dpart
->storage_uuid
= pp
->p_stor_uuid
;
98 dpart
->fstype
= pp
->p_fstype
;
100 bzero(&dpart
->fstype_uuid
, uuid_size
);
101 bzero(&dpart
->storage_uuid
, uuid_size
);
107 * Get the number of partitions
110 l64_getnumparts(disklabel_t lp
)
112 return(lp
.lab64
->d_npartitions
);
116 * Attempt to read a disk label from a device. 64 bit disklabels are
117 * sector-agnostic and begin at offset 0 on the device. 64 bit disklabels
118 * may only be used with GPT partitioning schemes.
120 * Returns NULL on sucess, and an error string on failure.
123 l64_readdisklabel(cdev_t dev
, struct diskslice
*sp
, disklabel_t
*lpp
,
124 struct disk_info
*info
)
127 struct disklabel64
*dlp
;
135 * XXX I/O size is subject to device DMA limitations
137 secsize
= info
->d_media_blksize
;
138 bpsize
= (sizeof(*dlp
) + secsize
- 1) & ~(secsize
- 1);
140 bp
= geteblk(bpsize
);
141 bp
->b_bio1
.bio_offset
= 0;
142 bp
->b_bio1
.bio_done
= biodone_sync
;
143 bp
->b_bio1
.bio_flags
|= BIO_SYNC
;
144 bp
->b_bcount
= bpsize
;
145 bp
->b_flags
&= ~B_INVAL
;
146 bp
->b_cmd
= BUF_CMD_READ
;
147 dev_dstrategy(dev
, &bp
->b_bio1
);
149 if (biowait(&bp
->b_bio1
, "labrd")) {
152 dlp
= (struct disklabel64
*)bp
->b_data
;
153 dlpcrcsize
= offsetof(struct disklabel64
,
154 d_partitions
[dlp
->d_npartitions
]) -
155 offsetof(struct disklabel64
, d_magic
);
156 savecrc
= dlp
->d_crc
;
158 if (dlp
->d_magic
!= DISKMAGIC64
) {
159 msg
= "no disk label";
160 } else if (dlp
->d_npartitions
> MAXPARTITIONS64
) {
161 msg
= "disklabel64 corrupted, too many partitions";
162 } else if (savecrc
!= crc32(&dlp
->d_magic
, dlpcrcsize
)) {
163 msg
= "disklabel64 corrupted, bad CRC";
165 dlp
->d_crc
= savecrc
;
166 (*lpp
).lab64
= kmalloc(sizeof(*dlp
),
167 M_DEVBUF
, M_WAITOK
|M_ZERO
);
168 *(*lpp
).lab64
= *dlp
;
172 bp
->b_flags
|= B_INVAL
| B_AGE
;
178 * If everything is good, copy olpx to nlpx. Check to see if any
179 * open partitions would change.
182 l64_setdisklabel(disklabel_t olpx
, disklabel_t nlpx
, struct diskslices
*ssp
,
183 struct diskslice
*sp
, u_int32_t
*openmask
)
185 struct disklabel64
*olp
, *nlp
;
186 struct partition64
*opp
, *npp
;
195 slicebsize
= (uint64_t)sp
->ds_size
* ssp
->dss_secsize
;
197 if (nlp
->d_magic
!= DISKMAGIC64
)
199 if (nlp
->d_npartitions
> MAXPARTITIONS64
)
201 savecrc
= nlp
->d_crc
;
203 nlpcrcsize
= offsetof(struct disklabel64
,
204 d_partitions
[nlp
->d_npartitions
]) -
205 offsetof(struct disklabel64
, d_magic
);
206 if (crc32(&nlp
->d_magic
, nlpcrcsize
) != savecrc
) {
207 nlp
->d_crc
= savecrc
;
210 nlp
->d_crc
= savecrc
;
213 * Check if open partitions have changed
216 while (i
< MAXPARTITIONS64
) {
217 if (openmask
[i
>> 5] == 0) {
221 if ((openmask
[i
>> 5] & (1 << (i
& 31))) == 0) {
225 if (nlp
->d_npartitions
<= i
)
227 opp
= &olp
->d_partitions
[i
];
228 npp
= &nlp
->d_partitions
[i
];
229 if (npp
->p_boffset
!= opp
->p_boffset
||
230 npp
->p_bsize
< opp
->p_bsize
) {
235 * Do not allow p_type_uuid or p_stor_uuid to change if
236 * the partition is currently open.
238 if (bcmp(&npp
->p_type_uuid
, &opp
->p_type_uuid
,
239 sizeof(npp
->p_type_uuid
)) != 0) {
242 if (bcmp(&npp
->p_stor_uuid
, &opp
->p_stor_uuid
,
243 sizeof(npp
->p_stor_uuid
)) != 0) {
250 * Make sure the label and partition offsets and sizes are sane.
252 if (nlp
->d_total_size
> slicebsize
)
254 if (nlp
->d_total_size
& (ssp
->dss_secsize
- 1))
256 if (nlp
->d_bbase
& (ssp
->dss_secsize
- 1))
258 if (nlp
->d_pbase
& (ssp
->dss_secsize
- 1))
260 if (nlp
->d_pstop
& (ssp
->dss_secsize
- 1))
262 if (nlp
->d_abase
& (ssp
->dss_secsize
- 1))
265 for (i
= 0; i
< nlp
->d_npartitions
; ++i
) {
266 npp
= &nlp
->d_partitions
[i
];
267 if (npp
->p_bsize
== 0) {
268 if (npp
->p_boffset
!= 0)
272 if (npp
->p_boffset
& (ssp
->dss_secsize
- 1))
274 if (npp
->p_bsize
& (ssp
->dss_secsize
- 1))
276 if (npp
->p_boffset
< nlp
->d_pbase
)
278 if (npp
->p_boffset
+ npp
->p_bsize
> nlp
->d_total_size
)
283 * Structurally we may add code to make modifications above in the
284 * future, so regenerate the crc anyway.
287 nlp
->d_crc
= crc32(&nlp
->d_magic
, nlpcrcsize
);
294 * Write disk label back to device after modification.
297 l64_writedisklabel(cdev_t dev
, struct diskslices
*ssp
,
298 struct diskslice
*sp
, disklabel_t lpx
)
300 struct disklabel64
*lp
;
301 struct disklabel64
*dlp
;
310 * XXX I/O size is subject to device DMA limitations
312 secsize
= ssp
->dss_secsize
;
313 bpsize
= (sizeof(*lp
) + secsize
- 1) & ~(secsize
- 1);
315 bp
= geteblk(bpsize
);
316 bp
->b_bio1
.bio_offset
= 0;
317 bp
->b_bio1
.bio_done
= biodone_sync
;
318 bp
->b_bio1
.bio_flags
|= BIO_SYNC
;
319 bp
->b_bcount
= bpsize
;
322 * Because our I/O is larger then the label, and because we do not
323 * write the d_reserved0[] area, do a read-modify-write.
325 bp
->b_flags
&= ~B_INVAL
;
326 bp
->b_cmd
= BUF_CMD_READ
;
327 KKASSERT(dkpart(dev
) == WHOLE_SLICE_PART
);
328 dev_dstrategy(dev
, &bp
->b_bio1
);
329 error
= biowait(&bp
->b_bio1
, "labrd");
333 dlp
= (void *)bp
->b_data
;
334 bcopy(&lp
->d_magic
, &dlp
->d_magic
,
335 sizeof(*lp
) - offsetof(struct disklabel64
, d_magic
));
336 bp
->b_cmd
= BUF_CMD_WRITE
;
337 bp
->b_bio1
.bio_done
= biodone_sync
;
338 bp
->b_bio1
.bio_flags
|= BIO_SYNC
;
339 KKASSERT(dkpart(dev
) == WHOLE_SLICE_PART
);
340 dev_dstrategy(dev
, &bp
->b_bio1
);
341 error
= biowait(&bp
->b_bio1
, "labwr");
343 bp
->b_flags
|= B_INVAL
| B_AGE
;
349 * Create a disklabel based on a disk_info structure for the purposes of
350 * DSO_COMPATLABEL - cases where no real label exists on the storage medium.
352 * If a diskslice is passed, the label is truncated to the slice.
354 * NOTE! This is not a legal label because d_bbase and d_pbase are both
358 l64_clone_label(struct disk_info
*info
, struct diskslice
*sp
)
360 struct disklabel64
*lp
;
362 uint32_t blksize
= info
->d_media_blksize
;
365 lp
= kmalloc(sizeof *lp
, M_DEVBUF
, M_WAITOK
| M_ZERO
);
368 lp
->d_total_size
= (uint64_t)sp
->ds_size
* blksize
;
370 lp
->d_total_size
= info
->d_media_blocks
* blksize
;
372 lp
->d_magic
= DISKMAGIC64
;
373 lp
->d_align
= blksize
;
374 lp
->d_npartitions
= MAXPARTITIONS64
;
375 lp
->d_pstop
= lp
->d_total_size
;
378 * Create a dummy 'c' part and a dummy 'a' part (if requested).
379 * Note that the 'c' part is really a hack. 64 bit disklabels
380 * do not use 'c' to mean the raw partition.
383 lp
->d_partitions
[2].p_boffset
= 0;
384 lp
->d_partitions
[2].p_bsize
= lp
->d_total_size
;
385 /* XXX SET FS TYPE */
387 if (info
->d_dsflags
& DSO_COMPATPARTA
) {
388 lp
->d_partitions
[0].p_boffset
= 0;
389 lp
->d_partitions
[0].p_bsize
= lp
->d_total_size
;
390 /* XXX SET FS TYPE */
393 lpcrcsize
= offsetof(struct disklabel64
,
394 d_partitions
[lp
->d_npartitions
]) -
395 offsetof(struct disklabel64
, d_magic
);
397 lp
->d_crc
= crc32(&lp
->d_magic
, lpcrcsize
);
403 * Create a virgin disklabel64 suitable for writing to the media.
405 * disklabel64 always reserves 32KB for a boot area and leaves room
406 * for up to RESPARTITIONS64 partitions.
409 l64_makevirginlabel(disklabel_t lpx
, struct diskslices
*ssp
,
410 struct diskslice
*sp
, struct disk_info
*info
)
412 struct disklabel64
*lp
= lpx
.lab64
;
413 struct partition64
*pp
;
416 uint64_t blkmask
; /* 64 bits so we can ~ */
420 * Setup the initial label. Use of a block size of at least 4KB
421 * for calculating the initial reserved areas to allow some degree
422 * of portability between media with different sector sizes.
424 * Note that the modified blksize is stored in d_align as a hint
425 * to the disklabeling program.
427 bzero(lp
, sizeof(*lp
));
428 if ((blksize
= info
->d_media_blksize
) < 4096)
430 blkmask
= blksize
- 1;
433 lp
->d_total_size
= (uint64_t)sp
->ds_size
* ssp
->dss_secsize
;
435 lp
->d_total_size
= info
->d_media_blocks
* info
->d_media_blksize
;
437 lp
->d_magic
= DISKMAGIC64
;
438 lp
->d_align
= blksize
;
439 lp
->d_npartitions
= MAXPARTITIONS64
;
440 kern_uuidgen(&lp
->d_stor_uuid
, 1);
442 ressize
= offsetof(struct disklabel64
, d_partitions
[RESPARTITIONS64
]);
443 ressize
= (ressize
+ (uint32_t)blkmask
) & ~blkmask
;
446 * NOTE: When calculating pbase take into account the slice offset
447 * so the partitions are at least 32K-aligned relative to the
448 * start of the physical disk. This will accomodate efficient
449 * access to 4096 byte physical sector drives.
451 lp
->d_bbase
= ressize
;
452 lp
->d_pbase
= lp
->d_bbase
+ ((32768 + blkmask
) & ~blkmask
);
453 lp
->d_pbase
= (lp
->d_pbase
+ PALIGN_MASK
) & ~(uint64_t)PALIGN_MASK
;
455 /* adjust for slice offset so we are physically aligned */
456 lp
->d_pbase
+= 32768 - (sp
->ds_offset
* info
->d_media_blksize
) % 32768;
458 lp
->d_pstop
= (lp
->d_total_size
- lp
->d_bbase
) & ~blkmask
;
459 lp
->d_abase
= lp
->d_pstop
;
462 * All partitions are left empty unless DSO_COMPATPARTA is set
465 if (info
->d_dsflags
& DSO_COMPATPARTA
) {
466 pp
= &lp
->d_partitions
[0];
467 pp
->p_boffset
= lp
->d_pbase
;
468 pp
->p_bsize
= lp
->d_pstop
- lp
->d_pbase
;
469 /* XXX SET FS TYPE */
472 lpcrcsize
= offsetof(struct disklabel64
,
473 d_partitions
[lp
->d_npartitions
]) -
474 offsetof(struct disklabel64
, d_magic
);
475 lp
->d_crc
= crc32(&lp
->d_magic
, lpcrcsize
);
479 * Set the number of blocks at the beginning of the slice which have
480 * been reserved for label operations. This area will be write-protected
481 * when accessed via the slice.
483 * For now just protect the label area proper. Do not protect the
484 * boot area. Note partitions in 64 bit disklabels do not overlap
485 * the disklabel or boot area.
488 l64_adjust_label_reserved(struct diskslices
*ssp
, int slice
,
489 struct diskslice
*sp
)
491 struct disklabel64
*lp
= sp
->ds_label
.lab64
;
493 sp
->ds_reserved
= lp
->d_bbase
/ ssp
->dss_secsize
;
496 struct disklabel_ops disklabel64_ops
= {
497 .labelsize
= sizeof(struct disklabel64
),
498 .op_readdisklabel
= l64_readdisklabel
,
499 .op_setdisklabel
= l64_setdisklabel
,
500 .op_writedisklabel
= l64_writedisklabel
,
501 .op_clone_label
= l64_clone_label
,
502 .op_adjust_label_reserved
= l64_adjust_label_reserved
,
503 .op_getpartbounds
= l64_getpartbounds
,
504 .op_loadpartinfo
= l64_loadpartinfo
,
505 .op_getnumparts
= l64_getnumparts
,
506 .op_makevirginlabel
= l64_makevirginlabel