xref: /freebsd/stand/common/part.c (revision eb69d1f144a6fcc765d1b9d44a5ae8082353e70b)
1 /*-
2  * Copyright (c) 2012 Andrey V. Elsukov <ae@FreeBSD.org>
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26 
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
29 
30 #include <stand.h>
31 #include <sys/param.h>
32 #include <sys/diskmbr.h>
33 #include <sys/disklabel.h>
34 #include <sys/endian.h>
35 #include <sys/gpt.h>
36 #include <sys/stddef.h>
37 #include <sys/queue.h>
38 #include <sys/vtoc.h>
39 
40 #include <crc32.h>
41 #include <part.h>
42 #include <uuid.h>
43 
44 #ifdef PART_DEBUG
45 #define	DEBUG(fmt, args...) printf("%s: " fmt "\n", __func__, ## args)
46 #else
47 #define	DEBUG(fmt, args...)
48 #endif
49 
50 #ifdef LOADER_GPT_SUPPORT
51 #define	MAXTBLSZ	64
52 static const uuid_t gpt_uuid_unused = GPT_ENT_TYPE_UNUSED;
53 static const uuid_t gpt_uuid_ms_basic_data = GPT_ENT_TYPE_MS_BASIC_DATA;
54 static const uuid_t gpt_uuid_freebsd_ufs = GPT_ENT_TYPE_FREEBSD_UFS;
55 static const uuid_t gpt_uuid_efi = GPT_ENT_TYPE_EFI;
56 static const uuid_t gpt_uuid_freebsd = GPT_ENT_TYPE_FREEBSD;
57 static const uuid_t gpt_uuid_freebsd_boot = GPT_ENT_TYPE_FREEBSD_BOOT;
58 static const uuid_t gpt_uuid_freebsd_nandfs = GPT_ENT_TYPE_FREEBSD_NANDFS;
59 static const uuid_t gpt_uuid_freebsd_swap = GPT_ENT_TYPE_FREEBSD_SWAP;
60 static const uuid_t gpt_uuid_freebsd_zfs = GPT_ENT_TYPE_FREEBSD_ZFS;
61 static const uuid_t gpt_uuid_freebsd_vinum = GPT_ENT_TYPE_FREEBSD_VINUM;
62 #endif
63 
64 struct pentry {
65 	struct ptable_entry	part;
66 	uint64_t		flags;
67 	union {
68 		uint8_t bsd;
69 		uint8_t	mbr;
70 		uuid_t	gpt;
71 		uint16_t vtoc8;
72 	} type;
73 	STAILQ_ENTRY(pentry)	entry;
74 };
75 
76 struct ptable {
77 	enum ptable_type	type;
78 	uint16_t		sectorsize;
79 	uint64_t		sectors;
80 
81 	STAILQ_HEAD(, pentry)	entries;
82 };
83 
84 static struct parttypes {
85 	enum partition_type	type;
86 	const char		*desc;
87 } ptypes[] = {
88 	{ PART_UNKNOWN,		"Unknown" },
89 	{ PART_EFI,		"EFI" },
90 	{ PART_FREEBSD,		"FreeBSD" },
91 	{ PART_FREEBSD_BOOT,	"FreeBSD boot" },
92 	{ PART_FREEBSD_NANDFS,	"FreeBSD nandfs" },
93 	{ PART_FREEBSD_UFS,	"FreeBSD UFS" },
94 	{ PART_FREEBSD_ZFS,	"FreeBSD ZFS" },
95 	{ PART_FREEBSD_SWAP,	"FreeBSD swap" },
96 	{ PART_FREEBSD_VINUM,	"FreeBSD vinum" },
97 	{ PART_LINUX,		"Linux" },
98 	{ PART_LINUX_SWAP,	"Linux swap" },
99 	{ PART_DOS,		"DOS/Windows" },
100 };
101 
102 const char *
103 parttype2str(enum partition_type type)
104 {
105 	size_t i;
106 
107 	for (i = 0; i < nitems(ptypes); i++)
108 		if (ptypes[i].type == type)
109 			return (ptypes[i].desc);
110 	return (ptypes[0].desc);
111 }
112 
113 #ifdef LOADER_GPT_SUPPORT
114 static void
115 uuid_letoh(uuid_t *uuid)
116 {
117 
118 	uuid->time_low = le32toh(uuid->time_low);
119 	uuid->time_mid = le16toh(uuid->time_mid);
120 	uuid->time_hi_and_version = le16toh(uuid->time_hi_and_version);
121 }
122 
123 static enum partition_type
124 gpt_parttype(uuid_t type)
125 {
126 
127 	if (uuid_equal(&type, &gpt_uuid_efi, NULL))
128 		return (PART_EFI);
129 	else if (uuid_equal(&type, &gpt_uuid_ms_basic_data, NULL))
130 		return (PART_DOS);
131 	else if (uuid_equal(&type, &gpt_uuid_freebsd_boot, NULL))
132 		return (PART_FREEBSD_BOOT);
133 	else if (uuid_equal(&type, &gpt_uuid_freebsd_ufs, NULL))
134 		return (PART_FREEBSD_UFS);
135 	else if (uuid_equal(&type, &gpt_uuid_freebsd_zfs, NULL))
136 		return (PART_FREEBSD_ZFS);
137 	else if (uuid_equal(&type, &gpt_uuid_freebsd_swap, NULL))
138 		return (PART_FREEBSD_SWAP);
139 	else if (uuid_equal(&type, &gpt_uuid_freebsd_vinum, NULL))
140 		return (PART_FREEBSD_VINUM);
141 	else if (uuid_equal(&type, &gpt_uuid_freebsd_nandfs, NULL))
142 		return (PART_FREEBSD_NANDFS);
143 	else if (uuid_equal(&type, &gpt_uuid_freebsd, NULL))
144 		return (PART_FREEBSD);
145 	return (PART_UNKNOWN);
146 }
147 
148 static struct gpt_hdr *
149 gpt_checkhdr(struct gpt_hdr *hdr, uint64_t lba_self, uint64_t lba_last,
150     uint16_t sectorsize)
151 {
152 	uint32_t sz, crc;
153 
154 	if (memcmp(hdr->hdr_sig, GPT_HDR_SIG, sizeof(hdr->hdr_sig)) != 0) {
155 		DEBUG("no GPT signature");
156 		return (NULL);
157 	}
158 	sz = le32toh(hdr->hdr_size);
159 	if (sz < 92 || sz > sectorsize) {
160 		DEBUG("invalid GPT header size: %d", sz);
161 		return (NULL);
162 	}
163 	crc = le32toh(hdr->hdr_crc_self);
164 	hdr->hdr_crc_self = 0;
165 	if (crc32(hdr, sz) != crc) {
166 		DEBUG("GPT header's CRC doesn't match");
167 		return (NULL);
168 	}
169 	hdr->hdr_crc_self = crc;
170 	hdr->hdr_revision = le32toh(hdr->hdr_revision);
171 	if (hdr->hdr_revision < GPT_HDR_REVISION) {
172 		DEBUG("unsupported GPT revision %d", hdr->hdr_revision);
173 		return (NULL);
174 	}
175 	hdr->hdr_lba_self = le64toh(hdr->hdr_lba_self);
176 	if (hdr->hdr_lba_self != lba_self) {
177 		DEBUG("self LBA doesn't match");
178 		return (NULL);
179 	}
180 	hdr->hdr_lba_alt = le64toh(hdr->hdr_lba_alt);
181 	if (hdr->hdr_lba_alt == hdr->hdr_lba_self) {
182 		DEBUG("invalid alternate LBA");
183 		return (NULL);
184 	}
185 	hdr->hdr_entries = le32toh(hdr->hdr_entries);
186 	hdr->hdr_entsz = le32toh(hdr->hdr_entsz);
187 	if (hdr->hdr_entries == 0 ||
188 	    hdr->hdr_entsz < sizeof(struct gpt_ent) ||
189 	    sectorsize % hdr->hdr_entsz != 0) {
190 		DEBUG("invalid entry size or number of entries");
191 		return (NULL);
192 	}
193 	hdr->hdr_lba_start = le64toh(hdr->hdr_lba_start);
194 	hdr->hdr_lba_end = le64toh(hdr->hdr_lba_end);
195 	hdr->hdr_lba_table = le64toh(hdr->hdr_lba_table);
196 	hdr->hdr_crc_table = le32toh(hdr->hdr_crc_table);
197 	uuid_letoh(&hdr->hdr_uuid);
198 	return (hdr);
199 }
200 
201 static int
202 gpt_checktbl(const struct gpt_hdr *hdr, uint8_t *tbl, size_t size,
203     uint64_t lba_last)
204 {
205 	struct gpt_ent *ent;
206 	uint32_t i, cnt;
207 
208 	cnt = size / hdr->hdr_entsz;
209 	if (hdr->hdr_entries <= cnt) {
210 		cnt = hdr->hdr_entries;
211 		/* Check CRC only when buffer size is enough for table. */
212 		if (hdr->hdr_crc_table !=
213 		    crc32(tbl, hdr->hdr_entries * hdr->hdr_entsz)) {
214 			DEBUG("GPT table's CRC doesn't match");
215 			return (-1);
216 		}
217 	}
218 	for (i = 0; i < cnt; i++) {
219 		ent = (struct gpt_ent *)(tbl + i * hdr->hdr_entsz);
220 		uuid_letoh(&ent->ent_type);
221 		if (uuid_equal(&ent->ent_type, &gpt_uuid_unused, NULL))
222 			continue;
223 		ent->ent_lba_start = le64toh(ent->ent_lba_start);
224 		ent->ent_lba_end = le64toh(ent->ent_lba_end);
225 	}
226 	return (0);
227 }
228 
229 static struct ptable *
230 ptable_gptread(struct ptable *table, void *dev, diskread_t dread)
231 {
232 	struct pentry *entry;
233 	struct gpt_hdr *phdr, hdr;
234 	struct gpt_ent *ent;
235 	uint8_t *buf, *tbl;
236 	uint64_t offset;
237 	int pri, sec;
238 	size_t size, i;
239 
240 	buf = malloc(table->sectorsize);
241 	if (buf == NULL)
242 		return (NULL);
243 	tbl = malloc(table->sectorsize * MAXTBLSZ);
244 	if (tbl == NULL) {
245 		free(buf);
246 		return (NULL);
247 	}
248 	/* Read the primary GPT header. */
249 	if (dread(dev, buf, 1, 1) != 0) {
250 		ptable_close(table);
251 		table = NULL;
252 		goto out;
253 	}
254 	pri = sec = 0;
255 	/* Check the primary GPT header. */
256 	phdr = gpt_checkhdr((struct gpt_hdr *)buf, 1, table->sectors - 1,
257 	    table->sectorsize);
258 	if (phdr != NULL) {
259 		/* Read the primary GPT table. */
260 		size = MIN(MAXTBLSZ,
261 		    howmany(phdr->hdr_entries * phdr->hdr_entsz,
262 		        table->sectorsize));
263 		if (dread(dev, tbl, size, phdr->hdr_lba_table) == 0 &&
264 		    gpt_checktbl(phdr, tbl, size * table->sectorsize,
265 		    table->sectors - 1) == 0) {
266 			memcpy(&hdr, phdr, sizeof(hdr));
267 			pri = 1;
268 		}
269 	}
270 	offset = pri ? hdr.hdr_lba_alt: table->sectors - 1;
271 	/* Read the backup GPT header. */
272 	if (dread(dev, buf, 1, offset) != 0)
273 		phdr = NULL;
274 	else
275 		phdr = gpt_checkhdr((struct gpt_hdr *)buf, offset,
276 		    table->sectors - 1, table->sectorsize);
277 	if (phdr != NULL) {
278 		/*
279 		 * Compare primary and backup headers.
280 		 * If they are equal, then we do not need to read backup
281 		 * table. If they are different, then prefer backup header
282 		 * and try to read backup table.
283 		 */
284 		if (pri == 0 ||
285 		    uuid_equal(&hdr.hdr_uuid, &phdr->hdr_uuid, NULL) == 0 ||
286 		    hdr.hdr_revision != phdr->hdr_revision ||
287 		    hdr.hdr_size != phdr->hdr_size ||
288 		    hdr.hdr_lba_start != phdr->hdr_lba_start ||
289 		    hdr.hdr_lba_end != phdr->hdr_lba_end ||
290 		    hdr.hdr_entries != phdr->hdr_entries ||
291 		    hdr.hdr_entsz != phdr->hdr_entsz ||
292 		    hdr.hdr_crc_table != phdr->hdr_crc_table) {
293 			/* Read the backup GPT table. */
294 			size = MIN(MAXTBLSZ,
295 				   howmany(phdr->hdr_entries * phdr->hdr_entsz,
296 				       table->sectorsize));
297 			if (dread(dev, tbl, size, phdr->hdr_lba_table) == 0 &&
298 			    gpt_checktbl(phdr, tbl, size * table->sectorsize,
299 			    table->sectors - 1) == 0) {
300 				memcpy(&hdr, phdr, sizeof(hdr));
301 				sec = 1;
302 			}
303 		}
304 	}
305 	if (pri == 0 && sec == 0) {
306 		/* Both primary and backup tables are invalid. */
307 		table->type = PTABLE_NONE;
308 		goto out;
309 	}
310 	DEBUG("GPT detected");
311 	size = MIN(hdr.hdr_entries * hdr.hdr_entsz,
312 	    MAXTBLSZ * table->sectorsize);
313 
314 	/*
315 	 * If the disk's sector count is smaller than the sector count recorded
316 	 * in the disk's GPT table header, set the table->sectors to the value
317 	 * recorded in GPT tables. This is done to work around buggy firmware
318 	 * that returns truncated disk sizes.
319 	 *
320 	 * Note, this is still not a foolproof way to get disk's size. For
321 	 * example, an image file can be truncated when copied to smaller media.
322 	 */
323 	if (hdr.hdr_lba_alt + 1 > table->sectors)
324 		table->sectors = hdr.hdr_lba_alt + 1;
325 
326 	for (i = 0; i < size / hdr.hdr_entsz; i++) {
327 		ent = (struct gpt_ent *)(tbl + i * hdr.hdr_entsz);
328 		if (uuid_equal(&ent->ent_type, &gpt_uuid_unused, NULL))
329 			continue;
330 
331 		/* Simple sanity checks. */
332 		if (ent->ent_lba_start < hdr.hdr_lba_start ||
333 		    ent->ent_lba_end > hdr.hdr_lba_end ||
334 		    ent->ent_lba_start > ent->ent_lba_end)
335 			continue;
336 
337 		entry = malloc(sizeof(*entry));
338 		if (entry == NULL)
339 			break;
340 		entry->part.start = ent->ent_lba_start;
341 		entry->part.end = ent->ent_lba_end;
342 		entry->part.index = i + 1;
343 		entry->part.type = gpt_parttype(ent->ent_type);
344 		entry->flags = le64toh(ent->ent_attr);
345 		memcpy(&entry->type.gpt, &ent->ent_type, sizeof(uuid_t));
346 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
347 		DEBUG("new GPT partition added");
348 	}
349 out:
350 	free(buf);
351 	free(tbl);
352 	return (table);
353 }
354 #endif /* LOADER_GPT_SUPPORT */
355 
356 #ifdef LOADER_MBR_SUPPORT
357 /* We do not need to support too many EBR partitions in the loader */
358 #define	MAXEBRENTRIES		8
359 static enum partition_type
360 mbr_parttype(uint8_t type)
361 {
362 
363 	switch (type) {
364 	case DOSPTYP_386BSD:
365 		return (PART_FREEBSD);
366 	case DOSPTYP_LINSWP:
367 		return (PART_LINUX_SWAP);
368 	case DOSPTYP_LINUX:
369 		return (PART_LINUX);
370 	case 0x01:
371 	case 0x04:
372 	case 0x06:
373 	case 0x07:
374 	case 0x0b:
375 	case 0x0c:
376 	case 0x0e:
377 		return (PART_DOS);
378 	}
379 	return (PART_UNKNOWN);
380 }
381 
382 static struct ptable *
383 ptable_ebrread(struct ptable *table, void *dev, diskread_t dread)
384 {
385 	struct dos_partition *dp;
386 	struct pentry *e1, *entry;
387 	uint32_t start, end, offset;
388 	u_char *buf;
389 	int i, index;
390 
391 	STAILQ_FOREACH(e1, &table->entries, entry) {
392 		if (e1->type.mbr == DOSPTYP_EXT ||
393 		    e1->type.mbr == DOSPTYP_EXTLBA)
394 			break;
395 	}
396 	if (e1 == NULL)
397 		return (table);
398 	index = 5;
399 	offset = e1->part.start;
400 	buf = malloc(table->sectorsize);
401 	if (buf == NULL)
402 		return (table);
403 	DEBUG("EBR detected");
404 	for (i = 0; i < MAXEBRENTRIES; i++) {
405 #if 0	/* Some BIOSes return an incorrect number of sectors */
406 		if (offset >= table->sectors)
407 			break;
408 #endif
409 		if (dread(dev, buf, 1, offset) != 0)
410 			break;
411 		dp = (struct dos_partition *)(buf + DOSPARTOFF);
412 		if (dp[0].dp_typ == 0)
413 			break;
414 		start = le32toh(dp[0].dp_start);
415 		if (dp[0].dp_typ == DOSPTYP_EXT &&
416 		    dp[1].dp_typ == 0) {
417 			offset = e1->part.start + start;
418 			continue;
419 		}
420 		end = le32toh(dp[0].dp_size);
421 		entry = malloc(sizeof(*entry));
422 		if (entry == NULL)
423 			break;
424 		entry->part.start = offset + start;
425 		entry->part.end = entry->part.start + end - 1;
426 		entry->part.index = index++;
427 		entry->part.type = mbr_parttype(dp[0].dp_typ);
428 		entry->flags = dp[0].dp_flag;
429 		entry->type.mbr = dp[0].dp_typ;
430 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
431 		DEBUG("new EBR partition added");
432 		if (dp[1].dp_typ == 0)
433 			break;
434 		offset = e1->part.start + le32toh(dp[1].dp_start);
435 	}
436 	free(buf);
437 	return (table);
438 }
439 #endif /* LOADER_MBR_SUPPORT */
440 
441 static enum partition_type
442 bsd_parttype(uint8_t type)
443 {
444 
445 	switch (type) {
446 	case FS_NANDFS:
447 		return (PART_FREEBSD_NANDFS);
448 	case FS_SWAP:
449 		return (PART_FREEBSD_SWAP);
450 	case FS_BSDFFS:
451 		return (PART_FREEBSD_UFS);
452 	case FS_VINUM:
453 		return (PART_FREEBSD_VINUM);
454 	case FS_ZFS:
455 		return (PART_FREEBSD_ZFS);
456 	}
457 	return (PART_UNKNOWN);
458 }
459 
460 static struct ptable *
461 ptable_bsdread(struct ptable *table, void *dev, diskread_t dread)
462 {
463 	struct disklabel *dl;
464 	struct partition *part;
465 	struct pentry *entry;
466 	uint8_t *buf;
467 	uint32_t raw_offset;
468 	int i;
469 
470 	if (table->sectorsize < sizeof(struct disklabel)) {
471 		DEBUG("Too small sectorsize");
472 		return (table);
473 	}
474 	buf = malloc(table->sectorsize);
475 	if (buf == NULL)
476 		return (table);
477 	if (dread(dev, buf, 1, 1) != 0) {
478 		DEBUG("read failed");
479 		ptable_close(table);
480 		table = NULL;
481 		goto out;
482 	}
483 	dl = (struct disklabel *)buf;
484 	if (le32toh(dl->d_magic) != DISKMAGIC &&
485 	    le32toh(dl->d_magic2) != DISKMAGIC)
486 		goto out;
487 	if (le32toh(dl->d_secsize) != table->sectorsize) {
488 		DEBUG("unsupported sector size");
489 		goto out;
490 	}
491 	dl->d_npartitions = le16toh(dl->d_npartitions);
492 	if (dl->d_npartitions > 20 || dl->d_npartitions < 8) {
493 		DEBUG("invalid number of partitions");
494 		goto out;
495 	}
496 	DEBUG("BSD detected");
497 	part = &dl->d_partitions[0];
498 	raw_offset = le32toh(part[RAW_PART].p_offset);
499 	for (i = 0; i < dl->d_npartitions; i++, part++) {
500 		if (i == RAW_PART)
501 			continue;
502 		if (part->p_size == 0)
503 			continue;
504 		entry = malloc(sizeof(*entry));
505 		if (entry == NULL)
506 			break;
507 		entry->part.start = le32toh(part->p_offset) - raw_offset;
508 		entry->part.end = entry->part.start +
509 		    le32toh(part->p_size) - 1;
510 		entry->part.type = bsd_parttype(part->p_fstype);
511 		entry->part.index = i; /* starts from zero */
512 		entry->type.bsd = part->p_fstype;
513 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
514 		DEBUG("new BSD partition added");
515 	}
516 	table->type = PTABLE_BSD;
517 out:
518 	free(buf);
519 	return (table);
520 }
521 
522 #ifdef LOADER_VTOC8_SUPPORT
523 static enum partition_type
524 vtoc8_parttype(uint16_t type)
525 {
526 
527 	switch (type) {
528 	case VTOC_TAG_FREEBSD_NANDFS:
529 		return (PART_FREEBSD_NANDFS);
530 	case VTOC_TAG_FREEBSD_SWAP:
531 		return (PART_FREEBSD_SWAP);
532 	case VTOC_TAG_FREEBSD_UFS:
533 		return (PART_FREEBSD_UFS);
534 	case VTOC_TAG_FREEBSD_VINUM:
535 		return (PART_FREEBSD_VINUM);
536 	case VTOC_TAG_FREEBSD_ZFS:
537 		return (PART_FREEBSD_ZFS);
538 	}
539 	return (PART_UNKNOWN);
540 }
541 
542 static struct ptable *
543 ptable_vtoc8read(struct ptable *table, void *dev, diskread_t dread)
544 {
545 	struct pentry *entry;
546 	struct vtoc8 *dl;
547 	uint8_t *buf;
548 	uint16_t sum, heads, sectors;
549 	int i;
550 
551 	if (table->sectorsize != sizeof(struct vtoc8))
552 		return (table);
553 	buf = malloc(table->sectorsize);
554 	if (buf == NULL)
555 		return (table);
556 	if (dread(dev, buf, 1, 0) != 0) {
557 		DEBUG("read failed");
558 		ptable_close(table);
559 		table = NULL;
560 		goto out;
561 	}
562 	dl = (struct vtoc8 *)buf;
563 	/* Check the sum */
564 	for (i = sum = 0; i < sizeof(struct vtoc8); i += sizeof(sum))
565 		sum ^= be16dec(buf + i);
566 	if (sum != 0) {
567 		DEBUG("incorrect checksum");
568 		goto out;
569 	}
570 	if (be16toh(dl->nparts) != VTOC8_NPARTS) {
571 		DEBUG("invalid number of entries");
572 		goto out;
573 	}
574 	sectors = be16toh(dl->nsecs);
575 	heads = be16toh(dl->nheads);
576 	if (sectors * heads == 0) {
577 		DEBUG("invalid geometry");
578 		goto out;
579 	}
580 	DEBUG("VTOC8 detected");
581 	for (i = 0; i < VTOC8_NPARTS; i++) {
582 		dl->part[i].tag = be16toh(dl->part[i].tag);
583 		if (i == VTOC_RAW_PART ||
584 		    dl->part[i].tag == VTOC_TAG_UNASSIGNED)
585 			continue;
586 		entry = malloc(sizeof(*entry));
587 		if (entry == NULL)
588 			break;
589 		entry->part.start = be32toh(dl->map[i].cyl) * heads * sectors;
590 		entry->part.end = be32toh(dl->map[i].nblks) +
591 		    entry->part.start - 1;
592 		entry->part.type = vtoc8_parttype(dl->part[i].tag);
593 		entry->part.index = i; /* starts from zero */
594 		entry->type.vtoc8 = dl->part[i].tag;
595 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
596 		DEBUG("new VTOC8 partition added");
597 	}
598 	table->type = PTABLE_VTOC8;
599 out:
600 	free(buf);
601 	return (table);
602 
603 }
604 #endif /* LOADER_VTOC8_SUPPORT */
605 
606 struct ptable *
607 ptable_open(void *dev, uint64_t sectors, uint16_t sectorsize,
608     diskread_t *dread)
609 {
610 	struct dos_partition *dp;
611 	struct ptable *table;
612 	uint8_t *buf;
613 	int i, count;
614 #ifdef LOADER_MBR_SUPPORT
615 	struct pentry *entry;
616 	uint32_t start, end;
617 	int has_ext;
618 #endif
619 	table = NULL;
620 	buf = malloc(sectorsize);
621 	if (buf == NULL)
622 		return (NULL);
623 	/* First, read the MBR. */
624 	if (dread(dev, buf, 1, DOSBBSECTOR) != 0) {
625 		DEBUG("read failed");
626 		goto out;
627 	}
628 
629 	table = malloc(sizeof(*table));
630 	if (table == NULL)
631 		goto out;
632 	table->sectors = sectors;
633 	table->sectorsize = sectorsize;
634 	table->type = PTABLE_NONE;
635 	STAILQ_INIT(&table->entries);
636 
637 #ifdef LOADER_VTOC8_SUPPORT
638 	if (be16dec(buf + offsetof(struct vtoc8, magic)) == VTOC_MAGIC) {
639 		if (ptable_vtoc8read(table, dev, dread) == NULL) {
640 			/* Read error. */
641 			table = NULL;
642 			goto out;
643 		} else if (table->type == PTABLE_VTOC8)
644 			goto out;
645 	}
646 #endif
647 	/* Check the BSD label. */
648 	if (ptable_bsdread(table, dev, dread) == NULL) { /* Read error. */
649 		table = NULL;
650 		goto out;
651 	} else if (table->type == PTABLE_BSD)
652 		goto out;
653 
654 #if defined(LOADER_GPT_SUPPORT) || defined(LOADER_MBR_SUPPORT)
655 	/* Check the MBR magic. */
656 	if (buf[DOSMAGICOFFSET] != 0x55 ||
657 	    buf[DOSMAGICOFFSET + 1] != 0xaa) {
658 		DEBUG("magic sequence not found");
659 #if defined(LOADER_GPT_SUPPORT)
660 		/* There is no PMBR, check that we have backup GPT */
661 		table->type = PTABLE_GPT;
662 		table = ptable_gptread(table, dev, dread);
663 #endif
664 		goto out;
665 	}
666 	/* Check that we have PMBR. Also do some validation. */
667 	dp = (struct dos_partition *)(buf + DOSPARTOFF);
668 	for (i = 0, count = 0; i < NDOSPART; i++) {
669 		if (dp[i].dp_flag != 0 && dp[i].dp_flag != 0x80) {
670 			DEBUG("invalid partition flag %x", dp[i].dp_flag);
671 			goto out;
672 		}
673 #ifdef LOADER_GPT_SUPPORT
674 		if (dp[i].dp_typ == DOSPTYP_PMBR) {
675 			table->type = PTABLE_GPT;
676 			DEBUG("PMBR detected");
677 		}
678 #endif
679 		if (dp[i].dp_typ != 0)
680 			count++;
681 	}
682 	/* Do we have some invalid values? */
683 	if (table->type == PTABLE_GPT && count > 1) {
684 		if (dp[1].dp_typ != DOSPTYP_HFS) {
685 			table->type = PTABLE_NONE;
686 			DEBUG("Incorrect PMBR, ignore it");
687 		} else {
688 			DEBUG("Bootcamp detected");
689 		}
690 	}
691 #ifdef LOADER_GPT_SUPPORT
692 	if (table->type == PTABLE_GPT) {
693 		table = ptable_gptread(table, dev, dread);
694 		goto out;
695 	}
696 #endif
697 #ifdef LOADER_MBR_SUPPORT
698 	/* Read MBR. */
699 	DEBUG("MBR detected");
700 	table->type = PTABLE_MBR;
701 	for (i = has_ext = 0; i < NDOSPART; i++) {
702 		if (dp[i].dp_typ == 0)
703 			continue;
704 		start = le32dec(&(dp[i].dp_start));
705 		end = le32dec(&(dp[i].dp_size));
706 		if (start == 0 || end == 0)
707 			continue;
708 #if 0	/* Some BIOSes return an incorrect number of sectors */
709 		if (start + end - 1 >= sectors)
710 			continue;	/* XXX: ignore */
711 #endif
712 		if (dp[i].dp_typ == DOSPTYP_EXT ||
713 		    dp[i].dp_typ == DOSPTYP_EXTLBA)
714 			has_ext = 1;
715 		entry = malloc(sizeof(*entry));
716 		if (entry == NULL)
717 			break;
718 		entry->part.start = start;
719 		entry->part.end = start + end - 1;
720 		entry->part.index = i + 1;
721 		entry->part.type = mbr_parttype(dp[i].dp_typ);
722 		entry->flags = dp[i].dp_flag;
723 		entry->type.mbr = dp[i].dp_typ;
724 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
725 		DEBUG("new MBR partition added");
726 	}
727 	if (has_ext) {
728 		table = ptable_ebrread(table, dev, dread);
729 		/* FALLTHROUGH */
730 	}
731 #endif /* LOADER_MBR_SUPPORT */
732 #endif /* LOADER_MBR_SUPPORT || LOADER_GPT_SUPPORT */
733 out:
734 	free(buf);
735 	return (table);
736 }
737 
738 void
739 ptable_close(struct ptable *table)
740 {
741 	struct pentry *entry;
742 
743 	while (!STAILQ_EMPTY(&table->entries)) {
744 		entry = STAILQ_FIRST(&table->entries);
745 		STAILQ_REMOVE_HEAD(&table->entries, entry);
746 		free(entry);
747 	}
748 	free(table);
749 }
750 
751 enum ptable_type
752 ptable_gettype(const struct ptable *table)
753 {
754 
755 	return (table->type);
756 }
757 
758 int
759 ptable_getsize(const struct ptable *table, uint64_t *sizep)
760 {
761 	uint64_t tmp = table->sectors * table->sectorsize;
762 
763 	if (tmp < table->sectors)
764 		return (EOVERFLOW);
765 
766 	if (sizep != NULL)
767 		*sizep = tmp;
768 	return (0);
769 }
770 
771 int
772 ptable_getpart(const struct ptable *table, struct ptable_entry *part, int index)
773 {
774 	struct pentry *entry;
775 
776 	if (part == NULL || table == NULL)
777 		return (EINVAL);
778 
779 	STAILQ_FOREACH(entry, &table->entries, entry) {
780 		if (entry->part.index != index)
781 			continue;
782 		memcpy(part, &entry->part, sizeof(*part));
783 		return (0);
784 	}
785 	return (ENOENT);
786 }
787 
788 /*
789  * Search for a slice with the following preferences:
790  *
791  * 1: Active FreeBSD slice
792  * 2: Non-active FreeBSD slice
793  * 3: Active Linux slice
794  * 4: non-active Linux slice
795  * 5: Active FAT/FAT32 slice
796  * 6: non-active FAT/FAT32 slice
797  */
798 #define	PREF_RAWDISK	0
799 #define	PREF_FBSD_ACT	1
800 #define	PREF_FBSD	2
801 #define	PREF_LINUX_ACT	3
802 #define	PREF_LINUX	4
803 #define	PREF_DOS_ACT	5
804 #define	PREF_DOS	6
805 #define	PREF_NONE	7
806 int
807 ptable_getbestpart(const struct ptable *table, struct ptable_entry *part)
808 {
809 	struct pentry *entry, *best;
810 	int pref, preflevel;
811 
812 	if (part == NULL || table == NULL)
813 		return (EINVAL);
814 
815 	best = NULL;
816 	preflevel = pref = PREF_NONE;
817 	STAILQ_FOREACH(entry, &table->entries, entry) {
818 #ifdef LOADER_MBR_SUPPORT
819 		if (table->type == PTABLE_MBR) {
820 			switch (entry->type.mbr) {
821 			case DOSPTYP_386BSD:
822 				pref = entry->flags & 0x80 ? PREF_FBSD_ACT:
823 				    PREF_FBSD;
824 				break;
825 			case DOSPTYP_LINUX:
826 				pref = entry->flags & 0x80 ? PREF_LINUX_ACT:
827 				    PREF_LINUX;
828 				break;
829 			case 0x01:		/* DOS/Windows */
830 			case 0x04:
831 			case 0x06:
832 			case 0x0c:
833 			case 0x0e:
834 			case DOSPTYP_FAT32:
835 				pref = entry->flags & 0x80 ? PREF_DOS_ACT:
836 				    PREF_DOS;
837 				break;
838 			default:
839 				pref = PREF_NONE;
840 			}
841 		}
842 #endif /* LOADER_MBR_SUPPORT */
843 #ifdef LOADER_GPT_SUPPORT
844 		if (table->type == PTABLE_GPT) {
845 			if (entry->part.type == PART_DOS)
846 				pref = PREF_DOS;
847 			else if (entry->part.type == PART_FREEBSD_UFS ||
848 			    entry->part.type == PART_FREEBSD_ZFS)
849 				pref = PREF_FBSD;
850 			else
851 				pref = PREF_NONE;
852 		}
853 #endif /* LOADER_GPT_SUPPORT */
854 		if (pref < preflevel) {
855 			preflevel = pref;
856 			best = entry;
857 		}
858 	}
859 	if (best != NULL) {
860 		memcpy(part, &best->part, sizeof(*part));
861 		return (0);
862 	}
863 	return (ENOENT);
864 }
865 
866 int
867 ptable_iterate(const struct ptable *table, void *arg, ptable_iterate_t *iter)
868 {
869 	struct pentry *entry;
870 	char name[32];
871 	int ret = 0;
872 
873 	name[0] = '\0';
874 	STAILQ_FOREACH(entry, &table->entries, entry) {
875 #ifdef LOADER_MBR_SUPPORT
876 		if (table->type == PTABLE_MBR)
877 			sprintf(name, "s%d", entry->part.index);
878 		else
879 #endif
880 #ifdef LOADER_GPT_SUPPORT
881 		if (table->type == PTABLE_GPT)
882 			sprintf(name, "p%d", entry->part.index);
883 		else
884 #endif
885 #ifdef LOADER_VTOC8_SUPPORT
886 		if (table->type == PTABLE_VTOC8)
887 			sprintf(name, "%c", (uint8_t) 'a' +
888 			    entry->part.index);
889 		else
890 #endif
891 		if (table->type == PTABLE_BSD)
892 			sprintf(name, "%c", (uint8_t) 'a' +
893 			    entry->part.index);
894 		if ((ret = iter(arg, name, &entry->part)) != 0)
895 			return (ret);
896 	}
897 	return (ret);
898 }
899