xref: /freebsd/stand/common/part.c (revision 42249ef2)
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 <fs/cd9660/iso.h>
41 
42 #include <crc32.h>
43 #include <part.h>
44 #include <uuid.h>
45 
46 #ifdef PART_DEBUG
47 #define	DPRINTF(fmt, args...) printf("%s: " fmt "\n", __func__, ## args)
48 #else
49 #define	DPRINTF(fmt, args...)	((void)0)
50 #endif
51 
52 #ifdef LOADER_GPT_SUPPORT
53 #define	MAXTBLSZ	64
54 static const uuid_t gpt_uuid_unused = GPT_ENT_TYPE_UNUSED;
55 static const uuid_t gpt_uuid_ms_basic_data = GPT_ENT_TYPE_MS_BASIC_DATA;
56 static const uuid_t gpt_uuid_freebsd_ufs = GPT_ENT_TYPE_FREEBSD_UFS;
57 static const uuid_t gpt_uuid_efi = GPT_ENT_TYPE_EFI;
58 static const uuid_t gpt_uuid_freebsd = GPT_ENT_TYPE_FREEBSD;
59 static const uuid_t gpt_uuid_freebsd_boot = GPT_ENT_TYPE_FREEBSD_BOOT;
60 static const uuid_t gpt_uuid_freebsd_swap = GPT_ENT_TYPE_FREEBSD_SWAP;
61 static const uuid_t gpt_uuid_freebsd_zfs = GPT_ENT_TYPE_FREEBSD_ZFS;
62 static const uuid_t gpt_uuid_freebsd_vinum = GPT_ENT_TYPE_FREEBSD_VINUM;
63 #endif
64 
65 struct pentry {
66 	struct ptable_entry	part;
67 	uint64_t		flags;
68 	union {
69 		uint8_t bsd;
70 		uint8_t	mbr;
71 		uuid_t	gpt;
72 		uint16_t vtoc8;
73 	} type;
74 	STAILQ_ENTRY(pentry)	entry;
75 };
76 
77 struct ptable {
78 	enum ptable_type	type;
79 	uint16_t		sectorsize;
80 	uint64_t		sectors;
81 
82 	STAILQ_HEAD(, pentry)	entries;
83 };
84 
85 static struct parttypes {
86 	enum partition_type	type;
87 	const char		*desc;
88 } ptypes[] = {
89 	{ PART_UNKNOWN,		"Unknown" },
90 	{ PART_EFI,		"EFI" },
91 	{ PART_FREEBSD,		"FreeBSD" },
92 	{ PART_FREEBSD_BOOT,	"FreeBSD boot" },
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 	{ PART_ISO9660,		"ISO9660" },
101 };
102 
103 const char *
104 parttype2str(enum partition_type type)
105 {
106 	size_t i;
107 
108 	for (i = 0; i < nitems(ptypes); i++)
109 		if (ptypes[i].type == type)
110 			return (ptypes[i].desc);
111 	return (ptypes[0].desc);
112 }
113 
114 #ifdef LOADER_GPT_SUPPORT
115 static void
116 uuid_letoh(uuid_t *uuid)
117 {
118 
119 	uuid->time_low = le32toh(uuid->time_low);
120 	uuid->time_mid = le16toh(uuid->time_mid);
121 	uuid->time_hi_and_version = le16toh(uuid->time_hi_and_version);
122 }
123 
124 static enum partition_type
125 gpt_parttype(uuid_t type)
126 {
127 
128 	if (uuid_equal(&type, &gpt_uuid_efi, NULL))
129 		return (PART_EFI);
130 	else if (uuid_equal(&type, &gpt_uuid_ms_basic_data, NULL))
131 		return (PART_DOS);
132 	else if (uuid_equal(&type, &gpt_uuid_freebsd_boot, NULL))
133 		return (PART_FREEBSD_BOOT);
134 	else if (uuid_equal(&type, &gpt_uuid_freebsd_ufs, NULL))
135 		return (PART_FREEBSD_UFS);
136 	else if (uuid_equal(&type, &gpt_uuid_freebsd_zfs, NULL))
137 		return (PART_FREEBSD_ZFS);
138 	else if (uuid_equal(&type, &gpt_uuid_freebsd_swap, NULL))
139 		return (PART_FREEBSD_SWAP);
140 	else if (uuid_equal(&type, &gpt_uuid_freebsd_vinum, NULL))
141 		return (PART_FREEBSD_VINUM);
142 	else if (uuid_equal(&type, &gpt_uuid_freebsd, NULL))
143 		return (PART_FREEBSD);
144 	return (PART_UNKNOWN);
145 }
146 
147 static struct gpt_hdr *
148 gpt_checkhdr(struct gpt_hdr *hdr, uint64_t lba_self, uint64_t lba_last,
149     uint16_t sectorsize)
150 {
151 	uint32_t sz, crc;
152 
153 	if (memcmp(hdr->hdr_sig, GPT_HDR_SIG, sizeof(hdr->hdr_sig)) != 0) {
154 		DPRINTF("no GPT signature");
155 		return (NULL);
156 	}
157 	sz = le32toh(hdr->hdr_size);
158 	if (sz < 92 || sz > sectorsize) {
159 		DPRINTF("invalid GPT header size: %d", sz);
160 		return (NULL);
161 	}
162 	crc = le32toh(hdr->hdr_crc_self);
163 	hdr->hdr_crc_self = 0;
164 	if (crc32(hdr, sz) != crc) {
165 		DPRINTF("GPT header's CRC doesn't match");
166 		return (NULL);
167 	}
168 	hdr->hdr_crc_self = crc;
169 	hdr->hdr_revision = le32toh(hdr->hdr_revision);
170 	if (hdr->hdr_revision < GPT_HDR_REVISION) {
171 		DPRINTF("unsupported GPT revision %d", hdr->hdr_revision);
172 		return (NULL);
173 	}
174 	hdr->hdr_lba_self = le64toh(hdr->hdr_lba_self);
175 	if (hdr->hdr_lba_self != lba_self) {
176 		DPRINTF("self LBA doesn't match");
177 		return (NULL);
178 	}
179 	hdr->hdr_lba_alt = le64toh(hdr->hdr_lba_alt);
180 	if (hdr->hdr_lba_alt == hdr->hdr_lba_self) {
181 		DPRINTF("invalid alternate LBA");
182 		return (NULL);
183 	}
184 	hdr->hdr_entries = le32toh(hdr->hdr_entries);
185 	hdr->hdr_entsz = le32toh(hdr->hdr_entsz);
186 	if (hdr->hdr_entries == 0 ||
187 	    hdr->hdr_entsz < sizeof(struct gpt_ent) ||
188 	    sectorsize % hdr->hdr_entsz != 0) {
189 		DPRINTF("invalid entry size or number of entries");
190 		return (NULL);
191 	}
192 	hdr->hdr_lba_start = le64toh(hdr->hdr_lba_start);
193 	hdr->hdr_lba_end = le64toh(hdr->hdr_lba_end);
194 	hdr->hdr_lba_table = le64toh(hdr->hdr_lba_table);
195 	hdr->hdr_crc_table = le32toh(hdr->hdr_crc_table);
196 	uuid_letoh(&hdr->hdr_uuid);
197 	return (hdr);
198 }
199 
200 static int
201 gpt_checktbl(const struct gpt_hdr *hdr, uint8_t *tbl, size_t size,
202     uint64_t lba_last)
203 {
204 	struct gpt_ent *ent;
205 	uint32_t i, cnt;
206 
207 	cnt = size / hdr->hdr_entsz;
208 	if (hdr->hdr_entries <= cnt) {
209 		cnt = hdr->hdr_entries;
210 		/* Check CRC only when buffer size is enough for table. */
211 		if (hdr->hdr_crc_table !=
212 		    crc32(tbl, hdr->hdr_entries * hdr->hdr_entsz)) {
213 			DPRINTF("GPT table's CRC doesn't match");
214 			return (-1);
215 		}
216 	}
217 	for (i = 0; i < cnt; i++) {
218 		ent = (struct gpt_ent *)(tbl + i * hdr->hdr_entsz);
219 		uuid_letoh(&ent->ent_type);
220 		if (uuid_equal(&ent->ent_type, &gpt_uuid_unused, NULL))
221 			continue;
222 		ent->ent_lba_start = le64toh(ent->ent_lba_start);
223 		ent->ent_lba_end = le64toh(ent->ent_lba_end);
224 	}
225 	return (0);
226 }
227 
228 static struct ptable *
229 ptable_gptread(struct ptable *table, void *dev, diskread_t dread)
230 {
231 	struct pentry *entry;
232 	struct gpt_hdr *phdr, hdr;
233 	struct gpt_ent *ent;
234 	uint8_t *buf, *tbl;
235 	uint64_t offset;
236 	int pri, sec;
237 	size_t size, i;
238 
239 	buf = malloc(table->sectorsize);
240 	if (buf == NULL)
241 		return (NULL);
242 	tbl = malloc(table->sectorsize * MAXTBLSZ);
243 	if (tbl == NULL) {
244 		free(buf);
245 		return (NULL);
246 	}
247 	/* Read the primary GPT header. */
248 	if (dread(dev, buf, 1, 1) != 0) {
249 		ptable_close(table);
250 		table = NULL;
251 		goto out;
252 	}
253 	pri = sec = 0;
254 	/* Check the primary GPT header. */
255 	phdr = gpt_checkhdr((struct gpt_hdr *)buf, 1, table->sectors - 1,
256 	    table->sectorsize);
257 	if (phdr != NULL) {
258 		/* Read the primary GPT table. */
259 		size = MIN(MAXTBLSZ,
260 		    howmany(phdr->hdr_entries * phdr->hdr_entsz,
261 		        table->sectorsize));
262 		if (dread(dev, tbl, size, phdr->hdr_lba_table) == 0 &&
263 		    gpt_checktbl(phdr, tbl, size * table->sectorsize,
264 		    table->sectors - 1) == 0) {
265 			memcpy(&hdr, phdr, sizeof(hdr));
266 			pri = 1;
267 		}
268 	}
269 	offset = pri ? hdr.hdr_lba_alt: table->sectors - 1;
270 	/* Read the backup GPT header. */
271 	if (dread(dev, buf, 1, offset) != 0)
272 		phdr = NULL;
273 	else
274 		phdr = gpt_checkhdr((struct gpt_hdr *)buf, offset,
275 		    table->sectors - 1, table->sectorsize);
276 	if (phdr != NULL) {
277 		/*
278 		 * Compare primary and backup headers.
279 		 * If they are equal, then we do not need to read backup
280 		 * table. If they are different, then prefer backup header
281 		 * and try to read backup table.
282 		 */
283 		if (pri == 0 ||
284 		    uuid_equal(&hdr.hdr_uuid, &phdr->hdr_uuid, NULL) == 0 ||
285 		    hdr.hdr_revision != phdr->hdr_revision ||
286 		    hdr.hdr_size != phdr->hdr_size ||
287 		    hdr.hdr_lba_start != phdr->hdr_lba_start ||
288 		    hdr.hdr_lba_end != phdr->hdr_lba_end ||
289 		    hdr.hdr_entries != phdr->hdr_entries ||
290 		    hdr.hdr_entsz != phdr->hdr_entsz ||
291 		    hdr.hdr_crc_table != phdr->hdr_crc_table) {
292 			/* Read the backup GPT table. */
293 			size = MIN(MAXTBLSZ,
294 				   howmany(phdr->hdr_entries * phdr->hdr_entsz,
295 				       table->sectorsize));
296 			if (dread(dev, tbl, size, phdr->hdr_lba_table) == 0 &&
297 			    gpt_checktbl(phdr, tbl, size * table->sectorsize,
298 			    table->sectors - 1) == 0) {
299 				memcpy(&hdr, phdr, sizeof(hdr));
300 				sec = 1;
301 			}
302 		}
303 	}
304 	if (pri == 0 && sec == 0) {
305 		/* Both primary and backup tables are invalid. */
306 		table->type = PTABLE_NONE;
307 		goto out;
308 	}
309 	DPRINTF("GPT detected");
310 	size = MIN(hdr.hdr_entries * hdr.hdr_entsz,
311 	    MAXTBLSZ * table->sectorsize);
312 
313 	/*
314 	 * If the disk's sector count is smaller than the sector count recorded
315 	 * in the disk's GPT table header, set the table->sectors to the value
316 	 * recorded in GPT tables. This is done to work around buggy firmware
317 	 * that returns truncated disk sizes.
318 	 *
319 	 * Note, this is still not a foolproof way to get disk's size. For
320 	 * example, an image file can be truncated when copied to smaller media.
321 	 */
322 	table->sectors = hdr.hdr_lba_alt + 1;
323 
324 	for (i = 0; i < size / hdr.hdr_entsz; i++) {
325 		ent = (struct gpt_ent *)(tbl + i * hdr.hdr_entsz);
326 		if (uuid_equal(&ent->ent_type, &gpt_uuid_unused, NULL))
327 			continue;
328 
329 		/* Simple sanity checks. */
330 		if (ent->ent_lba_start < hdr.hdr_lba_start ||
331 		    ent->ent_lba_end > hdr.hdr_lba_end ||
332 		    ent->ent_lba_start > ent->ent_lba_end)
333 			continue;
334 
335 		entry = malloc(sizeof(*entry));
336 		if (entry == NULL)
337 			break;
338 		entry->part.start = ent->ent_lba_start;
339 		entry->part.end = ent->ent_lba_end;
340 		entry->part.index = i + 1;
341 		entry->part.type = gpt_parttype(ent->ent_type);
342 		entry->flags = le64toh(ent->ent_attr);
343 		memcpy(&entry->type.gpt, &ent->ent_type, sizeof(uuid_t));
344 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
345 		DPRINTF("new GPT partition added");
346 	}
347 out:
348 	free(buf);
349 	free(tbl);
350 	return (table);
351 }
352 #endif /* LOADER_GPT_SUPPORT */
353 
354 #ifdef LOADER_MBR_SUPPORT
355 /* We do not need to support too many EBR partitions in the loader */
356 #define	MAXEBRENTRIES		8
357 static enum partition_type
358 mbr_parttype(uint8_t type)
359 {
360 
361 	switch (type) {
362 	case DOSPTYP_386BSD:
363 		return (PART_FREEBSD);
364 	case DOSPTYP_LINSWP:
365 		return (PART_LINUX_SWAP);
366 	case DOSPTYP_LINUX:
367 		return (PART_LINUX);
368 	case 0x01:
369 	case 0x04:
370 	case 0x06:
371 	case 0x07:
372 	case 0x0b:
373 	case 0x0c:
374 	case 0x0e:
375 		return (PART_DOS);
376 	}
377 	return (PART_UNKNOWN);
378 }
379 
380 static struct ptable *
381 ptable_ebrread(struct ptable *table, void *dev, diskread_t dread)
382 {
383 	struct dos_partition *dp;
384 	struct pentry *e1, *entry;
385 	uint32_t start, end, offset;
386 	u_char *buf;
387 	int i, index;
388 
389 	STAILQ_FOREACH(e1, &table->entries, entry) {
390 		if (e1->type.mbr == DOSPTYP_EXT ||
391 		    e1->type.mbr == DOSPTYP_EXTLBA)
392 			break;
393 	}
394 	if (e1 == NULL)
395 		return (table);
396 	index = 5;
397 	offset = e1->part.start;
398 	buf = malloc(table->sectorsize);
399 	if (buf == NULL)
400 		return (table);
401 	DPRINTF("EBR detected");
402 	for (i = 0; i < MAXEBRENTRIES; i++) {
403 #if 0	/* Some BIOSes return an incorrect number of sectors */
404 		if (offset >= table->sectors)
405 			break;
406 #endif
407 		if (dread(dev, buf, 1, offset) != 0)
408 			break;
409 		dp = (struct dos_partition *)(buf + DOSPARTOFF);
410 		if (dp[0].dp_typ == 0)
411 			break;
412 		start = le32toh(dp[0].dp_start);
413 		if (dp[0].dp_typ == DOSPTYP_EXT &&
414 		    dp[1].dp_typ == 0) {
415 			offset = e1->part.start + start;
416 			continue;
417 		}
418 		end = le32toh(dp[0].dp_size);
419 		entry = malloc(sizeof(*entry));
420 		if (entry == NULL)
421 			break;
422 		entry->part.start = offset + start;
423 		entry->part.end = entry->part.start + end - 1;
424 		entry->part.index = index++;
425 		entry->part.type = mbr_parttype(dp[0].dp_typ);
426 		entry->flags = dp[0].dp_flag;
427 		entry->type.mbr = dp[0].dp_typ;
428 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
429 		DPRINTF("new EBR partition added");
430 		if (dp[1].dp_typ == 0)
431 			break;
432 		offset = e1->part.start + le32toh(dp[1].dp_start);
433 	}
434 	free(buf);
435 	return (table);
436 }
437 #endif /* LOADER_MBR_SUPPORT */
438 
439 static enum partition_type
440 bsd_parttype(uint8_t type)
441 {
442 
443 	switch (type) {
444 	case FS_SWAP:
445 		return (PART_FREEBSD_SWAP);
446 	case FS_BSDFFS:
447 		return (PART_FREEBSD_UFS);
448 	case FS_VINUM:
449 		return (PART_FREEBSD_VINUM);
450 	case FS_ZFS:
451 		return (PART_FREEBSD_ZFS);
452 	}
453 	return (PART_UNKNOWN);
454 }
455 
456 static struct ptable *
457 ptable_bsdread(struct ptable *table, void *dev, diskread_t dread)
458 {
459 	struct disklabel *dl;
460 	struct partition *part;
461 	struct pentry *entry;
462 	uint8_t *buf;
463 	uint32_t raw_offset;
464 	int i;
465 
466 	if (table->sectorsize < sizeof(struct disklabel)) {
467 		DPRINTF("Too small sectorsize");
468 		return (table);
469 	}
470 	buf = malloc(table->sectorsize);
471 	if (buf == NULL)
472 		return (table);
473 	if (dread(dev, buf, 1, 1) != 0) {
474 		DPRINTF("read failed");
475 		ptable_close(table);
476 		table = NULL;
477 		goto out;
478 	}
479 	dl = (struct disklabel *)buf;
480 	if (le32toh(dl->d_magic) != DISKMAGIC &&
481 	    le32toh(dl->d_magic2) != DISKMAGIC)
482 		goto out;
483 	if (le32toh(dl->d_secsize) != table->sectorsize) {
484 		DPRINTF("unsupported sector size");
485 		goto out;
486 	}
487 	dl->d_npartitions = le16toh(dl->d_npartitions);
488 	if (dl->d_npartitions > 20 || dl->d_npartitions < 8) {
489 		DPRINTF("invalid number of partitions");
490 		goto out;
491 	}
492 	DPRINTF("BSD detected");
493 	part = &dl->d_partitions[0];
494 	raw_offset = le32toh(part[RAW_PART].p_offset);
495 	for (i = 0; i < dl->d_npartitions; i++, part++) {
496 		if (i == RAW_PART)
497 			continue;
498 		if (part->p_size == 0)
499 			continue;
500 		entry = malloc(sizeof(*entry));
501 		if (entry == NULL)
502 			break;
503 		entry->part.start = le32toh(part->p_offset) - raw_offset;
504 		entry->part.end = entry->part.start +
505 		    le32toh(part->p_size) - 1;
506 		entry->part.type = bsd_parttype(part->p_fstype);
507 		entry->part.index = i; /* starts from zero */
508 		entry->type.bsd = part->p_fstype;
509 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
510 		DPRINTF("new BSD partition added");
511 	}
512 	table->type = PTABLE_BSD;
513 out:
514 	free(buf);
515 	return (table);
516 }
517 
518 #ifdef LOADER_VTOC8_SUPPORT
519 static enum partition_type
520 vtoc8_parttype(uint16_t type)
521 {
522 
523 	switch (type) {
524 	case VTOC_TAG_FREEBSD_SWAP:
525 		return (PART_FREEBSD_SWAP);
526 	case VTOC_TAG_FREEBSD_UFS:
527 		return (PART_FREEBSD_UFS);
528 	case VTOC_TAG_FREEBSD_VINUM:
529 		return (PART_FREEBSD_VINUM);
530 	case VTOC_TAG_FREEBSD_ZFS:
531 		return (PART_FREEBSD_ZFS);
532 	}
533 	return (PART_UNKNOWN);
534 }
535 
536 static struct ptable *
537 ptable_vtoc8read(struct ptable *table, void *dev, diskread_t dread)
538 {
539 	struct pentry *entry;
540 	struct vtoc8 *dl;
541 	uint8_t *buf;
542 	uint16_t sum, heads, sectors;
543 	int i;
544 
545 	if (table->sectorsize != sizeof(struct vtoc8))
546 		return (table);
547 	buf = malloc(table->sectorsize);
548 	if (buf == NULL)
549 		return (table);
550 	if (dread(dev, buf, 1, 0) != 0) {
551 		DPRINTF("read failed");
552 		ptable_close(table);
553 		table = NULL;
554 		goto out;
555 	}
556 	dl = (struct vtoc8 *)buf;
557 	/* Check the sum */
558 	for (i = sum = 0; i < sizeof(struct vtoc8); i += sizeof(sum))
559 		sum ^= be16dec(buf + i);
560 	if (sum != 0) {
561 		DPRINTF("incorrect checksum");
562 		goto out;
563 	}
564 	if (be16toh(dl->nparts) != VTOC8_NPARTS) {
565 		DPRINTF("invalid number of entries");
566 		goto out;
567 	}
568 	sectors = be16toh(dl->nsecs);
569 	heads = be16toh(dl->nheads);
570 	if (sectors * heads == 0) {
571 		DPRINTF("invalid geometry");
572 		goto out;
573 	}
574 	DPRINTF("VTOC8 detected");
575 	for (i = 0; i < VTOC8_NPARTS; i++) {
576 		dl->part[i].tag = be16toh(dl->part[i].tag);
577 		if (i == VTOC_RAW_PART ||
578 		    dl->part[i].tag == VTOC_TAG_UNASSIGNED)
579 			continue;
580 		entry = malloc(sizeof(*entry));
581 		if (entry == NULL)
582 			break;
583 		entry->part.start = be32toh(dl->map[i].cyl) * heads * sectors;
584 		entry->part.end = be32toh(dl->map[i].nblks) +
585 		    entry->part.start - 1;
586 		entry->part.type = vtoc8_parttype(dl->part[i].tag);
587 		entry->part.index = i; /* starts from zero */
588 		entry->type.vtoc8 = dl->part[i].tag;
589 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
590 		DPRINTF("new VTOC8 partition added");
591 	}
592 	table->type = PTABLE_VTOC8;
593 out:
594 	free(buf);
595 	return (table);
596 
597 }
598 #endif /* LOADER_VTOC8_SUPPORT */
599 
600 #define cdb2devb(bno)   ((bno) * ISO_DEFAULT_BLOCK_SIZE / table->sectorsize)
601 
602 static struct ptable *
603 ptable_iso9660read(struct ptable *table, void *dev, diskread_t dread)
604 {
605 	uint8_t *buf;
606 	struct iso_primary_descriptor *vd;
607 	struct pentry *entry;
608 
609 	buf = malloc(table->sectorsize);
610 	if (buf == NULL)
611 		return (table);
612 
613 	if (dread(dev, buf, 1, cdb2devb(16)) != 0) {
614 		DPRINTF("read failed");
615 		ptable_close(table);
616 		table = NULL;
617 		goto out;
618 	}
619 	vd = (struct iso_primary_descriptor *)buf;
620 	if (bcmp(vd->id, ISO_STANDARD_ID, sizeof vd->id) != 0)
621 		goto out;
622 
623 	entry = malloc(sizeof(*entry));
624 	if (entry == NULL)
625 		goto out;
626 	entry->part.start = 0;
627 	entry->part.end = table->sectors;
628 	entry->part.type = PART_ISO9660;
629 	entry->part.index = 0;
630 	STAILQ_INSERT_TAIL(&table->entries, entry, entry);
631 
632 	table->type = PTABLE_ISO9660;
633 
634 out:
635 	free(buf);
636 	return (table);
637 }
638 
639 struct ptable *
640 ptable_open(void *dev, uint64_t sectors, uint16_t sectorsize,
641     diskread_t *dread)
642 {
643 	struct dos_partition *dp;
644 	struct ptable *table;
645 	uint8_t *buf;
646 	int i, count;
647 #ifdef LOADER_MBR_SUPPORT
648 	struct pentry *entry;
649 	uint32_t start, end;
650 	int has_ext;
651 #endif
652 	table = NULL;
653 	buf = malloc(sectorsize);
654 	if (buf == NULL)
655 		return (NULL);
656 	/* First, read the MBR. */
657 	if (dread(dev, buf, 1, DOSBBSECTOR) != 0) {
658 		DPRINTF("read failed");
659 		goto out;
660 	}
661 
662 	table = malloc(sizeof(*table));
663 	if (table == NULL)
664 		goto out;
665 	table->sectors = sectors;
666 	table->sectorsize = sectorsize;
667 	table->type = PTABLE_NONE;
668 	STAILQ_INIT(&table->entries);
669 
670 	if (ptable_iso9660read(table, dev, dread) == NULL) {
671 		/* Read error. */
672 		table = NULL;
673 		goto out;
674 	} else if (table->type == PTABLE_ISO9660)
675 		goto out;
676 
677 #ifdef LOADER_VTOC8_SUPPORT
678 	if (be16dec(buf + offsetof(struct vtoc8, magic)) == VTOC_MAGIC) {
679 		if (ptable_vtoc8read(table, dev, dread) == NULL) {
680 			/* Read error. */
681 			table = NULL;
682 			goto out;
683 		} else if (table->type == PTABLE_VTOC8)
684 			goto out;
685 	}
686 #endif
687 	/* Check the BSD label. */
688 	if (ptable_bsdread(table, dev, dread) == NULL) { /* Read error. */
689 		table = NULL;
690 		goto out;
691 	} else if (table->type == PTABLE_BSD)
692 		goto out;
693 
694 #if defined(LOADER_GPT_SUPPORT) || defined(LOADER_MBR_SUPPORT)
695 	/* Check the MBR magic. */
696 	if (buf[DOSMAGICOFFSET] != 0x55 ||
697 	    buf[DOSMAGICOFFSET + 1] != 0xaa) {
698 		DPRINTF("magic sequence not found");
699 #if defined(LOADER_GPT_SUPPORT)
700 		/* There is no PMBR, check that we have backup GPT */
701 		table->type = PTABLE_GPT;
702 		table = ptable_gptread(table, dev, dread);
703 #endif
704 		goto out;
705 	}
706 	/* Check that we have PMBR. Also do some validation. */
707 	dp = (struct dos_partition *)(buf + DOSPARTOFF);
708 	for (i = 0, count = 0; i < NDOSPART; i++) {
709 		if (dp[i].dp_flag != 0 && dp[i].dp_flag != 0x80) {
710 			DPRINTF("invalid partition flag %x", dp[i].dp_flag);
711 			goto out;
712 		}
713 #ifdef LOADER_GPT_SUPPORT
714 		if (dp[i].dp_typ == DOSPTYP_PMBR) {
715 			table->type = PTABLE_GPT;
716 			DPRINTF("PMBR detected");
717 		}
718 #endif
719 		if (dp[i].dp_typ != 0)
720 			count++;
721 	}
722 	/* Do we have some invalid values? */
723 	if (table->type == PTABLE_GPT && count > 1) {
724 		if (dp[1].dp_typ != DOSPTYP_HFS) {
725 			table->type = PTABLE_NONE;
726 			DPRINTF("Incorrect PMBR, ignore it");
727 		} else {
728 			DPRINTF("Bootcamp detected");
729 		}
730 	}
731 #ifdef LOADER_GPT_SUPPORT
732 	if (table->type == PTABLE_GPT) {
733 		table = ptable_gptread(table, dev, dread);
734 		goto out;
735 	}
736 #endif
737 #ifdef LOADER_MBR_SUPPORT
738 	/* Read MBR. */
739 	DPRINTF("MBR detected");
740 	table->type = PTABLE_MBR;
741 	for (i = has_ext = 0; i < NDOSPART; i++) {
742 		if (dp[i].dp_typ == 0)
743 			continue;
744 		start = le32dec(&(dp[i].dp_start));
745 		end = le32dec(&(dp[i].dp_size));
746 		if (start == 0 || end == 0)
747 			continue;
748 #if 0	/* Some BIOSes return an incorrect number of sectors */
749 		if (start + end - 1 >= sectors)
750 			continue;	/* XXX: ignore */
751 #endif
752 		if (dp[i].dp_typ == DOSPTYP_EXT ||
753 		    dp[i].dp_typ == DOSPTYP_EXTLBA)
754 			has_ext = 1;
755 		entry = malloc(sizeof(*entry));
756 		if (entry == NULL)
757 			break;
758 		entry->part.start = start;
759 		entry->part.end = start + end - 1;
760 		entry->part.index = i + 1;
761 		entry->part.type = mbr_parttype(dp[i].dp_typ);
762 		entry->flags = dp[i].dp_flag;
763 		entry->type.mbr = dp[i].dp_typ;
764 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
765 		DPRINTF("new MBR partition added");
766 	}
767 	if (has_ext) {
768 		table = ptable_ebrread(table, dev, dread);
769 		/* FALLTHROUGH */
770 	}
771 #endif /* LOADER_MBR_SUPPORT */
772 #endif /* LOADER_MBR_SUPPORT || LOADER_GPT_SUPPORT */
773 out:
774 	free(buf);
775 	return (table);
776 }
777 
778 void
779 ptable_close(struct ptable *table)
780 {
781 	struct pentry *entry;
782 
783 	if (table == NULL)
784 		return;
785 
786 	while (!STAILQ_EMPTY(&table->entries)) {
787 		entry = STAILQ_FIRST(&table->entries);
788 		STAILQ_REMOVE_HEAD(&table->entries, entry);
789 		free(entry);
790 	}
791 	free(table);
792 }
793 
794 enum ptable_type
795 ptable_gettype(const struct ptable *table)
796 {
797 
798 	return (table->type);
799 }
800 
801 int
802 ptable_getsize(const struct ptable *table, uint64_t *sizep)
803 {
804 	uint64_t tmp = table->sectors * table->sectorsize;
805 
806 	if (tmp < table->sectors)
807 		return (EOVERFLOW);
808 
809 	if (sizep != NULL)
810 		*sizep = tmp;
811 	return (0);
812 }
813 
814 int
815 ptable_getpart(const struct ptable *table, struct ptable_entry *part, int index)
816 {
817 	struct pentry *entry;
818 
819 	if (part == NULL || table == NULL)
820 		return (EINVAL);
821 
822 	STAILQ_FOREACH(entry, &table->entries, entry) {
823 		if (entry->part.index != index)
824 			continue;
825 		memcpy(part, &entry->part, sizeof(*part));
826 		return (0);
827 	}
828 	return (ENOENT);
829 }
830 
831 /*
832  * Search for a slice with the following preferences:
833  *
834  * 1: Active FreeBSD slice
835  * 2: Non-active FreeBSD slice
836  * 3: Active Linux slice
837  * 4: non-active Linux slice
838  * 5: Active FAT/FAT32 slice
839  * 6: non-active FAT/FAT32 slice
840  */
841 #define	PREF_RAWDISK	0
842 #define	PREF_FBSD_ACT	1
843 #define	PREF_FBSD	2
844 #define	PREF_LINUX_ACT	3
845 #define	PREF_LINUX	4
846 #define	PREF_DOS_ACT	5
847 #define	PREF_DOS	6
848 #define	PREF_NONE	7
849 int
850 ptable_getbestpart(const struct ptable *table, struct ptable_entry *part)
851 {
852 	struct pentry *entry, *best;
853 	int pref, preflevel;
854 
855 	if (part == NULL || table == NULL)
856 		return (EINVAL);
857 
858 	best = NULL;
859 	preflevel = pref = PREF_NONE;
860 	STAILQ_FOREACH(entry, &table->entries, entry) {
861 #ifdef LOADER_MBR_SUPPORT
862 		if (table->type == PTABLE_MBR) {
863 			switch (entry->type.mbr) {
864 			case DOSPTYP_386BSD:
865 				pref = entry->flags & 0x80 ? PREF_FBSD_ACT:
866 				    PREF_FBSD;
867 				break;
868 			case DOSPTYP_LINUX:
869 				pref = entry->flags & 0x80 ? PREF_LINUX_ACT:
870 				    PREF_LINUX;
871 				break;
872 			case 0x01:		/* DOS/Windows */
873 			case 0x04:
874 			case 0x06:
875 			case 0x0c:
876 			case 0x0e:
877 			case DOSPTYP_FAT32:
878 				pref = entry->flags & 0x80 ? PREF_DOS_ACT:
879 				    PREF_DOS;
880 				break;
881 			default:
882 				pref = PREF_NONE;
883 			}
884 		}
885 #endif /* LOADER_MBR_SUPPORT */
886 #ifdef LOADER_GPT_SUPPORT
887 		if (table->type == PTABLE_GPT) {
888 			if (entry->part.type == PART_DOS)
889 				pref = PREF_DOS;
890 			else if (entry->part.type == PART_FREEBSD_UFS ||
891 			    entry->part.type == PART_FREEBSD_ZFS)
892 				pref = PREF_FBSD;
893 			else
894 				pref = PREF_NONE;
895 		}
896 #endif /* LOADER_GPT_SUPPORT */
897 		if (pref < preflevel) {
898 			preflevel = pref;
899 			best = entry;
900 		}
901 	}
902 	if (best != NULL) {
903 		memcpy(part, &best->part, sizeof(*part));
904 		return (0);
905 	}
906 	return (ENOENT);
907 }
908 
909 int
910 ptable_iterate(const struct ptable *table, void *arg, ptable_iterate_t *iter)
911 {
912 	struct pentry *entry;
913 	char name[32];
914 	int ret = 0;
915 
916 	name[0] = '\0';
917 	STAILQ_FOREACH(entry, &table->entries, entry) {
918 #ifdef LOADER_MBR_SUPPORT
919 		if (table->type == PTABLE_MBR)
920 			sprintf(name, "s%d", entry->part.index);
921 		else
922 #endif
923 #ifdef LOADER_GPT_SUPPORT
924 		if (table->type == PTABLE_GPT)
925 			sprintf(name, "p%d", entry->part.index);
926 		else
927 #endif
928 #ifdef LOADER_VTOC8_SUPPORT
929 		if (table->type == PTABLE_VTOC8)
930 			sprintf(name, "%c", (uint8_t) 'a' +
931 			    entry->part.index);
932 		else
933 #endif
934 		if (table->type == PTABLE_BSD)
935 			sprintf(name, "%c", (uint8_t) 'a' +
936 			    entry->part.index);
937 		if ((ret = iter(arg, name, &entry->part)) != 0)
938 			return (ret);
939 	}
940 	return (ret);
941 }
942