xref: /freebsd/stand/common/part.c (revision 76f2606181eabc8ab20aa7297bbddc6e78bb549d)
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	DEBUG(fmt, args...) printf("%s: " fmt "\n", __func__, ## args)
48 #else
49 #define	DEBUG(fmt, args...)
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_nandfs = GPT_ENT_TYPE_FREEBSD_NANDFS;
61 static const uuid_t gpt_uuid_freebsd_swap = GPT_ENT_TYPE_FREEBSD_SWAP;
62 static const uuid_t gpt_uuid_freebsd_zfs = GPT_ENT_TYPE_FREEBSD_ZFS;
63 static const uuid_t gpt_uuid_freebsd_vinum = GPT_ENT_TYPE_FREEBSD_VINUM;
64 #endif
65 
66 struct pentry {
67 	struct ptable_entry	part;
68 	uint64_t		flags;
69 	union {
70 		uint8_t bsd;
71 		uint8_t	mbr;
72 		uuid_t	gpt;
73 		uint16_t vtoc8;
74 	} type;
75 	STAILQ_ENTRY(pentry)	entry;
76 };
77 
78 struct ptable {
79 	enum ptable_type	type;
80 	uint16_t		sectorsize;
81 	uint64_t		sectors;
82 
83 	STAILQ_HEAD(, pentry)	entries;
84 };
85 
86 static struct parttypes {
87 	enum partition_type	type;
88 	const char		*desc;
89 } ptypes[] = {
90 	{ PART_UNKNOWN,		"Unknown" },
91 	{ PART_EFI,		"EFI" },
92 	{ PART_FREEBSD,		"FreeBSD" },
93 	{ PART_FREEBSD_BOOT,	"FreeBSD boot" },
94 	{ PART_FREEBSD_NANDFS,	"FreeBSD nandfs" },
95 	{ PART_FREEBSD_UFS,	"FreeBSD UFS" },
96 	{ PART_FREEBSD_ZFS,	"FreeBSD ZFS" },
97 	{ PART_FREEBSD_SWAP,	"FreeBSD swap" },
98 	{ PART_FREEBSD_VINUM,	"FreeBSD vinum" },
99 	{ PART_LINUX,		"Linux" },
100 	{ PART_LINUX_SWAP,	"Linux swap" },
101 	{ PART_DOS,		"DOS/Windows" },
102 	{ PART_ISO9660,		"ISO9660" },
103 };
104 
105 const char *
106 parttype2str(enum partition_type type)
107 {
108 	size_t i;
109 
110 	for (i = 0; i < nitems(ptypes); i++)
111 		if (ptypes[i].type == type)
112 			return (ptypes[i].desc);
113 	return (ptypes[0].desc);
114 }
115 
116 #ifdef LOADER_GPT_SUPPORT
117 static void
118 uuid_letoh(uuid_t *uuid)
119 {
120 
121 	uuid->time_low = le32toh(uuid->time_low);
122 	uuid->time_mid = le16toh(uuid->time_mid);
123 	uuid->time_hi_and_version = le16toh(uuid->time_hi_and_version);
124 }
125 
126 static enum partition_type
127 gpt_parttype(uuid_t type)
128 {
129 
130 	if (uuid_equal(&type, &gpt_uuid_efi, NULL))
131 		return (PART_EFI);
132 	else if (uuid_equal(&type, &gpt_uuid_ms_basic_data, NULL))
133 		return (PART_DOS);
134 	else if (uuid_equal(&type, &gpt_uuid_freebsd_boot, NULL))
135 		return (PART_FREEBSD_BOOT);
136 	else if (uuid_equal(&type, &gpt_uuid_freebsd_ufs, NULL))
137 		return (PART_FREEBSD_UFS);
138 	else if (uuid_equal(&type, &gpt_uuid_freebsd_zfs, NULL))
139 		return (PART_FREEBSD_ZFS);
140 	else if (uuid_equal(&type, &gpt_uuid_freebsd_swap, NULL))
141 		return (PART_FREEBSD_SWAP);
142 	else if (uuid_equal(&type, &gpt_uuid_freebsd_vinum, NULL))
143 		return (PART_FREEBSD_VINUM);
144 	else if (uuid_equal(&type, &gpt_uuid_freebsd_nandfs, NULL))
145 		return (PART_FREEBSD_NANDFS);
146 	else if (uuid_equal(&type, &gpt_uuid_freebsd, NULL))
147 		return (PART_FREEBSD);
148 	return (PART_UNKNOWN);
149 }
150 
151 static struct gpt_hdr *
152 gpt_checkhdr(struct gpt_hdr *hdr, uint64_t lba_self, uint64_t lba_last,
153     uint16_t sectorsize)
154 {
155 	uint32_t sz, crc;
156 
157 	if (memcmp(hdr->hdr_sig, GPT_HDR_SIG, sizeof(hdr->hdr_sig)) != 0) {
158 		DEBUG("no GPT signature");
159 		return (NULL);
160 	}
161 	sz = le32toh(hdr->hdr_size);
162 	if (sz < 92 || sz > sectorsize) {
163 		DEBUG("invalid GPT header size: %d", sz);
164 		return (NULL);
165 	}
166 	crc = le32toh(hdr->hdr_crc_self);
167 	hdr->hdr_crc_self = 0;
168 	if (crc32(hdr, sz) != crc) {
169 		DEBUG("GPT header's CRC doesn't match");
170 		return (NULL);
171 	}
172 	hdr->hdr_crc_self = crc;
173 	hdr->hdr_revision = le32toh(hdr->hdr_revision);
174 	if (hdr->hdr_revision < GPT_HDR_REVISION) {
175 		DEBUG("unsupported GPT revision %d", hdr->hdr_revision);
176 		return (NULL);
177 	}
178 	hdr->hdr_lba_self = le64toh(hdr->hdr_lba_self);
179 	if (hdr->hdr_lba_self != lba_self) {
180 		DEBUG("self LBA doesn't match");
181 		return (NULL);
182 	}
183 	hdr->hdr_lba_alt = le64toh(hdr->hdr_lba_alt);
184 	if (hdr->hdr_lba_alt == hdr->hdr_lba_self) {
185 		DEBUG("invalid alternate LBA");
186 		return (NULL);
187 	}
188 	hdr->hdr_entries = le32toh(hdr->hdr_entries);
189 	hdr->hdr_entsz = le32toh(hdr->hdr_entsz);
190 	if (hdr->hdr_entries == 0 ||
191 	    hdr->hdr_entsz < sizeof(struct gpt_ent) ||
192 	    sectorsize % hdr->hdr_entsz != 0) {
193 		DEBUG("invalid entry size or number of entries");
194 		return (NULL);
195 	}
196 	hdr->hdr_lba_start = le64toh(hdr->hdr_lba_start);
197 	hdr->hdr_lba_end = le64toh(hdr->hdr_lba_end);
198 	hdr->hdr_lba_table = le64toh(hdr->hdr_lba_table);
199 	hdr->hdr_crc_table = le32toh(hdr->hdr_crc_table);
200 	uuid_letoh(&hdr->hdr_uuid);
201 	return (hdr);
202 }
203 
204 static int
205 gpt_checktbl(const struct gpt_hdr *hdr, uint8_t *tbl, size_t size,
206     uint64_t lba_last)
207 {
208 	struct gpt_ent *ent;
209 	uint32_t i, cnt;
210 
211 	cnt = size / hdr->hdr_entsz;
212 	if (hdr->hdr_entries <= cnt) {
213 		cnt = hdr->hdr_entries;
214 		/* Check CRC only when buffer size is enough for table. */
215 		if (hdr->hdr_crc_table !=
216 		    crc32(tbl, hdr->hdr_entries * hdr->hdr_entsz)) {
217 			DEBUG("GPT table's CRC doesn't match");
218 			return (-1);
219 		}
220 	}
221 	for (i = 0; i < cnt; i++) {
222 		ent = (struct gpt_ent *)(tbl + i * hdr->hdr_entsz);
223 		uuid_letoh(&ent->ent_type);
224 		if (uuid_equal(&ent->ent_type, &gpt_uuid_unused, NULL))
225 			continue;
226 		ent->ent_lba_start = le64toh(ent->ent_lba_start);
227 		ent->ent_lba_end = le64toh(ent->ent_lba_end);
228 	}
229 	return (0);
230 }
231 
232 static struct ptable *
233 ptable_gptread(struct ptable *table, void *dev, diskread_t dread)
234 {
235 	struct pentry *entry;
236 	struct gpt_hdr *phdr, hdr;
237 	struct gpt_ent *ent;
238 	uint8_t *buf, *tbl;
239 	uint64_t offset;
240 	int pri, sec;
241 	size_t size, i;
242 
243 	buf = malloc(table->sectorsize);
244 	if (buf == NULL)
245 		return (NULL);
246 	tbl = malloc(table->sectorsize * MAXTBLSZ);
247 	if (tbl == NULL) {
248 		free(buf);
249 		return (NULL);
250 	}
251 	/* Read the primary GPT header. */
252 	if (dread(dev, buf, 1, 1) != 0) {
253 		ptable_close(table);
254 		table = NULL;
255 		goto out;
256 	}
257 	pri = sec = 0;
258 	/* Check the primary GPT header. */
259 	phdr = gpt_checkhdr((struct gpt_hdr *)buf, 1, table->sectors - 1,
260 	    table->sectorsize);
261 	if (phdr != NULL) {
262 		/* Read the primary GPT table. */
263 		size = MIN(MAXTBLSZ,
264 		    howmany(phdr->hdr_entries * phdr->hdr_entsz,
265 		        table->sectorsize));
266 		if (dread(dev, tbl, size, phdr->hdr_lba_table) == 0 &&
267 		    gpt_checktbl(phdr, tbl, size * table->sectorsize,
268 		    table->sectors - 1) == 0) {
269 			memcpy(&hdr, phdr, sizeof(hdr));
270 			pri = 1;
271 		}
272 	}
273 	offset = pri ? hdr.hdr_lba_alt: table->sectors - 1;
274 	/* Read the backup GPT header. */
275 	if (dread(dev, buf, 1, offset) != 0)
276 		phdr = NULL;
277 	else
278 		phdr = gpt_checkhdr((struct gpt_hdr *)buf, offset,
279 		    table->sectors - 1, table->sectorsize);
280 	if (phdr != NULL) {
281 		/*
282 		 * Compare primary and backup headers.
283 		 * If they are equal, then we do not need to read backup
284 		 * table. If they are different, then prefer backup header
285 		 * and try to read backup table.
286 		 */
287 		if (pri == 0 ||
288 		    uuid_equal(&hdr.hdr_uuid, &phdr->hdr_uuid, NULL) == 0 ||
289 		    hdr.hdr_revision != phdr->hdr_revision ||
290 		    hdr.hdr_size != phdr->hdr_size ||
291 		    hdr.hdr_lba_start != phdr->hdr_lba_start ||
292 		    hdr.hdr_lba_end != phdr->hdr_lba_end ||
293 		    hdr.hdr_entries != phdr->hdr_entries ||
294 		    hdr.hdr_entsz != phdr->hdr_entsz ||
295 		    hdr.hdr_crc_table != phdr->hdr_crc_table) {
296 			/* Read the backup GPT table. */
297 			size = MIN(MAXTBLSZ,
298 				   howmany(phdr->hdr_entries * phdr->hdr_entsz,
299 				       table->sectorsize));
300 			if (dread(dev, tbl, size, phdr->hdr_lba_table) == 0 &&
301 			    gpt_checktbl(phdr, tbl, size * table->sectorsize,
302 			    table->sectors - 1) == 0) {
303 				memcpy(&hdr, phdr, sizeof(hdr));
304 				sec = 1;
305 			}
306 		}
307 	}
308 	if (pri == 0 && sec == 0) {
309 		/* Both primary and backup tables are invalid. */
310 		table->type = PTABLE_NONE;
311 		goto out;
312 	}
313 	DEBUG("GPT detected");
314 	size = MIN(hdr.hdr_entries * hdr.hdr_entsz,
315 	    MAXTBLSZ * table->sectorsize);
316 
317 	/*
318 	 * If the disk's sector count is smaller than the sector count recorded
319 	 * in the disk's GPT table header, set the table->sectors to the value
320 	 * recorded in GPT tables. This is done to work around buggy firmware
321 	 * that returns truncated disk sizes.
322 	 *
323 	 * Note, this is still not a foolproof way to get disk's size. For
324 	 * example, an image file can be truncated when copied to smaller media.
325 	 */
326 	if (hdr.hdr_lba_alt + 1 > table->sectors)
327 		table->sectors = hdr.hdr_lba_alt + 1;
328 
329 	for (i = 0; i < size / hdr.hdr_entsz; i++) {
330 		ent = (struct gpt_ent *)(tbl + i * hdr.hdr_entsz);
331 		if (uuid_equal(&ent->ent_type, &gpt_uuid_unused, NULL))
332 			continue;
333 
334 		/* Simple sanity checks. */
335 		if (ent->ent_lba_start < hdr.hdr_lba_start ||
336 		    ent->ent_lba_end > hdr.hdr_lba_end ||
337 		    ent->ent_lba_start > ent->ent_lba_end)
338 			continue;
339 
340 		entry = malloc(sizeof(*entry));
341 		if (entry == NULL)
342 			break;
343 		entry->part.start = ent->ent_lba_start;
344 		entry->part.end = ent->ent_lba_end;
345 		entry->part.index = i + 1;
346 		entry->part.type = gpt_parttype(ent->ent_type);
347 		entry->flags = le64toh(ent->ent_attr);
348 		memcpy(&entry->type.gpt, &ent->ent_type, sizeof(uuid_t));
349 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
350 		DEBUG("new GPT partition added");
351 	}
352 out:
353 	free(buf);
354 	free(tbl);
355 	return (table);
356 }
357 #endif /* LOADER_GPT_SUPPORT */
358 
359 #ifdef LOADER_MBR_SUPPORT
360 /* We do not need to support too many EBR partitions in the loader */
361 #define	MAXEBRENTRIES		8
362 static enum partition_type
363 mbr_parttype(uint8_t type)
364 {
365 
366 	switch (type) {
367 	case DOSPTYP_386BSD:
368 		return (PART_FREEBSD);
369 	case DOSPTYP_LINSWP:
370 		return (PART_LINUX_SWAP);
371 	case DOSPTYP_LINUX:
372 		return (PART_LINUX);
373 	case 0x01:
374 	case 0x04:
375 	case 0x06:
376 	case 0x07:
377 	case 0x0b:
378 	case 0x0c:
379 	case 0x0e:
380 		return (PART_DOS);
381 	}
382 	return (PART_UNKNOWN);
383 }
384 
385 static struct ptable *
386 ptable_ebrread(struct ptable *table, void *dev, diskread_t dread)
387 {
388 	struct dos_partition *dp;
389 	struct pentry *e1, *entry;
390 	uint32_t start, end, offset;
391 	u_char *buf;
392 	int i, index;
393 
394 	STAILQ_FOREACH(e1, &table->entries, entry) {
395 		if (e1->type.mbr == DOSPTYP_EXT ||
396 		    e1->type.mbr == DOSPTYP_EXTLBA)
397 			break;
398 	}
399 	if (e1 == NULL)
400 		return (table);
401 	index = 5;
402 	offset = e1->part.start;
403 	buf = malloc(table->sectorsize);
404 	if (buf == NULL)
405 		return (table);
406 	DEBUG("EBR detected");
407 	for (i = 0; i < MAXEBRENTRIES; i++) {
408 #if 0	/* Some BIOSes return an incorrect number of sectors */
409 		if (offset >= table->sectors)
410 			break;
411 #endif
412 		if (dread(dev, buf, 1, offset) != 0)
413 			break;
414 		dp = (struct dos_partition *)(buf + DOSPARTOFF);
415 		if (dp[0].dp_typ == 0)
416 			break;
417 		start = le32toh(dp[0].dp_start);
418 		if (dp[0].dp_typ == DOSPTYP_EXT &&
419 		    dp[1].dp_typ == 0) {
420 			offset = e1->part.start + start;
421 			continue;
422 		}
423 		end = le32toh(dp[0].dp_size);
424 		entry = malloc(sizeof(*entry));
425 		if (entry == NULL)
426 			break;
427 		entry->part.start = offset + start;
428 		entry->part.end = entry->part.start + end - 1;
429 		entry->part.index = index++;
430 		entry->part.type = mbr_parttype(dp[0].dp_typ);
431 		entry->flags = dp[0].dp_flag;
432 		entry->type.mbr = dp[0].dp_typ;
433 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
434 		DEBUG("new EBR partition added");
435 		if (dp[1].dp_typ == 0)
436 			break;
437 		offset = e1->part.start + le32toh(dp[1].dp_start);
438 	}
439 	free(buf);
440 	return (table);
441 }
442 #endif /* LOADER_MBR_SUPPORT */
443 
444 static enum partition_type
445 bsd_parttype(uint8_t type)
446 {
447 
448 	switch (type) {
449 	case FS_NANDFS:
450 		return (PART_FREEBSD_NANDFS);
451 	case FS_SWAP:
452 		return (PART_FREEBSD_SWAP);
453 	case FS_BSDFFS:
454 		return (PART_FREEBSD_UFS);
455 	case FS_VINUM:
456 		return (PART_FREEBSD_VINUM);
457 	case FS_ZFS:
458 		return (PART_FREEBSD_ZFS);
459 	}
460 	return (PART_UNKNOWN);
461 }
462 
463 static struct ptable *
464 ptable_bsdread(struct ptable *table, void *dev, diskread_t dread)
465 {
466 	struct disklabel *dl;
467 	struct partition *part;
468 	struct pentry *entry;
469 	uint8_t *buf;
470 	uint32_t raw_offset;
471 	int i;
472 
473 	if (table->sectorsize < sizeof(struct disklabel)) {
474 		DEBUG("Too small sectorsize");
475 		return (table);
476 	}
477 	buf = malloc(table->sectorsize);
478 	if (buf == NULL)
479 		return (table);
480 	if (dread(dev, buf, 1, 1) != 0) {
481 		DEBUG("read failed");
482 		ptable_close(table);
483 		table = NULL;
484 		goto out;
485 	}
486 	dl = (struct disklabel *)buf;
487 	if (le32toh(dl->d_magic) != DISKMAGIC &&
488 	    le32toh(dl->d_magic2) != DISKMAGIC)
489 		goto out;
490 	if (le32toh(dl->d_secsize) != table->sectorsize) {
491 		DEBUG("unsupported sector size");
492 		goto out;
493 	}
494 	dl->d_npartitions = le16toh(dl->d_npartitions);
495 	if (dl->d_npartitions > 20 || dl->d_npartitions < 8) {
496 		DEBUG("invalid number of partitions");
497 		goto out;
498 	}
499 	DEBUG("BSD detected");
500 	part = &dl->d_partitions[0];
501 	raw_offset = le32toh(part[RAW_PART].p_offset);
502 	for (i = 0; i < dl->d_npartitions; i++, part++) {
503 		if (i == RAW_PART)
504 			continue;
505 		if (part->p_size == 0)
506 			continue;
507 		entry = malloc(sizeof(*entry));
508 		if (entry == NULL)
509 			break;
510 		entry->part.start = le32toh(part->p_offset) - raw_offset;
511 		entry->part.end = entry->part.start +
512 		    le32toh(part->p_size) - 1;
513 		entry->part.type = bsd_parttype(part->p_fstype);
514 		entry->part.index = i; /* starts from zero */
515 		entry->type.bsd = part->p_fstype;
516 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
517 		DEBUG("new BSD partition added");
518 	}
519 	table->type = PTABLE_BSD;
520 out:
521 	free(buf);
522 	return (table);
523 }
524 
525 #ifdef LOADER_VTOC8_SUPPORT
526 static enum partition_type
527 vtoc8_parttype(uint16_t type)
528 {
529 
530 	switch (type) {
531 	case VTOC_TAG_FREEBSD_NANDFS:
532 		return (PART_FREEBSD_NANDFS);
533 	case VTOC_TAG_FREEBSD_SWAP:
534 		return (PART_FREEBSD_SWAP);
535 	case VTOC_TAG_FREEBSD_UFS:
536 		return (PART_FREEBSD_UFS);
537 	case VTOC_TAG_FREEBSD_VINUM:
538 		return (PART_FREEBSD_VINUM);
539 	case VTOC_TAG_FREEBSD_ZFS:
540 		return (PART_FREEBSD_ZFS);
541 	}
542 	return (PART_UNKNOWN);
543 }
544 
545 static struct ptable *
546 ptable_vtoc8read(struct ptable *table, void *dev, diskread_t dread)
547 {
548 	struct pentry *entry;
549 	struct vtoc8 *dl;
550 	uint8_t *buf;
551 	uint16_t sum, heads, sectors;
552 	int i;
553 
554 	if (table->sectorsize != sizeof(struct vtoc8))
555 		return (table);
556 	buf = malloc(table->sectorsize);
557 	if (buf == NULL)
558 		return (table);
559 	if (dread(dev, buf, 1, 0) != 0) {
560 		DEBUG("read failed");
561 		ptable_close(table);
562 		table = NULL;
563 		goto out;
564 	}
565 	dl = (struct vtoc8 *)buf;
566 	/* Check the sum */
567 	for (i = sum = 0; i < sizeof(struct vtoc8); i += sizeof(sum))
568 		sum ^= be16dec(buf + i);
569 	if (sum != 0) {
570 		DEBUG("incorrect checksum");
571 		goto out;
572 	}
573 	if (be16toh(dl->nparts) != VTOC8_NPARTS) {
574 		DEBUG("invalid number of entries");
575 		goto out;
576 	}
577 	sectors = be16toh(dl->nsecs);
578 	heads = be16toh(dl->nheads);
579 	if (sectors * heads == 0) {
580 		DEBUG("invalid geometry");
581 		goto out;
582 	}
583 	DEBUG("VTOC8 detected");
584 	for (i = 0; i < VTOC8_NPARTS; i++) {
585 		dl->part[i].tag = be16toh(dl->part[i].tag);
586 		if (i == VTOC_RAW_PART ||
587 		    dl->part[i].tag == VTOC_TAG_UNASSIGNED)
588 			continue;
589 		entry = malloc(sizeof(*entry));
590 		if (entry == NULL)
591 			break;
592 		entry->part.start = be32toh(dl->map[i].cyl) * heads * sectors;
593 		entry->part.end = be32toh(dl->map[i].nblks) +
594 		    entry->part.start - 1;
595 		entry->part.type = vtoc8_parttype(dl->part[i].tag);
596 		entry->part.index = i; /* starts from zero */
597 		entry->type.vtoc8 = dl->part[i].tag;
598 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
599 		DEBUG("new VTOC8 partition added");
600 	}
601 	table->type = PTABLE_VTOC8;
602 out:
603 	free(buf);
604 	return (table);
605 
606 }
607 #endif /* LOADER_VTOC8_SUPPORT */
608 
609 #define cdb2devb(bno)   ((bno) * ISO_DEFAULT_BLOCK_SIZE / table->sectorsize)
610 
611 static struct ptable *
612 ptable_iso9660read(struct ptable *table, void *dev, diskread_t dread)
613 {
614 	uint8_t *buf;
615 	struct iso_primary_descriptor *vd;
616 	struct pentry *entry;
617 
618 	buf = malloc(table->sectorsize);
619 	if (buf == NULL)
620 		return (table);
621 
622 	if (dread(dev, buf, 1, cdb2devb(16)) != 0) {
623 		DEBUG("read failed");
624 		ptable_close(table);
625 		table = NULL;
626 		goto out;
627 	}
628 	vd = (struct iso_primary_descriptor *)buf;
629 	if (bcmp(vd->id, ISO_STANDARD_ID, sizeof vd->id) != 0)
630 		goto out;
631 
632 	entry = malloc(sizeof(*entry));
633 	if (entry == NULL)
634 		goto out;
635 	entry->part.start = 0;
636 	entry->part.end = table->sectors;
637 	entry->part.type = PART_ISO9660;
638 	entry->part.index = 0;
639 	STAILQ_INSERT_TAIL(&table->entries, entry, entry);
640 
641 	table->type = PTABLE_ISO9660;
642 
643 out:
644 	free(buf);
645 	return (table);
646 }
647 
648 struct ptable *
649 ptable_open(void *dev, uint64_t sectors, uint16_t sectorsize,
650     diskread_t *dread)
651 {
652 	struct dos_partition *dp;
653 	struct ptable *table;
654 	uint8_t *buf;
655 	int i, count;
656 #ifdef LOADER_MBR_SUPPORT
657 	struct pentry *entry;
658 	uint32_t start, end;
659 	int has_ext;
660 #endif
661 	table = NULL;
662 	buf = malloc(sectorsize);
663 	if (buf == NULL)
664 		return (NULL);
665 	/* First, read the MBR. */
666 	if (dread(dev, buf, 1, DOSBBSECTOR) != 0) {
667 		DEBUG("read failed");
668 		goto out;
669 	}
670 
671 	table = malloc(sizeof(*table));
672 	if (table == NULL)
673 		goto out;
674 	table->sectors = sectors;
675 	table->sectorsize = sectorsize;
676 	table->type = PTABLE_NONE;
677 	STAILQ_INIT(&table->entries);
678 
679 	if (ptable_iso9660read(table, dev, dread) != NULL) {
680 		if (table->type == PTABLE_ISO9660)
681 			goto out;
682 	}
683 
684 #ifdef LOADER_VTOC8_SUPPORT
685 	if (be16dec(buf + offsetof(struct vtoc8, magic)) == VTOC_MAGIC) {
686 		if (ptable_vtoc8read(table, dev, dread) == NULL) {
687 			/* Read error. */
688 			table = NULL;
689 			goto out;
690 		} else if (table->type == PTABLE_VTOC8)
691 			goto out;
692 	}
693 #endif
694 	/* Check the BSD label. */
695 	if (ptable_bsdread(table, dev, dread) == NULL) { /* Read error. */
696 		table = NULL;
697 		goto out;
698 	} else if (table->type == PTABLE_BSD)
699 		goto out;
700 
701 #if defined(LOADER_GPT_SUPPORT) || defined(LOADER_MBR_SUPPORT)
702 	/* Check the MBR magic. */
703 	if (buf[DOSMAGICOFFSET] != 0x55 ||
704 	    buf[DOSMAGICOFFSET + 1] != 0xaa) {
705 		DEBUG("magic sequence not found");
706 #if defined(LOADER_GPT_SUPPORT)
707 		/* There is no PMBR, check that we have backup GPT */
708 		table->type = PTABLE_GPT;
709 		table = ptable_gptread(table, dev, dread);
710 #endif
711 		goto out;
712 	}
713 	/* Check that we have PMBR. Also do some validation. */
714 	dp = (struct dos_partition *)(buf + DOSPARTOFF);
715 	for (i = 0, count = 0; i < NDOSPART; i++) {
716 		if (dp[i].dp_flag != 0 && dp[i].dp_flag != 0x80) {
717 			DEBUG("invalid partition flag %x", dp[i].dp_flag);
718 			goto out;
719 		}
720 #ifdef LOADER_GPT_SUPPORT
721 		if (dp[i].dp_typ == DOSPTYP_PMBR) {
722 			table->type = PTABLE_GPT;
723 			DEBUG("PMBR detected");
724 		}
725 #endif
726 		if (dp[i].dp_typ != 0)
727 			count++;
728 	}
729 	/* Do we have some invalid values? */
730 	if (table->type == PTABLE_GPT && count > 1) {
731 		if (dp[1].dp_typ != DOSPTYP_HFS) {
732 			table->type = PTABLE_NONE;
733 			DEBUG("Incorrect PMBR, ignore it");
734 		} else {
735 			DEBUG("Bootcamp detected");
736 		}
737 	}
738 #ifdef LOADER_GPT_SUPPORT
739 	if (table->type == PTABLE_GPT) {
740 		table = ptable_gptread(table, dev, dread);
741 		goto out;
742 	}
743 #endif
744 #ifdef LOADER_MBR_SUPPORT
745 	/* Read MBR. */
746 	DEBUG("MBR detected");
747 	table->type = PTABLE_MBR;
748 	for (i = has_ext = 0; i < NDOSPART; i++) {
749 		if (dp[i].dp_typ == 0)
750 			continue;
751 		start = le32dec(&(dp[i].dp_start));
752 		end = le32dec(&(dp[i].dp_size));
753 		if (start == 0 || end == 0)
754 			continue;
755 #if 0	/* Some BIOSes return an incorrect number of sectors */
756 		if (start + end - 1 >= sectors)
757 			continue;	/* XXX: ignore */
758 #endif
759 		if (dp[i].dp_typ == DOSPTYP_EXT ||
760 		    dp[i].dp_typ == DOSPTYP_EXTLBA)
761 			has_ext = 1;
762 		entry = malloc(sizeof(*entry));
763 		if (entry == NULL)
764 			break;
765 		entry->part.start = start;
766 		entry->part.end = start + end - 1;
767 		entry->part.index = i + 1;
768 		entry->part.type = mbr_parttype(dp[i].dp_typ);
769 		entry->flags = dp[i].dp_flag;
770 		entry->type.mbr = dp[i].dp_typ;
771 		STAILQ_INSERT_TAIL(&table->entries, entry, entry);
772 		DEBUG("new MBR partition added");
773 	}
774 	if (has_ext) {
775 		table = ptable_ebrread(table, dev, dread);
776 		/* FALLTHROUGH */
777 	}
778 #endif /* LOADER_MBR_SUPPORT */
779 #endif /* LOADER_MBR_SUPPORT || LOADER_GPT_SUPPORT */
780 out:
781 	free(buf);
782 	return (table);
783 }
784 
785 void
786 ptable_close(struct ptable *table)
787 {
788 	struct pentry *entry;
789 
790 	while (!STAILQ_EMPTY(&table->entries)) {
791 		entry = STAILQ_FIRST(&table->entries);
792 		STAILQ_REMOVE_HEAD(&table->entries, entry);
793 		free(entry);
794 	}
795 	free(table);
796 }
797 
798 enum ptable_type
799 ptable_gettype(const struct ptable *table)
800 {
801 
802 	return (table->type);
803 }
804 
805 int
806 ptable_getsize(const struct ptable *table, uint64_t *sizep)
807 {
808 	uint64_t tmp = table->sectors * table->sectorsize;
809 
810 	if (tmp < table->sectors)
811 		return (EOVERFLOW);
812 
813 	if (sizep != NULL)
814 		*sizep = tmp;
815 	return (0);
816 }
817 
818 int
819 ptable_getpart(const struct ptable *table, struct ptable_entry *part, int index)
820 {
821 	struct pentry *entry;
822 
823 	if (part == NULL || table == NULL)
824 		return (EINVAL);
825 
826 	STAILQ_FOREACH(entry, &table->entries, entry) {
827 		if (entry->part.index != index)
828 			continue;
829 		memcpy(part, &entry->part, sizeof(*part));
830 		return (0);
831 	}
832 	return (ENOENT);
833 }
834 
835 /*
836  * Search for a slice with the following preferences:
837  *
838  * 1: Active FreeBSD slice
839  * 2: Non-active FreeBSD slice
840  * 3: Active Linux slice
841  * 4: non-active Linux slice
842  * 5: Active FAT/FAT32 slice
843  * 6: non-active FAT/FAT32 slice
844  */
845 #define	PREF_RAWDISK	0
846 #define	PREF_FBSD_ACT	1
847 #define	PREF_FBSD	2
848 #define	PREF_LINUX_ACT	3
849 #define	PREF_LINUX	4
850 #define	PREF_DOS_ACT	5
851 #define	PREF_DOS	6
852 #define	PREF_NONE	7
853 int
854 ptable_getbestpart(const struct ptable *table, struct ptable_entry *part)
855 {
856 	struct pentry *entry, *best;
857 	int pref, preflevel;
858 
859 	if (part == NULL || table == NULL)
860 		return (EINVAL);
861 
862 	best = NULL;
863 	preflevel = pref = PREF_NONE;
864 	STAILQ_FOREACH(entry, &table->entries, entry) {
865 #ifdef LOADER_MBR_SUPPORT
866 		if (table->type == PTABLE_MBR) {
867 			switch (entry->type.mbr) {
868 			case DOSPTYP_386BSD:
869 				pref = entry->flags & 0x80 ? PREF_FBSD_ACT:
870 				    PREF_FBSD;
871 				break;
872 			case DOSPTYP_LINUX:
873 				pref = entry->flags & 0x80 ? PREF_LINUX_ACT:
874 				    PREF_LINUX;
875 				break;
876 			case 0x01:		/* DOS/Windows */
877 			case 0x04:
878 			case 0x06:
879 			case 0x0c:
880 			case 0x0e:
881 			case DOSPTYP_FAT32:
882 				pref = entry->flags & 0x80 ? PREF_DOS_ACT:
883 				    PREF_DOS;
884 				break;
885 			default:
886 				pref = PREF_NONE;
887 			}
888 		}
889 #endif /* LOADER_MBR_SUPPORT */
890 #ifdef LOADER_GPT_SUPPORT
891 		if (table->type == PTABLE_GPT) {
892 			if (entry->part.type == PART_DOS)
893 				pref = PREF_DOS;
894 			else if (entry->part.type == PART_FREEBSD_UFS ||
895 			    entry->part.type == PART_FREEBSD_ZFS)
896 				pref = PREF_FBSD;
897 			else
898 				pref = PREF_NONE;
899 		}
900 #endif /* LOADER_GPT_SUPPORT */
901 		if (pref < preflevel) {
902 			preflevel = pref;
903 			best = entry;
904 		}
905 	}
906 	if (best != NULL) {
907 		memcpy(part, &best->part, sizeof(*part));
908 		return (0);
909 	}
910 	return (ENOENT);
911 }
912 
913 int
914 ptable_iterate(const struct ptable *table, void *arg, ptable_iterate_t *iter)
915 {
916 	struct pentry *entry;
917 	char name[32];
918 	int ret = 0;
919 
920 	name[0] = '\0';
921 	STAILQ_FOREACH(entry, &table->entries, entry) {
922 #ifdef LOADER_MBR_SUPPORT
923 		if (table->type == PTABLE_MBR)
924 			sprintf(name, "s%d", entry->part.index);
925 		else
926 #endif
927 #ifdef LOADER_GPT_SUPPORT
928 		if (table->type == PTABLE_GPT)
929 			sprintf(name, "p%d", entry->part.index);
930 		else
931 #endif
932 #ifdef LOADER_VTOC8_SUPPORT
933 		if (table->type == PTABLE_VTOC8)
934 			sprintf(name, "%c", (uint8_t) 'a' +
935 			    entry->part.index);
936 		else
937 #endif
938 		if (table->type == PTABLE_BSD)
939 			sprintf(name, "%c", (uint8_t) 'a' +
940 			    entry->part.index);
941 		if ((ret = iter(arg, name, &entry->part)) != 0)
942 			return (ret);
943 	}
944 	return (ret);
945 }
946