[LogFS] Move assertion
[linux-2.6.git] / fs / partitions / efi.c
blob49cfd5f5423894df4c2f52ac48bac1f82a9b43a3
1 /************************************************************
2 * EFI GUID Partition Table handling
4 * http://www.uefi.org/specs/
5 * http://www.intel.com/technology/efi/
7 * efi.[ch] by Matt Domsch <Matt_Domsch@dell.com>
8 * Copyright 2000,2001,2002,2004 Dell Inc.
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License as published by
12 * the Free Software Foundation; either version 2 of the License, or
13 * (at your option) any later version.
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
20 * You should have received a copy of the GNU General Public License
21 * along with this program; if not, write to the Free Software
22 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
25 * TODO:
27 * Changelog:
28 * Mon Nov 09 2004 Matt Domsch <Matt_Domsch@dell.com>
29 * - test for valid PMBR and valid PGPT before ever reading
30 * AGPT, allow override with 'gpt' kernel command line option.
31 * - check for first/last_usable_lba outside of size of disk
33 * Tue Mar 26 2002 Matt Domsch <Matt_Domsch@dell.com>
34 * - Ported to 2.5.7-pre1 and 2.5.7-dj2
35 * - Applied patch to avoid fault in alternate header handling
36 * - cleaned up find_valid_gpt
37 * - On-disk structure and copy in memory is *always* LE now -
38 * swab fields as needed
39 * - remove print_gpt_header()
40 * - only use first max_p partition entries, to keep the kernel minor number
41 * and partition numbers tied.
43 * Mon Feb 04 2002 Matt Domsch <Matt_Domsch@dell.com>
44 * - Removed __PRIPTR_PREFIX - not being used
46 * Mon Jan 14 2002 Matt Domsch <Matt_Domsch@dell.com>
47 * - Ported to 2.5.2-pre11 + library crc32 patch Linus applied
49 * Thu Dec 6 2001 Matt Domsch <Matt_Domsch@dell.com>
50 * - Added compare_gpts().
51 * - moved le_efi_guid_to_cpus() back into this file. GPT is the only
52 * thing that keeps EFI GUIDs on disk.
53 * - Changed gpt structure names and members to be simpler and more Linux-like.
55 * Wed Oct 17 2001 Matt Domsch <Matt_Domsch@dell.com>
56 * - Removed CONFIG_DEVFS_VOLUMES_UUID code entirely per Martin Wilck
58 * Wed Oct 10 2001 Matt Domsch <Matt_Domsch@dell.com>
59 * - Changed function comments to DocBook style per Andreas Dilger suggestion.
61 * Mon Oct 08 2001 Matt Domsch <Matt_Domsch@dell.com>
62 * - Change read_lba() to use the page cache per Al Viro's work.
63 * - print u64s properly on all architectures
64 * - fixed debug_printk(), now Dprintk()
66 * Mon Oct 01 2001 Matt Domsch <Matt_Domsch@dell.com>
67 * - Style cleanups
68 * - made most functions static
69 * - Endianness addition
70 * - remove test for second alternate header, as it's not per spec,
71 * and is unnecessary. There's now a method to read/write the last
72 * sector of an odd-sized disk from user space. No tools have ever
73 * been released which used this code, so it's effectively dead.
74 * - Per Asit Mallick of Intel, added a test for a valid PMBR.
75 * - Added kernel command line option 'gpt' to override valid PMBR test.
77 * Wed Jun 6 2001 Martin Wilck <Martin.Wilck@Fujitsu-Siemens.com>
78 * - added devfs volume UUID support (/dev/volumes/uuids) for
79 * mounting file systems by the partition GUID.
81 * Tue Dec 5 2000 Matt Domsch <Matt_Domsch@dell.com>
82 * - Moved crc32() to linux/lib, added efi_crc32().
84 * Thu Nov 30 2000 Matt Domsch <Matt_Domsch@dell.com>
85 * - Replaced Intel's CRC32 function with an equivalent
86 * non-license-restricted version.
88 * Wed Oct 25 2000 Matt Domsch <Matt_Domsch@dell.com>
89 * - Fixed the last_lba() call to return the proper last block
91 * Thu Oct 12 2000 Matt Domsch <Matt_Domsch@dell.com>
92 * - Thanks to Andries Brouwer for his debugging assistance.
93 * - Code works, detects all the partitions.
95 ************************************************************/
96 #include <linux/crc32.h>
97 #include <linux/math64.h>
98 #include "check.h"
99 #include "efi.h"
101 /* This allows a kernel command line option 'gpt' to override
102 * the test for invalid PMBR. Not __initdata because reloading
103 * the partition tables happens after init too.
105 static int force_gpt;
106 static int __init
107 force_gpt_fn(char *str)
109 force_gpt = 1;
110 return 1;
112 __setup("gpt", force_gpt_fn);
116 * efi_crc32() - EFI version of crc32 function
117 * @buf: buffer to calculate crc32 of
118 * @len - length of buf
120 * Description: Returns EFI-style CRC32 value for @buf
122 * This function uses the little endian Ethernet polynomial
123 * but seeds the function with ~0, and xor's with ~0 at the end.
124 * Note, the EFI Specification, v1.02, has a reference to
125 * Dr. Dobbs Journal, May 1994 (actually it's in May 1992).
127 static inline u32
128 efi_crc32(const void *buf, unsigned long len)
130 return (crc32(~0L, buf, len) ^ ~0L);
134 * last_lba(): return number of last logical block of device
135 * @bdev: block device
137 * Description: Returns last LBA value on success, 0 on error.
138 * This is stored (by sd and ide-geometry) in
139 * the part[0] entry for this disk, and is the number of
140 * physical sectors available on the disk.
142 static u64
143 last_lba(struct block_device *bdev)
145 if (!bdev || !bdev->bd_inode)
146 return 0;
147 return div_u64(bdev->bd_inode->i_size,
148 bdev_logical_block_size(bdev)) - 1ULL;
151 static inline int
152 pmbr_part_valid(struct partition *part)
154 if (part->sys_ind == EFI_PMBR_OSTYPE_EFI_GPT &&
155 le32_to_cpu(part->start_sect) == 1UL)
156 return 1;
157 return 0;
161 * is_pmbr_valid(): test Protective MBR for validity
162 * @mbr: pointer to a legacy mbr structure
164 * Description: Returns 1 if PMBR is valid, 0 otherwise.
165 * Validity depends on two things:
166 * 1) MSDOS signature is in the last two bytes of the MBR
167 * 2) One partition of type 0xEE is found
169 static int
170 is_pmbr_valid(legacy_mbr *mbr)
172 int i;
173 if (!mbr || le16_to_cpu(mbr->signature) != MSDOS_MBR_SIGNATURE)
174 return 0;
175 for (i = 0; i < 4; i++)
176 if (pmbr_part_valid(&mbr->partition_record[i]))
177 return 1;
178 return 0;
182 * read_lba(): Read bytes from disk, starting at given LBA
183 * @bdev
184 * @lba
185 * @buffer
186 * @size_t
188 * Description: Reads @count bytes from @bdev into @buffer.
189 * Returns number of bytes read on success, 0 on error.
191 static size_t
192 read_lba(struct block_device *bdev, u64 lba, u8 * buffer, size_t count)
194 size_t totalreadcount = 0;
195 sector_t n = lba * (bdev_logical_block_size(bdev) / 512);
197 if (!bdev || !buffer || lba > last_lba(bdev))
198 return 0;
200 while (count) {
201 int copied = 512;
202 Sector sect;
203 unsigned char *data = read_dev_sector(bdev, n++, &sect);
204 if (!data)
205 break;
206 if (copied > count)
207 copied = count;
208 memcpy(buffer, data, copied);
209 put_dev_sector(sect);
210 buffer += copied;
211 totalreadcount +=copied;
212 count -= copied;
214 return totalreadcount;
218 * alloc_read_gpt_entries(): reads partition entries from disk
219 * @bdev
220 * @gpt - GPT header
222 * Description: Returns ptes on success, NULL on error.
223 * Allocates space for PTEs based on information found in @gpt.
224 * Notes: remember to free pte when you're done!
226 static gpt_entry *
227 alloc_read_gpt_entries(struct block_device *bdev, gpt_header *gpt)
229 size_t count;
230 gpt_entry *pte;
231 if (!bdev || !gpt)
232 return NULL;
234 count = le32_to_cpu(gpt->num_partition_entries) *
235 le32_to_cpu(gpt->sizeof_partition_entry);
236 if (!count)
237 return NULL;
238 pte = kzalloc(count, GFP_KERNEL);
239 if (!pte)
240 return NULL;
242 if (read_lba(bdev, le64_to_cpu(gpt->partition_entry_lba),
243 (u8 *) pte,
244 count) < count) {
245 kfree(pte);
246 pte=NULL;
247 return NULL;
249 return pte;
253 * alloc_read_gpt_header(): Allocates GPT header, reads into it from disk
254 * @bdev
255 * @lba is the Logical Block Address of the partition table
257 * Description: returns GPT header on success, NULL on error. Allocates
258 * and fills a GPT header starting at @ from @bdev.
259 * Note: remember to free gpt when finished with it.
261 static gpt_header *
262 alloc_read_gpt_header(struct block_device *bdev, u64 lba)
264 gpt_header *gpt;
265 unsigned ssz = bdev_logical_block_size(bdev);
267 if (!bdev)
268 return NULL;
270 gpt = kzalloc(ssz, GFP_KERNEL);
271 if (!gpt)
272 return NULL;
274 if (read_lba(bdev, lba, (u8 *) gpt, ssz) < ssz) {
275 kfree(gpt);
276 gpt=NULL;
277 return NULL;
280 return gpt;
284 * is_gpt_valid() - tests one GPT header and PTEs for validity
285 * @bdev
286 * @lba is the logical block address of the GPT header to test
287 * @gpt is a GPT header ptr, filled on return.
288 * @ptes is a PTEs ptr, filled on return.
290 * Description: returns 1 if valid, 0 on error.
291 * If valid, returns pointers to newly allocated GPT header and PTEs.
293 static int
294 is_gpt_valid(struct block_device *bdev, u64 lba,
295 gpt_header **gpt, gpt_entry **ptes)
297 u32 crc, origcrc;
298 u64 lastlba;
300 if (!bdev || !gpt || !ptes)
301 return 0;
302 if (!(*gpt = alloc_read_gpt_header(bdev, lba)))
303 return 0;
305 /* Check the GUID Partition Table signature */
306 if (le64_to_cpu((*gpt)->signature) != GPT_HEADER_SIGNATURE) {
307 pr_debug("GUID Partition Table Header signature is wrong:"
308 "%lld != %lld\n",
309 (unsigned long long)le64_to_cpu((*gpt)->signature),
310 (unsigned long long)GPT_HEADER_SIGNATURE);
311 goto fail;
314 /* Check the GUID Partition Table CRC */
315 origcrc = le32_to_cpu((*gpt)->header_crc32);
316 (*gpt)->header_crc32 = 0;
317 crc = efi_crc32((const unsigned char *) (*gpt), le32_to_cpu((*gpt)->header_size));
319 if (crc != origcrc) {
320 pr_debug("GUID Partition Table Header CRC is wrong: %x != %x\n",
321 crc, origcrc);
322 goto fail;
324 (*gpt)->header_crc32 = cpu_to_le32(origcrc);
326 /* Check that the my_lba entry points to the LBA that contains
327 * the GUID Partition Table */
328 if (le64_to_cpu((*gpt)->my_lba) != lba) {
329 pr_debug("GPT my_lba incorrect: %lld != %lld\n",
330 (unsigned long long)le64_to_cpu((*gpt)->my_lba),
331 (unsigned long long)lba);
332 goto fail;
335 /* Check the first_usable_lba and last_usable_lba are
336 * within the disk.
338 lastlba = last_lba(bdev);
339 if (le64_to_cpu((*gpt)->first_usable_lba) > lastlba) {
340 pr_debug("GPT: first_usable_lba incorrect: %lld > %lld\n",
341 (unsigned long long)le64_to_cpu((*gpt)->first_usable_lba),
342 (unsigned long long)lastlba);
343 goto fail;
345 if (le64_to_cpu((*gpt)->last_usable_lba) > lastlba) {
346 pr_debug("GPT: last_usable_lba incorrect: %lld > %lld\n",
347 (unsigned long long)le64_to_cpu((*gpt)->last_usable_lba),
348 (unsigned long long)lastlba);
349 goto fail;
352 if (!(*ptes = alloc_read_gpt_entries(bdev, *gpt)))
353 goto fail;
355 /* Check the GUID Partition Entry Array CRC */
356 crc = efi_crc32((const unsigned char *) (*ptes),
357 le32_to_cpu((*gpt)->num_partition_entries) *
358 le32_to_cpu((*gpt)->sizeof_partition_entry));
360 if (crc != le32_to_cpu((*gpt)->partition_entry_array_crc32)) {
361 pr_debug("GUID Partitition Entry Array CRC check failed.\n");
362 goto fail_ptes;
365 /* We're done, all's well */
366 return 1;
368 fail_ptes:
369 kfree(*ptes);
370 *ptes = NULL;
371 fail:
372 kfree(*gpt);
373 *gpt = NULL;
374 return 0;
378 * is_pte_valid() - tests one PTE for validity
379 * @pte is the pte to check
380 * @lastlba is last lba of the disk
382 * Description: returns 1 if valid, 0 on error.
384 static inline int
385 is_pte_valid(const gpt_entry *pte, const u64 lastlba)
387 if ((!efi_guidcmp(pte->partition_type_guid, NULL_GUID)) ||
388 le64_to_cpu(pte->starting_lba) > lastlba ||
389 le64_to_cpu(pte->ending_lba) > lastlba)
390 return 0;
391 return 1;
395 * compare_gpts() - Search disk for valid GPT headers and PTEs
396 * @pgpt is the primary GPT header
397 * @agpt is the alternate GPT header
398 * @lastlba is the last LBA number
399 * Description: Returns nothing. Sanity checks pgpt and agpt fields
400 * and prints warnings on discrepancies.
403 static void
404 compare_gpts(gpt_header *pgpt, gpt_header *agpt, u64 lastlba)
406 int error_found = 0;
407 if (!pgpt || !agpt)
408 return;
409 if (le64_to_cpu(pgpt->my_lba) != le64_to_cpu(agpt->alternate_lba)) {
410 printk(KERN_WARNING
411 "GPT:Primary header LBA != Alt. header alternate_lba\n");
412 printk(KERN_WARNING "GPT:%lld != %lld\n",
413 (unsigned long long)le64_to_cpu(pgpt->my_lba),
414 (unsigned long long)le64_to_cpu(agpt->alternate_lba));
415 error_found++;
417 if (le64_to_cpu(pgpt->alternate_lba) != le64_to_cpu(agpt->my_lba)) {
418 printk(KERN_WARNING
419 "GPT:Primary header alternate_lba != Alt. header my_lba\n");
420 printk(KERN_WARNING "GPT:%lld != %lld\n",
421 (unsigned long long)le64_to_cpu(pgpt->alternate_lba),
422 (unsigned long long)le64_to_cpu(agpt->my_lba));
423 error_found++;
425 if (le64_to_cpu(pgpt->first_usable_lba) !=
426 le64_to_cpu(agpt->first_usable_lba)) {
427 printk(KERN_WARNING "GPT:first_usable_lbas don't match.\n");
428 printk(KERN_WARNING "GPT:%lld != %lld\n",
429 (unsigned long long)le64_to_cpu(pgpt->first_usable_lba),
430 (unsigned long long)le64_to_cpu(agpt->first_usable_lba));
431 error_found++;
433 if (le64_to_cpu(pgpt->last_usable_lba) !=
434 le64_to_cpu(agpt->last_usable_lba)) {
435 printk(KERN_WARNING "GPT:last_usable_lbas don't match.\n");
436 printk(KERN_WARNING "GPT:%lld != %lld\n",
437 (unsigned long long)le64_to_cpu(pgpt->last_usable_lba),
438 (unsigned long long)le64_to_cpu(agpt->last_usable_lba));
439 error_found++;
441 if (efi_guidcmp(pgpt->disk_guid, agpt->disk_guid)) {
442 printk(KERN_WARNING "GPT:disk_guids don't match.\n");
443 error_found++;
445 if (le32_to_cpu(pgpt->num_partition_entries) !=
446 le32_to_cpu(agpt->num_partition_entries)) {
447 printk(KERN_WARNING "GPT:num_partition_entries don't match: "
448 "0x%x != 0x%x\n",
449 le32_to_cpu(pgpt->num_partition_entries),
450 le32_to_cpu(agpt->num_partition_entries));
451 error_found++;
453 if (le32_to_cpu(pgpt->sizeof_partition_entry) !=
454 le32_to_cpu(agpt->sizeof_partition_entry)) {
455 printk(KERN_WARNING
456 "GPT:sizeof_partition_entry values don't match: "
457 "0x%x != 0x%x\n",
458 le32_to_cpu(pgpt->sizeof_partition_entry),
459 le32_to_cpu(agpt->sizeof_partition_entry));
460 error_found++;
462 if (le32_to_cpu(pgpt->partition_entry_array_crc32) !=
463 le32_to_cpu(agpt->partition_entry_array_crc32)) {
464 printk(KERN_WARNING
465 "GPT:partition_entry_array_crc32 values don't match: "
466 "0x%x != 0x%x\n",
467 le32_to_cpu(pgpt->partition_entry_array_crc32),
468 le32_to_cpu(agpt->partition_entry_array_crc32));
469 error_found++;
471 if (le64_to_cpu(pgpt->alternate_lba) != lastlba) {
472 printk(KERN_WARNING
473 "GPT:Primary header thinks Alt. header is not at the end of the disk.\n");
474 printk(KERN_WARNING "GPT:%lld != %lld\n",
475 (unsigned long long)le64_to_cpu(pgpt->alternate_lba),
476 (unsigned long long)lastlba);
477 error_found++;
480 if (le64_to_cpu(agpt->my_lba) != lastlba) {
481 printk(KERN_WARNING
482 "GPT:Alternate GPT header not at the end of the disk.\n");
483 printk(KERN_WARNING "GPT:%lld != %lld\n",
484 (unsigned long long)le64_to_cpu(agpt->my_lba),
485 (unsigned long long)lastlba);
486 error_found++;
489 if (error_found)
490 printk(KERN_WARNING
491 "GPT: Use GNU Parted to correct GPT errors.\n");
492 return;
496 * find_valid_gpt() - Search disk for valid GPT headers and PTEs
497 * @bdev
498 * @gpt is a GPT header ptr, filled on return.
499 * @ptes is a PTEs ptr, filled on return.
500 * Description: Returns 1 if valid, 0 on error.
501 * If valid, returns pointers to newly allocated GPT header and PTEs.
502 * Validity depends on PMBR being valid (or being overridden by the
503 * 'gpt' kernel command line option) and finding either the Primary
504 * GPT header and PTEs valid, or the Alternate GPT header and PTEs
505 * valid. If the Primary GPT header is not valid, the Alternate GPT header
506 * is not checked unless the 'gpt' kernel command line option is passed.
507 * This protects against devices which misreport their size, and forces
508 * the user to decide to use the Alternate GPT.
510 static int
511 find_valid_gpt(struct block_device *bdev, gpt_header **gpt, gpt_entry **ptes)
513 int good_pgpt = 0, good_agpt = 0, good_pmbr = 0;
514 gpt_header *pgpt = NULL, *agpt = NULL;
515 gpt_entry *pptes = NULL, *aptes = NULL;
516 legacy_mbr *legacymbr;
517 u64 lastlba;
518 if (!bdev || !gpt || !ptes)
519 return 0;
521 lastlba = last_lba(bdev);
522 if (!force_gpt) {
523 /* This will be added to the EFI Spec. per Intel after v1.02. */
524 legacymbr = kzalloc(sizeof (*legacymbr), GFP_KERNEL);
525 if (legacymbr) {
526 read_lba(bdev, 0, (u8 *) legacymbr,
527 sizeof (*legacymbr));
528 good_pmbr = is_pmbr_valid(legacymbr);
529 kfree(legacymbr);
531 if (!good_pmbr)
532 goto fail;
535 good_pgpt = is_gpt_valid(bdev, GPT_PRIMARY_PARTITION_TABLE_LBA,
536 &pgpt, &pptes);
537 if (good_pgpt)
538 good_agpt = is_gpt_valid(bdev,
539 le64_to_cpu(pgpt->alternate_lba),
540 &agpt, &aptes);
541 if (!good_agpt && force_gpt)
542 good_agpt = is_gpt_valid(bdev, lastlba,
543 &agpt, &aptes);
545 /* The obviously unsuccessful case */
546 if (!good_pgpt && !good_agpt)
547 goto fail;
549 compare_gpts(pgpt, agpt, lastlba);
551 /* The good cases */
552 if (good_pgpt) {
553 *gpt = pgpt;
554 *ptes = pptes;
555 kfree(agpt);
556 kfree(aptes);
557 if (!good_agpt) {
558 printk(KERN_WARNING
559 "Alternate GPT is invalid, "
560 "using primary GPT.\n");
562 return 1;
564 else if (good_agpt) {
565 *gpt = agpt;
566 *ptes = aptes;
567 kfree(pgpt);
568 kfree(pptes);
569 printk(KERN_WARNING
570 "Primary GPT is invalid, using alternate GPT.\n");
571 return 1;
574 fail:
575 kfree(pgpt);
576 kfree(agpt);
577 kfree(pptes);
578 kfree(aptes);
579 *gpt = NULL;
580 *ptes = NULL;
581 return 0;
585 * efi_partition(struct parsed_partitions *state, struct block_device *bdev)
586 * @state
587 * @bdev
589 * Description: called from check.c, if the disk contains GPT
590 * partitions, sets up partition entries in the kernel.
592 * If the first block on the disk is a legacy MBR,
593 * it will get handled by msdos_partition().
594 * If it's a Protective MBR, we'll handle it here.
596 * We do not create a Linux partition for GPT, but
597 * only for the actual data partitions.
598 * Returns:
599 * -1 if unable to read the partition table
600 * 0 if this isn't our partition table
601 * 1 if successful
605 efi_partition(struct parsed_partitions *state, struct block_device *bdev)
607 gpt_header *gpt = NULL;
608 gpt_entry *ptes = NULL;
609 u32 i;
610 unsigned ssz = bdev_logical_block_size(bdev) / 512;
612 if (!find_valid_gpt(bdev, &gpt, &ptes) || !gpt || !ptes) {
613 kfree(gpt);
614 kfree(ptes);
615 return 0;
618 pr_debug("GUID Partition Table is valid! Yea!\n");
620 for (i = 0; i < le32_to_cpu(gpt->num_partition_entries) && i < state->limit-1; i++) {
621 u64 start = le64_to_cpu(ptes[i].starting_lba);
622 u64 size = le64_to_cpu(ptes[i].ending_lba) -
623 le64_to_cpu(ptes[i].starting_lba) + 1ULL;
625 if (!is_pte_valid(&ptes[i], last_lba(bdev)))
626 continue;
628 put_partition(state, i+1, start * ssz, size * ssz);
630 /* If this is a RAID volume, tell md */
631 if (!efi_guidcmp(ptes[i].partition_type_guid,
632 PARTITION_LINUX_RAID_GUID))
633 state->parts[i+1].flags = 1;
635 kfree(ptes);
636 kfree(gpt);
637 printk("\n");
638 return 1;