thp: comment reminder in destroy_compound_page
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / partitions / efi.c
blobac0ccb5026a2d107eb5e5e38c84e4bdfe51682ed
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/ctype.h>
98 #include <linux/math64.h>
99 #include <linux/slab.h>
100 #include "check.h"
101 #include "efi.h"
103 /* This allows a kernel command line option 'gpt' to override
104 * the test for invalid PMBR. Not __initdata because reloading
105 * the partition tables happens after init too.
107 static int force_gpt;
108 static int __init
109 force_gpt_fn(char *str)
111 force_gpt = 1;
112 return 1;
114 __setup("gpt", force_gpt_fn);
118 * efi_crc32() - EFI version of crc32 function
119 * @buf: buffer to calculate crc32 of
120 * @len - length of buf
122 * Description: Returns EFI-style CRC32 value for @buf
124 * This function uses the little endian Ethernet polynomial
125 * but seeds the function with ~0, and xor's with ~0 at the end.
126 * Note, the EFI Specification, v1.02, has a reference to
127 * Dr. Dobbs Journal, May 1994 (actually it's in May 1992).
129 static inline u32
130 efi_crc32(const void *buf, unsigned long len)
132 return (crc32(~0L, buf, len) ^ ~0L);
136 * last_lba(): return number of last logical block of device
137 * @bdev: block device
139 * Description: Returns last LBA value on success, 0 on error.
140 * This is stored (by sd and ide-geometry) in
141 * the part[0] entry for this disk, and is the number of
142 * physical sectors available on the disk.
144 static u64 last_lba(struct block_device *bdev)
146 if (!bdev || !bdev->bd_inode)
147 return 0;
148 return div_u64(bdev->bd_inode->i_size,
149 bdev_logical_block_size(bdev)) - 1ULL;
152 static inline int
153 pmbr_part_valid(struct partition *part)
155 if (part->sys_ind == EFI_PMBR_OSTYPE_EFI_GPT &&
156 le32_to_cpu(part->start_sect) == 1UL)
157 return 1;
158 return 0;
162 * is_pmbr_valid(): test Protective MBR for validity
163 * @mbr: pointer to a legacy mbr structure
165 * Description: Returns 1 if PMBR is valid, 0 otherwise.
166 * Validity depends on two things:
167 * 1) MSDOS signature is in the last two bytes of the MBR
168 * 2) One partition of type 0xEE is found
170 static int
171 is_pmbr_valid(legacy_mbr *mbr)
173 int i;
174 if (!mbr || le16_to_cpu(mbr->signature) != MSDOS_MBR_SIGNATURE)
175 return 0;
176 for (i = 0; i < 4; i++)
177 if (pmbr_part_valid(&mbr->partition_record[i]))
178 return 1;
179 return 0;
183 * read_lba(): Read bytes from disk, starting at given LBA
184 * @state
185 * @lba
186 * @buffer
187 * @size_t
189 * Description: Reads @count bytes from @state->bdev into @buffer.
190 * Returns number of bytes read on success, 0 on error.
192 static size_t read_lba(struct parsed_partitions *state,
193 u64 lba, u8 *buffer, size_t count)
195 size_t totalreadcount = 0;
196 struct block_device *bdev = state->bdev;
197 sector_t n = lba * (bdev_logical_block_size(bdev) / 512);
199 if (!buffer || lba > last_lba(bdev))
200 return 0;
202 while (count) {
203 int copied = 512;
204 Sector sect;
205 unsigned char *data = read_part_sector(state, n++, &sect);
206 if (!data)
207 break;
208 if (copied > count)
209 copied = count;
210 memcpy(buffer, data, copied);
211 put_dev_sector(sect);
212 buffer += copied;
213 totalreadcount +=copied;
214 count -= copied;
216 return totalreadcount;
220 * alloc_read_gpt_entries(): reads partition entries from disk
221 * @state
222 * @gpt - GPT header
224 * Description: Returns ptes on success, NULL on error.
225 * Allocates space for PTEs based on information found in @gpt.
226 * Notes: remember to free pte when you're done!
228 static gpt_entry *alloc_read_gpt_entries(struct parsed_partitions *state,
229 gpt_header *gpt)
231 size_t count;
232 gpt_entry *pte;
234 if (!gpt)
235 return NULL;
237 count = le32_to_cpu(gpt->num_partition_entries) *
238 le32_to_cpu(gpt->sizeof_partition_entry);
239 if (!count)
240 return NULL;
241 pte = kzalloc(count, GFP_KERNEL);
242 if (!pte)
243 return NULL;
245 if (read_lba(state, le64_to_cpu(gpt->partition_entry_lba),
246 (u8 *) pte,
247 count) < count) {
248 kfree(pte);
249 pte=NULL;
250 return NULL;
252 return pte;
256 * alloc_read_gpt_header(): Allocates GPT header, reads into it from disk
257 * @state
258 * @lba is the Logical Block Address of the partition table
260 * Description: returns GPT header on success, NULL on error. Allocates
261 * and fills a GPT header starting at @ from @state->bdev.
262 * Note: remember to free gpt when finished with it.
264 static gpt_header *alloc_read_gpt_header(struct parsed_partitions *state,
265 u64 lba)
267 gpt_header *gpt;
268 unsigned ssz = bdev_logical_block_size(state->bdev);
270 gpt = kzalloc(ssz, GFP_KERNEL);
271 if (!gpt)
272 return NULL;
274 if (read_lba(state, 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 * @state
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 is_gpt_valid(struct parsed_partitions *state, u64 lba,
294 gpt_header **gpt, gpt_entry **ptes)
296 u32 crc, origcrc;
297 u64 lastlba;
299 if (!ptes)
300 return 0;
301 if (!(*gpt = alloc_read_gpt_header(state, lba)))
302 return 0;
304 /* Check the GUID Partition Table signature */
305 if (le64_to_cpu((*gpt)->signature) != GPT_HEADER_SIGNATURE) {
306 pr_debug("GUID Partition Table Header signature is wrong:"
307 "%lld != %lld\n",
308 (unsigned long long)le64_to_cpu((*gpt)->signature),
309 (unsigned long long)GPT_HEADER_SIGNATURE);
310 goto fail;
313 /* Check the GUID Partition Table CRC */
314 origcrc = le32_to_cpu((*gpt)->header_crc32);
315 (*gpt)->header_crc32 = 0;
316 crc = efi_crc32((const unsigned char *) (*gpt), le32_to_cpu((*gpt)->header_size));
318 if (crc != origcrc) {
319 pr_debug("GUID Partition Table Header CRC is wrong: %x != %x\n",
320 crc, origcrc);
321 goto fail;
323 (*gpt)->header_crc32 = cpu_to_le32(origcrc);
325 /* Check that the my_lba entry points to the LBA that contains
326 * the GUID Partition Table */
327 if (le64_to_cpu((*gpt)->my_lba) != lba) {
328 pr_debug("GPT my_lba incorrect: %lld != %lld\n",
329 (unsigned long long)le64_to_cpu((*gpt)->my_lba),
330 (unsigned long long)lba);
331 goto fail;
334 /* Check the first_usable_lba and last_usable_lba are
335 * within the disk.
337 lastlba = last_lba(state->bdev);
338 if (le64_to_cpu((*gpt)->first_usable_lba) > lastlba) {
339 pr_debug("GPT: first_usable_lba incorrect: %lld > %lld\n",
340 (unsigned long long)le64_to_cpu((*gpt)->first_usable_lba),
341 (unsigned long long)lastlba);
342 goto fail;
344 if (le64_to_cpu((*gpt)->last_usable_lba) > lastlba) {
345 pr_debug("GPT: last_usable_lba incorrect: %lld > %lld\n",
346 (unsigned long long)le64_to_cpu((*gpt)->last_usable_lba),
347 (unsigned long long)lastlba);
348 goto fail;
351 if (!(*ptes = alloc_read_gpt_entries(state, *gpt)))
352 goto fail;
354 /* Check the GUID Partition Entry Array CRC */
355 crc = efi_crc32((const unsigned char *) (*ptes),
356 le32_to_cpu((*gpt)->num_partition_entries) *
357 le32_to_cpu((*gpt)->sizeof_partition_entry));
359 if (crc != le32_to_cpu((*gpt)->partition_entry_array_crc32)) {
360 pr_debug("GUID Partitition Entry Array CRC check failed.\n");
361 goto fail_ptes;
364 /* We're done, all's well */
365 return 1;
367 fail_ptes:
368 kfree(*ptes);
369 *ptes = NULL;
370 fail:
371 kfree(*gpt);
372 *gpt = NULL;
373 return 0;
377 * is_pte_valid() - tests one PTE for validity
378 * @pte is the pte to check
379 * @lastlba is last lba of the disk
381 * Description: returns 1 if valid, 0 on error.
383 static inline int
384 is_pte_valid(const gpt_entry *pte, const u64 lastlba)
386 if ((!efi_guidcmp(pte->partition_type_guid, NULL_GUID)) ||
387 le64_to_cpu(pte->starting_lba) > lastlba ||
388 le64_to_cpu(pte->ending_lba) > lastlba)
389 return 0;
390 return 1;
394 * compare_gpts() - Search disk for valid GPT headers and PTEs
395 * @pgpt is the primary GPT header
396 * @agpt is the alternate GPT header
397 * @lastlba is the last LBA number
398 * Description: Returns nothing. Sanity checks pgpt and agpt fields
399 * and prints warnings on discrepancies.
402 static void
403 compare_gpts(gpt_header *pgpt, gpt_header *agpt, u64 lastlba)
405 int error_found = 0;
406 if (!pgpt || !agpt)
407 return;
408 if (le64_to_cpu(pgpt->my_lba) != le64_to_cpu(agpt->alternate_lba)) {
409 printk(KERN_WARNING
410 "GPT:Primary header LBA != Alt. header alternate_lba\n");
411 printk(KERN_WARNING "GPT:%lld != %lld\n",
412 (unsigned long long)le64_to_cpu(pgpt->my_lba),
413 (unsigned long long)le64_to_cpu(agpt->alternate_lba));
414 error_found++;
416 if (le64_to_cpu(pgpt->alternate_lba) != le64_to_cpu(agpt->my_lba)) {
417 printk(KERN_WARNING
418 "GPT:Primary header alternate_lba != Alt. header my_lba\n");
419 printk(KERN_WARNING "GPT:%lld != %lld\n",
420 (unsigned long long)le64_to_cpu(pgpt->alternate_lba),
421 (unsigned long long)le64_to_cpu(agpt->my_lba));
422 error_found++;
424 if (le64_to_cpu(pgpt->first_usable_lba) !=
425 le64_to_cpu(agpt->first_usable_lba)) {
426 printk(KERN_WARNING "GPT:first_usable_lbas don't match.\n");
427 printk(KERN_WARNING "GPT:%lld != %lld\n",
428 (unsigned long long)le64_to_cpu(pgpt->first_usable_lba),
429 (unsigned long long)le64_to_cpu(agpt->first_usable_lba));
430 error_found++;
432 if (le64_to_cpu(pgpt->last_usable_lba) !=
433 le64_to_cpu(agpt->last_usable_lba)) {
434 printk(KERN_WARNING "GPT:last_usable_lbas don't match.\n");
435 printk(KERN_WARNING "GPT:%lld != %lld\n",
436 (unsigned long long)le64_to_cpu(pgpt->last_usable_lba),
437 (unsigned long long)le64_to_cpu(agpt->last_usable_lba));
438 error_found++;
440 if (efi_guidcmp(pgpt->disk_guid, agpt->disk_guid)) {
441 printk(KERN_WARNING "GPT:disk_guids don't match.\n");
442 error_found++;
444 if (le32_to_cpu(pgpt->num_partition_entries) !=
445 le32_to_cpu(agpt->num_partition_entries)) {
446 printk(KERN_WARNING "GPT:num_partition_entries don't match: "
447 "0x%x != 0x%x\n",
448 le32_to_cpu(pgpt->num_partition_entries),
449 le32_to_cpu(agpt->num_partition_entries));
450 error_found++;
452 if (le32_to_cpu(pgpt->sizeof_partition_entry) !=
453 le32_to_cpu(agpt->sizeof_partition_entry)) {
454 printk(KERN_WARNING
455 "GPT:sizeof_partition_entry values don't match: "
456 "0x%x != 0x%x\n",
457 le32_to_cpu(pgpt->sizeof_partition_entry),
458 le32_to_cpu(agpt->sizeof_partition_entry));
459 error_found++;
461 if (le32_to_cpu(pgpt->partition_entry_array_crc32) !=
462 le32_to_cpu(agpt->partition_entry_array_crc32)) {
463 printk(KERN_WARNING
464 "GPT:partition_entry_array_crc32 values don't match: "
465 "0x%x != 0x%x\n",
466 le32_to_cpu(pgpt->partition_entry_array_crc32),
467 le32_to_cpu(agpt->partition_entry_array_crc32));
468 error_found++;
470 if (le64_to_cpu(pgpt->alternate_lba) != lastlba) {
471 printk(KERN_WARNING
472 "GPT:Primary header thinks Alt. header is not at the end of the disk.\n");
473 printk(KERN_WARNING "GPT:%lld != %lld\n",
474 (unsigned long long)le64_to_cpu(pgpt->alternate_lba),
475 (unsigned long long)lastlba);
476 error_found++;
479 if (le64_to_cpu(agpt->my_lba) != lastlba) {
480 printk(KERN_WARNING
481 "GPT:Alternate GPT header not at the end of the disk.\n");
482 printk(KERN_WARNING "GPT:%lld != %lld\n",
483 (unsigned long long)le64_to_cpu(agpt->my_lba),
484 (unsigned long long)lastlba);
485 error_found++;
488 if (error_found)
489 printk(KERN_WARNING
490 "GPT: Use GNU Parted to correct GPT errors.\n");
491 return;
495 * find_valid_gpt() - Search disk for valid GPT headers and PTEs
496 * @state
497 * @gpt is a GPT header ptr, filled on return.
498 * @ptes is a PTEs ptr, filled on return.
499 * Description: Returns 1 if valid, 0 on error.
500 * If valid, returns pointers to newly allocated GPT header and PTEs.
501 * Validity depends on PMBR being valid (or being overridden by the
502 * 'gpt' kernel command line option) and finding either the Primary
503 * GPT header and PTEs valid, or the Alternate GPT header and PTEs
504 * valid. If the Primary GPT header is not valid, the Alternate GPT header
505 * is not checked unless the 'gpt' kernel command line option is passed.
506 * This protects against devices which misreport their size, and forces
507 * the user to decide to use the Alternate GPT.
509 static int find_valid_gpt(struct parsed_partitions *state, gpt_header **gpt,
510 gpt_entry **ptes)
512 int good_pgpt = 0, good_agpt = 0, good_pmbr = 0;
513 gpt_header *pgpt = NULL, *agpt = NULL;
514 gpt_entry *pptes = NULL, *aptes = NULL;
515 legacy_mbr *legacymbr;
516 u64 lastlba;
518 if (!ptes)
519 return 0;
521 lastlba = last_lba(state->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(state, 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(state, GPT_PRIMARY_PARTITION_TABLE_LBA,
536 &pgpt, &pptes);
537 if (good_pgpt)
538 good_agpt = is_gpt_valid(state,
539 le64_to_cpu(pgpt->alternate_lba),
540 &agpt, &aptes);
541 if (!good_agpt && force_gpt)
542 good_agpt = is_gpt_valid(state, lastlba, &agpt, &aptes);
544 /* The obviously unsuccessful case */
545 if (!good_pgpt && !good_agpt)
546 goto fail;
548 compare_gpts(pgpt, agpt, lastlba);
550 /* The good cases */
551 if (good_pgpt) {
552 *gpt = pgpt;
553 *ptes = pptes;
554 kfree(agpt);
555 kfree(aptes);
556 if (!good_agpt) {
557 printk(KERN_WARNING
558 "Alternate GPT is invalid, "
559 "using primary GPT.\n");
561 return 1;
563 else if (good_agpt) {
564 *gpt = agpt;
565 *ptes = aptes;
566 kfree(pgpt);
567 kfree(pptes);
568 printk(KERN_WARNING
569 "Primary GPT is invalid, using alternate GPT.\n");
570 return 1;
573 fail:
574 kfree(pgpt);
575 kfree(agpt);
576 kfree(pptes);
577 kfree(aptes);
578 *gpt = NULL;
579 *ptes = NULL;
580 return 0;
584 * efi_partition(struct parsed_partitions *state)
585 * @state
587 * Description: called from check.c, if the disk contains GPT
588 * partitions, sets up partition entries in the kernel.
590 * If the first block on the disk is a legacy MBR,
591 * it will get handled by msdos_partition().
592 * If it's a Protective MBR, we'll handle it here.
594 * We do not create a Linux partition for GPT, but
595 * only for the actual data partitions.
596 * Returns:
597 * -1 if unable to read the partition table
598 * 0 if this isn't our partition table
599 * 1 if successful
602 int efi_partition(struct parsed_partitions *state)
604 gpt_header *gpt = NULL;
605 gpt_entry *ptes = NULL;
606 u32 i;
607 unsigned ssz = bdev_logical_block_size(state->bdev) / 512;
608 u8 unparsed_guid[37];
610 if (!find_valid_gpt(state, &gpt, &ptes) || !gpt || !ptes) {
611 kfree(gpt);
612 kfree(ptes);
613 return 0;
616 pr_debug("GUID Partition Table is valid! Yea!\n");
618 for (i = 0; i < le32_to_cpu(gpt->num_partition_entries) && i < state->limit-1; i++) {
619 struct partition_meta_info *info;
620 unsigned label_count = 0;
621 unsigned label_max;
622 u64 start = le64_to_cpu(ptes[i].starting_lba);
623 u64 size = le64_to_cpu(ptes[i].ending_lba) -
624 le64_to_cpu(ptes[i].starting_lba) + 1ULL;
626 if (!is_pte_valid(&ptes[i], last_lba(state->bdev)))
627 continue;
629 put_partition(state, i+1, start * ssz, size * ssz);
631 /* If this is a RAID volume, tell md */
632 if (!efi_guidcmp(ptes[i].partition_type_guid,
633 PARTITION_LINUX_RAID_GUID))
634 state->parts[i + 1].flags = ADDPART_FLAG_RAID;
636 info = &state->parts[i + 1].info;
637 /* Instead of doing a manual swap to big endian, reuse the
638 * common ASCII hex format as the interim.
640 efi_guid_unparse(&ptes[i].unique_partition_guid, unparsed_guid);
641 part_pack_uuid(unparsed_guid, info->uuid);
643 /* Naively convert UTF16-LE to 7 bits. */
644 label_max = min(sizeof(info->volname) - 1,
645 sizeof(ptes[i].partition_name));
646 info->volname[label_max] = 0;
647 while (label_count < label_max) {
648 u8 c = ptes[i].partition_name[label_count] & 0xff;
649 if (c && !isprint(c))
650 c = '!';
651 info->volname[label_count] = c;
652 label_count++;
654 state->parts[i + 1].has_info = true;
656 kfree(ptes);
657 kfree(gpt);
658 strlcat(state->pp_buf, "\n", PAGE_SIZE);
659 return 1;