[PATCH] sched: fix cond_resched_softirq() offset
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / partitions / efi.c
blob1bea610078b35c34808245cc2214947bd369d672
1 /************************************************************
2 * EFI GUID Partition Table handling
3 * Per Intel EFI Specification v1.02
4 * http://developer.intel.com/technology/efi/efi.htm
5 * efi.[ch] by Matt Domsch <Matt_Domsch@dell.com>
6 * Copyright 2000,2001,2002,2004 Dell Inc.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with this program; if not, write to the Free Software
20 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 * TODO:
25 * Changelog:
26 * Mon Nov 09 2004 Matt Domsch <Matt_Domsch@dell.com>
27 * - test for valid PMBR and valid PGPT before ever reading
28 * AGPT, allow override with 'gpt' kernel command line option.
29 * - check for first/last_usable_lba outside of size of disk
31 * Tue Mar 26 2002 Matt Domsch <Matt_Domsch@dell.com>
32 * - Ported to 2.5.7-pre1 and 2.5.7-dj2
33 * - Applied patch to avoid fault in alternate header handling
34 * - cleaned up find_valid_gpt
35 * - On-disk structure and copy in memory is *always* LE now -
36 * swab fields as needed
37 * - remove print_gpt_header()
38 * - only use first max_p partition entries, to keep the kernel minor number
39 * and partition numbers tied.
41 * Mon Feb 04 2002 Matt Domsch <Matt_Domsch@dell.com>
42 * - Removed __PRIPTR_PREFIX - not being used
44 * Mon Jan 14 2002 Matt Domsch <Matt_Domsch@dell.com>
45 * - Ported to 2.5.2-pre11 + library crc32 patch Linus applied
47 * Thu Dec 6 2001 Matt Domsch <Matt_Domsch@dell.com>
48 * - Added compare_gpts().
49 * - moved le_efi_guid_to_cpus() back into this file. GPT is the only
50 * thing that keeps EFI GUIDs on disk.
51 * - Changed gpt structure names and members to be simpler and more Linux-like.
53 * Wed Oct 17 2001 Matt Domsch <Matt_Domsch@dell.com>
54 * - Removed CONFIG_DEVFS_VOLUMES_UUID code entirely per Martin Wilck
56 * Wed Oct 10 2001 Matt Domsch <Matt_Domsch@dell.com>
57 * - Changed function comments to DocBook style per Andreas Dilger suggestion.
59 * Mon Oct 08 2001 Matt Domsch <Matt_Domsch@dell.com>
60 * - Change read_lba() to use the page cache per Al Viro's work.
61 * - print u64s properly on all architectures
62 * - fixed debug_printk(), now Dprintk()
64 * Mon Oct 01 2001 Matt Domsch <Matt_Domsch@dell.com>
65 * - Style cleanups
66 * - made most functions static
67 * - Endianness addition
68 * - remove test for second alternate header, as it's not per spec,
69 * and is unnecessary. There's now a method to read/write the last
70 * sector of an odd-sized disk from user space. No tools have ever
71 * been released which used this code, so it's effectively dead.
72 * - Per Asit Mallick of Intel, added a test for a valid PMBR.
73 * - Added kernel command line option 'gpt' to override valid PMBR test.
75 * Wed Jun 6 2001 Martin Wilck <Martin.Wilck@Fujitsu-Siemens.com>
76 * - added devfs volume UUID support (/dev/volumes/uuids) for
77 * mounting file systems by the partition GUID.
79 * Tue Dec 5 2000 Matt Domsch <Matt_Domsch@dell.com>
80 * - Moved crc32() to linux/lib, added efi_crc32().
82 * Thu Nov 30 2000 Matt Domsch <Matt_Domsch@dell.com>
83 * - Replaced Intel's CRC32 function with an equivalent
84 * non-license-restricted version.
86 * Wed Oct 25 2000 Matt Domsch <Matt_Domsch@dell.com>
87 * - Fixed the last_lba() call to return the proper last block
89 * Thu Oct 12 2000 Matt Domsch <Matt_Domsch@dell.com>
90 * - Thanks to Andries Brouwer for his debugging assistance.
91 * - Code works, detects all the partitions.
93 ************************************************************/
94 #include <linux/crc32.h>
95 #include "check.h"
96 #include "efi.h"
98 #undef EFI_DEBUG
99 #ifdef EFI_DEBUG
100 #define Dprintk(x...) printk(KERN_DEBUG x)
101 #else
102 #define Dprintk(x...)
103 #endif
105 /* This allows a kernel command line option 'gpt' to override
106 * the test for invalid PMBR. Not __initdata because reloading
107 * the partition tables happens after init too.
109 static int force_gpt;
110 static int __init
111 force_gpt_fn(char *str)
113 force_gpt = 1;
114 return 1;
116 __setup("gpt", force_gpt_fn);
120 * efi_crc32() - EFI version of crc32 function
121 * @buf: buffer to calculate crc32 of
122 * @len - length of buf
124 * Description: Returns EFI-style CRC32 value for @buf
126 * This function uses the little endian Ethernet polynomial
127 * but seeds the function with ~0, and xor's with ~0 at the end.
128 * Note, the EFI Specification, v1.02, has a reference to
129 * Dr. Dobbs Journal, May 1994 (actually it's in May 1992).
131 static inline u32
132 efi_crc32(const void *buf, unsigned long len)
134 return (crc32(~0L, buf, len) ^ ~0L);
138 * last_lba(): return number of last logical block of device
139 * @bdev: block device
141 * Description: Returns last LBA value on success, 0 on error.
142 * This is stored (by sd and ide-geometry) in
143 * the part[0] entry for this disk, and is the number of
144 * physical sectors available on the disk.
146 static u64
147 last_lba(struct block_device *bdev)
149 if (!bdev || !bdev->bd_inode)
150 return 0;
151 return (bdev->bd_inode->i_size >> 9) - 1ULL;
154 static inline int
155 pmbr_part_valid(struct partition *part, u64 lastlba)
157 if (part->sys_ind == EFI_PMBR_OSTYPE_EFI_GPT &&
158 le32_to_cpu(part->start_sect) == 1UL)
159 return 1;
160 return 0;
164 * is_pmbr_valid(): test Protective MBR for validity
165 * @mbr: pointer to a legacy mbr structure
166 * @lastlba: last_lba for the whole device
168 * Description: Returns 1 if PMBR is valid, 0 otherwise.
169 * Validity depends on two things:
170 * 1) MSDOS signature is in the last two bytes of the MBR
171 * 2) One partition of type 0xEE is found
173 static int
174 is_pmbr_valid(legacy_mbr *mbr, u64 lastlba)
176 int i;
177 if (!mbr || le16_to_cpu(mbr->signature) != MSDOS_MBR_SIGNATURE)
178 return 0;
179 for (i = 0; i < 4; i++)
180 if (pmbr_part_valid(&mbr->partition_record[i], lastlba))
181 return 1;
182 return 0;
186 * read_lba(): Read bytes from disk, starting at given LBA
187 * @bdev
188 * @lba
189 * @buffer
190 * @size_t
192 * Description: Reads @count bytes from @bdev into @buffer.
193 * Returns number of bytes read on success, 0 on error.
195 static size_t
196 read_lba(struct block_device *bdev, u64 lba, u8 * buffer, size_t count)
198 size_t totalreadcount = 0;
200 if (!bdev || !buffer || lba > last_lba(bdev))
201 return 0;
203 while (count) {
204 int copied = 512;
205 Sector sect;
206 unsigned char *data = read_dev_sector(bdev, lba++, &sect);
207 if (!data)
208 break;
209 if (copied > count)
210 copied = count;
211 memcpy(buffer, data, copied);
212 put_dev_sector(sect);
213 buffer += copied;
214 totalreadcount +=copied;
215 count -= copied;
217 return totalreadcount;
221 * alloc_read_gpt_entries(): reads partition entries from disk
222 * @bdev
223 * @gpt - GPT header
225 * Description: Returns ptes on success, NULL on error.
226 * Allocates space for PTEs based on information found in @gpt.
227 * Notes: remember to free pte when you're done!
229 static gpt_entry *
230 alloc_read_gpt_entries(struct block_device *bdev, gpt_header *gpt)
232 size_t count;
233 gpt_entry *pte;
234 if (!bdev || !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(bdev, 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 * @bdev
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 @bdev.
262 * Note: remember to free gpt when finished with it.
264 static gpt_header *
265 alloc_read_gpt_header(struct block_device *bdev, u64 lba)
267 gpt_header *gpt;
268 if (!bdev)
269 return NULL;
271 gpt = kzalloc(sizeof (gpt_header), GFP_KERNEL);
272 if (!gpt)
273 return NULL;
275 if (read_lba(bdev, lba, (u8 *) gpt,
276 sizeof (gpt_header)) < sizeof (gpt_header)) {
277 kfree(gpt);
278 gpt=NULL;
279 return NULL;
282 return gpt;
286 * is_gpt_valid() - tests one GPT header and PTEs for validity
287 * @bdev
288 * @lba is the logical block address of the GPT header to test
289 * @gpt is a GPT header ptr, filled on return.
290 * @ptes is a PTEs ptr, filled on return.
292 * Description: returns 1 if valid, 0 on error.
293 * If valid, returns pointers to newly allocated GPT header and PTEs.
295 static int
296 is_gpt_valid(struct block_device *bdev, u64 lba,
297 gpt_header **gpt, gpt_entry **ptes)
299 u32 crc, origcrc;
300 u64 lastlba;
302 if (!bdev || !gpt || !ptes)
303 return 0;
304 if (!(*gpt = alloc_read_gpt_header(bdev, lba)))
305 return 0;
307 /* Check the GUID Partition Table signature */
308 if (le64_to_cpu((*gpt)->signature) != GPT_HEADER_SIGNATURE) {
309 Dprintk("GUID Partition Table Header signature is wrong:"
310 "%lld != %lld\n",
311 (unsigned long long)le64_to_cpu((*gpt)->signature),
312 (unsigned long long)GPT_HEADER_SIGNATURE);
313 goto fail;
316 /* Check the GUID Partition Table CRC */
317 origcrc = le32_to_cpu((*gpt)->header_crc32);
318 (*gpt)->header_crc32 = 0;
319 crc = efi_crc32((const unsigned char *) (*gpt), le32_to_cpu((*gpt)->header_size));
321 if (crc != origcrc) {
322 Dprintk
323 ("GUID Partition Table Header CRC is wrong: %x != %x\n",
324 crc, origcrc);
325 goto fail;
327 (*gpt)->header_crc32 = cpu_to_le32(origcrc);
329 /* Check that the my_lba entry points to the LBA that contains
330 * the GUID Partition Table */
331 if (le64_to_cpu((*gpt)->my_lba) != lba) {
332 Dprintk("GPT my_lba incorrect: %lld != %lld\n",
333 (unsigned long long)le64_to_cpu((*gpt)->my_lba),
334 (unsigned long long)lba);
335 goto fail;
338 /* Check the first_usable_lba and last_usable_lba are
339 * within the disk.
341 lastlba = last_lba(bdev);
342 if (le64_to_cpu((*gpt)->first_usable_lba) > lastlba) {
343 Dprintk("GPT: first_usable_lba incorrect: %lld > %lld\n",
344 (unsigned long long)le64_to_cpu((*gpt)->first_usable_lba),
345 (unsigned long long)lastlba);
346 goto fail;
348 if (le64_to_cpu((*gpt)->last_usable_lba) > lastlba) {
349 Dprintk("GPT: last_usable_lba incorrect: %lld > %lld\n",
350 (unsigned long long)le64_to_cpu((*gpt)->last_usable_lba),
351 (unsigned long long)lastlba);
352 goto fail;
355 if (!(*ptes = alloc_read_gpt_entries(bdev, *gpt)))
356 goto fail;
358 /* Check the GUID Partition Entry Array CRC */
359 crc = efi_crc32((const unsigned char *) (*ptes),
360 le32_to_cpu((*gpt)->num_partition_entries) *
361 le32_to_cpu((*gpt)->sizeof_partition_entry));
363 if (crc != le32_to_cpu((*gpt)->partition_entry_array_crc32)) {
364 Dprintk("GUID Partitition Entry Array CRC check failed.\n");
365 goto fail_ptes;
368 /* We're done, all's well */
369 return 1;
371 fail_ptes:
372 kfree(*ptes);
373 *ptes = NULL;
374 fail:
375 kfree(*gpt);
376 *gpt = NULL;
377 return 0;
381 * is_pte_valid() - tests one PTE for validity
382 * @pte is the pte to check
383 * @lastlba is last lba of the disk
385 * Description: returns 1 if valid, 0 on error.
387 static inline int
388 is_pte_valid(const gpt_entry *pte, const u64 lastlba)
390 if ((!efi_guidcmp(pte->partition_type_guid, NULL_GUID)) ||
391 le64_to_cpu(pte->starting_lba) > lastlba ||
392 le64_to_cpu(pte->ending_lba) > lastlba)
393 return 0;
394 return 1;
398 * compare_gpts() - Search disk for valid GPT headers and PTEs
399 * @pgpt is the primary GPT header
400 * @agpt is the alternate GPT header
401 * @lastlba is the last LBA number
402 * Description: Returns nothing. Sanity checks pgpt and agpt fields
403 * and prints warnings on discrepancies.
406 static void
407 compare_gpts(gpt_header *pgpt, gpt_header *agpt, u64 lastlba)
409 int error_found = 0;
410 if (!pgpt || !agpt)
411 return;
412 if (le64_to_cpu(pgpt->my_lba) != le64_to_cpu(agpt->alternate_lba)) {
413 printk(KERN_WARNING
414 "GPT:Primary header LBA != Alt. header alternate_lba\n");
415 printk(KERN_WARNING "GPT:%lld != %lld\n",
416 (unsigned long long)le64_to_cpu(pgpt->my_lba),
417 (unsigned long long)le64_to_cpu(agpt->alternate_lba));
418 error_found++;
420 if (le64_to_cpu(pgpt->alternate_lba) != le64_to_cpu(agpt->my_lba)) {
421 printk(KERN_WARNING
422 "GPT:Primary header alternate_lba != Alt. header my_lba\n");
423 printk(KERN_WARNING "GPT:%lld != %lld\n",
424 (unsigned long long)le64_to_cpu(pgpt->alternate_lba),
425 (unsigned long long)le64_to_cpu(agpt->my_lba));
426 error_found++;
428 if (le64_to_cpu(pgpt->first_usable_lba) !=
429 le64_to_cpu(agpt->first_usable_lba)) {
430 printk(KERN_WARNING "GPT:first_usable_lbas don't match.\n");
431 printk(KERN_WARNING "GPT:%lld != %lld\n",
432 (unsigned long long)le64_to_cpu(pgpt->first_usable_lba),
433 (unsigned long long)le64_to_cpu(agpt->first_usable_lba));
434 error_found++;
436 if (le64_to_cpu(pgpt->last_usable_lba) !=
437 le64_to_cpu(agpt->last_usable_lba)) {
438 printk(KERN_WARNING "GPT:last_usable_lbas don't match.\n");
439 printk(KERN_WARNING "GPT:%lld != %lld\n",
440 (unsigned long long)le64_to_cpu(pgpt->last_usable_lba),
441 (unsigned long long)le64_to_cpu(agpt->last_usable_lba));
442 error_found++;
444 if (efi_guidcmp(pgpt->disk_guid, agpt->disk_guid)) {
445 printk(KERN_WARNING "GPT:disk_guids don't match.\n");
446 error_found++;
448 if (le32_to_cpu(pgpt->num_partition_entries) !=
449 le32_to_cpu(agpt->num_partition_entries)) {
450 printk(KERN_WARNING "GPT:num_partition_entries don't match: "
451 "0x%x != 0x%x\n",
452 le32_to_cpu(pgpt->num_partition_entries),
453 le32_to_cpu(agpt->num_partition_entries));
454 error_found++;
456 if (le32_to_cpu(pgpt->sizeof_partition_entry) !=
457 le32_to_cpu(agpt->sizeof_partition_entry)) {
458 printk(KERN_WARNING
459 "GPT:sizeof_partition_entry values don't match: "
460 "0x%x != 0x%x\n",
461 le32_to_cpu(pgpt->sizeof_partition_entry),
462 le32_to_cpu(agpt->sizeof_partition_entry));
463 error_found++;
465 if (le32_to_cpu(pgpt->partition_entry_array_crc32) !=
466 le32_to_cpu(agpt->partition_entry_array_crc32)) {
467 printk(KERN_WARNING
468 "GPT:partition_entry_array_crc32 values don't match: "
469 "0x%x != 0x%x\n",
470 le32_to_cpu(pgpt->partition_entry_array_crc32),
471 le32_to_cpu(agpt->partition_entry_array_crc32));
472 error_found++;
474 if (le64_to_cpu(pgpt->alternate_lba) != lastlba) {
475 printk(KERN_WARNING
476 "GPT:Primary header thinks Alt. header is not at the end of the disk.\n");
477 printk(KERN_WARNING "GPT:%lld != %lld\n",
478 (unsigned long long)le64_to_cpu(pgpt->alternate_lba),
479 (unsigned long long)lastlba);
480 error_found++;
483 if (le64_to_cpu(agpt->my_lba) != lastlba) {
484 printk(KERN_WARNING
485 "GPT:Alternate GPT header not at the end of the disk.\n");
486 printk(KERN_WARNING "GPT:%lld != %lld\n",
487 (unsigned long long)le64_to_cpu(agpt->my_lba),
488 (unsigned long long)lastlba);
489 error_found++;
492 if (error_found)
493 printk(KERN_WARNING
494 "GPT: Use GNU Parted to correct GPT errors.\n");
495 return;
499 * find_valid_gpt() - Search disk for valid GPT headers and PTEs
500 * @bdev
501 * @gpt is a GPT header ptr, filled on return.
502 * @ptes is a PTEs ptr, filled on return.
503 * Description: Returns 1 if valid, 0 on error.
504 * If valid, returns pointers to newly allocated GPT header and PTEs.
505 * Validity depends on PMBR being valid (or being overridden by the
506 * 'gpt' kernel command line option) and finding either the Primary
507 * GPT header and PTEs valid, or the Alternate GPT header and PTEs
508 * valid. If the Primary GPT header is not valid, the Alternate GPT header
509 * is not checked unless the 'gpt' kernel command line option is passed.
510 * This protects against devices which misreport their size, and forces
511 * the user to decide to use the Alternate GPT.
513 static int
514 find_valid_gpt(struct block_device *bdev, gpt_header **gpt, gpt_entry **ptes)
516 int good_pgpt = 0, good_agpt = 0, good_pmbr = 0;
517 gpt_header *pgpt = NULL, *agpt = NULL;
518 gpt_entry *pptes = NULL, *aptes = NULL;
519 legacy_mbr *legacymbr = NULL;
520 u64 lastlba;
521 if (!bdev || !gpt || !ptes)
522 return 0;
524 lastlba = last_lba(bdev);
525 if (!force_gpt) {
526 /* This will be added to the EFI Spec. per Intel after v1.02. */
527 legacymbr = kzalloc(sizeof (*legacymbr), GFP_KERNEL);
528 if (legacymbr) {
529 read_lba(bdev, 0, (u8 *) legacymbr,
530 sizeof (*legacymbr));
531 good_pmbr = is_pmbr_valid(legacymbr, lastlba);
532 kfree(legacymbr);
533 legacymbr=NULL;
535 if (!good_pmbr)
536 goto fail;
539 good_pgpt = is_gpt_valid(bdev, GPT_PRIMARY_PARTITION_TABLE_LBA,
540 &pgpt, &pptes);
541 if (good_pgpt)
542 good_agpt = is_gpt_valid(bdev,
543 le64_to_cpu(pgpt->alternate_lba),
544 &agpt, &aptes);
545 if (!good_agpt && force_gpt)
546 good_agpt = is_gpt_valid(bdev, lastlba,
547 &agpt, &aptes);
549 /* The obviously unsuccessful case */
550 if (!good_pgpt && !good_agpt)
551 goto fail;
553 compare_gpts(pgpt, agpt, lastlba);
555 /* The good cases */
556 if (good_pgpt) {
557 *gpt = pgpt;
558 *ptes = pptes;
559 kfree(agpt);
560 kfree(aptes);
561 if (!good_agpt) {
562 printk(KERN_WARNING
563 "Alternate GPT is invalid, "
564 "using primary GPT.\n");
566 return 1;
568 else if (good_agpt) {
569 *gpt = agpt;
570 *ptes = aptes;
571 kfree(pgpt);
572 kfree(pptes);
573 printk(KERN_WARNING
574 "Primary GPT is invalid, using alternate GPT.\n");
575 return 1;
578 fail:
579 kfree(pgpt);
580 kfree(agpt);
581 kfree(pptes);
582 kfree(aptes);
583 *gpt = NULL;
584 *ptes = NULL;
585 return 0;
589 * efi_partition(struct parsed_partitions *state, struct block_device *bdev)
590 * @state
591 * @bdev
593 * Description: called from check.c, if the disk contains GPT
594 * partitions, sets up partition entries in the kernel.
596 * If the first block on the disk is a legacy MBR,
597 * it will get handled by msdos_partition().
598 * If it's a Protective MBR, we'll handle it here.
600 * We do not create a Linux partition for GPT, but
601 * only for the actual data partitions.
602 * Returns:
603 * -1 if unable to read the partition table
604 * 0 if this isn't our partition table
605 * 1 if successful
609 efi_partition(struct parsed_partitions *state, struct block_device *bdev)
611 gpt_header *gpt = NULL;
612 gpt_entry *ptes = NULL;
613 u32 i;
615 if (!find_valid_gpt(bdev, &gpt, &ptes) || !gpt || !ptes) {
616 kfree(gpt);
617 kfree(ptes);
618 return 0;
621 Dprintk("GUID Partition Table is valid! Yea!\n");
623 for (i = 0; i < le32_to_cpu(gpt->num_partition_entries) && i < state->limit-1; i++) {
624 if (!is_pte_valid(&ptes[i], last_lba(bdev)))
625 continue;
627 put_partition(state, i+1, le64_to_cpu(ptes[i].starting_lba),
628 (le64_to_cpu(ptes[i].ending_lba) -
629 le64_to_cpu(ptes[i].starting_lba) +
630 1ULL));
632 /* If this is a RAID volume, tell md */
633 if (!efi_guidcmp(ptes[i].partition_type_guid,
634 PARTITION_LINUX_RAID_GUID))
635 state->parts[i+1].flags = 1;
637 kfree(ptes);
638 kfree(gpt);
639 printk("\n");
640 return 1;