x86, AMD IOMMU: flush domain TLB when there is more than one page to flush
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / partitions / efi.c
blobe7b07006bc4173c5489c3bb1532f01e0473e4bf1
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)
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
167 * Description: Returns 1 if PMBR is valid, 0 otherwise.
168 * Validity depends on two things:
169 * 1) MSDOS signature is in the last two bytes of the MBR
170 * 2) One partition of type 0xEE is found
172 static int
173 is_pmbr_valid(legacy_mbr *mbr)
175 int i;
176 if (!mbr || le16_to_cpu(mbr->signature) != MSDOS_MBR_SIGNATURE)
177 return 0;
178 for (i = 0; i < 4; i++)
179 if (pmbr_part_valid(&mbr->partition_record[i]))
180 return 1;
181 return 0;
185 * read_lba(): Read bytes from disk, starting at given LBA
186 * @bdev
187 * @lba
188 * @buffer
189 * @size_t
191 * Description: Reads @count bytes from @bdev into @buffer.
192 * Returns number of bytes read on success, 0 on error.
194 static size_t
195 read_lba(struct block_device *bdev, u64 lba, u8 * buffer, size_t count)
197 size_t totalreadcount = 0;
199 if (!bdev || !buffer || lba > last_lba(bdev))
200 return 0;
202 while (count) {
203 int copied = 512;
204 Sector sect;
205 unsigned char *data = read_dev_sector(bdev, lba++, &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 * @bdev
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 *
229 alloc_read_gpt_entries(struct block_device *bdev, gpt_header *gpt)
231 size_t count;
232 gpt_entry *pte;
233 if (!bdev || !gpt)
234 return NULL;
236 count = le32_to_cpu(gpt->num_partition_entries) *
237 le32_to_cpu(gpt->sizeof_partition_entry);
238 if (!count)
239 return NULL;
240 pte = kzalloc(count, GFP_KERNEL);
241 if (!pte)
242 return NULL;
244 if (read_lba(bdev, le64_to_cpu(gpt->partition_entry_lba),
245 (u8 *) pte,
246 count) < count) {
247 kfree(pte);
248 pte=NULL;
249 return NULL;
251 return pte;
255 * alloc_read_gpt_header(): Allocates GPT header, reads into it from disk
256 * @bdev
257 * @lba is the Logical Block Address of the partition table
259 * Description: returns GPT header on success, NULL on error. Allocates
260 * and fills a GPT header starting at @ from @bdev.
261 * Note: remember to free gpt when finished with it.
263 static gpt_header *
264 alloc_read_gpt_header(struct block_device *bdev, u64 lba)
266 gpt_header *gpt;
267 if (!bdev)
268 return NULL;
270 gpt = kzalloc(sizeof (gpt_header), GFP_KERNEL);
271 if (!gpt)
272 return NULL;
274 if (read_lba(bdev, lba, (u8 *) gpt,
275 sizeof (gpt_header)) < sizeof (gpt_header)) {
276 kfree(gpt);
277 gpt=NULL;
278 return NULL;
281 return gpt;
285 * is_gpt_valid() - tests one GPT header and PTEs for validity
286 * @bdev
287 * @lba is the logical block address of the GPT header to test
288 * @gpt is a GPT header ptr, filled on return.
289 * @ptes is a PTEs ptr, filled on return.
291 * Description: returns 1 if valid, 0 on error.
292 * If valid, returns pointers to newly allocated GPT header and PTEs.
294 static int
295 is_gpt_valid(struct block_device *bdev, u64 lba,
296 gpt_header **gpt, gpt_entry **ptes)
298 u32 crc, origcrc;
299 u64 lastlba;
301 if (!bdev || !gpt || !ptes)
302 return 0;
303 if (!(*gpt = alloc_read_gpt_header(bdev, lba)))
304 return 0;
306 /* Check the GUID Partition Table signature */
307 if (le64_to_cpu((*gpt)->signature) != GPT_HEADER_SIGNATURE) {
308 Dprintk("GUID Partition Table Header signature is wrong:"
309 "%lld != %lld\n",
310 (unsigned long long)le64_to_cpu((*gpt)->signature),
311 (unsigned long long)GPT_HEADER_SIGNATURE);
312 goto fail;
315 /* Check the GUID Partition Table CRC */
316 origcrc = le32_to_cpu((*gpt)->header_crc32);
317 (*gpt)->header_crc32 = 0;
318 crc = efi_crc32((const unsigned char *) (*gpt), le32_to_cpu((*gpt)->header_size));
320 if (crc != origcrc) {
321 Dprintk
322 ("GUID Partition Table Header CRC is wrong: %x != %x\n",
323 crc, origcrc);
324 goto fail;
326 (*gpt)->header_crc32 = cpu_to_le32(origcrc);
328 /* Check that the my_lba entry points to the LBA that contains
329 * the GUID Partition Table */
330 if (le64_to_cpu((*gpt)->my_lba) != lba) {
331 Dprintk("GPT my_lba incorrect: %lld != %lld\n",
332 (unsigned long long)le64_to_cpu((*gpt)->my_lba),
333 (unsigned long long)lba);
334 goto fail;
337 /* Check the first_usable_lba and last_usable_lba are
338 * within the disk.
340 lastlba = last_lba(bdev);
341 if (le64_to_cpu((*gpt)->first_usable_lba) > lastlba) {
342 Dprintk("GPT: first_usable_lba incorrect: %lld > %lld\n",
343 (unsigned long long)le64_to_cpu((*gpt)->first_usable_lba),
344 (unsigned long long)lastlba);
345 goto fail;
347 if (le64_to_cpu((*gpt)->last_usable_lba) > lastlba) {
348 Dprintk("GPT: last_usable_lba incorrect: %lld > %lld\n",
349 (unsigned long long)le64_to_cpu((*gpt)->last_usable_lba),
350 (unsigned long long)lastlba);
351 goto fail;
354 if (!(*ptes = alloc_read_gpt_entries(bdev, *gpt)))
355 goto fail;
357 /* Check the GUID Partition Entry Array CRC */
358 crc = efi_crc32((const unsigned char *) (*ptes),
359 le32_to_cpu((*gpt)->num_partition_entries) *
360 le32_to_cpu((*gpt)->sizeof_partition_entry));
362 if (crc != le32_to_cpu((*gpt)->partition_entry_array_crc32)) {
363 Dprintk("GUID Partitition Entry Array CRC check failed.\n");
364 goto fail_ptes;
367 /* We're done, all's well */
368 return 1;
370 fail_ptes:
371 kfree(*ptes);
372 *ptes = NULL;
373 fail:
374 kfree(*gpt);
375 *gpt = NULL;
376 return 0;
380 * is_pte_valid() - tests one PTE for validity
381 * @pte is the pte to check
382 * @lastlba is last lba of the disk
384 * Description: returns 1 if valid, 0 on error.
386 static inline int
387 is_pte_valid(const gpt_entry *pte, const u64 lastlba)
389 if ((!efi_guidcmp(pte->partition_type_guid, NULL_GUID)) ||
390 le64_to_cpu(pte->starting_lba) > lastlba ||
391 le64_to_cpu(pte->ending_lba) > lastlba)
392 return 0;
393 return 1;
397 * compare_gpts() - Search disk for valid GPT headers and PTEs
398 * @pgpt is the primary GPT header
399 * @agpt is the alternate GPT header
400 * @lastlba is the last LBA number
401 * Description: Returns nothing. Sanity checks pgpt and agpt fields
402 * and prints warnings on discrepancies.
405 static void
406 compare_gpts(gpt_header *pgpt, gpt_header *agpt, u64 lastlba)
408 int error_found = 0;
409 if (!pgpt || !agpt)
410 return;
411 if (le64_to_cpu(pgpt->my_lba) != le64_to_cpu(agpt->alternate_lba)) {
412 printk(KERN_WARNING
413 "GPT:Primary header LBA != Alt. header alternate_lba\n");
414 printk(KERN_WARNING "GPT:%lld != %lld\n",
415 (unsigned long long)le64_to_cpu(pgpt->my_lba),
416 (unsigned long long)le64_to_cpu(agpt->alternate_lba));
417 error_found++;
419 if (le64_to_cpu(pgpt->alternate_lba) != le64_to_cpu(agpt->my_lba)) {
420 printk(KERN_WARNING
421 "GPT:Primary header alternate_lba != Alt. header my_lba\n");
422 printk(KERN_WARNING "GPT:%lld != %lld\n",
423 (unsigned long long)le64_to_cpu(pgpt->alternate_lba),
424 (unsigned long long)le64_to_cpu(agpt->my_lba));
425 error_found++;
427 if (le64_to_cpu(pgpt->first_usable_lba) !=
428 le64_to_cpu(agpt->first_usable_lba)) {
429 printk(KERN_WARNING "GPT:first_usable_lbas don't match.\n");
430 printk(KERN_WARNING "GPT:%lld != %lld\n",
431 (unsigned long long)le64_to_cpu(pgpt->first_usable_lba),
432 (unsigned long long)le64_to_cpu(agpt->first_usable_lba));
433 error_found++;
435 if (le64_to_cpu(pgpt->last_usable_lba) !=
436 le64_to_cpu(agpt->last_usable_lba)) {
437 printk(KERN_WARNING "GPT:last_usable_lbas don't match.\n");
438 printk(KERN_WARNING "GPT:%lld != %lld\n",
439 (unsigned long long)le64_to_cpu(pgpt->last_usable_lba),
440 (unsigned long long)le64_to_cpu(agpt->last_usable_lba));
441 error_found++;
443 if (efi_guidcmp(pgpt->disk_guid, agpt->disk_guid)) {
444 printk(KERN_WARNING "GPT:disk_guids don't match.\n");
445 error_found++;
447 if (le32_to_cpu(pgpt->num_partition_entries) !=
448 le32_to_cpu(agpt->num_partition_entries)) {
449 printk(KERN_WARNING "GPT:num_partition_entries don't match: "
450 "0x%x != 0x%x\n",
451 le32_to_cpu(pgpt->num_partition_entries),
452 le32_to_cpu(agpt->num_partition_entries));
453 error_found++;
455 if (le32_to_cpu(pgpt->sizeof_partition_entry) !=
456 le32_to_cpu(agpt->sizeof_partition_entry)) {
457 printk(KERN_WARNING
458 "GPT:sizeof_partition_entry values don't match: "
459 "0x%x != 0x%x\n",
460 le32_to_cpu(pgpt->sizeof_partition_entry),
461 le32_to_cpu(agpt->sizeof_partition_entry));
462 error_found++;
464 if (le32_to_cpu(pgpt->partition_entry_array_crc32) !=
465 le32_to_cpu(agpt->partition_entry_array_crc32)) {
466 printk(KERN_WARNING
467 "GPT:partition_entry_array_crc32 values don't match: "
468 "0x%x != 0x%x\n",
469 le32_to_cpu(pgpt->partition_entry_array_crc32),
470 le32_to_cpu(agpt->partition_entry_array_crc32));
471 error_found++;
473 if (le64_to_cpu(pgpt->alternate_lba) != lastlba) {
474 printk(KERN_WARNING
475 "GPT:Primary header thinks Alt. header is not at the end of the disk.\n");
476 printk(KERN_WARNING "GPT:%lld != %lld\n",
477 (unsigned long long)le64_to_cpu(pgpt->alternate_lba),
478 (unsigned long long)lastlba);
479 error_found++;
482 if (le64_to_cpu(agpt->my_lba) != lastlba) {
483 printk(KERN_WARNING
484 "GPT:Alternate GPT header not at the end of the disk.\n");
485 printk(KERN_WARNING "GPT:%lld != %lld\n",
486 (unsigned long long)le64_to_cpu(agpt->my_lba),
487 (unsigned long long)lastlba);
488 error_found++;
491 if (error_found)
492 printk(KERN_WARNING
493 "GPT: Use GNU Parted to correct GPT errors.\n");
494 return;
498 * find_valid_gpt() - Search disk for valid GPT headers and PTEs
499 * @bdev
500 * @gpt is a GPT header ptr, filled on return.
501 * @ptes is a PTEs ptr, filled on return.
502 * Description: Returns 1 if valid, 0 on error.
503 * If valid, returns pointers to newly allocated GPT header and PTEs.
504 * Validity depends on PMBR being valid (or being overridden by the
505 * 'gpt' kernel command line option) and finding either the Primary
506 * GPT header and PTEs valid, or the Alternate GPT header and PTEs
507 * valid. If the Primary GPT header is not valid, the Alternate GPT header
508 * is not checked unless the 'gpt' kernel command line option is passed.
509 * This protects against devices which misreport their size, and forces
510 * the user to decide to use the Alternate GPT.
512 static int
513 find_valid_gpt(struct block_device *bdev, gpt_header **gpt, gpt_entry **ptes)
515 int good_pgpt = 0, good_agpt = 0, good_pmbr = 0;
516 gpt_header *pgpt = NULL, *agpt = NULL;
517 gpt_entry *pptes = NULL, *aptes = NULL;
518 legacy_mbr *legacymbr;
519 u64 lastlba;
520 if (!bdev || !gpt || !ptes)
521 return 0;
523 lastlba = last_lba(bdev);
524 if (!force_gpt) {
525 /* This will be added to the EFI Spec. per Intel after v1.02. */
526 legacymbr = kzalloc(sizeof (*legacymbr), GFP_KERNEL);
527 if (legacymbr) {
528 read_lba(bdev, 0, (u8 *) legacymbr,
529 sizeof (*legacymbr));
530 good_pmbr = is_pmbr_valid(legacymbr);
531 kfree(legacymbr);
533 if (!good_pmbr)
534 goto fail;
537 good_pgpt = is_gpt_valid(bdev, GPT_PRIMARY_PARTITION_TABLE_LBA,
538 &pgpt, &pptes);
539 if (good_pgpt)
540 good_agpt = is_gpt_valid(bdev,
541 le64_to_cpu(pgpt->alternate_lba),
542 &agpt, &aptes);
543 if (!good_agpt && force_gpt)
544 good_agpt = is_gpt_valid(bdev, lastlba,
545 &agpt, &aptes);
547 /* The obviously unsuccessful case */
548 if (!good_pgpt && !good_agpt)
549 goto fail;
551 compare_gpts(pgpt, agpt, lastlba);
553 /* The good cases */
554 if (good_pgpt) {
555 *gpt = pgpt;
556 *ptes = pptes;
557 kfree(agpt);
558 kfree(aptes);
559 if (!good_agpt) {
560 printk(KERN_WARNING
561 "Alternate GPT is invalid, "
562 "using primary GPT.\n");
564 return 1;
566 else if (good_agpt) {
567 *gpt = agpt;
568 *ptes = aptes;
569 kfree(pgpt);
570 kfree(pptes);
571 printk(KERN_WARNING
572 "Primary GPT is invalid, using alternate GPT.\n");
573 return 1;
576 fail:
577 kfree(pgpt);
578 kfree(agpt);
579 kfree(pptes);
580 kfree(aptes);
581 *gpt = NULL;
582 *ptes = NULL;
583 return 0;
587 * efi_partition(struct parsed_partitions *state, struct block_device *bdev)
588 * @state
589 * @bdev
591 * Description: called from check.c, if the disk contains GPT
592 * partitions, sets up partition entries in the kernel.
594 * If the first block on the disk is a legacy MBR,
595 * it will get handled by msdos_partition().
596 * If it's a Protective MBR, we'll handle it here.
598 * We do not create a Linux partition for GPT, but
599 * only for the actual data partitions.
600 * Returns:
601 * -1 if unable to read the partition table
602 * 0 if this isn't our partition table
603 * 1 if successful
607 efi_partition(struct parsed_partitions *state, struct block_device *bdev)
609 gpt_header *gpt = NULL;
610 gpt_entry *ptes = NULL;
611 u32 i;
613 if (!find_valid_gpt(bdev, &gpt, &ptes) || !gpt || !ptes) {
614 kfree(gpt);
615 kfree(ptes);
616 return 0;
619 Dprintk("GUID Partition Table is valid! Yea!\n");
621 for (i = 0; i < le32_to_cpu(gpt->num_partition_entries) && i < state->limit-1; i++) {
622 if (!is_pte_valid(&ptes[i], last_lba(bdev)))
623 continue;
625 put_partition(state, i+1, le64_to_cpu(ptes[i].starting_lba),
626 (le64_to_cpu(ptes[i].ending_lba) -
627 le64_to_cpu(ptes[i].starting_lba) +
628 1ULL));
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;