1 /*
2 * openfs.c --- open an ext2 filesystem
3 *
4 * Copyright (C) 1993, 1994, 1995, 1996 Theodore Ts'o.
5 *
6 * %Begin-Header%
7 * This file may be redistributed under the terms of the GNU Library
8 * General Public License, version 2.
9 * %End-Header%
10 */
11
12 #include "config.h"
13 #include <stdio.h>
14 #include <string.h>
15 #if HAVE_UNISTD_H
16 #include <unistd.h>
17 #endif
18 #include <fcntl.h>
19 #include <time.h>
20 #if HAVE_SYS_STAT_H
21 #include <sys/stat.h>
22 #endif
23 #if HAVE_SYS_TYPES_H
24 #include <sys/types.h>
25 #endif
26 #ifdef HAVE_ERRNO_H
27 #include <errno.h>
28 #endif
29
30 #include "ext2_fs.h"
31
32 #include "ext2fs.h"
33 #include "e2image.h"
34
ext2fs_descriptor_block_loc2(ext2_filsys fs,blk64_t group_block,dgrp_t i)35 blk64_t ext2fs_descriptor_block_loc2(ext2_filsys fs, blk64_t group_block,
36 dgrp_t i)
37 {
38 int bg;
39 int has_super = 0, group_zero_adjust = 0;
40 blk64_t ret_blk;
41
42 /*
43 * On a bigalloc FS with 1K blocks, block 0 is reserved for non-ext4
44 * stuff, so adjust for that if we're being asked for group 0.
45 */
46 if (i == 0 && fs->blocksize == 1024 && EXT2FS_CLUSTER_RATIO(fs) > 1)
47 group_zero_adjust = 1;
48
49 if (!ext2fs_has_feature_meta_bg(fs->super) ||
50 (i < fs->super->s_first_meta_bg))
51 return group_block + i + 1 + group_zero_adjust;
52
53 bg = EXT2_DESC_PER_BLOCK(fs->super) * i;
54 if (ext2fs_bg_has_super(fs, bg))
55 has_super = 1;
56 ret_blk = ext2fs_group_first_block2(fs, bg);
57 /*
58 * If group_block is not the normal value, we're trying to use
59 * the backup group descriptors and superblock --- so use the
60 * alternate location of the second block group in the
61 * metablock group. Ideally we should be testing each bg
62 * descriptor block individually for correctness, but we don't
63 * have the infrastructure in place to do that.
64 */
65 if (group_block != fs->super->s_first_data_block &&
66 ((ret_blk + has_super + fs->super->s_blocks_per_group) <
67 ext2fs_blocks_count(fs->super))) {
68 ret_blk += fs->super->s_blocks_per_group;
69
70 /*
71 * If we're going to jump forward a block group, make sure
72 * that we adjust has_super to account for the next group's
73 * backup superblock (or lack thereof).
74 */
75 if (ext2fs_bg_has_super(fs, bg + 1))
76 has_super = 1;
77 else
78 has_super = 0;
79 }
80 return ret_blk + has_super + group_zero_adjust;
81 }
82
ext2fs_descriptor_block_loc(ext2_filsys fs,blk_t group_block,dgrp_t i)83 blk_t ext2fs_descriptor_block_loc(ext2_filsys fs, blk_t group_block, dgrp_t i)
84 {
85 return ext2fs_descriptor_block_loc2(fs, group_block, i);
86 }
87
ext2fs_open(const char * name,int flags,int superblock,unsigned int block_size,io_manager manager,ext2_filsys * ret_fs)88 errcode_t ext2fs_open(const char *name, int flags, int superblock,
89 unsigned int block_size, io_manager manager,
90 ext2_filsys *ret_fs)
91 {
92 return ext2fs_open2(name, 0, flags, superblock, block_size,
93 manager, ret_fs);
94 }
95
block_sha_map_free_entry(void * data)96 static void block_sha_map_free_entry(void *data)
97 {
98 free(data);
99 return;
100 }
101
102 /*
103 * Note: if superblock is non-zero, block-size must also be non-zero.
104 * Superblock and block_size can be zero to use the default size.
105 *
106 * Valid flags for ext2fs_open()
107 *
108 * EXT2_FLAG_RW - Open the filesystem for read/write.
109 * EXT2_FLAG_FORCE - Open the filesystem even if some of the
110 * features aren't supported.
111 * EXT2_FLAG_JOURNAL_DEV_OK - Open an ext3 journal device
112 * EXT2_FLAG_SKIP_MMP - Open without multi-mount protection check.
113 * EXT2_FLAG_64BITS - Allow 64-bit bitfields (needed for large
114 * filesystems)
115 */
ext2fs_open2(const char * name,const char * io_options,int flags,int superblock,unsigned int block_size,io_manager manager,ext2_filsys * ret_fs)116 errcode_t ext2fs_open2(const char *name, const char *io_options,
117 int flags, int superblock,
118 unsigned int block_size, io_manager manager,
119 ext2_filsys *ret_fs)
120 {
121 ext2_filsys fs;
122 errcode_t retval;
123 unsigned long i, first_meta_bg;
124 __u32 features;
125 unsigned int blocks_per_group, io_flags;
126 blk64_t group_block, blk;
127 char *dest, *cp;
128 int group_zero_adjust = 0;
129 unsigned int inode_size;
130 __u64 groups_cnt;
131 #ifdef WORDS_BIGENDIAN
132 unsigned int groups_per_block;
133 struct ext2_group_desc *gdp;
134 int j;
135 #endif
136 char *time_env;
137 int csum_retries = 0;
138
139 EXT2_CHECK_MAGIC(manager, EXT2_ET_MAGIC_IO_MANAGER);
140
141 retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &fs);
142 if (retval)
143 return retval;
144
145 memset(fs, 0, sizeof(struct struct_ext2_filsys));
146 fs->magic = EXT2_ET_MAGIC_EXT2FS_FILSYS;
147 fs->flags = flags;
148 /* don't overwrite sb backups unless flag is explicitly cleared */
149 fs->flags |= EXT2_FLAG_MASTER_SB_ONLY;
150 fs->umask = 022;
151
152 time_env = getenv("E2FSPROGS_FAKE_TIME");
153 if (time_env)
154 fs->now = strtoul(time_env, NULL, 0);
155
156 retval = ext2fs_get_mem(strlen(name)+1, &fs->device_name);
157 if (retval)
158 goto cleanup;
159 strcpy(fs->device_name, name);
160 cp = strchr(fs->device_name, '?');
161 if (!io_options && cp) {
162 *cp++ = 0;
163 io_options = cp;
164 }
165
166 io_flags = 0;
167 if (flags & EXT2_FLAG_RW)
168 io_flags |= IO_FLAG_RW;
169 if (flags & EXT2_FLAG_EXCLUSIVE)
170 io_flags |= IO_FLAG_EXCLUSIVE;
171 if (flags & EXT2_FLAG_DIRECT_IO)
172 io_flags |= IO_FLAG_DIRECT_IO;
173 if (flags & EXT2_FLAG_THREADS)
174 io_flags |= IO_FLAG_THREADS;
175 retval = manager->open(fs->device_name, io_flags, &fs->io);
176 if (retval)
177 goto cleanup;
178 if (io_options &&
179 (retval = io_channel_set_options(fs->io, io_options)))
180 goto cleanup;
181 fs->image_io = fs->io;
182 fs->io->app_data = fs;
183 retval = io_channel_alloc_buf(fs->io, -SUPERBLOCK_SIZE, &fs->super);
184 if (retval)
185 goto cleanup;
186 if (flags & EXT2_FLAG_IMAGE_FILE) {
187 retval = ext2fs_get_mem(sizeof(struct ext2_image_hdr),
188 &fs->image_header);
189 if (retval)
190 goto cleanup;
191 retval = io_channel_read_blk(fs->io, 0,
192 -(int)sizeof(struct ext2_image_hdr),
193 fs->image_header);
194 if (retval)
195 goto cleanup;
196 if (ext2fs_le32_to_cpu(fs->image_header->magic_number) != EXT2_ET_MAGIC_E2IMAGE)
197 return EXT2_ET_MAGIC_E2IMAGE;
198 superblock = 1;
199 block_size = ext2fs_le32_to_cpu(fs->image_header->fs_blocksize);
200 }
201
202 /*
203 * If the user specifies a specific block # for the
204 * superblock, then he/she must also specify the block size!
205 * Otherwise, read the master superblock located at offset
206 * SUPERBLOCK_OFFSET from the start of the partition.
207 *
208 * Note: we only save a backup copy of the superblock if we
209 * are reading the superblock from the primary superblock location.
210 */
211 if (superblock) {
212 if (!block_size) {
213 retval = EXT2_ET_INVALID_ARGUMENT;
214 goto cleanup;
215 }
216 io_channel_set_blksize(fs->io, block_size);
217 group_block = superblock;
218 fs->orig_super = 0;
219 } else {
220 io_channel_set_blksize(fs->io, SUPERBLOCK_OFFSET);
221 superblock = 1;
222 group_block = 0;
223 retval = ext2fs_get_mem(SUPERBLOCK_SIZE, &fs->orig_super);
224 if (retval)
225 goto cleanup;
226 }
227 retry:
228 retval = io_channel_read_blk(fs->io, superblock, -SUPERBLOCK_SIZE,
229 fs->super);
230 if (retval)
231 goto cleanup;
232 if (fs->orig_super)
233 memcpy(fs->orig_super, fs->super, SUPERBLOCK_SIZE);
234
235 if (!(fs->flags & EXT2_FLAG_IGNORE_CSUM_ERRORS)) {
236 retval = 0;
237 if (!ext2fs_verify_csum_type(fs, fs->super))
238 retval = EXT2_ET_UNKNOWN_CSUM;
239 if (!ext2fs_superblock_csum_verify(fs, fs->super)) {
240 if (csum_retries++ < 3)
241 goto retry;
242 retval = EXT2_ET_SB_CSUM_INVALID;
243 }
244 }
245
246 #ifdef WORDS_BIGENDIAN
247 fs->flags |= EXT2_FLAG_SWAP_BYTES;
248 ext2fs_swap_super(fs->super);
249 #else
250 if (fs->flags & EXT2_FLAG_SWAP_BYTES) {
251 retval = EXT2_ET_UNIMPLEMENTED;
252 goto cleanup;
253 }
254 #endif
255
256 if (fs->super->s_magic != EXT2_SUPER_MAGIC)
257 retval = EXT2_ET_BAD_MAGIC;
258 if (retval)
259 goto cleanup;
260
261 if (fs->super->s_rev_level > EXT2_LIB_CURRENT_REV) {
262 retval = EXT2_ET_REV_TOO_HIGH;
263 goto cleanup;
264 }
265
266 /*
267 * Check for feature set incompatibility
268 */
269 if (!(flags & EXT2_FLAG_FORCE)) {
270 features = fs->super->s_feature_incompat;
271 #ifdef EXT2_LIB_SOFTSUPP_INCOMPAT
272 if (flags & EXT2_FLAG_SOFTSUPP_FEATURES)
273 features &= ~EXT2_LIB_SOFTSUPP_INCOMPAT;
274 #endif
275 if (features & ~EXT2_LIB_FEATURE_INCOMPAT_SUPP) {
276 retval = EXT2_ET_UNSUPP_FEATURE;
277 goto cleanup;
278 }
279
280 features = fs->super->s_feature_ro_compat;
281 #ifdef EXT2_LIB_SOFTSUPP_RO_COMPAT
282 if (flags & EXT2_FLAG_SOFTSUPP_FEATURES)
283 features &= ~EXT2_LIB_SOFTSUPP_RO_COMPAT;
284 #endif
285 if ((flags & EXT2_FLAG_RW) &&
286 (features & ~EXT2_LIB_FEATURE_RO_COMPAT_SUPP)) {
287 retval = EXT2_ET_RO_UNSUPP_FEATURE;
288 goto cleanup;
289 }
290
291 if (!(flags & EXT2_FLAG_JOURNAL_DEV_OK) &&
292 ext2fs_has_feature_journal_dev(fs->super)) {
293 retval = EXT2_ET_UNSUPP_FEATURE;
294 goto cleanup;
295 }
296 }
297
298 if (fs->super->s_log_block_size >
299 (unsigned) (EXT2_MAX_BLOCK_LOG_SIZE - EXT2_MIN_BLOCK_LOG_SIZE)) {
300 retval = EXT2_ET_CORRUPT_SUPERBLOCK;
301 goto cleanup;
302 }
303
304 /*
305 * bigalloc requires cluster-aware bitfield operations, which at the
306 * moment means we need EXT2_FLAG_64BITS.
307 */
308 if (ext2fs_has_feature_bigalloc(fs->super) &&
309 !(flags & EXT2_FLAG_64BITS)) {
310 retval = EXT2_ET_CANT_USE_LEGACY_BITMAPS;
311 goto cleanup;
312 }
313
314 if (!ext2fs_has_feature_bigalloc(fs->super) &&
315 (fs->super->s_log_block_size != fs->super->s_log_cluster_size)) {
316 retval = EXT2_ET_CORRUPT_SUPERBLOCK;
317 goto cleanup;
318 }
319 fs->fragsize = fs->blocksize = EXT2_BLOCK_SIZE(fs->super);
320 inode_size = EXT2_INODE_SIZE(fs->super);
321 if ((inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
322 (inode_size > fs->blocksize) ||
323 (inode_size & (inode_size - 1))) {
324 retval = EXT2_ET_CORRUPT_SUPERBLOCK;
325 goto cleanup;
326 }
327
328 /* Enforce the block group descriptor size */
329 if (ext2fs_has_feature_64bit(fs->super)) {
330 if (fs->super->s_desc_size < EXT2_MIN_DESC_SIZE_64BIT) {
331 retval = EXT2_ET_BAD_DESC_SIZE;
332 goto cleanup;
333 }
334 }
335
336 fs->cluster_ratio_bits = fs->super->s_log_cluster_size -
337 fs->super->s_log_block_size;
338 if (EXT2_BLOCKS_PER_GROUP(fs->super) !=
339 EXT2_CLUSTERS_PER_GROUP(fs->super) << fs->cluster_ratio_bits) {
340 retval = EXT2_ET_CORRUPT_SUPERBLOCK;
341 goto cleanup;
342 }
343 fs->inode_blocks_per_group = ((EXT2_INODES_PER_GROUP(fs->super) *
344 EXT2_INODE_SIZE(fs->super) +
345 EXT2_BLOCK_SIZE(fs->super) - 1) /
346 EXT2_BLOCK_SIZE(fs->super));
347 if (block_size) {
348 if (block_size != fs->blocksize) {
349 retval = EXT2_ET_UNEXPECTED_BLOCK_SIZE;
350 goto cleanup;
351 }
352 }
353 /*
354 * Set the blocksize to the filesystem's blocksize.
355 */
356 io_channel_set_blksize(fs->io, fs->blocksize);
357
358 /*
359 * If this is an external journal device, don't try to read
360 * the group descriptors, because they're not there.
361 */
362 if (ext2fs_has_feature_journal_dev(fs->super)) {
363 fs->group_desc_count = 0;
364 *ret_fs = fs;
365 return 0;
366 }
367
368 if (EXT2_INODES_PER_GROUP(fs->super) == 0) {
369 retval = EXT2_ET_CORRUPT_SUPERBLOCK;
370 goto cleanup;
371 }
372 /* Precompute the FS UUID to seed other checksums */
373 ext2fs_init_csum_seed(fs);
374
375 /*
376 * Read group descriptors
377 */
378 blocks_per_group = EXT2_BLOCKS_PER_GROUP(fs->super);
379 if (blocks_per_group == 0 ||
380 blocks_per_group > EXT2_MAX_BLOCKS_PER_GROUP(fs->super) ||
381 fs->inode_blocks_per_group > EXT2_MAX_INODES_PER_GROUP(fs->super) ||
382 EXT2_DESC_PER_BLOCK(fs->super) == 0 ||
383 fs->super->s_first_data_block >= ext2fs_blocks_count(fs->super)) {
384 retval = EXT2_ET_CORRUPT_SUPERBLOCK;
385 goto cleanup;
386 }
387 groups_cnt = ext2fs_div64_ceil(ext2fs_blocks_count(fs->super) -
388 fs->super->s_first_data_block,
389 blocks_per_group);
390 if (groups_cnt >> 32) {
391 retval = EXT2_ET_CORRUPT_SUPERBLOCK;
392 goto cleanup;
393 }
394 fs->group_desc_count = groups_cnt;
395 if (!(flags & EXT2_FLAG_IGNORE_SB_ERRORS) &&
396 (__u64)fs->group_desc_count * EXT2_INODES_PER_GROUP(fs->super) !=
397 fs->super->s_inodes_count) {
398 retval = EXT2_ET_CORRUPT_SUPERBLOCK;
399 goto cleanup;
400 }
401 fs->desc_blocks = ext2fs_div_ceil(fs->group_desc_count,
402 EXT2_DESC_PER_BLOCK(fs->super));
403 if (flags & EXT2_FLAG_SUPER_ONLY)
404 goto skip_read_bg;
405 retval = ext2fs_get_array(fs->desc_blocks, fs->blocksize,
406 &fs->group_desc);
407 if (retval)
408 goto cleanup;
409 if (!group_block)
410 group_block = fs->super->s_first_data_block;
411 /*
412 * On a FS with a 1K blocksize, block 0 is reserved for bootloaders
413 * so we must increment block numbers to any group 0 items.
414 *
415 * However, we cannot touch group_block directly because in the meta_bg
416 * case, the ext2fs_descriptor_block_loc2() function will interpret
417 * group_block != s_first_data_block to mean that we want to access the
418 * backup group descriptors. This is not what we want if the caller
419 * set superblock == 0 (i.e. auto-detect the superblock), which is
420 * what's going on here.
421 */
422 if (group_block == 0 && fs->blocksize == 1024)
423 group_zero_adjust = 1;
424 dest = (char *) fs->group_desc;
425 #ifdef WORDS_BIGENDIAN
426 groups_per_block = EXT2_DESC_PER_BLOCK(fs->super);
427 #endif
428 if (ext2fs_has_feature_meta_bg(fs->super) &&
429 !(flags & EXT2_FLAG_IMAGE_FILE)) {
430 first_meta_bg = fs->super->s_first_meta_bg;
431 if (first_meta_bg > fs->desc_blocks)
432 first_meta_bg = fs->desc_blocks;
433 } else
434 first_meta_bg = fs->desc_blocks;
435 if (first_meta_bg) {
436 retval = io_channel_read_blk(fs->io, group_block +
437 group_zero_adjust + 1,
438 first_meta_bg, dest);
439 if (retval)
440 goto cleanup;
441 #ifdef WORDS_BIGENDIAN
442 gdp = (struct ext2_group_desc *) dest;
443 for (j=0; j < groups_per_block*first_meta_bg; j++) {
444 gdp = ext2fs_group_desc(fs, fs->group_desc, j);
445 if (gdp)
446 ext2fs_swap_group_desc2(fs, gdp);
447 }
448 #endif
449 dest += fs->blocksize*first_meta_bg;
450 }
451
452 for (i = first_meta_bg ; i < fs->desc_blocks; i++) {
453 blk = ext2fs_descriptor_block_loc2(fs, group_block, i);
454 io_channel_cache_readahead(fs->io, blk, 1);
455 }
456
457 for (i=first_meta_bg ; i < fs->desc_blocks; i++) {
458 blk = ext2fs_descriptor_block_loc2(fs, group_block, i);
459 retval = io_channel_read_blk64(fs->io, blk, 1, dest);
460 if (retval)
461 goto cleanup;
462 #ifdef WORDS_BIGENDIAN
463 for (j=0; j < groups_per_block; j++) {
464 gdp = ext2fs_group_desc(fs, fs->group_desc,
465 i * groups_per_block + j);
466 if (gdp)
467 ext2fs_swap_group_desc2(fs, gdp);
468 }
469 #endif
470 dest += fs->blocksize;
471 }
472
473 fs->stride = fs->super->s_raid_stride;
474
475 /*
476 * If recovery is from backup superblock, Clear _UNININT flags &
477 * reset bg_itable_unused to zero
478 */
479 if (superblock > 1 && ext2fs_has_group_desc_csum(fs)) {
480 dgrp_t group;
481
482 for (group = 0; group < fs->group_desc_count; group++) {
483 ext2fs_bg_flags_clear(fs, group, EXT2_BG_BLOCK_UNINIT);
484 ext2fs_bg_flags_clear(fs, group, EXT2_BG_INODE_UNINIT);
485 ext2fs_bg_itable_unused_set(fs, group, 0);
486 /* The checksum will be reset later, but fix it here
487 * anyway to avoid printing a lot of spurious errors. */
488 ext2fs_group_desc_csum_set(fs, group);
489 }
490 if (fs->flags & EXT2_FLAG_RW)
491 ext2fs_mark_super_dirty(fs);
492 }
493 skip_read_bg:
494 if (ext2fs_has_feature_mmp(fs->super) &&
495 !(flags & EXT2_FLAG_SKIP_MMP) &&
496 (flags & (EXT2_FLAG_RW | EXT2_FLAG_EXCLUSIVE))) {
497 retval = ext2fs_mmp_start(fs);
498 if (retval) {
499 fs->flags |= EXT2_FLAG_SKIP_MMP; /* just do cleanup */
500 ext2fs_mmp_stop(fs);
501 goto cleanup;
502 }
503 }
504
505 if (fs->flags & EXT2_FLAG_SHARE_DUP) {
506 fs->block_sha_map = ext2fs_hashmap_create(ext2fs_djb2_hash,
507 block_sha_map_free_entry, 4096);
508 if (!fs->block_sha_map) {
509 retval = EXT2_ET_NO_MEMORY;
510 goto cleanup;
511 }
512 ext2fs_set_feature_shared_blocks(fs->super);
513 }
514
515 if (ext2fs_has_feature_casefold(fs->super))
516 fs->encoding = ext2fs_load_nls_table(fs->super->s_encoding);
517
518 fs->flags &= ~EXT2_FLAG_NOFREE_ON_ERROR;
519 *ret_fs = fs;
520
521 return 0;
522 cleanup:
523 if (!(flags & EXT2_FLAG_NOFREE_ON_ERROR)) {
524 ext2fs_free(fs);
525 fs = NULL;
526 }
527 *ret_fs = fs;
528 return retval;
529 }
530
531 /*
532 * Set/get the filesystem data I/O channel.
533 *
534 * These functions are only valid if EXT2_FLAG_IMAGE_FILE is true.
535 */
ext2fs_get_data_io(ext2_filsys fs,io_channel * old_io)536 errcode_t ext2fs_get_data_io(ext2_filsys fs, io_channel *old_io)
537 {
538 if ((fs->flags & EXT2_FLAG_IMAGE_FILE) == 0)
539 return EXT2_ET_NOT_IMAGE_FILE;
540 if (old_io) {
541 *old_io = (fs->image_io == fs->io) ? 0 : fs->io;
542 }
543 return 0;
544 }
545
ext2fs_set_data_io(ext2_filsys fs,io_channel new_io)546 errcode_t ext2fs_set_data_io(ext2_filsys fs, io_channel new_io)
547 {
548 if ((fs->flags & EXT2_FLAG_IMAGE_FILE) == 0)
549 return EXT2_ET_NOT_IMAGE_FILE;
550 fs->io = new_io ? new_io : fs->image_io;
551 return 0;
552 }
553
ext2fs_rewrite_to_io(ext2_filsys fs,io_channel new_io)554 errcode_t ext2fs_rewrite_to_io(ext2_filsys fs, io_channel new_io)
555 {
556 errcode_t err;
557
558 if ((fs->flags & EXT2_FLAG_IMAGE_FILE) == 0)
559 return EXT2_ET_NOT_IMAGE_FILE;
560 err = io_channel_set_blksize(new_io, fs->blocksize);
561 if (err)
562 return err;
563 if ((new_io == fs->image_io) || (new_io == fs->io))
564 return 0;
565 if ((fs->image_io != fs->io) &&
566 fs->image_io)
567 io_channel_close(fs->image_io);
568 if (fs->io)
569 io_channel_close(fs->io);
570 fs->io = fs->image_io = new_io;
571 fs->flags |= EXT2_FLAG_DIRTY | EXT2_FLAG_RW |
572 EXT2_FLAG_BB_DIRTY | EXT2_FLAG_IB_DIRTY;
573 fs->flags &= ~EXT2_FLAG_IMAGE_FILE;
574 return 0;
575 }
576