]>
Commit | Line | Data |
---|---|---|
17390c04 TT |
1 | /* |
2 | * journal.c --- code for handling the "ext3" journal | |
3b5386dc TT |
3 | * |
4 | * Copyright (C) 2000 Andreas Dilger | |
5 | * Copyright (C) 2000 Theodore Ts'o | |
6 | * | |
7 | * Parts of the code are based on fs/jfs/journal.c by Stephen C. Tweedie | |
8 | * Copyright (C) 1999 Red Hat Software | |
9 | * | |
10 | * This file may be redistributed under the terms of the | |
11 | * GNU General Public License version 2 or at your discretion | |
12 | * any later version. | |
17390c04 TT |
13 | */ |
14 | ||
3b5386dc TT |
15 | #ifdef HAVE_SYS_MOUNT_H |
16 | #include <sys/mount.h> | |
17 | #define MNT_FL (MS_MGC_VAL | MS_RDONLY) | |
18 | #endif | |
19 | #ifdef HAVE_SYS_STAT_H | |
20 | #include <sys/stat.h> | |
21 | #endif | |
17390c04 | 22 | |
53ef44c4 | 23 | #define E2FSCK_INCLUDE_INLINE_FUNCS |
0e8a9560 | 24 | #include "jfs_user.h" |
3b5386dc TT |
25 | #include "problem.h" |
26 | #include "uuid/uuid.h" | |
17390c04 | 27 | |
8cf93332 | 28 | #ifdef CONFIG_JBD_DEBUG /* Enabled by configure --enable-jfs-debug */ |
3b5386dc | 29 | static int bh_count = 0; |
3b5386dc TT |
30 | #endif |
31 | ||
d1a2182a TT |
32 | /* |
33 | * Define USE_INODE_IO to use the inode_io.c / fileio.c codepaths. | |
34 | * This creates a larger static binary, and a smaller binary using | |
35 | * shared libraries. It's also probably slightly less CPU-efficient, | |
36 | * which is why it's not on by default. But, it's a good way of | |
37 | * testing the functions in inode_io.c and fileio.c. | |
38 | */ | |
39 | #undef USE_INODE_IO | |
40 | ||
b2f93192 TT |
41 | /* Kernel compatibility functions for handling the journal. These allow us |
42 | * to use the recovery.c file virtually unchanged from the kernel, so we | |
43 | * don't have to do much to keep kernel and user recovery in sync. | |
44 | */ | |
8cf93332 | 45 | int journal_bmap(journal_t *journal, blk_t block, unsigned long *phys) |
3b5386dc | 46 | { |
d1a2182a TT |
47 | #ifdef USE_INODE_IO |
48 | *phys = block; | |
49 | return 0; | |
50 | #else | |
8cf93332 TT |
51 | struct inode *inode = journal->j_inode; |
52 | errcode_t retval; | |
53 | blk_t pblk; | |
3b5386dc | 54 | |
8cf93332 TT |
55 | if (!inode) { |
56 | *phys = block; | |
57 | return 0; | |
58 | } | |
3b5386dc | 59 | |
8cf93332 TT |
60 | retval= ext2fs_bmap(inode->i_ctx->fs, inode->i_ino, |
61 | &inode->i_ext2, NULL, 0, block, &pblk); | |
62 | *phys = pblk; | |
63 | return (retval); | |
d1a2182a | 64 | #endif |
3b5386dc TT |
65 | } |
66 | ||
8cf93332 | 67 | struct buffer_head *getblk(kdev_t kdev, blk_t blocknr, int blocksize) |
3b5386dc TT |
68 | { |
69 | struct buffer_head *bh; | |
70 | ||
8cf93332 | 71 | bh = e2fsck_allocate_memory(kdev->k_ctx, sizeof(*bh), "block buffer"); |
3b5386dc TT |
72 | if (!bh) |
73 | return NULL; | |
74 | ||
75 | jfs_debug(4, "getblk for block %lu (%d bytes)(total %d)\n", | |
53ef44c4 | 76 | (unsigned long) blocknr, blocksize, ++bh_count); |
3b5386dc | 77 | |
8cf93332 TT |
78 | bh->b_ctx = kdev->k_ctx; |
79 | if (kdev->k_dev == K_DEV_FS) | |
80 | bh->b_io = kdev->k_ctx->fs->io; | |
81 | else | |
82 | bh->b_io = kdev->k_ctx->journal_io; | |
3b5386dc TT |
83 | bh->b_size = blocksize; |
84 | bh->b_blocknr = blocknr; | |
85 | ||
86 | return bh; | |
87 | } | |
88 | ||
0e8a9560 | 89 | void ll_rw_block(int rw, int nr, struct buffer_head *bhp[]) |
3b5386dc TT |
90 | { |
91 | int retval; | |
0e8a9560 | 92 | struct buffer_head *bh; |
3b5386dc | 93 | |
0e8a9560 TT |
94 | for (; nr > 0; --nr) { |
95 | bh = *bhp++; | |
96 | if (rw == READ && !bh->b_uptodate) { | |
97 | jfs_debug(3, "reading block %lu/%p\n", | |
53ef44c4 | 98 | (unsigned long) bh->b_blocknr, (void *) bh); |
8cf93332 | 99 | retval = io_channel_read_blk(bh->b_io, |
0e8a9560 TT |
100 | bh->b_blocknr, |
101 | 1, bh->b_data); | |
102 | if (retval) { | |
103 | com_err(bh->b_ctx->device_name, retval, | |
104 | "while reading block %ld\n", | |
105 | bh->b_blocknr); | |
106 | bh->b_err = retval; | |
107 | continue; | |
108 | } | |
109 | bh->b_uptodate = 1; | |
110 | } else if (rw == WRITE && bh->b_dirty) { | |
111 | jfs_debug(3, "writing block %lu/%p\n", | |
53ef44c4 | 112 | (unsigned long) bh->b_blocknr, (void *) bh); |
8cf93332 | 113 | retval = io_channel_write_blk(bh->b_io, |
0e8a9560 TT |
114 | bh->b_blocknr, |
115 | 1, bh->b_data); | |
116 | if (retval) { | |
117 | com_err(bh->b_ctx->device_name, retval, | |
118 | "while writing block %ld\n", | |
119 | bh->b_blocknr); | |
120 | bh->b_err = retval; | |
121 | continue; | |
122 | } | |
123 | bh->b_dirty = 0; | |
124 | bh->b_uptodate = 1; | |
125 | } else | |
126 | jfs_debug(3, "no-op %s for block %lu\n", | |
127 | rw == READ ? "read" : "write", | |
53ef44c4 | 128 | (unsigned long) bh->b_blocknr); |
0e8a9560 | 129 | } |
3b5386dc TT |
130 | } |
131 | ||
8cf93332 | 132 | void mark_buffer_dirty(struct buffer_head *bh) |
3b5386dc | 133 | { |
8cf93332 | 134 | bh->b_dirty = 1; |
3b5386dc TT |
135 | } |
136 | ||
6a6d3d44 TT |
137 | static void mark_buffer_clean(struct buffer_head * bh) |
138 | { | |
139 | bh->b_dirty = 0; | |
140 | } | |
141 | ||
3b5386dc TT |
142 | void brelse(struct buffer_head *bh) |
143 | { | |
144 | if (bh->b_dirty) | |
0e8a9560 | 145 | ll_rw_block(WRITE, 1, &bh); |
3b5386dc | 146 | jfs_debug(3, "freeing block %lu/%p (total %d)\n", |
53ef44c4 | 147 | (unsigned long) bh->b_blocknr, (void *) bh, --bh_count); |
3b5386dc TT |
148 | ext2fs_free_mem((void **) &bh); |
149 | } | |
150 | ||
151 | int buffer_uptodate(struct buffer_head *bh) | |
152 | { | |
153 | return bh->b_uptodate; | |
154 | } | |
155 | ||
15986f79 TT |
156 | void mark_buffer_uptodate(struct buffer_head *bh, int val) |
157 | { | |
158 | bh->b_uptodate = val; | |
159 | } | |
160 | ||
3b5386dc TT |
161 | void wait_on_buffer(struct buffer_head *bh) |
162 | { | |
163 | if (!bh->b_uptodate) | |
0e8a9560 | 164 | ll_rw_block(READ, 1, &bh); |
3b5386dc TT |
165 | } |
166 | ||
80bfaa3e | 167 | |
3b5386dc TT |
168 | static void e2fsck_clear_recover(e2fsck_t ctx, int error) |
169 | { | |
5dd8f963 | 170 | ctx->fs->super->s_feature_incompat &= ~EXT3_FEATURE_INCOMPAT_RECOVER; |
3b5386dc TT |
171 | |
172 | /* if we had an error doing journal recovery, we need a full fsck */ | |
173 | if (error) | |
5dd8f963 | 174 | ctx->fs->super->s_state &= ~EXT2_VALID_FS; |
3b5386dc TT |
175 | ext2fs_mark_super_dirty(ctx->fs); |
176 | } | |
177 | ||
d1a2182a | 178 | static errcode_t e2fsck_get_journal(e2fsck_t ctx, journal_t **ret_journal) |
3b5386dc | 179 | { |
d1a2182a TT |
180 | struct ext2_super_block *sb = ctx->fs->super; |
181 | struct ext2_super_block jsuper; | |
182 | struct problem_context pctx; | |
183 | struct buffer_head *bh; | |
184 | struct inode *j_inode = NULL; | |
185 | struct kdev_s *dev_fs = NULL, *dev_journal; | |
3e699064 | 186 | const char *journal_name = 0; |
d1a2182a | 187 | journal_t *journal = NULL; |
3e699064 TT |
188 | errcode_t retval = 0; |
189 | io_manager io_ptr = 0; | |
d1a2182a TT |
190 | unsigned long start = 0; |
191 | int free_journal_name = 0; | |
192 | int ext_journal = 0; | |
193 | ||
194 | clear_problem_context(&pctx); | |
195 | ||
196 | journal = e2fsck_allocate_memory(ctx, sizeof(journal_t), "journal"); | |
197 | if (!journal) { | |
3b5386dc TT |
198 | return EXT2_ET_NO_MEMORY; |
199 | } | |
200 | ||
8cf93332 TT |
201 | dev_fs = e2fsck_allocate_memory(ctx, 2*sizeof(struct kdev_s), "kdev"); |
202 | if (!dev_fs) { | |
203 | retval = EXT2_ET_NO_MEMORY; | |
204 | goto errout; | |
205 | } | |
206 | dev_journal = dev_fs+1; | |
207 | ||
8cf93332 TT |
208 | dev_fs->k_ctx = dev_journal->k_ctx = ctx; |
209 | dev_fs->k_dev = K_DEV_FS; | |
210 | dev_journal->k_dev = K_DEV_JOURNAL; | |
211 | ||
d1a2182a TT |
212 | journal->j_dev = dev_journal; |
213 | journal->j_fs_dev = dev_fs; | |
214 | journal->j_inode = NULL; | |
215 | journal->j_blocksize = ctx->fs->blocksize; | |
3b5386dc | 216 | |
d1a2182a TT |
217 | if (uuid_is_null(sb->s_journal_uuid)) { |
218 | if (!sb->s_journal_inum) | |
219 | return EXT2_ET_BAD_INODE_NUM; | |
220 | j_inode = e2fsck_allocate_memory(ctx, sizeof(*j_inode), | |
221 | "journal inode"); | |
222 | if (!j_inode) { | |
223 | retval = EXT2_ET_NO_MEMORY; | |
224 | goto errout; | |
225 | } | |
3b5386dc | 226 | |
d1a2182a TT |
227 | j_inode->i_ctx = ctx; |
228 | j_inode->i_ino = sb->s_journal_inum; | |
229 | ||
230 | if ((retval = ext2fs_read_inode(ctx->fs, | |
231 | sb->s_journal_inum, | |
232 | &j_inode->i_ext2))) | |
233 | goto errout; | |
234 | if (!j_inode->i_ext2.i_links_count || | |
235 | !LINUX_S_ISREG(j_inode->i_ext2.i_mode)) { | |
236 | retval = EXT2_ET_NO_JOURNAL; | |
237 | goto errout; | |
238 | } | |
239 | if (j_inode->i_ext2.i_size / journal->j_blocksize < | |
240 | JFS_MIN_JOURNAL_BLOCKS) { | |
241 | retval = EXT2_ET_JOURNAL_TOO_SMALL; | |
242 | goto errout; | |
243 | } | |
3b5386dc | 244 | |
d1a2182a | 245 | journal->j_maxlen = j_inode->i_ext2.i_size / journal->j_blocksize; |
3b5386dc | 246 | |
d1a2182a TT |
247 | #ifdef USE_INODE_IO |
248 | retval = ext2fs_inode_io_intern(ctx->fs, sb->s_journal_inum, | |
249 | &journal_name); | |
250 | if (retval) | |
251 | goto errout; | |
3b5386dc | 252 | |
d1a2182a TT |
253 | io_ptr = inode_io_manager; |
254 | #else | |
255 | journal->j_inode = j_inode; | |
256 | ctx->journal_io = ctx->fs->io; | |
257 | if ((retval = journal_bmap(journal, 0, &start)) != 0) | |
258 | goto errout; | |
259 | #endif | |
260 | } else { | |
261 | ext_journal = 1; | |
262 | journal_name = ctx->journal_name; | |
263 | if (!journal_name) { | |
264 | journal_name = ext2fs_find_block_device(sb->s_journal_dev); | |
265 | free_journal_name = 1; | |
266 | } | |
267 | ||
268 | if (!journal_name) { | |
269 | fix_problem(ctx, PR_0_CANT_FIND_JOURNAL, &pctx); | |
270 | return EXT2_ET_LOAD_EXT_JOURNAL; | |
271 | } | |
272 | ||
273 | jfs_debug(1, "Using journal file %s\n", journal_name); | |
274 | io_ptr = unix_io_manager; | |
3b5386dc TT |
275 | } |
276 | ||
d1a2182a TT |
277 | #if 0 |
278 | test_io_backing_manager = io_ptr; | |
adee8d75 | 279 | io_ptr = test_io_manager; |
adee8d75 | 280 | #endif |
d1a2182a TT |
281 | #ifndef USE_INODE_IO |
282 | if (ext_journal) | |
283 | #endif | |
284 | retval = io_ptr->open(journal_name, IO_FLAG_RW, | |
285 | &ctx->journal_io); | |
286 | if (free_journal_name) | |
6d222f32 TT |
287 | free((void *) journal_name); |
288 | if (retval) | |
d1a2182a | 289 | goto errout; |
adee8d75 | 290 | |
d1a2182a TT |
291 | io_channel_set_blksize(ctx->journal_io, ctx->fs->blocksize); |
292 | ||
293 | if (ext_journal) { | |
294 | if (ctx->fs->blocksize == 1024) | |
295 | start = 1; | |
296 | bh = getblk(dev_journal, start, ctx->fs->blocksize); | |
297 | if (!bh) { | |
298 | retval = EXT2_ET_NO_MEMORY; | |
299 | goto errout; | |
300 | } | |
301 | ll_rw_block(READ, 1, &bh); | |
302 | if ((retval = bh->b_err) != 0) | |
303 | goto errout; | |
304 | memcpy(&jsuper, start ? bh->b_data : bh->b_data + 1024, | |
305 | sizeof(jsuper)); | |
306 | brelse(bh); | |
adee8d75 | 307 | #ifdef EXT2FS_ENABLE_SWAPFS |
d1a2182a TT |
308 | if (jsuper.s_magic == ext2fs_swab16(EXT2_SUPER_MAGIC)) |
309 | ext2fs_swap_super(&jsuper); | |
adee8d75 | 310 | #endif |
d1a2182a TT |
311 | if (jsuper.s_magic != EXT2_SUPER_MAGIC || |
312 | !(jsuper.s_feature_incompat & EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) { | |
313 | fix_problem(ctx, PR_0_EXT_JOURNAL_BAD_SUPER, &pctx); | |
314 | retval = EXT2_ET_LOAD_EXT_JOURNAL; | |
315 | goto errout; | |
316 | } | |
317 | /* Make sure the journal UUID is correct */ | |
318 | if (memcmp(jsuper.s_uuid, ctx->fs->super->s_journal_uuid, | |
319 | sizeof(jsuper.s_uuid))) { | |
320 | fix_problem(ctx, PR_0_JOURNAL_BAD_UUID, &pctx); | |
321 | retval = EXT2_ET_LOAD_EXT_JOURNAL; | |
322 | goto errout; | |
323 | } | |
adee8d75 | 324 | |
d1a2182a TT |
325 | journal->j_maxlen = jsuper.s_blocks_count; |
326 | start++; | |
3b5386dc TT |
327 | } |
328 | ||
d1a2182a | 329 | if (!(bh = getblk(dev_journal, start, journal->j_blocksize))) { |
adee8d75 TT |
330 | retval = EXT2_ET_NO_MEMORY; |
331 | goto errout; | |
332 | } | |
adee8d75 | 333 | |
d1a2182a TT |
334 | journal->j_sb_buffer = bh; |
335 | journal->j_superblock = (journal_superblock_t *)bh->b_data; | |
336 | ||
337 | #ifdef USE_INODE_IO | |
338 | if (j_inode) | |
339 | ext2fs_free_mem((void **)&j_inode); | |
340 | #endif | |
341 | ||
342 | *ret_journal = journal; | |
adee8d75 TT |
343 | return 0; |
344 | ||
345 | errout: | |
d1a2182a TT |
346 | if (dev_fs) |
347 | ext2fs_free_mem((void **)&dev_fs); | |
348 | if (j_inode) | |
349 | ext2fs_free_mem((void **)&j_inode); | |
350 | if (journal) | |
351 | ext2fs_free_mem((void **)&journal); | |
adee8d75 | 352 | return retval; |
d1a2182a | 353 | |
3b5386dc TT |
354 | } |
355 | ||
6a6d3d44 TT |
356 | static errcode_t e2fsck_journal_fix_bad_inode(e2fsck_t ctx, |
357 | struct problem_context *pctx) | |
3b5386dc | 358 | { |
5dd8f963 TT |
359 | struct ext2_super_block *sb = ctx->fs->super; |
360 | int recover = ctx->fs->super->s_feature_incompat & | |
361 | EXT3_FEATURE_INCOMPAT_RECOVER; | |
362 | int has_journal = ctx->fs->super->s_feature_compat & | |
363 | EXT3_FEATURE_COMPAT_HAS_JOURNAL; | |
3b5386dc | 364 | |
5dd8f963 | 365 | if (has_journal || sb->s_journal_inum) { |
3b5386dc | 366 | /* The journal inode is bogus, remove and force full fsck */ |
7e92dfae | 367 | pctx->ino = sb->s_journal_inum; |
3b5386dc | 368 | if (fix_problem(ctx, PR_0_JOURNAL_BAD_INODE, pctx)) { |
5dd8f963 | 369 | if (has_journal && sb->s_journal_inum) |
3b5386dc TT |
370 | printf("*** ext3 journal has been deleted - " |
371 | "filesystem is now ext2 only ***\n\n"); | |
5dd8f963 TT |
372 | sb->s_feature_compat &= ~EXT3_FEATURE_COMPAT_HAS_JOURNAL; |
373 | sb->s_journal_inum = 0; | |
424cd2be | 374 | ctx->flags |= E2F_FLAG_JOURNAL_INODE; /* FIXME: todo */ |
3b5386dc TT |
375 | e2fsck_clear_recover(ctx, 1); |
376 | return 0; | |
377 | } | |
378 | return EXT2_ET_BAD_INODE_NUM; | |
379 | } else if (recover) { | |
380 | if (fix_problem(ctx, PR_0_JOURNAL_RECOVER_SET, pctx)) { | |
381 | e2fsck_clear_recover(ctx, 1); | |
382 | return 0; | |
383 | } | |
384 | return EXT2_ET_UNSUPP_FEATURE; | |
385 | } | |
386 | return 0; | |
387 | } | |
388 | ||
62e3e7fe TT |
389 | #define V1_SB_SIZE 0x0024 |
390 | static void clear_v2_journal_fields(journal_t *journal) | |
391 | { | |
8cf93332 | 392 | e2fsck_t ctx = journal->j_dev->k_ctx; |
62e3e7fe TT |
393 | struct problem_context pctx; |
394 | ||
395 | clear_problem_context(&pctx); | |
396 | ||
397 | if (!fix_problem(ctx, PR_0_CLEAR_V2_JOURNAL, &pctx)) | |
398 | return; | |
399 | ||
400 | memset(((char *) journal->j_superblock) + V1_SB_SIZE, 0, | |
401 | ctx->fs->blocksize-V1_SB_SIZE); | |
8cf93332 | 402 | mark_buffer_dirty(journal->j_sb_buffer); |
62e3e7fe TT |
403 | } |
404 | ||
405 | ||
6a6d3d44 | 406 | static errcode_t e2fsck_journal_load(journal_t *journal) |
3b5386dc | 407 | { |
8cf93332 | 408 | e2fsck_t ctx = journal->j_dev->k_ctx; |
3b5386dc TT |
409 | journal_superblock_t *jsb; |
410 | struct buffer_head *jbh = journal->j_sb_buffer; | |
411 | struct problem_context pctx; | |
412 | ||
413 | clear_problem_context(&pctx); | |
414 | ||
0e8a9560 | 415 | ll_rw_block(READ, 1, &jbh); |
3b5386dc TT |
416 | if (jbh->b_err) { |
417 | com_err(ctx->device_name, jbh->b_err, | |
418 | _("reading journal superblock\n")); | |
419 | return jbh->b_err; | |
420 | } | |
421 | ||
422 | jsb = journal->j_superblock; | |
423 | /* If we don't even have JFS_MAGIC, we probably have a wrong inode */ | |
424 | if (jsb->s_header.h_magic != htonl(JFS_MAGIC_NUMBER)) | |
425 | return e2fsck_journal_fix_bad_inode(ctx, &pctx); | |
426 | ||
0e8a9560 TT |
427 | switch (ntohl(jsb->s_header.h_blocktype)) { |
428 | case JFS_SUPERBLOCK_V1: | |
429 | journal->j_format_version = 1; | |
62e3e7fe TT |
430 | if (jsb->s_feature_compat || |
431 | jsb->s_feature_incompat || | |
432 | jsb->s_feature_ro_compat || | |
433 | jsb->s_nr_users) | |
434 | clear_v2_journal_fields(journal); | |
0e8a9560 TT |
435 | break; |
436 | ||
437 | case JFS_SUPERBLOCK_V2: | |
438 | journal->j_format_version = 2; | |
b3b3d465 | 439 | if (ntohl(jsb->s_nr_users) > 1 && |
62e3e7fe TT |
440 | (ctx->fs->io == ctx->journal_io)) |
441 | clear_v2_journal_fields(journal); | |
adee8d75 TT |
442 | if (ntohl(jsb->s_nr_users) > 1) { |
443 | fix_problem(ctx, PR_0_JOURNAL_UNSUPP_MULTIFS, &pctx); | |
444 | return EXT2_ET_JOURNAL_UNSUPP_VERSION; | |
445 | } | |
0e8a9560 | 446 | break; |
c7f23364 TT |
447 | |
448 | /* | |
449 | * These should never appear in a journal super block, so if | |
450 | * they do, the journal is badly corrupted. | |
451 | */ | |
452 | case JFS_DESCRIPTOR_BLOCK: | |
453 | case JFS_COMMIT_BLOCK: | |
454 | case JFS_REVOKE_BLOCK: | |
455 | return EXT2_ET_CORRUPT_SUPERBLOCK; | |
0e8a9560 TT |
456 | |
457 | /* If we don't understand the superblock major type, but there | |
458 | * is a magic number, then it is likely to be a new format we | |
459 | * just don't understand, so leave it alone. */ | |
460 | default: | |
c7f23364 | 461 | return EXT2_ET_JOURNAL_UNSUPP_VERSION; |
0e8a9560 TT |
462 | } |
463 | ||
2f686ace TT |
464 | if (JFS_HAS_INCOMPAT_FEATURE(journal, ~JFS_KNOWN_INCOMPAT_FEATURES)) |
465 | return EXT2_ET_UNSUPP_FEATURE; | |
424cd2be | 466 | |
2f686ace TT |
467 | if (JFS_HAS_RO_COMPAT_FEATURE(journal, ~JFS_KNOWN_ROCOMPAT_FEATURES)) |
468 | return EXT2_ET_RO_UNSUPP_FEATURE; | |
3b5386dc | 469 | |
0e8a9560 TT |
470 | /* We have now checked whether we know enough about the journal |
471 | * format to be able to proceed safely, so any other checks that | |
472 | * fail we should attempt to recover from. */ | |
473 | if (jsb->s_blocksize != htonl(journal->j_blocksize)) { | |
474 | com_err(ctx->program_name, EXT2_ET_CORRUPT_SUPERBLOCK, | |
475 | _("%s: no valid journal superblock found\n"), | |
476 | ctx->device_name); | |
477 | return EXT2_ET_CORRUPT_SUPERBLOCK; | |
3b5386dc TT |
478 | } |
479 | ||
480 | if (ntohl(jsb->s_maxlen) < journal->j_maxlen) | |
481 | journal->j_maxlen = ntohl(jsb->s_maxlen); | |
482 | else if (ntohl(jsb->s_maxlen) > journal->j_maxlen) { | |
0e8a9560 TT |
483 | com_err(ctx->program_name, EXT2_ET_CORRUPT_SUPERBLOCK, |
484 | _("%s: journal too short\n"), | |
485 | ctx->device_name); | |
3b5386dc TT |
486 | return EXT2_ET_CORRUPT_SUPERBLOCK; |
487 | } | |
488 | ||
489 | journal->j_tail_sequence = ntohl(jsb->s_sequence); | |
ecf1b776 | 490 | journal->j_transaction_sequence = journal->j_tail_sequence; |
3b5386dc TT |
491 | journal->j_tail = ntohl(jsb->s_start); |
492 | journal->j_first = ntohl(jsb->s_first); | |
493 | journal->j_last = ntohl(jsb->s_maxlen); | |
494 | ||
495 | return 0; | |
496 | } | |
497 | ||
53ef44c4 | 498 | static void e2fsck_journal_reset_super(e2fsck_t ctx, journal_superblock_t *jsb, |
6a6d3d44 | 499 | journal_t *journal) |
3b5386dc | 500 | { |
0e8a9560 | 501 | char *p; |
424cd2be TT |
502 | union { |
503 | uuid_t uuid; | |
504 | __u32 val[4]; | |
505 | } u; | |
506 | __u32 new_seq = 0; | |
507 | int i; | |
508 | ||
0e8a9560 TT |
509 | /* Leave a valid existing V1 superblock signature alone. |
510 | * Anything unrecognisable we overwrite with a new V2 | |
511 | * signature. */ | |
512 | ||
513 | if (jsb->s_header.h_magic != htonl(JFS_MAGIC_NUMBER) || | |
514 | jsb->s_header.h_blocktype != htonl(JFS_SUPERBLOCK_V1)) { | |
515 | jsb->s_header.h_magic = htonl(JFS_MAGIC_NUMBER); | |
516 | jsb->s_header.h_blocktype = htonl(JFS_SUPERBLOCK_V2); | |
517 | } | |
518 | ||
519 | /* Zero out everything else beyond the superblock header */ | |
520 | ||
521 | p = ((char *) jsb) + sizeof(journal_header_t); | |
522 | memset (p, 0, ctx->fs->blocksize-sizeof(journal_header_t)); | |
523 | ||
3b5386dc | 524 | jsb->s_blocksize = htonl(ctx->fs->blocksize); |
0e8a9560 TT |
525 | jsb->s_maxlen = htonl(journal->j_maxlen); |
526 | jsb->s_first = htonl(1); | |
0e8a9560 | 527 | |
424cd2be TT |
528 | /* Initialize the journal sequence number so that there is "no" |
529 | * chance we will find old "valid" transactions in the journal. | |
530 | * This avoids the need to zero the whole journal (slow to do, | |
531 | * and risky when we are just recovering the filesystem). | |
532 | */ | |
533 | uuid_generate(u.uuid); | |
534 | for (i = 0; i < 4; i ++) | |
535 | new_seq ^= u.val[i]; | |
536 | jsb->s_sequence = htonl(new_seq); | |
0e8a9560 | 537 | |
8cf93332 | 538 | mark_buffer_dirty(journal->j_sb_buffer); |
0e8a9560 | 539 | ll_rw_block(WRITE, 1, &journal->j_sb_buffer); |
3b5386dc TT |
540 | } |
541 | ||
6a6d3d44 TT |
542 | static errcode_t e2fsck_journal_fix_corrupt_super(e2fsck_t ctx, |
543 | journal_t *journal, | |
544 | struct problem_context *pctx) | |
3b5386dc | 545 | { |
5dd8f963 TT |
546 | struct ext2_super_block *sb = ctx->fs->super; |
547 | int recover = ctx->fs->super->s_feature_incompat & | |
548 | EXT3_FEATURE_INCOMPAT_RECOVER; | |
3b5386dc | 549 | |
5dd8f963 | 550 | if (sb->s_feature_compat & EXT3_FEATURE_COMPAT_HAS_JOURNAL) { |
3b5386dc | 551 | if (fix_problem(ctx, PR_0_JOURNAL_BAD_SUPER, pctx)) { |
b2f93192 TT |
552 | e2fsck_journal_reset_super(ctx, journal->j_superblock, |
553 | journal); | |
3b5386dc TT |
554 | journal->j_transaction_sequence = 1; |
555 | e2fsck_clear_recover(ctx, recover); | |
556 | return 0; | |
557 | } | |
558 | return EXT2_ET_CORRUPT_SUPERBLOCK; | |
559 | } else if (e2fsck_journal_fix_bad_inode(ctx, pctx)) | |
560 | return EXT2_ET_CORRUPT_SUPERBLOCK; | |
561 | ||
562 | return 0; | |
563 | } | |
564 | ||
6a6d3d44 TT |
565 | static void e2fsck_journal_release(e2fsck_t ctx, journal_t *journal, |
566 | int reset, int drop) | |
3b5386dc TT |
567 | { |
568 | journal_superblock_t *jsb; | |
569 | ||
6a6d3d44 TT |
570 | if (drop) |
571 | mark_buffer_clean(journal->j_sb_buffer); | |
572 | else if (!(ctx->options & E2F_OPT_READONLY)) { | |
3b5386dc TT |
573 | jsb = journal->j_superblock; |
574 | jsb->s_sequence = htonl(journal->j_transaction_sequence); | |
575 | if (reset) | |
576 | jsb->s_start = 0; /* this marks the journal as empty */ | |
8cf93332 | 577 | mark_buffer_dirty(journal->j_sb_buffer); |
3b5386dc TT |
578 | } |
579 | brelse(journal->j_sb_buffer); | |
580 | ||
6d222f32 TT |
581 | if (ctx->journal_io) { |
582 | if (ctx->fs && ctx->fs->io != ctx->journal_io) | |
583 | io_channel_close(ctx->journal_io); | |
584 | ctx->journal_io = 0; | |
585 | } | |
586 | ||
d1a2182a | 587 | #ifndef USE_INODE_IO |
3b5386dc | 588 | if (journal->j_inode) |
6a6d3d44 | 589 | ext2fs_free_mem((void **)&journal->j_inode); |
d1a2182a TT |
590 | #endif |
591 | if (journal->j_fs_dev) | |
592 | ext2fs_free_mem((void **)&journal->j_fs_dev); | |
3b5386dc TT |
593 | ext2fs_free_mem((void **)&journal); |
594 | } | |
595 | ||
9b565759 TT |
596 | /* |
597 | * This function makes sure that the superblock fields regarding the | |
598 | * journal are consistent. | |
599 | */ | |
3b5386dc TT |
600 | int e2fsck_check_ext3_journal(e2fsck_t ctx) |
601 | { | |
5dd8f963 | 602 | struct ext2_super_block *sb = ctx->fs->super; |
3b5386dc | 603 | journal_t *journal; |
5dd8f963 TT |
604 | int recover = ctx->fs->super->s_feature_incompat & |
605 | EXT3_FEATURE_INCOMPAT_RECOVER; | |
3b5386dc | 606 | struct problem_context pctx; |
d37066a9 | 607 | problem_t problem; |
d3f35b64 | 608 | int reset = 0, force_fsck = 0; |
3b5386dc TT |
609 | int retval; |
610 | ||
611 | /* If we don't have any journal features, don't do anything more */ | |
5dd8f963 TT |
612 | if (!(sb->s_feature_compat & EXT3_FEATURE_COMPAT_HAS_JOURNAL) && |
613 | !recover && sb->s_journal_inum == 0 && sb->s_journal_dev == 0 && | |
614 | uuid_is_null(sb->s_journal_uuid)) | |
9b565759 | 615 | return 0; |
3b5386dc TT |
616 | |
617 | clear_problem_context(&pctx); | |
5dd8f963 | 618 | pctx.num = sb->s_journal_inum; |
3b5386dc TT |
619 | |
620 | retval = e2fsck_get_journal(ctx, &journal); | |
621 | if (retval) { | |
f2d5c937 TT |
622 | if ((retval == EXT2_ET_BAD_INODE_NUM) || |
623 | (retval == EXT2_ET_JOURNAL_TOO_SMALL) || | |
624 | (retval == EXT2_ET_NO_JOURNAL)) | |
3b5386dc TT |
625 | return e2fsck_journal_fix_bad_inode(ctx, &pctx); |
626 | return retval; | |
627 | } | |
628 | ||
629 | retval = e2fsck_journal_load(journal); | |
630 | if (retval) { | |
c7f23364 | 631 | if ((retval == EXT2_ET_CORRUPT_SUPERBLOCK) || |
2f686ace TT |
632 | ((retval == EXT2_ET_UNSUPP_FEATURE) && |
633 | (!fix_problem(ctx, PR_0_JOURNAL_UNSUPP_INCOMPAT, | |
634 | &pctx))) || | |
635 | ((retval == EXT2_ET_RO_UNSUPP_FEATURE) && | |
636 | (!fix_problem(ctx, PR_0_JOURNAL_UNSUPP_ROCOMPAT, | |
637 | &pctx))) || | |
c7f23364 TT |
638 | ((retval == EXT2_ET_JOURNAL_UNSUPP_VERSION) && |
639 | (!fix_problem(ctx, PR_0_JOURNAL_UNSUPP_VERSION, &pctx)))) | |
6a6d3d44 TT |
640 | retval = e2fsck_journal_fix_corrupt_super(ctx, journal, |
641 | &pctx); | |
642 | e2fsck_journal_release(ctx, journal, 0, 1); | |
3b5386dc TT |
643 | return retval; |
644 | } | |
645 | ||
646 | /* | |
647 | * We want to make the flags consistent here. We will not leave with | |
648 | * needs_recovery set but has_journal clear. We can't get in a loop | |
649 | * with -y, -n, or -p, only if a user isn't making up their mind. | |
650 | */ | |
651 | no_has_journal: | |
5dd8f963 TT |
652 | if (!(sb->s_feature_compat & EXT3_FEATURE_COMPAT_HAS_JOURNAL)) { |
653 | recover = sb->s_feature_incompat & EXT3_FEATURE_INCOMPAT_RECOVER; | |
3b5386dc TT |
654 | pctx.str = "inode"; |
655 | if (fix_problem(ctx, PR_0_JOURNAL_HAS_JOURNAL, &pctx)) { | |
656 | if (recover && | |
657 | !fix_problem(ctx, PR_0_JOURNAL_RECOVER_SET, &pctx)) | |
658 | goto no_has_journal; | |
d3f35b64 TT |
659 | /* |
660 | * Need a full fsck if we are releasing a | |
8c2dc521 | 661 | * journal stored on a reserved inode. |
d3f35b64 TT |
662 | */ |
663 | force_fsck = recover || | |
664 | (sb->s_journal_inum < EXT2_FIRST_INODE(sb)); | |
665 | /* Clear all of the journal fields */ | |
5dd8f963 | 666 | sb->s_journal_inum = 0; |
d3f35b64 TT |
667 | sb->s_journal_dev = 0; |
668 | memset(sb->s_journal_uuid, 0, | |
669 | sizeof(sb->s_journal_uuid)); | |
670 | e2fsck_clear_recover(ctx, force_fsck); | |
3b5386dc | 671 | } else if (!(ctx->options & E2F_OPT_READONLY)) { |
5dd8f963 | 672 | sb->s_feature_compat |= EXT3_FEATURE_COMPAT_HAS_JOURNAL; |
3b5386dc TT |
673 | ext2fs_mark_super_dirty(ctx->fs); |
674 | } | |
675 | } | |
676 | ||
5dd8f963 TT |
677 | if (sb->s_feature_compat & EXT3_FEATURE_COMPAT_HAS_JOURNAL && |
678 | !(sb->s_feature_incompat & EXT3_FEATURE_INCOMPAT_RECOVER) && | |
3b5386dc | 679 | journal->j_superblock->s_start != 0) { |
d37066a9 TT |
680 | /* Print status information */ |
681 | fix_problem(ctx, PR_0_JOURNAL_RECOVERY_CLEAR, &pctx); | |
682 | if (ctx->superblock) | |
683 | problem = PR_0_JOURNAL_RUN_DEFAULT; | |
684 | else | |
685 | problem = PR_0_JOURNAL_RUN; | |
686 | if (fix_problem(ctx, problem, &pctx)) { | |
687 | ctx->options |= E2F_OPT_FORCE; | |
688 | sb->s_feature_incompat |= | |
689 | EXT3_FEATURE_INCOMPAT_RECOVER; | |
690 | ext2fs_mark_super_dirty(ctx->fs); | |
691 | } else if (fix_problem(ctx, | |
692 | PR_0_JOURNAL_RESET_JOURNAL, &pctx)) { | |
3b5386dc | 693 | reset = 1; |
d3f35b64 TT |
694 | sb->s_state &= ~EXT2_VALID_FS; |
695 | ext2fs_mark_super_dirty(ctx->fs); | |
696 | } | |
697 | /* | |
698 | * If the user answers no to the above question, we | |
699 | * ignore the fact that journal apparently has data; | |
700 | * accidentally replaying over valid data would be far | |
701 | * worse than skipping a questionable recovery. | |
702 | * | |
703 | * XXX should we abort with a fatal error here? What | |
704 | * will the ext3 kernel code do if a filesystem with | |
705 | * !NEEDS_RECOVERY but with a non-zero | |
706 | * journal->j_superblock->s_start is mounted? | |
707 | */ | |
3b5386dc TT |
708 | } |
709 | ||
6a6d3d44 | 710 | e2fsck_journal_release(ctx, journal, reset, 0); |
3b5386dc TT |
711 | return retval; |
712 | } | |
713 | ||
6a6d3d44 | 714 | static errcode_t recover_ext3_journal(e2fsck_t ctx) |
3b5386dc | 715 | { |
3b5386dc TT |
716 | journal_t *journal; |
717 | int retval; | |
718 | ||
8cf93332 | 719 | journal_init_revoke_caches(); |
3b5386dc TT |
720 | retval = e2fsck_get_journal(ctx, &journal); |
721 | if (retval) | |
ecf1b776 TT |
722 | return retval; |
723 | ||
3b5386dc TT |
724 | retval = e2fsck_journal_load(journal); |
725 | if (retval) | |
6a6d3d44 | 726 | goto errout; |
3b5386dc | 727 | |
0e8a9560 TT |
728 | retval = journal_init_revoke(journal, 1024); |
729 | if (retval) | |
6a6d3d44 | 730 | goto errout; |
0e8a9560 | 731 | |
3b5386dc | 732 | retval = -journal_recover(journal); |
c0a083fa TT |
733 | if (retval) |
734 | goto errout; | |
735 | ||
736 | if (journal->j_superblock->s_errno) { | |
737 | ctx->fs->super->s_state |= EXT2_ERROR_FS; | |
738 | ext2fs_mark_super_dirty(ctx->fs); | |
739 | journal->j_superblock->s_errno = 0; | |
8cf93332 | 740 | mark_buffer_dirty(journal->j_sb_buffer); |
c0a083fa TT |
741 | } |
742 | ||
6a6d3d44 | 743 | errout: |
d1a2182a | 744 | journal_destroy_revoke(journal); |
8cf93332 | 745 | journal_destroy_revoke_caches(); |
6a6d3d44 | 746 | e2fsck_journal_release(ctx, journal, 1, 0); |
3b5386dc TT |
747 | return retval; |
748 | } | |
749 | ||
80bfaa3e TT |
750 | int e2fsck_run_ext3_journal(e2fsck_t ctx) |
751 | { | |
752 | io_manager io_ptr = ctx->fs->io->manager; | |
753 | int blocksize = ctx->fs->blocksize; | |
ecf1b776 | 754 | errcode_t retval, recover_retval; |
8394902e TT |
755 | |
756 | printf(_("%s: recovering journal\n"), ctx->device_name); | |
ecf1b776 | 757 | if (ctx->options & E2F_OPT_READONLY) { |
8394902e | 758 | printf(_("%s: won't do journal recovery while read-only\n"), |
ecf1b776 TT |
759 | ctx->device_name); |
760 | return EXT2_ET_FILE_RO; | |
761 | } | |
762 | ||
0f2cfe25 | 763 | if (ctx->fs->flags & EXT2_FLAG_DIRTY) |
3c6b8977 | 764 | ext2fs_flush(ctx->fs); /* Force out any modifications */ |
d0515212 | 765 | |
ecf1b776 | 766 | recover_retval = recover_ext3_journal(ctx); |
80bfaa3e TT |
767 | |
768 | /* | |
769 | * Reload the filesystem context to get up-to-date data from disk | |
770 | * because journal recovery will change the filesystem under us. | |
771 | */ | |
772 | ext2fs_close(ctx->fs); | |
ecf1b776 | 773 | retval = ext2fs_open(ctx->filesystem_name, EXT2_FLAG_RW, |
80bfaa3e TT |
774 | ctx->superblock, blocksize, io_ptr, |
775 | &ctx->fs); | |
776 | ||
777 | if (retval) { | |
778 | com_err(ctx->program_name, retval, | |
779 | _("while trying to re-open %s"), | |
780 | ctx->device_name); | |
99a2cc96 | 781 | fatal_error(ctx, 0); |
80bfaa3e TT |
782 | } |
783 | ctx->fs->priv_data = ctx; | |
17390c04 | 784 | |
ecf1b776 TT |
785 | /* Set the superblock flags */ |
786 | e2fsck_clear_recover(ctx, recover_retval); | |
787 | return recover_retval; | |
17390c04 | 788 | } |
773fd8a1 TT |
789 | |
790 | /* | |
791 | * This function will move the journal inode from a visible file in | |
792 | * the filesystem directory hierarchy to the reserved inode if necessary. | |
793 | */ | |
546a1ff1 | 794 | static const char * const journal_names[] = { |
773fd8a1 TT |
795 | ".journal", "journal", ".journal.dat", "journal.dat", 0 }; |
796 | ||
797 | void e2fsck_move_ext3_journal(e2fsck_t ctx) | |
798 | { | |
799 | struct ext2_super_block *sb = ctx->fs->super; | |
800 | struct problem_context pctx; | |
801 | struct ext2_inode inode; | |
802 | ext2_filsys fs = ctx->fs; | |
803 | ext2_ino_t ino; | |
804 | errcode_t retval; | |
805 | const char * const * cpp; | |
806 | int group, mount_flags; | |
807 | ||
808 | /* | |
809 | * If the filesystem is opened read-only, or there is no | |
810 | * journal, or the journal is already in the hidden inode, | |
811 | * then do nothing. | |
812 | */ | |
813 | if ((ctx->options & E2F_OPT_READONLY) || | |
814 | (sb->s_journal_inum == 0) || | |
815 | (sb->s_journal_inum == EXT2_JOURNAL_INO) || | |
816 | !(sb->s_feature_compat & EXT3_FEATURE_COMPAT_HAS_JOURNAL)) | |
817 | return; | |
818 | ||
819 | /* | |
820 | * If the filesystem is mounted, or we can't tell whether | |
821 | * or not it's mounted, do nothing. | |
822 | */ | |
823 | retval = ext2fs_check_if_mounted(ctx->filesystem_name, &mount_flags); | |
824 | if (retval || (mount_flags & EXT2_MF_MOUNTED)) | |
825 | return; | |
826 | ||
827 | /* | |
828 | * If we can't find the name of the journal inode, then do | |
829 | * nothing. | |
830 | */ | |
831 | for (cpp = journal_names; *cpp; cpp++) { | |
832 | retval = ext2fs_lookup(fs, EXT2_ROOT_INO, *cpp, | |
833 | strlen(*cpp), 0, &ino); | |
834 | if ((retval == 0) && (ino == sb->s_journal_inum)) | |
835 | break; | |
836 | } | |
837 | if (*cpp == 0) | |
838 | return; | |
839 | ||
840 | /* | |
841 | * The inode had better have only one link and not be readable. | |
842 | */ | |
843 | if (ext2fs_read_inode(fs, ino, &inode) != 0) | |
844 | return; | |
845 | if (inode.i_links_count != 1) | |
846 | return; | |
847 | ||
848 | /* We need the inode bitmap to be loaded */ | |
849 | retval = ext2fs_read_bitmaps(fs); | |
850 | if (retval) | |
851 | return; | |
852 | ||
853 | clear_problem_context(&pctx); | |
854 | pctx.str = *cpp; | |
855 | if (!fix_problem(ctx, PR_0_MOVE_JOURNAL, &pctx)) | |
856 | return; | |
857 | ||
858 | /* | |
859 | * OK, we've done all the checks, let's actually move the | |
860 | * journal inode. Errors at this point mean we need to force | |
861 | * an ext2 filesystem check. | |
862 | */ | |
863 | if ((retval = ext2fs_unlink(fs, EXT2_ROOT_INO, *cpp, ino, 0)) != 0) | |
864 | goto err_out; | |
865 | if ((retval = ext2fs_write_inode(fs, EXT2_JOURNAL_INO, &inode)) != 0) | |
866 | goto err_out; | |
867 | sb->s_journal_inum = EXT2_JOURNAL_INO; | |
868 | ext2fs_mark_super_dirty(fs); | |
869 | inode.i_links_count = 0; | |
870 | inode.i_dtime = time(0); | |
871 | if ((retval = ext2fs_write_inode(fs, ino, &inode)) != 0) | |
872 | goto err_out; | |
873 | ||
874 | group = ext2fs_group_of_ino(fs, ino); | |
875 | ext2fs_unmark_inode_bitmap(fs->inode_map, ino); | |
876 | ext2fs_mark_ib_dirty(fs); | |
877 | fs->group_desc[group].bg_free_inodes_count++; | |
878 | fs->super->s_free_inodes_count++; | |
879 | return; | |
880 | ||
881 | err_out: | |
882 | pctx.errcode = retval; | |
883 | fix_problem(ctx, PR_0_ERR_MOVE_JOURNAL, &pctx); | |
884 | fs->super->s_state &= ~EXT2_VALID_FS; | |
885 | ext2fs_mark_super_dirty(fs); | |
886 | return; | |
887 | } | |
888 |