Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/ext3/super.c | |
3 | * | |
4 | * Copyright (C) 1992, 1993, 1994, 1995 | |
5 | * Remy Card (card@masi.ibp.fr) | |
6 | * Laboratoire MASI - Institut Blaise Pascal | |
7 | * Universite Pierre et Marie Curie (Paris VI) | |
8 | * | |
9 | * from | |
10 | * | |
11 | * linux/fs/minix/inode.c | |
12 | * | |
13 | * Copyright (C) 1991, 1992 Linus Torvalds | |
14 | * | |
15 | * Big-endian to little-endian byte-swapping/bitmaps by | |
16 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
17 | */ | |
18 | ||
1da177e4 | 19 | #include <linux/module.h> |
1da177e4 LT |
20 | #include <linux/blkdev.h> |
21 | #include <linux/parser.h> | |
a5694255 | 22 | #include <linux/exportfs.h> |
4613ad18 | 23 | #include <linux/statfs.h> |
1da177e4 LT |
24 | #include <linux/random.h> |
25 | #include <linux/mount.h> | |
1da177e4 | 26 | #include <linux/quotaops.h> |
8fc2751b | 27 | #include <linux/seq_file.h> |
3fc74269 | 28 | #include <linux/log2.h> |
d71bc6db | 29 | #include <linux/cleancache.h> |
381be254 | 30 | |
1da177e4 | 31 | #include <asm/uaccess.h> |
381be254 | 32 | |
4613ad18 AV |
33 | #define CREATE_TRACE_POINTS |
34 | ||
35 | #include "ext3.h" | |
1da177e4 LT |
36 | #include "xattr.h" |
37 | #include "acl.h" | |
381be254 | 38 | #include "namei.h" |
1da177e4 | 39 | |
bbae8bcc LT |
40 | #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED |
41 | #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA | |
42 | #else | |
43 | #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA | |
44 | #endif | |
45 | ||
71b96257 JL |
46 | static int ext3_load_journal(struct super_block *, struct ext3_super_block *, |
47 | unsigned long journal_devnum); | |
1da177e4 | 48 | static int ext3_create_journal(struct super_block *, struct ext3_super_block *, |
eee194e7 | 49 | unsigned int); |
c4be0c1d TS |
50 | static int ext3_commit_super(struct super_block *sb, |
51 | struct ext3_super_block *es, | |
1da177e4 LT |
52 | int sync); |
53 | static void ext3_mark_recovery_complete(struct super_block * sb, | |
54 | struct ext3_super_block * es); | |
55 | static void ext3_clear_journal_err(struct super_block * sb, | |
56 | struct ext3_super_block * es); | |
57 | static int ext3_sync_fs(struct super_block *sb, int wait); | |
58 | static const char *ext3_decode_error(struct super_block * sb, int errno, | |
59 | char nbuf[16]); | |
60 | static int ext3_remount (struct super_block * sb, int * flags, char * data); | |
726c3342 | 61 | static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf); |
c4be0c1d | 62 | static int ext3_unfreeze(struct super_block *sb); |
c4be0c1d | 63 | static int ext3_freeze(struct super_block *sb); |
1da177e4 | 64 | |
ae6ddcc5 | 65 | /* |
1da177e4 | 66 | * Wrappers for journal_start/end. |
1da177e4 LT |
67 | */ |
68 | handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks) | |
69 | { | |
70 | journal_t *journal; | |
71 | ||
72 | if (sb->s_flags & MS_RDONLY) | |
73 | return ERR_PTR(-EROFS); | |
74 | ||
75 | /* Special case here: if the journal has aborted behind our | |
76 | * backs (eg. EIO in the commit thread), then we still need to | |
77 | * take the FS itself readonly cleanly. */ | |
78 | journal = EXT3_SB(sb)->s_journal; | |
79 | if (is_journal_aborted(journal)) { | |
e05b6b52 | 80 | ext3_abort(sb, __func__, |
1da177e4 LT |
81 | "Detected aborted journal"); |
82 | return ERR_PTR(-EROFS); | |
83 | } | |
84 | ||
85 | return journal_start(journal, nblocks); | |
86 | } | |
87 | ||
1da177e4 LT |
88 | int __ext3_journal_stop(const char *where, handle_t *handle) |
89 | { | |
90 | struct super_block *sb; | |
91 | int err; | |
92 | int rc; | |
93 | ||
94 | sb = handle->h_transaction->t_journal->j_private; | |
95 | err = handle->h_err; | |
96 | rc = journal_stop(handle); | |
97 | ||
98 | if (!err) | |
99 | err = rc; | |
100 | if (err) | |
101 | __ext3_std_error(sb, where, err); | |
102 | return err; | |
103 | } | |
104 | ||
105 | void ext3_journal_abort_handle(const char *caller, const char *err_fn, | |
106 | struct buffer_head *bh, handle_t *handle, int err) | |
107 | { | |
108 | char nbuf[16]; | |
109 | const char *errstr = ext3_decode_error(NULL, err, nbuf); | |
110 | ||
111 | if (bh) | |
112 | BUFFER_TRACE(bh, "abort"); | |
113 | ||
114 | if (!handle->h_err) | |
115 | handle->h_err = err; | |
116 | ||
117 | if (is_handle_aborted(handle)) | |
118 | return; | |
119 | ||
4cf46b67 AF |
120 | printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n", |
121 | caller, errstr, err_fn); | |
1da177e4 LT |
122 | |
123 | journal_abort_handle(handle); | |
124 | } | |
125 | ||
4cf46b67 AF |
126 | void ext3_msg(struct super_block *sb, const char *prefix, |
127 | const char *fmt, ...) | |
128 | { | |
99fbb1e2 | 129 | struct va_format vaf; |
4cf46b67 AF |
130 | va_list args; |
131 | ||
132 | va_start(args, fmt); | |
99fbb1e2 JP |
133 | |
134 | vaf.fmt = fmt; | |
135 | vaf.va = &args; | |
136 | ||
137 | printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf); | |
138 | ||
4cf46b67 AF |
139 | va_end(args); |
140 | } | |
141 | ||
1da177e4 LT |
142 | /* Deal with the reporting of failure conditions on a filesystem such as |
143 | * inconsistencies detected or read IO failures. | |
144 | * | |
145 | * On ext2, we can store the error state of the filesystem in the | |
146 | * superblock. That is not possible on ext3, because we may have other | |
147 | * write ordering constraints on the superblock which prevent us from | |
148 | * writing it out straight away; and given that the journal is about to | |
149 | * be aborted, we can't rely on the current, or future, transactions to | |
150 | * write out the superblock safely. | |
151 | * | |
152 | * We'll just use the journal_abort() error code to record an error in | |
d6b198bc | 153 | * the journal instead. On recovery, the journal will complain about |
1da177e4 LT |
154 | * that error until we've noted it down and cleared it. |
155 | */ | |
156 | ||
157 | static void ext3_handle_error(struct super_block *sb) | |
158 | { | |
159 | struct ext3_super_block *es = EXT3_SB(sb)->s_es; | |
160 | ||
161 | EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS; | |
162 | es->s_state |= cpu_to_le16(EXT3_ERROR_FS); | |
163 | ||
164 | if (sb->s_flags & MS_RDONLY) | |
165 | return; | |
166 | ||
7543fc7b | 167 | if (!test_opt (sb, ERRORS_CONT)) { |
1da177e4 LT |
168 | journal_t *journal = EXT3_SB(sb)->s_journal; |
169 | ||
e3c96435 | 170 | set_opt(EXT3_SB(sb)->s_mount_opt, ABORT); |
1da177e4 LT |
171 | if (journal) |
172 | journal_abort(journal, -EIO); | |
173 | } | |
7543fc7b | 174 | if (test_opt (sb, ERRORS_RO)) { |
4cf46b67 AF |
175 | ext3_msg(sb, KERN_CRIT, |
176 | "error: remounting filesystem read-only"); | |
7543fc7b VA |
177 | sb->s_flags |= MS_RDONLY; |
178 | } | |
179 | ext3_commit_super(sb, es, 1); | |
1da177e4 | 180 | if (test_opt(sb, ERRORS_PANIC)) |
4cf46b67 | 181 | panic("EXT3-fs (%s): panic forced after error\n", |
1da177e4 | 182 | sb->s_id); |
1da177e4 LT |
183 | } |
184 | ||
99fbb1e2 JP |
185 | void ext3_error(struct super_block *sb, const char *function, |
186 | const char *fmt, ...) | |
1da177e4 | 187 | { |
99fbb1e2 | 188 | struct va_format vaf; |
1da177e4 LT |
189 | va_list args; |
190 | ||
191 | va_start(args, fmt); | |
99fbb1e2 JP |
192 | |
193 | vaf.fmt = fmt; | |
194 | vaf.va = &args; | |
195 | ||
196 | printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n", | |
197 | sb->s_id, function, &vaf); | |
198 | ||
1da177e4 LT |
199 | va_end(args); |
200 | ||
201 | ext3_handle_error(sb); | |
202 | } | |
203 | ||
204 | static const char *ext3_decode_error(struct super_block * sb, int errno, | |
205 | char nbuf[16]) | |
206 | { | |
207 | char *errstr = NULL; | |
208 | ||
209 | switch (errno) { | |
210 | case -EIO: | |
211 | errstr = "IO failure"; | |
212 | break; | |
213 | case -ENOMEM: | |
214 | errstr = "Out of memory"; | |
215 | break; | |
216 | case -EROFS: | |
217 | if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT) | |
218 | errstr = "Journal has aborted"; | |
219 | else | |
220 | errstr = "Readonly filesystem"; | |
221 | break; | |
222 | default: | |
223 | /* If the caller passed in an extra buffer for unknown | |
224 | * errors, textualise them now. Else we just return | |
225 | * NULL. */ | |
226 | if (nbuf) { | |
227 | /* Check for truncated error codes... */ | |
228 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
229 | errstr = nbuf; | |
230 | } | |
231 | break; | |
232 | } | |
233 | ||
234 | return errstr; | |
235 | } | |
236 | ||
237 | /* __ext3_std_error decodes expected errors from journaling functions | |
238 | * automatically and invokes the appropriate error response. */ | |
239 | ||
240 | void __ext3_std_error (struct super_block * sb, const char * function, | |
241 | int errno) | |
242 | { | |
243 | char nbuf[16]; | |
30121624 | 244 | const char *errstr; |
1da177e4 | 245 | |
30121624 ST |
246 | /* Special case: if the error is EROFS, and we're not already |
247 | * inside a transaction, then there's really no point in logging | |
248 | * an error. */ | |
249 | if (errno == -EROFS && journal_current_handle() == NULL && | |
250 | (sb->s_flags & MS_RDONLY)) | |
251 | return; | |
252 | ||
253 | errstr = ext3_decode_error(sb, errno, nbuf); | |
4cf46b67 | 254 | ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr); |
1da177e4 LT |
255 | |
256 | ext3_handle_error(sb); | |
257 | } | |
258 | ||
259 | /* | |
260 | * ext3_abort is a much stronger failure handler than ext3_error. The | |
261 | * abort function may be used to deal with unrecoverable failures such | |
262 | * as journal IO errors or ENOMEM at a critical moment in log management. | |
263 | * | |
264 | * We unconditionally force the filesystem into an ABORT|READONLY state, | |
265 | * unless the error response on the fs has been set to panic in which | |
266 | * case we take the easy way out and panic immediately. | |
267 | */ | |
268 | ||
99fbb1e2 JP |
269 | void ext3_abort(struct super_block *sb, const char *function, |
270 | const char *fmt, ...) | |
1da177e4 | 271 | { |
99fbb1e2 | 272 | struct va_format vaf; |
1da177e4 LT |
273 | va_list args; |
274 | ||
1da177e4 | 275 | va_start(args, fmt); |
99fbb1e2 JP |
276 | |
277 | vaf.fmt = fmt; | |
278 | vaf.va = &args; | |
279 | ||
280 | printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n", | |
281 | sb->s_id, function, &vaf); | |
282 | ||
1da177e4 LT |
283 | va_end(args); |
284 | ||
285 | if (test_opt(sb, ERRORS_PANIC)) | |
4cf46b67 | 286 | panic("EXT3-fs: panic from previous error\n"); |
1da177e4 LT |
287 | |
288 | if (sb->s_flags & MS_RDONLY) | |
289 | return; | |
290 | ||
4cf46b67 AF |
291 | ext3_msg(sb, KERN_CRIT, |
292 | "error: remounting filesystem read-only"); | |
1da177e4 LT |
293 | EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS; |
294 | sb->s_flags |= MS_RDONLY; | |
e3c96435 | 295 | set_opt(EXT3_SB(sb)->s_mount_opt, ABORT); |
44d6f787 HK |
296 | if (EXT3_SB(sb)->s_journal) |
297 | journal_abort(EXT3_SB(sb)->s_journal, -EIO); | |
1da177e4 LT |
298 | } |
299 | ||
99fbb1e2 JP |
300 | void ext3_warning(struct super_block *sb, const char *function, |
301 | const char *fmt, ...) | |
1da177e4 | 302 | { |
99fbb1e2 | 303 | struct va_format vaf; |
1da177e4 LT |
304 | va_list args; |
305 | ||
306 | va_start(args, fmt); | |
99fbb1e2 JP |
307 | |
308 | vaf.fmt = fmt; | |
309 | vaf.va = &args; | |
310 | ||
311 | printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n", | |
312 | sb->s_id, function, &vaf); | |
313 | ||
1da177e4 LT |
314 | va_end(args); |
315 | } | |
316 | ||
317 | void ext3_update_dynamic_rev(struct super_block *sb) | |
318 | { | |
319 | struct ext3_super_block *es = EXT3_SB(sb)->s_es; | |
320 | ||
321 | if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV) | |
322 | return; | |
323 | ||
4cf46b67 AF |
324 | ext3_msg(sb, KERN_WARNING, |
325 | "warning: updating to rev %d because of " | |
326 | "new feature flag, running e2fsck is recommended", | |
327 | EXT3_DYNAMIC_REV); | |
1da177e4 LT |
328 | |
329 | es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO); | |
330 | es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE); | |
331 | es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV); | |
332 | /* leave es->s_feature_*compat flags alone */ | |
333 | /* es->s_uuid will be set by e2fsck if empty */ | |
334 | ||
335 | /* | |
336 | * The rest of the superblock fields should be zero, and if not it | |
337 | * means they are likely already in use, so leave them alone. We | |
338 | * can leave it up to e2fsck to clean up any inconsistencies there. | |
339 | */ | |
340 | } | |
341 | ||
342 | /* | |
343 | * Open the external journal device | |
344 | */ | |
4cf46b67 | 345 | static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb) |
1da177e4 LT |
346 | { |
347 | struct block_device *bdev; | |
348 | char b[BDEVNAME_SIZE]; | |
349 | ||
d4d77629 | 350 | bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); |
1da177e4 LT |
351 | if (IS_ERR(bdev)) |
352 | goto fail; | |
353 | return bdev; | |
354 | ||
355 | fail: | |
4cf46b67 AF |
356 | ext3_msg(sb, "error: failed to open journal device %s: %ld", |
357 | __bdevname(dev, b), PTR_ERR(bdev)); | |
358 | ||
1da177e4 LT |
359 | return NULL; |
360 | } | |
361 | ||
362 | /* | |
363 | * Release the journal device | |
364 | */ | |
365 | static int ext3_blkdev_put(struct block_device *bdev) | |
366 | { | |
e525fd89 | 367 | return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); |
1da177e4 LT |
368 | } |
369 | ||
370 | static int ext3_blkdev_remove(struct ext3_sb_info *sbi) | |
371 | { | |
372 | struct block_device *bdev; | |
373 | int ret = -ENODEV; | |
374 | ||
375 | bdev = sbi->journal_bdev; | |
376 | if (bdev) { | |
377 | ret = ext3_blkdev_put(bdev); | |
378 | sbi->journal_bdev = NULL; | |
379 | } | |
380 | return ret; | |
381 | } | |
382 | ||
383 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
384 | { | |
385 | return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode; | |
386 | } | |
387 | ||
388 | static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi) | |
389 | { | |
390 | struct list_head *l; | |
391 | ||
4cf46b67 | 392 | ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d", |
1da177e4 LT |
393 | le32_to_cpu(sbi->s_es->s_last_orphan)); |
394 | ||
4cf46b67 | 395 | ext3_msg(sb, KERN_ERR, "sb_info orphan list:"); |
1da177e4 LT |
396 | list_for_each(l, &sbi->s_orphan) { |
397 | struct inode *inode = orphan_list_entry(l); | |
4cf46b67 | 398 | ext3_msg(sb, KERN_ERR, " " |
eee194e7 | 399 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", |
1da177e4 | 400 | inode->i_sb->s_id, inode->i_ino, inode, |
ae6ddcc5 | 401 | inode->i_mode, inode->i_nlink, |
1da177e4 LT |
402 | NEXT_ORPHAN(inode)); |
403 | } | |
404 | } | |
405 | ||
406 | static void ext3_put_super (struct super_block * sb) | |
407 | { | |
408 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
409 | struct ext3_super_block *es = sbi->s_es; | |
44d6f787 | 410 | int i, err; |
1da177e4 | 411 | |
e0ccfd95 | 412 | dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); |
1da177e4 | 413 | ext3_xattr_put_super(sb); |
44d6f787 HK |
414 | err = journal_destroy(sbi->s_journal); |
415 | sbi->s_journal = NULL; | |
416 | if (err < 0) | |
2d7c820e | 417 | ext3_abort(sb, __func__, "Couldn't clean up the journal"); |
44d6f787 | 418 | |
1da177e4 LT |
419 | if (!(sb->s_flags & MS_RDONLY)) { |
420 | EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
421 | es->s_state = cpu_to_le16(sbi->s_mount_state); | |
422 | BUFFER_TRACE(sbi->s_sbh, "marking dirty"); | |
423 | mark_buffer_dirty(sbi->s_sbh); | |
424 | ext3_commit_super(sb, es, 1); | |
425 | } | |
426 | ||
427 | for (i = 0; i < sbi->s_gdb_count; i++) | |
428 | brelse(sbi->s_group_desc[i]); | |
429 | kfree(sbi->s_group_desc); | |
430 | percpu_counter_destroy(&sbi->s_freeblocks_counter); | |
431 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
432 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
433 | brelse(sbi->s_sbh); | |
434 | #ifdef CONFIG_QUOTA | |
435 | for (i = 0; i < MAXQUOTAS; i++) | |
436 | kfree(sbi->s_qf_names[i]); | |
437 | #endif | |
438 | ||
439 | /* Debugging code just in case the in-memory inode orphan list | |
440 | * isn't empty. The on-disk one can be non-empty if we've | |
441 | * detected an error and taken the fs readonly, but the | |
442 | * in-memory list had better be clean by this point. */ | |
443 | if (!list_empty(&sbi->s_orphan)) | |
444 | dump_orphan_list(sb, sbi); | |
445 | J_ASSERT(list_empty(&sbi->s_orphan)); | |
446 | ||
f98393a6 | 447 | invalidate_bdev(sb->s_bdev); |
1da177e4 LT |
448 | if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { |
449 | /* | |
450 | * Invalidate the journal device's buffers. We don't want them | |
451 | * floating about in memory - the physical journal device may | |
452 | * hotswapped, and it breaks the `ro-after' testing code. | |
453 | */ | |
454 | sync_blockdev(sbi->journal_bdev); | |
f98393a6 | 455 | invalidate_bdev(sbi->journal_bdev); |
1da177e4 LT |
456 | ext3_blkdev_remove(sbi); |
457 | } | |
458 | sb->s_fs_info = NULL; | |
5df096d6 | 459 | kfree(sbi->s_blockgroup_lock); |
1da177e4 | 460 | kfree(sbi); |
1da177e4 LT |
461 | } |
462 | ||
e18b890b | 463 | static struct kmem_cache *ext3_inode_cachep; |
1da177e4 LT |
464 | |
465 | /* | |
466 | * Called inside transaction, so use GFP_NOFS | |
467 | */ | |
468 | static struct inode *ext3_alloc_inode(struct super_block *sb) | |
469 | { | |
470 | struct ext3_inode_info *ei; | |
471 | ||
e6b4f8da | 472 | ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS); |
1da177e4 LT |
473 | if (!ei) |
474 | return NULL; | |
1da177e4 LT |
475 | ei->i_block_alloc_info = NULL; |
476 | ei->vfs_inode.i_version = 1; | |
fe8bc91c JK |
477 | atomic_set(&ei->i_datasync_tid, 0); |
478 | atomic_set(&ei->i_sync_tid, 0); | |
1da177e4 LT |
479 | return &ei->vfs_inode; |
480 | } | |
481 | ||
785c4bcc LC |
482 | static int ext3_drop_inode(struct inode *inode) |
483 | { | |
484 | int drop = generic_drop_inode(inode); | |
485 | ||
486 | trace_ext3_drop_inode(inode, drop); | |
487 | return drop; | |
488 | } | |
489 | ||
fa0d7e3d NP |
490 | static void ext3_i_callback(struct rcu_head *head) |
491 | { | |
492 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
fa0d7e3d NP |
493 | kmem_cache_free(ext3_inode_cachep, EXT3_I(inode)); |
494 | } | |
495 | ||
1da177e4 LT |
496 | static void ext3_destroy_inode(struct inode *inode) |
497 | { | |
9f7dd93d VA |
498 | if (!list_empty(&(EXT3_I(inode)->i_orphan))) { |
499 | printk("EXT3 Inode %p: orphan list check failed!\n", | |
500 | EXT3_I(inode)); | |
501 | print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, | |
502 | EXT3_I(inode), sizeof(struct ext3_inode_info), | |
503 | false); | |
504 | dump_stack(); | |
505 | } | |
fa0d7e3d | 506 | call_rcu(&inode->i_rcu, ext3_i_callback); |
1da177e4 LT |
507 | } |
508 | ||
51cc5068 | 509 | static void init_once(void *foo) |
1da177e4 LT |
510 | { |
511 | struct ext3_inode_info *ei = (struct ext3_inode_info *) foo; | |
512 | ||
a35afb83 | 513 | INIT_LIST_HEAD(&ei->i_orphan); |
1da177e4 | 514 | #ifdef CONFIG_EXT3_FS_XATTR |
a35afb83 | 515 | init_rwsem(&ei->xattr_sem); |
1da177e4 | 516 | #endif |
a35afb83 CL |
517 | mutex_init(&ei->truncate_mutex); |
518 | inode_init_once(&ei->vfs_inode); | |
1da177e4 | 519 | } |
ae6ddcc5 | 520 | |
1da177e4 LT |
521 | static int init_inodecache(void) |
522 | { | |
523 | ext3_inode_cachep = kmem_cache_create("ext3_inode_cache", | |
524 | sizeof(struct ext3_inode_info), | |
fffb60f9 PJ |
525 | 0, (SLAB_RECLAIM_ACCOUNT| |
526 | SLAB_MEM_SPREAD), | |
20c2df83 | 527 | init_once); |
1da177e4 LT |
528 | if (ext3_inode_cachep == NULL) |
529 | return -ENOMEM; | |
530 | return 0; | |
531 | } | |
532 | ||
533 | static void destroy_inodecache(void) | |
534 | { | |
8c0a8537 KS |
535 | /* |
536 | * Make sure all delayed rcu free inodes are flushed before we | |
537 | * destroy cache. | |
538 | */ | |
539 | rcu_barrier(); | |
1a1d92c1 | 540 | kmem_cache_destroy(ext3_inode_cachep); |
1da177e4 LT |
541 | } |
542 | ||
6cd37cda | 543 | static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb) |
8fc2751b | 544 | { |
6cd37cda | 545 | #if defined(CONFIG_QUOTA) |
275abf5b | 546 | struct ext3_sb_info *sbi = EXT3_SB(sb); |
8fc2751b | 547 | |
1aeec434 JK |
548 | if (sbi->s_jquota_fmt) { |
549 | char *fmtname = ""; | |
550 | ||
551 | switch (sbi->s_jquota_fmt) { | |
552 | case QFMT_VFS_OLD: | |
553 | fmtname = "vfsold"; | |
554 | break; | |
555 | case QFMT_VFS_V0: | |
556 | fmtname = "vfsv0"; | |
557 | break; | |
558 | case QFMT_VFS_V1: | |
559 | fmtname = "vfsv1"; | |
560 | break; | |
561 | } | |
562 | seq_printf(seq, ",jqfmt=%s", fmtname); | |
563 | } | |
8fc2751b MB |
564 | |
565 | if (sbi->s_qf_names[USRQUOTA]) | |
566 | seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); | |
567 | ||
568 | if (sbi->s_qf_names[GRPQUOTA]) | |
569 | seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); | |
570 | ||
e3c96435 | 571 | if (test_opt(sb, USRQUOTA)) |
8fc2751b MB |
572 | seq_puts(seq, ",usrquota"); |
573 | ||
e3c96435 | 574 | if (test_opt(sb, GRPQUOTA)) |
8fc2751b MB |
575 | seq_puts(seq, ",grpquota"); |
576 | #endif | |
6cd37cda PO |
577 | } |
578 | ||
3c4cec65 JK |
579 | static char *data_mode_string(unsigned long mode) |
580 | { | |
581 | switch (mode) { | |
582 | case EXT3_MOUNT_JOURNAL_DATA: | |
583 | return "journal"; | |
584 | case EXT3_MOUNT_ORDERED_DATA: | |
585 | return "ordered"; | |
586 | case EXT3_MOUNT_WRITEBACK_DATA: | |
587 | return "writeback"; | |
588 | } | |
589 | return "unknown"; | |
590 | } | |
591 | ||
571beed8 MS |
592 | /* |
593 | * Show an option if | |
594 | * - it's set to a non-default value OR | |
595 | * - if the per-sb default is different from the global default | |
596 | */ | |
34c80b1d | 597 | static int ext3_show_options(struct seq_file *seq, struct dentry *root) |
6cd37cda | 598 | { |
34c80b1d | 599 | struct super_block *sb = root->d_sb; |
571beed8 MS |
600 | struct ext3_sb_info *sbi = EXT3_SB(sb); |
601 | struct ext3_super_block *es = sbi->s_es; | |
602 | unsigned long def_mount_opts; | |
603 | ||
604 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
605 | ||
606 | if (sbi->s_sb_block != 1) | |
607 | seq_printf(seq, ",sb=%lu", sbi->s_sb_block); | |
608 | if (test_opt(sb, MINIX_DF)) | |
609 | seq_puts(seq, ",minixdf"); | |
610 | if (test_opt(sb, GRPID)) | |
611 | seq_puts(seq, ",grpid"); | |
612 | if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS)) | |
613 | seq_puts(seq, ",nogrpid"); | |
1523299d | 614 | if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT3_DEF_RESUID)) || |
571beed8 | 615 | le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) { |
1523299d EB |
616 | seq_printf(seq, ",resuid=%u", |
617 | from_kuid_munged(&init_user_ns, sbi->s_resuid)); | |
571beed8 | 618 | } |
1523299d | 619 | if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT3_DEF_RESGID)) || |
571beed8 | 620 | le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) { |
1523299d EB |
621 | seq_printf(seq, ",resgid=%u", |
622 | from_kgid_munged(&init_user_ns, sbi->s_resgid)); | |
571beed8 | 623 | } |
1eca93f9 | 624 | if (test_opt(sb, ERRORS_RO)) { |
571beed8 MS |
625 | int def_errors = le16_to_cpu(es->s_errors); |
626 | ||
627 | if (def_errors == EXT3_ERRORS_PANIC || | |
1eca93f9 AK |
628 | def_errors == EXT3_ERRORS_CONTINUE) { |
629 | seq_puts(seq, ",errors=remount-ro"); | |
571beed8 MS |
630 | } |
631 | } | |
1eca93f9 AK |
632 | if (test_opt(sb, ERRORS_CONT)) |
633 | seq_puts(seq, ",errors=continue"); | |
571beed8 MS |
634 | if (test_opt(sb, ERRORS_PANIC)) |
635 | seq_puts(seq, ",errors=panic"); | |
636 | if (test_opt(sb, NO_UID32)) | |
637 | seq_puts(seq, ",nouid32"); | |
638 | if (test_opt(sb, DEBUG)) | |
639 | seq_puts(seq, ",debug"); | |
571beed8 MS |
640 | #ifdef CONFIG_EXT3_FS_XATTR |
641 | if (test_opt(sb, XATTR_USER)) | |
642 | seq_puts(seq, ",user_xattr"); | |
643 | if (!test_opt(sb, XATTR_USER) && | |
644 | (def_mount_opts & EXT3_DEFM_XATTR_USER)) { | |
645 | seq_puts(seq, ",nouser_xattr"); | |
646 | } | |
647 | #endif | |
648 | #ifdef CONFIG_EXT3_FS_POSIX_ACL | |
649 | if (test_opt(sb, POSIX_ACL)) | |
650 | seq_puts(seq, ",acl"); | |
651 | if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL)) | |
652 | seq_puts(seq, ",noacl"); | |
653 | #endif | |
654 | if (!test_opt(sb, RESERVATION)) | |
655 | seq_puts(seq, ",noreservation"); | |
656 | if (sbi->s_commit_interval) { | |
657 | seq_printf(seq, ",commit=%u", | |
658 | (unsigned) (sbi->s_commit_interval / HZ)); | |
659 | } | |
0636c73e ES |
660 | |
661 | /* | |
662 | * Always display barrier state so it's clear what the status is. | |
663 | */ | |
664 | seq_puts(seq, ",barrier="); | |
665 | seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0"); | |
e3c96435 | 666 | seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS))); |
0e4fb5e2 HK |
667 | if (test_opt(sb, DATA_ERR_ABORT)) |
668 | seq_puts(seq, ",data_err=abort"); | |
669 | ||
dee1d3b6 ES |
670 | if (test_opt(sb, NOLOAD)) |
671 | seq_puts(seq, ",norecovery"); | |
672 | ||
6cd37cda | 673 | ext3_show_quota_options(seq, sb); |
8fc2751b MB |
674 | |
675 | return 0; | |
676 | } | |
1da177e4 | 677 | |
fdb36673 | 678 | |
74af0baa CH |
679 | static struct inode *ext3_nfs_get_inode(struct super_block *sb, |
680 | u64 ino, u32 generation) | |
fdb36673 | 681 | { |
fdb36673 | 682 | struct inode *inode; |
fdb36673 N |
683 | |
684 | if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO) | |
685 | return ERR_PTR(-ESTALE); | |
686 | if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count)) | |
687 | return ERR_PTR(-ESTALE); | |
688 | ||
689 | /* iget isn't really right if the inode is currently unallocated!! | |
690 | * | |
691 | * ext3_read_inode will return a bad_inode if the inode had been | |
692 | * deleted, so we should be safe. | |
693 | * | |
694 | * Currently we don't know the generation for parent directory, so | |
695 | * a generation of 0 means "accept any" | |
696 | */ | |
473043dc DH |
697 | inode = ext3_iget(sb, ino); |
698 | if (IS_ERR(inode)) | |
699 | return ERR_CAST(inode); | |
700 | if (generation && inode->i_generation != generation) { | |
fdb36673 N |
701 | iput(inode); |
702 | return ERR_PTR(-ESTALE); | |
703 | } | |
74af0baa CH |
704 | |
705 | return inode; | |
706 | } | |
707 | ||
708 | static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
709 | int fh_len, int fh_type) | |
710 | { | |
711 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
712 | ext3_nfs_get_inode); | |
713 | } | |
714 | ||
715 | static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid, | |
716 | int fh_len, int fh_type) | |
717 | { | |
718 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
719 | ext3_nfs_get_inode); | |
fdb36673 N |
720 | } |
721 | ||
6b082b53 TO |
722 | /* |
723 | * Try to release metadata pages (indirect blocks, directories) which are | |
724 | * mapped via the block device. Since these pages could have journal heads | |
725 | * which would prevent try_to_free_buffers() from freeing them, we must use | |
726 | * jbd layer's try_to_free_buffers() function to release them. | |
727 | */ | |
728 | static int bdev_try_to_free_page(struct super_block *sb, struct page *page, | |
729 | gfp_t wait) | |
730 | { | |
731 | journal_t *journal = EXT3_SB(sb)->s_journal; | |
732 | ||
733 | WARN_ON(PageChecked(page)); | |
734 | if (!page_has_buffers(page)) | |
735 | return 0; | |
736 | if (journal) | |
737 | return journal_try_to_free_buffers(journal, page, | |
738 | wait & ~__GFP_WAIT); | |
739 | return try_to_free_buffers(page); | |
740 | } | |
741 | ||
8fc2751b | 742 | #ifdef CONFIG_QUOTA |
1da177e4 LT |
743 | #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group") |
744 | #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) | |
745 | ||
1da177e4 LT |
746 | static int ext3_write_dquot(struct dquot *dquot); |
747 | static int ext3_acquire_dquot(struct dquot *dquot); | |
748 | static int ext3_release_dquot(struct dquot *dquot); | |
749 | static int ext3_mark_dquot_dirty(struct dquot *dquot); | |
750 | static int ext3_write_info(struct super_block *sb, int type); | |
2fd83a4f | 751 | static int ext3_quota_on(struct super_block *sb, int type, int format_id, |
f00c9e44 | 752 | struct path *path); |
1da177e4 LT |
753 | static int ext3_quota_on_mount(struct super_block *sb, int type); |
754 | static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data, | |
755 | size_t len, loff_t off); | |
756 | static ssize_t ext3_quota_write(struct super_block *sb, int type, | |
757 | const char *data, size_t len, loff_t off); | |
758 | ||
61e225dc | 759 | static const struct dquot_operations ext3_quota_operations = { |
1da177e4 LT |
760 | .write_dquot = ext3_write_dquot, |
761 | .acquire_dquot = ext3_acquire_dquot, | |
762 | .release_dquot = ext3_release_dquot, | |
763 | .mark_dirty = ext3_mark_dquot_dirty, | |
157091a2 JK |
764 | .write_info = ext3_write_info, |
765 | .alloc_dquot = dquot_alloc, | |
766 | .destroy_dquot = dquot_destroy, | |
1da177e4 LT |
767 | }; |
768 | ||
0d54b217 | 769 | static const struct quotactl_ops ext3_qctl_operations = { |
1da177e4 | 770 | .quota_on = ext3_quota_on, |
287a8095 CH |
771 | .quota_off = dquot_quota_off, |
772 | .quota_sync = dquot_quota_sync, | |
773 | .get_info = dquot_get_dqinfo, | |
774 | .set_info = dquot_set_dqinfo, | |
775 | .get_dqblk = dquot_get_dqblk, | |
776 | .set_dqblk = dquot_set_dqblk | |
1da177e4 LT |
777 | }; |
778 | #endif | |
779 | ||
ee9b6d61 | 780 | static const struct super_operations ext3_sops = { |
1da177e4 LT |
781 | .alloc_inode = ext3_alloc_inode, |
782 | .destroy_inode = ext3_destroy_inode, | |
1da177e4 LT |
783 | .write_inode = ext3_write_inode, |
784 | .dirty_inode = ext3_dirty_inode, | |
785c4bcc | 785 | .drop_inode = ext3_drop_inode, |
ac14a95b | 786 | .evict_inode = ext3_evict_inode, |
1da177e4 | 787 | .put_super = ext3_put_super, |
1da177e4 | 788 | .sync_fs = ext3_sync_fs, |
c4be0c1d TS |
789 | .freeze_fs = ext3_freeze, |
790 | .unfreeze_fs = ext3_unfreeze, | |
1da177e4 LT |
791 | .statfs = ext3_statfs, |
792 | .remount_fs = ext3_remount, | |
8fc2751b | 793 | .show_options = ext3_show_options, |
1da177e4 LT |
794 | #ifdef CONFIG_QUOTA |
795 | .quota_read = ext3_quota_read, | |
796 | .quota_write = ext3_quota_write, | |
797 | #endif | |
6b082b53 | 798 | .bdev_try_to_free_page = bdev_try_to_free_page, |
1da177e4 LT |
799 | }; |
800 | ||
39655164 | 801 | static const struct export_operations ext3_export_ops = { |
74af0baa CH |
802 | .fh_to_dentry = ext3_fh_to_dentry, |
803 | .fh_to_parent = ext3_fh_to_parent, | |
1da177e4 LT |
804 | .get_parent = ext3_get_parent, |
805 | }; | |
806 | ||
807 | enum { | |
808 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
809 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
2860b733 | 810 | Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov, |
1da177e4 | 811 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, |
ade1a29e | 812 | Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh, |
71b96257 | 813 | Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev, |
1da177e4 | 814 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, |
0e4fb5e2 | 815 | Opt_data_err_abort, Opt_data_err_ignore, |
1da177e4 | 816 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, |
1aeec434 | 817 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, |
0636c73e ES |
818 | Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, |
819 | Opt_resize, Opt_usrquota, Opt_grpquota | |
1da177e4 LT |
820 | }; |
821 | ||
a447c093 | 822 | static const match_table_t tokens = { |
1da177e4 LT |
823 | {Opt_bsd_df, "bsddf"}, |
824 | {Opt_minix_df, "minixdf"}, | |
825 | {Opt_grpid, "grpid"}, | |
826 | {Opt_grpid, "bsdgroups"}, | |
827 | {Opt_nogrpid, "nogrpid"}, | |
828 | {Opt_nogrpid, "sysvgroups"}, | |
829 | {Opt_resgid, "resgid=%u"}, | |
830 | {Opt_resuid, "resuid=%u"}, | |
831 | {Opt_sb, "sb=%u"}, | |
832 | {Opt_err_cont, "errors=continue"}, | |
833 | {Opt_err_panic, "errors=panic"}, | |
834 | {Opt_err_ro, "errors=remount-ro"}, | |
835 | {Opt_nouid32, "nouid32"}, | |
836 | {Opt_nocheck, "nocheck"}, | |
837 | {Opt_nocheck, "check=none"}, | |
1da177e4 LT |
838 | {Opt_debug, "debug"}, |
839 | {Opt_oldalloc, "oldalloc"}, | |
840 | {Opt_orlov, "orlov"}, | |
841 | {Opt_user_xattr, "user_xattr"}, | |
842 | {Opt_nouser_xattr, "nouser_xattr"}, | |
843 | {Opt_acl, "acl"}, | |
844 | {Opt_noacl, "noacl"}, | |
845 | {Opt_reservation, "reservation"}, | |
846 | {Opt_noreservation, "noreservation"}, | |
847 | {Opt_noload, "noload"}, | |
dee1d3b6 | 848 | {Opt_noload, "norecovery"}, |
1da177e4 | 849 | {Opt_nobh, "nobh"}, |
ade1a29e | 850 | {Opt_bh, "bh"}, |
1da177e4 LT |
851 | {Opt_commit, "commit=%u"}, |
852 | {Opt_journal_update, "journal=update"}, | |
853 | {Opt_journal_inum, "journal=%u"}, | |
71b96257 | 854 | {Opt_journal_dev, "journal_dev=%u"}, |
1da177e4 LT |
855 | {Opt_abort, "abort"}, |
856 | {Opt_data_journal, "data=journal"}, | |
857 | {Opt_data_ordered, "data=ordered"}, | |
858 | {Opt_data_writeback, "data=writeback"}, | |
0e4fb5e2 HK |
859 | {Opt_data_err_abort, "data_err=abort"}, |
860 | {Opt_data_err_ignore, "data_err=ignore"}, | |
1da177e4 LT |
861 | {Opt_offusrjquota, "usrjquota="}, |
862 | {Opt_usrjquota, "usrjquota=%s"}, | |
863 | {Opt_offgrpjquota, "grpjquota="}, | |
864 | {Opt_grpjquota, "grpjquota=%s"}, | |
865 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
866 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
1aeec434 | 867 | {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, |
8fc2751b | 868 | {Opt_grpquota, "grpquota"}, |
1f54587b JK |
869 | {Opt_noquota, "noquota"}, |
870 | {Opt_quota, "quota"}, | |
8fc2751b | 871 | {Opt_usrquota, "usrquota"}, |
1da177e4 | 872 | {Opt_barrier, "barrier=%u"}, |
0636c73e ES |
873 | {Opt_barrier, "barrier"}, |
874 | {Opt_nobarrier, "nobarrier"}, | |
1da177e4 | 875 | {Opt_resize, "resize"}, |
92587216 | 876 | {Opt_err, NULL}, |
1da177e4 LT |
877 | }; |
878 | ||
4cf46b67 | 879 | static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb) |
1da177e4 | 880 | { |
e9ad5620 DK |
881 | ext3_fsblk_t sb_block; |
882 | char *options = (char *) *data; | |
1da177e4 LT |
883 | |
884 | if (!options || strncmp(options, "sb=", 3) != 0) | |
885 | return 1; /* Default location */ | |
886 | options += 3; | |
43d23f90 | 887 | /*todo: use simple_strtoll with >32bit ext3 */ |
1da177e4 LT |
888 | sb_block = simple_strtoul(options, &options, 0); |
889 | if (*options && *options != ',') { | |
4cf46b67 | 890 | ext3_msg(sb, "error: invalid sb specification: %s", |
1da177e4 LT |
891 | (char *) *data); |
892 | return 1; | |
893 | } | |
894 | if (*options == ',') | |
895 | options++; | |
896 | *data = (void *) options; | |
897 | return sb_block; | |
898 | } | |
899 | ||
e1f5c67a DM |
900 | #ifdef CONFIG_QUOTA |
901 | static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) | |
902 | { | |
903 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
904 | char *qname; | |
905 | ||
906 | if (sb_any_quota_loaded(sb) && | |
907 | !sbi->s_qf_names[qtype]) { | |
908 | ext3_msg(sb, KERN_ERR, | |
909 | "Cannot change journaled " | |
910 | "quota options when quota turned on"); | |
911 | return 0; | |
912 | } | |
913 | qname = match_strdup(args); | |
914 | if (!qname) { | |
915 | ext3_msg(sb, KERN_ERR, | |
916 | "Not enough memory for storing quotafile name"); | |
917 | return 0; | |
918 | } | |
919 | if (sbi->s_qf_names[qtype] && | |
920 | strcmp(sbi->s_qf_names[qtype], qname)) { | |
921 | ext3_msg(sb, KERN_ERR, | |
922 | "%s quota file already specified", QTYPE2NAME(qtype)); | |
923 | kfree(qname); | |
924 | return 0; | |
925 | } | |
926 | sbi->s_qf_names[qtype] = qname; | |
927 | if (strchr(sbi->s_qf_names[qtype], '/')) { | |
928 | ext3_msg(sb, KERN_ERR, | |
929 | "quotafile must be on filesystem root"); | |
930 | kfree(sbi->s_qf_names[qtype]); | |
931 | sbi->s_qf_names[qtype] = NULL; | |
932 | return 0; | |
933 | } | |
934 | set_opt(sbi->s_mount_opt, QUOTA); | |
935 | return 1; | |
936 | } | |
937 | ||
938 | static int clear_qf_name(struct super_block *sb, int qtype) { | |
939 | ||
940 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
941 | ||
942 | if (sb_any_quota_loaded(sb) && | |
943 | sbi->s_qf_names[qtype]) { | |
944 | ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options" | |
945 | " when quota turned on"); | |
946 | return 0; | |
947 | } | |
948 | /* | |
949 | * The space will be released later when all options are confirmed | |
950 | * to be correct | |
951 | */ | |
952 | sbi->s_qf_names[qtype] = NULL; | |
953 | return 1; | |
954 | } | |
955 | #endif | |
956 | ||
71b96257 | 957 | static int parse_options (char *options, struct super_block *sb, |
eee194e7 | 958 | unsigned int *inum, unsigned long *journal_devnum, |
43d23f90 | 959 | ext3_fsblk_t *n_blocks_count, int is_remount) |
1da177e4 LT |
960 | { |
961 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
962 | char * p; | |
963 | substring_t args[MAX_OPT_ARGS]; | |
964 | int data_opt = 0; | |
965 | int option; | |
1523299d EB |
966 | kuid_t uid; |
967 | kgid_t gid; | |
1da177e4 | 968 | #ifdef CONFIG_QUOTA |
e1f5c67a | 969 | int qfmt; |
1da177e4 LT |
970 | #endif |
971 | ||
972 | if (!options) | |
973 | return 1; | |
974 | ||
975 | while ((p = strsep (&options, ",")) != NULL) { | |
976 | int token; | |
977 | if (!*p) | |
978 | continue; | |
0636c73e ES |
979 | /* |
980 | * Initialize args struct so we know whether arg was | |
981 | * found; some options take optional arguments. | |
982 | */ | |
76f59b3b | 983 | args[0].to = args[0].from = NULL; |
1da177e4 LT |
984 | token = match_token(p, tokens, args); |
985 | switch (token) { | |
986 | case Opt_bsd_df: | |
987 | clear_opt (sbi->s_mount_opt, MINIX_DF); | |
988 | break; | |
989 | case Opt_minix_df: | |
990 | set_opt (sbi->s_mount_opt, MINIX_DF); | |
991 | break; | |
992 | case Opt_grpid: | |
993 | set_opt (sbi->s_mount_opt, GRPID); | |
994 | break; | |
995 | case Opt_nogrpid: | |
996 | clear_opt (sbi->s_mount_opt, GRPID); | |
997 | break; | |
998 | case Opt_resuid: | |
999 | if (match_int(&args[0], &option)) | |
1000 | return 0; | |
1523299d EB |
1001 | uid = make_kuid(current_user_ns(), option); |
1002 | if (!uid_valid(uid)) { | |
1003 | ext3_msg(sb, KERN_ERR, "Invalid uid value %d", option); | |
66415f6b | 1004 | return 0; |
1523299d EB |
1005 | |
1006 | } | |
1007 | sbi->s_resuid = uid; | |
1da177e4 LT |
1008 | break; |
1009 | case Opt_resgid: | |
1010 | if (match_int(&args[0], &option)) | |
1011 | return 0; | |
1523299d EB |
1012 | gid = make_kgid(current_user_ns(), option); |
1013 | if (!gid_valid(gid)) { | |
1014 | ext3_msg(sb, KERN_ERR, "Invalid gid value %d", option); | |
66415f6b | 1015 | return 0; |
1523299d EB |
1016 | } |
1017 | sbi->s_resgid = gid; | |
1da177e4 LT |
1018 | break; |
1019 | case Opt_sb: | |
1020 | /* handled by get_sb_block() instead of here */ | |
1021 | /* *sb_block = match_int(&args[0]); */ | |
1022 | break; | |
1023 | case Opt_err_panic: | |
1024 | clear_opt (sbi->s_mount_opt, ERRORS_CONT); | |
1025 | clear_opt (sbi->s_mount_opt, ERRORS_RO); | |
1026 | set_opt (sbi->s_mount_opt, ERRORS_PANIC); | |
1027 | break; | |
1028 | case Opt_err_ro: | |
1029 | clear_opt (sbi->s_mount_opt, ERRORS_CONT); | |
1030 | clear_opt (sbi->s_mount_opt, ERRORS_PANIC); | |
1031 | set_opt (sbi->s_mount_opt, ERRORS_RO); | |
1032 | break; | |
1033 | case Opt_err_cont: | |
1034 | clear_opt (sbi->s_mount_opt, ERRORS_RO); | |
1035 | clear_opt (sbi->s_mount_opt, ERRORS_PANIC); | |
1036 | set_opt (sbi->s_mount_opt, ERRORS_CONT); | |
1037 | break; | |
1038 | case Opt_nouid32: | |
1039 | set_opt (sbi->s_mount_opt, NO_UID32); | |
1040 | break; | |
1da177e4 LT |
1041 | case Opt_nocheck: |
1042 | clear_opt (sbi->s_mount_opt, CHECK); | |
1043 | break; | |
1044 | case Opt_debug: | |
1045 | set_opt (sbi->s_mount_opt, DEBUG); | |
1046 | break; | |
1047 | case Opt_oldalloc: | |
fbc85402 LC |
1048 | ext3_msg(sb, KERN_WARNING, |
1049 | "Ignoring deprecated oldalloc option"); | |
1da177e4 LT |
1050 | break; |
1051 | case Opt_orlov: | |
fbc85402 LC |
1052 | ext3_msg(sb, KERN_WARNING, |
1053 | "Ignoring deprecated orlov option"); | |
1da177e4 LT |
1054 | break; |
1055 | #ifdef CONFIG_EXT3_FS_XATTR | |
1056 | case Opt_user_xattr: | |
1057 | set_opt (sbi->s_mount_opt, XATTR_USER); | |
1058 | break; | |
1059 | case Opt_nouser_xattr: | |
1060 | clear_opt (sbi->s_mount_opt, XATTR_USER); | |
1061 | break; | |
1062 | #else | |
1063 | case Opt_user_xattr: | |
1064 | case Opt_nouser_xattr: | |
4cf46b67 AF |
1065 | ext3_msg(sb, KERN_INFO, |
1066 | "(no)user_xattr options not supported"); | |
1da177e4 LT |
1067 | break; |
1068 | #endif | |
1069 | #ifdef CONFIG_EXT3_FS_POSIX_ACL | |
1070 | case Opt_acl: | |
1071 | set_opt(sbi->s_mount_opt, POSIX_ACL); | |
1072 | break; | |
1073 | case Opt_noacl: | |
1074 | clear_opt(sbi->s_mount_opt, POSIX_ACL); | |
1075 | break; | |
1076 | #else | |
1077 | case Opt_acl: | |
1078 | case Opt_noacl: | |
4cf46b67 AF |
1079 | ext3_msg(sb, KERN_INFO, |
1080 | "(no)acl options not supported"); | |
1da177e4 LT |
1081 | break; |
1082 | #endif | |
1083 | case Opt_reservation: | |
1084 | set_opt(sbi->s_mount_opt, RESERVATION); | |
1085 | break; | |
1086 | case Opt_noreservation: | |
1087 | clear_opt(sbi->s_mount_opt, RESERVATION); | |
1088 | break; | |
1089 | case Opt_journal_update: | |
1090 | /* @@@ FIXME */ | |
1091 | /* Eventually we will want to be able to create | |
1092 | a journal file here. For now, only allow the | |
1093 | user to specify an existing inode to be the | |
1094 | journal file. */ | |
1095 | if (is_remount) { | |
4cf46b67 AF |
1096 | ext3_msg(sb, KERN_ERR, "error: cannot specify " |
1097 | "journal on remount"); | |
1da177e4 LT |
1098 | return 0; |
1099 | } | |
1100 | set_opt (sbi->s_mount_opt, UPDATE_JOURNAL); | |
1101 | break; | |
1102 | case Opt_journal_inum: | |
1103 | if (is_remount) { | |
4cf46b67 AF |
1104 | ext3_msg(sb, KERN_ERR, "error: cannot specify " |
1105 | "journal on remount"); | |
1da177e4 LT |
1106 | return 0; |
1107 | } | |
1108 | if (match_int(&args[0], &option)) | |
1109 | return 0; | |
1110 | *inum = option; | |
1111 | break; | |
71b96257 JL |
1112 | case Opt_journal_dev: |
1113 | if (is_remount) { | |
4cf46b67 AF |
1114 | ext3_msg(sb, KERN_ERR, "error: cannot specify " |
1115 | "journal on remount"); | |
71b96257 JL |
1116 | return 0; |
1117 | } | |
1118 | if (match_int(&args[0], &option)) | |
1119 | return 0; | |
1120 | *journal_devnum = option; | |
1121 | break; | |
1da177e4 LT |
1122 | case Opt_noload: |
1123 | set_opt (sbi->s_mount_opt, NOLOAD); | |
1124 | break; | |
1125 | case Opt_commit: | |
1126 | if (match_int(&args[0], &option)) | |
1127 | return 0; | |
1128 | if (option < 0) | |
1129 | return 0; | |
1130 | if (option == 0) | |
1131 | option = JBD_DEFAULT_MAX_COMMIT_AGE; | |
1132 | sbi->s_commit_interval = HZ * option; | |
1133 | break; | |
1134 | case Opt_data_journal: | |
1135 | data_opt = EXT3_MOUNT_JOURNAL_DATA; | |
1136 | goto datacheck; | |
1137 | case Opt_data_ordered: | |
1138 | data_opt = EXT3_MOUNT_ORDERED_DATA; | |
1139 | goto datacheck; | |
1140 | case Opt_data_writeback: | |
1141 | data_opt = EXT3_MOUNT_WRITEBACK_DATA; | |
1142 | datacheck: | |
1143 | if (is_remount) { | |
e3c96435 | 1144 | if (test_opt(sb, DATA_FLAGS) == data_opt) |
3c4cec65 | 1145 | break; |
4cf46b67 AF |
1146 | ext3_msg(sb, KERN_ERR, |
1147 | "error: cannot change " | |
3c4cec65 JK |
1148 | "data mode on remount. The filesystem " |
1149 | "is mounted in data=%s mode and you " | |
4cf46b67 | 1150 | "try to remount it in data=%s mode.", |
e3c96435 DM |
1151 | data_mode_string(test_opt(sb, |
1152 | DATA_FLAGS)), | |
3c4cec65 JK |
1153 | data_mode_string(data_opt)); |
1154 | return 0; | |
1da177e4 | 1155 | } else { |
e3c96435 | 1156 | clear_opt(sbi->s_mount_opt, DATA_FLAGS); |
1da177e4 LT |
1157 | sbi->s_mount_opt |= data_opt; |
1158 | } | |
1159 | break; | |
0e4fb5e2 HK |
1160 | case Opt_data_err_abort: |
1161 | set_opt(sbi->s_mount_opt, DATA_ERR_ABORT); | |
1162 | break; | |
1163 | case Opt_data_err_ignore: | |
1164 | clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT); | |
1165 | break; | |
1da177e4 LT |
1166 | #ifdef CONFIG_QUOTA |
1167 | case Opt_usrjquota: | |
e1f5c67a | 1168 | if (!set_qf_name(sb, USRQUOTA, &args[0])) |
1da177e4 | 1169 | return 0; |
e1f5c67a DM |
1170 | break; |
1171 | case Opt_grpjquota: | |
1172 | if (!set_qf_name(sb, GRPQUOTA, &args[0])) | |
1da177e4 | 1173 | return 0; |
1da177e4 LT |
1174 | break; |
1175 | case Opt_offusrjquota: | |
e1f5c67a DM |
1176 | if (!clear_qf_name(sb, USRQUOTA)) |
1177 | return 0; | |
1178 | break; | |
1da177e4 | 1179 | case Opt_offgrpjquota: |
e1f5c67a | 1180 | if (!clear_qf_name(sb, GRPQUOTA)) |
1da177e4 | 1181 | return 0; |
1da177e4 LT |
1182 | break; |
1183 | case Opt_jqfmt_vfsold: | |
d06bf1d2 JK |
1184 | qfmt = QFMT_VFS_OLD; |
1185 | goto set_qf_format; | |
1da177e4 | 1186 | case Opt_jqfmt_vfsv0: |
d06bf1d2 | 1187 | qfmt = QFMT_VFS_V0; |
1aeec434 JK |
1188 | goto set_qf_format; |
1189 | case Opt_jqfmt_vfsv1: | |
1190 | qfmt = QFMT_VFS_V1; | |
d06bf1d2 | 1191 | set_qf_format: |
ee0d5ffe | 1192 | if (sb_any_quota_loaded(sb) && |
d06bf1d2 | 1193 | sbi->s_jquota_fmt != qfmt) { |
4cf46b67 | 1194 | ext3_msg(sb, KERN_ERR, "error: cannot change " |
d06bf1d2 | 1195 | "journaled quota options when " |
4cf46b67 | 1196 | "quota turned on."); |
d06bf1d2 JK |
1197 | return 0; |
1198 | } | |
1199 | sbi->s_jquota_fmt = qfmt; | |
1da177e4 | 1200 | break; |
1f54587b | 1201 | case Opt_quota: |
8fc2751b MB |
1202 | case Opt_usrquota: |
1203 | set_opt(sbi->s_mount_opt, QUOTA); | |
1204 | set_opt(sbi->s_mount_opt, USRQUOTA); | |
1205 | break; | |
1206 | case Opt_grpquota: | |
1f54587b | 1207 | set_opt(sbi->s_mount_opt, QUOTA); |
8fc2751b | 1208 | set_opt(sbi->s_mount_opt, GRPQUOTA); |
1f54587b JK |
1209 | break; |
1210 | case Opt_noquota: | |
ee0d5ffe | 1211 | if (sb_any_quota_loaded(sb)) { |
4cf46b67 AF |
1212 | ext3_msg(sb, KERN_ERR, "error: cannot change " |
1213 | "quota options when quota turned on."); | |
1f54587b JK |
1214 | return 0; |
1215 | } | |
1216 | clear_opt(sbi->s_mount_opt, QUOTA); | |
8fc2751b MB |
1217 | clear_opt(sbi->s_mount_opt, USRQUOTA); |
1218 | clear_opt(sbi->s_mount_opt, GRPQUOTA); | |
1f54587b | 1219 | break; |
1da177e4 | 1220 | #else |
8fc2751b MB |
1221 | case Opt_quota: |
1222 | case Opt_usrquota: | |
1223 | case Opt_grpquota: | |
4cf46b67 AF |
1224 | ext3_msg(sb, KERN_ERR, |
1225 | "error: quota options not supported."); | |
fa1ff1e0 | 1226 | break; |
1da177e4 LT |
1227 | case Opt_usrjquota: |
1228 | case Opt_grpjquota: | |
1229 | case Opt_offusrjquota: | |
1230 | case Opt_offgrpjquota: | |
1231 | case Opt_jqfmt_vfsold: | |
1232 | case Opt_jqfmt_vfsv0: | |
1aeec434 | 1233 | case Opt_jqfmt_vfsv1: |
4cf46b67 AF |
1234 | ext3_msg(sb, KERN_ERR, |
1235 | "error: journaled quota options not " | |
1236 | "supported."); | |
1da177e4 | 1237 | break; |
1f54587b JK |
1238 | case Opt_noquota: |
1239 | break; | |
1da177e4 LT |
1240 | #endif |
1241 | case Opt_abort: | |
1242 | set_opt(sbi->s_mount_opt, ABORT); | |
1243 | break; | |
0636c73e ES |
1244 | case Opt_nobarrier: |
1245 | clear_opt(sbi->s_mount_opt, BARRIER); | |
1246 | break; | |
1da177e4 | 1247 | case Opt_barrier: |
0636c73e ES |
1248 | if (args[0].from) { |
1249 | if (match_int(&args[0], &option)) | |
1250 | return 0; | |
1251 | } else | |
1252 | option = 1; /* No argument, default to 1 */ | |
1da177e4 LT |
1253 | if (option) |
1254 | set_opt(sbi->s_mount_opt, BARRIER); | |
1255 | else | |
1256 | clear_opt(sbi->s_mount_opt, BARRIER); | |
1257 | break; | |
1258 | case Opt_ignore: | |
1259 | break; | |
1260 | case Opt_resize: | |
08c6a96f | 1261 | if (!is_remount) { |
4cf46b67 AF |
1262 | ext3_msg(sb, KERN_ERR, |
1263 | "error: resize option only available " | |
1264 | "for remount"); | |
1da177e4 LT |
1265 | return 0; |
1266 | } | |
c7f1721e KZ |
1267 | if (match_int(&args[0], &option) != 0) |
1268 | return 0; | |
1da177e4 LT |
1269 | *n_blocks_count = option; |
1270 | break; | |
1271 | case Opt_nobh: | |
4c4d3901 CH |
1272 | ext3_msg(sb, KERN_WARNING, |
1273 | "warning: ignoring deprecated nobh option"); | |
1da177e4 | 1274 | break; |
ade1a29e | 1275 | case Opt_bh: |
4c4d3901 CH |
1276 | ext3_msg(sb, KERN_WARNING, |
1277 | "warning: ignoring deprecated bh option"); | |
ade1a29e | 1278 | break; |
1da177e4 | 1279 | default: |
4cf46b67 AF |
1280 | ext3_msg(sb, KERN_ERR, |
1281 | "error: unrecognized mount option \"%s\" " | |
1282 | "or missing value", p); | |
1da177e4 LT |
1283 | return 0; |
1284 | } | |
1285 | } | |
1286 | #ifdef CONFIG_QUOTA | |
8fc2751b | 1287 | if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { |
e3c96435 | 1288 | if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) |
8fc2751b | 1289 | clear_opt(sbi->s_mount_opt, USRQUOTA); |
e3c96435 | 1290 | if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) |
8fc2751b MB |
1291 | clear_opt(sbi->s_mount_opt, GRPQUOTA); |
1292 | ||
e1f5c67a | 1293 | if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { |
4cf46b67 AF |
1294 | ext3_msg(sb, KERN_ERR, "error: old and new quota " |
1295 | "format mixing."); | |
8fc2751b MB |
1296 | return 0; |
1297 | } | |
1298 | ||
1299 | if (!sbi->s_jquota_fmt) { | |
4cf46b67 AF |
1300 | ext3_msg(sb, KERN_ERR, "error: journaled quota format " |
1301 | "not specified."); | |
8fc2751b MB |
1302 | return 0; |
1303 | } | |
1304 | } else { | |
1305 | if (sbi->s_jquota_fmt) { | |
4cf46b67 | 1306 | ext3_msg(sb, KERN_ERR, "error: journaled quota format " |
99aeaf63 | 1307 | "specified with no journaling " |
4cf46b67 | 1308 | "enabled."); |
8fc2751b MB |
1309 | return 0; |
1310 | } | |
1da177e4 LT |
1311 | } |
1312 | #endif | |
1da177e4 LT |
1313 | return 1; |
1314 | } | |
1315 | ||
1316 | static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es, | |
1317 | int read_only) | |
1318 | { | |
1319 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
1320 | int res = 0; | |
1321 | ||
1322 | if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) { | |
4cf46b67 AF |
1323 | ext3_msg(sb, KERN_ERR, |
1324 | "error: revision level too high, " | |
1325 | "forcing read-only mode"); | |
1da177e4 LT |
1326 | res = MS_RDONLY; |
1327 | } | |
1328 | if (read_only) | |
1329 | return res; | |
1330 | if (!(sbi->s_mount_state & EXT3_VALID_FS)) | |
4cf46b67 AF |
1331 | ext3_msg(sb, KERN_WARNING, |
1332 | "warning: mounting unchecked fs, " | |
1333 | "running e2fsck is recommended"); | |
1da177e4 | 1334 | else if ((sbi->s_mount_state & EXT3_ERROR_FS)) |
4cf46b67 AF |
1335 | ext3_msg(sb, KERN_WARNING, |
1336 | "warning: mounting fs with errors, " | |
1337 | "running e2fsck is recommended"); | |
df0d6b8f | 1338 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && |
1da177e4 | 1339 | le16_to_cpu(es->s_mnt_count) >= |
df0d6b8f | 1340 | le16_to_cpu(es->s_max_mnt_count)) |
4cf46b67 AF |
1341 | ext3_msg(sb, KERN_WARNING, |
1342 | "warning: maximal mount count reached, " | |
1343 | "running e2fsck is recommended"); | |
1da177e4 LT |
1344 | else if (le32_to_cpu(es->s_checkinterval) && |
1345 | (le32_to_cpu(es->s_lastcheck) + | |
1346 | le32_to_cpu(es->s_checkinterval) <= get_seconds())) | |
4cf46b67 AF |
1347 | ext3_msg(sb, KERN_WARNING, |
1348 | "warning: checktime reached, " | |
1349 | "running e2fsck is recommended"); | |
1da177e4 LT |
1350 | #if 0 |
1351 | /* @@@ We _will_ want to clear the valid bit if we find | |
1352 | inconsistencies, to force a fsck at reboot. But for | |
1353 | a plain journaled filesystem we can keep it set as | |
1354 | valid forever! :) */ | |
e7f23ebd | 1355 | es->s_state &= cpu_to_le16(~EXT3_VALID_FS); |
1da177e4 | 1356 | #endif |
df0d6b8f | 1357 | if (!le16_to_cpu(es->s_max_mnt_count)) |
1da177e4 | 1358 | es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT); |
50e8a289 | 1359 | le16_add_cpu(&es->s_mnt_count, 1); |
1da177e4 LT |
1360 | es->s_mtime = cpu_to_le32(get_seconds()); |
1361 | ext3_update_dynamic_rev(sb); | |
1362 | EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
1363 | ||
1364 | ext3_commit_super(sb, es, 1); | |
1365 | if (test_opt(sb, DEBUG)) | |
4cf46b67 AF |
1366 | ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, " |
1367 | "bpg=%lu, ipg=%lu, mo=%04lx]", | |
1da177e4 LT |
1368 | sb->s_blocksize, |
1369 | sbi->s_groups_count, | |
1370 | EXT3_BLOCKS_PER_GROUP(sb), | |
1371 | EXT3_INODES_PER_GROUP(sb), | |
1372 | sbi->s_mount_opt); | |
1373 | ||
1da177e4 LT |
1374 | if (EXT3_SB(sb)->s_journal->j_inode == NULL) { |
1375 | char b[BDEVNAME_SIZE]; | |
4cf46b67 | 1376 | ext3_msg(sb, KERN_INFO, "using external journal on %s", |
1da177e4 LT |
1377 | bdevname(EXT3_SB(sb)->s_journal->j_dev, b)); |
1378 | } else { | |
4cf46b67 | 1379 | ext3_msg(sb, KERN_INFO, "using internal journal"); |
1da177e4 | 1380 | } |
d71bc6db | 1381 | cleancache_init_fs(sb); |
1da177e4 LT |
1382 | return res; |
1383 | } | |
1384 | ||
1385 | /* Called at mount-time, super-block is locked */ | |
197cd65a | 1386 | static int ext3_check_descriptors(struct super_block *sb) |
1da177e4 LT |
1387 | { |
1388 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
1da177e4 LT |
1389 | int i; |
1390 | ||
1391 | ext3_debug ("Checking group descriptors"); | |
1392 | ||
197cd65a AM |
1393 | for (i = 0; i < sbi->s_groups_count; i++) { |
1394 | struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL); | |
1eaafeae AM |
1395 | ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i); |
1396 | ext3_fsblk_t last_block; | |
197cd65a | 1397 | |
855565e8 ES |
1398 | if (i == sbi->s_groups_count - 1) |
1399 | last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1; | |
1400 | else | |
1401 | last_block = first_block + | |
1402 | (EXT3_BLOCKS_PER_GROUP(sb) - 1); | |
1403 | ||
855565e8 ES |
1404 | if (le32_to_cpu(gdp->bg_block_bitmap) < first_block || |
1405 | le32_to_cpu(gdp->bg_block_bitmap) > last_block) | |
1da177e4 LT |
1406 | { |
1407 | ext3_error (sb, "ext3_check_descriptors", | |
1408 | "Block bitmap for group %d" | |
1409 | " not in group (block %lu)!", | |
1410 | i, (unsigned long) | |
1411 | le32_to_cpu(gdp->bg_block_bitmap)); | |
1412 | return 0; | |
1413 | } | |
855565e8 ES |
1414 | if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block || |
1415 | le32_to_cpu(gdp->bg_inode_bitmap) > last_block) | |
1da177e4 LT |
1416 | { |
1417 | ext3_error (sb, "ext3_check_descriptors", | |
1418 | "Inode bitmap for group %d" | |
1419 | " not in group (block %lu)!", | |
1420 | i, (unsigned long) | |
1421 | le32_to_cpu(gdp->bg_inode_bitmap)); | |
1422 | return 0; | |
1423 | } | |
855565e8 | 1424 | if (le32_to_cpu(gdp->bg_inode_table) < first_block || |
780dcdb2 | 1425 | le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 > |
855565e8 | 1426 | last_block) |
1da177e4 LT |
1427 | { |
1428 | ext3_error (sb, "ext3_check_descriptors", | |
1429 | "Inode table for group %d" | |
1430 | " not in group (block %lu)!", | |
1431 | i, (unsigned long) | |
1432 | le32_to_cpu(gdp->bg_inode_table)); | |
1433 | return 0; | |
1434 | } | |
1da177e4 LT |
1435 | } |
1436 | ||
1437 | sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb)); | |
1438 | sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb)); | |
1439 | return 1; | |
1440 | } | |
1441 | ||
1442 | ||
1443 | /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at | |
1444 | * the superblock) which were deleted from all directories, but held open by | |
1445 | * a process at the time of a crash. We walk the list and try to delete these | |
1446 | * inodes at recovery time (only with a read-write filesystem). | |
1447 | * | |
1448 | * In order to keep the orphan inode chain consistent during traversal (in | |
1449 | * case of crash during recovery), we link each inode into the superblock | |
1450 | * orphan list_head and handle it the same way as an inode deletion during | |
1451 | * normal operation (which journals the operations for us). | |
1452 | * | |
1453 | * We only do an iget() and an iput() on each inode, which is very safe if we | |
1454 | * accidentally point at an in-use or already deleted inode. The worst that | |
1455 | * can happen in this case is that we get a "bit already cleared" message from | |
1456 | * ext3_free_inode(). The only reason we would point at a wrong inode is if | |
1457 | * e2fsck was run on this filesystem, and it must have already done the orphan | |
1458 | * inode cleanup for us, so we can safely abort without any further action. | |
1459 | */ | |
1460 | static void ext3_orphan_cleanup (struct super_block * sb, | |
1461 | struct ext3_super_block * es) | |
1462 | { | |
1463 | unsigned int s_flags = sb->s_flags; | |
1464 | int nr_orphans = 0, nr_truncates = 0; | |
1465 | #ifdef CONFIG_QUOTA | |
1466 | int i; | |
1467 | #endif | |
1468 | if (!es->s_last_orphan) { | |
1469 | jbd_debug(4, "no orphan inodes to clean up\n"); | |
1470 | return; | |
1471 | } | |
1472 | ||
a8f48a95 | 1473 | if (bdev_read_only(sb->s_bdev)) { |
4cf46b67 AF |
1474 | ext3_msg(sb, KERN_ERR, "error: write access " |
1475 | "unavailable, skipping orphan cleanup."); | |
a8f48a95 ES |
1476 | return; |
1477 | } | |
1478 | ||
ce654b37 AG |
1479 | /* Check if feature set allows readwrite operations */ |
1480 | if (EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) { | |
1481 | ext3_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " | |
1482 | "unknown ROCOMPAT features"); | |
1483 | return; | |
a8f48a95 ES |
1484 | } |
1485 | ||
1da177e4 | 1486 | if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) { |
e4230108 ES |
1487 | /* don't clear list on RO mount w/ errors */ |
1488 | if (es->s_last_orphan && !(s_flags & MS_RDONLY)) { | |
1da177e4 LT |
1489 | jbd_debug(1, "Errors on filesystem, " |
1490 | "clearing orphan list.\n"); | |
e4230108 ES |
1491 | es->s_last_orphan = 0; |
1492 | } | |
1da177e4 LT |
1493 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); |
1494 | return; | |
1495 | } | |
1496 | ||
1497 | if (s_flags & MS_RDONLY) { | |
4cf46b67 | 1498 | ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); |
1da177e4 LT |
1499 | sb->s_flags &= ~MS_RDONLY; |
1500 | } | |
1501 | #ifdef CONFIG_QUOTA | |
1502 | /* Needed for iput() to work correctly and not trash data */ | |
1503 | sb->s_flags |= MS_ACTIVE; | |
1504 | /* Turn on quotas so that they are updated correctly */ | |
1505 | for (i = 0; i < MAXQUOTAS; i++) { | |
1506 | if (EXT3_SB(sb)->s_qf_names[i]) { | |
1507 | int ret = ext3_quota_on_mount(sb, i); | |
1508 | if (ret < 0) | |
4cf46b67 AF |
1509 | ext3_msg(sb, KERN_ERR, |
1510 | "error: cannot turn on journaled " | |
1511 | "quota: %d", ret); | |
1da177e4 LT |
1512 | } |
1513 | } | |
1514 | #endif | |
1515 | ||
1516 | while (es->s_last_orphan) { | |
1517 | struct inode *inode; | |
1518 | ||
473043dc DH |
1519 | inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); |
1520 | if (IS_ERR(inode)) { | |
1da177e4 LT |
1521 | es->s_last_orphan = 0; |
1522 | break; | |
1523 | } | |
1524 | ||
1525 | list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan); | |
871a2931 | 1526 | dquot_initialize(inode); |
1da177e4 LT |
1527 | if (inode->i_nlink) { |
1528 | printk(KERN_DEBUG | |
eee194e7 | 1529 | "%s: truncating inode %lu to %Ld bytes\n", |
e05b6b52 | 1530 | __func__, inode->i_ino, inode->i_size); |
eee194e7 | 1531 | jbd_debug(2, "truncating inode %lu to %Ld bytes\n", |
1da177e4 LT |
1532 | inode->i_ino, inode->i_size); |
1533 | ext3_truncate(inode); | |
1534 | nr_truncates++; | |
1535 | } else { | |
1536 | printk(KERN_DEBUG | |
eee194e7 | 1537 | "%s: deleting unreferenced inode %lu\n", |
e05b6b52 | 1538 | __func__, inode->i_ino); |
eee194e7 | 1539 | jbd_debug(2, "deleting unreferenced inode %lu\n", |
1da177e4 LT |
1540 | inode->i_ino); |
1541 | nr_orphans++; | |
1542 | } | |
1543 | iput(inode); /* The delete magic happens here! */ | |
1544 | } | |
1545 | ||
1546 | #define PLURAL(x) (x), ((x)==1) ? "" : "s" | |
1547 | ||
1548 | if (nr_orphans) | |
4cf46b67 AF |
1549 | ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted", |
1550 | PLURAL(nr_orphans)); | |
1da177e4 | 1551 | if (nr_truncates) |
4cf46b67 AF |
1552 | ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up", |
1553 | PLURAL(nr_truncates)); | |
1da177e4 LT |
1554 | #ifdef CONFIG_QUOTA |
1555 | /* Turn quotas off */ | |
1556 | for (i = 0; i < MAXQUOTAS; i++) { | |
1557 | if (sb_dqopt(sb)->files[i]) | |
287a8095 | 1558 | dquot_quota_off(sb, i); |
1da177e4 LT |
1559 | } |
1560 | #endif | |
1561 | sb->s_flags = s_flags; /* Restore MS_RDONLY status */ | |
1562 | } | |
1563 | ||
1da177e4 LT |
1564 | /* |
1565 | * Maximal file size. There is a direct, and {,double-,triple-}indirect | |
1566 | * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks. | |
1567 | * We need to be 1 filesystem block less than the 2^32 sector limit. | |
1568 | */ | |
1569 | static loff_t ext3_max_size(int bits) | |
1570 | { | |
1571 | loff_t res = EXT3_NDIR_BLOCKS; | |
fe7fdc37 AK |
1572 | int meta_blocks; |
1573 | loff_t upper_limit; | |
1574 | ||
1575 | /* This is calculated to be the largest file size for a | |
1576 | * dense, file such that the total number of | |
1da177e4 | 1577 | * sectors in the file, including data and all indirect blocks, |
fe7fdc37 AK |
1578 | * does not exceed 2^32 -1 |
1579 | * __u32 i_blocks representing the total number of | |
1580 | * 512 bytes blocks of the file | |
1581 | */ | |
1582 | upper_limit = (1LL << 32) - 1; | |
1583 | ||
1584 | /* total blocks in file system block size */ | |
1585 | upper_limit >>= (bits - 9); | |
1586 | ||
1587 | ||
1588 | /* indirect blocks */ | |
1589 | meta_blocks = 1; | |
1590 | /* double indirect blocks */ | |
1591 | meta_blocks += 1 + (1LL << (bits-2)); | |
1592 | /* tripple indirect blocks */ | |
1593 | meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); | |
1594 | ||
1595 | upper_limit -= meta_blocks; | |
1596 | upper_limit <<= bits; | |
1da177e4 LT |
1597 | |
1598 | res += 1LL << (bits-2); | |
1599 | res += 1LL << (2*(bits-2)); | |
1600 | res += 1LL << (3*(bits-2)); | |
1601 | res <<= bits; | |
1602 | if (res > upper_limit) | |
1603 | res = upper_limit; | |
fe7fdc37 AK |
1604 | |
1605 | if (res > MAX_LFS_FILESIZE) | |
1606 | res = MAX_LFS_FILESIZE; | |
1607 | ||
1da177e4 LT |
1608 | return res; |
1609 | } | |
1610 | ||
43d23f90 MC |
1611 | static ext3_fsblk_t descriptor_loc(struct super_block *sb, |
1612 | ext3_fsblk_t logic_sb_block, | |
1da177e4 LT |
1613 | int nr) |
1614 | { | |
1615 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
43d23f90 | 1616 | unsigned long bg, first_meta_bg; |
1da177e4 LT |
1617 | int has_super = 0; |
1618 | ||
1da177e4 LT |
1619 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); |
1620 | ||
1621 | if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) || | |
1622 | nr < first_meta_bg) | |
1623 | return (logic_sb_block + nr + 1); | |
1624 | bg = sbi->s_desc_per_block * nr; | |
1625 | if (ext3_bg_has_super(sb, bg)) | |
1626 | has_super = 1; | |
43d23f90 | 1627 | return (has_super + ext3_group_first_block_no(sb, bg)); |
1da177e4 LT |
1628 | } |
1629 | ||
1630 | ||
1631 | static int ext3_fill_super (struct super_block *sb, void *data, int silent) | |
1632 | { | |
1633 | struct buffer_head * bh; | |
1634 | struct ext3_super_block *es = NULL; | |
1635 | struct ext3_sb_info *sbi; | |
43d23f90 | 1636 | ext3_fsblk_t block; |
4cf46b67 | 1637 | ext3_fsblk_t sb_block = get_sb_block(&data, sb); |
43d23f90 | 1638 | ext3_fsblk_t logic_sb_block; |
1da177e4 | 1639 | unsigned long offset = 0; |
eee194e7 | 1640 | unsigned int journal_inum = 0; |
71b96257 | 1641 | unsigned long journal_devnum = 0; |
1da177e4 LT |
1642 | unsigned long def_mount_opts; |
1643 | struct inode *root; | |
1644 | int blocksize; | |
1645 | int hblock; | |
1646 | int db_count; | |
1647 | int i; | |
1648 | int needs_recovery; | |
473043dc | 1649 | int ret = -EINVAL; |
1da177e4 | 1650 | __le32 features; |
833f4077 | 1651 | int err; |
1da177e4 | 1652 | |
f8314dc6 | 1653 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); |
1da177e4 LT |
1654 | if (!sbi) |
1655 | return -ENOMEM; | |
5df096d6 PE |
1656 | |
1657 | sbi->s_blockgroup_lock = | |
1658 | kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); | |
1659 | if (!sbi->s_blockgroup_lock) { | |
1660 | kfree(sbi); | |
1661 | return -ENOMEM; | |
1662 | } | |
1da177e4 | 1663 | sb->s_fs_info = sbi; |
571beed8 | 1664 | sbi->s_sb_block = sb_block; |
1da177e4 | 1665 | |
1da177e4 LT |
1666 | blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE); |
1667 | if (!blocksize) { | |
4cf46b67 | 1668 | ext3_msg(sb, KERN_ERR, "error: unable to set blocksize"); |
1da177e4 LT |
1669 | goto out_fail; |
1670 | } | |
1671 | ||
1672 | /* | |
1673 | * The ext3 superblock will not be buffer aligned for other than 1kB | |
1674 | * block sizes. We need to calculate the offset from buffer start. | |
1675 | */ | |
1676 | if (blocksize != EXT3_MIN_BLOCK_SIZE) { | |
1677 | logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize; | |
1678 | offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize; | |
1679 | } else { | |
1680 | logic_sb_block = sb_block; | |
1681 | } | |
1682 | ||
1683 | if (!(bh = sb_bread(sb, logic_sb_block))) { | |
4cf46b67 | 1684 | ext3_msg(sb, KERN_ERR, "error: unable to read superblock"); |
1da177e4 LT |
1685 | goto out_fail; |
1686 | } | |
1687 | /* | |
1688 | * Note: s_es must be initialized as soon as possible because | |
1689 | * some ext3 macro-instructions depend on its value | |
1690 | */ | |
57e94d86 | 1691 | es = (struct ext3_super_block *) (bh->b_data + offset); |
1da177e4 LT |
1692 | sbi->s_es = es; |
1693 | sb->s_magic = le16_to_cpu(es->s_magic); | |
1694 | if (sb->s_magic != EXT3_SUPER_MAGIC) | |
1695 | goto cantfind_ext3; | |
1696 | ||
1697 | /* Set defaults before we parse the mount options */ | |
1698 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
1699 | if (def_mount_opts & EXT3_DEFM_DEBUG) | |
1700 | set_opt(sbi->s_mount_opt, DEBUG); | |
1701 | if (def_mount_opts & EXT3_DEFM_BSDGROUPS) | |
1702 | set_opt(sbi->s_mount_opt, GRPID); | |
1703 | if (def_mount_opts & EXT3_DEFM_UID16) | |
1704 | set_opt(sbi->s_mount_opt, NO_UID32); | |
2e7842b8 | 1705 | #ifdef CONFIG_EXT3_FS_XATTR |
1da177e4 LT |
1706 | if (def_mount_opts & EXT3_DEFM_XATTR_USER) |
1707 | set_opt(sbi->s_mount_opt, XATTR_USER); | |
2e7842b8 HD |
1708 | #endif |
1709 | #ifdef CONFIG_EXT3_FS_POSIX_ACL | |
1da177e4 LT |
1710 | if (def_mount_opts & EXT3_DEFM_ACL) |
1711 | set_opt(sbi->s_mount_opt, POSIX_ACL); | |
2e7842b8 | 1712 | #endif |
1da177e4 | 1713 | if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA) |
e3c96435 | 1714 | set_opt(sbi->s_mount_opt, JOURNAL_DATA); |
1da177e4 | 1715 | else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED) |
e3c96435 | 1716 | set_opt(sbi->s_mount_opt, ORDERED_DATA); |
1da177e4 | 1717 | else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK) |
e3c96435 | 1718 | set_opt(sbi->s_mount_opt, WRITEBACK_DATA); |
1da177e4 LT |
1719 | |
1720 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC) | |
1721 | set_opt(sbi->s_mount_opt, ERRORS_PANIC); | |
1eca93f9 | 1722 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE) |
2245d7c2 | 1723 | set_opt(sbi->s_mount_opt, ERRORS_CONT); |
1eca93f9 AK |
1724 | else |
1725 | set_opt(sbi->s_mount_opt, ERRORS_RO); | |
1da177e4 | 1726 | |
1523299d EB |
1727 | sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); |
1728 | sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); | |
1da177e4 | 1729 | |
00eacd66 CH |
1730 | /* enable barriers by default */ |
1731 | set_opt(sbi->s_mount_opt, BARRIER); | |
1da177e4 LT |
1732 | set_opt(sbi->s_mount_opt, RESERVATION); |
1733 | ||
71b96257 JL |
1734 | if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum, |
1735 | NULL, 0)) | |
1da177e4 LT |
1736 | goto failed_mount; |
1737 | ||
1738 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
e3c96435 | 1739 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
1da177e4 LT |
1740 | |
1741 | if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV && | |
1742 | (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) || | |
1743 | EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) || | |
1744 | EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U))) | |
4cf46b67 AF |
1745 | ext3_msg(sb, KERN_WARNING, |
1746 | "warning: feature flags set on rev 0 fs, " | |
1747 | "running e2fsck is recommended"); | |
1da177e4 LT |
1748 | /* |
1749 | * Check feature flags regardless of the revision level, since we | |
1750 | * previously didn't change the revision level when setting the flags, | |
1751 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
1752 | */ | |
1753 | features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP); | |
1754 | if (features) { | |
4cf46b67 AF |
1755 | ext3_msg(sb, KERN_ERR, |
1756 | "error: couldn't mount because of unsupported " | |
1757 | "optional features (%x)", le32_to_cpu(features)); | |
1da177e4 LT |
1758 | goto failed_mount; |
1759 | } | |
1760 | features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP); | |
1761 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
4cf46b67 AF |
1762 | ext3_msg(sb, KERN_ERR, |
1763 | "error: couldn't mount RDWR because of unsupported " | |
1764 | "optional features (%x)", le32_to_cpu(features)); | |
1da177e4 LT |
1765 | goto failed_mount; |
1766 | } | |
1767 | blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); | |
1768 | ||
1769 | if (blocksize < EXT3_MIN_BLOCK_SIZE || | |
1770 | blocksize > EXT3_MAX_BLOCK_SIZE) { | |
4cf46b67 AF |
1771 | ext3_msg(sb, KERN_ERR, |
1772 | "error: couldn't mount because of unsupported " | |
1773 | "filesystem blocksize %d", blocksize); | |
1da177e4 LT |
1774 | goto failed_mount; |
1775 | } | |
1776 | ||
e1defc4f | 1777 | hblock = bdev_logical_block_size(sb->s_bdev); |
1da177e4 LT |
1778 | if (sb->s_blocksize != blocksize) { |
1779 | /* | |
1780 | * Make sure the blocksize for the filesystem is larger | |
1781 | * than the hardware sectorsize for the machine. | |
1782 | */ | |
1783 | if (blocksize < hblock) { | |
4cf46b67 AF |
1784 | ext3_msg(sb, KERN_ERR, |
1785 | "error: fsblocksize %d too small for " | |
1786 | "hardware sectorsize %d", blocksize, hblock); | |
1da177e4 LT |
1787 | goto failed_mount; |
1788 | } | |
1789 | ||
1790 | brelse (bh); | |
0f0a89eb | 1791 | if (!sb_set_blocksize(sb, blocksize)) { |
4cf46b67 AF |
1792 | ext3_msg(sb, KERN_ERR, |
1793 | "error: bad blocksize %d", blocksize); | |
0f0a89eb TS |
1794 | goto out_fail; |
1795 | } | |
1da177e4 LT |
1796 | logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize; |
1797 | offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize; | |
1798 | bh = sb_bread(sb, logic_sb_block); | |
1799 | if (!bh) { | |
4cf46b67 AF |
1800 | ext3_msg(sb, KERN_ERR, |
1801 | "error: can't read superblock on 2nd try"); | |
1da177e4 LT |
1802 | goto failed_mount; |
1803 | } | |
57e94d86 | 1804 | es = (struct ext3_super_block *)(bh->b_data + offset); |
1da177e4 LT |
1805 | sbi->s_es = es; |
1806 | if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) { | |
4cf46b67 AF |
1807 | ext3_msg(sb, KERN_ERR, |
1808 | "error: magic mismatch"); | |
1da177e4 LT |
1809 | goto failed_mount; |
1810 | } | |
1811 | } | |
1812 | ||
1813 | sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits); | |
1814 | ||
1815 | if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) { | |
1816 | sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE; | |
1817 | sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO; | |
1818 | } else { | |
1819 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
1820 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
1821 | if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) || | |
3fc74269 | 1822 | (!is_power_of_2(sbi->s_inode_size)) || |
1da177e4 | 1823 | (sbi->s_inode_size > blocksize)) { |
4cf46b67 AF |
1824 | ext3_msg(sb, KERN_ERR, |
1825 | "error: unsupported inode size: %d", | |
1da177e4 LT |
1826 | sbi->s_inode_size); |
1827 | goto failed_mount; | |
1828 | } | |
1829 | } | |
1830 | sbi->s_frag_size = EXT3_MIN_FRAG_SIZE << | |
1831 | le32_to_cpu(es->s_log_frag_size); | |
1832 | if (blocksize != sbi->s_frag_size) { | |
4cf46b67 AF |
1833 | ext3_msg(sb, KERN_ERR, |
1834 | "error: fragsize %lu != blocksize %u (unsupported)", | |
1da177e4 LT |
1835 | sbi->s_frag_size, blocksize); |
1836 | goto failed_mount; | |
1837 | } | |
1838 | sbi->s_frags_per_block = 1; | |
1839 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); | |
1840 | sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group); | |
1841 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); | |
b47b6f38 | 1842 | if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0) |
1da177e4 LT |
1843 | goto cantfind_ext3; |
1844 | sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb); | |
1845 | if (sbi->s_inodes_per_block == 0) | |
1846 | goto cantfind_ext3; | |
1847 | sbi->s_itb_per_group = sbi->s_inodes_per_group / | |
1848 | sbi->s_inodes_per_block; | |
1849 | sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc); | |
1850 | sbi->s_sbh = bh; | |
1851 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
f0d1b0b3 DH |
1852 | sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb)); |
1853 | sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb)); | |
1da177e4 LT |
1854 | for (i=0; i < 4; i++) |
1855 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); | |
1856 | sbi->s_def_hash_version = es->s_def_hash_version; | |
5e1f8c9e TT |
1857 | i = le32_to_cpu(es->s_flags); |
1858 | if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
1859 | sbi->s_hash_unsigned = 3; | |
1860 | else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
1861 | #ifdef __CHAR_UNSIGNED__ | |
1862 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
1863 | sbi->s_hash_unsigned = 3; | |
1864 | #else | |
1865 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
1866 | #endif | |
5e1f8c9e | 1867 | } |
1da177e4 LT |
1868 | |
1869 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
4cf46b67 AF |
1870 | ext3_msg(sb, KERN_ERR, |
1871 | "#blocks per group too big: %lu", | |
1da177e4 LT |
1872 | sbi->s_blocks_per_group); |
1873 | goto failed_mount; | |
1874 | } | |
1875 | if (sbi->s_frags_per_group > blocksize * 8) { | |
4cf46b67 AF |
1876 | ext3_msg(sb, KERN_ERR, |
1877 | "error: #fragments per group too big: %lu", | |
1da177e4 LT |
1878 | sbi->s_frags_per_group); |
1879 | goto failed_mount; | |
1880 | } | |
1881 | if (sbi->s_inodes_per_group > blocksize * 8) { | |
4cf46b67 AF |
1882 | ext3_msg(sb, KERN_ERR, |
1883 | "error: #inodes per group too big: %lu", | |
1da177e4 LT |
1884 | sbi->s_inodes_per_group); |
1885 | goto failed_mount; | |
1886 | } | |
1887 | ||
ad692bf3 DW |
1888 | err = generic_check_addressable(sb->s_blocksize_bits, |
1889 | le32_to_cpu(es->s_blocks_count)); | |
1890 | if (err) { | |
4cf46b67 AF |
1891 | ext3_msg(sb, KERN_ERR, |
1892 | "error: filesystem is too large to mount safely"); | |
fcd5df35 | 1893 | if (sizeof(sector_t) < 8) |
4cf46b67 AF |
1894 | ext3_msg(sb, KERN_ERR, |
1895 | "error: CONFIG_LBDAF not enabled"); | |
ad692bf3 | 1896 | ret = err; |
fcd5df35 MC |
1897 | goto failed_mount; |
1898 | } | |
1899 | ||
1da177e4 LT |
1900 | if (EXT3_BLOCKS_PER_GROUP(sb) == 0) |
1901 | goto cantfind_ext3; | |
855565e8 ES |
1902 | sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) - |
1903 | le32_to_cpu(es->s_first_data_block) - 1) | |
1904 | / EXT3_BLOCKS_PER_GROUP(sb)) + 1; | |
81a4e320 | 1905 | db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb)); |
1da177e4 LT |
1906 | sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *), |
1907 | GFP_KERNEL); | |
1908 | if (sbi->s_group_desc == NULL) { | |
4cf46b67 AF |
1909 | ext3_msg(sb, KERN_ERR, |
1910 | "error: not enough memory"); | |
4569cd1b | 1911 | ret = -ENOMEM; |
1da177e4 LT |
1912 | goto failed_mount; |
1913 | } | |
1914 | ||
5df096d6 | 1915 | bgl_lock_init(sbi->s_blockgroup_lock); |
1da177e4 LT |
1916 | |
1917 | for (i = 0; i < db_count; i++) { | |
1918 | block = descriptor_loc(sb, logic_sb_block, i); | |
1919 | sbi->s_group_desc[i] = sb_bread(sb, block); | |
1920 | if (!sbi->s_group_desc[i]) { | |
4cf46b67 AF |
1921 | ext3_msg(sb, KERN_ERR, |
1922 | "error: can't read group descriptor %d", i); | |
1da177e4 LT |
1923 | db_count = i; |
1924 | goto failed_mount2; | |
1925 | } | |
1926 | } | |
1927 | if (!ext3_check_descriptors (sb)) { | |
4cf46b67 AF |
1928 | ext3_msg(sb, KERN_ERR, |
1929 | "error: group descriptors corrupted"); | |
1da177e4 LT |
1930 | goto failed_mount2; |
1931 | } | |
1932 | sbi->s_gdb_count = db_count; | |
1933 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
1934 | spin_lock_init(&sbi->s_next_gen_lock); | |
0216bfcf | 1935 | |
1da177e4 LT |
1936 | /* per fileystem reservation list head & lock */ |
1937 | spin_lock_init(&sbi->s_rsv_window_lock); | |
1938 | sbi->s_rsv_window_root = RB_ROOT; | |
1939 | /* Add a single, static dummy reservation to the start of the | |
1940 | * reservation window list --- it gives us a placeholder for | |
1941 | * append-at-start-of-list which makes the allocation logic | |
1942 | * _much_ simpler. */ | |
1943 | sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED; | |
1944 | sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED; | |
1945 | sbi->s_rsv_window_head.rsv_alloc_hit = 0; | |
1946 | sbi->s_rsv_window_head.rsv_goal_size = 0; | |
1947 | ext3_rsv_window_add(sb, &sbi->s_rsv_window_head); | |
1948 | ||
1949 | /* | |
1950 | * set up enough so that it can read an inode | |
1951 | */ | |
1952 | sb->s_op = &ext3_sops; | |
1953 | sb->s_export_op = &ext3_export_ops; | |
1954 | sb->s_xattr = ext3_xattr_handlers; | |
1955 | #ifdef CONFIG_QUOTA | |
1956 | sb->s_qcop = &ext3_qctl_operations; | |
1957 | sb->dq_op = &ext3_quota_operations; | |
1958 | #endif | |
03cb5f03 | 1959 | memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); |
1da177e4 | 1960 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ |
b8a052d0 | 1961 | mutex_init(&sbi->s_orphan_lock); |
96d2a495 | 1962 | mutex_init(&sbi->s_resize_lock); |
1da177e4 LT |
1963 | |
1964 | sb->s_root = NULL; | |
1965 | ||
1966 | needs_recovery = (es->s_last_orphan != 0 || | |
1967 | EXT3_HAS_INCOMPAT_FEATURE(sb, | |
1968 | EXT3_FEATURE_INCOMPAT_RECOVER)); | |
1969 | ||
1970 | /* | |
1971 | * The first inode we look at is the journal inode. Don't try | |
1972 | * root first: it may be modified in the journal! | |
1973 | */ | |
1974 | if (!test_opt(sb, NOLOAD) && | |
1975 | EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) { | |
71b96257 | 1976 | if (ext3_load_journal(sb, es, journal_devnum)) |
41d1a636 | 1977 | goto failed_mount2; |
1da177e4 LT |
1978 | } else if (journal_inum) { |
1979 | if (ext3_create_journal(sb, es, journal_inum)) | |
41d1a636 | 1980 | goto failed_mount2; |
1da177e4 LT |
1981 | } else { |
1982 | if (!silent) | |
4cf46b67 AF |
1983 | ext3_msg(sb, KERN_ERR, |
1984 | "error: no journal found. " | |
1985 | "mounting ext3 over ext2?"); | |
41d1a636 DM |
1986 | goto failed_mount2; |
1987 | } | |
1988 | err = percpu_counter_init(&sbi->s_freeblocks_counter, | |
1989 | ext3_count_free_blocks(sb)); | |
1990 | if (!err) { | |
1991 | err = percpu_counter_init(&sbi->s_freeinodes_counter, | |
1992 | ext3_count_free_inodes(sb)); | |
1993 | } | |
1994 | if (!err) { | |
1995 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
1996 | ext3_count_dirs(sb)); | |
1997 | } | |
1998 | if (err) { | |
1999 | ext3_msg(sb, KERN_ERR, "error: insufficient memory"); | |
4569cd1b | 2000 | ret = err; |
0216bfcf | 2001 | goto failed_mount3; |
1da177e4 LT |
2002 | } |
2003 | ||
2004 | /* We have now updated the journal if required, so we can | |
2005 | * validate the data journaling mode. */ | |
2006 | switch (test_opt(sb, DATA_FLAGS)) { | |
2007 | case 0: | |
2008 | /* No mode set, assume a default based on the journal | |
2009 | capabilities: ORDERED_DATA if the journal can | |
2010 | cope, else JOURNAL_DATA */ | |
2011 | if (journal_check_available_features | |
2012 | (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) | |
bbae8bcc | 2013 | set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE); |
1da177e4 LT |
2014 | else |
2015 | set_opt(sbi->s_mount_opt, JOURNAL_DATA); | |
2016 | break; | |
2017 | ||
2018 | case EXT3_MOUNT_ORDERED_DATA: | |
2019 | case EXT3_MOUNT_WRITEBACK_DATA: | |
2020 | if (!journal_check_available_features | |
2021 | (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) { | |
4cf46b67 AF |
2022 | ext3_msg(sb, KERN_ERR, |
2023 | "error: journal does not support " | |
2024 | "requested data journaling mode"); | |
41d1a636 | 2025 | goto failed_mount3; |
1da177e4 LT |
2026 | } |
2027 | default: | |
2028 | break; | |
2029 | } | |
2030 | ||
1da177e4 LT |
2031 | /* |
2032 | * The journal_load will have done any necessary log recovery, | |
2033 | * so we can safely mount the rest of the filesystem now. | |
2034 | */ | |
2035 | ||
473043dc DH |
2036 | root = ext3_iget(sb, EXT3_ROOT_INO); |
2037 | if (IS_ERR(root)) { | |
4cf46b67 | 2038 | ext3_msg(sb, KERN_ERR, "error: get root inode failed"); |
473043dc | 2039 | ret = PTR_ERR(root); |
41d1a636 | 2040 | goto failed_mount3; |
1da177e4 LT |
2041 | } |
2042 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
473043dc | 2043 | iput(root); |
4cf46b67 | 2044 | ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck"); |
41d1a636 | 2045 | goto failed_mount3; |
1da177e4 | 2046 | } |
48fde701 | 2047 | sb->s_root = d_make_root(root); |
473043dc | 2048 | if (!sb->s_root) { |
4cf46b67 | 2049 | ext3_msg(sb, KERN_ERR, "error: get root dentry failed"); |
473043dc | 2050 | ret = -ENOMEM; |
41d1a636 | 2051 | goto failed_mount3; |
473043dc | 2052 | } |
1da177e4 | 2053 | |
f007dbf8 ES |
2054 | if (ext3_setup_super(sb, es, sb->s_flags & MS_RDONLY)) |
2055 | sb->s_flags |= MS_RDONLY; | |
ed505ee4 | 2056 | |
1da177e4 LT |
2057 | EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS; |
2058 | ext3_orphan_cleanup(sb, es); | |
2059 | EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS; | |
63894ab9 EG |
2060 | if (needs_recovery) { |
2061 | ext3_mark_recovery_complete(sb, es); | |
4cf46b67 | 2062 | ext3_msg(sb, KERN_INFO, "recovery complete"); |
63894ab9 | 2063 | } |
4cf46b67 | 2064 | ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode", |
1da177e4 LT |
2065 | test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal": |
2066 | test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered": | |
2067 | "writeback"); | |
2068 | ||
1da177e4 LT |
2069 | return 0; |
2070 | ||
2071 | cantfind_ext3: | |
2072 | if (!silent) | |
4cf46b67 AF |
2073 | ext3_msg(sb, KERN_INFO, |
2074 | "error: can't find ext3 filesystem on dev %s.", | |
1da177e4 LT |
2075 | sb->s_id); |
2076 | goto failed_mount; | |
2077 | ||
0216bfcf MC |
2078 | failed_mount3: |
2079 | percpu_counter_destroy(&sbi->s_freeblocks_counter); | |
2080 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
2081 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
41d1a636 | 2082 | journal_destroy(sbi->s_journal); |
1da177e4 LT |
2083 | failed_mount2: |
2084 | for (i = 0; i < db_count; i++) | |
2085 | brelse(sbi->s_group_desc[i]); | |
2086 | kfree(sbi->s_group_desc); | |
2087 | failed_mount: | |
2088 | #ifdef CONFIG_QUOTA | |
2089 | for (i = 0; i < MAXQUOTAS; i++) | |
2090 | kfree(sbi->s_qf_names[i]); | |
2091 | #endif | |
2092 | ext3_blkdev_remove(sbi); | |
2093 | brelse(bh); | |
2094 | out_fail: | |
2095 | sb->s_fs_info = NULL; | |
de5ce037 | 2096 | kfree(sbi->s_blockgroup_lock); |
1da177e4 | 2097 | kfree(sbi); |
473043dc | 2098 | return ret; |
1da177e4 LT |
2099 | } |
2100 | ||
2101 | /* | |
2102 | * Setup any per-fs journal parameters now. We'll do this both on | |
2103 | * initial mount, once the journal has been initialised but before we've | |
ae6ddcc5 | 2104 | * done any recovery; and again on any subsequent remount. |
1da177e4 LT |
2105 | */ |
2106 | static void ext3_init_journal_params(struct super_block *sb, journal_t *journal) | |
2107 | { | |
2108 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
2109 | ||
2110 | if (sbi->s_commit_interval) | |
2111 | journal->j_commit_interval = sbi->s_commit_interval; | |
2112 | /* We could also set up an ext3-specific default for the commit | |
2113 | * interval here, but for now we'll just fall back to the jbd | |
2114 | * default. */ | |
2115 | ||
2116 | spin_lock(&journal->j_state_lock); | |
2117 | if (test_opt(sb, BARRIER)) | |
2118 | journal->j_flags |= JFS_BARRIER; | |
2119 | else | |
2120 | journal->j_flags &= ~JFS_BARRIER; | |
0e4fb5e2 HK |
2121 | if (test_opt(sb, DATA_ERR_ABORT)) |
2122 | journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR; | |
2123 | else | |
2124 | journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR; | |
1da177e4 LT |
2125 | spin_unlock(&journal->j_state_lock); |
2126 | } | |
2127 | ||
eee194e7 ES |
2128 | static journal_t *ext3_get_journal(struct super_block *sb, |
2129 | unsigned int journal_inum) | |
1da177e4 LT |
2130 | { |
2131 | struct inode *journal_inode; | |
2132 | journal_t *journal; | |
2133 | ||
2134 | /* First, test for the existence of a valid inode on disk. Bad | |
2135 | * things happen if we iget() an unused inode, as the subsequent | |
2136 | * iput() will try to delete it. */ | |
2137 | ||
473043dc DH |
2138 | journal_inode = ext3_iget(sb, journal_inum); |
2139 | if (IS_ERR(journal_inode)) { | |
4cf46b67 | 2140 | ext3_msg(sb, KERN_ERR, "error: no journal found"); |
1da177e4 LT |
2141 | return NULL; |
2142 | } | |
2143 | if (!journal_inode->i_nlink) { | |
2144 | make_bad_inode(journal_inode); | |
2145 | iput(journal_inode); | |
4cf46b67 | 2146 | ext3_msg(sb, KERN_ERR, "error: journal inode is deleted"); |
1da177e4 LT |
2147 | return NULL; |
2148 | } | |
2149 | ||
2150 | jbd_debug(2, "Journal inode found at %p: %Ld bytes\n", | |
2151 | journal_inode, journal_inode->i_size); | |
473043dc | 2152 | if (!S_ISREG(journal_inode->i_mode)) { |
4cf46b67 | 2153 | ext3_msg(sb, KERN_ERR, "error: invalid journal inode"); |
1da177e4 LT |
2154 | iput(journal_inode); |
2155 | return NULL; | |
2156 | } | |
2157 | ||
2158 | journal = journal_init_inode(journal_inode); | |
2159 | if (!journal) { | |
4cf46b67 | 2160 | ext3_msg(sb, KERN_ERR, "error: could not load journal inode"); |
1da177e4 LT |
2161 | iput(journal_inode); |
2162 | return NULL; | |
2163 | } | |
2164 | journal->j_private = sb; | |
2165 | ext3_init_journal_params(sb, journal); | |
2166 | return journal; | |
2167 | } | |
2168 | ||
2169 | static journal_t *ext3_get_dev_journal(struct super_block *sb, | |
2170 | dev_t j_dev) | |
2171 | { | |
2172 | struct buffer_head * bh; | |
2173 | journal_t *journal; | |
43d23f90 | 2174 | ext3_fsblk_t start; |
1c2bf374 | 2175 | ext3_fsblk_t len; |
1da177e4 | 2176 | int hblock, blocksize; |
43d23f90 | 2177 | ext3_fsblk_t sb_block; |
1da177e4 LT |
2178 | unsigned long offset; |
2179 | struct ext3_super_block * es; | |
2180 | struct block_device *bdev; | |
2181 | ||
4cf46b67 | 2182 | bdev = ext3_blkdev_get(j_dev, sb); |
1da177e4 LT |
2183 | if (bdev == NULL) |
2184 | return NULL; | |
2185 | ||
1da177e4 | 2186 | blocksize = sb->s_blocksize; |
e1defc4f | 2187 | hblock = bdev_logical_block_size(bdev); |
1da177e4 | 2188 | if (blocksize < hblock) { |
4cf46b67 AF |
2189 | ext3_msg(sb, KERN_ERR, |
2190 | "error: blocksize too small for journal device"); | |
1da177e4 LT |
2191 | goto out_bdev; |
2192 | } | |
2193 | ||
2194 | sb_block = EXT3_MIN_BLOCK_SIZE / blocksize; | |
2195 | offset = EXT3_MIN_BLOCK_SIZE % blocksize; | |
2196 | set_blocksize(bdev, blocksize); | |
2197 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
4cf46b67 AF |
2198 | ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of " |
2199 | "external journal"); | |
1da177e4 LT |
2200 | goto out_bdev; |
2201 | } | |
2202 | ||
57e94d86 | 2203 | es = (struct ext3_super_block *) (bh->b_data + offset); |
1da177e4 LT |
2204 | if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) || |
2205 | !(le32_to_cpu(es->s_feature_incompat) & | |
2206 | EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) { | |
4cf46b67 AF |
2207 | ext3_msg(sb, KERN_ERR, "error: external journal has " |
2208 | "bad superblock"); | |
1da177e4 LT |
2209 | brelse(bh); |
2210 | goto out_bdev; | |
2211 | } | |
2212 | ||
2213 | if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { | |
4cf46b67 | 2214 | ext3_msg(sb, KERN_ERR, "error: journal UUID does not match"); |
1da177e4 LT |
2215 | brelse(bh); |
2216 | goto out_bdev; | |
2217 | } | |
2218 | ||
2219 | len = le32_to_cpu(es->s_blocks_count); | |
2220 | start = sb_block + 1; | |
2221 | brelse(bh); /* we're done with the superblock */ | |
2222 | ||
2223 | journal = journal_init_dev(bdev, sb->s_bdev, | |
2224 | start, len, blocksize); | |
2225 | if (!journal) { | |
4cf46b67 AF |
2226 | ext3_msg(sb, KERN_ERR, |
2227 | "error: failed to create device journal"); | |
1da177e4 LT |
2228 | goto out_bdev; |
2229 | } | |
2230 | journal->j_private = sb; | |
d03e1292 ZL |
2231 | if (!bh_uptodate_or_lock(journal->j_sb_buffer)) { |
2232 | if (bh_submit_read(journal->j_sb_buffer)) { | |
2233 | ext3_msg(sb, KERN_ERR, "I/O error on journal device"); | |
2234 | goto out_journal; | |
2235 | } | |
1da177e4 LT |
2236 | } |
2237 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
4cf46b67 AF |
2238 | ext3_msg(sb, KERN_ERR, |
2239 | "error: external journal has more than one " | |
2240 | "user (unsupported) - %d", | |
1da177e4 LT |
2241 | be32_to_cpu(journal->j_superblock->s_nr_users)); |
2242 | goto out_journal; | |
2243 | } | |
2244 | EXT3_SB(sb)->journal_bdev = bdev; | |
2245 | ext3_init_journal_params(sb, journal); | |
2246 | return journal; | |
2247 | out_journal: | |
2248 | journal_destroy(journal); | |
2249 | out_bdev: | |
2250 | ext3_blkdev_put(bdev); | |
2251 | return NULL; | |
2252 | } | |
2253 | ||
71b96257 JL |
2254 | static int ext3_load_journal(struct super_block *sb, |
2255 | struct ext3_super_block *es, | |
2256 | unsigned long journal_devnum) | |
1da177e4 LT |
2257 | { |
2258 | journal_t *journal; | |
eee194e7 | 2259 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); |
71b96257 | 2260 | dev_t journal_dev; |
1da177e4 LT |
2261 | int err = 0; |
2262 | int really_read_only; | |
2263 | ||
71b96257 JL |
2264 | if (journal_devnum && |
2265 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
4cf46b67 AF |
2266 | ext3_msg(sb, KERN_INFO, "external journal device major/minor " |
2267 | "numbers have changed"); | |
71b96257 JL |
2268 | journal_dev = new_decode_dev(journal_devnum); |
2269 | } else | |
2270 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
2271 | ||
1da177e4 LT |
2272 | really_read_only = bdev_read_only(sb->s_bdev); |
2273 | ||
2274 | /* | |
2275 | * Are we loading a blank journal or performing recovery after a | |
2276 | * crash? For recovery, we need to check in advance whether we | |
2277 | * can get read-write access to the device. | |
2278 | */ | |
2279 | ||
2280 | if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) { | |
2281 | if (sb->s_flags & MS_RDONLY) { | |
4cf46b67 AF |
2282 | ext3_msg(sb, KERN_INFO, |
2283 | "recovery required on readonly filesystem"); | |
1da177e4 | 2284 | if (really_read_only) { |
4cf46b67 AF |
2285 | ext3_msg(sb, KERN_ERR, "error: write access " |
2286 | "unavailable, cannot proceed"); | |
1da177e4 LT |
2287 | return -EROFS; |
2288 | } | |
4cf46b67 AF |
2289 | ext3_msg(sb, KERN_INFO, |
2290 | "write access will be enabled during recovery"); | |
1da177e4 LT |
2291 | } |
2292 | } | |
2293 | ||
2294 | if (journal_inum && journal_dev) { | |
4cf46b67 AF |
2295 | ext3_msg(sb, KERN_ERR, "error: filesystem has both journal " |
2296 | "and inode journals"); | |
1da177e4 LT |
2297 | return -EINVAL; |
2298 | } | |
2299 | ||
2300 | if (journal_inum) { | |
2301 | if (!(journal = ext3_get_journal(sb, journal_inum))) | |
2302 | return -EINVAL; | |
2303 | } else { | |
2304 | if (!(journal = ext3_get_dev_journal(sb, journal_dev))) | |
2305 | return -EINVAL; | |
2306 | } | |
2307 | ||
0636c73e ES |
2308 | if (!(journal->j_flags & JFS_BARRIER)) |
2309 | printk(KERN_INFO "EXT3-fs: barriers not enabled\n"); | |
2310 | ||
1da177e4 LT |
2311 | if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { |
2312 | err = journal_update_format(journal); | |
2313 | if (err) { | |
4cf46b67 | 2314 | ext3_msg(sb, KERN_ERR, "error updating journal"); |
1da177e4 LT |
2315 | journal_destroy(journal); |
2316 | return err; | |
2317 | } | |
2318 | } | |
2319 | ||
2320 | if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) | |
2321 | err = journal_wipe(journal, !really_read_only); | |
2322 | if (!err) | |
2323 | err = journal_load(journal); | |
2324 | ||
2325 | if (err) { | |
4cf46b67 | 2326 | ext3_msg(sb, KERN_ERR, "error loading journal"); |
1da177e4 LT |
2327 | journal_destroy(journal); |
2328 | return err; | |
2329 | } | |
2330 | ||
2331 | EXT3_SB(sb)->s_journal = journal; | |
2332 | ext3_clear_journal_err(sb, es); | |
71b96257 | 2333 | |
31d710a7 | 2334 | if (!really_read_only && journal_devnum && |
71b96257 JL |
2335 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { |
2336 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
71b96257 JL |
2337 | |
2338 | /* Make sure we flush the recovery flag to disk. */ | |
2339 | ext3_commit_super(sb, es, 1); | |
2340 | } | |
2341 | ||
1da177e4 LT |
2342 | return 0; |
2343 | } | |
2344 | ||
4cf46b67 AF |
2345 | static int ext3_create_journal(struct super_block *sb, |
2346 | struct ext3_super_block *es, | |
eee194e7 | 2347 | unsigned int journal_inum) |
1da177e4 LT |
2348 | { |
2349 | journal_t *journal; | |
952d9de1 | 2350 | int err; |
1da177e4 LT |
2351 | |
2352 | if (sb->s_flags & MS_RDONLY) { | |
4cf46b67 AF |
2353 | ext3_msg(sb, KERN_ERR, |
2354 | "error: readonly filesystem when trying to " | |
2355 | "create journal"); | |
1da177e4 LT |
2356 | return -EROFS; |
2357 | } | |
2358 | ||
952d9de1 BP |
2359 | journal = ext3_get_journal(sb, journal_inum); |
2360 | if (!journal) | |
1da177e4 LT |
2361 | return -EINVAL; |
2362 | ||
4cf46b67 | 2363 | ext3_msg(sb, KERN_INFO, "creating new journal on inode %u", |
1da177e4 LT |
2364 | journal_inum); |
2365 | ||
952d9de1 BP |
2366 | err = journal_create(journal); |
2367 | if (err) { | |
4cf46b67 | 2368 | ext3_msg(sb, KERN_ERR, "error creating journal"); |
1da177e4 LT |
2369 | journal_destroy(journal); |
2370 | return -EIO; | |
2371 | } | |
2372 | ||
2373 | EXT3_SB(sb)->s_journal = journal; | |
2374 | ||
2375 | ext3_update_dynamic_rev(sb); | |
2376 | EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
2377 | EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL); | |
2378 | ||
2379 | es->s_journal_inum = cpu_to_le32(journal_inum); | |
1da177e4 LT |
2380 | |
2381 | /* Make sure we flush the recovery flag to disk. */ | |
2382 | ext3_commit_super(sb, es, 1); | |
2383 | ||
2384 | return 0; | |
2385 | } | |
2386 | ||
c4be0c1d TS |
2387 | static int ext3_commit_super(struct super_block *sb, |
2388 | struct ext3_super_block *es, | |
1da177e4 LT |
2389 | int sync) |
2390 | { | |
2391 | struct buffer_head *sbh = EXT3_SB(sb)->s_sbh; | |
c4be0c1d | 2392 | int error = 0; |
1da177e4 LT |
2393 | |
2394 | if (!sbh) | |
c4be0c1d | 2395 | return error; |
dff6825e DW |
2396 | |
2397 | if (buffer_write_io_error(sbh)) { | |
2398 | /* | |
2399 | * Oh, dear. A previous attempt to write the | |
2400 | * superblock failed. This could happen because the | |
2401 | * USB device was yanked out. Or it could happen to | |
2402 | * be a transient write error and maybe the block will | |
2403 | * be remapped. Nothing we can do but to retry the | |
2404 | * write and hope for the best. | |
2405 | */ | |
2406 | ext3_msg(sb, KERN_ERR, "previous I/O error to " | |
2407 | "superblock detected"); | |
2408 | clear_buffer_write_io_error(sbh); | |
2409 | set_buffer_uptodate(sbh); | |
2410 | } | |
96ec2e0a TT |
2411 | /* |
2412 | * If the file system is mounted read-only, don't update the | |
2413 | * superblock write time. This avoids updating the superblock | |
2414 | * write time when we are mounting the root file system | |
2415 | * read/only but we need to replay the journal; at that point, | |
2416 | * for people who are east of GMT and who make their clock | |
2417 | * tick in localtime for Windows bug-for-bug compatibility, | |
2418 | * the clock is set in the future, and this will cause e2fsck | |
2419 | * to complain and force a full file system check. | |
2420 | */ | |
2421 | if (!(sb->s_flags & MS_RDONLY)) | |
2422 | es->s_wtime = cpu_to_le32(get_seconds()); | |
1da177e4 LT |
2423 | es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb)); |
2424 | es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb)); | |
2425 | BUFFER_TRACE(sbh, "marking dirty"); | |
2426 | mark_buffer_dirty(sbh); | |
dff6825e | 2427 | if (sync) { |
c4be0c1d | 2428 | error = sync_dirty_buffer(sbh); |
dff6825e DW |
2429 | if (buffer_write_io_error(sbh)) { |
2430 | ext3_msg(sb, KERN_ERR, "I/O error while writing " | |
2431 | "superblock"); | |
2432 | clear_buffer_write_io_error(sbh); | |
2433 | set_buffer_uptodate(sbh); | |
2434 | } | |
2435 | } | |
c4be0c1d | 2436 | return error; |
1da177e4 LT |
2437 | } |
2438 | ||
2439 | ||
2440 | /* | |
2441 | * Have we just finished recovery? If so, and if we are mounting (or | |
2442 | * remounting) the filesystem readonly, then we will end up with a | |
2443 | * consistent fs on disk. Record that fact. | |
2444 | */ | |
2445 | static void ext3_mark_recovery_complete(struct super_block * sb, | |
2446 | struct ext3_super_block * es) | |
2447 | { | |
2448 | journal_t *journal = EXT3_SB(sb)->s_journal; | |
2449 | ||
2450 | journal_lock_updates(journal); | |
2d7c820e HK |
2451 | if (journal_flush(journal) < 0) |
2452 | goto out; | |
2453 | ||
1da177e4 LT |
2454 | if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) && |
2455 | sb->s_flags & MS_RDONLY) { | |
2456 | EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
1da177e4 LT |
2457 | ext3_commit_super(sb, es, 1); |
2458 | } | |
2d7c820e HK |
2459 | |
2460 | out: | |
1da177e4 LT |
2461 | journal_unlock_updates(journal); |
2462 | } | |
2463 | ||
2464 | /* | |
2465 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
2466 | * has recorded an error from a previous lifetime, move that error to the | |
2467 | * main filesystem now. | |
2468 | */ | |
4cf46b67 AF |
2469 | static void ext3_clear_journal_err(struct super_block *sb, |
2470 | struct ext3_super_block *es) | |
1da177e4 LT |
2471 | { |
2472 | journal_t *journal; | |
2473 | int j_errno; | |
2474 | const char *errstr; | |
2475 | ||
2476 | journal = EXT3_SB(sb)->s_journal; | |
2477 | ||
2478 | /* | |
2479 | * Now check for any error status which may have been recorded in the | |
2480 | * journal by a prior ext3_error() or ext3_abort() | |
2481 | */ | |
2482 | ||
2483 | j_errno = journal_errno(journal); | |
2484 | if (j_errno) { | |
2485 | char nbuf[16]; | |
2486 | ||
2487 | errstr = ext3_decode_error(sb, j_errno, nbuf); | |
e05b6b52 | 2488 | ext3_warning(sb, __func__, "Filesystem error recorded " |
1da177e4 | 2489 | "from previous mount: %s", errstr); |
e05b6b52 | 2490 | ext3_warning(sb, __func__, "Marking fs in need of " |
1da177e4 LT |
2491 | "filesystem check."); |
2492 | ||
2493 | EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS; | |
2494 | es->s_state |= cpu_to_le16(EXT3_ERROR_FS); | |
2495 | ext3_commit_super (sb, es, 1); | |
2496 | ||
2497 | journal_clear_err(journal); | |
2498 | } | |
2499 | } | |
2500 | ||
2501 | /* | |
2502 | * Force the running and committing transactions to commit, | |
2503 | * and wait on the commit. | |
2504 | */ | |
2505 | int ext3_force_commit(struct super_block *sb) | |
2506 | { | |
2507 | journal_t *journal; | |
2508 | int ret; | |
2509 | ||
2510 | if (sb->s_flags & MS_RDONLY) | |
2511 | return 0; | |
2512 | ||
2513 | journal = EXT3_SB(sb)->s_journal; | |
1da177e4 LT |
2514 | ret = ext3_journal_force_commit(journal); |
2515 | return ret; | |
2516 | } | |
2517 | ||
1da177e4 LT |
2518 | static int ext3_sync_fs(struct super_block *sb, int wait) |
2519 | { | |
02ac597c | 2520 | tid_t target; |
c87591b7 | 2521 | |
785c4bcc | 2522 | trace_ext3_sync_fs(sb, wait); |
a1177825 JK |
2523 | /* |
2524 | * Writeback quota in non-journalled quota case - journalled quota has | |
2525 | * no dirty dquots | |
2526 | */ | |
2527 | dquot_writeback_dquots(sb, -1); | |
02ac597c JK |
2528 | if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) { |
2529 | if (wait) | |
2530 | log_wait_commit(EXT3_SB(sb)->s_journal, target); | |
2531 | } | |
1da177e4 LT |
2532 | return 0; |
2533 | } | |
2534 | ||
2535 | /* | |
2536 | * LVM calls this function before a (read-only) snapshot is created. This | |
2537 | * gives us a chance to flush the journal completely and mark the fs clean. | |
2538 | */ | |
c4be0c1d | 2539 | static int ext3_freeze(struct super_block *sb) |
1da177e4 | 2540 | { |
c4be0c1d TS |
2541 | int error = 0; |
2542 | journal_t *journal; | |
1da177e4 LT |
2543 | |
2544 | if (!(sb->s_flags & MS_RDONLY)) { | |
c4be0c1d | 2545 | journal = EXT3_SB(sb)->s_journal; |
1da177e4 LT |
2546 | |
2547 | /* Now we set up the journal barrier. */ | |
2548 | journal_lock_updates(journal); | |
2d7c820e HK |
2549 | |
2550 | /* | |
2551 | * We don't want to clear needs_recovery flag when we failed | |
2552 | * to flush the journal. | |
2553 | */ | |
c4be0c1d TS |
2554 | error = journal_flush(journal); |
2555 | if (error < 0) | |
2556 | goto out; | |
1da177e4 LT |
2557 | |
2558 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
2559 | EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
c4be0c1d TS |
2560 | error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1); |
2561 | if (error) | |
2562 | goto out; | |
1da177e4 | 2563 | } |
c4be0c1d TS |
2564 | return 0; |
2565 | ||
2566 | out: | |
2567 | journal_unlock_updates(journal); | |
2568 | return error; | |
1da177e4 LT |
2569 | } |
2570 | ||
2571 | /* | |
2572 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
2573 | * flag here, even though the filesystem is not technically dirty yet. | |
2574 | */ | |
c4be0c1d | 2575 | static int ext3_unfreeze(struct super_block *sb) |
1da177e4 LT |
2576 | { |
2577 | if (!(sb->s_flags & MS_RDONLY)) { | |
1da177e4 LT |
2578 | /* Reser the needs_recovery flag before the fs is unlocked. */ |
2579 | EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
2580 | ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1); | |
1da177e4 LT |
2581 | journal_unlock_updates(EXT3_SB(sb)->s_journal); |
2582 | } | |
c4be0c1d | 2583 | return 0; |
1da177e4 LT |
2584 | } |
2585 | ||
2586 | static int ext3_remount (struct super_block * sb, int * flags, char * data) | |
2587 | { | |
2588 | struct ext3_super_block * es; | |
2589 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
43d23f90 | 2590 | ext3_fsblk_t n_blocks_count = 0; |
08c6a96f JK |
2591 | unsigned long old_sb_flags; |
2592 | struct ext3_mount_options old_opts; | |
c79d967d | 2593 | int enable_quota = 0; |
08c6a96f JK |
2594 | int err; |
2595 | #ifdef CONFIG_QUOTA | |
2596 | int i; | |
2597 | #endif | |
2598 | ||
2599 | /* Store the original options */ | |
2600 | old_sb_flags = sb->s_flags; | |
2601 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
2602 | old_opts.s_resuid = sbi->s_resuid; | |
2603 | old_opts.s_resgid = sbi->s_resgid; | |
2604 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
2605 | #ifdef CONFIG_QUOTA | |
2606 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
2607 | for (i = 0; i < MAXQUOTAS; i++) | |
2608 | old_opts.s_qf_names[i] = sbi->s_qf_names[i]; | |
2609 | #endif | |
1da177e4 LT |
2610 | |
2611 | /* | |
2612 | * Allow the "check" option to be passed as a remount option. | |
2613 | */ | |
71b96257 | 2614 | if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) { |
08c6a96f JK |
2615 | err = -EINVAL; |
2616 | goto restore_opts; | |
2617 | } | |
1da177e4 | 2618 | |
e3c96435 | 2619 | if (test_opt(sb, ABORT)) |
e05b6b52 | 2620 | ext3_abort(sb, __func__, "Abort forced by user"); |
1da177e4 LT |
2621 | |
2622 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
e3c96435 | 2623 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
1da177e4 LT |
2624 | |
2625 | es = sbi->s_es; | |
2626 | ||
2627 | ext3_init_journal_params(sb, sbi->s_journal); | |
2628 | ||
2629 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) || | |
2630 | n_blocks_count > le32_to_cpu(es->s_blocks_count)) { | |
e3c96435 | 2631 | if (test_opt(sb, ABORT)) { |
08c6a96f JK |
2632 | err = -EROFS; |
2633 | goto restore_opts; | |
2634 | } | |
1da177e4 LT |
2635 | |
2636 | if (*flags & MS_RDONLY) { | |
0f0dd62f CH |
2637 | err = dquot_suspend(sb, -1); |
2638 | if (err < 0) | |
c79d967d | 2639 | goto restore_opts; |
c79d967d | 2640 | |
1da177e4 LT |
2641 | /* |
2642 | * First of all, the unconditional stuff we have to do | |
2643 | * to disable replay of the journal when we next remount | |
2644 | */ | |
2645 | sb->s_flags |= MS_RDONLY; | |
2646 | ||
2647 | /* | |
2648 | * OK, test if we are remounting a valid rw partition | |
2649 | * readonly, and if so set the rdonly flag and then | |
2650 | * mark the partition as valid again. | |
2651 | */ | |
2652 | if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) && | |
2653 | (sbi->s_mount_state & EXT3_VALID_FS)) | |
2654 | es->s_state = cpu_to_le16(sbi->s_mount_state); | |
2655 | ||
2656 | ext3_mark_recovery_complete(sb, es); | |
2657 | } else { | |
2658 | __le32 ret; | |
2659 | if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb, | |
2660 | ~EXT3_FEATURE_RO_COMPAT_SUPP))) { | |
4cf46b67 AF |
2661 | ext3_msg(sb, KERN_WARNING, |
2662 | "warning: couldn't remount RDWR " | |
2663 | "because of unsupported optional " | |
2664 | "features (%x)", le32_to_cpu(ret)); | |
08c6a96f JK |
2665 | err = -EROFS; |
2666 | goto restore_opts; | |
1da177e4 | 2667 | } |
ea9a05a1 ES |
2668 | |
2669 | /* | |
2670 | * If we have an unprocessed orphan list hanging | |
2671 | * around from a previously readonly bdev mount, | |
1cde201d | 2672 | * require a full umount & mount for now. |
ea9a05a1 ES |
2673 | */ |
2674 | if (es->s_last_orphan) { | |
4cf46b67 | 2675 | ext3_msg(sb, KERN_WARNING, "warning: couldn't " |
ea9a05a1 ES |
2676 | "remount RDWR because of unprocessed " |
2677 | "orphan inode list. Please " | |
1cde201d | 2678 | "umount & mount instead."); |
ea9a05a1 ES |
2679 | err = -EINVAL; |
2680 | goto restore_opts; | |
2681 | } | |
2682 | ||
1da177e4 LT |
2683 | /* |
2684 | * Mounting a RDONLY partition read-write, so reread | |
2685 | * and store the current valid flag. (It may have | |
2686 | * been changed by e2fsck since we originally mounted | |
2687 | * the partition.) | |
2688 | */ | |
2689 | ext3_clear_journal_err(sb, es); | |
2690 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
a4e4de36 | 2691 | if ((err = ext3_group_extend(sb, es, n_blocks_count))) |
08c6a96f | 2692 | goto restore_opts; |
1da177e4 LT |
2693 | if (!ext3_setup_super (sb, es, 0)) |
2694 | sb->s_flags &= ~MS_RDONLY; | |
c79d967d | 2695 | enable_quota = 1; |
1da177e4 LT |
2696 | } |
2697 | } | |
08c6a96f JK |
2698 | #ifdef CONFIG_QUOTA |
2699 | /* Release old quota file names */ | |
2700 | for (i = 0; i < MAXQUOTAS; i++) | |
2701 | if (old_opts.s_qf_names[i] && | |
2702 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
2703 | kfree(old_opts.s_qf_names[i]); | |
2704 | #endif | |
c79d967d | 2705 | if (enable_quota) |
0f0dd62f | 2706 | dquot_resume(sb, -1); |
1da177e4 | 2707 | return 0; |
08c6a96f JK |
2708 | restore_opts: |
2709 | sb->s_flags = old_sb_flags; | |
2710 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
2711 | sbi->s_resuid = old_opts.s_resuid; | |
2712 | sbi->s_resgid = old_opts.s_resgid; | |
2713 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
2714 | #ifdef CONFIG_QUOTA | |
2715 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
2716 | for (i = 0; i < MAXQUOTAS; i++) { | |
2717 | if (sbi->s_qf_names[i] && | |
2718 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
2719 | kfree(sbi->s_qf_names[i]); | |
2720 | sbi->s_qf_names[i] = old_opts.s_qf_names[i]; | |
2721 | } | |
2722 | #endif | |
2723 | return err; | |
1da177e4 LT |
2724 | } |
2725 | ||
726c3342 | 2726 | static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf) |
1da177e4 | 2727 | { |
726c3342 | 2728 | struct super_block *sb = dentry->d_sb; |
09fe316a AT |
2729 | struct ext3_sb_info *sbi = EXT3_SB(sb); |
2730 | struct ext3_super_block *es = sbi->s_es; | |
50ee0a32 | 2731 | u64 fsid; |
1da177e4 | 2732 | |
a71ce8c6 BP |
2733 | if (test_opt(sb, MINIX_DF)) { |
2734 | sbi->s_overhead_last = 0; | |
2735 | } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) { | |
2736 | unsigned long ngroups = sbi->s_groups_count, i; | |
2737 | ext3_fsblk_t overhead = 0; | |
1da177e4 LT |
2738 | smp_rmb(); |
2739 | ||
2740 | /* | |
a71ce8c6 BP |
2741 | * Compute the overhead (FS structures). This is constant |
2742 | * for a given filesystem unless the number of block groups | |
2743 | * changes so we cache the previous value until it does. | |
1da177e4 LT |
2744 | */ |
2745 | ||
2746 | /* | |
2747 | * All of the blocks before first_data_block are | |
2748 | * overhead | |
2749 | */ | |
2750 | overhead = le32_to_cpu(es->s_first_data_block); | |
2751 | ||
2752 | /* | |
2753 | * Add the overhead attributed to the superblock and | |
2754 | * block group descriptors. If the sparse superblocks | |
2755 | * feature is turned on, then not all groups have this. | |
2756 | */ | |
2757 | for (i = 0; i < ngroups; i++) { | |
2758 | overhead += ext3_bg_has_super(sb, i) + | |
2759 | ext3_bg_num_gdb(sb, i); | |
2760 | cond_resched(); | |
2761 | } | |
2762 | ||
2763 | /* | |
2764 | * Every block group has an inode bitmap, a block | |
2765 | * bitmap, and an inode table. | |
2766 | */ | |
a71ce8c6 BP |
2767 | overhead += ngroups * (2 + sbi->s_itb_per_group); |
2768 | sbi->s_overhead_last = overhead; | |
2769 | smp_wmb(); | |
2770 | sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count); | |
1da177e4 LT |
2771 | } |
2772 | ||
2773 | buf->f_type = EXT3_SUPER_MAGIC; | |
2774 | buf->f_bsize = sb->s_blocksize; | |
a71ce8c6 | 2775 | buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last; |
52d9f3b4 | 2776 | buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter); |
1da177e4 LT |
2777 | buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count); |
2778 | if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count)) | |
2779 | buf->f_bavail = 0; | |
2780 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
52d9f3b4 | 2781 | buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); |
1da177e4 | 2782 | buf->f_namelen = EXT3_NAME_LEN; |
50ee0a32 PE |
2783 | fsid = le64_to_cpup((void *)es->s_uuid) ^ |
2784 | le64_to_cpup((void *)es->s_uuid + sizeof(u64)); | |
2785 | buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; | |
2786 | buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; | |
1da177e4 LT |
2787 | return 0; |
2788 | } | |
2789 | ||
2790 | /* Helper function for writing quotas on sync - we need to start transaction before quota file | |
2791 | * is locked for write. Otherwise the are possible deadlocks: | |
2792 | * Process 1 Process 2 | |
2793 | * ext3_create() quota_sync() | |
2794 | * journal_start() write_dquot() | |
871a2931 | 2795 | * dquot_initialize() down(dqio_mutex) |
d3be915f | 2796 | * down(dqio_mutex) journal_start() |
1da177e4 LT |
2797 | * |
2798 | */ | |
2799 | ||
2800 | #ifdef CONFIG_QUOTA | |
2801 | ||
2802 | static inline struct inode *dquot_to_inode(struct dquot *dquot) | |
2803 | { | |
4c376dca | 2804 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; |
1da177e4 LT |
2805 | } |
2806 | ||
1da177e4 LT |
2807 | static int ext3_write_dquot(struct dquot *dquot) |
2808 | { | |
2809 | int ret, err; | |
2810 | handle_t *handle; | |
2811 | struct inode *inode; | |
2812 | ||
2813 | inode = dquot_to_inode(dquot); | |
2814 | handle = ext3_journal_start(inode, | |
1f54587b | 2815 | EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); |
1da177e4 LT |
2816 | if (IS_ERR(handle)) |
2817 | return PTR_ERR(handle); | |
2818 | ret = dquot_commit(dquot); | |
2819 | err = ext3_journal_stop(handle); | |
2820 | if (!ret) | |
2821 | ret = err; | |
2822 | return ret; | |
2823 | } | |
2824 | ||
2825 | static int ext3_acquire_dquot(struct dquot *dquot) | |
2826 | { | |
2827 | int ret, err; | |
2828 | handle_t *handle; | |
2829 | ||
2830 | handle = ext3_journal_start(dquot_to_inode(dquot), | |
1f54587b | 2831 | EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb)); |
1da177e4 LT |
2832 | if (IS_ERR(handle)) |
2833 | return PTR_ERR(handle); | |
2834 | ret = dquot_acquire(dquot); | |
2835 | err = ext3_journal_stop(handle); | |
2836 | if (!ret) | |
2837 | ret = err; | |
2838 | return ret; | |
2839 | } | |
2840 | ||
2841 | static int ext3_release_dquot(struct dquot *dquot) | |
2842 | { | |
2843 | int ret, err; | |
2844 | handle_t *handle; | |
2845 | ||
2846 | handle = ext3_journal_start(dquot_to_inode(dquot), | |
1f54587b | 2847 | EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb)); |
9c3013e9 JK |
2848 | if (IS_ERR(handle)) { |
2849 | /* Release dquot anyway to avoid endless cycle in dqput() */ | |
2850 | dquot_release(dquot); | |
1da177e4 | 2851 | return PTR_ERR(handle); |
9c3013e9 | 2852 | } |
1da177e4 LT |
2853 | ret = dquot_release(dquot); |
2854 | err = ext3_journal_stop(handle); | |
2855 | if (!ret) | |
2856 | ret = err; | |
2857 | return ret; | |
2858 | } | |
2859 | ||
2860 | static int ext3_mark_dquot_dirty(struct dquot *dquot) | |
2861 | { | |
99aeaf63 | 2862 | /* Are we journaling quotas? */ |
1da177e4 LT |
2863 | if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || |
2864 | EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { | |
2865 | dquot_mark_dquot_dirty(dquot); | |
2866 | return ext3_write_dquot(dquot); | |
2867 | } else { | |
2868 | return dquot_mark_dquot_dirty(dquot); | |
2869 | } | |
2870 | } | |
2871 | ||
2872 | static int ext3_write_info(struct super_block *sb, int type) | |
2873 | { | |
2874 | int ret, err; | |
2875 | handle_t *handle; | |
2876 | ||
2877 | /* Data block + inode block */ | |
2878 | handle = ext3_journal_start(sb->s_root->d_inode, 2); | |
2879 | if (IS_ERR(handle)) | |
2880 | return PTR_ERR(handle); | |
2881 | ret = dquot_commit_info(sb, type); | |
2882 | err = ext3_journal_stop(handle); | |
2883 | if (!ret) | |
2884 | ret = err; | |
2885 | return ret; | |
2886 | } | |
2887 | ||
2888 | /* | |
2889 | * Turn on quotas during mount time - we need to find | |
2890 | * the quota file and such... | |
2891 | */ | |
2892 | static int ext3_quota_on_mount(struct super_block *sb, int type) | |
2893 | { | |
287a8095 CH |
2894 | return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type], |
2895 | EXT3_SB(sb)->s_jquota_fmt, type); | |
1da177e4 LT |
2896 | } |
2897 | ||
2898 | /* | |
2899 | * Standard function to be called on quota_on | |
2900 | */ | |
2901 | static int ext3_quota_on(struct super_block *sb, int type, int format_id, | |
f00c9e44 | 2902 | struct path *path) |
1da177e4 LT |
2903 | { |
2904 | int err; | |
1da177e4 | 2905 | |
1f54587b JK |
2906 | if (!test_opt(sb, QUOTA)) |
2907 | return -EINVAL; | |
9cfe7b90 | 2908 | |
1da177e4 | 2909 | /* Quotafile not on the same filesystem? */ |
d8c9584e | 2910 | if (path->dentry->d_sb != sb) |
1da177e4 | 2911 | return -EXDEV; |
9cfe7b90 JK |
2912 | /* Journaling quota? */ |
2913 | if (EXT3_SB(sb)->s_qf_names[type]) { | |
2914 | /* Quotafile not of fs root? */ | |
f00c9e44 | 2915 | if (path->dentry->d_parent != sb->s_root) |
4cf46b67 AF |
2916 | ext3_msg(sb, KERN_WARNING, |
2917 | "warning: Quota file not on filesystem root. " | |
2918 | "Journaled quota will not work."); | |
9cfe7b90 JK |
2919 | } |
2920 | ||
2921 | /* | |
2922 | * When we journal data on quota file, we have to flush journal to see | |
2923 | * all updates to the file when we bypass pagecache... | |
2924 | */ | |
f00c9e44 | 2925 | if (ext3_should_journal_data(path->dentry->d_inode)) { |
9cfe7b90 JK |
2926 | /* |
2927 | * We don't need to lock updates but journal_flush() could | |
2928 | * otherwise be livelocked... | |
2929 | */ | |
2930 | journal_lock_updates(EXT3_SB(sb)->s_journal); | |
2d7c820e | 2931 | err = journal_flush(EXT3_SB(sb)->s_journal); |
9cfe7b90 | 2932 | journal_unlock_updates(EXT3_SB(sb)->s_journal); |
f00c9e44 | 2933 | if (err) |
2d7c820e | 2934 | return err; |
9cfe7b90 JK |
2935 | } |
2936 | ||
f00c9e44 | 2937 | return dquot_quota_on(sb, type, format_id, path); |
1da177e4 LT |
2938 | } |
2939 | ||
2940 | /* Read data from quotafile - avoid pagecache and such because we cannot afford | |
2941 | * acquiring the locks... As quota files are never truncated and quota code | |
25985edc | 2942 | * itself serializes the operations (and no one else should touch the files) |
1da177e4 LT |
2943 | * we don't have to be afraid of races */ |
2944 | static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data, | |
2945 | size_t len, loff_t off) | |
2946 | { | |
2947 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
2948 | sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb); | |
2949 | int err = 0; | |
2950 | int offset = off & (sb->s_blocksize - 1); | |
2951 | int tocopy; | |
2952 | size_t toread; | |
2953 | struct buffer_head *bh; | |
2954 | loff_t i_size = i_size_read(inode); | |
2955 | ||
2956 | if (off > i_size) | |
2957 | return 0; | |
2958 | if (off+len > i_size) | |
2959 | len = i_size-off; | |
2960 | toread = len; | |
2961 | while (toread > 0) { | |
2962 | tocopy = sb->s_blocksize - offset < toread ? | |
2963 | sb->s_blocksize - offset : toread; | |
2964 | bh = ext3_bread(NULL, inode, blk, 0, &err); | |
2965 | if (err) | |
2966 | return err; | |
2967 | if (!bh) /* A hole? */ | |
2968 | memset(data, 0, tocopy); | |
2969 | else | |
2970 | memcpy(data, bh->b_data+offset, tocopy); | |
2971 | brelse(bh); | |
2972 | offset = 0; | |
2973 | toread -= tocopy; | |
2974 | data += tocopy; | |
2975 | blk++; | |
2976 | } | |
2977 | return len; | |
2978 | } | |
2979 | ||
2980 | /* Write to quotafile (we know the transaction is already started and has | |
2981 | * enough credits) */ | |
2982 | static ssize_t ext3_quota_write(struct super_block *sb, int type, | |
2983 | const char *data, size_t len, loff_t off) | |
2984 | { | |
2985 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
2986 | sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb); | |
2987 | int err = 0; | |
2988 | int offset = off & (sb->s_blocksize - 1); | |
1da177e4 | 2989 | int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL; |
1da177e4 LT |
2990 | struct buffer_head *bh; |
2991 | handle_t *handle = journal_current_handle(); | |
2992 | ||
9c3013e9 | 2993 | if (!handle) { |
4cf46b67 AF |
2994 | ext3_msg(sb, KERN_WARNING, |
2995 | "warning: quota write (off=%llu, len=%llu)" | |
2996 | " cancelled because transaction is not started.", | |
9c3013e9 JK |
2997 | (unsigned long long)off, (unsigned long long)len); |
2998 | return -EIO; | |
2999 | } | |
e5472147 DM |
3000 | |
3001 | /* | |
3002 | * Since we account only one data block in transaction credits, | |
3003 | * then it is impossible to cross a block boundary. | |
3004 | */ | |
3005 | if (sb->s_blocksize - offset < len) { | |
3006 | ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" | |
3007 | " cancelled because not block aligned", | |
3008 | (unsigned long long)off, (unsigned long long)len); | |
3009 | return -EIO; | |
3010 | } | |
e5472147 DM |
3011 | bh = ext3_bread(handle, inode, blk, 1, &err); |
3012 | if (!bh) | |
3013 | goto out; | |
3014 | if (journal_quota) { | |
3015 | err = ext3_journal_get_write_access(handle, bh); | |
3016 | if (err) { | |
3017 | brelse(bh); | |
1da177e4 | 3018 | goto out; |
1da177e4 | 3019 | } |
1da177e4 | 3020 | } |
e5472147 DM |
3021 | lock_buffer(bh); |
3022 | memcpy(bh->b_data+offset, data, len); | |
3023 | flush_dcache_page(bh->b_page); | |
3024 | unlock_buffer(bh); | |
3025 | if (journal_quota) | |
3026 | err = ext3_journal_dirty_metadata(handle, bh); | |
3027 | else { | |
3028 | /* Always do at least ordered writes for quotas */ | |
3029 | err = ext3_journal_dirty_data(handle, bh); | |
3030 | mark_buffer_dirty(bh); | |
3031 | } | |
3032 | brelse(bh); | |
1da177e4 | 3033 | out: |
905c3937 | 3034 | if (err) |
1da177e4 | 3035 | return err; |
e5472147 DM |
3036 | if (inode->i_size < off + len) { |
3037 | i_size_write(inode, off + len); | |
1da177e4 LT |
3038 | EXT3_I(inode)->i_disksize = inode->i_size; |
3039 | } | |
3040 | inode->i_version++; | |
3041 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
3042 | ext3_mark_inode_dirty(handle, inode); | |
e5472147 | 3043 | return len; |
1da177e4 LT |
3044 | } |
3045 | ||
3046 | #endif | |
3047 | ||
152a0836 AV |
3048 | static struct dentry *ext3_mount(struct file_system_type *fs_type, |
3049 | int flags, const char *dev_name, void *data) | |
1da177e4 | 3050 | { |
152a0836 | 3051 | return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super); |
1da177e4 LT |
3052 | } |
3053 | ||
3054 | static struct file_system_type ext3_fs_type = { | |
3055 | .owner = THIS_MODULE, | |
3056 | .name = "ext3", | |
152a0836 | 3057 | .mount = ext3_mount, |
1da177e4 LT |
3058 | .kill_sb = kill_block_super, |
3059 | .fs_flags = FS_REQUIRES_DEV, | |
3060 | }; | |
3061 | ||
3062 | static int __init init_ext3_fs(void) | |
3063 | { | |
3064 | int err = init_ext3_xattr(); | |
3065 | if (err) | |
3066 | return err; | |
3067 | err = init_inodecache(); | |
3068 | if (err) | |
3069 | goto out1; | |
3070 | err = register_filesystem(&ext3_fs_type); | |
3071 | if (err) | |
3072 | goto out; | |
3073 | return 0; | |
3074 | out: | |
3075 | destroy_inodecache(); | |
3076 | out1: | |
e9ad5620 | 3077 | exit_ext3_xattr(); |
1da177e4 LT |
3078 | return err; |
3079 | } | |
3080 | ||
3081 | static void __exit exit_ext3_fs(void) | |
3082 | { | |
3083 | unregister_filesystem(&ext3_fs_type); | |
3084 | destroy_inodecache(); | |
3085 | exit_ext3_xattr(); | |
3086 | } | |
3087 | ||
3088 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
3089 | MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); | |
3090 | MODULE_LICENSE("GPL"); | |
3091 | module_init(init_ext3_fs) | |
3092 | module_exit(exit_ext3_fs) |