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 LT |
19 | #include <linux/module.h> |
20 | #include <linux/string.h> | |
21 | #include <linux/fs.h> | |
22 | #include <linux/time.h> | |
23 | #include <linux/jbd.h> | |
24 | #include <linux/ext3_fs.h> | |
25 | #include <linux/ext3_jbd.h> | |
26 | #include <linux/slab.h> | |
27 | #include <linux/init.h> | |
28 | #include <linux/blkdev.h> | |
29 | #include <linux/parser.h> | |
30 | #include <linux/smp_lock.h> | |
31 | #include <linux/buffer_head.h> | |
a5694255 | 32 | #include <linux/exportfs.h> |
1da177e4 LT |
33 | #include <linux/vfs.h> |
34 | #include <linux/random.h> | |
35 | #include <linux/mount.h> | |
36 | #include <linux/namei.h> | |
37 | #include <linux/quotaops.h> | |
8fc2751b | 38 | #include <linux/seq_file.h> |
3fc74269 | 39 | #include <linux/log2.h> |
381be254 | 40 | |
1da177e4 | 41 | #include <asm/uaccess.h> |
381be254 | 42 | |
1da177e4 LT |
43 | #include "xattr.h" |
44 | #include "acl.h" | |
381be254 | 45 | #include "namei.h" |
1da177e4 | 46 | |
bbae8bcc LT |
47 | #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED |
48 | #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA | |
49 | #else | |
50 | #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA | |
51 | #endif | |
52 | ||
71b96257 JL |
53 | static int ext3_load_journal(struct super_block *, struct ext3_super_block *, |
54 | unsigned long journal_devnum); | |
1da177e4 | 55 | static int ext3_create_journal(struct super_block *, struct ext3_super_block *, |
eee194e7 | 56 | unsigned int); |
c4be0c1d TS |
57 | static int ext3_commit_super(struct super_block *sb, |
58 | struct ext3_super_block *es, | |
1da177e4 LT |
59 | int sync); |
60 | static void ext3_mark_recovery_complete(struct super_block * sb, | |
61 | struct ext3_super_block * es); | |
62 | static void ext3_clear_journal_err(struct super_block * sb, | |
63 | struct ext3_super_block * es); | |
64 | static int ext3_sync_fs(struct super_block *sb, int wait); | |
65 | static const char *ext3_decode_error(struct super_block * sb, int errno, | |
66 | char nbuf[16]); | |
67 | static int ext3_remount (struct super_block * sb, int * flags, char * data); | |
726c3342 | 68 | static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf); |
c4be0c1d | 69 | static int ext3_unfreeze(struct super_block *sb); |
c4be0c1d | 70 | static int ext3_freeze(struct super_block *sb); |
1da177e4 | 71 | |
ae6ddcc5 | 72 | /* |
1da177e4 LT |
73 | * Wrappers for journal_start/end. |
74 | * | |
75 | * The only special thing we need to do here is to make sure that all | |
76 | * journal_end calls result in the superblock being marked dirty, so | |
77 | * that sync() will call the filesystem's write_super callback if | |
ae6ddcc5 | 78 | * appropriate. |
1da177e4 LT |
79 | */ |
80 | handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks) | |
81 | { | |
82 | journal_t *journal; | |
83 | ||
84 | if (sb->s_flags & MS_RDONLY) | |
85 | return ERR_PTR(-EROFS); | |
86 | ||
87 | /* Special case here: if the journal has aborted behind our | |
88 | * backs (eg. EIO in the commit thread), then we still need to | |
89 | * take the FS itself readonly cleanly. */ | |
90 | journal = EXT3_SB(sb)->s_journal; | |
91 | if (is_journal_aborted(journal)) { | |
e05b6b52 | 92 | ext3_abort(sb, __func__, |
1da177e4 LT |
93 | "Detected aborted journal"); |
94 | return ERR_PTR(-EROFS); | |
95 | } | |
96 | ||
97 | return journal_start(journal, nblocks); | |
98 | } | |
99 | ||
ae6ddcc5 | 100 | /* |
1da177e4 LT |
101 | * The only special thing we need to do here is to make sure that all |
102 | * journal_stop calls result in the superblock being marked dirty, so | |
103 | * that sync() will call the filesystem's write_super callback if | |
ae6ddcc5 | 104 | * appropriate. |
1da177e4 LT |
105 | */ |
106 | int __ext3_journal_stop(const char *where, handle_t *handle) | |
107 | { | |
108 | struct super_block *sb; | |
109 | int err; | |
110 | int rc; | |
111 | ||
112 | sb = handle->h_transaction->t_journal->j_private; | |
113 | err = handle->h_err; | |
114 | rc = journal_stop(handle); | |
115 | ||
116 | if (!err) | |
117 | err = rc; | |
118 | if (err) | |
119 | __ext3_std_error(sb, where, err); | |
120 | return err; | |
121 | } | |
122 | ||
123 | void ext3_journal_abort_handle(const char *caller, const char *err_fn, | |
124 | struct buffer_head *bh, handle_t *handle, int err) | |
125 | { | |
126 | char nbuf[16]; | |
127 | const char *errstr = ext3_decode_error(NULL, err, nbuf); | |
128 | ||
129 | if (bh) | |
130 | BUFFER_TRACE(bh, "abort"); | |
131 | ||
132 | if (!handle->h_err) | |
133 | handle->h_err = err; | |
134 | ||
135 | if (is_handle_aborted(handle)) | |
136 | return; | |
137 | ||
4cf46b67 AF |
138 | printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n", |
139 | caller, errstr, err_fn); | |
1da177e4 LT |
140 | |
141 | journal_abort_handle(handle); | |
142 | } | |
143 | ||
4cf46b67 AF |
144 | void ext3_msg(struct super_block *sb, const char *prefix, |
145 | const char *fmt, ...) | |
146 | { | |
147 | va_list args; | |
148 | ||
149 | va_start(args, fmt); | |
150 | printk("%sEXT3-fs (%s): ", prefix, sb->s_id); | |
151 | vprintk(fmt, args); | |
152 | printk("\n"); | |
153 | va_end(args); | |
154 | } | |
155 | ||
1da177e4 LT |
156 | /* Deal with the reporting of failure conditions on a filesystem such as |
157 | * inconsistencies detected or read IO failures. | |
158 | * | |
159 | * On ext2, we can store the error state of the filesystem in the | |
160 | * superblock. That is not possible on ext3, because we may have other | |
161 | * write ordering constraints on the superblock which prevent us from | |
162 | * writing it out straight away; and given that the journal is about to | |
163 | * be aborted, we can't rely on the current, or future, transactions to | |
164 | * write out the superblock safely. | |
165 | * | |
166 | * We'll just use the journal_abort() error code to record an error in | |
167 | * the journal instead. On recovery, the journal will compain about | |
168 | * that error until we've noted it down and cleared it. | |
169 | */ | |
170 | ||
171 | static void ext3_handle_error(struct super_block *sb) | |
172 | { | |
173 | struct ext3_super_block *es = EXT3_SB(sb)->s_es; | |
174 | ||
175 | EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS; | |
176 | es->s_state |= cpu_to_le16(EXT3_ERROR_FS); | |
177 | ||
178 | if (sb->s_flags & MS_RDONLY) | |
179 | return; | |
180 | ||
7543fc7b | 181 | if (!test_opt (sb, ERRORS_CONT)) { |
1da177e4 LT |
182 | journal_t *journal = EXT3_SB(sb)->s_journal; |
183 | ||
184 | EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT; | |
185 | if (journal) | |
186 | journal_abort(journal, -EIO); | |
187 | } | |
7543fc7b | 188 | if (test_opt (sb, ERRORS_RO)) { |
4cf46b67 AF |
189 | ext3_msg(sb, KERN_CRIT, |
190 | "error: remounting filesystem read-only"); | |
7543fc7b VA |
191 | sb->s_flags |= MS_RDONLY; |
192 | } | |
193 | ext3_commit_super(sb, es, 1); | |
1da177e4 | 194 | if (test_opt(sb, ERRORS_PANIC)) |
4cf46b67 | 195 | panic("EXT3-fs (%s): panic forced after error\n", |
1da177e4 | 196 | sb->s_id); |
1da177e4 LT |
197 | } |
198 | ||
199 | void ext3_error (struct super_block * sb, const char * function, | |
200 | const char * fmt, ...) | |
201 | { | |
202 | va_list args; | |
203 | ||
204 | va_start(args, fmt); | |
205 | printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function); | |
206 | vprintk(fmt, args); | |
207 | printk("\n"); | |
208 | va_end(args); | |
209 | ||
210 | ext3_handle_error(sb); | |
211 | } | |
212 | ||
213 | static const char *ext3_decode_error(struct super_block * sb, int errno, | |
214 | char nbuf[16]) | |
215 | { | |
216 | char *errstr = NULL; | |
217 | ||
218 | switch (errno) { | |
219 | case -EIO: | |
220 | errstr = "IO failure"; | |
221 | break; | |
222 | case -ENOMEM: | |
223 | errstr = "Out of memory"; | |
224 | break; | |
225 | case -EROFS: | |
226 | if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT) | |
227 | errstr = "Journal has aborted"; | |
228 | else | |
229 | errstr = "Readonly filesystem"; | |
230 | break; | |
231 | default: | |
232 | /* If the caller passed in an extra buffer for unknown | |
233 | * errors, textualise them now. Else we just return | |
234 | * NULL. */ | |
235 | if (nbuf) { | |
236 | /* Check for truncated error codes... */ | |
237 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
238 | errstr = nbuf; | |
239 | } | |
240 | break; | |
241 | } | |
242 | ||
243 | return errstr; | |
244 | } | |
245 | ||
246 | /* __ext3_std_error decodes expected errors from journaling functions | |
247 | * automatically and invokes the appropriate error response. */ | |
248 | ||
249 | void __ext3_std_error (struct super_block * sb, const char * function, | |
250 | int errno) | |
251 | { | |
252 | char nbuf[16]; | |
30121624 | 253 | const char *errstr; |
1da177e4 | 254 | |
30121624 ST |
255 | /* Special case: if the error is EROFS, and we're not already |
256 | * inside a transaction, then there's really no point in logging | |
257 | * an error. */ | |
258 | if (errno == -EROFS && journal_current_handle() == NULL && | |
259 | (sb->s_flags & MS_RDONLY)) | |
260 | return; | |
261 | ||
262 | errstr = ext3_decode_error(sb, errno, nbuf); | |
4cf46b67 | 263 | ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr); |
1da177e4 LT |
264 | |
265 | ext3_handle_error(sb); | |
266 | } | |
267 | ||
268 | /* | |
269 | * ext3_abort is a much stronger failure handler than ext3_error. The | |
270 | * abort function may be used to deal with unrecoverable failures such | |
271 | * as journal IO errors or ENOMEM at a critical moment in log management. | |
272 | * | |
273 | * We unconditionally force the filesystem into an ABORT|READONLY state, | |
274 | * unless the error response on the fs has been set to panic in which | |
275 | * case we take the easy way out and panic immediately. | |
276 | */ | |
277 | ||
278 | void ext3_abort (struct super_block * sb, const char * function, | |
279 | const char * fmt, ...) | |
280 | { | |
281 | va_list args; | |
282 | ||
1da177e4 | 283 | va_start(args, fmt); |
4cf46b67 | 284 | printk(KERN_CRIT "EXT3-fs (%s): error: %s: ", sb->s_id, function); |
1da177e4 LT |
285 | vprintk(fmt, args); |
286 | printk("\n"); | |
287 | va_end(args); | |
288 | ||
289 | if (test_opt(sb, ERRORS_PANIC)) | |
4cf46b67 | 290 | panic("EXT3-fs: panic from previous error\n"); |
1da177e4 LT |
291 | |
292 | if (sb->s_flags & MS_RDONLY) | |
293 | return; | |
294 | ||
4cf46b67 AF |
295 | ext3_msg(sb, KERN_CRIT, |
296 | "error: remounting filesystem read-only"); | |
1da177e4 LT |
297 | EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS; |
298 | sb->s_flags |= MS_RDONLY; | |
299 | EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT; | |
44d6f787 HK |
300 | if (EXT3_SB(sb)->s_journal) |
301 | journal_abort(EXT3_SB(sb)->s_journal, -EIO); | |
1da177e4 LT |
302 | } |
303 | ||
304 | void ext3_warning (struct super_block * sb, const char * function, | |
305 | const char * fmt, ...) | |
306 | { | |
307 | va_list args; | |
308 | ||
309 | va_start(args, fmt); | |
4cf46b67 | 310 | printk(KERN_WARNING "EXT3-fs (%s): warning: %s: ", |
1da177e4 LT |
311 | sb->s_id, function); |
312 | vprintk(fmt, args); | |
313 | printk("\n"); | |
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 | ||
350 | bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE); | |
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 | { | |
367 | bd_release(bdev); | |
9a1c3542 | 368 | return blkdev_put(bdev, FMODE_READ|FMODE_WRITE); |
1da177e4 LT |
369 | } |
370 | ||
371 | static int ext3_blkdev_remove(struct ext3_sb_info *sbi) | |
372 | { | |
373 | struct block_device *bdev; | |
374 | int ret = -ENODEV; | |
375 | ||
376 | bdev = sbi->journal_bdev; | |
377 | if (bdev) { | |
378 | ret = ext3_blkdev_put(bdev); | |
379 | sbi->journal_bdev = NULL; | |
380 | } | |
381 | return ret; | |
382 | } | |
383 | ||
384 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
385 | { | |
386 | return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode; | |
387 | } | |
388 | ||
389 | static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi) | |
390 | { | |
391 | struct list_head *l; | |
392 | ||
4cf46b67 | 393 | ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d", |
1da177e4 LT |
394 | le32_to_cpu(sbi->s_es->s_last_orphan)); |
395 | ||
4cf46b67 | 396 | ext3_msg(sb, KERN_ERR, "sb_info orphan list:"); |
1da177e4 LT |
397 | list_for_each(l, &sbi->s_orphan) { |
398 | struct inode *inode = orphan_list_entry(l); | |
4cf46b67 | 399 | ext3_msg(sb, KERN_ERR, " " |
eee194e7 | 400 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", |
1da177e4 | 401 | inode->i_sb->s_id, inode->i_ino, inode, |
ae6ddcc5 | 402 | inode->i_mode, inode->i_nlink, |
1da177e4 LT |
403 | NEXT_ORPHAN(inode)); |
404 | } | |
405 | } | |
406 | ||
407 | static void ext3_put_super (struct super_block * sb) | |
408 | { | |
409 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
410 | struct ext3_super_block *es = sbi->s_es; | |
44d6f787 | 411 | int i, err; |
1da177e4 | 412 | |
6cfd0148 CH |
413 | lock_kernel(); |
414 | ||
1da177e4 | 415 | ext3_xattr_put_super(sb); |
44d6f787 HK |
416 | err = journal_destroy(sbi->s_journal); |
417 | sbi->s_journal = NULL; | |
418 | if (err < 0) | |
2d7c820e | 419 | ext3_abort(sb, __func__, "Couldn't clean up the journal"); |
44d6f787 | 420 | |
1da177e4 LT |
421 | if (!(sb->s_flags & MS_RDONLY)) { |
422 | EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
423 | es->s_state = cpu_to_le16(sbi->s_mount_state); | |
424 | BUFFER_TRACE(sbi->s_sbh, "marking dirty"); | |
425 | mark_buffer_dirty(sbi->s_sbh); | |
426 | ext3_commit_super(sb, es, 1); | |
427 | } | |
428 | ||
429 | for (i = 0; i < sbi->s_gdb_count; i++) | |
430 | brelse(sbi->s_group_desc[i]); | |
431 | kfree(sbi->s_group_desc); | |
432 | percpu_counter_destroy(&sbi->s_freeblocks_counter); | |
433 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
434 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
435 | brelse(sbi->s_sbh); | |
436 | #ifdef CONFIG_QUOTA | |
437 | for (i = 0; i < MAXQUOTAS; i++) | |
438 | kfree(sbi->s_qf_names[i]); | |
439 | #endif | |
440 | ||
441 | /* Debugging code just in case the in-memory inode orphan list | |
442 | * isn't empty. The on-disk one can be non-empty if we've | |
443 | * detected an error and taken the fs readonly, but the | |
444 | * in-memory list had better be clean by this point. */ | |
445 | if (!list_empty(&sbi->s_orphan)) | |
446 | dump_orphan_list(sb, sbi); | |
447 | J_ASSERT(list_empty(&sbi->s_orphan)); | |
448 | ||
f98393a6 | 449 | invalidate_bdev(sb->s_bdev); |
1da177e4 LT |
450 | if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { |
451 | /* | |
452 | * Invalidate the journal device's buffers. We don't want them | |
453 | * floating about in memory - the physical journal device may | |
454 | * hotswapped, and it breaks the `ro-after' testing code. | |
455 | */ | |
456 | sync_blockdev(sbi->journal_bdev); | |
f98393a6 | 457 | invalidate_bdev(sbi->journal_bdev); |
1da177e4 LT |
458 | ext3_blkdev_remove(sbi); |
459 | } | |
460 | sb->s_fs_info = NULL; | |
5df096d6 | 461 | kfree(sbi->s_blockgroup_lock); |
1da177e4 | 462 | kfree(sbi); |
6cfd0148 CH |
463 | |
464 | unlock_kernel(); | |
1da177e4 LT |
465 | } |
466 | ||
e18b890b | 467 | static struct kmem_cache *ext3_inode_cachep; |
1da177e4 LT |
468 | |
469 | /* | |
470 | * Called inside transaction, so use GFP_NOFS | |
471 | */ | |
472 | static struct inode *ext3_alloc_inode(struct super_block *sb) | |
473 | { | |
474 | struct ext3_inode_info *ei; | |
475 | ||
e6b4f8da | 476 | ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS); |
1da177e4 LT |
477 | if (!ei) |
478 | return NULL; | |
1da177e4 LT |
479 | ei->i_block_alloc_info = NULL; |
480 | ei->vfs_inode.i_version = 1; | |
fe8bc91c JK |
481 | atomic_set(&ei->i_datasync_tid, 0); |
482 | atomic_set(&ei->i_sync_tid, 0); | |
1da177e4 LT |
483 | return &ei->vfs_inode; |
484 | } | |
485 | ||
486 | static void ext3_destroy_inode(struct inode *inode) | |
487 | { | |
9f7dd93d VA |
488 | if (!list_empty(&(EXT3_I(inode)->i_orphan))) { |
489 | printk("EXT3 Inode %p: orphan list check failed!\n", | |
490 | EXT3_I(inode)); | |
491 | print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, | |
492 | EXT3_I(inode), sizeof(struct ext3_inode_info), | |
493 | false); | |
494 | dump_stack(); | |
495 | } | |
1da177e4 LT |
496 | kmem_cache_free(ext3_inode_cachep, EXT3_I(inode)); |
497 | } | |
498 | ||
51cc5068 | 499 | static void init_once(void *foo) |
1da177e4 LT |
500 | { |
501 | struct ext3_inode_info *ei = (struct ext3_inode_info *) foo; | |
502 | ||
a35afb83 | 503 | INIT_LIST_HEAD(&ei->i_orphan); |
1da177e4 | 504 | #ifdef CONFIG_EXT3_FS_XATTR |
a35afb83 | 505 | init_rwsem(&ei->xattr_sem); |
1da177e4 | 506 | #endif |
a35afb83 CL |
507 | mutex_init(&ei->truncate_mutex); |
508 | inode_init_once(&ei->vfs_inode); | |
1da177e4 | 509 | } |
ae6ddcc5 | 510 | |
1da177e4 LT |
511 | static int init_inodecache(void) |
512 | { | |
513 | ext3_inode_cachep = kmem_cache_create("ext3_inode_cache", | |
514 | sizeof(struct ext3_inode_info), | |
fffb60f9 PJ |
515 | 0, (SLAB_RECLAIM_ACCOUNT| |
516 | SLAB_MEM_SPREAD), | |
20c2df83 | 517 | init_once); |
1da177e4 LT |
518 | if (ext3_inode_cachep == NULL) |
519 | return -ENOMEM; | |
520 | return 0; | |
521 | } | |
522 | ||
523 | static void destroy_inodecache(void) | |
524 | { | |
1a1d92c1 | 525 | kmem_cache_destroy(ext3_inode_cachep); |
1da177e4 LT |
526 | } |
527 | ||
528 | static void ext3_clear_inode(struct inode *inode) | |
529 | { | |
530 | struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info; | |
1da177e4 LT |
531 | ext3_discard_reservation(inode); |
532 | EXT3_I(inode)->i_block_alloc_info = NULL; | |
e6022603 AM |
533 | if (unlikely(rsv)) |
534 | kfree(rsv); | |
1da177e4 LT |
535 | } |
536 | ||
6cd37cda | 537 | static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb) |
8fc2751b | 538 | { |
6cd37cda | 539 | #if defined(CONFIG_QUOTA) |
275abf5b | 540 | struct ext3_sb_info *sbi = EXT3_SB(sb); |
8fc2751b | 541 | |
1aeec434 JK |
542 | if (sbi->s_jquota_fmt) { |
543 | char *fmtname = ""; | |
544 | ||
545 | switch (sbi->s_jquota_fmt) { | |
546 | case QFMT_VFS_OLD: | |
547 | fmtname = "vfsold"; | |
548 | break; | |
549 | case QFMT_VFS_V0: | |
550 | fmtname = "vfsv0"; | |
551 | break; | |
552 | case QFMT_VFS_V1: | |
553 | fmtname = "vfsv1"; | |
554 | break; | |
555 | } | |
556 | seq_printf(seq, ",jqfmt=%s", fmtname); | |
557 | } | |
8fc2751b MB |
558 | |
559 | if (sbi->s_qf_names[USRQUOTA]) | |
560 | seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); | |
561 | ||
562 | if (sbi->s_qf_names[GRPQUOTA]) | |
563 | seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); | |
564 | ||
565 | if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) | |
566 | seq_puts(seq, ",usrquota"); | |
567 | ||
568 | if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) | |
569 | seq_puts(seq, ",grpquota"); | |
570 | #endif | |
6cd37cda PO |
571 | } |
572 | ||
3c4cec65 JK |
573 | static char *data_mode_string(unsigned long mode) |
574 | { | |
575 | switch (mode) { | |
576 | case EXT3_MOUNT_JOURNAL_DATA: | |
577 | return "journal"; | |
578 | case EXT3_MOUNT_ORDERED_DATA: | |
579 | return "ordered"; | |
580 | case EXT3_MOUNT_WRITEBACK_DATA: | |
581 | return "writeback"; | |
582 | } | |
583 | return "unknown"; | |
584 | } | |
585 | ||
571beed8 MS |
586 | /* |
587 | * Show an option if | |
588 | * - it's set to a non-default value OR | |
589 | * - if the per-sb default is different from the global default | |
590 | */ | |
6cd37cda PO |
591 | static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs) |
592 | { | |
593 | struct super_block *sb = vfs->mnt_sb; | |
571beed8 MS |
594 | struct ext3_sb_info *sbi = EXT3_SB(sb); |
595 | struct ext3_super_block *es = sbi->s_es; | |
596 | unsigned long def_mount_opts; | |
597 | ||
598 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
599 | ||
600 | if (sbi->s_sb_block != 1) | |
601 | seq_printf(seq, ",sb=%lu", sbi->s_sb_block); | |
602 | if (test_opt(sb, MINIX_DF)) | |
603 | seq_puts(seq, ",minixdf"); | |
604 | if (test_opt(sb, GRPID)) | |
605 | seq_puts(seq, ",grpid"); | |
606 | if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS)) | |
607 | seq_puts(seq, ",nogrpid"); | |
608 | if (sbi->s_resuid != EXT3_DEF_RESUID || | |
609 | le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) { | |
610 | seq_printf(seq, ",resuid=%u", sbi->s_resuid); | |
611 | } | |
612 | if (sbi->s_resgid != EXT3_DEF_RESGID || | |
613 | le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) { | |
614 | seq_printf(seq, ",resgid=%u", sbi->s_resgid); | |
615 | } | |
1eca93f9 | 616 | if (test_opt(sb, ERRORS_RO)) { |
571beed8 MS |
617 | int def_errors = le16_to_cpu(es->s_errors); |
618 | ||
619 | if (def_errors == EXT3_ERRORS_PANIC || | |
1eca93f9 AK |
620 | def_errors == EXT3_ERRORS_CONTINUE) { |
621 | seq_puts(seq, ",errors=remount-ro"); | |
571beed8 MS |
622 | } |
623 | } | |
1eca93f9 AK |
624 | if (test_opt(sb, ERRORS_CONT)) |
625 | seq_puts(seq, ",errors=continue"); | |
571beed8 MS |
626 | if (test_opt(sb, ERRORS_PANIC)) |
627 | seq_puts(seq, ",errors=panic"); | |
628 | if (test_opt(sb, NO_UID32)) | |
629 | seq_puts(seq, ",nouid32"); | |
630 | if (test_opt(sb, DEBUG)) | |
631 | seq_puts(seq, ",debug"); | |
632 | if (test_opt(sb, OLDALLOC)) | |
633 | seq_puts(seq, ",oldalloc"); | |
634 | #ifdef CONFIG_EXT3_FS_XATTR | |
635 | if (test_opt(sb, XATTR_USER)) | |
636 | seq_puts(seq, ",user_xattr"); | |
637 | if (!test_opt(sb, XATTR_USER) && | |
638 | (def_mount_opts & EXT3_DEFM_XATTR_USER)) { | |
639 | seq_puts(seq, ",nouser_xattr"); | |
640 | } | |
641 | #endif | |
642 | #ifdef CONFIG_EXT3_FS_POSIX_ACL | |
643 | if (test_opt(sb, POSIX_ACL)) | |
644 | seq_puts(seq, ",acl"); | |
645 | if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL)) | |
646 | seq_puts(seq, ",noacl"); | |
647 | #endif | |
648 | if (!test_opt(sb, RESERVATION)) | |
649 | seq_puts(seq, ",noreservation"); | |
650 | if (sbi->s_commit_interval) { | |
651 | seq_printf(seq, ",commit=%u", | |
652 | (unsigned) (sbi->s_commit_interval / HZ)); | |
653 | } | |
654 | if (test_opt(sb, BARRIER)) | |
655 | seq_puts(seq, ",barrier=1"); | |
656 | if (test_opt(sb, NOBH)) | |
657 | seq_puts(seq, ",nobh"); | |
6cd37cda | 658 | |
3c4cec65 JK |
659 | seq_printf(seq, ",data=%s", data_mode_string(sbi->s_mount_opt & |
660 | EXT3_MOUNT_DATA_FLAGS)); | |
0e4fb5e2 HK |
661 | if (test_opt(sb, DATA_ERR_ABORT)) |
662 | seq_puts(seq, ",data_err=abort"); | |
663 | ||
dee1d3b6 ES |
664 | if (test_opt(sb, NOLOAD)) |
665 | seq_puts(seq, ",norecovery"); | |
666 | ||
6cd37cda | 667 | ext3_show_quota_options(seq, sb); |
8fc2751b MB |
668 | |
669 | return 0; | |
670 | } | |
1da177e4 | 671 | |
fdb36673 | 672 | |
74af0baa CH |
673 | static struct inode *ext3_nfs_get_inode(struct super_block *sb, |
674 | u64 ino, u32 generation) | |
fdb36673 | 675 | { |
fdb36673 | 676 | struct inode *inode; |
fdb36673 N |
677 | |
678 | if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO) | |
679 | return ERR_PTR(-ESTALE); | |
680 | if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count)) | |
681 | return ERR_PTR(-ESTALE); | |
682 | ||
683 | /* iget isn't really right if the inode is currently unallocated!! | |
684 | * | |
685 | * ext3_read_inode will return a bad_inode if the inode had been | |
686 | * deleted, so we should be safe. | |
687 | * | |
688 | * Currently we don't know the generation for parent directory, so | |
689 | * a generation of 0 means "accept any" | |
690 | */ | |
473043dc DH |
691 | inode = ext3_iget(sb, ino); |
692 | if (IS_ERR(inode)) | |
693 | return ERR_CAST(inode); | |
694 | if (generation && inode->i_generation != generation) { | |
fdb36673 N |
695 | iput(inode); |
696 | return ERR_PTR(-ESTALE); | |
697 | } | |
74af0baa CH |
698 | |
699 | return inode; | |
700 | } | |
701 | ||
702 | static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
703 | int fh_len, int fh_type) | |
704 | { | |
705 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
706 | ext3_nfs_get_inode); | |
707 | } | |
708 | ||
709 | static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid, | |
710 | int fh_len, int fh_type) | |
711 | { | |
712 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
713 | ext3_nfs_get_inode); | |
fdb36673 N |
714 | } |
715 | ||
6b082b53 TO |
716 | /* |
717 | * Try to release metadata pages (indirect blocks, directories) which are | |
718 | * mapped via the block device. Since these pages could have journal heads | |
719 | * which would prevent try_to_free_buffers() from freeing them, we must use | |
720 | * jbd layer's try_to_free_buffers() function to release them. | |
721 | */ | |
722 | static int bdev_try_to_free_page(struct super_block *sb, struct page *page, | |
723 | gfp_t wait) | |
724 | { | |
725 | journal_t *journal = EXT3_SB(sb)->s_journal; | |
726 | ||
727 | WARN_ON(PageChecked(page)); | |
728 | if (!page_has_buffers(page)) | |
729 | return 0; | |
730 | if (journal) | |
731 | return journal_try_to_free_buffers(journal, page, | |
732 | wait & ~__GFP_WAIT); | |
733 | return try_to_free_buffers(page); | |
734 | } | |
735 | ||
8fc2751b | 736 | #ifdef CONFIG_QUOTA |
1da177e4 LT |
737 | #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group") |
738 | #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) | |
739 | ||
1da177e4 LT |
740 | static int ext3_write_dquot(struct dquot *dquot); |
741 | static int ext3_acquire_dquot(struct dquot *dquot); | |
742 | static int ext3_release_dquot(struct dquot *dquot); | |
743 | static int ext3_mark_dquot_dirty(struct dquot *dquot); | |
744 | static int ext3_write_info(struct super_block *sb, int type); | |
2fd83a4f JK |
745 | static int ext3_quota_on(struct super_block *sb, int type, int format_id, |
746 | char *path, int remount); | |
1da177e4 LT |
747 | static int ext3_quota_on_mount(struct super_block *sb, int type); |
748 | static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data, | |
749 | size_t len, loff_t off); | |
750 | static ssize_t ext3_quota_write(struct super_block *sb, int type, | |
751 | const char *data, size_t len, loff_t off); | |
752 | ||
61e225dc | 753 | static const struct dquot_operations ext3_quota_operations = { |
a219ce37 JK |
754 | .initialize = dquot_initialize, |
755 | .drop = dquot_drop, | |
1da177e4 LT |
756 | .alloc_space = dquot_alloc_space, |
757 | .alloc_inode = dquot_alloc_inode, | |
758 | .free_space = dquot_free_space, | |
759 | .free_inode = dquot_free_inode, | |
760 | .transfer = dquot_transfer, | |
761 | .write_dquot = ext3_write_dquot, | |
762 | .acquire_dquot = ext3_acquire_dquot, | |
763 | .release_dquot = ext3_release_dquot, | |
764 | .mark_dirty = ext3_mark_dquot_dirty, | |
157091a2 JK |
765 | .write_info = ext3_write_info, |
766 | .alloc_dquot = dquot_alloc, | |
767 | .destroy_dquot = dquot_destroy, | |
1da177e4 LT |
768 | }; |
769 | ||
0d54b217 | 770 | static const struct quotactl_ops ext3_qctl_operations = { |
1da177e4 LT |
771 | .quota_on = ext3_quota_on, |
772 | .quota_off = vfs_quota_off, | |
773 | .quota_sync = vfs_quota_sync, | |
774 | .get_info = vfs_get_dqinfo, | |
775 | .set_info = vfs_set_dqinfo, | |
776 | .get_dqblk = vfs_get_dqblk, | |
777 | .set_dqblk = vfs_set_dqblk | |
778 | }; | |
779 | #endif | |
780 | ||
ee9b6d61 | 781 | static const struct super_operations ext3_sops = { |
1da177e4 LT |
782 | .alloc_inode = ext3_alloc_inode, |
783 | .destroy_inode = ext3_destroy_inode, | |
1da177e4 LT |
784 | .write_inode = ext3_write_inode, |
785 | .dirty_inode = ext3_dirty_inode, | |
786 | .delete_inode = ext3_delete_inode, | |
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, | |
793 | .clear_inode = ext3_clear_inode, | |
8fc2751b | 794 | .show_options = ext3_show_options, |
1da177e4 LT |
795 | #ifdef CONFIG_QUOTA |
796 | .quota_read = ext3_quota_read, | |
797 | .quota_write = ext3_quota_write, | |
798 | #endif | |
6b082b53 | 799 | .bdev_try_to_free_page = bdev_try_to_free_page, |
1da177e4 LT |
800 | }; |
801 | ||
39655164 | 802 | static const struct export_operations ext3_export_ops = { |
74af0baa CH |
803 | .fh_to_dentry = ext3_fh_to_dentry, |
804 | .fh_to_parent = ext3_fh_to_parent, | |
1da177e4 LT |
805 | .get_parent = ext3_get_parent, |
806 | }; | |
807 | ||
808 | enum { | |
809 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
810 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
2860b733 | 811 | Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov, |
1da177e4 | 812 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, |
ade1a29e | 813 | Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh, |
71b96257 | 814 | Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev, |
1da177e4 | 815 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, |
0e4fb5e2 | 816 | Opt_data_err_abort, Opt_data_err_ignore, |
1da177e4 | 817 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, |
1aeec434 JK |
818 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, |
819 | Opt_noquota, Opt_ignore, Opt_barrier, Opt_err, Opt_resize, | |
820 | Opt_usrquota, Opt_grpquota | |
1da177e4 LT |
821 | }; |
822 | ||
a447c093 | 823 | static const match_table_t tokens = { |
1da177e4 LT |
824 | {Opt_bsd_df, "bsddf"}, |
825 | {Opt_minix_df, "minixdf"}, | |
826 | {Opt_grpid, "grpid"}, | |
827 | {Opt_grpid, "bsdgroups"}, | |
828 | {Opt_nogrpid, "nogrpid"}, | |
829 | {Opt_nogrpid, "sysvgroups"}, | |
830 | {Opt_resgid, "resgid=%u"}, | |
831 | {Opt_resuid, "resuid=%u"}, | |
832 | {Opt_sb, "sb=%u"}, | |
833 | {Opt_err_cont, "errors=continue"}, | |
834 | {Opt_err_panic, "errors=panic"}, | |
835 | {Opt_err_ro, "errors=remount-ro"}, | |
836 | {Opt_nouid32, "nouid32"}, | |
837 | {Opt_nocheck, "nocheck"}, | |
838 | {Opt_nocheck, "check=none"}, | |
1da177e4 LT |
839 | {Opt_debug, "debug"}, |
840 | {Opt_oldalloc, "oldalloc"}, | |
841 | {Opt_orlov, "orlov"}, | |
842 | {Opt_user_xattr, "user_xattr"}, | |
843 | {Opt_nouser_xattr, "nouser_xattr"}, | |
844 | {Opt_acl, "acl"}, | |
845 | {Opt_noacl, "noacl"}, | |
846 | {Opt_reservation, "reservation"}, | |
847 | {Opt_noreservation, "noreservation"}, | |
848 | {Opt_noload, "noload"}, | |
dee1d3b6 | 849 | {Opt_noload, "norecovery"}, |
1da177e4 | 850 | {Opt_nobh, "nobh"}, |
ade1a29e | 851 | {Opt_bh, "bh"}, |
1da177e4 LT |
852 | {Opt_commit, "commit=%u"}, |
853 | {Opt_journal_update, "journal=update"}, | |
854 | {Opt_journal_inum, "journal=%u"}, | |
71b96257 | 855 | {Opt_journal_dev, "journal_dev=%u"}, |
1da177e4 LT |
856 | {Opt_abort, "abort"}, |
857 | {Opt_data_journal, "data=journal"}, | |
858 | {Opt_data_ordered, "data=ordered"}, | |
859 | {Opt_data_writeback, "data=writeback"}, | |
0e4fb5e2 HK |
860 | {Opt_data_err_abort, "data_err=abort"}, |
861 | {Opt_data_err_ignore, "data_err=ignore"}, | |
1da177e4 LT |
862 | {Opt_offusrjquota, "usrjquota="}, |
863 | {Opt_usrjquota, "usrjquota=%s"}, | |
864 | {Opt_offgrpjquota, "grpjquota="}, | |
865 | {Opt_grpjquota, "grpjquota=%s"}, | |
866 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
867 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
1aeec434 | 868 | {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, |
8fc2751b | 869 | {Opt_grpquota, "grpquota"}, |
1f54587b JK |
870 | {Opt_noquota, "noquota"}, |
871 | {Opt_quota, "quota"}, | |
8fc2751b | 872 | {Opt_usrquota, "usrquota"}, |
1da177e4 | 873 | {Opt_barrier, "barrier=%u"}, |
1da177e4 | 874 | {Opt_resize, "resize"}, |
92587216 | 875 | {Opt_err, NULL}, |
1da177e4 LT |
876 | }; |
877 | ||
4cf46b67 | 878 | static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb) |
1da177e4 | 879 | { |
e9ad5620 DK |
880 | ext3_fsblk_t sb_block; |
881 | char *options = (char *) *data; | |
1da177e4 LT |
882 | |
883 | if (!options || strncmp(options, "sb=", 3) != 0) | |
884 | return 1; /* Default location */ | |
885 | options += 3; | |
43d23f90 | 886 | /*todo: use simple_strtoll with >32bit ext3 */ |
1da177e4 LT |
887 | sb_block = simple_strtoul(options, &options, 0); |
888 | if (*options && *options != ',') { | |
4cf46b67 | 889 | ext3_msg(sb, "error: invalid sb specification: %s", |
1da177e4 LT |
890 | (char *) *data); |
891 | return 1; | |
892 | } | |
893 | if (*options == ',') | |
894 | options++; | |
895 | *data = (void *) options; | |
896 | return sb_block; | |
897 | } | |
898 | ||
71b96257 | 899 | static int parse_options (char *options, struct super_block *sb, |
eee194e7 | 900 | unsigned int *inum, unsigned long *journal_devnum, |
43d23f90 | 901 | ext3_fsblk_t *n_blocks_count, int is_remount) |
1da177e4 LT |
902 | { |
903 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
904 | char * p; | |
905 | substring_t args[MAX_OPT_ARGS]; | |
906 | int data_opt = 0; | |
907 | int option; | |
908 | #ifdef CONFIG_QUOTA | |
d06bf1d2 | 909 | int qtype, qfmt; |
1da177e4 LT |
910 | char *qname; |
911 | #endif | |
912 | ||
913 | if (!options) | |
914 | return 1; | |
915 | ||
916 | while ((p = strsep (&options, ",")) != NULL) { | |
917 | int token; | |
918 | if (!*p) | |
919 | continue; | |
920 | ||
921 | token = match_token(p, tokens, args); | |
922 | switch (token) { | |
923 | case Opt_bsd_df: | |
924 | clear_opt (sbi->s_mount_opt, MINIX_DF); | |
925 | break; | |
926 | case Opt_minix_df: | |
927 | set_opt (sbi->s_mount_opt, MINIX_DF); | |
928 | break; | |
929 | case Opt_grpid: | |
930 | set_opt (sbi->s_mount_opt, GRPID); | |
931 | break; | |
932 | case Opt_nogrpid: | |
933 | clear_opt (sbi->s_mount_opt, GRPID); | |
934 | break; | |
935 | case Opt_resuid: | |
936 | if (match_int(&args[0], &option)) | |
937 | return 0; | |
938 | sbi->s_resuid = option; | |
939 | break; | |
940 | case Opt_resgid: | |
941 | if (match_int(&args[0], &option)) | |
942 | return 0; | |
943 | sbi->s_resgid = option; | |
944 | break; | |
945 | case Opt_sb: | |
946 | /* handled by get_sb_block() instead of here */ | |
947 | /* *sb_block = match_int(&args[0]); */ | |
948 | break; | |
949 | case Opt_err_panic: | |
950 | clear_opt (sbi->s_mount_opt, ERRORS_CONT); | |
951 | clear_opt (sbi->s_mount_opt, ERRORS_RO); | |
952 | set_opt (sbi->s_mount_opt, ERRORS_PANIC); | |
953 | break; | |
954 | case Opt_err_ro: | |
955 | clear_opt (sbi->s_mount_opt, ERRORS_CONT); | |
956 | clear_opt (sbi->s_mount_opt, ERRORS_PANIC); | |
957 | set_opt (sbi->s_mount_opt, ERRORS_RO); | |
958 | break; | |
959 | case Opt_err_cont: | |
960 | clear_opt (sbi->s_mount_opt, ERRORS_RO); | |
961 | clear_opt (sbi->s_mount_opt, ERRORS_PANIC); | |
962 | set_opt (sbi->s_mount_opt, ERRORS_CONT); | |
963 | break; | |
964 | case Opt_nouid32: | |
965 | set_opt (sbi->s_mount_opt, NO_UID32); | |
966 | break; | |
1da177e4 LT |
967 | case Opt_nocheck: |
968 | clear_opt (sbi->s_mount_opt, CHECK); | |
969 | break; | |
970 | case Opt_debug: | |
971 | set_opt (sbi->s_mount_opt, DEBUG); | |
972 | break; | |
973 | case Opt_oldalloc: | |
974 | set_opt (sbi->s_mount_opt, OLDALLOC); | |
975 | break; | |
976 | case Opt_orlov: | |
977 | clear_opt (sbi->s_mount_opt, OLDALLOC); | |
978 | break; | |
979 | #ifdef CONFIG_EXT3_FS_XATTR | |
980 | case Opt_user_xattr: | |
981 | set_opt (sbi->s_mount_opt, XATTR_USER); | |
982 | break; | |
983 | case Opt_nouser_xattr: | |
984 | clear_opt (sbi->s_mount_opt, XATTR_USER); | |
985 | break; | |
986 | #else | |
987 | case Opt_user_xattr: | |
988 | case Opt_nouser_xattr: | |
4cf46b67 AF |
989 | ext3_msg(sb, KERN_INFO, |
990 | "(no)user_xattr options not supported"); | |
1da177e4 LT |
991 | break; |
992 | #endif | |
993 | #ifdef CONFIG_EXT3_FS_POSIX_ACL | |
994 | case Opt_acl: | |
995 | set_opt(sbi->s_mount_opt, POSIX_ACL); | |
996 | break; | |
997 | case Opt_noacl: | |
998 | clear_opt(sbi->s_mount_opt, POSIX_ACL); | |
999 | break; | |
1000 | #else | |
1001 | case Opt_acl: | |
1002 | case Opt_noacl: | |
4cf46b67 AF |
1003 | ext3_msg(sb, KERN_INFO, |
1004 | "(no)acl options not supported"); | |
1da177e4 LT |
1005 | break; |
1006 | #endif | |
1007 | case Opt_reservation: | |
1008 | set_opt(sbi->s_mount_opt, RESERVATION); | |
1009 | break; | |
1010 | case Opt_noreservation: | |
1011 | clear_opt(sbi->s_mount_opt, RESERVATION); | |
1012 | break; | |
1013 | case Opt_journal_update: | |
1014 | /* @@@ FIXME */ | |
1015 | /* Eventually we will want to be able to create | |
1016 | a journal file here. For now, only allow the | |
1017 | user to specify an existing inode to be the | |
1018 | journal file. */ | |
1019 | if (is_remount) { | |
4cf46b67 AF |
1020 | ext3_msg(sb, KERN_ERR, "error: cannot specify " |
1021 | "journal on remount"); | |
1da177e4 LT |
1022 | return 0; |
1023 | } | |
1024 | set_opt (sbi->s_mount_opt, UPDATE_JOURNAL); | |
1025 | break; | |
1026 | case Opt_journal_inum: | |
1027 | if (is_remount) { | |
4cf46b67 AF |
1028 | ext3_msg(sb, KERN_ERR, "error: cannot specify " |
1029 | "journal on remount"); | |
1da177e4 LT |
1030 | return 0; |
1031 | } | |
1032 | if (match_int(&args[0], &option)) | |
1033 | return 0; | |
1034 | *inum = option; | |
1035 | break; | |
71b96257 JL |
1036 | case Opt_journal_dev: |
1037 | if (is_remount) { | |
4cf46b67 AF |
1038 | ext3_msg(sb, KERN_ERR, "error: cannot specify " |
1039 | "journal on remount"); | |
71b96257 JL |
1040 | return 0; |
1041 | } | |
1042 | if (match_int(&args[0], &option)) | |
1043 | return 0; | |
1044 | *journal_devnum = option; | |
1045 | break; | |
1da177e4 LT |
1046 | case Opt_noload: |
1047 | set_opt (sbi->s_mount_opt, NOLOAD); | |
1048 | break; | |
1049 | case Opt_commit: | |
1050 | if (match_int(&args[0], &option)) | |
1051 | return 0; | |
1052 | if (option < 0) | |
1053 | return 0; | |
1054 | if (option == 0) | |
1055 | option = JBD_DEFAULT_MAX_COMMIT_AGE; | |
1056 | sbi->s_commit_interval = HZ * option; | |
1057 | break; | |
1058 | case Opt_data_journal: | |
1059 | data_opt = EXT3_MOUNT_JOURNAL_DATA; | |
1060 | goto datacheck; | |
1061 | case Opt_data_ordered: | |
1062 | data_opt = EXT3_MOUNT_ORDERED_DATA; | |
1063 | goto datacheck; | |
1064 | case Opt_data_writeback: | |
1065 | data_opt = EXT3_MOUNT_WRITEBACK_DATA; | |
1066 | datacheck: | |
1067 | if (is_remount) { | |
1068 | if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS) | |
3c4cec65 JK |
1069 | == data_opt) |
1070 | break; | |
4cf46b67 AF |
1071 | ext3_msg(sb, KERN_ERR, |
1072 | "error: cannot change " | |
3c4cec65 JK |
1073 | "data mode on remount. The filesystem " |
1074 | "is mounted in data=%s mode and you " | |
4cf46b67 | 1075 | "try to remount it in data=%s mode.", |
3c4cec65 JK |
1076 | data_mode_string(sbi->s_mount_opt & |
1077 | EXT3_MOUNT_DATA_FLAGS), | |
1078 | data_mode_string(data_opt)); | |
1079 | return 0; | |
1da177e4 LT |
1080 | } else { |
1081 | sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS; | |
1082 | sbi->s_mount_opt |= data_opt; | |
1083 | } | |
1084 | break; | |
0e4fb5e2 HK |
1085 | case Opt_data_err_abort: |
1086 | set_opt(sbi->s_mount_opt, DATA_ERR_ABORT); | |
1087 | break; | |
1088 | case Opt_data_err_ignore: | |
1089 | clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT); | |
1090 | break; | |
1da177e4 LT |
1091 | #ifdef CONFIG_QUOTA |
1092 | case Opt_usrjquota: | |
1093 | qtype = USRQUOTA; | |
1094 | goto set_qf_name; | |
1095 | case Opt_grpjquota: | |
1096 | qtype = GRPQUOTA; | |
1097 | set_qf_name: | |
ee0d5ffe | 1098 | if (sb_any_quota_loaded(sb) && |
d06bf1d2 | 1099 | !sbi->s_qf_names[qtype]) { |
4cf46b67 AF |
1100 | ext3_msg(sb, KERN_ERR, |
1101 | "error: cannot change journaled " | |
1102 | "quota options when quota turned on."); | |
1da177e4 LT |
1103 | return 0; |
1104 | } | |
1105 | qname = match_strdup(&args[0]); | |
1106 | if (!qname) { | |
4cf46b67 AF |
1107 | ext3_msg(sb, KERN_ERR, |
1108 | "error: not enough memory for " | |
1109 | "storing quotafile name."); | |
1da177e4 LT |
1110 | return 0; |
1111 | } | |
1112 | if (sbi->s_qf_names[qtype] && | |
1113 | strcmp(sbi->s_qf_names[qtype], qname)) { | |
4cf46b67 AF |
1114 | ext3_msg(sb, KERN_ERR, |
1115 | "error: %s quota file already " | |
1116 | "specified.", QTYPE2NAME(qtype)); | |
1da177e4 LT |
1117 | kfree(qname); |
1118 | return 0; | |
1119 | } | |
1120 | sbi->s_qf_names[qtype] = qname; | |
1121 | if (strchr(sbi->s_qf_names[qtype], '/')) { | |
4cf46b67 AF |
1122 | ext3_msg(sb, KERN_ERR, |
1123 | "error: quotafile must be on " | |
1124 | "filesystem root."); | |
1da177e4 LT |
1125 | kfree(sbi->s_qf_names[qtype]); |
1126 | sbi->s_qf_names[qtype] = NULL; | |
1127 | return 0; | |
1128 | } | |
1f54587b | 1129 | set_opt(sbi->s_mount_opt, QUOTA); |
1da177e4 LT |
1130 | break; |
1131 | case Opt_offusrjquota: | |
1132 | qtype = USRQUOTA; | |
1133 | goto clear_qf_name; | |
1134 | case Opt_offgrpjquota: | |
1135 | qtype = GRPQUOTA; | |
1136 | clear_qf_name: | |
ee0d5ffe | 1137 | if (sb_any_quota_loaded(sb) && |
d06bf1d2 | 1138 | sbi->s_qf_names[qtype]) { |
4cf46b67 | 1139 | ext3_msg(sb, KERN_ERR, "error: cannot change " |
99aeaf63 | 1140 | "journaled quota options when " |
4cf46b67 | 1141 | "quota turned on."); |
1da177e4 LT |
1142 | return 0; |
1143 | } | |
08c6a96f JK |
1144 | /* |
1145 | * The space will be released later when all options | |
1146 | * are confirmed to be correct | |
1147 | */ | |
1da177e4 LT |
1148 | sbi->s_qf_names[qtype] = NULL; |
1149 | break; | |
1150 | case Opt_jqfmt_vfsold: | |
d06bf1d2 JK |
1151 | qfmt = QFMT_VFS_OLD; |
1152 | goto set_qf_format; | |
1da177e4 | 1153 | case Opt_jqfmt_vfsv0: |
d06bf1d2 | 1154 | qfmt = QFMT_VFS_V0; |
1aeec434 JK |
1155 | goto set_qf_format; |
1156 | case Opt_jqfmt_vfsv1: | |
1157 | qfmt = QFMT_VFS_V1; | |
d06bf1d2 | 1158 | set_qf_format: |
ee0d5ffe | 1159 | if (sb_any_quota_loaded(sb) && |
d06bf1d2 | 1160 | sbi->s_jquota_fmt != qfmt) { |
4cf46b67 | 1161 | ext3_msg(sb, KERN_ERR, "error: cannot change " |
d06bf1d2 | 1162 | "journaled quota options when " |
4cf46b67 | 1163 | "quota turned on."); |
d06bf1d2 JK |
1164 | return 0; |
1165 | } | |
1166 | sbi->s_jquota_fmt = qfmt; | |
1da177e4 | 1167 | break; |
1f54587b | 1168 | case Opt_quota: |
8fc2751b MB |
1169 | case Opt_usrquota: |
1170 | set_opt(sbi->s_mount_opt, QUOTA); | |
1171 | set_opt(sbi->s_mount_opt, USRQUOTA); | |
1172 | break; | |
1173 | case Opt_grpquota: | |
1f54587b | 1174 | set_opt(sbi->s_mount_opt, QUOTA); |
8fc2751b | 1175 | set_opt(sbi->s_mount_opt, GRPQUOTA); |
1f54587b JK |
1176 | break; |
1177 | case Opt_noquota: | |
ee0d5ffe | 1178 | if (sb_any_quota_loaded(sb)) { |
4cf46b67 AF |
1179 | ext3_msg(sb, KERN_ERR, "error: cannot change " |
1180 | "quota options when quota turned on."); | |
1f54587b JK |
1181 | return 0; |
1182 | } | |
1183 | clear_opt(sbi->s_mount_opt, QUOTA); | |
8fc2751b MB |
1184 | clear_opt(sbi->s_mount_opt, USRQUOTA); |
1185 | clear_opt(sbi->s_mount_opt, GRPQUOTA); | |
1f54587b | 1186 | break; |
1da177e4 | 1187 | #else |
8fc2751b MB |
1188 | case Opt_quota: |
1189 | case Opt_usrquota: | |
1190 | case Opt_grpquota: | |
4cf46b67 AF |
1191 | ext3_msg(sb, KERN_ERR, |
1192 | "error: quota options not supported."); | |
fa1ff1e0 | 1193 | break; |
1da177e4 LT |
1194 | case Opt_usrjquota: |
1195 | case Opt_grpjquota: | |
1196 | case Opt_offusrjquota: | |
1197 | case Opt_offgrpjquota: | |
1198 | case Opt_jqfmt_vfsold: | |
1199 | case Opt_jqfmt_vfsv0: | |
1aeec434 | 1200 | case Opt_jqfmt_vfsv1: |
4cf46b67 AF |
1201 | ext3_msg(sb, KERN_ERR, |
1202 | "error: journaled quota options not " | |
1203 | "supported."); | |
1da177e4 | 1204 | break; |
1f54587b JK |
1205 | case Opt_noquota: |
1206 | break; | |
1da177e4 LT |
1207 | #endif |
1208 | case Opt_abort: | |
1209 | set_opt(sbi->s_mount_opt, ABORT); | |
1210 | break; | |
1211 | case Opt_barrier: | |
1212 | if (match_int(&args[0], &option)) | |
1213 | return 0; | |
1214 | if (option) | |
1215 | set_opt(sbi->s_mount_opt, BARRIER); | |
1216 | else | |
1217 | clear_opt(sbi->s_mount_opt, BARRIER); | |
1218 | break; | |
1219 | case Opt_ignore: | |
1220 | break; | |
1221 | case Opt_resize: | |
08c6a96f | 1222 | if (!is_remount) { |
4cf46b67 AF |
1223 | ext3_msg(sb, KERN_ERR, |
1224 | "error: resize option only available " | |
1225 | "for remount"); | |
1da177e4 LT |
1226 | return 0; |
1227 | } | |
c7f1721e KZ |
1228 | if (match_int(&args[0], &option) != 0) |
1229 | return 0; | |
1da177e4 LT |
1230 | *n_blocks_count = option; |
1231 | break; | |
1232 | case Opt_nobh: | |
1233 | set_opt(sbi->s_mount_opt, NOBH); | |
1234 | break; | |
ade1a29e BP |
1235 | case Opt_bh: |
1236 | clear_opt(sbi->s_mount_opt, NOBH); | |
1237 | break; | |
1da177e4 | 1238 | default: |
4cf46b67 AF |
1239 | ext3_msg(sb, KERN_ERR, |
1240 | "error: unrecognized mount option \"%s\" " | |
1241 | "or missing value", p); | |
1da177e4 LT |
1242 | return 0; |
1243 | } | |
1244 | } | |
1245 | #ifdef CONFIG_QUOTA | |
8fc2751b MB |
1246 | if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { |
1247 | if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) && | |
1248 | sbi->s_qf_names[USRQUOTA]) | |
1249 | clear_opt(sbi->s_mount_opt, USRQUOTA); | |
1250 | ||
1251 | if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) && | |
1252 | sbi->s_qf_names[GRPQUOTA]) | |
1253 | clear_opt(sbi->s_mount_opt, GRPQUOTA); | |
1254 | ||
1255 | if ((sbi->s_qf_names[USRQUOTA] && | |
1256 | (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) || | |
1257 | (sbi->s_qf_names[GRPQUOTA] && | |
1258 | (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) { | |
4cf46b67 AF |
1259 | ext3_msg(sb, KERN_ERR, "error: old and new quota " |
1260 | "format mixing."); | |
8fc2751b MB |
1261 | return 0; |
1262 | } | |
1263 | ||
1264 | if (!sbi->s_jquota_fmt) { | |
4cf46b67 AF |
1265 | ext3_msg(sb, KERN_ERR, "error: journaled quota format " |
1266 | "not specified."); | |
8fc2751b MB |
1267 | return 0; |
1268 | } | |
1269 | } else { | |
1270 | if (sbi->s_jquota_fmt) { | |
4cf46b67 | 1271 | ext3_msg(sb, KERN_ERR, "error: journaled quota format " |
99aeaf63 | 1272 | "specified with no journaling " |
4cf46b67 | 1273 | "enabled."); |
8fc2751b MB |
1274 | return 0; |
1275 | } | |
1da177e4 LT |
1276 | } |
1277 | #endif | |
1da177e4 LT |
1278 | return 1; |
1279 | } | |
1280 | ||
1281 | static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es, | |
1282 | int read_only) | |
1283 | { | |
1284 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
1285 | int res = 0; | |
1286 | ||
1287 | if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) { | |
4cf46b67 AF |
1288 | ext3_msg(sb, KERN_ERR, |
1289 | "error: revision level too high, " | |
1290 | "forcing read-only mode"); | |
1da177e4 LT |
1291 | res = MS_RDONLY; |
1292 | } | |
1293 | if (read_only) | |
1294 | return res; | |
1295 | if (!(sbi->s_mount_state & EXT3_VALID_FS)) | |
4cf46b67 AF |
1296 | ext3_msg(sb, KERN_WARNING, |
1297 | "warning: mounting unchecked fs, " | |
1298 | "running e2fsck is recommended"); | |
1da177e4 | 1299 | else if ((sbi->s_mount_state & EXT3_ERROR_FS)) |
4cf46b67 AF |
1300 | ext3_msg(sb, KERN_WARNING, |
1301 | "warning: mounting fs with errors, " | |
1302 | "running e2fsck is recommended"); | |
1da177e4 LT |
1303 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 && |
1304 | le16_to_cpu(es->s_mnt_count) >= | |
1305 | (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) | |
4cf46b67 AF |
1306 | ext3_msg(sb, KERN_WARNING, |
1307 | "warning: maximal mount count reached, " | |
1308 | "running e2fsck is recommended"); | |
1da177e4 LT |
1309 | else if (le32_to_cpu(es->s_checkinterval) && |
1310 | (le32_to_cpu(es->s_lastcheck) + | |
1311 | le32_to_cpu(es->s_checkinterval) <= get_seconds())) | |
4cf46b67 AF |
1312 | ext3_msg(sb, KERN_WARNING, |
1313 | "warning: checktime reached, " | |
1314 | "running e2fsck is recommended"); | |
1da177e4 LT |
1315 | #if 0 |
1316 | /* @@@ We _will_ want to clear the valid bit if we find | |
1317 | inconsistencies, to force a fsck at reboot. But for | |
1318 | a plain journaled filesystem we can keep it set as | |
1319 | valid forever! :) */ | |
e7f23ebd | 1320 | es->s_state &= cpu_to_le16(~EXT3_VALID_FS); |
1da177e4 LT |
1321 | #endif |
1322 | if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) | |
1323 | es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT); | |
50e8a289 | 1324 | le16_add_cpu(&es->s_mnt_count, 1); |
1da177e4 LT |
1325 | es->s_mtime = cpu_to_le32(get_seconds()); |
1326 | ext3_update_dynamic_rev(sb); | |
1327 | EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
1328 | ||
1329 | ext3_commit_super(sb, es, 1); | |
1330 | if (test_opt(sb, DEBUG)) | |
4cf46b67 AF |
1331 | ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, " |
1332 | "bpg=%lu, ipg=%lu, mo=%04lx]", | |
1da177e4 LT |
1333 | sb->s_blocksize, |
1334 | sbi->s_groups_count, | |
1335 | EXT3_BLOCKS_PER_GROUP(sb), | |
1336 | EXT3_INODES_PER_GROUP(sb), | |
1337 | sbi->s_mount_opt); | |
1338 | ||
1da177e4 LT |
1339 | if (EXT3_SB(sb)->s_journal->j_inode == NULL) { |
1340 | char b[BDEVNAME_SIZE]; | |
4cf46b67 | 1341 | ext3_msg(sb, KERN_INFO, "using external journal on %s", |
1da177e4 LT |
1342 | bdevname(EXT3_SB(sb)->s_journal->j_dev, b)); |
1343 | } else { | |
4cf46b67 | 1344 | ext3_msg(sb, KERN_INFO, "using internal journal"); |
1da177e4 | 1345 | } |
1da177e4 LT |
1346 | return res; |
1347 | } | |
1348 | ||
1349 | /* Called at mount-time, super-block is locked */ | |
197cd65a | 1350 | static int ext3_check_descriptors(struct super_block *sb) |
1da177e4 LT |
1351 | { |
1352 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
1da177e4 LT |
1353 | int i; |
1354 | ||
1355 | ext3_debug ("Checking group descriptors"); | |
1356 | ||
197cd65a AM |
1357 | for (i = 0; i < sbi->s_groups_count; i++) { |
1358 | struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL); | |
1eaafeae AM |
1359 | ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i); |
1360 | ext3_fsblk_t last_block; | |
197cd65a | 1361 | |
855565e8 ES |
1362 | if (i == sbi->s_groups_count - 1) |
1363 | last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1; | |
1364 | else | |
1365 | last_block = first_block + | |
1366 | (EXT3_BLOCKS_PER_GROUP(sb) - 1); | |
1367 | ||
855565e8 ES |
1368 | if (le32_to_cpu(gdp->bg_block_bitmap) < first_block || |
1369 | le32_to_cpu(gdp->bg_block_bitmap) > last_block) | |
1da177e4 LT |
1370 | { |
1371 | ext3_error (sb, "ext3_check_descriptors", | |
1372 | "Block bitmap for group %d" | |
1373 | " not in group (block %lu)!", | |
1374 | i, (unsigned long) | |
1375 | le32_to_cpu(gdp->bg_block_bitmap)); | |
1376 | return 0; | |
1377 | } | |
855565e8 ES |
1378 | if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block || |
1379 | le32_to_cpu(gdp->bg_inode_bitmap) > last_block) | |
1da177e4 LT |
1380 | { |
1381 | ext3_error (sb, "ext3_check_descriptors", | |
1382 | "Inode bitmap for group %d" | |
1383 | " not in group (block %lu)!", | |
1384 | i, (unsigned long) | |
1385 | le32_to_cpu(gdp->bg_inode_bitmap)); | |
1386 | return 0; | |
1387 | } | |
855565e8 | 1388 | if (le32_to_cpu(gdp->bg_inode_table) < first_block || |
780dcdb2 | 1389 | le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 > |
855565e8 | 1390 | last_block) |
1da177e4 LT |
1391 | { |
1392 | ext3_error (sb, "ext3_check_descriptors", | |
1393 | "Inode table for group %d" | |
1394 | " not in group (block %lu)!", | |
1395 | i, (unsigned long) | |
1396 | le32_to_cpu(gdp->bg_inode_table)); | |
1397 | return 0; | |
1398 | } | |
1da177e4 LT |
1399 | } |
1400 | ||
1401 | sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb)); | |
1402 | sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb)); | |
1403 | return 1; | |
1404 | } | |
1405 | ||
1406 | ||
1407 | /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at | |
1408 | * the superblock) which were deleted from all directories, but held open by | |
1409 | * a process at the time of a crash. We walk the list and try to delete these | |
1410 | * inodes at recovery time (only with a read-write filesystem). | |
1411 | * | |
1412 | * In order to keep the orphan inode chain consistent during traversal (in | |
1413 | * case of crash during recovery), we link each inode into the superblock | |
1414 | * orphan list_head and handle it the same way as an inode deletion during | |
1415 | * normal operation (which journals the operations for us). | |
1416 | * | |
1417 | * We only do an iget() and an iput() on each inode, which is very safe if we | |
1418 | * accidentally point at an in-use or already deleted inode. The worst that | |
1419 | * can happen in this case is that we get a "bit already cleared" message from | |
1420 | * ext3_free_inode(). The only reason we would point at a wrong inode is if | |
1421 | * e2fsck was run on this filesystem, and it must have already done the orphan | |
1422 | * inode cleanup for us, so we can safely abort without any further action. | |
1423 | */ | |
1424 | static void ext3_orphan_cleanup (struct super_block * sb, | |
1425 | struct ext3_super_block * es) | |
1426 | { | |
1427 | unsigned int s_flags = sb->s_flags; | |
1428 | int nr_orphans = 0, nr_truncates = 0; | |
1429 | #ifdef CONFIG_QUOTA | |
1430 | int i; | |
1431 | #endif | |
1432 | if (!es->s_last_orphan) { | |
1433 | jbd_debug(4, "no orphan inodes to clean up\n"); | |
1434 | return; | |
1435 | } | |
1436 | ||
a8f48a95 | 1437 | if (bdev_read_only(sb->s_bdev)) { |
4cf46b67 AF |
1438 | ext3_msg(sb, KERN_ERR, "error: write access " |
1439 | "unavailable, skipping orphan cleanup."); | |
a8f48a95 ES |
1440 | return; |
1441 | } | |
1442 | ||
1da177e4 LT |
1443 | if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) { |
1444 | if (es->s_last_orphan) | |
1445 | jbd_debug(1, "Errors on filesystem, " | |
1446 | "clearing orphan list.\n"); | |
1447 | es->s_last_orphan = 0; | |
1448 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); | |
1449 | return; | |
1450 | } | |
1451 | ||
1452 | if (s_flags & MS_RDONLY) { | |
4cf46b67 | 1453 | ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); |
1da177e4 LT |
1454 | sb->s_flags &= ~MS_RDONLY; |
1455 | } | |
1456 | #ifdef CONFIG_QUOTA | |
1457 | /* Needed for iput() to work correctly and not trash data */ | |
1458 | sb->s_flags |= MS_ACTIVE; | |
1459 | /* Turn on quotas so that they are updated correctly */ | |
1460 | for (i = 0; i < MAXQUOTAS; i++) { | |
1461 | if (EXT3_SB(sb)->s_qf_names[i]) { | |
1462 | int ret = ext3_quota_on_mount(sb, i); | |
1463 | if (ret < 0) | |
4cf46b67 AF |
1464 | ext3_msg(sb, KERN_ERR, |
1465 | "error: cannot turn on journaled " | |
1466 | "quota: %d", ret); | |
1da177e4 LT |
1467 | } |
1468 | } | |
1469 | #endif | |
1470 | ||
1471 | while (es->s_last_orphan) { | |
1472 | struct inode *inode; | |
1473 | ||
473043dc DH |
1474 | inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); |
1475 | if (IS_ERR(inode)) { | |
1da177e4 LT |
1476 | es->s_last_orphan = 0; |
1477 | break; | |
1478 | } | |
1479 | ||
1480 | list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan); | |
81a05227 | 1481 | vfs_dq_init(inode); |
1da177e4 LT |
1482 | if (inode->i_nlink) { |
1483 | printk(KERN_DEBUG | |
eee194e7 | 1484 | "%s: truncating inode %lu to %Ld bytes\n", |
e05b6b52 | 1485 | __func__, inode->i_ino, inode->i_size); |
eee194e7 | 1486 | jbd_debug(2, "truncating inode %lu to %Ld bytes\n", |
1da177e4 LT |
1487 | inode->i_ino, inode->i_size); |
1488 | ext3_truncate(inode); | |
1489 | nr_truncates++; | |
1490 | } else { | |
1491 | printk(KERN_DEBUG | |
eee194e7 | 1492 | "%s: deleting unreferenced inode %lu\n", |
e05b6b52 | 1493 | __func__, inode->i_ino); |
eee194e7 | 1494 | jbd_debug(2, "deleting unreferenced inode %lu\n", |
1da177e4 LT |
1495 | inode->i_ino); |
1496 | nr_orphans++; | |
1497 | } | |
1498 | iput(inode); /* The delete magic happens here! */ | |
1499 | } | |
1500 | ||
1501 | #define PLURAL(x) (x), ((x)==1) ? "" : "s" | |
1502 | ||
1503 | if (nr_orphans) | |
4cf46b67 AF |
1504 | ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted", |
1505 | PLURAL(nr_orphans)); | |
1da177e4 | 1506 | if (nr_truncates) |
4cf46b67 AF |
1507 | ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up", |
1508 | PLURAL(nr_truncates)); | |
1da177e4 LT |
1509 | #ifdef CONFIG_QUOTA |
1510 | /* Turn quotas off */ | |
1511 | for (i = 0; i < MAXQUOTAS; i++) { | |
1512 | if (sb_dqopt(sb)->files[i]) | |
2fd83a4f | 1513 | vfs_quota_off(sb, i, 0); |
1da177e4 LT |
1514 | } |
1515 | #endif | |
1516 | sb->s_flags = s_flags; /* Restore MS_RDONLY status */ | |
1517 | } | |
1518 | ||
1da177e4 LT |
1519 | /* |
1520 | * Maximal file size. There is a direct, and {,double-,triple-}indirect | |
1521 | * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks. | |
1522 | * We need to be 1 filesystem block less than the 2^32 sector limit. | |
1523 | */ | |
1524 | static loff_t ext3_max_size(int bits) | |
1525 | { | |
1526 | loff_t res = EXT3_NDIR_BLOCKS; | |
fe7fdc37 AK |
1527 | int meta_blocks; |
1528 | loff_t upper_limit; | |
1529 | ||
1530 | /* This is calculated to be the largest file size for a | |
1531 | * dense, file such that the total number of | |
1da177e4 | 1532 | * sectors in the file, including data and all indirect blocks, |
fe7fdc37 AK |
1533 | * does not exceed 2^32 -1 |
1534 | * __u32 i_blocks representing the total number of | |
1535 | * 512 bytes blocks of the file | |
1536 | */ | |
1537 | upper_limit = (1LL << 32) - 1; | |
1538 | ||
1539 | /* total blocks in file system block size */ | |
1540 | upper_limit >>= (bits - 9); | |
1541 | ||
1542 | ||
1543 | /* indirect blocks */ | |
1544 | meta_blocks = 1; | |
1545 | /* double indirect blocks */ | |
1546 | meta_blocks += 1 + (1LL << (bits-2)); | |
1547 | /* tripple indirect blocks */ | |
1548 | meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); | |
1549 | ||
1550 | upper_limit -= meta_blocks; | |
1551 | upper_limit <<= bits; | |
1da177e4 LT |
1552 | |
1553 | res += 1LL << (bits-2); | |
1554 | res += 1LL << (2*(bits-2)); | |
1555 | res += 1LL << (3*(bits-2)); | |
1556 | res <<= bits; | |
1557 | if (res > upper_limit) | |
1558 | res = upper_limit; | |
fe7fdc37 AK |
1559 | |
1560 | if (res > MAX_LFS_FILESIZE) | |
1561 | res = MAX_LFS_FILESIZE; | |
1562 | ||
1da177e4 LT |
1563 | return res; |
1564 | } | |
1565 | ||
43d23f90 MC |
1566 | static ext3_fsblk_t descriptor_loc(struct super_block *sb, |
1567 | ext3_fsblk_t logic_sb_block, | |
1da177e4 LT |
1568 | int nr) |
1569 | { | |
1570 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
43d23f90 | 1571 | unsigned long bg, first_meta_bg; |
1da177e4 LT |
1572 | int has_super = 0; |
1573 | ||
1da177e4 LT |
1574 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); |
1575 | ||
1576 | if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) || | |
1577 | nr < first_meta_bg) | |
1578 | return (logic_sb_block + nr + 1); | |
1579 | bg = sbi->s_desc_per_block * nr; | |
1580 | if (ext3_bg_has_super(sb, bg)) | |
1581 | has_super = 1; | |
43d23f90 | 1582 | return (has_super + ext3_group_first_block_no(sb, bg)); |
1da177e4 LT |
1583 | } |
1584 | ||
1585 | ||
1586 | static int ext3_fill_super (struct super_block *sb, void *data, int silent) | |
1587 | { | |
1588 | struct buffer_head * bh; | |
1589 | struct ext3_super_block *es = NULL; | |
1590 | struct ext3_sb_info *sbi; | |
43d23f90 | 1591 | ext3_fsblk_t block; |
4cf46b67 | 1592 | ext3_fsblk_t sb_block = get_sb_block(&data, sb); |
43d23f90 | 1593 | ext3_fsblk_t logic_sb_block; |
1da177e4 | 1594 | unsigned long offset = 0; |
eee194e7 | 1595 | unsigned int journal_inum = 0; |
71b96257 | 1596 | unsigned long journal_devnum = 0; |
1da177e4 LT |
1597 | unsigned long def_mount_opts; |
1598 | struct inode *root; | |
1599 | int blocksize; | |
1600 | int hblock; | |
1601 | int db_count; | |
1602 | int i; | |
1603 | int needs_recovery; | |
473043dc | 1604 | int ret = -EINVAL; |
1da177e4 | 1605 | __le32 features; |
833f4077 | 1606 | int err; |
1da177e4 | 1607 | |
f8314dc6 | 1608 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); |
1da177e4 LT |
1609 | if (!sbi) |
1610 | return -ENOMEM; | |
5df096d6 PE |
1611 | |
1612 | sbi->s_blockgroup_lock = | |
1613 | kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); | |
1614 | if (!sbi->s_blockgroup_lock) { | |
1615 | kfree(sbi); | |
1616 | return -ENOMEM; | |
1617 | } | |
1da177e4 | 1618 | sb->s_fs_info = sbi; |
1da177e4 LT |
1619 | sbi->s_mount_opt = 0; |
1620 | sbi->s_resuid = EXT3_DEF_RESUID; | |
1621 | sbi->s_resgid = EXT3_DEF_RESGID; | |
571beed8 | 1622 | sbi->s_sb_block = sb_block; |
1da177e4 LT |
1623 | |
1624 | unlock_kernel(); | |
1625 | ||
1626 | blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE); | |
1627 | if (!blocksize) { | |
4cf46b67 | 1628 | ext3_msg(sb, KERN_ERR, "error: unable to set blocksize"); |
1da177e4 LT |
1629 | goto out_fail; |
1630 | } | |
1631 | ||
1632 | /* | |
1633 | * The ext3 superblock will not be buffer aligned for other than 1kB | |
1634 | * block sizes. We need to calculate the offset from buffer start. | |
1635 | */ | |
1636 | if (blocksize != EXT3_MIN_BLOCK_SIZE) { | |
1637 | logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize; | |
1638 | offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize; | |
1639 | } else { | |
1640 | logic_sb_block = sb_block; | |
1641 | } | |
1642 | ||
1643 | if (!(bh = sb_bread(sb, logic_sb_block))) { | |
4cf46b67 | 1644 | ext3_msg(sb, KERN_ERR, "error: unable to read superblock"); |
1da177e4 LT |
1645 | goto out_fail; |
1646 | } | |
1647 | /* | |
1648 | * Note: s_es must be initialized as soon as possible because | |
1649 | * some ext3 macro-instructions depend on its value | |
1650 | */ | |
1651 | es = (struct ext3_super_block *) (((char *)bh->b_data) + offset); | |
1652 | sbi->s_es = es; | |
1653 | sb->s_magic = le16_to_cpu(es->s_magic); | |
1654 | if (sb->s_magic != EXT3_SUPER_MAGIC) | |
1655 | goto cantfind_ext3; | |
1656 | ||
1657 | /* Set defaults before we parse the mount options */ | |
1658 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
1659 | if (def_mount_opts & EXT3_DEFM_DEBUG) | |
1660 | set_opt(sbi->s_mount_opt, DEBUG); | |
1661 | if (def_mount_opts & EXT3_DEFM_BSDGROUPS) | |
1662 | set_opt(sbi->s_mount_opt, GRPID); | |
1663 | if (def_mount_opts & EXT3_DEFM_UID16) | |
1664 | set_opt(sbi->s_mount_opt, NO_UID32); | |
2e7842b8 | 1665 | #ifdef CONFIG_EXT3_FS_XATTR |
1da177e4 LT |
1666 | if (def_mount_opts & EXT3_DEFM_XATTR_USER) |
1667 | set_opt(sbi->s_mount_opt, XATTR_USER); | |
2e7842b8 HD |
1668 | #endif |
1669 | #ifdef CONFIG_EXT3_FS_POSIX_ACL | |
1da177e4 LT |
1670 | if (def_mount_opts & EXT3_DEFM_ACL) |
1671 | set_opt(sbi->s_mount_opt, POSIX_ACL); | |
2e7842b8 | 1672 | #endif |
1da177e4 LT |
1673 | if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA) |
1674 | sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA; | |
1675 | else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED) | |
1676 | sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA; | |
1677 | else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK) | |
1678 | sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA; | |
1679 | ||
1680 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC) | |
1681 | set_opt(sbi->s_mount_opt, ERRORS_PANIC); | |
1eca93f9 | 1682 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE) |
2245d7c2 | 1683 | set_opt(sbi->s_mount_opt, ERRORS_CONT); |
1eca93f9 AK |
1684 | else |
1685 | set_opt(sbi->s_mount_opt, ERRORS_RO); | |
1da177e4 LT |
1686 | |
1687 | sbi->s_resuid = le16_to_cpu(es->s_def_resuid); | |
1688 | sbi->s_resgid = le16_to_cpu(es->s_def_resgid); | |
1689 | ||
1690 | set_opt(sbi->s_mount_opt, RESERVATION); | |
1691 | ||
71b96257 JL |
1692 | if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum, |
1693 | NULL, 0)) | |
1da177e4 LT |
1694 | goto failed_mount; |
1695 | ||
1696 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
1697 | ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); | |
1698 | ||
1699 | if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV && | |
1700 | (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) || | |
1701 | EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) || | |
1702 | EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U))) | |
4cf46b67 AF |
1703 | ext3_msg(sb, KERN_WARNING, |
1704 | "warning: feature flags set on rev 0 fs, " | |
1705 | "running e2fsck is recommended"); | |
1da177e4 LT |
1706 | /* |
1707 | * Check feature flags regardless of the revision level, since we | |
1708 | * previously didn't change the revision level when setting the flags, | |
1709 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
1710 | */ | |
1711 | features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP); | |
1712 | if (features) { | |
4cf46b67 AF |
1713 | ext3_msg(sb, KERN_ERR, |
1714 | "error: couldn't mount because of unsupported " | |
1715 | "optional features (%x)", le32_to_cpu(features)); | |
1da177e4 LT |
1716 | goto failed_mount; |
1717 | } | |
1718 | features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP); | |
1719 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
4cf46b67 AF |
1720 | ext3_msg(sb, KERN_ERR, |
1721 | "error: couldn't mount RDWR because of unsupported " | |
1722 | "optional features (%x)", le32_to_cpu(features)); | |
1da177e4 LT |
1723 | goto failed_mount; |
1724 | } | |
1725 | blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); | |
1726 | ||
1727 | if (blocksize < EXT3_MIN_BLOCK_SIZE || | |
1728 | blocksize > EXT3_MAX_BLOCK_SIZE) { | |
4cf46b67 AF |
1729 | ext3_msg(sb, KERN_ERR, |
1730 | "error: couldn't mount because of unsupported " | |
1731 | "filesystem blocksize %d", blocksize); | |
1da177e4 LT |
1732 | goto failed_mount; |
1733 | } | |
1734 | ||
e1defc4f | 1735 | hblock = bdev_logical_block_size(sb->s_bdev); |
1da177e4 LT |
1736 | if (sb->s_blocksize != blocksize) { |
1737 | /* | |
1738 | * Make sure the blocksize for the filesystem is larger | |
1739 | * than the hardware sectorsize for the machine. | |
1740 | */ | |
1741 | if (blocksize < hblock) { | |
4cf46b67 AF |
1742 | ext3_msg(sb, KERN_ERR, |
1743 | "error: fsblocksize %d too small for " | |
1744 | "hardware sectorsize %d", blocksize, hblock); | |
1da177e4 LT |
1745 | goto failed_mount; |
1746 | } | |
1747 | ||
1748 | brelse (bh); | |
0f0a89eb | 1749 | if (!sb_set_blocksize(sb, blocksize)) { |
4cf46b67 AF |
1750 | ext3_msg(sb, KERN_ERR, |
1751 | "error: bad blocksize %d", blocksize); | |
0f0a89eb TS |
1752 | goto out_fail; |
1753 | } | |
1da177e4 LT |
1754 | logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize; |
1755 | offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize; | |
1756 | bh = sb_bread(sb, logic_sb_block); | |
1757 | if (!bh) { | |
4cf46b67 AF |
1758 | ext3_msg(sb, KERN_ERR, |
1759 | "error: can't read superblock on 2nd try"); | |
1da177e4 LT |
1760 | goto failed_mount; |
1761 | } | |
1762 | es = (struct ext3_super_block *)(((char *)bh->b_data) + offset); | |
1763 | sbi->s_es = es; | |
1764 | if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) { | |
4cf46b67 AF |
1765 | ext3_msg(sb, KERN_ERR, |
1766 | "error: magic mismatch"); | |
1da177e4 LT |
1767 | goto failed_mount; |
1768 | } | |
1769 | } | |
1770 | ||
1771 | sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits); | |
1772 | ||
1773 | if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) { | |
1774 | sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE; | |
1775 | sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO; | |
1776 | } else { | |
1777 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
1778 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
1779 | if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) || | |
3fc74269 | 1780 | (!is_power_of_2(sbi->s_inode_size)) || |
1da177e4 | 1781 | (sbi->s_inode_size > blocksize)) { |
4cf46b67 AF |
1782 | ext3_msg(sb, KERN_ERR, |
1783 | "error: unsupported inode size: %d", | |
1da177e4 LT |
1784 | sbi->s_inode_size); |
1785 | goto failed_mount; | |
1786 | } | |
1787 | } | |
1788 | sbi->s_frag_size = EXT3_MIN_FRAG_SIZE << | |
1789 | le32_to_cpu(es->s_log_frag_size); | |
1790 | if (blocksize != sbi->s_frag_size) { | |
4cf46b67 AF |
1791 | ext3_msg(sb, KERN_ERR, |
1792 | "error: fragsize %lu != blocksize %u (unsupported)", | |
1da177e4 LT |
1793 | sbi->s_frag_size, blocksize); |
1794 | goto failed_mount; | |
1795 | } | |
1796 | sbi->s_frags_per_block = 1; | |
1797 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); | |
1798 | sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group); | |
1799 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); | |
b47b6f38 | 1800 | if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0) |
1da177e4 LT |
1801 | goto cantfind_ext3; |
1802 | sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb); | |
1803 | if (sbi->s_inodes_per_block == 0) | |
1804 | goto cantfind_ext3; | |
1805 | sbi->s_itb_per_group = sbi->s_inodes_per_group / | |
1806 | sbi->s_inodes_per_block; | |
1807 | sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc); | |
1808 | sbi->s_sbh = bh; | |
1809 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
f0d1b0b3 DH |
1810 | sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb)); |
1811 | sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb)); | |
1da177e4 LT |
1812 | for (i=0; i < 4; i++) |
1813 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); | |
1814 | sbi->s_def_hash_version = es->s_def_hash_version; | |
5e1f8c9e TT |
1815 | i = le32_to_cpu(es->s_flags); |
1816 | if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
1817 | sbi->s_hash_unsigned = 3; | |
1818 | else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
1819 | #ifdef __CHAR_UNSIGNED__ | |
1820 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
1821 | sbi->s_hash_unsigned = 3; | |
1822 | #else | |
1823 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
1824 | #endif | |
5e1f8c9e | 1825 | } |
1da177e4 LT |
1826 | |
1827 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
4cf46b67 AF |
1828 | ext3_msg(sb, KERN_ERR, |
1829 | "#blocks per group too big: %lu", | |
1da177e4 LT |
1830 | sbi->s_blocks_per_group); |
1831 | goto failed_mount; | |
1832 | } | |
1833 | if (sbi->s_frags_per_group > blocksize * 8) { | |
4cf46b67 AF |
1834 | ext3_msg(sb, KERN_ERR, |
1835 | "error: #fragments per group too big: %lu", | |
1da177e4 LT |
1836 | sbi->s_frags_per_group); |
1837 | goto failed_mount; | |
1838 | } | |
1839 | if (sbi->s_inodes_per_group > blocksize * 8) { | |
4cf46b67 AF |
1840 | ext3_msg(sb, KERN_ERR, |
1841 | "error: #inodes per group too big: %lu", | |
1da177e4 LT |
1842 | sbi->s_inodes_per_group); |
1843 | goto failed_mount; | |
1844 | } | |
1845 | ||
fcd5df35 MC |
1846 | if (le32_to_cpu(es->s_blocks_count) > |
1847 | (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) { | |
4cf46b67 AF |
1848 | ext3_msg(sb, KERN_ERR, |
1849 | "error: filesystem is too large to mount safely"); | |
fcd5df35 | 1850 | if (sizeof(sector_t) < 8) |
4cf46b67 AF |
1851 | ext3_msg(sb, KERN_ERR, |
1852 | "error: CONFIG_LBDAF not enabled"); | |
fcd5df35 MC |
1853 | goto failed_mount; |
1854 | } | |
1855 | ||
1da177e4 LT |
1856 | if (EXT3_BLOCKS_PER_GROUP(sb) == 0) |
1857 | goto cantfind_ext3; | |
855565e8 ES |
1858 | sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) - |
1859 | le32_to_cpu(es->s_first_data_block) - 1) | |
1860 | / EXT3_BLOCKS_PER_GROUP(sb)) + 1; | |
1da177e4 LT |
1861 | db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) / |
1862 | EXT3_DESC_PER_BLOCK(sb); | |
1863 | sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *), | |
1864 | GFP_KERNEL); | |
1865 | if (sbi->s_group_desc == NULL) { | |
4cf46b67 AF |
1866 | ext3_msg(sb, KERN_ERR, |
1867 | "error: not enough memory"); | |
1da177e4 LT |
1868 | goto failed_mount; |
1869 | } | |
1870 | ||
5df096d6 | 1871 | bgl_lock_init(sbi->s_blockgroup_lock); |
1da177e4 LT |
1872 | |
1873 | for (i = 0; i < db_count; i++) { | |
1874 | block = descriptor_loc(sb, logic_sb_block, i); | |
1875 | sbi->s_group_desc[i] = sb_bread(sb, block); | |
1876 | if (!sbi->s_group_desc[i]) { | |
4cf46b67 AF |
1877 | ext3_msg(sb, KERN_ERR, |
1878 | "error: can't read group descriptor %d", i); | |
1da177e4 LT |
1879 | db_count = i; |
1880 | goto failed_mount2; | |
1881 | } | |
1882 | } | |
1883 | if (!ext3_check_descriptors (sb)) { | |
4cf46b67 AF |
1884 | ext3_msg(sb, KERN_ERR, |
1885 | "error: group descriptors corrupted"); | |
1da177e4 LT |
1886 | goto failed_mount2; |
1887 | } | |
1888 | sbi->s_gdb_count = db_count; | |
1889 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
1890 | spin_lock_init(&sbi->s_next_gen_lock); | |
0216bfcf | 1891 | |
833f4077 PZ |
1892 | err = percpu_counter_init(&sbi->s_freeblocks_counter, |
1893 | ext3_count_free_blocks(sb)); | |
1894 | if (!err) { | |
1895 | err = percpu_counter_init(&sbi->s_freeinodes_counter, | |
1896 | ext3_count_free_inodes(sb)); | |
1897 | } | |
1898 | if (!err) { | |
1899 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
1900 | ext3_count_dirs(sb)); | |
1901 | } | |
1902 | if (err) { | |
4cf46b67 | 1903 | ext3_msg(sb, KERN_ERR, "error: insufficient memory"); |
833f4077 PZ |
1904 | goto failed_mount3; |
1905 | } | |
0216bfcf | 1906 | |
1da177e4 LT |
1907 | /* per fileystem reservation list head & lock */ |
1908 | spin_lock_init(&sbi->s_rsv_window_lock); | |
1909 | sbi->s_rsv_window_root = RB_ROOT; | |
1910 | /* Add a single, static dummy reservation to the start of the | |
1911 | * reservation window list --- it gives us a placeholder for | |
1912 | * append-at-start-of-list which makes the allocation logic | |
1913 | * _much_ simpler. */ | |
1914 | sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED; | |
1915 | sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED; | |
1916 | sbi->s_rsv_window_head.rsv_alloc_hit = 0; | |
1917 | sbi->s_rsv_window_head.rsv_goal_size = 0; | |
1918 | ext3_rsv_window_add(sb, &sbi->s_rsv_window_head); | |
1919 | ||
1920 | /* | |
1921 | * set up enough so that it can read an inode | |
1922 | */ | |
1923 | sb->s_op = &ext3_sops; | |
1924 | sb->s_export_op = &ext3_export_ops; | |
1925 | sb->s_xattr = ext3_xattr_handlers; | |
1926 | #ifdef CONFIG_QUOTA | |
1927 | sb->s_qcop = &ext3_qctl_operations; | |
1928 | sb->dq_op = &ext3_quota_operations; | |
1929 | #endif | |
1930 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ | |
b8a052d0 | 1931 | mutex_init(&sbi->s_orphan_lock); |
1da177e4 LT |
1932 | |
1933 | sb->s_root = NULL; | |
1934 | ||
1935 | needs_recovery = (es->s_last_orphan != 0 || | |
1936 | EXT3_HAS_INCOMPAT_FEATURE(sb, | |
1937 | EXT3_FEATURE_INCOMPAT_RECOVER)); | |
1938 | ||
1939 | /* | |
1940 | * The first inode we look at is the journal inode. Don't try | |
1941 | * root first: it may be modified in the journal! | |
1942 | */ | |
1943 | if (!test_opt(sb, NOLOAD) && | |
1944 | EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) { | |
71b96257 | 1945 | if (ext3_load_journal(sb, es, journal_devnum)) |
0216bfcf | 1946 | goto failed_mount3; |
1da177e4 LT |
1947 | } else if (journal_inum) { |
1948 | if (ext3_create_journal(sb, es, journal_inum)) | |
0216bfcf | 1949 | goto failed_mount3; |
1da177e4 LT |
1950 | } else { |
1951 | if (!silent) | |
4cf46b67 AF |
1952 | ext3_msg(sb, KERN_ERR, |
1953 | "error: no journal found. " | |
1954 | "mounting ext3 over ext2?"); | |
0216bfcf | 1955 | goto failed_mount3; |
1da177e4 LT |
1956 | } |
1957 | ||
1958 | /* We have now updated the journal if required, so we can | |
1959 | * validate the data journaling mode. */ | |
1960 | switch (test_opt(sb, DATA_FLAGS)) { | |
1961 | case 0: | |
1962 | /* No mode set, assume a default based on the journal | |
1963 | capabilities: ORDERED_DATA if the journal can | |
1964 | cope, else JOURNAL_DATA */ | |
1965 | if (journal_check_available_features | |
1966 | (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) | |
bbae8bcc | 1967 | set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE); |
1da177e4 LT |
1968 | else |
1969 | set_opt(sbi->s_mount_opt, JOURNAL_DATA); | |
1970 | break; | |
1971 | ||
1972 | case EXT3_MOUNT_ORDERED_DATA: | |
1973 | case EXT3_MOUNT_WRITEBACK_DATA: | |
1974 | if (!journal_check_available_features | |
1975 | (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) { | |
4cf46b67 AF |
1976 | ext3_msg(sb, KERN_ERR, |
1977 | "error: journal does not support " | |
1978 | "requested data journaling mode"); | |
0216bfcf | 1979 | goto failed_mount4; |
1da177e4 LT |
1980 | } |
1981 | default: | |
1982 | break; | |
1983 | } | |
1984 | ||
1985 | if (test_opt(sb, NOBH)) { | |
1da177e4 | 1986 | if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) { |
4cf46b67 AF |
1987 | ext3_msg(sb, KERN_WARNING, |
1988 | "warning: ignoring nobh option - " | |
1989 | "it is supported only with writeback mode"); | |
1da177e4 LT |
1990 | clear_opt(sbi->s_mount_opt, NOBH); |
1991 | } | |
1992 | } | |
1993 | /* | |
1994 | * The journal_load will have done any necessary log recovery, | |
1995 | * so we can safely mount the rest of the filesystem now. | |
1996 | */ | |
1997 | ||
473043dc DH |
1998 | root = ext3_iget(sb, EXT3_ROOT_INO); |
1999 | if (IS_ERR(root)) { | |
4cf46b67 | 2000 | ext3_msg(sb, KERN_ERR, "error: get root inode failed"); |
473043dc | 2001 | ret = PTR_ERR(root); |
0216bfcf | 2002 | goto failed_mount4; |
1da177e4 LT |
2003 | } |
2004 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
473043dc | 2005 | iput(root); |
4cf46b67 | 2006 | ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck"); |
0216bfcf | 2007 | goto failed_mount4; |
1da177e4 | 2008 | } |
473043dc DH |
2009 | sb->s_root = d_alloc_root(root); |
2010 | if (!sb->s_root) { | |
4cf46b67 | 2011 | ext3_msg(sb, KERN_ERR, "error: get root dentry failed"); |
473043dc DH |
2012 | iput(root); |
2013 | ret = -ENOMEM; | |
2014 | goto failed_mount4; | |
2015 | } | |
1da177e4 LT |
2016 | |
2017 | ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY); | |
ed505ee4 | 2018 | |
1da177e4 LT |
2019 | EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS; |
2020 | ext3_orphan_cleanup(sb, es); | |
2021 | EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS; | |
2022 | if (needs_recovery) | |
4cf46b67 | 2023 | ext3_msg(sb, KERN_INFO, "recovery complete"); |
1da177e4 | 2024 | ext3_mark_recovery_complete(sb, es); |
4cf46b67 | 2025 | ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode", |
1da177e4 LT |
2026 | test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal": |
2027 | test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered": | |
2028 | "writeback"); | |
2029 | ||
1da177e4 LT |
2030 | lock_kernel(); |
2031 | return 0; | |
2032 | ||
2033 | cantfind_ext3: | |
2034 | if (!silent) | |
4cf46b67 AF |
2035 | ext3_msg(sb, KERN_INFO, |
2036 | "error: can't find ext3 filesystem on dev %s.", | |
1da177e4 LT |
2037 | sb->s_id); |
2038 | goto failed_mount; | |
2039 | ||
0216bfcf | 2040 | failed_mount4: |
1da177e4 | 2041 | journal_destroy(sbi->s_journal); |
0216bfcf MC |
2042 | failed_mount3: |
2043 | percpu_counter_destroy(&sbi->s_freeblocks_counter); | |
2044 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
2045 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
1da177e4 LT |
2046 | failed_mount2: |
2047 | for (i = 0; i < db_count; i++) | |
2048 | brelse(sbi->s_group_desc[i]); | |
2049 | kfree(sbi->s_group_desc); | |
2050 | failed_mount: | |
2051 | #ifdef CONFIG_QUOTA | |
2052 | for (i = 0; i < MAXQUOTAS; i++) | |
2053 | kfree(sbi->s_qf_names[i]); | |
2054 | #endif | |
2055 | ext3_blkdev_remove(sbi); | |
2056 | brelse(bh); | |
2057 | out_fail: | |
2058 | sb->s_fs_info = NULL; | |
de5ce037 | 2059 | kfree(sbi->s_blockgroup_lock); |
1da177e4 LT |
2060 | kfree(sbi); |
2061 | lock_kernel(); | |
473043dc | 2062 | return ret; |
1da177e4 LT |
2063 | } |
2064 | ||
2065 | /* | |
2066 | * Setup any per-fs journal parameters now. We'll do this both on | |
2067 | * initial mount, once the journal has been initialised but before we've | |
ae6ddcc5 | 2068 | * done any recovery; and again on any subsequent remount. |
1da177e4 LT |
2069 | */ |
2070 | static void ext3_init_journal_params(struct super_block *sb, journal_t *journal) | |
2071 | { | |
2072 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
2073 | ||
2074 | if (sbi->s_commit_interval) | |
2075 | journal->j_commit_interval = sbi->s_commit_interval; | |
2076 | /* We could also set up an ext3-specific default for the commit | |
2077 | * interval here, but for now we'll just fall back to the jbd | |
2078 | * default. */ | |
2079 | ||
2080 | spin_lock(&journal->j_state_lock); | |
2081 | if (test_opt(sb, BARRIER)) | |
2082 | journal->j_flags |= JFS_BARRIER; | |
2083 | else | |
2084 | journal->j_flags &= ~JFS_BARRIER; | |
0e4fb5e2 HK |
2085 | if (test_opt(sb, DATA_ERR_ABORT)) |
2086 | journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR; | |
2087 | else | |
2088 | journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR; | |
1da177e4 LT |
2089 | spin_unlock(&journal->j_state_lock); |
2090 | } | |
2091 | ||
eee194e7 ES |
2092 | static journal_t *ext3_get_journal(struct super_block *sb, |
2093 | unsigned int journal_inum) | |
1da177e4 LT |
2094 | { |
2095 | struct inode *journal_inode; | |
2096 | journal_t *journal; | |
2097 | ||
2098 | /* First, test for the existence of a valid inode on disk. Bad | |
2099 | * things happen if we iget() an unused inode, as the subsequent | |
2100 | * iput() will try to delete it. */ | |
2101 | ||
473043dc DH |
2102 | journal_inode = ext3_iget(sb, journal_inum); |
2103 | if (IS_ERR(journal_inode)) { | |
4cf46b67 | 2104 | ext3_msg(sb, KERN_ERR, "error: no journal found"); |
1da177e4 LT |
2105 | return NULL; |
2106 | } | |
2107 | if (!journal_inode->i_nlink) { | |
2108 | make_bad_inode(journal_inode); | |
2109 | iput(journal_inode); | |
4cf46b67 | 2110 | ext3_msg(sb, KERN_ERR, "error: journal inode is deleted"); |
1da177e4 LT |
2111 | return NULL; |
2112 | } | |
2113 | ||
2114 | jbd_debug(2, "Journal inode found at %p: %Ld bytes\n", | |
2115 | journal_inode, journal_inode->i_size); | |
473043dc | 2116 | if (!S_ISREG(journal_inode->i_mode)) { |
4cf46b67 | 2117 | ext3_msg(sb, KERN_ERR, "error: invalid journal inode"); |
1da177e4 LT |
2118 | iput(journal_inode); |
2119 | return NULL; | |
2120 | } | |
2121 | ||
2122 | journal = journal_init_inode(journal_inode); | |
2123 | if (!journal) { | |
4cf46b67 | 2124 | ext3_msg(sb, KERN_ERR, "error: could not load journal inode"); |
1da177e4 LT |
2125 | iput(journal_inode); |
2126 | return NULL; | |
2127 | } | |
2128 | journal->j_private = sb; | |
2129 | ext3_init_journal_params(sb, journal); | |
2130 | return journal; | |
2131 | } | |
2132 | ||
2133 | static journal_t *ext3_get_dev_journal(struct super_block *sb, | |
2134 | dev_t j_dev) | |
2135 | { | |
2136 | struct buffer_head * bh; | |
2137 | journal_t *journal; | |
43d23f90 | 2138 | ext3_fsblk_t start; |
1c2bf374 | 2139 | ext3_fsblk_t len; |
1da177e4 | 2140 | int hblock, blocksize; |
43d23f90 | 2141 | ext3_fsblk_t sb_block; |
1da177e4 LT |
2142 | unsigned long offset; |
2143 | struct ext3_super_block * es; | |
2144 | struct block_device *bdev; | |
2145 | ||
4cf46b67 | 2146 | bdev = ext3_blkdev_get(j_dev, sb); |
1da177e4 LT |
2147 | if (bdev == NULL) |
2148 | return NULL; | |
2149 | ||
2150 | if (bd_claim(bdev, sb)) { | |
4cf46b67 AF |
2151 | ext3_msg(sb, KERN_ERR, |
2152 | "error: failed to claim external journal device"); | |
9a1c3542 | 2153 | blkdev_put(bdev, FMODE_READ|FMODE_WRITE); |
1da177e4 LT |
2154 | return NULL; |
2155 | } | |
2156 | ||
2157 | blocksize = sb->s_blocksize; | |
e1defc4f | 2158 | hblock = bdev_logical_block_size(bdev); |
1da177e4 | 2159 | if (blocksize < hblock) { |
4cf46b67 AF |
2160 | ext3_msg(sb, KERN_ERR, |
2161 | "error: blocksize too small for journal device"); | |
1da177e4 LT |
2162 | goto out_bdev; |
2163 | } | |
2164 | ||
2165 | sb_block = EXT3_MIN_BLOCK_SIZE / blocksize; | |
2166 | offset = EXT3_MIN_BLOCK_SIZE % blocksize; | |
2167 | set_blocksize(bdev, blocksize); | |
2168 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
4cf46b67 AF |
2169 | ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of " |
2170 | "external journal"); | |
1da177e4 LT |
2171 | goto out_bdev; |
2172 | } | |
2173 | ||
2174 | es = (struct ext3_super_block *) (((char *)bh->b_data) + offset); | |
2175 | if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) || | |
2176 | !(le32_to_cpu(es->s_feature_incompat) & | |
2177 | EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) { | |
4cf46b67 AF |
2178 | ext3_msg(sb, KERN_ERR, "error: external journal has " |
2179 | "bad superblock"); | |
1da177e4 LT |
2180 | brelse(bh); |
2181 | goto out_bdev; | |
2182 | } | |
2183 | ||
2184 | if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { | |
4cf46b67 | 2185 | ext3_msg(sb, KERN_ERR, "error: journal UUID does not match"); |
1da177e4 LT |
2186 | brelse(bh); |
2187 | goto out_bdev; | |
2188 | } | |
2189 | ||
2190 | len = le32_to_cpu(es->s_blocks_count); | |
2191 | start = sb_block + 1; | |
2192 | brelse(bh); /* we're done with the superblock */ | |
2193 | ||
2194 | journal = journal_init_dev(bdev, sb->s_bdev, | |
2195 | start, len, blocksize); | |
2196 | if (!journal) { | |
4cf46b67 AF |
2197 | ext3_msg(sb, KERN_ERR, |
2198 | "error: failed to create device journal"); | |
1da177e4 LT |
2199 | goto out_bdev; |
2200 | } | |
2201 | journal->j_private = sb; | |
2202 | ll_rw_block(READ, 1, &journal->j_sb_buffer); | |
2203 | wait_on_buffer(journal->j_sb_buffer); | |
2204 | if (!buffer_uptodate(journal->j_sb_buffer)) { | |
4cf46b67 | 2205 | ext3_msg(sb, KERN_ERR, "I/O error on journal device"); |
1da177e4 LT |
2206 | goto out_journal; |
2207 | } | |
2208 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
4cf46b67 AF |
2209 | ext3_msg(sb, KERN_ERR, |
2210 | "error: external journal has more than one " | |
2211 | "user (unsupported) - %d", | |
1da177e4 LT |
2212 | be32_to_cpu(journal->j_superblock->s_nr_users)); |
2213 | goto out_journal; | |
2214 | } | |
2215 | EXT3_SB(sb)->journal_bdev = bdev; | |
2216 | ext3_init_journal_params(sb, journal); | |
2217 | return journal; | |
2218 | out_journal: | |
2219 | journal_destroy(journal); | |
2220 | out_bdev: | |
2221 | ext3_blkdev_put(bdev); | |
2222 | return NULL; | |
2223 | } | |
2224 | ||
71b96257 JL |
2225 | static int ext3_load_journal(struct super_block *sb, |
2226 | struct ext3_super_block *es, | |
2227 | unsigned long journal_devnum) | |
1da177e4 LT |
2228 | { |
2229 | journal_t *journal; | |
eee194e7 | 2230 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); |
71b96257 | 2231 | dev_t journal_dev; |
1da177e4 LT |
2232 | int err = 0; |
2233 | int really_read_only; | |
2234 | ||
71b96257 JL |
2235 | if (journal_devnum && |
2236 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
4cf46b67 AF |
2237 | ext3_msg(sb, KERN_INFO, "external journal device major/minor " |
2238 | "numbers have changed"); | |
71b96257 JL |
2239 | journal_dev = new_decode_dev(journal_devnum); |
2240 | } else | |
2241 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
2242 | ||
1da177e4 LT |
2243 | really_read_only = bdev_read_only(sb->s_bdev); |
2244 | ||
2245 | /* | |
2246 | * Are we loading a blank journal or performing recovery after a | |
2247 | * crash? For recovery, we need to check in advance whether we | |
2248 | * can get read-write access to the device. | |
2249 | */ | |
2250 | ||
2251 | if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) { | |
2252 | if (sb->s_flags & MS_RDONLY) { | |
4cf46b67 AF |
2253 | ext3_msg(sb, KERN_INFO, |
2254 | "recovery required on readonly filesystem"); | |
1da177e4 | 2255 | if (really_read_only) { |
4cf46b67 AF |
2256 | ext3_msg(sb, KERN_ERR, "error: write access " |
2257 | "unavailable, cannot proceed"); | |
1da177e4 LT |
2258 | return -EROFS; |
2259 | } | |
4cf46b67 AF |
2260 | ext3_msg(sb, KERN_INFO, |
2261 | "write access will be enabled during recovery"); | |
1da177e4 LT |
2262 | } |
2263 | } | |
2264 | ||
2265 | if (journal_inum && journal_dev) { | |
4cf46b67 AF |
2266 | ext3_msg(sb, KERN_ERR, "error: filesystem has both journal " |
2267 | "and inode journals"); | |
1da177e4 LT |
2268 | return -EINVAL; |
2269 | } | |
2270 | ||
2271 | if (journal_inum) { | |
2272 | if (!(journal = ext3_get_journal(sb, journal_inum))) | |
2273 | return -EINVAL; | |
2274 | } else { | |
2275 | if (!(journal = ext3_get_dev_journal(sb, journal_dev))) | |
2276 | return -EINVAL; | |
2277 | } | |
2278 | ||
2279 | if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { | |
2280 | err = journal_update_format(journal); | |
2281 | if (err) { | |
4cf46b67 | 2282 | ext3_msg(sb, KERN_ERR, "error updating journal"); |
1da177e4 LT |
2283 | journal_destroy(journal); |
2284 | return err; | |
2285 | } | |
2286 | } | |
2287 | ||
2288 | if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) | |
2289 | err = journal_wipe(journal, !really_read_only); | |
2290 | if (!err) | |
2291 | err = journal_load(journal); | |
2292 | ||
2293 | if (err) { | |
4cf46b67 | 2294 | ext3_msg(sb, KERN_ERR, "error loading journal"); |
1da177e4 LT |
2295 | journal_destroy(journal); |
2296 | return err; | |
2297 | } | |
2298 | ||
2299 | EXT3_SB(sb)->s_journal = journal; | |
2300 | ext3_clear_journal_err(sb, es); | |
71b96257 JL |
2301 | |
2302 | if (journal_devnum && | |
2303 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
2304 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
71b96257 JL |
2305 | |
2306 | /* Make sure we flush the recovery flag to disk. */ | |
2307 | ext3_commit_super(sb, es, 1); | |
2308 | } | |
2309 | ||
1da177e4 LT |
2310 | return 0; |
2311 | } | |
2312 | ||
4cf46b67 AF |
2313 | static int ext3_create_journal(struct super_block *sb, |
2314 | struct ext3_super_block *es, | |
eee194e7 | 2315 | unsigned int journal_inum) |
1da177e4 LT |
2316 | { |
2317 | journal_t *journal; | |
952d9de1 | 2318 | int err; |
1da177e4 LT |
2319 | |
2320 | if (sb->s_flags & MS_RDONLY) { | |
4cf46b67 AF |
2321 | ext3_msg(sb, KERN_ERR, |
2322 | "error: readonly filesystem when trying to " | |
2323 | "create journal"); | |
1da177e4 LT |
2324 | return -EROFS; |
2325 | } | |
2326 | ||
952d9de1 BP |
2327 | journal = ext3_get_journal(sb, journal_inum); |
2328 | if (!journal) | |
1da177e4 LT |
2329 | return -EINVAL; |
2330 | ||
4cf46b67 | 2331 | ext3_msg(sb, KERN_INFO, "creating new journal on inode %u", |
1da177e4 LT |
2332 | journal_inum); |
2333 | ||
952d9de1 BP |
2334 | err = journal_create(journal); |
2335 | if (err) { | |
4cf46b67 | 2336 | ext3_msg(sb, KERN_ERR, "error creating journal"); |
1da177e4 LT |
2337 | journal_destroy(journal); |
2338 | return -EIO; | |
2339 | } | |
2340 | ||
2341 | EXT3_SB(sb)->s_journal = journal; | |
2342 | ||
2343 | ext3_update_dynamic_rev(sb); | |
2344 | EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
2345 | EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL); | |
2346 | ||
2347 | es->s_journal_inum = cpu_to_le32(journal_inum); | |
1da177e4 LT |
2348 | |
2349 | /* Make sure we flush the recovery flag to disk. */ | |
2350 | ext3_commit_super(sb, es, 1); | |
2351 | ||
2352 | return 0; | |
2353 | } | |
2354 | ||
c4be0c1d TS |
2355 | static int ext3_commit_super(struct super_block *sb, |
2356 | struct ext3_super_block *es, | |
1da177e4 LT |
2357 | int sync) |
2358 | { | |
2359 | struct buffer_head *sbh = EXT3_SB(sb)->s_sbh; | |
c4be0c1d | 2360 | int error = 0; |
1da177e4 LT |
2361 | |
2362 | if (!sbh) | |
c4be0c1d | 2363 | return error; |
96ec2e0a TT |
2364 | /* |
2365 | * If the file system is mounted read-only, don't update the | |
2366 | * superblock write time. This avoids updating the superblock | |
2367 | * write time when we are mounting the root file system | |
2368 | * read/only but we need to replay the journal; at that point, | |
2369 | * for people who are east of GMT and who make their clock | |
2370 | * tick in localtime for Windows bug-for-bug compatibility, | |
2371 | * the clock is set in the future, and this will cause e2fsck | |
2372 | * to complain and force a full file system check. | |
2373 | */ | |
2374 | if (!(sb->s_flags & MS_RDONLY)) | |
2375 | es->s_wtime = cpu_to_le32(get_seconds()); | |
1da177e4 LT |
2376 | es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb)); |
2377 | es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb)); | |
2378 | BUFFER_TRACE(sbh, "marking dirty"); | |
2379 | mark_buffer_dirty(sbh); | |
2380 | if (sync) | |
c4be0c1d TS |
2381 | error = sync_dirty_buffer(sbh); |
2382 | return error; | |
1da177e4 LT |
2383 | } |
2384 | ||
2385 | ||
2386 | /* | |
2387 | * Have we just finished recovery? If so, and if we are mounting (or | |
2388 | * remounting) the filesystem readonly, then we will end up with a | |
2389 | * consistent fs on disk. Record that fact. | |
2390 | */ | |
2391 | static void ext3_mark_recovery_complete(struct super_block * sb, | |
2392 | struct ext3_super_block * es) | |
2393 | { | |
2394 | journal_t *journal = EXT3_SB(sb)->s_journal; | |
2395 | ||
2396 | journal_lock_updates(journal); | |
2d7c820e HK |
2397 | if (journal_flush(journal) < 0) |
2398 | goto out; | |
2399 | ||
1da177e4 LT |
2400 | if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) && |
2401 | sb->s_flags & MS_RDONLY) { | |
2402 | EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
1da177e4 LT |
2403 | ext3_commit_super(sb, es, 1); |
2404 | } | |
2d7c820e HK |
2405 | |
2406 | out: | |
1da177e4 LT |
2407 | journal_unlock_updates(journal); |
2408 | } | |
2409 | ||
2410 | /* | |
2411 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
2412 | * has recorded an error from a previous lifetime, move that error to the | |
2413 | * main filesystem now. | |
2414 | */ | |
4cf46b67 AF |
2415 | static void ext3_clear_journal_err(struct super_block *sb, |
2416 | struct ext3_super_block *es) | |
1da177e4 LT |
2417 | { |
2418 | journal_t *journal; | |
2419 | int j_errno; | |
2420 | const char *errstr; | |
2421 | ||
2422 | journal = EXT3_SB(sb)->s_journal; | |
2423 | ||
2424 | /* | |
2425 | * Now check for any error status which may have been recorded in the | |
2426 | * journal by a prior ext3_error() or ext3_abort() | |
2427 | */ | |
2428 | ||
2429 | j_errno = journal_errno(journal); | |
2430 | if (j_errno) { | |
2431 | char nbuf[16]; | |
2432 | ||
2433 | errstr = ext3_decode_error(sb, j_errno, nbuf); | |
e05b6b52 | 2434 | ext3_warning(sb, __func__, "Filesystem error recorded " |
1da177e4 | 2435 | "from previous mount: %s", errstr); |
e05b6b52 | 2436 | ext3_warning(sb, __func__, "Marking fs in need of " |
1da177e4 LT |
2437 | "filesystem check."); |
2438 | ||
2439 | EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS; | |
2440 | es->s_state |= cpu_to_le16(EXT3_ERROR_FS); | |
2441 | ext3_commit_super (sb, es, 1); | |
2442 | ||
2443 | journal_clear_err(journal); | |
2444 | } | |
2445 | } | |
2446 | ||
2447 | /* | |
2448 | * Force the running and committing transactions to commit, | |
2449 | * and wait on the commit. | |
2450 | */ | |
2451 | int ext3_force_commit(struct super_block *sb) | |
2452 | { | |
2453 | journal_t *journal; | |
2454 | int ret; | |
2455 | ||
2456 | if (sb->s_flags & MS_RDONLY) | |
2457 | return 0; | |
2458 | ||
2459 | journal = EXT3_SB(sb)->s_journal; | |
1da177e4 LT |
2460 | ret = ext3_journal_force_commit(journal); |
2461 | return ret; | |
2462 | } | |
2463 | ||
1da177e4 LT |
2464 | static int ext3_sync_fs(struct super_block *sb, int wait) |
2465 | { | |
02ac597c | 2466 | tid_t target; |
c87591b7 | 2467 | |
02ac597c JK |
2468 | if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) { |
2469 | if (wait) | |
2470 | log_wait_commit(EXT3_SB(sb)->s_journal, target); | |
2471 | } | |
1da177e4 LT |
2472 | return 0; |
2473 | } | |
2474 | ||
2475 | /* | |
2476 | * LVM calls this function before a (read-only) snapshot is created. This | |
2477 | * gives us a chance to flush the journal completely and mark the fs clean. | |
2478 | */ | |
c4be0c1d | 2479 | static int ext3_freeze(struct super_block *sb) |
1da177e4 | 2480 | { |
c4be0c1d TS |
2481 | int error = 0; |
2482 | journal_t *journal; | |
1da177e4 LT |
2483 | |
2484 | if (!(sb->s_flags & MS_RDONLY)) { | |
c4be0c1d | 2485 | journal = EXT3_SB(sb)->s_journal; |
1da177e4 LT |
2486 | |
2487 | /* Now we set up the journal barrier. */ | |
2488 | journal_lock_updates(journal); | |
2d7c820e HK |
2489 | |
2490 | /* | |
2491 | * We don't want to clear needs_recovery flag when we failed | |
2492 | * to flush the journal. | |
2493 | */ | |
c4be0c1d TS |
2494 | error = journal_flush(journal); |
2495 | if (error < 0) | |
2496 | goto out; | |
1da177e4 LT |
2497 | |
2498 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
2499 | EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
c4be0c1d TS |
2500 | error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1); |
2501 | if (error) | |
2502 | goto out; | |
1da177e4 | 2503 | } |
c4be0c1d TS |
2504 | return 0; |
2505 | ||
2506 | out: | |
2507 | journal_unlock_updates(journal); | |
2508 | return error; | |
1da177e4 LT |
2509 | } |
2510 | ||
2511 | /* | |
2512 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
2513 | * flag here, even though the filesystem is not technically dirty yet. | |
2514 | */ | |
c4be0c1d | 2515 | static int ext3_unfreeze(struct super_block *sb) |
1da177e4 LT |
2516 | { |
2517 | if (!(sb->s_flags & MS_RDONLY)) { | |
2518 | lock_super(sb); | |
2519 | /* Reser the needs_recovery flag before the fs is unlocked. */ | |
2520 | EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); | |
2521 | ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1); | |
2522 | unlock_super(sb); | |
2523 | journal_unlock_updates(EXT3_SB(sb)->s_journal); | |
2524 | } | |
c4be0c1d | 2525 | return 0; |
1da177e4 LT |
2526 | } |
2527 | ||
2528 | static int ext3_remount (struct super_block * sb, int * flags, char * data) | |
2529 | { | |
2530 | struct ext3_super_block * es; | |
2531 | struct ext3_sb_info *sbi = EXT3_SB(sb); | |
43d23f90 | 2532 | ext3_fsblk_t n_blocks_count = 0; |
08c6a96f JK |
2533 | unsigned long old_sb_flags; |
2534 | struct ext3_mount_options old_opts; | |
2535 | int err; | |
2536 | #ifdef CONFIG_QUOTA | |
2537 | int i; | |
2538 | #endif | |
2539 | ||
337eb00a AIB |
2540 | lock_kernel(); |
2541 | ||
08c6a96f | 2542 | /* Store the original options */ |
bbd6851a | 2543 | lock_super(sb); |
08c6a96f JK |
2544 | old_sb_flags = sb->s_flags; |
2545 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
2546 | old_opts.s_resuid = sbi->s_resuid; | |
2547 | old_opts.s_resgid = sbi->s_resgid; | |
2548 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
2549 | #ifdef CONFIG_QUOTA | |
2550 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
2551 | for (i = 0; i < MAXQUOTAS; i++) | |
2552 | old_opts.s_qf_names[i] = sbi->s_qf_names[i]; | |
2553 | #endif | |
1da177e4 LT |
2554 | |
2555 | /* | |
2556 | * Allow the "check" option to be passed as a remount option. | |
2557 | */ | |
71b96257 | 2558 | if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) { |
08c6a96f JK |
2559 | err = -EINVAL; |
2560 | goto restore_opts; | |
2561 | } | |
1da177e4 LT |
2562 | |
2563 | if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) | |
e05b6b52 | 2564 | ext3_abort(sb, __func__, "Abort forced by user"); |
1da177e4 LT |
2565 | |
2566 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
2567 | ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); | |
2568 | ||
2569 | es = sbi->s_es; | |
2570 | ||
2571 | ext3_init_journal_params(sb, sbi->s_journal); | |
2572 | ||
2573 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) || | |
2574 | n_blocks_count > le32_to_cpu(es->s_blocks_count)) { | |
08c6a96f JK |
2575 | if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) { |
2576 | err = -EROFS; | |
2577 | goto restore_opts; | |
2578 | } | |
1da177e4 LT |
2579 | |
2580 | if (*flags & MS_RDONLY) { | |
2581 | /* | |
2582 | * First of all, the unconditional stuff we have to do | |
2583 | * to disable replay of the journal when we next remount | |
2584 | */ | |
2585 | sb->s_flags |= MS_RDONLY; | |
2586 | ||
2587 | /* | |
2588 | * OK, test if we are remounting a valid rw partition | |
2589 | * readonly, and if so set the rdonly flag and then | |
2590 | * mark the partition as valid again. | |
2591 | */ | |
2592 | if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) && | |
2593 | (sbi->s_mount_state & EXT3_VALID_FS)) | |
2594 | es->s_state = cpu_to_le16(sbi->s_mount_state); | |
2595 | ||
2596 | ext3_mark_recovery_complete(sb, es); | |
2597 | } else { | |
2598 | __le32 ret; | |
2599 | if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb, | |
2600 | ~EXT3_FEATURE_RO_COMPAT_SUPP))) { | |
4cf46b67 AF |
2601 | ext3_msg(sb, KERN_WARNING, |
2602 | "warning: couldn't remount RDWR " | |
2603 | "because of unsupported optional " | |
2604 | "features (%x)", le32_to_cpu(ret)); | |
08c6a96f JK |
2605 | err = -EROFS; |
2606 | goto restore_opts; | |
1da177e4 | 2607 | } |
ea9a05a1 ES |
2608 | |
2609 | /* | |
2610 | * If we have an unprocessed orphan list hanging | |
2611 | * around from a previously readonly bdev mount, | |
2612 | * require a full umount/remount for now. | |
2613 | */ | |
2614 | if (es->s_last_orphan) { | |
4cf46b67 | 2615 | ext3_msg(sb, KERN_WARNING, "warning: couldn't " |
ea9a05a1 ES |
2616 | "remount RDWR because of unprocessed " |
2617 | "orphan inode list. Please " | |
4cf46b67 | 2618 | "umount/remount instead."); |
ea9a05a1 ES |
2619 | err = -EINVAL; |
2620 | goto restore_opts; | |
2621 | } | |
2622 | ||
1da177e4 LT |
2623 | /* |
2624 | * Mounting a RDONLY partition read-write, so reread | |
2625 | * and store the current valid flag. (It may have | |
2626 | * been changed by e2fsck since we originally mounted | |
2627 | * the partition.) | |
2628 | */ | |
2629 | ext3_clear_journal_err(sb, es); | |
2630 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
a4e4de36 | 2631 | if ((err = ext3_group_extend(sb, es, n_blocks_count))) |
08c6a96f | 2632 | goto restore_opts; |
1da177e4 LT |
2633 | if (!ext3_setup_super (sb, es, 0)) |
2634 | sb->s_flags &= ~MS_RDONLY; | |
2635 | } | |
2636 | } | |
08c6a96f JK |
2637 | #ifdef CONFIG_QUOTA |
2638 | /* Release old quota file names */ | |
2639 | for (i = 0; i < MAXQUOTAS; i++) | |
2640 | if (old_opts.s_qf_names[i] && | |
2641 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
2642 | kfree(old_opts.s_qf_names[i]); | |
2643 | #endif | |
bbd6851a | 2644 | unlock_super(sb); |
337eb00a | 2645 | unlock_kernel(); |
1da177e4 | 2646 | return 0; |
08c6a96f JK |
2647 | restore_opts: |
2648 | sb->s_flags = old_sb_flags; | |
2649 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
2650 | sbi->s_resuid = old_opts.s_resuid; | |
2651 | sbi->s_resgid = old_opts.s_resgid; | |
2652 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
2653 | #ifdef CONFIG_QUOTA | |
2654 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
2655 | for (i = 0; i < MAXQUOTAS; i++) { | |
2656 | if (sbi->s_qf_names[i] && | |
2657 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
2658 | kfree(sbi->s_qf_names[i]); | |
2659 | sbi->s_qf_names[i] = old_opts.s_qf_names[i]; | |
2660 | } | |
2661 | #endif | |
bbd6851a | 2662 | unlock_super(sb); |
337eb00a | 2663 | unlock_kernel(); |
08c6a96f | 2664 | return err; |
1da177e4 LT |
2665 | } |
2666 | ||
726c3342 | 2667 | static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf) |
1da177e4 | 2668 | { |
726c3342 | 2669 | struct super_block *sb = dentry->d_sb; |
09fe316a AT |
2670 | struct ext3_sb_info *sbi = EXT3_SB(sb); |
2671 | struct ext3_super_block *es = sbi->s_es; | |
50ee0a32 | 2672 | u64 fsid; |
1da177e4 | 2673 | |
a71ce8c6 BP |
2674 | if (test_opt(sb, MINIX_DF)) { |
2675 | sbi->s_overhead_last = 0; | |
2676 | } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) { | |
2677 | unsigned long ngroups = sbi->s_groups_count, i; | |
2678 | ext3_fsblk_t overhead = 0; | |
1da177e4 LT |
2679 | smp_rmb(); |
2680 | ||
2681 | /* | |
a71ce8c6 BP |
2682 | * Compute the overhead (FS structures). This is constant |
2683 | * for a given filesystem unless the number of block groups | |
2684 | * changes so we cache the previous value until it does. | |
1da177e4 LT |
2685 | */ |
2686 | ||
2687 | /* | |
2688 | * All of the blocks before first_data_block are | |
2689 | * overhead | |
2690 | */ | |
2691 | overhead = le32_to_cpu(es->s_first_data_block); | |
2692 | ||
2693 | /* | |
2694 | * Add the overhead attributed to the superblock and | |
2695 | * block group descriptors. If the sparse superblocks | |
2696 | * feature is turned on, then not all groups have this. | |
2697 | */ | |
2698 | for (i = 0; i < ngroups; i++) { | |
2699 | overhead += ext3_bg_has_super(sb, i) + | |
2700 | ext3_bg_num_gdb(sb, i); | |
2701 | cond_resched(); | |
2702 | } | |
2703 | ||
2704 | /* | |
2705 | * Every block group has an inode bitmap, a block | |
2706 | * bitmap, and an inode table. | |
2707 | */ | |
a71ce8c6 BP |
2708 | overhead += ngroups * (2 + sbi->s_itb_per_group); |
2709 | sbi->s_overhead_last = overhead; | |
2710 | smp_wmb(); | |
2711 | sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count); | |
1da177e4 LT |
2712 | } |
2713 | ||
2714 | buf->f_type = EXT3_SUPER_MAGIC; | |
2715 | buf->f_bsize = sb->s_blocksize; | |
a71ce8c6 | 2716 | buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last; |
52d9f3b4 | 2717 | buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter); |
1da177e4 LT |
2718 | buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count); |
2719 | if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count)) | |
2720 | buf->f_bavail = 0; | |
2721 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
52d9f3b4 | 2722 | buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); |
1da177e4 | 2723 | buf->f_namelen = EXT3_NAME_LEN; |
50ee0a32 PE |
2724 | fsid = le64_to_cpup((void *)es->s_uuid) ^ |
2725 | le64_to_cpup((void *)es->s_uuid + sizeof(u64)); | |
2726 | buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; | |
2727 | buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; | |
1da177e4 LT |
2728 | return 0; |
2729 | } | |
2730 | ||
2731 | /* Helper function for writing quotas on sync - we need to start transaction before quota file | |
2732 | * is locked for write. Otherwise the are possible deadlocks: | |
2733 | * Process 1 Process 2 | |
2734 | * ext3_create() quota_sync() | |
2735 | * journal_start() write_dquot() | |
81a05227 | 2736 | * vfs_dq_init() down(dqio_mutex) |
d3be915f | 2737 | * down(dqio_mutex) journal_start() |
1da177e4 LT |
2738 | * |
2739 | */ | |
2740 | ||
2741 | #ifdef CONFIG_QUOTA | |
2742 | ||
2743 | static inline struct inode *dquot_to_inode(struct dquot *dquot) | |
2744 | { | |
2745 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]; | |
2746 | } | |
2747 | ||
1da177e4 LT |
2748 | static int ext3_write_dquot(struct dquot *dquot) |
2749 | { | |
2750 | int ret, err; | |
2751 | handle_t *handle; | |
2752 | struct inode *inode; | |
2753 | ||
2754 | inode = dquot_to_inode(dquot); | |
2755 | handle = ext3_journal_start(inode, | |
1f54587b | 2756 | EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); |
1da177e4 LT |
2757 | if (IS_ERR(handle)) |
2758 | return PTR_ERR(handle); | |
2759 | ret = dquot_commit(dquot); | |
2760 | err = ext3_journal_stop(handle); | |
2761 | if (!ret) | |
2762 | ret = err; | |
2763 | return ret; | |
2764 | } | |
2765 | ||
2766 | static int ext3_acquire_dquot(struct dquot *dquot) | |
2767 | { | |
2768 | int ret, err; | |
2769 | handle_t *handle; | |
2770 | ||
2771 | handle = ext3_journal_start(dquot_to_inode(dquot), | |
1f54587b | 2772 | EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb)); |
1da177e4 LT |
2773 | if (IS_ERR(handle)) |
2774 | return PTR_ERR(handle); | |
2775 | ret = dquot_acquire(dquot); | |
2776 | err = ext3_journal_stop(handle); | |
2777 | if (!ret) | |
2778 | ret = err; | |
2779 | return ret; | |
2780 | } | |
2781 | ||
2782 | static int ext3_release_dquot(struct dquot *dquot) | |
2783 | { | |
2784 | int ret, err; | |
2785 | handle_t *handle; | |
2786 | ||
2787 | handle = ext3_journal_start(dquot_to_inode(dquot), | |
1f54587b | 2788 | EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb)); |
9c3013e9 JK |
2789 | if (IS_ERR(handle)) { |
2790 | /* Release dquot anyway to avoid endless cycle in dqput() */ | |
2791 | dquot_release(dquot); | |
1da177e4 | 2792 | return PTR_ERR(handle); |
9c3013e9 | 2793 | } |
1da177e4 LT |
2794 | ret = dquot_release(dquot); |
2795 | err = ext3_journal_stop(handle); | |
2796 | if (!ret) | |
2797 | ret = err; | |
2798 | return ret; | |
2799 | } | |
2800 | ||
2801 | static int ext3_mark_dquot_dirty(struct dquot *dquot) | |
2802 | { | |
99aeaf63 | 2803 | /* Are we journaling quotas? */ |
1da177e4 LT |
2804 | if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || |
2805 | EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { | |
2806 | dquot_mark_dquot_dirty(dquot); | |
2807 | return ext3_write_dquot(dquot); | |
2808 | } else { | |
2809 | return dquot_mark_dquot_dirty(dquot); | |
2810 | } | |
2811 | } | |
2812 | ||
2813 | static int ext3_write_info(struct super_block *sb, int type) | |
2814 | { | |
2815 | int ret, err; | |
2816 | handle_t *handle; | |
2817 | ||
2818 | /* Data block + inode block */ | |
2819 | handle = ext3_journal_start(sb->s_root->d_inode, 2); | |
2820 | if (IS_ERR(handle)) | |
2821 | return PTR_ERR(handle); | |
2822 | ret = dquot_commit_info(sb, type); | |
2823 | err = ext3_journal_stop(handle); | |
2824 | if (!ret) | |
2825 | ret = err; | |
2826 | return ret; | |
2827 | } | |
2828 | ||
2829 | /* | |
2830 | * Turn on quotas during mount time - we need to find | |
2831 | * the quota file and such... | |
2832 | */ | |
2833 | static int ext3_quota_on_mount(struct super_block *sb, int type) | |
2834 | { | |
84de856e CH |
2835 | return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type], |
2836 | EXT3_SB(sb)->s_jquota_fmt, type); | |
1da177e4 LT |
2837 | } |
2838 | ||
2839 | /* | |
2840 | * Standard function to be called on quota_on | |
2841 | */ | |
2842 | static int ext3_quota_on(struct super_block *sb, int type, int format_id, | |
8264613d | 2843 | char *name, int remount) |
1da177e4 LT |
2844 | { |
2845 | int err; | |
8264613d | 2846 | struct path path; |
1da177e4 | 2847 | |
1f54587b JK |
2848 | if (!test_opt(sb, QUOTA)) |
2849 | return -EINVAL; | |
8264613d | 2850 | /* When remounting, no checks are needed and in fact, name is NULL */ |
9cfe7b90 | 2851 | if (remount) |
8264613d | 2852 | return vfs_quota_on(sb, type, format_id, name, remount); |
9cfe7b90 | 2853 | |
8264613d | 2854 | err = kern_path(name, LOOKUP_FOLLOW, &path); |
1da177e4 LT |
2855 | if (err) |
2856 | return err; | |
9cfe7b90 | 2857 | |
1da177e4 | 2858 | /* Quotafile not on the same filesystem? */ |
8264613d AV |
2859 | if (path.mnt->mnt_sb != sb) { |
2860 | path_put(&path); | |
1da177e4 LT |
2861 | return -EXDEV; |
2862 | } | |
9cfe7b90 JK |
2863 | /* Journaling quota? */ |
2864 | if (EXT3_SB(sb)->s_qf_names[type]) { | |
2865 | /* Quotafile not of fs root? */ | |
8264613d | 2866 | if (path.dentry->d_parent != sb->s_root) |
4cf46b67 AF |
2867 | ext3_msg(sb, KERN_WARNING, |
2868 | "warning: Quota file not on filesystem root. " | |
2869 | "Journaled quota will not work."); | |
9cfe7b90 JK |
2870 | } |
2871 | ||
2872 | /* | |
2873 | * When we journal data on quota file, we have to flush journal to see | |
2874 | * all updates to the file when we bypass pagecache... | |
2875 | */ | |
8264613d | 2876 | if (ext3_should_journal_data(path.dentry->d_inode)) { |
9cfe7b90 JK |
2877 | /* |
2878 | * We don't need to lock updates but journal_flush() could | |
2879 | * otherwise be livelocked... | |
2880 | */ | |
2881 | journal_lock_updates(EXT3_SB(sb)->s_journal); | |
2d7c820e | 2882 | err = journal_flush(EXT3_SB(sb)->s_journal); |
9cfe7b90 | 2883 | journal_unlock_updates(EXT3_SB(sb)->s_journal); |
2d7c820e | 2884 | if (err) { |
12e1ec9f | 2885 | path_put(&path); |
2d7c820e HK |
2886 | return err; |
2887 | } | |
9cfe7b90 JK |
2888 | } |
2889 | ||
8264613d AV |
2890 | err = vfs_quota_on_path(sb, type, format_id, &path); |
2891 | path_put(&path); | |
77e69dac | 2892 | return err; |
1da177e4 LT |
2893 | } |
2894 | ||
2895 | /* Read data from quotafile - avoid pagecache and such because we cannot afford | |
2896 | * acquiring the locks... As quota files are never truncated and quota code | |
2897 | * itself serializes the operations (and noone else should touch the files) | |
2898 | * we don't have to be afraid of races */ | |
2899 | static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data, | |
2900 | size_t len, loff_t off) | |
2901 | { | |
2902 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
2903 | sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb); | |
2904 | int err = 0; | |
2905 | int offset = off & (sb->s_blocksize - 1); | |
2906 | int tocopy; | |
2907 | size_t toread; | |
2908 | struct buffer_head *bh; | |
2909 | loff_t i_size = i_size_read(inode); | |
2910 | ||
2911 | if (off > i_size) | |
2912 | return 0; | |
2913 | if (off+len > i_size) | |
2914 | len = i_size-off; | |
2915 | toread = len; | |
2916 | while (toread > 0) { | |
2917 | tocopy = sb->s_blocksize - offset < toread ? | |
2918 | sb->s_blocksize - offset : toread; | |
2919 | bh = ext3_bread(NULL, inode, blk, 0, &err); | |
2920 | if (err) | |
2921 | return err; | |
2922 | if (!bh) /* A hole? */ | |
2923 | memset(data, 0, tocopy); | |
2924 | else | |
2925 | memcpy(data, bh->b_data+offset, tocopy); | |
2926 | brelse(bh); | |
2927 | offset = 0; | |
2928 | toread -= tocopy; | |
2929 | data += tocopy; | |
2930 | blk++; | |
2931 | } | |
2932 | return len; | |
2933 | } | |
2934 | ||
2935 | /* Write to quotafile (we know the transaction is already started and has | |
2936 | * enough credits) */ | |
2937 | static ssize_t ext3_quota_write(struct super_block *sb, int type, | |
2938 | const char *data, size_t len, loff_t off) | |
2939 | { | |
2940 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
2941 | sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb); | |
2942 | int err = 0; | |
2943 | int offset = off & (sb->s_blocksize - 1); | |
2944 | int tocopy; | |
2945 | int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL; | |
2946 | size_t towrite = len; | |
2947 | struct buffer_head *bh; | |
2948 | handle_t *handle = journal_current_handle(); | |
2949 | ||
9c3013e9 | 2950 | if (!handle) { |
4cf46b67 AF |
2951 | ext3_msg(sb, KERN_WARNING, |
2952 | "warning: quota write (off=%llu, len=%llu)" | |
2953 | " cancelled because transaction is not started.", | |
9c3013e9 JK |
2954 | (unsigned long long)off, (unsigned long long)len); |
2955 | return -EIO; | |
2956 | } | |
5c81a419 | 2957 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); |
1da177e4 LT |
2958 | while (towrite > 0) { |
2959 | tocopy = sb->s_blocksize - offset < towrite ? | |
2960 | sb->s_blocksize - offset : towrite; | |
2961 | bh = ext3_bread(handle, inode, blk, 1, &err); | |
2962 | if (!bh) | |
2963 | goto out; | |
2964 | if (journal_quota) { | |
2965 | err = ext3_journal_get_write_access(handle, bh); | |
2966 | if (err) { | |
2967 | brelse(bh); | |
2968 | goto out; | |
2969 | } | |
2970 | } | |
2971 | lock_buffer(bh); | |
2972 | memcpy(bh->b_data+offset, data, tocopy); | |
2973 | flush_dcache_page(bh->b_page); | |
2974 | unlock_buffer(bh); | |
2975 | if (journal_quota) | |
2976 | err = ext3_journal_dirty_metadata(handle, bh); | |
2977 | else { | |
2978 | /* Always do at least ordered writes for quotas */ | |
2979 | err = ext3_journal_dirty_data(handle, bh); | |
2980 | mark_buffer_dirty(bh); | |
2981 | } | |
2982 | brelse(bh); | |
2983 | if (err) | |
2984 | goto out; | |
2985 | offset = 0; | |
2986 | towrite -= tocopy; | |
2987 | data += tocopy; | |
2988 | blk++; | |
2989 | } | |
2990 | out: | |
f5c8f7da JK |
2991 | if (len == towrite) { |
2992 | mutex_unlock(&inode->i_mutex); | |
1da177e4 | 2993 | return err; |
f5c8f7da | 2994 | } |
1da177e4 LT |
2995 | if (inode->i_size < off+len-towrite) { |
2996 | i_size_write(inode, off+len-towrite); | |
2997 | EXT3_I(inode)->i_disksize = inode->i_size; | |
2998 | } | |
2999 | inode->i_version++; | |
3000 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
3001 | ext3_mark_inode_dirty(handle, inode); | |
1b1dcc1b | 3002 | mutex_unlock(&inode->i_mutex); |
1da177e4 LT |
3003 | return len - towrite; |
3004 | } | |
3005 | ||
3006 | #endif | |
3007 | ||
454e2398 DH |
3008 | static int ext3_get_sb(struct file_system_type *fs_type, |
3009 | int flags, const char *dev_name, void *data, struct vfsmount *mnt) | |
1da177e4 | 3010 | { |
454e2398 | 3011 | return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt); |
1da177e4 LT |
3012 | } |
3013 | ||
3014 | static struct file_system_type ext3_fs_type = { | |
3015 | .owner = THIS_MODULE, | |
3016 | .name = "ext3", | |
3017 | .get_sb = ext3_get_sb, | |
3018 | .kill_sb = kill_block_super, | |
3019 | .fs_flags = FS_REQUIRES_DEV, | |
3020 | }; | |
3021 | ||
3022 | static int __init init_ext3_fs(void) | |
3023 | { | |
3024 | int err = init_ext3_xattr(); | |
3025 | if (err) | |
3026 | return err; | |
3027 | err = init_inodecache(); | |
3028 | if (err) | |
3029 | goto out1; | |
3030 | err = register_filesystem(&ext3_fs_type); | |
3031 | if (err) | |
3032 | goto out; | |
3033 | return 0; | |
3034 | out: | |
3035 | destroy_inodecache(); | |
3036 | out1: | |
e9ad5620 | 3037 | exit_ext3_xattr(); |
1da177e4 LT |
3038 | return err; |
3039 | } | |
3040 | ||
3041 | static void __exit exit_ext3_fs(void) | |
3042 | { | |
3043 | unregister_filesystem(&ext3_fs_type); | |
3044 | destroy_inodecache(); | |
3045 | exit_ext3_xattr(); | |
3046 | } | |
3047 | ||
3048 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
3049 | MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); | |
3050 | MODULE_LICENSE("GPL"); | |
3051 | module_init(init_ext3_fs) | |
3052 | module_exit(exit_ext3_fs) |