Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/super.c |
ac27a0ec DK |
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 | ||
19 | #include <linux/module.h> | |
20 | #include <linux/string.h> | |
21 | #include <linux/fs.h> | |
22 | #include <linux/time.h> | |
dab291af | 23 | #include <linux/jbd2.h> |
617ba13b | 24 | #include <linux/ext4_fs.h> |
dab291af | 25 | #include <linux/ext4_jbd2.h> |
ac27a0ec DK |
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> | |
32 | #include <linux/vfs.h> | |
33 | #include <linux/random.h> | |
34 | #include <linux/mount.h> | |
35 | #include <linux/namei.h> | |
36 | #include <linux/quotaops.h> | |
37 | #include <linux/seq_file.h> | |
38 | ||
39 | #include <asm/uaccess.h> | |
40 | ||
41 | #include "xattr.h" | |
42 | #include "acl.h" | |
43 | #include "namei.h" | |
44 | ||
617ba13b | 45 | static int ext4_load_journal(struct super_block *, struct ext4_super_block *, |
ac27a0ec | 46 | unsigned long journal_devnum); |
617ba13b | 47 | static int ext4_create_journal(struct super_block *, struct ext4_super_block *, |
ac27a0ec | 48 | unsigned int); |
617ba13b MC |
49 | static void ext4_commit_super (struct super_block * sb, |
50 | struct ext4_super_block * es, | |
ac27a0ec | 51 | int sync); |
617ba13b MC |
52 | static void ext4_mark_recovery_complete(struct super_block * sb, |
53 | struct ext4_super_block * es); | |
54 | static void ext4_clear_journal_err(struct super_block * sb, | |
55 | struct ext4_super_block * es); | |
56 | static int ext4_sync_fs(struct super_block *sb, int wait); | |
57 | static const char *ext4_decode_error(struct super_block * sb, int errno, | |
ac27a0ec | 58 | char nbuf[16]); |
617ba13b MC |
59 | static int ext4_remount (struct super_block * sb, int * flags, char * data); |
60 | static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf); | |
61 | static void ext4_unlockfs(struct super_block *sb); | |
62 | static void ext4_write_super (struct super_block * sb); | |
63 | static void ext4_write_super_lockfs(struct super_block *sb); | |
ac27a0ec | 64 | |
bd81d8ee | 65 | |
8fadc143 AR |
66 | ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, |
67 | struct ext4_group_desc *bg) | |
bd81d8ee LV |
68 | { |
69 | return le32_to_cpu(bg->bg_block_bitmap) | | |
8fadc143 AR |
70 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
71 | (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); | |
bd81d8ee LV |
72 | } |
73 | ||
8fadc143 AR |
74 | ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, |
75 | struct ext4_group_desc *bg) | |
bd81d8ee LV |
76 | { |
77 | return le32_to_cpu(bg->bg_inode_bitmap) | | |
8fadc143 AR |
78 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
79 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); | |
bd81d8ee LV |
80 | } |
81 | ||
8fadc143 AR |
82 | ext4_fsblk_t ext4_inode_table(struct super_block *sb, |
83 | struct ext4_group_desc *bg) | |
bd81d8ee LV |
84 | { |
85 | return le32_to_cpu(bg->bg_inode_table) | | |
8fadc143 AR |
86 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
87 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); | |
bd81d8ee LV |
88 | } |
89 | ||
8fadc143 AR |
90 | void ext4_block_bitmap_set(struct super_block *sb, |
91 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee LV |
92 | { |
93 | bg->bg_block_bitmap = cpu_to_le32((u32)blk); | |
8fadc143 AR |
94 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
95 | bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
96 | } |
97 | ||
8fadc143 AR |
98 | void ext4_inode_bitmap_set(struct super_block *sb, |
99 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee LV |
100 | { |
101 | bg->bg_inode_bitmap = cpu_to_le32((u32)blk); | |
8fadc143 AR |
102 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
103 | bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
104 | } |
105 | ||
8fadc143 AR |
106 | void ext4_inode_table_set(struct super_block *sb, |
107 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee LV |
108 | { |
109 | bg->bg_inode_table = cpu_to_le32((u32)blk); | |
8fadc143 AR |
110 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
111 | bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
112 | } |
113 | ||
ac27a0ec | 114 | /* |
dab291af | 115 | * Wrappers for jbd2_journal_start/end. |
ac27a0ec DK |
116 | * |
117 | * The only special thing we need to do here is to make sure that all | |
118 | * journal_end calls result in the superblock being marked dirty, so | |
119 | * that sync() will call the filesystem's write_super callback if | |
120 | * appropriate. | |
121 | */ | |
617ba13b | 122 | handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks) |
ac27a0ec DK |
123 | { |
124 | journal_t *journal; | |
125 | ||
126 | if (sb->s_flags & MS_RDONLY) | |
127 | return ERR_PTR(-EROFS); | |
128 | ||
129 | /* Special case here: if the journal has aborted behind our | |
130 | * backs (eg. EIO in the commit thread), then we still need to | |
131 | * take the FS itself readonly cleanly. */ | |
617ba13b | 132 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 133 | if (is_journal_aborted(journal)) { |
617ba13b | 134 | ext4_abort(sb, __FUNCTION__, |
ac27a0ec DK |
135 | "Detected aborted journal"); |
136 | return ERR_PTR(-EROFS); | |
137 | } | |
138 | ||
dab291af | 139 | return jbd2_journal_start(journal, nblocks); |
ac27a0ec DK |
140 | } |
141 | ||
142 | /* | |
143 | * The only special thing we need to do here is to make sure that all | |
dab291af | 144 | * jbd2_journal_stop calls result in the superblock being marked dirty, so |
ac27a0ec DK |
145 | * that sync() will call the filesystem's write_super callback if |
146 | * appropriate. | |
147 | */ | |
617ba13b | 148 | int __ext4_journal_stop(const char *where, handle_t *handle) |
ac27a0ec DK |
149 | { |
150 | struct super_block *sb; | |
151 | int err; | |
152 | int rc; | |
153 | ||
154 | sb = handle->h_transaction->t_journal->j_private; | |
155 | err = handle->h_err; | |
dab291af | 156 | rc = jbd2_journal_stop(handle); |
ac27a0ec DK |
157 | |
158 | if (!err) | |
159 | err = rc; | |
160 | if (err) | |
617ba13b | 161 | __ext4_std_error(sb, where, err); |
ac27a0ec DK |
162 | return err; |
163 | } | |
164 | ||
617ba13b | 165 | void ext4_journal_abort_handle(const char *caller, const char *err_fn, |
ac27a0ec DK |
166 | struct buffer_head *bh, handle_t *handle, int err) |
167 | { | |
168 | char nbuf[16]; | |
617ba13b | 169 | const char *errstr = ext4_decode_error(NULL, err, nbuf); |
ac27a0ec DK |
170 | |
171 | if (bh) | |
172 | BUFFER_TRACE(bh, "abort"); | |
173 | ||
174 | if (!handle->h_err) | |
175 | handle->h_err = err; | |
176 | ||
177 | if (is_handle_aborted(handle)) | |
178 | return; | |
179 | ||
180 | printk(KERN_ERR "%s: aborting transaction: %s in %s\n", | |
181 | caller, errstr, err_fn); | |
182 | ||
dab291af | 183 | jbd2_journal_abort_handle(handle); |
ac27a0ec DK |
184 | } |
185 | ||
186 | /* Deal with the reporting of failure conditions on a filesystem such as | |
187 | * inconsistencies detected or read IO failures. | |
188 | * | |
189 | * On ext2, we can store the error state of the filesystem in the | |
617ba13b | 190 | * superblock. That is not possible on ext4, because we may have other |
ac27a0ec DK |
191 | * write ordering constraints on the superblock which prevent us from |
192 | * writing it out straight away; and given that the journal is about to | |
193 | * be aborted, we can't rely on the current, or future, transactions to | |
194 | * write out the superblock safely. | |
195 | * | |
dab291af | 196 | * We'll just use the jbd2_journal_abort() error code to record an error in |
ac27a0ec DK |
197 | * the journal instead. On recovery, the journal will compain about |
198 | * that error until we've noted it down and cleared it. | |
199 | */ | |
200 | ||
617ba13b | 201 | static void ext4_handle_error(struct super_block *sb) |
ac27a0ec | 202 | { |
617ba13b | 203 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 204 | |
617ba13b MC |
205 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
206 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
ac27a0ec DK |
207 | |
208 | if (sb->s_flags & MS_RDONLY) | |
209 | return; | |
210 | ||
211 | if (!test_opt (sb, ERRORS_CONT)) { | |
617ba13b | 212 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 213 | |
617ba13b | 214 | EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT; |
ac27a0ec | 215 | if (journal) |
dab291af | 216 | jbd2_journal_abort(journal, -EIO); |
ac27a0ec DK |
217 | } |
218 | if (test_opt (sb, ERRORS_RO)) { | |
219 | printk (KERN_CRIT "Remounting filesystem read-only\n"); | |
220 | sb->s_flags |= MS_RDONLY; | |
221 | } | |
617ba13b | 222 | ext4_commit_super(sb, es, 1); |
ac27a0ec | 223 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 224 | panic("EXT4-fs (device %s): panic forced after error\n", |
ac27a0ec DK |
225 | sb->s_id); |
226 | } | |
227 | ||
617ba13b | 228 | void ext4_error (struct super_block * sb, const char * function, |
ac27a0ec DK |
229 | const char * fmt, ...) |
230 | { | |
231 | va_list args; | |
232 | ||
233 | va_start(args, fmt); | |
617ba13b | 234 | printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function); |
ac27a0ec DK |
235 | vprintk(fmt, args); |
236 | printk("\n"); | |
237 | va_end(args); | |
238 | ||
617ba13b | 239 | ext4_handle_error(sb); |
ac27a0ec DK |
240 | } |
241 | ||
617ba13b | 242 | static const char *ext4_decode_error(struct super_block * sb, int errno, |
ac27a0ec DK |
243 | char nbuf[16]) |
244 | { | |
245 | char *errstr = NULL; | |
246 | ||
247 | switch (errno) { | |
248 | case -EIO: | |
249 | errstr = "IO failure"; | |
250 | break; | |
251 | case -ENOMEM: | |
252 | errstr = "Out of memory"; | |
253 | break; | |
254 | case -EROFS: | |
dab291af | 255 | if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT) |
ac27a0ec DK |
256 | errstr = "Journal has aborted"; |
257 | else | |
258 | errstr = "Readonly filesystem"; | |
259 | break; | |
260 | default: | |
261 | /* If the caller passed in an extra buffer for unknown | |
262 | * errors, textualise them now. Else we just return | |
263 | * NULL. */ | |
264 | if (nbuf) { | |
265 | /* Check for truncated error codes... */ | |
266 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
267 | errstr = nbuf; | |
268 | } | |
269 | break; | |
270 | } | |
271 | ||
272 | return errstr; | |
273 | } | |
274 | ||
617ba13b | 275 | /* __ext4_std_error decodes expected errors from journaling functions |
ac27a0ec DK |
276 | * automatically and invokes the appropriate error response. */ |
277 | ||
617ba13b | 278 | void __ext4_std_error (struct super_block * sb, const char * function, |
ac27a0ec DK |
279 | int errno) |
280 | { | |
281 | char nbuf[16]; | |
282 | const char *errstr; | |
283 | ||
284 | /* Special case: if the error is EROFS, and we're not already | |
285 | * inside a transaction, then there's really no point in logging | |
286 | * an error. */ | |
287 | if (errno == -EROFS && journal_current_handle() == NULL && | |
288 | (sb->s_flags & MS_RDONLY)) | |
289 | return; | |
290 | ||
617ba13b MC |
291 | errstr = ext4_decode_error(sb, errno, nbuf); |
292 | printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n", | |
ac27a0ec DK |
293 | sb->s_id, function, errstr); |
294 | ||
617ba13b | 295 | ext4_handle_error(sb); |
ac27a0ec DK |
296 | } |
297 | ||
298 | /* | |
617ba13b | 299 | * ext4_abort is a much stronger failure handler than ext4_error. The |
ac27a0ec DK |
300 | * abort function may be used to deal with unrecoverable failures such |
301 | * as journal IO errors or ENOMEM at a critical moment in log management. | |
302 | * | |
303 | * We unconditionally force the filesystem into an ABORT|READONLY state, | |
304 | * unless the error response on the fs has been set to panic in which | |
305 | * case we take the easy way out and panic immediately. | |
306 | */ | |
307 | ||
617ba13b | 308 | void ext4_abort (struct super_block * sb, const char * function, |
ac27a0ec DK |
309 | const char * fmt, ...) |
310 | { | |
311 | va_list args; | |
312 | ||
617ba13b | 313 | printk (KERN_CRIT "ext4_abort called.\n"); |
ac27a0ec DK |
314 | |
315 | va_start(args, fmt); | |
617ba13b | 316 | printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function); |
ac27a0ec DK |
317 | vprintk(fmt, args); |
318 | printk("\n"); | |
319 | va_end(args); | |
320 | ||
321 | if (test_opt(sb, ERRORS_PANIC)) | |
617ba13b | 322 | panic("EXT4-fs panic from previous error\n"); |
ac27a0ec DK |
323 | |
324 | if (sb->s_flags & MS_RDONLY) | |
325 | return; | |
326 | ||
327 | printk(KERN_CRIT "Remounting filesystem read-only\n"); | |
617ba13b | 328 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
ac27a0ec | 329 | sb->s_flags |= MS_RDONLY; |
617ba13b | 330 | EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT; |
dab291af | 331 | jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); |
ac27a0ec DK |
332 | } |
333 | ||
617ba13b | 334 | void ext4_warning (struct super_block * sb, const char * function, |
ac27a0ec DK |
335 | const char * fmt, ...) |
336 | { | |
337 | va_list args; | |
338 | ||
339 | va_start(args, fmt); | |
617ba13b | 340 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ", |
ac27a0ec DK |
341 | sb->s_id, function); |
342 | vprintk(fmt, args); | |
343 | printk("\n"); | |
344 | va_end(args); | |
345 | } | |
346 | ||
617ba13b | 347 | void ext4_update_dynamic_rev(struct super_block *sb) |
ac27a0ec | 348 | { |
617ba13b | 349 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 350 | |
617ba13b | 351 | if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) |
ac27a0ec DK |
352 | return; |
353 | ||
617ba13b | 354 | ext4_warning(sb, __FUNCTION__, |
ac27a0ec DK |
355 | "updating to rev %d because of new feature flag, " |
356 | "running e2fsck is recommended", | |
617ba13b | 357 | EXT4_DYNAMIC_REV); |
ac27a0ec | 358 | |
617ba13b MC |
359 | es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); |
360 | es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); | |
361 | es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); | |
ac27a0ec DK |
362 | /* leave es->s_feature_*compat flags alone */ |
363 | /* es->s_uuid will be set by e2fsck if empty */ | |
364 | ||
365 | /* | |
366 | * The rest of the superblock fields should be zero, and if not it | |
367 | * means they are likely already in use, so leave them alone. We | |
368 | * can leave it up to e2fsck to clean up any inconsistencies there. | |
369 | */ | |
370 | } | |
371 | ||
372 | /* | |
373 | * Open the external journal device | |
374 | */ | |
617ba13b | 375 | static struct block_device *ext4_blkdev_get(dev_t dev) |
ac27a0ec DK |
376 | { |
377 | struct block_device *bdev; | |
378 | char b[BDEVNAME_SIZE]; | |
379 | ||
380 | bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE); | |
381 | if (IS_ERR(bdev)) | |
382 | goto fail; | |
383 | return bdev; | |
384 | ||
385 | fail: | |
617ba13b | 386 | printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n", |
ac27a0ec DK |
387 | __bdevname(dev, b), PTR_ERR(bdev)); |
388 | return NULL; | |
389 | } | |
390 | ||
391 | /* | |
392 | * Release the journal device | |
393 | */ | |
617ba13b | 394 | static int ext4_blkdev_put(struct block_device *bdev) |
ac27a0ec DK |
395 | { |
396 | bd_release(bdev); | |
397 | return blkdev_put(bdev); | |
398 | } | |
399 | ||
617ba13b | 400 | static int ext4_blkdev_remove(struct ext4_sb_info *sbi) |
ac27a0ec DK |
401 | { |
402 | struct block_device *bdev; | |
403 | int ret = -ENODEV; | |
404 | ||
405 | bdev = sbi->journal_bdev; | |
406 | if (bdev) { | |
617ba13b | 407 | ret = ext4_blkdev_put(bdev); |
ac27a0ec DK |
408 | sbi->journal_bdev = NULL; |
409 | } | |
410 | return ret; | |
411 | } | |
412 | ||
413 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
414 | { | |
617ba13b | 415 | return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec DK |
416 | } |
417 | ||
617ba13b | 418 | static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) |
ac27a0ec DK |
419 | { |
420 | struct list_head *l; | |
421 | ||
422 | printk(KERN_ERR "sb orphan head is %d\n", | |
423 | le32_to_cpu(sbi->s_es->s_last_orphan)); | |
424 | ||
425 | printk(KERN_ERR "sb_info orphan list:\n"); | |
426 | list_for_each(l, &sbi->s_orphan) { | |
427 | struct inode *inode = orphan_list_entry(l); | |
428 | printk(KERN_ERR " " | |
429 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", | |
430 | inode->i_sb->s_id, inode->i_ino, inode, | |
431 | inode->i_mode, inode->i_nlink, | |
432 | NEXT_ORPHAN(inode)); | |
433 | } | |
434 | } | |
435 | ||
617ba13b | 436 | static void ext4_put_super (struct super_block * sb) |
ac27a0ec | 437 | { |
617ba13b MC |
438 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
439 | struct ext4_super_block *es = sbi->s_es; | |
ac27a0ec DK |
440 | int i; |
441 | ||
a86c6181 | 442 | ext4_ext_release(sb); |
617ba13b | 443 | ext4_xattr_put_super(sb); |
dab291af | 444 | jbd2_journal_destroy(sbi->s_journal); |
ac27a0ec | 445 | if (!(sb->s_flags & MS_RDONLY)) { |
617ba13b | 446 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
ac27a0ec DK |
447 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
448 | BUFFER_TRACE(sbi->s_sbh, "marking dirty"); | |
449 | mark_buffer_dirty(sbi->s_sbh); | |
617ba13b | 450 | ext4_commit_super(sb, es, 1); |
ac27a0ec DK |
451 | } |
452 | ||
453 | for (i = 0; i < sbi->s_gdb_count; i++) | |
454 | brelse(sbi->s_group_desc[i]); | |
455 | kfree(sbi->s_group_desc); | |
456 | percpu_counter_destroy(&sbi->s_freeblocks_counter); | |
457 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
458 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
459 | brelse(sbi->s_sbh); | |
460 | #ifdef CONFIG_QUOTA | |
461 | for (i = 0; i < MAXQUOTAS; i++) | |
462 | kfree(sbi->s_qf_names[i]); | |
463 | #endif | |
464 | ||
465 | /* Debugging code just in case the in-memory inode orphan list | |
466 | * isn't empty. The on-disk one can be non-empty if we've | |
467 | * detected an error and taken the fs readonly, but the | |
468 | * in-memory list had better be clean by this point. */ | |
469 | if (!list_empty(&sbi->s_orphan)) | |
470 | dump_orphan_list(sb, sbi); | |
471 | J_ASSERT(list_empty(&sbi->s_orphan)); | |
472 | ||
473 | invalidate_bdev(sb->s_bdev, 0); | |
474 | if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { | |
475 | /* | |
476 | * Invalidate the journal device's buffers. We don't want them | |
477 | * floating about in memory - the physical journal device may | |
478 | * hotswapped, and it breaks the `ro-after' testing code. | |
479 | */ | |
480 | sync_blockdev(sbi->journal_bdev); | |
481 | invalidate_bdev(sbi->journal_bdev, 0); | |
617ba13b | 482 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
483 | } |
484 | sb->s_fs_info = NULL; | |
485 | kfree(sbi); | |
486 | return; | |
487 | } | |
488 | ||
617ba13b | 489 | static kmem_cache_t *ext4_inode_cachep; |
ac27a0ec DK |
490 | |
491 | /* | |
492 | * Called inside transaction, so use GFP_NOFS | |
493 | */ | |
617ba13b | 494 | static struct inode *ext4_alloc_inode(struct super_block *sb) |
ac27a0ec | 495 | { |
617ba13b | 496 | struct ext4_inode_info *ei; |
ac27a0ec | 497 | |
617ba13b | 498 | ei = kmem_cache_alloc(ext4_inode_cachep, SLAB_NOFS); |
ac27a0ec DK |
499 | if (!ei) |
500 | return NULL; | |
617ba13b MC |
501 | #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL |
502 | ei->i_acl = EXT4_ACL_NOT_CACHED; | |
503 | ei->i_default_acl = EXT4_ACL_NOT_CACHED; | |
ac27a0ec DK |
504 | #endif |
505 | ei->i_block_alloc_info = NULL; | |
506 | ei->vfs_inode.i_version = 1; | |
a86c6181 | 507 | memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); |
ac27a0ec DK |
508 | return &ei->vfs_inode; |
509 | } | |
510 | ||
617ba13b | 511 | static void ext4_destroy_inode(struct inode *inode) |
ac27a0ec | 512 | { |
617ba13b | 513 | kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); |
ac27a0ec DK |
514 | } |
515 | ||
516 | static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags) | |
517 | { | |
617ba13b | 518 | struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; |
ac27a0ec DK |
519 | |
520 | if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == | |
521 | SLAB_CTOR_CONSTRUCTOR) { | |
522 | INIT_LIST_HEAD(&ei->i_orphan); | |
617ba13b | 523 | #ifdef CONFIG_EXT4DEV_FS_XATTR |
ac27a0ec DK |
524 | init_rwsem(&ei->xattr_sem); |
525 | #endif | |
526 | mutex_init(&ei->truncate_mutex); | |
527 | inode_init_once(&ei->vfs_inode); | |
528 | } | |
529 | } | |
530 | ||
531 | static int init_inodecache(void) | |
532 | { | |
617ba13b MC |
533 | ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", |
534 | sizeof(struct ext4_inode_info), | |
ac27a0ec DK |
535 | 0, (SLAB_RECLAIM_ACCOUNT| |
536 | SLAB_MEM_SPREAD), | |
537 | init_once, NULL); | |
617ba13b | 538 | if (ext4_inode_cachep == NULL) |
ac27a0ec DK |
539 | return -ENOMEM; |
540 | return 0; | |
541 | } | |
542 | ||
543 | static void destroy_inodecache(void) | |
544 | { | |
617ba13b | 545 | kmem_cache_destroy(ext4_inode_cachep); |
ac27a0ec DK |
546 | } |
547 | ||
617ba13b | 548 | static void ext4_clear_inode(struct inode *inode) |
ac27a0ec | 549 | { |
617ba13b MC |
550 | struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info; |
551 | #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
552 | if (EXT4_I(inode)->i_acl && | |
553 | EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) { | |
554 | posix_acl_release(EXT4_I(inode)->i_acl); | |
555 | EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED; | |
556 | } | |
557 | if (EXT4_I(inode)->i_default_acl && | |
558 | EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) { | |
559 | posix_acl_release(EXT4_I(inode)->i_default_acl); | |
560 | EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED; | |
ac27a0ec DK |
561 | } |
562 | #endif | |
617ba13b MC |
563 | ext4_discard_reservation(inode); |
564 | EXT4_I(inode)->i_block_alloc_info = NULL; | |
ac27a0ec DK |
565 | if (unlikely(rsv)) |
566 | kfree(rsv); | |
567 | } | |
568 | ||
617ba13b | 569 | static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb) |
ac27a0ec DK |
570 | { |
571 | #if defined(CONFIG_QUOTA) | |
617ba13b | 572 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
573 | |
574 | if (sbi->s_jquota_fmt) | |
575 | seq_printf(seq, ",jqfmt=%s", | |
576 | (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0"); | |
577 | ||
578 | if (sbi->s_qf_names[USRQUOTA]) | |
579 | seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); | |
580 | ||
581 | if (sbi->s_qf_names[GRPQUOTA]) | |
582 | seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); | |
583 | ||
617ba13b | 584 | if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) |
ac27a0ec DK |
585 | seq_puts(seq, ",usrquota"); |
586 | ||
617ba13b | 587 | if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) |
ac27a0ec DK |
588 | seq_puts(seq, ",grpquota"); |
589 | #endif | |
590 | } | |
591 | ||
617ba13b | 592 | static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs) |
ac27a0ec DK |
593 | { |
594 | struct super_block *sb = vfs->mnt_sb; | |
595 | ||
617ba13b | 596 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) |
ac27a0ec | 597 | seq_puts(seq, ",data=journal"); |
617ba13b | 598 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) |
ac27a0ec | 599 | seq_puts(seq, ",data=ordered"); |
617ba13b | 600 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) |
ac27a0ec DK |
601 | seq_puts(seq, ",data=writeback"); |
602 | ||
617ba13b | 603 | ext4_show_quota_options(seq, sb); |
ac27a0ec DK |
604 | |
605 | return 0; | |
606 | } | |
607 | ||
608 | ||
617ba13b | 609 | static struct dentry *ext4_get_dentry(struct super_block *sb, void *vobjp) |
ac27a0ec DK |
610 | { |
611 | __u32 *objp = vobjp; | |
612 | unsigned long ino = objp[0]; | |
613 | __u32 generation = objp[1]; | |
614 | struct inode *inode; | |
615 | struct dentry *result; | |
616 | ||
617ba13b | 617 | if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) |
ac27a0ec | 618 | return ERR_PTR(-ESTALE); |
617ba13b | 619 | if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) |
ac27a0ec DK |
620 | return ERR_PTR(-ESTALE); |
621 | ||
622 | /* iget isn't really right if the inode is currently unallocated!! | |
623 | * | |
617ba13b | 624 | * ext4_read_inode will return a bad_inode if the inode had been |
ac27a0ec DK |
625 | * deleted, so we should be safe. |
626 | * | |
627 | * Currently we don't know the generation for parent directory, so | |
628 | * a generation of 0 means "accept any" | |
629 | */ | |
630 | inode = iget(sb, ino); | |
631 | if (inode == NULL) | |
632 | return ERR_PTR(-ENOMEM); | |
633 | if (is_bad_inode(inode) || | |
634 | (generation && inode->i_generation != generation)) { | |
635 | iput(inode); | |
636 | return ERR_PTR(-ESTALE); | |
637 | } | |
638 | /* now to find a dentry. | |
639 | * If possible, get a well-connected one | |
640 | */ | |
641 | result = d_alloc_anon(inode); | |
642 | if (!result) { | |
643 | iput(inode); | |
644 | return ERR_PTR(-ENOMEM); | |
645 | } | |
646 | return result; | |
647 | } | |
648 | ||
649 | #ifdef CONFIG_QUOTA | |
650 | #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group") | |
651 | #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) | |
652 | ||
617ba13b MC |
653 | static int ext4_dquot_initialize(struct inode *inode, int type); |
654 | static int ext4_dquot_drop(struct inode *inode); | |
655 | static int ext4_write_dquot(struct dquot *dquot); | |
656 | static int ext4_acquire_dquot(struct dquot *dquot); | |
657 | static int ext4_release_dquot(struct dquot *dquot); | |
658 | static int ext4_mark_dquot_dirty(struct dquot *dquot); | |
659 | static int ext4_write_info(struct super_block *sb, int type); | |
660 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path); | |
661 | static int ext4_quota_on_mount(struct super_block *sb, int type); | |
662 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, | |
ac27a0ec | 663 | size_t len, loff_t off); |
617ba13b | 664 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
665 | const char *data, size_t len, loff_t off); |
666 | ||
617ba13b MC |
667 | static struct dquot_operations ext4_quota_operations = { |
668 | .initialize = ext4_dquot_initialize, | |
669 | .drop = ext4_dquot_drop, | |
ac27a0ec DK |
670 | .alloc_space = dquot_alloc_space, |
671 | .alloc_inode = dquot_alloc_inode, | |
672 | .free_space = dquot_free_space, | |
673 | .free_inode = dquot_free_inode, | |
674 | .transfer = dquot_transfer, | |
617ba13b MC |
675 | .write_dquot = ext4_write_dquot, |
676 | .acquire_dquot = ext4_acquire_dquot, | |
677 | .release_dquot = ext4_release_dquot, | |
678 | .mark_dirty = ext4_mark_dquot_dirty, | |
679 | .write_info = ext4_write_info | |
ac27a0ec DK |
680 | }; |
681 | ||
617ba13b MC |
682 | static struct quotactl_ops ext4_qctl_operations = { |
683 | .quota_on = ext4_quota_on, | |
ac27a0ec DK |
684 | .quota_off = vfs_quota_off, |
685 | .quota_sync = vfs_quota_sync, | |
686 | .get_info = vfs_get_dqinfo, | |
687 | .set_info = vfs_set_dqinfo, | |
688 | .get_dqblk = vfs_get_dqblk, | |
689 | .set_dqblk = vfs_set_dqblk | |
690 | }; | |
691 | #endif | |
692 | ||
617ba13b MC |
693 | static struct super_operations ext4_sops = { |
694 | .alloc_inode = ext4_alloc_inode, | |
695 | .destroy_inode = ext4_destroy_inode, | |
696 | .read_inode = ext4_read_inode, | |
697 | .write_inode = ext4_write_inode, | |
698 | .dirty_inode = ext4_dirty_inode, | |
699 | .delete_inode = ext4_delete_inode, | |
700 | .put_super = ext4_put_super, | |
701 | .write_super = ext4_write_super, | |
702 | .sync_fs = ext4_sync_fs, | |
703 | .write_super_lockfs = ext4_write_super_lockfs, | |
704 | .unlockfs = ext4_unlockfs, | |
705 | .statfs = ext4_statfs, | |
706 | .remount_fs = ext4_remount, | |
707 | .clear_inode = ext4_clear_inode, | |
708 | .show_options = ext4_show_options, | |
ac27a0ec | 709 | #ifdef CONFIG_QUOTA |
617ba13b MC |
710 | .quota_read = ext4_quota_read, |
711 | .quota_write = ext4_quota_write, | |
ac27a0ec DK |
712 | #endif |
713 | }; | |
714 | ||
617ba13b MC |
715 | static struct export_operations ext4_export_ops = { |
716 | .get_parent = ext4_get_parent, | |
717 | .get_dentry = ext4_get_dentry, | |
ac27a0ec DK |
718 | }; |
719 | ||
720 | enum { | |
721 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
722 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
723 | Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov, | |
724 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, | |
725 | Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh, | |
726 | Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev, | |
727 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, | |
728 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, | |
729 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota, | |
730 | Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota, | |
a86c6181 | 731 | Opt_grpquota, Opt_extents, |
ac27a0ec DK |
732 | }; |
733 | ||
734 | static match_table_t tokens = { | |
735 | {Opt_bsd_df, "bsddf"}, | |
736 | {Opt_minix_df, "minixdf"}, | |
737 | {Opt_grpid, "grpid"}, | |
738 | {Opt_grpid, "bsdgroups"}, | |
739 | {Opt_nogrpid, "nogrpid"}, | |
740 | {Opt_nogrpid, "sysvgroups"}, | |
741 | {Opt_resgid, "resgid=%u"}, | |
742 | {Opt_resuid, "resuid=%u"}, | |
743 | {Opt_sb, "sb=%u"}, | |
744 | {Opt_err_cont, "errors=continue"}, | |
745 | {Opt_err_panic, "errors=panic"}, | |
746 | {Opt_err_ro, "errors=remount-ro"}, | |
747 | {Opt_nouid32, "nouid32"}, | |
748 | {Opt_nocheck, "nocheck"}, | |
749 | {Opt_nocheck, "check=none"}, | |
750 | {Opt_debug, "debug"}, | |
751 | {Opt_oldalloc, "oldalloc"}, | |
752 | {Opt_orlov, "orlov"}, | |
753 | {Opt_user_xattr, "user_xattr"}, | |
754 | {Opt_nouser_xattr, "nouser_xattr"}, | |
755 | {Opt_acl, "acl"}, | |
756 | {Opt_noacl, "noacl"}, | |
757 | {Opt_reservation, "reservation"}, | |
758 | {Opt_noreservation, "noreservation"}, | |
759 | {Opt_noload, "noload"}, | |
760 | {Opt_nobh, "nobh"}, | |
761 | {Opt_bh, "bh"}, | |
762 | {Opt_commit, "commit=%u"}, | |
763 | {Opt_journal_update, "journal=update"}, | |
764 | {Opt_journal_inum, "journal=%u"}, | |
765 | {Opt_journal_dev, "journal_dev=%u"}, | |
766 | {Opt_abort, "abort"}, | |
767 | {Opt_data_journal, "data=journal"}, | |
768 | {Opt_data_ordered, "data=ordered"}, | |
769 | {Opt_data_writeback, "data=writeback"}, | |
770 | {Opt_offusrjquota, "usrjquota="}, | |
771 | {Opt_usrjquota, "usrjquota=%s"}, | |
772 | {Opt_offgrpjquota, "grpjquota="}, | |
773 | {Opt_grpjquota, "grpjquota=%s"}, | |
774 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
775 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
776 | {Opt_grpquota, "grpquota"}, | |
777 | {Opt_noquota, "noquota"}, | |
778 | {Opt_quota, "quota"}, | |
779 | {Opt_usrquota, "usrquota"}, | |
780 | {Opt_barrier, "barrier=%u"}, | |
a86c6181 | 781 | {Opt_extents, "extents"}, |
ac27a0ec DK |
782 | {Opt_err, NULL}, |
783 | {Opt_resize, "resize"}, | |
784 | }; | |
785 | ||
617ba13b | 786 | static ext4_fsblk_t get_sb_block(void **data) |
ac27a0ec | 787 | { |
617ba13b | 788 | ext4_fsblk_t sb_block; |
ac27a0ec DK |
789 | char *options = (char *) *data; |
790 | ||
791 | if (!options || strncmp(options, "sb=", 3) != 0) | |
792 | return 1; /* Default location */ | |
793 | options += 3; | |
617ba13b | 794 | /*todo: use simple_strtoll with >32bit ext4 */ |
ac27a0ec DK |
795 | sb_block = simple_strtoul(options, &options, 0); |
796 | if (*options && *options != ',') { | |
617ba13b | 797 | printk("EXT4-fs: Invalid sb specification: %s\n", |
ac27a0ec DK |
798 | (char *) *data); |
799 | return 1; | |
800 | } | |
801 | if (*options == ',') | |
802 | options++; | |
803 | *data = (void *) options; | |
804 | return sb_block; | |
805 | } | |
806 | ||
807 | static int parse_options (char *options, struct super_block *sb, | |
808 | unsigned int *inum, unsigned long *journal_devnum, | |
617ba13b | 809 | ext4_fsblk_t *n_blocks_count, int is_remount) |
ac27a0ec | 810 | { |
617ba13b | 811 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
812 | char * p; |
813 | substring_t args[MAX_OPT_ARGS]; | |
814 | int data_opt = 0; | |
815 | int option; | |
816 | #ifdef CONFIG_QUOTA | |
817 | int qtype; | |
818 | char *qname; | |
819 | #endif | |
820 | ||
821 | if (!options) | |
822 | return 1; | |
823 | ||
824 | while ((p = strsep (&options, ",")) != NULL) { | |
825 | int token; | |
826 | if (!*p) | |
827 | continue; | |
828 | ||
829 | token = match_token(p, tokens, args); | |
830 | switch (token) { | |
831 | case Opt_bsd_df: | |
832 | clear_opt (sbi->s_mount_opt, MINIX_DF); | |
833 | break; | |
834 | case Opt_minix_df: | |
835 | set_opt (sbi->s_mount_opt, MINIX_DF); | |
836 | break; | |
837 | case Opt_grpid: | |
838 | set_opt (sbi->s_mount_opt, GRPID); | |
839 | break; | |
840 | case Opt_nogrpid: | |
841 | clear_opt (sbi->s_mount_opt, GRPID); | |
842 | break; | |
843 | case Opt_resuid: | |
844 | if (match_int(&args[0], &option)) | |
845 | return 0; | |
846 | sbi->s_resuid = option; | |
847 | break; | |
848 | case Opt_resgid: | |
849 | if (match_int(&args[0], &option)) | |
850 | return 0; | |
851 | sbi->s_resgid = option; | |
852 | break; | |
853 | case Opt_sb: | |
854 | /* handled by get_sb_block() instead of here */ | |
855 | /* *sb_block = match_int(&args[0]); */ | |
856 | break; | |
857 | case Opt_err_panic: | |
858 | clear_opt (sbi->s_mount_opt, ERRORS_CONT); | |
859 | clear_opt (sbi->s_mount_opt, ERRORS_RO); | |
860 | set_opt (sbi->s_mount_opt, ERRORS_PANIC); | |
861 | break; | |
862 | case Opt_err_ro: | |
863 | clear_opt (sbi->s_mount_opt, ERRORS_CONT); | |
864 | clear_opt (sbi->s_mount_opt, ERRORS_PANIC); | |
865 | set_opt (sbi->s_mount_opt, ERRORS_RO); | |
866 | break; | |
867 | case Opt_err_cont: | |
868 | clear_opt (sbi->s_mount_opt, ERRORS_RO); | |
869 | clear_opt (sbi->s_mount_opt, ERRORS_PANIC); | |
870 | set_opt (sbi->s_mount_opt, ERRORS_CONT); | |
871 | break; | |
872 | case Opt_nouid32: | |
873 | set_opt (sbi->s_mount_opt, NO_UID32); | |
874 | break; | |
875 | case Opt_nocheck: | |
876 | clear_opt (sbi->s_mount_opt, CHECK); | |
877 | break; | |
878 | case Opt_debug: | |
879 | set_opt (sbi->s_mount_opt, DEBUG); | |
880 | break; | |
881 | case Opt_oldalloc: | |
882 | set_opt (sbi->s_mount_opt, OLDALLOC); | |
883 | break; | |
884 | case Opt_orlov: | |
885 | clear_opt (sbi->s_mount_opt, OLDALLOC); | |
886 | break; | |
617ba13b | 887 | #ifdef CONFIG_EXT4DEV_FS_XATTR |
ac27a0ec DK |
888 | case Opt_user_xattr: |
889 | set_opt (sbi->s_mount_opt, XATTR_USER); | |
890 | break; | |
891 | case Opt_nouser_xattr: | |
892 | clear_opt (sbi->s_mount_opt, XATTR_USER); | |
893 | break; | |
894 | #else | |
895 | case Opt_user_xattr: | |
896 | case Opt_nouser_xattr: | |
617ba13b | 897 | printk("EXT4 (no)user_xattr options not supported\n"); |
ac27a0ec DK |
898 | break; |
899 | #endif | |
617ba13b | 900 | #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL |
ac27a0ec DK |
901 | case Opt_acl: |
902 | set_opt(sbi->s_mount_opt, POSIX_ACL); | |
903 | break; | |
904 | case Opt_noacl: | |
905 | clear_opt(sbi->s_mount_opt, POSIX_ACL); | |
906 | break; | |
907 | #else | |
908 | case Opt_acl: | |
909 | case Opt_noacl: | |
617ba13b | 910 | printk("EXT4 (no)acl options not supported\n"); |
ac27a0ec DK |
911 | break; |
912 | #endif | |
913 | case Opt_reservation: | |
914 | set_opt(sbi->s_mount_opt, RESERVATION); | |
915 | break; | |
916 | case Opt_noreservation: | |
917 | clear_opt(sbi->s_mount_opt, RESERVATION); | |
918 | break; | |
919 | case Opt_journal_update: | |
920 | /* @@@ FIXME */ | |
921 | /* Eventually we will want to be able to create | |
922 | a journal file here. For now, only allow the | |
923 | user to specify an existing inode to be the | |
924 | journal file. */ | |
925 | if (is_remount) { | |
617ba13b | 926 | printk(KERN_ERR "EXT4-fs: cannot specify " |
ac27a0ec DK |
927 | "journal on remount\n"); |
928 | return 0; | |
929 | } | |
930 | set_opt (sbi->s_mount_opt, UPDATE_JOURNAL); | |
931 | break; | |
932 | case Opt_journal_inum: | |
933 | if (is_remount) { | |
617ba13b | 934 | printk(KERN_ERR "EXT4-fs: cannot specify " |
ac27a0ec DK |
935 | "journal on remount\n"); |
936 | return 0; | |
937 | } | |
938 | if (match_int(&args[0], &option)) | |
939 | return 0; | |
940 | *inum = option; | |
941 | break; | |
942 | case Opt_journal_dev: | |
943 | if (is_remount) { | |
617ba13b | 944 | printk(KERN_ERR "EXT4-fs: cannot specify " |
ac27a0ec DK |
945 | "journal on remount\n"); |
946 | return 0; | |
947 | } | |
948 | if (match_int(&args[0], &option)) | |
949 | return 0; | |
950 | *journal_devnum = option; | |
951 | break; | |
952 | case Opt_noload: | |
953 | set_opt (sbi->s_mount_opt, NOLOAD); | |
954 | break; | |
955 | case Opt_commit: | |
956 | if (match_int(&args[0], &option)) | |
957 | return 0; | |
958 | if (option < 0) | |
959 | return 0; | |
960 | if (option == 0) | |
961 | option = JBD_DEFAULT_MAX_COMMIT_AGE; | |
962 | sbi->s_commit_interval = HZ * option; | |
963 | break; | |
964 | case Opt_data_journal: | |
617ba13b | 965 | data_opt = EXT4_MOUNT_JOURNAL_DATA; |
ac27a0ec DK |
966 | goto datacheck; |
967 | case Opt_data_ordered: | |
617ba13b | 968 | data_opt = EXT4_MOUNT_ORDERED_DATA; |
ac27a0ec DK |
969 | goto datacheck; |
970 | case Opt_data_writeback: | |
617ba13b | 971 | data_opt = EXT4_MOUNT_WRITEBACK_DATA; |
ac27a0ec DK |
972 | datacheck: |
973 | if (is_remount) { | |
617ba13b | 974 | if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS) |
ac27a0ec DK |
975 | != data_opt) { |
976 | printk(KERN_ERR | |
617ba13b | 977 | "EXT4-fs: cannot change data " |
ac27a0ec DK |
978 | "mode on remount\n"); |
979 | return 0; | |
980 | } | |
981 | } else { | |
617ba13b | 982 | sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS; |
ac27a0ec DK |
983 | sbi->s_mount_opt |= data_opt; |
984 | } | |
985 | break; | |
986 | #ifdef CONFIG_QUOTA | |
987 | case Opt_usrjquota: | |
988 | qtype = USRQUOTA; | |
989 | goto set_qf_name; | |
990 | case Opt_grpjquota: | |
991 | qtype = GRPQUOTA; | |
992 | set_qf_name: | |
993 | if (sb_any_quota_enabled(sb)) { | |
994 | printk(KERN_ERR | |
617ba13b | 995 | "EXT4-fs: Cannot change journalled " |
ac27a0ec DK |
996 | "quota options when quota turned on.\n"); |
997 | return 0; | |
998 | } | |
999 | qname = match_strdup(&args[0]); | |
1000 | if (!qname) { | |
1001 | printk(KERN_ERR | |
617ba13b | 1002 | "EXT4-fs: not enough memory for " |
ac27a0ec DK |
1003 | "storing quotafile name.\n"); |
1004 | return 0; | |
1005 | } | |
1006 | if (sbi->s_qf_names[qtype] && | |
1007 | strcmp(sbi->s_qf_names[qtype], qname)) { | |
1008 | printk(KERN_ERR | |
617ba13b | 1009 | "EXT4-fs: %s quota file already " |
ac27a0ec DK |
1010 | "specified.\n", QTYPE2NAME(qtype)); |
1011 | kfree(qname); | |
1012 | return 0; | |
1013 | } | |
1014 | sbi->s_qf_names[qtype] = qname; | |
1015 | if (strchr(sbi->s_qf_names[qtype], '/')) { | |
1016 | printk(KERN_ERR | |
617ba13b | 1017 | "EXT4-fs: quotafile must be on " |
ac27a0ec DK |
1018 | "filesystem root.\n"); |
1019 | kfree(sbi->s_qf_names[qtype]); | |
1020 | sbi->s_qf_names[qtype] = NULL; | |
1021 | return 0; | |
1022 | } | |
1023 | set_opt(sbi->s_mount_opt, QUOTA); | |
1024 | break; | |
1025 | case Opt_offusrjquota: | |
1026 | qtype = USRQUOTA; | |
1027 | goto clear_qf_name; | |
1028 | case Opt_offgrpjquota: | |
1029 | qtype = GRPQUOTA; | |
1030 | clear_qf_name: | |
1031 | if (sb_any_quota_enabled(sb)) { | |
617ba13b | 1032 | printk(KERN_ERR "EXT4-fs: Cannot change " |
ac27a0ec DK |
1033 | "journalled quota options when " |
1034 | "quota turned on.\n"); | |
1035 | return 0; | |
1036 | } | |
1037 | /* | |
1038 | * The space will be released later when all options | |
1039 | * are confirmed to be correct | |
1040 | */ | |
1041 | sbi->s_qf_names[qtype] = NULL; | |
1042 | break; | |
1043 | case Opt_jqfmt_vfsold: | |
1044 | sbi->s_jquota_fmt = QFMT_VFS_OLD; | |
1045 | break; | |
1046 | case Opt_jqfmt_vfsv0: | |
1047 | sbi->s_jquota_fmt = QFMT_VFS_V0; | |
1048 | break; | |
1049 | case Opt_quota: | |
1050 | case Opt_usrquota: | |
1051 | set_opt(sbi->s_mount_opt, QUOTA); | |
1052 | set_opt(sbi->s_mount_opt, USRQUOTA); | |
1053 | break; | |
1054 | case Opt_grpquota: | |
1055 | set_opt(sbi->s_mount_opt, QUOTA); | |
1056 | set_opt(sbi->s_mount_opt, GRPQUOTA); | |
1057 | break; | |
1058 | case Opt_noquota: | |
1059 | if (sb_any_quota_enabled(sb)) { | |
617ba13b | 1060 | printk(KERN_ERR "EXT4-fs: Cannot change quota " |
ac27a0ec DK |
1061 | "options when quota turned on.\n"); |
1062 | return 0; | |
1063 | } | |
1064 | clear_opt(sbi->s_mount_opt, QUOTA); | |
1065 | clear_opt(sbi->s_mount_opt, USRQUOTA); | |
1066 | clear_opt(sbi->s_mount_opt, GRPQUOTA); | |
1067 | break; | |
1068 | #else | |
1069 | case Opt_quota: | |
1070 | case Opt_usrquota: | |
1071 | case Opt_grpquota: | |
1072 | case Opt_usrjquota: | |
1073 | case Opt_grpjquota: | |
1074 | case Opt_offusrjquota: | |
1075 | case Opt_offgrpjquota: | |
1076 | case Opt_jqfmt_vfsold: | |
1077 | case Opt_jqfmt_vfsv0: | |
1078 | printk(KERN_ERR | |
617ba13b | 1079 | "EXT4-fs: journalled quota options not " |
ac27a0ec DK |
1080 | "supported.\n"); |
1081 | break; | |
1082 | case Opt_noquota: | |
1083 | break; | |
1084 | #endif | |
1085 | case Opt_abort: | |
1086 | set_opt(sbi->s_mount_opt, ABORT); | |
1087 | break; | |
1088 | case Opt_barrier: | |
1089 | if (match_int(&args[0], &option)) | |
1090 | return 0; | |
1091 | if (option) | |
1092 | set_opt(sbi->s_mount_opt, BARRIER); | |
1093 | else | |
1094 | clear_opt(sbi->s_mount_opt, BARRIER); | |
1095 | break; | |
1096 | case Opt_ignore: | |
1097 | break; | |
1098 | case Opt_resize: | |
1099 | if (!is_remount) { | |
617ba13b | 1100 | printk("EXT4-fs: resize option only available " |
ac27a0ec DK |
1101 | "for remount\n"); |
1102 | return 0; | |
1103 | } | |
1104 | if (match_int(&args[0], &option) != 0) | |
1105 | return 0; | |
1106 | *n_blocks_count = option; | |
1107 | break; | |
1108 | case Opt_nobh: | |
1109 | set_opt(sbi->s_mount_opt, NOBH); | |
1110 | break; | |
1111 | case Opt_bh: | |
1112 | clear_opt(sbi->s_mount_opt, NOBH); | |
1113 | break; | |
a86c6181 AT |
1114 | case Opt_extents: |
1115 | set_opt (sbi->s_mount_opt, EXTENTS); | |
1116 | break; | |
ac27a0ec DK |
1117 | default: |
1118 | printk (KERN_ERR | |
617ba13b | 1119 | "EXT4-fs: Unrecognized mount option \"%s\" " |
ac27a0ec DK |
1120 | "or missing value\n", p); |
1121 | return 0; | |
1122 | } | |
1123 | } | |
1124 | #ifdef CONFIG_QUOTA | |
1125 | if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { | |
617ba13b | 1126 | if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) && |
ac27a0ec DK |
1127 | sbi->s_qf_names[USRQUOTA]) |
1128 | clear_opt(sbi->s_mount_opt, USRQUOTA); | |
1129 | ||
617ba13b | 1130 | if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) && |
ac27a0ec DK |
1131 | sbi->s_qf_names[GRPQUOTA]) |
1132 | clear_opt(sbi->s_mount_opt, GRPQUOTA); | |
1133 | ||
1134 | if ((sbi->s_qf_names[USRQUOTA] && | |
617ba13b | 1135 | (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) || |
ac27a0ec | 1136 | (sbi->s_qf_names[GRPQUOTA] && |
617ba13b MC |
1137 | (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) { |
1138 | printk(KERN_ERR "EXT4-fs: old and new quota " | |
ac27a0ec DK |
1139 | "format mixing.\n"); |
1140 | return 0; | |
1141 | } | |
1142 | ||
1143 | if (!sbi->s_jquota_fmt) { | |
617ba13b | 1144 | printk(KERN_ERR "EXT4-fs: journalled quota format " |
ac27a0ec DK |
1145 | "not specified.\n"); |
1146 | return 0; | |
1147 | } | |
1148 | } else { | |
1149 | if (sbi->s_jquota_fmt) { | |
617ba13b | 1150 | printk(KERN_ERR "EXT4-fs: journalled quota format " |
ac27a0ec DK |
1151 | "specified with no journalling " |
1152 | "enabled.\n"); | |
1153 | return 0; | |
1154 | } | |
1155 | } | |
1156 | #endif | |
1157 | return 1; | |
1158 | } | |
1159 | ||
617ba13b | 1160 | static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, |
ac27a0ec DK |
1161 | int read_only) |
1162 | { | |
617ba13b | 1163 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
1164 | int res = 0; |
1165 | ||
617ba13b MC |
1166 | if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { |
1167 | printk (KERN_ERR "EXT4-fs warning: revision level too high, " | |
ac27a0ec DK |
1168 | "forcing read-only mode\n"); |
1169 | res = MS_RDONLY; | |
1170 | } | |
1171 | if (read_only) | |
1172 | return res; | |
617ba13b MC |
1173 | if (!(sbi->s_mount_state & EXT4_VALID_FS)) |
1174 | printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, " | |
ac27a0ec | 1175 | "running e2fsck is recommended\n"); |
617ba13b | 1176 | else if ((sbi->s_mount_state & EXT4_ERROR_FS)) |
ac27a0ec | 1177 | printk (KERN_WARNING |
617ba13b | 1178 | "EXT4-fs warning: mounting fs with errors, " |
ac27a0ec DK |
1179 | "running e2fsck is recommended\n"); |
1180 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 && | |
1181 | le16_to_cpu(es->s_mnt_count) >= | |
1182 | (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) | |
1183 | printk (KERN_WARNING | |
617ba13b | 1184 | "EXT4-fs warning: maximal mount count reached, " |
ac27a0ec DK |
1185 | "running e2fsck is recommended\n"); |
1186 | else if (le32_to_cpu(es->s_checkinterval) && | |
1187 | (le32_to_cpu(es->s_lastcheck) + | |
1188 | le32_to_cpu(es->s_checkinterval) <= get_seconds())) | |
1189 | printk (KERN_WARNING | |
617ba13b | 1190 | "EXT4-fs warning: checktime reached, " |
ac27a0ec DK |
1191 | "running e2fsck is recommended\n"); |
1192 | #if 0 | |
1193 | /* @@@ We _will_ want to clear the valid bit if we find | |
1194 | inconsistencies, to force a fsck at reboot. But for | |
1195 | a plain journaled filesystem we can keep it set as | |
1196 | valid forever! :) */ | |
617ba13b | 1197 | es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS); |
ac27a0ec DK |
1198 | #endif |
1199 | if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) | |
617ba13b | 1200 | es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); |
ac27a0ec DK |
1201 | es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1); |
1202 | es->s_mtime = cpu_to_le32(get_seconds()); | |
617ba13b MC |
1203 | ext4_update_dynamic_rev(sb); |
1204 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
ac27a0ec | 1205 | |
617ba13b | 1206 | ext4_commit_super(sb, es, 1); |
ac27a0ec | 1207 | if (test_opt(sb, DEBUG)) |
617ba13b | 1208 | printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, " |
ac27a0ec DK |
1209 | "bpg=%lu, ipg=%lu, mo=%04lx]\n", |
1210 | sb->s_blocksize, | |
1211 | sbi->s_groups_count, | |
617ba13b MC |
1212 | EXT4_BLOCKS_PER_GROUP(sb), |
1213 | EXT4_INODES_PER_GROUP(sb), | |
ac27a0ec DK |
1214 | sbi->s_mount_opt); |
1215 | ||
617ba13b MC |
1216 | printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id); |
1217 | if (EXT4_SB(sb)->s_journal->j_inode == NULL) { | |
ac27a0ec DK |
1218 | char b[BDEVNAME_SIZE]; |
1219 | ||
1220 | printk("external journal on %s\n", | |
617ba13b | 1221 | bdevname(EXT4_SB(sb)->s_journal->j_dev, b)); |
ac27a0ec DK |
1222 | } else { |
1223 | printk("internal journal\n"); | |
1224 | } | |
1225 | return res; | |
1226 | } | |
1227 | ||
1228 | /* Called at mount-time, super-block is locked */ | |
617ba13b | 1229 | static int ext4_check_descriptors (struct super_block * sb) |
ac27a0ec | 1230 | { |
617ba13b MC |
1231 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
1232 | ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); | |
1233 | ext4_fsblk_t last_block; | |
bd81d8ee LV |
1234 | ext4_fsblk_t block_bitmap; |
1235 | ext4_fsblk_t inode_bitmap; | |
1236 | ext4_fsblk_t inode_table; | |
617ba13b | 1237 | struct ext4_group_desc * gdp = NULL; |
ac27a0ec DK |
1238 | int desc_block = 0; |
1239 | int i; | |
1240 | ||
617ba13b | 1241 | ext4_debug ("Checking group descriptors"); |
ac27a0ec DK |
1242 | |
1243 | for (i = 0; i < sbi->s_groups_count; i++) | |
1244 | { | |
1245 | if (i == sbi->s_groups_count - 1) | |
bd81d8ee | 1246 | last_block = ext4_blocks_count(sbi->s_es) - 1; |
ac27a0ec DK |
1247 | else |
1248 | last_block = first_block + | |
617ba13b | 1249 | (EXT4_BLOCKS_PER_GROUP(sb) - 1); |
ac27a0ec | 1250 | |
617ba13b MC |
1251 | if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0) |
1252 | gdp = (struct ext4_group_desc *) | |
ac27a0ec | 1253 | sbi->s_group_desc[desc_block++]->b_data; |
8fadc143 | 1254 | block_bitmap = ext4_block_bitmap(sb, gdp); |
bd81d8ee | 1255 | if (block_bitmap < first_block || block_bitmap > last_block) |
ac27a0ec | 1256 | { |
617ba13b | 1257 | ext4_error (sb, "ext4_check_descriptors", |
ac27a0ec | 1258 | "Block bitmap for group %d" |
2ae02107 | 1259 | " not in group (block %llu)!", |
bd81d8ee | 1260 | i, block_bitmap); |
ac27a0ec DK |
1261 | return 0; |
1262 | } | |
8fadc143 | 1263 | inode_bitmap = ext4_inode_bitmap(sb, gdp); |
bd81d8ee | 1264 | if (inode_bitmap < first_block || inode_bitmap > last_block) |
ac27a0ec | 1265 | { |
617ba13b | 1266 | ext4_error (sb, "ext4_check_descriptors", |
ac27a0ec | 1267 | "Inode bitmap for group %d" |
2ae02107 | 1268 | " not in group (block %llu)!", |
bd81d8ee | 1269 | i, inode_bitmap); |
ac27a0ec DK |
1270 | return 0; |
1271 | } | |
8fadc143 | 1272 | inode_table = ext4_inode_table(sb, gdp); |
bd81d8ee LV |
1273 | if (inode_table < first_block || |
1274 | inode_table + sbi->s_itb_per_group > last_block) | |
ac27a0ec | 1275 | { |
617ba13b | 1276 | ext4_error (sb, "ext4_check_descriptors", |
ac27a0ec | 1277 | "Inode table for group %d" |
2ae02107 | 1278 | " not in group (block %llu)!", |
bd81d8ee | 1279 | i, inode_table); |
ac27a0ec DK |
1280 | return 0; |
1281 | } | |
617ba13b | 1282 | first_block += EXT4_BLOCKS_PER_GROUP(sb); |
0d1ee42f AR |
1283 | gdp = (struct ext4_group_desc *) |
1284 | ((__u8 *)gdp + EXT4_DESC_SIZE(sb)); | |
ac27a0ec DK |
1285 | } |
1286 | ||
bd81d8ee | 1287 | ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb)); |
617ba13b | 1288 | sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb)); |
ac27a0ec DK |
1289 | return 1; |
1290 | } | |
1291 | ||
1292 | ||
617ba13b | 1293 | /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at |
ac27a0ec DK |
1294 | * the superblock) which were deleted from all directories, but held open by |
1295 | * a process at the time of a crash. We walk the list and try to delete these | |
1296 | * inodes at recovery time (only with a read-write filesystem). | |
1297 | * | |
1298 | * In order to keep the orphan inode chain consistent during traversal (in | |
1299 | * case of crash during recovery), we link each inode into the superblock | |
1300 | * orphan list_head and handle it the same way as an inode deletion during | |
1301 | * normal operation (which journals the operations for us). | |
1302 | * | |
1303 | * We only do an iget() and an iput() on each inode, which is very safe if we | |
1304 | * accidentally point at an in-use or already deleted inode. The worst that | |
1305 | * can happen in this case is that we get a "bit already cleared" message from | |
617ba13b | 1306 | * ext4_free_inode(). The only reason we would point at a wrong inode is if |
ac27a0ec DK |
1307 | * e2fsck was run on this filesystem, and it must have already done the orphan |
1308 | * inode cleanup for us, so we can safely abort without any further action. | |
1309 | */ | |
617ba13b MC |
1310 | static void ext4_orphan_cleanup (struct super_block * sb, |
1311 | struct ext4_super_block * es) | |
ac27a0ec DK |
1312 | { |
1313 | unsigned int s_flags = sb->s_flags; | |
1314 | int nr_orphans = 0, nr_truncates = 0; | |
1315 | #ifdef CONFIG_QUOTA | |
1316 | int i; | |
1317 | #endif | |
1318 | if (!es->s_last_orphan) { | |
1319 | jbd_debug(4, "no orphan inodes to clean up\n"); | |
1320 | return; | |
1321 | } | |
1322 | ||
617ba13b | 1323 | if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { |
ac27a0ec DK |
1324 | if (es->s_last_orphan) |
1325 | jbd_debug(1, "Errors on filesystem, " | |
1326 | "clearing orphan list.\n"); | |
1327 | es->s_last_orphan = 0; | |
1328 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); | |
1329 | return; | |
1330 | } | |
1331 | ||
1332 | if (s_flags & MS_RDONLY) { | |
617ba13b | 1333 | printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n", |
ac27a0ec DK |
1334 | sb->s_id); |
1335 | sb->s_flags &= ~MS_RDONLY; | |
1336 | } | |
1337 | #ifdef CONFIG_QUOTA | |
1338 | /* Needed for iput() to work correctly and not trash data */ | |
1339 | sb->s_flags |= MS_ACTIVE; | |
1340 | /* Turn on quotas so that they are updated correctly */ | |
1341 | for (i = 0; i < MAXQUOTAS; i++) { | |
617ba13b MC |
1342 | if (EXT4_SB(sb)->s_qf_names[i]) { |
1343 | int ret = ext4_quota_on_mount(sb, i); | |
ac27a0ec DK |
1344 | if (ret < 0) |
1345 | printk(KERN_ERR | |
617ba13b | 1346 | "EXT4-fs: Cannot turn on journalled " |
ac27a0ec DK |
1347 | "quota: error %d\n", ret); |
1348 | } | |
1349 | } | |
1350 | #endif | |
1351 | ||
1352 | while (es->s_last_orphan) { | |
1353 | struct inode *inode; | |
1354 | ||
1355 | if (!(inode = | |
617ba13b | 1356 | ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) { |
ac27a0ec DK |
1357 | es->s_last_orphan = 0; |
1358 | break; | |
1359 | } | |
1360 | ||
617ba13b | 1361 | list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); |
ac27a0ec DK |
1362 | DQUOT_INIT(inode); |
1363 | if (inode->i_nlink) { | |
1364 | printk(KERN_DEBUG | |
1365 | "%s: truncating inode %lu to %Ld bytes\n", | |
1366 | __FUNCTION__, inode->i_ino, inode->i_size); | |
1367 | jbd_debug(2, "truncating inode %lu to %Ld bytes\n", | |
1368 | inode->i_ino, inode->i_size); | |
617ba13b | 1369 | ext4_truncate(inode); |
ac27a0ec DK |
1370 | nr_truncates++; |
1371 | } else { | |
1372 | printk(KERN_DEBUG | |
1373 | "%s: deleting unreferenced inode %lu\n", | |
1374 | __FUNCTION__, inode->i_ino); | |
1375 | jbd_debug(2, "deleting unreferenced inode %lu\n", | |
1376 | inode->i_ino); | |
1377 | nr_orphans++; | |
1378 | } | |
1379 | iput(inode); /* The delete magic happens here! */ | |
1380 | } | |
1381 | ||
1382 | #define PLURAL(x) (x), ((x)==1) ? "" : "s" | |
1383 | ||
1384 | if (nr_orphans) | |
617ba13b | 1385 | printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n", |
ac27a0ec DK |
1386 | sb->s_id, PLURAL(nr_orphans)); |
1387 | if (nr_truncates) | |
617ba13b | 1388 | printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n", |
ac27a0ec DK |
1389 | sb->s_id, PLURAL(nr_truncates)); |
1390 | #ifdef CONFIG_QUOTA | |
1391 | /* Turn quotas off */ | |
1392 | for (i = 0; i < MAXQUOTAS; i++) { | |
1393 | if (sb_dqopt(sb)->files[i]) | |
1394 | vfs_quota_off(sb, i); | |
1395 | } | |
1396 | #endif | |
1397 | sb->s_flags = s_flags; /* Restore MS_RDONLY status */ | |
1398 | } | |
1399 | ||
1400 | #define log2(n) ffz(~(n)) | |
1401 | ||
1402 | /* | |
1403 | * Maximal file size. There is a direct, and {,double-,triple-}indirect | |
1404 | * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks. | |
1405 | * We need to be 1 filesystem block less than the 2^32 sector limit. | |
1406 | */ | |
617ba13b | 1407 | static loff_t ext4_max_size(int bits) |
ac27a0ec | 1408 | { |
617ba13b | 1409 | loff_t res = EXT4_NDIR_BLOCKS; |
ac27a0ec DK |
1410 | /* This constant is calculated to be the largest file size for a |
1411 | * dense, 4k-blocksize file such that the total number of | |
1412 | * sectors in the file, including data and all indirect blocks, | |
1413 | * does not exceed 2^32. */ | |
1414 | const loff_t upper_limit = 0x1ff7fffd000LL; | |
1415 | ||
1416 | res += 1LL << (bits-2); | |
1417 | res += 1LL << (2*(bits-2)); | |
1418 | res += 1LL << (3*(bits-2)); | |
1419 | res <<= bits; | |
1420 | if (res > upper_limit) | |
1421 | res = upper_limit; | |
1422 | return res; | |
1423 | } | |
1424 | ||
617ba13b MC |
1425 | static ext4_fsblk_t descriptor_loc(struct super_block *sb, |
1426 | ext4_fsblk_t logic_sb_block, | |
ac27a0ec DK |
1427 | int nr) |
1428 | { | |
617ba13b | 1429 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
1430 | unsigned long bg, first_meta_bg; |
1431 | int has_super = 0; | |
1432 | ||
1433 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); | |
1434 | ||
617ba13b | 1435 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec DK |
1436 | nr < first_meta_bg) |
1437 | return (logic_sb_block + nr + 1); | |
1438 | bg = sbi->s_desc_per_block * nr; | |
617ba13b | 1439 | if (ext4_bg_has_super(sb, bg)) |
ac27a0ec | 1440 | has_super = 1; |
617ba13b | 1441 | return (has_super + ext4_group_first_block_no(sb, bg)); |
ac27a0ec DK |
1442 | } |
1443 | ||
1444 | ||
617ba13b | 1445 | static int ext4_fill_super (struct super_block *sb, void *data, int silent) |
ac27a0ec DK |
1446 | { |
1447 | struct buffer_head * bh; | |
617ba13b MC |
1448 | struct ext4_super_block *es = NULL; |
1449 | struct ext4_sb_info *sbi; | |
1450 | ext4_fsblk_t block; | |
1451 | ext4_fsblk_t sb_block = get_sb_block(&data); | |
1452 | ext4_fsblk_t logic_sb_block; | |
ac27a0ec DK |
1453 | unsigned long offset = 0; |
1454 | unsigned int journal_inum = 0; | |
1455 | unsigned long journal_devnum = 0; | |
1456 | unsigned long def_mount_opts; | |
1457 | struct inode *root; | |
1458 | int blocksize; | |
1459 | int hblock; | |
1460 | int db_count; | |
1461 | int i; | |
1462 | int needs_recovery; | |
1463 | __le32 features; | |
bd81d8ee | 1464 | __u64 blocks_count; |
ac27a0ec DK |
1465 | |
1466 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
1467 | if (!sbi) | |
1468 | return -ENOMEM; | |
1469 | sb->s_fs_info = sbi; | |
1470 | sbi->s_mount_opt = 0; | |
617ba13b MC |
1471 | sbi->s_resuid = EXT4_DEF_RESUID; |
1472 | sbi->s_resgid = EXT4_DEF_RESGID; | |
ac27a0ec DK |
1473 | |
1474 | unlock_kernel(); | |
1475 | ||
617ba13b | 1476 | blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); |
ac27a0ec | 1477 | if (!blocksize) { |
617ba13b | 1478 | printk(KERN_ERR "EXT4-fs: unable to set blocksize\n"); |
ac27a0ec DK |
1479 | goto out_fail; |
1480 | } | |
1481 | ||
1482 | /* | |
617ba13b | 1483 | * The ext4 superblock will not be buffer aligned for other than 1kB |
ac27a0ec DK |
1484 | * block sizes. We need to calculate the offset from buffer start. |
1485 | */ | |
617ba13b | 1486 | if (blocksize != EXT4_MIN_BLOCK_SIZE) { |
3a5b2ecd MC |
1487 | logic_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
1488 | offset = sector_div(logic_sb_block, blocksize); | |
ac27a0ec DK |
1489 | } else { |
1490 | logic_sb_block = sb_block; | |
1491 | } | |
1492 | ||
1493 | if (!(bh = sb_bread(sb, logic_sb_block))) { | |
617ba13b | 1494 | printk (KERN_ERR "EXT4-fs: unable to read superblock\n"); |
ac27a0ec DK |
1495 | goto out_fail; |
1496 | } | |
1497 | /* | |
1498 | * Note: s_es must be initialized as soon as possible because | |
617ba13b | 1499 | * some ext4 macro-instructions depend on its value |
ac27a0ec | 1500 | */ |
617ba13b | 1501 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
ac27a0ec DK |
1502 | sbi->s_es = es; |
1503 | sb->s_magic = le16_to_cpu(es->s_magic); | |
617ba13b MC |
1504 | if (sb->s_magic != EXT4_SUPER_MAGIC) |
1505 | goto cantfind_ext4; | |
ac27a0ec DK |
1506 | |
1507 | /* Set defaults before we parse the mount options */ | |
1508 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
617ba13b | 1509 | if (def_mount_opts & EXT4_DEFM_DEBUG) |
ac27a0ec | 1510 | set_opt(sbi->s_mount_opt, DEBUG); |
617ba13b | 1511 | if (def_mount_opts & EXT4_DEFM_BSDGROUPS) |
ac27a0ec | 1512 | set_opt(sbi->s_mount_opt, GRPID); |
617ba13b | 1513 | if (def_mount_opts & EXT4_DEFM_UID16) |
ac27a0ec | 1514 | set_opt(sbi->s_mount_opt, NO_UID32); |
617ba13b | 1515 | if (def_mount_opts & EXT4_DEFM_XATTR_USER) |
ac27a0ec | 1516 | set_opt(sbi->s_mount_opt, XATTR_USER); |
617ba13b | 1517 | if (def_mount_opts & EXT4_DEFM_ACL) |
ac27a0ec | 1518 | set_opt(sbi->s_mount_opt, POSIX_ACL); |
617ba13b MC |
1519 | if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) |
1520 | sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; | |
1521 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) | |
1522 | sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA; | |
1523 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) | |
1524 | sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA; | |
1525 | ||
1526 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) | |
ac27a0ec | 1527 | set_opt(sbi->s_mount_opt, ERRORS_PANIC); |
617ba13b | 1528 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO) |
ac27a0ec DK |
1529 | set_opt(sbi->s_mount_opt, ERRORS_RO); |
1530 | ||
1531 | sbi->s_resuid = le16_to_cpu(es->s_def_resuid); | |
1532 | sbi->s_resgid = le16_to_cpu(es->s_def_resgid); | |
1533 | ||
1534 | set_opt(sbi->s_mount_opt, RESERVATION); | |
1535 | ||
1536 | if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum, | |
1537 | NULL, 0)) | |
1538 | goto failed_mount; | |
1539 | ||
1540 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
617ba13b | 1541 | ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec | 1542 | |
617ba13b MC |
1543 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && |
1544 | (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || | |
1545 | EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || | |
1546 | EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) | |
ac27a0ec | 1547 | printk(KERN_WARNING |
617ba13b | 1548 | "EXT4-fs warning: feature flags set on rev 0 fs, " |
ac27a0ec DK |
1549 | "running e2fsck is recommended\n"); |
1550 | /* | |
1551 | * Check feature flags regardless of the revision level, since we | |
1552 | * previously didn't change the revision level when setting the flags, | |
1553 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
1554 | */ | |
617ba13b | 1555 | features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP); |
ac27a0ec | 1556 | if (features) { |
617ba13b | 1557 | printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of " |
ac27a0ec DK |
1558 | "unsupported optional features (%x).\n", |
1559 | sb->s_id, le32_to_cpu(features)); | |
1560 | goto failed_mount; | |
1561 | } | |
617ba13b | 1562 | features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP); |
ac27a0ec | 1563 | if (!(sb->s_flags & MS_RDONLY) && features) { |
617ba13b | 1564 | printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of " |
ac27a0ec DK |
1565 | "unsupported optional features (%x).\n", |
1566 | sb->s_id, le32_to_cpu(features)); | |
1567 | goto failed_mount; | |
1568 | } | |
1569 | blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); | |
1570 | ||
617ba13b MC |
1571 | if (blocksize < EXT4_MIN_BLOCK_SIZE || |
1572 | blocksize > EXT4_MAX_BLOCK_SIZE) { | |
ac27a0ec | 1573 | printk(KERN_ERR |
617ba13b | 1574 | "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n", |
ac27a0ec DK |
1575 | blocksize, sb->s_id); |
1576 | goto failed_mount; | |
1577 | } | |
1578 | ||
1579 | hblock = bdev_hardsect_size(sb->s_bdev); | |
1580 | if (sb->s_blocksize != blocksize) { | |
1581 | /* | |
1582 | * Make sure the blocksize for the filesystem is larger | |
1583 | * than the hardware sectorsize for the machine. | |
1584 | */ | |
1585 | if (blocksize < hblock) { | |
617ba13b | 1586 | printk(KERN_ERR "EXT4-fs: blocksize %d too small for " |
ac27a0ec DK |
1587 | "device blocksize %d.\n", blocksize, hblock); |
1588 | goto failed_mount; | |
1589 | } | |
1590 | ||
1591 | brelse (bh); | |
1592 | sb_set_blocksize(sb, blocksize); | |
3a5b2ecd MC |
1593 | logic_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
1594 | offset = sector_div(logic_sb_block, blocksize); | |
ac27a0ec DK |
1595 | bh = sb_bread(sb, logic_sb_block); |
1596 | if (!bh) { | |
1597 | printk(KERN_ERR | |
617ba13b | 1598 | "EXT4-fs: Can't read superblock on 2nd try.\n"); |
ac27a0ec DK |
1599 | goto failed_mount; |
1600 | } | |
617ba13b | 1601 | es = (struct ext4_super_block *)(((char *)bh->b_data) + offset); |
ac27a0ec | 1602 | sbi->s_es = es; |
617ba13b | 1603 | if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { |
ac27a0ec | 1604 | printk (KERN_ERR |
617ba13b | 1605 | "EXT4-fs: Magic mismatch, very weird !\n"); |
ac27a0ec DK |
1606 | goto failed_mount; |
1607 | } | |
1608 | } | |
1609 | ||
617ba13b | 1610 | sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits); |
ac27a0ec | 1611 | |
617ba13b MC |
1612 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { |
1613 | sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; | |
1614 | sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; | |
ac27a0ec DK |
1615 | } else { |
1616 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
1617 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
617ba13b | 1618 | if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || |
ac27a0ec DK |
1619 | (sbi->s_inode_size & (sbi->s_inode_size - 1)) || |
1620 | (sbi->s_inode_size > blocksize)) { | |
1621 | printk (KERN_ERR | |
617ba13b | 1622 | "EXT4-fs: unsupported inode size: %d\n", |
ac27a0ec DK |
1623 | sbi->s_inode_size); |
1624 | goto failed_mount; | |
1625 | } | |
1626 | } | |
617ba13b | 1627 | sbi->s_frag_size = EXT4_MIN_FRAG_SIZE << |
ac27a0ec DK |
1628 | le32_to_cpu(es->s_log_frag_size); |
1629 | if (blocksize != sbi->s_frag_size) { | |
1630 | printk(KERN_ERR | |
617ba13b | 1631 | "EXT4-fs: fragsize %lu != blocksize %u (unsupported)\n", |
ac27a0ec DK |
1632 | sbi->s_frag_size, blocksize); |
1633 | goto failed_mount; | |
1634 | } | |
0d1ee42f AR |
1635 | sbi->s_desc_size = le16_to_cpu(es->s_desc_size); |
1636 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { | |
8fadc143 | 1637 | if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || |
0d1ee42f AR |
1638 | sbi->s_desc_size > EXT4_MAX_DESC_SIZE || |
1639 | sbi->s_desc_size & (sbi->s_desc_size - 1)) { | |
1640 | printk(KERN_ERR | |
8fadc143 | 1641 | "EXT4-fs: unsupported descriptor size %lu\n", |
0d1ee42f AR |
1642 | sbi->s_desc_size); |
1643 | goto failed_mount; | |
1644 | } | |
1645 | } else | |
1646 | sbi->s_desc_size = EXT4_MIN_DESC_SIZE; | |
ac27a0ec DK |
1647 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); |
1648 | sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group); | |
1649 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); | |
617ba13b MC |
1650 | if (EXT4_INODE_SIZE(sb) == 0) |
1651 | goto cantfind_ext4; | |
1652 | sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); | |
ac27a0ec | 1653 | if (sbi->s_inodes_per_block == 0) |
617ba13b | 1654 | goto cantfind_ext4; |
ac27a0ec DK |
1655 | sbi->s_itb_per_group = sbi->s_inodes_per_group / |
1656 | sbi->s_inodes_per_block; | |
0d1ee42f | 1657 | sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); |
ac27a0ec DK |
1658 | sbi->s_sbh = bh; |
1659 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
617ba13b MC |
1660 | sbi->s_addr_per_block_bits = log2(EXT4_ADDR_PER_BLOCK(sb)); |
1661 | sbi->s_desc_per_block_bits = log2(EXT4_DESC_PER_BLOCK(sb)); | |
ac27a0ec DK |
1662 | for (i=0; i < 4; i++) |
1663 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); | |
1664 | sbi->s_def_hash_version = es->s_def_hash_version; | |
1665 | ||
1666 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
1667 | printk (KERN_ERR | |
617ba13b | 1668 | "EXT4-fs: #blocks per group too big: %lu\n", |
ac27a0ec DK |
1669 | sbi->s_blocks_per_group); |
1670 | goto failed_mount; | |
1671 | } | |
1672 | if (sbi->s_frags_per_group > blocksize * 8) { | |
1673 | printk (KERN_ERR | |
617ba13b | 1674 | "EXT4-fs: #fragments per group too big: %lu\n", |
ac27a0ec DK |
1675 | sbi->s_frags_per_group); |
1676 | goto failed_mount; | |
1677 | } | |
1678 | if (sbi->s_inodes_per_group > blocksize * 8) { | |
1679 | printk (KERN_ERR | |
617ba13b | 1680 | "EXT4-fs: #inodes per group too big: %lu\n", |
ac27a0ec DK |
1681 | sbi->s_inodes_per_group); |
1682 | goto failed_mount; | |
1683 | } | |
1684 | ||
bd81d8ee | 1685 | if (ext4_blocks_count(es) > |
ac27a0ec | 1686 | (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) { |
617ba13b | 1687 | printk(KERN_ERR "EXT4-fs: filesystem on %s:" |
ac27a0ec DK |
1688 | " too large to mount safely\n", sb->s_id); |
1689 | if (sizeof(sector_t) < 8) | |
617ba13b | 1690 | printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not " |
ac27a0ec DK |
1691 | "enabled\n"); |
1692 | goto failed_mount; | |
1693 | } | |
1694 | ||
617ba13b MC |
1695 | if (EXT4_BLOCKS_PER_GROUP(sb) == 0) |
1696 | goto cantfind_ext4; | |
bd81d8ee LV |
1697 | blocks_count = (ext4_blocks_count(es) - |
1698 | le32_to_cpu(es->s_first_data_block) + | |
1699 | EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
1700 | do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); | |
1701 | sbi->s_groups_count = blocks_count; | |
617ba13b MC |
1702 | db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / |
1703 | EXT4_DESC_PER_BLOCK(sb); | |
ac27a0ec DK |
1704 | sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *), |
1705 | GFP_KERNEL); | |
1706 | if (sbi->s_group_desc == NULL) { | |
617ba13b | 1707 | printk (KERN_ERR "EXT4-fs: not enough memory\n"); |
ac27a0ec DK |
1708 | goto failed_mount; |
1709 | } | |
1710 | ||
1711 | bgl_lock_init(&sbi->s_blockgroup_lock); | |
1712 | ||
1713 | for (i = 0; i < db_count; i++) { | |
1714 | block = descriptor_loc(sb, logic_sb_block, i); | |
1715 | sbi->s_group_desc[i] = sb_bread(sb, block); | |
1716 | if (!sbi->s_group_desc[i]) { | |
617ba13b | 1717 | printk (KERN_ERR "EXT4-fs: " |
ac27a0ec DK |
1718 | "can't read group descriptor %d\n", i); |
1719 | db_count = i; | |
1720 | goto failed_mount2; | |
1721 | } | |
1722 | } | |
617ba13b MC |
1723 | if (!ext4_check_descriptors (sb)) { |
1724 | printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n"); | |
ac27a0ec DK |
1725 | goto failed_mount2; |
1726 | } | |
1727 | sbi->s_gdb_count = db_count; | |
1728 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
1729 | spin_lock_init(&sbi->s_next_gen_lock); | |
1730 | ||
1731 | percpu_counter_init(&sbi->s_freeblocks_counter, | |
617ba13b | 1732 | ext4_count_free_blocks(sb)); |
ac27a0ec | 1733 | percpu_counter_init(&sbi->s_freeinodes_counter, |
617ba13b | 1734 | ext4_count_free_inodes(sb)); |
ac27a0ec | 1735 | percpu_counter_init(&sbi->s_dirs_counter, |
617ba13b | 1736 | ext4_count_dirs(sb)); |
ac27a0ec DK |
1737 | |
1738 | /* per fileystem reservation list head & lock */ | |
1739 | spin_lock_init(&sbi->s_rsv_window_lock); | |
1740 | sbi->s_rsv_window_root = RB_ROOT; | |
1741 | /* Add a single, static dummy reservation to the start of the | |
1742 | * reservation window list --- it gives us a placeholder for | |
1743 | * append-at-start-of-list which makes the allocation logic | |
1744 | * _much_ simpler. */ | |
617ba13b MC |
1745 | sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED; |
1746 | sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED; | |
ac27a0ec DK |
1747 | sbi->s_rsv_window_head.rsv_alloc_hit = 0; |
1748 | sbi->s_rsv_window_head.rsv_goal_size = 0; | |
617ba13b | 1749 | ext4_rsv_window_add(sb, &sbi->s_rsv_window_head); |
ac27a0ec DK |
1750 | |
1751 | /* | |
1752 | * set up enough so that it can read an inode | |
1753 | */ | |
617ba13b MC |
1754 | sb->s_op = &ext4_sops; |
1755 | sb->s_export_op = &ext4_export_ops; | |
1756 | sb->s_xattr = ext4_xattr_handlers; | |
ac27a0ec | 1757 | #ifdef CONFIG_QUOTA |
617ba13b MC |
1758 | sb->s_qcop = &ext4_qctl_operations; |
1759 | sb->dq_op = &ext4_quota_operations; | |
ac27a0ec DK |
1760 | #endif |
1761 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ | |
1762 | ||
1763 | sb->s_root = NULL; | |
1764 | ||
1765 | needs_recovery = (es->s_last_orphan != 0 || | |
617ba13b MC |
1766 | EXT4_HAS_INCOMPAT_FEATURE(sb, |
1767 | EXT4_FEATURE_INCOMPAT_RECOVER)); | |
ac27a0ec DK |
1768 | |
1769 | /* | |
1770 | * The first inode we look at is the journal inode. Don't try | |
1771 | * root first: it may be modified in the journal! | |
1772 | */ | |
1773 | if (!test_opt(sb, NOLOAD) && | |
617ba13b MC |
1774 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
1775 | if (ext4_load_journal(sb, es, journal_devnum)) | |
ac27a0ec DK |
1776 | goto failed_mount3; |
1777 | } else if (journal_inum) { | |
617ba13b | 1778 | if (ext4_create_journal(sb, es, journal_inum)) |
ac27a0ec DK |
1779 | goto failed_mount3; |
1780 | } else { | |
1781 | if (!silent) | |
1782 | printk (KERN_ERR | |
617ba13b | 1783 | "ext4: No journal on filesystem on %s\n", |
ac27a0ec DK |
1784 | sb->s_id); |
1785 | goto failed_mount3; | |
1786 | } | |
1787 | ||
1788 | /* We have now updated the journal if required, so we can | |
1789 | * validate the data journaling mode. */ | |
1790 | switch (test_opt(sb, DATA_FLAGS)) { | |
1791 | case 0: | |
1792 | /* No mode set, assume a default based on the journal | |
1793 | capabilities: ORDERED_DATA if the journal can | |
1794 | cope, else JOURNAL_DATA */ | |
dab291af MC |
1795 | if (jbd2_journal_check_available_features |
1796 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) | |
ac27a0ec DK |
1797 | set_opt(sbi->s_mount_opt, ORDERED_DATA); |
1798 | else | |
1799 | set_opt(sbi->s_mount_opt, JOURNAL_DATA); | |
1800 | break; | |
1801 | ||
617ba13b MC |
1802 | case EXT4_MOUNT_ORDERED_DATA: |
1803 | case EXT4_MOUNT_WRITEBACK_DATA: | |
dab291af MC |
1804 | if (!jbd2_journal_check_available_features |
1805 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { | |
617ba13b | 1806 | printk(KERN_ERR "EXT4-fs: Journal does not support " |
ac27a0ec DK |
1807 | "requested data journaling mode\n"); |
1808 | goto failed_mount4; | |
1809 | } | |
1810 | default: | |
1811 | break; | |
1812 | } | |
1813 | ||
1814 | if (test_opt(sb, NOBH)) { | |
617ba13b MC |
1815 | if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) { |
1816 | printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - " | |
ac27a0ec DK |
1817 | "its supported only with writeback mode\n"); |
1818 | clear_opt(sbi->s_mount_opt, NOBH); | |
1819 | } | |
1820 | } | |
1821 | /* | |
dab291af | 1822 | * The jbd2_journal_load will have done any necessary log recovery, |
ac27a0ec DK |
1823 | * so we can safely mount the rest of the filesystem now. |
1824 | */ | |
1825 | ||
617ba13b | 1826 | root = iget(sb, EXT4_ROOT_INO); |
ac27a0ec DK |
1827 | sb->s_root = d_alloc_root(root); |
1828 | if (!sb->s_root) { | |
617ba13b | 1829 | printk(KERN_ERR "EXT4-fs: get root inode failed\n"); |
ac27a0ec DK |
1830 | iput(root); |
1831 | goto failed_mount4; | |
1832 | } | |
1833 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
1834 | dput(sb->s_root); | |
1835 | sb->s_root = NULL; | |
617ba13b | 1836 | printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n"); |
ac27a0ec DK |
1837 | goto failed_mount4; |
1838 | } | |
1839 | ||
617ba13b | 1840 | ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY); |
ac27a0ec DK |
1841 | /* |
1842 | * akpm: core read_super() calls in here with the superblock locked. | |
1843 | * That deadlocks, because orphan cleanup needs to lock the superblock | |
1844 | * in numerous places. Here we just pop the lock - it's relatively | |
1845 | * harmless, because we are now ready to accept write_super() requests, | |
1846 | * and aviro says that's the only reason for hanging onto the | |
1847 | * superblock lock. | |
1848 | */ | |
617ba13b MC |
1849 | EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; |
1850 | ext4_orphan_cleanup(sb, es); | |
1851 | EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; | |
ac27a0ec | 1852 | if (needs_recovery) |
617ba13b MC |
1853 | printk (KERN_INFO "EXT4-fs: recovery complete.\n"); |
1854 | ext4_mark_recovery_complete(sb, es); | |
1855 | printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n", | |
1856 | test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal": | |
1857 | test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered": | |
ac27a0ec DK |
1858 | "writeback"); |
1859 | ||
a86c6181 AT |
1860 | ext4_ext_init(sb); |
1861 | ||
ac27a0ec DK |
1862 | lock_kernel(); |
1863 | return 0; | |
1864 | ||
617ba13b | 1865 | cantfind_ext4: |
ac27a0ec | 1866 | if (!silent) |
617ba13b | 1867 | printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n", |
ac27a0ec DK |
1868 | sb->s_id); |
1869 | goto failed_mount; | |
1870 | ||
1871 | failed_mount4: | |
dab291af | 1872 | jbd2_journal_destroy(sbi->s_journal); |
ac27a0ec DK |
1873 | failed_mount3: |
1874 | percpu_counter_destroy(&sbi->s_freeblocks_counter); | |
1875 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
1876 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
1877 | failed_mount2: | |
1878 | for (i = 0; i < db_count; i++) | |
1879 | brelse(sbi->s_group_desc[i]); | |
1880 | kfree(sbi->s_group_desc); | |
1881 | failed_mount: | |
1882 | #ifdef CONFIG_QUOTA | |
1883 | for (i = 0; i < MAXQUOTAS; i++) | |
1884 | kfree(sbi->s_qf_names[i]); | |
1885 | #endif | |
617ba13b | 1886 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
1887 | brelse(bh); |
1888 | out_fail: | |
1889 | sb->s_fs_info = NULL; | |
1890 | kfree(sbi); | |
1891 | lock_kernel(); | |
1892 | return -EINVAL; | |
1893 | } | |
1894 | ||
1895 | /* | |
1896 | * Setup any per-fs journal parameters now. We'll do this both on | |
1897 | * initial mount, once the journal has been initialised but before we've | |
1898 | * done any recovery; and again on any subsequent remount. | |
1899 | */ | |
617ba13b | 1900 | static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) |
ac27a0ec | 1901 | { |
617ba13b | 1902 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
1903 | |
1904 | if (sbi->s_commit_interval) | |
1905 | journal->j_commit_interval = sbi->s_commit_interval; | |
617ba13b | 1906 | /* We could also set up an ext4-specific default for the commit |
ac27a0ec DK |
1907 | * interval here, but for now we'll just fall back to the jbd |
1908 | * default. */ | |
1909 | ||
1910 | spin_lock(&journal->j_state_lock); | |
1911 | if (test_opt(sb, BARRIER)) | |
dab291af | 1912 | journal->j_flags |= JBD2_BARRIER; |
ac27a0ec | 1913 | else |
dab291af | 1914 | journal->j_flags &= ~JBD2_BARRIER; |
ac27a0ec DK |
1915 | spin_unlock(&journal->j_state_lock); |
1916 | } | |
1917 | ||
617ba13b | 1918 | static journal_t *ext4_get_journal(struct super_block *sb, |
ac27a0ec DK |
1919 | unsigned int journal_inum) |
1920 | { | |
1921 | struct inode *journal_inode; | |
1922 | journal_t *journal; | |
1923 | ||
1924 | /* First, test for the existence of a valid inode on disk. Bad | |
1925 | * things happen if we iget() an unused inode, as the subsequent | |
1926 | * iput() will try to delete it. */ | |
1927 | ||
1928 | journal_inode = iget(sb, journal_inum); | |
1929 | if (!journal_inode) { | |
617ba13b | 1930 | printk(KERN_ERR "EXT4-fs: no journal found.\n"); |
ac27a0ec DK |
1931 | return NULL; |
1932 | } | |
1933 | if (!journal_inode->i_nlink) { | |
1934 | make_bad_inode(journal_inode); | |
1935 | iput(journal_inode); | |
617ba13b | 1936 | printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n"); |
ac27a0ec DK |
1937 | return NULL; |
1938 | } | |
1939 | ||
1940 | jbd_debug(2, "Journal inode found at %p: %Ld bytes\n", | |
1941 | journal_inode, journal_inode->i_size); | |
1942 | if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) { | |
617ba13b | 1943 | printk(KERN_ERR "EXT4-fs: invalid journal inode.\n"); |
ac27a0ec DK |
1944 | iput(journal_inode); |
1945 | return NULL; | |
1946 | } | |
1947 | ||
dab291af | 1948 | journal = jbd2_journal_init_inode(journal_inode); |
ac27a0ec | 1949 | if (!journal) { |
617ba13b | 1950 | printk(KERN_ERR "EXT4-fs: Could not load journal inode\n"); |
ac27a0ec DK |
1951 | iput(journal_inode); |
1952 | return NULL; | |
1953 | } | |
1954 | journal->j_private = sb; | |
617ba13b | 1955 | ext4_init_journal_params(sb, journal); |
ac27a0ec DK |
1956 | return journal; |
1957 | } | |
1958 | ||
617ba13b | 1959 | static journal_t *ext4_get_dev_journal(struct super_block *sb, |
ac27a0ec DK |
1960 | dev_t j_dev) |
1961 | { | |
1962 | struct buffer_head * bh; | |
1963 | journal_t *journal; | |
617ba13b MC |
1964 | ext4_fsblk_t start; |
1965 | ext4_fsblk_t len; | |
ac27a0ec | 1966 | int hblock, blocksize; |
617ba13b | 1967 | ext4_fsblk_t sb_block; |
ac27a0ec | 1968 | unsigned long offset; |
617ba13b | 1969 | struct ext4_super_block * es; |
ac27a0ec DK |
1970 | struct block_device *bdev; |
1971 | ||
617ba13b | 1972 | bdev = ext4_blkdev_get(j_dev); |
ac27a0ec DK |
1973 | if (bdev == NULL) |
1974 | return NULL; | |
1975 | ||
1976 | if (bd_claim(bdev, sb)) { | |
1977 | printk(KERN_ERR | |
617ba13b | 1978 | "EXT4: failed to claim external journal device.\n"); |
ac27a0ec DK |
1979 | blkdev_put(bdev); |
1980 | return NULL; | |
1981 | } | |
1982 | ||
1983 | blocksize = sb->s_blocksize; | |
1984 | hblock = bdev_hardsect_size(bdev); | |
1985 | if (blocksize < hblock) { | |
1986 | printk(KERN_ERR | |
617ba13b | 1987 | "EXT4-fs: blocksize too small for journal device.\n"); |
ac27a0ec DK |
1988 | goto out_bdev; |
1989 | } | |
1990 | ||
617ba13b MC |
1991 | sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; |
1992 | offset = EXT4_MIN_BLOCK_SIZE % blocksize; | |
ac27a0ec DK |
1993 | set_blocksize(bdev, blocksize); |
1994 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
617ba13b | 1995 | printk(KERN_ERR "EXT4-fs: couldn't read superblock of " |
ac27a0ec DK |
1996 | "external journal\n"); |
1997 | goto out_bdev; | |
1998 | } | |
1999 | ||
617ba13b MC |
2000 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
2001 | if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || | |
ac27a0ec | 2002 | !(le32_to_cpu(es->s_feature_incompat) & |
617ba13b MC |
2003 | EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { |
2004 | printk(KERN_ERR "EXT4-fs: external journal has " | |
ac27a0ec DK |
2005 | "bad superblock\n"); |
2006 | brelse(bh); | |
2007 | goto out_bdev; | |
2008 | } | |
2009 | ||
617ba13b MC |
2010 | if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { |
2011 | printk(KERN_ERR "EXT4-fs: journal UUID does not match\n"); | |
ac27a0ec DK |
2012 | brelse(bh); |
2013 | goto out_bdev; | |
2014 | } | |
2015 | ||
bd81d8ee | 2016 | len = ext4_blocks_count(es); |
ac27a0ec DK |
2017 | start = sb_block + 1; |
2018 | brelse(bh); /* we're done with the superblock */ | |
2019 | ||
dab291af | 2020 | journal = jbd2_journal_init_dev(bdev, sb->s_bdev, |
ac27a0ec DK |
2021 | start, len, blocksize); |
2022 | if (!journal) { | |
617ba13b | 2023 | printk(KERN_ERR "EXT4-fs: failed to create device journal\n"); |
ac27a0ec DK |
2024 | goto out_bdev; |
2025 | } | |
2026 | journal->j_private = sb; | |
2027 | ll_rw_block(READ, 1, &journal->j_sb_buffer); | |
2028 | wait_on_buffer(journal->j_sb_buffer); | |
2029 | if (!buffer_uptodate(journal->j_sb_buffer)) { | |
617ba13b | 2030 | printk(KERN_ERR "EXT4-fs: I/O error on journal device\n"); |
ac27a0ec DK |
2031 | goto out_journal; |
2032 | } | |
2033 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
617ba13b | 2034 | printk(KERN_ERR "EXT4-fs: External journal has more than one " |
ac27a0ec DK |
2035 | "user (unsupported) - %d\n", |
2036 | be32_to_cpu(journal->j_superblock->s_nr_users)); | |
2037 | goto out_journal; | |
2038 | } | |
617ba13b MC |
2039 | EXT4_SB(sb)->journal_bdev = bdev; |
2040 | ext4_init_journal_params(sb, journal); | |
ac27a0ec DK |
2041 | return journal; |
2042 | out_journal: | |
dab291af | 2043 | jbd2_journal_destroy(journal); |
ac27a0ec | 2044 | out_bdev: |
617ba13b | 2045 | ext4_blkdev_put(bdev); |
ac27a0ec DK |
2046 | return NULL; |
2047 | } | |
2048 | ||
617ba13b MC |
2049 | static int ext4_load_journal(struct super_block *sb, |
2050 | struct ext4_super_block *es, | |
ac27a0ec DK |
2051 | unsigned long journal_devnum) |
2052 | { | |
2053 | journal_t *journal; | |
2054 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); | |
2055 | dev_t journal_dev; | |
2056 | int err = 0; | |
2057 | int really_read_only; | |
2058 | ||
2059 | if (journal_devnum && | |
2060 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
617ba13b | 2061 | printk(KERN_INFO "EXT4-fs: external journal device major/minor " |
ac27a0ec DK |
2062 | "numbers have changed\n"); |
2063 | journal_dev = new_decode_dev(journal_devnum); | |
2064 | } else | |
2065 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
2066 | ||
2067 | really_read_only = bdev_read_only(sb->s_bdev); | |
2068 | ||
2069 | /* | |
2070 | * Are we loading a blank journal or performing recovery after a | |
2071 | * crash? For recovery, we need to check in advance whether we | |
2072 | * can get read-write access to the device. | |
2073 | */ | |
2074 | ||
617ba13b | 2075 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { |
ac27a0ec | 2076 | if (sb->s_flags & MS_RDONLY) { |
617ba13b | 2077 | printk(KERN_INFO "EXT4-fs: INFO: recovery " |
ac27a0ec DK |
2078 | "required on readonly filesystem.\n"); |
2079 | if (really_read_only) { | |
617ba13b | 2080 | printk(KERN_ERR "EXT4-fs: write access " |
ac27a0ec DK |
2081 | "unavailable, cannot proceed.\n"); |
2082 | return -EROFS; | |
2083 | } | |
617ba13b | 2084 | printk (KERN_INFO "EXT4-fs: write access will " |
ac27a0ec DK |
2085 | "be enabled during recovery.\n"); |
2086 | } | |
2087 | } | |
2088 | ||
2089 | if (journal_inum && journal_dev) { | |
617ba13b | 2090 | printk(KERN_ERR "EXT4-fs: filesystem has both journal " |
ac27a0ec DK |
2091 | "and inode journals!\n"); |
2092 | return -EINVAL; | |
2093 | } | |
2094 | ||
2095 | if (journal_inum) { | |
617ba13b | 2096 | if (!(journal = ext4_get_journal(sb, journal_inum))) |
ac27a0ec DK |
2097 | return -EINVAL; |
2098 | } else { | |
617ba13b | 2099 | if (!(journal = ext4_get_dev_journal(sb, journal_dev))) |
ac27a0ec DK |
2100 | return -EINVAL; |
2101 | } | |
2102 | ||
2103 | if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { | |
dab291af | 2104 | err = jbd2_journal_update_format(journal); |
ac27a0ec | 2105 | if (err) { |
617ba13b | 2106 | printk(KERN_ERR "EXT4-fs: error updating journal.\n"); |
dab291af | 2107 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
2108 | return err; |
2109 | } | |
2110 | } | |
2111 | ||
617ba13b | 2112 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) |
dab291af | 2113 | err = jbd2_journal_wipe(journal, !really_read_only); |
ac27a0ec | 2114 | if (!err) |
dab291af | 2115 | err = jbd2_journal_load(journal); |
ac27a0ec DK |
2116 | |
2117 | if (err) { | |
617ba13b | 2118 | printk(KERN_ERR "EXT4-fs: error loading journal.\n"); |
dab291af | 2119 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
2120 | return err; |
2121 | } | |
2122 | ||
617ba13b MC |
2123 | EXT4_SB(sb)->s_journal = journal; |
2124 | ext4_clear_journal_err(sb, es); | |
ac27a0ec DK |
2125 | |
2126 | if (journal_devnum && | |
2127 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
2128 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
2129 | sb->s_dirt = 1; | |
2130 | ||
2131 | /* Make sure we flush the recovery flag to disk. */ | |
617ba13b | 2132 | ext4_commit_super(sb, es, 1); |
ac27a0ec DK |
2133 | } |
2134 | ||
2135 | return 0; | |
2136 | } | |
2137 | ||
617ba13b MC |
2138 | static int ext4_create_journal(struct super_block * sb, |
2139 | struct ext4_super_block * es, | |
ac27a0ec DK |
2140 | unsigned int journal_inum) |
2141 | { | |
2142 | journal_t *journal; | |
2143 | ||
2144 | if (sb->s_flags & MS_RDONLY) { | |
617ba13b | 2145 | printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to " |
ac27a0ec DK |
2146 | "create journal.\n"); |
2147 | return -EROFS; | |
2148 | } | |
2149 | ||
617ba13b | 2150 | if (!(journal = ext4_get_journal(sb, journal_inum))) |
ac27a0ec DK |
2151 | return -EINVAL; |
2152 | ||
617ba13b | 2153 | printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n", |
ac27a0ec DK |
2154 | journal_inum); |
2155 | ||
dab291af | 2156 | if (jbd2_journal_create(journal)) { |
617ba13b | 2157 | printk(KERN_ERR "EXT4-fs: error creating journal.\n"); |
dab291af | 2158 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
2159 | return -EIO; |
2160 | } | |
2161 | ||
617ba13b | 2162 | EXT4_SB(sb)->s_journal = journal; |
ac27a0ec | 2163 | |
617ba13b MC |
2164 | ext4_update_dynamic_rev(sb); |
2165 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
2166 | EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL); | |
ac27a0ec DK |
2167 | |
2168 | es->s_journal_inum = cpu_to_le32(journal_inum); | |
2169 | sb->s_dirt = 1; | |
2170 | ||
2171 | /* Make sure we flush the recovery flag to disk. */ | |
617ba13b | 2172 | ext4_commit_super(sb, es, 1); |
ac27a0ec DK |
2173 | |
2174 | return 0; | |
2175 | } | |
2176 | ||
617ba13b MC |
2177 | static void ext4_commit_super (struct super_block * sb, |
2178 | struct ext4_super_block * es, | |
ac27a0ec DK |
2179 | int sync) |
2180 | { | |
617ba13b | 2181 | struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; |
ac27a0ec DK |
2182 | |
2183 | if (!sbh) | |
2184 | return; | |
2185 | es->s_wtime = cpu_to_le32(get_seconds()); | |
bd81d8ee | 2186 | ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb)); |
617ba13b | 2187 | es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb)); |
ac27a0ec DK |
2188 | BUFFER_TRACE(sbh, "marking dirty"); |
2189 | mark_buffer_dirty(sbh); | |
2190 | if (sync) | |
2191 | sync_dirty_buffer(sbh); | |
2192 | } | |
2193 | ||
2194 | ||
2195 | /* | |
2196 | * Have we just finished recovery? If so, and if we are mounting (or | |
2197 | * remounting) the filesystem readonly, then we will end up with a | |
2198 | * consistent fs on disk. Record that fact. | |
2199 | */ | |
617ba13b MC |
2200 | static void ext4_mark_recovery_complete(struct super_block * sb, |
2201 | struct ext4_super_block * es) | |
ac27a0ec | 2202 | { |
617ba13b | 2203 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 2204 | |
dab291af MC |
2205 | jbd2_journal_lock_updates(journal); |
2206 | jbd2_journal_flush(journal); | |
617ba13b | 2207 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && |
ac27a0ec | 2208 | sb->s_flags & MS_RDONLY) { |
617ba13b | 2209 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
ac27a0ec | 2210 | sb->s_dirt = 0; |
617ba13b | 2211 | ext4_commit_super(sb, es, 1); |
ac27a0ec | 2212 | } |
dab291af | 2213 | jbd2_journal_unlock_updates(journal); |
ac27a0ec DK |
2214 | } |
2215 | ||
2216 | /* | |
2217 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
2218 | * has recorded an error from a previous lifetime, move that error to the | |
2219 | * main filesystem now. | |
2220 | */ | |
617ba13b MC |
2221 | static void ext4_clear_journal_err(struct super_block * sb, |
2222 | struct ext4_super_block * es) | |
ac27a0ec DK |
2223 | { |
2224 | journal_t *journal; | |
2225 | int j_errno; | |
2226 | const char *errstr; | |
2227 | ||
617ba13b | 2228 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec DK |
2229 | |
2230 | /* | |
2231 | * Now check for any error status which may have been recorded in the | |
617ba13b | 2232 | * journal by a prior ext4_error() or ext4_abort() |
ac27a0ec DK |
2233 | */ |
2234 | ||
dab291af | 2235 | j_errno = jbd2_journal_errno(journal); |
ac27a0ec DK |
2236 | if (j_errno) { |
2237 | char nbuf[16]; | |
2238 | ||
617ba13b MC |
2239 | errstr = ext4_decode_error(sb, j_errno, nbuf); |
2240 | ext4_warning(sb, __FUNCTION__, "Filesystem error recorded " | |
ac27a0ec | 2241 | "from previous mount: %s", errstr); |
617ba13b | 2242 | ext4_warning(sb, __FUNCTION__, "Marking fs in need of " |
ac27a0ec DK |
2243 | "filesystem check."); |
2244 | ||
617ba13b MC |
2245 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
2246 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
2247 | ext4_commit_super (sb, es, 1); | |
ac27a0ec | 2248 | |
dab291af | 2249 | jbd2_journal_clear_err(journal); |
ac27a0ec DK |
2250 | } |
2251 | } | |
2252 | ||
2253 | /* | |
2254 | * Force the running and committing transactions to commit, | |
2255 | * and wait on the commit. | |
2256 | */ | |
617ba13b | 2257 | int ext4_force_commit(struct super_block *sb) |
ac27a0ec DK |
2258 | { |
2259 | journal_t *journal; | |
2260 | int ret; | |
2261 | ||
2262 | if (sb->s_flags & MS_RDONLY) | |
2263 | return 0; | |
2264 | ||
617ba13b | 2265 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 2266 | sb->s_dirt = 0; |
617ba13b | 2267 | ret = ext4_journal_force_commit(journal); |
ac27a0ec DK |
2268 | return ret; |
2269 | } | |
2270 | ||
2271 | /* | |
617ba13b | 2272 | * Ext4 always journals updates to the superblock itself, so we don't |
ac27a0ec DK |
2273 | * have to propagate any other updates to the superblock on disk at this |
2274 | * point. Just start an async writeback to get the buffers on their way | |
2275 | * to the disk. | |
2276 | * | |
2277 | * This implicitly triggers the writebehind on sync(). | |
2278 | */ | |
2279 | ||
617ba13b | 2280 | static void ext4_write_super (struct super_block * sb) |
ac27a0ec DK |
2281 | { |
2282 | if (mutex_trylock(&sb->s_lock) != 0) | |
2283 | BUG(); | |
2284 | sb->s_dirt = 0; | |
2285 | } | |
2286 | ||
617ba13b | 2287 | static int ext4_sync_fs(struct super_block *sb, int wait) |
ac27a0ec DK |
2288 | { |
2289 | tid_t target; | |
2290 | ||
2291 | sb->s_dirt = 0; | |
dab291af | 2292 | if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) { |
ac27a0ec | 2293 | if (wait) |
dab291af | 2294 | jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target); |
ac27a0ec DK |
2295 | } |
2296 | return 0; | |
2297 | } | |
2298 | ||
2299 | /* | |
2300 | * LVM calls this function before a (read-only) snapshot is created. This | |
2301 | * gives us a chance to flush the journal completely and mark the fs clean. | |
2302 | */ | |
617ba13b | 2303 | static void ext4_write_super_lockfs(struct super_block *sb) |
ac27a0ec DK |
2304 | { |
2305 | sb->s_dirt = 0; | |
2306 | ||
2307 | if (!(sb->s_flags & MS_RDONLY)) { | |
617ba13b | 2308 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec DK |
2309 | |
2310 | /* Now we set up the journal barrier. */ | |
dab291af MC |
2311 | jbd2_journal_lock_updates(journal); |
2312 | jbd2_journal_flush(journal); | |
ac27a0ec DK |
2313 | |
2314 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
617ba13b MC |
2315 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
2316 | ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1); | |
ac27a0ec DK |
2317 | } |
2318 | } | |
2319 | ||
2320 | /* | |
2321 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
2322 | * flag here, even though the filesystem is not technically dirty yet. | |
2323 | */ | |
617ba13b | 2324 | static void ext4_unlockfs(struct super_block *sb) |
ac27a0ec DK |
2325 | { |
2326 | if (!(sb->s_flags & MS_RDONLY)) { | |
2327 | lock_super(sb); | |
2328 | /* Reser the needs_recovery flag before the fs is unlocked. */ | |
617ba13b MC |
2329 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
2330 | ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1); | |
ac27a0ec | 2331 | unlock_super(sb); |
dab291af | 2332 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); |
ac27a0ec DK |
2333 | } |
2334 | } | |
2335 | ||
617ba13b | 2336 | static int ext4_remount (struct super_block * sb, int * flags, char * data) |
ac27a0ec | 2337 | { |
617ba13b MC |
2338 | struct ext4_super_block * es; |
2339 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2340 | ext4_fsblk_t n_blocks_count = 0; | |
ac27a0ec | 2341 | unsigned long old_sb_flags; |
617ba13b | 2342 | struct ext4_mount_options old_opts; |
ac27a0ec DK |
2343 | int err; |
2344 | #ifdef CONFIG_QUOTA | |
2345 | int i; | |
2346 | #endif | |
2347 | ||
2348 | /* Store the original options */ | |
2349 | old_sb_flags = sb->s_flags; | |
2350 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
2351 | old_opts.s_resuid = sbi->s_resuid; | |
2352 | old_opts.s_resgid = sbi->s_resgid; | |
2353 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
2354 | #ifdef CONFIG_QUOTA | |
2355 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
2356 | for (i = 0; i < MAXQUOTAS; i++) | |
2357 | old_opts.s_qf_names[i] = sbi->s_qf_names[i]; | |
2358 | #endif | |
2359 | ||
2360 | /* | |
2361 | * Allow the "check" option to be passed as a remount option. | |
2362 | */ | |
2363 | if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) { | |
2364 | err = -EINVAL; | |
2365 | goto restore_opts; | |
2366 | } | |
2367 | ||
617ba13b MC |
2368 | if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) |
2369 | ext4_abort(sb, __FUNCTION__, "Abort forced by user"); | |
ac27a0ec DK |
2370 | |
2371 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
617ba13b | 2372 | ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec DK |
2373 | |
2374 | es = sbi->s_es; | |
2375 | ||
617ba13b | 2376 | ext4_init_journal_params(sb, sbi->s_journal); |
ac27a0ec DK |
2377 | |
2378 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) || | |
bd81d8ee | 2379 | n_blocks_count > ext4_blocks_count(es)) { |
617ba13b | 2380 | if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) { |
ac27a0ec DK |
2381 | err = -EROFS; |
2382 | goto restore_opts; | |
2383 | } | |
2384 | ||
2385 | if (*flags & MS_RDONLY) { | |
2386 | /* | |
2387 | * First of all, the unconditional stuff we have to do | |
2388 | * to disable replay of the journal when we next remount | |
2389 | */ | |
2390 | sb->s_flags |= MS_RDONLY; | |
2391 | ||
2392 | /* | |
2393 | * OK, test if we are remounting a valid rw partition | |
2394 | * readonly, and if so set the rdonly flag and then | |
2395 | * mark the partition as valid again. | |
2396 | */ | |
617ba13b MC |
2397 | if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && |
2398 | (sbi->s_mount_state & EXT4_VALID_FS)) | |
ac27a0ec DK |
2399 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
2400 | ||
617ba13b | 2401 | ext4_mark_recovery_complete(sb, es); |
ac27a0ec DK |
2402 | } else { |
2403 | __le32 ret; | |
617ba13b MC |
2404 | if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb, |
2405 | ~EXT4_FEATURE_RO_COMPAT_SUPP))) { | |
2406 | printk(KERN_WARNING "EXT4-fs: %s: couldn't " | |
ac27a0ec DK |
2407 | "remount RDWR because of unsupported " |
2408 | "optional features (%x).\n", | |
2409 | sb->s_id, le32_to_cpu(ret)); | |
2410 | err = -EROFS; | |
2411 | goto restore_opts; | |
2412 | } | |
2413 | /* | |
2414 | * Mounting a RDONLY partition read-write, so reread | |
2415 | * and store the current valid flag. (It may have | |
2416 | * been changed by e2fsck since we originally mounted | |
2417 | * the partition.) | |
2418 | */ | |
617ba13b | 2419 | ext4_clear_journal_err(sb, es); |
ac27a0ec | 2420 | sbi->s_mount_state = le16_to_cpu(es->s_state); |
617ba13b | 2421 | if ((err = ext4_group_extend(sb, es, n_blocks_count))) |
ac27a0ec | 2422 | goto restore_opts; |
617ba13b | 2423 | if (!ext4_setup_super (sb, es, 0)) |
ac27a0ec DK |
2424 | sb->s_flags &= ~MS_RDONLY; |
2425 | } | |
2426 | } | |
2427 | #ifdef CONFIG_QUOTA | |
2428 | /* Release old quota file names */ | |
2429 | for (i = 0; i < MAXQUOTAS; i++) | |
2430 | if (old_opts.s_qf_names[i] && | |
2431 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
2432 | kfree(old_opts.s_qf_names[i]); | |
2433 | #endif | |
2434 | return 0; | |
2435 | restore_opts: | |
2436 | sb->s_flags = old_sb_flags; | |
2437 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
2438 | sbi->s_resuid = old_opts.s_resuid; | |
2439 | sbi->s_resgid = old_opts.s_resgid; | |
2440 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
2441 | #ifdef CONFIG_QUOTA | |
2442 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
2443 | for (i = 0; i < MAXQUOTAS; i++) { | |
2444 | if (sbi->s_qf_names[i] && | |
2445 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
2446 | kfree(sbi->s_qf_names[i]); | |
2447 | sbi->s_qf_names[i] = old_opts.s_qf_names[i]; | |
2448 | } | |
2449 | #endif | |
2450 | return err; | |
2451 | } | |
2452 | ||
617ba13b | 2453 | static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf) |
ac27a0ec DK |
2454 | { |
2455 | struct super_block *sb = dentry->d_sb; | |
617ba13b MC |
2456 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2457 | struct ext4_super_block *es = sbi->s_es; | |
2458 | ext4_fsblk_t overhead; | |
ac27a0ec DK |
2459 | int i; |
2460 | ||
2461 | if (test_opt (sb, MINIX_DF)) | |
2462 | overhead = 0; | |
2463 | else { | |
2464 | unsigned long ngroups; | |
617ba13b | 2465 | ngroups = EXT4_SB(sb)->s_groups_count; |
ac27a0ec DK |
2466 | smp_rmb(); |
2467 | ||
2468 | /* | |
2469 | * Compute the overhead (FS structures) | |
2470 | */ | |
2471 | ||
2472 | /* | |
2473 | * All of the blocks before first_data_block are | |
2474 | * overhead | |
2475 | */ | |
2476 | overhead = le32_to_cpu(es->s_first_data_block); | |
2477 | ||
2478 | /* | |
2479 | * Add the overhead attributed to the superblock and | |
2480 | * block group descriptors. If the sparse superblocks | |
2481 | * feature is turned on, then not all groups have this. | |
2482 | */ | |
2483 | for (i = 0; i < ngroups; i++) { | |
617ba13b MC |
2484 | overhead += ext4_bg_has_super(sb, i) + |
2485 | ext4_bg_num_gdb(sb, i); | |
ac27a0ec DK |
2486 | cond_resched(); |
2487 | } | |
2488 | ||
2489 | /* | |
2490 | * Every block group has an inode bitmap, a block | |
2491 | * bitmap, and an inode table. | |
2492 | */ | |
617ba13b | 2493 | overhead += (ngroups * (2 + EXT4_SB(sb)->s_itb_per_group)); |
ac27a0ec DK |
2494 | } |
2495 | ||
617ba13b | 2496 | buf->f_type = EXT4_SUPER_MAGIC; |
ac27a0ec | 2497 | buf->f_bsize = sb->s_blocksize; |
bd81d8ee | 2498 | buf->f_blocks = ext4_blocks_count(es) - overhead; |
ac27a0ec | 2499 | buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter); |
bd81d8ee LV |
2500 | buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); |
2501 | if (buf->f_bfree < ext4_r_blocks_count(es)) | |
ac27a0ec DK |
2502 | buf->f_bavail = 0; |
2503 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
2504 | buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter); | |
617ba13b | 2505 | buf->f_namelen = EXT4_NAME_LEN; |
ac27a0ec DK |
2506 | return 0; |
2507 | } | |
2508 | ||
2509 | /* Helper function for writing quotas on sync - we need to start transaction before quota file | |
2510 | * is locked for write. Otherwise the are possible deadlocks: | |
2511 | * Process 1 Process 2 | |
617ba13b | 2512 | * ext4_create() quota_sync() |
dab291af | 2513 | * jbd2_journal_start() write_dquot() |
ac27a0ec | 2514 | * DQUOT_INIT() down(dqio_mutex) |
dab291af | 2515 | * down(dqio_mutex) jbd2_journal_start() |
ac27a0ec DK |
2516 | * |
2517 | */ | |
2518 | ||
2519 | #ifdef CONFIG_QUOTA | |
2520 | ||
2521 | static inline struct inode *dquot_to_inode(struct dquot *dquot) | |
2522 | { | |
2523 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]; | |
2524 | } | |
2525 | ||
617ba13b | 2526 | static int ext4_dquot_initialize(struct inode *inode, int type) |
ac27a0ec DK |
2527 | { |
2528 | handle_t *handle; | |
2529 | int ret, err; | |
2530 | ||
2531 | /* We may create quota structure so we need to reserve enough blocks */ | |
617ba13b | 2532 | handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb)); |
ac27a0ec DK |
2533 | if (IS_ERR(handle)) |
2534 | return PTR_ERR(handle); | |
2535 | ret = dquot_initialize(inode, type); | |
617ba13b | 2536 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
2537 | if (!ret) |
2538 | ret = err; | |
2539 | return ret; | |
2540 | } | |
2541 | ||
617ba13b | 2542 | static int ext4_dquot_drop(struct inode *inode) |
ac27a0ec DK |
2543 | { |
2544 | handle_t *handle; | |
2545 | int ret, err; | |
2546 | ||
2547 | /* We may delete quota structure so we need to reserve enough blocks */ | |
617ba13b | 2548 | handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb)); |
ac27a0ec DK |
2549 | if (IS_ERR(handle)) |
2550 | return PTR_ERR(handle); | |
2551 | ret = dquot_drop(inode); | |
617ba13b | 2552 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
2553 | if (!ret) |
2554 | ret = err; | |
2555 | return ret; | |
2556 | } | |
2557 | ||
617ba13b | 2558 | static int ext4_write_dquot(struct dquot *dquot) |
ac27a0ec DK |
2559 | { |
2560 | int ret, err; | |
2561 | handle_t *handle; | |
2562 | struct inode *inode; | |
2563 | ||
2564 | inode = dquot_to_inode(dquot); | |
617ba13b MC |
2565 | handle = ext4_journal_start(inode, |
2566 | EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); | |
ac27a0ec DK |
2567 | if (IS_ERR(handle)) |
2568 | return PTR_ERR(handle); | |
2569 | ret = dquot_commit(dquot); | |
617ba13b | 2570 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
2571 | if (!ret) |
2572 | ret = err; | |
2573 | return ret; | |
2574 | } | |
2575 | ||
617ba13b | 2576 | static int ext4_acquire_dquot(struct dquot *dquot) |
ac27a0ec DK |
2577 | { |
2578 | int ret, err; | |
2579 | handle_t *handle; | |
2580 | ||
617ba13b MC |
2581 | handle = ext4_journal_start(dquot_to_inode(dquot), |
2582 | EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); | |
ac27a0ec DK |
2583 | if (IS_ERR(handle)) |
2584 | return PTR_ERR(handle); | |
2585 | ret = dquot_acquire(dquot); | |
617ba13b | 2586 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
2587 | if (!ret) |
2588 | ret = err; | |
2589 | return ret; | |
2590 | } | |
2591 | ||
617ba13b | 2592 | static int ext4_release_dquot(struct dquot *dquot) |
ac27a0ec DK |
2593 | { |
2594 | int ret, err; | |
2595 | handle_t *handle; | |
2596 | ||
617ba13b MC |
2597 | handle = ext4_journal_start(dquot_to_inode(dquot), |
2598 | EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); | |
ac27a0ec DK |
2599 | if (IS_ERR(handle)) |
2600 | return PTR_ERR(handle); | |
2601 | ret = dquot_release(dquot); | |
617ba13b | 2602 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
2603 | if (!ret) |
2604 | ret = err; | |
2605 | return ret; | |
2606 | } | |
2607 | ||
617ba13b | 2608 | static int ext4_mark_dquot_dirty(struct dquot *dquot) |
ac27a0ec DK |
2609 | { |
2610 | /* Are we journalling quotas? */ | |
617ba13b MC |
2611 | if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || |
2612 | EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { | |
ac27a0ec | 2613 | dquot_mark_dquot_dirty(dquot); |
617ba13b | 2614 | return ext4_write_dquot(dquot); |
ac27a0ec DK |
2615 | } else { |
2616 | return dquot_mark_dquot_dirty(dquot); | |
2617 | } | |
2618 | } | |
2619 | ||
617ba13b | 2620 | static int ext4_write_info(struct super_block *sb, int type) |
ac27a0ec DK |
2621 | { |
2622 | int ret, err; | |
2623 | handle_t *handle; | |
2624 | ||
2625 | /* Data block + inode block */ | |
617ba13b | 2626 | handle = ext4_journal_start(sb->s_root->d_inode, 2); |
ac27a0ec DK |
2627 | if (IS_ERR(handle)) |
2628 | return PTR_ERR(handle); | |
2629 | ret = dquot_commit_info(sb, type); | |
617ba13b | 2630 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
2631 | if (!ret) |
2632 | ret = err; | |
2633 | return ret; | |
2634 | } | |
2635 | ||
2636 | /* | |
2637 | * Turn on quotas during mount time - we need to find | |
2638 | * the quota file and such... | |
2639 | */ | |
617ba13b | 2640 | static int ext4_quota_on_mount(struct super_block *sb, int type) |
ac27a0ec | 2641 | { |
617ba13b MC |
2642 | return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], |
2643 | EXT4_SB(sb)->s_jquota_fmt, type); | |
ac27a0ec DK |
2644 | } |
2645 | ||
2646 | /* | |
2647 | * Standard function to be called on quota_on | |
2648 | */ | |
617ba13b | 2649 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
ac27a0ec DK |
2650 | char *path) |
2651 | { | |
2652 | int err; | |
2653 | struct nameidata nd; | |
2654 | ||
2655 | if (!test_opt(sb, QUOTA)) | |
2656 | return -EINVAL; | |
2657 | /* Not journalling quota? */ | |
617ba13b MC |
2658 | if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] && |
2659 | !EXT4_SB(sb)->s_qf_names[GRPQUOTA]) | |
ac27a0ec DK |
2660 | return vfs_quota_on(sb, type, format_id, path); |
2661 | err = path_lookup(path, LOOKUP_FOLLOW, &nd); | |
2662 | if (err) | |
2663 | return err; | |
2664 | /* Quotafile not on the same filesystem? */ | |
2665 | if (nd.mnt->mnt_sb != sb) { | |
2666 | path_release(&nd); | |
2667 | return -EXDEV; | |
2668 | } | |
2669 | /* Quotafile not of fs root? */ | |
2670 | if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode) | |
2671 | printk(KERN_WARNING | |
617ba13b | 2672 | "EXT4-fs: Quota file not on filesystem root. " |
ac27a0ec DK |
2673 | "Journalled quota will not work.\n"); |
2674 | path_release(&nd); | |
2675 | return vfs_quota_on(sb, type, format_id, path); | |
2676 | } | |
2677 | ||
2678 | /* Read data from quotafile - avoid pagecache and such because we cannot afford | |
2679 | * acquiring the locks... As quota files are never truncated and quota code | |
2680 | * itself serializes the operations (and noone else should touch the files) | |
2681 | * we don't have to be afraid of races */ | |
617ba13b | 2682 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, |
ac27a0ec DK |
2683 | size_t len, loff_t off) |
2684 | { | |
2685 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
617ba13b | 2686 | sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
2687 | int err = 0; |
2688 | int offset = off & (sb->s_blocksize - 1); | |
2689 | int tocopy; | |
2690 | size_t toread; | |
2691 | struct buffer_head *bh; | |
2692 | loff_t i_size = i_size_read(inode); | |
2693 | ||
2694 | if (off > i_size) | |
2695 | return 0; | |
2696 | if (off+len > i_size) | |
2697 | len = i_size-off; | |
2698 | toread = len; | |
2699 | while (toread > 0) { | |
2700 | tocopy = sb->s_blocksize - offset < toread ? | |
2701 | sb->s_blocksize - offset : toread; | |
617ba13b | 2702 | bh = ext4_bread(NULL, inode, blk, 0, &err); |
ac27a0ec DK |
2703 | if (err) |
2704 | return err; | |
2705 | if (!bh) /* A hole? */ | |
2706 | memset(data, 0, tocopy); | |
2707 | else | |
2708 | memcpy(data, bh->b_data+offset, tocopy); | |
2709 | brelse(bh); | |
2710 | offset = 0; | |
2711 | toread -= tocopy; | |
2712 | data += tocopy; | |
2713 | blk++; | |
2714 | } | |
2715 | return len; | |
2716 | } | |
2717 | ||
2718 | /* Write to quotafile (we know the transaction is already started and has | |
2719 | * enough credits) */ | |
617ba13b | 2720 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
2721 | const char *data, size_t len, loff_t off) |
2722 | { | |
2723 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
617ba13b | 2724 | sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
2725 | int err = 0; |
2726 | int offset = off & (sb->s_blocksize - 1); | |
2727 | int tocopy; | |
617ba13b | 2728 | int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL; |
ac27a0ec DK |
2729 | size_t towrite = len; |
2730 | struct buffer_head *bh; | |
2731 | handle_t *handle = journal_current_handle(); | |
2732 | ||
2733 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); | |
2734 | while (towrite > 0) { | |
2735 | tocopy = sb->s_blocksize - offset < towrite ? | |
2736 | sb->s_blocksize - offset : towrite; | |
617ba13b | 2737 | bh = ext4_bread(handle, inode, blk, 1, &err); |
ac27a0ec DK |
2738 | if (!bh) |
2739 | goto out; | |
2740 | if (journal_quota) { | |
617ba13b | 2741 | err = ext4_journal_get_write_access(handle, bh); |
ac27a0ec DK |
2742 | if (err) { |
2743 | brelse(bh); | |
2744 | goto out; | |
2745 | } | |
2746 | } | |
2747 | lock_buffer(bh); | |
2748 | memcpy(bh->b_data+offset, data, tocopy); | |
2749 | flush_dcache_page(bh->b_page); | |
2750 | unlock_buffer(bh); | |
2751 | if (journal_quota) | |
617ba13b | 2752 | err = ext4_journal_dirty_metadata(handle, bh); |
ac27a0ec DK |
2753 | else { |
2754 | /* Always do at least ordered writes for quotas */ | |
617ba13b | 2755 | err = ext4_journal_dirty_data(handle, bh); |
ac27a0ec DK |
2756 | mark_buffer_dirty(bh); |
2757 | } | |
2758 | brelse(bh); | |
2759 | if (err) | |
2760 | goto out; | |
2761 | offset = 0; | |
2762 | towrite -= tocopy; | |
2763 | data += tocopy; | |
2764 | blk++; | |
2765 | } | |
2766 | out: | |
2767 | if (len == towrite) | |
2768 | return err; | |
2769 | if (inode->i_size < off+len-towrite) { | |
2770 | i_size_write(inode, off+len-towrite); | |
617ba13b | 2771 | EXT4_I(inode)->i_disksize = inode->i_size; |
ac27a0ec DK |
2772 | } |
2773 | inode->i_version++; | |
2774 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
617ba13b | 2775 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec DK |
2776 | mutex_unlock(&inode->i_mutex); |
2777 | return len - towrite; | |
2778 | } | |
2779 | ||
2780 | #endif | |
2781 | ||
617ba13b | 2782 | static int ext4_get_sb(struct file_system_type *fs_type, |
ac27a0ec DK |
2783 | int flags, const char *dev_name, void *data, struct vfsmount *mnt) |
2784 | { | |
617ba13b | 2785 | return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt); |
ac27a0ec DK |
2786 | } |
2787 | ||
617ba13b | 2788 | static struct file_system_type ext4dev_fs_type = { |
ac27a0ec | 2789 | .owner = THIS_MODULE, |
617ba13b MC |
2790 | .name = "ext4dev", |
2791 | .get_sb = ext4_get_sb, | |
ac27a0ec DK |
2792 | .kill_sb = kill_block_super, |
2793 | .fs_flags = FS_REQUIRES_DEV, | |
2794 | }; | |
2795 | ||
617ba13b | 2796 | static int __init init_ext4_fs(void) |
ac27a0ec | 2797 | { |
617ba13b | 2798 | int err = init_ext4_xattr(); |
ac27a0ec DK |
2799 | if (err) |
2800 | return err; | |
2801 | err = init_inodecache(); | |
2802 | if (err) | |
2803 | goto out1; | |
617ba13b | 2804 | err = register_filesystem(&ext4dev_fs_type); |
ac27a0ec DK |
2805 | if (err) |
2806 | goto out; | |
2807 | return 0; | |
2808 | out: | |
2809 | destroy_inodecache(); | |
2810 | out1: | |
617ba13b | 2811 | exit_ext4_xattr(); |
ac27a0ec DK |
2812 | return err; |
2813 | } | |
2814 | ||
617ba13b | 2815 | static void __exit exit_ext4_fs(void) |
ac27a0ec | 2816 | { |
617ba13b | 2817 | unregister_filesystem(&ext4dev_fs_type); |
ac27a0ec | 2818 | destroy_inodecache(); |
617ba13b | 2819 | exit_ext4_xattr(); |
ac27a0ec DK |
2820 | } |
2821 | ||
2822 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
617ba13b | 2823 | MODULE_DESCRIPTION("Fourth Extended Filesystem with extents"); |
ac27a0ec | 2824 | MODULE_LICENSE("GPL"); |
617ba13b MC |
2825 | module_init(init_ext4_fs) |
2826 | module_exit(exit_ext4_fs) |