?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[PARISC] fix section mismatches in arch/parisc/kernel
[linux-2.6/mini2440.git] / fs / ext3 / super.c
blob6e3062913a92f52544a35169fb1ab6e9a44e7831
1 /*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
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
45 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
46 unsigned long journal_devnum);
47 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
48 unsigned int);
49 static void ext3_commit_super (struct super_block * sb,
50 struct ext3_super_block * es,
51 int sync);
52 static void ext3_mark_recovery_complete(struct super_block * sb,
53 struct ext3_super_block * es);
54 static void ext3_clear_journal_err(struct super_block * sb,
55 struct ext3_super_block * es);
56 static int ext3_sync_fs(struct super_block *sb, int wait);
57 static const char *ext3_decode_error(struct super_block * sb, int errno,
58 char nbuf[16]);
59 static int ext3_remount (struct super_block * sb, int * flags, char * data);
60 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
61 static void ext3_unlockfs(struct super_block *sb);
62 static void ext3_write_super (struct super_block * sb);
63 static void ext3_write_super_lockfs(struct super_block *sb);
64
65 /*
66 * Wrappers for journal_start/end.
67 *
68 * The only special thing we need to do here is to make sure that all
69 * journal_end calls result in the superblock being marked dirty, so
70 * that sync() will call the filesystem's write_super callback if
71 * appropriate.
72 */
73 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
74 {
75 journal_t *journal;
76
77 if (sb->s_flags & MS_RDONLY)
78 return ERR_PTR(-EROFS);
79
80 /* Special case here: if the journal has aborted behind our
81 * backs (eg. EIO in the commit thread), then we still need to
82 * take the FS itself readonly cleanly. */
83 journal = EXT3_SB(sb)->s_journal;
84 if (is_journal_aborted(journal)) {
85 ext3_abort(sb, __FUNCTION__,
86 "Detected aborted journal");
87 return ERR_PTR(-EROFS);
88 }
89
90 return journal_start(journal, nblocks);
91 }
92
93 /*
94 * The only special thing we need to do here is to make sure that all
95 * journal_stop calls result in the superblock being marked dirty, so
96 * that sync() will call the filesystem's write_super callback if
97 * appropriate.
98 */
99 int __ext3_journal_stop(const char *where, handle_t *handle)
100 {
101 struct super_block *sb;
102 int err;
103 int rc;
104
105 sb = handle->h_transaction->t_journal->j_private;
106 err = handle->h_err;
107 rc = journal_stop(handle);
108
109 if (!err)
110 err = rc;
111 if (err)
112 __ext3_std_error(sb, where, err);
113 return err;
114 }
115
116 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
117 struct buffer_head *bh, handle_t *handle, int err)
118 {
119 char nbuf[16];
120 const char *errstr = ext3_decode_error(NULL, err, nbuf);
121
122 if (bh)
123 BUFFER_TRACE(bh, "abort");
124
125 if (!handle->h_err)
126 handle->h_err = err;
127
128 if (is_handle_aborted(handle))
129 return;
130
131 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
132 caller, errstr, err_fn);
133
134 journal_abort_handle(handle);
135 }
136
137 /* Deal with the reporting of failure conditions on a filesystem such as
138 * inconsistencies detected or read IO failures.
139 *
140 * On ext2, we can store the error state of the filesystem in the
141 * superblock. That is not possible on ext3, because we may have other
142 * write ordering constraints on the superblock which prevent us from
143 * writing it out straight away; and given that the journal is about to
144 * be aborted, we can't rely on the current, or future, transactions to
145 * write out the superblock safely.
146 *
147 * We'll just use the journal_abort() error code to record an error in
148 * the journal instead. On recovery, the journal will compain about
149 * that error until we've noted it down and cleared it.
150 */
151
152 static void ext3_handle_error(struct super_block *sb)
153 {
154 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
155
156 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
157 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
158
159 if (sb->s_flags & MS_RDONLY)
160 return;
161
162 if (!test_opt (sb, ERRORS_CONT)) {
163 journal_t *journal = EXT3_SB(sb)->s_journal;
164
165 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
166 if (journal)
167 journal_abort(journal, -EIO);
168 }
169 if (test_opt (sb, ERRORS_RO)) {
170 printk (KERN_CRIT "Remounting filesystem read-only\n");
171 sb->s_flags |= MS_RDONLY;
172 }
173 ext3_commit_super(sb, es, 1);
174 if (test_opt(sb, ERRORS_PANIC))
175 panic("EXT3-fs (device %s): panic forced after error\n",
176 sb->s_id);
177 }
178
179 void ext3_error (struct super_block * sb, const char * function,
180 const char * fmt, ...)
181 {
182 va_list args;
183
184 va_start(args, fmt);
185 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
186 vprintk(fmt, args);
187 printk("\n");
188 va_end(args);
189
190 ext3_handle_error(sb);
191 }
192
193 static const char *ext3_decode_error(struct super_block * sb, int errno,
194 char nbuf[16])
195 {
196 char *errstr = NULL;
197
198 switch (errno) {
199 case -EIO:
200 errstr = "IO failure";
201 break;
202 case -ENOMEM:
203 errstr = "Out of memory";
204 break;
205 case -EROFS:
206 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
207 errstr = "Journal has aborted";
208 else
209 errstr = "Readonly filesystem";
210 break;
211 default:
212 /* If the caller passed in an extra buffer for unknown
213 * errors, textualise them now. Else we just return
214 * NULL. */
215 if (nbuf) {
216 /* Check for truncated error codes... */
217 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
218 errstr = nbuf;
219 }
220 break;
221 }
222
223 return errstr;
224 }
225
226 /* __ext3_std_error decodes expected errors from journaling functions
227 * automatically and invokes the appropriate error response. */
228
229 void __ext3_std_error (struct super_block * sb, const char * function,
230 int errno)
231 {
232 char nbuf[16];
233 const char *errstr;
234
235 /* Special case: if the error is EROFS, and we're not already
236 * inside a transaction, then there's really no point in logging
237 * an error. */
238 if (errno == -EROFS && journal_current_handle() == NULL &&
239 (sb->s_flags & MS_RDONLY))
240 return;
241
242 errstr = ext3_decode_error(sb, errno, nbuf);
243 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
244 sb->s_id, function, errstr);
245
246 ext3_handle_error(sb);
247 }
248
249 /*
250 * ext3_abort is a much stronger failure handler than ext3_error. The
251 * abort function may be used to deal with unrecoverable failures such
252 * as journal IO errors or ENOMEM at a critical moment in log management.
253 *
254 * We unconditionally force the filesystem into an ABORT|READONLY state,
255 * unless the error response on the fs has been set to panic in which
256 * case we take the easy way out and panic immediately.
257 */
258
259 void ext3_abort (struct super_block * sb, const char * function,
260 const char * fmt, ...)
261 {
262 va_list args;
263
264 printk (KERN_CRIT "ext3_abort called.\n");
265
266 va_start(args, fmt);
267 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
268 vprintk(fmt, args);
269 printk("\n");
270 va_end(args);
271
272 if (test_opt(sb, ERRORS_PANIC))
273 panic("EXT3-fs panic from previous error\n");
274
275 if (sb->s_flags & MS_RDONLY)
276 return;
277
278 printk(KERN_CRIT "Remounting filesystem read-only\n");
279 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
280 sb->s_flags |= MS_RDONLY;
281 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
282 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
283 }
284
285 void ext3_warning (struct super_block * sb, const char * function,
286 const char * fmt, ...)
287 {
288 va_list args;
289
290 va_start(args, fmt);
291 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
292 sb->s_id, function);
293 vprintk(fmt, args);
294 printk("\n");
295 va_end(args);
296 }
297
298 void ext3_update_dynamic_rev(struct super_block *sb)
299 {
300 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
301
302 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
303 return;
304
305 ext3_warning(sb, __FUNCTION__,
306 "updating to rev %d because of new feature flag, "
307 "running e2fsck is recommended",
308 EXT3_DYNAMIC_REV);
309
310 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
311 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
312 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
313 /* leave es->s_feature_*compat flags alone */
314 /* es->s_uuid will be set by e2fsck if empty */
315
316 /*
317 * The rest of the superblock fields should be zero, and if not it
318 * means they are likely already in use, so leave them alone. We
319 * can leave it up to e2fsck to clean up any inconsistencies there.
320 */
321 }
322
323 /*
324 * Open the external journal device
325 */
326 static struct block_device *ext3_blkdev_get(dev_t dev)
327 {
328 struct block_device *bdev;
329 char b[BDEVNAME_SIZE];
330
331 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
332 if (IS_ERR(bdev))
333 goto fail;
334 return bdev;
335
336 fail:
337 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
338 __bdevname(dev, b), PTR_ERR(bdev));
339 return NULL;
340 }
341
342 /*
343 * Release the journal device
344 */
345 static int ext3_blkdev_put(struct block_device *bdev)
346 {
347 bd_release(bdev);
348 return blkdev_put(bdev);
349 }
350
351 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
352 {
353 struct block_device *bdev;
354 int ret = -ENODEV;
355
356 bdev = sbi->journal_bdev;
357 if (bdev) {
358 ret = ext3_blkdev_put(bdev);
359 sbi->journal_bdev = NULL;
360 }
361 return ret;
362 }
363
364 static inline struct inode *orphan_list_entry(struct list_head *l)
365 {
366 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
367 }
368
369 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
370 {
371 struct list_head *l;
372
373 printk(KERN_ERR "sb orphan head is %d\n",
374 le32_to_cpu(sbi->s_es->s_last_orphan));
375
376 printk(KERN_ERR "sb_info orphan list:\n");
377 list_for_each(l, &sbi->s_orphan) {
378 struct inode *inode = orphan_list_entry(l);
379 printk(KERN_ERR " "
380 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
381 inode->i_sb->s_id, inode->i_ino, inode,
382 inode->i_mode, inode->i_nlink,
383 NEXT_ORPHAN(inode));
384 }
385 }
386
387 static void ext3_put_super (struct super_block * sb)
388 {
389 struct ext3_sb_info *sbi = EXT3_SB(sb);
390 struct ext3_super_block *es = sbi->s_es;
391 int i;
392
393 ext3_xattr_put_super(sb);
394 journal_destroy(sbi->s_journal);
395 if (!(sb->s_flags & MS_RDONLY)) {
396 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
397 es->s_state = cpu_to_le16(sbi->s_mount_state);
398 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
399 mark_buffer_dirty(sbi->s_sbh);
400 ext3_commit_super(sb, es, 1);
401 }
402
403 for (i = 0; i < sbi->s_gdb_count; i++)
404 brelse(sbi->s_group_desc[i]);
405 kfree(sbi->s_group_desc);
406 percpu_counter_destroy(&sbi->s_freeblocks_counter);
407 percpu_counter_destroy(&sbi->s_freeinodes_counter);
408 percpu_counter_destroy(&sbi->s_dirs_counter);
409 brelse(sbi->s_sbh);
410 #ifdef CONFIG_QUOTA
411 for (i = 0; i < MAXQUOTAS; i++)
412 kfree(sbi->s_qf_names[i]);
413 #endif
414
415 /* Debugging code just in case the in-memory inode orphan list
416 * isn't empty. The on-disk one can be non-empty if we've
417 * detected an error and taken the fs readonly, but the
418 * in-memory list had better be clean by this point. */
419 if (!list_empty(&sbi->s_orphan))
420 dump_orphan_list(sb, sbi);
421 J_ASSERT(list_empty(&sbi->s_orphan));
422
423 invalidate_bdev(sb->s_bdev);
424 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
425 /*
426 * Invalidate the journal device's buffers. We don't want them
427 * floating about in memory - the physical journal device may
428 * hotswapped, and it breaks the `ro-after' testing code.
429 */
430 sync_blockdev(sbi->journal_bdev);
431 invalidate_bdev(sbi->journal_bdev);
432 ext3_blkdev_remove(sbi);
433 }
434 sb->s_fs_info = NULL;
435 kfree(sbi);
436 return;
437 }
438
439 static struct kmem_cache *ext3_inode_cachep;
440
441 /*
442 * Called inside transaction, so use GFP_NOFS
443 */
444 static struct inode *ext3_alloc_inode(struct super_block *sb)
445 {
446 struct ext3_inode_info *ei;
447
448 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
449 if (!ei)
450 return NULL;
451 #ifdef CONFIG_EXT3_FS_POSIX_ACL
452 ei->i_acl = EXT3_ACL_NOT_CACHED;
453 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
454 #endif
455 ei->i_block_alloc_info = NULL;
456 ei->vfs_inode.i_version = 1;
457 return &ei->vfs_inode;
458 }
459
460 static void ext3_destroy_inode(struct inode *inode)
461 {
462 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
463 }
464
465 static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags)
466 {
467 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
468
469 INIT_LIST_HEAD(&ei->i_orphan);
470 #ifdef CONFIG_EXT3_FS_XATTR
471 init_rwsem(&ei->xattr_sem);
472 #endif
473 mutex_init(&ei->truncate_mutex);
474 inode_init_once(&ei->vfs_inode);
475 }
476
477 static int init_inodecache(void)
478 {
479 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
480 sizeof(struct ext3_inode_info),
481 0, (SLAB_RECLAIM_ACCOUNT|
482 SLAB_MEM_SPREAD),
483 init_once, NULL);
484 if (ext3_inode_cachep == NULL)
485 return -ENOMEM;
486 return 0;
487 }
488
489 static void destroy_inodecache(void)
490 {
491 kmem_cache_destroy(ext3_inode_cachep);
492 }
493
494 static void ext3_clear_inode(struct inode *inode)
495 {
496 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
497 #ifdef CONFIG_EXT3_FS_POSIX_ACL
498 if (EXT3_I(inode)->i_acl &&
499 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
500 posix_acl_release(EXT3_I(inode)->i_acl);
501 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
502 }
503 if (EXT3_I(inode)->i_default_acl &&
504 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
505 posix_acl_release(EXT3_I(inode)->i_default_acl);
506 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
507 }
508 #endif
509 ext3_discard_reservation(inode);
510 EXT3_I(inode)->i_block_alloc_info = NULL;
511 if (unlikely(rsv))
512 kfree(rsv);
513 }
514
515 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
516 {
517 #if defined(CONFIG_QUOTA)
518 struct ext3_sb_info *sbi = EXT3_SB(sb);
519
520 if (sbi->s_jquota_fmt)
521 seq_printf(seq, ",jqfmt=%s",
522 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
523
524 if (sbi->s_qf_names[USRQUOTA])
525 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
526
527 if (sbi->s_qf_names[GRPQUOTA])
528 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
529
530 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
531 seq_puts(seq, ",usrquota");
532
533 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
534 seq_puts(seq, ",grpquota");
535 #endif
536 }
537
538 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
539 {
540 struct super_block *sb = vfs->mnt_sb;
541
542 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
543 seq_puts(seq, ",data=journal");
544 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
545 seq_puts(seq, ",data=ordered");
546 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
547 seq_puts(seq, ",data=writeback");
548
549 ext3_show_quota_options(seq, sb);
550
551 return 0;
552 }
553
554
555 static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
556 {
557 __u32 *objp = vobjp;
558 unsigned long ino = objp[0];
559 __u32 generation = objp[1];
560 struct inode *inode;
561 struct dentry *result;
562
563 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
564 return ERR_PTR(-ESTALE);
565 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
566 return ERR_PTR(-ESTALE);
567
568 /* iget isn't really right if the inode is currently unallocated!!
569 *
570 * ext3_read_inode will return a bad_inode if the inode had been
571 * deleted, so we should be safe.
572 *
573 * Currently we don't know the generation for parent directory, so
574 * a generation of 0 means "accept any"
575 */
576 inode = iget(sb, ino);
577 if (inode == NULL)
578 return ERR_PTR(-ENOMEM);
579 if (is_bad_inode(inode) ||
580 (generation && inode->i_generation != generation)) {
581 iput(inode);
582 return ERR_PTR(-ESTALE);
583 }
584 /* now to find a dentry.
585 * If possible, get a well-connected one
586 */
587 result = d_alloc_anon(inode);
588 if (!result) {
589 iput(inode);
590 return ERR_PTR(-ENOMEM);
591 }
592 return result;
593 }
594
595 #ifdef CONFIG_QUOTA
596 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
597 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
598
599 static int ext3_dquot_initialize(struct inode *inode, int type);
600 static int ext3_dquot_drop(struct inode *inode);
601 static int ext3_write_dquot(struct dquot *dquot);
602 static int ext3_acquire_dquot(struct dquot *dquot);
603 static int ext3_release_dquot(struct dquot *dquot);
604 static int ext3_mark_dquot_dirty(struct dquot *dquot);
605 static int ext3_write_info(struct super_block *sb, int type);
606 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
607 static int ext3_quota_on_mount(struct super_block *sb, int type);
608 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
609 size_t len, loff_t off);
610 static ssize_t ext3_quota_write(struct super_block *sb, int type,
611 const char *data, size_t len, loff_t off);
612
613 static struct dquot_operations ext3_quota_operations = {
614 .initialize = ext3_dquot_initialize,
615 .drop = ext3_dquot_drop,
616 .alloc_space = dquot_alloc_space,
617 .alloc_inode = dquot_alloc_inode,
618 .free_space = dquot_free_space,
619 .free_inode = dquot_free_inode,
620 .transfer = dquot_transfer,
621 .write_dquot = ext3_write_dquot,
622 .acquire_dquot = ext3_acquire_dquot,
623 .release_dquot = ext3_release_dquot,
624 .mark_dirty = ext3_mark_dquot_dirty,
625 .write_info = ext3_write_info
626 };
627
628 static struct quotactl_ops ext3_qctl_operations = {
629 .quota_on = ext3_quota_on,
630 .quota_off = vfs_quota_off,
631 .quota_sync = vfs_quota_sync,
632 .get_info = vfs_get_dqinfo,
633 .set_info = vfs_set_dqinfo,
634 .get_dqblk = vfs_get_dqblk,
635 .set_dqblk = vfs_set_dqblk
636 };
637 #endif
638
639 static const struct super_operations ext3_sops = {
640 .alloc_inode = ext3_alloc_inode,
641 .destroy_inode = ext3_destroy_inode,
642 .read_inode = ext3_read_inode,
643 .write_inode = ext3_write_inode,
644 .dirty_inode = ext3_dirty_inode,
645 .delete_inode = ext3_delete_inode,
646 .put_super = ext3_put_super,
647 .write_super = ext3_write_super,
648 .sync_fs = ext3_sync_fs,
649 .write_super_lockfs = ext3_write_super_lockfs,
650 .unlockfs = ext3_unlockfs,
651 .statfs = ext3_statfs,
652 .remount_fs = ext3_remount,
653 .clear_inode = ext3_clear_inode,
654 .show_options = ext3_show_options,
655 #ifdef CONFIG_QUOTA
656 .quota_read = ext3_quota_read,
657 .quota_write = ext3_quota_write,
658 #endif
659 };
660
661 static struct export_operations ext3_export_ops = {
662 .get_parent = ext3_get_parent,
663 .get_dentry = ext3_get_dentry,
664 };
665
666 enum {
667 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
668 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
669 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
670 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
671 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
672 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
673 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
674 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
675 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
676 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
677 Opt_grpquota
678 };
679
680 static match_table_t tokens = {
681 {Opt_bsd_df, "bsddf"},
682 {Opt_minix_df, "minixdf"},
683 {Opt_grpid, "grpid"},
684 {Opt_grpid, "bsdgroups"},
685 {Opt_nogrpid, "nogrpid"},
686 {Opt_nogrpid, "sysvgroups"},
687 {Opt_resgid, "resgid=%u"},
688 {Opt_resuid, "resuid=%u"},
689 {Opt_sb, "sb=%u"},
690 {Opt_err_cont, "errors=continue"},
691 {Opt_err_panic, "errors=panic"},
692 {Opt_err_ro, "errors=remount-ro"},
693 {Opt_nouid32, "nouid32"},
694 {Opt_nocheck, "nocheck"},
695 {Opt_nocheck, "check=none"},
696 {Opt_debug, "debug"},
697 {Opt_oldalloc, "oldalloc"},
698 {Opt_orlov, "orlov"},
699 {Opt_user_xattr, "user_xattr"},
700 {Opt_nouser_xattr, "nouser_xattr"},
701 {Opt_acl, "acl"},
702 {Opt_noacl, "noacl"},
703 {Opt_reservation, "reservation"},
704 {Opt_noreservation, "noreservation"},
705 {Opt_noload, "noload"},
706 {Opt_nobh, "nobh"},
707 {Opt_bh, "bh"},
708 {Opt_commit, "commit=%u"},
709 {Opt_journal_update, "journal=update"},
710 {Opt_journal_inum, "journal=%u"},
711 {Opt_journal_dev, "journal_dev=%u"},
712 {Opt_abort, "abort"},
713 {Opt_data_journal, "data=journal"},
714 {Opt_data_ordered, "data=ordered"},
715 {Opt_data_writeback, "data=writeback"},
716 {Opt_offusrjquota, "usrjquota="},
717 {Opt_usrjquota, "usrjquota=%s"},
718 {Opt_offgrpjquota, "grpjquota="},
719 {Opt_grpjquota, "grpjquota=%s"},
720 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
721 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
722 {Opt_grpquota, "grpquota"},
723 {Opt_noquota, "noquota"},
724 {Opt_quota, "quota"},
725 {Opt_usrquota, "usrquota"},
726 {Opt_barrier, "barrier=%u"},
727 {Opt_err, NULL},
728 {Opt_resize, "resize"},
729 };
730
731 static ext3_fsblk_t get_sb_block(void **data)
732 {
733 ext3_fsblk_t sb_block;
734 char *options = (char *) *data;
735
736 if (!options || strncmp(options, "sb=", 3) != 0)
737 return 1; /* Default location */
738 options += 3;
739 /*todo: use simple_strtoll with >32bit ext3 */
740 sb_block = simple_strtoul(options, &options, 0);
741 if (*options && *options != ',') {
742 printk("EXT3-fs: Invalid sb specification: %s\n",
743 (char *) *data);
744 return 1;
745 }
746 if (*options == ',')
747 options++;
748 *data = (void *) options;
749 return sb_block;
750 }
751
752 static int parse_options (char *options, struct super_block *sb,
753 unsigned int *inum, unsigned long *journal_devnum,
754 ext3_fsblk_t *n_blocks_count, int is_remount)
755 {
756 struct ext3_sb_info *sbi = EXT3_SB(sb);
757 char * p;
758 substring_t args[MAX_OPT_ARGS];
759 int data_opt = 0;
760 int option;
761 #ifdef CONFIG_QUOTA
762 int qtype;
763 char *qname;
764 #endif
765
766 if (!options)
767 return 1;
768
769 while ((p = strsep (&options, ",")) != NULL) {
770 int token;
771 if (!*p)
772 continue;
773
774 token = match_token(p, tokens, args);
775 switch (token) {
776 case Opt_bsd_df:
777 clear_opt (sbi->s_mount_opt, MINIX_DF);
778 break;
779 case Opt_minix_df:
780 set_opt (sbi->s_mount_opt, MINIX_DF);
781 break;
782 case Opt_grpid:
783 set_opt (sbi->s_mount_opt, GRPID);
784 break;
785 case Opt_nogrpid:
786 clear_opt (sbi->s_mount_opt, GRPID);
787 break;
788 case Opt_resuid:
789 if (match_int(&args[0], &option))
790 return 0;
791 sbi->s_resuid = option;
792 break;
793 case Opt_resgid:
794 if (match_int(&args[0], &option))
795 return 0;
796 sbi->s_resgid = option;
797 break;
798 case Opt_sb:
799 /* handled by get_sb_block() instead of here */
800 /* *sb_block = match_int(&args[0]); */
801 break;
802 case Opt_err_panic:
803 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
804 clear_opt (sbi->s_mount_opt, ERRORS_RO);
805 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
806 break;
807 case Opt_err_ro:
808 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
809 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
810 set_opt (sbi->s_mount_opt, ERRORS_RO);
811 break;
812 case Opt_err_cont:
813 clear_opt (sbi->s_mount_opt, ERRORS_RO);
814 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
815 set_opt (sbi->s_mount_opt, ERRORS_CONT);
816 break;
817 case Opt_nouid32:
818 set_opt (sbi->s_mount_opt, NO_UID32);
819 break;
820 case Opt_nocheck:
821 clear_opt (sbi->s_mount_opt, CHECK);
822 break;
823 case Opt_debug:
824 set_opt (sbi->s_mount_opt, DEBUG);
825 break;
826 case Opt_oldalloc:
827 set_opt (sbi->s_mount_opt, OLDALLOC);
828 break;
829 case Opt_orlov:
830 clear_opt (sbi->s_mount_opt, OLDALLOC);
831 break;
832 #ifdef CONFIG_EXT3_FS_XATTR
833 case Opt_user_xattr:
834 set_opt (sbi->s_mount_opt, XATTR_USER);
835 break;
836 case Opt_nouser_xattr:
837 clear_opt (sbi->s_mount_opt, XATTR_USER);
838 break;
839 #else
840 case Opt_user_xattr:
841 case Opt_nouser_xattr:
842 printk("EXT3 (no)user_xattr options not supported\n");
843 break;
844 #endif
845 #ifdef CONFIG_EXT3_FS_POSIX_ACL
846 case Opt_acl:
847 set_opt(sbi->s_mount_opt, POSIX_ACL);
848 break;
849 case Opt_noacl:
850 clear_opt(sbi->s_mount_opt, POSIX_ACL);
851 break;
852 #else
853 case Opt_acl:
854 case Opt_noacl:
855 printk("EXT3 (no)acl options not supported\n");
856 break;
857 #endif
858 case Opt_reservation:
859 set_opt(sbi->s_mount_opt, RESERVATION);
860 break;
861 case Opt_noreservation:
862 clear_opt(sbi->s_mount_opt, RESERVATION);
863 break;
864 case Opt_journal_update:
865 /* @@@ FIXME */
866 /* Eventually we will want to be able to create
867 a journal file here. For now, only allow the
868 user to specify an existing inode to be the
869 journal file. */
870 if (is_remount) {
871 printk(KERN_ERR "EXT3-fs: cannot specify "
872 "journal on remount\n");
873 return 0;
874 }
875 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
876 break;
877 case Opt_journal_inum:
878 if (is_remount) {
879 printk(KERN_ERR "EXT3-fs: cannot specify "
880 "journal on remount\n");
881 return 0;
882 }
883 if (match_int(&args[0], &option))
884 return 0;
885 *inum = option;
886 break;
887 case Opt_journal_dev:
888 if (is_remount) {
889 printk(KERN_ERR "EXT3-fs: cannot specify "
890 "journal on remount\n");
891 return 0;
892 }
893 if (match_int(&args[0], &option))
894 return 0;
895 *journal_devnum = option;
896 break;
897 case Opt_noload:
898 set_opt (sbi->s_mount_opt, NOLOAD);
899 break;
900 case Opt_commit:
901 if (match_int(&args[0], &option))
902 return 0;
903 if (option < 0)
904 return 0;
905 if (option == 0)
906 option = JBD_DEFAULT_MAX_COMMIT_AGE;
907 sbi->s_commit_interval = HZ * option;
908 break;
909 case Opt_data_journal:
910 data_opt = EXT3_MOUNT_JOURNAL_DATA;
911 goto datacheck;
912 case Opt_data_ordered:
913 data_opt = EXT3_MOUNT_ORDERED_DATA;
914 goto datacheck;
915 case Opt_data_writeback:
916 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
917 datacheck:
918 if (is_remount) {
919 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
920 != data_opt) {
921 printk(KERN_ERR
922 "EXT3-fs: cannot change data "
923 "mode on remount\n");
924 return 0;
925 }
926 } else {
927 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
928 sbi->s_mount_opt |= data_opt;
929 }
930 break;
931 #ifdef CONFIG_QUOTA
932 case Opt_usrjquota:
933 qtype = USRQUOTA;
934 goto set_qf_name;
935 case Opt_grpjquota:
936 qtype = GRPQUOTA;
937 set_qf_name:
938 if (sb_any_quota_enabled(sb)) {
939 printk(KERN_ERR
940 "EXT3-fs: Cannot change journalled "
941 "quota options when quota turned on.\n");
942 return 0;
943 }
944 qname = match_strdup(&args[0]);
945 if (!qname) {
946 printk(KERN_ERR
947 "EXT3-fs: not enough memory for "
948 "storing quotafile name.\n");
949 return 0;
950 }
951 if (sbi->s_qf_names[qtype] &&
952 strcmp(sbi->s_qf_names[qtype], qname)) {
953 printk(KERN_ERR
954 "EXT3-fs: %s quota file already "
955 "specified.\n", QTYPE2NAME(qtype));
956 kfree(qname);
957 return 0;
958 }
959 sbi->s_qf_names[qtype] = qname;
960 if (strchr(sbi->s_qf_names[qtype], '/')) {
961 printk(KERN_ERR
962 "EXT3-fs: quotafile must be on "
963 "filesystem root.\n");
964 kfree(sbi->s_qf_names[qtype]);
965 sbi->s_qf_names[qtype] = NULL;
966 return 0;
967 }
968 set_opt(sbi->s_mount_opt, QUOTA);
969 break;
970 case Opt_offusrjquota:
971 qtype = USRQUOTA;
972 goto clear_qf_name;
973 case Opt_offgrpjquota:
974 qtype = GRPQUOTA;
975 clear_qf_name:
976 if (sb_any_quota_enabled(sb)) {
977 printk(KERN_ERR "EXT3-fs: Cannot change "
978 "journalled quota options when "
979 "quota turned on.\n");
980 return 0;
981 }
982 /*
983 * The space will be released later when all options
984 * are confirmed to be correct
985 */
986 sbi->s_qf_names[qtype] = NULL;
987 break;
988 case Opt_jqfmt_vfsold:
989 sbi->s_jquota_fmt = QFMT_VFS_OLD;
990 break;
991 case Opt_jqfmt_vfsv0:
992 sbi->s_jquota_fmt = QFMT_VFS_V0;
993 break;
994 case Opt_quota:
995 case Opt_usrquota:
996 set_opt(sbi->s_mount_opt, QUOTA);
997 set_opt(sbi->s_mount_opt, USRQUOTA);
998 break;
999 case Opt_grpquota:
1000 set_opt(sbi->s_mount_opt, QUOTA);
1001 set_opt(sbi->s_mount_opt, GRPQUOTA);
1002 break;
1003 case Opt_noquota:
1004 if (sb_any_quota_enabled(sb)) {
1005 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1006 "options when quota turned on.\n");
1007 return 0;
1008 }
1009 clear_opt(sbi->s_mount_opt, QUOTA);
1010 clear_opt(sbi->s_mount_opt, USRQUOTA);
1011 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1012 break;
1013 #else
1014 case Opt_quota:
1015 case Opt_usrquota:
1016 case Opt_grpquota:
1017 case Opt_usrjquota:
1018 case Opt_grpjquota:
1019 case Opt_offusrjquota:
1020 case Opt_offgrpjquota:
1021 case Opt_jqfmt_vfsold:
1022 case Opt_jqfmt_vfsv0:
1023 printk(KERN_ERR
1024 "EXT3-fs: journalled quota options not "
1025 "supported.\n");
1026 break;
1027 case Opt_noquota:
1028 break;
1029 #endif
1030 case Opt_abort:
1031 set_opt(sbi->s_mount_opt, ABORT);
1032 break;
1033 case Opt_barrier:
1034 if (match_int(&args[0], &option))
1035 return 0;
1036 if (option)
1037 set_opt(sbi->s_mount_opt, BARRIER);
1038 else
1039 clear_opt(sbi->s_mount_opt, BARRIER);
1040 break;
1041 case Opt_ignore:
1042 break;
1043 case Opt_resize:
1044 if (!is_remount) {
1045 printk("EXT3-fs: resize option only available "
1046 "for remount\n");
1047 return 0;
1048 }
1049 if (match_int(&args[0], &option) != 0)
1050 return 0;
1051 *n_blocks_count = option;
1052 break;
1053 case Opt_nobh:
1054 set_opt(sbi->s_mount_opt, NOBH);
1055 break;
1056 case Opt_bh:
1057 clear_opt(sbi->s_mount_opt, NOBH);
1058 break;
1059 default:
1060 printk (KERN_ERR
1061 "EXT3-fs: Unrecognized mount option \"%s\" "
1062 "or missing value\n", p);
1063 return 0;
1064 }
1065 }
1066 #ifdef CONFIG_QUOTA
1067 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1068 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1069 sbi->s_qf_names[USRQUOTA])
1070 clear_opt(sbi->s_mount_opt, USRQUOTA);
1071
1072 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1073 sbi->s_qf_names[GRPQUOTA])
1074 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1075
1076 if ((sbi->s_qf_names[USRQUOTA] &&
1077 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1078 (sbi->s_qf_names[GRPQUOTA] &&
1079 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1080 printk(KERN_ERR "EXT3-fs: old and new quota "
1081 "format mixing.\n");
1082 return 0;
1083 }
1084
1085 if (!sbi->s_jquota_fmt) {
1086 printk(KERN_ERR "EXT3-fs: journalled quota format "
1087 "not specified.\n");
1088 return 0;
1089 }
1090 } else {
1091 if (sbi->s_jquota_fmt) {
1092 printk(KERN_ERR "EXT3-fs: journalled quota format "
1093 "specified with no journalling "
1094 "enabled.\n");
1095 return 0;
1096 }
1097 }
1098 #endif
1099 return 1;
1100 }
1101
1102 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1103 int read_only)
1104 {
1105 struct ext3_sb_info *sbi = EXT3_SB(sb);
1106 int res = 0;
1107
1108 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1109 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1110 "forcing read-only mode\n");
1111 res = MS_RDONLY;
1112 }
1113 if (read_only)
1114 return res;
1115 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1116 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1117 "running e2fsck is recommended\n");
1118 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1119 printk (KERN_WARNING
1120 "EXT3-fs warning: mounting fs with errors, "
1121 "running e2fsck is recommended\n");
1122 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1123 le16_to_cpu(es->s_mnt_count) >=
1124 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1125 printk (KERN_WARNING
1126 "EXT3-fs warning: maximal mount count reached, "
1127 "running e2fsck is recommended\n");
1128 else if (le32_to_cpu(es->s_checkinterval) &&
1129 (le32_to_cpu(es->s_lastcheck) +
1130 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1131 printk (KERN_WARNING
1132 "EXT3-fs warning: checktime reached, "
1133 "running e2fsck is recommended\n");
1134 #if 0
1135 /* @@@ We _will_ want to clear the valid bit if we find
1136 inconsistencies, to force a fsck at reboot. But for
1137 a plain journaled filesystem we can keep it set as
1138 valid forever! :) */
1139 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1140 #endif
1141 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1142 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1143 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1144 es->s_mtime = cpu_to_le32(get_seconds());
1145 ext3_update_dynamic_rev(sb);
1146 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1147
1148 ext3_commit_super(sb, es, 1);
1149 if (test_opt(sb, DEBUG))
1150 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1151 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1152 sb->s_blocksize,
1153 sbi->s_groups_count,
1154 EXT3_BLOCKS_PER_GROUP(sb),
1155 EXT3_INODES_PER_GROUP(sb),
1156 sbi->s_mount_opt);
1157
1158 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1159 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1160 char b[BDEVNAME_SIZE];
1161
1162 printk("external journal on %s\n",
1163 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1164 } else {
1165 printk("internal journal\n");
1166 }
1167 return res;
1168 }
1169
1170 /* Called at mount-time, super-block is locked */
1171 static int ext3_check_descriptors (struct super_block * sb)
1172 {
1173 struct ext3_sb_info *sbi = EXT3_SB(sb);
1174 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1175 ext3_fsblk_t last_block;
1176 struct ext3_group_desc * gdp = NULL;
1177 int desc_block = 0;
1178 int i;
1179
1180 ext3_debug ("Checking group descriptors");
1181
1182 for (i = 0; i < sbi->s_groups_count; i++)
1183 {
1184 if (i == sbi->s_groups_count - 1)
1185 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1186 else
1187 last_block = first_block +
1188 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1189
1190 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1191 gdp = (struct ext3_group_desc *)
1192 sbi->s_group_desc[desc_block++]->b_data;
1193 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1194 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1195 {
1196 ext3_error (sb, "ext3_check_descriptors",
1197 "Block bitmap for group %d"
1198 " not in group (block %lu)!",
1199 i, (unsigned long)
1200 le32_to_cpu(gdp->bg_block_bitmap));
1201 return 0;
1202 }
1203 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1204 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1205 {
1206 ext3_error (sb, "ext3_check_descriptors",
1207 "Inode bitmap for group %d"
1208 " not in group (block %lu)!",
1209 i, (unsigned long)
1210 le32_to_cpu(gdp->bg_inode_bitmap));
1211 return 0;
1212 }
1213 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1214 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >
1215 last_block)
1216 {
1217 ext3_error (sb, "ext3_check_descriptors",
1218 "Inode table for group %d"
1219 " not in group (block %lu)!",
1220 i, (unsigned long)
1221 le32_to_cpu(gdp->bg_inode_table));
1222 return 0;
1223 }
1224 first_block += EXT3_BLOCKS_PER_GROUP(sb);
1225 gdp++;
1226 }
1227
1228 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1229 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1230 return 1;
1231 }
1232
1233
1234 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1235 * the superblock) which were deleted from all directories, but held open by
1236 * a process at the time of a crash. We walk the list and try to delete these
1237 * inodes at recovery time (only with a read-write filesystem).
1238 *
1239 * In order to keep the orphan inode chain consistent during traversal (in
1240 * case of crash during recovery), we link each inode into the superblock
1241 * orphan list_head and handle it the same way as an inode deletion during
1242 * normal operation (which journals the operations for us).
1243 *
1244 * We only do an iget() and an iput() on each inode, which is very safe if we
1245 * accidentally point at an in-use or already deleted inode. The worst that
1246 * can happen in this case is that we get a "bit already cleared" message from
1247 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1248 * e2fsck was run on this filesystem, and it must have already done the orphan
1249 * inode cleanup for us, so we can safely abort without any further action.
1250 */
1251 static void ext3_orphan_cleanup (struct super_block * sb,
1252 struct ext3_super_block * es)
1253 {
1254 unsigned int s_flags = sb->s_flags;
1255 int nr_orphans = 0, nr_truncates = 0;
1256 #ifdef CONFIG_QUOTA
1257 int i;
1258 #endif
1259 if (!es->s_last_orphan) {
1260 jbd_debug(4, "no orphan inodes to clean up\n");
1261 return;
1262 }
1263
1264 if (bdev_read_only(sb->s_bdev)) {
1265 printk(KERN_ERR "EXT3-fs: write access "
1266 "unavailable, skipping orphan cleanup.\n");
1267 return;
1268 }
1269
1270 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1271 if (es->s_last_orphan)
1272 jbd_debug(1, "Errors on filesystem, "
1273 "clearing orphan list.\n");
1274 es->s_last_orphan = 0;
1275 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1276 return;
1277 }
1278
1279 if (s_flags & MS_RDONLY) {
1280 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1281 sb->s_id);
1282 sb->s_flags &= ~MS_RDONLY;
1283 }
1284 #ifdef CONFIG_QUOTA
1285 /* Needed for iput() to work correctly and not trash data */
1286 sb->s_flags |= MS_ACTIVE;
1287 /* Turn on quotas so that they are updated correctly */
1288 for (i = 0; i < MAXQUOTAS; i++) {
1289 if (EXT3_SB(sb)->s_qf_names[i]) {
1290 int ret = ext3_quota_on_mount(sb, i);
1291 if (ret < 0)
1292 printk(KERN_ERR
1293 "EXT3-fs: Cannot turn on journalled "
1294 "quota: error %d\n", ret);
1295 }
1296 }
1297 #endif
1298
1299 while (es->s_last_orphan) {
1300 struct inode *inode;
1301
1302 if (!(inode =
1303 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1304 es->s_last_orphan = 0;
1305 break;
1306 }
1307
1308 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1309 DQUOT_INIT(inode);
1310 if (inode->i_nlink) {
1311 printk(KERN_DEBUG
1312 "%s: truncating inode %lu to %Ld bytes\n",
1313 __FUNCTION__, inode->i_ino, inode->i_size);
1314 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1315 inode->i_ino, inode->i_size);
1316 ext3_truncate(inode);
1317 nr_truncates++;
1318 } else {
1319 printk(KERN_DEBUG
1320 "%s: deleting unreferenced inode %lu\n",
1321 __FUNCTION__, inode->i_ino);
1322 jbd_debug(2, "deleting unreferenced inode %lu\n",
1323 inode->i_ino);
1324 nr_orphans++;
1325 }
1326 iput(inode); /* The delete magic happens here! */
1327 }
1328
1329 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1330
1331 if (nr_orphans)
1332 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1333 sb->s_id, PLURAL(nr_orphans));
1334 if (nr_truncates)
1335 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1336 sb->s_id, PLURAL(nr_truncates));
1337 #ifdef CONFIG_QUOTA
1338 /* Turn quotas off */
1339 for (i = 0; i < MAXQUOTAS; i++) {
1340 if (sb_dqopt(sb)->files[i])
1341 vfs_quota_off(sb, i);
1342 }
1343 #endif
1344 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1345 }
1346
1347 /*
1348 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1349 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1350 * We need to be 1 filesystem block less than the 2^32 sector limit.
1351 */
1352 static loff_t ext3_max_size(int bits)
1353 {
1354 loff_t res = EXT3_NDIR_BLOCKS;
1355 /* This constant is calculated to be the largest file size for a
1356 * dense, 4k-blocksize file such that the total number of
1357 * sectors in the file, including data and all indirect blocks,
1358 * does not exceed 2^32. */
1359 const loff_t upper_limit = 0x1ff7fffd000LL;
1360
1361 res += 1LL << (bits-2);
1362 res += 1LL << (2*(bits-2));
1363 res += 1LL << (3*(bits-2));
1364 res <<= bits;
1365 if (res > upper_limit)
1366 res = upper_limit;
1367 return res;
1368 }
1369
1370 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1371 ext3_fsblk_t logic_sb_block,
1372 int nr)
1373 {
1374 struct ext3_sb_info *sbi = EXT3_SB(sb);
1375 unsigned long bg, first_meta_bg;
1376 int has_super = 0;
1377
1378 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1379
1380 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1381 nr < first_meta_bg)
1382 return (logic_sb_block + nr + 1);
1383 bg = sbi->s_desc_per_block * nr;
1384 if (ext3_bg_has_super(sb, bg))
1385 has_super = 1;
1386 return (has_super + ext3_group_first_block_no(sb, bg));
1387 }
1388
1389
1390 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1391 {
1392 struct buffer_head * bh;
1393 struct ext3_super_block *es = NULL;
1394 struct ext3_sb_info *sbi;
1395 ext3_fsblk_t block;
1396 ext3_fsblk_t sb_block = get_sb_block(&data);
1397 ext3_fsblk_t logic_sb_block;
1398 unsigned long offset = 0;
1399 unsigned int journal_inum = 0;
1400 unsigned long journal_devnum = 0;
1401 unsigned long def_mount_opts;
1402 struct inode *root;
1403 int blocksize;
1404 int hblock;
1405 int db_count;
1406 int i;
1407 int needs_recovery;
1408 __le32 features;
1409
1410 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1411 if (!sbi)
1412 return -ENOMEM;
1413 sb->s_fs_info = sbi;
1414 sbi->s_mount_opt = 0;
1415 sbi->s_resuid = EXT3_DEF_RESUID;
1416 sbi->s_resgid = EXT3_DEF_RESGID;
1417
1418 unlock_kernel();
1419
1420 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1421 if (!blocksize) {
1422 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1423 goto out_fail;
1424 }
1425
1426 /*
1427 * The ext3 superblock will not be buffer aligned for other than 1kB
1428 * block sizes. We need to calculate the offset from buffer start.
1429 */
1430 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1431 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1432 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1433 } else {
1434 logic_sb_block = sb_block;
1435 }
1436
1437 if (!(bh = sb_bread(sb, logic_sb_block))) {
1438 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1439 goto out_fail;
1440 }
1441 /*
1442 * Note: s_es must be initialized as soon as possible because
1443 * some ext3 macro-instructions depend on its value
1444 */
1445 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1446 sbi->s_es = es;
1447 sb->s_magic = le16_to_cpu(es->s_magic);
1448 if (sb->s_magic != EXT3_SUPER_MAGIC)
1449 goto cantfind_ext3;
1450
1451 /* Set defaults before we parse the mount options */
1452 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1453 if (def_mount_opts & EXT3_DEFM_DEBUG)
1454 set_opt(sbi->s_mount_opt, DEBUG);
1455 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1456 set_opt(sbi->s_mount_opt, GRPID);
1457 if (def_mount_opts & EXT3_DEFM_UID16)
1458 set_opt(sbi->s_mount_opt, NO_UID32);
1459 #ifdef CONFIG_EXT3_FS_XATTR
1460 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1461 set_opt(sbi->s_mount_opt, XATTR_USER);
1462 #endif
1463 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1464 if (def_mount_opts & EXT3_DEFM_ACL)
1465 set_opt(sbi->s_mount_opt, POSIX_ACL);
1466 #endif
1467 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1468 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1469 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1470 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1471 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1472 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1473
1474 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1475 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1476 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1477 set_opt(sbi->s_mount_opt, ERRORS_RO);
1478 else
1479 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1480
1481 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1482 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1483
1484 set_opt(sbi->s_mount_opt, RESERVATION);
1485
1486 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1487 NULL, 0))
1488 goto failed_mount;
1489
1490 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1491 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1492
1493 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1494 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1495 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1496 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1497 printk(KERN_WARNING
1498 "EXT3-fs warning: feature flags set on rev 0 fs, "
1499 "running e2fsck is recommended\n");
1500 /*
1501 * Check feature flags regardless of the revision level, since we
1502 * previously didn't change the revision level when setting the flags,
1503 * so there is a chance incompat flags are set on a rev 0 filesystem.
1504 */
1505 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1506 if (features) {
1507 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1508 "unsupported optional features (%x).\n",
1509 sb->s_id, le32_to_cpu(features));
1510 goto failed_mount;
1511 }
1512 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1513 if (!(sb->s_flags & MS_RDONLY) && features) {
1514 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1515 "unsupported optional features (%x).\n",
1516 sb->s_id, le32_to_cpu(features));
1517 goto failed_mount;
1518 }
1519 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1520
1521 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1522 blocksize > EXT3_MAX_BLOCK_SIZE) {
1523 printk(KERN_ERR
1524 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1525 blocksize, sb->s_id);
1526 goto failed_mount;
1527 }
1528
1529 hblock = bdev_hardsect_size(sb->s_bdev);
1530 if (sb->s_blocksize != blocksize) {
1531 /*
1532 * Make sure the blocksize for the filesystem is larger
1533 * than the hardware sectorsize for the machine.
1534 */
1535 if (blocksize < hblock) {
1536 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1537 "device blocksize %d.\n", blocksize, hblock);
1538 goto failed_mount;
1539 }
1540
1541 brelse (bh);
1542 sb_set_blocksize(sb, blocksize);
1543 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1544 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1545 bh = sb_bread(sb, logic_sb_block);
1546 if (!bh) {
1547 printk(KERN_ERR
1548 "EXT3-fs: Can't read superblock on 2nd try.\n");
1549 goto failed_mount;
1550 }
1551 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1552 sbi->s_es = es;
1553 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1554 printk (KERN_ERR
1555 "EXT3-fs: Magic mismatch, very weird !\n");
1556 goto failed_mount;
1557 }
1558 }
1559
1560 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1561
1562 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1563 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1564 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1565 } else {
1566 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1567 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1568 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1569 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1570 (sbi->s_inode_size > blocksize)) {
1571 printk (KERN_ERR
1572 "EXT3-fs: unsupported inode size: %d\n",
1573 sbi->s_inode_size);
1574 goto failed_mount;
1575 }
1576 }
1577 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1578 le32_to_cpu(es->s_log_frag_size);
1579 if (blocksize != sbi->s_frag_size) {
1580 printk(KERN_ERR
1581 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1582 sbi->s_frag_size, blocksize);
1583 goto failed_mount;
1584 }
1585 sbi->s_frags_per_block = 1;
1586 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1587 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1588 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1589 if (EXT3_INODE_SIZE(sb) == 0)
1590 goto cantfind_ext3;
1591 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1592 if (sbi->s_inodes_per_block == 0)
1593 goto cantfind_ext3;
1594 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1595 sbi->s_inodes_per_block;
1596 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1597 sbi->s_sbh = bh;
1598 sbi->s_mount_state = le16_to_cpu(es->s_state);
1599 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1600 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1601 for (i=0; i < 4; i++)
1602 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1603 sbi->s_def_hash_version = es->s_def_hash_version;
1604
1605 if (sbi->s_blocks_per_group > blocksize * 8) {
1606 printk (KERN_ERR
1607 "EXT3-fs: #blocks per group too big: %lu\n",
1608 sbi->s_blocks_per_group);
1609 goto failed_mount;
1610 }
1611 if (sbi->s_frags_per_group > blocksize * 8) {
1612 printk (KERN_ERR
1613 "EXT3-fs: #fragments per group too big: %lu\n",
1614 sbi->s_frags_per_group);
1615 goto failed_mount;
1616 }
1617 if (sbi->s_inodes_per_group > blocksize * 8) {
1618 printk (KERN_ERR
1619 "EXT3-fs: #inodes per group too big: %lu\n",
1620 sbi->s_inodes_per_group);
1621 goto failed_mount;
1622 }
1623
1624 if (le32_to_cpu(es->s_blocks_count) >
1625 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1626 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1627 " too large to mount safely\n", sb->s_id);
1628 if (sizeof(sector_t) < 8)
1629 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1630 "enabled\n");
1631 goto failed_mount;
1632 }
1633
1634 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1635 goto cantfind_ext3;
1636 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1637 le32_to_cpu(es->s_first_data_block) - 1)
1638 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1639 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1640 EXT3_DESC_PER_BLOCK(sb);
1641 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1642 GFP_KERNEL);
1643 if (sbi->s_group_desc == NULL) {
1644 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1645 goto failed_mount;
1646 }
1647
1648 bgl_lock_init(&sbi->s_blockgroup_lock);
1649
1650 for (i = 0; i < db_count; i++) {
1651 block = descriptor_loc(sb, logic_sb_block, i);
1652 sbi->s_group_desc[i] = sb_bread(sb, block);
1653 if (!sbi->s_group_desc[i]) {
1654 printk (KERN_ERR "EXT3-fs: "
1655 "can't read group descriptor %d\n", i);
1656 db_count = i;
1657 goto failed_mount2;
1658 }
1659 }
1660 if (!ext3_check_descriptors (sb)) {
1661 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1662 goto failed_mount2;
1663 }
1664 sbi->s_gdb_count = db_count;
1665 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1666 spin_lock_init(&sbi->s_next_gen_lock);
1667
1668 percpu_counter_init(&sbi->s_freeblocks_counter,
1669 ext3_count_free_blocks(sb));
1670 percpu_counter_init(&sbi->s_freeinodes_counter,
1671 ext3_count_free_inodes(sb));
1672 percpu_counter_init(&sbi->s_dirs_counter,
1673 ext3_count_dirs(sb));
1674
1675 /* per fileystem reservation list head & lock */
1676 spin_lock_init(&sbi->s_rsv_window_lock);
1677 sbi->s_rsv_window_root = RB_ROOT;
1678 /* Add a single, static dummy reservation to the start of the
1679 * reservation window list --- it gives us a placeholder for
1680 * append-at-start-of-list which makes the allocation logic
1681 * _much_ simpler. */
1682 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1683 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1684 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1685 sbi->s_rsv_window_head.rsv_goal_size = 0;
1686 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1687
1688 /*
1689 * set up enough so that it can read an inode
1690 */
1691 sb->s_op = &ext3_sops;
1692 sb->s_export_op = &ext3_export_ops;
1693 sb->s_xattr = ext3_xattr_handlers;
1694 #ifdef CONFIG_QUOTA
1695 sb->s_qcop = &ext3_qctl_operations;
1696 sb->dq_op = &ext3_quota_operations;
1697 #endif
1698 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1699
1700 sb->s_root = NULL;
1701
1702 needs_recovery = (es->s_last_orphan != 0 ||
1703 EXT3_HAS_INCOMPAT_FEATURE(sb,
1704 EXT3_FEATURE_INCOMPAT_RECOVER));
1705
1706 /*
1707 * The first inode we look at is the journal inode. Don't try
1708 * root first: it may be modified in the journal!
1709 */
1710 if (!test_opt(sb, NOLOAD) &&
1711 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1712 if (ext3_load_journal(sb, es, journal_devnum))
1713 goto failed_mount3;
1714 } else if (journal_inum) {
1715 if (ext3_create_journal(sb, es, journal_inum))
1716 goto failed_mount3;
1717 } else {
1718 if (!silent)
1719 printk (KERN_ERR
1720 "ext3: No journal on filesystem on %s\n",
1721 sb->s_id);
1722 goto failed_mount3;
1723 }
1724
1725 /* We have now updated the journal if required, so we can
1726 * validate the data journaling mode. */
1727 switch (test_opt(sb, DATA_FLAGS)) {
1728 case 0:
1729 /* No mode set, assume a default based on the journal
1730 capabilities: ORDERED_DATA if the journal can
1731 cope, else JOURNAL_DATA */
1732 if (journal_check_available_features
1733 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1734 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1735 else
1736 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1737 break;
1738
1739 case EXT3_MOUNT_ORDERED_DATA:
1740 case EXT3_MOUNT_WRITEBACK_DATA:
1741 if (!journal_check_available_features
1742 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1743 printk(KERN_ERR "EXT3-fs: Journal does not support "
1744 "requested data journaling mode\n");
1745 goto failed_mount4;
1746 }
1747 default:
1748 break;
1749 }
1750
1751 if (test_opt(sb, NOBH)) {
1752 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1753 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1754 "its supported only with writeback mode\n");
1755 clear_opt(sbi->s_mount_opt, NOBH);
1756 }
1757 }
1758 /*
1759 * The journal_load will have done any necessary log recovery,
1760 * so we can safely mount the rest of the filesystem now.
1761 */
1762
1763 root = iget(sb, EXT3_ROOT_INO);
1764 sb->s_root = d_alloc_root(root);
1765 if (!sb->s_root) {
1766 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1767 iput(root);
1768 goto failed_mount4;
1769 }
1770 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1771 dput(sb->s_root);
1772 sb->s_root = NULL;
1773 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1774 goto failed_mount4;
1775 }
1776
1777 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1778 /*
1779 * akpm: core read_super() calls in here with the superblock locked.
1780 * That deadlocks, because orphan cleanup needs to lock the superblock
1781 * in numerous places. Here we just pop the lock - it's relatively
1782 * harmless, because we are now ready to accept write_super() requests,
1783 * and aviro says that's the only reason for hanging onto the
1784 * superblock lock.
1785 */
1786 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1787 ext3_orphan_cleanup(sb, es);
1788 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1789 if (needs_recovery)
1790 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1791 ext3_mark_recovery_complete(sb, es);
1792 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1793 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1794 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1795 "writeback");
1796
1797 lock_kernel();
1798 return 0;
1799
1800 cantfind_ext3:
1801 if (!silent)
1802 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1803 sb->s_id);
1804 goto failed_mount;
1805
1806 failed_mount4:
1807 journal_destroy(sbi->s_journal);
1808 failed_mount3:
1809 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1810 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1811 percpu_counter_destroy(&sbi->s_dirs_counter);
1812 failed_mount2:
1813 for (i = 0; i < db_count; i++)
1814 brelse(sbi->s_group_desc[i]);
1815 kfree(sbi->s_group_desc);
1816 failed_mount:
1817 #ifdef CONFIG_QUOTA
1818 for (i = 0; i < MAXQUOTAS; i++)
1819 kfree(sbi->s_qf_names[i]);
1820 #endif
1821 ext3_blkdev_remove(sbi);
1822 brelse(bh);
1823 out_fail:
1824 sb->s_fs_info = NULL;
1825 kfree(sbi);
1826 lock_kernel();
1827 return -EINVAL;
1828 }
1829
1830 /*
1831 * Setup any per-fs journal parameters now. We'll do this both on
1832 * initial mount, once the journal has been initialised but before we've
1833 * done any recovery; and again on any subsequent remount.
1834 */
1835 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1836 {
1837 struct ext3_sb_info *sbi = EXT3_SB(sb);
1838
1839 if (sbi->s_commit_interval)
1840 journal->j_commit_interval = sbi->s_commit_interval;
1841 /* We could also set up an ext3-specific default for the commit
1842 * interval here, but for now we'll just fall back to the jbd
1843 * default. */
1844
1845 spin_lock(&journal->j_state_lock);
1846 if (test_opt(sb, BARRIER))
1847 journal->j_flags |= JFS_BARRIER;
1848 else
1849 journal->j_flags &= ~JFS_BARRIER;
1850 spin_unlock(&journal->j_state_lock);
1851 }
1852
1853 static journal_t *ext3_get_journal(struct super_block *sb,
1854 unsigned int journal_inum)
1855 {
1856 struct inode *journal_inode;
1857 journal_t *journal;
1858
1859 /* First, test for the existence of a valid inode on disk. Bad
1860 * things happen if we iget() an unused inode, as the subsequent
1861 * iput() will try to delete it. */
1862
1863 journal_inode = iget(sb, journal_inum);
1864 if (!journal_inode) {
1865 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1866 return NULL;
1867 }
1868 if (!journal_inode->i_nlink) {
1869 make_bad_inode(journal_inode);
1870 iput(journal_inode);
1871 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1872 return NULL;
1873 }
1874
1875 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1876 journal_inode, journal_inode->i_size);
1877 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1878 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1879 iput(journal_inode);
1880 return NULL;
1881 }
1882
1883 journal = journal_init_inode(journal_inode);
1884 if (!journal) {
1885 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1886 iput(journal_inode);
1887 return NULL;
1888 }
1889 journal->j_private = sb;
1890 ext3_init_journal_params(sb, journal);
1891 return journal;
1892 }
1893
1894 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1895 dev_t j_dev)
1896 {
1897 struct buffer_head * bh;
1898 journal_t *journal;
1899 ext3_fsblk_t start;
1900 ext3_fsblk_t len;
1901 int hblock, blocksize;
1902 ext3_fsblk_t sb_block;
1903 unsigned long offset;
1904 struct ext3_super_block * es;
1905 struct block_device *bdev;
1906
1907 bdev = ext3_blkdev_get(j_dev);
1908 if (bdev == NULL)
1909 return NULL;
1910
1911 if (bd_claim(bdev, sb)) {
1912 printk(KERN_ERR
1913 "EXT3: failed to claim external journal device.\n");
1914 blkdev_put(bdev);
1915 return NULL;
1916 }
1917
1918 blocksize = sb->s_blocksize;
1919 hblock = bdev_hardsect_size(bdev);
1920 if (blocksize < hblock) {
1921 printk(KERN_ERR
1922 "EXT3-fs: blocksize too small for journal device.\n");
1923 goto out_bdev;
1924 }
1925
1926 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1927 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1928 set_blocksize(bdev, blocksize);
1929 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1930 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1931 "external journal\n");
1932 goto out_bdev;
1933 }
1934
1935 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1936 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1937 !(le32_to_cpu(es->s_feature_incompat) &
1938 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1939 printk(KERN_ERR "EXT3-fs: external journal has "
1940 "bad superblock\n");
1941 brelse(bh);
1942 goto out_bdev;
1943 }
1944
1945 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1946 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1947 brelse(bh);
1948 goto out_bdev;
1949 }
1950
1951 len = le32_to_cpu(es->s_blocks_count);
1952 start = sb_block + 1;
1953 brelse(bh); /* we're done with the superblock */
1954
1955 journal = journal_init_dev(bdev, sb->s_bdev,
1956 start, len, blocksize);
1957 if (!journal) {
1958 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1959 goto out_bdev;
1960 }
1961 journal->j_private = sb;
1962 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1963 wait_on_buffer(journal->j_sb_buffer);
1964 if (!buffer_uptodate(journal->j_sb_buffer)) {
1965 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1966 goto out_journal;
1967 }
1968 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1969 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1970 "user (unsupported) - %d\n",
1971 be32_to_cpu(journal->j_superblock->s_nr_users));
1972 goto out_journal;
1973 }
1974 EXT3_SB(sb)->journal_bdev = bdev;
1975 ext3_init_journal_params(sb, journal);
1976 return journal;
1977 out_journal:
1978 journal_destroy(journal);
1979 out_bdev:
1980 ext3_blkdev_put(bdev);
1981 return NULL;
1982 }
1983
1984 static int ext3_load_journal(struct super_block *sb,
1985 struct ext3_super_block *es,
1986 unsigned long journal_devnum)
1987 {
1988 journal_t *journal;
1989 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
1990 dev_t journal_dev;
1991 int err = 0;
1992 int really_read_only;
1993
1994 if (journal_devnum &&
1995 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
1996 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
1997 "numbers have changed\n");
1998 journal_dev = new_decode_dev(journal_devnum);
1999 } else
2000 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2001
2002 really_read_only = bdev_read_only(sb->s_bdev);
2003
2004 /*
2005 * Are we loading a blank journal or performing recovery after a
2006 * crash? For recovery, we need to check in advance whether we
2007 * can get read-write access to the device.
2008 */
2009
2010 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2011 if (sb->s_flags & MS_RDONLY) {
2012 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2013 "required on readonly filesystem.\n");
2014 if (really_read_only) {
2015 printk(KERN_ERR "EXT3-fs: write access "
2016 "unavailable, cannot proceed.\n");
2017 return -EROFS;
2018 }
2019 printk (KERN_INFO "EXT3-fs: write access will "
2020 "be enabled during recovery.\n");
2021 }
2022 }
2023
2024 if (journal_inum && journal_dev) {
2025 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2026 "and inode journals!\n");
2027 return -EINVAL;
2028 }
2029
2030 if (journal_inum) {
2031 if (!(journal = ext3_get_journal(sb, journal_inum)))
2032 return -EINVAL;
2033 } else {
2034 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2035 return -EINVAL;
2036 }
2037
2038 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2039 err = journal_update_format(journal);
2040 if (err) {
2041 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2042 journal_destroy(journal);
2043 return err;
2044 }
2045 }
2046
2047 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2048 err = journal_wipe(journal, !really_read_only);
2049 if (!err)
2050 err = journal_load(journal);
2051
2052 if (err) {
2053 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2054 journal_destroy(journal);
2055 return err;
2056 }
2057
2058 EXT3_SB(sb)->s_journal = journal;
2059 ext3_clear_journal_err(sb, es);
2060
2061 if (journal_devnum &&
2062 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2063 es->s_journal_dev = cpu_to_le32(journal_devnum);
2064 sb->s_dirt = 1;
2065
2066 /* Make sure we flush the recovery flag to disk. */
2067 ext3_commit_super(sb, es, 1);
2068 }
2069
2070 return 0;
2071 }
2072
2073 static int ext3_create_journal(struct super_block * sb,
2074 struct ext3_super_block * es,
2075 unsigned int journal_inum)
2076 {
2077 journal_t *journal;
2078
2079 if (sb->s_flags & MS_RDONLY) {
2080 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2081 "create journal.\n");
2082 return -EROFS;
2083 }
2084
2085 if (!(journal = ext3_get_journal(sb, journal_inum)))
2086 return -EINVAL;
2087
2088 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2089 journal_inum);
2090
2091 if (journal_create(journal)) {
2092 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2093 journal_destroy(journal);
2094 return -EIO;
2095 }
2096
2097 EXT3_SB(sb)->s_journal = journal;
2098
2099 ext3_update_dynamic_rev(sb);
2100 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2101 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2102
2103 es->s_journal_inum = cpu_to_le32(journal_inum);
2104 sb->s_dirt = 1;
2105
2106 /* Make sure we flush the recovery flag to disk. */
2107 ext3_commit_super(sb, es, 1);
2108
2109 return 0;
2110 }
2111
2112 static void ext3_commit_super (struct super_block * sb,
2113 struct ext3_super_block * es,
2114 int sync)
2115 {
2116 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2117
2118 if (!sbh)
2119 return;
2120 es->s_wtime = cpu_to_le32(get_seconds());
2121 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2122 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2123 BUFFER_TRACE(sbh, "marking dirty");
2124 mark_buffer_dirty(sbh);
2125 if (sync)
2126 sync_dirty_buffer(sbh);
2127 }
2128
2129
2130 /*
2131 * Have we just finished recovery? If so, and if we are mounting (or
2132 * remounting) the filesystem readonly, then we will end up with a
2133 * consistent fs on disk. Record that fact.
2134 */
2135 static void ext3_mark_recovery_complete(struct super_block * sb,
2136 struct ext3_super_block * es)
2137 {
2138 journal_t *journal = EXT3_SB(sb)->s_journal;
2139
2140 journal_lock_updates(journal);
2141 journal_flush(journal);
2142 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2143 sb->s_flags & MS_RDONLY) {
2144 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2145 sb->s_dirt = 0;
2146 ext3_commit_super(sb, es, 1);
2147 }
2148 journal_unlock_updates(journal);
2149 }
2150
2151 /*
2152 * If we are mounting (or read-write remounting) a filesystem whose journal
2153 * has recorded an error from a previous lifetime, move that error to the
2154 * main filesystem now.
2155 */
2156 static void ext3_clear_journal_err(struct super_block * sb,
2157 struct ext3_super_block * es)
2158 {
2159 journal_t *journal;
2160 int j_errno;
2161 const char *errstr;
2162
2163 journal = EXT3_SB(sb)->s_journal;
2164
2165 /*
2166 * Now check for any error status which may have been recorded in the
2167 * journal by a prior ext3_error() or ext3_abort()
2168 */
2169
2170 j_errno = journal_errno(journal);
2171 if (j_errno) {
2172 char nbuf[16];
2173
2174 errstr = ext3_decode_error(sb, j_errno, nbuf);
2175 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2176 "from previous mount: %s", errstr);
2177 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2178 "filesystem check.");
2179
2180 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2181 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2182 ext3_commit_super (sb, es, 1);
2183
2184 journal_clear_err(journal);
2185 }
2186 }
2187
2188 /*
2189 * Force the running and committing transactions to commit,
2190 * and wait on the commit.
2191 */
2192 int ext3_force_commit(struct super_block *sb)
2193 {
2194 journal_t *journal;
2195 int ret;
2196
2197 if (sb->s_flags & MS_RDONLY)
2198 return 0;
2199
2200 journal = EXT3_SB(sb)->s_journal;
2201 sb->s_dirt = 0;
2202 ret = ext3_journal_force_commit(journal);
2203 return ret;
2204 }
2205
2206 /*
2207 * Ext3 always journals updates to the superblock itself, so we don't
2208 * have to propagate any other updates to the superblock on disk at this
2209 * point. Just start an async writeback to get the buffers on their way
2210 * to the disk.
2211 *
2212 * This implicitly triggers the writebehind on sync().
2213 */
2214
2215 static void ext3_write_super (struct super_block * sb)
2216 {
2217 if (mutex_trylock(&sb->s_lock) != 0)
2218 BUG();
2219 sb->s_dirt = 0;
2220 }
2221
2222 static int ext3_sync_fs(struct super_block *sb, int wait)
2223 {
2224 tid_t target;
2225
2226 sb->s_dirt = 0;
2227 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2228 if (wait)
2229 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2230 }
2231 return 0;
2232 }
2233
2234 /*
2235 * LVM calls this function before a (read-only) snapshot is created. This
2236 * gives us a chance to flush the journal completely and mark the fs clean.
2237 */
2238 static void ext3_write_super_lockfs(struct super_block *sb)
2239 {
2240 sb->s_dirt = 0;
2241
2242 if (!(sb->s_flags & MS_RDONLY)) {
2243 journal_t *journal = EXT3_SB(sb)->s_journal;
2244
2245 /* Now we set up the journal barrier. */
2246 journal_lock_updates(journal);
2247 journal_flush(journal);
2248
2249 /* Journal blocked and flushed, clear needs_recovery flag. */
2250 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2251 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2252 }
2253 }
2254
2255 /*
2256 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2257 * flag here, even though the filesystem is not technically dirty yet.
2258 */
2259 static void ext3_unlockfs(struct super_block *sb)
2260 {
2261 if (!(sb->s_flags & MS_RDONLY)) {
2262 lock_super(sb);
2263 /* Reser the needs_recovery flag before the fs is unlocked. */
2264 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2265 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2266 unlock_super(sb);
2267 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2268 }
2269 }
2270
2271 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2272 {
2273 struct ext3_super_block * es;
2274 struct ext3_sb_info *sbi = EXT3_SB(sb);
2275 ext3_fsblk_t n_blocks_count = 0;
2276 unsigned long old_sb_flags;
2277 struct ext3_mount_options old_opts;
2278 int err;
2279 #ifdef CONFIG_QUOTA
2280 int i;
2281 #endif
2282
2283 /* Store the original options */
2284 old_sb_flags = sb->s_flags;
2285 old_opts.s_mount_opt = sbi->s_mount_opt;
2286 old_opts.s_resuid = sbi->s_resuid;
2287 old_opts.s_resgid = sbi->s_resgid;
2288 old_opts.s_commit_interval = sbi->s_commit_interval;
2289 #ifdef CONFIG_QUOTA
2290 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2291 for (i = 0; i < MAXQUOTAS; i++)
2292 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2293 #endif
2294
2295 /*
2296 * Allow the "check" option to be passed as a remount option.
2297 */
2298 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2299 err = -EINVAL;
2300 goto restore_opts;
2301 }
2302
2303 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2304 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2305
2306 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2307 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2308
2309 es = sbi->s_es;
2310
2311 ext3_init_journal_params(sb, sbi->s_journal);
2312
2313 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2314 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2315 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2316 err = -EROFS;
2317 goto restore_opts;
2318 }
2319
2320 if (*flags & MS_RDONLY) {
2321 /*
2322 * First of all, the unconditional stuff we have to do
2323 * to disable replay of the journal when we next remount
2324 */
2325 sb->s_flags |= MS_RDONLY;
2326
2327 /*
2328 * OK, test if we are remounting a valid rw partition
2329 * readonly, and if so set the rdonly flag and then
2330 * mark the partition as valid again.
2331 */
2332 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2333 (sbi->s_mount_state & EXT3_VALID_FS))
2334 es->s_state = cpu_to_le16(sbi->s_mount_state);
2335
2336 ext3_mark_recovery_complete(sb, es);
2337 } else {
2338 __le32 ret;
2339 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2340 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2341 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2342 "remount RDWR because of unsupported "
2343 "optional features (%x).\n",
2344 sb->s_id, le32_to_cpu(ret));
2345 err = -EROFS;
2346 goto restore_opts;
2347 }
2348
2349 /*
2350 * If we have an unprocessed orphan list hanging
2351 * around from a previously readonly bdev mount,
2352 * require a full umount/remount for now.
2353 */
2354 if (es->s_last_orphan) {
2355 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2356 "remount RDWR because of unprocessed "
2357 "orphan inode list. Please "
2358 "umount/remount instead.\n",
2359 sb->s_id);
2360 err = -EINVAL;
2361 goto restore_opts;
2362 }
2363
2364 /*
2365 * Mounting a RDONLY partition read-write, so reread
2366 * and store the current valid flag. (It may have
2367 * been changed by e2fsck since we originally mounted
2368 * the partition.)
2369 */
2370 ext3_clear_journal_err(sb, es);
2371 sbi->s_mount_state = le16_to_cpu(es->s_state);
2372 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2373 goto restore_opts;
2374 if (!ext3_setup_super (sb, es, 0))
2375 sb->s_flags &= ~MS_RDONLY;
2376 }
2377 }
2378 #ifdef CONFIG_QUOTA
2379 /* Release old quota file names */
2380 for (i = 0; i < MAXQUOTAS; i++)
2381 if (old_opts.s_qf_names[i] &&
2382 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2383 kfree(old_opts.s_qf_names[i]);
2384 #endif
2385 return 0;
2386 restore_opts:
2387 sb->s_flags = old_sb_flags;
2388 sbi->s_mount_opt = old_opts.s_mount_opt;
2389 sbi->s_resuid = old_opts.s_resuid;
2390 sbi->s_resgid = old_opts.s_resgid;
2391 sbi->s_commit_interval = old_opts.s_commit_interval;
2392 #ifdef CONFIG_QUOTA
2393 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2394 for (i = 0; i < MAXQUOTAS; i++) {
2395 if (sbi->s_qf_names[i] &&
2396 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2397 kfree(sbi->s_qf_names[i]);
2398 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2399 }
2400 #endif
2401 return err;
2402 }
2403
2404 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2405 {
2406 struct super_block *sb = dentry->d_sb;
2407 struct ext3_sb_info *sbi = EXT3_SB(sb);
2408 struct ext3_super_block *es = sbi->s_es;
2409 ext3_fsblk_t overhead;
2410 int i;
2411 u64 fsid;
2412
2413 if (test_opt (sb, MINIX_DF))
2414 overhead = 0;
2415 else {
2416 unsigned long ngroups;
2417 ngroups = EXT3_SB(sb)->s_groups_count;
2418 smp_rmb();
2419
2420 /*
2421 * Compute the overhead (FS structures)
2422 */
2423
2424 /*
2425 * All of the blocks before first_data_block are
2426 * overhead
2427 */
2428 overhead = le32_to_cpu(es->s_first_data_block);
2429
2430 /*
2431 * Add the overhead attributed to the superblock and
2432 * block group descriptors. If the sparse superblocks
2433 * feature is turned on, then not all groups have this.
2434 */
2435 for (i = 0; i < ngroups; i++) {
2436 overhead += ext3_bg_has_super(sb, i) +
2437 ext3_bg_num_gdb(sb, i);
2438 cond_resched();
2439 }
2440
2441 /*
2442 * Every block group has an inode bitmap, a block
2443 * bitmap, and an inode table.
2444 */
2445 overhead += (ngroups * (2 + EXT3_SB(sb)->s_itb_per_group));
2446 }
2447
2448 buf->f_type = EXT3_SUPER_MAGIC;
2449 buf->f_bsize = sb->s_blocksize;
2450 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
2451 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
2452 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2453 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2454 buf->f_bavail = 0;
2455 buf->f_files = le32_to_cpu(es->s_inodes_count);
2456 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
2457 buf->f_namelen = EXT3_NAME_LEN;
2458 fsid = le64_to_cpup((void *)es->s_uuid) ^
2459 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2460 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2461 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2462 return 0;
2463 }
2464
2465 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2466 * is locked for write. Otherwise the are possible deadlocks:
2467 * Process 1 Process 2
2468 * ext3_create() quota_sync()
2469 * journal_start() write_dquot()
2470 * DQUOT_INIT() down(dqio_mutex)
2471 * down(dqio_mutex) journal_start()
2472 *
2473 */
2474
2475 #ifdef CONFIG_QUOTA
2476
2477 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2478 {
2479 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2480 }
2481
2482 static int ext3_dquot_initialize(struct inode *inode, int type)
2483 {
2484 handle_t *handle;
2485 int ret, err;
2486
2487 /* We may create quota structure so we need to reserve enough blocks */
2488 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2489 if (IS_ERR(handle))
2490 return PTR_ERR(handle);
2491 ret = dquot_initialize(inode, type);
2492 err = ext3_journal_stop(handle);
2493 if (!ret)
2494 ret = err;
2495 return ret;
2496 }
2497
2498 static int ext3_dquot_drop(struct inode *inode)
2499 {
2500 handle_t *handle;
2501 int ret, err;
2502
2503 /* We may delete quota structure so we need to reserve enough blocks */
2504 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2505 if (IS_ERR(handle))
2506 return PTR_ERR(handle);
2507 ret = dquot_drop(inode);
2508 err = ext3_journal_stop(handle);
2509 if (!ret)
2510 ret = err;
2511 return ret;
2512 }
2513
2514 static int ext3_write_dquot(struct dquot *dquot)
2515 {
2516 int ret, err;
2517 handle_t *handle;
2518 struct inode *inode;
2519
2520 inode = dquot_to_inode(dquot);
2521 handle = ext3_journal_start(inode,
2522 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2523 if (IS_ERR(handle))
2524 return PTR_ERR(handle);
2525 ret = dquot_commit(dquot);
2526 err = ext3_journal_stop(handle);
2527 if (!ret)
2528 ret = err;
2529 return ret;
2530 }
2531
2532 static int ext3_acquire_dquot(struct dquot *dquot)
2533 {
2534 int ret, err;
2535 handle_t *handle;
2536
2537 handle = ext3_journal_start(dquot_to_inode(dquot),
2538 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2539 if (IS_ERR(handle))
2540 return PTR_ERR(handle);
2541 ret = dquot_acquire(dquot);
2542 err = ext3_journal_stop(handle);
2543 if (!ret)
2544 ret = err;
2545 return ret;
2546 }
2547
2548 static int ext3_release_dquot(struct dquot *dquot)
2549 {
2550 int ret, err;
2551 handle_t *handle;
2552
2553 handle = ext3_journal_start(dquot_to_inode(dquot),
2554 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2555 if (IS_ERR(handle))
2556 return PTR_ERR(handle);
2557 ret = dquot_release(dquot);
2558 err = ext3_journal_stop(handle);
2559 if (!ret)
2560 ret = err;
2561 return ret;
2562 }
2563
2564 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2565 {
2566 /* Are we journalling quotas? */
2567 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2568 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2569 dquot_mark_dquot_dirty(dquot);
2570 return ext3_write_dquot(dquot);
2571 } else {
2572 return dquot_mark_dquot_dirty(dquot);
2573 }
2574 }
2575
2576 static int ext3_write_info(struct super_block *sb, int type)
2577 {
2578 int ret, err;
2579 handle_t *handle;
2580
2581 /* Data block + inode block */
2582 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2583 if (IS_ERR(handle))
2584 return PTR_ERR(handle);
2585 ret = dquot_commit_info(sb, type);
2586 err = ext3_journal_stop(handle);
2587 if (!ret)
2588 ret = err;
2589 return ret;
2590 }
2591
2592 /*
2593 * Turn on quotas during mount time - we need to find
2594 * the quota file and such...
2595 */
2596 static int ext3_quota_on_mount(struct super_block *sb, int type)
2597 {
2598 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2599 EXT3_SB(sb)->s_jquota_fmt, type);
2600 }
2601
2602 /*
2603 * Standard function to be called on quota_on
2604 */
2605 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2606 char *path)
2607 {
2608 int err;
2609 struct nameidata nd;
2610
2611 if (!test_opt(sb, QUOTA))
2612 return -EINVAL;
2613 /* Not journalling quota? */
2614 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2615 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2616 return vfs_quota_on(sb, type, format_id, path);
2617 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2618 if (err)
2619 return err;
2620 /* Quotafile not on the same filesystem? */
2621 if (nd.mnt->mnt_sb != sb) {
2622 path_release(&nd);
2623 return -EXDEV;
2624 }
2625 /* Quotafile not of fs root? */
2626 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2627 printk(KERN_WARNING
2628 "EXT3-fs: Quota file not on filesystem root. "
2629 "Journalled quota will not work.\n");
2630 path_release(&nd);
2631 return vfs_quota_on(sb, type, format_id, path);
2632 }
2633
2634 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2635 * acquiring the locks... As quota files are never truncated and quota code
2636 * itself serializes the operations (and noone else should touch the files)
2637 * we don't have to be afraid of races */
2638 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2639 size_t len, loff_t off)
2640 {
2641 struct inode *inode = sb_dqopt(sb)->files[type];
2642 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2643 int err = 0;
2644 int offset = off & (sb->s_blocksize - 1);
2645 int tocopy;
2646 size_t toread;
2647 struct buffer_head *bh;
2648 loff_t i_size = i_size_read(inode);
2649
2650 if (off > i_size)
2651 return 0;
2652 if (off+len > i_size)
2653 len = i_size-off;
2654 toread = len;
2655 while (toread > 0) {
2656 tocopy = sb->s_blocksize - offset < toread ?
2657 sb->s_blocksize - offset : toread;
2658 bh = ext3_bread(NULL, inode, blk, 0, &err);
2659 if (err)
2660 return err;
2661 if (!bh) /* A hole? */
2662 memset(data, 0, tocopy);
2663 else
2664 memcpy(data, bh->b_data+offset, tocopy);
2665 brelse(bh);
2666 offset = 0;
2667 toread -= tocopy;
2668 data += tocopy;
2669 blk++;
2670 }
2671 return len;
2672 }
2673
2674 /* Write to quotafile (we know the transaction is already started and has
2675 * enough credits) */
2676 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2677 const char *data, size_t len, loff_t off)
2678 {
2679 struct inode *inode = sb_dqopt(sb)->files[type];
2680 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2681 int err = 0;
2682 int offset = off & (sb->s_blocksize - 1);
2683 int tocopy;
2684 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2685 size_t towrite = len;
2686 struct buffer_head *bh;
2687 handle_t *handle = journal_current_handle();
2688
2689 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2690 while (towrite > 0) {
2691 tocopy = sb->s_blocksize - offset < towrite ?
2692 sb->s_blocksize - offset : towrite;
2693 bh = ext3_bread(handle, inode, blk, 1, &err);
2694 if (!bh)
2695 goto out;
2696 if (journal_quota) {
2697 err = ext3_journal_get_write_access(handle, bh);
2698 if (err) {
2699 brelse(bh);
2700 goto out;
2701 }
2702 }
2703 lock_buffer(bh);
2704 memcpy(bh->b_data+offset, data, tocopy);
2705 flush_dcache_page(bh->b_page);
2706 unlock_buffer(bh);
2707 if (journal_quota)
2708 err = ext3_journal_dirty_metadata(handle, bh);
2709 else {
2710 /* Always do at least ordered writes for quotas */
2711 err = ext3_journal_dirty_data(handle, bh);
2712 mark_buffer_dirty(bh);
2713 }
2714 brelse(bh);
2715 if (err)
2716 goto out;
2717 offset = 0;
2718 towrite -= tocopy;
2719 data += tocopy;
2720 blk++;
2721 }
2722 out:
2723 if (len == towrite)
2724 return err;
2725 if (inode->i_size < off+len-towrite) {
2726 i_size_write(inode, off+len-towrite);
2727 EXT3_I(inode)->i_disksize = inode->i_size;
2728 }
2729 inode->i_version++;
2730 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2731 ext3_mark_inode_dirty(handle, inode);
2732 mutex_unlock(&inode->i_mutex);
2733 return len - towrite;
2734 }
2735
2736 #endif
2737
2738 static int ext3_get_sb(struct file_system_type *fs_type,
2739 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2740 {
2741 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2742 }
2743
2744 static struct file_system_type ext3_fs_type = {
2745 .owner = THIS_MODULE,
2746 .name = "ext3",
2747 .get_sb = ext3_get_sb,
2748 .kill_sb = kill_block_super,
2749 .fs_flags = FS_REQUIRES_DEV,
2750 };
2751
2752 static int __init init_ext3_fs(void)
2753 {
2754 int err = init_ext3_xattr();
2755 if (err)
2756 return err;
2757 err = init_inodecache();
2758 if (err)
2759 goto out1;
2760 err = register_filesystem(&ext3_fs_type);
2761 if (err)
2762 goto out;
2763 return 0;
2764 out:
2765 destroy_inodecache();
2766 out1:
2767 exit_ext3_xattr();
2768 return err;
2769 }
2770
2771 static void __exit exit_ext3_fs(void)
2772 {
2773 unregister_filesystem(&ext3_fs_type);
2774 destroy_inodecache();
2775 exit_ext3_xattr();
2776 }
2777
2778 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2779 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2780 MODULE_LICENSE("GPL");
2781 module_init(init_ext3_fs)
2782 module_exit(exit_ext3_fs)
Support for the Chinese Samsung S3C2440 based development boards