2 * linux/fs/ext3/super.c
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)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.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/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
41 #include <asm/uaccess.h>
47 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
48 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
50 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
53 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
54 unsigned long journal_devnum);
55 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
57 static int ext3_commit_super(struct super_block *sb,
58 struct ext3_super_block *es,
60 static void ext3_mark_recovery_complete(struct super_block * sb,
61 struct ext3_super_block * es);
62 static void ext3_clear_journal_err(struct super_block * sb,
63 struct ext3_super_block * es);
64 static int ext3_sync_fs(struct super_block *sb, int wait);
65 static const char *ext3_decode_error(struct super_block * sb, int errno,
67 static int ext3_remount (struct super_block * sb, int * flags, char * data);
68 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
69 static int ext3_unfreeze(struct super_block *sb);
70 static int ext3_freeze(struct super_block *sb);
73 * Wrappers for journal_start/end.
75 * The only special thing we need to do here is to make sure that all
76 * journal_end calls result in the superblock being marked dirty, so
77 * that sync() will call the filesystem's write_super callback if
80 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
84 if (sb->s_flags & MS_RDONLY)
85 return ERR_PTR(-EROFS);
87 /* Special case here: if the journal has aborted behind our
88 * backs (eg. EIO in the commit thread), then we still need to
89 * take the FS itself readonly cleanly. */
90 journal = EXT3_SB(sb)->s_journal;
91 if (is_journal_aborted(journal)) {
92 ext3_abort(sb, __func__,
93 "Detected aborted journal");
94 return ERR_PTR(-EROFS);
97 return journal_start(journal, nblocks);
101 * The only special thing we need to do here is to make sure that all
102 * journal_stop calls result in the superblock being marked dirty, so
103 * that sync() will call the filesystem's write_super callback if
106 int __ext3_journal_stop(const char *where, handle_t *handle)
108 struct super_block *sb;
112 sb = handle->h_transaction->t_journal->j_private;
114 rc = journal_stop(handle);
119 __ext3_std_error(sb, where, err);
123 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
124 struct buffer_head *bh, handle_t *handle, int err)
127 const char *errstr = ext3_decode_error(NULL, err, nbuf);
130 BUFFER_TRACE(bh, "abort");
135 if (is_handle_aborted(handle))
138 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
139 caller, errstr, err_fn);
141 journal_abort_handle(handle);
144 void ext3_msg(struct super_block *sb, const char *prefix,
145 const char *fmt, ...)
150 printk("%sEXT3-fs (%s): ", prefix, sb->s_id);
156 /* Deal with the reporting of failure conditions on a filesystem such as
157 * inconsistencies detected or read IO failures.
159 * On ext2, we can store the error state of the filesystem in the
160 * superblock. That is not possible on ext3, because we may have other
161 * write ordering constraints on the superblock which prevent us from
162 * writing it out straight away; and given that the journal is about to
163 * be aborted, we can't rely on the current, or future, transactions to
164 * write out the superblock safely.
166 * We'll just use the journal_abort() error code to record an error in
167 * the journal instead. On recovery, the journal will compain about
168 * that error until we've noted it down and cleared it.
171 static void ext3_handle_error(struct super_block *sb)
173 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
175 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
176 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
178 if (sb->s_flags & MS_RDONLY)
181 if (!test_opt (sb, ERRORS_CONT)) {
182 journal_t *journal = EXT3_SB(sb)->s_journal;
184 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
186 journal_abort(journal, -EIO);
188 if (test_opt (sb, ERRORS_RO)) {
189 ext3_msg(sb, KERN_CRIT,
190 "error: remounting filesystem read-only");
191 sb->s_flags |= MS_RDONLY;
193 ext3_commit_super(sb, es, 1);
194 if (test_opt(sb, ERRORS_PANIC))
195 panic("EXT3-fs (%s): panic forced after error\n",
199 void ext3_error (struct super_block * sb, const char * function,
200 const char * fmt, ...)
205 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
210 ext3_handle_error(sb);
213 static const char *ext3_decode_error(struct super_block * sb, int errno,
220 errstr = "IO failure";
223 errstr = "Out of memory";
226 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
227 errstr = "Journal has aborted";
229 errstr = "Readonly filesystem";
232 /* If the caller passed in an extra buffer for unknown
233 * errors, textualise them now. Else we just return
236 /* Check for truncated error codes... */
237 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
246 /* __ext3_std_error decodes expected errors from journaling functions
247 * automatically and invokes the appropriate error response. */
249 void __ext3_std_error (struct super_block * sb, const char * function,
255 /* Special case: if the error is EROFS, and we're not already
256 * inside a transaction, then there's really no point in logging
258 if (errno == -EROFS && journal_current_handle() == NULL &&
259 (sb->s_flags & MS_RDONLY))
262 errstr = ext3_decode_error(sb, errno, nbuf);
263 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
265 ext3_handle_error(sb);
269 * ext3_abort is a much stronger failure handler than ext3_error. The
270 * abort function may be used to deal with unrecoverable failures such
271 * as journal IO errors or ENOMEM at a critical moment in log management.
273 * We unconditionally force the filesystem into an ABORT|READONLY state,
274 * unless the error response on the fs has been set to panic in which
275 * case we take the easy way out and panic immediately.
278 void ext3_abort (struct super_block * sb, const char * function,
279 const char * fmt, ...)
284 printk(KERN_CRIT "EXT3-fs (%s): error: %s: ", sb->s_id, function);
289 if (test_opt(sb, ERRORS_PANIC))
290 panic("EXT3-fs: panic from previous error\n");
292 if (sb->s_flags & MS_RDONLY)
295 ext3_msg(sb, KERN_CRIT,
296 "error: remounting filesystem read-only");
297 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
298 sb->s_flags |= MS_RDONLY;
299 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
300 if (EXT3_SB(sb)->s_journal)
301 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
304 void ext3_warning (struct super_block * sb, const char * function,
305 const char * fmt, ...)
310 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: ",
317 void ext3_update_dynamic_rev(struct super_block *sb)
319 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
321 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
324 ext3_msg(sb, KERN_WARNING,
325 "warning: updating to rev %d because of "
326 "new feature flag, running e2fsck is recommended",
329 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
330 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
331 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
332 /* leave es->s_feature_*compat flags alone */
333 /* es->s_uuid will be set by e2fsck if empty */
336 * The rest of the superblock fields should be zero, and if not it
337 * means they are likely already in use, so leave them alone. We
338 * can leave it up to e2fsck to clean up any inconsistencies there.
343 * Open the external journal device
345 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
347 struct block_device *bdev;
348 char b[BDEVNAME_SIZE];
350 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
356 ext3_msg(sb, "error: failed to open journal device %s: %ld",
357 __bdevname(dev, b), PTR_ERR(bdev));
363 * Release the journal device
365 static int ext3_blkdev_put(struct block_device *bdev)
368 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
371 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
373 struct block_device *bdev;
376 bdev = sbi->journal_bdev;
378 ret = ext3_blkdev_put(bdev);
379 sbi->journal_bdev = NULL;
384 static inline struct inode *orphan_list_entry(struct list_head *l)
386 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
389 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
393 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
394 le32_to_cpu(sbi->s_es->s_last_orphan));
396 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
397 list_for_each(l, &sbi->s_orphan) {
398 struct inode *inode = orphan_list_entry(l);
399 ext3_msg(sb, KERN_ERR, " "
400 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
401 inode->i_sb->s_id, inode->i_ino, inode,
402 inode->i_mode, inode->i_nlink,
407 static void ext3_put_super (struct super_block * sb)
409 struct ext3_sb_info *sbi = EXT3_SB(sb);
410 struct ext3_super_block *es = sbi->s_es;
415 ext3_xattr_put_super(sb);
416 err = journal_destroy(sbi->s_journal);
417 sbi->s_journal = NULL;
419 ext3_abort(sb, __func__, "Couldn't clean up the journal");
421 if (!(sb->s_flags & MS_RDONLY)) {
422 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
423 es->s_state = cpu_to_le16(sbi->s_mount_state);
424 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
425 mark_buffer_dirty(sbi->s_sbh);
426 ext3_commit_super(sb, es, 1);
429 for (i = 0; i < sbi->s_gdb_count; i++)
430 brelse(sbi->s_group_desc[i]);
431 kfree(sbi->s_group_desc);
432 percpu_counter_destroy(&sbi->s_freeblocks_counter);
433 percpu_counter_destroy(&sbi->s_freeinodes_counter);
434 percpu_counter_destroy(&sbi->s_dirs_counter);
437 for (i = 0; i < MAXQUOTAS; i++)
438 kfree(sbi->s_qf_names[i]);
441 /* Debugging code just in case the in-memory inode orphan list
442 * isn't empty. The on-disk one can be non-empty if we've
443 * detected an error and taken the fs readonly, but the
444 * in-memory list had better be clean by this point. */
445 if (!list_empty(&sbi->s_orphan))
446 dump_orphan_list(sb, sbi);
447 J_ASSERT(list_empty(&sbi->s_orphan));
449 invalidate_bdev(sb->s_bdev);
450 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
452 * Invalidate the journal device's buffers. We don't want them
453 * floating about in memory - the physical journal device may
454 * hotswapped, and it breaks the `ro-after' testing code.
456 sync_blockdev(sbi->journal_bdev);
457 invalidate_bdev(sbi->journal_bdev);
458 ext3_blkdev_remove(sbi);
460 sb->s_fs_info = NULL;
461 kfree(sbi->s_blockgroup_lock);
467 static struct kmem_cache *ext3_inode_cachep;
470 * Called inside transaction, so use GFP_NOFS
472 static struct inode *ext3_alloc_inode(struct super_block *sb)
474 struct ext3_inode_info *ei;
476 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
479 ei->i_block_alloc_info = NULL;
480 ei->vfs_inode.i_version = 1;
481 atomic_set(&ei->i_datasync_tid, 0);
482 atomic_set(&ei->i_sync_tid, 0);
483 return &ei->vfs_inode;
486 static void ext3_destroy_inode(struct inode *inode)
488 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
489 printk("EXT3 Inode %p: orphan list check failed!\n",
491 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
492 EXT3_I(inode), sizeof(struct ext3_inode_info),
496 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
499 static void init_once(void *foo)
501 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
503 INIT_LIST_HEAD(&ei->i_orphan);
504 #ifdef CONFIG_EXT3_FS_XATTR
505 init_rwsem(&ei->xattr_sem);
507 mutex_init(&ei->truncate_mutex);
508 inode_init_once(&ei->vfs_inode);
511 static int init_inodecache(void)
513 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
514 sizeof(struct ext3_inode_info),
515 0, (SLAB_RECLAIM_ACCOUNT|
518 if (ext3_inode_cachep == NULL)
523 static void destroy_inodecache(void)
525 kmem_cache_destroy(ext3_inode_cachep);
528 static void ext3_clear_inode(struct inode *inode)
530 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
531 ext3_discard_reservation(inode);
532 EXT3_I(inode)->i_block_alloc_info = NULL;
537 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
539 #if defined(CONFIG_QUOTA)
540 struct ext3_sb_info *sbi = EXT3_SB(sb);
542 if (sbi->s_jquota_fmt) {
545 switch (sbi->s_jquota_fmt) {
556 seq_printf(seq, ",jqfmt=%s", fmtname);
559 if (sbi->s_qf_names[USRQUOTA])
560 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
562 if (sbi->s_qf_names[GRPQUOTA])
563 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
565 if (test_opt(sb, USRQUOTA))
566 seq_puts(seq, ",usrquota");
568 if (test_opt(sb, GRPQUOTA))
569 seq_puts(seq, ",grpquota");
573 static char *data_mode_string(unsigned long mode)
576 case EXT3_MOUNT_JOURNAL_DATA:
578 case EXT3_MOUNT_ORDERED_DATA:
580 case EXT3_MOUNT_WRITEBACK_DATA:
588 * - it's set to a non-default value OR
589 * - if the per-sb default is different from the global default
591 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
593 struct super_block *sb = vfs->mnt_sb;
594 struct ext3_sb_info *sbi = EXT3_SB(sb);
595 struct ext3_super_block *es = sbi->s_es;
596 unsigned long def_mount_opts;
598 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
600 if (sbi->s_sb_block != 1)
601 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
602 if (test_opt(sb, MINIX_DF))
603 seq_puts(seq, ",minixdf");
604 if (test_opt(sb, GRPID))
605 seq_puts(seq, ",grpid");
606 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
607 seq_puts(seq, ",nogrpid");
608 if (sbi->s_resuid != EXT3_DEF_RESUID ||
609 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
610 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
612 if (sbi->s_resgid != EXT3_DEF_RESGID ||
613 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
614 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
616 if (test_opt(sb, ERRORS_RO)) {
617 int def_errors = le16_to_cpu(es->s_errors);
619 if (def_errors == EXT3_ERRORS_PANIC ||
620 def_errors == EXT3_ERRORS_CONTINUE) {
621 seq_puts(seq, ",errors=remount-ro");
624 if (test_opt(sb, ERRORS_CONT))
625 seq_puts(seq, ",errors=continue");
626 if (test_opt(sb, ERRORS_PANIC))
627 seq_puts(seq, ",errors=panic");
628 if (test_opt(sb, NO_UID32))
629 seq_puts(seq, ",nouid32");
630 if (test_opt(sb, DEBUG))
631 seq_puts(seq, ",debug");
632 if (test_opt(sb, OLDALLOC))
633 seq_puts(seq, ",oldalloc");
634 #ifdef CONFIG_EXT3_FS_XATTR
635 if (test_opt(sb, XATTR_USER))
636 seq_puts(seq, ",user_xattr");
637 if (!test_opt(sb, XATTR_USER) &&
638 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
639 seq_puts(seq, ",nouser_xattr");
642 #ifdef CONFIG_EXT3_FS_POSIX_ACL
643 if (test_opt(sb, POSIX_ACL))
644 seq_puts(seq, ",acl");
645 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
646 seq_puts(seq, ",noacl");
648 if (!test_opt(sb, RESERVATION))
649 seq_puts(seq, ",noreservation");
650 if (sbi->s_commit_interval) {
651 seq_printf(seq, ",commit=%u",
652 (unsigned) (sbi->s_commit_interval / HZ));
654 if (test_opt(sb, BARRIER))
655 seq_puts(seq, ",barrier=1");
656 if (test_opt(sb, NOBH))
657 seq_puts(seq, ",nobh");
659 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
660 if (test_opt(sb, DATA_ERR_ABORT))
661 seq_puts(seq, ",data_err=abort");
663 if (test_opt(sb, NOLOAD))
664 seq_puts(seq, ",norecovery");
666 ext3_show_quota_options(seq, sb);
672 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
673 u64 ino, u32 generation)
677 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
678 return ERR_PTR(-ESTALE);
679 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
680 return ERR_PTR(-ESTALE);
682 /* iget isn't really right if the inode is currently unallocated!!
684 * ext3_read_inode will return a bad_inode if the inode had been
685 * deleted, so we should be safe.
687 * Currently we don't know the generation for parent directory, so
688 * a generation of 0 means "accept any"
690 inode = ext3_iget(sb, ino);
692 return ERR_CAST(inode);
693 if (generation && inode->i_generation != generation) {
695 return ERR_PTR(-ESTALE);
701 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
702 int fh_len, int fh_type)
704 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
708 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
709 int fh_len, int fh_type)
711 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
716 * Try to release metadata pages (indirect blocks, directories) which are
717 * mapped via the block device. Since these pages could have journal heads
718 * which would prevent try_to_free_buffers() from freeing them, we must use
719 * jbd layer's try_to_free_buffers() function to release them.
721 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
724 journal_t *journal = EXT3_SB(sb)->s_journal;
726 WARN_ON(PageChecked(page));
727 if (!page_has_buffers(page))
730 return journal_try_to_free_buffers(journal, page,
732 return try_to_free_buffers(page);
736 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
737 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
739 static int ext3_write_dquot(struct dquot *dquot);
740 static int ext3_acquire_dquot(struct dquot *dquot);
741 static int ext3_release_dquot(struct dquot *dquot);
742 static int ext3_mark_dquot_dirty(struct dquot *dquot);
743 static int ext3_write_info(struct super_block *sb, int type);
744 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
745 char *path, int remount);
746 static int ext3_quota_on_mount(struct super_block *sb, int type);
747 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
748 size_t len, loff_t off);
749 static ssize_t ext3_quota_write(struct super_block *sb, int type,
750 const char *data, size_t len, loff_t off);
752 static const struct dquot_operations ext3_quota_operations = {
753 .initialize = dquot_initialize,
755 .alloc_inode = dquot_alloc_inode,
756 .free_inode = dquot_free_inode,
757 .transfer = dquot_transfer,
758 .write_dquot = ext3_write_dquot,
759 .acquire_dquot = ext3_acquire_dquot,
760 .release_dquot = ext3_release_dquot,
761 .mark_dirty = ext3_mark_dquot_dirty,
762 .write_info = ext3_write_info,
763 .alloc_dquot = dquot_alloc,
764 .destroy_dquot = dquot_destroy,
767 static const struct quotactl_ops ext3_qctl_operations = {
768 .quota_on = ext3_quota_on,
769 .quota_off = vfs_quota_off,
770 .quota_sync = vfs_quota_sync,
771 .get_info = vfs_get_dqinfo,
772 .set_info = vfs_set_dqinfo,
773 .get_dqblk = vfs_get_dqblk,
774 .set_dqblk = vfs_set_dqblk
778 static const struct super_operations ext3_sops = {
779 .alloc_inode = ext3_alloc_inode,
780 .destroy_inode = ext3_destroy_inode,
781 .write_inode = ext3_write_inode,
782 .dirty_inode = ext3_dirty_inode,
783 .delete_inode = ext3_delete_inode,
784 .put_super = ext3_put_super,
785 .sync_fs = ext3_sync_fs,
786 .freeze_fs = ext3_freeze,
787 .unfreeze_fs = ext3_unfreeze,
788 .statfs = ext3_statfs,
789 .remount_fs = ext3_remount,
790 .clear_inode = ext3_clear_inode,
791 .show_options = ext3_show_options,
793 .quota_read = ext3_quota_read,
794 .quota_write = ext3_quota_write,
796 .bdev_try_to_free_page = bdev_try_to_free_page,
799 static const struct export_operations ext3_export_ops = {
800 .fh_to_dentry = ext3_fh_to_dentry,
801 .fh_to_parent = ext3_fh_to_parent,
802 .get_parent = ext3_get_parent,
806 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
807 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
808 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
809 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
810 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
811 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
812 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
813 Opt_data_err_abort, Opt_data_err_ignore,
814 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
815 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
816 Opt_noquota, Opt_ignore, Opt_barrier, Opt_err, Opt_resize,
817 Opt_usrquota, Opt_grpquota
820 static const match_table_t tokens = {
821 {Opt_bsd_df, "bsddf"},
822 {Opt_minix_df, "minixdf"},
823 {Opt_grpid, "grpid"},
824 {Opt_grpid, "bsdgroups"},
825 {Opt_nogrpid, "nogrpid"},
826 {Opt_nogrpid, "sysvgroups"},
827 {Opt_resgid, "resgid=%u"},
828 {Opt_resuid, "resuid=%u"},
830 {Opt_err_cont, "errors=continue"},
831 {Opt_err_panic, "errors=panic"},
832 {Opt_err_ro, "errors=remount-ro"},
833 {Opt_nouid32, "nouid32"},
834 {Opt_nocheck, "nocheck"},
835 {Opt_nocheck, "check=none"},
836 {Opt_debug, "debug"},
837 {Opt_oldalloc, "oldalloc"},
838 {Opt_orlov, "orlov"},
839 {Opt_user_xattr, "user_xattr"},
840 {Opt_nouser_xattr, "nouser_xattr"},
842 {Opt_noacl, "noacl"},
843 {Opt_reservation, "reservation"},
844 {Opt_noreservation, "noreservation"},
845 {Opt_noload, "noload"},
846 {Opt_noload, "norecovery"},
849 {Opt_commit, "commit=%u"},
850 {Opt_journal_update, "journal=update"},
851 {Opt_journal_inum, "journal=%u"},
852 {Opt_journal_dev, "journal_dev=%u"},
853 {Opt_abort, "abort"},
854 {Opt_data_journal, "data=journal"},
855 {Opt_data_ordered, "data=ordered"},
856 {Opt_data_writeback, "data=writeback"},
857 {Opt_data_err_abort, "data_err=abort"},
858 {Opt_data_err_ignore, "data_err=ignore"},
859 {Opt_offusrjquota, "usrjquota="},
860 {Opt_usrjquota, "usrjquota=%s"},
861 {Opt_offgrpjquota, "grpjquota="},
862 {Opt_grpjquota, "grpjquota=%s"},
863 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
864 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
865 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
866 {Opt_grpquota, "grpquota"},
867 {Opt_noquota, "noquota"},
868 {Opt_quota, "quota"},
869 {Opt_usrquota, "usrquota"},
870 {Opt_barrier, "barrier=%u"},
871 {Opt_resize, "resize"},
875 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
877 ext3_fsblk_t sb_block;
878 char *options = (char *) *data;
880 if (!options || strncmp(options, "sb=", 3) != 0)
881 return 1; /* Default location */
883 /*todo: use simple_strtoll with >32bit ext3 */
884 sb_block = simple_strtoul(options, &options, 0);
885 if (*options && *options != ',') {
886 ext3_msg(sb, "error: invalid sb specification: %s",
892 *data = (void *) options;
897 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
899 struct ext3_sb_info *sbi = EXT3_SB(sb);
902 if (sb_any_quota_loaded(sb) &&
903 !sbi->s_qf_names[qtype]) {
904 ext3_msg(sb, KERN_ERR,
905 "Cannot change journaled "
906 "quota options when quota turned on");
909 qname = match_strdup(args);
911 ext3_msg(sb, KERN_ERR,
912 "Not enough memory for storing quotafile name");
915 if (sbi->s_qf_names[qtype] &&
916 strcmp(sbi->s_qf_names[qtype], qname)) {
917 ext3_msg(sb, KERN_ERR,
918 "%s quota file already specified", QTYPE2NAME(qtype));
922 sbi->s_qf_names[qtype] = qname;
923 if (strchr(sbi->s_qf_names[qtype], '/')) {
924 ext3_msg(sb, KERN_ERR,
925 "quotafile must be on filesystem root");
926 kfree(sbi->s_qf_names[qtype]);
927 sbi->s_qf_names[qtype] = NULL;
930 set_opt(sbi->s_mount_opt, QUOTA);
934 static int clear_qf_name(struct super_block *sb, int qtype) {
936 struct ext3_sb_info *sbi = EXT3_SB(sb);
938 if (sb_any_quota_loaded(sb) &&
939 sbi->s_qf_names[qtype]) {
940 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
941 " when quota turned on");
945 * The space will be released later when all options are confirmed
948 sbi->s_qf_names[qtype] = NULL;
953 static int parse_options (char *options, struct super_block *sb,
954 unsigned int *inum, unsigned long *journal_devnum,
955 ext3_fsblk_t *n_blocks_count, int is_remount)
957 struct ext3_sb_info *sbi = EXT3_SB(sb);
959 substring_t args[MAX_OPT_ARGS];
969 while ((p = strsep (&options, ",")) != NULL) {
974 token = match_token(p, tokens, args);
977 clear_opt (sbi->s_mount_opt, MINIX_DF);
980 set_opt (sbi->s_mount_opt, MINIX_DF);
983 set_opt (sbi->s_mount_opt, GRPID);
986 clear_opt (sbi->s_mount_opt, GRPID);
989 if (match_int(&args[0], &option))
991 sbi->s_resuid = option;
994 if (match_int(&args[0], &option))
996 sbi->s_resgid = option;
999 /* handled by get_sb_block() instead of here */
1000 /* *sb_block = match_int(&args[0]); */
1003 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1004 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1005 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1008 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1009 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1010 set_opt (sbi->s_mount_opt, ERRORS_RO);
1013 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1014 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1015 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1018 set_opt (sbi->s_mount_opt, NO_UID32);
1021 clear_opt (sbi->s_mount_opt, CHECK);
1024 set_opt (sbi->s_mount_opt, DEBUG);
1027 set_opt (sbi->s_mount_opt, OLDALLOC);
1030 clear_opt (sbi->s_mount_opt, OLDALLOC);
1032 #ifdef CONFIG_EXT3_FS_XATTR
1033 case Opt_user_xattr:
1034 set_opt (sbi->s_mount_opt, XATTR_USER);
1036 case Opt_nouser_xattr:
1037 clear_opt (sbi->s_mount_opt, XATTR_USER);
1040 case Opt_user_xattr:
1041 case Opt_nouser_xattr:
1042 ext3_msg(sb, KERN_INFO,
1043 "(no)user_xattr options not supported");
1046 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1048 set_opt(sbi->s_mount_opt, POSIX_ACL);
1051 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1056 ext3_msg(sb, KERN_INFO,
1057 "(no)acl options not supported");
1060 case Opt_reservation:
1061 set_opt(sbi->s_mount_opt, RESERVATION);
1063 case Opt_noreservation:
1064 clear_opt(sbi->s_mount_opt, RESERVATION);
1066 case Opt_journal_update:
1068 /* Eventually we will want to be able to create
1069 a journal file here. For now, only allow the
1070 user to specify an existing inode to be the
1073 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1074 "journal on remount");
1077 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1079 case Opt_journal_inum:
1081 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1082 "journal on remount");
1085 if (match_int(&args[0], &option))
1089 case Opt_journal_dev:
1091 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1092 "journal on remount");
1095 if (match_int(&args[0], &option))
1097 *journal_devnum = option;
1100 set_opt (sbi->s_mount_opt, NOLOAD);
1103 if (match_int(&args[0], &option))
1108 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1109 sbi->s_commit_interval = HZ * option;
1111 case Opt_data_journal:
1112 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1114 case Opt_data_ordered:
1115 data_opt = EXT3_MOUNT_ORDERED_DATA;
1117 case Opt_data_writeback:
1118 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1121 if (test_opt(sb, DATA_FLAGS) == data_opt)
1123 ext3_msg(sb, KERN_ERR,
1124 "error: cannot change "
1125 "data mode on remount. The filesystem "
1126 "is mounted in data=%s mode and you "
1127 "try to remount it in data=%s mode.",
1128 data_mode_string(test_opt(sb,
1130 data_mode_string(data_opt));
1133 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1134 sbi->s_mount_opt |= data_opt;
1137 case Opt_data_err_abort:
1138 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1140 case Opt_data_err_ignore:
1141 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1145 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1149 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1152 case Opt_offusrjquota:
1153 if (!clear_qf_name(sb, USRQUOTA))
1156 case Opt_offgrpjquota:
1157 if (!clear_qf_name(sb, GRPQUOTA))
1160 case Opt_jqfmt_vfsold:
1161 qfmt = QFMT_VFS_OLD;
1163 case Opt_jqfmt_vfsv0:
1166 case Opt_jqfmt_vfsv1:
1169 if (sb_any_quota_loaded(sb) &&
1170 sbi->s_jquota_fmt != qfmt) {
1171 ext3_msg(sb, KERN_ERR, "error: cannot change "
1172 "journaled quota options when "
1173 "quota turned on.");
1176 sbi->s_jquota_fmt = qfmt;
1180 set_opt(sbi->s_mount_opt, QUOTA);
1181 set_opt(sbi->s_mount_opt, USRQUOTA);
1184 set_opt(sbi->s_mount_opt, QUOTA);
1185 set_opt(sbi->s_mount_opt, GRPQUOTA);
1188 if (sb_any_quota_loaded(sb)) {
1189 ext3_msg(sb, KERN_ERR, "error: cannot change "
1190 "quota options when quota turned on.");
1193 clear_opt(sbi->s_mount_opt, QUOTA);
1194 clear_opt(sbi->s_mount_opt, USRQUOTA);
1195 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1201 ext3_msg(sb, KERN_ERR,
1202 "error: quota options not supported.");
1206 case Opt_offusrjquota:
1207 case Opt_offgrpjquota:
1208 case Opt_jqfmt_vfsold:
1209 case Opt_jqfmt_vfsv0:
1210 case Opt_jqfmt_vfsv1:
1211 ext3_msg(sb, KERN_ERR,
1212 "error: journaled quota options not "
1219 set_opt(sbi->s_mount_opt, ABORT);
1222 if (match_int(&args[0], &option))
1225 set_opt(sbi->s_mount_opt, BARRIER);
1227 clear_opt(sbi->s_mount_opt, BARRIER);
1233 ext3_msg(sb, KERN_ERR,
1234 "error: resize option only available "
1238 if (match_int(&args[0], &option) != 0)
1240 *n_blocks_count = option;
1243 set_opt(sbi->s_mount_opt, NOBH);
1246 clear_opt(sbi->s_mount_opt, NOBH);
1249 ext3_msg(sb, KERN_ERR,
1250 "error: unrecognized mount option \"%s\" "
1251 "or missing value", p);
1256 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1257 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1258 clear_opt(sbi->s_mount_opt, USRQUOTA);
1259 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1260 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1262 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1263 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1268 if (!sbi->s_jquota_fmt) {
1269 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1274 if (sbi->s_jquota_fmt) {
1275 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1276 "specified with no journaling "
1285 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1288 struct ext3_sb_info *sbi = EXT3_SB(sb);
1291 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1292 ext3_msg(sb, KERN_ERR,
1293 "error: revision level too high, "
1294 "forcing read-only mode");
1299 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1300 ext3_msg(sb, KERN_WARNING,
1301 "warning: mounting unchecked fs, "
1302 "running e2fsck is recommended");
1303 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1304 ext3_msg(sb, KERN_WARNING,
1305 "warning: mounting fs with errors, "
1306 "running e2fsck is recommended");
1307 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1308 le16_to_cpu(es->s_mnt_count) >=
1309 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1310 ext3_msg(sb, KERN_WARNING,
1311 "warning: maximal mount count reached, "
1312 "running e2fsck is recommended");
1313 else if (le32_to_cpu(es->s_checkinterval) &&
1314 (le32_to_cpu(es->s_lastcheck) +
1315 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1316 ext3_msg(sb, KERN_WARNING,
1317 "warning: checktime reached, "
1318 "running e2fsck is recommended");
1320 /* @@@ We _will_ want to clear the valid bit if we find
1321 inconsistencies, to force a fsck at reboot. But for
1322 a plain journaled filesystem we can keep it set as
1323 valid forever! :) */
1324 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1326 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1327 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1328 le16_add_cpu(&es->s_mnt_count, 1);
1329 es->s_mtime = cpu_to_le32(get_seconds());
1330 ext3_update_dynamic_rev(sb);
1331 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1333 ext3_commit_super(sb, es, 1);
1334 if (test_opt(sb, DEBUG))
1335 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1336 "bpg=%lu, ipg=%lu, mo=%04lx]",
1338 sbi->s_groups_count,
1339 EXT3_BLOCKS_PER_GROUP(sb),
1340 EXT3_INODES_PER_GROUP(sb),
1343 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1344 char b[BDEVNAME_SIZE];
1345 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1346 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1348 ext3_msg(sb, KERN_INFO, "using internal journal");
1353 /* Called at mount-time, super-block is locked */
1354 static int ext3_check_descriptors(struct super_block *sb)
1356 struct ext3_sb_info *sbi = EXT3_SB(sb);
1359 ext3_debug ("Checking group descriptors");
1361 for (i = 0; i < sbi->s_groups_count; i++) {
1362 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1363 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1364 ext3_fsblk_t last_block;
1366 if (i == sbi->s_groups_count - 1)
1367 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1369 last_block = first_block +
1370 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1372 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1373 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1375 ext3_error (sb, "ext3_check_descriptors",
1376 "Block bitmap for group %d"
1377 " not in group (block %lu)!",
1379 le32_to_cpu(gdp->bg_block_bitmap));
1382 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1383 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1385 ext3_error (sb, "ext3_check_descriptors",
1386 "Inode bitmap for group %d"
1387 " not in group (block %lu)!",
1389 le32_to_cpu(gdp->bg_inode_bitmap));
1392 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1393 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1396 ext3_error (sb, "ext3_check_descriptors",
1397 "Inode table for group %d"
1398 " not in group (block %lu)!",
1400 le32_to_cpu(gdp->bg_inode_table));
1405 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1406 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1411 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1412 * the superblock) which were deleted from all directories, but held open by
1413 * a process at the time of a crash. We walk the list and try to delete these
1414 * inodes at recovery time (only with a read-write filesystem).
1416 * In order to keep the orphan inode chain consistent during traversal (in
1417 * case of crash during recovery), we link each inode into the superblock
1418 * orphan list_head and handle it the same way as an inode deletion during
1419 * normal operation (which journals the operations for us).
1421 * We only do an iget() and an iput() on each inode, which is very safe if we
1422 * accidentally point at an in-use or already deleted inode. The worst that
1423 * can happen in this case is that we get a "bit already cleared" message from
1424 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1425 * e2fsck was run on this filesystem, and it must have already done the orphan
1426 * inode cleanup for us, so we can safely abort without any further action.
1428 static void ext3_orphan_cleanup (struct super_block * sb,
1429 struct ext3_super_block * es)
1431 unsigned int s_flags = sb->s_flags;
1432 int nr_orphans = 0, nr_truncates = 0;
1436 if (!es->s_last_orphan) {
1437 jbd_debug(4, "no orphan inodes to clean up\n");
1441 if (bdev_read_only(sb->s_bdev)) {
1442 ext3_msg(sb, KERN_ERR, "error: write access "
1443 "unavailable, skipping orphan cleanup.");
1447 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1448 if (es->s_last_orphan)
1449 jbd_debug(1, "Errors on filesystem, "
1450 "clearing orphan list.\n");
1451 es->s_last_orphan = 0;
1452 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1456 if (s_flags & MS_RDONLY) {
1457 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1458 sb->s_flags &= ~MS_RDONLY;
1461 /* Needed for iput() to work correctly and not trash data */
1462 sb->s_flags |= MS_ACTIVE;
1463 /* Turn on quotas so that they are updated correctly */
1464 for (i = 0; i < MAXQUOTAS; i++) {
1465 if (EXT3_SB(sb)->s_qf_names[i]) {
1466 int ret = ext3_quota_on_mount(sb, i);
1468 ext3_msg(sb, KERN_ERR,
1469 "error: cannot turn on journaled "
1475 while (es->s_last_orphan) {
1476 struct inode *inode;
1478 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1479 if (IS_ERR(inode)) {
1480 es->s_last_orphan = 0;
1484 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1486 if (inode->i_nlink) {
1488 "%s: truncating inode %lu to %Ld bytes\n",
1489 __func__, inode->i_ino, inode->i_size);
1490 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1491 inode->i_ino, inode->i_size);
1492 ext3_truncate(inode);
1496 "%s: deleting unreferenced inode %lu\n",
1497 __func__, inode->i_ino);
1498 jbd_debug(2, "deleting unreferenced inode %lu\n",
1502 iput(inode); /* The delete magic happens here! */
1505 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1508 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1509 PLURAL(nr_orphans));
1511 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1512 PLURAL(nr_truncates));
1514 /* Turn quotas off */
1515 for (i = 0; i < MAXQUOTAS; i++) {
1516 if (sb_dqopt(sb)->files[i])
1517 vfs_quota_off(sb, i, 0);
1520 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1524 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1525 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1526 * We need to be 1 filesystem block less than the 2^32 sector limit.
1528 static loff_t ext3_max_size(int bits)
1530 loff_t res = EXT3_NDIR_BLOCKS;
1534 /* This is calculated to be the largest file size for a
1535 * dense, file such that the total number of
1536 * sectors in the file, including data and all indirect blocks,
1537 * does not exceed 2^32 -1
1538 * __u32 i_blocks representing the total number of
1539 * 512 bytes blocks of the file
1541 upper_limit = (1LL << 32) - 1;
1543 /* total blocks in file system block size */
1544 upper_limit >>= (bits - 9);
1547 /* indirect blocks */
1549 /* double indirect blocks */
1550 meta_blocks += 1 + (1LL << (bits-2));
1551 /* tripple indirect blocks */
1552 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1554 upper_limit -= meta_blocks;
1555 upper_limit <<= bits;
1557 res += 1LL << (bits-2);
1558 res += 1LL << (2*(bits-2));
1559 res += 1LL << (3*(bits-2));
1561 if (res > upper_limit)
1564 if (res > MAX_LFS_FILESIZE)
1565 res = MAX_LFS_FILESIZE;
1570 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1571 ext3_fsblk_t logic_sb_block,
1574 struct ext3_sb_info *sbi = EXT3_SB(sb);
1575 unsigned long bg, first_meta_bg;
1578 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1580 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1582 return (logic_sb_block + nr + 1);
1583 bg = sbi->s_desc_per_block * nr;
1584 if (ext3_bg_has_super(sb, bg))
1586 return (has_super + ext3_group_first_block_no(sb, bg));
1590 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1592 struct buffer_head * bh;
1593 struct ext3_super_block *es = NULL;
1594 struct ext3_sb_info *sbi;
1596 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1597 ext3_fsblk_t logic_sb_block;
1598 unsigned long offset = 0;
1599 unsigned int journal_inum = 0;
1600 unsigned long journal_devnum = 0;
1601 unsigned long def_mount_opts;
1612 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1616 sbi->s_blockgroup_lock =
1617 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1618 if (!sbi->s_blockgroup_lock) {
1622 sb->s_fs_info = sbi;
1623 sbi->s_mount_opt = 0;
1624 sbi->s_resuid = EXT3_DEF_RESUID;
1625 sbi->s_resgid = EXT3_DEF_RESGID;
1626 sbi->s_sb_block = sb_block;
1630 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1632 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1637 * The ext3 superblock will not be buffer aligned for other than 1kB
1638 * block sizes. We need to calculate the offset from buffer start.
1640 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1641 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1642 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1644 logic_sb_block = sb_block;
1647 if (!(bh = sb_bread(sb, logic_sb_block))) {
1648 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1652 * Note: s_es must be initialized as soon as possible because
1653 * some ext3 macro-instructions depend on its value
1655 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1657 sb->s_magic = le16_to_cpu(es->s_magic);
1658 if (sb->s_magic != EXT3_SUPER_MAGIC)
1661 /* Set defaults before we parse the mount options */
1662 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1663 if (def_mount_opts & EXT3_DEFM_DEBUG)
1664 set_opt(sbi->s_mount_opt, DEBUG);
1665 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1666 set_opt(sbi->s_mount_opt, GRPID);
1667 if (def_mount_opts & EXT3_DEFM_UID16)
1668 set_opt(sbi->s_mount_opt, NO_UID32);
1669 #ifdef CONFIG_EXT3_FS_XATTR
1670 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1671 set_opt(sbi->s_mount_opt, XATTR_USER);
1673 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1674 if (def_mount_opts & EXT3_DEFM_ACL)
1675 set_opt(sbi->s_mount_opt, POSIX_ACL);
1677 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1678 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1679 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1680 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1681 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1682 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1684 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1685 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1686 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1687 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1689 set_opt(sbi->s_mount_opt, ERRORS_RO);
1691 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1692 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1694 set_opt(sbi->s_mount_opt, RESERVATION);
1696 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1700 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1701 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1703 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1704 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1705 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1706 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1707 ext3_msg(sb, KERN_WARNING,
1708 "warning: feature flags set on rev 0 fs, "
1709 "running e2fsck is recommended");
1711 * Check feature flags regardless of the revision level, since we
1712 * previously didn't change the revision level when setting the flags,
1713 * so there is a chance incompat flags are set on a rev 0 filesystem.
1715 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1717 ext3_msg(sb, KERN_ERR,
1718 "error: couldn't mount because of unsupported "
1719 "optional features (%x)", le32_to_cpu(features));
1722 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1723 if (!(sb->s_flags & MS_RDONLY) && features) {
1724 ext3_msg(sb, KERN_ERR,
1725 "error: couldn't mount RDWR because of unsupported "
1726 "optional features (%x)", le32_to_cpu(features));
1729 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1731 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1732 blocksize > EXT3_MAX_BLOCK_SIZE) {
1733 ext3_msg(sb, KERN_ERR,
1734 "error: couldn't mount because of unsupported "
1735 "filesystem blocksize %d", blocksize);
1739 hblock = bdev_logical_block_size(sb->s_bdev);
1740 if (sb->s_blocksize != blocksize) {
1742 * Make sure the blocksize for the filesystem is larger
1743 * than the hardware sectorsize for the machine.
1745 if (blocksize < hblock) {
1746 ext3_msg(sb, KERN_ERR,
1747 "error: fsblocksize %d too small for "
1748 "hardware sectorsize %d", blocksize, hblock);
1753 if (!sb_set_blocksize(sb, blocksize)) {
1754 ext3_msg(sb, KERN_ERR,
1755 "error: bad blocksize %d", blocksize);
1758 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1759 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1760 bh = sb_bread(sb, logic_sb_block);
1762 ext3_msg(sb, KERN_ERR,
1763 "error: can't read superblock on 2nd try");
1766 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1768 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1769 ext3_msg(sb, KERN_ERR,
1770 "error: magic mismatch");
1775 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1777 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1778 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1779 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1781 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1782 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1783 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1784 (!is_power_of_2(sbi->s_inode_size)) ||
1785 (sbi->s_inode_size > blocksize)) {
1786 ext3_msg(sb, KERN_ERR,
1787 "error: unsupported inode size: %d",
1792 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1793 le32_to_cpu(es->s_log_frag_size);
1794 if (blocksize != sbi->s_frag_size) {
1795 ext3_msg(sb, KERN_ERR,
1796 "error: fragsize %lu != blocksize %u (unsupported)",
1797 sbi->s_frag_size, blocksize);
1800 sbi->s_frags_per_block = 1;
1801 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1802 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1803 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1804 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1806 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1807 if (sbi->s_inodes_per_block == 0)
1809 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1810 sbi->s_inodes_per_block;
1811 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1813 sbi->s_mount_state = le16_to_cpu(es->s_state);
1814 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1815 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1816 for (i=0; i < 4; i++)
1817 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1818 sbi->s_def_hash_version = es->s_def_hash_version;
1819 i = le32_to_cpu(es->s_flags);
1820 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1821 sbi->s_hash_unsigned = 3;
1822 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1823 #ifdef __CHAR_UNSIGNED__
1824 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1825 sbi->s_hash_unsigned = 3;
1827 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1831 if (sbi->s_blocks_per_group > blocksize * 8) {
1832 ext3_msg(sb, KERN_ERR,
1833 "#blocks per group too big: %lu",
1834 sbi->s_blocks_per_group);
1837 if (sbi->s_frags_per_group > blocksize * 8) {
1838 ext3_msg(sb, KERN_ERR,
1839 "error: #fragments per group too big: %lu",
1840 sbi->s_frags_per_group);
1843 if (sbi->s_inodes_per_group > blocksize * 8) {
1844 ext3_msg(sb, KERN_ERR,
1845 "error: #inodes per group too big: %lu",
1846 sbi->s_inodes_per_group);
1850 if (le32_to_cpu(es->s_blocks_count) >
1851 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1852 ext3_msg(sb, KERN_ERR,
1853 "error: filesystem is too large to mount safely");
1854 if (sizeof(sector_t) < 8)
1855 ext3_msg(sb, KERN_ERR,
1856 "error: CONFIG_LBDAF not enabled");
1860 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1862 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1863 le32_to_cpu(es->s_first_data_block) - 1)
1864 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1865 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1866 EXT3_DESC_PER_BLOCK(sb);
1867 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1869 if (sbi->s_group_desc == NULL) {
1870 ext3_msg(sb, KERN_ERR,
1871 "error: not enough memory");
1875 bgl_lock_init(sbi->s_blockgroup_lock);
1877 for (i = 0; i < db_count; i++) {
1878 block = descriptor_loc(sb, logic_sb_block, i);
1879 sbi->s_group_desc[i] = sb_bread(sb, block);
1880 if (!sbi->s_group_desc[i]) {
1881 ext3_msg(sb, KERN_ERR,
1882 "error: can't read group descriptor %d", i);
1887 if (!ext3_check_descriptors (sb)) {
1888 ext3_msg(sb, KERN_ERR,
1889 "error: group descriptors corrupted");
1892 sbi->s_gdb_count = db_count;
1893 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1894 spin_lock_init(&sbi->s_next_gen_lock);
1896 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1897 ext3_count_free_blocks(sb));
1899 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1900 ext3_count_free_inodes(sb));
1903 err = percpu_counter_init(&sbi->s_dirs_counter,
1904 ext3_count_dirs(sb));
1907 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
1911 /* per fileystem reservation list head & lock */
1912 spin_lock_init(&sbi->s_rsv_window_lock);
1913 sbi->s_rsv_window_root = RB_ROOT;
1914 /* Add a single, static dummy reservation to the start of the
1915 * reservation window list --- it gives us a placeholder for
1916 * append-at-start-of-list which makes the allocation logic
1917 * _much_ simpler. */
1918 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1919 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1920 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1921 sbi->s_rsv_window_head.rsv_goal_size = 0;
1922 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1925 * set up enough so that it can read an inode
1927 sb->s_op = &ext3_sops;
1928 sb->s_export_op = &ext3_export_ops;
1929 sb->s_xattr = ext3_xattr_handlers;
1931 sb->s_qcop = &ext3_qctl_operations;
1932 sb->dq_op = &ext3_quota_operations;
1934 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1935 mutex_init(&sbi->s_orphan_lock);
1936 mutex_init(&sbi->s_resize_lock);
1940 needs_recovery = (es->s_last_orphan != 0 ||
1941 EXT3_HAS_INCOMPAT_FEATURE(sb,
1942 EXT3_FEATURE_INCOMPAT_RECOVER));
1945 * The first inode we look at is the journal inode. Don't try
1946 * root first: it may be modified in the journal!
1948 if (!test_opt(sb, NOLOAD) &&
1949 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1950 if (ext3_load_journal(sb, es, journal_devnum))
1952 } else if (journal_inum) {
1953 if (ext3_create_journal(sb, es, journal_inum))
1957 ext3_msg(sb, KERN_ERR,
1958 "error: no journal found. "
1959 "mounting ext3 over ext2?");
1963 /* We have now updated the journal if required, so we can
1964 * validate the data journaling mode. */
1965 switch (test_opt(sb, DATA_FLAGS)) {
1967 /* No mode set, assume a default based on the journal
1968 capabilities: ORDERED_DATA if the journal can
1969 cope, else JOURNAL_DATA */
1970 if (journal_check_available_features
1971 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1972 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1974 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1977 case EXT3_MOUNT_ORDERED_DATA:
1978 case EXT3_MOUNT_WRITEBACK_DATA:
1979 if (!journal_check_available_features
1980 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1981 ext3_msg(sb, KERN_ERR,
1982 "error: journal does not support "
1983 "requested data journaling mode");
1990 if (test_opt(sb, NOBH)) {
1991 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1992 ext3_msg(sb, KERN_WARNING,
1993 "warning: ignoring nobh option - "
1994 "it is supported only with writeback mode");
1995 clear_opt(sbi->s_mount_opt, NOBH);
1999 * The journal_load will have done any necessary log recovery,
2000 * so we can safely mount the rest of the filesystem now.
2003 root = ext3_iget(sb, EXT3_ROOT_INO);
2005 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
2006 ret = PTR_ERR(root);
2009 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2011 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
2014 sb->s_root = d_alloc_root(root);
2016 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2022 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2024 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2025 ext3_orphan_cleanup(sb, es);
2026 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2028 ext3_msg(sb, KERN_INFO, "recovery complete");
2029 ext3_mark_recovery_complete(sb, es);
2030 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2031 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2032 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2040 ext3_msg(sb, KERN_INFO,
2041 "error: can't find ext3 filesystem on dev %s.",
2046 journal_destroy(sbi->s_journal);
2048 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2049 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2050 percpu_counter_destroy(&sbi->s_dirs_counter);
2052 for (i = 0; i < db_count; i++)
2053 brelse(sbi->s_group_desc[i]);
2054 kfree(sbi->s_group_desc);
2057 for (i = 0; i < MAXQUOTAS; i++)
2058 kfree(sbi->s_qf_names[i]);
2060 ext3_blkdev_remove(sbi);
2063 sb->s_fs_info = NULL;
2064 kfree(sbi->s_blockgroup_lock);
2071 * Setup any per-fs journal parameters now. We'll do this both on
2072 * initial mount, once the journal has been initialised but before we've
2073 * done any recovery; and again on any subsequent remount.
2075 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2077 struct ext3_sb_info *sbi = EXT3_SB(sb);
2079 if (sbi->s_commit_interval)
2080 journal->j_commit_interval = sbi->s_commit_interval;
2081 /* We could also set up an ext3-specific default for the commit
2082 * interval here, but for now we'll just fall back to the jbd
2085 spin_lock(&journal->j_state_lock);
2086 if (test_opt(sb, BARRIER))
2087 journal->j_flags |= JFS_BARRIER;
2089 journal->j_flags &= ~JFS_BARRIER;
2090 if (test_opt(sb, DATA_ERR_ABORT))
2091 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2093 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2094 spin_unlock(&journal->j_state_lock);
2097 static journal_t *ext3_get_journal(struct super_block *sb,
2098 unsigned int journal_inum)
2100 struct inode *journal_inode;
2103 /* First, test for the existence of a valid inode on disk. Bad
2104 * things happen if we iget() an unused inode, as the subsequent
2105 * iput() will try to delete it. */
2107 journal_inode = ext3_iget(sb, journal_inum);
2108 if (IS_ERR(journal_inode)) {
2109 ext3_msg(sb, KERN_ERR, "error: no journal found");
2112 if (!journal_inode->i_nlink) {
2113 make_bad_inode(journal_inode);
2114 iput(journal_inode);
2115 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2119 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2120 journal_inode, journal_inode->i_size);
2121 if (!S_ISREG(journal_inode->i_mode)) {
2122 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2123 iput(journal_inode);
2127 journal = journal_init_inode(journal_inode);
2129 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2130 iput(journal_inode);
2133 journal->j_private = sb;
2134 ext3_init_journal_params(sb, journal);
2138 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2141 struct buffer_head * bh;
2145 int hblock, blocksize;
2146 ext3_fsblk_t sb_block;
2147 unsigned long offset;
2148 struct ext3_super_block * es;
2149 struct block_device *bdev;
2151 bdev = ext3_blkdev_get(j_dev, sb);
2155 if (bd_claim(bdev, sb)) {
2156 ext3_msg(sb, KERN_ERR,
2157 "error: failed to claim external journal device");
2158 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2162 blocksize = sb->s_blocksize;
2163 hblock = bdev_logical_block_size(bdev);
2164 if (blocksize < hblock) {
2165 ext3_msg(sb, KERN_ERR,
2166 "error: blocksize too small for journal device");
2170 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2171 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2172 set_blocksize(bdev, blocksize);
2173 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2174 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2175 "external journal");
2179 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2180 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2181 !(le32_to_cpu(es->s_feature_incompat) &
2182 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2183 ext3_msg(sb, KERN_ERR, "error: external journal has "
2189 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2190 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2195 len = le32_to_cpu(es->s_blocks_count);
2196 start = sb_block + 1;
2197 brelse(bh); /* we're done with the superblock */
2199 journal = journal_init_dev(bdev, sb->s_bdev,
2200 start, len, blocksize);
2202 ext3_msg(sb, KERN_ERR,
2203 "error: failed to create device journal");
2206 journal->j_private = sb;
2207 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2208 wait_on_buffer(journal->j_sb_buffer);
2209 if (!buffer_uptodate(journal->j_sb_buffer)) {
2210 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2213 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2214 ext3_msg(sb, KERN_ERR,
2215 "error: external journal has more than one "
2216 "user (unsupported) - %d",
2217 be32_to_cpu(journal->j_superblock->s_nr_users));
2220 EXT3_SB(sb)->journal_bdev = bdev;
2221 ext3_init_journal_params(sb, journal);
2224 journal_destroy(journal);
2226 ext3_blkdev_put(bdev);
2230 static int ext3_load_journal(struct super_block *sb,
2231 struct ext3_super_block *es,
2232 unsigned long journal_devnum)
2235 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2238 int really_read_only;
2240 if (journal_devnum &&
2241 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2242 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2243 "numbers have changed");
2244 journal_dev = new_decode_dev(journal_devnum);
2246 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2248 really_read_only = bdev_read_only(sb->s_bdev);
2251 * Are we loading a blank journal or performing recovery after a
2252 * crash? For recovery, we need to check in advance whether we
2253 * can get read-write access to the device.
2256 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2257 if (sb->s_flags & MS_RDONLY) {
2258 ext3_msg(sb, KERN_INFO,
2259 "recovery required on readonly filesystem");
2260 if (really_read_only) {
2261 ext3_msg(sb, KERN_ERR, "error: write access "
2262 "unavailable, cannot proceed");
2265 ext3_msg(sb, KERN_INFO,
2266 "write access will be enabled during recovery");
2270 if (journal_inum && journal_dev) {
2271 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2272 "and inode journals");
2277 if (!(journal = ext3_get_journal(sb, journal_inum)))
2280 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2284 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2285 err = journal_update_format(journal);
2287 ext3_msg(sb, KERN_ERR, "error updating journal");
2288 journal_destroy(journal);
2293 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2294 err = journal_wipe(journal, !really_read_only);
2296 err = journal_load(journal);
2299 ext3_msg(sb, KERN_ERR, "error loading journal");
2300 journal_destroy(journal);
2304 EXT3_SB(sb)->s_journal = journal;
2305 ext3_clear_journal_err(sb, es);
2307 if (journal_devnum &&
2308 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2309 es->s_journal_dev = cpu_to_le32(journal_devnum);
2311 /* Make sure we flush the recovery flag to disk. */
2312 ext3_commit_super(sb, es, 1);
2318 static int ext3_create_journal(struct super_block *sb,
2319 struct ext3_super_block *es,
2320 unsigned int journal_inum)
2325 if (sb->s_flags & MS_RDONLY) {
2326 ext3_msg(sb, KERN_ERR,
2327 "error: readonly filesystem when trying to "
2332 journal = ext3_get_journal(sb, journal_inum);
2336 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2339 err = journal_create(journal);
2341 ext3_msg(sb, KERN_ERR, "error creating journal");
2342 journal_destroy(journal);
2346 EXT3_SB(sb)->s_journal = journal;
2348 ext3_update_dynamic_rev(sb);
2349 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2350 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2352 es->s_journal_inum = cpu_to_le32(journal_inum);
2354 /* Make sure we flush the recovery flag to disk. */
2355 ext3_commit_super(sb, es, 1);
2360 static int ext3_commit_super(struct super_block *sb,
2361 struct ext3_super_block *es,
2364 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2370 * If the file system is mounted read-only, don't update the
2371 * superblock write time. This avoids updating the superblock
2372 * write time when we are mounting the root file system
2373 * read/only but we need to replay the journal; at that point,
2374 * for people who are east of GMT and who make their clock
2375 * tick in localtime for Windows bug-for-bug compatibility,
2376 * the clock is set in the future, and this will cause e2fsck
2377 * to complain and force a full file system check.
2379 if (!(sb->s_flags & MS_RDONLY))
2380 es->s_wtime = cpu_to_le32(get_seconds());
2381 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2382 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2383 BUFFER_TRACE(sbh, "marking dirty");
2384 mark_buffer_dirty(sbh);
2386 error = sync_dirty_buffer(sbh);
2392 * Have we just finished recovery? If so, and if we are mounting (or
2393 * remounting) the filesystem readonly, then we will end up with a
2394 * consistent fs on disk. Record that fact.
2396 static void ext3_mark_recovery_complete(struct super_block * sb,
2397 struct ext3_super_block * es)
2399 journal_t *journal = EXT3_SB(sb)->s_journal;
2401 journal_lock_updates(journal);
2402 if (journal_flush(journal) < 0)
2405 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2406 sb->s_flags & MS_RDONLY) {
2407 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2408 ext3_commit_super(sb, es, 1);
2412 journal_unlock_updates(journal);
2416 * If we are mounting (or read-write remounting) a filesystem whose journal
2417 * has recorded an error from a previous lifetime, move that error to the
2418 * main filesystem now.
2420 static void ext3_clear_journal_err(struct super_block *sb,
2421 struct ext3_super_block *es)
2427 journal = EXT3_SB(sb)->s_journal;
2430 * Now check for any error status which may have been recorded in the
2431 * journal by a prior ext3_error() or ext3_abort()
2434 j_errno = journal_errno(journal);
2438 errstr = ext3_decode_error(sb, j_errno, nbuf);
2439 ext3_warning(sb, __func__, "Filesystem error recorded "
2440 "from previous mount: %s", errstr);
2441 ext3_warning(sb, __func__, "Marking fs in need of "
2442 "filesystem check.");
2444 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2445 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2446 ext3_commit_super (sb, es, 1);
2448 journal_clear_err(journal);
2453 * Force the running and committing transactions to commit,
2454 * and wait on the commit.
2456 int ext3_force_commit(struct super_block *sb)
2461 if (sb->s_flags & MS_RDONLY)
2464 journal = EXT3_SB(sb)->s_journal;
2465 ret = ext3_journal_force_commit(journal);
2469 static int ext3_sync_fs(struct super_block *sb, int wait)
2473 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2475 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2481 * LVM calls this function before a (read-only) snapshot is created. This
2482 * gives us a chance to flush the journal completely and mark the fs clean.
2484 static int ext3_freeze(struct super_block *sb)
2489 if (!(sb->s_flags & MS_RDONLY)) {
2490 journal = EXT3_SB(sb)->s_journal;
2492 /* Now we set up the journal barrier. */
2493 journal_lock_updates(journal);
2496 * We don't want to clear needs_recovery flag when we failed
2497 * to flush the journal.
2499 error = journal_flush(journal);
2503 /* Journal blocked and flushed, clear needs_recovery flag. */
2504 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2505 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2512 journal_unlock_updates(journal);
2517 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2518 * flag here, even though the filesystem is not technically dirty yet.
2520 static int ext3_unfreeze(struct super_block *sb)
2522 if (!(sb->s_flags & MS_RDONLY)) {
2524 /* Reser the needs_recovery flag before the fs is unlocked. */
2525 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2526 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2528 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2533 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2535 struct ext3_super_block * es;
2536 struct ext3_sb_info *sbi = EXT3_SB(sb);
2537 ext3_fsblk_t n_blocks_count = 0;
2538 unsigned long old_sb_flags;
2539 struct ext3_mount_options old_opts;
2547 /* Store the original options */
2549 old_sb_flags = sb->s_flags;
2550 old_opts.s_mount_opt = sbi->s_mount_opt;
2551 old_opts.s_resuid = sbi->s_resuid;
2552 old_opts.s_resgid = sbi->s_resgid;
2553 old_opts.s_commit_interval = sbi->s_commit_interval;
2555 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2556 for (i = 0; i < MAXQUOTAS; i++)
2557 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2561 * Allow the "check" option to be passed as a remount option.
2563 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2568 if (test_opt(sb, ABORT))
2569 ext3_abort(sb, __func__, "Abort forced by user");
2571 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2572 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2576 ext3_init_journal_params(sb, sbi->s_journal);
2578 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2579 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2580 if (test_opt(sb, ABORT)) {
2585 if (*flags & MS_RDONLY) {
2587 * First of all, the unconditional stuff we have to do
2588 * to disable replay of the journal when we next remount
2590 sb->s_flags |= MS_RDONLY;
2593 * OK, test if we are remounting a valid rw partition
2594 * readonly, and if so set the rdonly flag and then
2595 * mark the partition as valid again.
2597 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2598 (sbi->s_mount_state & EXT3_VALID_FS))
2599 es->s_state = cpu_to_le16(sbi->s_mount_state);
2601 ext3_mark_recovery_complete(sb, es);
2604 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2605 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2606 ext3_msg(sb, KERN_WARNING,
2607 "warning: couldn't remount RDWR "
2608 "because of unsupported optional "
2609 "features (%x)", le32_to_cpu(ret));
2615 * If we have an unprocessed orphan list hanging
2616 * around from a previously readonly bdev mount,
2617 * require a full umount/remount for now.
2619 if (es->s_last_orphan) {
2620 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2621 "remount RDWR because of unprocessed "
2622 "orphan inode list. Please "
2623 "umount/remount instead.");
2629 * Mounting a RDONLY partition read-write, so reread
2630 * and store the current valid flag. (It may have
2631 * been changed by e2fsck since we originally mounted
2634 ext3_clear_journal_err(sb, es);
2635 sbi->s_mount_state = le16_to_cpu(es->s_state);
2636 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2638 if (!ext3_setup_super (sb, es, 0))
2639 sb->s_flags &= ~MS_RDONLY;
2643 /* Release old quota file names */
2644 for (i = 0; i < MAXQUOTAS; i++)
2645 if (old_opts.s_qf_names[i] &&
2646 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2647 kfree(old_opts.s_qf_names[i]);
2653 sb->s_flags = old_sb_flags;
2654 sbi->s_mount_opt = old_opts.s_mount_opt;
2655 sbi->s_resuid = old_opts.s_resuid;
2656 sbi->s_resgid = old_opts.s_resgid;
2657 sbi->s_commit_interval = old_opts.s_commit_interval;
2659 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2660 for (i = 0; i < MAXQUOTAS; i++) {
2661 if (sbi->s_qf_names[i] &&
2662 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2663 kfree(sbi->s_qf_names[i]);
2664 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2672 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2674 struct super_block *sb = dentry->d_sb;
2675 struct ext3_sb_info *sbi = EXT3_SB(sb);
2676 struct ext3_super_block *es = sbi->s_es;
2679 if (test_opt(sb, MINIX_DF)) {
2680 sbi->s_overhead_last = 0;
2681 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2682 unsigned long ngroups = sbi->s_groups_count, i;
2683 ext3_fsblk_t overhead = 0;
2687 * Compute the overhead (FS structures). This is constant
2688 * for a given filesystem unless the number of block groups
2689 * changes so we cache the previous value until it does.
2693 * All of the blocks before first_data_block are
2696 overhead = le32_to_cpu(es->s_first_data_block);
2699 * Add the overhead attributed to the superblock and
2700 * block group descriptors. If the sparse superblocks
2701 * feature is turned on, then not all groups have this.
2703 for (i = 0; i < ngroups; i++) {
2704 overhead += ext3_bg_has_super(sb, i) +
2705 ext3_bg_num_gdb(sb, i);
2710 * Every block group has an inode bitmap, a block
2711 * bitmap, and an inode table.
2713 overhead += ngroups * (2 + sbi->s_itb_per_group);
2714 sbi->s_overhead_last = overhead;
2716 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2719 buf->f_type = EXT3_SUPER_MAGIC;
2720 buf->f_bsize = sb->s_blocksize;
2721 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2722 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2723 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2724 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2726 buf->f_files = le32_to_cpu(es->s_inodes_count);
2727 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2728 buf->f_namelen = EXT3_NAME_LEN;
2729 fsid = le64_to_cpup((void *)es->s_uuid) ^
2730 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2731 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2732 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2736 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2737 * is locked for write. Otherwise the are possible deadlocks:
2738 * Process 1 Process 2
2739 * ext3_create() quota_sync()
2740 * journal_start() write_dquot()
2741 * vfs_dq_init() down(dqio_mutex)
2742 * down(dqio_mutex) journal_start()
2748 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2750 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2753 static int ext3_write_dquot(struct dquot *dquot)
2757 struct inode *inode;
2759 inode = dquot_to_inode(dquot);
2760 handle = ext3_journal_start(inode,
2761 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2763 return PTR_ERR(handle);
2764 ret = dquot_commit(dquot);
2765 err = ext3_journal_stop(handle);
2771 static int ext3_acquire_dquot(struct dquot *dquot)
2776 handle = ext3_journal_start(dquot_to_inode(dquot),
2777 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2779 return PTR_ERR(handle);
2780 ret = dquot_acquire(dquot);
2781 err = ext3_journal_stop(handle);
2787 static int ext3_release_dquot(struct dquot *dquot)
2792 handle = ext3_journal_start(dquot_to_inode(dquot),
2793 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2794 if (IS_ERR(handle)) {
2795 /* Release dquot anyway to avoid endless cycle in dqput() */
2796 dquot_release(dquot);
2797 return PTR_ERR(handle);
2799 ret = dquot_release(dquot);
2800 err = ext3_journal_stop(handle);
2806 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2808 /* Are we journaling quotas? */
2809 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2810 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2811 dquot_mark_dquot_dirty(dquot);
2812 return ext3_write_dquot(dquot);
2814 return dquot_mark_dquot_dirty(dquot);
2818 static int ext3_write_info(struct super_block *sb, int type)
2823 /* Data block + inode block */
2824 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2826 return PTR_ERR(handle);
2827 ret = dquot_commit_info(sb, type);
2828 err = ext3_journal_stop(handle);
2835 * Turn on quotas during mount time - we need to find
2836 * the quota file and such...
2838 static int ext3_quota_on_mount(struct super_block *sb, int type)
2840 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2841 EXT3_SB(sb)->s_jquota_fmt, type);
2845 * Standard function to be called on quota_on
2847 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2848 char *name, int remount)
2853 if (!test_opt(sb, QUOTA))
2855 /* When remounting, no checks are needed and in fact, name is NULL */
2857 return vfs_quota_on(sb, type, format_id, name, remount);
2859 err = kern_path(name, LOOKUP_FOLLOW, &path);
2863 /* Quotafile not on the same filesystem? */
2864 if (path.mnt->mnt_sb != sb) {
2868 /* Journaling quota? */
2869 if (EXT3_SB(sb)->s_qf_names[type]) {
2870 /* Quotafile not of fs root? */
2871 if (path.dentry->d_parent != sb->s_root)
2872 ext3_msg(sb, KERN_WARNING,
2873 "warning: Quota file not on filesystem root. "
2874 "Journaled quota will not work.");
2878 * When we journal data on quota file, we have to flush journal to see
2879 * all updates to the file when we bypass pagecache...
2881 if (ext3_should_journal_data(path.dentry->d_inode)) {
2883 * We don't need to lock updates but journal_flush() could
2884 * otherwise be livelocked...
2886 journal_lock_updates(EXT3_SB(sb)->s_journal);
2887 err = journal_flush(EXT3_SB(sb)->s_journal);
2888 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2895 err = vfs_quota_on_path(sb, type, format_id, &path);
2900 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2901 * acquiring the locks... As quota files are never truncated and quota code
2902 * itself serializes the operations (and noone else should touch the files)
2903 * we don't have to be afraid of races */
2904 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2905 size_t len, loff_t off)
2907 struct inode *inode = sb_dqopt(sb)->files[type];
2908 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2910 int offset = off & (sb->s_blocksize - 1);
2913 struct buffer_head *bh;
2914 loff_t i_size = i_size_read(inode);
2918 if (off+len > i_size)
2921 while (toread > 0) {
2922 tocopy = sb->s_blocksize - offset < toread ?
2923 sb->s_blocksize - offset : toread;
2924 bh = ext3_bread(NULL, inode, blk, 0, &err);
2927 if (!bh) /* A hole? */
2928 memset(data, 0, tocopy);
2930 memcpy(data, bh->b_data+offset, tocopy);
2940 /* Write to quotafile (we know the transaction is already started and has
2941 * enough credits) */
2942 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2943 const char *data, size_t len, loff_t off)
2945 struct inode *inode = sb_dqopt(sb)->files[type];
2946 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2948 int offset = off & (sb->s_blocksize - 1);
2949 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2950 struct buffer_head *bh;
2951 handle_t *handle = journal_current_handle();
2954 ext3_msg(sb, KERN_WARNING,
2955 "warning: quota write (off=%llu, len=%llu)"
2956 " cancelled because transaction is not started.",
2957 (unsigned long long)off, (unsigned long long)len);
2962 * Since we account only one data block in transaction credits,
2963 * then it is impossible to cross a block boundary.
2965 if (sb->s_blocksize - offset < len) {
2966 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2967 " cancelled because not block aligned",
2968 (unsigned long long)off, (unsigned long long)len);
2971 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2972 bh = ext3_bread(handle, inode, blk, 1, &err);
2975 if (journal_quota) {
2976 err = ext3_journal_get_write_access(handle, bh);
2983 memcpy(bh->b_data+offset, data, len);
2984 flush_dcache_page(bh->b_page);
2987 err = ext3_journal_dirty_metadata(handle, bh);
2989 /* Always do at least ordered writes for quotas */
2990 err = ext3_journal_dirty_data(handle, bh);
2991 mark_buffer_dirty(bh);
2996 mutex_unlock(&inode->i_mutex);
2999 if (inode->i_size < off + len) {
3000 i_size_write(inode, off + len);
3001 EXT3_I(inode)->i_disksize = inode->i_size;
3004 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3005 ext3_mark_inode_dirty(handle, inode);
3006 mutex_unlock(&inode->i_mutex);
3012 static int ext3_get_sb(struct file_system_type *fs_type,
3013 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3015 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
3018 static struct file_system_type ext3_fs_type = {
3019 .owner = THIS_MODULE,
3021 .get_sb = ext3_get_sb,
3022 .kill_sb = kill_block_super,
3023 .fs_flags = FS_REQUIRES_DEV,
3026 static int __init init_ext3_fs(void)
3028 int err = init_ext3_xattr();
3031 err = init_inodecache();
3034 err = register_filesystem(&ext3_fs_type);
3039 destroy_inodecache();
3045 static void __exit exit_ext3_fs(void)
3047 unregister_filesystem(&ext3_fs_type);
3048 destroy_inodecache();
3052 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3053 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3054 MODULE_LICENSE("GPL");
3055 module_init(init_ext3_fs)
3056 module_exit(exit_ext3_fs)