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 complain 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;
533 ext3_discard_reservation(inode);
534 EXT3_I(inode)->i_block_alloc_info = NULL;
539 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
541 #if defined(CONFIG_QUOTA)
542 struct ext3_sb_info *sbi = EXT3_SB(sb);
544 if (sbi->s_jquota_fmt) {
547 switch (sbi->s_jquota_fmt) {
558 seq_printf(seq, ",jqfmt=%s", fmtname);
561 if (sbi->s_qf_names[USRQUOTA])
562 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
564 if (sbi->s_qf_names[GRPQUOTA])
565 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
567 if (test_opt(sb, USRQUOTA))
568 seq_puts(seq, ",usrquota");
570 if (test_opt(sb, GRPQUOTA))
571 seq_puts(seq, ",grpquota");
575 static char *data_mode_string(unsigned long mode)
578 case EXT3_MOUNT_JOURNAL_DATA:
580 case EXT3_MOUNT_ORDERED_DATA:
582 case EXT3_MOUNT_WRITEBACK_DATA:
590 * - it's set to a non-default value OR
591 * - if the per-sb default is different from the global default
593 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
595 struct super_block *sb = vfs->mnt_sb;
596 struct ext3_sb_info *sbi = EXT3_SB(sb);
597 struct ext3_super_block *es = sbi->s_es;
598 unsigned long def_mount_opts;
600 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
602 if (sbi->s_sb_block != 1)
603 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
604 if (test_opt(sb, MINIX_DF))
605 seq_puts(seq, ",minixdf");
606 if (test_opt(sb, GRPID))
607 seq_puts(seq, ",grpid");
608 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
609 seq_puts(seq, ",nogrpid");
610 if (sbi->s_resuid != EXT3_DEF_RESUID ||
611 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
612 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
614 if (sbi->s_resgid != EXT3_DEF_RESGID ||
615 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
616 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
618 if (test_opt(sb, ERRORS_RO)) {
619 int def_errors = le16_to_cpu(es->s_errors);
621 if (def_errors == EXT3_ERRORS_PANIC ||
622 def_errors == EXT3_ERRORS_CONTINUE) {
623 seq_puts(seq, ",errors=remount-ro");
626 if (test_opt(sb, ERRORS_CONT))
627 seq_puts(seq, ",errors=continue");
628 if (test_opt(sb, ERRORS_PANIC))
629 seq_puts(seq, ",errors=panic");
630 if (test_opt(sb, NO_UID32))
631 seq_puts(seq, ",nouid32");
632 if (test_opt(sb, DEBUG))
633 seq_puts(seq, ",debug");
634 if (test_opt(sb, OLDALLOC))
635 seq_puts(seq, ",oldalloc");
636 #ifdef CONFIG_EXT3_FS_XATTR
637 if (test_opt(sb, XATTR_USER))
638 seq_puts(seq, ",user_xattr");
639 if (!test_opt(sb, XATTR_USER) &&
640 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
641 seq_puts(seq, ",nouser_xattr");
644 #ifdef CONFIG_EXT3_FS_POSIX_ACL
645 if (test_opt(sb, POSIX_ACL))
646 seq_puts(seq, ",acl");
647 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
648 seq_puts(seq, ",noacl");
650 if (!test_opt(sb, RESERVATION))
651 seq_puts(seq, ",noreservation");
652 if (sbi->s_commit_interval) {
653 seq_printf(seq, ",commit=%u",
654 (unsigned) (sbi->s_commit_interval / HZ));
658 * Always display barrier state so it's clear what the status is.
660 seq_puts(seq, ",barrier=");
661 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
662 if (test_opt(sb, NOBH))
663 seq_puts(seq, ",nobh");
665 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
666 if (test_opt(sb, DATA_ERR_ABORT))
667 seq_puts(seq, ",data_err=abort");
669 if (test_opt(sb, NOLOAD))
670 seq_puts(seq, ",norecovery");
672 ext3_show_quota_options(seq, sb);
678 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
679 u64 ino, u32 generation)
683 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
684 return ERR_PTR(-ESTALE);
685 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
686 return ERR_PTR(-ESTALE);
688 /* iget isn't really right if the inode is currently unallocated!!
690 * ext3_read_inode will return a bad_inode if the inode had been
691 * deleted, so we should be safe.
693 * Currently we don't know the generation for parent directory, so
694 * a generation of 0 means "accept any"
696 inode = ext3_iget(sb, ino);
698 return ERR_CAST(inode);
699 if (generation && inode->i_generation != generation) {
701 return ERR_PTR(-ESTALE);
707 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
708 int fh_len, int fh_type)
710 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
714 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
715 int fh_len, int fh_type)
717 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
722 * Try to release metadata pages (indirect blocks, directories) which are
723 * mapped via the block device. Since these pages could have journal heads
724 * which would prevent try_to_free_buffers() from freeing them, we must use
725 * jbd layer's try_to_free_buffers() function to release them.
727 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
730 journal_t *journal = EXT3_SB(sb)->s_journal;
732 WARN_ON(PageChecked(page));
733 if (!page_has_buffers(page))
736 return journal_try_to_free_buffers(journal, page,
738 return try_to_free_buffers(page);
742 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
743 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
745 static int ext3_write_dquot(struct dquot *dquot);
746 static int ext3_acquire_dquot(struct dquot *dquot);
747 static int ext3_release_dquot(struct dquot *dquot);
748 static int ext3_mark_dquot_dirty(struct dquot *dquot);
749 static int ext3_write_info(struct super_block *sb, int type);
750 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
751 char *path, int remount);
752 static int ext3_quota_on_mount(struct super_block *sb, int type);
753 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
754 size_t len, loff_t off);
755 static ssize_t ext3_quota_write(struct super_block *sb, int type,
756 const char *data, size_t len, loff_t off);
758 static const struct dquot_operations ext3_quota_operations = {
759 .write_dquot = ext3_write_dquot,
760 .acquire_dquot = ext3_acquire_dquot,
761 .release_dquot = ext3_release_dquot,
762 .mark_dirty = ext3_mark_dquot_dirty,
763 .write_info = ext3_write_info,
764 .alloc_dquot = dquot_alloc,
765 .destroy_dquot = dquot_destroy,
768 static const struct quotactl_ops ext3_qctl_operations = {
769 .quota_on = ext3_quota_on,
770 .quota_off = vfs_quota_off,
771 .quota_sync = vfs_quota_sync,
772 .get_info = vfs_get_dqinfo,
773 .set_info = vfs_set_dqinfo,
774 .get_dqblk = vfs_get_dqblk,
775 .set_dqblk = vfs_set_dqblk
779 static const struct super_operations ext3_sops = {
780 .alloc_inode = ext3_alloc_inode,
781 .destroy_inode = ext3_destroy_inode,
782 .write_inode = ext3_write_inode,
783 .dirty_inode = ext3_dirty_inode,
784 .delete_inode = ext3_delete_inode,
785 .put_super = ext3_put_super,
786 .sync_fs = ext3_sync_fs,
787 .freeze_fs = ext3_freeze,
788 .unfreeze_fs = ext3_unfreeze,
789 .statfs = ext3_statfs,
790 .remount_fs = ext3_remount,
791 .clear_inode = ext3_clear_inode,
792 .show_options = ext3_show_options,
794 .quota_read = ext3_quota_read,
795 .quota_write = ext3_quota_write,
797 .bdev_try_to_free_page = bdev_try_to_free_page,
800 static const struct export_operations ext3_export_ops = {
801 .fh_to_dentry = ext3_fh_to_dentry,
802 .fh_to_parent = ext3_fh_to_parent,
803 .get_parent = ext3_get_parent,
807 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
808 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
809 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
810 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
811 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
812 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
813 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
814 Opt_data_err_abort, Opt_data_err_ignore,
815 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
816 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
817 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
818 Opt_resize, Opt_usrquota, Opt_grpquota
821 static const match_table_t tokens = {
822 {Opt_bsd_df, "bsddf"},
823 {Opt_minix_df, "minixdf"},
824 {Opt_grpid, "grpid"},
825 {Opt_grpid, "bsdgroups"},
826 {Opt_nogrpid, "nogrpid"},
827 {Opt_nogrpid, "sysvgroups"},
828 {Opt_resgid, "resgid=%u"},
829 {Opt_resuid, "resuid=%u"},
831 {Opt_err_cont, "errors=continue"},
832 {Opt_err_panic, "errors=panic"},
833 {Opt_err_ro, "errors=remount-ro"},
834 {Opt_nouid32, "nouid32"},
835 {Opt_nocheck, "nocheck"},
836 {Opt_nocheck, "check=none"},
837 {Opt_debug, "debug"},
838 {Opt_oldalloc, "oldalloc"},
839 {Opt_orlov, "orlov"},
840 {Opt_user_xattr, "user_xattr"},
841 {Opt_nouser_xattr, "nouser_xattr"},
843 {Opt_noacl, "noacl"},
844 {Opt_reservation, "reservation"},
845 {Opt_noreservation, "noreservation"},
846 {Opt_noload, "noload"},
847 {Opt_noload, "norecovery"},
850 {Opt_commit, "commit=%u"},
851 {Opt_journal_update, "journal=update"},
852 {Opt_journal_inum, "journal=%u"},
853 {Opt_journal_dev, "journal_dev=%u"},
854 {Opt_abort, "abort"},
855 {Opt_data_journal, "data=journal"},
856 {Opt_data_ordered, "data=ordered"},
857 {Opt_data_writeback, "data=writeback"},
858 {Opt_data_err_abort, "data_err=abort"},
859 {Opt_data_err_ignore, "data_err=ignore"},
860 {Opt_offusrjquota, "usrjquota="},
861 {Opt_usrjquota, "usrjquota=%s"},
862 {Opt_offgrpjquota, "grpjquota="},
863 {Opt_grpjquota, "grpjquota=%s"},
864 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
865 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
866 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
867 {Opt_grpquota, "grpquota"},
868 {Opt_noquota, "noquota"},
869 {Opt_quota, "quota"},
870 {Opt_usrquota, "usrquota"},
871 {Opt_barrier, "barrier=%u"},
872 {Opt_barrier, "barrier"},
873 {Opt_nobarrier, "nobarrier"},
874 {Opt_resize, "resize"},
878 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
880 ext3_fsblk_t sb_block;
881 char *options = (char *) *data;
883 if (!options || strncmp(options, "sb=", 3) != 0)
884 return 1; /* Default location */
886 /*todo: use simple_strtoll with >32bit ext3 */
887 sb_block = simple_strtoul(options, &options, 0);
888 if (*options && *options != ',') {
889 ext3_msg(sb, "error: invalid sb specification: %s",
895 *data = (void *) options;
900 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
902 struct ext3_sb_info *sbi = EXT3_SB(sb);
905 if (sb_any_quota_loaded(sb) &&
906 !sbi->s_qf_names[qtype]) {
907 ext3_msg(sb, KERN_ERR,
908 "Cannot change journaled "
909 "quota options when quota turned on");
912 qname = match_strdup(args);
914 ext3_msg(sb, KERN_ERR,
915 "Not enough memory for storing quotafile name");
918 if (sbi->s_qf_names[qtype] &&
919 strcmp(sbi->s_qf_names[qtype], qname)) {
920 ext3_msg(sb, KERN_ERR,
921 "%s quota file already specified", QTYPE2NAME(qtype));
925 sbi->s_qf_names[qtype] = qname;
926 if (strchr(sbi->s_qf_names[qtype], '/')) {
927 ext3_msg(sb, KERN_ERR,
928 "quotafile must be on filesystem root");
929 kfree(sbi->s_qf_names[qtype]);
930 sbi->s_qf_names[qtype] = NULL;
933 set_opt(sbi->s_mount_opt, QUOTA);
937 static int clear_qf_name(struct super_block *sb, int qtype) {
939 struct ext3_sb_info *sbi = EXT3_SB(sb);
941 if (sb_any_quota_loaded(sb) &&
942 sbi->s_qf_names[qtype]) {
943 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
944 " when quota turned on");
948 * The space will be released later when all options are confirmed
951 sbi->s_qf_names[qtype] = NULL;
956 static int parse_options (char *options, struct super_block *sb,
957 unsigned int *inum, unsigned long *journal_devnum,
958 ext3_fsblk_t *n_blocks_count, int is_remount)
960 struct ext3_sb_info *sbi = EXT3_SB(sb);
962 substring_t args[MAX_OPT_ARGS];
972 while ((p = strsep (&options, ",")) != NULL) {
977 * Initialize args struct so we know whether arg was
978 * found; some options take optional arguments.
980 args[0].to = args[0].from = 0;
981 token = match_token(p, tokens, args);
984 clear_opt (sbi->s_mount_opt, MINIX_DF);
987 set_opt (sbi->s_mount_opt, MINIX_DF);
990 set_opt (sbi->s_mount_opt, GRPID);
993 clear_opt (sbi->s_mount_opt, GRPID);
996 if (match_int(&args[0], &option))
998 sbi->s_resuid = option;
1001 if (match_int(&args[0], &option))
1003 sbi->s_resgid = option;
1006 /* handled by get_sb_block() instead of here */
1007 /* *sb_block = match_int(&args[0]); */
1010 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1011 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1012 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1015 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1016 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1017 set_opt (sbi->s_mount_opt, ERRORS_RO);
1020 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1021 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1022 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1025 set_opt (sbi->s_mount_opt, NO_UID32);
1028 clear_opt (sbi->s_mount_opt, CHECK);
1031 set_opt (sbi->s_mount_opt, DEBUG);
1034 set_opt (sbi->s_mount_opt, OLDALLOC);
1037 clear_opt (sbi->s_mount_opt, OLDALLOC);
1039 #ifdef CONFIG_EXT3_FS_XATTR
1040 case Opt_user_xattr:
1041 set_opt (sbi->s_mount_opt, XATTR_USER);
1043 case Opt_nouser_xattr:
1044 clear_opt (sbi->s_mount_opt, XATTR_USER);
1047 case Opt_user_xattr:
1048 case Opt_nouser_xattr:
1049 ext3_msg(sb, KERN_INFO,
1050 "(no)user_xattr options not supported");
1053 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1055 set_opt(sbi->s_mount_opt, POSIX_ACL);
1058 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1063 ext3_msg(sb, KERN_INFO,
1064 "(no)acl options not supported");
1067 case Opt_reservation:
1068 set_opt(sbi->s_mount_opt, RESERVATION);
1070 case Opt_noreservation:
1071 clear_opt(sbi->s_mount_opt, RESERVATION);
1073 case Opt_journal_update:
1075 /* Eventually we will want to be able to create
1076 a journal file here. For now, only allow the
1077 user to specify an existing inode to be the
1080 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1081 "journal on remount");
1084 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1086 case Opt_journal_inum:
1088 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1089 "journal on remount");
1092 if (match_int(&args[0], &option))
1096 case Opt_journal_dev:
1098 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1099 "journal on remount");
1102 if (match_int(&args[0], &option))
1104 *journal_devnum = option;
1107 set_opt (sbi->s_mount_opt, NOLOAD);
1110 if (match_int(&args[0], &option))
1115 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1116 sbi->s_commit_interval = HZ * option;
1118 case Opt_data_journal:
1119 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1121 case Opt_data_ordered:
1122 data_opt = EXT3_MOUNT_ORDERED_DATA;
1124 case Opt_data_writeback:
1125 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1128 if (test_opt(sb, DATA_FLAGS) == data_opt)
1130 ext3_msg(sb, KERN_ERR,
1131 "error: cannot change "
1132 "data mode on remount. The filesystem "
1133 "is mounted in data=%s mode and you "
1134 "try to remount it in data=%s mode.",
1135 data_mode_string(test_opt(sb,
1137 data_mode_string(data_opt));
1140 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1141 sbi->s_mount_opt |= data_opt;
1144 case Opt_data_err_abort:
1145 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1147 case Opt_data_err_ignore:
1148 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1152 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1156 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1159 case Opt_offusrjquota:
1160 if (!clear_qf_name(sb, USRQUOTA))
1163 case Opt_offgrpjquota:
1164 if (!clear_qf_name(sb, GRPQUOTA))
1167 case Opt_jqfmt_vfsold:
1168 qfmt = QFMT_VFS_OLD;
1170 case Opt_jqfmt_vfsv0:
1173 case Opt_jqfmt_vfsv1:
1176 if (sb_any_quota_loaded(sb) &&
1177 sbi->s_jquota_fmt != qfmt) {
1178 ext3_msg(sb, KERN_ERR, "error: cannot change "
1179 "journaled quota options when "
1180 "quota turned on.");
1183 sbi->s_jquota_fmt = qfmt;
1187 set_opt(sbi->s_mount_opt, QUOTA);
1188 set_opt(sbi->s_mount_opt, USRQUOTA);
1191 set_opt(sbi->s_mount_opt, QUOTA);
1192 set_opt(sbi->s_mount_opt, GRPQUOTA);
1195 if (sb_any_quota_loaded(sb)) {
1196 ext3_msg(sb, KERN_ERR, "error: cannot change "
1197 "quota options when quota turned on.");
1200 clear_opt(sbi->s_mount_opt, QUOTA);
1201 clear_opt(sbi->s_mount_opt, USRQUOTA);
1202 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1208 ext3_msg(sb, KERN_ERR,
1209 "error: quota options not supported.");
1213 case Opt_offusrjquota:
1214 case Opt_offgrpjquota:
1215 case Opt_jqfmt_vfsold:
1216 case Opt_jqfmt_vfsv0:
1217 case Opt_jqfmt_vfsv1:
1218 ext3_msg(sb, KERN_ERR,
1219 "error: journaled quota options not "
1226 set_opt(sbi->s_mount_opt, ABORT);
1229 clear_opt(sbi->s_mount_opt, BARRIER);
1233 if (match_int(&args[0], &option))
1236 option = 1; /* No argument, default to 1 */
1238 set_opt(sbi->s_mount_opt, BARRIER);
1240 clear_opt(sbi->s_mount_opt, BARRIER);
1246 ext3_msg(sb, KERN_ERR,
1247 "error: resize option only available "
1251 if (match_int(&args[0], &option) != 0)
1253 *n_blocks_count = option;
1256 set_opt(sbi->s_mount_opt, NOBH);
1259 clear_opt(sbi->s_mount_opt, NOBH);
1262 ext3_msg(sb, KERN_ERR,
1263 "error: unrecognized mount option \"%s\" "
1264 "or missing value", p);
1269 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1270 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1271 clear_opt(sbi->s_mount_opt, USRQUOTA);
1272 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1273 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1275 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1276 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1281 if (!sbi->s_jquota_fmt) {
1282 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1287 if (sbi->s_jquota_fmt) {
1288 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1289 "specified with no journaling "
1298 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1301 struct ext3_sb_info *sbi = EXT3_SB(sb);
1304 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1305 ext3_msg(sb, KERN_ERR,
1306 "error: revision level too high, "
1307 "forcing read-only mode");
1312 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1313 ext3_msg(sb, KERN_WARNING,
1314 "warning: mounting unchecked fs, "
1315 "running e2fsck is recommended");
1316 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1317 ext3_msg(sb, KERN_WARNING,
1318 "warning: mounting fs with errors, "
1319 "running e2fsck is recommended");
1320 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1321 le16_to_cpu(es->s_mnt_count) >=
1322 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1323 ext3_msg(sb, KERN_WARNING,
1324 "warning: maximal mount count reached, "
1325 "running e2fsck is recommended");
1326 else if (le32_to_cpu(es->s_checkinterval) &&
1327 (le32_to_cpu(es->s_lastcheck) +
1328 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1329 ext3_msg(sb, KERN_WARNING,
1330 "warning: checktime reached, "
1331 "running e2fsck is recommended");
1333 /* @@@ We _will_ want to clear the valid bit if we find
1334 inconsistencies, to force a fsck at reboot. But for
1335 a plain journaled filesystem we can keep it set as
1336 valid forever! :) */
1337 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1339 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1340 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1341 le16_add_cpu(&es->s_mnt_count, 1);
1342 es->s_mtime = cpu_to_le32(get_seconds());
1343 ext3_update_dynamic_rev(sb);
1344 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1346 ext3_commit_super(sb, es, 1);
1347 if (test_opt(sb, DEBUG))
1348 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1349 "bpg=%lu, ipg=%lu, mo=%04lx]",
1351 sbi->s_groups_count,
1352 EXT3_BLOCKS_PER_GROUP(sb),
1353 EXT3_INODES_PER_GROUP(sb),
1356 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1357 char b[BDEVNAME_SIZE];
1358 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1359 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1361 ext3_msg(sb, KERN_INFO, "using internal journal");
1366 /* Called at mount-time, super-block is locked */
1367 static int ext3_check_descriptors(struct super_block *sb)
1369 struct ext3_sb_info *sbi = EXT3_SB(sb);
1372 ext3_debug ("Checking group descriptors");
1374 for (i = 0; i < sbi->s_groups_count; i++) {
1375 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1376 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1377 ext3_fsblk_t last_block;
1379 if (i == sbi->s_groups_count - 1)
1380 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1382 last_block = first_block +
1383 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1385 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1386 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1388 ext3_error (sb, "ext3_check_descriptors",
1389 "Block bitmap for group %d"
1390 " not in group (block %lu)!",
1392 le32_to_cpu(gdp->bg_block_bitmap));
1395 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1396 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1398 ext3_error (sb, "ext3_check_descriptors",
1399 "Inode bitmap for group %d"
1400 " not in group (block %lu)!",
1402 le32_to_cpu(gdp->bg_inode_bitmap));
1405 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1406 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1409 ext3_error (sb, "ext3_check_descriptors",
1410 "Inode table for group %d"
1411 " not in group (block %lu)!",
1413 le32_to_cpu(gdp->bg_inode_table));
1418 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1419 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1424 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1425 * the superblock) which were deleted from all directories, but held open by
1426 * a process at the time of a crash. We walk the list and try to delete these
1427 * inodes at recovery time (only with a read-write filesystem).
1429 * In order to keep the orphan inode chain consistent during traversal (in
1430 * case of crash during recovery), we link each inode into the superblock
1431 * orphan list_head and handle it the same way as an inode deletion during
1432 * normal operation (which journals the operations for us).
1434 * We only do an iget() and an iput() on each inode, which is very safe if we
1435 * accidentally point at an in-use or already deleted inode. The worst that
1436 * can happen in this case is that we get a "bit already cleared" message from
1437 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1438 * e2fsck was run on this filesystem, and it must have already done the orphan
1439 * inode cleanup for us, so we can safely abort without any further action.
1441 static void ext3_orphan_cleanup (struct super_block * sb,
1442 struct ext3_super_block * es)
1444 unsigned int s_flags = sb->s_flags;
1445 int nr_orphans = 0, nr_truncates = 0;
1449 if (!es->s_last_orphan) {
1450 jbd_debug(4, "no orphan inodes to clean up\n");
1454 if (bdev_read_only(sb->s_bdev)) {
1455 ext3_msg(sb, KERN_ERR, "error: write access "
1456 "unavailable, skipping orphan cleanup.");
1460 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1461 if (es->s_last_orphan)
1462 jbd_debug(1, "Errors on filesystem, "
1463 "clearing orphan list.\n");
1464 es->s_last_orphan = 0;
1465 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1469 if (s_flags & MS_RDONLY) {
1470 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1471 sb->s_flags &= ~MS_RDONLY;
1474 /* Needed for iput() to work correctly and not trash data */
1475 sb->s_flags |= MS_ACTIVE;
1476 /* Turn on quotas so that they are updated correctly */
1477 for (i = 0; i < MAXQUOTAS; i++) {
1478 if (EXT3_SB(sb)->s_qf_names[i]) {
1479 int ret = ext3_quota_on_mount(sb, i);
1481 ext3_msg(sb, KERN_ERR,
1482 "error: cannot turn on journaled "
1488 while (es->s_last_orphan) {
1489 struct inode *inode;
1491 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1492 if (IS_ERR(inode)) {
1493 es->s_last_orphan = 0;
1497 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1498 dquot_initialize(inode);
1499 if (inode->i_nlink) {
1501 "%s: truncating inode %lu to %Ld bytes\n",
1502 __func__, inode->i_ino, inode->i_size);
1503 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1504 inode->i_ino, inode->i_size);
1505 ext3_truncate(inode);
1509 "%s: deleting unreferenced inode %lu\n",
1510 __func__, inode->i_ino);
1511 jbd_debug(2, "deleting unreferenced inode %lu\n",
1515 iput(inode); /* The delete magic happens here! */
1518 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1521 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1522 PLURAL(nr_orphans));
1524 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1525 PLURAL(nr_truncates));
1527 /* Turn quotas off */
1528 for (i = 0; i < MAXQUOTAS; i++) {
1529 if (sb_dqopt(sb)->files[i])
1530 vfs_quota_off(sb, i, 0);
1533 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1537 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1538 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1539 * We need to be 1 filesystem block less than the 2^32 sector limit.
1541 static loff_t ext3_max_size(int bits)
1543 loff_t res = EXT3_NDIR_BLOCKS;
1547 /* This is calculated to be the largest file size for a
1548 * dense, file such that the total number of
1549 * sectors in the file, including data and all indirect blocks,
1550 * does not exceed 2^32 -1
1551 * __u32 i_blocks representing the total number of
1552 * 512 bytes blocks of the file
1554 upper_limit = (1LL << 32) - 1;
1556 /* total blocks in file system block size */
1557 upper_limit >>= (bits - 9);
1560 /* indirect blocks */
1562 /* double indirect blocks */
1563 meta_blocks += 1 + (1LL << (bits-2));
1564 /* tripple indirect blocks */
1565 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1567 upper_limit -= meta_blocks;
1568 upper_limit <<= bits;
1570 res += 1LL << (bits-2);
1571 res += 1LL << (2*(bits-2));
1572 res += 1LL << (3*(bits-2));
1574 if (res > upper_limit)
1577 if (res > MAX_LFS_FILESIZE)
1578 res = MAX_LFS_FILESIZE;
1583 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1584 ext3_fsblk_t logic_sb_block,
1587 struct ext3_sb_info *sbi = EXT3_SB(sb);
1588 unsigned long bg, first_meta_bg;
1591 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1593 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1595 return (logic_sb_block + nr + 1);
1596 bg = sbi->s_desc_per_block * nr;
1597 if (ext3_bg_has_super(sb, bg))
1599 return (has_super + ext3_group_first_block_no(sb, bg));
1603 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1605 struct buffer_head * bh;
1606 struct ext3_super_block *es = NULL;
1607 struct ext3_sb_info *sbi;
1609 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1610 ext3_fsblk_t logic_sb_block;
1611 unsigned long offset = 0;
1612 unsigned int journal_inum = 0;
1613 unsigned long journal_devnum = 0;
1614 unsigned long def_mount_opts;
1625 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1629 sbi->s_blockgroup_lock =
1630 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1631 if (!sbi->s_blockgroup_lock) {
1635 sb->s_fs_info = sbi;
1636 sbi->s_mount_opt = 0;
1637 sbi->s_resuid = EXT3_DEF_RESUID;
1638 sbi->s_resgid = EXT3_DEF_RESGID;
1639 sbi->s_sb_block = sb_block;
1643 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1645 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1650 * The ext3 superblock will not be buffer aligned for other than 1kB
1651 * block sizes. We need to calculate the offset from buffer start.
1653 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1654 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1655 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1657 logic_sb_block = sb_block;
1660 if (!(bh = sb_bread(sb, logic_sb_block))) {
1661 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1665 * Note: s_es must be initialized as soon as possible because
1666 * some ext3 macro-instructions depend on its value
1668 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1670 sb->s_magic = le16_to_cpu(es->s_magic);
1671 if (sb->s_magic != EXT3_SUPER_MAGIC)
1674 /* Set defaults before we parse the mount options */
1675 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1676 if (def_mount_opts & EXT3_DEFM_DEBUG)
1677 set_opt(sbi->s_mount_opt, DEBUG);
1678 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1679 set_opt(sbi->s_mount_opt, GRPID);
1680 if (def_mount_opts & EXT3_DEFM_UID16)
1681 set_opt(sbi->s_mount_opt, NO_UID32);
1682 #ifdef CONFIG_EXT3_FS_XATTR
1683 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1684 set_opt(sbi->s_mount_opt, XATTR_USER);
1686 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1687 if (def_mount_opts & EXT3_DEFM_ACL)
1688 set_opt(sbi->s_mount_opt, POSIX_ACL);
1690 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1691 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1692 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1693 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1694 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1695 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1697 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1698 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1699 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1700 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1702 set_opt(sbi->s_mount_opt, ERRORS_RO);
1704 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1705 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1707 set_opt(sbi->s_mount_opt, RESERVATION);
1709 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1713 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1714 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1716 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1717 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1718 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1719 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1720 ext3_msg(sb, KERN_WARNING,
1721 "warning: feature flags set on rev 0 fs, "
1722 "running e2fsck is recommended");
1724 * Check feature flags regardless of the revision level, since we
1725 * previously didn't change the revision level when setting the flags,
1726 * so there is a chance incompat flags are set on a rev 0 filesystem.
1728 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1730 ext3_msg(sb, KERN_ERR,
1731 "error: couldn't mount because of unsupported "
1732 "optional features (%x)", le32_to_cpu(features));
1735 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1736 if (!(sb->s_flags & MS_RDONLY) && features) {
1737 ext3_msg(sb, KERN_ERR,
1738 "error: couldn't mount RDWR because of unsupported "
1739 "optional features (%x)", le32_to_cpu(features));
1742 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1744 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1745 blocksize > EXT3_MAX_BLOCK_SIZE) {
1746 ext3_msg(sb, KERN_ERR,
1747 "error: couldn't mount because of unsupported "
1748 "filesystem blocksize %d", blocksize);
1752 hblock = bdev_logical_block_size(sb->s_bdev);
1753 if (sb->s_blocksize != blocksize) {
1755 * Make sure the blocksize for the filesystem is larger
1756 * than the hardware sectorsize for the machine.
1758 if (blocksize < hblock) {
1759 ext3_msg(sb, KERN_ERR,
1760 "error: fsblocksize %d too small for "
1761 "hardware sectorsize %d", blocksize, hblock);
1766 if (!sb_set_blocksize(sb, blocksize)) {
1767 ext3_msg(sb, KERN_ERR,
1768 "error: bad blocksize %d", blocksize);
1771 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1772 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1773 bh = sb_bread(sb, logic_sb_block);
1775 ext3_msg(sb, KERN_ERR,
1776 "error: can't read superblock on 2nd try");
1779 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1781 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1782 ext3_msg(sb, KERN_ERR,
1783 "error: magic mismatch");
1788 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1790 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1791 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1792 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1794 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1795 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1796 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1797 (!is_power_of_2(sbi->s_inode_size)) ||
1798 (sbi->s_inode_size > blocksize)) {
1799 ext3_msg(sb, KERN_ERR,
1800 "error: unsupported inode size: %d",
1805 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1806 le32_to_cpu(es->s_log_frag_size);
1807 if (blocksize != sbi->s_frag_size) {
1808 ext3_msg(sb, KERN_ERR,
1809 "error: fragsize %lu != blocksize %u (unsupported)",
1810 sbi->s_frag_size, blocksize);
1813 sbi->s_frags_per_block = 1;
1814 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1815 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1816 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1817 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1819 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1820 if (sbi->s_inodes_per_block == 0)
1822 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1823 sbi->s_inodes_per_block;
1824 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1826 sbi->s_mount_state = le16_to_cpu(es->s_state);
1827 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1828 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1829 for (i=0; i < 4; i++)
1830 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1831 sbi->s_def_hash_version = es->s_def_hash_version;
1832 i = le32_to_cpu(es->s_flags);
1833 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1834 sbi->s_hash_unsigned = 3;
1835 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1836 #ifdef __CHAR_UNSIGNED__
1837 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1838 sbi->s_hash_unsigned = 3;
1840 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1844 if (sbi->s_blocks_per_group > blocksize * 8) {
1845 ext3_msg(sb, KERN_ERR,
1846 "#blocks per group too big: %lu",
1847 sbi->s_blocks_per_group);
1850 if (sbi->s_frags_per_group > blocksize * 8) {
1851 ext3_msg(sb, KERN_ERR,
1852 "error: #fragments per group too big: %lu",
1853 sbi->s_frags_per_group);
1856 if (sbi->s_inodes_per_group > blocksize * 8) {
1857 ext3_msg(sb, KERN_ERR,
1858 "error: #inodes per group too big: %lu",
1859 sbi->s_inodes_per_group);
1863 if (le32_to_cpu(es->s_blocks_count) >
1864 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1865 ext3_msg(sb, KERN_ERR,
1866 "error: filesystem is too large to mount safely");
1867 if (sizeof(sector_t) < 8)
1868 ext3_msg(sb, KERN_ERR,
1869 "error: CONFIG_LBDAF not enabled");
1873 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1875 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1876 le32_to_cpu(es->s_first_data_block) - 1)
1877 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1878 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1879 EXT3_DESC_PER_BLOCK(sb);
1880 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1882 if (sbi->s_group_desc == NULL) {
1883 ext3_msg(sb, KERN_ERR,
1884 "error: not enough memory");
1888 bgl_lock_init(sbi->s_blockgroup_lock);
1890 for (i = 0; i < db_count; i++) {
1891 block = descriptor_loc(sb, logic_sb_block, i);
1892 sbi->s_group_desc[i] = sb_bread(sb, block);
1893 if (!sbi->s_group_desc[i]) {
1894 ext3_msg(sb, KERN_ERR,
1895 "error: can't read group descriptor %d", i);
1900 if (!ext3_check_descriptors (sb)) {
1901 ext3_msg(sb, KERN_ERR,
1902 "error: group descriptors corrupted");
1905 sbi->s_gdb_count = db_count;
1906 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1907 spin_lock_init(&sbi->s_next_gen_lock);
1909 /* per fileystem reservation list head & lock */
1910 spin_lock_init(&sbi->s_rsv_window_lock);
1911 sbi->s_rsv_window_root = RB_ROOT;
1912 /* Add a single, static dummy reservation to the start of the
1913 * reservation window list --- it gives us a placeholder for
1914 * append-at-start-of-list which makes the allocation logic
1915 * _much_ simpler. */
1916 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1917 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1918 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1919 sbi->s_rsv_window_head.rsv_goal_size = 0;
1920 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1923 * set up enough so that it can read an inode
1925 sb->s_op = &ext3_sops;
1926 sb->s_export_op = &ext3_export_ops;
1927 sb->s_xattr = ext3_xattr_handlers;
1929 sb->s_qcop = &ext3_qctl_operations;
1930 sb->dq_op = &ext3_quota_operations;
1932 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1933 mutex_init(&sbi->s_orphan_lock);
1934 mutex_init(&sbi->s_resize_lock);
1938 needs_recovery = (es->s_last_orphan != 0 ||
1939 EXT3_HAS_INCOMPAT_FEATURE(sb,
1940 EXT3_FEATURE_INCOMPAT_RECOVER));
1943 * The first inode we look at is the journal inode. Don't try
1944 * root first: it may be modified in the journal!
1946 if (!test_opt(sb, NOLOAD) &&
1947 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1948 if (ext3_load_journal(sb, es, journal_devnum))
1950 } else if (journal_inum) {
1951 if (ext3_create_journal(sb, es, journal_inum))
1955 ext3_msg(sb, KERN_ERR,
1956 "error: no journal found. "
1957 "mounting ext3 over ext2?");
1960 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1961 ext3_count_free_blocks(sb));
1963 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1964 ext3_count_free_inodes(sb));
1967 err = percpu_counter_init(&sbi->s_dirs_counter,
1968 ext3_count_dirs(sb));
1971 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
1975 /* We have now updated the journal if required, so we can
1976 * validate the data journaling mode. */
1977 switch (test_opt(sb, DATA_FLAGS)) {
1979 /* No mode set, assume a default based on the journal
1980 capabilities: ORDERED_DATA if the journal can
1981 cope, else JOURNAL_DATA */
1982 if (journal_check_available_features
1983 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1984 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1986 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1989 case EXT3_MOUNT_ORDERED_DATA:
1990 case EXT3_MOUNT_WRITEBACK_DATA:
1991 if (!journal_check_available_features
1992 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1993 ext3_msg(sb, KERN_ERR,
1994 "error: journal does not support "
1995 "requested data journaling mode");
2002 if (test_opt(sb, NOBH)) {
2003 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
2004 ext3_msg(sb, KERN_WARNING,
2005 "warning: ignoring nobh option - "
2006 "it is supported only with writeback mode");
2007 clear_opt(sbi->s_mount_opt, NOBH);
2011 * The journal_load will have done any necessary log recovery,
2012 * so we can safely mount the rest of the filesystem now.
2015 root = ext3_iget(sb, EXT3_ROOT_INO);
2017 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
2018 ret = PTR_ERR(root);
2021 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2023 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
2026 sb->s_root = d_alloc_root(root);
2028 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2034 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2036 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2037 ext3_orphan_cleanup(sb, es);
2038 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2040 ext3_msg(sb, KERN_INFO, "recovery complete");
2041 ext3_mark_recovery_complete(sb, es);
2042 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2043 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2044 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2052 ext3_msg(sb, KERN_INFO,
2053 "error: can't find ext3 filesystem on dev %s.",
2058 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2059 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2060 percpu_counter_destroy(&sbi->s_dirs_counter);
2061 journal_destroy(sbi->s_journal);
2063 for (i = 0; i < db_count; i++)
2064 brelse(sbi->s_group_desc[i]);
2065 kfree(sbi->s_group_desc);
2068 for (i = 0; i < MAXQUOTAS; i++)
2069 kfree(sbi->s_qf_names[i]);
2071 ext3_blkdev_remove(sbi);
2074 sb->s_fs_info = NULL;
2075 kfree(sbi->s_blockgroup_lock);
2082 * Setup any per-fs journal parameters now. We'll do this both on
2083 * initial mount, once the journal has been initialised but before we've
2084 * done any recovery; and again on any subsequent remount.
2086 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2088 struct ext3_sb_info *sbi = EXT3_SB(sb);
2090 if (sbi->s_commit_interval)
2091 journal->j_commit_interval = sbi->s_commit_interval;
2092 /* We could also set up an ext3-specific default for the commit
2093 * interval here, but for now we'll just fall back to the jbd
2096 spin_lock(&journal->j_state_lock);
2097 if (test_opt(sb, BARRIER))
2098 journal->j_flags |= JFS_BARRIER;
2100 journal->j_flags &= ~JFS_BARRIER;
2101 if (test_opt(sb, DATA_ERR_ABORT))
2102 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2104 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2105 spin_unlock(&journal->j_state_lock);
2108 static journal_t *ext3_get_journal(struct super_block *sb,
2109 unsigned int journal_inum)
2111 struct inode *journal_inode;
2114 /* First, test for the existence of a valid inode on disk. Bad
2115 * things happen if we iget() an unused inode, as the subsequent
2116 * iput() will try to delete it. */
2118 journal_inode = ext3_iget(sb, journal_inum);
2119 if (IS_ERR(journal_inode)) {
2120 ext3_msg(sb, KERN_ERR, "error: no journal found");
2123 if (!journal_inode->i_nlink) {
2124 make_bad_inode(journal_inode);
2125 iput(journal_inode);
2126 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2130 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2131 journal_inode, journal_inode->i_size);
2132 if (!S_ISREG(journal_inode->i_mode)) {
2133 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2134 iput(journal_inode);
2138 journal = journal_init_inode(journal_inode);
2140 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2141 iput(journal_inode);
2144 journal->j_private = sb;
2145 ext3_init_journal_params(sb, journal);
2149 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2152 struct buffer_head * bh;
2156 int hblock, blocksize;
2157 ext3_fsblk_t sb_block;
2158 unsigned long offset;
2159 struct ext3_super_block * es;
2160 struct block_device *bdev;
2162 bdev = ext3_blkdev_get(j_dev, sb);
2166 if (bd_claim(bdev, sb)) {
2167 ext3_msg(sb, KERN_ERR,
2168 "error: failed to claim external journal device");
2169 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2173 blocksize = sb->s_blocksize;
2174 hblock = bdev_logical_block_size(bdev);
2175 if (blocksize < hblock) {
2176 ext3_msg(sb, KERN_ERR,
2177 "error: blocksize too small for journal device");
2181 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2182 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2183 set_blocksize(bdev, blocksize);
2184 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2185 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2186 "external journal");
2190 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2191 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2192 !(le32_to_cpu(es->s_feature_incompat) &
2193 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2194 ext3_msg(sb, KERN_ERR, "error: external journal has "
2200 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2201 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2206 len = le32_to_cpu(es->s_blocks_count);
2207 start = sb_block + 1;
2208 brelse(bh); /* we're done with the superblock */
2210 journal = journal_init_dev(bdev, sb->s_bdev,
2211 start, len, blocksize);
2213 ext3_msg(sb, KERN_ERR,
2214 "error: failed to create device journal");
2217 journal->j_private = sb;
2218 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2219 wait_on_buffer(journal->j_sb_buffer);
2220 if (!buffer_uptodate(journal->j_sb_buffer)) {
2221 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2224 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2225 ext3_msg(sb, KERN_ERR,
2226 "error: external journal has more than one "
2227 "user (unsupported) - %d",
2228 be32_to_cpu(journal->j_superblock->s_nr_users));
2231 EXT3_SB(sb)->journal_bdev = bdev;
2232 ext3_init_journal_params(sb, journal);
2235 journal_destroy(journal);
2237 ext3_blkdev_put(bdev);
2241 static int ext3_load_journal(struct super_block *sb,
2242 struct ext3_super_block *es,
2243 unsigned long journal_devnum)
2246 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2249 int really_read_only;
2251 if (journal_devnum &&
2252 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2253 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2254 "numbers have changed");
2255 journal_dev = new_decode_dev(journal_devnum);
2257 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2259 really_read_only = bdev_read_only(sb->s_bdev);
2262 * Are we loading a blank journal or performing recovery after a
2263 * crash? For recovery, we need to check in advance whether we
2264 * can get read-write access to the device.
2267 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2268 if (sb->s_flags & MS_RDONLY) {
2269 ext3_msg(sb, KERN_INFO,
2270 "recovery required on readonly filesystem");
2271 if (really_read_only) {
2272 ext3_msg(sb, KERN_ERR, "error: write access "
2273 "unavailable, cannot proceed");
2276 ext3_msg(sb, KERN_INFO,
2277 "write access will be enabled during recovery");
2281 if (journal_inum && journal_dev) {
2282 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2283 "and inode journals");
2288 if (!(journal = ext3_get_journal(sb, journal_inum)))
2291 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2295 if (!(journal->j_flags & JFS_BARRIER))
2296 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2298 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2299 err = journal_update_format(journal);
2301 ext3_msg(sb, KERN_ERR, "error updating journal");
2302 journal_destroy(journal);
2307 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2308 err = journal_wipe(journal, !really_read_only);
2310 err = journal_load(journal);
2313 ext3_msg(sb, KERN_ERR, "error loading journal");
2314 journal_destroy(journal);
2318 EXT3_SB(sb)->s_journal = journal;
2319 ext3_clear_journal_err(sb, es);
2321 if (journal_devnum &&
2322 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2323 es->s_journal_dev = cpu_to_le32(journal_devnum);
2325 /* Make sure we flush the recovery flag to disk. */
2326 ext3_commit_super(sb, es, 1);
2332 static int ext3_create_journal(struct super_block *sb,
2333 struct ext3_super_block *es,
2334 unsigned int journal_inum)
2339 if (sb->s_flags & MS_RDONLY) {
2340 ext3_msg(sb, KERN_ERR,
2341 "error: readonly filesystem when trying to "
2346 journal = ext3_get_journal(sb, journal_inum);
2350 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2353 err = journal_create(journal);
2355 ext3_msg(sb, KERN_ERR, "error creating journal");
2356 journal_destroy(journal);
2360 EXT3_SB(sb)->s_journal = journal;
2362 ext3_update_dynamic_rev(sb);
2363 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2364 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2366 es->s_journal_inum = cpu_to_le32(journal_inum);
2368 /* Make sure we flush the recovery flag to disk. */
2369 ext3_commit_super(sb, es, 1);
2374 static int ext3_commit_super(struct super_block *sb,
2375 struct ext3_super_block *es,
2378 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2384 * If the file system is mounted read-only, don't update the
2385 * superblock write time. This avoids updating the superblock
2386 * write time when we are mounting the root file system
2387 * read/only but we need to replay the journal; at that point,
2388 * for people who are east of GMT and who make their clock
2389 * tick in localtime for Windows bug-for-bug compatibility,
2390 * the clock is set in the future, and this will cause e2fsck
2391 * to complain and force a full file system check.
2393 if (!(sb->s_flags & MS_RDONLY))
2394 es->s_wtime = cpu_to_le32(get_seconds());
2395 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2396 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2397 BUFFER_TRACE(sbh, "marking dirty");
2398 mark_buffer_dirty(sbh);
2400 error = sync_dirty_buffer(sbh);
2406 * Have we just finished recovery? If so, and if we are mounting (or
2407 * remounting) the filesystem readonly, then we will end up with a
2408 * consistent fs on disk. Record that fact.
2410 static void ext3_mark_recovery_complete(struct super_block * sb,
2411 struct ext3_super_block * es)
2413 journal_t *journal = EXT3_SB(sb)->s_journal;
2415 journal_lock_updates(journal);
2416 if (journal_flush(journal) < 0)
2419 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2420 sb->s_flags & MS_RDONLY) {
2421 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2422 ext3_commit_super(sb, es, 1);
2426 journal_unlock_updates(journal);
2430 * If we are mounting (or read-write remounting) a filesystem whose journal
2431 * has recorded an error from a previous lifetime, move that error to the
2432 * main filesystem now.
2434 static void ext3_clear_journal_err(struct super_block *sb,
2435 struct ext3_super_block *es)
2441 journal = EXT3_SB(sb)->s_journal;
2444 * Now check for any error status which may have been recorded in the
2445 * journal by a prior ext3_error() or ext3_abort()
2448 j_errno = journal_errno(journal);
2452 errstr = ext3_decode_error(sb, j_errno, nbuf);
2453 ext3_warning(sb, __func__, "Filesystem error recorded "
2454 "from previous mount: %s", errstr);
2455 ext3_warning(sb, __func__, "Marking fs in need of "
2456 "filesystem check.");
2458 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2459 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2460 ext3_commit_super (sb, es, 1);
2462 journal_clear_err(journal);
2467 * Force the running and committing transactions to commit,
2468 * and wait on the commit.
2470 int ext3_force_commit(struct super_block *sb)
2475 if (sb->s_flags & MS_RDONLY)
2478 journal = EXT3_SB(sb)->s_journal;
2479 ret = ext3_journal_force_commit(journal);
2483 static int ext3_sync_fs(struct super_block *sb, int wait)
2487 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2489 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2495 * LVM calls this function before a (read-only) snapshot is created. This
2496 * gives us a chance to flush the journal completely and mark the fs clean.
2498 static int ext3_freeze(struct super_block *sb)
2503 if (!(sb->s_flags & MS_RDONLY)) {
2504 journal = EXT3_SB(sb)->s_journal;
2506 /* Now we set up the journal barrier. */
2507 journal_lock_updates(journal);
2510 * We don't want to clear needs_recovery flag when we failed
2511 * to flush the journal.
2513 error = journal_flush(journal);
2517 /* Journal blocked and flushed, clear needs_recovery flag. */
2518 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2519 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2526 journal_unlock_updates(journal);
2531 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2532 * flag here, even though the filesystem is not technically dirty yet.
2534 static int ext3_unfreeze(struct super_block *sb)
2536 if (!(sb->s_flags & MS_RDONLY)) {
2538 /* Reser the needs_recovery flag before the fs is unlocked. */
2539 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2540 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2542 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2547 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2549 struct ext3_super_block * es;
2550 struct ext3_sb_info *sbi = EXT3_SB(sb);
2551 ext3_fsblk_t n_blocks_count = 0;
2552 unsigned long old_sb_flags;
2553 struct ext3_mount_options old_opts;
2561 /* Store the original options */
2563 old_sb_flags = sb->s_flags;
2564 old_opts.s_mount_opt = sbi->s_mount_opt;
2565 old_opts.s_resuid = sbi->s_resuid;
2566 old_opts.s_resgid = sbi->s_resgid;
2567 old_opts.s_commit_interval = sbi->s_commit_interval;
2569 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2570 for (i = 0; i < MAXQUOTAS; i++)
2571 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2575 * Allow the "check" option to be passed as a remount option.
2577 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2582 if (test_opt(sb, ABORT))
2583 ext3_abort(sb, __func__, "Abort forced by user");
2585 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2586 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2590 ext3_init_journal_params(sb, sbi->s_journal);
2592 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2593 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2594 if (test_opt(sb, ABORT)) {
2599 if (*flags & MS_RDONLY) {
2601 * First of all, the unconditional stuff we have to do
2602 * to disable replay of the journal when we next remount
2604 sb->s_flags |= MS_RDONLY;
2607 * OK, test if we are remounting a valid rw partition
2608 * readonly, and if so set the rdonly flag and then
2609 * mark the partition as valid again.
2611 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2612 (sbi->s_mount_state & EXT3_VALID_FS))
2613 es->s_state = cpu_to_le16(sbi->s_mount_state);
2615 ext3_mark_recovery_complete(sb, es);
2618 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2619 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2620 ext3_msg(sb, KERN_WARNING,
2621 "warning: couldn't remount RDWR "
2622 "because of unsupported optional "
2623 "features (%x)", le32_to_cpu(ret));
2629 * If we have an unprocessed orphan list hanging
2630 * around from a previously readonly bdev mount,
2631 * require a full umount/remount for now.
2633 if (es->s_last_orphan) {
2634 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2635 "remount RDWR because of unprocessed "
2636 "orphan inode list. Please "
2637 "umount/remount instead.");
2643 * Mounting a RDONLY partition read-write, so reread
2644 * and store the current valid flag. (It may have
2645 * been changed by e2fsck since we originally mounted
2648 ext3_clear_journal_err(sb, es);
2649 sbi->s_mount_state = le16_to_cpu(es->s_state);
2650 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2652 if (!ext3_setup_super (sb, es, 0))
2653 sb->s_flags &= ~MS_RDONLY;
2657 /* Release old quota file names */
2658 for (i = 0; i < MAXQUOTAS; i++)
2659 if (old_opts.s_qf_names[i] &&
2660 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2661 kfree(old_opts.s_qf_names[i]);
2667 sb->s_flags = old_sb_flags;
2668 sbi->s_mount_opt = old_opts.s_mount_opt;
2669 sbi->s_resuid = old_opts.s_resuid;
2670 sbi->s_resgid = old_opts.s_resgid;
2671 sbi->s_commit_interval = old_opts.s_commit_interval;
2673 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2674 for (i = 0; i < MAXQUOTAS; i++) {
2675 if (sbi->s_qf_names[i] &&
2676 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2677 kfree(sbi->s_qf_names[i]);
2678 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2686 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2688 struct super_block *sb = dentry->d_sb;
2689 struct ext3_sb_info *sbi = EXT3_SB(sb);
2690 struct ext3_super_block *es = sbi->s_es;
2693 if (test_opt(sb, MINIX_DF)) {
2694 sbi->s_overhead_last = 0;
2695 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2696 unsigned long ngroups = sbi->s_groups_count, i;
2697 ext3_fsblk_t overhead = 0;
2701 * Compute the overhead (FS structures). This is constant
2702 * for a given filesystem unless the number of block groups
2703 * changes so we cache the previous value until it does.
2707 * All of the blocks before first_data_block are
2710 overhead = le32_to_cpu(es->s_first_data_block);
2713 * Add the overhead attributed to the superblock and
2714 * block group descriptors. If the sparse superblocks
2715 * feature is turned on, then not all groups have this.
2717 for (i = 0; i < ngroups; i++) {
2718 overhead += ext3_bg_has_super(sb, i) +
2719 ext3_bg_num_gdb(sb, i);
2724 * Every block group has an inode bitmap, a block
2725 * bitmap, and an inode table.
2727 overhead += ngroups * (2 + sbi->s_itb_per_group);
2728 sbi->s_overhead_last = overhead;
2730 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2733 buf->f_type = EXT3_SUPER_MAGIC;
2734 buf->f_bsize = sb->s_blocksize;
2735 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2736 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2737 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2738 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2740 buf->f_files = le32_to_cpu(es->s_inodes_count);
2741 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2742 buf->f_namelen = EXT3_NAME_LEN;
2743 fsid = le64_to_cpup((void *)es->s_uuid) ^
2744 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2745 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2746 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2750 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2751 * is locked for write. Otherwise the are possible deadlocks:
2752 * Process 1 Process 2
2753 * ext3_create() quota_sync()
2754 * journal_start() write_dquot()
2755 * dquot_initialize() down(dqio_mutex)
2756 * down(dqio_mutex) journal_start()
2762 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2764 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2767 static int ext3_write_dquot(struct dquot *dquot)
2771 struct inode *inode;
2773 inode = dquot_to_inode(dquot);
2774 handle = ext3_journal_start(inode,
2775 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2777 return PTR_ERR(handle);
2778 ret = dquot_commit(dquot);
2779 err = ext3_journal_stop(handle);
2785 static int ext3_acquire_dquot(struct dquot *dquot)
2790 handle = ext3_journal_start(dquot_to_inode(dquot),
2791 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2793 return PTR_ERR(handle);
2794 ret = dquot_acquire(dquot);
2795 err = ext3_journal_stop(handle);
2801 static int ext3_release_dquot(struct dquot *dquot)
2806 handle = ext3_journal_start(dquot_to_inode(dquot),
2807 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2808 if (IS_ERR(handle)) {
2809 /* Release dquot anyway to avoid endless cycle in dqput() */
2810 dquot_release(dquot);
2811 return PTR_ERR(handle);
2813 ret = dquot_release(dquot);
2814 err = ext3_journal_stop(handle);
2820 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2822 /* Are we journaling quotas? */
2823 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2824 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2825 dquot_mark_dquot_dirty(dquot);
2826 return ext3_write_dquot(dquot);
2828 return dquot_mark_dquot_dirty(dquot);
2832 static int ext3_write_info(struct super_block *sb, int type)
2837 /* Data block + inode block */
2838 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2840 return PTR_ERR(handle);
2841 ret = dquot_commit_info(sb, type);
2842 err = ext3_journal_stop(handle);
2849 * Turn on quotas during mount time - we need to find
2850 * the quota file and such...
2852 static int ext3_quota_on_mount(struct super_block *sb, int type)
2854 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2855 EXT3_SB(sb)->s_jquota_fmt, type);
2859 * Standard function to be called on quota_on
2861 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2862 char *name, int remount)
2867 if (!test_opt(sb, QUOTA))
2869 /* When remounting, no checks are needed and in fact, name is NULL */
2871 return vfs_quota_on(sb, type, format_id, name, remount);
2873 err = kern_path(name, LOOKUP_FOLLOW, &path);
2877 /* Quotafile not on the same filesystem? */
2878 if (path.mnt->mnt_sb != sb) {
2882 /* Journaling quota? */
2883 if (EXT3_SB(sb)->s_qf_names[type]) {
2884 /* Quotafile not of fs root? */
2885 if (path.dentry->d_parent != sb->s_root)
2886 ext3_msg(sb, KERN_WARNING,
2887 "warning: Quota file not on filesystem root. "
2888 "Journaled quota will not work.");
2892 * When we journal data on quota file, we have to flush journal to see
2893 * all updates to the file when we bypass pagecache...
2895 if (ext3_should_journal_data(path.dentry->d_inode)) {
2897 * We don't need to lock updates but journal_flush() could
2898 * otherwise be livelocked...
2900 journal_lock_updates(EXT3_SB(sb)->s_journal);
2901 err = journal_flush(EXT3_SB(sb)->s_journal);
2902 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2909 err = vfs_quota_on_path(sb, type, format_id, &path);
2914 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2915 * acquiring the locks... As quota files are never truncated and quota code
2916 * itself serializes the operations (and noone else should touch the files)
2917 * we don't have to be afraid of races */
2918 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2919 size_t len, loff_t off)
2921 struct inode *inode = sb_dqopt(sb)->files[type];
2922 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2924 int offset = off & (sb->s_blocksize - 1);
2927 struct buffer_head *bh;
2928 loff_t i_size = i_size_read(inode);
2932 if (off+len > i_size)
2935 while (toread > 0) {
2936 tocopy = sb->s_blocksize - offset < toread ?
2937 sb->s_blocksize - offset : toread;
2938 bh = ext3_bread(NULL, inode, blk, 0, &err);
2941 if (!bh) /* A hole? */
2942 memset(data, 0, tocopy);
2944 memcpy(data, bh->b_data+offset, tocopy);
2954 /* Write to quotafile (we know the transaction is already started and has
2955 * enough credits) */
2956 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2957 const char *data, size_t len, loff_t off)
2959 struct inode *inode = sb_dqopt(sb)->files[type];
2960 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2962 int offset = off & (sb->s_blocksize - 1);
2963 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2964 struct buffer_head *bh;
2965 handle_t *handle = journal_current_handle();
2968 ext3_msg(sb, KERN_WARNING,
2969 "warning: quota write (off=%llu, len=%llu)"
2970 " cancelled because transaction is not started.",
2971 (unsigned long long)off, (unsigned long long)len);
2976 * Since we account only one data block in transaction credits,
2977 * then it is impossible to cross a block boundary.
2979 if (sb->s_blocksize - offset < len) {
2980 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2981 " cancelled because not block aligned",
2982 (unsigned long long)off, (unsigned long long)len);
2985 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2986 bh = ext3_bread(handle, inode, blk, 1, &err);
2989 if (journal_quota) {
2990 err = ext3_journal_get_write_access(handle, bh);
2997 memcpy(bh->b_data+offset, data, len);
2998 flush_dcache_page(bh->b_page);
3001 err = ext3_journal_dirty_metadata(handle, bh);
3003 /* Always do at least ordered writes for quotas */
3004 err = ext3_journal_dirty_data(handle, bh);
3005 mark_buffer_dirty(bh);
3010 mutex_unlock(&inode->i_mutex);
3013 if (inode->i_size < off + len) {
3014 i_size_write(inode, off + len);
3015 EXT3_I(inode)->i_disksize = inode->i_size;
3018 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3019 ext3_mark_inode_dirty(handle, inode);
3020 mutex_unlock(&inode->i_mutex);
3026 static int ext3_get_sb(struct file_system_type *fs_type,
3027 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3029 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
3032 static struct file_system_type ext3_fs_type = {
3033 .owner = THIS_MODULE,
3035 .get_sb = ext3_get_sb,
3036 .kill_sb = kill_block_super,
3037 .fs_flags = FS_REQUIRES_DEV,
3040 static int __init init_ext3_fs(void)
3042 int err = init_ext3_xattr();
3045 err = init_inodecache();
3048 err = register_filesystem(&ext3_fs_type);
3053 destroy_inodecache();
3059 static void __exit exit_ext3_fs(void)
3061 unregister_filesystem(&ext3_fs_type);
3062 destroy_inodecache();
3066 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3067 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3068 MODULE_LICENSE("GPL");
3069 module_init(init_ext3_fs)
3070 module_exit(exit_ext3_fs)