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/blkdev.h>
21 #include <linux/parser.h>
22 #include <linux/exportfs.h>
23 #include <linux/statfs.h>
24 #include <linux/random.h>
25 #include <linux/mount.h>
26 #include <linux/quotaops.h>
27 #include <linux/seq_file.h>
28 #include <linux/log2.h>
29 #include <linux/cleancache.h>
31 #include <asm/uaccess.h>
33 #define CREATE_TRACE_POINTS
40 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
41 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
43 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
46 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
47 unsigned long journal_devnum);
48 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
50 static int ext3_commit_super(struct super_block *sb,
51 struct ext3_super_block *es,
53 static void ext3_mark_recovery_complete(struct super_block * sb,
54 struct ext3_super_block * es);
55 static void ext3_clear_journal_err(struct super_block * sb,
56 struct ext3_super_block * es);
57 static int ext3_sync_fs(struct super_block *sb, int wait);
58 static const char *ext3_decode_error(struct super_block * sb, int errno,
60 static int ext3_remount (struct super_block * sb, int * flags, char * data);
61 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
62 static int ext3_unfreeze(struct super_block *sb);
63 static int ext3_freeze(struct super_block *sb);
66 * Wrappers for journal_start/end.
68 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
72 if (sb->s_flags & MS_RDONLY)
73 return ERR_PTR(-EROFS);
75 /* Special case here: if the journal has aborted behind our
76 * backs (eg. EIO in the commit thread), then we still need to
77 * take the FS itself readonly cleanly. */
78 journal = EXT3_SB(sb)->s_journal;
79 if (is_journal_aborted(journal)) {
80 ext3_abort(sb, __func__,
81 "Detected aborted journal");
82 return ERR_PTR(-EROFS);
85 return journal_start(journal, nblocks);
88 int __ext3_journal_stop(const char *where, handle_t *handle)
90 struct super_block *sb;
94 sb = handle->h_transaction->t_journal->j_private;
96 rc = journal_stop(handle);
101 __ext3_std_error(sb, where, err);
105 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
106 struct buffer_head *bh, handle_t *handle, int err)
109 const char *errstr = ext3_decode_error(NULL, err, nbuf);
112 BUFFER_TRACE(bh, "abort");
117 if (is_handle_aborted(handle))
120 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
121 caller, errstr, err_fn);
123 journal_abort_handle(handle);
126 void ext3_msg(struct super_block *sb, const char *prefix,
127 const char *fmt, ...)
129 struct va_format vaf;
137 printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
142 /* Deal with the reporting of failure conditions on a filesystem such as
143 * inconsistencies detected or read IO failures.
145 * On ext2, we can store the error state of the filesystem in the
146 * superblock. That is not possible on ext3, because we may have other
147 * write ordering constraints on the superblock which prevent us from
148 * writing it out straight away; and given that the journal is about to
149 * be aborted, we can't rely on the current, or future, transactions to
150 * write out the superblock safely.
152 * We'll just use the journal_abort() error code to record an error in
153 * the journal instead. On recovery, the journal will complain about
154 * that error until we've noted it down and cleared it.
157 static void ext3_handle_error(struct super_block *sb)
159 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
161 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
162 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
164 if (sb->s_flags & MS_RDONLY)
167 if (!test_opt (sb, ERRORS_CONT)) {
168 journal_t *journal = EXT3_SB(sb)->s_journal;
170 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
172 journal_abort(journal, -EIO);
174 if (test_opt (sb, ERRORS_RO)) {
175 ext3_msg(sb, KERN_CRIT,
176 "error: remounting filesystem read-only");
177 sb->s_flags |= MS_RDONLY;
179 ext3_commit_super(sb, es, 1);
180 if (test_opt(sb, ERRORS_PANIC))
181 panic("EXT3-fs (%s): panic forced after error\n",
185 void ext3_error(struct super_block *sb, const char *function,
186 const char *fmt, ...)
188 struct va_format vaf;
196 printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
197 sb->s_id, function, &vaf);
201 ext3_handle_error(sb);
204 static const char *ext3_decode_error(struct super_block * sb, int errno,
211 errstr = "IO failure";
214 errstr = "Out of memory";
217 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
218 errstr = "Journal has aborted";
220 errstr = "Readonly filesystem";
223 /* If the caller passed in an extra buffer for unknown
224 * errors, textualise them now. Else we just return
227 /* Check for truncated error codes... */
228 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
237 /* __ext3_std_error decodes expected errors from journaling functions
238 * automatically and invokes the appropriate error response. */
240 void __ext3_std_error (struct super_block * sb, const char * function,
246 /* Special case: if the error is EROFS, and we're not already
247 * inside a transaction, then there's really no point in logging
249 if (errno == -EROFS && journal_current_handle() == NULL &&
250 (sb->s_flags & MS_RDONLY))
253 errstr = ext3_decode_error(sb, errno, nbuf);
254 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
256 ext3_handle_error(sb);
260 * ext3_abort is a much stronger failure handler than ext3_error. The
261 * abort function may be used to deal with unrecoverable failures such
262 * as journal IO errors or ENOMEM at a critical moment in log management.
264 * We unconditionally force the filesystem into an ABORT|READONLY state,
265 * unless the error response on the fs has been set to panic in which
266 * case we take the easy way out and panic immediately.
269 void ext3_abort(struct super_block *sb, const char *function,
270 const char *fmt, ...)
272 struct va_format vaf;
280 printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
281 sb->s_id, function, &vaf);
285 if (test_opt(sb, ERRORS_PANIC))
286 panic("EXT3-fs: panic from previous error\n");
288 if (sb->s_flags & MS_RDONLY)
291 ext3_msg(sb, KERN_CRIT,
292 "error: remounting filesystem read-only");
293 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
294 sb->s_flags |= MS_RDONLY;
295 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
296 if (EXT3_SB(sb)->s_journal)
297 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
300 void ext3_warning(struct super_block *sb, const char *function,
301 const char *fmt, ...)
303 struct va_format vaf;
311 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
312 sb->s_id, function, &vaf);
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 = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
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)
367 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
370 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
372 struct block_device *bdev;
375 bdev = sbi->journal_bdev;
377 ret = ext3_blkdev_put(bdev);
378 sbi->journal_bdev = NULL;
383 static inline struct inode *orphan_list_entry(struct list_head *l)
385 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
388 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
392 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
393 le32_to_cpu(sbi->s_es->s_last_orphan));
395 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
396 list_for_each(l, &sbi->s_orphan) {
397 struct inode *inode = orphan_list_entry(l);
398 ext3_msg(sb, KERN_ERR, " "
399 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
400 inode->i_sb->s_id, inode->i_ino, inode,
401 inode->i_mode, inode->i_nlink,
406 static void ext3_put_super (struct super_block * sb)
408 struct ext3_sb_info *sbi = EXT3_SB(sb);
409 struct ext3_super_block *es = sbi->s_es;
412 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
413 ext3_xattr_put_super(sb);
414 err = journal_destroy(sbi->s_journal);
415 sbi->s_journal = NULL;
417 ext3_abort(sb, __func__, "Couldn't clean up the journal");
419 if (!(sb->s_flags & MS_RDONLY)) {
420 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
421 es->s_state = cpu_to_le16(sbi->s_mount_state);
422 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
423 mark_buffer_dirty(sbi->s_sbh);
424 ext3_commit_super(sb, es, 1);
427 for (i = 0; i < sbi->s_gdb_count; i++)
428 brelse(sbi->s_group_desc[i]);
429 kfree(sbi->s_group_desc);
430 percpu_counter_destroy(&sbi->s_freeblocks_counter);
431 percpu_counter_destroy(&sbi->s_freeinodes_counter);
432 percpu_counter_destroy(&sbi->s_dirs_counter);
435 for (i = 0; i < MAXQUOTAS; i++)
436 kfree(sbi->s_qf_names[i]);
439 /* Debugging code just in case the in-memory inode orphan list
440 * isn't empty. The on-disk one can be non-empty if we've
441 * detected an error and taken the fs readonly, but the
442 * in-memory list had better be clean by this point. */
443 if (!list_empty(&sbi->s_orphan))
444 dump_orphan_list(sb, sbi);
445 J_ASSERT(list_empty(&sbi->s_orphan));
447 invalidate_bdev(sb->s_bdev);
448 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
450 * Invalidate the journal device's buffers. We don't want them
451 * floating about in memory - the physical journal device may
452 * hotswapped, and it breaks the `ro-after' testing code.
454 sync_blockdev(sbi->journal_bdev);
455 invalidate_bdev(sbi->journal_bdev);
456 ext3_blkdev_remove(sbi);
458 sb->s_fs_info = NULL;
459 kfree(sbi->s_blockgroup_lock);
463 static struct kmem_cache *ext3_inode_cachep;
466 * Called inside transaction, so use GFP_NOFS
468 static struct inode *ext3_alloc_inode(struct super_block *sb)
470 struct ext3_inode_info *ei;
472 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
475 ei->i_block_alloc_info = NULL;
476 ei->vfs_inode.i_version = 1;
477 atomic_set(&ei->i_datasync_tid, 0);
478 atomic_set(&ei->i_sync_tid, 0);
479 return &ei->vfs_inode;
482 static int ext3_drop_inode(struct inode *inode)
484 int drop = generic_drop_inode(inode);
486 trace_ext3_drop_inode(inode, drop);
490 static void ext3_i_callback(struct rcu_head *head)
492 struct inode *inode = container_of(head, struct inode, i_rcu);
493 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
496 static void ext3_destroy_inode(struct inode *inode)
498 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
499 printk("EXT3 Inode %p: orphan list check failed!\n",
501 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
502 EXT3_I(inode), sizeof(struct ext3_inode_info),
506 call_rcu(&inode->i_rcu, ext3_i_callback);
509 static void init_once(void *foo)
511 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
513 INIT_LIST_HEAD(&ei->i_orphan);
514 #ifdef CONFIG_EXT3_FS_XATTR
515 init_rwsem(&ei->xattr_sem);
517 mutex_init(&ei->truncate_mutex);
518 inode_init_once(&ei->vfs_inode);
521 static int init_inodecache(void)
523 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
524 sizeof(struct ext3_inode_info),
525 0, (SLAB_RECLAIM_ACCOUNT|
528 if (ext3_inode_cachep == NULL)
533 static void destroy_inodecache(void)
535 kmem_cache_destroy(ext3_inode_cachep);
538 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
540 #if defined(CONFIG_QUOTA)
541 struct ext3_sb_info *sbi = EXT3_SB(sb);
543 if (sbi->s_jquota_fmt) {
546 switch (sbi->s_jquota_fmt) {
557 seq_printf(seq, ",jqfmt=%s", fmtname);
560 if (sbi->s_qf_names[USRQUOTA])
561 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
563 if (sbi->s_qf_names[GRPQUOTA])
564 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
566 if (test_opt(sb, USRQUOTA))
567 seq_puts(seq, ",usrquota");
569 if (test_opt(sb, GRPQUOTA))
570 seq_puts(seq, ",grpquota");
574 static char *data_mode_string(unsigned long mode)
577 case EXT3_MOUNT_JOURNAL_DATA:
579 case EXT3_MOUNT_ORDERED_DATA:
581 case EXT3_MOUNT_WRITEBACK_DATA:
589 * - it's set to a non-default value OR
590 * - if the per-sb default is different from the global default
592 static int ext3_show_options(struct seq_file *seq, struct dentry *root)
594 struct super_block *sb = root->d_sb;
595 struct ext3_sb_info *sbi = EXT3_SB(sb);
596 struct ext3_super_block *es = sbi->s_es;
597 unsigned long def_mount_opts;
599 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
601 if (sbi->s_sb_block != 1)
602 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
603 if (test_opt(sb, MINIX_DF))
604 seq_puts(seq, ",minixdf");
605 if (test_opt(sb, GRPID))
606 seq_puts(seq, ",grpid");
607 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
608 seq_puts(seq, ",nogrpid");
609 if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT3_DEF_RESUID)) ||
610 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
611 seq_printf(seq, ",resuid=%u",
612 from_kuid_munged(&init_user_ns, sbi->s_resuid));
614 if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT3_DEF_RESGID)) ||
615 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
616 seq_printf(seq, ",resgid=%u",
617 from_kgid_munged(&init_user_ns, sbi->s_resgid));
619 if (test_opt(sb, ERRORS_RO)) {
620 int def_errors = le16_to_cpu(es->s_errors);
622 if (def_errors == EXT3_ERRORS_PANIC ||
623 def_errors == EXT3_ERRORS_CONTINUE) {
624 seq_puts(seq, ",errors=remount-ro");
627 if (test_opt(sb, ERRORS_CONT))
628 seq_puts(seq, ",errors=continue");
629 if (test_opt(sb, ERRORS_PANIC))
630 seq_puts(seq, ",errors=panic");
631 if (test_opt(sb, NO_UID32))
632 seq_puts(seq, ",nouid32");
633 if (test_opt(sb, DEBUG))
634 seq_puts(seq, ",debug");
635 #ifdef CONFIG_EXT3_FS_XATTR
636 if (test_opt(sb, XATTR_USER))
637 seq_puts(seq, ",user_xattr");
638 if (!test_opt(sb, XATTR_USER) &&
639 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
640 seq_puts(seq, ",nouser_xattr");
643 #ifdef CONFIG_EXT3_FS_POSIX_ACL
644 if (test_opt(sb, POSIX_ACL))
645 seq_puts(seq, ",acl");
646 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
647 seq_puts(seq, ",noacl");
649 if (!test_opt(sb, RESERVATION))
650 seq_puts(seq, ",noreservation");
651 if (sbi->s_commit_interval) {
652 seq_printf(seq, ",commit=%u",
653 (unsigned) (sbi->s_commit_interval / HZ));
657 * Always display barrier state so it's clear what the status is.
659 seq_puts(seq, ",barrier=");
660 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
661 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
662 if (test_opt(sb, DATA_ERR_ABORT))
663 seq_puts(seq, ",data_err=abort");
665 if (test_opt(sb, NOLOAD))
666 seq_puts(seq, ",norecovery");
668 ext3_show_quota_options(seq, sb);
674 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
675 u64 ino, u32 generation)
679 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
680 return ERR_PTR(-ESTALE);
681 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
682 return ERR_PTR(-ESTALE);
684 /* iget isn't really right if the inode is currently unallocated!!
686 * ext3_read_inode will return a bad_inode if the inode had been
687 * deleted, so we should be safe.
689 * Currently we don't know the generation for parent directory, so
690 * a generation of 0 means "accept any"
692 inode = ext3_iget(sb, ino);
694 return ERR_CAST(inode);
695 if (generation && inode->i_generation != generation) {
697 return ERR_PTR(-ESTALE);
703 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
704 int fh_len, int fh_type)
706 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
710 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
711 int fh_len, int fh_type)
713 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
718 * Try to release metadata pages (indirect blocks, directories) which are
719 * mapped via the block device. Since these pages could have journal heads
720 * which would prevent try_to_free_buffers() from freeing them, we must use
721 * jbd layer's try_to_free_buffers() function to release them.
723 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
726 journal_t *journal = EXT3_SB(sb)->s_journal;
728 WARN_ON(PageChecked(page));
729 if (!page_has_buffers(page))
732 return journal_try_to_free_buffers(journal, page,
734 return try_to_free_buffers(page);
738 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
739 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
741 static int ext3_write_dquot(struct dquot *dquot);
742 static int ext3_acquire_dquot(struct dquot *dquot);
743 static int ext3_release_dquot(struct dquot *dquot);
744 static int ext3_mark_dquot_dirty(struct dquot *dquot);
745 static int ext3_write_info(struct super_block *sb, int type);
746 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
748 static int ext3_quota_on_mount(struct super_block *sb, int type);
749 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
750 size_t len, loff_t off);
751 static ssize_t ext3_quota_write(struct super_block *sb, int type,
752 const char *data, size_t len, loff_t off);
754 static const struct dquot_operations ext3_quota_operations = {
755 .write_dquot = ext3_write_dquot,
756 .acquire_dquot = ext3_acquire_dquot,
757 .release_dquot = ext3_release_dquot,
758 .mark_dirty = ext3_mark_dquot_dirty,
759 .write_info = ext3_write_info,
760 .alloc_dquot = dquot_alloc,
761 .destroy_dquot = dquot_destroy,
764 static const struct quotactl_ops ext3_qctl_operations = {
765 .quota_on = ext3_quota_on,
766 .quota_off = dquot_quota_off,
767 .quota_sync = dquot_quota_sync,
768 .get_info = dquot_get_dqinfo,
769 .set_info = dquot_set_dqinfo,
770 .get_dqblk = dquot_get_dqblk,
771 .set_dqblk = dquot_set_dqblk
775 static const struct super_operations ext3_sops = {
776 .alloc_inode = ext3_alloc_inode,
777 .destroy_inode = ext3_destroy_inode,
778 .write_inode = ext3_write_inode,
779 .dirty_inode = ext3_dirty_inode,
780 .drop_inode = ext3_drop_inode,
781 .evict_inode = ext3_evict_inode,
782 .put_super = ext3_put_super,
783 .sync_fs = ext3_sync_fs,
784 .freeze_fs = ext3_freeze,
785 .unfreeze_fs = ext3_unfreeze,
786 .statfs = ext3_statfs,
787 .remount_fs = ext3_remount,
788 .show_options = ext3_show_options,
790 .quota_read = ext3_quota_read,
791 .quota_write = ext3_quota_write,
793 .bdev_try_to_free_page = bdev_try_to_free_page,
796 static const struct export_operations ext3_export_ops = {
797 .fh_to_dentry = ext3_fh_to_dentry,
798 .fh_to_parent = ext3_fh_to_parent,
799 .get_parent = ext3_get_parent,
803 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
804 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
805 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
806 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
807 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
808 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
809 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
810 Opt_data_err_abort, Opt_data_err_ignore,
811 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
812 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
813 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
814 Opt_resize, Opt_usrquota, Opt_grpquota
817 static const match_table_t tokens = {
818 {Opt_bsd_df, "bsddf"},
819 {Opt_minix_df, "minixdf"},
820 {Opt_grpid, "grpid"},
821 {Opt_grpid, "bsdgroups"},
822 {Opt_nogrpid, "nogrpid"},
823 {Opt_nogrpid, "sysvgroups"},
824 {Opt_resgid, "resgid=%u"},
825 {Opt_resuid, "resuid=%u"},
827 {Opt_err_cont, "errors=continue"},
828 {Opt_err_panic, "errors=panic"},
829 {Opt_err_ro, "errors=remount-ro"},
830 {Opt_nouid32, "nouid32"},
831 {Opt_nocheck, "nocheck"},
832 {Opt_nocheck, "check=none"},
833 {Opt_debug, "debug"},
834 {Opt_oldalloc, "oldalloc"},
835 {Opt_orlov, "orlov"},
836 {Opt_user_xattr, "user_xattr"},
837 {Opt_nouser_xattr, "nouser_xattr"},
839 {Opt_noacl, "noacl"},
840 {Opt_reservation, "reservation"},
841 {Opt_noreservation, "noreservation"},
842 {Opt_noload, "noload"},
843 {Opt_noload, "norecovery"},
846 {Opt_commit, "commit=%u"},
847 {Opt_journal_update, "journal=update"},
848 {Opt_journal_inum, "journal=%u"},
849 {Opt_journal_dev, "journal_dev=%u"},
850 {Opt_abort, "abort"},
851 {Opt_data_journal, "data=journal"},
852 {Opt_data_ordered, "data=ordered"},
853 {Opt_data_writeback, "data=writeback"},
854 {Opt_data_err_abort, "data_err=abort"},
855 {Opt_data_err_ignore, "data_err=ignore"},
856 {Opt_offusrjquota, "usrjquota="},
857 {Opt_usrjquota, "usrjquota=%s"},
858 {Opt_offgrpjquota, "grpjquota="},
859 {Opt_grpjquota, "grpjquota=%s"},
860 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
861 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
862 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
863 {Opt_grpquota, "grpquota"},
864 {Opt_noquota, "noquota"},
865 {Opt_quota, "quota"},
866 {Opt_usrquota, "usrquota"},
867 {Opt_barrier, "barrier=%u"},
868 {Opt_barrier, "barrier"},
869 {Opt_nobarrier, "nobarrier"},
870 {Opt_resize, "resize"},
874 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
876 ext3_fsblk_t sb_block;
877 char *options = (char *) *data;
879 if (!options || strncmp(options, "sb=", 3) != 0)
880 return 1; /* Default location */
882 /*todo: use simple_strtoll with >32bit ext3 */
883 sb_block = simple_strtoul(options, &options, 0);
884 if (*options && *options != ',') {
885 ext3_msg(sb, "error: invalid sb specification: %s",
891 *data = (void *) options;
896 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
898 struct ext3_sb_info *sbi = EXT3_SB(sb);
901 if (sb_any_quota_loaded(sb) &&
902 !sbi->s_qf_names[qtype]) {
903 ext3_msg(sb, KERN_ERR,
904 "Cannot change journaled "
905 "quota options when quota turned on");
908 qname = match_strdup(args);
910 ext3_msg(sb, KERN_ERR,
911 "Not enough memory for storing quotafile name");
914 if (sbi->s_qf_names[qtype] &&
915 strcmp(sbi->s_qf_names[qtype], qname)) {
916 ext3_msg(sb, KERN_ERR,
917 "%s quota file already specified", QTYPE2NAME(qtype));
921 sbi->s_qf_names[qtype] = qname;
922 if (strchr(sbi->s_qf_names[qtype], '/')) {
923 ext3_msg(sb, KERN_ERR,
924 "quotafile must be on filesystem root");
925 kfree(sbi->s_qf_names[qtype]);
926 sbi->s_qf_names[qtype] = NULL;
929 set_opt(sbi->s_mount_opt, QUOTA);
933 static int clear_qf_name(struct super_block *sb, int qtype) {
935 struct ext3_sb_info *sbi = EXT3_SB(sb);
937 if (sb_any_quota_loaded(sb) &&
938 sbi->s_qf_names[qtype]) {
939 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
940 " when quota turned on");
944 * The space will be released later when all options are confirmed
947 sbi->s_qf_names[qtype] = NULL;
952 static int parse_options (char *options, struct super_block *sb,
953 unsigned int *inum, unsigned long *journal_devnum,
954 ext3_fsblk_t *n_blocks_count, int is_remount)
956 struct ext3_sb_info *sbi = EXT3_SB(sb);
958 substring_t args[MAX_OPT_ARGS];
970 while ((p = strsep (&options, ",")) != NULL) {
975 * Initialize args struct so we know whether arg was
976 * found; some options take optional arguments.
978 args[0].to = args[0].from = 0;
979 token = match_token(p, tokens, args);
982 clear_opt (sbi->s_mount_opt, MINIX_DF);
985 set_opt (sbi->s_mount_opt, MINIX_DF);
988 set_opt (sbi->s_mount_opt, GRPID);
991 clear_opt (sbi->s_mount_opt, GRPID);
994 if (match_int(&args[0], &option))
996 uid = make_kuid(current_user_ns(), option);
997 if (!uid_valid(uid)) {
998 ext3_msg(sb, KERN_ERR, "Invalid uid value %d", option);
1002 sbi->s_resuid = uid;
1005 if (match_int(&args[0], &option))
1007 gid = make_kgid(current_user_ns(), option);
1008 if (!gid_valid(gid)) {
1009 ext3_msg(sb, KERN_ERR, "Invalid gid value %d", option);
1012 sbi->s_resgid = gid;
1015 /* handled by get_sb_block() instead of here */
1016 /* *sb_block = match_int(&args[0]); */
1019 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1020 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1021 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1024 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1025 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1026 set_opt (sbi->s_mount_opt, ERRORS_RO);
1029 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1030 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1031 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1034 set_opt (sbi->s_mount_opt, NO_UID32);
1037 clear_opt (sbi->s_mount_opt, CHECK);
1040 set_opt (sbi->s_mount_opt, DEBUG);
1043 ext3_msg(sb, KERN_WARNING,
1044 "Ignoring deprecated oldalloc option");
1047 ext3_msg(sb, KERN_WARNING,
1048 "Ignoring deprecated orlov option");
1050 #ifdef CONFIG_EXT3_FS_XATTR
1051 case Opt_user_xattr:
1052 set_opt (sbi->s_mount_opt, XATTR_USER);
1054 case Opt_nouser_xattr:
1055 clear_opt (sbi->s_mount_opt, XATTR_USER);
1058 case Opt_user_xattr:
1059 case Opt_nouser_xattr:
1060 ext3_msg(sb, KERN_INFO,
1061 "(no)user_xattr options not supported");
1064 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1066 set_opt(sbi->s_mount_opt, POSIX_ACL);
1069 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1074 ext3_msg(sb, KERN_INFO,
1075 "(no)acl options not supported");
1078 case Opt_reservation:
1079 set_opt(sbi->s_mount_opt, RESERVATION);
1081 case Opt_noreservation:
1082 clear_opt(sbi->s_mount_opt, RESERVATION);
1084 case Opt_journal_update:
1086 /* Eventually we will want to be able to create
1087 a journal file here. For now, only allow the
1088 user to specify an existing inode to be the
1091 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1092 "journal on remount");
1095 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1097 case Opt_journal_inum:
1099 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1100 "journal on remount");
1103 if (match_int(&args[0], &option))
1107 case Opt_journal_dev:
1109 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1110 "journal on remount");
1113 if (match_int(&args[0], &option))
1115 *journal_devnum = option;
1118 set_opt (sbi->s_mount_opt, NOLOAD);
1121 if (match_int(&args[0], &option))
1126 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1127 sbi->s_commit_interval = HZ * option;
1129 case Opt_data_journal:
1130 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1132 case Opt_data_ordered:
1133 data_opt = EXT3_MOUNT_ORDERED_DATA;
1135 case Opt_data_writeback:
1136 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1139 if (test_opt(sb, DATA_FLAGS) == data_opt)
1141 ext3_msg(sb, KERN_ERR,
1142 "error: cannot change "
1143 "data mode on remount. The filesystem "
1144 "is mounted in data=%s mode and you "
1145 "try to remount it in data=%s mode.",
1146 data_mode_string(test_opt(sb,
1148 data_mode_string(data_opt));
1151 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1152 sbi->s_mount_opt |= data_opt;
1155 case Opt_data_err_abort:
1156 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1158 case Opt_data_err_ignore:
1159 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1163 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1167 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1170 case Opt_offusrjquota:
1171 if (!clear_qf_name(sb, USRQUOTA))
1174 case Opt_offgrpjquota:
1175 if (!clear_qf_name(sb, GRPQUOTA))
1178 case Opt_jqfmt_vfsold:
1179 qfmt = QFMT_VFS_OLD;
1181 case Opt_jqfmt_vfsv0:
1184 case Opt_jqfmt_vfsv1:
1187 if (sb_any_quota_loaded(sb) &&
1188 sbi->s_jquota_fmt != qfmt) {
1189 ext3_msg(sb, KERN_ERR, "error: cannot change "
1190 "journaled quota options when "
1191 "quota turned on.");
1194 sbi->s_jquota_fmt = qfmt;
1198 set_opt(sbi->s_mount_opt, QUOTA);
1199 set_opt(sbi->s_mount_opt, USRQUOTA);
1202 set_opt(sbi->s_mount_opt, QUOTA);
1203 set_opt(sbi->s_mount_opt, GRPQUOTA);
1206 if (sb_any_quota_loaded(sb)) {
1207 ext3_msg(sb, KERN_ERR, "error: cannot change "
1208 "quota options when quota turned on.");
1211 clear_opt(sbi->s_mount_opt, QUOTA);
1212 clear_opt(sbi->s_mount_opt, USRQUOTA);
1213 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1219 ext3_msg(sb, KERN_ERR,
1220 "error: quota options not supported.");
1224 case Opt_offusrjquota:
1225 case Opt_offgrpjquota:
1226 case Opt_jqfmt_vfsold:
1227 case Opt_jqfmt_vfsv0:
1228 case Opt_jqfmt_vfsv1:
1229 ext3_msg(sb, KERN_ERR,
1230 "error: journaled quota options not "
1237 set_opt(sbi->s_mount_opt, ABORT);
1240 clear_opt(sbi->s_mount_opt, BARRIER);
1244 if (match_int(&args[0], &option))
1247 option = 1; /* No argument, default to 1 */
1249 set_opt(sbi->s_mount_opt, BARRIER);
1251 clear_opt(sbi->s_mount_opt, BARRIER);
1257 ext3_msg(sb, KERN_ERR,
1258 "error: resize option only available "
1262 if (match_int(&args[0], &option) != 0)
1264 *n_blocks_count = option;
1267 ext3_msg(sb, KERN_WARNING,
1268 "warning: ignoring deprecated nobh option");
1271 ext3_msg(sb, KERN_WARNING,
1272 "warning: ignoring deprecated bh option");
1275 ext3_msg(sb, KERN_ERR,
1276 "error: unrecognized mount option \"%s\" "
1277 "or missing value", p);
1282 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1283 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1284 clear_opt(sbi->s_mount_opt, USRQUOTA);
1285 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1286 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1288 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1289 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1294 if (!sbi->s_jquota_fmt) {
1295 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1300 if (sbi->s_jquota_fmt) {
1301 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1302 "specified with no journaling "
1311 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1314 struct ext3_sb_info *sbi = EXT3_SB(sb);
1317 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1318 ext3_msg(sb, KERN_ERR,
1319 "error: revision level too high, "
1320 "forcing read-only mode");
1325 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1326 ext3_msg(sb, KERN_WARNING,
1327 "warning: mounting unchecked fs, "
1328 "running e2fsck is recommended");
1329 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1330 ext3_msg(sb, KERN_WARNING,
1331 "warning: mounting fs with errors, "
1332 "running e2fsck is recommended");
1333 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1334 le16_to_cpu(es->s_mnt_count) >=
1335 le16_to_cpu(es->s_max_mnt_count))
1336 ext3_msg(sb, KERN_WARNING,
1337 "warning: maximal mount count reached, "
1338 "running e2fsck is recommended");
1339 else if (le32_to_cpu(es->s_checkinterval) &&
1340 (le32_to_cpu(es->s_lastcheck) +
1341 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1342 ext3_msg(sb, KERN_WARNING,
1343 "warning: checktime reached, "
1344 "running e2fsck is recommended");
1346 /* @@@ We _will_ want to clear the valid bit if we find
1347 inconsistencies, to force a fsck at reboot. But for
1348 a plain journaled filesystem we can keep it set as
1349 valid forever! :) */
1350 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1352 if (!le16_to_cpu(es->s_max_mnt_count))
1353 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1354 le16_add_cpu(&es->s_mnt_count, 1);
1355 es->s_mtime = cpu_to_le32(get_seconds());
1356 ext3_update_dynamic_rev(sb);
1357 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1359 ext3_commit_super(sb, es, 1);
1360 if (test_opt(sb, DEBUG))
1361 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1362 "bpg=%lu, ipg=%lu, mo=%04lx]",
1364 sbi->s_groups_count,
1365 EXT3_BLOCKS_PER_GROUP(sb),
1366 EXT3_INODES_PER_GROUP(sb),
1369 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1370 char b[BDEVNAME_SIZE];
1371 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1372 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1374 ext3_msg(sb, KERN_INFO, "using internal journal");
1376 cleancache_init_fs(sb);
1380 /* Called at mount-time, super-block is locked */
1381 static int ext3_check_descriptors(struct super_block *sb)
1383 struct ext3_sb_info *sbi = EXT3_SB(sb);
1386 ext3_debug ("Checking group descriptors");
1388 for (i = 0; i < sbi->s_groups_count; i++) {
1389 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1390 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1391 ext3_fsblk_t last_block;
1393 if (i == sbi->s_groups_count - 1)
1394 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1396 last_block = first_block +
1397 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1399 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1400 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1402 ext3_error (sb, "ext3_check_descriptors",
1403 "Block bitmap for group %d"
1404 " not in group (block %lu)!",
1406 le32_to_cpu(gdp->bg_block_bitmap));
1409 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1410 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1412 ext3_error (sb, "ext3_check_descriptors",
1413 "Inode bitmap for group %d"
1414 " not in group (block %lu)!",
1416 le32_to_cpu(gdp->bg_inode_bitmap));
1419 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1420 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1423 ext3_error (sb, "ext3_check_descriptors",
1424 "Inode table for group %d"
1425 " not in group (block %lu)!",
1427 le32_to_cpu(gdp->bg_inode_table));
1432 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1433 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1438 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1439 * the superblock) which were deleted from all directories, but held open by
1440 * a process at the time of a crash. We walk the list and try to delete these
1441 * inodes at recovery time (only with a read-write filesystem).
1443 * In order to keep the orphan inode chain consistent during traversal (in
1444 * case of crash during recovery), we link each inode into the superblock
1445 * orphan list_head and handle it the same way as an inode deletion during
1446 * normal operation (which journals the operations for us).
1448 * We only do an iget() and an iput() on each inode, which is very safe if we
1449 * accidentally point at an in-use or already deleted inode. The worst that
1450 * can happen in this case is that we get a "bit already cleared" message from
1451 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1452 * e2fsck was run on this filesystem, and it must have already done the orphan
1453 * inode cleanup for us, so we can safely abort without any further action.
1455 static void ext3_orphan_cleanup (struct super_block * sb,
1456 struct ext3_super_block * es)
1458 unsigned int s_flags = sb->s_flags;
1459 int nr_orphans = 0, nr_truncates = 0;
1463 if (!es->s_last_orphan) {
1464 jbd_debug(4, "no orphan inodes to clean up\n");
1468 if (bdev_read_only(sb->s_bdev)) {
1469 ext3_msg(sb, KERN_ERR, "error: write access "
1470 "unavailable, skipping orphan cleanup.");
1474 /* Check if feature set allows readwrite operations */
1475 if (EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) {
1476 ext3_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
1477 "unknown ROCOMPAT features");
1481 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1482 /* don't clear list on RO mount w/ errors */
1483 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
1484 jbd_debug(1, "Errors on filesystem, "
1485 "clearing orphan list.\n");
1486 es->s_last_orphan = 0;
1488 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1492 if (s_flags & MS_RDONLY) {
1493 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1494 sb->s_flags &= ~MS_RDONLY;
1497 /* Needed for iput() to work correctly and not trash data */
1498 sb->s_flags |= MS_ACTIVE;
1499 /* Turn on quotas so that they are updated correctly */
1500 for (i = 0; i < MAXQUOTAS; i++) {
1501 if (EXT3_SB(sb)->s_qf_names[i]) {
1502 int ret = ext3_quota_on_mount(sb, i);
1504 ext3_msg(sb, KERN_ERR,
1505 "error: cannot turn on journaled "
1511 while (es->s_last_orphan) {
1512 struct inode *inode;
1514 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1515 if (IS_ERR(inode)) {
1516 es->s_last_orphan = 0;
1520 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1521 dquot_initialize(inode);
1522 if (inode->i_nlink) {
1524 "%s: truncating inode %lu to %Ld bytes\n",
1525 __func__, inode->i_ino, inode->i_size);
1526 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1527 inode->i_ino, inode->i_size);
1528 ext3_truncate(inode);
1532 "%s: deleting unreferenced inode %lu\n",
1533 __func__, inode->i_ino);
1534 jbd_debug(2, "deleting unreferenced inode %lu\n",
1538 iput(inode); /* The delete magic happens here! */
1541 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1544 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1545 PLURAL(nr_orphans));
1547 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1548 PLURAL(nr_truncates));
1550 /* Turn quotas off */
1551 for (i = 0; i < MAXQUOTAS; i++) {
1552 if (sb_dqopt(sb)->files[i])
1553 dquot_quota_off(sb, i);
1556 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1560 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1561 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1562 * We need to be 1 filesystem block less than the 2^32 sector limit.
1564 static loff_t ext3_max_size(int bits)
1566 loff_t res = EXT3_NDIR_BLOCKS;
1570 /* This is calculated to be the largest file size for a
1571 * dense, file such that the total number of
1572 * sectors in the file, including data and all indirect blocks,
1573 * does not exceed 2^32 -1
1574 * __u32 i_blocks representing the total number of
1575 * 512 bytes blocks of the file
1577 upper_limit = (1LL << 32) - 1;
1579 /* total blocks in file system block size */
1580 upper_limit >>= (bits - 9);
1583 /* indirect blocks */
1585 /* double indirect blocks */
1586 meta_blocks += 1 + (1LL << (bits-2));
1587 /* tripple indirect blocks */
1588 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1590 upper_limit -= meta_blocks;
1591 upper_limit <<= bits;
1593 res += 1LL << (bits-2);
1594 res += 1LL << (2*(bits-2));
1595 res += 1LL << (3*(bits-2));
1597 if (res > upper_limit)
1600 if (res > MAX_LFS_FILESIZE)
1601 res = MAX_LFS_FILESIZE;
1606 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1607 ext3_fsblk_t logic_sb_block,
1610 struct ext3_sb_info *sbi = EXT3_SB(sb);
1611 unsigned long bg, first_meta_bg;
1614 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1616 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1618 return (logic_sb_block + nr + 1);
1619 bg = sbi->s_desc_per_block * nr;
1620 if (ext3_bg_has_super(sb, bg))
1622 return (has_super + ext3_group_first_block_no(sb, bg));
1626 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1628 struct buffer_head * bh;
1629 struct ext3_super_block *es = NULL;
1630 struct ext3_sb_info *sbi;
1632 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1633 ext3_fsblk_t logic_sb_block;
1634 unsigned long offset = 0;
1635 unsigned int journal_inum = 0;
1636 unsigned long journal_devnum = 0;
1637 unsigned long def_mount_opts;
1648 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1652 sbi->s_blockgroup_lock =
1653 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1654 if (!sbi->s_blockgroup_lock) {
1658 sb->s_fs_info = sbi;
1659 sbi->s_mount_opt = 0;
1660 sbi->s_resuid = make_kuid(&init_user_ns, EXT3_DEF_RESUID);
1661 sbi->s_resgid = make_kgid(&init_user_ns, EXT3_DEF_RESGID);
1662 sbi->s_sb_block = sb_block;
1664 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1666 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1671 * The ext3 superblock will not be buffer aligned for other than 1kB
1672 * block sizes. We need to calculate the offset from buffer start.
1674 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1675 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1676 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1678 logic_sb_block = sb_block;
1681 if (!(bh = sb_bread(sb, logic_sb_block))) {
1682 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1686 * Note: s_es must be initialized as soon as possible because
1687 * some ext3 macro-instructions depend on its value
1689 es = (struct ext3_super_block *) (bh->b_data + offset);
1691 sb->s_magic = le16_to_cpu(es->s_magic);
1692 if (sb->s_magic != EXT3_SUPER_MAGIC)
1695 /* Set defaults before we parse the mount options */
1696 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1697 if (def_mount_opts & EXT3_DEFM_DEBUG)
1698 set_opt(sbi->s_mount_opt, DEBUG);
1699 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1700 set_opt(sbi->s_mount_opt, GRPID);
1701 if (def_mount_opts & EXT3_DEFM_UID16)
1702 set_opt(sbi->s_mount_opt, NO_UID32);
1703 #ifdef CONFIG_EXT3_FS_XATTR
1704 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1705 set_opt(sbi->s_mount_opt, XATTR_USER);
1707 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1708 if (def_mount_opts & EXT3_DEFM_ACL)
1709 set_opt(sbi->s_mount_opt, POSIX_ACL);
1711 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1712 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1713 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1714 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1715 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1716 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1718 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1719 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1720 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1721 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1723 set_opt(sbi->s_mount_opt, ERRORS_RO);
1725 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
1726 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
1728 /* enable barriers by default */
1729 set_opt(sbi->s_mount_opt, BARRIER);
1730 set_opt(sbi->s_mount_opt, RESERVATION);
1732 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1736 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1737 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1739 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1740 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1741 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1742 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1743 ext3_msg(sb, KERN_WARNING,
1744 "warning: feature flags set on rev 0 fs, "
1745 "running e2fsck is recommended");
1747 * Check feature flags regardless of the revision level, since we
1748 * previously didn't change the revision level when setting the flags,
1749 * so there is a chance incompat flags are set on a rev 0 filesystem.
1751 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1753 ext3_msg(sb, KERN_ERR,
1754 "error: couldn't mount because of unsupported "
1755 "optional features (%x)", le32_to_cpu(features));
1758 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1759 if (!(sb->s_flags & MS_RDONLY) && features) {
1760 ext3_msg(sb, KERN_ERR,
1761 "error: couldn't mount RDWR because of unsupported "
1762 "optional features (%x)", le32_to_cpu(features));
1765 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1767 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1768 blocksize > EXT3_MAX_BLOCK_SIZE) {
1769 ext3_msg(sb, KERN_ERR,
1770 "error: couldn't mount because of unsupported "
1771 "filesystem blocksize %d", blocksize);
1775 hblock = bdev_logical_block_size(sb->s_bdev);
1776 if (sb->s_blocksize != blocksize) {
1778 * Make sure the blocksize for the filesystem is larger
1779 * than the hardware sectorsize for the machine.
1781 if (blocksize < hblock) {
1782 ext3_msg(sb, KERN_ERR,
1783 "error: fsblocksize %d too small for "
1784 "hardware sectorsize %d", blocksize, hblock);
1789 if (!sb_set_blocksize(sb, blocksize)) {
1790 ext3_msg(sb, KERN_ERR,
1791 "error: bad blocksize %d", blocksize);
1794 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1795 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1796 bh = sb_bread(sb, logic_sb_block);
1798 ext3_msg(sb, KERN_ERR,
1799 "error: can't read superblock on 2nd try");
1802 es = (struct ext3_super_block *)(bh->b_data + offset);
1804 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1805 ext3_msg(sb, KERN_ERR,
1806 "error: magic mismatch");
1811 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1813 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1814 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1815 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1817 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1818 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1819 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1820 (!is_power_of_2(sbi->s_inode_size)) ||
1821 (sbi->s_inode_size > blocksize)) {
1822 ext3_msg(sb, KERN_ERR,
1823 "error: unsupported inode size: %d",
1828 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1829 le32_to_cpu(es->s_log_frag_size);
1830 if (blocksize != sbi->s_frag_size) {
1831 ext3_msg(sb, KERN_ERR,
1832 "error: fragsize %lu != blocksize %u (unsupported)",
1833 sbi->s_frag_size, blocksize);
1836 sbi->s_frags_per_block = 1;
1837 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1838 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1839 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1840 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1842 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1843 if (sbi->s_inodes_per_block == 0)
1845 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1846 sbi->s_inodes_per_block;
1847 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1849 sbi->s_mount_state = le16_to_cpu(es->s_state);
1850 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1851 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1852 for (i=0; i < 4; i++)
1853 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1854 sbi->s_def_hash_version = es->s_def_hash_version;
1855 i = le32_to_cpu(es->s_flags);
1856 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1857 sbi->s_hash_unsigned = 3;
1858 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1859 #ifdef __CHAR_UNSIGNED__
1860 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1861 sbi->s_hash_unsigned = 3;
1863 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1867 if (sbi->s_blocks_per_group > blocksize * 8) {
1868 ext3_msg(sb, KERN_ERR,
1869 "#blocks per group too big: %lu",
1870 sbi->s_blocks_per_group);
1873 if (sbi->s_frags_per_group > blocksize * 8) {
1874 ext3_msg(sb, KERN_ERR,
1875 "error: #fragments per group too big: %lu",
1876 sbi->s_frags_per_group);
1879 if (sbi->s_inodes_per_group > blocksize * 8) {
1880 ext3_msg(sb, KERN_ERR,
1881 "error: #inodes per group too big: %lu",
1882 sbi->s_inodes_per_group);
1886 err = generic_check_addressable(sb->s_blocksize_bits,
1887 le32_to_cpu(es->s_blocks_count));
1889 ext3_msg(sb, KERN_ERR,
1890 "error: filesystem is too large to mount safely");
1891 if (sizeof(sector_t) < 8)
1892 ext3_msg(sb, KERN_ERR,
1893 "error: CONFIG_LBDAF not enabled");
1898 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1900 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1901 le32_to_cpu(es->s_first_data_block) - 1)
1902 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1903 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
1904 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1906 if (sbi->s_group_desc == NULL) {
1907 ext3_msg(sb, KERN_ERR,
1908 "error: not enough memory");
1913 bgl_lock_init(sbi->s_blockgroup_lock);
1915 for (i = 0; i < db_count; i++) {
1916 block = descriptor_loc(sb, logic_sb_block, i);
1917 sbi->s_group_desc[i] = sb_bread(sb, block);
1918 if (!sbi->s_group_desc[i]) {
1919 ext3_msg(sb, KERN_ERR,
1920 "error: can't read group descriptor %d", i);
1925 if (!ext3_check_descriptors (sb)) {
1926 ext3_msg(sb, KERN_ERR,
1927 "error: group descriptors corrupted");
1930 sbi->s_gdb_count = db_count;
1931 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1932 spin_lock_init(&sbi->s_next_gen_lock);
1934 /* per fileystem reservation list head & lock */
1935 spin_lock_init(&sbi->s_rsv_window_lock);
1936 sbi->s_rsv_window_root = RB_ROOT;
1937 /* Add a single, static dummy reservation to the start of the
1938 * reservation window list --- it gives us a placeholder for
1939 * append-at-start-of-list which makes the allocation logic
1940 * _much_ simpler. */
1941 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1942 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1943 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1944 sbi->s_rsv_window_head.rsv_goal_size = 0;
1945 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1948 * set up enough so that it can read an inode
1950 sb->s_op = &ext3_sops;
1951 sb->s_export_op = &ext3_export_ops;
1952 sb->s_xattr = ext3_xattr_handlers;
1954 sb->s_qcop = &ext3_qctl_operations;
1955 sb->dq_op = &ext3_quota_operations;
1957 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
1958 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1959 mutex_init(&sbi->s_orphan_lock);
1960 mutex_init(&sbi->s_resize_lock);
1964 needs_recovery = (es->s_last_orphan != 0 ||
1965 EXT3_HAS_INCOMPAT_FEATURE(sb,
1966 EXT3_FEATURE_INCOMPAT_RECOVER));
1969 * The first inode we look at is the journal inode. Don't try
1970 * root first: it may be modified in the journal!
1972 if (!test_opt(sb, NOLOAD) &&
1973 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1974 if (ext3_load_journal(sb, es, journal_devnum))
1976 } else if (journal_inum) {
1977 if (ext3_create_journal(sb, es, journal_inum))
1981 ext3_msg(sb, KERN_ERR,
1982 "error: no journal found. "
1983 "mounting ext3 over ext2?");
1986 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1987 ext3_count_free_blocks(sb));
1989 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1990 ext3_count_free_inodes(sb));
1993 err = percpu_counter_init(&sbi->s_dirs_counter,
1994 ext3_count_dirs(sb));
1997 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
2002 /* We have now updated the journal if required, so we can
2003 * validate the data journaling mode. */
2004 switch (test_opt(sb, DATA_FLAGS)) {
2006 /* No mode set, assume a default based on the journal
2007 capabilities: ORDERED_DATA if the journal can
2008 cope, else JOURNAL_DATA */
2009 if (journal_check_available_features
2010 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
2011 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
2013 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2016 case EXT3_MOUNT_ORDERED_DATA:
2017 case EXT3_MOUNT_WRITEBACK_DATA:
2018 if (!journal_check_available_features
2019 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
2020 ext3_msg(sb, KERN_ERR,
2021 "error: journal does not support "
2022 "requested data journaling mode");
2030 * The journal_load will have done any necessary log recovery,
2031 * so we can safely mount the rest of the filesystem now.
2034 root = ext3_iget(sb, EXT3_ROOT_INO);
2036 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
2037 ret = PTR_ERR(root);
2040 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2042 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
2045 sb->s_root = d_make_root(root);
2047 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2052 if (ext3_setup_super(sb, es, sb->s_flags & MS_RDONLY))
2053 sb->s_flags |= MS_RDONLY;
2055 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2056 ext3_orphan_cleanup(sb, es);
2057 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2058 if (needs_recovery) {
2059 ext3_mark_recovery_complete(sb, es);
2060 ext3_msg(sb, KERN_INFO, "recovery complete");
2062 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2063 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2064 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2071 ext3_msg(sb, KERN_INFO,
2072 "error: can't find ext3 filesystem on dev %s.",
2077 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2078 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2079 percpu_counter_destroy(&sbi->s_dirs_counter);
2080 journal_destroy(sbi->s_journal);
2082 for (i = 0; i < db_count; i++)
2083 brelse(sbi->s_group_desc[i]);
2084 kfree(sbi->s_group_desc);
2087 for (i = 0; i < MAXQUOTAS; i++)
2088 kfree(sbi->s_qf_names[i]);
2090 ext3_blkdev_remove(sbi);
2093 sb->s_fs_info = NULL;
2094 kfree(sbi->s_blockgroup_lock);
2100 * Setup any per-fs journal parameters now. We'll do this both on
2101 * initial mount, once the journal has been initialised but before we've
2102 * done any recovery; and again on any subsequent remount.
2104 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2106 struct ext3_sb_info *sbi = EXT3_SB(sb);
2108 if (sbi->s_commit_interval)
2109 journal->j_commit_interval = sbi->s_commit_interval;
2110 /* We could also set up an ext3-specific default for the commit
2111 * interval here, but for now we'll just fall back to the jbd
2114 spin_lock(&journal->j_state_lock);
2115 if (test_opt(sb, BARRIER))
2116 journal->j_flags |= JFS_BARRIER;
2118 journal->j_flags &= ~JFS_BARRIER;
2119 if (test_opt(sb, DATA_ERR_ABORT))
2120 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2122 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2123 spin_unlock(&journal->j_state_lock);
2126 static journal_t *ext3_get_journal(struct super_block *sb,
2127 unsigned int journal_inum)
2129 struct inode *journal_inode;
2132 /* First, test for the existence of a valid inode on disk. Bad
2133 * things happen if we iget() an unused inode, as the subsequent
2134 * iput() will try to delete it. */
2136 journal_inode = ext3_iget(sb, journal_inum);
2137 if (IS_ERR(journal_inode)) {
2138 ext3_msg(sb, KERN_ERR, "error: no journal found");
2141 if (!journal_inode->i_nlink) {
2142 make_bad_inode(journal_inode);
2143 iput(journal_inode);
2144 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2148 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2149 journal_inode, journal_inode->i_size);
2150 if (!S_ISREG(journal_inode->i_mode)) {
2151 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2152 iput(journal_inode);
2156 journal = journal_init_inode(journal_inode);
2158 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2159 iput(journal_inode);
2162 journal->j_private = sb;
2163 ext3_init_journal_params(sb, journal);
2167 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2170 struct buffer_head * bh;
2174 int hblock, blocksize;
2175 ext3_fsblk_t sb_block;
2176 unsigned long offset;
2177 struct ext3_super_block * es;
2178 struct block_device *bdev;
2180 bdev = ext3_blkdev_get(j_dev, sb);
2184 blocksize = sb->s_blocksize;
2185 hblock = bdev_logical_block_size(bdev);
2186 if (blocksize < hblock) {
2187 ext3_msg(sb, KERN_ERR,
2188 "error: blocksize too small for journal device");
2192 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2193 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2194 set_blocksize(bdev, blocksize);
2195 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2196 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2197 "external journal");
2201 es = (struct ext3_super_block *) (bh->b_data + offset);
2202 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2203 !(le32_to_cpu(es->s_feature_incompat) &
2204 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2205 ext3_msg(sb, KERN_ERR, "error: external journal has "
2211 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2212 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2217 len = le32_to_cpu(es->s_blocks_count);
2218 start = sb_block + 1;
2219 brelse(bh); /* we're done with the superblock */
2221 journal = journal_init_dev(bdev, sb->s_bdev,
2222 start, len, blocksize);
2224 ext3_msg(sb, KERN_ERR,
2225 "error: failed to create device journal");
2228 journal->j_private = sb;
2229 if (!bh_uptodate_or_lock(journal->j_sb_buffer)) {
2230 if (bh_submit_read(journal->j_sb_buffer)) {
2231 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2235 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2236 ext3_msg(sb, KERN_ERR,
2237 "error: external journal has more than one "
2238 "user (unsupported) - %d",
2239 be32_to_cpu(journal->j_superblock->s_nr_users));
2242 EXT3_SB(sb)->journal_bdev = bdev;
2243 ext3_init_journal_params(sb, journal);
2246 journal_destroy(journal);
2248 ext3_blkdev_put(bdev);
2252 static int ext3_load_journal(struct super_block *sb,
2253 struct ext3_super_block *es,
2254 unsigned long journal_devnum)
2257 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2260 int really_read_only;
2262 if (journal_devnum &&
2263 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2264 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2265 "numbers have changed");
2266 journal_dev = new_decode_dev(journal_devnum);
2268 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2270 really_read_only = bdev_read_only(sb->s_bdev);
2273 * Are we loading a blank journal or performing recovery after a
2274 * crash? For recovery, we need to check in advance whether we
2275 * can get read-write access to the device.
2278 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2279 if (sb->s_flags & MS_RDONLY) {
2280 ext3_msg(sb, KERN_INFO,
2281 "recovery required on readonly filesystem");
2282 if (really_read_only) {
2283 ext3_msg(sb, KERN_ERR, "error: write access "
2284 "unavailable, cannot proceed");
2287 ext3_msg(sb, KERN_INFO,
2288 "write access will be enabled during recovery");
2292 if (journal_inum && journal_dev) {
2293 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2294 "and inode journals");
2299 if (!(journal = ext3_get_journal(sb, journal_inum)))
2302 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2306 if (!(journal->j_flags & JFS_BARRIER))
2307 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2309 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2310 err = journal_update_format(journal);
2312 ext3_msg(sb, KERN_ERR, "error updating journal");
2313 journal_destroy(journal);
2318 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2319 err = journal_wipe(journal, !really_read_only);
2321 err = journal_load(journal);
2324 ext3_msg(sb, KERN_ERR, "error loading journal");
2325 journal_destroy(journal);
2329 EXT3_SB(sb)->s_journal = journal;
2330 ext3_clear_journal_err(sb, es);
2332 if (!really_read_only && journal_devnum &&
2333 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2334 es->s_journal_dev = cpu_to_le32(journal_devnum);
2336 /* Make sure we flush the recovery flag to disk. */
2337 ext3_commit_super(sb, es, 1);
2343 static int ext3_create_journal(struct super_block *sb,
2344 struct ext3_super_block *es,
2345 unsigned int journal_inum)
2350 if (sb->s_flags & MS_RDONLY) {
2351 ext3_msg(sb, KERN_ERR,
2352 "error: readonly filesystem when trying to "
2357 journal = ext3_get_journal(sb, journal_inum);
2361 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2364 err = journal_create(journal);
2366 ext3_msg(sb, KERN_ERR, "error creating journal");
2367 journal_destroy(journal);
2371 EXT3_SB(sb)->s_journal = journal;
2373 ext3_update_dynamic_rev(sb);
2374 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2375 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2377 es->s_journal_inum = cpu_to_le32(journal_inum);
2379 /* Make sure we flush the recovery flag to disk. */
2380 ext3_commit_super(sb, es, 1);
2385 static int ext3_commit_super(struct super_block *sb,
2386 struct ext3_super_block *es,
2389 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2395 if (buffer_write_io_error(sbh)) {
2397 * Oh, dear. A previous attempt to write the
2398 * superblock failed. This could happen because the
2399 * USB device was yanked out. Or it could happen to
2400 * be a transient write error and maybe the block will
2401 * be remapped. Nothing we can do but to retry the
2402 * write and hope for the best.
2404 ext3_msg(sb, KERN_ERR, "previous I/O error to "
2405 "superblock detected");
2406 clear_buffer_write_io_error(sbh);
2407 set_buffer_uptodate(sbh);
2410 * If the file system is mounted read-only, don't update the
2411 * superblock write time. This avoids updating the superblock
2412 * write time when we are mounting the root file system
2413 * read/only but we need to replay the journal; at that point,
2414 * for people who are east of GMT and who make their clock
2415 * tick in localtime for Windows bug-for-bug compatibility,
2416 * the clock is set in the future, and this will cause e2fsck
2417 * to complain and force a full file system check.
2419 if (!(sb->s_flags & MS_RDONLY))
2420 es->s_wtime = cpu_to_le32(get_seconds());
2421 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2422 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2423 BUFFER_TRACE(sbh, "marking dirty");
2424 mark_buffer_dirty(sbh);
2426 error = sync_dirty_buffer(sbh);
2427 if (buffer_write_io_error(sbh)) {
2428 ext3_msg(sb, KERN_ERR, "I/O error while writing "
2430 clear_buffer_write_io_error(sbh);
2431 set_buffer_uptodate(sbh);
2439 * Have we just finished recovery? If so, and if we are mounting (or
2440 * remounting) the filesystem readonly, then we will end up with a
2441 * consistent fs on disk. Record that fact.
2443 static void ext3_mark_recovery_complete(struct super_block * sb,
2444 struct ext3_super_block * es)
2446 journal_t *journal = EXT3_SB(sb)->s_journal;
2448 journal_lock_updates(journal);
2449 if (journal_flush(journal) < 0)
2452 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2453 sb->s_flags & MS_RDONLY) {
2454 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2455 ext3_commit_super(sb, es, 1);
2459 journal_unlock_updates(journal);
2463 * If we are mounting (or read-write remounting) a filesystem whose journal
2464 * has recorded an error from a previous lifetime, move that error to the
2465 * main filesystem now.
2467 static void ext3_clear_journal_err(struct super_block *sb,
2468 struct ext3_super_block *es)
2474 journal = EXT3_SB(sb)->s_journal;
2477 * Now check for any error status which may have been recorded in the
2478 * journal by a prior ext3_error() or ext3_abort()
2481 j_errno = journal_errno(journal);
2485 errstr = ext3_decode_error(sb, j_errno, nbuf);
2486 ext3_warning(sb, __func__, "Filesystem error recorded "
2487 "from previous mount: %s", errstr);
2488 ext3_warning(sb, __func__, "Marking fs in need of "
2489 "filesystem check.");
2491 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2492 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2493 ext3_commit_super (sb, es, 1);
2495 journal_clear_err(journal);
2500 * Force the running and committing transactions to commit,
2501 * and wait on the commit.
2503 int ext3_force_commit(struct super_block *sb)
2508 if (sb->s_flags & MS_RDONLY)
2511 journal = EXT3_SB(sb)->s_journal;
2512 ret = ext3_journal_force_commit(journal);
2516 static int ext3_sync_fs(struct super_block *sb, int wait)
2520 trace_ext3_sync_fs(sb, wait);
2522 * Writeback quota in non-journalled quota case - journalled quota has
2525 dquot_writeback_dquots(sb, -1);
2526 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2528 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2534 * LVM calls this function before a (read-only) snapshot is created. This
2535 * gives us a chance to flush the journal completely and mark the fs clean.
2537 static int ext3_freeze(struct super_block *sb)
2542 if (!(sb->s_flags & MS_RDONLY)) {
2543 journal = EXT3_SB(sb)->s_journal;
2545 /* Now we set up the journal barrier. */
2546 journal_lock_updates(journal);
2549 * We don't want to clear needs_recovery flag when we failed
2550 * to flush the journal.
2552 error = journal_flush(journal);
2556 /* Journal blocked and flushed, clear needs_recovery flag. */
2557 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2558 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2565 journal_unlock_updates(journal);
2570 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2571 * flag here, even though the filesystem is not technically dirty yet.
2573 static int ext3_unfreeze(struct super_block *sb)
2575 if (!(sb->s_flags & MS_RDONLY)) {
2577 /* Reser the needs_recovery flag before the fs is unlocked. */
2578 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2579 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2581 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2586 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2588 struct ext3_super_block * es;
2589 struct ext3_sb_info *sbi = EXT3_SB(sb);
2590 ext3_fsblk_t n_blocks_count = 0;
2591 unsigned long old_sb_flags;
2592 struct ext3_mount_options old_opts;
2593 int enable_quota = 0;
2599 /* Store the original options */
2601 old_sb_flags = sb->s_flags;
2602 old_opts.s_mount_opt = sbi->s_mount_opt;
2603 old_opts.s_resuid = sbi->s_resuid;
2604 old_opts.s_resgid = sbi->s_resgid;
2605 old_opts.s_commit_interval = sbi->s_commit_interval;
2607 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2608 for (i = 0; i < MAXQUOTAS; i++)
2609 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2613 * Allow the "check" option to be passed as a remount option.
2615 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2620 if (test_opt(sb, ABORT))
2621 ext3_abort(sb, __func__, "Abort forced by user");
2623 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2624 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2628 ext3_init_journal_params(sb, sbi->s_journal);
2630 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2631 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2632 if (test_opt(sb, ABORT)) {
2637 if (*flags & MS_RDONLY) {
2638 err = dquot_suspend(sb, -1);
2643 * First of all, the unconditional stuff we have to do
2644 * to disable replay of the journal when we next remount
2646 sb->s_flags |= MS_RDONLY;
2649 * OK, test if we are remounting a valid rw partition
2650 * readonly, and if so set the rdonly flag and then
2651 * mark the partition as valid again.
2653 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2654 (sbi->s_mount_state & EXT3_VALID_FS))
2655 es->s_state = cpu_to_le16(sbi->s_mount_state);
2657 ext3_mark_recovery_complete(sb, es);
2660 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2661 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2662 ext3_msg(sb, KERN_WARNING,
2663 "warning: couldn't remount RDWR "
2664 "because of unsupported optional "
2665 "features (%x)", le32_to_cpu(ret));
2671 * If we have an unprocessed orphan list hanging
2672 * around from a previously readonly bdev mount,
2673 * require a full umount & mount for now.
2675 if (es->s_last_orphan) {
2676 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2677 "remount RDWR because of unprocessed "
2678 "orphan inode list. Please "
2679 "umount & mount instead.");
2685 * Mounting a RDONLY partition read-write, so reread
2686 * and store the current valid flag. (It may have
2687 * been changed by e2fsck since we originally mounted
2690 ext3_clear_journal_err(sb, es);
2691 sbi->s_mount_state = le16_to_cpu(es->s_state);
2692 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2694 if (!ext3_setup_super (sb, es, 0))
2695 sb->s_flags &= ~MS_RDONLY;
2700 /* Release old quota file names */
2701 for (i = 0; i < MAXQUOTAS; i++)
2702 if (old_opts.s_qf_names[i] &&
2703 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2704 kfree(old_opts.s_qf_names[i]);
2709 dquot_resume(sb, -1);
2712 sb->s_flags = old_sb_flags;
2713 sbi->s_mount_opt = old_opts.s_mount_opt;
2714 sbi->s_resuid = old_opts.s_resuid;
2715 sbi->s_resgid = old_opts.s_resgid;
2716 sbi->s_commit_interval = old_opts.s_commit_interval;
2718 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2719 for (i = 0; i < MAXQUOTAS; i++) {
2720 if (sbi->s_qf_names[i] &&
2721 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2722 kfree(sbi->s_qf_names[i]);
2723 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2730 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2732 struct super_block *sb = dentry->d_sb;
2733 struct ext3_sb_info *sbi = EXT3_SB(sb);
2734 struct ext3_super_block *es = sbi->s_es;
2737 if (test_opt(sb, MINIX_DF)) {
2738 sbi->s_overhead_last = 0;
2739 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2740 unsigned long ngroups = sbi->s_groups_count, i;
2741 ext3_fsblk_t overhead = 0;
2745 * Compute the overhead (FS structures). This is constant
2746 * for a given filesystem unless the number of block groups
2747 * changes so we cache the previous value until it does.
2751 * All of the blocks before first_data_block are
2754 overhead = le32_to_cpu(es->s_first_data_block);
2757 * Add the overhead attributed to the superblock and
2758 * block group descriptors. If the sparse superblocks
2759 * feature is turned on, then not all groups have this.
2761 for (i = 0; i < ngroups; i++) {
2762 overhead += ext3_bg_has_super(sb, i) +
2763 ext3_bg_num_gdb(sb, i);
2768 * Every block group has an inode bitmap, a block
2769 * bitmap, and an inode table.
2771 overhead += ngroups * (2 + sbi->s_itb_per_group);
2772 sbi->s_overhead_last = overhead;
2774 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2777 buf->f_type = EXT3_SUPER_MAGIC;
2778 buf->f_bsize = sb->s_blocksize;
2779 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2780 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2781 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2782 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2784 buf->f_files = le32_to_cpu(es->s_inodes_count);
2785 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2786 buf->f_namelen = EXT3_NAME_LEN;
2787 fsid = le64_to_cpup((void *)es->s_uuid) ^
2788 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2789 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2790 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2794 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2795 * is locked for write. Otherwise the are possible deadlocks:
2796 * Process 1 Process 2
2797 * ext3_create() quota_sync()
2798 * journal_start() write_dquot()
2799 * dquot_initialize() down(dqio_mutex)
2800 * down(dqio_mutex) journal_start()
2806 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2808 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2811 static int ext3_write_dquot(struct dquot *dquot)
2815 struct inode *inode;
2817 inode = dquot_to_inode(dquot);
2818 handle = ext3_journal_start(inode,
2819 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2821 return PTR_ERR(handle);
2822 ret = dquot_commit(dquot);
2823 err = ext3_journal_stop(handle);
2829 static int ext3_acquire_dquot(struct dquot *dquot)
2834 handle = ext3_journal_start(dquot_to_inode(dquot),
2835 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2837 return PTR_ERR(handle);
2838 ret = dquot_acquire(dquot);
2839 err = ext3_journal_stop(handle);
2845 static int ext3_release_dquot(struct dquot *dquot)
2850 handle = ext3_journal_start(dquot_to_inode(dquot),
2851 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2852 if (IS_ERR(handle)) {
2853 /* Release dquot anyway to avoid endless cycle in dqput() */
2854 dquot_release(dquot);
2855 return PTR_ERR(handle);
2857 ret = dquot_release(dquot);
2858 err = ext3_journal_stop(handle);
2864 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2866 /* Are we journaling quotas? */
2867 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2868 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2869 dquot_mark_dquot_dirty(dquot);
2870 return ext3_write_dquot(dquot);
2872 return dquot_mark_dquot_dirty(dquot);
2876 static int ext3_write_info(struct super_block *sb, int type)
2881 /* Data block + inode block */
2882 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2884 return PTR_ERR(handle);
2885 ret = dquot_commit_info(sb, type);
2886 err = ext3_journal_stop(handle);
2893 * Turn on quotas during mount time - we need to find
2894 * the quota file and such...
2896 static int ext3_quota_on_mount(struct super_block *sb, int type)
2898 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2899 EXT3_SB(sb)->s_jquota_fmt, type);
2903 * Standard function to be called on quota_on
2905 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2910 if (!test_opt(sb, QUOTA))
2913 /* Quotafile not on the same filesystem? */
2914 if (path->dentry->d_sb != sb)
2916 /* Journaling quota? */
2917 if (EXT3_SB(sb)->s_qf_names[type]) {
2918 /* Quotafile not of fs root? */
2919 if (path->dentry->d_parent != sb->s_root)
2920 ext3_msg(sb, KERN_WARNING,
2921 "warning: Quota file not on filesystem root. "
2922 "Journaled quota will not work.");
2926 * When we journal data on quota file, we have to flush journal to see
2927 * all updates to the file when we bypass pagecache...
2929 if (ext3_should_journal_data(path->dentry->d_inode)) {
2931 * We don't need to lock updates but journal_flush() could
2932 * otherwise be livelocked...
2934 journal_lock_updates(EXT3_SB(sb)->s_journal);
2935 err = journal_flush(EXT3_SB(sb)->s_journal);
2936 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2941 return dquot_quota_on(sb, type, format_id, path);
2944 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2945 * acquiring the locks... As quota files are never truncated and quota code
2946 * itself serializes the operations (and no one else should touch the files)
2947 * we don't have to be afraid of races */
2948 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2949 size_t len, loff_t off)
2951 struct inode *inode = sb_dqopt(sb)->files[type];
2952 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2954 int offset = off & (sb->s_blocksize - 1);
2957 struct buffer_head *bh;
2958 loff_t i_size = i_size_read(inode);
2962 if (off+len > i_size)
2965 while (toread > 0) {
2966 tocopy = sb->s_blocksize - offset < toread ?
2967 sb->s_blocksize - offset : toread;
2968 bh = ext3_bread(NULL, inode, blk, 0, &err);
2971 if (!bh) /* A hole? */
2972 memset(data, 0, tocopy);
2974 memcpy(data, bh->b_data+offset, tocopy);
2984 /* Write to quotafile (we know the transaction is already started and has
2985 * enough credits) */
2986 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2987 const char *data, size_t len, loff_t off)
2989 struct inode *inode = sb_dqopt(sb)->files[type];
2990 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2992 int offset = off & (sb->s_blocksize - 1);
2993 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2994 struct buffer_head *bh;
2995 handle_t *handle = journal_current_handle();
2998 ext3_msg(sb, KERN_WARNING,
2999 "warning: quota write (off=%llu, len=%llu)"
3000 " cancelled because transaction is not started.",
3001 (unsigned long long)off, (unsigned long long)len);
3006 * Since we account only one data block in transaction credits,
3007 * then it is impossible to cross a block boundary.
3009 if (sb->s_blocksize - offset < len) {
3010 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3011 " cancelled because not block aligned",
3012 (unsigned long long)off, (unsigned long long)len);
3015 bh = ext3_bread(handle, inode, blk, 1, &err);
3018 if (journal_quota) {
3019 err = ext3_journal_get_write_access(handle, bh);
3026 memcpy(bh->b_data+offset, data, len);
3027 flush_dcache_page(bh->b_page);
3030 err = ext3_journal_dirty_metadata(handle, bh);
3032 /* Always do at least ordered writes for quotas */
3033 err = ext3_journal_dirty_data(handle, bh);
3034 mark_buffer_dirty(bh);
3040 if (inode->i_size < off + len) {
3041 i_size_write(inode, off + len);
3042 EXT3_I(inode)->i_disksize = inode->i_size;
3045 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3046 ext3_mark_inode_dirty(handle, inode);
3052 static struct dentry *ext3_mount(struct file_system_type *fs_type,
3053 int flags, const char *dev_name, void *data)
3055 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
3058 static struct file_system_type ext3_fs_type = {
3059 .owner = THIS_MODULE,
3061 .mount = ext3_mount,
3062 .kill_sb = kill_block_super,
3063 .fs_flags = FS_REQUIRES_DEV,
3066 static int __init init_ext3_fs(void)
3068 int err = init_ext3_xattr();
3071 err = init_inodecache();
3074 err = register_filesystem(&ext3_fs_type);
3079 destroy_inodecache();
3085 static void __exit exit_ext3_fs(void)
3087 unregister_filesystem(&ext3_fs_type);
3088 destroy_inodecache();
3092 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3093 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3094 MODULE_LICENSE("GPL");
3095 module_init(init_ext3_fs)
3096 module_exit(exit_ext3_fs)