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>
30 #include <linux/namei.h>
32 #include <asm/uaccess.h>
34 #define CREATE_TRACE_POINTS
41 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
42 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
44 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
47 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
49 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
51 static int ext3_commit_super(struct super_block *sb,
52 struct ext3_super_block *es,
54 static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56 static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58 static int ext3_sync_fs(struct super_block *sb, int wait);
59 static const char *ext3_decode_error(struct super_block * sb, int errno,
61 static int ext3_remount (struct super_block * sb, int * flags, char * data);
62 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
63 static int ext3_unfreeze(struct super_block *sb);
64 static int ext3_freeze(struct super_block *sb);
67 * Wrappers for journal_start/end.
69 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
73 if (sb->s_flags & MS_RDONLY)
74 return ERR_PTR(-EROFS);
76 /* Special case here: if the journal has aborted behind our
77 * backs (eg. EIO in the commit thread), then we still need to
78 * take the FS itself readonly cleanly. */
79 journal = EXT3_SB(sb)->s_journal;
80 if (is_journal_aborted(journal)) {
81 ext3_abort(sb, __func__,
82 "Detected aborted journal");
83 return ERR_PTR(-EROFS);
86 return journal_start(journal, nblocks);
89 int __ext3_journal_stop(const char *where, handle_t *handle)
91 struct super_block *sb;
95 sb = handle->h_transaction->t_journal->j_private;
97 rc = journal_stop(handle);
102 __ext3_std_error(sb, where, err);
106 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
107 struct buffer_head *bh, handle_t *handle, int err)
110 const char *errstr = ext3_decode_error(NULL, err, nbuf);
113 BUFFER_TRACE(bh, "abort");
118 if (is_handle_aborted(handle))
121 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
122 caller, errstr, err_fn);
124 journal_abort_handle(handle);
127 void ext3_msg(struct super_block *sb, const char *prefix,
128 const char *fmt, ...)
130 struct va_format vaf;
138 printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
143 /* Deal with the reporting of failure conditions on a filesystem such as
144 * inconsistencies detected or read IO failures.
146 * On ext2, we can store the error state of the filesystem in the
147 * superblock. That is not possible on ext3, because we may have other
148 * write ordering constraints on the superblock which prevent us from
149 * writing it out straight away; and given that the journal is about to
150 * be aborted, we can't rely on the current, or future, transactions to
151 * write out the superblock safely.
153 * We'll just use the journal_abort() error code to record an error in
154 * the journal instead. On recovery, the journal will complain about
155 * that error until we've noted it down and cleared it.
158 static void ext3_handle_error(struct super_block *sb)
160 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
162 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
163 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
165 if (sb->s_flags & MS_RDONLY)
168 if (!test_opt (sb, ERRORS_CONT)) {
169 journal_t *journal = EXT3_SB(sb)->s_journal;
171 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
173 journal_abort(journal, -EIO);
175 if (test_opt (sb, ERRORS_RO)) {
176 ext3_msg(sb, KERN_CRIT,
177 "error: remounting filesystem read-only");
179 * Make sure updated value of ->s_mount_state will be visible
180 * before ->s_flags update.
183 sb->s_flags |= MS_RDONLY;
185 ext3_commit_super(sb, es, 1);
186 if (test_opt(sb, ERRORS_PANIC))
187 panic("EXT3-fs (%s): panic forced after error\n",
191 void ext3_error(struct super_block *sb, const char *function,
192 const char *fmt, ...)
194 struct va_format vaf;
202 printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
203 sb->s_id, function, &vaf);
207 ext3_handle_error(sb);
210 static const char *ext3_decode_error(struct super_block * sb, int errno,
217 errstr = "IO failure";
220 errstr = "Out of memory";
223 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
224 errstr = "Journal has aborted";
226 errstr = "Readonly filesystem";
229 /* If the caller passed in an extra buffer for unknown
230 * errors, textualise them now. Else we just return
233 /* Check for truncated error codes... */
234 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
243 /* __ext3_std_error decodes expected errors from journaling functions
244 * automatically and invokes the appropriate error response. */
246 void __ext3_std_error (struct super_block * sb, const char * function,
252 /* Special case: if the error is EROFS, and we're not already
253 * inside a transaction, then there's really no point in logging
255 if (errno == -EROFS && journal_current_handle() == NULL &&
256 (sb->s_flags & MS_RDONLY))
259 errstr = ext3_decode_error(sb, errno, nbuf);
260 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
262 ext3_handle_error(sb);
266 * ext3_abort is a much stronger failure handler than ext3_error. The
267 * abort function may be used to deal with unrecoverable failures such
268 * as journal IO errors or ENOMEM at a critical moment in log management.
270 * We unconditionally force the filesystem into an ABORT|READONLY state,
271 * unless the error response on the fs has been set to panic in which
272 * case we take the easy way out and panic immediately.
275 void ext3_abort(struct super_block *sb, const char *function,
276 const char *fmt, ...)
278 struct va_format vaf;
286 printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
287 sb->s_id, function, &vaf);
291 if (test_opt(sb, ERRORS_PANIC))
292 panic("EXT3-fs: panic from previous error\n");
294 if (sb->s_flags & MS_RDONLY)
297 ext3_msg(sb, KERN_CRIT,
298 "error: remounting filesystem read-only");
299 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
300 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
302 * Make sure updated value of ->s_mount_state will be visible
303 * before ->s_flags update.
306 sb->s_flags |= MS_RDONLY;
308 if (EXT3_SB(sb)->s_journal)
309 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
312 void ext3_warning(struct super_block *sb, const char *function,
313 const char *fmt, ...)
315 struct va_format vaf;
323 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
324 sb->s_id, function, &vaf);
329 void ext3_update_dynamic_rev(struct super_block *sb)
331 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
333 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
336 ext3_msg(sb, KERN_WARNING,
337 "warning: updating to rev %d because of "
338 "new feature flag, running e2fsck is recommended",
341 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
342 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
343 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
344 /* leave es->s_feature_*compat flags alone */
345 /* es->s_uuid will be set by e2fsck if empty */
348 * The rest of the superblock fields should be zero, and if not it
349 * means they are likely already in use, so leave them alone. We
350 * can leave it up to e2fsck to clean up any inconsistencies there.
355 * Open the external journal device
357 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
359 struct block_device *bdev;
360 char b[BDEVNAME_SIZE];
362 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
368 ext3_msg(sb, KERN_ERR, "error: failed to open journal device %s: %ld",
369 __bdevname(dev, b), PTR_ERR(bdev));
375 * Release the journal device
377 static void ext3_blkdev_put(struct block_device *bdev)
379 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
382 static void ext3_blkdev_remove(struct ext3_sb_info *sbi)
384 struct block_device *bdev;
385 bdev = sbi->journal_bdev;
387 ext3_blkdev_put(bdev);
388 sbi->journal_bdev = NULL;
392 static inline struct inode *orphan_list_entry(struct list_head *l)
394 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
397 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
401 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
402 le32_to_cpu(sbi->s_es->s_last_orphan));
404 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
405 list_for_each(l, &sbi->s_orphan) {
406 struct inode *inode = orphan_list_entry(l);
407 ext3_msg(sb, KERN_ERR, " "
408 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
409 inode->i_sb->s_id, inode->i_ino, inode,
410 inode->i_mode, inode->i_nlink,
415 static void ext3_put_super (struct super_block * sb)
417 struct ext3_sb_info *sbi = EXT3_SB(sb);
418 struct ext3_super_block *es = sbi->s_es;
421 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
422 ext3_xattr_put_super(sb);
423 err = journal_destroy(sbi->s_journal);
424 sbi->s_journal = NULL;
426 ext3_abort(sb, __func__, "Couldn't clean up the journal");
428 if (!(sb->s_flags & MS_RDONLY)) {
429 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
430 es->s_state = cpu_to_le16(sbi->s_mount_state);
431 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
432 mark_buffer_dirty(sbi->s_sbh);
433 ext3_commit_super(sb, es, 1);
436 for (i = 0; i < sbi->s_gdb_count; i++)
437 brelse(sbi->s_group_desc[i]);
438 kfree(sbi->s_group_desc);
439 percpu_counter_destroy(&sbi->s_freeblocks_counter);
440 percpu_counter_destroy(&sbi->s_freeinodes_counter);
441 percpu_counter_destroy(&sbi->s_dirs_counter);
444 for (i = 0; i < EXT3_MAXQUOTAS; i++)
445 kfree(sbi->s_qf_names[i]);
448 /* Debugging code just in case the in-memory inode orphan list
449 * isn't empty. The on-disk one can be non-empty if we've
450 * detected an error and taken the fs readonly, but the
451 * in-memory list had better be clean by this point. */
452 if (!list_empty(&sbi->s_orphan))
453 dump_orphan_list(sb, sbi);
454 J_ASSERT(list_empty(&sbi->s_orphan));
456 invalidate_bdev(sb->s_bdev);
457 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
459 * Invalidate the journal device's buffers. We don't want them
460 * floating about in memory - the physical journal device may
461 * hotswapped, and it breaks the `ro-after' testing code.
463 sync_blockdev(sbi->journal_bdev);
464 invalidate_bdev(sbi->journal_bdev);
465 ext3_blkdev_remove(sbi);
467 sb->s_fs_info = NULL;
468 kfree(sbi->s_blockgroup_lock);
472 static struct kmem_cache *ext3_inode_cachep;
475 * Called inside transaction, so use GFP_NOFS
477 static struct inode *ext3_alloc_inode(struct super_block *sb)
479 struct ext3_inode_info *ei;
481 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
484 ei->i_block_alloc_info = NULL;
485 ei->vfs_inode.i_version = 1;
486 atomic_set(&ei->i_datasync_tid, 0);
487 atomic_set(&ei->i_sync_tid, 0);
488 return &ei->vfs_inode;
491 static int ext3_drop_inode(struct inode *inode)
493 int drop = generic_drop_inode(inode);
495 trace_ext3_drop_inode(inode, drop);
499 static void ext3_i_callback(struct rcu_head *head)
501 struct inode *inode = container_of(head, struct inode, i_rcu);
502 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
505 static void ext3_destroy_inode(struct inode *inode)
507 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
508 printk("EXT3 Inode %p: orphan list check failed!\n",
510 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
511 EXT3_I(inode), sizeof(struct ext3_inode_info),
515 call_rcu(&inode->i_rcu, ext3_i_callback);
518 static void init_once(void *foo)
520 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
522 INIT_LIST_HEAD(&ei->i_orphan);
523 #ifdef CONFIG_EXT3_FS_XATTR
524 init_rwsem(&ei->xattr_sem);
526 mutex_init(&ei->truncate_mutex);
527 inode_init_once(&ei->vfs_inode);
530 static int __init init_inodecache(void)
532 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
533 sizeof(struct ext3_inode_info),
534 0, (SLAB_RECLAIM_ACCOUNT|
537 if (ext3_inode_cachep == NULL)
542 static void destroy_inodecache(void)
545 * Make sure all delayed rcu free inodes are flushed before we
549 kmem_cache_destroy(ext3_inode_cachep);
552 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
554 #if defined(CONFIG_QUOTA)
555 struct ext3_sb_info *sbi = EXT3_SB(sb);
557 if (sbi->s_jquota_fmt) {
560 switch (sbi->s_jquota_fmt) {
571 seq_printf(seq, ",jqfmt=%s", fmtname);
574 if (sbi->s_qf_names[USRQUOTA])
575 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
577 if (sbi->s_qf_names[GRPQUOTA])
578 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
580 if (test_opt(sb, USRQUOTA))
581 seq_puts(seq, ",usrquota");
583 if (test_opt(sb, GRPQUOTA))
584 seq_puts(seq, ",grpquota");
588 static char *data_mode_string(unsigned long mode)
591 case EXT3_MOUNT_JOURNAL_DATA:
593 case EXT3_MOUNT_ORDERED_DATA:
595 case EXT3_MOUNT_WRITEBACK_DATA:
603 * - it's set to a non-default value OR
604 * - if the per-sb default is different from the global default
606 static int ext3_show_options(struct seq_file *seq, struct dentry *root)
608 struct super_block *sb = root->d_sb;
609 struct ext3_sb_info *sbi = EXT3_SB(sb);
610 struct ext3_super_block *es = sbi->s_es;
611 unsigned long def_mount_opts;
613 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
615 if (sbi->s_sb_block != 1)
616 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
617 if (test_opt(sb, MINIX_DF))
618 seq_puts(seq, ",minixdf");
619 if (test_opt(sb, GRPID))
620 seq_puts(seq, ",grpid");
621 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
622 seq_puts(seq, ",nogrpid");
623 if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT3_DEF_RESUID)) ||
624 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
625 seq_printf(seq, ",resuid=%u",
626 from_kuid_munged(&init_user_ns, sbi->s_resuid));
628 if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT3_DEF_RESGID)) ||
629 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
630 seq_printf(seq, ",resgid=%u",
631 from_kgid_munged(&init_user_ns, sbi->s_resgid));
633 if (test_opt(sb, ERRORS_RO)) {
634 int def_errors = le16_to_cpu(es->s_errors);
636 if (def_errors == EXT3_ERRORS_PANIC ||
637 def_errors == EXT3_ERRORS_CONTINUE) {
638 seq_puts(seq, ",errors=remount-ro");
641 if (test_opt(sb, ERRORS_CONT))
642 seq_puts(seq, ",errors=continue");
643 if (test_opt(sb, ERRORS_PANIC))
644 seq_puts(seq, ",errors=panic");
645 if (test_opt(sb, NO_UID32))
646 seq_puts(seq, ",nouid32");
647 if (test_opt(sb, DEBUG))
648 seq_puts(seq, ",debug");
649 #ifdef CONFIG_EXT3_FS_XATTR
650 if (test_opt(sb, XATTR_USER))
651 seq_puts(seq, ",user_xattr");
652 if (!test_opt(sb, XATTR_USER) &&
653 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
654 seq_puts(seq, ",nouser_xattr");
657 #ifdef CONFIG_EXT3_FS_POSIX_ACL
658 if (test_opt(sb, POSIX_ACL))
659 seq_puts(seq, ",acl");
660 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
661 seq_puts(seq, ",noacl");
663 if (!test_opt(sb, RESERVATION))
664 seq_puts(seq, ",noreservation");
665 if (sbi->s_commit_interval) {
666 seq_printf(seq, ",commit=%u",
667 (unsigned) (sbi->s_commit_interval / HZ));
671 * Always display barrier state so it's clear what the status is.
673 seq_puts(seq, ",barrier=");
674 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
675 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
676 if (test_opt(sb, DATA_ERR_ABORT))
677 seq_puts(seq, ",data_err=abort");
679 if (test_opt(sb, NOLOAD))
680 seq_puts(seq, ",norecovery");
682 ext3_show_quota_options(seq, sb);
688 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
689 u64 ino, u32 generation)
693 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
694 return ERR_PTR(-ESTALE);
695 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
696 return ERR_PTR(-ESTALE);
698 /* iget isn't really right if the inode is currently unallocated!!
700 * ext3_read_inode will return a bad_inode if the inode had been
701 * deleted, so we should be safe.
703 * Currently we don't know the generation for parent directory, so
704 * a generation of 0 means "accept any"
706 inode = ext3_iget(sb, ino);
708 return ERR_CAST(inode);
709 if (generation && inode->i_generation != generation) {
711 return ERR_PTR(-ESTALE);
717 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
718 int fh_len, int fh_type)
720 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
724 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
725 int fh_len, int fh_type)
727 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
732 * Try to release metadata pages (indirect blocks, directories) which are
733 * mapped via the block device. Since these pages could have journal heads
734 * which would prevent try_to_free_buffers() from freeing them, we must use
735 * jbd layer's try_to_free_buffers() function to release them.
737 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
740 journal_t *journal = EXT3_SB(sb)->s_journal;
742 WARN_ON(PageChecked(page));
743 if (!page_has_buffers(page))
746 return journal_try_to_free_buffers(journal, page,
748 return try_to_free_buffers(page);
752 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
753 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
755 static int ext3_write_dquot(struct dquot *dquot);
756 static int ext3_acquire_dquot(struct dquot *dquot);
757 static int ext3_release_dquot(struct dquot *dquot);
758 static int ext3_mark_dquot_dirty(struct dquot *dquot);
759 static int ext3_write_info(struct super_block *sb, int type);
760 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
762 static int ext3_quota_on_mount(struct super_block *sb, int type);
763 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
764 size_t len, loff_t off);
765 static ssize_t ext3_quota_write(struct super_block *sb, int type,
766 const char *data, size_t len, loff_t off);
768 static const struct dquot_operations ext3_quota_operations = {
769 .write_dquot = ext3_write_dquot,
770 .acquire_dquot = ext3_acquire_dquot,
771 .release_dquot = ext3_release_dquot,
772 .mark_dirty = ext3_mark_dquot_dirty,
773 .write_info = ext3_write_info,
774 .alloc_dquot = dquot_alloc,
775 .destroy_dquot = dquot_destroy,
778 static const struct quotactl_ops ext3_qctl_operations = {
779 .quota_on = ext3_quota_on,
780 .quota_off = dquot_quota_off,
781 .quota_sync = dquot_quota_sync,
782 .get_info = dquot_get_dqinfo,
783 .set_info = dquot_set_dqinfo,
784 .get_dqblk = dquot_get_dqblk,
785 .set_dqblk = dquot_set_dqblk
789 static const struct super_operations ext3_sops = {
790 .alloc_inode = ext3_alloc_inode,
791 .destroy_inode = ext3_destroy_inode,
792 .write_inode = ext3_write_inode,
793 .dirty_inode = ext3_dirty_inode,
794 .drop_inode = ext3_drop_inode,
795 .evict_inode = ext3_evict_inode,
796 .put_super = ext3_put_super,
797 .sync_fs = ext3_sync_fs,
798 .freeze_fs = ext3_freeze,
799 .unfreeze_fs = ext3_unfreeze,
800 .statfs = ext3_statfs,
801 .remount_fs = ext3_remount,
802 .show_options = ext3_show_options,
804 .quota_read = ext3_quota_read,
805 .quota_write = ext3_quota_write,
807 .bdev_try_to_free_page = bdev_try_to_free_page,
810 static const struct export_operations ext3_export_ops = {
811 .fh_to_dentry = ext3_fh_to_dentry,
812 .fh_to_parent = ext3_fh_to_parent,
813 .get_parent = ext3_get_parent,
817 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
818 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
819 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
820 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
821 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
822 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
824 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
825 Opt_data_err_abort, Opt_data_err_ignore,
826 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
827 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
828 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
829 Opt_resize, Opt_usrquota, Opt_grpquota
832 static const match_table_t tokens = {
833 {Opt_bsd_df, "bsddf"},
834 {Opt_minix_df, "minixdf"},
835 {Opt_grpid, "grpid"},
836 {Opt_grpid, "bsdgroups"},
837 {Opt_nogrpid, "nogrpid"},
838 {Opt_nogrpid, "sysvgroups"},
839 {Opt_resgid, "resgid=%u"},
840 {Opt_resuid, "resuid=%u"},
842 {Opt_err_cont, "errors=continue"},
843 {Opt_err_panic, "errors=panic"},
844 {Opt_err_ro, "errors=remount-ro"},
845 {Opt_nouid32, "nouid32"},
846 {Opt_nocheck, "nocheck"},
847 {Opt_nocheck, "check=none"},
848 {Opt_debug, "debug"},
849 {Opt_oldalloc, "oldalloc"},
850 {Opt_orlov, "orlov"},
851 {Opt_user_xattr, "user_xattr"},
852 {Opt_nouser_xattr, "nouser_xattr"},
854 {Opt_noacl, "noacl"},
855 {Opt_reservation, "reservation"},
856 {Opt_noreservation, "noreservation"},
857 {Opt_noload, "noload"},
858 {Opt_noload, "norecovery"},
861 {Opt_commit, "commit=%u"},
862 {Opt_journal_update, "journal=update"},
863 {Opt_journal_inum, "journal=%u"},
864 {Opt_journal_dev, "journal_dev=%u"},
865 {Opt_journal_path, "journal_path=%s"},
866 {Opt_abort, "abort"},
867 {Opt_data_journal, "data=journal"},
868 {Opt_data_ordered, "data=ordered"},
869 {Opt_data_writeback, "data=writeback"},
870 {Opt_data_err_abort, "data_err=abort"},
871 {Opt_data_err_ignore, "data_err=ignore"},
872 {Opt_offusrjquota, "usrjquota="},
873 {Opt_usrjquota, "usrjquota=%s"},
874 {Opt_offgrpjquota, "grpjquota="},
875 {Opt_grpjquota, "grpjquota=%s"},
876 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
877 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
878 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
879 {Opt_grpquota, "grpquota"},
880 {Opt_noquota, "noquota"},
881 {Opt_quota, "quota"},
882 {Opt_usrquota, "usrquota"},
883 {Opt_barrier, "barrier=%u"},
884 {Opt_barrier, "barrier"},
885 {Opt_nobarrier, "nobarrier"},
886 {Opt_resize, "resize"},
890 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
892 ext3_fsblk_t sb_block;
893 char *options = (char *) *data;
895 if (!options || strncmp(options, "sb=", 3) != 0)
896 return 1; /* Default location */
898 /*todo: use simple_strtoll with >32bit ext3 */
899 sb_block = simple_strtoul(options, &options, 0);
900 if (*options && *options != ',') {
901 ext3_msg(sb, KERN_ERR, "error: invalid sb specification: %s",
907 *data = (void *) options;
912 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
914 struct ext3_sb_info *sbi = EXT3_SB(sb);
917 if (sb_any_quota_loaded(sb) &&
918 !sbi->s_qf_names[qtype]) {
919 ext3_msg(sb, KERN_ERR,
920 "Cannot change journaled "
921 "quota options when quota turned on");
924 qname = match_strdup(args);
926 ext3_msg(sb, KERN_ERR,
927 "Not enough memory for storing quotafile name");
930 if (sbi->s_qf_names[qtype]) {
931 int same = !strcmp(sbi->s_qf_names[qtype], qname);
935 ext3_msg(sb, KERN_ERR,
936 "%s quota file already specified",
941 if (strchr(qname, '/')) {
942 ext3_msg(sb, KERN_ERR,
943 "quotafile must be on filesystem root");
947 sbi->s_qf_names[qtype] = qname;
948 set_opt(sbi->s_mount_opt, QUOTA);
952 static int clear_qf_name(struct super_block *sb, int qtype) {
954 struct ext3_sb_info *sbi = EXT3_SB(sb);
956 if (sb_any_quota_loaded(sb) &&
957 sbi->s_qf_names[qtype]) {
958 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
959 " when quota turned on");
962 if (sbi->s_qf_names[qtype]) {
963 kfree(sbi->s_qf_names[qtype]);
964 sbi->s_qf_names[qtype] = NULL;
970 static int parse_options (char *options, struct super_block *sb,
971 unsigned int *inum, unsigned long *journal_devnum,
972 ext3_fsblk_t *n_blocks_count, int is_remount)
974 struct ext3_sb_info *sbi = EXT3_SB(sb);
976 substring_t args[MAX_OPT_ARGS];
982 struct inode *journal_inode;
993 while ((p = strsep (&options, ",")) != NULL) {
998 * Initialize args struct so we know whether arg was
999 * found; some options take optional arguments.
1001 args[0].to = args[0].from = NULL;
1002 token = match_token(p, tokens, args);
1005 clear_opt (sbi->s_mount_opt, MINIX_DF);
1008 set_opt (sbi->s_mount_opt, MINIX_DF);
1011 set_opt (sbi->s_mount_opt, GRPID);
1014 clear_opt (sbi->s_mount_opt, GRPID);
1017 if (match_int(&args[0], &option))
1019 uid = make_kuid(current_user_ns(), option);
1020 if (!uid_valid(uid)) {
1021 ext3_msg(sb, KERN_ERR, "Invalid uid value %d", option);
1025 sbi->s_resuid = uid;
1028 if (match_int(&args[0], &option))
1030 gid = make_kgid(current_user_ns(), option);
1031 if (!gid_valid(gid)) {
1032 ext3_msg(sb, KERN_ERR, "Invalid gid value %d", option);
1035 sbi->s_resgid = gid;
1038 /* handled by get_sb_block() instead of here */
1039 /* *sb_block = match_int(&args[0]); */
1042 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1043 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1044 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1047 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1048 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1049 set_opt (sbi->s_mount_opt, ERRORS_RO);
1052 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1053 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1054 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1057 set_opt (sbi->s_mount_opt, NO_UID32);
1060 clear_opt (sbi->s_mount_opt, CHECK);
1063 set_opt (sbi->s_mount_opt, DEBUG);
1066 ext3_msg(sb, KERN_WARNING,
1067 "Ignoring deprecated oldalloc option");
1070 ext3_msg(sb, KERN_WARNING,
1071 "Ignoring deprecated orlov option");
1073 #ifdef CONFIG_EXT3_FS_XATTR
1074 case Opt_user_xattr:
1075 set_opt (sbi->s_mount_opt, XATTR_USER);
1077 case Opt_nouser_xattr:
1078 clear_opt (sbi->s_mount_opt, XATTR_USER);
1081 case Opt_user_xattr:
1082 case Opt_nouser_xattr:
1083 ext3_msg(sb, KERN_INFO,
1084 "(no)user_xattr options not supported");
1087 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1089 set_opt(sbi->s_mount_opt, POSIX_ACL);
1092 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1097 ext3_msg(sb, KERN_INFO,
1098 "(no)acl options not supported");
1101 case Opt_reservation:
1102 set_opt(sbi->s_mount_opt, RESERVATION);
1104 case Opt_noreservation:
1105 clear_opt(sbi->s_mount_opt, RESERVATION);
1107 case Opt_journal_update:
1109 /* Eventually we will want to be able to create
1110 a journal file here. For now, only allow the
1111 user to specify an existing inode to be the
1114 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1115 "journal on remount");
1118 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1120 case Opt_journal_inum:
1122 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1123 "journal on remount");
1126 if (match_int(&args[0], &option))
1130 case Opt_journal_dev:
1132 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1133 "journal on remount");
1136 if (match_int(&args[0], &option))
1138 *journal_devnum = option;
1140 case Opt_journal_path:
1142 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1143 "journal on remount");
1147 journal_path = match_strdup(&args[0]);
1148 if (!journal_path) {
1149 ext3_msg(sb, KERN_ERR, "error: could not dup "
1150 "journal device string");
1154 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1156 ext3_msg(sb, KERN_ERR, "error: could not find "
1157 "journal device path: error %d", error);
1158 kfree(journal_path);
1162 journal_inode = path.dentry->d_inode;
1163 if (!S_ISBLK(journal_inode->i_mode)) {
1164 ext3_msg(sb, KERN_ERR, "error: journal path %s "
1165 "is not a block device", journal_path);
1167 kfree(journal_path);
1171 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1173 kfree(journal_path);
1176 set_opt (sbi->s_mount_opt, NOLOAD);
1179 if (match_int(&args[0], &option))
1184 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1185 sbi->s_commit_interval = HZ * option;
1187 case Opt_data_journal:
1188 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1190 case Opt_data_ordered:
1191 data_opt = EXT3_MOUNT_ORDERED_DATA;
1193 case Opt_data_writeback:
1194 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1197 if (test_opt(sb, DATA_FLAGS) == data_opt)
1199 ext3_msg(sb, KERN_ERR,
1200 "error: cannot change "
1201 "data mode on remount. The filesystem "
1202 "is mounted in data=%s mode and you "
1203 "try to remount it in data=%s mode.",
1204 data_mode_string(test_opt(sb,
1206 data_mode_string(data_opt));
1209 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1210 sbi->s_mount_opt |= data_opt;
1213 case Opt_data_err_abort:
1214 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1216 case Opt_data_err_ignore:
1217 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1221 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1225 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1228 case Opt_offusrjquota:
1229 if (!clear_qf_name(sb, USRQUOTA))
1232 case Opt_offgrpjquota:
1233 if (!clear_qf_name(sb, GRPQUOTA))
1236 case Opt_jqfmt_vfsold:
1237 qfmt = QFMT_VFS_OLD;
1239 case Opt_jqfmt_vfsv0:
1242 case Opt_jqfmt_vfsv1:
1245 if (sb_any_quota_loaded(sb) &&
1246 sbi->s_jquota_fmt != qfmt) {
1247 ext3_msg(sb, KERN_ERR, "error: cannot change "
1248 "journaled quota options when "
1249 "quota turned on.");
1252 sbi->s_jquota_fmt = qfmt;
1256 set_opt(sbi->s_mount_opt, QUOTA);
1257 set_opt(sbi->s_mount_opt, USRQUOTA);
1260 set_opt(sbi->s_mount_opt, QUOTA);
1261 set_opt(sbi->s_mount_opt, GRPQUOTA);
1264 if (sb_any_quota_loaded(sb)) {
1265 ext3_msg(sb, KERN_ERR, "error: cannot change "
1266 "quota options when quota turned on.");
1269 clear_opt(sbi->s_mount_opt, QUOTA);
1270 clear_opt(sbi->s_mount_opt, USRQUOTA);
1271 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1277 ext3_msg(sb, KERN_ERR,
1278 "error: quota options not supported.");
1282 case Opt_offusrjquota:
1283 case Opt_offgrpjquota:
1284 case Opt_jqfmt_vfsold:
1285 case Opt_jqfmt_vfsv0:
1286 case Opt_jqfmt_vfsv1:
1287 ext3_msg(sb, KERN_ERR,
1288 "error: journaled quota options not "
1295 set_opt(sbi->s_mount_opt, ABORT);
1298 clear_opt(sbi->s_mount_opt, BARRIER);
1302 if (match_int(&args[0], &option))
1305 option = 1; /* No argument, default to 1 */
1307 set_opt(sbi->s_mount_opt, BARRIER);
1309 clear_opt(sbi->s_mount_opt, BARRIER);
1315 ext3_msg(sb, KERN_ERR,
1316 "error: resize option only available "
1320 if (match_int(&args[0], &option) != 0)
1322 *n_blocks_count = option;
1325 ext3_msg(sb, KERN_WARNING,
1326 "warning: ignoring deprecated nobh option");
1329 ext3_msg(sb, KERN_WARNING,
1330 "warning: ignoring deprecated bh option");
1333 ext3_msg(sb, KERN_ERR,
1334 "error: unrecognized mount option \"%s\" "
1335 "or missing value", p);
1340 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1341 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1342 clear_opt(sbi->s_mount_opt, USRQUOTA);
1343 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1344 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1346 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1347 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1352 if (!sbi->s_jquota_fmt) {
1353 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1362 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1365 struct ext3_sb_info *sbi = EXT3_SB(sb);
1368 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1369 ext3_msg(sb, KERN_ERR,
1370 "error: revision level too high, "
1371 "forcing read-only mode");
1376 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1377 ext3_msg(sb, KERN_WARNING,
1378 "warning: mounting unchecked fs, "
1379 "running e2fsck is recommended");
1380 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1381 ext3_msg(sb, KERN_WARNING,
1382 "warning: mounting fs with errors, "
1383 "running e2fsck is recommended");
1384 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1385 le16_to_cpu(es->s_mnt_count) >=
1386 le16_to_cpu(es->s_max_mnt_count))
1387 ext3_msg(sb, KERN_WARNING,
1388 "warning: maximal mount count reached, "
1389 "running e2fsck is recommended");
1390 else if (le32_to_cpu(es->s_checkinterval) &&
1391 (le32_to_cpu(es->s_lastcheck) +
1392 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1393 ext3_msg(sb, KERN_WARNING,
1394 "warning: checktime reached, "
1395 "running e2fsck is recommended");
1397 /* @@@ We _will_ want to clear the valid bit if we find
1398 inconsistencies, to force a fsck at reboot. But for
1399 a plain journaled filesystem we can keep it set as
1400 valid forever! :) */
1401 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1403 if (!le16_to_cpu(es->s_max_mnt_count))
1404 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1405 le16_add_cpu(&es->s_mnt_count, 1);
1406 es->s_mtime = cpu_to_le32(get_seconds());
1407 ext3_update_dynamic_rev(sb);
1408 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1410 ext3_commit_super(sb, es, 1);
1411 if (test_opt(sb, DEBUG))
1412 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1413 "bpg=%lu, ipg=%lu, mo=%04lx]",
1415 sbi->s_groups_count,
1416 EXT3_BLOCKS_PER_GROUP(sb),
1417 EXT3_INODES_PER_GROUP(sb),
1420 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1421 char b[BDEVNAME_SIZE];
1422 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1423 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1425 ext3_msg(sb, KERN_INFO, "using internal journal");
1427 cleancache_init_fs(sb);
1431 /* Called at mount-time, super-block is locked */
1432 static int ext3_check_descriptors(struct super_block *sb)
1434 struct ext3_sb_info *sbi = EXT3_SB(sb);
1437 ext3_debug ("Checking group descriptors");
1439 for (i = 0; i < sbi->s_groups_count; i++) {
1440 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1441 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1442 ext3_fsblk_t last_block;
1444 if (i == sbi->s_groups_count - 1)
1445 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1447 last_block = first_block +
1448 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1450 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1451 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1453 ext3_error (sb, "ext3_check_descriptors",
1454 "Block bitmap for group %d"
1455 " not in group (block %lu)!",
1457 le32_to_cpu(gdp->bg_block_bitmap));
1460 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1461 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1463 ext3_error (sb, "ext3_check_descriptors",
1464 "Inode bitmap for group %d"
1465 " not in group (block %lu)!",
1467 le32_to_cpu(gdp->bg_inode_bitmap));
1470 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1471 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1474 ext3_error (sb, "ext3_check_descriptors",
1475 "Inode table for group %d"
1476 " not in group (block %lu)!",
1478 le32_to_cpu(gdp->bg_inode_table));
1483 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1484 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1489 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1490 * the superblock) which were deleted from all directories, but held open by
1491 * a process at the time of a crash. We walk the list and try to delete these
1492 * inodes at recovery time (only with a read-write filesystem).
1494 * In order to keep the orphan inode chain consistent during traversal (in
1495 * case of crash during recovery), we link each inode into the superblock
1496 * orphan list_head and handle it the same way as an inode deletion during
1497 * normal operation (which journals the operations for us).
1499 * We only do an iget() and an iput() on each inode, which is very safe if we
1500 * accidentally point at an in-use or already deleted inode. The worst that
1501 * can happen in this case is that we get a "bit already cleared" message from
1502 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1503 * e2fsck was run on this filesystem, and it must have already done the orphan
1504 * inode cleanup for us, so we can safely abort without any further action.
1506 static void ext3_orphan_cleanup (struct super_block * sb,
1507 struct ext3_super_block * es)
1509 unsigned int s_flags = sb->s_flags;
1510 int nr_orphans = 0, nr_truncates = 0;
1514 if (!es->s_last_orphan) {
1515 jbd_debug(4, "no orphan inodes to clean up\n");
1519 if (bdev_read_only(sb->s_bdev)) {
1520 ext3_msg(sb, KERN_ERR, "error: write access "
1521 "unavailable, skipping orphan cleanup.");
1525 /* Check if feature set allows readwrite operations */
1526 if (EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) {
1527 ext3_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
1528 "unknown ROCOMPAT features");
1532 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1533 /* don't clear list on RO mount w/ errors */
1534 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
1535 jbd_debug(1, "Errors on filesystem, "
1536 "clearing orphan list.\n");
1537 es->s_last_orphan = 0;
1539 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1543 if (s_flags & MS_RDONLY) {
1544 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1545 sb->s_flags &= ~MS_RDONLY;
1548 /* Needed for iput() to work correctly and not trash data */
1549 sb->s_flags |= MS_ACTIVE;
1550 /* Turn on quotas so that they are updated correctly */
1551 for (i = 0; i < EXT3_MAXQUOTAS; i++) {
1552 if (EXT3_SB(sb)->s_qf_names[i]) {
1553 int ret = ext3_quota_on_mount(sb, i);
1555 ext3_msg(sb, KERN_ERR,
1556 "error: cannot turn on journaled "
1562 while (es->s_last_orphan) {
1563 struct inode *inode;
1565 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1566 if (IS_ERR(inode)) {
1567 es->s_last_orphan = 0;
1571 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1572 dquot_initialize(inode);
1573 if (inode->i_nlink) {
1575 "%s: truncating inode %lu to %Ld bytes\n",
1576 __func__, inode->i_ino, inode->i_size);
1577 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1578 inode->i_ino, inode->i_size);
1579 ext3_truncate(inode);
1583 "%s: deleting unreferenced inode %lu\n",
1584 __func__, inode->i_ino);
1585 jbd_debug(2, "deleting unreferenced inode %lu\n",
1589 iput(inode); /* The delete magic happens here! */
1592 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1595 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1596 PLURAL(nr_orphans));
1598 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1599 PLURAL(nr_truncates));
1601 /* Turn quotas off */
1602 for (i = 0; i < EXT3_MAXQUOTAS; i++) {
1603 if (sb_dqopt(sb)->files[i])
1604 dquot_quota_off(sb, i);
1607 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1611 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1612 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1613 * We need to be 1 filesystem block less than the 2^32 sector limit.
1615 static loff_t ext3_max_size(int bits)
1617 loff_t res = EXT3_NDIR_BLOCKS;
1621 /* This is calculated to be the largest file size for a
1622 * dense, file such that the total number of
1623 * sectors in the file, including data and all indirect blocks,
1624 * does not exceed 2^32 -1
1625 * __u32 i_blocks representing the total number of
1626 * 512 bytes blocks of the file
1628 upper_limit = (1LL << 32) - 1;
1630 /* total blocks in file system block size */
1631 upper_limit >>= (bits - 9);
1634 /* indirect blocks */
1636 /* double indirect blocks */
1637 meta_blocks += 1 + (1LL << (bits-2));
1638 /* tripple indirect blocks */
1639 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1641 upper_limit -= meta_blocks;
1642 upper_limit <<= bits;
1644 res += 1LL << (bits-2);
1645 res += 1LL << (2*(bits-2));
1646 res += 1LL << (3*(bits-2));
1648 if (res > upper_limit)
1651 if (res > MAX_LFS_FILESIZE)
1652 res = MAX_LFS_FILESIZE;
1657 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1658 ext3_fsblk_t logic_sb_block,
1661 struct ext3_sb_info *sbi = EXT3_SB(sb);
1662 unsigned long bg, first_meta_bg;
1665 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1667 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1669 return (logic_sb_block + nr + 1);
1670 bg = sbi->s_desc_per_block * nr;
1671 if (ext3_bg_has_super(sb, bg))
1673 return (has_super + ext3_group_first_block_no(sb, bg));
1677 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1679 struct buffer_head * bh;
1680 struct ext3_super_block *es = NULL;
1681 struct ext3_sb_info *sbi;
1683 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1684 ext3_fsblk_t logic_sb_block;
1685 unsigned long offset = 0;
1686 unsigned int journal_inum = 0;
1687 unsigned long journal_devnum = 0;
1688 unsigned long def_mount_opts;
1699 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1703 sbi->s_blockgroup_lock =
1704 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1705 if (!sbi->s_blockgroup_lock) {
1709 sb->s_fs_info = sbi;
1710 sbi->s_sb_block = sb_block;
1712 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1714 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1719 * The ext3 superblock will not be buffer aligned for other than 1kB
1720 * block sizes. We need to calculate the offset from buffer start.
1722 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1723 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1724 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1726 logic_sb_block = sb_block;
1729 if (!(bh = sb_bread(sb, logic_sb_block))) {
1730 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1734 * Note: s_es must be initialized as soon as possible because
1735 * some ext3 macro-instructions depend on its value
1737 es = (struct ext3_super_block *) (bh->b_data + offset);
1739 sb->s_magic = le16_to_cpu(es->s_magic);
1740 if (sb->s_magic != EXT3_SUPER_MAGIC)
1743 /* Set defaults before we parse the mount options */
1744 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1745 if (def_mount_opts & EXT3_DEFM_DEBUG)
1746 set_opt(sbi->s_mount_opt, DEBUG);
1747 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1748 set_opt(sbi->s_mount_opt, GRPID);
1749 if (def_mount_opts & EXT3_DEFM_UID16)
1750 set_opt(sbi->s_mount_opt, NO_UID32);
1751 #ifdef CONFIG_EXT3_FS_XATTR
1752 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1753 set_opt(sbi->s_mount_opt, XATTR_USER);
1755 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1756 if (def_mount_opts & EXT3_DEFM_ACL)
1757 set_opt(sbi->s_mount_opt, POSIX_ACL);
1759 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1760 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1761 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1762 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1763 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1764 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1766 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1767 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1768 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1769 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1771 set_opt(sbi->s_mount_opt, ERRORS_RO);
1773 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
1774 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
1776 /* enable barriers by default */
1777 set_opt(sbi->s_mount_opt, BARRIER);
1778 set_opt(sbi->s_mount_opt, RESERVATION);
1780 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1784 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1785 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1787 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1788 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1789 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1790 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1791 ext3_msg(sb, KERN_WARNING,
1792 "warning: feature flags set on rev 0 fs, "
1793 "running e2fsck is recommended");
1795 * Check feature flags regardless of the revision level, since we
1796 * previously didn't change the revision level when setting the flags,
1797 * so there is a chance incompat flags are set on a rev 0 filesystem.
1799 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1801 ext3_msg(sb, KERN_ERR,
1802 "error: couldn't mount because of unsupported "
1803 "optional features (%x)", le32_to_cpu(features));
1806 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1807 if (!(sb->s_flags & MS_RDONLY) && features) {
1808 ext3_msg(sb, KERN_ERR,
1809 "error: couldn't mount RDWR because of unsupported "
1810 "optional features (%x)", le32_to_cpu(features));
1813 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1815 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1816 blocksize > EXT3_MAX_BLOCK_SIZE) {
1817 ext3_msg(sb, KERN_ERR,
1818 "error: couldn't mount because of unsupported "
1819 "filesystem blocksize %d", blocksize);
1823 hblock = bdev_logical_block_size(sb->s_bdev);
1824 if (sb->s_blocksize != blocksize) {
1826 * Make sure the blocksize for the filesystem is larger
1827 * than the hardware sectorsize for the machine.
1829 if (blocksize < hblock) {
1830 ext3_msg(sb, KERN_ERR,
1831 "error: fsblocksize %d too small for "
1832 "hardware sectorsize %d", blocksize, hblock);
1837 if (!sb_set_blocksize(sb, blocksize)) {
1838 ext3_msg(sb, KERN_ERR,
1839 "error: bad blocksize %d", blocksize);
1842 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1843 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1844 bh = sb_bread(sb, logic_sb_block);
1846 ext3_msg(sb, KERN_ERR,
1847 "error: can't read superblock on 2nd try");
1850 es = (struct ext3_super_block *)(bh->b_data + offset);
1852 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1853 ext3_msg(sb, KERN_ERR,
1854 "error: magic mismatch");
1859 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1861 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1862 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1863 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1865 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1866 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1867 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1868 (!is_power_of_2(sbi->s_inode_size)) ||
1869 (sbi->s_inode_size > blocksize)) {
1870 ext3_msg(sb, KERN_ERR,
1871 "error: unsupported inode size: %d",
1876 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1877 le32_to_cpu(es->s_log_frag_size);
1878 if (blocksize != sbi->s_frag_size) {
1879 ext3_msg(sb, KERN_ERR,
1880 "error: fragsize %lu != blocksize %u (unsupported)",
1881 sbi->s_frag_size, blocksize);
1884 sbi->s_frags_per_block = 1;
1885 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1886 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1887 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1888 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1890 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1891 if (sbi->s_inodes_per_block == 0)
1893 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1894 sbi->s_inodes_per_block;
1895 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1897 sbi->s_mount_state = le16_to_cpu(es->s_state);
1898 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1899 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1900 for (i=0; i < 4; i++)
1901 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1902 sbi->s_def_hash_version = es->s_def_hash_version;
1903 i = le32_to_cpu(es->s_flags);
1904 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1905 sbi->s_hash_unsigned = 3;
1906 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1907 #ifdef __CHAR_UNSIGNED__
1908 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1909 sbi->s_hash_unsigned = 3;
1911 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1915 if (sbi->s_blocks_per_group > blocksize * 8) {
1916 ext3_msg(sb, KERN_ERR,
1917 "#blocks per group too big: %lu",
1918 sbi->s_blocks_per_group);
1921 if (sbi->s_frags_per_group > blocksize * 8) {
1922 ext3_msg(sb, KERN_ERR,
1923 "error: #fragments per group too big: %lu",
1924 sbi->s_frags_per_group);
1927 if (sbi->s_inodes_per_group > blocksize * 8) {
1928 ext3_msg(sb, KERN_ERR,
1929 "error: #inodes per group too big: %lu",
1930 sbi->s_inodes_per_group);
1934 err = generic_check_addressable(sb->s_blocksize_bits,
1935 le32_to_cpu(es->s_blocks_count));
1937 ext3_msg(sb, KERN_ERR,
1938 "error: filesystem is too large to mount safely");
1939 if (sizeof(sector_t) < 8)
1940 ext3_msg(sb, KERN_ERR,
1941 "error: CONFIG_LBDAF not enabled");
1946 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1948 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1949 le32_to_cpu(es->s_first_data_block) - 1)
1950 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1951 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
1952 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1954 if (sbi->s_group_desc == NULL) {
1955 ext3_msg(sb, KERN_ERR,
1956 "error: not enough memory");
1961 bgl_lock_init(sbi->s_blockgroup_lock);
1963 for (i = 0; i < db_count; i++) {
1964 block = descriptor_loc(sb, logic_sb_block, i);
1965 sbi->s_group_desc[i] = sb_bread(sb, block);
1966 if (!sbi->s_group_desc[i]) {
1967 ext3_msg(sb, KERN_ERR,
1968 "error: can't read group descriptor %d", i);
1973 if (!ext3_check_descriptors (sb)) {
1974 ext3_msg(sb, KERN_ERR,
1975 "error: group descriptors corrupted");
1978 sbi->s_gdb_count = db_count;
1979 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1980 spin_lock_init(&sbi->s_next_gen_lock);
1982 /* per fileystem reservation list head & lock */
1983 spin_lock_init(&sbi->s_rsv_window_lock);
1984 sbi->s_rsv_window_root = RB_ROOT;
1985 /* Add a single, static dummy reservation to the start of the
1986 * reservation window list --- it gives us a placeholder for
1987 * append-at-start-of-list which makes the allocation logic
1988 * _much_ simpler. */
1989 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1990 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1991 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1992 sbi->s_rsv_window_head.rsv_goal_size = 0;
1993 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1996 * set up enough so that it can read an inode
1998 sb->s_op = &ext3_sops;
1999 sb->s_export_op = &ext3_export_ops;
2000 sb->s_xattr = ext3_xattr_handlers;
2002 sb->s_qcop = &ext3_qctl_operations;
2003 sb->dq_op = &ext3_quota_operations;
2005 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
2006 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2007 mutex_init(&sbi->s_orphan_lock);
2008 mutex_init(&sbi->s_resize_lock);
2012 needs_recovery = (es->s_last_orphan != 0 ||
2013 EXT3_HAS_INCOMPAT_FEATURE(sb,
2014 EXT3_FEATURE_INCOMPAT_RECOVER));
2017 * The first inode we look at is the journal inode. Don't try
2018 * root first: it may be modified in the journal!
2020 if (!test_opt(sb, NOLOAD) &&
2021 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
2022 if (ext3_load_journal(sb, es, journal_devnum))
2024 } else if (journal_inum) {
2025 if (ext3_create_journal(sb, es, journal_inum))
2029 ext3_msg(sb, KERN_ERR,
2030 "error: no journal found. "
2031 "mounting ext3 over ext2?");
2034 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2035 ext3_count_free_blocks(sb), GFP_KERNEL);
2037 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2038 ext3_count_free_inodes(sb), GFP_KERNEL);
2041 err = percpu_counter_init(&sbi->s_dirs_counter,
2042 ext3_count_dirs(sb), GFP_KERNEL);
2045 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
2050 /* We have now updated the journal if required, so we can
2051 * validate the data journaling mode. */
2052 switch (test_opt(sb, DATA_FLAGS)) {
2054 /* No mode set, assume a default based on the journal
2055 capabilities: ORDERED_DATA if the journal can
2056 cope, else JOURNAL_DATA */
2057 if (journal_check_available_features
2058 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
2059 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
2061 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2064 case EXT3_MOUNT_ORDERED_DATA:
2065 case EXT3_MOUNT_WRITEBACK_DATA:
2066 if (!journal_check_available_features
2067 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
2068 ext3_msg(sb, KERN_ERR,
2069 "error: journal does not support "
2070 "requested data journaling mode");
2078 * The journal_load will have done any necessary log recovery,
2079 * so we can safely mount the rest of the filesystem now.
2082 root = ext3_iget(sb, EXT3_ROOT_INO);
2084 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
2085 ret = PTR_ERR(root);
2088 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2090 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
2093 sb->s_root = d_make_root(root);
2095 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2100 if (ext3_setup_super(sb, es, sb->s_flags & MS_RDONLY))
2101 sb->s_flags |= MS_RDONLY;
2103 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2104 ext3_orphan_cleanup(sb, es);
2105 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2106 if (needs_recovery) {
2107 ext3_mark_recovery_complete(sb, es);
2108 ext3_msg(sb, KERN_INFO, "recovery complete");
2110 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2111 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2112 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2119 ext3_msg(sb, KERN_INFO,
2120 "error: can't find ext3 filesystem on dev %s.",
2125 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2126 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2127 percpu_counter_destroy(&sbi->s_dirs_counter);
2128 journal_destroy(sbi->s_journal);
2130 for (i = 0; i < db_count; i++)
2131 brelse(sbi->s_group_desc[i]);
2132 kfree(sbi->s_group_desc);
2135 for (i = 0; i < EXT3_MAXQUOTAS; i++)
2136 kfree(sbi->s_qf_names[i]);
2138 ext3_blkdev_remove(sbi);
2141 sb->s_fs_info = NULL;
2142 kfree(sbi->s_blockgroup_lock);
2148 * Setup any per-fs journal parameters now. We'll do this both on
2149 * initial mount, once the journal has been initialised but before we've
2150 * done any recovery; and again on any subsequent remount.
2152 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2154 struct ext3_sb_info *sbi = EXT3_SB(sb);
2156 if (sbi->s_commit_interval)
2157 journal->j_commit_interval = sbi->s_commit_interval;
2158 /* We could also set up an ext3-specific default for the commit
2159 * interval here, but for now we'll just fall back to the jbd
2162 spin_lock(&journal->j_state_lock);
2163 if (test_opt(sb, BARRIER))
2164 journal->j_flags |= JFS_BARRIER;
2166 journal->j_flags &= ~JFS_BARRIER;
2167 if (test_opt(sb, DATA_ERR_ABORT))
2168 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2170 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2171 spin_unlock(&journal->j_state_lock);
2174 static journal_t *ext3_get_journal(struct super_block *sb,
2175 unsigned int journal_inum)
2177 struct inode *journal_inode;
2180 /* First, test for the existence of a valid inode on disk. Bad
2181 * things happen if we iget() an unused inode, as the subsequent
2182 * iput() will try to delete it. */
2184 journal_inode = ext3_iget(sb, journal_inum);
2185 if (IS_ERR(journal_inode)) {
2186 ext3_msg(sb, KERN_ERR, "error: no journal found");
2189 if (!journal_inode->i_nlink) {
2190 make_bad_inode(journal_inode);
2191 iput(journal_inode);
2192 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2196 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2197 journal_inode, journal_inode->i_size);
2198 if (!S_ISREG(journal_inode->i_mode)) {
2199 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2200 iput(journal_inode);
2204 journal = journal_init_inode(journal_inode);
2206 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2207 iput(journal_inode);
2210 journal->j_private = sb;
2211 ext3_init_journal_params(sb, journal);
2215 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2218 struct buffer_head * bh;
2222 int hblock, blocksize;
2223 ext3_fsblk_t sb_block;
2224 unsigned long offset;
2225 struct ext3_super_block * es;
2226 struct block_device *bdev;
2228 bdev = ext3_blkdev_get(j_dev, sb);
2232 blocksize = sb->s_blocksize;
2233 hblock = bdev_logical_block_size(bdev);
2234 if (blocksize < hblock) {
2235 ext3_msg(sb, KERN_ERR,
2236 "error: blocksize too small for journal device");
2240 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2241 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2242 set_blocksize(bdev, blocksize);
2243 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2244 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2245 "external journal");
2249 es = (struct ext3_super_block *) (bh->b_data + offset);
2250 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2251 !(le32_to_cpu(es->s_feature_incompat) &
2252 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2253 ext3_msg(sb, KERN_ERR, "error: external journal has "
2259 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2260 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2265 len = le32_to_cpu(es->s_blocks_count);
2266 start = sb_block + 1;
2267 brelse(bh); /* we're done with the superblock */
2269 journal = journal_init_dev(bdev, sb->s_bdev,
2270 start, len, blocksize);
2272 ext3_msg(sb, KERN_ERR,
2273 "error: failed to create device journal");
2276 journal->j_private = sb;
2277 if (!bh_uptodate_or_lock(journal->j_sb_buffer)) {
2278 if (bh_submit_read(journal->j_sb_buffer)) {
2279 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2283 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2284 ext3_msg(sb, KERN_ERR,
2285 "error: external journal has more than one "
2286 "user (unsupported) - %d",
2287 be32_to_cpu(journal->j_superblock->s_nr_users));
2290 EXT3_SB(sb)->journal_bdev = bdev;
2291 ext3_init_journal_params(sb, journal);
2294 journal_destroy(journal);
2296 ext3_blkdev_put(bdev);
2300 static int ext3_load_journal(struct super_block *sb,
2301 struct ext3_super_block *es,
2302 unsigned long journal_devnum)
2305 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2308 int really_read_only;
2310 if (journal_devnum &&
2311 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2312 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2313 "numbers have changed");
2314 journal_dev = new_decode_dev(journal_devnum);
2316 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2318 really_read_only = bdev_read_only(sb->s_bdev);
2321 * Are we loading a blank journal or performing recovery after a
2322 * crash? For recovery, we need to check in advance whether we
2323 * can get read-write access to the device.
2326 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2327 if (sb->s_flags & MS_RDONLY) {
2328 ext3_msg(sb, KERN_INFO,
2329 "recovery required on readonly filesystem");
2330 if (really_read_only) {
2331 ext3_msg(sb, KERN_ERR, "error: write access "
2332 "unavailable, cannot proceed");
2335 ext3_msg(sb, KERN_INFO,
2336 "write access will be enabled during recovery");
2340 if (journal_inum && journal_dev) {
2341 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2342 "and inode journals");
2347 if (!(journal = ext3_get_journal(sb, journal_inum)))
2350 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2354 if (!(journal->j_flags & JFS_BARRIER))
2355 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2357 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2358 err = journal_update_format(journal);
2360 ext3_msg(sb, KERN_ERR, "error updating journal");
2361 journal_destroy(journal);
2366 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2367 err = journal_wipe(journal, !really_read_only);
2369 err = journal_load(journal);
2372 ext3_msg(sb, KERN_ERR, "error loading journal");
2373 journal_destroy(journal);
2377 EXT3_SB(sb)->s_journal = journal;
2378 ext3_clear_journal_err(sb, es);
2380 if (!really_read_only && journal_devnum &&
2381 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2382 es->s_journal_dev = cpu_to_le32(journal_devnum);
2384 /* Make sure we flush the recovery flag to disk. */
2385 ext3_commit_super(sb, es, 1);
2391 static int ext3_create_journal(struct super_block *sb,
2392 struct ext3_super_block *es,
2393 unsigned int journal_inum)
2398 if (sb->s_flags & MS_RDONLY) {
2399 ext3_msg(sb, KERN_ERR,
2400 "error: readonly filesystem when trying to "
2405 journal = ext3_get_journal(sb, journal_inum);
2409 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2412 err = journal_create(journal);
2414 ext3_msg(sb, KERN_ERR, "error creating journal");
2415 journal_destroy(journal);
2419 EXT3_SB(sb)->s_journal = journal;
2421 ext3_update_dynamic_rev(sb);
2422 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2423 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2425 es->s_journal_inum = cpu_to_le32(journal_inum);
2427 /* Make sure we flush the recovery flag to disk. */
2428 ext3_commit_super(sb, es, 1);
2433 static int ext3_commit_super(struct super_block *sb,
2434 struct ext3_super_block *es,
2437 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2443 if (buffer_write_io_error(sbh)) {
2445 * Oh, dear. A previous attempt to write the
2446 * superblock failed. This could happen because the
2447 * USB device was yanked out. Or it could happen to
2448 * be a transient write error and maybe the block will
2449 * be remapped. Nothing we can do but to retry the
2450 * write and hope for the best.
2452 ext3_msg(sb, KERN_ERR, "previous I/O error to "
2453 "superblock detected");
2454 clear_buffer_write_io_error(sbh);
2455 set_buffer_uptodate(sbh);
2458 * If the file system is mounted read-only, don't update the
2459 * superblock write time. This avoids updating the superblock
2460 * write time when we are mounting the root file system
2461 * read/only but we need to replay the journal; at that point,
2462 * for people who are east of GMT and who make their clock
2463 * tick in localtime for Windows bug-for-bug compatibility,
2464 * the clock is set in the future, and this will cause e2fsck
2465 * to complain and force a full file system check.
2467 if (!(sb->s_flags & MS_RDONLY))
2468 es->s_wtime = cpu_to_le32(get_seconds());
2469 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2470 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2471 BUFFER_TRACE(sbh, "marking dirty");
2472 mark_buffer_dirty(sbh);
2474 error = sync_dirty_buffer(sbh);
2475 if (buffer_write_io_error(sbh)) {
2476 ext3_msg(sb, KERN_ERR, "I/O error while writing "
2478 clear_buffer_write_io_error(sbh);
2479 set_buffer_uptodate(sbh);
2487 * Have we just finished recovery? If so, and if we are mounting (or
2488 * remounting) the filesystem readonly, then we will end up with a
2489 * consistent fs on disk. Record that fact.
2491 static void ext3_mark_recovery_complete(struct super_block * sb,
2492 struct ext3_super_block * es)
2494 journal_t *journal = EXT3_SB(sb)->s_journal;
2496 journal_lock_updates(journal);
2497 if (journal_flush(journal) < 0)
2500 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2501 sb->s_flags & MS_RDONLY) {
2502 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2503 ext3_commit_super(sb, es, 1);
2507 journal_unlock_updates(journal);
2511 * If we are mounting (or read-write remounting) a filesystem whose journal
2512 * has recorded an error from a previous lifetime, move that error to the
2513 * main filesystem now.
2515 static void ext3_clear_journal_err(struct super_block *sb,
2516 struct ext3_super_block *es)
2522 journal = EXT3_SB(sb)->s_journal;
2525 * Now check for any error status which may have been recorded in the
2526 * journal by a prior ext3_error() or ext3_abort()
2529 j_errno = journal_errno(journal);
2533 errstr = ext3_decode_error(sb, j_errno, nbuf);
2534 ext3_warning(sb, __func__, "Filesystem error recorded "
2535 "from previous mount: %s", errstr);
2536 ext3_warning(sb, __func__, "Marking fs in need of "
2537 "filesystem check.");
2539 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2540 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2541 ext3_commit_super (sb, es, 1);
2543 journal_clear_err(journal);
2548 * Force the running and committing transactions to commit,
2549 * and wait on the commit.
2551 int ext3_force_commit(struct super_block *sb)
2556 if (sb->s_flags & MS_RDONLY)
2559 journal = EXT3_SB(sb)->s_journal;
2560 ret = ext3_journal_force_commit(journal);
2564 static int ext3_sync_fs(struct super_block *sb, int wait)
2568 trace_ext3_sync_fs(sb, wait);
2570 * Writeback quota in non-journalled quota case - journalled quota has
2573 dquot_writeback_dquots(sb, -1);
2574 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2576 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2582 * LVM calls this function before a (read-only) snapshot is created. This
2583 * gives us a chance to flush the journal completely and mark the fs clean.
2585 static int ext3_freeze(struct super_block *sb)
2590 if (!(sb->s_flags & MS_RDONLY)) {
2591 journal = EXT3_SB(sb)->s_journal;
2593 /* Now we set up the journal barrier. */
2594 journal_lock_updates(journal);
2597 * We don't want to clear needs_recovery flag when we failed
2598 * to flush the journal.
2600 error = journal_flush(journal);
2604 /* Journal blocked and flushed, clear needs_recovery flag. */
2605 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2606 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2613 journal_unlock_updates(journal);
2618 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2619 * flag here, even though the filesystem is not technically dirty yet.
2621 static int ext3_unfreeze(struct super_block *sb)
2623 if (!(sb->s_flags & MS_RDONLY)) {
2624 /* Reser the needs_recovery flag before the fs is unlocked. */
2625 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2626 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2627 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2632 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2634 struct ext3_super_block * es;
2635 struct ext3_sb_info *sbi = EXT3_SB(sb);
2636 ext3_fsblk_t n_blocks_count = 0;
2637 unsigned long old_sb_flags;
2638 struct ext3_mount_options old_opts;
2639 int enable_quota = 0;
2645 sync_filesystem(sb);
2647 /* Store the original options */
2648 old_sb_flags = sb->s_flags;
2649 old_opts.s_mount_opt = sbi->s_mount_opt;
2650 old_opts.s_resuid = sbi->s_resuid;
2651 old_opts.s_resgid = sbi->s_resgid;
2652 old_opts.s_commit_interval = sbi->s_commit_interval;
2654 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2655 for (i = 0; i < EXT3_MAXQUOTAS; i++)
2656 if (sbi->s_qf_names[i]) {
2657 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
2659 if (!old_opts.s_qf_names[i]) {
2662 for (j = 0; j < i; j++)
2663 kfree(old_opts.s_qf_names[j]);
2667 old_opts.s_qf_names[i] = NULL;
2671 * Allow the "check" option to be passed as a remount option.
2673 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2678 if (test_opt(sb, ABORT))
2679 ext3_abort(sb, __func__, "Abort forced by user");
2681 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2682 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2686 ext3_init_journal_params(sb, sbi->s_journal);
2688 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2689 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2690 if (test_opt(sb, ABORT)) {
2695 if (*flags & MS_RDONLY) {
2696 err = dquot_suspend(sb, -1);
2701 * First of all, the unconditional stuff we have to do
2702 * to disable replay of the journal when we next remount
2704 sb->s_flags |= MS_RDONLY;
2707 * OK, test if we are remounting a valid rw partition
2708 * readonly, and if so set the rdonly flag and then
2709 * mark the partition as valid again.
2711 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2712 (sbi->s_mount_state & EXT3_VALID_FS))
2713 es->s_state = cpu_to_le16(sbi->s_mount_state);
2715 ext3_mark_recovery_complete(sb, es);
2718 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2719 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2720 ext3_msg(sb, KERN_WARNING,
2721 "warning: couldn't remount RDWR "
2722 "because of unsupported optional "
2723 "features (%x)", le32_to_cpu(ret));
2729 * If we have an unprocessed orphan list hanging
2730 * around from a previously readonly bdev mount,
2731 * require a full umount & mount for now.
2733 if (es->s_last_orphan) {
2734 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2735 "remount RDWR because of unprocessed "
2736 "orphan inode list. Please "
2737 "umount & mount instead.");
2743 * Mounting a RDONLY partition read-write, so reread
2744 * and store the current valid flag. (It may have
2745 * been changed by e2fsck since we originally mounted
2748 ext3_clear_journal_err(sb, es);
2749 sbi->s_mount_state = le16_to_cpu(es->s_state);
2750 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2752 if (!ext3_setup_super (sb, es, 0))
2753 sb->s_flags &= ~MS_RDONLY;
2758 /* Release old quota file names */
2759 for (i = 0; i < EXT3_MAXQUOTAS; i++)
2760 kfree(old_opts.s_qf_names[i]);
2763 dquot_resume(sb, -1);
2766 sb->s_flags = old_sb_flags;
2767 sbi->s_mount_opt = old_opts.s_mount_opt;
2768 sbi->s_resuid = old_opts.s_resuid;
2769 sbi->s_resgid = old_opts.s_resgid;
2770 sbi->s_commit_interval = old_opts.s_commit_interval;
2772 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2773 for (i = 0; i < EXT3_MAXQUOTAS; i++) {
2774 kfree(sbi->s_qf_names[i]);
2775 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2781 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2783 struct super_block *sb = dentry->d_sb;
2784 struct ext3_sb_info *sbi = EXT3_SB(sb);
2785 struct ext3_super_block *es = sbi->s_es;
2788 if (test_opt(sb, MINIX_DF)) {
2789 sbi->s_overhead_last = 0;
2790 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2791 unsigned long ngroups = sbi->s_groups_count, i;
2792 ext3_fsblk_t overhead = 0;
2796 * Compute the overhead (FS structures). This is constant
2797 * for a given filesystem unless the number of block groups
2798 * changes so we cache the previous value until it does.
2802 * All of the blocks before first_data_block are
2805 overhead = le32_to_cpu(es->s_first_data_block);
2808 * Add the overhead attributed to the superblock and
2809 * block group descriptors. If the sparse superblocks
2810 * feature is turned on, then not all groups have this.
2812 for (i = 0; i < ngroups; i++) {
2813 overhead += ext3_bg_has_super(sb, i) +
2814 ext3_bg_num_gdb(sb, i);
2819 * Every block group has an inode bitmap, a block
2820 * bitmap, and an inode table.
2822 overhead += ngroups * (2 + sbi->s_itb_per_group);
2824 /* Add the internal journal blocks as well */
2825 if (sbi->s_journal && !sbi->journal_bdev)
2826 overhead += sbi->s_journal->j_maxlen;
2828 sbi->s_overhead_last = overhead;
2830 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2833 buf->f_type = EXT3_SUPER_MAGIC;
2834 buf->f_bsize = sb->s_blocksize;
2835 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2836 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2837 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2838 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2840 buf->f_files = le32_to_cpu(es->s_inodes_count);
2841 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2842 buf->f_namelen = EXT3_NAME_LEN;
2843 fsid = le64_to_cpup((void *)es->s_uuid) ^
2844 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2845 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2846 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2850 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2851 * is locked for write. Otherwise the are possible deadlocks:
2852 * Process 1 Process 2
2853 * ext3_create() quota_sync()
2854 * journal_start() write_dquot()
2855 * dquot_initialize() down(dqio_mutex)
2856 * down(dqio_mutex) journal_start()
2862 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2864 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
2867 static int ext3_write_dquot(struct dquot *dquot)
2871 struct inode *inode;
2873 inode = dquot_to_inode(dquot);
2874 handle = ext3_journal_start(inode,
2875 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2877 return PTR_ERR(handle);
2878 ret = dquot_commit(dquot);
2879 err = ext3_journal_stop(handle);
2885 static int ext3_acquire_dquot(struct dquot *dquot)
2890 handle = ext3_journal_start(dquot_to_inode(dquot),
2891 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2893 return PTR_ERR(handle);
2894 ret = dquot_acquire(dquot);
2895 err = ext3_journal_stop(handle);
2901 static int ext3_release_dquot(struct dquot *dquot)
2906 handle = ext3_journal_start(dquot_to_inode(dquot),
2907 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2908 if (IS_ERR(handle)) {
2909 /* Release dquot anyway to avoid endless cycle in dqput() */
2910 dquot_release(dquot);
2911 return PTR_ERR(handle);
2913 ret = dquot_release(dquot);
2914 err = ext3_journal_stop(handle);
2920 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2922 /* Are we journaling quotas? */
2923 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2924 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2925 dquot_mark_dquot_dirty(dquot);
2926 return ext3_write_dquot(dquot);
2928 return dquot_mark_dquot_dirty(dquot);
2932 static int ext3_write_info(struct super_block *sb, int type)
2937 /* Data block + inode block */
2938 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2940 return PTR_ERR(handle);
2941 ret = dquot_commit_info(sb, type);
2942 err = ext3_journal_stop(handle);
2949 * Turn on quotas during mount time - we need to find
2950 * the quota file and such...
2952 static int ext3_quota_on_mount(struct super_block *sb, int type)
2954 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2955 EXT3_SB(sb)->s_jquota_fmt, type);
2959 * Standard function to be called on quota_on
2961 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2966 if (!test_opt(sb, QUOTA))
2969 /* Quotafile not on the same filesystem? */
2970 if (path->dentry->d_sb != sb)
2972 /* Journaling quota? */
2973 if (EXT3_SB(sb)->s_qf_names[type]) {
2974 /* Quotafile not of fs root? */
2975 if (path->dentry->d_parent != sb->s_root)
2976 ext3_msg(sb, KERN_WARNING,
2977 "warning: Quota file not on filesystem root. "
2978 "Journaled quota will not work.");
2982 * When we journal data on quota file, we have to flush journal to see
2983 * all updates to the file when we bypass pagecache...
2985 if (ext3_should_journal_data(path->dentry->d_inode)) {
2987 * We don't need to lock updates but journal_flush() could
2988 * otherwise be livelocked...
2990 journal_lock_updates(EXT3_SB(sb)->s_journal);
2991 err = journal_flush(EXT3_SB(sb)->s_journal);
2992 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2997 return dquot_quota_on(sb, type, format_id, path);
3000 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3001 * acquiring the locks... As quota files are never truncated and quota code
3002 * itself serializes the operations (and no one else should touch the files)
3003 * we don't have to be afraid of races */
3004 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
3005 size_t len, loff_t off)
3007 struct inode *inode = sb_dqopt(sb)->files[type];
3008 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
3010 int offset = off & (sb->s_blocksize - 1);
3013 struct buffer_head *bh;
3014 loff_t i_size = i_size_read(inode);
3018 if (off+len > i_size)
3021 while (toread > 0) {
3022 tocopy = sb->s_blocksize - offset < toread ?
3023 sb->s_blocksize - offset : toread;
3024 bh = ext3_bread(NULL, inode, blk, 0, &err);
3027 if (!bh) /* A hole? */
3028 memset(data, 0, tocopy);
3030 memcpy(data, bh->b_data+offset, tocopy);
3040 /* Write to quotafile (we know the transaction is already started and has
3041 * enough credits) */
3042 static ssize_t ext3_quota_write(struct super_block *sb, int type,
3043 const char *data, size_t len, loff_t off)
3045 struct inode *inode = sb_dqopt(sb)->files[type];
3046 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
3048 int offset = off & (sb->s_blocksize - 1);
3049 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
3050 struct buffer_head *bh;
3051 handle_t *handle = journal_current_handle();
3054 ext3_msg(sb, KERN_WARNING,
3055 "warning: quota write (off=%llu, len=%llu)"
3056 " cancelled because transaction is not started.",
3057 (unsigned long long)off, (unsigned long long)len);
3062 * Since we account only one data block in transaction credits,
3063 * then it is impossible to cross a block boundary.
3065 if (sb->s_blocksize - offset < len) {
3066 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3067 " cancelled because not block aligned",
3068 (unsigned long long)off, (unsigned long long)len);
3071 bh = ext3_bread(handle, inode, blk, 1, &err);
3074 if (journal_quota) {
3075 err = ext3_journal_get_write_access(handle, bh);
3082 memcpy(bh->b_data+offset, data, len);
3083 flush_dcache_page(bh->b_page);
3086 err = ext3_journal_dirty_metadata(handle, bh);
3088 /* Always do at least ordered writes for quotas */
3089 err = ext3_journal_dirty_data(handle, bh);
3090 mark_buffer_dirty(bh);
3096 if (inode->i_size < off + len) {
3097 i_size_write(inode, off + len);
3098 EXT3_I(inode)->i_disksize = inode->i_size;
3101 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3102 ext3_mark_inode_dirty(handle, inode);
3108 static struct dentry *ext3_mount(struct file_system_type *fs_type,
3109 int flags, const char *dev_name, void *data)
3111 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
3114 static struct file_system_type ext3_fs_type = {
3115 .owner = THIS_MODULE,
3117 .mount = ext3_mount,
3118 .kill_sb = kill_block_super,
3119 .fs_flags = FS_REQUIRES_DEV,
3121 MODULE_ALIAS_FS("ext3");
3123 static int __init init_ext3_fs(void)
3125 int err = init_ext3_xattr();
3128 err = init_inodecache();
3131 err = register_filesystem(&ext3_fs_type);
3136 destroy_inodecache();
3142 static void __exit exit_ext3_fs(void)
3144 unregister_filesystem(&ext3_fs_type);
3145 destroy_inodecache();
3149 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3150 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3151 MODULE_LICENSE("GPL");
3152 module_init(init_ext3_fs)
3153 module_exit(exit_ext3_fs)