2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/ext4_fs.h>
25 #include <linux/ext4_jbd2.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
39 #include <asm/uaccess.h>
45 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
46 unsigned long journal_devnum);
47 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
49 static void ext4_commit_super (struct super_block * sb,
50 struct ext4_super_block * es,
52 static void ext4_mark_recovery_complete(struct super_block * sb,
53 struct ext4_super_block * es);
54 static void ext4_clear_journal_err(struct super_block * sb,
55 struct ext4_super_block * es);
56 static int ext4_sync_fs(struct super_block *sb, int wait);
57 static const char *ext4_decode_error(struct super_block * sb, int errno,
59 static int ext4_remount (struct super_block * sb, int * flags, char * data);
60 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
61 static void ext4_unlockfs(struct super_block *sb);
62 static void ext4_write_super (struct super_block * sb);
63 static void ext4_write_super_lockfs(struct super_block *sb);
66 ext4_fsblk_t ext4_block_bitmap(struct ext4_group_desc *bg)
68 return le32_to_cpu(bg->bg_block_bitmap) |
69 ((ext4_fsblk_t)le16_to_cpu(bg->bg_block_bitmap_hi) << 32);
72 ext4_fsblk_t ext4_inode_bitmap(struct ext4_group_desc *bg)
74 return le32_to_cpu(bg->bg_inode_bitmap) |
75 ((ext4_fsblk_t)le16_to_cpu(bg->bg_inode_bitmap_hi) << 32);
78 ext4_fsblk_t ext4_inode_table(struct ext4_group_desc *bg)
80 return le32_to_cpu(bg->bg_inode_table) |
81 ((ext4_fsblk_t)le16_to_cpu(bg->bg_inode_table_hi) << 32);
84 void ext4_block_bitmap_set(struct ext4_group_desc *bg, ext4_fsblk_t blk)
86 bg->bg_block_bitmap = cpu_to_le32((u32)blk);
87 bg->bg_block_bitmap_hi = cpu_to_le16(blk >> 32);
90 void ext4_inode_bitmap_set(struct ext4_group_desc *bg, ext4_fsblk_t blk)
92 bg->bg_inode_bitmap = cpu_to_le32((u32)blk);
93 bg->bg_inode_bitmap_hi = cpu_to_le16(blk >> 32);
96 void ext4_inode_table_set(struct ext4_group_desc *bg, ext4_fsblk_t blk)
98 bg->bg_inode_table = cpu_to_le32((u32)blk);
99 bg->bg_inode_table_hi = cpu_to_le16(blk >> 32);
103 * Wrappers for jbd2_journal_start/end.
105 * The only special thing we need to do here is to make sure that all
106 * journal_end calls result in the superblock being marked dirty, so
107 * that sync() will call the filesystem's write_super callback if
110 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
114 if (sb->s_flags & MS_RDONLY)
115 return ERR_PTR(-EROFS);
117 /* Special case here: if the journal has aborted behind our
118 * backs (eg. EIO in the commit thread), then we still need to
119 * take the FS itself readonly cleanly. */
120 journal = EXT4_SB(sb)->s_journal;
121 if (is_journal_aborted(journal)) {
122 ext4_abort(sb, __FUNCTION__,
123 "Detected aborted journal");
124 return ERR_PTR(-EROFS);
127 return jbd2_journal_start(journal, nblocks);
131 * The only special thing we need to do here is to make sure that all
132 * jbd2_journal_stop calls result in the superblock being marked dirty, so
133 * that sync() will call the filesystem's write_super callback if
136 int __ext4_journal_stop(const char *where, handle_t *handle)
138 struct super_block *sb;
142 sb = handle->h_transaction->t_journal->j_private;
144 rc = jbd2_journal_stop(handle);
149 __ext4_std_error(sb, where, err);
153 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
154 struct buffer_head *bh, handle_t *handle, int err)
157 const char *errstr = ext4_decode_error(NULL, err, nbuf);
160 BUFFER_TRACE(bh, "abort");
165 if (is_handle_aborted(handle))
168 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
169 caller, errstr, err_fn);
171 jbd2_journal_abort_handle(handle);
174 /* Deal with the reporting of failure conditions on a filesystem such as
175 * inconsistencies detected or read IO failures.
177 * On ext2, we can store the error state of the filesystem in the
178 * superblock. That is not possible on ext4, because we may have other
179 * write ordering constraints on the superblock which prevent us from
180 * writing it out straight away; and given that the journal is about to
181 * be aborted, we can't rely on the current, or future, transactions to
182 * write out the superblock safely.
184 * We'll just use the jbd2_journal_abort() error code to record an error in
185 * the journal instead. On recovery, the journal will compain about
186 * that error until we've noted it down and cleared it.
189 static void ext4_handle_error(struct super_block *sb)
191 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
193 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
194 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
196 if (sb->s_flags & MS_RDONLY)
199 if (!test_opt (sb, ERRORS_CONT)) {
200 journal_t *journal = EXT4_SB(sb)->s_journal;
202 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
204 jbd2_journal_abort(journal, -EIO);
206 if (test_opt (sb, ERRORS_RO)) {
207 printk (KERN_CRIT "Remounting filesystem read-only\n");
208 sb->s_flags |= MS_RDONLY;
210 ext4_commit_super(sb, es, 1);
211 if (test_opt(sb, ERRORS_PANIC))
212 panic("EXT4-fs (device %s): panic forced after error\n",
216 void ext4_error (struct super_block * sb, const char * function,
217 const char * fmt, ...)
222 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
227 ext4_handle_error(sb);
230 static const char *ext4_decode_error(struct super_block * sb, int errno,
237 errstr = "IO failure";
240 errstr = "Out of memory";
243 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
244 errstr = "Journal has aborted";
246 errstr = "Readonly filesystem";
249 /* If the caller passed in an extra buffer for unknown
250 * errors, textualise them now. Else we just return
253 /* Check for truncated error codes... */
254 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
263 /* __ext4_std_error decodes expected errors from journaling functions
264 * automatically and invokes the appropriate error response. */
266 void __ext4_std_error (struct super_block * sb, const char * function,
272 /* Special case: if the error is EROFS, and we're not already
273 * inside a transaction, then there's really no point in logging
275 if (errno == -EROFS && journal_current_handle() == NULL &&
276 (sb->s_flags & MS_RDONLY))
279 errstr = ext4_decode_error(sb, errno, nbuf);
280 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
281 sb->s_id, function, errstr);
283 ext4_handle_error(sb);
287 * ext4_abort is a much stronger failure handler than ext4_error. The
288 * abort function may be used to deal with unrecoverable failures such
289 * as journal IO errors or ENOMEM at a critical moment in log management.
291 * We unconditionally force the filesystem into an ABORT|READONLY state,
292 * unless the error response on the fs has been set to panic in which
293 * case we take the easy way out and panic immediately.
296 void ext4_abort (struct super_block * sb, const char * function,
297 const char * fmt, ...)
301 printk (KERN_CRIT "ext4_abort called.\n");
304 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
309 if (test_opt(sb, ERRORS_PANIC))
310 panic("EXT4-fs panic from previous error\n");
312 if (sb->s_flags & MS_RDONLY)
315 printk(KERN_CRIT "Remounting filesystem read-only\n");
316 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
317 sb->s_flags |= MS_RDONLY;
318 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
319 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
322 void ext4_warning (struct super_block * sb, const char * function,
323 const char * fmt, ...)
328 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
335 void ext4_update_dynamic_rev(struct super_block *sb)
337 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
339 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
342 ext4_warning(sb, __FUNCTION__,
343 "updating to rev %d because of new feature flag, "
344 "running e2fsck is recommended",
347 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
348 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
349 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
350 /* leave es->s_feature_*compat flags alone */
351 /* es->s_uuid will be set by e2fsck if empty */
354 * The rest of the superblock fields should be zero, and if not it
355 * means they are likely already in use, so leave them alone. We
356 * can leave it up to e2fsck to clean up any inconsistencies there.
361 * Open the external journal device
363 static struct block_device *ext4_blkdev_get(dev_t dev)
365 struct block_device *bdev;
366 char b[BDEVNAME_SIZE];
368 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
374 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
375 __bdevname(dev, b), PTR_ERR(bdev));
380 * Release the journal device
382 static int ext4_blkdev_put(struct block_device *bdev)
385 return blkdev_put(bdev);
388 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
390 struct block_device *bdev;
393 bdev = sbi->journal_bdev;
395 ret = ext4_blkdev_put(bdev);
396 sbi->journal_bdev = NULL;
401 static inline struct inode *orphan_list_entry(struct list_head *l)
403 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
406 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
410 printk(KERN_ERR "sb orphan head is %d\n",
411 le32_to_cpu(sbi->s_es->s_last_orphan));
413 printk(KERN_ERR "sb_info orphan list:\n");
414 list_for_each(l, &sbi->s_orphan) {
415 struct inode *inode = orphan_list_entry(l);
417 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
418 inode->i_sb->s_id, inode->i_ino, inode,
419 inode->i_mode, inode->i_nlink,
424 static void ext4_put_super (struct super_block * sb)
426 struct ext4_sb_info *sbi = EXT4_SB(sb);
427 struct ext4_super_block *es = sbi->s_es;
430 ext4_ext_release(sb);
431 ext4_xattr_put_super(sb);
432 jbd2_journal_destroy(sbi->s_journal);
433 if (!(sb->s_flags & MS_RDONLY)) {
434 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
435 es->s_state = cpu_to_le16(sbi->s_mount_state);
436 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
437 mark_buffer_dirty(sbi->s_sbh);
438 ext4_commit_super(sb, es, 1);
441 for (i = 0; i < sbi->s_gdb_count; i++)
442 brelse(sbi->s_group_desc[i]);
443 kfree(sbi->s_group_desc);
444 percpu_counter_destroy(&sbi->s_freeblocks_counter);
445 percpu_counter_destroy(&sbi->s_freeinodes_counter);
446 percpu_counter_destroy(&sbi->s_dirs_counter);
449 for (i = 0; i < MAXQUOTAS; i++)
450 kfree(sbi->s_qf_names[i]);
453 /* Debugging code just in case the in-memory inode orphan list
454 * isn't empty. The on-disk one can be non-empty if we've
455 * detected an error and taken the fs readonly, but the
456 * in-memory list had better be clean by this point. */
457 if (!list_empty(&sbi->s_orphan))
458 dump_orphan_list(sb, sbi);
459 J_ASSERT(list_empty(&sbi->s_orphan));
461 invalidate_bdev(sb->s_bdev, 0);
462 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
464 * Invalidate the journal device's buffers. We don't want them
465 * floating about in memory - the physical journal device may
466 * hotswapped, and it breaks the `ro-after' testing code.
468 sync_blockdev(sbi->journal_bdev);
469 invalidate_bdev(sbi->journal_bdev, 0);
470 ext4_blkdev_remove(sbi);
472 sb->s_fs_info = NULL;
477 static kmem_cache_t *ext4_inode_cachep;
480 * Called inside transaction, so use GFP_NOFS
482 static struct inode *ext4_alloc_inode(struct super_block *sb)
484 struct ext4_inode_info *ei;
486 ei = kmem_cache_alloc(ext4_inode_cachep, SLAB_NOFS);
489 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
490 ei->i_acl = EXT4_ACL_NOT_CACHED;
491 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
493 ei->i_block_alloc_info = NULL;
494 ei->vfs_inode.i_version = 1;
495 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
496 return &ei->vfs_inode;
499 static void ext4_destroy_inode(struct inode *inode)
501 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
504 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
506 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
508 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
509 SLAB_CTOR_CONSTRUCTOR) {
510 INIT_LIST_HEAD(&ei->i_orphan);
511 #ifdef CONFIG_EXT4DEV_FS_XATTR
512 init_rwsem(&ei->xattr_sem);
514 mutex_init(&ei->truncate_mutex);
515 inode_init_once(&ei->vfs_inode);
519 static int init_inodecache(void)
521 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
522 sizeof(struct ext4_inode_info),
523 0, (SLAB_RECLAIM_ACCOUNT|
526 if (ext4_inode_cachep == NULL)
531 static void destroy_inodecache(void)
533 kmem_cache_destroy(ext4_inode_cachep);
536 static void ext4_clear_inode(struct inode *inode)
538 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
539 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
540 if (EXT4_I(inode)->i_acl &&
541 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
542 posix_acl_release(EXT4_I(inode)->i_acl);
543 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
545 if (EXT4_I(inode)->i_default_acl &&
546 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
547 posix_acl_release(EXT4_I(inode)->i_default_acl);
548 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
551 ext4_discard_reservation(inode);
552 EXT4_I(inode)->i_block_alloc_info = NULL;
557 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
559 #if defined(CONFIG_QUOTA)
560 struct ext4_sb_info *sbi = EXT4_SB(sb);
562 if (sbi->s_jquota_fmt)
563 seq_printf(seq, ",jqfmt=%s",
564 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
566 if (sbi->s_qf_names[USRQUOTA])
567 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
569 if (sbi->s_qf_names[GRPQUOTA])
570 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
572 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
573 seq_puts(seq, ",usrquota");
575 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
576 seq_puts(seq, ",grpquota");
580 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
582 struct super_block *sb = vfs->mnt_sb;
584 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
585 seq_puts(seq, ",data=journal");
586 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
587 seq_puts(seq, ",data=ordered");
588 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
589 seq_puts(seq, ",data=writeback");
591 ext4_show_quota_options(seq, sb);
597 static struct dentry *ext4_get_dentry(struct super_block *sb, void *vobjp)
600 unsigned long ino = objp[0];
601 __u32 generation = objp[1];
603 struct dentry *result;
605 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
606 return ERR_PTR(-ESTALE);
607 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
608 return ERR_PTR(-ESTALE);
610 /* iget isn't really right if the inode is currently unallocated!!
612 * ext4_read_inode will return a bad_inode if the inode had been
613 * deleted, so we should be safe.
615 * Currently we don't know the generation for parent directory, so
616 * a generation of 0 means "accept any"
618 inode = iget(sb, ino);
620 return ERR_PTR(-ENOMEM);
621 if (is_bad_inode(inode) ||
622 (generation && inode->i_generation != generation)) {
624 return ERR_PTR(-ESTALE);
626 /* now to find a dentry.
627 * If possible, get a well-connected one
629 result = d_alloc_anon(inode);
632 return ERR_PTR(-ENOMEM);
638 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
639 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
641 static int ext4_dquot_initialize(struct inode *inode, int type);
642 static int ext4_dquot_drop(struct inode *inode);
643 static int ext4_write_dquot(struct dquot *dquot);
644 static int ext4_acquire_dquot(struct dquot *dquot);
645 static int ext4_release_dquot(struct dquot *dquot);
646 static int ext4_mark_dquot_dirty(struct dquot *dquot);
647 static int ext4_write_info(struct super_block *sb, int type);
648 static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
649 static int ext4_quota_on_mount(struct super_block *sb, int type);
650 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
651 size_t len, loff_t off);
652 static ssize_t ext4_quota_write(struct super_block *sb, int type,
653 const char *data, size_t len, loff_t off);
655 static struct dquot_operations ext4_quota_operations = {
656 .initialize = ext4_dquot_initialize,
657 .drop = ext4_dquot_drop,
658 .alloc_space = dquot_alloc_space,
659 .alloc_inode = dquot_alloc_inode,
660 .free_space = dquot_free_space,
661 .free_inode = dquot_free_inode,
662 .transfer = dquot_transfer,
663 .write_dquot = ext4_write_dquot,
664 .acquire_dquot = ext4_acquire_dquot,
665 .release_dquot = ext4_release_dquot,
666 .mark_dirty = ext4_mark_dquot_dirty,
667 .write_info = ext4_write_info
670 static struct quotactl_ops ext4_qctl_operations = {
671 .quota_on = ext4_quota_on,
672 .quota_off = vfs_quota_off,
673 .quota_sync = vfs_quota_sync,
674 .get_info = vfs_get_dqinfo,
675 .set_info = vfs_set_dqinfo,
676 .get_dqblk = vfs_get_dqblk,
677 .set_dqblk = vfs_set_dqblk
681 static struct super_operations ext4_sops = {
682 .alloc_inode = ext4_alloc_inode,
683 .destroy_inode = ext4_destroy_inode,
684 .read_inode = ext4_read_inode,
685 .write_inode = ext4_write_inode,
686 .dirty_inode = ext4_dirty_inode,
687 .delete_inode = ext4_delete_inode,
688 .put_super = ext4_put_super,
689 .write_super = ext4_write_super,
690 .sync_fs = ext4_sync_fs,
691 .write_super_lockfs = ext4_write_super_lockfs,
692 .unlockfs = ext4_unlockfs,
693 .statfs = ext4_statfs,
694 .remount_fs = ext4_remount,
695 .clear_inode = ext4_clear_inode,
696 .show_options = ext4_show_options,
698 .quota_read = ext4_quota_read,
699 .quota_write = ext4_quota_write,
703 static struct export_operations ext4_export_ops = {
704 .get_parent = ext4_get_parent,
705 .get_dentry = ext4_get_dentry,
709 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
710 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
711 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
712 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
713 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
714 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
715 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
716 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
717 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
718 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
719 Opt_grpquota, Opt_extents,
722 static match_table_t tokens = {
723 {Opt_bsd_df, "bsddf"},
724 {Opt_minix_df, "minixdf"},
725 {Opt_grpid, "grpid"},
726 {Opt_grpid, "bsdgroups"},
727 {Opt_nogrpid, "nogrpid"},
728 {Opt_nogrpid, "sysvgroups"},
729 {Opt_resgid, "resgid=%u"},
730 {Opt_resuid, "resuid=%u"},
732 {Opt_err_cont, "errors=continue"},
733 {Opt_err_panic, "errors=panic"},
734 {Opt_err_ro, "errors=remount-ro"},
735 {Opt_nouid32, "nouid32"},
736 {Opt_nocheck, "nocheck"},
737 {Opt_nocheck, "check=none"},
738 {Opt_debug, "debug"},
739 {Opt_oldalloc, "oldalloc"},
740 {Opt_orlov, "orlov"},
741 {Opt_user_xattr, "user_xattr"},
742 {Opt_nouser_xattr, "nouser_xattr"},
744 {Opt_noacl, "noacl"},
745 {Opt_reservation, "reservation"},
746 {Opt_noreservation, "noreservation"},
747 {Opt_noload, "noload"},
750 {Opt_commit, "commit=%u"},
751 {Opt_journal_update, "journal=update"},
752 {Opt_journal_inum, "journal=%u"},
753 {Opt_journal_dev, "journal_dev=%u"},
754 {Opt_abort, "abort"},
755 {Opt_data_journal, "data=journal"},
756 {Opt_data_ordered, "data=ordered"},
757 {Opt_data_writeback, "data=writeback"},
758 {Opt_offusrjquota, "usrjquota="},
759 {Opt_usrjquota, "usrjquota=%s"},
760 {Opt_offgrpjquota, "grpjquota="},
761 {Opt_grpjquota, "grpjquota=%s"},
762 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
763 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
764 {Opt_grpquota, "grpquota"},
765 {Opt_noquota, "noquota"},
766 {Opt_quota, "quota"},
767 {Opt_usrquota, "usrquota"},
768 {Opt_barrier, "barrier=%u"},
769 {Opt_extents, "extents"},
771 {Opt_resize, "resize"},
774 static ext4_fsblk_t get_sb_block(void **data)
776 ext4_fsblk_t sb_block;
777 char *options = (char *) *data;
779 if (!options || strncmp(options, "sb=", 3) != 0)
780 return 1; /* Default location */
782 /*todo: use simple_strtoll with >32bit ext4 */
783 sb_block = simple_strtoul(options, &options, 0);
784 if (*options && *options != ',') {
785 printk("EXT4-fs: Invalid sb specification: %s\n",
791 *data = (void *) options;
795 static int parse_options (char *options, struct super_block *sb,
796 unsigned int *inum, unsigned long *journal_devnum,
797 ext4_fsblk_t *n_blocks_count, int is_remount)
799 struct ext4_sb_info *sbi = EXT4_SB(sb);
801 substring_t args[MAX_OPT_ARGS];
812 while ((p = strsep (&options, ",")) != NULL) {
817 token = match_token(p, tokens, args);
820 clear_opt (sbi->s_mount_opt, MINIX_DF);
823 set_opt (sbi->s_mount_opt, MINIX_DF);
826 set_opt (sbi->s_mount_opt, GRPID);
829 clear_opt (sbi->s_mount_opt, GRPID);
832 if (match_int(&args[0], &option))
834 sbi->s_resuid = option;
837 if (match_int(&args[0], &option))
839 sbi->s_resgid = option;
842 /* handled by get_sb_block() instead of here */
843 /* *sb_block = match_int(&args[0]); */
846 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
847 clear_opt (sbi->s_mount_opt, ERRORS_RO);
848 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
851 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
852 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
853 set_opt (sbi->s_mount_opt, ERRORS_RO);
856 clear_opt (sbi->s_mount_opt, ERRORS_RO);
857 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
858 set_opt (sbi->s_mount_opt, ERRORS_CONT);
861 set_opt (sbi->s_mount_opt, NO_UID32);
864 clear_opt (sbi->s_mount_opt, CHECK);
867 set_opt (sbi->s_mount_opt, DEBUG);
870 set_opt (sbi->s_mount_opt, OLDALLOC);
873 clear_opt (sbi->s_mount_opt, OLDALLOC);
875 #ifdef CONFIG_EXT4DEV_FS_XATTR
877 set_opt (sbi->s_mount_opt, XATTR_USER);
879 case Opt_nouser_xattr:
880 clear_opt (sbi->s_mount_opt, XATTR_USER);
884 case Opt_nouser_xattr:
885 printk("EXT4 (no)user_xattr options not supported\n");
888 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
890 set_opt(sbi->s_mount_opt, POSIX_ACL);
893 clear_opt(sbi->s_mount_opt, POSIX_ACL);
898 printk("EXT4 (no)acl options not supported\n");
901 case Opt_reservation:
902 set_opt(sbi->s_mount_opt, RESERVATION);
904 case Opt_noreservation:
905 clear_opt(sbi->s_mount_opt, RESERVATION);
907 case Opt_journal_update:
909 /* Eventually we will want to be able to create
910 a journal file here. For now, only allow the
911 user to specify an existing inode to be the
914 printk(KERN_ERR "EXT4-fs: cannot specify "
915 "journal on remount\n");
918 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
920 case Opt_journal_inum:
922 printk(KERN_ERR "EXT4-fs: cannot specify "
923 "journal on remount\n");
926 if (match_int(&args[0], &option))
930 case Opt_journal_dev:
932 printk(KERN_ERR "EXT4-fs: cannot specify "
933 "journal on remount\n");
936 if (match_int(&args[0], &option))
938 *journal_devnum = option;
941 set_opt (sbi->s_mount_opt, NOLOAD);
944 if (match_int(&args[0], &option))
949 option = JBD_DEFAULT_MAX_COMMIT_AGE;
950 sbi->s_commit_interval = HZ * option;
952 case Opt_data_journal:
953 data_opt = EXT4_MOUNT_JOURNAL_DATA;
955 case Opt_data_ordered:
956 data_opt = EXT4_MOUNT_ORDERED_DATA;
958 case Opt_data_writeback:
959 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
962 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
965 "EXT4-fs: cannot change data "
966 "mode on remount\n");
970 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
971 sbi->s_mount_opt |= data_opt;
981 if (sb_any_quota_enabled(sb)) {
983 "EXT4-fs: Cannot change journalled "
984 "quota options when quota turned on.\n");
987 qname = match_strdup(&args[0]);
990 "EXT4-fs: not enough memory for "
991 "storing quotafile name.\n");
994 if (sbi->s_qf_names[qtype] &&
995 strcmp(sbi->s_qf_names[qtype], qname)) {
997 "EXT4-fs: %s quota file already "
998 "specified.\n", QTYPE2NAME(qtype));
1002 sbi->s_qf_names[qtype] = qname;
1003 if (strchr(sbi->s_qf_names[qtype], '/')) {
1005 "EXT4-fs: quotafile must be on "
1006 "filesystem root.\n");
1007 kfree(sbi->s_qf_names[qtype]);
1008 sbi->s_qf_names[qtype] = NULL;
1011 set_opt(sbi->s_mount_opt, QUOTA);
1013 case Opt_offusrjquota:
1016 case Opt_offgrpjquota:
1019 if (sb_any_quota_enabled(sb)) {
1020 printk(KERN_ERR "EXT4-fs: Cannot change "
1021 "journalled quota options when "
1022 "quota turned on.\n");
1026 * The space will be released later when all options
1027 * are confirmed to be correct
1029 sbi->s_qf_names[qtype] = NULL;
1031 case Opt_jqfmt_vfsold:
1032 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1034 case Opt_jqfmt_vfsv0:
1035 sbi->s_jquota_fmt = QFMT_VFS_V0;
1039 set_opt(sbi->s_mount_opt, QUOTA);
1040 set_opt(sbi->s_mount_opt, USRQUOTA);
1043 set_opt(sbi->s_mount_opt, QUOTA);
1044 set_opt(sbi->s_mount_opt, GRPQUOTA);
1047 if (sb_any_quota_enabled(sb)) {
1048 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1049 "options when quota turned on.\n");
1052 clear_opt(sbi->s_mount_opt, QUOTA);
1053 clear_opt(sbi->s_mount_opt, USRQUOTA);
1054 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1062 case Opt_offusrjquota:
1063 case Opt_offgrpjquota:
1064 case Opt_jqfmt_vfsold:
1065 case Opt_jqfmt_vfsv0:
1067 "EXT4-fs: journalled quota options not "
1074 set_opt(sbi->s_mount_opt, ABORT);
1077 if (match_int(&args[0], &option))
1080 set_opt(sbi->s_mount_opt, BARRIER);
1082 clear_opt(sbi->s_mount_opt, BARRIER);
1088 printk("EXT4-fs: resize option only available "
1092 if (match_int(&args[0], &option) != 0)
1094 *n_blocks_count = option;
1097 set_opt(sbi->s_mount_opt, NOBH);
1100 clear_opt(sbi->s_mount_opt, NOBH);
1103 set_opt (sbi->s_mount_opt, EXTENTS);
1107 "EXT4-fs: Unrecognized mount option \"%s\" "
1108 "or missing value\n", p);
1113 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1114 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1115 sbi->s_qf_names[USRQUOTA])
1116 clear_opt(sbi->s_mount_opt, USRQUOTA);
1118 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1119 sbi->s_qf_names[GRPQUOTA])
1120 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1122 if ((sbi->s_qf_names[USRQUOTA] &&
1123 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1124 (sbi->s_qf_names[GRPQUOTA] &&
1125 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1126 printk(KERN_ERR "EXT4-fs: old and new quota "
1127 "format mixing.\n");
1131 if (!sbi->s_jquota_fmt) {
1132 printk(KERN_ERR "EXT4-fs: journalled quota format "
1133 "not specified.\n");
1137 if (sbi->s_jquota_fmt) {
1138 printk(KERN_ERR "EXT4-fs: journalled quota format "
1139 "specified with no journalling "
1148 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1151 struct ext4_sb_info *sbi = EXT4_SB(sb);
1154 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1155 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1156 "forcing read-only mode\n");
1161 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1162 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1163 "running e2fsck is recommended\n");
1164 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1165 printk (KERN_WARNING
1166 "EXT4-fs warning: mounting fs with errors, "
1167 "running e2fsck is recommended\n");
1168 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1169 le16_to_cpu(es->s_mnt_count) >=
1170 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1171 printk (KERN_WARNING
1172 "EXT4-fs warning: maximal mount count reached, "
1173 "running e2fsck is recommended\n");
1174 else if (le32_to_cpu(es->s_checkinterval) &&
1175 (le32_to_cpu(es->s_lastcheck) +
1176 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1177 printk (KERN_WARNING
1178 "EXT4-fs warning: checktime reached, "
1179 "running e2fsck is recommended\n");
1181 /* @@@ We _will_ want to clear the valid bit if we find
1182 inconsistencies, to force a fsck at reboot. But for
1183 a plain journaled filesystem we can keep it set as
1184 valid forever! :) */
1185 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1187 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1188 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1189 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1190 es->s_mtime = cpu_to_le32(get_seconds());
1191 ext4_update_dynamic_rev(sb);
1192 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1194 ext4_commit_super(sb, es, 1);
1195 if (test_opt(sb, DEBUG))
1196 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1197 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1199 sbi->s_groups_count,
1200 EXT4_BLOCKS_PER_GROUP(sb),
1201 EXT4_INODES_PER_GROUP(sb),
1204 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1205 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1206 char b[BDEVNAME_SIZE];
1208 printk("external journal on %s\n",
1209 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1211 printk("internal journal\n");
1216 /* Called at mount-time, super-block is locked */
1217 static int ext4_check_descriptors (struct super_block * sb)
1219 struct ext4_sb_info *sbi = EXT4_SB(sb);
1220 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1221 ext4_fsblk_t last_block;
1222 ext4_fsblk_t block_bitmap;
1223 ext4_fsblk_t inode_bitmap;
1224 ext4_fsblk_t inode_table;
1225 struct ext4_group_desc * gdp = NULL;
1229 ext4_debug ("Checking group descriptors");
1231 for (i = 0; i < sbi->s_groups_count; i++)
1233 if (i == sbi->s_groups_count - 1)
1234 last_block = ext4_blocks_count(sbi->s_es) - 1;
1236 last_block = first_block +
1237 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1239 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1240 gdp = (struct ext4_group_desc *)
1241 sbi->s_group_desc[desc_block++]->b_data;
1242 block_bitmap = ext4_block_bitmap(gdp);
1243 if (block_bitmap < first_block || block_bitmap > last_block)
1245 ext4_error (sb, "ext4_check_descriptors",
1246 "Block bitmap for group %d"
1247 " not in group (block %llu)!",
1251 inode_bitmap = ext4_inode_bitmap(gdp);
1252 if (inode_bitmap < first_block || inode_bitmap > last_block)
1254 ext4_error (sb, "ext4_check_descriptors",
1255 "Inode bitmap for group %d"
1256 " not in group (block %llu)!",
1260 inode_table = ext4_inode_table(gdp);
1261 if (inode_table < first_block ||
1262 inode_table + sbi->s_itb_per_group > last_block)
1264 ext4_error (sb, "ext4_check_descriptors",
1265 "Inode table for group %d"
1266 " not in group (block %llu)!",
1270 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1271 gdp = (struct ext4_group_desc *)
1272 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
1275 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1276 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1281 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1282 * the superblock) which were deleted from all directories, but held open by
1283 * a process at the time of a crash. We walk the list and try to delete these
1284 * inodes at recovery time (only with a read-write filesystem).
1286 * In order to keep the orphan inode chain consistent during traversal (in
1287 * case of crash during recovery), we link each inode into the superblock
1288 * orphan list_head and handle it the same way as an inode deletion during
1289 * normal operation (which journals the operations for us).
1291 * We only do an iget() and an iput() on each inode, which is very safe if we
1292 * accidentally point at an in-use or already deleted inode. The worst that
1293 * can happen in this case is that we get a "bit already cleared" message from
1294 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1295 * e2fsck was run on this filesystem, and it must have already done the orphan
1296 * inode cleanup for us, so we can safely abort without any further action.
1298 static void ext4_orphan_cleanup (struct super_block * sb,
1299 struct ext4_super_block * es)
1301 unsigned int s_flags = sb->s_flags;
1302 int nr_orphans = 0, nr_truncates = 0;
1306 if (!es->s_last_orphan) {
1307 jbd_debug(4, "no orphan inodes to clean up\n");
1311 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1312 if (es->s_last_orphan)
1313 jbd_debug(1, "Errors on filesystem, "
1314 "clearing orphan list.\n");
1315 es->s_last_orphan = 0;
1316 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1320 if (s_flags & MS_RDONLY) {
1321 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1323 sb->s_flags &= ~MS_RDONLY;
1326 /* Needed for iput() to work correctly and not trash data */
1327 sb->s_flags |= MS_ACTIVE;
1328 /* Turn on quotas so that they are updated correctly */
1329 for (i = 0; i < MAXQUOTAS; i++) {
1330 if (EXT4_SB(sb)->s_qf_names[i]) {
1331 int ret = ext4_quota_on_mount(sb, i);
1334 "EXT4-fs: Cannot turn on journalled "
1335 "quota: error %d\n", ret);
1340 while (es->s_last_orphan) {
1341 struct inode *inode;
1344 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1345 es->s_last_orphan = 0;
1349 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1351 if (inode->i_nlink) {
1353 "%s: truncating inode %lu to %Ld bytes\n",
1354 __FUNCTION__, inode->i_ino, inode->i_size);
1355 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1356 inode->i_ino, inode->i_size);
1357 ext4_truncate(inode);
1361 "%s: deleting unreferenced inode %lu\n",
1362 __FUNCTION__, inode->i_ino);
1363 jbd_debug(2, "deleting unreferenced inode %lu\n",
1367 iput(inode); /* The delete magic happens here! */
1370 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1373 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1374 sb->s_id, PLURAL(nr_orphans));
1376 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1377 sb->s_id, PLURAL(nr_truncates));
1379 /* Turn quotas off */
1380 for (i = 0; i < MAXQUOTAS; i++) {
1381 if (sb_dqopt(sb)->files[i])
1382 vfs_quota_off(sb, i);
1385 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1388 #define log2(n) ffz(~(n))
1391 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1392 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1393 * We need to be 1 filesystem block less than the 2^32 sector limit.
1395 static loff_t ext4_max_size(int bits)
1397 loff_t res = EXT4_NDIR_BLOCKS;
1398 /* This constant is calculated to be the largest file size for a
1399 * dense, 4k-blocksize file such that the total number of
1400 * sectors in the file, including data and all indirect blocks,
1401 * does not exceed 2^32. */
1402 const loff_t upper_limit = 0x1ff7fffd000LL;
1404 res += 1LL << (bits-2);
1405 res += 1LL << (2*(bits-2));
1406 res += 1LL << (3*(bits-2));
1408 if (res > upper_limit)
1413 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1414 ext4_fsblk_t logic_sb_block,
1417 struct ext4_sb_info *sbi = EXT4_SB(sb);
1418 unsigned long bg, first_meta_bg;
1421 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1423 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1425 return (logic_sb_block + nr + 1);
1426 bg = sbi->s_desc_per_block * nr;
1427 if (ext4_bg_has_super(sb, bg))
1429 return (has_super + ext4_group_first_block_no(sb, bg));
1433 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1435 struct buffer_head * bh;
1436 struct ext4_super_block *es = NULL;
1437 struct ext4_sb_info *sbi;
1439 ext4_fsblk_t sb_block = get_sb_block(&data);
1440 ext4_fsblk_t logic_sb_block;
1441 unsigned long offset = 0;
1442 unsigned int journal_inum = 0;
1443 unsigned long journal_devnum = 0;
1444 unsigned long def_mount_opts;
1454 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1457 sb->s_fs_info = sbi;
1458 sbi->s_mount_opt = 0;
1459 sbi->s_resuid = EXT4_DEF_RESUID;
1460 sbi->s_resgid = EXT4_DEF_RESGID;
1464 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1466 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1471 * The ext4 superblock will not be buffer aligned for other than 1kB
1472 * block sizes. We need to calculate the offset from buffer start.
1474 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1475 logic_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1476 offset = sector_div(logic_sb_block, blocksize);
1478 logic_sb_block = sb_block;
1481 if (!(bh = sb_bread(sb, logic_sb_block))) {
1482 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1486 * Note: s_es must be initialized as soon as possible because
1487 * some ext4 macro-instructions depend on its value
1489 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1491 sb->s_magic = le16_to_cpu(es->s_magic);
1492 if (sb->s_magic != EXT4_SUPER_MAGIC)
1495 /* Set defaults before we parse the mount options */
1496 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1497 if (def_mount_opts & EXT4_DEFM_DEBUG)
1498 set_opt(sbi->s_mount_opt, DEBUG);
1499 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1500 set_opt(sbi->s_mount_opt, GRPID);
1501 if (def_mount_opts & EXT4_DEFM_UID16)
1502 set_opt(sbi->s_mount_opt, NO_UID32);
1503 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1504 set_opt(sbi->s_mount_opt, XATTR_USER);
1505 if (def_mount_opts & EXT4_DEFM_ACL)
1506 set_opt(sbi->s_mount_opt, POSIX_ACL);
1507 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1508 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1509 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1510 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1511 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1512 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1514 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1515 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1516 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
1517 set_opt(sbi->s_mount_opt, ERRORS_RO);
1519 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1520 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1522 set_opt(sbi->s_mount_opt, RESERVATION);
1524 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1528 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1529 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1531 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1532 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1533 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1534 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1536 "EXT4-fs warning: feature flags set on rev 0 fs, "
1537 "running e2fsck is recommended\n");
1539 * Check feature flags regardless of the revision level, since we
1540 * previously didn't change the revision level when setting the flags,
1541 * so there is a chance incompat flags are set on a rev 0 filesystem.
1543 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1545 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1546 "unsupported optional features (%x).\n",
1547 sb->s_id, le32_to_cpu(features));
1550 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1551 if (!(sb->s_flags & MS_RDONLY) && features) {
1552 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1553 "unsupported optional features (%x).\n",
1554 sb->s_id, le32_to_cpu(features));
1557 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1559 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1560 blocksize > EXT4_MAX_BLOCK_SIZE) {
1562 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1563 blocksize, sb->s_id);
1567 hblock = bdev_hardsect_size(sb->s_bdev);
1568 if (sb->s_blocksize != blocksize) {
1570 * Make sure the blocksize for the filesystem is larger
1571 * than the hardware sectorsize for the machine.
1573 if (blocksize < hblock) {
1574 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
1575 "device blocksize %d.\n", blocksize, hblock);
1580 sb_set_blocksize(sb, blocksize);
1581 logic_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1582 offset = sector_div(logic_sb_block, blocksize);
1583 bh = sb_bread(sb, logic_sb_block);
1586 "EXT4-fs: Can't read superblock on 2nd try.\n");
1589 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1591 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1593 "EXT4-fs: Magic mismatch, very weird !\n");
1598 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1600 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1601 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1602 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1604 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1605 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1606 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1607 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1608 (sbi->s_inode_size > blocksize)) {
1610 "EXT4-fs: unsupported inode size: %d\n",
1615 sbi->s_frag_size = EXT4_MIN_FRAG_SIZE <<
1616 le32_to_cpu(es->s_log_frag_size);
1617 if (blocksize != sbi->s_frag_size) {
1619 "EXT4-fs: fragsize %lu != blocksize %u (unsupported)\n",
1620 sbi->s_frag_size, blocksize);
1623 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1624 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
1625 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE ||
1626 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1627 sbi->s_desc_size & (sbi->s_desc_size - 1)) {
1629 "EXT4-fs: unsupported descriptor size %ld\n",
1634 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
1635 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1636 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1637 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1638 if (EXT4_INODE_SIZE(sb) == 0)
1640 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
1641 if (sbi->s_inodes_per_block == 0)
1643 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1644 sbi->s_inodes_per_block;
1645 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
1647 sbi->s_mount_state = le16_to_cpu(es->s_state);
1648 sbi->s_addr_per_block_bits = log2(EXT4_ADDR_PER_BLOCK(sb));
1649 sbi->s_desc_per_block_bits = log2(EXT4_DESC_PER_BLOCK(sb));
1650 for (i=0; i < 4; i++)
1651 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1652 sbi->s_def_hash_version = es->s_def_hash_version;
1654 if (sbi->s_blocks_per_group > blocksize * 8) {
1656 "EXT4-fs: #blocks per group too big: %lu\n",
1657 sbi->s_blocks_per_group);
1660 if (sbi->s_frags_per_group > blocksize * 8) {
1662 "EXT4-fs: #fragments per group too big: %lu\n",
1663 sbi->s_frags_per_group);
1666 if (sbi->s_inodes_per_group > blocksize * 8) {
1668 "EXT4-fs: #inodes per group too big: %lu\n",
1669 sbi->s_inodes_per_group);
1673 if (ext4_blocks_count(es) >
1674 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1675 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
1676 " too large to mount safely\n", sb->s_id);
1677 if (sizeof(sector_t) < 8)
1678 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
1683 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1685 blocks_count = (ext4_blocks_count(es) -
1686 le32_to_cpu(es->s_first_data_block) +
1687 EXT4_BLOCKS_PER_GROUP(sb) - 1);
1688 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
1689 sbi->s_groups_count = blocks_count;
1690 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1691 EXT4_DESC_PER_BLOCK(sb);
1692 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1694 if (sbi->s_group_desc == NULL) {
1695 printk (KERN_ERR "EXT4-fs: not enough memory\n");
1699 bgl_lock_init(&sbi->s_blockgroup_lock);
1701 for (i = 0; i < db_count; i++) {
1702 block = descriptor_loc(sb, logic_sb_block, i);
1703 sbi->s_group_desc[i] = sb_bread(sb, block);
1704 if (!sbi->s_group_desc[i]) {
1705 printk (KERN_ERR "EXT4-fs: "
1706 "can't read group descriptor %d\n", i);
1711 if (!ext4_check_descriptors (sb)) {
1712 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
1715 sbi->s_gdb_count = db_count;
1716 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1717 spin_lock_init(&sbi->s_next_gen_lock);
1719 percpu_counter_init(&sbi->s_freeblocks_counter,
1720 ext4_count_free_blocks(sb));
1721 percpu_counter_init(&sbi->s_freeinodes_counter,
1722 ext4_count_free_inodes(sb));
1723 percpu_counter_init(&sbi->s_dirs_counter,
1724 ext4_count_dirs(sb));
1726 /* per fileystem reservation list head & lock */
1727 spin_lock_init(&sbi->s_rsv_window_lock);
1728 sbi->s_rsv_window_root = RB_ROOT;
1729 /* Add a single, static dummy reservation to the start of the
1730 * reservation window list --- it gives us a placeholder for
1731 * append-at-start-of-list which makes the allocation logic
1732 * _much_ simpler. */
1733 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1734 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1735 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1736 sbi->s_rsv_window_head.rsv_goal_size = 0;
1737 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
1740 * set up enough so that it can read an inode
1742 sb->s_op = &ext4_sops;
1743 sb->s_export_op = &ext4_export_ops;
1744 sb->s_xattr = ext4_xattr_handlers;
1746 sb->s_qcop = &ext4_qctl_operations;
1747 sb->dq_op = &ext4_quota_operations;
1749 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1753 needs_recovery = (es->s_last_orphan != 0 ||
1754 EXT4_HAS_INCOMPAT_FEATURE(sb,
1755 EXT4_FEATURE_INCOMPAT_RECOVER));
1758 * The first inode we look at is the journal inode. Don't try
1759 * root first: it may be modified in the journal!
1761 if (!test_opt(sb, NOLOAD) &&
1762 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1763 if (ext4_load_journal(sb, es, journal_devnum))
1765 } else if (journal_inum) {
1766 if (ext4_create_journal(sb, es, journal_inum))
1771 "ext4: No journal on filesystem on %s\n",
1776 /* We have now updated the journal if required, so we can
1777 * validate the data journaling mode. */
1778 switch (test_opt(sb, DATA_FLAGS)) {
1780 /* No mode set, assume a default based on the journal
1781 capabilities: ORDERED_DATA if the journal can
1782 cope, else JOURNAL_DATA */
1783 if (jbd2_journal_check_available_features
1784 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
1785 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1787 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1790 case EXT4_MOUNT_ORDERED_DATA:
1791 case EXT4_MOUNT_WRITEBACK_DATA:
1792 if (!jbd2_journal_check_available_features
1793 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
1794 printk(KERN_ERR "EXT4-fs: Journal does not support "
1795 "requested data journaling mode\n");
1802 if (test_opt(sb, NOBH)) {
1803 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1804 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
1805 "its supported only with writeback mode\n");
1806 clear_opt(sbi->s_mount_opt, NOBH);
1810 * The jbd2_journal_load will have done any necessary log recovery,
1811 * so we can safely mount the rest of the filesystem now.
1814 root = iget(sb, EXT4_ROOT_INO);
1815 sb->s_root = d_alloc_root(root);
1817 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1821 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1824 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
1828 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1830 * akpm: core read_super() calls in here with the superblock locked.
1831 * That deadlocks, because orphan cleanup needs to lock the superblock
1832 * in numerous places. Here we just pop the lock - it's relatively
1833 * harmless, because we are now ready to accept write_super() requests,
1834 * and aviro says that's the only reason for hanging onto the
1837 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
1838 ext4_orphan_cleanup(sb, es);
1839 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
1841 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
1842 ext4_mark_recovery_complete(sb, es);
1843 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
1844 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
1845 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
1855 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
1860 jbd2_journal_destroy(sbi->s_journal);
1862 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1863 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1864 percpu_counter_destroy(&sbi->s_dirs_counter);
1866 for (i = 0; i < db_count; i++)
1867 brelse(sbi->s_group_desc[i]);
1868 kfree(sbi->s_group_desc);
1871 for (i = 0; i < MAXQUOTAS; i++)
1872 kfree(sbi->s_qf_names[i]);
1874 ext4_blkdev_remove(sbi);
1877 sb->s_fs_info = NULL;
1884 * Setup any per-fs journal parameters now. We'll do this both on
1885 * initial mount, once the journal has been initialised but before we've
1886 * done any recovery; and again on any subsequent remount.
1888 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
1890 struct ext4_sb_info *sbi = EXT4_SB(sb);
1892 if (sbi->s_commit_interval)
1893 journal->j_commit_interval = sbi->s_commit_interval;
1894 /* We could also set up an ext4-specific default for the commit
1895 * interval here, but for now we'll just fall back to the jbd
1898 spin_lock(&journal->j_state_lock);
1899 if (test_opt(sb, BARRIER))
1900 journal->j_flags |= JBD2_BARRIER;
1902 journal->j_flags &= ~JBD2_BARRIER;
1903 spin_unlock(&journal->j_state_lock);
1906 static journal_t *ext4_get_journal(struct super_block *sb,
1907 unsigned int journal_inum)
1909 struct inode *journal_inode;
1912 /* First, test for the existence of a valid inode on disk. Bad
1913 * things happen if we iget() an unused inode, as the subsequent
1914 * iput() will try to delete it. */
1916 journal_inode = iget(sb, journal_inum);
1917 if (!journal_inode) {
1918 printk(KERN_ERR "EXT4-fs: no journal found.\n");
1921 if (!journal_inode->i_nlink) {
1922 make_bad_inode(journal_inode);
1923 iput(journal_inode);
1924 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
1928 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1929 journal_inode, journal_inode->i_size);
1930 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1931 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
1932 iput(journal_inode);
1936 journal = jbd2_journal_init_inode(journal_inode);
1938 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
1939 iput(journal_inode);
1942 journal->j_private = sb;
1943 ext4_init_journal_params(sb, journal);
1947 static journal_t *ext4_get_dev_journal(struct super_block *sb,
1950 struct buffer_head * bh;
1954 int hblock, blocksize;
1955 ext4_fsblk_t sb_block;
1956 unsigned long offset;
1957 struct ext4_super_block * es;
1958 struct block_device *bdev;
1960 bdev = ext4_blkdev_get(j_dev);
1964 if (bd_claim(bdev, sb)) {
1966 "EXT4: failed to claim external journal device.\n");
1971 blocksize = sb->s_blocksize;
1972 hblock = bdev_hardsect_size(bdev);
1973 if (blocksize < hblock) {
1975 "EXT4-fs: blocksize too small for journal device.\n");
1979 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
1980 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
1981 set_blocksize(bdev, blocksize);
1982 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1983 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
1984 "external journal\n");
1988 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1989 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
1990 !(le32_to_cpu(es->s_feature_incompat) &
1991 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1992 printk(KERN_ERR "EXT4-fs: external journal has "
1993 "bad superblock\n");
1998 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1999 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2004 len = ext4_blocks_count(es);
2005 start = sb_block + 1;
2006 brelse(bh); /* we're done with the superblock */
2008 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2009 start, len, blocksize);
2011 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2014 journal->j_private = sb;
2015 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2016 wait_on_buffer(journal->j_sb_buffer);
2017 if (!buffer_uptodate(journal->j_sb_buffer)) {
2018 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2021 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2022 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2023 "user (unsupported) - %d\n",
2024 be32_to_cpu(journal->j_superblock->s_nr_users));
2027 EXT4_SB(sb)->journal_bdev = bdev;
2028 ext4_init_journal_params(sb, journal);
2031 jbd2_journal_destroy(journal);
2033 ext4_blkdev_put(bdev);
2037 static int ext4_load_journal(struct super_block *sb,
2038 struct ext4_super_block *es,
2039 unsigned long journal_devnum)
2042 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2045 int really_read_only;
2047 if (journal_devnum &&
2048 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2049 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2050 "numbers have changed\n");
2051 journal_dev = new_decode_dev(journal_devnum);
2053 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2055 really_read_only = bdev_read_only(sb->s_bdev);
2058 * Are we loading a blank journal or performing recovery after a
2059 * crash? For recovery, we need to check in advance whether we
2060 * can get read-write access to the device.
2063 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2064 if (sb->s_flags & MS_RDONLY) {
2065 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2066 "required on readonly filesystem.\n");
2067 if (really_read_only) {
2068 printk(KERN_ERR "EXT4-fs: write access "
2069 "unavailable, cannot proceed.\n");
2072 printk (KERN_INFO "EXT4-fs: write access will "
2073 "be enabled during recovery.\n");
2077 if (journal_inum && journal_dev) {
2078 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2079 "and inode journals!\n");
2084 if (!(journal = ext4_get_journal(sb, journal_inum)))
2087 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2091 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2092 err = jbd2_journal_update_format(journal);
2094 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2095 jbd2_journal_destroy(journal);
2100 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2101 err = jbd2_journal_wipe(journal, !really_read_only);
2103 err = jbd2_journal_load(journal);
2106 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2107 jbd2_journal_destroy(journal);
2111 EXT4_SB(sb)->s_journal = journal;
2112 ext4_clear_journal_err(sb, es);
2114 if (journal_devnum &&
2115 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2116 es->s_journal_dev = cpu_to_le32(journal_devnum);
2119 /* Make sure we flush the recovery flag to disk. */
2120 ext4_commit_super(sb, es, 1);
2126 static int ext4_create_journal(struct super_block * sb,
2127 struct ext4_super_block * es,
2128 unsigned int journal_inum)
2132 if (sb->s_flags & MS_RDONLY) {
2133 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2134 "create journal.\n");
2138 if (!(journal = ext4_get_journal(sb, journal_inum)))
2141 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2144 if (jbd2_journal_create(journal)) {
2145 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2146 jbd2_journal_destroy(journal);
2150 EXT4_SB(sb)->s_journal = journal;
2152 ext4_update_dynamic_rev(sb);
2153 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2154 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2156 es->s_journal_inum = cpu_to_le32(journal_inum);
2159 /* Make sure we flush the recovery flag to disk. */
2160 ext4_commit_super(sb, es, 1);
2165 static void ext4_commit_super (struct super_block * sb,
2166 struct ext4_super_block * es,
2169 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2173 es->s_wtime = cpu_to_le32(get_seconds());
2174 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2175 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2176 BUFFER_TRACE(sbh, "marking dirty");
2177 mark_buffer_dirty(sbh);
2179 sync_dirty_buffer(sbh);
2184 * Have we just finished recovery? If so, and if we are mounting (or
2185 * remounting) the filesystem readonly, then we will end up with a
2186 * consistent fs on disk. Record that fact.
2188 static void ext4_mark_recovery_complete(struct super_block * sb,
2189 struct ext4_super_block * es)
2191 journal_t *journal = EXT4_SB(sb)->s_journal;
2193 jbd2_journal_lock_updates(journal);
2194 jbd2_journal_flush(journal);
2195 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2196 sb->s_flags & MS_RDONLY) {
2197 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2199 ext4_commit_super(sb, es, 1);
2201 jbd2_journal_unlock_updates(journal);
2205 * If we are mounting (or read-write remounting) a filesystem whose journal
2206 * has recorded an error from a previous lifetime, move that error to the
2207 * main filesystem now.
2209 static void ext4_clear_journal_err(struct super_block * sb,
2210 struct ext4_super_block * es)
2216 journal = EXT4_SB(sb)->s_journal;
2219 * Now check for any error status which may have been recorded in the
2220 * journal by a prior ext4_error() or ext4_abort()
2223 j_errno = jbd2_journal_errno(journal);
2227 errstr = ext4_decode_error(sb, j_errno, nbuf);
2228 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2229 "from previous mount: %s", errstr);
2230 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2231 "filesystem check.");
2233 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2234 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2235 ext4_commit_super (sb, es, 1);
2237 jbd2_journal_clear_err(journal);
2242 * Force the running and committing transactions to commit,
2243 * and wait on the commit.
2245 int ext4_force_commit(struct super_block *sb)
2250 if (sb->s_flags & MS_RDONLY)
2253 journal = EXT4_SB(sb)->s_journal;
2255 ret = ext4_journal_force_commit(journal);
2260 * Ext4 always journals updates to the superblock itself, so we don't
2261 * have to propagate any other updates to the superblock on disk at this
2262 * point. Just start an async writeback to get the buffers on their way
2265 * This implicitly triggers the writebehind on sync().
2268 static void ext4_write_super (struct super_block * sb)
2270 if (mutex_trylock(&sb->s_lock) != 0)
2275 static int ext4_sync_fs(struct super_block *sb, int wait)
2280 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2282 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2288 * LVM calls this function before a (read-only) snapshot is created. This
2289 * gives us a chance to flush the journal completely and mark the fs clean.
2291 static void ext4_write_super_lockfs(struct super_block *sb)
2295 if (!(sb->s_flags & MS_RDONLY)) {
2296 journal_t *journal = EXT4_SB(sb)->s_journal;
2298 /* Now we set up the journal barrier. */
2299 jbd2_journal_lock_updates(journal);
2300 jbd2_journal_flush(journal);
2302 /* Journal blocked and flushed, clear needs_recovery flag. */
2303 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2304 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2309 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2310 * flag here, even though the filesystem is not technically dirty yet.
2312 static void ext4_unlockfs(struct super_block *sb)
2314 if (!(sb->s_flags & MS_RDONLY)) {
2316 /* Reser the needs_recovery flag before the fs is unlocked. */
2317 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2318 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2320 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2324 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2326 struct ext4_super_block * es;
2327 struct ext4_sb_info *sbi = EXT4_SB(sb);
2328 ext4_fsblk_t n_blocks_count = 0;
2329 unsigned long old_sb_flags;
2330 struct ext4_mount_options old_opts;
2336 /* Store the original options */
2337 old_sb_flags = sb->s_flags;
2338 old_opts.s_mount_opt = sbi->s_mount_opt;
2339 old_opts.s_resuid = sbi->s_resuid;
2340 old_opts.s_resgid = sbi->s_resgid;
2341 old_opts.s_commit_interval = sbi->s_commit_interval;
2343 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2344 for (i = 0; i < MAXQUOTAS; i++)
2345 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2349 * Allow the "check" option to be passed as a remount option.
2351 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2356 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2357 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2359 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2360 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2364 ext4_init_journal_params(sb, sbi->s_journal);
2366 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2367 n_blocks_count > ext4_blocks_count(es)) {
2368 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2373 if (*flags & MS_RDONLY) {
2375 * First of all, the unconditional stuff we have to do
2376 * to disable replay of the journal when we next remount
2378 sb->s_flags |= MS_RDONLY;
2381 * OK, test if we are remounting a valid rw partition
2382 * readonly, and if so set the rdonly flag and then
2383 * mark the partition as valid again.
2385 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2386 (sbi->s_mount_state & EXT4_VALID_FS))
2387 es->s_state = cpu_to_le16(sbi->s_mount_state);
2389 ext4_mark_recovery_complete(sb, es);
2392 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2393 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2394 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2395 "remount RDWR because of unsupported "
2396 "optional features (%x).\n",
2397 sb->s_id, le32_to_cpu(ret));
2402 * Mounting a RDONLY partition read-write, so reread
2403 * and store the current valid flag. (It may have
2404 * been changed by e2fsck since we originally mounted
2407 ext4_clear_journal_err(sb, es);
2408 sbi->s_mount_state = le16_to_cpu(es->s_state);
2409 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2411 if (!ext4_setup_super (sb, es, 0))
2412 sb->s_flags &= ~MS_RDONLY;
2416 /* Release old quota file names */
2417 for (i = 0; i < MAXQUOTAS; i++)
2418 if (old_opts.s_qf_names[i] &&
2419 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2420 kfree(old_opts.s_qf_names[i]);
2424 sb->s_flags = old_sb_flags;
2425 sbi->s_mount_opt = old_opts.s_mount_opt;
2426 sbi->s_resuid = old_opts.s_resuid;
2427 sbi->s_resgid = old_opts.s_resgid;
2428 sbi->s_commit_interval = old_opts.s_commit_interval;
2430 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2431 for (i = 0; i < MAXQUOTAS; i++) {
2432 if (sbi->s_qf_names[i] &&
2433 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2434 kfree(sbi->s_qf_names[i]);
2435 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2441 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2443 struct super_block *sb = dentry->d_sb;
2444 struct ext4_sb_info *sbi = EXT4_SB(sb);
2445 struct ext4_super_block *es = sbi->s_es;
2446 ext4_fsblk_t overhead;
2449 if (test_opt (sb, MINIX_DF))
2452 unsigned long ngroups;
2453 ngroups = EXT4_SB(sb)->s_groups_count;
2457 * Compute the overhead (FS structures)
2461 * All of the blocks before first_data_block are
2464 overhead = le32_to_cpu(es->s_first_data_block);
2467 * Add the overhead attributed to the superblock and
2468 * block group descriptors. If the sparse superblocks
2469 * feature is turned on, then not all groups have this.
2471 for (i = 0; i < ngroups; i++) {
2472 overhead += ext4_bg_has_super(sb, i) +
2473 ext4_bg_num_gdb(sb, i);
2478 * Every block group has an inode bitmap, a block
2479 * bitmap, and an inode table.
2481 overhead += (ngroups * (2 + EXT4_SB(sb)->s_itb_per_group));
2484 buf->f_type = EXT4_SUPER_MAGIC;
2485 buf->f_bsize = sb->s_blocksize;
2486 buf->f_blocks = ext4_blocks_count(es) - overhead;
2487 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
2488 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2489 if (buf->f_bfree < ext4_r_blocks_count(es))
2491 buf->f_files = le32_to_cpu(es->s_inodes_count);
2492 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
2493 buf->f_namelen = EXT4_NAME_LEN;
2497 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2498 * is locked for write. Otherwise the are possible deadlocks:
2499 * Process 1 Process 2
2500 * ext4_create() quota_sync()
2501 * jbd2_journal_start() write_dquot()
2502 * DQUOT_INIT() down(dqio_mutex)
2503 * down(dqio_mutex) jbd2_journal_start()
2509 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2511 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2514 static int ext4_dquot_initialize(struct inode *inode, int type)
2519 /* We may create quota structure so we need to reserve enough blocks */
2520 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
2522 return PTR_ERR(handle);
2523 ret = dquot_initialize(inode, type);
2524 err = ext4_journal_stop(handle);
2530 static int ext4_dquot_drop(struct inode *inode)
2535 /* We may delete quota structure so we need to reserve enough blocks */
2536 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2538 return PTR_ERR(handle);
2539 ret = dquot_drop(inode);
2540 err = ext4_journal_stop(handle);
2546 static int ext4_write_dquot(struct dquot *dquot)
2550 struct inode *inode;
2552 inode = dquot_to_inode(dquot);
2553 handle = ext4_journal_start(inode,
2554 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2556 return PTR_ERR(handle);
2557 ret = dquot_commit(dquot);
2558 err = ext4_journal_stop(handle);
2564 static int ext4_acquire_dquot(struct dquot *dquot)
2569 handle = ext4_journal_start(dquot_to_inode(dquot),
2570 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2572 return PTR_ERR(handle);
2573 ret = dquot_acquire(dquot);
2574 err = ext4_journal_stop(handle);
2580 static int ext4_release_dquot(struct dquot *dquot)
2585 handle = ext4_journal_start(dquot_to_inode(dquot),
2586 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2588 return PTR_ERR(handle);
2589 ret = dquot_release(dquot);
2590 err = ext4_journal_stop(handle);
2596 static int ext4_mark_dquot_dirty(struct dquot *dquot)
2598 /* Are we journalling quotas? */
2599 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2600 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2601 dquot_mark_dquot_dirty(dquot);
2602 return ext4_write_dquot(dquot);
2604 return dquot_mark_dquot_dirty(dquot);
2608 static int ext4_write_info(struct super_block *sb, int type)
2613 /* Data block + inode block */
2614 handle = ext4_journal_start(sb->s_root->d_inode, 2);
2616 return PTR_ERR(handle);
2617 ret = dquot_commit_info(sb, type);
2618 err = ext4_journal_stop(handle);
2625 * Turn on quotas during mount time - we need to find
2626 * the quota file and such...
2628 static int ext4_quota_on_mount(struct super_block *sb, int type)
2630 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2631 EXT4_SB(sb)->s_jquota_fmt, type);
2635 * Standard function to be called on quota_on
2637 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
2641 struct nameidata nd;
2643 if (!test_opt(sb, QUOTA))
2645 /* Not journalling quota? */
2646 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2647 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
2648 return vfs_quota_on(sb, type, format_id, path);
2649 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2652 /* Quotafile not on the same filesystem? */
2653 if (nd.mnt->mnt_sb != sb) {
2657 /* Quotafile not of fs root? */
2658 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2660 "EXT4-fs: Quota file not on filesystem root. "
2661 "Journalled quota will not work.\n");
2663 return vfs_quota_on(sb, type, format_id, path);
2666 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2667 * acquiring the locks... As quota files are never truncated and quota code
2668 * itself serializes the operations (and noone else should touch the files)
2669 * we don't have to be afraid of races */
2670 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
2671 size_t len, loff_t off)
2673 struct inode *inode = sb_dqopt(sb)->files[type];
2674 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2676 int offset = off & (sb->s_blocksize - 1);
2679 struct buffer_head *bh;
2680 loff_t i_size = i_size_read(inode);
2684 if (off+len > i_size)
2687 while (toread > 0) {
2688 tocopy = sb->s_blocksize - offset < toread ?
2689 sb->s_blocksize - offset : toread;
2690 bh = ext4_bread(NULL, inode, blk, 0, &err);
2693 if (!bh) /* A hole? */
2694 memset(data, 0, tocopy);
2696 memcpy(data, bh->b_data+offset, tocopy);
2706 /* Write to quotafile (we know the transaction is already started and has
2707 * enough credits) */
2708 static ssize_t ext4_quota_write(struct super_block *sb, int type,
2709 const char *data, size_t len, loff_t off)
2711 struct inode *inode = sb_dqopt(sb)->files[type];
2712 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2714 int offset = off & (sb->s_blocksize - 1);
2716 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
2717 size_t towrite = len;
2718 struct buffer_head *bh;
2719 handle_t *handle = journal_current_handle();
2721 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2722 while (towrite > 0) {
2723 tocopy = sb->s_blocksize - offset < towrite ?
2724 sb->s_blocksize - offset : towrite;
2725 bh = ext4_bread(handle, inode, blk, 1, &err);
2728 if (journal_quota) {
2729 err = ext4_journal_get_write_access(handle, bh);
2736 memcpy(bh->b_data+offset, data, tocopy);
2737 flush_dcache_page(bh->b_page);
2740 err = ext4_journal_dirty_metadata(handle, bh);
2742 /* Always do at least ordered writes for quotas */
2743 err = ext4_journal_dirty_data(handle, bh);
2744 mark_buffer_dirty(bh);
2757 if (inode->i_size < off+len-towrite) {
2758 i_size_write(inode, off+len-towrite);
2759 EXT4_I(inode)->i_disksize = inode->i_size;
2762 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2763 ext4_mark_inode_dirty(handle, inode);
2764 mutex_unlock(&inode->i_mutex);
2765 return len - towrite;
2770 static int ext4_get_sb(struct file_system_type *fs_type,
2771 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2773 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
2776 static struct file_system_type ext4dev_fs_type = {
2777 .owner = THIS_MODULE,
2779 .get_sb = ext4_get_sb,
2780 .kill_sb = kill_block_super,
2781 .fs_flags = FS_REQUIRES_DEV,
2784 static int __init init_ext4_fs(void)
2786 int err = init_ext4_xattr();
2789 err = init_inodecache();
2792 err = register_filesystem(&ext4dev_fs_type);
2797 destroy_inodecache();
2803 static void __exit exit_ext4_fs(void)
2805 unregister_filesystem(&ext4dev_fs_type);
2806 destroy_inodecache();
2810 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2811 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
2812 MODULE_LICENSE("GPL");
2813 module_init(init_ext4_fs)
2814 module_exit(exit_ext4_fs)