2 * linux/fs/ext3/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/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>
38 #include <linux/log2.h>
40 #include <asm/uaccess.h>
46 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
47 unsigned long journal_devnum);
48 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
50 static void ext3_commit_super (struct super_block * sb,
51 struct ext3_super_block * es,
53 static void ext3_mark_recovery_complete(struct super_block * sb,
54 struct ext3_super_block * es);
55 static void ext3_clear_journal_err(struct super_block * sb,
56 struct ext3_super_block * es);
57 static int ext3_sync_fs(struct super_block *sb, int wait);
58 static const char *ext3_decode_error(struct super_block * sb, int errno,
60 static int ext3_remount (struct super_block * sb, int * flags, char * data);
61 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
62 static void ext3_unlockfs(struct super_block *sb);
63 static void ext3_write_super (struct super_block * sb);
64 static void ext3_write_super_lockfs(struct super_block *sb);
67 * Wrappers for journal_start/end.
69 * The only special thing we need to do here is to make sure that all
70 * journal_end calls result in the superblock being marked dirty, so
71 * that sync() will call the filesystem's write_super callback if
74 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
78 if (sb->s_flags & MS_RDONLY)
79 return ERR_PTR(-EROFS);
81 /* Special case here: if the journal has aborted behind our
82 * backs (eg. EIO in the commit thread), then we still need to
83 * take the FS itself readonly cleanly. */
84 journal = EXT3_SB(sb)->s_journal;
85 if (is_journal_aborted(journal)) {
86 ext3_abort(sb, __FUNCTION__,
87 "Detected aborted journal");
88 return ERR_PTR(-EROFS);
91 return journal_start(journal, nblocks);
95 * The only special thing we need to do here is to make sure that all
96 * journal_stop calls result in the superblock being marked dirty, so
97 * that sync() will call the filesystem's write_super callback if
100 int __ext3_journal_stop(const char *where, handle_t *handle)
102 struct super_block *sb;
106 sb = handle->h_transaction->t_journal->j_private;
108 rc = journal_stop(handle);
113 __ext3_std_error(sb, where, err);
117 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
118 struct buffer_head *bh, handle_t *handle, int err)
121 const char *errstr = ext3_decode_error(NULL, err, nbuf);
124 BUFFER_TRACE(bh, "abort");
129 if (is_handle_aborted(handle))
132 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
133 caller, errstr, err_fn);
135 journal_abort_handle(handle);
138 /* Deal with the reporting of failure conditions on a filesystem such as
139 * inconsistencies detected or read IO failures.
141 * On ext2, we can store the error state of the filesystem in the
142 * superblock. That is not possible on ext3, because we may have other
143 * write ordering constraints on the superblock which prevent us from
144 * writing it out straight away; and given that the journal is about to
145 * be aborted, we can't rely on the current, or future, transactions to
146 * write out the superblock safely.
148 * We'll just use the journal_abort() error code to record an error in
149 * the journal instead. On recovery, the journal will compain about
150 * that error until we've noted it down and cleared it.
153 static void ext3_handle_error(struct super_block *sb)
155 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
158 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160 if (sb->s_flags & MS_RDONLY)
163 if (!test_opt (sb, ERRORS_CONT)) {
164 journal_t *journal = EXT3_SB(sb)->s_journal;
166 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
168 journal_abort(journal, -EIO);
170 if (test_opt (sb, ERRORS_RO)) {
171 printk (KERN_CRIT "Remounting filesystem read-only\n");
172 sb->s_flags |= MS_RDONLY;
174 ext3_commit_super(sb, es, 1);
175 if (test_opt(sb, ERRORS_PANIC))
176 panic("EXT3-fs (device %s): panic forced after error\n",
180 void ext3_error (struct super_block * sb, const char * function,
181 const char * fmt, ...)
186 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
191 ext3_handle_error(sb);
194 static const char *ext3_decode_error(struct super_block * sb, int errno,
201 errstr = "IO failure";
204 errstr = "Out of memory";
207 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
208 errstr = "Journal has aborted";
210 errstr = "Readonly filesystem";
213 /* If the caller passed in an extra buffer for unknown
214 * errors, textualise them now. Else we just return
217 /* Check for truncated error codes... */
218 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
227 /* __ext3_std_error decodes expected errors from journaling functions
228 * automatically and invokes the appropriate error response. */
230 void __ext3_std_error (struct super_block * sb, const char * function,
236 /* Special case: if the error is EROFS, and we're not already
237 * inside a transaction, then there's really no point in logging
239 if (errno == -EROFS && journal_current_handle() == NULL &&
240 (sb->s_flags & MS_RDONLY))
243 errstr = ext3_decode_error(sb, errno, nbuf);
244 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
245 sb->s_id, function, errstr);
247 ext3_handle_error(sb);
251 * ext3_abort is a much stronger failure handler than ext3_error. The
252 * abort function may be used to deal with unrecoverable failures such
253 * as journal IO errors or ENOMEM at a critical moment in log management.
255 * We unconditionally force the filesystem into an ABORT|READONLY state,
256 * unless the error response on the fs has been set to panic in which
257 * case we take the easy way out and panic immediately.
260 void ext3_abort (struct super_block * sb, const char * function,
261 const char * fmt, ...)
265 printk (KERN_CRIT "ext3_abort called.\n");
268 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
273 if (test_opt(sb, ERRORS_PANIC))
274 panic("EXT3-fs panic from previous error\n");
276 if (sb->s_flags & MS_RDONLY)
279 printk(KERN_CRIT "Remounting filesystem read-only\n");
280 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
281 sb->s_flags |= MS_RDONLY;
282 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
283 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
286 void ext3_warning (struct super_block * sb, const char * function,
287 const char * fmt, ...)
292 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
299 void ext3_update_dynamic_rev(struct super_block *sb)
301 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
303 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
306 ext3_warning(sb, __FUNCTION__,
307 "updating to rev %d because of new feature flag, "
308 "running e2fsck is recommended",
311 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
312 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
313 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
314 /* leave es->s_feature_*compat flags alone */
315 /* es->s_uuid will be set by e2fsck if empty */
318 * The rest of the superblock fields should be zero, and if not it
319 * means they are likely already in use, so leave them alone. We
320 * can leave it up to e2fsck to clean up any inconsistencies there.
325 * Open the external journal device
327 static struct block_device *ext3_blkdev_get(dev_t dev)
329 struct block_device *bdev;
330 char b[BDEVNAME_SIZE];
332 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
338 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
339 __bdevname(dev, b), PTR_ERR(bdev));
344 * Release the journal device
346 static int ext3_blkdev_put(struct block_device *bdev)
349 return blkdev_put(bdev);
352 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
354 struct block_device *bdev;
357 bdev = sbi->journal_bdev;
359 ret = ext3_blkdev_put(bdev);
360 sbi->journal_bdev = NULL;
365 static inline struct inode *orphan_list_entry(struct list_head *l)
367 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
370 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
374 printk(KERN_ERR "sb orphan head is %d\n",
375 le32_to_cpu(sbi->s_es->s_last_orphan));
377 printk(KERN_ERR "sb_info orphan list:\n");
378 list_for_each(l, &sbi->s_orphan) {
379 struct inode *inode = orphan_list_entry(l);
381 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
382 inode->i_sb->s_id, inode->i_ino, inode,
383 inode->i_mode, inode->i_nlink,
388 static void ext3_put_super (struct super_block * sb)
390 struct ext3_sb_info *sbi = EXT3_SB(sb);
391 struct ext3_super_block *es = sbi->s_es;
394 ext3_xattr_put_super(sb);
395 journal_destroy(sbi->s_journal);
396 if (!(sb->s_flags & MS_RDONLY)) {
397 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
398 es->s_state = cpu_to_le16(sbi->s_mount_state);
399 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
400 mark_buffer_dirty(sbi->s_sbh);
401 ext3_commit_super(sb, es, 1);
404 for (i = 0; i < sbi->s_gdb_count; i++)
405 brelse(sbi->s_group_desc[i]);
406 kfree(sbi->s_group_desc);
407 percpu_counter_destroy(&sbi->s_freeblocks_counter);
408 percpu_counter_destroy(&sbi->s_freeinodes_counter);
409 percpu_counter_destroy(&sbi->s_dirs_counter);
412 for (i = 0; i < MAXQUOTAS; i++)
413 kfree(sbi->s_qf_names[i]);
416 /* Debugging code just in case the in-memory inode orphan list
417 * isn't empty. The on-disk one can be non-empty if we've
418 * detected an error and taken the fs readonly, but the
419 * in-memory list had better be clean by this point. */
420 if (!list_empty(&sbi->s_orphan))
421 dump_orphan_list(sb, sbi);
422 J_ASSERT(list_empty(&sbi->s_orphan));
424 invalidate_bdev(sb->s_bdev);
425 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
427 * Invalidate the journal device's buffers. We don't want them
428 * floating about in memory - the physical journal device may
429 * hotswapped, and it breaks the `ro-after' testing code.
431 sync_blockdev(sbi->journal_bdev);
432 invalidate_bdev(sbi->journal_bdev);
433 ext3_blkdev_remove(sbi);
435 sb->s_fs_info = NULL;
440 static struct kmem_cache *ext3_inode_cachep;
443 * Called inside transaction, so use GFP_NOFS
445 static struct inode *ext3_alloc_inode(struct super_block *sb)
447 struct ext3_inode_info *ei;
449 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
452 #ifdef CONFIG_EXT3_FS_POSIX_ACL
453 ei->i_acl = EXT3_ACL_NOT_CACHED;
454 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
456 ei->i_block_alloc_info = NULL;
457 ei->vfs_inode.i_version = 1;
458 return &ei->vfs_inode;
461 static void ext3_destroy_inode(struct inode *inode)
463 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
464 printk("EXT3 Inode %p: orphan list check failed!\n",
466 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
467 EXT3_I(inode), sizeof(struct ext3_inode_info),
471 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
474 static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags)
476 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
478 INIT_LIST_HEAD(&ei->i_orphan);
479 #ifdef CONFIG_EXT3_FS_XATTR
480 init_rwsem(&ei->xattr_sem);
482 mutex_init(&ei->truncate_mutex);
483 inode_init_once(&ei->vfs_inode);
486 static int init_inodecache(void)
488 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
489 sizeof(struct ext3_inode_info),
490 0, (SLAB_RECLAIM_ACCOUNT|
493 if (ext3_inode_cachep == NULL)
498 static void destroy_inodecache(void)
500 kmem_cache_destroy(ext3_inode_cachep);
503 static void ext3_clear_inode(struct inode *inode)
505 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
506 #ifdef CONFIG_EXT3_FS_POSIX_ACL
507 if (EXT3_I(inode)->i_acl &&
508 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
509 posix_acl_release(EXT3_I(inode)->i_acl);
510 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
512 if (EXT3_I(inode)->i_default_acl &&
513 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
514 posix_acl_release(EXT3_I(inode)->i_default_acl);
515 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
518 ext3_discard_reservation(inode);
519 EXT3_I(inode)->i_block_alloc_info = NULL;
524 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
526 #if defined(CONFIG_QUOTA)
527 struct ext3_sb_info *sbi = EXT3_SB(sb);
529 if (sbi->s_jquota_fmt)
530 seq_printf(seq, ",jqfmt=%s",
531 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
533 if (sbi->s_qf_names[USRQUOTA])
534 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
536 if (sbi->s_qf_names[GRPQUOTA])
537 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
539 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
540 seq_puts(seq, ",usrquota");
542 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
543 seq_puts(seq, ",grpquota");
547 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
549 struct super_block *sb = vfs->mnt_sb;
551 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
552 seq_puts(seq, ",data=journal");
553 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
554 seq_puts(seq, ",data=ordered");
555 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
556 seq_puts(seq, ",data=writeback");
558 ext3_show_quota_options(seq, sb);
564 static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
567 unsigned long ino = objp[0];
568 __u32 generation = objp[1];
570 struct dentry *result;
572 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
573 return ERR_PTR(-ESTALE);
574 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
575 return ERR_PTR(-ESTALE);
577 /* iget isn't really right if the inode is currently unallocated!!
579 * ext3_read_inode will return a bad_inode if the inode had been
580 * deleted, so we should be safe.
582 * Currently we don't know the generation for parent directory, so
583 * a generation of 0 means "accept any"
585 inode = iget(sb, ino);
587 return ERR_PTR(-ENOMEM);
588 if (is_bad_inode(inode) ||
589 (generation && inode->i_generation != generation)) {
591 return ERR_PTR(-ESTALE);
593 /* now to find a dentry.
594 * If possible, get a well-connected one
596 result = d_alloc_anon(inode);
599 return ERR_PTR(-ENOMEM);
605 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
606 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
608 static int ext3_dquot_initialize(struct inode *inode, int type);
609 static int ext3_dquot_drop(struct inode *inode);
610 static int ext3_write_dquot(struct dquot *dquot);
611 static int ext3_acquire_dquot(struct dquot *dquot);
612 static int ext3_release_dquot(struct dquot *dquot);
613 static int ext3_mark_dquot_dirty(struct dquot *dquot);
614 static int ext3_write_info(struct super_block *sb, int type);
615 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
616 static int ext3_quota_on_mount(struct super_block *sb, int type);
617 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
618 size_t len, loff_t off);
619 static ssize_t ext3_quota_write(struct super_block *sb, int type,
620 const char *data, size_t len, loff_t off);
622 static struct dquot_operations ext3_quota_operations = {
623 .initialize = ext3_dquot_initialize,
624 .drop = ext3_dquot_drop,
625 .alloc_space = dquot_alloc_space,
626 .alloc_inode = dquot_alloc_inode,
627 .free_space = dquot_free_space,
628 .free_inode = dquot_free_inode,
629 .transfer = dquot_transfer,
630 .write_dquot = ext3_write_dquot,
631 .acquire_dquot = ext3_acquire_dquot,
632 .release_dquot = ext3_release_dquot,
633 .mark_dirty = ext3_mark_dquot_dirty,
634 .write_info = ext3_write_info
637 static struct quotactl_ops ext3_qctl_operations = {
638 .quota_on = ext3_quota_on,
639 .quota_off = vfs_quota_off,
640 .quota_sync = vfs_quota_sync,
641 .get_info = vfs_get_dqinfo,
642 .set_info = vfs_set_dqinfo,
643 .get_dqblk = vfs_get_dqblk,
644 .set_dqblk = vfs_set_dqblk
648 static const struct super_operations ext3_sops = {
649 .alloc_inode = ext3_alloc_inode,
650 .destroy_inode = ext3_destroy_inode,
651 .read_inode = ext3_read_inode,
652 .write_inode = ext3_write_inode,
653 .dirty_inode = ext3_dirty_inode,
654 .delete_inode = ext3_delete_inode,
655 .put_super = ext3_put_super,
656 .write_super = ext3_write_super,
657 .sync_fs = ext3_sync_fs,
658 .write_super_lockfs = ext3_write_super_lockfs,
659 .unlockfs = ext3_unlockfs,
660 .statfs = ext3_statfs,
661 .remount_fs = ext3_remount,
662 .clear_inode = ext3_clear_inode,
663 .show_options = ext3_show_options,
665 .quota_read = ext3_quota_read,
666 .quota_write = ext3_quota_write,
670 static struct export_operations ext3_export_ops = {
671 .get_parent = ext3_get_parent,
672 .get_dentry = ext3_get_dentry,
676 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
677 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
678 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
679 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
680 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
681 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
682 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
683 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
684 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
685 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
689 static match_table_t tokens = {
690 {Opt_bsd_df, "bsddf"},
691 {Opt_minix_df, "minixdf"},
692 {Opt_grpid, "grpid"},
693 {Opt_grpid, "bsdgroups"},
694 {Opt_nogrpid, "nogrpid"},
695 {Opt_nogrpid, "sysvgroups"},
696 {Opt_resgid, "resgid=%u"},
697 {Opt_resuid, "resuid=%u"},
699 {Opt_err_cont, "errors=continue"},
700 {Opt_err_panic, "errors=panic"},
701 {Opt_err_ro, "errors=remount-ro"},
702 {Opt_nouid32, "nouid32"},
703 {Opt_nocheck, "nocheck"},
704 {Opt_nocheck, "check=none"},
705 {Opt_debug, "debug"},
706 {Opt_oldalloc, "oldalloc"},
707 {Opt_orlov, "orlov"},
708 {Opt_user_xattr, "user_xattr"},
709 {Opt_nouser_xattr, "nouser_xattr"},
711 {Opt_noacl, "noacl"},
712 {Opt_reservation, "reservation"},
713 {Opt_noreservation, "noreservation"},
714 {Opt_noload, "noload"},
717 {Opt_commit, "commit=%u"},
718 {Opt_journal_update, "journal=update"},
719 {Opt_journal_inum, "journal=%u"},
720 {Opt_journal_dev, "journal_dev=%u"},
721 {Opt_abort, "abort"},
722 {Opt_data_journal, "data=journal"},
723 {Opt_data_ordered, "data=ordered"},
724 {Opt_data_writeback, "data=writeback"},
725 {Opt_offusrjquota, "usrjquota="},
726 {Opt_usrjquota, "usrjquota=%s"},
727 {Opt_offgrpjquota, "grpjquota="},
728 {Opt_grpjquota, "grpjquota=%s"},
729 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
730 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
731 {Opt_grpquota, "grpquota"},
732 {Opt_noquota, "noquota"},
733 {Opt_quota, "quota"},
734 {Opt_usrquota, "usrquota"},
735 {Opt_barrier, "barrier=%u"},
737 {Opt_resize, "resize"},
740 static ext3_fsblk_t get_sb_block(void **data)
742 ext3_fsblk_t sb_block;
743 char *options = (char *) *data;
745 if (!options || strncmp(options, "sb=", 3) != 0)
746 return 1; /* Default location */
748 /*todo: use simple_strtoll with >32bit ext3 */
749 sb_block = simple_strtoul(options, &options, 0);
750 if (*options && *options != ',') {
751 printk("EXT3-fs: Invalid sb specification: %s\n",
757 *data = (void *) options;
761 static int parse_options (char *options, struct super_block *sb,
762 unsigned int *inum, unsigned long *journal_devnum,
763 ext3_fsblk_t *n_blocks_count, int is_remount)
765 struct ext3_sb_info *sbi = EXT3_SB(sb);
767 substring_t args[MAX_OPT_ARGS];
778 while ((p = strsep (&options, ",")) != NULL) {
783 token = match_token(p, tokens, args);
786 clear_opt (sbi->s_mount_opt, MINIX_DF);
789 set_opt (sbi->s_mount_opt, MINIX_DF);
792 set_opt (sbi->s_mount_opt, GRPID);
795 clear_opt (sbi->s_mount_opt, GRPID);
798 if (match_int(&args[0], &option))
800 sbi->s_resuid = option;
803 if (match_int(&args[0], &option))
805 sbi->s_resgid = option;
808 /* handled by get_sb_block() instead of here */
809 /* *sb_block = match_int(&args[0]); */
812 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
813 clear_opt (sbi->s_mount_opt, ERRORS_RO);
814 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
817 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
818 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
819 set_opt (sbi->s_mount_opt, ERRORS_RO);
822 clear_opt (sbi->s_mount_opt, ERRORS_RO);
823 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
824 set_opt (sbi->s_mount_opt, ERRORS_CONT);
827 set_opt (sbi->s_mount_opt, NO_UID32);
830 clear_opt (sbi->s_mount_opt, CHECK);
833 set_opt (sbi->s_mount_opt, DEBUG);
836 set_opt (sbi->s_mount_opt, OLDALLOC);
839 clear_opt (sbi->s_mount_opt, OLDALLOC);
841 #ifdef CONFIG_EXT3_FS_XATTR
843 set_opt (sbi->s_mount_opt, XATTR_USER);
845 case Opt_nouser_xattr:
846 clear_opt (sbi->s_mount_opt, XATTR_USER);
850 case Opt_nouser_xattr:
851 printk("EXT3 (no)user_xattr options not supported\n");
854 #ifdef CONFIG_EXT3_FS_POSIX_ACL
856 set_opt(sbi->s_mount_opt, POSIX_ACL);
859 clear_opt(sbi->s_mount_opt, POSIX_ACL);
864 printk("EXT3 (no)acl options not supported\n");
867 case Opt_reservation:
868 set_opt(sbi->s_mount_opt, RESERVATION);
870 case Opt_noreservation:
871 clear_opt(sbi->s_mount_opt, RESERVATION);
873 case Opt_journal_update:
875 /* Eventually we will want to be able to create
876 a journal file here. For now, only allow the
877 user to specify an existing inode to be the
880 printk(KERN_ERR "EXT3-fs: cannot specify "
881 "journal on remount\n");
884 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
886 case Opt_journal_inum:
888 printk(KERN_ERR "EXT3-fs: cannot specify "
889 "journal on remount\n");
892 if (match_int(&args[0], &option))
896 case Opt_journal_dev:
898 printk(KERN_ERR "EXT3-fs: cannot specify "
899 "journal on remount\n");
902 if (match_int(&args[0], &option))
904 *journal_devnum = option;
907 set_opt (sbi->s_mount_opt, NOLOAD);
910 if (match_int(&args[0], &option))
915 option = JBD_DEFAULT_MAX_COMMIT_AGE;
916 sbi->s_commit_interval = HZ * option;
918 case Opt_data_journal:
919 data_opt = EXT3_MOUNT_JOURNAL_DATA;
921 case Opt_data_ordered:
922 data_opt = EXT3_MOUNT_ORDERED_DATA;
924 case Opt_data_writeback:
925 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
928 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
931 "EXT3-fs: cannot change data "
932 "mode on remount\n");
936 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
937 sbi->s_mount_opt |= data_opt;
947 if (sb_any_quota_enabled(sb)) {
949 "EXT3-fs: Cannot change journalled "
950 "quota options when quota turned on.\n");
953 qname = match_strdup(&args[0]);
956 "EXT3-fs: not enough memory for "
957 "storing quotafile name.\n");
960 if (sbi->s_qf_names[qtype] &&
961 strcmp(sbi->s_qf_names[qtype], qname)) {
963 "EXT3-fs: %s quota file already "
964 "specified.\n", QTYPE2NAME(qtype));
968 sbi->s_qf_names[qtype] = qname;
969 if (strchr(sbi->s_qf_names[qtype], '/')) {
971 "EXT3-fs: quotafile must be on "
972 "filesystem root.\n");
973 kfree(sbi->s_qf_names[qtype]);
974 sbi->s_qf_names[qtype] = NULL;
977 set_opt(sbi->s_mount_opt, QUOTA);
979 case Opt_offusrjquota:
982 case Opt_offgrpjquota:
985 if (sb_any_quota_enabled(sb)) {
986 printk(KERN_ERR "EXT3-fs: Cannot change "
987 "journalled quota options when "
988 "quota turned on.\n");
992 * The space will be released later when all options
993 * are confirmed to be correct
995 sbi->s_qf_names[qtype] = NULL;
997 case Opt_jqfmt_vfsold:
998 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1000 case Opt_jqfmt_vfsv0:
1001 sbi->s_jquota_fmt = QFMT_VFS_V0;
1005 set_opt(sbi->s_mount_opt, QUOTA);
1006 set_opt(sbi->s_mount_opt, USRQUOTA);
1009 set_opt(sbi->s_mount_opt, QUOTA);
1010 set_opt(sbi->s_mount_opt, GRPQUOTA);
1013 if (sb_any_quota_enabled(sb)) {
1014 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1015 "options when quota turned on.\n");
1018 clear_opt(sbi->s_mount_opt, QUOTA);
1019 clear_opt(sbi->s_mount_opt, USRQUOTA);
1020 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1028 case Opt_offusrjquota:
1029 case Opt_offgrpjquota:
1030 case Opt_jqfmt_vfsold:
1031 case Opt_jqfmt_vfsv0:
1033 "EXT3-fs: journalled quota options not "
1040 set_opt(sbi->s_mount_opt, ABORT);
1043 if (match_int(&args[0], &option))
1046 set_opt(sbi->s_mount_opt, BARRIER);
1048 clear_opt(sbi->s_mount_opt, BARRIER);
1054 printk("EXT3-fs: resize option only available "
1058 if (match_int(&args[0], &option) != 0)
1060 *n_blocks_count = option;
1063 set_opt(sbi->s_mount_opt, NOBH);
1066 clear_opt(sbi->s_mount_opt, NOBH);
1070 "EXT3-fs: Unrecognized mount option \"%s\" "
1071 "or missing value\n", p);
1076 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1077 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1078 sbi->s_qf_names[USRQUOTA])
1079 clear_opt(sbi->s_mount_opt, USRQUOTA);
1081 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1082 sbi->s_qf_names[GRPQUOTA])
1083 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1085 if ((sbi->s_qf_names[USRQUOTA] &&
1086 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1087 (sbi->s_qf_names[GRPQUOTA] &&
1088 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1089 printk(KERN_ERR "EXT3-fs: old and new quota "
1090 "format mixing.\n");
1094 if (!sbi->s_jquota_fmt) {
1095 printk(KERN_ERR "EXT3-fs: journalled quota format "
1096 "not specified.\n");
1100 if (sbi->s_jquota_fmt) {
1101 printk(KERN_ERR "EXT3-fs: journalled quota format "
1102 "specified with no journalling "
1111 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1114 struct ext3_sb_info *sbi = EXT3_SB(sb);
1117 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1118 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1119 "forcing read-only mode\n");
1124 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1125 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1126 "running e2fsck is recommended\n");
1127 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1128 printk (KERN_WARNING
1129 "EXT3-fs warning: mounting fs with errors, "
1130 "running e2fsck is recommended\n");
1131 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1132 le16_to_cpu(es->s_mnt_count) >=
1133 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1134 printk (KERN_WARNING
1135 "EXT3-fs warning: maximal mount count reached, "
1136 "running e2fsck is recommended\n");
1137 else if (le32_to_cpu(es->s_checkinterval) &&
1138 (le32_to_cpu(es->s_lastcheck) +
1139 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1140 printk (KERN_WARNING
1141 "EXT3-fs warning: checktime reached, "
1142 "running e2fsck is recommended\n");
1144 /* @@@ We _will_ want to clear the valid bit if we find
1145 inconsistencies, to force a fsck at reboot. But for
1146 a plain journaled filesystem we can keep it set as
1147 valid forever! :) */
1148 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1150 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1151 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1152 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1153 es->s_mtime = cpu_to_le32(get_seconds());
1154 ext3_update_dynamic_rev(sb);
1155 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1157 ext3_commit_super(sb, es, 1);
1158 if (test_opt(sb, DEBUG))
1159 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1160 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1162 sbi->s_groups_count,
1163 EXT3_BLOCKS_PER_GROUP(sb),
1164 EXT3_INODES_PER_GROUP(sb),
1167 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1168 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1169 char b[BDEVNAME_SIZE];
1171 printk("external journal on %s\n",
1172 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1174 printk("internal journal\n");
1179 /* Called at mount-time, super-block is locked */
1180 static int ext3_check_descriptors (struct super_block * sb)
1182 struct ext3_sb_info *sbi = EXT3_SB(sb);
1183 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1184 ext3_fsblk_t last_block;
1185 struct ext3_group_desc * gdp = NULL;
1189 ext3_debug ("Checking group descriptors");
1191 for (i = 0; i < sbi->s_groups_count; i++)
1193 if (i == sbi->s_groups_count - 1)
1194 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1196 last_block = first_block +
1197 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1199 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1200 gdp = (struct ext3_group_desc *)
1201 sbi->s_group_desc[desc_block++]->b_data;
1202 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1203 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1205 ext3_error (sb, "ext3_check_descriptors",
1206 "Block bitmap for group %d"
1207 " not in group (block %lu)!",
1209 le32_to_cpu(gdp->bg_block_bitmap));
1212 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1213 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1215 ext3_error (sb, "ext3_check_descriptors",
1216 "Inode bitmap for group %d"
1217 " not in group (block %lu)!",
1219 le32_to_cpu(gdp->bg_inode_bitmap));
1222 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1223 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >
1226 ext3_error (sb, "ext3_check_descriptors",
1227 "Inode table for group %d"
1228 " not in group (block %lu)!",
1230 le32_to_cpu(gdp->bg_inode_table));
1233 first_block += EXT3_BLOCKS_PER_GROUP(sb);
1237 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1238 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1243 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1244 * the superblock) which were deleted from all directories, but held open by
1245 * a process at the time of a crash. We walk the list and try to delete these
1246 * inodes at recovery time (only with a read-write filesystem).
1248 * In order to keep the orphan inode chain consistent during traversal (in
1249 * case of crash during recovery), we link each inode into the superblock
1250 * orphan list_head and handle it the same way as an inode deletion during
1251 * normal operation (which journals the operations for us).
1253 * We only do an iget() and an iput() on each inode, which is very safe if we
1254 * accidentally point at an in-use or already deleted inode. The worst that
1255 * can happen in this case is that we get a "bit already cleared" message from
1256 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1257 * e2fsck was run on this filesystem, and it must have already done the orphan
1258 * inode cleanup for us, so we can safely abort without any further action.
1260 static void ext3_orphan_cleanup (struct super_block * sb,
1261 struct ext3_super_block * es)
1263 unsigned int s_flags = sb->s_flags;
1264 int nr_orphans = 0, nr_truncates = 0;
1268 if (!es->s_last_orphan) {
1269 jbd_debug(4, "no orphan inodes to clean up\n");
1273 if (bdev_read_only(sb->s_bdev)) {
1274 printk(KERN_ERR "EXT3-fs: write access "
1275 "unavailable, skipping orphan cleanup.\n");
1279 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1280 if (es->s_last_orphan)
1281 jbd_debug(1, "Errors on filesystem, "
1282 "clearing orphan list.\n");
1283 es->s_last_orphan = 0;
1284 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1288 if (s_flags & MS_RDONLY) {
1289 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1291 sb->s_flags &= ~MS_RDONLY;
1294 /* Needed for iput() to work correctly and not trash data */
1295 sb->s_flags |= MS_ACTIVE;
1296 /* Turn on quotas so that they are updated correctly */
1297 for (i = 0; i < MAXQUOTAS; i++) {
1298 if (EXT3_SB(sb)->s_qf_names[i]) {
1299 int ret = ext3_quota_on_mount(sb, i);
1302 "EXT3-fs: Cannot turn on journalled "
1303 "quota: error %d\n", ret);
1308 while (es->s_last_orphan) {
1309 struct inode *inode;
1312 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1313 es->s_last_orphan = 0;
1317 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1319 if (inode->i_nlink) {
1321 "%s: truncating inode %lu to %Ld bytes\n",
1322 __FUNCTION__, inode->i_ino, inode->i_size);
1323 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1324 inode->i_ino, inode->i_size);
1325 ext3_truncate(inode);
1329 "%s: deleting unreferenced inode %lu\n",
1330 __FUNCTION__, inode->i_ino);
1331 jbd_debug(2, "deleting unreferenced inode %lu\n",
1335 iput(inode); /* The delete magic happens here! */
1338 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1341 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1342 sb->s_id, PLURAL(nr_orphans));
1344 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1345 sb->s_id, PLURAL(nr_truncates));
1347 /* Turn quotas off */
1348 for (i = 0; i < MAXQUOTAS; i++) {
1349 if (sb_dqopt(sb)->files[i])
1350 vfs_quota_off(sb, i);
1353 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1357 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1358 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1359 * We need to be 1 filesystem block less than the 2^32 sector limit.
1361 static loff_t ext3_max_size(int bits)
1363 loff_t res = EXT3_NDIR_BLOCKS;
1364 /* This constant is calculated to be the largest file size for a
1365 * dense, 4k-blocksize file such that the total number of
1366 * sectors in the file, including data and all indirect blocks,
1367 * does not exceed 2^32. */
1368 const loff_t upper_limit = 0x1ff7fffd000LL;
1370 res += 1LL << (bits-2);
1371 res += 1LL << (2*(bits-2));
1372 res += 1LL << (3*(bits-2));
1374 if (res > upper_limit)
1379 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1380 ext3_fsblk_t logic_sb_block,
1383 struct ext3_sb_info *sbi = EXT3_SB(sb);
1384 unsigned long bg, first_meta_bg;
1387 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1389 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1391 return (logic_sb_block + nr + 1);
1392 bg = sbi->s_desc_per_block * nr;
1393 if (ext3_bg_has_super(sb, bg))
1395 return (has_super + ext3_group_first_block_no(sb, bg));
1399 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1401 struct buffer_head * bh;
1402 struct ext3_super_block *es = NULL;
1403 struct ext3_sb_info *sbi;
1405 ext3_fsblk_t sb_block = get_sb_block(&data);
1406 ext3_fsblk_t logic_sb_block;
1407 unsigned long offset = 0;
1408 unsigned int journal_inum = 0;
1409 unsigned long journal_devnum = 0;
1410 unsigned long def_mount_opts;
1419 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1422 sb->s_fs_info = sbi;
1423 sbi->s_mount_opt = 0;
1424 sbi->s_resuid = EXT3_DEF_RESUID;
1425 sbi->s_resgid = EXT3_DEF_RESGID;
1429 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1431 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1436 * The ext3 superblock will not be buffer aligned for other than 1kB
1437 * block sizes. We need to calculate the offset from buffer start.
1439 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1440 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1441 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1443 logic_sb_block = sb_block;
1446 if (!(bh = sb_bread(sb, logic_sb_block))) {
1447 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1451 * Note: s_es must be initialized as soon as possible because
1452 * some ext3 macro-instructions depend on its value
1454 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1456 sb->s_magic = le16_to_cpu(es->s_magic);
1457 if (sb->s_magic != EXT3_SUPER_MAGIC)
1460 /* Set defaults before we parse the mount options */
1461 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1462 if (def_mount_opts & EXT3_DEFM_DEBUG)
1463 set_opt(sbi->s_mount_opt, DEBUG);
1464 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1465 set_opt(sbi->s_mount_opt, GRPID);
1466 if (def_mount_opts & EXT3_DEFM_UID16)
1467 set_opt(sbi->s_mount_opt, NO_UID32);
1468 #ifdef CONFIG_EXT3_FS_XATTR
1469 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1470 set_opt(sbi->s_mount_opt, XATTR_USER);
1472 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1473 if (def_mount_opts & EXT3_DEFM_ACL)
1474 set_opt(sbi->s_mount_opt, POSIX_ACL);
1476 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1477 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1478 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1479 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1480 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1481 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1483 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1484 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1485 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1486 set_opt(sbi->s_mount_opt, ERRORS_RO);
1488 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1490 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1491 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1493 set_opt(sbi->s_mount_opt, RESERVATION);
1495 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1499 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1500 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1502 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1503 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1504 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1505 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1507 "EXT3-fs warning: feature flags set on rev 0 fs, "
1508 "running e2fsck is recommended\n");
1510 * Check feature flags regardless of the revision level, since we
1511 * previously didn't change the revision level when setting the flags,
1512 * so there is a chance incompat flags are set on a rev 0 filesystem.
1514 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1516 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1517 "unsupported optional features (%x).\n",
1518 sb->s_id, le32_to_cpu(features));
1521 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1522 if (!(sb->s_flags & MS_RDONLY) && features) {
1523 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1524 "unsupported optional features (%x).\n",
1525 sb->s_id, le32_to_cpu(features));
1528 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1530 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1531 blocksize > EXT3_MAX_BLOCK_SIZE) {
1533 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1534 blocksize, sb->s_id);
1538 hblock = bdev_hardsect_size(sb->s_bdev);
1539 if (sb->s_blocksize != blocksize) {
1541 * Make sure the blocksize for the filesystem is larger
1542 * than the hardware sectorsize for the machine.
1544 if (blocksize < hblock) {
1545 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1546 "device blocksize %d.\n", blocksize, hblock);
1551 sb_set_blocksize(sb, blocksize);
1552 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1553 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1554 bh = sb_bread(sb, logic_sb_block);
1557 "EXT3-fs: Can't read superblock on 2nd try.\n");
1560 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1562 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1564 "EXT3-fs: Magic mismatch, very weird !\n");
1569 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1571 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1572 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1573 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1575 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1576 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1577 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1578 (!is_power_of_2(sbi->s_inode_size)) ||
1579 (sbi->s_inode_size > blocksize)) {
1581 "EXT3-fs: unsupported inode size: %d\n",
1586 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1587 le32_to_cpu(es->s_log_frag_size);
1588 if (blocksize != sbi->s_frag_size) {
1590 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1591 sbi->s_frag_size, blocksize);
1594 sbi->s_frags_per_block = 1;
1595 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1596 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1597 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1598 if (EXT3_INODE_SIZE(sb) == 0)
1600 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1601 if (sbi->s_inodes_per_block == 0)
1603 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1604 sbi->s_inodes_per_block;
1605 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1607 sbi->s_mount_state = le16_to_cpu(es->s_state);
1608 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1609 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1610 for (i=0; i < 4; i++)
1611 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1612 sbi->s_def_hash_version = es->s_def_hash_version;
1614 if (sbi->s_blocks_per_group > blocksize * 8) {
1616 "EXT3-fs: #blocks per group too big: %lu\n",
1617 sbi->s_blocks_per_group);
1620 if (sbi->s_frags_per_group > blocksize * 8) {
1622 "EXT3-fs: #fragments per group too big: %lu\n",
1623 sbi->s_frags_per_group);
1626 if (sbi->s_inodes_per_group > blocksize * 8) {
1628 "EXT3-fs: #inodes per group too big: %lu\n",
1629 sbi->s_inodes_per_group);
1633 if (le32_to_cpu(es->s_blocks_count) >
1634 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1635 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1636 " too large to mount safely\n", sb->s_id);
1637 if (sizeof(sector_t) < 8)
1638 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1643 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1645 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1646 le32_to_cpu(es->s_first_data_block) - 1)
1647 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1648 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1649 EXT3_DESC_PER_BLOCK(sb);
1650 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1652 if (sbi->s_group_desc == NULL) {
1653 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1657 bgl_lock_init(&sbi->s_blockgroup_lock);
1659 for (i = 0; i < db_count; i++) {
1660 block = descriptor_loc(sb, logic_sb_block, i);
1661 sbi->s_group_desc[i] = sb_bread(sb, block);
1662 if (!sbi->s_group_desc[i]) {
1663 printk (KERN_ERR "EXT3-fs: "
1664 "can't read group descriptor %d\n", i);
1669 if (!ext3_check_descriptors (sb)) {
1670 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1673 sbi->s_gdb_count = db_count;
1674 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1675 spin_lock_init(&sbi->s_next_gen_lock);
1677 percpu_counter_init(&sbi->s_freeblocks_counter,
1678 ext3_count_free_blocks(sb));
1679 percpu_counter_init(&sbi->s_freeinodes_counter,
1680 ext3_count_free_inodes(sb));
1681 percpu_counter_init(&sbi->s_dirs_counter,
1682 ext3_count_dirs(sb));
1684 /* per fileystem reservation list head & lock */
1685 spin_lock_init(&sbi->s_rsv_window_lock);
1686 sbi->s_rsv_window_root = RB_ROOT;
1687 /* Add a single, static dummy reservation to the start of the
1688 * reservation window list --- it gives us a placeholder for
1689 * append-at-start-of-list which makes the allocation logic
1690 * _much_ simpler. */
1691 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1692 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1693 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1694 sbi->s_rsv_window_head.rsv_goal_size = 0;
1695 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1698 * set up enough so that it can read an inode
1700 sb->s_op = &ext3_sops;
1701 sb->s_export_op = &ext3_export_ops;
1702 sb->s_xattr = ext3_xattr_handlers;
1704 sb->s_qcop = &ext3_qctl_operations;
1705 sb->dq_op = &ext3_quota_operations;
1707 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1711 needs_recovery = (es->s_last_orphan != 0 ||
1712 EXT3_HAS_INCOMPAT_FEATURE(sb,
1713 EXT3_FEATURE_INCOMPAT_RECOVER));
1716 * The first inode we look at is the journal inode. Don't try
1717 * root first: it may be modified in the journal!
1719 if (!test_opt(sb, NOLOAD) &&
1720 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1721 if (ext3_load_journal(sb, es, journal_devnum))
1723 } else if (journal_inum) {
1724 if (ext3_create_journal(sb, es, journal_inum))
1729 "ext3: No journal on filesystem on %s\n",
1734 /* We have now updated the journal if required, so we can
1735 * validate the data journaling mode. */
1736 switch (test_opt(sb, DATA_FLAGS)) {
1738 /* No mode set, assume a default based on the journal
1739 capabilities: ORDERED_DATA if the journal can
1740 cope, else JOURNAL_DATA */
1741 if (journal_check_available_features
1742 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1743 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1745 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1748 case EXT3_MOUNT_ORDERED_DATA:
1749 case EXT3_MOUNT_WRITEBACK_DATA:
1750 if (!journal_check_available_features
1751 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1752 printk(KERN_ERR "EXT3-fs: Journal does not support "
1753 "requested data journaling mode\n");
1760 if (test_opt(sb, NOBH)) {
1761 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1762 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1763 "its supported only with writeback mode\n");
1764 clear_opt(sbi->s_mount_opt, NOBH);
1768 * The journal_load will have done any necessary log recovery,
1769 * so we can safely mount the rest of the filesystem now.
1772 root = iget(sb, EXT3_ROOT_INO);
1773 sb->s_root = d_alloc_root(root);
1775 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1779 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1782 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1786 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1788 * akpm: core read_super() calls in here with the superblock locked.
1789 * That deadlocks, because orphan cleanup needs to lock the superblock
1790 * in numerous places. Here we just pop the lock - it's relatively
1791 * harmless, because we are now ready to accept write_super() requests,
1792 * and aviro says that's the only reason for hanging onto the
1795 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1796 ext3_orphan_cleanup(sb, es);
1797 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1799 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1800 ext3_mark_recovery_complete(sb, es);
1801 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1802 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1803 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1811 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1816 journal_destroy(sbi->s_journal);
1818 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1819 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1820 percpu_counter_destroy(&sbi->s_dirs_counter);
1822 for (i = 0; i < db_count; i++)
1823 brelse(sbi->s_group_desc[i]);
1824 kfree(sbi->s_group_desc);
1827 for (i = 0; i < MAXQUOTAS; i++)
1828 kfree(sbi->s_qf_names[i]);
1830 ext3_blkdev_remove(sbi);
1833 sb->s_fs_info = NULL;
1840 * Setup any per-fs journal parameters now. We'll do this both on
1841 * initial mount, once the journal has been initialised but before we've
1842 * done any recovery; and again on any subsequent remount.
1844 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1846 struct ext3_sb_info *sbi = EXT3_SB(sb);
1848 if (sbi->s_commit_interval)
1849 journal->j_commit_interval = sbi->s_commit_interval;
1850 /* We could also set up an ext3-specific default for the commit
1851 * interval here, but for now we'll just fall back to the jbd
1854 spin_lock(&journal->j_state_lock);
1855 if (test_opt(sb, BARRIER))
1856 journal->j_flags |= JFS_BARRIER;
1858 journal->j_flags &= ~JFS_BARRIER;
1859 spin_unlock(&journal->j_state_lock);
1862 static journal_t *ext3_get_journal(struct super_block *sb,
1863 unsigned int journal_inum)
1865 struct inode *journal_inode;
1868 /* First, test for the existence of a valid inode on disk. Bad
1869 * things happen if we iget() an unused inode, as the subsequent
1870 * iput() will try to delete it. */
1872 journal_inode = iget(sb, journal_inum);
1873 if (!journal_inode) {
1874 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1877 if (!journal_inode->i_nlink) {
1878 make_bad_inode(journal_inode);
1879 iput(journal_inode);
1880 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1884 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1885 journal_inode, journal_inode->i_size);
1886 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1887 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1888 iput(journal_inode);
1892 journal = journal_init_inode(journal_inode);
1894 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1895 iput(journal_inode);
1898 journal->j_private = sb;
1899 ext3_init_journal_params(sb, journal);
1903 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1906 struct buffer_head * bh;
1910 int hblock, blocksize;
1911 ext3_fsblk_t sb_block;
1912 unsigned long offset;
1913 struct ext3_super_block * es;
1914 struct block_device *bdev;
1916 bdev = ext3_blkdev_get(j_dev);
1920 if (bd_claim(bdev, sb)) {
1922 "EXT3: failed to claim external journal device.\n");
1927 blocksize = sb->s_blocksize;
1928 hblock = bdev_hardsect_size(bdev);
1929 if (blocksize < hblock) {
1931 "EXT3-fs: blocksize too small for journal device.\n");
1935 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1936 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1937 set_blocksize(bdev, blocksize);
1938 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1939 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1940 "external journal\n");
1944 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1945 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1946 !(le32_to_cpu(es->s_feature_incompat) &
1947 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1948 printk(KERN_ERR "EXT3-fs: external journal has "
1949 "bad superblock\n");
1954 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1955 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1960 len = le32_to_cpu(es->s_blocks_count);
1961 start = sb_block + 1;
1962 brelse(bh); /* we're done with the superblock */
1964 journal = journal_init_dev(bdev, sb->s_bdev,
1965 start, len, blocksize);
1967 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1970 journal->j_private = sb;
1971 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1972 wait_on_buffer(journal->j_sb_buffer);
1973 if (!buffer_uptodate(journal->j_sb_buffer)) {
1974 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1977 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1978 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1979 "user (unsupported) - %d\n",
1980 be32_to_cpu(journal->j_superblock->s_nr_users));
1983 EXT3_SB(sb)->journal_bdev = bdev;
1984 ext3_init_journal_params(sb, journal);
1987 journal_destroy(journal);
1989 ext3_blkdev_put(bdev);
1993 static int ext3_load_journal(struct super_block *sb,
1994 struct ext3_super_block *es,
1995 unsigned long journal_devnum)
1998 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2001 int really_read_only;
2003 if (journal_devnum &&
2004 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2005 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2006 "numbers have changed\n");
2007 journal_dev = new_decode_dev(journal_devnum);
2009 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2011 really_read_only = bdev_read_only(sb->s_bdev);
2014 * Are we loading a blank journal or performing recovery after a
2015 * crash? For recovery, we need to check in advance whether we
2016 * can get read-write access to the device.
2019 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2020 if (sb->s_flags & MS_RDONLY) {
2021 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2022 "required on readonly filesystem.\n");
2023 if (really_read_only) {
2024 printk(KERN_ERR "EXT3-fs: write access "
2025 "unavailable, cannot proceed.\n");
2028 printk (KERN_INFO "EXT3-fs: write access will "
2029 "be enabled during recovery.\n");
2033 if (journal_inum && journal_dev) {
2034 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2035 "and inode journals!\n");
2040 if (!(journal = ext3_get_journal(sb, journal_inum)))
2043 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2047 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2048 err = journal_update_format(journal);
2050 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2051 journal_destroy(journal);
2056 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2057 err = journal_wipe(journal, !really_read_only);
2059 err = journal_load(journal);
2062 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2063 journal_destroy(journal);
2067 EXT3_SB(sb)->s_journal = journal;
2068 ext3_clear_journal_err(sb, es);
2070 if (journal_devnum &&
2071 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2072 es->s_journal_dev = cpu_to_le32(journal_devnum);
2075 /* Make sure we flush the recovery flag to disk. */
2076 ext3_commit_super(sb, es, 1);
2082 static int ext3_create_journal(struct super_block * sb,
2083 struct ext3_super_block * es,
2084 unsigned int journal_inum)
2089 if (sb->s_flags & MS_RDONLY) {
2090 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2091 "create journal.\n");
2095 journal = ext3_get_journal(sb, journal_inum);
2099 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2102 err = journal_create(journal);
2104 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2105 journal_destroy(journal);
2109 EXT3_SB(sb)->s_journal = journal;
2111 ext3_update_dynamic_rev(sb);
2112 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2113 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2115 es->s_journal_inum = cpu_to_le32(journal_inum);
2118 /* Make sure we flush the recovery flag to disk. */
2119 ext3_commit_super(sb, es, 1);
2124 static void ext3_commit_super (struct super_block * sb,
2125 struct ext3_super_block * es,
2128 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2132 es->s_wtime = cpu_to_le32(get_seconds());
2133 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2134 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2135 BUFFER_TRACE(sbh, "marking dirty");
2136 mark_buffer_dirty(sbh);
2138 sync_dirty_buffer(sbh);
2143 * Have we just finished recovery? If so, and if we are mounting (or
2144 * remounting) the filesystem readonly, then we will end up with a
2145 * consistent fs on disk. Record that fact.
2147 static void ext3_mark_recovery_complete(struct super_block * sb,
2148 struct ext3_super_block * es)
2150 journal_t *journal = EXT3_SB(sb)->s_journal;
2152 journal_lock_updates(journal);
2153 journal_flush(journal);
2155 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2156 sb->s_flags & MS_RDONLY) {
2157 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2159 ext3_commit_super(sb, es, 1);
2162 journal_unlock_updates(journal);
2166 * If we are mounting (or read-write remounting) a filesystem whose journal
2167 * has recorded an error from a previous lifetime, move that error to the
2168 * main filesystem now.
2170 static void ext3_clear_journal_err(struct super_block * sb,
2171 struct ext3_super_block * es)
2177 journal = EXT3_SB(sb)->s_journal;
2180 * Now check for any error status which may have been recorded in the
2181 * journal by a prior ext3_error() or ext3_abort()
2184 j_errno = journal_errno(journal);
2188 errstr = ext3_decode_error(sb, j_errno, nbuf);
2189 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2190 "from previous mount: %s", errstr);
2191 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2192 "filesystem check.");
2194 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2195 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2196 ext3_commit_super (sb, es, 1);
2198 journal_clear_err(journal);
2203 * Force the running and committing transactions to commit,
2204 * and wait on the commit.
2206 int ext3_force_commit(struct super_block *sb)
2211 if (sb->s_flags & MS_RDONLY)
2214 journal = EXT3_SB(sb)->s_journal;
2216 ret = ext3_journal_force_commit(journal);
2221 * Ext3 always journals updates to the superblock itself, so we don't
2222 * have to propagate any other updates to the superblock on disk at this
2223 * point. Just start an async writeback to get the buffers on their way
2226 * This implicitly triggers the writebehind on sync().
2229 static void ext3_write_super (struct super_block * sb)
2231 if (mutex_trylock(&sb->s_lock) != 0)
2236 static int ext3_sync_fs(struct super_block *sb, int wait)
2241 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2243 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2249 * LVM calls this function before a (read-only) snapshot is created. This
2250 * gives us a chance to flush the journal completely and mark the fs clean.
2252 static void ext3_write_super_lockfs(struct super_block *sb)
2256 if (!(sb->s_flags & MS_RDONLY)) {
2257 journal_t *journal = EXT3_SB(sb)->s_journal;
2259 /* Now we set up the journal barrier. */
2260 journal_lock_updates(journal);
2261 journal_flush(journal);
2263 /* Journal blocked and flushed, clear needs_recovery flag. */
2264 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2265 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2270 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2271 * flag here, even though the filesystem is not technically dirty yet.
2273 static void ext3_unlockfs(struct super_block *sb)
2275 if (!(sb->s_flags & MS_RDONLY)) {
2277 /* Reser the needs_recovery flag before the fs is unlocked. */
2278 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2279 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2281 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2285 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2287 struct ext3_super_block * es;
2288 struct ext3_sb_info *sbi = EXT3_SB(sb);
2289 ext3_fsblk_t n_blocks_count = 0;
2290 unsigned long old_sb_flags;
2291 struct ext3_mount_options old_opts;
2297 /* Store the original options */
2298 old_sb_flags = sb->s_flags;
2299 old_opts.s_mount_opt = sbi->s_mount_opt;
2300 old_opts.s_resuid = sbi->s_resuid;
2301 old_opts.s_resgid = sbi->s_resgid;
2302 old_opts.s_commit_interval = sbi->s_commit_interval;
2304 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2305 for (i = 0; i < MAXQUOTAS; i++)
2306 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2310 * Allow the "check" option to be passed as a remount option.
2312 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2317 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2318 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2320 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2321 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2325 ext3_init_journal_params(sb, sbi->s_journal);
2327 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2328 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2329 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2334 if (*flags & MS_RDONLY) {
2336 * First of all, the unconditional stuff we have to do
2337 * to disable replay of the journal when we next remount
2339 sb->s_flags |= MS_RDONLY;
2342 * OK, test if we are remounting a valid rw partition
2343 * readonly, and if so set the rdonly flag and then
2344 * mark the partition as valid again.
2346 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2347 (sbi->s_mount_state & EXT3_VALID_FS))
2348 es->s_state = cpu_to_le16(sbi->s_mount_state);
2351 * We have to unlock super so that we can wait for
2355 ext3_mark_recovery_complete(sb, es);
2359 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2360 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2361 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2362 "remount RDWR because of unsupported "
2363 "optional features (%x).\n",
2364 sb->s_id, le32_to_cpu(ret));
2370 * If we have an unprocessed orphan list hanging
2371 * around from a previously readonly bdev mount,
2372 * require a full umount/remount for now.
2374 if (es->s_last_orphan) {
2375 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2376 "remount RDWR because of unprocessed "
2377 "orphan inode list. Please "
2378 "umount/remount instead.\n",
2385 * Mounting a RDONLY partition read-write, so reread
2386 * and store the current valid flag. (It may have
2387 * been changed by e2fsck since we originally mounted
2390 ext3_clear_journal_err(sb, es);
2391 sbi->s_mount_state = le16_to_cpu(es->s_state);
2392 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2394 if (!ext3_setup_super (sb, es, 0))
2395 sb->s_flags &= ~MS_RDONLY;
2399 /* Release old quota file names */
2400 for (i = 0; i < MAXQUOTAS; i++)
2401 if (old_opts.s_qf_names[i] &&
2402 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2403 kfree(old_opts.s_qf_names[i]);
2407 sb->s_flags = old_sb_flags;
2408 sbi->s_mount_opt = old_opts.s_mount_opt;
2409 sbi->s_resuid = old_opts.s_resuid;
2410 sbi->s_resgid = old_opts.s_resgid;
2411 sbi->s_commit_interval = old_opts.s_commit_interval;
2413 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2414 for (i = 0; i < MAXQUOTAS; i++) {
2415 if (sbi->s_qf_names[i] &&
2416 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2417 kfree(sbi->s_qf_names[i]);
2418 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2424 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2426 struct super_block *sb = dentry->d_sb;
2427 struct ext3_sb_info *sbi = EXT3_SB(sb);
2428 struct ext3_super_block *es = sbi->s_es;
2431 if (test_opt(sb, MINIX_DF)) {
2432 sbi->s_overhead_last = 0;
2433 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2434 unsigned long ngroups = sbi->s_groups_count, i;
2435 ext3_fsblk_t overhead = 0;
2439 * Compute the overhead (FS structures). This is constant
2440 * for a given filesystem unless the number of block groups
2441 * changes so we cache the previous value until it does.
2445 * All of the blocks before first_data_block are
2448 overhead = le32_to_cpu(es->s_first_data_block);
2451 * Add the overhead attributed to the superblock and
2452 * block group descriptors. If the sparse superblocks
2453 * feature is turned on, then not all groups have this.
2455 for (i = 0; i < ngroups; i++) {
2456 overhead += ext3_bg_has_super(sb, i) +
2457 ext3_bg_num_gdb(sb, i);
2462 * Every block group has an inode bitmap, a block
2463 * bitmap, and an inode table.
2465 overhead += ngroups * (2 + sbi->s_itb_per_group);
2466 sbi->s_overhead_last = overhead;
2468 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2471 buf->f_type = EXT3_SUPER_MAGIC;
2472 buf->f_bsize = sb->s_blocksize;
2473 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2474 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
2475 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2476 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2477 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2479 buf->f_files = le32_to_cpu(es->s_inodes_count);
2480 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
2481 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2482 buf->f_namelen = EXT3_NAME_LEN;
2483 fsid = le64_to_cpup((void *)es->s_uuid) ^
2484 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2485 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2486 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2490 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2491 * is locked for write. Otherwise the are possible deadlocks:
2492 * Process 1 Process 2
2493 * ext3_create() quota_sync()
2494 * journal_start() write_dquot()
2495 * DQUOT_INIT() down(dqio_mutex)
2496 * down(dqio_mutex) journal_start()
2502 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2504 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2507 static int ext3_dquot_initialize(struct inode *inode, int type)
2512 /* We may create quota structure so we need to reserve enough blocks */
2513 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2515 return PTR_ERR(handle);
2516 ret = dquot_initialize(inode, type);
2517 err = ext3_journal_stop(handle);
2523 static int ext3_dquot_drop(struct inode *inode)
2528 /* We may delete quota structure so we need to reserve enough blocks */
2529 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2531 return PTR_ERR(handle);
2532 ret = dquot_drop(inode);
2533 err = ext3_journal_stop(handle);
2539 static int ext3_write_dquot(struct dquot *dquot)
2543 struct inode *inode;
2545 inode = dquot_to_inode(dquot);
2546 handle = ext3_journal_start(inode,
2547 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2549 return PTR_ERR(handle);
2550 ret = dquot_commit(dquot);
2551 err = ext3_journal_stop(handle);
2557 static int ext3_acquire_dquot(struct dquot *dquot)
2562 handle = ext3_journal_start(dquot_to_inode(dquot),
2563 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2565 return PTR_ERR(handle);
2566 ret = dquot_acquire(dquot);
2567 err = ext3_journal_stop(handle);
2573 static int ext3_release_dquot(struct dquot *dquot)
2578 handle = ext3_journal_start(dquot_to_inode(dquot),
2579 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2581 return PTR_ERR(handle);
2582 ret = dquot_release(dquot);
2583 err = ext3_journal_stop(handle);
2589 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2591 /* Are we journalling quotas? */
2592 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2593 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2594 dquot_mark_dquot_dirty(dquot);
2595 return ext3_write_dquot(dquot);
2597 return dquot_mark_dquot_dirty(dquot);
2601 static int ext3_write_info(struct super_block *sb, int type)
2606 /* Data block + inode block */
2607 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2609 return PTR_ERR(handle);
2610 ret = dquot_commit_info(sb, type);
2611 err = ext3_journal_stop(handle);
2618 * Turn on quotas during mount time - we need to find
2619 * the quota file and such...
2621 static int ext3_quota_on_mount(struct super_block *sb, int type)
2623 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2624 EXT3_SB(sb)->s_jquota_fmt, type);
2628 * Standard function to be called on quota_on
2630 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2634 struct nameidata nd;
2636 if (!test_opt(sb, QUOTA))
2638 /* Not journalling quota? */
2639 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2640 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2641 return vfs_quota_on(sb, type, format_id, path);
2642 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2645 /* Quotafile not on the same filesystem? */
2646 if (nd.mnt->mnt_sb != sb) {
2650 /* Quotafile not of fs root? */
2651 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2653 "EXT3-fs: Quota file not on filesystem root. "
2654 "Journalled quota will not work.\n");
2656 return vfs_quota_on(sb, type, format_id, path);
2659 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2660 * acquiring the locks... As quota files are never truncated and quota code
2661 * itself serializes the operations (and noone else should touch the files)
2662 * we don't have to be afraid of races */
2663 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2664 size_t len, loff_t off)
2666 struct inode *inode = sb_dqopt(sb)->files[type];
2667 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2669 int offset = off & (sb->s_blocksize - 1);
2672 struct buffer_head *bh;
2673 loff_t i_size = i_size_read(inode);
2677 if (off+len > i_size)
2680 while (toread > 0) {
2681 tocopy = sb->s_blocksize - offset < toread ?
2682 sb->s_blocksize - offset : toread;
2683 bh = ext3_bread(NULL, inode, blk, 0, &err);
2686 if (!bh) /* A hole? */
2687 memset(data, 0, tocopy);
2689 memcpy(data, bh->b_data+offset, tocopy);
2699 /* Write to quotafile (we know the transaction is already started and has
2700 * enough credits) */
2701 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2702 const char *data, size_t len, loff_t off)
2704 struct inode *inode = sb_dqopt(sb)->files[type];
2705 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2707 int offset = off & (sb->s_blocksize - 1);
2709 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2710 size_t towrite = len;
2711 struct buffer_head *bh;
2712 handle_t *handle = journal_current_handle();
2714 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2715 while (towrite > 0) {
2716 tocopy = sb->s_blocksize - offset < towrite ?
2717 sb->s_blocksize - offset : towrite;
2718 bh = ext3_bread(handle, inode, blk, 1, &err);
2721 if (journal_quota) {
2722 err = ext3_journal_get_write_access(handle, bh);
2729 memcpy(bh->b_data+offset, data, tocopy);
2730 flush_dcache_page(bh->b_page);
2733 err = ext3_journal_dirty_metadata(handle, bh);
2735 /* Always do at least ordered writes for quotas */
2736 err = ext3_journal_dirty_data(handle, bh);
2737 mark_buffer_dirty(bh);
2750 if (inode->i_size < off+len-towrite) {
2751 i_size_write(inode, off+len-towrite);
2752 EXT3_I(inode)->i_disksize = inode->i_size;
2755 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2756 ext3_mark_inode_dirty(handle, inode);
2757 mutex_unlock(&inode->i_mutex);
2758 return len - towrite;
2763 static int ext3_get_sb(struct file_system_type *fs_type,
2764 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2766 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2769 static struct file_system_type ext3_fs_type = {
2770 .owner = THIS_MODULE,
2772 .get_sb = ext3_get_sb,
2773 .kill_sb = kill_block_super,
2774 .fs_flags = FS_REQUIRES_DEV,
2777 static int __init init_ext3_fs(void)
2779 int err = init_ext3_xattr();
2782 err = init_inodecache();
2785 err = register_filesystem(&ext3_fs_type);
2790 destroy_inodecache();
2796 static void __exit exit_ext3_fs(void)
2798 unregister_filesystem(&ext3_fs_type);
2799 destroy_inodecache();
2803 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2804 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2805 MODULE_LICENSE("GPL");
2806 module_init(init_ext3_fs)
2807 module_exit(exit_ext3_fs)