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>
39 #include <asm/uaccess.h>
45 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
46 unsigned long journal_devnum);
47 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
49 static void ext3_commit_super (struct super_block * sb,
50 struct ext3_super_block * es,
52 static void ext3_mark_recovery_complete(struct super_block * sb,
53 struct ext3_super_block * es);
54 static void ext3_clear_journal_err(struct super_block * sb,
55 struct ext3_super_block * es);
56 static int ext3_sync_fs(struct super_block *sb, int wait);
57 static const char *ext3_decode_error(struct super_block * sb, int errno,
59 static int ext3_remount (struct super_block * sb, int * flags, char * data);
60 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
61 static void ext3_unlockfs(struct super_block *sb);
62 static void ext3_write_super (struct super_block * sb);
63 static void ext3_write_super_lockfs(struct super_block *sb);
66 * Wrappers for journal_start/end.
68 * The only special thing we need to do here is to make sure that all
69 * journal_end calls result in the superblock being marked dirty, so
70 * that sync() will call the filesystem's write_super callback if
73 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
77 if (sb->s_flags & MS_RDONLY)
78 return ERR_PTR(-EROFS);
80 /* Special case here: if the journal has aborted behind our
81 * backs (eg. EIO in the commit thread), then we still need to
82 * take the FS itself readonly cleanly. */
83 journal = EXT3_SB(sb)->s_journal;
84 if (is_journal_aborted(journal)) {
85 ext3_abort(sb, __FUNCTION__,
86 "Detected aborted journal");
87 return ERR_PTR(-EROFS);
90 return journal_start(journal, nblocks);
94 * The only special thing we need to do here is to make sure that all
95 * journal_stop calls result in the superblock being marked dirty, so
96 * that sync() will call the filesystem's write_super callback if
99 int __ext3_journal_stop(const char *where, handle_t *handle)
101 struct super_block *sb;
105 sb = handle->h_transaction->t_journal->j_private;
107 rc = journal_stop(handle);
112 __ext3_std_error(sb, where, err);
116 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
117 struct buffer_head *bh, handle_t *handle, int err)
120 const char *errstr = ext3_decode_error(NULL, err, nbuf);
123 BUFFER_TRACE(bh, "abort");
128 if (is_handle_aborted(handle))
131 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
132 caller, errstr, err_fn);
134 journal_abort_handle(handle);
137 /* Deal with the reporting of failure conditions on a filesystem such as
138 * inconsistencies detected or read IO failures.
140 * On ext2, we can store the error state of the filesystem in the
141 * superblock. That is not possible on ext3, because we may have other
142 * write ordering constraints on the superblock which prevent us from
143 * writing it out straight away; and given that the journal is about to
144 * be aborted, we can't rely on the current, or future, transactions to
145 * write out the superblock safely.
147 * We'll just use the journal_abort() error code to record an error in
148 * the journal instead. On recovery, the journal will compain about
149 * that error until we've noted it down and cleared it.
152 static void ext3_handle_error(struct super_block *sb)
154 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
156 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
157 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
159 if (sb->s_flags & MS_RDONLY)
162 if (test_opt (sb, ERRORS_RO)) {
163 printk (KERN_CRIT "Remounting filesystem read-only\n");
164 sb->s_flags |= MS_RDONLY;
166 journal_t *journal = EXT3_SB(sb)->s_journal;
168 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
170 journal_abort(journal, -EIO);
172 if (test_opt(sb, ERRORS_PANIC))
173 panic("EXT3-fs (device %s): panic forced after error\n",
175 ext3_commit_super(sb, es, 1);
178 void ext3_error (struct super_block * sb, const char * function,
179 const char * fmt, ...)
184 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
189 ext3_handle_error(sb);
192 static const char *ext3_decode_error(struct super_block * sb, int errno,
199 errstr = "IO failure";
202 errstr = "Out of memory";
205 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
206 errstr = "Journal has aborted";
208 errstr = "Readonly filesystem";
211 /* If the caller passed in an extra buffer for unknown
212 * errors, textualise them now. Else we just return
215 /* Check for truncated error codes... */
216 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
225 /* __ext3_std_error decodes expected errors from journaling functions
226 * automatically and invokes the appropriate error response. */
228 void __ext3_std_error (struct super_block * sb, const char * function,
234 /* Special case: if the error is EROFS, and we're not already
235 * inside a transaction, then there's really no point in logging
237 if (errno == -EROFS && journal_current_handle() == NULL &&
238 (sb->s_flags & MS_RDONLY))
241 errstr = ext3_decode_error(sb, errno, nbuf);
242 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
243 sb->s_id, function, errstr);
245 ext3_handle_error(sb);
249 * ext3_abort is a much stronger failure handler than ext3_error. The
250 * abort function may be used to deal with unrecoverable failures such
251 * as journal IO errors or ENOMEM at a critical moment in log management.
253 * We unconditionally force the filesystem into an ABORT|READONLY state,
254 * unless the error response on the fs has been set to panic in which
255 * case we take the easy way out and panic immediately.
258 void ext3_abort (struct super_block * sb, const char * function,
259 const char * fmt, ...)
263 printk (KERN_CRIT "ext3_abort called.\n");
266 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
271 if (test_opt(sb, ERRORS_PANIC))
272 panic("EXT3-fs panic from previous error\n");
274 if (sb->s_flags & MS_RDONLY)
277 printk(KERN_CRIT "Remounting filesystem read-only\n");
278 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
279 sb->s_flags |= MS_RDONLY;
280 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
281 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
284 void ext3_warning (struct super_block * sb, const char * function,
285 const char * fmt, ...)
290 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
297 void ext3_update_dynamic_rev(struct super_block *sb)
299 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
301 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
304 ext3_warning(sb, __FUNCTION__,
305 "updating to rev %d because of new feature flag, "
306 "running e2fsck is recommended",
309 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
310 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
311 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
312 /* leave es->s_feature_*compat flags alone */
313 /* es->s_uuid will be set by e2fsck if empty */
316 * The rest of the superblock fields should be zero, and if not it
317 * means they are likely already in use, so leave them alone. We
318 * can leave it up to e2fsck to clean up any inconsistencies there.
323 * Open the external journal device
325 static struct block_device *ext3_blkdev_get(dev_t dev)
327 struct block_device *bdev;
328 char b[BDEVNAME_SIZE];
330 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
336 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
337 __bdevname(dev, b), PTR_ERR(bdev));
342 * Release the journal device
344 static int ext3_blkdev_put(struct block_device *bdev)
347 return blkdev_put(bdev);
350 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
352 struct block_device *bdev;
355 bdev = sbi->journal_bdev;
357 ret = ext3_blkdev_put(bdev);
358 sbi->journal_bdev = NULL;
363 static inline struct inode *orphan_list_entry(struct list_head *l)
365 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
368 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
372 printk(KERN_ERR "sb orphan head is %d\n",
373 le32_to_cpu(sbi->s_es->s_last_orphan));
375 printk(KERN_ERR "sb_info orphan list:\n");
376 list_for_each(l, &sbi->s_orphan) {
377 struct inode *inode = orphan_list_entry(l);
379 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
380 inode->i_sb->s_id, inode->i_ino, inode,
381 inode->i_mode, inode->i_nlink,
386 static void ext3_put_super (struct super_block * sb)
388 struct ext3_sb_info *sbi = EXT3_SB(sb);
389 struct ext3_super_block *es = sbi->s_es;
392 ext3_xattr_put_super(sb);
393 journal_destroy(sbi->s_journal);
394 if (!(sb->s_flags & MS_RDONLY)) {
395 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
396 es->s_state = cpu_to_le16(sbi->s_mount_state);
397 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
398 mark_buffer_dirty(sbi->s_sbh);
399 ext3_commit_super(sb, es, 1);
402 for (i = 0; i < sbi->s_gdb_count; i++)
403 brelse(sbi->s_group_desc[i]);
404 kfree(sbi->s_group_desc);
405 percpu_counter_destroy(&sbi->s_freeblocks_counter);
406 percpu_counter_destroy(&sbi->s_freeinodes_counter);
407 percpu_counter_destroy(&sbi->s_dirs_counter);
410 for (i = 0; i < MAXQUOTAS; i++)
411 kfree(sbi->s_qf_names[i]);
414 /* Debugging code just in case the in-memory inode orphan list
415 * isn't empty. The on-disk one can be non-empty if we've
416 * detected an error and taken the fs readonly, but the
417 * in-memory list had better be clean by this point. */
418 if (!list_empty(&sbi->s_orphan))
419 dump_orphan_list(sb, sbi);
420 J_ASSERT(list_empty(&sbi->s_orphan));
422 invalidate_bdev(sb->s_bdev, 0);
423 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
425 * Invalidate the journal device's buffers. We don't want them
426 * floating about in memory - the physical journal device may
427 * hotswapped, and it breaks the `ro-after' testing code.
429 sync_blockdev(sbi->journal_bdev);
430 invalidate_bdev(sbi->journal_bdev, 0);
431 ext3_blkdev_remove(sbi);
433 sb->s_fs_info = NULL;
438 static kmem_cache_t *ext3_inode_cachep;
441 * Called inside transaction, so use GFP_NOFS
443 static struct inode *ext3_alloc_inode(struct super_block *sb)
445 struct ext3_inode_info *ei;
447 ei = kmem_cache_alloc(ext3_inode_cachep, SLAB_NOFS);
450 #ifdef CONFIG_EXT3_FS_POSIX_ACL
451 ei->i_acl = EXT3_ACL_NOT_CACHED;
452 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
454 ei->i_block_alloc_info = NULL;
455 ei->vfs_inode.i_version = 1;
456 return &ei->vfs_inode;
459 static void ext3_destroy_inode(struct inode *inode)
461 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
464 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
466 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
468 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
469 SLAB_CTOR_CONSTRUCTOR) {
470 INIT_LIST_HEAD(&ei->i_orphan);
471 #ifdef CONFIG_EXT3_FS_XATTR
472 init_rwsem(&ei->xattr_sem);
474 mutex_init(&ei->truncate_mutex);
475 inode_init_once(&ei->vfs_inode);
479 static int init_inodecache(void)
481 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
482 sizeof(struct ext3_inode_info),
483 0, (SLAB_RECLAIM_ACCOUNT|
486 if (ext3_inode_cachep == NULL)
491 static void destroy_inodecache(void)
493 if (kmem_cache_destroy(ext3_inode_cachep))
494 printk(KERN_INFO "ext3_inode_cache: not all structures were freed\n");
497 static void ext3_clear_inode(struct inode *inode)
499 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
500 #ifdef CONFIG_EXT3_FS_POSIX_ACL
501 if (EXT3_I(inode)->i_acl &&
502 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
503 posix_acl_release(EXT3_I(inode)->i_acl);
504 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
506 if (EXT3_I(inode)->i_default_acl &&
507 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
508 posix_acl_release(EXT3_I(inode)->i_default_acl);
509 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
512 ext3_discard_reservation(inode);
513 EXT3_I(inode)->i_block_alloc_info = NULL;
518 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
520 #if defined(CONFIG_QUOTA)
521 struct ext3_sb_info *sbi = EXT3_SB(sb);
523 if (sbi->s_jquota_fmt)
524 seq_printf(seq, ",jqfmt=%s",
525 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
527 if (sbi->s_qf_names[USRQUOTA])
528 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
530 if (sbi->s_qf_names[GRPQUOTA])
531 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
533 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
534 seq_puts(seq, ",usrquota");
536 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
537 seq_puts(seq, ",grpquota");
541 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
543 struct super_block *sb = vfs->mnt_sb;
545 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
546 seq_puts(seq, ",data=journal");
547 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
548 seq_puts(seq, ",data=ordered");
549 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
550 seq_puts(seq, ",data=writeback");
552 ext3_show_quota_options(seq, sb);
558 static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
561 unsigned long ino = objp[0];
562 __u32 generation = objp[1];
564 struct dentry *result;
566 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
567 return ERR_PTR(-ESTALE);
568 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
569 return ERR_PTR(-ESTALE);
571 /* iget isn't really right if the inode is currently unallocated!!
573 * ext3_read_inode will return a bad_inode if the inode had been
574 * deleted, so we should be safe.
576 * Currently we don't know the generation for parent directory, so
577 * a generation of 0 means "accept any"
579 inode = iget(sb, ino);
581 return ERR_PTR(-ENOMEM);
582 if (is_bad_inode(inode) ||
583 (generation && inode->i_generation != generation)) {
585 return ERR_PTR(-ESTALE);
587 /* now to find a dentry.
588 * If possible, get a well-connected one
590 result = d_alloc_anon(inode);
593 return ERR_PTR(-ENOMEM);
599 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
600 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
602 static int ext3_dquot_initialize(struct inode *inode, int type);
603 static int ext3_dquot_drop(struct inode *inode);
604 static int ext3_write_dquot(struct dquot *dquot);
605 static int ext3_acquire_dquot(struct dquot *dquot);
606 static int ext3_release_dquot(struct dquot *dquot);
607 static int ext3_mark_dquot_dirty(struct dquot *dquot);
608 static int ext3_write_info(struct super_block *sb, int type);
609 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
610 static int ext3_quota_on_mount(struct super_block *sb, int type);
611 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
612 size_t len, loff_t off);
613 static ssize_t ext3_quota_write(struct super_block *sb, int type,
614 const char *data, size_t len, loff_t off);
616 static struct dquot_operations ext3_quota_operations = {
617 .initialize = ext3_dquot_initialize,
618 .drop = ext3_dquot_drop,
619 .alloc_space = dquot_alloc_space,
620 .alloc_inode = dquot_alloc_inode,
621 .free_space = dquot_free_space,
622 .free_inode = dquot_free_inode,
623 .transfer = dquot_transfer,
624 .write_dquot = ext3_write_dquot,
625 .acquire_dquot = ext3_acquire_dquot,
626 .release_dquot = ext3_release_dquot,
627 .mark_dirty = ext3_mark_dquot_dirty,
628 .write_info = ext3_write_info
631 static struct quotactl_ops ext3_qctl_operations = {
632 .quota_on = ext3_quota_on,
633 .quota_off = vfs_quota_off,
634 .quota_sync = vfs_quota_sync,
635 .get_info = vfs_get_dqinfo,
636 .set_info = vfs_set_dqinfo,
637 .get_dqblk = vfs_get_dqblk,
638 .set_dqblk = vfs_set_dqblk
642 static struct super_operations ext3_sops = {
643 .alloc_inode = ext3_alloc_inode,
644 .destroy_inode = ext3_destroy_inode,
645 .read_inode = ext3_read_inode,
646 .write_inode = ext3_write_inode,
647 .dirty_inode = ext3_dirty_inode,
648 .delete_inode = ext3_delete_inode,
649 .put_super = ext3_put_super,
650 .write_super = ext3_write_super,
651 .sync_fs = ext3_sync_fs,
652 .write_super_lockfs = ext3_write_super_lockfs,
653 .unlockfs = ext3_unlockfs,
654 .statfs = ext3_statfs,
655 .remount_fs = ext3_remount,
656 .clear_inode = ext3_clear_inode,
657 .show_options = ext3_show_options,
659 .quota_read = ext3_quota_read,
660 .quota_write = ext3_quota_write,
664 static struct export_operations ext3_export_ops = {
665 .get_parent = ext3_get_parent,
666 .get_dentry = ext3_get_dentry,
670 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
671 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
672 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
673 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
674 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
675 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
676 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
677 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
678 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
679 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
680 Opt_grpquota, Opt_tagxid
683 static match_table_t tokens = {
684 {Opt_bsd_df, "bsddf"},
685 {Opt_minix_df, "minixdf"},
686 {Opt_grpid, "grpid"},
687 {Opt_grpid, "bsdgroups"},
688 {Opt_nogrpid, "nogrpid"},
689 {Opt_nogrpid, "sysvgroups"},
690 {Opt_resgid, "resgid=%u"},
691 {Opt_resuid, "resuid=%u"},
693 {Opt_err_cont, "errors=continue"},
694 {Opt_err_panic, "errors=panic"},
695 {Opt_err_ro, "errors=remount-ro"},
696 {Opt_nouid32, "nouid32"},
697 {Opt_nocheck, "nocheck"},
698 {Opt_nocheck, "check=none"},
699 {Opt_debug, "debug"},
700 {Opt_oldalloc, "oldalloc"},
701 {Opt_orlov, "orlov"},
702 {Opt_user_xattr, "user_xattr"},
703 {Opt_nouser_xattr, "nouser_xattr"},
705 {Opt_noacl, "noacl"},
706 {Opt_reservation, "reservation"},
707 {Opt_noreservation, "noreservation"},
708 {Opt_noload, "noload"},
711 {Opt_commit, "commit=%u"},
712 {Opt_journal_update, "journal=update"},
713 {Opt_journal_inum, "journal=%u"},
714 {Opt_journal_dev, "journal_dev=%u"},
715 {Opt_abort, "abort"},
716 {Opt_data_journal, "data=journal"},
717 {Opt_data_ordered, "data=ordered"},
718 {Opt_data_writeback, "data=writeback"},
719 {Opt_offusrjquota, "usrjquota="},
720 {Opt_usrjquota, "usrjquota=%s"},
721 {Opt_offgrpjquota, "grpjquota="},
722 {Opt_grpjquota, "grpjquota=%s"},
723 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
724 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
725 {Opt_grpquota, "grpquota"},
726 {Opt_noquota, "noquota"},
727 {Opt_quota, "quota"},
728 {Opt_usrquota, "usrquota"},
729 {Opt_barrier, "barrier=%u"},
730 {Opt_tagxid, "tagxid"},
732 {Opt_resize, "resize"},
735 static ext3_fsblk_t get_sb_block(void **data)
737 ext3_fsblk_t sb_block;
738 char *options = (char *) *data;
740 if (!options || strncmp(options, "sb=", 3) != 0)
741 return 1; /* Default location */
743 /*todo: use simple_strtoll with >32bit ext3 */
744 sb_block = simple_strtoul(options, &options, 0);
745 if (*options && *options != ',') {
746 printk("EXT3-fs: Invalid sb specification: %s\n",
752 *data = (void *) options;
756 static int parse_options (char *options, struct super_block *sb,
757 unsigned int *inum, unsigned long *journal_devnum,
758 ext3_fsblk_t *n_blocks_count, int is_remount)
760 struct ext3_sb_info *sbi = EXT3_SB(sb);
762 substring_t args[MAX_OPT_ARGS];
773 while ((p = strsep (&options, ",")) != NULL) {
778 token = match_token(p, tokens, args);
781 clear_opt (sbi->s_mount_opt, MINIX_DF);
784 set_opt (sbi->s_mount_opt, MINIX_DF);
787 set_opt (sbi->s_mount_opt, GRPID);
790 clear_opt (sbi->s_mount_opt, GRPID);
793 if (match_int(&args[0], &option))
795 sbi->s_resuid = option;
798 if (match_int(&args[0], &option))
800 sbi->s_resgid = option;
803 /* handled by get_sb_block() instead of here */
804 /* *sb_block = match_int(&args[0]); */
807 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
808 clear_opt (sbi->s_mount_opt, ERRORS_RO);
809 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
812 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
813 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
814 set_opt (sbi->s_mount_opt, ERRORS_RO);
817 clear_opt (sbi->s_mount_opt, ERRORS_RO);
818 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
819 set_opt (sbi->s_mount_opt, ERRORS_CONT);
822 set_opt (sbi->s_mount_opt, NO_UID32);
824 #ifndef CONFIG_INOXID_NONE
826 set_opt (sbi->s_mount_opt, TAGXID);
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 (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1274 if (es->s_last_orphan)
1275 jbd_debug(1, "Errors on filesystem, "
1276 "clearing orphan list.\n");
1277 es->s_last_orphan = 0;
1278 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1282 if (s_flags & MS_RDONLY) {
1283 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1285 sb->s_flags &= ~MS_RDONLY;
1288 /* Needed for iput() to work correctly and not trash data */
1289 sb->s_flags |= MS_ACTIVE;
1290 /* Turn on quotas so that they are updated correctly */
1291 for (i = 0; i < MAXQUOTAS; i++) {
1292 if (EXT3_SB(sb)->s_qf_names[i]) {
1293 int ret = ext3_quota_on_mount(sb, i);
1296 "EXT3-fs: Cannot turn on journalled "
1297 "quota: error %d\n", ret);
1302 while (es->s_last_orphan) {
1303 struct inode *inode;
1306 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1307 es->s_last_orphan = 0;
1311 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1313 if (inode->i_nlink) {
1315 "%s: truncating inode %lu to %Ld bytes\n",
1316 __FUNCTION__, inode->i_ino, inode->i_size);
1317 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1318 inode->i_ino, inode->i_size);
1319 ext3_truncate(inode);
1323 "%s: deleting unreferenced inode %lu\n",
1324 __FUNCTION__, inode->i_ino);
1325 jbd_debug(2, "deleting unreferenced inode %lu\n",
1329 iput(inode); /* The delete magic happens here! */
1332 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1335 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1336 sb->s_id, PLURAL(nr_orphans));
1338 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1339 sb->s_id, PLURAL(nr_truncates));
1341 /* Turn quotas off */
1342 for (i = 0; i < MAXQUOTAS; i++) {
1343 if (sb_dqopt(sb)->files[i])
1344 vfs_quota_off(sb, i);
1347 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1350 #define log2(n) ffz(~(n))
1353 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1354 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1355 * We need to be 1 filesystem block less than the 2^32 sector limit.
1357 static loff_t ext3_max_size(int bits)
1359 loff_t res = EXT3_NDIR_BLOCKS;
1360 /* This constant is calculated to be the largest file size for a
1361 * dense, 4k-blocksize file such that the total number of
1362 * sectors in the file, including data and all indirect blocks,
1363 * does not exceed 2^32. */
1364 const loff_t upper_limit = 0x1ff7fffd000LL;
1366 res += 1LL << (bits-2);
1367 res += 1LL << (2*(bits-2));
1368 res += 1LL << (3*(bits-2));
1370 if (res > upper_limit)
1375 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1376 ext3_fsblk_t logic_sb_block,
1379 struct ext3_sb_info *sbi = EXT3_SB(sb);
1380 unsigned long bg, first_meta_bg;
1383 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1385 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1387 return (logic_sb_block + nr + 1);
1388 bg = sbi->s_desc_per_block * nr;
1389 if (ext3_bg_has_super(sb, bg))
1391 return (has_super + ext3_group_first_block_no(sb, bg));
1395 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1397 struct buffer_head * bh;
1398 struct ext3_super_block *es = NULL;
1399 struct ext3_sb_info *sbi;
1401 ext3_fsblk_t sb_block = get_sb_block(&data);
1402 ext3_fsblk_t logic_sb_block;
1403 unsigned long offset = 0;
1404 unsigned int journal_inum = 0;
1405 unsigned long journal_devnum = 0;
1406 unsigned long def_mount_opts;
1415 sbi = kmalloc(sizeof(*sbi), GFP_KERNEL);
1418 sb->s_fs_info = sbi;
1419 memset(sbi, 0, sizeof(*sbi));
1420 sbi->s_mount_opt = 0;
1421 sbi->s_resuid = EXT3_DEF_RESUID;
1422 sbi->s_resgid = EXT3_DEF_RESGID;
1426 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1428 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1433 * The ext3 superblock will not be buffer aligned for other than 1kB
1434 * block sizes. We need to calculate the offset from buffer start.
1436 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1437 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1438 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1440 logic_sb_block = sb_block;
1443 if (!(bh = sb_bread(sb, logic_sb_block))) {
1444 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1448 * Note: s_es must be initialized as soon as possible because
1449 * some ext3 macro-instructions depend on its value
1451 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1453 sb->s_magic = le16_to_cpu(es->s_magic);
1454 if (sb->s_magic != EXT3_SUPER_MAGIC)
1457 /* Set defaults before we parse the mount options */
1458 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1459 if (def_mount_opts & EXT3_DEFM_DEBUG)
1460 set_opt(sbi->s_mount_opt, DEBUG);
1461 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1462 set_opt(sbi->s_mount_opt, GRPID);
1463 if (def_mount_opts & EXT3_DEFM_UID16)
1464 set_opt(sbi->s_mount_opt, NO_UID32);
1465 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1466 set_opt(sbi->s_mount_opt, XATTR_USER);
1467 if (def_mount_opts & EXT3_DEFM_ACL)
1468 set_opt(sbi->s_mount_opt, POSIX_ACL);
1469 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1470 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1471 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1472 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1473 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1474 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1476 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1477 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1478 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1479 set_opt(sbi->s_mount_opt, ERRORS_RO);
1481 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1482 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1484 set_opt(sbi->s_mount_opt, RESERVATION);
1486 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1490 if (EXT3_SB(sb)->s_mount_opt & EXT3_MOUNT_TAGXID)
1491 sb->s_flags |= MS_TAGXID;
1493 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1494 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1496 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1497 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1498 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1499 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1501 "EXT3-fs warning: feature flags set on rev 0 fs, "
1502 "running e2fsck is recommended\n");
1504 * Check feature flags regardless of the revision level, since we
1505 * previously didn't change the revision level when setting the flags,
1506 * so there is a chance incompat flags are set on a rev 0 filesystem.
1508 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1510 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1511 "unsupported optional features (%x).\n",
1512 sb->s_id, le32_to_cpu(features));
1515 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1516 if (!(sb->s_flags & MS_RDONLY) && features) {
1517 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1518 "unsupported optional features (%x).\n",
1519 sb->s_id, le32_to_cpu(features));
1522 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1524 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1525 blocksize > EXT3_MAX_BLOCK_SIZE) {
1527 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1528 blocksize, sb->s_id);
1532 hblock = bdev_hardsect_size(sb->s_bdev);
1533 if (sb->s_blocksize != blocksize) {
1535 * Make sure the blocksize for the filesystem is larger
1536 * than the hardware sectorsize for the machine.
1538 if (blocksize < hblock) {
1539 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1540 "device blocksize %d.\n", blocksize, hblock);
1545 sb_set_blocksize(sb, blocksize);
1546 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1547 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1548 bh = sb_bread(sb, logic_sb_block);
1551 "EXT3-fs: Can't read superblock on 2nd try.\n");
1554 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1556 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1558 "EXT3-fs: Magic mismatch, very weird !\n");
1563 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1565 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1566 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1567 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1569 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1570 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1571 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1572 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1573 (sbi->s_inode_size > blocksize)) {
1575 "EXT3-fs: unsupported inode size: %d\n",
1580 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1581 le32_to_cpu(es->s_log_frag_size);
1582 if (blocksize != sbi->s_frag_size) {
1584 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1585 sbi->s_frag_size, blocksize);
1588 sbi->s_frags_per_block = 1;
1589 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1590 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1591 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1592 if (EXT3_INODE_SIZE(sb) == 0)
1594 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1595 if (sbi->s_inodes_per_block == 0)
1597 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1598 sbi->s_inodes_per_block;
1599 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1601 sbi->s_mount_state = le16_to_cpu(es->s_state);
1602 sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb));
1603 sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb));
1604 for (i=0; i < 4; i++)
1605 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1606 sbi->s_def_hash_version = es->s_def_hash_version;
1608 if (sbi->s_blocks_per_group > blocksize * 8) {
1610 "EXT3-fs: #blocks per group too big: %lu\n",
1611 sbi->s_blocks_per_group);
1614 if (sbi->s_frags_per_group > blocksize * 8) {
1616 "EXT3-fs: #fragments per group too big: %lu\n",
1617 sbi->s_frags_per_group);
1620 if (sbi->s_inodes_per_group > blocksize * 8) {
1622 "EXT3-fs: #inodes per group too big: %lu\n",
1623 sbi->s_inodes_per_group);
1627 if (le32_to_cpu(es->s_blocks_count) >
1628 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1629 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1630 " too large to mount safely\n", sb->s_id);
1631 if (sizeof(sector_t) < 8)
1632 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1637 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1639 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1640 le32_to_cpu(es->s_first_data_block) - 1)
1641 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1642 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1643 EXT3_DESC_PER_BLOCK(sb);
1644 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1646 if (sbi->s_group_desc == NULL) {
1647 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1651 bgl_lock_init(&sbi->s_blockgroup_lock);
1653 for (i = 0; i < db_count; i++) {
1654 block = descriptor_loc(sb, logic_sb_block, i);
1655 sbi->s_group_desc[i] = sb_bread(sb, block);
1656 if (!sbi->s_group_desc[i]) {
1657 printk (KERN_ERR "EXT3-fs: "
1658 "can't read group descriptor %d\n", i);
1663 if (!ext3_check_descriptors (sb)) {
1664 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1667 sbi->s_gdb_count = db_count;
1668 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1669 spin_lock_init(&sbi->s_next_gen_lock);
1671 percpu_counter_init(&sbi->s_freeblocks_counter,
1672 ext3_count_free_blocks(sb));
1673 percpu_counter_init(&sbi->s_freeinodes_counter,
1674 ext3_count_free_inodes(sb));
1675 percpu_counter_init(&sbi->s_dirs_counter,
1676 ext3_count_dirs(sb));
1678 /* per fileystem reservation list head & lock */
1679 spin_lock_init(&sbi->s_rsv_window_lock);
1680 sbi->s_rsv_window_root = RB_ROOT;
1681 /* Add a single, static dummy reservation to the start of the
1682 * reservation window list --- it gives us a placeholder for
1683 * append-at-start-of-list which makes the allocation logic
1684 * _much_ simpler. */
1685 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1686 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1687 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1688 sbi->s_rsv_window_head.rsv_goal_size = 0;
1689 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1692 * set up enough so that it can read an inode
1694 sb->s_op = &ext3_sops;
1695 sb->s_export_op = &ext3_export_ops;
1696 sb->s_xattr = ext3_xattr_handlers;
1698 sb->s_qcop = &ext3_qctl_operations;
1699 sb->dq_op = &ext3_quota_operations;
1701 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1705 needs_recovery = (es->s_last_orphan != 0 ||
1706 EXT3_HAS_INCOMPAT_FEATURE(sb,
1707 EXT3_FEATURE_INCOMPAT_RECOVER));
1710 * The first inode we look at is the journal inode. Don't try
1711 * root first: it may be modified in the journal!
1713 if (!test_opt(sb, NOLOAD) &&
1714 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1715 if (ext3_load_journal(sb, es, journal_devnum))
1717 } else if (journal_inum) {
1718 if (ext3_create_journal(sb, es, journal_inum))
1723 "ext3: No journal on filesystem on %s\n",
1728 /* We have now updated the journal if required, so we can
1729 * validate the data journaling mode. */
1730 switch (test_opt(sb, DATA_FLAGS)) {
1732 /* No mode set, assume a default based on the journal
1733 capabilities: ORDERED_DATA if the journal can
1734 cope, else JOURNAL_DATA */
1735 if (journal_check_available_features
1736 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1737 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1739 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1742 case EXT3_MOUNT_ORDERED_DATA:
1743 case EXT3_MOUNT_WRITEBACK_DATA:
1744 if (!journal_check_available_features
1745 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1746 printk(KERN_ERR "EXT3-fs: Journal does not support "
1747 "requested data journaling mode\n");
1754 if (test_opt(sb, NOBH)) {
1755 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1756 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1757 "its supported only with writeback mode\n");
1758 clear_opt(sbi->s_mount_opt, NOBH);
1762 * The journal_load will have done any necessary log recovery,
1763 * so we can safely mount the rest of the filesystem now.
1766 root = iget(sb, EXT3_ROOT_INO);
1767 sb->s_root = d_alloc_root(root);
1769 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1773 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1776 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1780 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1782 * akpm: core read_super() calls in here with the superblock locked.
1783 * That deadlocks, because orphan cleanup needs to lock the superblock
1784 * in numerous places. Here we just pop the lock - it's relatively
1785 * harmless, because we are now ready to accept write_super() requests,
1786 * and aviro says that's the only reason for hanging onto the
1789 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1790 ext3_orphan_cleanup(sb, es);
1791 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1793 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1794 ext3_mark_recovery_complete(sb, es);
1795 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1796 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1797 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1805 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1810 journal_destroy(sbi->s_journal);
1812 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1813 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1814 percpu_counter_destroy(&sbi->s_dirs_counter);
1816 for (i = 0; i < db_count; i++)
1817 brelse(sbi->s_group_desc[i]);
1818 kfree(sbi->s_group_desc);
1821 for (i = 0; i < MAXQUOTAS; i++)
1822 kfree(sbi->s_qf_names[i]);
1824 ext3_blkdev_remove(sbi);
1827 sb->s_fs_info = NULL;
1834 * Setup any per-fs journal parameters now. We'll do this both on
1835 * initial mount, once the journal has been initialised but before we've
1836 * done any recovery; and again on any subsequent remount.
1838 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1840 struct ext3_sb_info *sbi = EXT3_SB(sb);
1842 if (sbi->s_commit_interval)
1843 journal->j_commit_interval = sbi->s_commit_interval;
1844 /* We could also set up an ext3-specific default for the commit
1845 * interval here, but for now we'll just fall back to the jbd
1848 spin_lock(&journal->j_state_lock);
1849 if (test_opt(sb, BARRIER))
1850 journal->j_flags |= JFS_BARRIER;
1852 journal->j_flags &= ~JFS_BARRIER;
1853 spin_unlock(&journal->j_state_lock);
1856 static journal_t *ext3_get_journal(struct super_block *sb,
1857 unsigned int journal_inum)
1859 struct inode *journal_inode;
1862 /* First, test for the existence of a valid inode on disk. Bad
1863 * things happen if we iget() an unused inode, as the subsequent
1864 * iput() will try to delete it. */
1866 journal_inode = iget(sb, journal_inum);
1867 if (!journal_inode) {
1868 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1871 if (!journal_inode->i_nlink) {
1872 make_bad_inode(journal_inode);
1873 iput(journal_inode);
1874 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1878 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1879 journal_inode, journal_inode->i_size);
1880 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1881 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1882 iput(journal_inode);
1886 journal = journal_init_inode(journal_inode);
1888 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1889 iput(journal_inode);
1892 journal->j_private = sb;
1893 ext3_init_journal_params(sb, journal);
1897 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1900 struct buffer_head * bh;
1904 int hblock, blocksize;
1905 ext3_fsblk_t sb_block;
1906 unsigned long offset;
1907 struct ext3_super_block * es;
1908 struct block_device *bdev;
1910 bdev = ext3_blkdev_get(j_dev);
1914 if (bd_claim(bdev, sb)) {
1916 "EXT3: failed to claim external journal device.\n");
1921 blocksize = sb->s_blocksize;
1922 hblock = bdev_hardsect_size(bdev);
1923 if (blocksize < hblock) {
1925 "EXT3-fs: blocksize too small for journal device.\n");
1929 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1930 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1931 set_blocksize(bdev, blocksize);
1932 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1933 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1934 "external journal\n");
1938 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1939 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1940 !(le32_to_cpu(es->s_feature_incompat) &
1941 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1942 printk(KERN_ERR "EXT3-fs: external journal has "
1943 "bad superblock\n");
1948 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1949 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1954 len = le32_to_cpu(es->s_blocks_count);
1955 start = sb_block + 1;
1956 brelse(bh); /* we're done with the superblock */
1958 journal = journal_init_dev(bdev, sb->s_bdev,
1959 start, len, blocksize);
1961 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1964 journal->j_private = sb;
1965 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1966 wait_on_buffer(journal->j_sb_buffer);
1967 if (!buffer_uptodate(journal->j_sb_buffer)) {
1968 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1971 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1972 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1973 "user (unsupported) - %d\n",
1974 be32_to_cpu(journal->j_superblock->s_nr_users));
1977 EXT3_SB(sb)->journal_bdev = bdev;
1978 ext3_init_journal_params(sb, journal);
1981 journal_destroy(journal);
1983 ext3_blkdev_put(bdev);
1987 static int ext3_load_journal(struct super_block *sb,
1988 struct ext3_super_block *es,
1989 unsigned long journal_devnum)
1992 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
1995 int really_read_only;
1997 if (journal_devnum &&
1998 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
1999 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2000 "numbers have changed\n");
2001 journal_dev = new_decode_dev(journal_devnum);
2003 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2005 really_read_only = bdev_read_only(sb->s_bdev);
2008 * Are we loading a blank journal or performing recovery after a
2009 * crash? For recovery, we need to check in advance whether we
2010 * can get read-write access to the device.
2013 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2014 if (sb->s_flags & MS_RDONLY) {
2015 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2016 "required on readonly filesystem.\n");
2017 if (really_read_only) {
2018 printk(KERN_ERR "EXT3-fs: write access "
2019 "unavailable, cannot proceed.\n");
2022 printk (KERN_INFO "EXT3-fs: write access will "
2023 "be enabled during recovery.\n");
2027 if (journal_inum && journal_dev) {
2028 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2029 "and inode journals!\n");
2034 if (!(journal = ext3_get_journal(sb, journal_inum)))
2037 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2041 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2042 err = journal_update_format(journal);
2044 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2045 journal_destroy(journal);
2050 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2051 err = journal_wipe(journal, !really_read_only);
2053 err = journal_load(journal);
2056 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2057 journal_destroy(journal);
2061 EXT3_SB(sb)->s_journal = journal;
2062 ext3_clear_journal_err(sb, es);
2064 if (journal_devnum &&
2065 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2066 es->s_journal_dev = cpu_to_le32(journal_devnum);
2069 /* Make sure we flush the recovery flag to disk. */
2070 ext3_commit_super(sb, es, 1);
2076 static int ext3_create_journal(struct super_block * sb,
2077 struct ext3_super_block * es,
2078 unsigned int journal_inum)
2082 if (sb->s_flags & MS_RDONLY) {
2083 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2084 "create journal.\n");
2088 if (!(journal = ext3_get_journal(sb, journal_inum)))
2091 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2094 if (journal_create(journal)) {
2095 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2096 journal_destroy(journal);
2100 EXT3_SB(sb)->s_journal = journal;
2102 ext3_update_dynamic_rev(sb);
2103 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2104 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2106 es->s_journal_inum = cpu_to_le32(journal_inum);
2109 /* Make sure we flush the recovery flag to disk. */
2110 ext3_commit_super(sb, es, 1);
2115 static void ext3_commit_super (struct super_block * sb,
2116 struct ext3_super_block * es,
2119 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2123 es->s_wtime = cpu_to_le32(get_seconds());
2124 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2125 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2126 BUFFER_TRACE(sbh, "marking dirty");
2127 mark_buffer_dirty(sbh);
2129 sync_dirty_buffer(sbh);
2134 * Have we just finished recovery? If so, and if we are mounting (or
2135 * remounting) the filesystem readonly, then we will end up with a
2136 * consistent fs on disk. Record that fact.
2138 static void ext3_mark_recovery_complete(struct super_block * sb,
2139 struct ext3_super_block * es)
2141 journal_t *journal = EXT3_SB(sb)->s_journal;
2143 journal_lock_updates(journal);
2144 journal_flush(journal);
2145 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2146 sb->s_flags & MS_RDONLY) {
2147 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2149 ext3_commit_super(sb, es, 1);
2151 journal_unlock_updates(journal);
2155 * If we are mounting (or read-write remounting) a filesystem whose journal
2156 * has recorded an error from a previous lifetime, move that error to the
2157 * main filesystem now.
2159 static void ext3_clear_journal_err(struct super_block * sb,
2160 struct ext3_super_block * es)
2166 journal = EXT3_SB(sb)->s_journal;
2169 * Now check for any error status which may have been recorded in the
2170 * journal by a prior ext3_error() or ext3_abort()
2173 j_errno = journal_errno(journal);
2177 errstr = ext3_decode_error(sb, j_errno, nbuf);
2178 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2179 "from previous mount: %s", errstr);
2180 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2181 "filesystem check.");
2183 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2184 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2185 ext3_commit_super (sb, es, 1);
2187 journal_clear_err(journal);
2192 * Force the running and committing transactions to commit,
2193 * and wait on the commit.
2195 int ext3_force_commit(struct super_block *sb)
2200 if (sb->s_flags & MS_RDONLY)
2203 journal = EXT3_SB(sb)->s_journal;
2205 ret = ext3_journal_force_commit(journal);
2210 * Ext3 always journals updates to the superblock itself, so we don't
2211 * have to propagate any other updates to the superblock on disk at this
2212 * point. Just start an async writeback to get the buffers on their way
2215 * This implicitly triggers the writebehind on sync().
2218 static void ext3_write_super (struct super_block * sb)
2220 if (mutex_trylock(&sb->s_lock) != 0)
2225 static int ext3_sync_fs(struct super_block *sb, int wait)
2230 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2232 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2238 * LVM calls this function before a (read-only) snapshot is created. This
2239 * gives us a chance to flush the journal completely and mark the fs clean.
2241 static void ext3_write_super_lockfs(struct super_block *sb)
2245 if (!(sb->s_flags & MS_RDONLY)) {
2246 journal_t *journal = EXT3_SB(sb)->s_journal;
2248 /* Now we set up the journal barrier. */
2249 journal_lock_updates(journal);
2250 journal_flush(journal);
2252 /* Journal blocked and flushed, clear needs_recovery flag. */
2253 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2254 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2259 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2260 * flag here, even though the filesystem is not technically dirty yet.
2262 static void ext3_unlockfs(struct super_block *sb)
2264 if (!(sb->s_flags & MS_RDONLY)) {
2266 /* Reser the needs_recovery flag before the fs is unlocked. */
2267 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2268 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2270 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2274 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2276 struct ext3_super_block * es;
2277 struct ext3_sb_info *sbi = EXT3_SB(sb);
2278 ext3_fsblk_t n_blocks_count = 0;
2279 unsigned long old_sb_flags;
2280 struct ext3_mount_options old_opts;
2286 /* Store the original options */
2287 old_sb_flags = sb->s_flags;
2288 old_opts.s_mount_opt = sbi->s_mount_opt;
2289 old_opts.s_resuid = sbi->s_resuid;
2290 old_opts.s_resgid = sbi->s_resgid;
2291 old_opts.s_commit_interval = sbi->s_commit_interval;
2293 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2294 for (i = 0; i < MAXQUOTAS; i++)
2295 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2299 * Allow the "check" option to be passed as a remount option.
2301 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2306 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2307 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2308 if ((sbi->s_mount_opt & EXT3_MOUNT_TAGXID) &&
2309 !(sb->s_flags & MS_TAGXID)) {
2310 printk("EXT3-fs: %s: tagxid not permitted on remount.\n",
2315 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2316 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2320 ext3_init_journal_params(sb, sbi->s_journal);
2322 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2323 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2324 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2329 if (*flags & MS_RDONLY) {
2331 * First of all, the unconditional stuff we have to do
2332 * to disable replay of the journal when we next remount
2334 sb->s_flags |= MS_RDONLY;
2337 * OK, test if we are remounting a valid rw partition
2338 * readonly, and if so set the rdonly flag and then
2339 * mark the partition as valid again.
2341 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2342 (sbi->s_mount_state & EXT3_VALID_FS))
2343 es->s_state = cpu_to_le16(sbi->s_mount_state);
2345 ext3_mark_recovery_complete(sb, es);
2348 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2349 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2350 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2351 "remount RDWR because of unsupported "
2352 "optional features (%x).\n",
2353 sb->s_id, le32_to_cpu(ret));
2358 * Mounting a RDONLY partition read-write, so reread
2359 * and store the current valid flag. (It may have
2360 * been changed by e2fsck since we originally mounted
2363 ext3_clear_journal_err(sb, es);
2364 sbi->s_mount_state = le16_to_cpu(es->s_state);
2365 if ((ret = ext3_group_extend(sb, es, n_blocks_count))) {
2369 if (!ext3_setup_super (sb, es, 0))
2370 sb->s_flags &= ~MS_RDONLY;
2374 /* Release old quota file names */
2375 for (i = 0; i < MAXQUOTAS; i++)
2376 if (old_opts.s_qf_names[i] &&
2377 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2378 kfree(old_opts.s_qf_names[i]);
2382 sb->s_flags = old_sb_flags;
2383 sbi->s_mount_opt = old_opts.s_mount_opt;
2384 sbi->s_resuid = old_opts.s_resuid;
2385 sbi->s_resgid = old_opts.s_resgid;
2386 sbi->s_commit_interval = old_opts.s_commit_interval;
2388 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2389 for (i = 0; i < MAXQUOTAS; i++) {
2390 if (sbi->s_qf_names[i] &&
2391 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2392 kfree(sbi->s_qf_names[i]);
2393 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2399 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2401 struct super_block *sb = dentry->d_sb;
2402 struct ext3_sb_info *sbi = EXT3_SB(sb);
2403 struct ext3_super_block *es = sbi->s_es;
2404 ext3_fsblk_t overhead;
2407 if (test_opt (sb, MINIX_DF))
2410 unsigned long ngroups;
2411 ngroups = EXT3_SB(sb)->s_groups_count;
2415 * Compute the overhead (FS structures)
2419 * All of the blocks before first_data_block are
2422 overhead = le32_to_cpu(es->s_first_data_block);
2425 * Add the overhead attributed to the superblock and
2426 * block group descriptors. If the sparse superblocks
2427 * feature is turned on, then not all groups have this.
2429 for (i = 0; i < ngroups; i++) {
2430 overhead += ext3_bg_has_super(sb, i) +
2431 ext3_bg_num_gdb(sb, i);
2436 * Every block group has an inode bitmap, a block
2437 * bitmap, and an inode table.
2439 overhead += (ngroups * (2 + EXT3_SB(sb)->s_itb_per_group));
2442 buf->f_type = EXT3_SUPER_MAGIC;
2443 buf->f_bsize = sb->s_blocksize;
2444 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
2445 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
2446 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2447 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2449 buf->f_files = le32_to_cpu(es->s_inodes_count);
2450 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
2451 buf->f_namelen = EXT3_NAME_LEN;
2455 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2456 * is locked for write. Otherwise the are possible deadlocks:
2457 * Process 1 Process 2
2458 * ext3_create() quota_sync()
2459 * journal_start() write_dquot()
2460 * DQUOT_INIT() down(dqio_mutex)
2461 * down(dqio_mutex) journal_start()
2467 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2469 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2472 static int ext3_dquot_initialize(struct inode *inode, int type)
2477 /* We may create quota structure so we need to reserve enough blocks */
2478 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2480 return PTR_ERR(handle);
2481 ret = dquot_initialize(inode, type);
2482 err = ext3_journal_stop(handle);
2488 static int ext3_dquot_drop(struct inode *inode)
2493 /* We may delete quota structure so we need to reserve enough blocks */
2494 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2496 return PTR_ERR(handle);
2497 ret = dquot_drop(inode);
2498 err = ext3_journal_stop(handle);
2504 static int ext3_write_dquot(struct dquot *dquot)
2508 struct inode *inode;
2510 inode = dquot_to_inode(dquot);
2511 handle = ext3_journal_start(inode,
2512 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2514 return PTR_ERR(handle);
2515 ret = dquot_commit(dquot);
2516 err = ext3_journal_stop(handle);
2522 static int ext3_acquire_dquot(struct dquot *dquot)
2527 handle = ext3_journal_start(dquot_to_inode(dquot),
2528 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2530 return PTR_ERR(handle);
2531 ret = dquot_acquire(dquot);
2532 err = ext3_journal_stop(handle);
2538 static int ext3_release_dquot(struct dquot *dquot)
2543 handle = ext3_journal_start(dquot_to_inode(dquot),
2544 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2546 return PTR_ERR(handle);
2547 ret = dquot_release(dquot);
2548 err = ext3_journal_stop(handle);
2554 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2556 /* Are we journalling quotas? */
2557 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2558 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2559 dquot_mark_dquot_dirty(dquot);
2560 return ext3_write_dquot(dquot);
2562 return dquot_mark_dquot_dirty(dquot);
2566 static int ext3_write_info(struct super_block *sb, int type)
2571 /* Data block + inode block */
2572 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2574 return PTR_ERR(handle);
2575 ret = dquot_commit_info(sb, type);
2576 err = ext3_journal_stop(handle);
2583 * Turn on quotas during mount time - we need to find
2584 * the quota file and such...
2586 static int ext3_quota_on_mount(struct super_block *sb, int type)
2588 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2589 EXT3_SB(sb)->s_jquota_fmt, type);
2593 * Standard function to be called on quota_on
2595 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2599 struct nameidata nd;
2601 if (!test_opt(sb, QUOTA))
2603 /* Not journalling quota? */
2604 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2605 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2606 return vfs_quota_on(sb, type, format_id, path);
2607 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2610 /* Quotafile not on the same filesystem? */
2611 if (nd.mnt->mnt_sb != sb) {
2615 /* Quotafile not of fs root? */
2616 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2618 "EXT3-fs: Quota file not on filesystem root. "
2619 "Journalled quota will not work.\n");
2621 return vfs_quota_on(sb, type, format_id, path);
2624 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2625 * acquiring the locks... As quota files are never truncated and quota code
2626 * itself serializes the operations (and noone else should touch the files)
2627 * we don't have to be afraid of races */
2628 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2629 size_t len, loff_t off)
2631 struct inode *inode = sb_dqopt(sb)->files[type];
2632 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2634 int offset = off & (sb->s_blocksize - 1);
2637 struct buffer_head *bh;
2638 loff_t i_size = i_size_read(inode);
2642 if (off+len > i_size)
2645 while (toread > 0) {
2646 tocopy = sb->s_blocksize - offset < toread ?
2647 sb->s_blocksize - offset : toread;
2648 bh = ext3_bread(NULL, inode, blk, 0, &err);
2651 if (!bh) /* A hole? */
2652 memset(data, 0, tocopy);
2654 memcpy(data, bh->b_data+offset, tocopy);
2664 /* Write to quotafile (we know the transaction is already started and has
2665 * enough credits) */
2666 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2667 const char *data, size_t len, loff_t off)
2669 struct inode *inode = sb_dqopt(sb)->files[type];
2670 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2672 int offset = off & (sb->s_blocksize - 1);
2674 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2675 size_t towrite = len;
2676 struct buffer_head *bh;
2677 handle_t *handle = journal_current_handle();
2679 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2680 while (towrite > 0) {
2681 tocopy = sb->s_blocksize - offset < towrite ?
2682 sb->s_blocksize - offset : towrite;
2683 bh = ext3_bread(handle, inode, blk, 1, &err);
2686 if (journal_quota) {
2687 err = ext3_journal_get_write_access(handle, bh);
2694 memcpy(bh->b_data+offset, data, tocopy);
2695 flush_dcache_page(bh->b_page);
2698 err = ext3_journal_dirty_metadata(handle, bh);
2700 /* Always do at least ordered writes for quotas */
2701 err = ext3_journal_dirty_data(handle, bh);
2702 mark_buffer_dirty(bh);
2715 if (inode->i_size < off+len-towrite) {
2716 i_size_write(inode, off+len-towrite);
2717 EXT3_I(inode)->i_disksize = inode->i_size;
2720 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2721 ext3_mark_inode_dirty(handle, inode);
2722 mutex_unlock(&inode->i_mutex);
2723 return len - towrite;
2728 static int ext3_get_sb(struct file_system_type *fs_type,
2729 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2731 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2734 static struct file_system_type ext3_fs_type = {
2735 .owner = THIS_MODULE,
2737 .get_sb = ext3_get_sb,
2738 .kill_sb = kill_block_super,
2739 .fs_flags = FS_REQUIRES_DEV,
2742 static int __init init_ext3_fs(void)
2744 int err = init_ext3_xattr();
2747 err = init_inodecache();
2750 err = register_filesystem(&ext3_fs_type);
2755 destroy_inodecache();
2761 static void __exit exit_ext3_fs(void)
2763 unregister_filesystem(&ext3_fs_type);
2764 destroy_inodecache();
2768 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2769 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2770 MODULE_LICENSE("GPL");
2771 module_init(init_ext3_fs)
2772 module_exit(exit_ext3_fs)