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/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
41 #include <asm/uaccess.h>
47 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
49 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
51 static void ext3_commit_super (struct super_block * sb,
52 struct ext3_super_block * es,
54 static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56 static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58 static int ext3_sync_fs(struct super_block *sb, int wait);
59 static const char *ext3_decode_error(struct super_block * sb, int errno,
61 static int ext3_remount (struct super_block * sb, int * flags, char * data);
62 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
63 static void ext3_unlockfs(struct super_block *sb);
64 static void ext3_write_super (struct super_block * sb);
65 static void ext3_write_super_lockfs(struct super_block *sb);
68 * Wrappers for journal_start/end.
70 * The only special thing we need to do here is to make sure that all
71 * journal_end calls result in the superblock being marked dirty, so
72 * that sync() will call the filesystem's write_super callback if
75 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
79 if (sb->s_flags & MS_RDONLY)
80 return ERR_PTR(-EROFS);
82 /* Special case here: if the journal has aborted behind our
83 * backs (eg. EIO in the commit thread), then we still need to
84 * take the FS itself readonly cleanly. */
85 journal = EXT3_SB(sb)->s_journal;
86 if (is_journal_aborted(journal)) {
87 ext3_abort(sb, __FUNCTION__,
88 "Detected aborted journal");
89 return ERR_PTR(-EROFS);
92 return journal_start(journal, nblocks);
96 * The only special thing we need to do here is to make sure that all
97 * journal_stop calls result in the superblock being marked dirty, so
98 * that sync() will call the filesystem's write_super callback if
101 int __ext3_journal_stop(const char *where, handle_t *handle)
103 struct super_block *sb;
107 sb = handle->h_transaction->t_journal->j_private;
109 rc = journal_stop(handle);
114 __ext3_std_error(sb, where, err);
118 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
119 struct buffer_head *bh, handle_t *handle, int err)
122 const char *errstr = ext3_decode_error(NULL, err, nbuf);
125 BUFFER_TRACE(bh, "abort");
130 if (is_handle_aborted(handle))
133 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
134 caller, errstr, err_fn);
136 journal_abort_handle(handle);
139 /* Deal with the reporting of failure conditions on a filesystem such as
140 * inconsistencies detected or read IO failures.
142 * On ext2, we can store the error state of the filesystem in the
143 * superblock. That is not possible on ext3, because we may have other
144 * write ordering constraints on the superblock which prevent us from
145 * writing it out straight away; and given that the journal is about to
146 * be aborted, we can't rely on the current, or future, transactions to
147 * write out the superblock safely.
149 * We'll just use the journal_abort() error code to record an error in
150 * the journal instead. On recovery, the journal will compain about
151 * that error until we've noted it down and cleared it.
154 static void ext3_handle_error(struct super_block *sb)
156 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
158 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
159 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
161 if (sb->s_flags & MS_RDONLY)
164 if (!test_opt (sb, ERRORS_CONT)) {
165 journal_t *journal = EXT3_SB(sb)->s_journal;
167 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
169 journal_abort(journal, -EIO);
171 if (test_opt (sb, ERRORS_RO)) {
172 printk (KERN_CRIT "Remounting filesystem read-only\n");
173 sb->s_flags |= MS_RDONLY;
175 ext3_commit_super(sb, es, 1);
176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
181 void ext3_error (struct super_block * sb, const char * function,
182 const char * fmt, ...)
187 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
192 ext3_handle_error(sb);
195 static const char *ext3_decode_error(struct super_block * sb, int errno,
202 errstr = "IO failure";
205 errstr = "Out of memory";
208 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
209 errstr = "Journal has aborted";
211 errstr = "Readonly filesystem";
214 /* If the caller passed in an extra buffer for unknown
215 * errors, textualise them now. Else we just return
218 /* Check for truncated error codes... */
219 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
228 /* __ext3_std_error decodes expected errors from journaling functions
229 * automatically and invokes the appropriate error response. */
231 void __ext3_std_error (struct super_block * sb, const char * function,
237 /* Special case: if the error is EROFS, and we're not already
238 * inside a transaction, then there's really no point in logging
240 if (errno == -EROFS && journal_current_handle() == NULL &&
241 (sb->s_flags & MS_RDONLY))
244 errstr = ext3_decode_error(sb, errno, nbuf);
245 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
246 sb->s_id, function, errstr);
248 ext3_handle_error(sb);
252 * ext3_abort is a much stronger failure handler than ext3_error. The
253 * abort function may be used to deal with unrecoverable failures such
254 * as journal IO errors or ENOMEM at a critical moment in log management.
256 * We unconditionally force the filesystem into an ABORT|READONLY state,
257 * unless the error response on the fs has been set to panic in which
258 * case we take the easy way out and panic immediately.
261 void ext3_abort (struct super_block * sb, const char * function,
262 const char * fmt, ...)
266 printk (KERN_CRIT "ext3_abort called.\n");
269 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
274 if (test_opt(sb, ERRORS_PANIC))
275 panic("EXT3-fs panic from previous error\n");
277 if (sb->s_flags & MS_RDONLY)
280 printk(KERN_CRIT "Remounting filesystem read-only\n");
281 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
282 sb->s_flags |= MS_RDONLY;
283 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
284 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
287 void ext3_warning (struct super_block * sb, const char * function,
288 const char * fmt, ...)
293 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
300 void ext3_update_dynamic_rev(struct super_block *sb)
302 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
304 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
307 ext3_warning(sb, __FUNCTION__,
308 "updating to rev %d because of new feature flag, "
309 "running e2fsck is recommended",
312 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
313 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
314 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
315 /* leave es->s_feature_*compat flags alone */
316 /* es->s_uuid will be set by e2fsck if empty */
319 * The rest of the superblock fields should be zero, and if not it
320 * means they are likely already in use, so leave them alone. We
321 * can leave it up to e2fsck to clean up any inconsistencies there.
326 * Open the external journal device
328 static struct block_device *ext3_blkdev_get(dev_t dev)
330 struct block_device *bdev;
331 char b[BDEVNAME_SIZE];
333 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
339 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
340 __bdevname(dev, b), PTR_ERR(bdev));
345 * Release the journal device
347 static int ext3_blkdev_put(struct block_device *bdev)
350 return blkdev_put(bdev);
353 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
355 struct block_device *bdev;
358 bdev = sbi->journal_bdev;
360 ret = ext3_blkdev_put(bdev);
361 sbi->journal_bdev = NULL;
366 static inline struct inode *orphan_list_entry(struct list_head *l)
368 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
371 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
375 printk(KERN_ERR "sb orphan head is %d\n",
376 le32_to_cpu(sbi->s_es->s_last_orphan));
378 printk(KERN_ERR "sb_info orphan list:\n");
379 list_for_each(l, &sbi->s_orphan) {
380 struct inode *inode = orphan_list_entry(l);
382 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
383 inode->i_sb->s_id, inode->i_ino, inode,
384 inode->i_mode, inode->i_nlink,
389 static void ext3_put_super (struct super_block * sb)
391 struct ext3_sb_info *sbi = EXT3_SB(sb);
392 struct ext3_super_block *es = sbi->s_es;
395 ext3_xattr_put_super(sb);
396 journal_destroy(sbi->s_journal);
397 if (!(sb->s_flags & MS_RDONLY)) {
398 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
399 es->s_state = cpu_to_le16(sbi->s_mount_state);
400 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
401 mark_buffer_dirty(sbi->s_sbh);
402 ext3_commit_super(sb, es, 1);
405 for (i = 0; i < sbi->s_gdb_count; i++)
406 brelse(sbi->s_group_desc[i]);
407 kfree(sbi->s_group_desc);
408 percpu_counter_destroy(&sbi->s_freeblocks_counter);
409 percpu_counter_destroy(&sbi->s_freeinodes_counter);
410 percpu_counter_destroy(&sbi->s_dirs_counter);
413 for (i = 0; i < MAXQUOTAS; i++)
414 kfree(sbi->s_qf_names[i]);
417 /* Debugging code just in case the in-memory inode orphan list
418 * isn't empty. The on-disk one can be non-empty if we've
419 * detected an error and taken the fs readonly, but the
420 * in-memory list had better be clean by this point. */
421 if (!list_empty(&sbi->s_orphan))
422 dump_orphan_list(sb, sbi);
423 J_ASSERT(list_empty(&sbi->s_orphan));
425 invalidate_bdev(sb->s_bdev);
426 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
428 * Invalidate the journal device's buffers. We don't want them
429 * floating about in memory - the physical journal device may
430 * hotswapped, and it breaks the `ro-after' testing code.
432 sync_blockdev(sbi->journal_bdev);
433 invalidate_bdev(sbi->journal_bdev);
434 ext3_blkdev_remove(sbi);
436 sb->s_fs_info = NULL;
441 static struct kmem_cache *ext3_inode_cachep;
444 * Called inside transaction, so use GFP_NOFS
446 static struct inode *ext3_alloc_inode(struct super_block *sb)
448 struct ext3_inode_info *ei;
450 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
453 #ifdef CONFIG_EXT3_FS_POSIX_ACL
454 ei->i_acl = EXT3_ACL_NOT_CACHED;
455 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
457 ei->i_block_alloc_info = NULL;
458 ei->vfs_inode.i_version = 1;
459 return &ei->vfs_inode;
462 static void ext3_destroy_inode(struct inode *inode)
464 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
465 printk("EXT3 Inode %p: orphan list check failed!\n",
467 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
468 EXT3_I(inode), sizeof(struct ext3_inode_info),
472 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
475 static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags)
477 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
479 INIT_LIST_HEAD(&ei->i_orphan);
480 #ifdef CONFIG_EXT3_FS_XATTR
481 init_rwsem(&ei->xattr_sem);
483 mutex_init(&ei->truncate_mutex);
484 inode_init_once(&ei->vfs_inode);
487 static int init_inodecache(void)
489 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
490 sizeof(struct ext3_inode_info),
491 0, (SLAB_RECLAIM_ACCOUNT|
494 if (ext3_inode_cachep == NULL)
499 static void destroy_inodecache(void)
501 kmem_cache_destroy(ext3_inode_cachep);
504 static void ext3_clear_inode(struct inode *inode)
506 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
507 #ifdef CONFIG_EXT3_FS_POSIX_ACL
508 if (EXT3_I(inode)->i_acl &&
509 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
510 posix_acl_release(EXT3_I(inode)->i_acl);
511 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
513 if (EXT3_I(inode)->i_default_acl &&
514 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
515 posix_acl_release(EXT3_I(inode)->i_default_acl);
516 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
519 ext3_discard_reservation(inode);
520 EXT3_I(inode)->i_block_alloc_info = NULL;
525 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
527 #if defined(CONFIG_QUOTA)
528 struct ext3_sb_info *sbi = EXT3_SB(sb);
530 if (sbi->s_jquota_fmt)
531 seq_printf(seq, ",jqfmt=%s",
532 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
534 if (sbi->s_qf_names[USRQUOTA])
535 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
537 if (sbi->s_qf_names[GRPQUOTA])
538 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
540 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
541 seq_puts(seq, ",usrquota");
543 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
544 seq_puts(seq, ",grpquota");
548 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
550 struct super_block *sb = vfs->mnt_sb;
552 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
553 seq_puts(seq, ",data=journal");
554 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
555 seq_puts(seq, ",data=ordered");
556 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
557 seq_puts(seq, ",data=writeback");
559 ext3_show_quota_options(seq, sb);
565 static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
568 unsigned long ino = objp[0];
569 __u32 generation = objp[1];
571 struct dentry *result;
573 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
574 return ERR_PTR(-ESTALE);
575 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
576 return ERR_PTR(-ESTALE);
578 /* iget isn't really right if the inode is currently unallocated!!
580 * ext3_read_inode will return a bad_inode if the inode had been
581 * deleted, so we should be safe.
583 * Currently we don't know the generation for parent directory, so
584 * a generation of 0 means "accept any"
586 inode = iget(sb, ino);
588 return ERR_PTR(-ENOMEM);
589 if (is_bad_inode(inode) ||
590 (generation && inode->i_generation != generation)) {
592 return ERR_PTR(-ESTALE);
594 /* now to find a dentry.
595 * If possible, get a well-connected one
597 result = d_alloc_anon(inode);
600 return ERR_PTR(-ENOMEM);
606 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
607 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
609 static int ext3_dquot_initialize(struct inode *inode, int type);
610 static int ext3_dquot_drop(struct inode *inode);
611 static int ext3_write_dquot(struct dquot *dquot);
612 static int ext3_acquire_dquot(struct dquot *dquot);
613 static int ext3_release_dquot(struct dquot *dquot);
614 static int ext3_mark_dquot_dirty(struct dquot *dquot);
615 static int ext3_write_info(struct super_block *sb, int type);
616 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
617 static int ext3_quota_on_mount(struct super_block *sb, int type);
618 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
619 size_t len, loff_t off);
620 static ssize_t ext3_quota_write(struct super_block *sb, int type,
621 const char *data, size_t len, loff_t off);
623 static struct dquot_operations ext3_quota_operations = {
624 .initialize = ext3_dquot_initialize,
625 .drop = ext3_dquot_drop,
626 .alloc_space = dquot_alloc_space,
627 .alloc_inode = dquot_alloc_inode,
628 .free_space = dquot_free_space,
629 .free_inode = dquot_free_inode,
630 .transfer = dquot_transfer,
631 .write_dquot = ext3_write_dquot,
632 .acquire_dquot = ext3_acquire_dquot,
633 .release_dquot = ext3_release_dquot,
634 .mark_dirty = ext3_mark_dquot_dirty,
635 .write_info = ext3_write_info
638 static struct quotactl_ops ext3_qctl_operations = {
639 .quota_on = ext3_quota_on,
640 .quota_off = vfs_quota_off,
641 .quota_sync = vfs_quota_sync,
642 .get_info = vfs_get_dqinfo,
643 .set_info = vfs_set_dqinfo,
644 .get_dqblk = vfs_get_dqblk,
645 .set_dqblk = vfs_set_dqblk
649 static const struct super_operations ext3_sops = {
650 .alloc_inode = ext3_alloc_inode,
651 .destroy_inode = ext3_destroy_inode,
652 .read_inode = ext3_read_inode,
653 .write_inode = ext3_write_inode,
654 .dirty_inode = ext3_dirty_inode,
655 .delete_inode = ext3_delete_inode,
656 .put_super = ext3_put_super,
657 .write_super = ext3_write_super,
658 .sync_fs = ext3_sync_fs,
659 .write_super_lockfs = ext3_write_super_lockfs,
660 .unlockfs = ext3_unlockfs,
661 .statfs = ext3_statfs,
662 .remount_fs = ext3_remount,
663 .clear_inode = ext3_clear_inode,
664 .show_options = ext3_show_options,
666 .quota_read = ext3_quota_read,
667 .quota_write = ext3_quota_write,
671 static struct export_operations ext3_export_ops = {
672 .get_parent = ext3_get_parent,
673 .get_dentry = ext3_get_dentry,
677 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
678 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
679 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
680 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
681 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
682 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
683 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
684 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
685 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
686 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
690 static match_table_t tokens = {
691 {Opt_bsd_df, "bsddf"},
692 {Opt_minix_df, "minixdf"},
693 {Opt_grpid, "grpid"},
694 {Opt_grpid, "bsdgroups"},
695 {Opt_nogrpid, "nogrpid"},
696 {Opt_nogrpid, "sysvgroups"},
697 {Opt_resgid, "resgid=%u"},
698 {Opt_resuid, "resuid=%u"},
700 {Opt_err_cont, "errors=continue"},
701 {Opt_err_panic, "errors=panic"},
702 {Opt_err_ro, "errors=remount-ro"},
703 {Opt_nouid32, "nouid32"},
704 {Opt_nocheck, "nocheck"},
705 {Opt_nocheck, "check=none"},
706 {Opt_debug, "debug"},
707 {Opt_oldalloc, "oldalloc"},
708 {Opt_orlov, "orlov"},
709 {Opt_user_xattr, "user_xattr"},
710 {Opt_nouser_xattr, "nouser_xattr"},
712 {Opt_noacl, "noacl"},
713 {Opt_reservation, "reservation"},
714 {Opt_noreservation, "noreservation"},
715 {Opt_noload, "noload"},
718 {Opt_commit, "commit=%u"},
719 {Opt_journal_update, "journal=update"},
720 {Opt_journal_inum, "journal=%u"},
721 {Opt_journal_dev, "journal_dev=%u"},
722 {Opt_abort, "abort"},
723 {Opt_data_journal, "data=journal"},
724 {Opt_data_ordered, "data=ordered"},
725 {Opt_data_writeback, "data=writeback"},
726 {Opt_offusrjquota, "usrjquota="},
727 {Opt_usrjquota, "usrjquota=%s"},
728 {Opt_offgrpjquota, "grpjquota="},
729 {Opt_grpjquota, "grpjquota=%s"},
730 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
731 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
732 {Opt_grpquota, "grpquota"},
733 {Opt_noquota, "noquota"},
734 {Opt_quota, "quota"},
735 {Opt_usrquota, "usrquota"},
736 {Opt_barrier, "barrier=%u"},
738 {Opt_resize, "resize"},
741 static ext3_fsblk_t get_sb_block(void **data)
743 ext3_fsblk_t sb_block;
744 char *options = (char *) *data;
746 if (!options || strncmp(options, "sb=", 3) != 0)
747 return 1; /* Default location */
749 /*todo: use simple_strtoll with >32bit ext3 */
750 sb_block = simple_strtoul(options, &options, 0);
751 if (*options && *options != ',') {
752 printk("EXT3-fs: Invalid sb specification: %s\n",
758 *data = (void *) options;
762 static int parse_options (char *options, struct super_block *sb,
763 unsigned int *inum, unsigned long *journal_devnum,
764 ext3_fsblk_t *n_blocks_count, int is_remount)
766 struct ext3_sb_info *sbi = EXT3_SB(sb);
768 substring_t args[MAX_OPT_ARGS];
779 while ((p = strsep (&options, ",")) != NULL) {
784 token = match_token(p, tokens, args);
787 clear_opt (sbi->s_mount_opt, MINIX_DF);
790 set_opt (sbi->s_mount_opt, MINIX_DF);
793 set_opt (sbi->s_mount_opt, GRPID);
796 clear_opt (sbi->s_mount_opt, GRPID);
799 if (match_int(&args[0], &option))
801 sbi->s_resuid = option;
804 if (match_int(&args[0], &option))
806 sbi->s_resgid = option;
809 /* handled by get_sb_block() instead of here */
810 /* *sb_block = match_int(&args[0]); */
813 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
814 clear_opt (sbi->s_mount_opt, ERRORS_RO);
815 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
818 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
819 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
820 set_opt (sbi->s_mount_opt, ERRORS_RO);
823 clear_opt (sbi->s_mount_opt, ERRORS_RO);
824 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
825 set_opt (sbi->s_mount_opt, ERRORS_CONT);
828 set_opt (sbi->s_mount_opt, NO_UID32);
831 clear_opt (sbi->s_mount_opt, CHECK);
834 set_opt (sbi->s_mount_opt, DEBUG);
837 set_opt (sbi->s_mount_opt, OLDALLOC);
840 clear_opt (sbi->s_mount_opt, OLDALLOC);
842 #ifdef CONFIG_EXT3_FS_XATTR
844 set_opt (sbi->s_mount_opt, XATTR_USER);
846 case Opt_nouser_xattr:
847 clear_opt (sbi->s_mount_opt, XATTR_USER);
851 case Opt_nouser_xattr:
852 printk("EXT3 (no)user_xattr options not supported\n");
855 #ifdef CONFIG_EXT3_FS_POSIX_ACL
857 set_opt(sbi->s_mount_opt, POSIX_ACL);
860 clear_opt(sbi->s_mount_opt, POSIX_ACL);
865 printk("EXT3 (no)acl options not supported\n");
868 case Opt_reservation:
869 set_opt(sbi->s_mount_opt, RESERVATION);
871 case Opt_noreservation:
872 clear_opt(sbi->s_mount_opt, RESERVATION);
874 case Opt_journal_update:
876 /* Eventually we will want to be able to create
877 a journal file here. For now, only allow the
878 user to specify an existing inode to be the
881 printk(KERN_ERR "EXT3-fs: cannot specify "
882 "journal on remount\n");
885 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
887 case Opt_journal_inum:
889 printk(KERN_ERR "EXT3-fs: cannot specify "
890 "journal on remount\n");
893 if (match_int(&args[0], &option))
897 case Opt_journal_dev:
899 printk(KERN_ERR "EXT3-fs: cannot specify "
900 "journal on remount\n");
903 if (match_int(&args[0], &option))
905 *journal_devnum = option;
908 set_opt (sbi->s_mount_opt, NOLOAD);
911 if (match_int(&args[0], &option))
916 option = JBD_DEFAULT_MAX_COMMIT_AGE;
917 sbi->s_commit_interval = HZ * option;
919 case Opt_data_journal:
920 data_opt = EXT3_MOUNT_JOURNAL_DATA;
922 case Opt_data_ordered:
923 data_opt = EXT3_MOUNT_ORDERED_DATA;
925 case Opt_data_writeback:
926 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
929 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
932 "EXT3-fs: cannot change data "
933 "mode on remount\n");
937 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
938 sbi->s_mount_opt |= data_opt;
948 if (sb_any_quota_enabled(sb)) {
950 "EXT3-fs: Cannot change journalled "
951 "quota options when quota turned on.\n");
954 qname = match_strdup(&args[0]);
957 "EXT3-fs: not enough memory for "
958 "storing quotafile name.\n");
961 if (sbi->s_qf_names[qtype] &&
962 strcmp(sbi->s_qf_names[qtype], qname)) {
964 "EXT3-fs: %s quota file already "
965 "specified.\n", QTYPE2NAME(qtype));
969 sbi->s_qf_names[qtype] = qname;
970 if (strchr(sbi->s_qf_names[qtype], '/')) {
972 "EXT3-fs: quotafile must be on "
973 "filesystem root.\n");
974 kfree(sbi->s_qf_names[qtype]);
975 sbi->s_qf_names[qtype] = NULL;
978 set_opt(sbi->s_mount_opt, QUOTA);
980 case Opt_offusrjquota:
983 case Opt_offgrpjquota:
986 if (sb_any_quota_enabled(sb)) {
987 printk(KERN_ERR "EXT3-fs: Cannot change "
988 "journalled quota options when "
989 "quota turned on.\n");
993 * The space will be released later when all options
994 * are confirmed to be correct
996 sbi->s_qf_names[qtype] = NULL;
998 case Opt_jqfmt_vfsold:
999 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1001 case Opt_jqfmt_vfsv0:
1002 sbi->s_jquota_fmt = QFMT_VFS_V0;
1006 set_opt(sbi->s_mount_opt, QUOTA);
1007 set_opt(sbi->s_mount_opt, USRQUOTA);
1010 set_opt(sbi->s_mount_opt, QUOTA);
1011 set_opt(sbi->s_mount_opt, GRPQUOTA);
1014 if (sb_any_quota_enabled(sb)) {
1015 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1016 "options when quota turned on.\n");
1019 clear_opt(sbi->s_mount_opt, QUOTA);
1020 clear_opt(sbi->s_mount_opt, USRQUOTA);
1021 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1029 case Opt_offusrjquota:
1030 case Opt_offgrpjquota:
1031 case Opt_jqfmt_vfsold:
1032 case Opt_jqfmt_vfsv0:
1034 "EXT3-fs: journalled quota options not "
1041 set_opt(sbi->s_mount_opt, ABORT);
1044 if (match_int(&args[0], &option))
1047 set_opt(sbi->s_mount_opt, BARRIER);
1049 clear_opt(sbi->s_mount_opt, BARRIER);
1055 printk("EXT3-fs: resize option only available "
1059 if (match_int(&args[0], &option) != 0)
1061 *n_blocks_count = option;
1064 set_opt(sbi->s_mount_opt, NOBH);
1067 clear_opt(sbi->s_mount_opt, NOBH);
1071 "EXT3-fs: Unrecognized mount option \"%s\" "
1072 "or missing value\n", p);
1077 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1078 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1079 sbi->s_qf_names[USRQUOTA])
1080 clear_opt(sbi->s_mount_opt, USRQUOTA);
1082 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1083 sbi->s_qf_names[GRPQUOTA])
1084 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1086 if ((sbi->s_qf_names[USRQUOTA] &&
1087 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1088 (sbi->s_qf_names[GRPQUOTA] &&
1089 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1090 printk(KERN_ERR "EXT3-fs: old and new quota "
1091 "format mixing.\n");
1095 if (!sbi->s_jquota_fmt) {
1096 printk(KERN_ERR "EXT3-fs: journalled quota format "
1097 "not specified.\n");
1101 if (sbi->s_jquota_fmt) {
1102 printk(KERN_ERR "EXT3-fs: journalled quota format "
1103 "specified with no journalling "
1112 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1115 struct ext3_sb_info *sbi = EXT3_SB(sb);
1118 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1119 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1120 "forcing read-only mode\n");
1125 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1126 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1127 "running e2fsck is recommended\n");
1128 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1129 printk (KERN_WARNING
1130 "EXT3-fs warning: mounting fs with errors, "
1131 "running e2fsck is recommended\n");
1132 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1133 le16_to_cpu(es->s_mnt_count) >=
1134 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1135 printk (KERN_WARNING
1136 "EXT3-fs warning: maximal mount count reached, "
1137 "running e2fsck is recommended\n");
1138 else if (le32_to_cpu(es->s_checkinterval) &&
1139 (le32_to_cpu(es->s_lastcheck) +
1140 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1141 printk (KERN_WARNING
1142 "EXT3-fs warning: checktime reached, "
1143 "running e2fsck is recommended\n");
1145 /* @@@ We _will_ want to clear the valid bit if we find
1146 inconsistencies, to force a fsck at reboot. But for
1147 a plain journaled filesystem we can keep it set as
1148 valid forever! :) */
1149 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1151 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1152 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1153 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1154 es->s_mtime = cpu_to_le32(get_seconds());
1155 ext3_update_dynamic_rev(sb);
1156 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1158 ext3_commit_super(sb, es, 1);
1159 if (test_opt(sb, DEBUG))
1160 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1161 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1163 sbi->s_groups_count,
1164 EXT3_BLOCKS_PER_GROUP(sb),
1165 EXT3_INODES_PER_GROUP(sb),
1168 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1169 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1170 char b[BDEVNAME_SIZE];
1172 printk("external journal on %s\n",
1173 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1175 printk("internal journal\n");
1180 /* Called at mount-time, super-block is locked */
1181 static int ext3_check_descriptors (struct super_block * sb)
1183 struct ext3_sb_info *sbi = EXT3_SB(sb);
1184 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1185 ext3_fsblk_t last_block;
1186 struct ext3_group_desc * gdp = NULL;
1190 ext3_debug ("Checking group descriptors");
1192 for (i = 0; i < sbi->s_groups_count; i++)
1194 if (i == sbi->s_groups_count - 1)
1195 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1197 last_block = first_block +
1198 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1200 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1201 gdp = (struct ext3_group_desc *)
1202 sbi->s_group_desc[desc_block++]->b_data;
1203 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1204 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1206 ext3_error (sb, "ext3_check_descriptors",
1207 "Block bitmap for group %d"
1208 " not in group (block %lu)!",
1210 le32_to_cpu(gdp->bg_block_bitmap));
1213 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1214 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1216 ext3_error (sb, "ext3_check_descriptors",
1217 "Inode bitmap for group %d"
1218 " not in group (block %lu)!",
1220 le32_to_cpu(gdp->bg_inode_bitmap));
1223 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1224 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >
1227 ext3_error (sb, "ext3_check_descriptors",
1228 "Inode table for group %d"
1229 " not in group (block %lu)!",
1231 le32_to_cpu(gdp->bg_inode_table));
1234 first_block += EXT3_BLOCKS_PER_GROUP(sb);
1238 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1239 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1244 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1245 * the superblock) which were deleted from all directories, but held open by
1246 * a process at the time of a crash. We walk the list and try to delete these
1247 * inodes at recovery time (only with a read-write filesystem).
1249 * In order to keep the orphan inode chain consistent during traversal (in
1250 * case of crash during recovery), we link each inode into the superblock
1251 * orphan list_head and handle it the same way as an inode deletion during
1252 * normal operation (which journals the operations for us).
1254 * We only do an iget() and an iput() on each inode, which is very safe if we
1255 * accidentally point at an in-use or already deleted inode. The worst that
1256 * can happen in this case is that we get a "bit already cleared" message from
1257 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1258 * e2fsck was run on this filesystem, and it must have already done the orphan
1259 * inode cleanup for us, so we can safely abort without any further action.
1261 static void ext3_orphan_cleanup (struct super_block * sb,
1262 struct ext3_super_block * es)
1264 unsigned int s_flags = sb->s_flags;
1265 int nr_orphans = 0, nr_truncates = 0;
1269 if (!es->s_last_orphan) {
1270 jbd_debug(4, "no orphan inodes to clean up\n");
1274 if (bdev_read_only(sb->s_bdev)) {
1275 printk(KERN_ERR "EXT3-fs: write access "
1276 "unavailable, skipping orphan cleanup.\n");
1280 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1281 if (es->s_last_orphan)
1282 jbd_debug(1, "Errors on filesystem, "
1283 "clearing orphan list.\n");
1284 es->s_last_orphan = 0;
1285 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1289 if (s_flags & MS_RDONLY) {
1290 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1292 sb->s_flags &= ~MS_RDONLY;
1295 /* Needed for iput() to work correctly and not trash data */
1296 sb->s_flags |= MS_ACTIVE;
1297 /* Turn on quotas so that they are updated correctly */
1298 for (i = 0; i < MAXQUOTAS; i++) {
1299 if (EXT3_SB(sb)->s_qf_names[i]) {
1300 int ret = ext3_quota_on_mount(sb, i);
1303 "EXT3-fs: Cannot turn on journalled "
1304 "quota: error %d\n", ret);
1309 while (es->s_last_orphan) {
1310 struct inode *inode;
1313 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1314 es->s_last_orphan = 0;
1318 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1320 if (inode->i_nlink) {
1322 "%s: truncating inode %lu to %Ld bytes\n",
1323 __FUNCTION__, inode->i_ino, inode->i_size);
1324 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1325 inode->i_ino, inode->i_size);
1326 ext3_truncate(inode);
1330 "%s: deleting unreferenced inode %lu\n",
1331 __FUNCTION__, inode->i_ino);
1332 jbd_debug(2, "deleting unreferenced inode %lu\n",
1336 iput(inode); /* The delete magic happens here! */
1339 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1342 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1343 sb->s_id, PLURAL(nr_orphans));
1345 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1346 sb->s_id, PLURAL(nr_truncates));
1348 /* Turn quotas off */
1349 for (i = 0; i < MAXQUOTAS; i++) {
1350 if (sb_dqopt(sb)->files[i])
1351 vfs_quota_off(sb, i);
1354 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1358 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1359 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1360 * We need to be 1 filesystem block less than the 2^32 sector limit.
1362 static loff_t ext3_max_size(int bits)
1364 loff_t res = EXT3_NDIR_BLOCKS;
1365 /* This constant is calculated to be the largest file size for a
1366 * dense, 4k-blocksize file such that the total number of
1367 * sectors in the file, including data and all indirect blocks,
1368 * does not exceed 2^32. */
1369 const loff_t upper_limit = 0x1ff7fffd000LL;
1371 res += 1LL << (bits-2);
1372 res += 1LL << (2*(bits-2));
1373 res += 1LL << (3*(bits-2));
1375 if (res > upper_limit)
1380 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1381 ext3_fsblk_t logic_sb_block,
1384 struct ext3_sb_info *sbi = EXT3_SB(sb);
1385 unsigned long bg, first_meta_bg;
1388 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1390 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1392 return (logic_sb_block + nr + 1);
1393 bg = sbi->s_desc_per_block * nr;
1394 if (ext3_bg_has_super(sb, bg))
1396 return (has_super + ext3_group_first_block_no(sb, bg));
1400 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1402 struct buffer_head * bh;
1403 struct ext3_super_block *es = NULL;
1404 struct ext3_sb_info *sbi;
1406 ext3_fsblk_t sb_block = get_sb_block(&data);
1407 ext3_fsblk_t logic_sb_block;
1408 unsigned long offset = 0;
1409 unsigned int journal_inum = 0;
1410 unsigned long journal_devnum = 0;
1411 unsigned long def_mount_opts;
1420 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1423 sb->s_fs_info = sbi;
1424 sbi->s_mount_opt = 0;
1425 sbi->s_resuid = EXT3_DEF_RESUID;
1426 sbi->s_resgid = EXT3_DEF_RESGID;
1430 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1432 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1437 * The ext3 superblock will not be buffer aligned for other than 1kB
1438 * block sizes. We need to calculate the offset from buffer start.
1440 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1441 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1442 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1444 logic_sb_block = sb_block;
1447 if (!(bh = sb_bread(sb, logic_sb_block))) {
1448 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1452 * Note: s_es must be initialized as soon as possible because
1453 * some ext3 macro-instructions depend on its value
1455 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1457 sb->s_magic = le16_to_cpu(es->s_magic);
1458 if (sb->s_magic != EXT3_SUPER_MAGIC)
1461 /* Set defaults before we parse the mount options */
1462 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1463 if (def_mount_opts & EXT3_DEFM_DEBUG)
1464 set_opt(sbi->s_mount_opt, DEBUG);
1465 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1466 set_opt(sbi->s_mount_opt, GRPID);
1467 if (def_mount_opts & EXT3_DEFM_UID16)
1468 set_opt(sbi->s_mount_opt, NO_UID32);
1469 #ifdef CONFIG_EXT3_FS_XATTR
1470 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1471 set_opt(sbi->s_mount_opt, XATTR_USER);
1473 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1474 if (def_mount_opts & EXT3_DEFM_ACL)
1475 set_opt(sbi->s_mount_opt, POSIX_ACL);
1477 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1478 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1479 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1480 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1481 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1482 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1484 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1485 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1486 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1487 set_opt(sbi->s_mount_opt, ERRORS_RO);
1489 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1491 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1492 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1494 set_opt(sbi->s_mount_opt, RESERVATION);
1496 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1500 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1501 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1503 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1504 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1505 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1506 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1508 "EXT3-fs warning: feature flags set on rev 0 fs, "
1509 "running e2fsck is recommended\n");
1511 * Check feature flags regardless of the revision level, since we
1512 * previously didn't change the revision level when setting the flags,
1513 * so there is a chance incompat flags are set on a rev 0 filesystem.
1515 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1517 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1518 "unsupported optional features (%x).\n",
1519 sb->s_id, le32_to_cpu(features));
1522 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1523 if (!(sb->s_flags & MS_RDONLY) && features) {
1524 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1525 "unsupported optional features (%x).\n",
1526 sb->s_id, le32_to_cpu(features));
1529 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1531 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1532 blocksize > EXT3_MAX_BLOCK_SIZE) {
1534 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1535 blocksize, sb->s_id);
1539 hblock = bdev_hardsect_size(sb->s_bdev);
1540 if (sb->s_blocksize != blocksize) {
1542 * Make sure the blocksize for the filesystem is larger
1543 * than the hardware sectorsize for the machine.
1545 if (blocksize < hblock) {
1546 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1547 "device blocksize %d.\n", blocksize, hblock);
1552 sb_set_blocksize(sb, blocksize);
1553 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1554 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1555 bh = sb_bread(sb, logic_sb_block);
1558 "EXT3-fs: Can't read superblock on 2nd try.\n");
1561 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1563 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1565 "EXT3-fs: Magic mismatch, very weird !\n");
1570 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1572 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1573 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1574 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1576 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1577 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1578 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1579 (!is_power_of_2(sbi->s_inode_size)) ||
1580 (sbi->s_inode_size > blocksize)) {
1582 "EXT3-fs: unsupported inode size: %d\n",
1587 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1588 le32_to_cpu(es->s_log_frag_size);
1589 if (blocksize != sbi->s_frag_size) {
1591 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1592 sbi->s_frag_size, blocksize);
1595 sbi->s_frags_per_block = 1;
1596 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1597 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1598 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1599 if (EXT3_INODE_SIZE(sb) == 0)
1601 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1602 if (sbi->s_inodes_per_block == 0)
1604 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1605 sbi->s_inodes_per_block;
1606 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1608 sbi->s_mount_state = le16_to_cpu(es->s_state);
1609 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1610 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1611 for (i=0; i < 4; i++)
1612 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1613 sbi->s_def_hash_version = es->s_def_hash_version;
1615 if (sbi->s_blocks_per_group > blocksize * 8) {
1617 "EXT3-fs: #blocks per group too big: %lu\n",
1618 sbi->s_blocks_per_group);
1621 if (sbi->s_frags_per_group > blocksize * 8) {
1623 "EXT3-fs: #fragments per group too big: %lu\n",
1624 sbi->s_frags_per_group);
1627 if (sbi->s_inodes_per_group > blocksize * 8) {
1629 "EXT3-fs: #inodes per group too big: %lu\n",
1630 sbi->s_inodes_per_group);
1634 if (le32_to_cpu(es->s_blocks_count) >
1635 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1636 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1637 " too large to mount safely\n", sb->s_id);
1638 if (sizeof(sector_t) < 8)
1639 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1644 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1646 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1647 le32_to_cpu(es->s_first_data_block) - 1)
1648 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1649 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1650 EXT3_DESC_PER_BLOCK(sb);
1651 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1653 if (sbi->s_group_desc == NULL) {
1654 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1658 bgl_lock_init(&sbi->s_blockgroup_lock);
1660 for (i = 0; i < db_count; i++) {
1661 block = descriptor_loc(sb, logic_sb_block, i);
1662 sbi->s_group_desc[i] = sb_bread(sb, block);
1663 if (!sbi->s_group_desc[i]) {
1664 printk (KERN_ERR "EXT3-fs: "
1665 "can't read group descriptor %d\n", i);
1670 if (!ext3_check_descriptors (sb)) {
1671 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1674 sbi->s_gdb_count = db_count;
1675 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1676 spin_lock_init(&sbi->s_next_gen_lock);
1678 percpu_counter_init(&sbi->s_freeblocks_counter,
1679 ext3_count_free_blocks(sb));
1680 percpu_counter_init(&sbi->s_freeinodes_counter,
1681 ext3_count_free_inodes(sb));
1682 percpu_counter_init(&sbi->s_dirs_counter,
1683 ext3_count_dirs(sb));
1685 /* per fileystem reservation list head & lock */
1686 spin_lock_init(&sbi->s_rsv_window_lock);
1687 sbi->s_rsv_window_root = RB_ROOT;
1688 /* Add a single, static dummy reservation to the start of the
1689 * reservation window list --- it gives us a placeholder for
1690 * append-at-start-of-list which makes the allocation logic
1691 * _much_ simpler. */
1692 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1693 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1694 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1695 sbi->s_rsv_window_head.rsv_goal_size = 0;
1696 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1699 * set up enough so that it can read an inode
1701 sb->s_op = &ext3_sops;
1702 sb->s_export_op = &ext3_export_ops;
1703 sb->s_xattr = ext3_xattr_handlers;
1705 sb->s_qcop = &ext3_qctl_operations;
1706 sb->dq_op = &ext3_quota_operations;
1708 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1712 needs_recovery = (es->s_last_orphan != 0 ||
1713 EXT3_HAS_INCOMPAT_FEATURE(sb,
1714 EXT3_FEATURE_INCOMPAT_RECOVER));
1717 * The first inode we look at is the journal inode. Don't try
1718 * root first: it may be modified in the journal!
1720 if (!test_opt(sb, NOLOAD) &&
1721 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1722 if (ext3_load_journal(sb, es, journal_devnum))
1724 } else if (journal_inum) {
1725 if (ext3_create_journal(sb, es, journal_inum))
1730 "ext3: No journal on filesystem on %s\n",
1735 /* We have now updated the journal if required, so we can
1736 * validate the data journaling mode. */
1737 switch (test_opt(sb, DATA_FLAGS)) {
1739 /* No mode set, assume a default based on the journal
1740 capabilities: ORDERED_DATA if the journal can
1741 cope, else JOURNAL_DATA */
1742 if (journal_check_available_features
1743 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1744 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1746 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1749 case EXT3_MOUNT_ORDERED_DATA:
1750 case EXT3_MOUNT_WRITEBACK_DATA:
1751 if (!journal_check_available_features
1752 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1753 printk(KERN_ERR "EXT3-fs: Journal does not support "
1754 "requested data journaling mode\n");
1761 if (test_opt(sb, NOBH)) {
1762 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1763 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1764 "its supported only with writeback mode\n");
1765 clear_opt(sbi->s_mount_opt, NOBH);
1769 * The journal_load will have done any necessary log recovery,
1770 * so we can safely mount the rest of the filesystem now.
1773 root = iget(sb, EXT3_ROOT_INO);
1774 sb->s_root = d_alloc_root(root);
1776 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1780 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1783 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1787 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1789 * akpm: core read_super() calls in here with the superblock locked.
1790 * That deadlocks, because orphan cleanup needs to lock the superblock
1791 * in numerous places. Here we just pop the lock - it's relatively
1792 * harmless, because we are now ready to accept write_super() requests,
1793 * and aviro says that's the only reason for hanging onto the
1796 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1797 ext3_orphan_cleanup(sb, es);
1798 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1800 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1801 ext3_mark_recovery_complete(sb, es);
1802 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1803 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1804 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1812 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1817 journal_destroy(sbi->s_journal);
1819 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1820 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1821 percpu_counter_destroy(&sbi->s_dirs_counter);
1823 for (i = 0; i < db_count; i++)
1824 brelse(sbi->s_group_desc[i]);
1825 kfree(sbi->s_group_desc);
1828 for (i = 0; i < MAXQUOTAS; i++)
1829 kfree(sbi->s_qf_names[i]);
1831 ext3_blkdev_remove(sbi);
1834 sb->s_fs_info = NULL;
1841 * Setup any per-fs journal parameters now. We'll do this both on
1842 * initial mount, once the journal has been initialised but before we've
1843 * done any recovery; and again on any subsequent remount.
1845 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1847 struct ext3_sb_info *sbi = EXT3_SB(sb);
1849 if (sbi->s_commit_interval)
1850 journal->j_commit_interval = sbi->s_commit_interval;
1851 /* We could also set up an ext3-specific default for the commit
1852 * interval here, but for now we'll just fall back to the jbd
1855 spin_lock(&journal->j_state_lock);
1856 if (test_opt(sb, BARRIER))
1857 journal->j_flags |= JFS_BARRIER;
1859 journal->j_flags &= ~JFS_BARRIER;
1860 spin_unlock(&journal->j_state_lock);
1863 static journal_t *ext3_get_journal(struct super_block *sb,
1864 unsigned int journal_inum)
1866 struct inode *journal_inode;
1869 /* First, test for the existence of a valid inode on disk. Bad
1870 * things happen if we iget() an unused inode, as the subsequent
1871 * iput() will try to delete it. */
1873 journal_inode = iget(sb, journal_inum);
1874 if (!journal_inode) {
1875 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1878 if (!journal_inode->i_nlink) {
1879 make_bad_inode(journal_inode);
1880 iput(journal_inode);
1881 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1885 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1886 journal_inode, journal_inode->i_size);
1887 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1888 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1889 iput(journal_inode);
1893 journal = journal_init_inode(journal_inode);
1895 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1896 iput(journal_inode);
1899 journal->j_private = sb;
1900 ext3_init_journal_params(sb, journal);
1904 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1907 struct buffer_head * bh;
1911 int hblock, blocksize;
1912 ext3_fsblk_t sb_block;
1913 unsigned long offset;
1914 struct ext3_super_block * es;
1915 struct block_device *bdev;
1917 bdev = ext3_blkdev_get(j_dev);
1921 if (bd_claim(bdev, sb)) {
1923 "EXT3: failed to claim external journal device.\n");
1928 blocksize = sb->s_blocksize;
1929 hblock = bdev_hardsect_size(bdev);
1930 if (blocksize < hblock) {
1932 "EXT3-fs: blocksize too small for journal device.\n");
1936 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1937 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1938 set_blocksize(bdev, blocksize);
1939 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1940 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1941 "external journal\n");
1945 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1946 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1947 !(le32_to_cpu(es->s_feature_incompat) &
1948 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1949 printk(KERN_ERR "EXT3-fs: external journal has "
1950 "bad superblock\n");
1955 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1956 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1961 len = le32_to_cpu(es->s_blocks_count);
1962 start = sb_block + 1;
1963 brelse(bh); /* we're done with the superblock */
1965 journal = journal_init_dev(bdev, sb->s_bdev,
1966 start, len, blocksize);
1968 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1971 journal->j_private = sb;
1972 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1973 wait_on_buffer(journal->j_sb_buffer);
1974 if (!buffer_uptodate(journal->j_sb_buffer)) {
1975 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1978 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1979 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1980 "user (unsupported) - %d\n",
1981 be32_to_cpu(journal->j_superblock->s_nr_users));
1984 EXT3_SB(sb)->journal_bdev = bdev;
1985 ext3_init_journal_params(sb, journal);
1988 journal_destroy(journal);
1990 ext3_blkdev_put(bdev);
1994 static int ext3_load_journal(struct super_block *sb,
1995 struct ext3_super_block *es,
1996 unsigned long journal_devnum)
1999 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2002 int really_read_only;
2004 if (journal_devnum &&
2005 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2006 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2007 "numbers have changed\n");
2008 journal_dev = new_decode_dev(journal_devnum);
2010 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2012 really_read_only = bdev_read_only(sb->s_bdev);
2015 * Are we loading a blank journal or performing recovery after a
2016 * crash? For recovery, we need to check in advance whether we
2017 * can get read-write access to the device.
2020 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2021 if (sb->s_flags & MS_RDONLY) {
2022 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2023 "required on readonly filesystem.\n");
2024 if (really_read_only) {
2025 printk(KERN_ERR "EXT3-fs: write access "
2026 "unavailable, cannot proceed.\n");
2029 printk (KERN_INFO "EXT3-fs: write access will "
2030 "be enabled during recovery.\n");
2034 if (journal_inum && journal_dev) {
2035 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2036 "and inode journals!\n");
2041 if (!(journal = ext3_get_journal(sb, journal_inum)))
2044 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2048 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2049 err = journal_update_format(journal);
2051 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2052 journal_destroy(journal);
2057 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2058 err = journal_wipe(journal, !really_read_only);
2060 err = journal_load(journal);
2063 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2064 journal_destroy(journal);
2068 EXT3_SB(sb)->s_journal = journal;
2069 ext3_clear_journal_err(sb, es);
2071 if (journal_devnum &&
2072 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2073 es->s_journal_dev = cpu_to_le32(journal_devnum);
2076 /* Make sure we flush the recovery flag to disk. */
2077 ext3_commit_super(sb, es, 1);
2083 static int ext3_create_journal(struct super_block * sb,
2084 struct ext3_super_block * es,
2085 unsigned int journal_inum)
2090 if (sb->s_flags & MS_RDONLY) {
2091 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2092 "create journal.\n");
2096 journal = ext3_get_journal(sb, journal_inum);
2100 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2103 err = journal_create(journal);
2105 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2106 journal_destroy(journal);
2110 EXT3_SB(sb)->s_journal = journal;
2112 ext3_update_dynamic_rev(sb);
2113 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2114 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2116 es->s_journal_inum = cpu_to_le32(journal_inum);
2119 /* Make sure we flush the recovery flag to disk. */
2120 ext3_commit_super(sb, es, 1);
2125 static void ext3_commit_super (struct super_block * sb,
2126 struct ext3_super_block * es,
2129 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2133 es->s_wtime = cpu_to_le32(get_seconds());
2134 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2135 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2136 BUFFER_TRACE(sbh, "marking dirty");
2137 mark_buffer_dirty(sbh);
2139 sync_dirty_buffer(sbh);
2144 * Have we just finished recovery? If so, and if we are mounting (or
2145 * remounting) the filesystem readonly, then we will end up with a
2146 * consistent fs on disk. Record that fact.
2148 static void ext3_mark_recovery_complete(struct super_block * sb,
2149 struct ext3_super_block * es)
2151 journal_t *journal = EXT3_SB(sb)->s_journal;
2153 journal_lock_updates(journal);
2154 journal_flush(journal);
2156 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2157 sb->s_flags & MS_RDONLY) {
2158 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2160 ext3_commit_super(sb, es, 1);
2163 journal_unlock_updates(journal);
2167 * If we are mounting (or read-write remounting) a filesystem whose journal
2168 * has recorded an error from a previous lifetime, move that error to the
2169 * main filesystem now.
2171 static void ext3_clear_journal_err(struct super_block * sb,
2172 struct ext3_super_block * es)
2178 journal = EXT3_SB(sb)->s_journal;
2181 * Now check for any error status which may have been recorded in the
2182 * journal by a prior ext3_error() or ext3_abort()
2185 j_errno = journal_errno(journal);
2189 errstr = ext3_decode_error(sb, j_errno, nbuf);
2190 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2191 "from previous mount: %s", errstr);
2192 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2193 "filesystem check.");
2195 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2196 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2197 ext3_commit_super (sb, es, 1);
2199 journal_clear_err(journal);
2204 * Force the running and committing transactions to commit,
2205 * and wait on the commit.
2207 int ext3_force_commit(struct super_block *sb)
2212 if (sb->s_flags & MS_RDONLY)
2215 journal = EXT3_SB(sb)->s_journal;
2217 ret = ext3_journal_force_commit(journal);
2222 * Ext3 always journals updates to the superblock itself, so we don't
2223 * have to propagate any other updates to the superblock on disk at this
2224 * point. Just start an async writeback to get the buffers on their way
2227 * This implicitly triggers the writebehind on sync().
2230 static void ext3_write_super (struct super_block * sb)
2232 if (mutex_trylock(&sb->s_lock) != 0)
2237 static int ext3_sync_fs(struct super_block *sb, int wait)
2242 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2244 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2250 * LVM calls this function before a (read-only) snapshot is created. This
2251 * gives us a chance to flush the journal completely and mark the fs clean.
2253 static void ext3_write_super_lockfs(struct super_block *sb)
2257 if (!(sb->s_flags & MS_RDONLY)) {
2258 journal_t *journal = EXT3_SB(sb)->s_journal;
2260 /* Now we set up the journal barrier. */
2261 journal_lock_updates(journal);
2262 journal_flush(journal);
2264 /* Journal blocked and flushed, clear needs_recovery flag. */
2265 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2266 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2271 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2272 * flag here, even though the filesystem is not technically dirty yet.
2274 static void ext3_unlockfs(struct super_block *sb)
2276 if (!(sb->s_flags & MS_RDONLY)) {
2278 /* Reser the needs_recovery flag before the fs is unlocked. */
2279 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2280 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2282 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2286 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2288 struct ext3_super_block * es;
2289 struct ext3_sb_info *sbi = EXT3_SB(sb);
2290 ext3_fsblk_t n_blocks_count = 0;
2291 unsigned long old_sb_flags;
2292 struct ext3_mount_options old_opts;
2298 /* Store the original options */
2299 old_sb_flags = sb->s_flags;
2300 old_opts.s_mount_opt = sbi->s_mount_opt;
2301 old_opts.s_resuid = sbi->s_resuid;
2302 old_opts.s_resgid = sbi->s_resgid;
2303 old_opts.s_commit_interval = sbi->s_commit_interval;
2305 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2306 for (i = 0; i < MAXQUOTAS; i++)
2307 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2311 * Allow the "check" option to be passed as a remount option.
2313 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2318 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2319 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2321 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2322 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2326 ext3_init_journal_params(sb, sbi->s_journal);
2328 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2329 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2330 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2335 if (*flags & MS_RDONLY) {
2337 * First of all, the unconditional stuff we have to do
2338 * to disable replay of the journal when we next remount
2340 sb->s_flags |= MS_RDONLY;
2343 * OK, test if we are remounting a valid rw partition
2344 * readonly, and if so set the rdonly flag and then
2345 * mark the partition as valid again.
2347 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2348 (sbi->s_mount_state & EXT3_VALID_FS))
2349 es->s_state = cpu_to_le16(sbi->s_mount_state);
2352 * We have to unlock super so that we can wait for
2356 ext3_mark_recovery_complete(sb, es);
2360 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2361 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2362 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2363 "remount RDWR because of unsupported "
2364 "optional features (%x).\n",
2365 sb->s_id, le32_to_cpu(ret));
2371 * If we have an unprocessed orphan list hanging
2372 * around from a previously readonly bdev mount,
2373 * require a full umount/remount for now.
2375 if (es->s_last_orphan) {
2376 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2377 "remount RDWR because of unprocessed "
2378 "orphan inode list. Please "
2379 "umount/remount instead.\n",
2386 * Mounting a RDONLY partition read-write, so reread
2387 * and store the current valid flag. (It may have
2388 * been changed by e2fsck since we originally mounted
2391 ext3_clear_journal_err(sb, es);
2392 sbi->s_mount_state = le16_to_cpu(es->s_state);
2393 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2395 if (!ext3_setup_super (sb, es, 0))
2396 sb->s_flags &= ~MS_RDONLY;
2400 /* Release old quota file names */
2401 for (i = 0; i < MAXQUOTAS; i++)
2402 if (old_opts.s_qf_names[i] &&
2403 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2404 kfree(old_opts.s_qf_names[i]);
2408 sb->s_flags = old_sb_flags;
2409 sbi->s_mount_opt = old_opts.s_mount_opt;
2410 sbi->s_resuid = old_opts.s_resuid;
2411 sbi->s_resgid = old_opts.s_resgid;
2412 sbi->s_commit_interval = old_opts.s_commit_interval;
2414 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2415 for (i = 0; i < MAXQUOTAS; i++) {
2416 if (sbi->s_qf_names[i] &&
2417 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2418 kfree(sbi->s_qf_names[i]);
2419 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2425 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2427 struct super_block *sb = dentry->d_sb;
2428 struct ext3_sb_info *sbi = EXT3_SB(sb);
2429 struct ext3_super_block *es = sbi->s_es;
2432 if (test_opt(sb, MINIX_DF)) {
2433 sbi->s_overhead_last = 0;
2434 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2435 unsigned long ngroups = sbi->s_groups_count, i;
2436 ext3_fsblk_t overhead = 0;
2440 * Compute the overhead (FS structures). This is constant
2441 * for a given filesystem unless the number of block groups
2442 * changes so we cache the previous value until it does.
2446 * All of the blocks before first_data_block are
2449 overhead = le32_to_cpu(es->s_first_data_block);
2452 * Add the overhead attributed to the superblock and
2453 * block group descriptors. If the sparse superblocks
2454 * feature is turned on, then not all groups have this.
2456 for (i = 0; i < ngroups; i++) {
2457 overhead += ext3_bg_has_super(sb, i) +
2458 ext3_bg_num_gdb(sb, i);
2463 * Every block group has an inode bitmap, a block
2464 * bitmap, and an inode table.
2466 overhead += ngroups * (2 + sbi->s_itb_per_group);
2467 sbi->s_overhead_last = overhead;
2469 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2472 buf->f_type = EXT3_SUPER_MAGIC;
2473 buf->f_bsize = sb->s_blocksize;
2474 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2475 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
2476 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2477 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2478 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2480 buf->f_files = le32_to_cpu(es->s_inodes_count);
2481 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
2482 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2483 buf->f_namelen = EXT3_NAME_LEN;
2484 fsid = le64_to_cpup((void *)es->s_uuid) ^
2485 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2486 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2487 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2491 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2492 * is locked for write. Otherwise the are possible deadlocks:
2493 * Process 1 Process 2
2494 * ext3_create() quota_sync()
2495 * journal_start() write_dquot()
2496 * DQUOT_INIT() down(dqio_mutex)
2497 * down(dqio_mutex) journal_start()
2503 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2505 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2508 static int ext3_dquot_initialize(struct inode *inode, int type)
2513 /* We may create quota structure so we need to reserve enough blocks */
2514 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2516 return PTR_ERR(handle);
2517 ret = dquot_initialize(inode, type);
2518 err = ext3_journal_stop(handle);
2524 static int ext3_dquot_drop(struct inode *inode)
2529 /* We may delete quota structure so we need to reserve enough blocks */
2530 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2532 return PTR_ERR(handle);
2533 ret = dquot_drop(inode);
2534 err = ext3_journal_stop(handle);
2540 static int ext3_write_dquot(struct dquot *dquot)
2544 struct inode *inode;
2546 inode = dquot_to_inode(dquot);
2547 handle = ext3_journal_start(inode,
2548 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2550 return PTR_ERR(handle);
2551 ret = dquot_commit(dquot);
2552 err = ext3_journal_stop(handle);
2558 static int ext3_acquire_dquot(struct dquot *dquot)
2563 handle = ext3_journal_start(dquot_to_inode(dquot),
2564 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2566 return PTR_ERR(handle);
2567 ret = dquot_acquire(dquot);
2568 err = ext3_journal_stop(handle);
2574 static int ext3_release_dquot(struct dquot *dquot)
2579 handle = ext3_journal_start(dquot_to_inode(dquot),
2580 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2582 return PTR_ERR(handle);
2583 ret = dquot_release(dquot);
2584 err = ext3_journal_stop(handle);
2590 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2592 /* Are we journalling quotas? */
2593 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2594 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2595 dquot_mark_dquot_dirty(dquot);
2596 return ext3_write_dquot(dquot);
2598 return dquot_mark_dquot_dirty(dquot);
2602 static int ext3_write_info(struct super_block *sb, int type)
2607 /* Data block + inode block */
2608 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2610 return PTR_ERR(handle);
2611 ret = dquot_commit_info(sb, type);
2612 err = ext3_journal_stop(handle);
2619 * Turn on quotas during mount time - we need to find
2620 * the quota file and such...
2622 static int ext3_quota_on_mount(struct super_block *sb, int type)
2624 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2625 EXT3_SB(sb)->s_jquota_fmt, type);
2629 * Standard function to be called on quota_on
2631 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2635 struct nameidata nd;
2637 if (!test_opt(sb, QUOTA))
2639 /* Not journalling quota? */
2640 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2641 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2642 return vfs_quota_on(sb, type, format_id, path);
2643 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2646 /* Quotafile not on the same filesystem? */
2647 if (nd.mnt->mnt_sb != sb) {
2651 /* Quotafile not of fs root? */
2652 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2654 "EXT3-fs: Quota file not on filesystem root. "
2655 "Journalled quota will not work.\n");
2657 return vfs_quota_on(sb, type, format_id, path);
2660 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2661 * acquiring the locks... As quota files are never truncated and quota code
2662 * itself serializes the operations (and noone else should touch the files)
2663 * we don't have to be afraid of races */
2664 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2665 size_t len, loff_t off)
2667 struct inode *inode = sb_dqopt(sb)->files[type];
2668 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2670 int offset = off & (sb->s_blocksize - 1);
2673 struct buffer_head *bh;
2674 loff_t i_size = i_size_read(inode);
2678 if (off+len > i_size)
2681 while (toread > 0) {
2682 tocopy = sb->s_blocksize - offset < toread ?
2683 sb->s_blocksize - offset : toread;
2684 bh = ext3_bread(NULL, inode, blk, 0, &err);
2687 if (!bh) /* A hole? */
2688 memset(data, 0, tocopy);
2690 memcpy(data, bh->b_data+offset, tocopy);
2700 /* Write to quotafile (we know the transaction is already started and has
2701 * enough credits) */
2702 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2703 const char *data, size_t len, loff_t off)
2705 struct inode *inode = sb_dqopt(sb)->files[type];
2706 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2708 int offset = off & (sb->s_blocksize - 1);
2710 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2711 size_t towrite = len;
2712 struct buffer_head *bh;
2713 handle_t *handle = journal_current_handle();
2715 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2716 while (towrite > 0) {
2717 tocopy = sb->s_blocksize - offset < towrite ?
2718 sb->s_blocksize - offset : towrite;
2719 bh = ext3_bread(handle, inode, blk, 1, &err);
2722 if (journal_quota) {
2723 err = ext3_journal_get_write_access(handle, bh);
2730 memcpy(bh->b_data+offset, data, tocopy);
2731 flush_dcache_page(bh->b_page);
2734 err = ext3_journal_dirty_metadata(handle, bh);
2736 /* Always do at least ordered writes for quotas */
2737 err = ext3_journal_dirty_data(handle, bh);
2738 mark_buffer_dirty(bh);
2751 if (inode->i_size < off+len-towrite) {
2752 i_size_write(inode, off+len-towrite);
2753 EXT3_I(inode)->i_disksize = inode->i_size;
2756 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2757 ext3_mark_inode_dirty(handle, inode);
2758 mutex_unlock(&inode->i_mutex);
2759 return len - towrite;
2764 static int ext3_get_sb(struct file_system_type *fs_type,
2765 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2767 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2770 static struct file_system_type ext3_fs_type = {
2771 .owner = THIS_MODULE,
2773 .get_sb = ext3_get_sb,
2774 .kill_sb = kill_block_super,
2775 .fs_flags = FS_REQUIRES_DEV,
2778 static int __init init_ext3_fs(void)
2780 int err = init_ext3_xattr();
2783 err = init_inodecache();
2786 err = register_filesystem(&ext3_fs_type);
2791 destroy_inodecache();
2797 static void __exit exit_ext3_fs(void)
2799 unregister_filesystem(&ext3_fs_type);
2800 destroy_inodecache();
2804 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2805 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2806 MODULE_LICENSE("GPL");
2807 module_init(init_ext3_fs)
2808 module_exit(exit_ext3_fs)