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_CONT)) {
163 journal_t *journal = EXT3_SB(sb)->s_journal;
165 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
167 journal_abort(journal, -EIO);
169 if (test_opt (sb, ERRORS_RO)) {
170 printk (KERN_CRIT "Remounting filesystem read-only\n");
171 sb->s_flags |= MS_RDONLY;
173 ext3_commit_super(sb, es, 1);
174 if (test_opt(sb, ERRORS_PANIC))
175 panic("EXT3-fs (device %s): panic forced after error\n",
179 void ext3_error (struct super_block * sb, const char * function,
180 const char * fmt, ...)
185 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
190 ext3_handle_error(sb);
193 static const char *ext3_decode_error(struct super_block * sb, int errno,
200 errstr = "IO failure";
203 errstr = "Out of memory";
206 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
207 errstr = "Journal has aborted";
209 errstr = "Readonly filesystem";
212 /* If the caller passed in an extra buffer for unknown
213 * errors, textualise them now. Else we just return
216 /* Check for truncated error codes... */
217 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
226 /* __ext3_std_error decodes expected errors from journaling functions
227 * automatically and invokes the appropriate error response. */
229 void __ext3_std_error (struct super_block * sb, const char * function,
235 /* Special case: if the error is EROFS, and we're not already
236 * inside a transaction, then there's really no point in logging
238 if (errno == -EROFS && journal_current_handle() == NULL &&
239 (sb->s_flags & MS_RDONLY))
242 errstr = ext3_decode_error(sb, errno, nbuf);
243 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
244 sb->s_id, function, errstr);
246 ext3_handle_error(sb);
250 * ext3_abort is a much stronger failure handler than ext3_error. The
251 * abort function may be used to deal with unrecoverable failures such
252 * as journal IO errors or ENOMEM at a critical moment in log management.
254 * We unconditionally force the filesystem into an ABORT|READONLY state,
255 * unless the error response on the fs has been set to panic in which
256 * case we take the easy way out and panic immediately.
259 void ext3_abort (struct super_block * sb, const char * function,
260 const char * fmt, ...)
264 printk (KERN_CRIT "ext3_abort called.\n");
267 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
272 if (test_opt(sb, ERRORS_PANIC))
273 panic("EXT3-fs panic from previous error\n");
275 if (sb->s_flags & MS_RDONLY)
278 printk(KERN_CRIT "Remounting filesystem read-only\n");
279 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
280 sb->s_flags |= MS_RDONLY;
281 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
282 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
285 void ext3_warning (struct super_block * sb, const char * function,
286 const char * fmt, ...)
291 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
298 void ext3_update_dynamic_rev(struct super_block *sb)
300 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
302 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
305 ext3_warning(sb, __FUNCTION__,
306 "updating to rev %d because of new feature flag, "
307 "running e2fsck is recommended",
310 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
311 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
312 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
313 /* leave es->s_feature_*compat flags alone */
314 /* es->s_uuid will be set by e2fsck if empty */
317 * The rest of the superblock fields should be zero, and if not it
318 * means they are likely already in use, so leave them alone. We
319 * can leave it up to e2fsck to clean up any inconsistencies there.
324 * Open the external journal device
326 static struct block_device *ext3_blkdev_get(dev_t dev)
328 struct block_device *bdev;
329 char b[BDEVNAME_SIZE];
331 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
337 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
338 __bdevname(dev, b), PTR_ERR(bdev));
343 * Release the journal device
345 static int ext3_blkdev_put(struct block_device *bdev)
348 return blkdev_put(bdev);
351 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
353 struct block_device *bdev;
356 bdev = sbi->journal_bdev;
358 ret = ext3_blkdev_put(bdev);
359 sbi->journal_bdev = NULL;
364 static inline struct inode *orphan_list_entry(struct list_head *l)
366 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
369 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
373 printk(KERN_ERR "sb orphan head is %d\n",
374 le32_to_cpu(sbi->s_es->s_last_orphan));
376 printk(KERN_ERR "sb_info orphan list:\n");
377 list_for_each(l, &sbi->s_orphan) {
378 struct inode *inode = orphan_list_entry(l);
380 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
381 inode->i_sb->s_id, inode->i_ino, inode,
382 inode->i_mode, inode->i_nlink,
387 static void ext3_put_super (struct super_block * sb)
389 struct ext3_sb_info *sbi = EXT3_SB(sb);
390 struct ext3_super_block *es = sbi->s_es;
393 ext3_xattr_put_super(sb);
394 journal_destroy(sbi->s_journal);
395 if (!(sb->s_flags & MS_RDONLY)) {
396 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
397 es->s_state = cpu_to_le16(sbi->s_mount_state);
398 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
399 mark_buffer_dirty(sbi->s_sbh);
400 ext3_commit_super(sb, es, 1);
403 for (i = 0; i < sbi->s_gdb_count; i++)
404 brelse(sbi->s_group_desc[i]);
405 kfree(sbi->s_group_desc);
406 percpu_counter_destroy(&sbi->s_freeblocks_counter);
407 percpu_counter_destroy(&sbi->s_freeinodes_counter);
408 percpu_counter_destroy(&sbi->s_dirs_counter);
411 for (i = 0; i < MAXQUOTAS; i++)
412 kfree(sbi->s_qf_names[i]);
415 /* Debugging code just in case the in-memory inode orphan list
416 * isn't empty. The on-disk one can be non-empty if we've
417 * detected an error and taken the fs readonly, but the
418 * in-memory list had better be clean by this point. */
419 if (!list_empty(&sbi->s_orphan))
420 dump_orphan_list(sb, sbi);
421 J_ASSERT(list_empty(&sbi->s_orphan));
423 invalidate_bdev(sb->s_bdev, 0);
424 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
426 * Invalidate the journal device's buffers. We don't want them
427 * floating about in memory - the physical journal device may
428 * hotswapped, and it breaks the `ro-after' testing code.
430 sync_blockdev(sbi->journal_bdev);
431 invalidate_bdev(sbi->journal_bdev, 0);
432 ext3_blkdev_remove(sbi);
434 sb->s_fs_info = NULL;
439 static kmem_cache_t *ext3_inode_cachep;
442 * Called inside transaction, so use GFP_NOFS
444 static struct inode *ext3_alloc_inode(struct super_block *sb)
446 struct ext3_inode_info *ei;
448 ei = kmem_cache_alloc(ext3_inode_cachep, SLAB_NOFS);
451 #ifdef CONFIG_EXT3_FS_POSIX_ACL
452 ei->i_acl = EXT3_ACL_NOT_CACHED;
453 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
455 ei->i_block_alloc_info = NULL;
456 ei->vfs_inode.i_version = 1;
457 return &ei->vfs_inode;
460 static void ext3_destroy_inode(struct inode *inode)
462 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
465 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
467 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
469 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
470 SLAB_CTOR_CONSTRUCTOR) {
471 INIT_LIST_HEAD(&ei->i_orphan);
472 #ifdef CONFIG_EXT3_FS_XATTR
473 init_rwsem(&ei->xattr_sem);
475 mutex_init(&ei->truncate_mutex);
476 inode_init_once(&ei->vfs_inode);
480 static int init_inodecache(void)
482 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
483 sizeof(struct ext3_inode_info),
484 0, (SLAB_RECLAIM_ACCOUNT|
487 if (ext3_inode_cachep == NULL)
492 static void destroy_inodecache(void)
494 if (kmem_cache_destroy(ext3_inode_cachep))
495 printk(KERN_INFO "ext3_inode_cache: not all structures were freed\n");
498 static void ext3_clear_inode(struct inode *inode)
500 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
501 #ifdef CONFIG_EXT3_FS_POSIX_ACL
502 if (EXT3_I(inode)->i_acl &&
503 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
504 posix_acl_release(EXT3_I(inode)->i_acl);
505 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
507 if (EXT3_I(inode)->i_default_acl &&
508 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
509 posix_acl_release(EXT3_I(inode)->i_default_acl);
510 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
513 ext3_discard_reservation(inode);
514 EXT3_I(inode)->i_block_alloc_info = NULL;
519 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
521 #if defined(CONFIG_QUOTA)
522 struct ext3_sb_info *sbi = EXT3_SB(sb);
524 if (sbi->s_jquota_fmt)
525 seq_printf(seq, ",jqfmt=%s",
526 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
528 if (sbi->s_qf_names[USRQUOTA])
529 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
531 if (sbi->s_qf_names[GRPQUOTA])
532 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
534 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
535 seq_puts(seq, ",usrquota");
537 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
538 seq_puts(seq, ",grpquota");
542 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
544 struct super_block *sb = vfs->mnt_sb;
546 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
547 seq_puts(seq, ",data=journal");
548 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
549 seq_puts(seq, ",data=ordered");
550 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
551 seq_puts(seq, ",data=writeback");
553 ext3_show_quota_options(seq, sb);
559 static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
562 unsigned long ino = objp[0];
563 __u32 generation = objp[1];
565 struct dentry *result;
567 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
568 return ERR_PTR(-ESTALE);
569 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
570 return ERR_PTR(-ESTALE);
572 /* iget isn't really right if the inode is currently unallocated!!
574 * ext3_read_inode will return a bad_inode if the inode had been
575 * deleted, so we should be safe.
577 * Currently we don't know the generation for parent directory, so
578 * a generation of 0 means "accept any"
580 inode = iget(sb, ino);
582 return ERR_PTR(-ENOMEM);
583 if (is_bad_inode(inode) ||
584 (generation && inode->i_generation != generation)) {
586 return ERR_PTR(-ESTALE);
588 /* now to find a dentry.
589 * If possible, get a well-connected one
591 result = d_alloc_anon(inode);
594 return ERR_PTR(-ENOMEM);
600 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
601 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
603 static int ext3_dquot_initialize(struct inode *inode, int type);
604 static int ext3_dquot_drop(struct inode *inode);
605 static int ext3_write_dquot(struct dquot *dquot);
606 static int ext3_acquire_dquot(struct dquot *dquot);
607 static int ext3_release_dquot(struct dquot *dquot);
608 static int ext3_mark_dquot_dirty(struct dquot *dquot);
609 static int ext3_write_info(struct super_block *sb, int type);
610 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
611 static int ext3_quota_on_mount(struct super_block *sb, int type);
612 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
613 size_t len, loff_t off);
614 static ssize_t ext3_quota_write(struct super_block *sb, int type,
615 const char *data, size_t len, loff_t off);
617 static struct dquot_operations ext3_quota_operations = {
618 .initialize = ext3_dquot_initialize,
619 .drop = ext3_dquot_drop,
620 .alloc_space = dquot_alloc_space,
621 .alloc_inode = dquot_alloc_inode,
622 .free_space = dquot_free_space,
623 .free_inode = dquot_free_inode,
624 .transfer = dquot_transfer,
625 .write_dquot = ext3_write_dquot,
626 .acquire_dquot = ext3_acquire_dquot,
627 .release_dquot = ext3_release_dquot,
628 .mark_dirty = ext3_mark_dquot_dirty,
629 .write_info = ext3_write_info
632 static struct quotactl_ops ext3_qctl_operations = {
633 .quota_on = ext3_quota_on,
634 .quota_off = vfs_quota_off,
635 .quota_sync = vfs_quota_sync,
636 .get_info = vfs_get_dqinfo,
637 .set_info = vfs_set_dqinfo,
638 .get_dqblk = vfs_get_dqblk,
639 .set_dqblk = vfs_set_dqblk
643 static struct super_operations ext3_sops = {
644 .alloc_inode = ext3_alloc_inode,
645 .destroy_inode = ext3_destroy_inode,
646 .read_inode = ext3_read_inode,
647 .write_inode = ext3_write_inode,
648 .dirty_inode = ext3_dirty_inode,
649 .delete_inode = ext3_delete_inode,
650 .put_super = ext3_put_super,
651 .write_super = ext3_write_super,
652 .sync_fs = ext3_sync_fs,
653 .write_super_lockfs = ext3_write_super_lockfs,
654 .unlockfs = ext3_unlockfs,
655 .statfs = ext3_statfs,
656 .remount_fs = ext3_remount,
657 .clear_inode = ext3_clear_inode,
658 .show_options = ext3_show_options,
660 .quota_read = ext3_quota_read,
661 .quota_write = ext3_quota_write,
665 static struct export_operations ext3_export_ops = {
666 .get_parent = ext3_get_parent,
667 .get_dentry = ext3_get_dentry,
671 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
672 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
673 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
674 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
675 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
676 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
677 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
678 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
679 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
680 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
684 static match_table_t tokens = {
685 {Opt_bsd_df, "bsddf"},
686 {Opt_minix_df, "minixdf"},
687 {Opt_grpid, "grpid"},
688 {Opt_grpid, "bsdgroups"},
689 {Opt_nogrpid, "nogrpid"},
690 {Opt_nogrpid, "sysvgroups"},
691 {Opt_resgid, "resgid=%u"},
692 {Opt_resuid, "resuid=%u"},
694 {Opt_err_cont, "errors=continue"},
695 {Opt_err_panic, "errors=panic"},
696 {Opt_err_ro, "errors=remount-ro"},
697 {Opt_nouid32, "nouid32"},
698 {Opt_nocheck, "nocheck"},
699 {Opt_nocheck, "check=none"},
700 {Opt_debug, "debug"},
701 {Opt_oldalloc, "oldalloc"},
702 {Opt_orlov, "orlov"},
703 {Opt_user_xattr, "user_xattr"},
704 {Opt_nouser_xattr, "nouser_xattr"},
706 {Opt_noacl, "noacl"},
707 {Opt_reservation, "reservation"},
708 {Opt_noreservation, "noreservation"},
709 {Opt_noload, "noload"},
712 {Opt_commit, "commit=%u"},
713 {Opt_journal_update, "journal=update"},
714 {Opt_journal_inum, "journal=%u"},
715 {Opt_journal_dev, "journal_dev=%u"},
716 {Opt_abort, "abort"},
717 {Opt_data_journal, "data=journal"},
718 {Opt_data_ordered, "data=ordered"},
719 {Opt_data_writeback, "data=writeback"},
720 {Opt_offusrjquota, "usrjquota="},
721 {Opt_usrjquota, "usrjquota=%s"},
722 {Opt_offgrpjquota, "grpjquota="},
723 {Opt_grpjquota, "grpjquota=%s"},
724 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
725 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
726 {Opt_grpquota, "grpquota"},
727 {Opt_noquota, "noquota"},
728 {Opt_quota, "quota"},
729 {Opt_usrquota, "usrquota"},
730 {Opt_barrier, "barrier=%u"},
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);
825 clear_opt (sbi->s_mount_opt, CHECK);
828 set_opt (sbi->s_mount_opt, DEBUG);
831 set_opt (sbi->s_mount_opt, OLDALLOC);
834 clear_opt (sbi->s_mount_opt, OLDALLOC);
836 #ifdef CONFIG_EXT3_FS_XATTR
838 set_opt (sbi->s_mount_opt, XATTR_USER);
840 case Opt_nouser_xattr:
841 clear_opt (sbi->s_mount_opt, XATTR_USER);
845 case Opt_nouser_xattr:
846 printk("EXT3 (no)user_xattr options not supported\n");
849 #ifdef CONFIG_EXT3_FS_POSIX_ACL
851 set_opt(sbi->s_mount_opt, POSIX_ACL);
854 clear_opt(sbi->s_mount_opt, POSIX_ACL);
859 printk("EXT3 (no)acl options not supported\n");
862 case Opt_reservation:
863 set_opt(sbi->s_mount_opt, RESERVATION);
865 case Opt_noreservation:
866 clear_opt(sbi->s_mount_opt, RESERVATION);
868 case Opt_journal_update:
870 /* Eventually we will want to be able to create
871 a journal file here. For now, only allow the
872 user to specify an existing inode to be the
875 printk(KERN_ERR "EXT3-fs: cannot specify "
876 "journal on remount\n");
879 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
881 case Opt_journal_inum:
883 printk(KERN_ERR "EXT3-fs: cannot specify "
884 "journal on remount\n");
887 if (match_int(&args[0], &option))
891 case Opt_journal_dev:
893 printk(KERN_ERR "EXT3-fs: cannot specify "
894 "journal on remount\n");
897 if (match_int(&args[0], &option))
899 *journal_devnum = option;
902 set_opt (sbi->s_mount_opt, NOLOAD);
905 if (match_int(&args[0], &option))
910 option = JBD_DEFAULT_MAX_COMMIT_AGE;
911 sbi->s_commit_interval = HZ * option;
913 case Opt_data_journal:
914 data_opt = EXT3_MOUNT_JOURNAL_DATA;
916 case Opt_data_ordered:
917 data_opt = EXT3_MOUNT_ORDERED_DATA;
919 case Opt_data_writeback:
920 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
923 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
926 "EXT3-fs: cannot change data "
927 "mode on remount\n");
931 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
932 sbi->s_mount_opt |= data_opt;
942 if (sb_any_quota_enabled(sb)) {
944 "EXT3-fs: Cannot change journalled "
945 "quota options when quota turned on.\n");
948 qname = match_strdup(&args[0]);
951 "EXT3-fs: not enough memory for "
952 "storing quotafile name.\n");
955 if (sbi->s_qf_names[qtype] &&
956 strcmp(sbi->s_qf_names[qtype], qname)) {
958 "EXT3-fs: %s quota file already "
959 "specified.\n", QTYPE2NAME(qtype));
963 sbi->s_qf_names[qtype] = qname;
964 if (strchr(sbi->s_qf_names[qtype], '/')) {
966 "EXT3-fs: quotafile must be on "
967 "filesystem root.\n");
968 kfree(sbi->s_qf_names[qtype]);
969 sbi->s_qf_names[qtype] = NULL;
972 set_opt(sbi->s_mount_opt, QUOTA);
974 case Opt_offusrjquota:
977 case Opt_offgrpjquota:
980 if (sb_any_quota_enabled(sb)) {
981 printk(KERN_ERR "EXT3-fs: Cannot change "
982 "journalled quota options when "
983 "quota turned on.\n");
987 * The space will be released later when all options
988 * are confirmed to be correct
990 sbi->s_qf_names[qtype] = NULL;
992 case Opt_jqfmt_vfsold:
993 sbi->s_jquota_fmt = QFMT_VFS_OLD;
995 case Opt_jqfmt_vfsv0:
996 sbi->s_jquota_fmt = QFMT_VFS_V0;
1000 set_opt(sbi->s_mount_opt, QUOTA);
1001 set_opt(sbi->s_mount_opt, USRQUOTA);
1004 set_opt(sbi->s_mount_opt, QUOTA);
1005 set_opt(sbi->s_mount_opt, GRPQUOTA);
1008 if (sb_any_quota_enabled(sb)) {
1009 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1010 "options when quota turned on.\n");
1013 clear_opt(sbi->s_mount_opt, QUOTA);
1014 clear_opt(sbi->s_mount_opt, USRQUOTA);
1015 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1023 case Opt_offusrjquota:
1024 case Opt_offgrpjquota:
1025 case Opt_jqfmt_vfsold:
1026 case Opt_jqfmt_vfsv0:
1028 "EXT3-fs: journalled quota options not "
1035 set_opt(sbi->s_mount_opt, ABORT);
1038 if (match_int(&args[0], &option))
1041 set_opt(sbi->s_mount_opt, BARRIER);
1043 clear_opt(sbi->s_mount_opt, BARRIER);
1049 printk("EXT3-fs: resize option only available "
1053 if (match_int(&args[0], &option) != 0)
1055 *n_blocks_count = option;
1058 set_opt(sbi->s_mount_opt, NOBH);
1061 clear_opt(sbi->s_mount_opt, NOBH);
1065 "EXT3-fs: Unrecognized mount option \"%s\" "
1066 "or missing value\n", p);
1071 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1072 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1073 sbi->s_qf_names[USRQUOTA])
1074 clear_opt(sbi->s_mount_opt, USRQUOTA);
1076 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1077 sbi->s_qf_names[GRPQUOTA])
1078 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1080 if ((sbi->s_qf_names[USRQUOTA] &&
1081 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1082 (sbi->s_qf_names[GRPQUOTA] &&
1083 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1084 printk(KERN_ERR "EXT3-fs: old and new quota "
1085 "format mixing.\n");
1089 if (!sbi->s_jquota_fmt) {
1090 printk(KERN_ERR "EXT3-fs: journalled quota format "
1091 "not specified.\n");
1095 if (sbi->s_jquota_fmt) {
1096 printk(KERN_ERR "EXT3-fs: journalled quota format "
1097 "specified with no journalling "
1106 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1109 struct ext3_sb_info *sbi = EXT3_SB(sb);
1112 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1113 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1114 "forcing read-only mode\n");
1119 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1120 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1121 "running e2fsck is recommended\n");
1122 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1123 printk (KERN_WARNING
1124 "EXT3-fs warning: mounting fs with errors, "
1125 "running e2fsck is recommended\n");
1126 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1127 le16_to_cpu(es->s_mnt_count) >=
1128 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1129 printk (KERN_WARNING
1130 "EXT3-fs warning: maximal mount count reached, "
1131 "running e2fsck is recommended\n");
1132 else if (le32_to_cpu(es->s_checkinterval) &&
1133 (le32_to_cpu(es->s_lastcheck) +
1134 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1135 printk (KERN_WARNING
1136 "EXT3-fs warning: checktime reached, "
1137 "running e2fsck is recommended\n");
1139 /* @@@ We _will_ want to clear the valid bit if we find
1140 inconsistencies, to force a fsck at reboot. But for
1141 a plain journaled filesystem we can keep it set as
1142 valid forever! :) */
1143 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1145 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1146 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1147 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1148 es->s_mtime = cpu_to_le32(get_seconds());
1149 ext3_update_dynamic_rev(sb);
1150 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1152 ext3_commit_super(sb, es, 1);
1153 if (test_opt(sb, DEBUG))
1154 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1155 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1157 sbi->s_groups_count,
1158 EXT3_BLOCKS_PER_GROUP(sb),
1159 EXT3_INODES_PER_GROUP(sb),
1162 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1163 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1164 char b[BDEVNAME_SIZE];
1166 printk("external journal on %s\n",
1167 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1169 printk("internal journal\n");
1174 /* Called at mount-time, super-block is locked */
1175 static int ext3_check_descriptors (struct super_block * sb)
1177 struct ext3_sb_info *sbi = EXT3_SB(sb);
1178 ext3_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1179 ext3_fsblk_t last_block;
1180 struct ext3_group_desc * gdp = NULL;
1184 ext3_debug ("Checking group descriptors");
1186 for (i = 0; i < sbi->s_groups_count; i++)
1188 if (i == sbi->s_groups_count - 1)
1189 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1191 last_block = first_block +
1192 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1194 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1195 gdp = (struct ext3_group_desc *)
1196 sbi->s_group_desc[desc_block++]->b_data;
1197 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1198 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1200 ext3_error (sb, "ext3_check_descriptors",
1201 "Block bitmap for group %d"
1202 " not in group (block %lu)!",
1204 le32_to_cpu(gdp->bg_block_bitmap));
1207 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1208 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1210 ext3_error (sb, "ext3_check_descriptors",
1211 "Inode bitmap for group %d"
1212 " not in group (block %lu)!",
1214 le32_to_cpu(gdp->bg_inode_bitmap));
1217 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1218 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >
1221 ext3_error (sb, "ext3_check_descriptors",
1222 "Inode table for group %d"
1223 " not in group (block %lu)!",
1225 le32_to_cpu(gdp->bg_inode_table));
1228 first_block += EXT3_BLOCKS_PER_GROUP(sb);
1232 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1233 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1238 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1239 * the superblock) which were deleted from all directories, but held open by
1240 * a process at the time of a crash. We walk the list and try to delete these
1241 * inodes at recovery time (only with a read-write filesystem).
1243 * In order to keep the orphan inode chain consistent during traversal (in
1244 * case of crash during recovery), we link each inode into the superblock
1245 * orphan list_head and handle it the same way as an inode deletion during
1246 * normal operation (which journals the operations for us).
1248 * We only do an iget() and an iput() on each inode, which is very safe if we
1249 * accidentally point at an in-use or already deleted inode. The worst that
1250 * can happen in this case is that we get a "bit already cleared" message from
1251 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1252 * e2fsck was run on this filesystem, and it must have already done the orphan
1253 * inode cleanup for us, so we can safely abort without any further action.
1255 static void ext3_orphan_cleanup (struct super_block * sb,
1256 struct ext3_super_block * es)
1258 unsigned int s_flags = sb->s_flags;
1259 int nr_orphans = 0, nr_truncates = 0;
1263 if (!es->s_last_orphan) {
1264 jbd_debug(4, "no orphan inodes to clean up\n");
1268 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1269 if (es->s_last_orphan)
1270 jbd_debug(1, "Errors on filesystem, "
1271 "clearing orphan list.\n");
1272 es->s_last_orphan = 0;
1273 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1277 if (s_flags & MS_RDONLY) {
1278 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1280 sb->s_flags &= ~MS_RDONLY;
1283 /* Needed for iput() to work correctly and not trash data */
1284 sb->s_flags |= MS_ACTIVE;
1285 /* Turn on quotas so that they are updated correctly */
1286 for (i = 0; i < MAXQUOTAS; i++) {
1287 if (EXT3_SB(sb)->s_qf_names[i]) {
1288 int ret = ext3_quota_on_mount(sb, i);
1291 "EXT3-fs: Cannot turn on journalled "
1292 "quota: error %d\n", ret);
1297 while (es->s_last_orphan) {
1298 struct inode *inode;
1301 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1302 es->s_last_orphan = 0;
1306 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1308 if (inode->i_nlink) {
1310 "%s: truncating inode %lu to %Ld bytes\n",
1311 __FUNCTION__, inode->i_ino, inode->i_size);
1312 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1313 inode->i_ino, inode->i_size);
1314 ext3_truncate(inode);
1318 "%s: deleting unreferenced inode %lu\n",
1319 __FUNCTION__, inode->i_ino);
1320 jbd_debug(2, "deleting unreferenced inode %lu\n",
1324 iput(inode); /* The delete magic happens here! */
1327 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1330 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1331 sb->s_id, PLURAL(nr_orphans));
1333 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1334 sb->s_id, PLURAL(nr_truncates));
1336 /* Turn quotas off */
1337 for (i = 0; i < MAXQUOTAS; i++) {
1338 if (sb_dqopt(sb)->files[i])
1339 vfs_quota_off(sb, i);
1342 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1345 #define log2(n) ffz(~(n))
1348 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1349 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1350 * We need to be 1 filesystem block less than the 2^32 sector limit.
1352 static loff_t ext3_max_size(int bits)
1354 loff_t res = EXT3_NDIR_BLOCKS;
1355 /* This constant is calculated to be the largest file size for a
1356 * dense, 4k-blocksize file such that the total number of
1357 * sectors in the file, including data and all indirect blocks,
1358 * does not exceed 2^32. */
1359 const loff_t upper_limit = 0x1ff7fffd000LL;
1361 res += 1LL << (bits-2);
1362 res += 1LL << (2*(bits-2));
1363 res += 1LL << (3*(bits-2));
1365 if (res > upper_limit)
1370 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1371 ext3_fsblk_t logic_sb_block,
1374 struct ext3_sb_info *sbi = EXT3_SB(sb);
1375 unsigned long bg, first_meta_bg;
1378 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1380 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1382 return (logic_sb_block + nr + 1);
1383 bg = sbi->s_desc_per_block * nr;
1384 if (ext3_bg_has_super(sb, bg))
1386 return (has_super + ext3_group_first_block_no(sb, bg));
1390 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1392 struct buffer_head * bh;
1393 struct ext3_super_block *es = NULL;
1394 struct ext3_sb_info *sbi;
1396 ext3_fsblk_t sb_block = get_sb_block(&data);
1397 ext3_fsblk_t logic_sb_block;
1398 unsigned long offset = 0;
1399 unsigned int journal_inum = 0;
1400 unsigned long journal_devnum = 0;
1401 unsigned long def_mount_opts;
1410 sbi = kmalloc(sizeof(*sbi), GFP_KERNEL);
1413 sb->s_fs_info = sbi;
1414 memset(sbi, 0, sizeof(*sbi));
1415 sbi->s_mount_opt = 0;
1416 sbi->s_resuid = EXT3_DEF_RESUID;
1417 sbi->s_resgid = EXT3_DEF_RESGID;
1421 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1423 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1428 * The ext3 superblock will not be buffer aligned for other than 1kB
1429 * block sizes. We need to calculate the offset from buffer start.
1431 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1432 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1433 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1435 logic_sb_block = sb_block;
1438 if (!(bh = sb_bread(sb, logic_sb_block))) {
1439 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1443 * Note: s_es must be initialized as soon as possible because
1444 * some ext3 macro-instructions depend on its value
1446 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1448 sb->s_magic = le16_to_cpu(es->s_magic);
1449 if (sb->s_magic != EXT3_SUPER_MAGIC)
1452 /* Set defaults before we parse the mount options */
1453 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1454 if (def_mount_opts & EXT3_DEFM_DEBUG)
1455 set_opt(sbi->s_mount_opt, DEBUG);
1456 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1457 set_opt(sbi->s_mount_opt, GRPID);
1458 if (def_mount_opts & EXT3_DEFM_UID16)
1459 set_opt(sbi->s_mount_opt, NO_UID32);
1460 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1461 set_opt(sbi->s_mount_opt, XATTR_USER);
1462 if (def_mount_opts & EXT3_DEFM_ACL)
1463 set_opt(sbi->s_mount_opt, POSIX_ACL);
1464 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1465 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1466 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1467 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1468 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1469 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1471 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1472 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1473 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1474 set_opt(sbi->s_mount_opt, ERRORS_RO);
1476 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1477 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1479 set_opt(sbi->s_mount_opt, RESERVATION);
1481 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1485 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1486 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1488 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1489 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1490 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1491 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1493 "EXT3-fs warning: feature flags set on rev 0 fs, "
1494 "running e2fsck is recommended\n");
1496 * Check feature flags regardless of the revision level, since we
1497 * previously didn't change the revision level when setting the flags,
1498 * so there is a chance incompat flags are set on a rev 0 filesystem.
1500 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1502 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1503 "unsupported optional features (%x).\n",
1504 sb->s_id, le32_to_cpu(features));
1507 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1508 if (!(sb->s_flags & MS_RDONLY) && features) {
1509 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1510 "unsupported optional features (%x).\n",
1511 sb->s_id, le32_to_cpu(features));
1514 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1516 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1517 blocksize > EXT3_MAX_BLOCK_SIZE) {
1519 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1520 blocksize, sb->s_id);
1524 hblock = bdev_hardsect_size(sb->s_bdev);
1525 if (sb->s_blocksize != blocksize) {
1527 * Make sure the blocksize for the filesystem is larger
1528 * than the hardware sectorsize for the machine.
1530 if (blocksize < hblock) {
1531 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1532 "device blocksize %d.\n", blocksize, hblock);
1537 sb_set_blocksize(sb, blocksize);
1538 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1539 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1540 bh = sb_bread(sb, logic_sb_block);
1543 "EXT3-fs: Can't read superblock on 2nd try.\n");
1546 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1548 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1550 "EXT3-fs: Magic mismatch, very weird !\n");
1555 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1557 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1558 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1559 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1561 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1562 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1563 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1564 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1565 (sbi->s_inode_size > blocksize)) {
1567 "EXT3-fs: unsupported inode size: %d\n",
1572 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1573 le32_to_cpu(es->s_log_frag_size);
1574 if (blocksize != sbi->s_frag_size) {
1576 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1577 sbi->s_frag_size, blocksize);
1580 sbi->s_frags_per_block = 1;
1581 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1582 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1583 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1584 if (EXT3_INODE_SIZE(sb) == 0)
1586 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1587 if (sbi->s_inodes_per_block == 0)
1589 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1590 sbi->s_inodes_per_block;
1591 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1593 sbi->s_mount_state = le16_to_cpu(es->s_state);
1594 sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb));
1595 sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb));
1596 for (i=0; i < 4; i++)
1597 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1598 sbi->s_def_hash_version = es->s_def_hash_version;
1600 if (sbi->s_blocks_per_group > blocksize * 8) {
1602 "EXT3-fs: #blocks per group too big: %lu\n",
1603 sbi->s_blocks_per_group);
1606 if (sbi->s_frags_per_group > blocksize * 8) {
1608 "EXT3-fs: #fragments per group too big: %lu\n",
1609 sbi->s_frags_per_group);
1612 if (sbi->s_inodes_per_group > blocksize * 8) {
1614 "EXT3-fs: #inodes per group too big: %lu\n",
1615 sbi->s_inodes_per_group);
1619 if (le32_to_cpu(es->s_blocks_count) >
1620 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1621 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1622 " too large to mount safely\n", sb->s_id);
1623 if (sizeof(sector_t) < 8)
1624 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1629 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1631 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1632 le32_to_cpu(es->s_first_data_block) - 1)
1633 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1634 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1635 EXT3_DESC_PER_BLOCK(sb);
1636 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1638 if (sbi->s_group_desc == NULL) {
1639 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1643 bgl_lock_init(&sbi->s_blockgroup_lock);
1645 for (i = 0; i < db_count; i++) {
1646 block = descriptor_loc(sb, logic_sb_block, i);
1647 sbi->s_group_desc[i] = sb_bread(sb, block);
1648 if (!sbi->s_group_desc[i]) {
1649 printk (KERN_ERR "EXT3-fs: "
1650 "can't read group descriptor %d\n", i);
1655 if (!ext3_check_descriptors (sb)) {
1656 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1659 sbi->s_gdb_count = db_count;
1660 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1661 spin_lock_init(&sbi->s_next_gen_lock);
1663 percpu_counter_init(&sbi->s_freeblocks_counter,
1664 ext3_count_free_blocks(sb));
1665 percpu_counter_init(&sbi->s_freeinodes_counter,
1666 ext3_count_free_inodes(sb));
1667 percpu_counter_init(&sbi->s_dirs_counter,
1668 ext3_count_dirs(sb));
1670 /* per fileystem reservation list head & lock */
1671 spin_lock_init(&sbi->s_rsv_window_lock);
1672 sbi->s_rsv_window_root = RB_ROOT;
1673 /* Add a single, static dummy reservation to the start of the
1674 * reservation window list --- it gives us a placeholder for
1675 * append-at-start-of-list which makes the allocation logic
1676 * _much_ simpler. */
1677 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1678 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1679 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1680 sbi->s_rsv_window_head.rsv_goal_size = 0;
1681 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1684 * set up enough so that it can read an inode
1686 sb->s_op = &ext3_sops;
1687 sb->s_export_op = &ext3_export_ops;
1688 sb->s_xattr = ext3_xattr_handlers;
1690 sb->s_qcop = &ext3_qctl_operations;
1691 sb->dq_op = &ext3_quota_operations;
1693 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1697 needs_recovery = (es->s_last_orphan != 0 ||
1698 EXT3_HAS_INCOMPAT_FEATURE(sb,
1699 EXT3_FEATURE_INCOMPAT_RECOVER));
1702 * The first inode we look at is the journal inode. Don't try
1703 * root first: it may be modified in the journal!
1705 if (!test_opt(sb, NOLOAD) &&
1706 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1707 if (ext3_load_journal(sb, es, journal_devnum))
1709 } else if (journal_inum) {
1710 if (ext3_create_journal(sb, es, journal_inum))
1715 "ext3: No journal on filesystem on %s\n",
1720 /* We have now updated the journal if required, so we can
1721 * validate the data journaling mode. */
1722 switch (test_opt(sb, DATA_FLAGS)) {
1724 /* No mode set, assume a default based on the journal
1725 capabilities: ORDERED_DATA if the journal can
1726 cope, else JOURNAL_DATA */
1727 if (journal_check_available_features
1728 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1729 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1731 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1734 case EXT3_MOUNT_ORDERED_DATA:
1735 case EXT3_MOUNT_WRITEBACK_DATA:
1736 if (!journal_check_available_features
1737 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1738 printk(KERN_ERR "EXT3-fs: Journal does not support "
1739 "requested data journaling mode\n");
1746 if (test_opt(sb, NOBH)) {
1747 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1748 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1749 "its supported only with writeback mode\n");
1750 clear_opt(sbi->s_mount_opt, NOBH);
1754 * The journal_load will have done any necessary log recovery,
1755 * so we can safely mount the rest of the filesystem now.
1758 root = iget(sb, EXT3_ROOT_INO);
1759 sb->s_root = d_alloc_root(root);
1761 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1765 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1768 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1772 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1774 * akpm: core read_super() calls in here with the superblock locked.
1775 * That deadlocks, because orphan cleanup needs to lock the superblock
1776 * in numerous places. Here we just pop the lock - it's relatively
1777 * harmless, because we are now ready to accept write_super() requests,
1778 * and aviro says that's the only reason for hanging onto the
1781 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1782 ext3_orphan_cleanup(sb, es);
1783 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1785 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1786 ext3_mark_recovery_complete(sb, es);
1787 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1788 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1789 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1797 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1802 journal_destroy(sbi->s_journal);
1804 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1805 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1806 percpu_counter_destroy(&sbi->s_dirs_counter);
1808 for (i = 0; i < db_count; i++)
1809 brelse(sbi->s_group_desc[i]);
1810 kfree(sbi->s_group_desc);
1813 for (i = 0; i < MAXQUOTAS; i++)
1814 kfree(sbi->s_qf_names[i]);
1816 ext3_blkdev_remove(sbi);
1819 sb->s_fs_info = NULL;
1826 * Setup any per-fs journal parameters now. We'll do this both on
1827 * initial mount, once the journal has been initialised but before we've
1828 * done any recovery; and again on any subsequent remount.
1830 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1832 struct ext3_sb_info *sbi = EXT3_SB(sb);
1834 if (sbi->s_commit_interval)
1835 journal->j_commit_interval = sbi->s_commit_interval;
1836 /* We could also set up an ext3-specific default for the commit
1837 * interval here, but for now we'll just fall back to the jbd
1840 spin_lock(&journal->j_state_lock);
1841 if (test_opt(sb, BARRIER))
1842 journal->j_flags |= JFS_BARRIER;
1844 journal->j_flags &= ~JFS_BARRIER;
1845 spin_unlock(&journal->j_state_lock);
1848 static journal_t *ext3_get_journal(struct super_block *sb,
1849 unsigned int journal_inum)
1851 struct inode *journal_inode;
1854 /* First, test for the existence of a valid inode on disk. Bad
1855 * things happen if we iget() an unused inode, as the subsequent
1856 * iput() will try to delete it. */
1858 journal_inode = iget(sb, journal_inum);
1859 if (!journal_inode) {
1860 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1863 if (!journal_inode->i_nlink) {
1864 make_bad_inode(journal_inode);
1865 iput(journal_inode);
1866 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1870 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1871 journal_inode, journal_inode->i_size);
1872 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1873 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1874 iput(journal_inode);
1878 journal = journal_init_inode(journal_inode);
1880 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1881 iput(journal_inode);
1884 journal->j_private = sb;
1885 ext3_init_journal_params(sb, journal);
1889 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1892 struct buffer_head * bh;
1896 int hblock, blocksize;
1897 ext3_fsblk_t sb_block;
1898 unsigned long offset;
1899 struct ext3_super_block * es;
1900 struct block_device *bdev;
1902 bdev = ext3_blkdev_get(j_dev);
1906 if (bd_claim(bdev, sb)) {
1908 "EXT3: failed to claim external journal device.\n");
1913 blocksize = sb->s_blocksize;
1914 hblock = bdev_hardsect_size(bdev);
1915 if (blocksize < hblock) {
1917 "EXT3-fs: blocksize too small for journal device.\n");
1921 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1922 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1923 set_blocksize(bdev, blocksize);
1924 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1925 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1926 "external journal\n");
1930 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1931 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1932 !(le32_to_cpu(es->s_feature_incompat) &
1933 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1934 printk(KERN_ERR "EXT3-fs: external journal has "
1935 "bad superblock\n");
1940 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1941 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1946 len = le32_to_cpu(es->s_blocks_count);
1947 start = sb_block + 1;
1948 brelse(bh); /* we're done with the superblock */
1950 journal = journal_init_dev(bdev, sb->s_bdev,
1951 start, len, blocksize);
1953 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1956 journal->j_private = sb;
1957 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1958 wait_on_buffer(journal->j_sb_buffer);
1959 if (!buffer_uptodate(journal->j_sb_buffer)) {
1960 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1963 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1964 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1965 "user (unsupported) - %d\n",
1966 be32_to_cpu(journal->j_superblock->s_nr_users));
1969 EXT3_SB(sb)->journal_bdev = bdev;
1970 ext3_init_journal_params(sb, journal);
1973 journal_destroy(journal);
1975 ext3_blkdev_put(bdev);
1979 static int ext3_load_journal(struct super_block *sb,
1980 struct ext3_super_block *es,
1981 unsigned long journal_devnum)
1984 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
1987 int really_read_only;
1989 if (journal_devnum &&
1990 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
1991 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
1992 "numbers have changed\n");
1993 journal_dev = new_decode_dev(journal_devnum);
1995 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
1997 really_read_only = bdev_read_only(sb->s_bdev);
2000 * Are we loading a blank journal or performing recovery after a
2001 * crash? For recovery, we need to check in advance whether we
2002 * can get read-write access to the device.
2005 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2006 if (sb->s_flags & MS_RDONLY) {
2007 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2008 "required on readonly filesystem.\n");
2009 if (really_read_only) {
2010 printk(KERN_ERR "EXT3-fs: write access "
2011 "unavailable, cannot proceed.\n");
2014 printk (KERN_INFO "EXT3-fs: write access will "
2015 "be enabled during recovery.\n");
2019 if (journal_inum && journal_dev) {
2020 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2021 "and inode journals!\n");
2026 if (!(journal = ext3_get_journal(sb, journal_inum)))
2029 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2033 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2034 err = journal_update_format(journal);
2036 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2037 journal_destroy(journal);
2042 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2043 err = journal_wipe(journal, !really_read_only);
2045 err = journal_load(journal);
2048 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2049 journal_destroy(journal);
2053 EXT3_SB(sb)->s_journal = journal;
2054 ext3_clear_journal_err(sb, es);
2056 if (journal_devnum &&
2057 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2058 es->s_journal_dev = cpu_to_le32(journal_devnum);
2061 /* Make sure we flush the recovery flag to disk. */
2062 ext3_commit_super(sb, es, 1);
2068 static int ext3_create_journal(struct super_block * sb,
2069 struct ext3_super_block * es,
2070 unsigned int journal_inum)
2074 if (sb->s_flags & MS_RDONLY) {
2075 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2076 "create journal.\n");
2080 if (!(journal = ext3_get_journal(sb, journal_inum)))
2083 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2086 if (journal_create(journal)) {
2087 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2088 journal_destroy(journal);
2092 EXT3_SB(sb)->s_journal = journal;
2094 ext3_update_dynamic_rev(sb);
2095 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2096 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2098 es->s_journal_inum = cpu_to_le32(journal_inum);
2101 /* Make sure we flush the recovery flag to disk. */
2102 ext3_commit_super(sb, es, 1);
2107 static void ext3_commit_super (struct super_block * sb,
2108 struct ext3_super_block * es,
2111 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2115 es->s_wtime = cpu_to_le32(get_seconds());
2116 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2117 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2118 BUFFER_TRACE(sbh, "marking dirty");
2119 mark_buffer_dirty(sbh);
2121 sync_dirty_buffer(sbh);
2126 * Have we just finished recovery? If so, and if we are mounting (or
2127 * remounting) the filesystem readonly, then we will end up with a
2128 * consistent fs on disk. Record that fact.
2130 static void ext3_mark_recovery_complete(struct super_block * sb,
2131 struct ext3_super_block * es)
2133 journal_t *journal = EXT3_SB(sb)->s_journal;
2135 journal_lock_updates(journal);
2136 journal_flush(journal);
2137 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2138 sb->s_flags & MS_RDONLY) {
2139 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2141 ext3_commit_super(sb, es, 1);
2143 journal_unlock_updates(journal);
2147 * If we are mounting (or read-write remounting) a filesystem whose journal
2148 * has recorded an error from a previous lifetime, move that error to the
2149 * main filesystem now.
2151 static void ext3_clear_journal_err(struct super_block * sb,
2152 struct ext3_super_block * es)
2158 journal = EXT3_SB(sb)->s_journal;
2161 * Now check for any error status which may have been recorded in the
2162 * journal by a prior ext3_error() or ext3_abort()
2165 j_errno = journal_errno(journal);
2169 errstr = ext3_decode_error(sb, j_errno, nbuf);
2170 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2171 "from previous mount: %s", errstr);
2172 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2173 "filesystem check.");
2175 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2176 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2177 ext3_commit_super (sb, es, 1);
2179 journal_clear_err(journal);
2184 * Force the running and committing transactions to commit,
2185 * and wait on the commit.
2187 int ext3_force_commit(struct super_block *sb)
2192 if (sb->s_flags & MS_RDONLY)
2195 journal = EXT3_SB(sb)->s_journal;
2197 ret = ext3_journal_force_commit(journal);
2202 * Ext3 always journals updates to the superblock itself, so we don't
2203 * have to propagate any other updates to the superblock on disk at this
2204 * point. Just start an async writeback to get the buffers on their way
2207 * This implicitly triggers the writebehind on sync().
2210 static void ext3_write_super (struct super_block * sb)
2212 if (mutex_trylock(&sb->s_lock) != 0)
2217 static int ext3_sync_fs(struct super_block *sb, int wait)
2222 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2224 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2230 * LVM calls this function before a (read-only) snapshot is created. This
2231 * gives us a chance to flush the journal completely and mark the fs clean.
2233 static void ext3_write_super_lockfs(struct super_block *sb)
2237 if (!(sb->s_flags & MS_RDONLY)) {
2238 journal_t *journal = EXT3_SB(sb)->s_journal;
2240 /* Now we set up the journal barrier. */
2241 journal_lock_updates(journal);
2242 journal_flush(journal);
2244 /* Journal blocked and flushed, clear needs_recovery flag. */
2245 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2246 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2251 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2252 * flag here, even though the filesystem is not technically dirty yet.
2254 static void ext3_unlockfs(struct super_block *sb)
2256 if (!(sb->s_flags & MS_RDONLY)) {
2258 /* Reser the needs_recovery flag before the fs is unlocked. */
2259 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2260 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2262 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2266 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2268 struct ext3_super_block * es;
2269 struct ext3_sb_info *sbi = EXT3_SB(sb);
2270 ext3_fsblk_t n_blocks_count = 0;
2271 unsigned long old_sb_flags;
2272 struct ext3_mount_options old_opts;
2278 /* Store the original options */
2279 old_sb_flags = sb->s_flags;
2280 old_opts.s_mount_opt = sbi->s_mount_opt;
2281 old_opts.s_resuid = sbi->s_resuid;
2282 old_opts.s_resgid = sbi->s_resgid;
2283 old_opts.s_commit_interval = sbi->s_commit_interval;
2285 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2286 for (i = 0; i < MAXQUOTAS; i++)
2287 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2291 * Allow the "check" option to be passed as a remount option.
2293 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2298 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2299 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2301 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2302 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2306 ext3_init_journal_params(sb, sbi->s_journal);
2308 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2309 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2310 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2315 if (*flags & MS_RDONLY) {
2317 * First of all, the unconditional stuff we have to do
2318 * to disable replay of the journal when we next remount
2320 sb->s_flags |= MS_RDONLY;
2323 * OK, test if we are remounting a valid rw partition
2324 * readonly, and if so set the rdonly flag and then
2325 * mark the partition as valid again.
2327 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2328 (sbi->s_mount_state & EXT3_VALID_FS))
2329 es->s_state = cpu_to_le16(sbi->s_mount_state);
2331 ext3_mark_recovery_complete(sb, es);
2334 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2335 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2336 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2337 "remount RDWR because of unsupported "
2338 "optional features (%x).\n",
2339 sb->s_id, le32_to_cpu(ret));
2344 * Mounting a RDONLY partition read-write, so reread
2345 * and store the current valid flag. (It may have
2346 * been changed by e2fsck since we originally mounted
2349 ext3_clear_journal_err(sb, es);
2350 sbi->s_mount_state = le16_to_cpu(es->s_state);
2351 if ((ret = ext3_group_extend(sb, es, n_blocks_count))) {
2355 if (!ext3_setup_super (sb, es, 0))
2356 sb->s_flags &= ~MS_RDONLY;
2360 /* Release old quota file names */
2361 for (i = 0; i < MAXQUOTAS; i++)
2362 if (old_opts.s_qf_names[i] &&
2363 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2364 kfree(old_opts.s_qf_names[i]);
2368 sb->s_flags = old_sb_flags;
2369 sbi->s_mount_opt = old_opts.s_mount_opt;
2370 sbi->s_resuid = old_opts.s_resuid;
2371 sbi->s_resgid = old_opts.s_resgid;
2372 sbi->s_commit_interval = old_opts.s_commit_interval;
2374 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2375 for (i = 0; i < MAXQUOTAS; i++) {
2376 if (sbi->s_qf_names[i] &&
2377 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2378 kfree(sbi->s_qf_names[i]);
2379 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2385 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2387 struct super_block *sb = dentry->d_sb;
2388 struct ext3_sb_info *sbi = EXT3_SB(sb);
2389 struct ext3_super_block *es = sbi->s_es;
2390 ext3_fsblk_t overhead;
2393 if (test_opt (sb, MINIX_DF))
2396 unsigned long ngroups;
2397 ngroups = EXT3_SB(sb)->s_groups_count;
2401 * Compute the overhead (FS structures)
2405 * All of the blocks before first_data_block are
2408 overhead = le32_to_cpu(es->s_first_data_block);
2411 * Add the overhead attributed to the superblock and
2412 * block group descriptors. If the sparse superblocks
2413 * feature is turned on, then not all groups have this.
2415 for (i = 0; i < ngroups; i++) {
2416 overhead += ext3_bg_has_super(sb, i) +
2417 ext3_bg_num_gdb(sb, i);
2422 * Every block group has an inode bitmap, a block
2423 * bitmap, and an inode table.
2425 overhead += (ngroups * (2 + EXT3_SB(sb)->s_itb_per_group));
2428 buf->f_type = EXT3_SUPER_MAGIC;
2429 buf->f_bsize = sb->s_blocksize;
2430 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
2431 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
2432 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2433 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2435 buf->f_files = le32_to_cpu(es->s_inodes_count);
2436 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
2437 buf->f_namelen = EXT3_NAME_LEN;
2441 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2442 * is locked for write. Otherwise the are possible deadlocks:
2443 * Process 1 Process 2
2444 * ext3_create() quota_sync()
2445 * journal_start() write_dquot()
2446 * DQUOT_INIT() down(dqio_mutex)
2447 * down(dqio_mutex) journal_start()
2453 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2455 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2458 static int ext3_dquot_initialize(struct inode *inode, int type)
2463 /* We may create quota structure so we need to reserve enough blocks */
2464 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2466 return PTR_ERR(handle);
2467 ret = dquot_initialize(inode, type);
2468 err = ext3_journal_stop(handle);
2474 static int ext3_dquot_drop(struct inode *inode)
2479 /* We may delete quota structure so we need to reserve enough blocks */
2480 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2482 return PTR_ERR(handle);
2483 ret = dquot_drop(inode);
2484 err = ext3_journal_stop(handle);
2490 static int ext3_write_dquot(struct dquot *dquot)
2494 struct inode *inode;
2496 inode = dquot_to_inode(dquot);
2497 handle = ext3_journal_start(inode,
2498 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2500 return PTR_ERR(handle);
2501 ret = dquot_commit(dquot);
2502 err = ext3_journal_stop(handle);
2508 static int ext3_acquire_dquot(struct dquot *dquot)
2513 handle = ext3_journal_start(dquot_to_inode(dquot),
2514 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2516 return PTR_ERR(handle);
2517 ret = dquot_acquire(dquot);
2518 err = ext3_journal_stop(handle);
2524 static int ext3_release_dquot(struct dquot *dquot)
2529 handle = ext3_journal_start(dquot_to_inode(dquot),
2530 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2532 return PTR_ERR(handle);
2533 ret = dquot_release(dquot);
2534 err = ext3_journal_stop(handle);
2540 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2542 /* Are we journalling quotas? */
2543 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2544 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2545 dquot_mark_dquot_dirty(dquot);
2546 return ext3_write_dquot(dquot);
2548 return dquot_mark_dquot_dirty(dquot);
2552 static int ext3_write_info(struct super_block *sb, int type)
2557 /* Data block + inode block */
2558 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2560 return PTR_ERR(handle);
2561 ret = dquot_commit_info(sb, type);
2562 err = ext3_journal_stop(handle);
2569 * Turn on quotas during mount time - we need to find
2570 * the quota file and such...
2572 static int ext3_quota_on_mount(struct super_block *sb, int type)
2574 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2575 EXT3_SB(sb)->s_jquota_fmt, type);
2579 * Standard function to be called on quota_on
2581 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2585 struct nameidata nd;
2587 if (!test_opt(sb, QUOTA))
2589 /* Not journalling quota? */
2590 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2591 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2592 return vfs_quota_on(sb, type, format_id, path);
2593 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2596 /* Quotafile not on the same filesystem? */
2597 if (nd.mnt->mnt_sb != sb) {
2601 /* Quotafile not of fs root? */
2602 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2604 "EXT3-fs: Quota file not on filesystem root. "
2605 "Journalled quota will not work.\n");
2607 return vfs_quota_on(sb, type, format_id, path);
2610 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2611 * acquiring the locks... As quota files are never truncated and quota code
2612 * itself serializes the operations (and noone else should touch the files)
2613 * we don't have to be afraid of races */
2614 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2615 size_t len, loff_t off)
2617 struct inode *inode = sb_dqopt(sb)->files[type];
2618 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2620 int offset = off & (sb->s_blocksize - 1);
2623 struct buffer_head *bh;
2624 loff_t i_size = i_size_read(inode);
2628 if (off+len > i_size)
2631 while (toread > 0) {
2632 tocopy = sb->s_blocksize - offset < toread ?
2633 sb->s_blocksize - offset : toread;
2634 bh = ext3_bread(NULL, inode, blk, 0, &err);
2637 if (!bh) /* A hole? */
2638 memset(data, 0, tocopy);
2640 memcpy(data, bh->b_data+offset, tocopy);
2650 /* Write to quotafile (we know the transaction is already started and has
2651 * enough credits) */
2652 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2653 const char *data, size_t len, loff_t off)
2655 struct inode *inode = sb_dqopt(sb)->files[type];
2656 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2658 int offset = off & (sb->s_blocksize - 1);
2660 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2661 size_t towrite = len;
2662 struct buffer_head *bh;
2663 handle_t *handle = journal_current_handle();
2665 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2666 while (towrite > 0) {
2667 tocopy = sb->s_blocksize - offset < towrite ?
2668 sb->s_blocksize - offset : towrite;
2669 bh = ext3_bread(handle, inode, blk, 1, &err);
2672 if (journal_quota) {
2673 err = ext3_journal_get_write_access(handle, bh);
2680 memcpy(bh->b_data+offset, data, tocopy);
2681 flush_dcache_page(bh->b_page);
2684 err = ext3_journal_dirty_metadata(handle, bh);
2686 /* Always do at least ordered writes for quotas */
2687 err = ext3_journal_dirty_data(handle, bh);
2688 mark_buffer_dirty(bh);
2701 if (inode->i_size < off+len-towrite) {
2702 i_size_write(inode, off+len-towrite);
2703 EXT3_I(inode)->i_disksize = inode->i_size;
2706 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2707 ext3_mark_inode_dirty(handle, inode);
2708 mutex_unlock(&inode->i_mutex);
2709 return len - towrite;
2714 static int ext3_get_sb(struct file_system_type *fs_type,
2715 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2717 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2720 static struct file_system_type ext3_fs_type = {
2721 .owner = THIS_MODULE,
2723 .get_sb = ext3_get_sb,
2724 .kill_sb = kill_block_super,
2725 .fs_flags = FS_REQUIRES_DEV,
2728 static int __init init_ext3_fs(void)
2730 int err = init_ext3_xattr();
2733 err = init_inodecache();
2736 err = register_filesystem(&ext3_fs_type);
2741 destroy_inodecache();
2747 static void __exit exit_ext3_fs(void)
2749 unregister_filesystem(&ext3_fs_type);
2750 destroy_inodecache();
2754 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2755 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2756 MODULE_LICENSE("GPL");
2757 module_init(init_ext3_fs)
2758 module_exit(exit_ext3_fs)