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, __func__,
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 if (EXT3_SB(sb)->s_journal)
285 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
288 void ext3_warning (struct super_block * sb, const char * function,
289 const char * fmt, ...)
294 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
301 void ext3_update_dynamic_rev(struct super_block *sb)
303 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
305 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
308 ext3_warning(sb, __func__,
309 "updating to rev %d because of new feature flag, "
310 "running e2fsck is recommended",
313 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
314 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
315 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
316 /* leave es->s_feature_*compat flags alone */
317 /* es->s_uuid will be set by e2fsck if empty */
320 * The rest of the superblock fields should be zero, and if not it
321 * means they are likely already in use, so leave them alone. We
322 * can leave it up to e2fsck to clean up any inconsistencies there.
327 * Open the external journal device
329 static struct block_device *ext3_blkdev_get(dev_t dev)
331 struct block_device *bdev;
332 char b[BDEVNAME_SIZE];
334 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
340 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
341 __bdevname(dev, b), PTR_ERR(bdev));
346 * Release the journal device
348 static int ext3_blkdev_put(struct block_device *bdev)
351 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
354 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
356 struct block_device *bdev;
359 bdev = sbi->journal_bdev;
361 ret = ext3_blkdev_put(bdev);
362 sbi->journal_bdev = NULL;
367 static inline struct inode *orphan_list_entry(struct list_head *l)
369 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
372 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
376 printk(KERN_ERR "sb orphan head is %d\n",
377 le32_to_cpu(sbi->s_es->s_last_orphan));
379 printk(KERN_ERR "sb_info orphan list:\n");
380 list_for_each(l, &sbi->s_orphan) {
381 struct inode *inode = orphan_list_entry(l);
383 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
384 inode->i_sb->s_id, inode->i_ino, inode,
385 inode->i_mode, inode->i_nlink,
390 static void ext3_put_super (struct super_block * sb)
392 struct ext3_sb_info *sbi = EXT3_SB(sb);
393 struct ext3_super_block *es = sbi->s_es;
396 ext3_xattr_put_super(sb);
397 err = journal_destroy(sbi->s_journal);
398 sbi->s_journal = NULL;
400 ext3_abort(sb, __func__, "Couldn't clean up the journal");
402 if (!(sb->s_flags & MS_RDONLY)) {
403 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
404 es->s_state = cpu_to_le16(sbi->s_mount_state);
405 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
406 mark_buffer_dirty(sbi->s_sbh);
407 ext3_commit_super(sb, es, 1);
410 for (i = 0; i < sbi->s_gdb_count; i++)
411 brelse(sbi->s_group_desc[i]);
412 kfree(sbi->s_group_desc);
413 percpu_counter_destroy(&sbi->s_freeblocks_counter);
414 percpu_counter_destroy(&sbi->s_freeinodes_counter);
415 percpu_counter_destroy(&sbi->s_dirs_counter);
418 for (i = 0; i < MAXQUOTAS; i++)
419 kfree(sbi->s_qf_names[i]);
422 /* Debugging code just in case the in-memory inode orphan list
423 * isn't empty. The on-disk one can be non-empty if we've
424 * detected an error and taken the fs readonly, but the
425 * in-memory list had better be clean by this point. */
426 if (!list_empty(&sbi->s_orphan))
427 dump_orphan_list(sb, sbi);
428 J_ASSERT(list_empty(&sbi->s_orphan));
430 invalidate_bdev(sb->s_bdev);
431 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
433 * Invalidate the journal device's buffers. We don't want them
434 * floating about in memory - the physical journal device may
435 * hotswapped, and it breaks the `ro-after' testing code.
437 sync_blockdev(sbi->journal_bdev);
438 invalidate_bdev(sbi->journal_bdev);
439 ext3_blkdev_remove(sbi);
441 sb->s_fs_info = NULL;
446 static struct kmem_cache *ext3_inode_cachep;
449 * Called inside transaction, so use GFP_NOFS
451 static struct inode *ext3_alloc_inode(struct super_block *sb)
453 struct ext3_inode_info *ei;
455 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
458 #ifdef CONFIG_EXT3_FS_POSIX_ACL
459 ei->i_acl = EXT3_ACL_NOT_CACHED;
460 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
462 ei->i_block_alloc_info = NULL;
463 ei->vfs_inode.i_version = 1;
464 return &ei->vfs_inode;
467 static void ext3_destroy_inode(struct inode *inode)
469 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
470 printk("EXT3 Inode %p: orphan list check failed!\n",
472 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
473 EXT3_I(inode), sizeof(struct ext3_inode_info),
477 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
480 static void init_once(void *foo)
482 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
484 INIT_LIST_HEAD(&ei->i_orphan);
485 #ifdef CONFIG_EXT3_FS_XATTR
486 init_rwsem(&ei->xattr_sem);
488 mutex_init(&ei->truncate_mutex);
489 inode_init_once(&ei->vfs_inode);
492 static int init_inodecache(void)
494 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
495 sizeof(struct ext3_inode_info),
496 0, (SLAB_RECLAIM_ACCOUNT|
499 if (ext3_inode_cachep == NULL)
504 static void destroy_inodecache(void)
506 kmem_cache_destroy(ext3_inode_cachep);
509 static void ext3_clear_inode(struct inode *inode)
511 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
512 #ifdef CONFIG_EXT3_FS_POSIX_ACL
513 if (EXT3_I(inode)->i_acl &&
514 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
515 posix_acl_release(EXT3_I(inode)->i_acl);
516 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
518 if (EXT3_I(inode)->i_default_acl &&
519 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
520 posix_acl_release(EXT3_I(inode)->i_default_acl);
521 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
524 ext3_discard_reservation(inode);
525 EXT3_I(inode)->i_block_alloc_info = NULL;
530 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
532 #if defined(CONFIG_QUOTA)
533 struct ext3_sb_info *sbi = EXT3_SB(sb);
535 if (sbi->s_jquota_fmt)
536 seq_printf(seq, ",jqfmt=%s",
537 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
539 if (sbi->s_qf_names[USRQUOTA])
540 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
542 if (sbi->s_qf_names[GRPQUOTA])
543 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
545 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
546 seq_puts(seq, ",usrquota");
548 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
549 seq_puts(seq, ",grpquota");
555 * - it's set to a non-default value OR
556 * - if the per-sb default is different from the global default
558 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
560 struct super_block *sb = vfs->mnt_sb;
561 struct ext3_sb_info *sbi = EXT3_SB(sb);
562 struct ext3_super_block *es = sbi->s_es;
563 unsigned long def_mount_opts;
565 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
567 if (sbi->s_sb_block != 1)
568 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
569 if (test_opt(sb, MINIX_DF))
570 seq_puts(seq, ",minixdf");
571 if (test_opt(sb, GRPID))
572 seq_puts(seq, ",grpid");
573 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
574 seq_puts(seq, ",nogrpid");
575 if (sbi->s_resuid != EXT3_DEF_RESUID ||
576 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
577 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
579 if (sbi->s_resgid != EXT3_DEF_RESGID ||
580 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
581 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
583 if (test_opt(sb, ERRORS_RO)) {
584 int def_errors = le16_to_cpu(es->s_errors);
586 if (def_errors == EXT3_ERRORS_PANIC ||
587 def_errors == EXT3_ERRORS_CONTINUE) {
588 seq_puts(seq, ",errors=remount-ro");
591 if (test_opt(sb, ERRORS_CONT))
592 seq_puts(seq, ",errors=continue");
593 if (test_opt(sb, ERRORS_PANIC))
594 seq_puts(seq, ",errors=panic");
595 if (test_opt(sb, NO_UID32))
596 seq_puts(seq, ",nouid32");
597 if (test_opt(sb, DEBUG))
598 seq_puts(seq, ",debug");
599 if (test_opt(sb, OLDALLOC))
600 seq_puts(seq, ",oldalloc");
601 #ifdef CONFIG_EXT3_FS_XATTR
602 if (test_opt(sb, XATTR_USER))
603 seq_puts(seq, ",user_xattr");
604 if (!test_opt(sb, XATTR_USER) &&
605 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
606 seq_puts(seq, ",nouser_xattr");
609 #ifdef CONFIG_EXT3_FS_POSIX_ACL
610 if (test_opt(sb, POSIX_ACL))
611 seq_puts(seq, ",acl");
612 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
613 seq_puts(seq, ",noacl");
615 if (!test_opt(sb, RESERVATION))
616 seq_puts(seq, ",noreservation");
617 if (sbi->s_commit_interval) {
618 seq_printf(seq, ",commit=%u",
619 (unsigned) (sbi->s_commit_interval / HZ));
621 if (test_opt(sb, BARRIER))
622 seq_puts(seq, ",barrier=1");
623 if (test_opt(sb, NOBH))
624 seq_puts(seq, ",nobh");
626 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
627 seq_puts(seq, ",data=journal");
628 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
629 seq_puts(seq, ",data=ordered");
630 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
631 seq_puts(seq, ",data=writeback");
633 if (test_opt(sb, DATA_ERR_ABORT))
634 seq_puts(seq, ",data_err=abort");
636 ext3_show_quota_options(seq, sb);
642 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
643 u64 ino, u32 generation)
647 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
648 return ERR_PTR(-ESTALE);
649 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
650 return ERR_PTR(-ESTALE);
652 /* iget isn't really right if the inode is currently unallocated!!
654 * ext3_read_inode will return a bad_inode if the inode had been
655 * deleted, so we should be safe.
657 * Currently we don't know the generation for parent directory, so
658 * a generation of 0 means "accept any"
660 inode = ext3_iget(sb, ino);
662 return ERR_CAST(inode);
663 if (generation && inode->i_generation != generation) {
665 return ERR_PTR(-ESTALE);
671 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
672 int fh_len, int fh_type)
674 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
678 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
679 int fh_len, int fh_type)
681 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
686 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
687 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
689 static int ext3_dquot_initialize(struct inode *inode, int type);
690 static int ext3_dquot_drop(struct inode *inode);
691 static int ext3_write_dquot(struct dquot *dquot);
692 static int ext3_acquire_dquot(struct dquot *dquot);
693 static int ext3_release_dquot(struct dquot *dquot);
694 static int ext3_mark_dquot_dirty(struct dquot *dquot);
695 static int ext3_write_info(struct super_block *sb, int type);
696 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
697 char *path, int remount);
698 static int ext3_quota_on_mount(struct super_block *sb, int type);
699 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
700 size_t len, loff_t off);
701 static ssize_t ext3_quota_write(struct super_block *sb, int type,
702 const char *data, size_t len, loff_t off);
704 static struct dquot_operations ext3_quota_operations = {
705 .initialize = ext3_dquot_initialize,
706 .drop = ext3_dquot_drop,
707 .alloc_space = dquot_alloc_space,
708 .alloc_inode = dquot_alloc_inode,
709 .free_space = dquot_free_space,
710 .free_inode = dquot_free_inode,
711 .transfer = dquot_transfer,
712 .write_dquot = ext3_write_dquot,
713 .acquire_dquot = ext3_acquire_dquot,
714 .release_dquot = ext3_release_dquot,
715 .mark_dirty = ext3_mark_dquot_dirty,
716 .write_info = ext3_write_info
719 static struct quotactl_ops ext3_qctl_operations = {
720 .quota_on = ext3_quota_on,
721 .quota_off = vfs_quota_off,
722 .quota_sync = vfs_quota_sync,
723 .get_info = vfs_get_dqinfo,
724 .set_info = vfs_set_dqinfo,
725 .get_dqblk = vfs_get_dqblk,
726 .set_dqblk = vfs_set_dqblk
730 static const struct super_operations ext3_sops = {
731 .alloc_inode = ext3_alloc_inode,
732 .destroy_inode = ext3_destroy_inode,
733 .write_inode = ext3_write_inode,
734 .dirty_inode = ext3_dirty_inode,
735 .delete_inode = ext3_delete_inode,
736 .put_super = ext3_put_super,
737 .write_super = ext3_write_super,
738 .sync_fs = ext3_sync_fs,
739 .write_super_lockfs = ext3_write_super_lockfs,
740 .unlockfs = ext3_unlockfs,
741 .statfs = ext3_statfs,
742 .remount_fs = ext3_remount,
743 .clear_inode = ext3_clear_inode,
744 .show_options = ext3_show_options,
746 .quota_read = ext3_quota_read,
747 .quota_write = ext3_quota_write,
751 static const struct export_operations ext3_export_ops = {
752 .fh_to_dentry = ext3_fh_to_dentry,
753 .fh_to_parent = ext3_fh_to_parent,
754 .get_parent = ext3_get_parent,
758 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
759 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
760 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
761 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
762 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
763 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
764 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
765 Opt_data_err_abort, Opt_data_err_ignore,
766 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
767 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
768 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
772 static const match_table_t tokens = {
773 {Opt_bsd_df, "bsddf"},
774 {Opt_minix_df, "minixdf"},
775 {Opt_grpid, "grpid"},
776 {Opt_grpid, "bsdgroups"},
777 {Opt_nogrpid, "nogrpid"},
778 {Opt_nogrpid, "sysvgroups"},
779 {Opt_resgid, "resgid=%u"},
780 {Opt_resuid, "resuid=%u"},
782 {Opt_err_cont, "errors=continue"},
783 {Opt_err_panic, "errors=panic"},
784 {Opt_err_ro, "errors=remount-ro"},
785 {Opt_nouid32, "nouid32"},
786 {Opt_nocheck, "nocheck"},
787 {Opt_nocheck, "check=none"},
788 {Opt_debug, "debug"},
789 {Opt_oldalloc, "oldalloc"},
790 {Opt_orlov, "orlov"},
791 {Opt_user_xattr, "user_xattr"},
792 {Opt_nouser_xattr, "nouser_xattr"},
794 {Opt_noacl, "noacl"},
795 {Opt_reservation, "reservation"},
796 {Opt_noreservation, "noreservation"},
797 {Opt_noload, "noload"},
800 {Opt_commit, "commit=%u"},
801 {Opt_journal_update, "journal=update"},
802 {Opt_journal_inum, "journal=%u"},
803 {Opt_journal_dev, "journal_dev=%u"},
804 {Opt_abort, "abort"},
805 {Opt_data_journal, "data=journal"},
806 {Opt_data_ordered, "data=ordered"},
807 {Opt_data_writeback, "data=writeback"},
808 {Opt_data_err_abort, "data_err=abort"},
809 {Opt_data_err_ignore, "data_err=ignore"},
810 {Opt_offusrjquota, "usrjquota="},
811 {Opt_usrjquota, "usrjquota=%s"},
812 {Opt_offgrpjquota, "grpjquota="},
813 {Opt_grpjquota, "grpjquota=%s"},
814 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
815 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
816 {Opt_grpquota, "grpquota"},
817 {Opt_noquota, "noquota"},
818 {Opt_quota, "quota"},
819 {Opt_usrquota, "usrquota"},
820 {Opt_barrier, "barrier=%u"},
821 {Opt_resize, "resize"},
825 static ext3_fsblk_t get_sb_block(void **data)
827 ext3_fsblk_t sb_block;
828 char *options = (char *) *data;
830 if (!options || strncmp(options, "sb=", 3) != 0)
831 return 1; /* Default location */
833 /*todo: use simple_strtoll with >32bit ext3 */
834 sb_block = simple_strtoul(options, &options, 0);
835 if (*options && *options != ',') {
836 printk("EXT3-fs: Invalid sb specification: %s\n",
842 *data = (void *) options;
846 static int parse_options (char *options, struct super_block *sb,
847 unsigned int *inum, unsigned long *journal_devnum,
848 ext3_fsblk_t *n_blocks_count, int is_remount)
850 struct ext3_sb_info *sbi = EXT3_SB(sb);
852 substring_t args[MAX_OPT_ARGS];
863 while ((p = strsep (&options, ",")) != NULL) {
868 token = match_token(p, tokens, args);
871 clear_opt (sbi->s_mount_opt, MINIX_DF);
874 set_opt (sbi->s_mount_opt, MINIX_DF);
877 set_opt (sbi->s_mount_opt, GRPID);
880 clear_opt (sbi->s_mount_opt, GRPID);
883 if (match_int(&args[0], &option))
885 sbi->s_resuid = option;
888 if (match_int(&args[0], &option))
890 sbi->s_resgid = option;
893 /* handled by get_sb_block() instead of here */
894 /* *sb_block = match_int(&args[0]); */
897 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
898 clear_opt (sbi->s_mount_opt, ERRORS_RO);
899 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
902 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
903 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
904 set_opt (sbi->s_mount_opt, ERRORS_RO);
907 clear_opt (sbi->s_mount_opt, ERRORS_RO);
908 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
909 set_opt (sbi->s_mount_opt, ERRORS_CONT);
912 set_opt (sbi->s_mount_opt, NO_UID32);
915 clear_opt (sbi->s_mount_opt, CHECK);
918 set_opt (sbi->s_mount_opt, DEBUG);
921 set_opt (sbi->s_mount_opt, OLDALLOC);
924 clear_opt (sbi->s_mount_opt, OLDALLOC);
926 #ifdef CONFIG_EXT3_FS_XATTR
928 set_opt (sbi->s_mount_opt, XATTR_USER);
930 case Opt_nouser_xattr:
931 clear_opt (sbi->s_mount_opt, XATTR_USER);
935 case Opt_nouser_xattr:
936 printk("EXT3 (no)user_xattr options not supported\n");
939 #ifdef CONFIG_EXT3_FS_POSIX_ACL
941 set_opt(sbi->s_mount_opt, POSIX_ACL);
944 clear_opt(sbi->s_mount_opt, POSIX_ACL);
949 printk("EXT3 (no)acl options not supported\n");
952 case Opt_reservation:
953 set_opt(sbi->s_mount_opt, RESERVATION);
955 case Opt_noreservation:
956 clear_opt(sbi->s_mount_opt, RESERVATION);
958 case Opt_journal_update:
960 /* Eventually we will want to be able to create
961 a journal file here. For now, only allow the
962 user to specify an existing inode to be the
965 printk(KERN_ERR "EXT3-fs: cannot specify "
966 "journal on remount\n");
969 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
971 case Opt_journal_inum:
973 printk(KERN_ERR "EXT3-fs: cannot specify "
974 "journal on remount\n");
977 if (match_int(&args[0], &option))
981 case Opt_journal_dev:
983 printk(KERN_ERR "EXT3-fs: cannot specify "
984 "journal on remount\n");
987 if (match_int(&args[0], &option))
989 *journal_devnum = option;
992 set_opt (sbi->s_mount_opt, NOLOAD);
995 if (match_int(&args[0], &option))
1000 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1001 sbi->s_commit_interval = HZ * option;
1003 case Opt_data_journal:
1004 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1006 case Opt_data_ordered:
1007 data_opt = EXT3_MOUNT_ORDERED_DATA;
1009 case Opt_data_writeback:
1010 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1013 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1016 "EXT3-fs: cannot change data "
1017 "mode on remount\n");
1021 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1022 sbi->s_mount_opt |= data_opt;
1025 case Opt_data_err_abort:
1026 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1028 case Opt_data_err_ignore:
1029 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1038 if (sb_any_quota_loaded(sb) &&
1039 !sbi->s_qf_names[qtype]) {
1041 "EXT3-fs: Cannot change journaled "
1042 "quota options when quota turned on.\n");
1045 qname = match_strdup(&args[0]);
1048 "EXT3-fs: not enough memory for "
1049 "storing quotafile name.\n");
1052 if (sbi->s_qf_names[qtype] &&
1053 strcmp(sbi->s_qf_names[qtype], qname)) {
1055 "EXT3-fs: %s quota file already "
1056 "specified.\n", QTYPE2NAME(qtype));
1060 sbi->s_qf_names[qtype] = qname;
1061 if (strchr(sbi->s_qf_names[qtype], '/')) {
1063 "EXT3-fs: quotafile must be on "
1064 "filesystem root.\n");
1065 kfree(sbi->s_qf_names[qtype]);
1066 sbi->s_qf_names[qtype] = NULL;
1069 set_opt(sbi->s_mount_opt, QUOTA);
1071 case Opt_offusrjquota:
1074 case Opt_offgrpjquota:
1077 if (sb_any_quota_loaded(sb) &&
1078 sbi->s_qf_names[qtype]) {
1079 printk(KERN_ERR "EXT3-fs: Cannot change "
1080 "journaled quota options when "
1081 "quota turned on.\n");
1085 * The space will be released later when all options
1086 * are confirmed to be correct
1088 sbi->s_qf_names[qtype] = NULL;
1090 case Opt_jqfmt_vfsold:
1091 qfmt = QFMT_VFS_OLD;
1093 case Opt_jqfmt_vfsv0:
1096 if (sb_any_quota_loaded(sb) &&
1097 sbi->s_jquota_fmt != qfmt) {
1098 printk(KERN_ERR "EXT3-fs: Cannot change "
1099 "journaled quota options when "
1100 "quota turned on.\n");
1103 sbi->s_jquota_fmt = qfmt;
1107 set_opt(sbi->s_mount_opt, QUOTA);
1108 set_opt(sbi->s_mount_opt, USRQUOTA);
1111 set_opt(sbi->s_mount_opt, QUOTA);
1112 set_opt(sbi->s_mount_opt, GRPQUOTA);
1115 if (sb_any_quota_loaded(sb)) {
1116 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1117 "options when quota turned on.\n");
1120 clear_opt(sbi->s_mount_opt, QUOTA);
1121 clear_opt(sbi->s_mount_opt, USRQUOTA);
1122 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1129 "EXT3-fs: quota options not supported.\n");
1133 case Opt_offusrjquota:
1134 case Opt_offgrpjquota:
1135 case Opt_jqfmt_vfsold:
1136 case Opt_jqfmt_vfsv0:
1138 "EXT3-fs: journaled quota options not "
1145 set_opt(sbi->s_mount_opt, ABORT);
1148 if (match_int(&args[0], &option))
1151 set_opt(sbi->s_mount_opt, BARRIER);
1153 clear_opt(sbi->s_mount_opt, BARRIER);
1159 printk("EXT3-fs: resize option only available "
1163 if (match_int(&args[0], &option) != 0)
1165 *n_blocks_count = option;
1168 set_opt(sbi->s_mount_opt, NOBH);
1171 clear_opt(sbi->s_mount_opt, NOBH);
1175 "EXT3-fs: Unrecognized mount option \"%s\" "
1176 "or missing value\n", p);
1181 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1182 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1183 sbi->s_qf_names[USRQUOTA])
1184 clear_opt(sbi->s_mount_opt, USRQUOTA);
1186 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1187 sbi->s_qf_names[GRPQUOTA])
1188 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1190 if ((sbi->s_qf_names[USRQUOTA] &&
1191 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1192 (sbi->s_qf_names[GRPQUOTA] &&
1193 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1194 printk(KERN_ERR "EXT3-fs: old and new quota "
1195 "format mixing.\n");
1199 if (!sbi->s_jquota_fmt) {
1200 printk(KERN_ERR "EXT3-fs: journaled quota format "
1201 "not specified.\n");
1205 if (sbi->s_jquota_fmt) {
1206 printk(KERN_ERR "EXT3-fs: journaled quota format "
1207 "specified with no journaling "
1216 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1219 struct ext3_sb_info *sbi = EXT3_SB(sb);
1222 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1223 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1224 "forcing read-only mode\n");
1229 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1230 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1231 "running e2fsck is recommended\n");
1232 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1233 printk (KERN_WARNING
1234 "EXT3-fs warning: mounting fs with errors, "
1235 "running e2fsck is recommended\n");
1236 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1237 le16_to_cpu(es->s_mnt_count) >=
1238 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1239 printk (KERN_WARNING
1240 "EXT3-fs warning: maximal mount count reached, "
1241 "running e2fsck is recommended\n");
1242 else if (le32_to_cpu(es->s_checkinterval) &&
1243 (le32_to_cpu(es->s_lastcheck) +
1244 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1245 printk (KERN_WARNING
1246 "EXT3-fs warning: checktime reached, "
1247 "running e2fsck is recommended\n");
1249 /* @@@ We _will_ want to clear the valid bit if we find
1250 inconsistencies, to force a fsck at reboot. But for
1251 a plain journaled filesystem we can keep it set as
1252 valid forever! :) */
1253 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1255 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1256 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1257 le16_add_cpu(&es->s_mnt_count, 1);
1258 es->s_mtime = cpu_to_le32(get_seconds());
1259 ext3_update_dynamic_rev(sb);
1260 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1262 ext3_commit_super(sb, es, 1);
1263 if (test_opt(sb, DEBUG))
1264 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1265 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1267 sbi->s_groups_count,
1268 EXT3_BLOCKS_PER_GROUP(sb),
1269 EXT3_INODES_PER_GROUP(sb),
1272 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1273 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1274 char b[BDEVNAME_SIZE];
1276 printk("external journal on %s\n",
1277 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1279 printk("internal journal\n");
1284 /* Called at mount-time, super-block is locked */
1285 static int ext3_check_descriptors(struct super_block *sb)
1287 struct ext3_sb_info *sbi = EXT3_SB(sb);
1290 ext3_debug ("Checking group descriptors");
1292 for (i = 0; i < sbi->s_groups_count; i++) {
1293 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1294 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1295 ext3_fsblk_t last_block;
1297 if (i == sbi->s_groups_count - 1)
1298 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1300 last_block = first_block +
1301 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1303 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1304 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1306 ext3_error (sb, "ext3_check_descriptors",
1307 "Block bitmap for group %d"
1308 " not in group (block %lu)!",
1310 le32_to_cpu(gdp->bg_block_bitmap));
1313 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1314 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1316 ext3_error (sb, "ext3_check_descriptors",
1317 "Inode bitmap for group %d"
1318 " not in group (block %lu)!",
1320 le32_to_cpu(gdp->bg_inode_bitmap));
1323 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1324 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1327 ext3_error (sb, "ext3_check_descriptors",
1328 "Inode table for group %d"
1329 " not in group (block %lu)!",
1331 le32_to_cpu(gdp->bg_inode_table));
1336 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1337 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1342 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1343 * the superblock) which were deleted from all directories, but held open by
1344 * a process at the time of a crash. We walk the list and try to delete these
1345 * inodes at recovery time (only with a read-write filesystem).
1347 * In order to keep the orphan inode chain consistent during traversal (in
1348 * case of crash during recovery), we link each inode into the superblock
1349 * orphan list_head and handle it the same way as an inode deletion during
1350 * normal operation (which journals the operations for us).
1352 * We only do an iget() and an iput() on each inode, which is very safe if we
1353 * accidentally point at an in-use or already deleted inode. The worst that
1354 * can happen in this case is that we get a "bit already cleared" message from
1355 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1356 * e2fsck was run on this filesystem, and it must have already done the orphan
1357 * inode cleanup for us, so we can safely abort without any further action.
1359 static void ext3_orphan_cleanup (struct super_block * sb,
1360 struct ext3_super_block * es)
1362 unsigned int s_flags = sb->s_flags;
1363 int nr_orphans = 0, nr_truncates = 0;
1367 if (!es->s_last_orphan) {
1368 jbd_debug(4, "no orphan inodes to clean up\n");
1372 if (bdev_read_only(sb->s_bdev)) {
1373 printk(KERN_ERR "EXT3-fs: write access "
1374 "unavailable, skipping orphan cleanup.\n");
1378 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1379 if (es->s_last_orphan)
1380 jbd_debug(1, "Errors on filesystem, "
1381 "clearing orphan list.\n");
1382 es->s_last_orphan = 0;
1383 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1387 if (s_flags & MS_RDONLY) {
1388 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1390 sb->s_flags &= ~MS_RDONLY;
1393 /* Needed for iput() to work correctly and not trash data */
1394 sb->s_flags |= MS_ACTIVE;
1395 /* Turn on quotas so that they are updated correctly */
1396 for (i = 0; i < MAXQUOTAS; i++) {
1397 if (EXT3_SB(sb)->s_qf_names[i]) {
1398 int ret = ext3_quota_on_mount(sb, i);
1401 "EXT3-fs: Cannot turn on journaled "
1402 "quota: error %d\n", ret);
1407 while (es->s_last_orphan) {
1408 struct inode *inode;
1410 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1411 if (IS_ERR(inode)) {
1412 es->s_last_orphan = 0;
1416 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1418 if (inode->i_nlink) {
1420 "%s: truncating inode %lu to %Ld bytes\n",
1421 __func__, inode->i_ino, inode->i_size);
1422 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1423 inode->i_ino, inode->i_size);
1424 ext3_truncate(inode);
1428 "%s: deleting unreferenced inode %lu\n",
1429 __func__, inode->i_ino);
1430 jbd_debug(2, "deleting unreferenced inode %lu\n",
1434 iput(inode); /* The delete magic happens here! */
1437 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1440 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1441 sb->s_id, PLURAL(nr_orphans));
1443 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1444 sb->s_id, PLURAL(nr_truncates));
1446 /* Turn quotas off */
1447 for (i = 0; i < MAXQUOTAS; i++) {
1448 if (sb_dqopt(sb)->files[i])
1449 vfs_quota_off(sb, i, 0);
1452 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1456 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1457 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1458 * We need to be 1 filesystem block less than the 2^32 sector limit.
1460 static loff_t ext3_max_size(int bits)
1462 loff_t res = EXT3_NDIR_BLOCKS;
1466 /* This is calculated to be the largest file size for a
1467 * dense, file such that the total number of
1468 * sectors in the file, including data and all indirect blocks,
1469 * does not exceed 2^32 -1
1470 * __u32 i_blocks representing the total number of
1471 * 512 bytes blocks of the file
1473 upper_limit = (1LL << 32) - 1;
1475 /* total blocks in file system block size */
1476 upper_limit >>= (bits - 9);
1479 /* indirect blocks */
1481 /* double indirect blocks */
1482 meta_blocks += 1 + (1LL << (bits-2));
1483 /* tripple indirect blocks */
1484 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1486 upper_limit -= meta_blocks;
1487 upper_limit <<= bits;
1489 res += 1LL << (bits-2);
1490 res += 1LL << (2*(bits-2));
1491 res += 1LL << (3*(bits-2));
1493 if (res > upper_limit)
1496 if (res > MAX_LFS_FILESIZE)
1497 res = MAX_LFS_FILESIZE;
1502 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1503 ext3_fsblk_t logic_sb_block,
1506 struct ext3_sb_info *sbi = EXT3_SB(sb);
1507 unsigned long bg, first_meta_bg;
1510 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1512 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1514 return (logic_sb_block + nr + 1);
1515 bg = sbi->s_desc_per_block * nr;
1516 if (ext3_bg_has_super(sb, bg))
1518 return (has_super + ext3_group_first_block_no(sb, bg));
1522 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1524 struct buffer_head * bh;
1525 struct ext3_super_block *es = NULL;
1526 struct ext3_sb_info *sbi;
1528 ext3_fsblk_t sb_block = get_sb_block(&data);
1529 ext3_fsblk_t logic_sb_block;
1530 unsigned long offset = 0;
1531 unsigned int journal_inum = 0;
1532 unsigned long journal_devnum = 0;
1533 unsigned long def_mount_opts;
1544 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1547 sb->s_fs_info = sbi;
1548 sbi->s_mount_opt = 0;
1549 sbi->s_resuid = EXT3_DEF_RESUID;
1550 sbi->s_resgid = EXT3_DEF_RESGID;
1551 sbi->s_sb_block = sb_block;
1555 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1557 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1562 * The ext3 superblock will not be buffer aligned for other than 1kB
1563 * block sizes. We need to calculate the offset from buffer start.
1565 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1566 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1567 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1569 logic_sb_block = sb_block;
1572 if (!(bh = sb_bread(sb, logic_sb_block))) {
1573 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1577 * Note: s_es must be initialized as soon as possible because
1578 * some ext3 macro-instructions depend on its value
1580 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1582 sb->s_magic = le16_to_cpu(es->s_magic);
1583 if (sb->s_magic != EXT3_SUPER_MAGIC)
1586 /* Set defaults before we parse the mount options */
1587 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1588 if (def_mount_opts & EXT3_DEFM_DEBUG)
1589 set_opt(sbi->s_mount_opt, DEBUG);
1590 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1591 set_opt(sbi->s_mount_opt, GRPID);
1592 if (def_mount_opts & EXT3_DEFM_UID16)
1593 set_opt(sbi->s_mount_opt, NO_UID32);
1594 #ifdef CONFIG_EXT3_FS_XATTR
1595 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1596 set_opt(sbi->s_mount_opt, XATTR_USER);
1598 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1599 if (def_mount_opts & EXT3_DEFM_ACL)
1600 set_opt(sbi->s_mount_opt, POSIX_ACL);
1602 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1603 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1604 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1605 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1606 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1607 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1609 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1610 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1611 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1612 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1614 set_opt(sbi->s_mount_opt, ERRORS_RO);
1616 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1617 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1619 set_opt(sbi->s_mount_opt, RESERVATION);
1621 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1625 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1626 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1628 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1629 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1630 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1631 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1633 "EXT3-fs warning: feature flags set on rev 0 fs, "
1634 "running e2fsck is recommended\n");
1636 * Check feature flags regardless of the revision level, since we
1637 * previously didn't change the revision level when setting the flags,
1638 * so there is a chance incompat flags are set on a rev 0 filesystem.
1640 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1642 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1643 "unsupported optional features (%x).\n",
1644 sb->s_id, le32_to_cpu(features));
1647 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1648 if (!(sb->s_flags & MS_RDONLY) && features) {
1649 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1650 "unsupported optional features (%x).\n",
1651 sb->s_id, le32_to_cpu(features));
1654 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1656 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1657 blocksize > EXT3_MAX_BLOCK_SIZE) {
1659 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1660 blocksize, sb->s_id);
1664 hblock = bdev_hardsect_size(sb->s_bdev);
1665 if (sb->s_blocksize != blocksize) {
1667 * Make sure the blocksize for the filesystem is larger
1668 * than the hardware sectorsize for the machine.
1670 if (blocksize < hblock) {
1671 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1672 "device blocksize %d.\n", blocksize, hblock);
1677 if (!sb_set_blocksize(sb, blocksize)) {
1678 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1682 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1683 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1684 bh = sb_bread(sb, logic_sb_block);
1687 "EXT3-fs: Can't read superblock on 2nd try.\n");
1690 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1692 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1694 "EXT3-fs: Magic mismatch, very weird !\n");
1699 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1701 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1702 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1703 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1705 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1706 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1707 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1708 (!is_power_of_2(sbi->s_inode_size)) ||
1709 (sbi->s_inode_size > blocksize)) {
1711 "EXT3-fs: unsupported inode size: %d\n",
1716 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1717 le32_to_cpu(es->s_log_frag_size);
1718 if (blocksize != sbi->s_frag_size) {
1720 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1721 sbi->s_frag_size, blocksize);
1724 sbi->s_frags_per_block = 1;
1725 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1726 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1727 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1728 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1730 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1731 if (sbi->s_inodes_per_block == 0)
1733 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1734 sbi->s_inodes_per_block;
1735 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1737 sbi->s_mount_state = le16_to_cpu(es->s_state);
1738 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1739 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1740 for (i=0; i < 4; i++)
1741 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1742 sbi->s_def_hash_version = es->s_def_hash_version;
1744 if (sbi->s_blocks_per_group > blocksize * 8) {
1746 "EXT3-fs: #blocks per group too big: %lu\n",
1747 sbi->s_blocks_per_group);
1750 if (sbi->s_frags_per_group > blocksize * 8) {
1752 "EXT3-fs: #fragments per group too big: %lu\n",
1753 sbi->s_frags_per_group);
1756 if (sbi->s_inodes_per_group > blocksize * 8) {
1758 "EXT3-fs: #inodes per group too big: %lu\n",
1759 sbi->s_inodes_per_group);
1763 if (le32_to_cpu(es->s_blocks_count) >
1764 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1765 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1766 " too large to mount safely\n", sb->s_id);
1767 if (sizeof(sector_t) < 8)
1768 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1773 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1775 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1776 le32_to_cpu(es->s_first_data_block) - 1)
1777 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1778 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1779 EXT3_DESC_PER_BLOCK(sb);
1780 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1782 if (sbi->s_group_desc == NULL) {
1783 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1787 bgl_lock_init(&sbi->s_blockgroup_lock);
1789 for (i = 0; i < db_count; i++) {
1790 block = descriptor_loc(sb, logic_sb_block, i);
1791 sbi->s_group_desc[i] = sb_bread(sb, block);
1792 if (!sbi->s_group_desc[i]) {
1793 printk (KERN_ERR "EXT3-fs: "
1794 "can't read group descriptor %d\n", i);
1799 if (!ext3_check_descriptors (sb)) {
1800 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1803 sbi->s_gdb_count = db_count;
1804 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1805 spin_lock_init(&sbi->s_next_gen_lock);
1807 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1808 ext3_count_free_blocks(sb));
1810 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1811 ext3_count_free_inodes(sb));
1814 err = percpu_counter_init(&sbi->s_dirs_counter,
1815 ext3_count_dirs(sb));
1818 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1822 /* per fileystem reservation list head & lock */
1823 spin_lock_init(&sbi->s_rsv_window_lock);
1824 sbi->s_rsv_window_root = RB_ROOT;
1825 /* Add a single, static dummy reservation to the start of the
1826 * reservation window list --- it gives us a placeholder for
1827 * append-at-start-of-list which makes the allocation logic
1828 * _much_ simpler. */
1829 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1830 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1831 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1832 sbi->s_rsv_window_head.rsv_goal_size = 0;
1833 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1836 * set up enough so that it can read an inode
1838 sb->s_op = &ext3_sops;
1839 sb->s_export_op = &ext3_export_ops;
1840 sb->s_xattr = ext3_xattr_handlers;
1842 sb->s_qcop = &ext3_qctl_operations;
1843 sb->dq_op = &ext3_quota_operations;
1845 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1849 needs_recovery = (es->s_last_orphan != 0 ||
1850 EXT3_HAS_INCOMPAT_FEATURE(sb,
1851 EXT3_FEATURE_INCOMPAT_RECOVER));
1854 * The first inode we look at is the journal inode. Don't try
1855 * root first: it may be modified in the journal!
1857 if (!test_opt(sb, NOLOAD) &&
1858 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1859 if (ext3_load_journal(sb, es, journal_devnum))
1861 } else if (journal_inum) {
1862 if (ext3_create_journal(sb, es, journal_inum))
1867 "ext3: No journal on filesystem on %s\n",
1872 /* We have now updated the journal if required, so we can
1873 * validate the data journaling mode. */
1874 switch (test_opt(sb, DATA_FLAGS)) {
1876 /* No mode set, assume a default based on the journal
1877 capabilities: ORDERED_DATA if the journal can
1878 cope, else JOURNAL_DATA */
1879 if (journal_check_available_features
1880 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1881 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1883 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1886 case EXT3_MOUNT_ORDERED_DATA:
1887 case EXT3_MOUNT_WRITEBACK_DATA:
1888 if (!journal_check_available_features
1889 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1890 printk(KERN_ERR "EXT3-fs: Journal does not support "
1891 "requested data journaling mode\n");
1898 if (test_opt(sb, NOBH)) {
1899 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1900 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1901 "its supported only with writeback mode\n");
1902 clear_opt(sbi->s_mount_opt, NOBH);
1906 * The journal_load will have done any necessary log recovery,
1907 * so we can safely mount the rest of the filesystem now.
1910 root = ext3_iget(sb, EXT3_ROOT_INO);
1912 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1913 ret = PTR_ERR(root);
1916 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1918 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1921 sb->s_root = d_alloc_root(root);
1923 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1929 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1931 * akpm: core read_super() calls in here with the superblock locked.
1932 * That deadlocks, because orphan cleanup needs to lock the superblock
1933 * in numerous places. Here we just pop the lock - it's relatively
1934 * harmless, because we are now ready to accept write_super() requests,
1935 * and aviro says that's the only reason for hanging onto the
1938 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1939 ext3_orphan_cleanup(sb, es);
1940 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1942 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1943 ext3_mark_recovery_complete(sb, es);
1944 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1945 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1946 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1954 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1959 journal_destroy(sbi->s_journal);
1961 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1962 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1963 percpu_counter_destroy(&sbi->s_dirs_counter);
1965 for (i = 0; i < db_count; i++)
1966 brelse(sbi->s_group_desc[i]);
1967 kfree(sbi->s_group_desc);
1970 for (i = 0; i < MAXQUOTAS; i++)
1971 kfree(sbi->s_qf_names[i]);
1973 ext3_blkdev_remove(sbi);
1976 sb->s_fs_info = NULL;
1983 * Setup any per-fs journal parameters now. We'll do this both on
1984 * initial mount, once the journal has been initialised but before we've
1985 * done any recovery; and again on any subsequent remount.
1987 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1989 struct ext3_sb_info *sbi = EXT3_SB(sb);
1991 if (sbi->s_commit_interval)
1992 journal->j_commit_interval = sbi->s_commit_interval;
1993 /* We could also set up an ext3-specific default for the commit
1994 * interval here, but for now we'll just fall back to the jbd
1997 spin_lock(&journal->j_state_lock);
1998 if (test_opt(sb, BARRIER))
1999 journal->j_flags |= JFS_BARRIER;
2001 journal->j_flags &= ~JFS_BARRIER;
2002 if (test_opt(sb, DATA_ERR_ABORT))
2003 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2005 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2006 spin_unlock(&journal->j_state_lock);
2009 static journal_t *ext3_get_journal(struct super_block *sb,
2010 unsigned int journal_inum)
2012 struct inode *journal_inode;
2015 /* First, test for the existence of a valid inode on disk. Bad
2016 * things happen if we iget() an unused inode, as the subsequent
2017 * iput() will try to delete it. */
2019 journal_inode = ext3_iget(sb, journal_inum);
2020 if (IS_ERR(journal_inode)) {
2021 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2024 if (!journal_inode->i_nlink) {
2025 make_bad_inode(journal_inode);
2026 iput(journal_inode);
2027 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2031 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2032 journal_inode, journal_inode->i_size);
2033 if (!S_ISREG(journal_inode->i_mode)) {
2034 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2035 iput(journal_inode);
2039 journal = journal_init_inode(journal_inode);
2041 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2042 iput(journal_inode);
2045 journal->j_private = sb;
2046 ext3_init_journal_params(sb, journal);
2050 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2053 struct buffer_head * bh;
2057 int hblock, blocksize;
2058 ext3_fsblk_t sb_block;
2059 unsigned long offset;
2060 struct ext3_super_block * es;
2061 struct block_device *bdev;
2063 bdev = ext3_blkdev_get(j_dev);
2067 if (bd_claim(bdev, sb)) {
2069 "EXT3: failed to claim external journal device.\n");
2070 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2074 blocksize = sb->s_blocksize;
2075 hblock = bdev_hardsect_size(bdev);
2076 if (blocksize < hblock) {
2078 "EXT3-fs: blocksize too small for journal device.\n");
2082 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2083 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2084 set_blocksize(bdev, blocksize);
2085 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2086 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2087 "external journal\n");
2091 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2092 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2093 !(le32_to_cpu(es->s_feature_incompat) &
2094 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2095 printk(KERN_ERR "EXT3-fs: external journal has "
2096 "bad superblock\n");
2101 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2102 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2107 len = le32_to_cpu(es->s_blocks_count);
2108 start = sb_block + 1;
2109 brelse(bh); /* we're done with the superblock */
2111 journal = journal_init_dev(bdev, sb->s_bdev,
2112 start, len, blocksize);
2114 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2117 journal->j_private = sb;
2118 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2119 wait_on_buffer(journal->j_sb_buffer);
2120 if (!buffer_uptodate(journal->j_sb_buffer)) {
2121 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2124 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2125 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2126 "user (unsupported) - %d\n",
2127 be32_to_cpu(journal->j_superblock->s_nr_users));
2130 EXT3_SB(sb)->journal_bdev = bdev;
2131 ext3_init_journal_params(sb, journal);
2134 journal_destroy(journal);
2136 ext3_blkdev_put(bdev);
2140 static int ext3_load_journal(struct super_block *sb,
2141 struct ext3_super_block *es,
2142 unsigned long journal_devnum)
2145 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2148 int really_read_only;
2150 if (journal_devnum &&
2151 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2152 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2153 "numbers have changed\n");
2154 journal_dev = new_decode_dev(journal_devnum);
2156 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2158 really_read_only = bdev_read_only(sb->s_bdev);
2161 * Are we loading a blank journal or performing recovery after a
2162 * crash? For recovery, we need to check in advance whether we
2163 * can get read-write access to the device.
2166 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2167 if (sb->s_flags & MS_RDONLY) {
2168 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2169 "required on readonly filesystem.\n");
2170 if (really_read_only) {
2171 printk(KERN_ERR "EXT3-fs: write access "
2172 "unavailable, cannot proceed.\n");
2175 printk (KERN_INFO "EXT3-fs: write access will "
2176 "be enabled during recovery.\n");
2180 if (journal_inum && journal_dev) {
2181 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2182 "and inode journals!\n");
2187 if (!(journal = ext3_get_journal(sb, journal_inum)))
2190 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2194 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2195 err = journal_update_format(journal);
2197 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2198 journal_destroy(journal);
2203 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2204 err = journal_wipe(journal, !really_read_only);
2206 err = journal_load(journal);
2209 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2210 journal_destroy(journal);
2214 EXT3_SB(sb)->s_journal = journal;
2215 ext3_clear_journal_err(sb, es);
2217 if (journal_devnum &&
2218 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2219 es->s_journal_dev = cpu_to_le32(journal_devnum);
2222 /* Make sure we flush the recovery flag to disk. */
2223 ext3_commit_super(sb, es, 1);
2229 static int ext3_create_journal(struct super_block * sb,
2230 struct ext3_super_block * es,
2231 unsigned int journal_inum)
2236 if (sb->s_flags & MS_RDONLY) {
2237 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2238 "create journal.\n");
2242 journal = ext3_get_journal(sb, journal_inum);
2246 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2249 err = journal_create(journal);
2251 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2252 journal_destroy(journal);
2256 EXT3_SB(sb)->s_journal = journal;
2258 ext3_update_dynamic_rev(sb);
2259 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2260 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2262 es->s_journal_inum = cpu_to_le32(journal_inum);
2265 /* Make sure we flush the recovery flag to disk. */
2266 ext3_commit_super(sb, es, 1);
2271 static void ext3_commit_super (struct super_block * sb,
2272 struct ext3_super_block * es,
2275 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2279 es->s_wtime = cpu_to_le32(get_seconds());
2280 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2281 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2282 BUFFER_TRACE(sbh, "marking dirty");
2283 mark_buffer_dirty(sbh);
2285 sync_dirty_buffer(sbh);
2290 * Have we just finished recovery? If so, and if we are mounting (or
2291 * remounting) the filesystem readonly, then we will end up with a
2292 * consistent fs on disk. Record that fact.
2294 static void ext3_mark_recovery_complete(struct super_block * sb,
2295 struct ext3_super_block * es)
2297 journal_t *journal = EXT3_SB(sb)->s_journal;
2299 journal_lock_updates(journal);
2300 if (journal_flush(journal) < 0)
2304 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2305 sb->s_flags & MS_RDONLY) {
2306 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2308 ext3_commit_super(sb, es, 1);
2313 journal_unlock_updates(journal);
2317 * If we are mounting (or read-write remounting) a filesystem whose journal
2318 * has recorded an error from a previous lifetime, move that error to the
2319 * main filesystem now.
2321 static void ext3_clear_journal_err(struct super_block * sb,
2322 struct ext3_super_block * es)
2328 journal = EXT3_SB(sb)->s_journal;
2331 * Now check for any error status which may have been recorded in the
2332 * journal by a prior ext3_error() or ext3_abort()
2335 j_errno = journal_errno(journal);
2339 errstr = ext3_decode_error(sb, j_errno, nbuf);
2340 ext3_warning(sb, __func__, "Filesystem error recorded "
2341 "from previous mount: %s", errstr);
2342 ext3_warning(sb, __func__, "Marking fs in need of "
2343 "filesystem check.");
2345 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2346 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2347 ext3_commit_super (sb, es, 1);
2349 journal_clear_err(journal);
2354 * Force the running and committing transactions to commit,
2355 * and wait on the commit.
2357 int ext3_force_commit(struct super_block *sb)
2362 if (sb->s_flags & MS_RDONLY)
2365 journal = EXT3_SB(sb)->s_journal;
2367 ret = ext3_journal_force_commit(journal);
2372 * Ext3 always journals updates to the superblock itself, so we don't
2373 * have to propagate any other updates to the superblock on disk at this
2374 * point. (We can probably nuke this function altogether, and remove
2375 * any mention to sb->s_dirt in all of fs/ext3; eventual cleanup...)
2377 static void ext3_write_super (struct super_block * sb)
2379 if (mutex_trylock(&sb->s_lock) != 0)
2384 static int ext3_sync_fs(struct super_block *sb, int wait)
2388 ext3_force_commit(sb);
2390 journal_start_commit(EXT3_SB(sb)->s_journal, NULL);
2396 * LVM calls this function before a (read-only) snapshot is created. This
2397 * gives us a chance to flush the journal completely and mark the fs clean.
2399 static void ext3_write_super_lockfs(struct super_block *sb)
2403 if (!(sb->s_flags & MS_RDONLY)) {
2404 journal_t *journal = EXT3_SB(sb)->s_journal;
2406 /* Now we set up the journal barrier. */
2407 journal_lock_updates(journal);
2410 * We don't want to clear needs_recovery flag when we failed
2411 * to flush the journal.
2413 if (journal_flush(journal) < 0)
2416 /* Journal blocked and flushed, clear needs_recovery flag. */
2417 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2418 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2423 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2424 * flag here, even though the filesystem is not technically dirty yet.
2426 static void ext3_unlockfs(struct super_block *sb)
2428 if (!(sb->s_flags & MS_RDONLY)) {
2430 /* Reser the needs_recovery flag before the fs is unlocked. */
2431 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2432 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2434 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2438 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2440 struct ext3_super_block * es;
2441 struct ext3_sb_info *sbi = EXT3_SB(sb);
2442 ext3_fsblk_t n_blocks_count = 0;
2443 unsigned long old_sb_flags;
2444 struct ext3_mount_options old_opts;
2450 /* Store the original options */
2451 old_sb_flags = sb->s_flags;
2452 old_opts.s_mount_opt = sbi->s_mount_opt;
2453 old_opts.s_resuid = sbi->s_resuid;
2454 old_opts.s_resgid = sbi->s_resgid;
2455 old_opts.s_commit_interval = sbi->s_commit_interval;
2457 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2458 for (i = 0; i < MAXQUOTAS; i++)
2459 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2463 * Allow the "check" option to be passed as a remount option.
2465 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2470 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2471 ext3_abort(sb, __func__, "Abort forced by user");
2473 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2474 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2478 ext3_init_journal_params(sb, sbi->s_journal);
2480 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2481 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2482 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2487 if (*flags & MS_RDONLY) {
2489 * First of all, the unconditional stuff we have to do
2490 * to disable replay of the journal when we next remount
2492 sb->s_flags |= MS_RDONLY;
2495 * OK, test if we are remounting a valid rw partition
2496 * readonly, and if so set the rdonly flag and then
2497 * mark the partition as valid again.
2499 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2500 (sbi->s_mount_state & EXT3_VALID_FS))
2501 es->s_state = cpu_to_le16(sbi->s_mount_state);
2504 * We have to unlock super so that we can wait for
2508 ext3_mark_recovery_complete(sb, es);
2512 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2513 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2514 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2515 "remount RDWR because of unsupported "
2516 "optional features (%x).\n",
2517 sb->s_id, le32_to_cpu(ret));
2523 * If we have an unprocessed orphan list hanging
2524 * around from a previously readonly bdev mount,
2525 * require a full umount/remount for now.
2527 if (es->s_last_orphan) {
2528 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2529 "remount RDWR because of unprocessed "
2530 "orphan inode list. Please "
2531 "umount/remount instead.\n",
2538 * Mounting a RDONLY partition read-write, so reread
2539 * and store the current valid flag. (It may have
2540 * been changed by e2fsck since we originally mounted
2543 ext3_clear_journal_err(sb, es);
2544 sbi->s_mount_state = le16_to_cpu(es->s_state);
2545 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2547 if (!ext3_setup_super (sb, es, 0))
2548 sb->s_flags &= ~MS_RDONLY;
2552 /* Release old quota file names */
2553 for (i = 0; i < MAXQUOTAS; i++)
2554 if (old_opts.s_qf_names[i] &&
2555 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2556 kfree(old_opts.s_qf_names[i]);
2560 sb->s_flags = old_sb_flags;
2561 sbi->s_mount_opt = old_opts.s_mount_opt;
2562 sbi->s_resuid = old_opts.s_resuid;
2563 sbi->s_resgid = old_opts.s_resgid;
2564 sbi->s_commit_interval = old_opts.s_commit_interval;
2566 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2567 for (i = 0; i < MAXQUOTAS; i++) {
2568 if (sbi->s_qf_names[i] &&
2569 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2570 kfree(sbi->s_qf_names[i]);
2571 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2577 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2579 struct super_block *sb = dentry->d_sb;
2580 struct ext3_sb_info *sbi = EXT3_SB(sb);
2581 struct ext3_super_block *es = sbi->s_es;
2584 if (test_opt(sb, MINIX_DF)) {
2585 sbi->s_overhead_last = 0;
2586 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2587 unsigned long ngroups = sbi->s_groups_count, i;
2588 ext3_fsblk_t overhead = 0;
2592 * Compute the overhead (FS structures). This is constant
2593 * for a given filesystem unless the number of block groups
2594 * changes so we cache the previous value until it does.
2598 * All of the blocks before first_data_block are
2601 overhead = le32_to_cpu(es->s_first_data_block);
2604 * Add the overhead attributed to the superblock and
2605 * block group descriptors. If the sparse superblocks
2606 * feature is turned on, then not all groups have this.
2608 for (i = 0; i < ngroups; i++) {
2609 overhead += ext3_bg_has_super(sb, i) +
2610 ext3_bg_num_gdb(sb, i);
2615 * Every block group has an inode bitmap, a block
2616 * bitmap, and an inode table.
2618 overhead += ngroups * (2 + sbi->s_itb_per_group);
2619 sbi->s_overhead_last = overhead;
2621 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2624 buf->f_type = EXT3_SUPER_MAGIC;
2625 buf->f_bsize = sb->s_blocksize;
2626 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2627 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2628 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2629 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2630 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2632 buf->f_files = le32_to_cpu(es->s_inodes_count);
2633 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2634 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2635 buf->f_namelen = EXT3_NAME_LEN;
2636 fsid = le64_to_cpup((void *)es->s_uuid) ^
2637 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2638 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2639 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2643 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2644 * is locked for write. Otherwise the are possible deadlocks:
2645 * Process 1 Process 2
2646 * ext3_create() quota_sync()
2647 * journal_start() write_dquot()
2648 * DQUOT_INIT() down(dqio_mutex)
2649 * down(dqio_mutex) journal_start()
2655 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2657 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2660 static int ext3_dquot_initialize(struct inode *inode, int type)
2665 /* We may create quota structure so we need to reserve enough blocks */
2666 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2668 return PTR_ERR(handle);
2669 ret = dquot_initialize(inode, type);
2670 err = ext3_journal_stop(handle);
2676 static int ext3_dquot_drop(struct inode *inode)
2681 /* We may delete quota structure so we need to reserve enough blocks */
2682 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2683 if (IS_ERR(handle)) {
2685 * We call dquot_drop() anyway to at least release references
2686 * to quota structures so that umount does not hang.
2689 return PTR_ERR(handle);
2691 ret = dquot_drop(inode);
2692 err = ext3_journal_stop(handle);
2698 static int ext3_write_dquot(struct dquot *dquot)
2702 struct inode *inode;
2704 inode = dquot_to_inode(dquot);
2705 handle = ext3_journal_start(inode,
2706 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2708 return PTR_ERR(handle);
2709 ret = dquot_commit(dquot);
2710 err = ext3_journal_stop(handle);
2716 static int ext3_acquire_dquot(struct dquot *dquot)
2721 handle = ext3_journal_start(dquot_to_inode(dquot),
2722 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2724 return PTR_ERR(handle);
2725 ret = dquot_acquire(dquot);
2726 err = ext3_journal_stop(handle);
2732 static int ext3_release_dquot(struct dquot *dquot)
2737 handle = ext3_journal_start(dquot_to_inode(dquot),
2738 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2739 if (IS_ERR(handle)) {
2740 /* Release dquot anyway to avoid endless cycle in dqput() */
2741 dquot_release(dquot);
2742 return PTR_ERR(handle);
2744 ret = dquot_release(dquot);
2745 err = ext3_journal_stop(handle);
2751 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2753 /* Are we journaling quotas? */
2754 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2755 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2756 dquot_mark_dquot_dirty(dquot);
2757 return ext3_write_dquot(dquot);
2759 return dquot_mark_dquot_dirty(dquot);
2763 static int ext3_write_info(struct super_block *sb, int type)
2768 /* Data block + inode block */
2769 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2771 return PTR_ERR(handle);
2772 ret = dquot_commit_info(sb, type);
2773 err = ext3_journal_stop(handle);
2780 * Turn on quotas during mount time - we need to find
2781 * the quota file and such...
2783 static int ext3_quota_on_mount(struct super_block *sb, int type)
2785 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2786 EXT3_SB(sb)->s_jquota_fmt, type);
2790 * Standard function to be called on quota_on
2792 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2793 char *name, int remount)
2798 if (!test_opt(sb, QUOTA))
2800 /* When remounting, no checks are needed and in fact, name is NULL */
2802 return vfs_quota_on(sb, type, format_id, name, remount);
2804 err = kern_path(name, LOOKUP_FOLLOW, &path);
2808 /* Quotafile not on the same filesystem? */
2809 if (path.mnt->mnt_sb != sb) {
2813 /* Journaling quota? */
2814 if (EXT3_SB(sb)->s_qf_names[type]) {
2815 /* Quotafile not of fs root? */
2816 if (path.dentry->d_parent != sb->s_root)
2818 "EXT3-fs: Quota file not on filesystem root. "
2819 "Journaled quota will not work.\n");
2823 * When we journal data on quota file, we have to flush journal to see
2824 * all updates to the file when we bypass pagecache...
2826 if (ext3_should_journal_data(path.dentry->d_inode)) {
2828 * We don't need to lock updates but journal_flush() could
2829 * otherwise be livelocked...
2831 journal_lock_updates(EXT3_SB(sb)->s_journal);
2832 err = journal_flush(EXT3_SB(sb)->s_journal);
2833 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2840 err = vfs_quota_on_path(sb, type, format_id, &path);
2845 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2846 * acquiring the locks... As quota files are never truncated and quota code
2847 * itself serializes the operations (and noone else should touch the files)
2848 * we don't have to be afraid of races */
2849 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2850 size_t len, loff_t off)
2852 struct inode *inode = sb_dqopt(sb)->files[type];
2853 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2855 int offset = off & (sb->s_blocksize - 1);
2858 struct buffer_head *bh;
2859 loff_t i_size = i_size_read(inode);
2863 if (off+len > i_size)
2866 while (toread > 0) {
2867 tocopy = sb->s_blocksize - offset < toread ?
2868 sb->s_blocksize - offset : toread;
2869 bh = ext3_bread(NULL, inode, blk, 0, &err);
2872 if (!bh) /* A hole? */
2873 memset(data, 0, tocopy);
2875 memcpy(data, bh->b_data+offset, tocopy);
2885 /* Write to quotafile (we know the transaction is already started and has
2886 * enough credits) */
2887 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2888 const char *data, size_t len, loff_t off)
2890 struct inode *inode = sb_dqopt(sb)->files[type];
2891 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2893 int offset = off & (sb->s_blocksize - 1);
2895 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2896 size_t towrite = len;
2897 struct buffer_head *bh;
2898 handle_t *handle = journal_current_handle();
2901 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2902 " cancelled because transaction is not started.\n",
2903 (unsigned long long)off, (unsigned long long)len);
2906 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2907 while (towrite > 0) {
2908 tocopy = sb->s_blocksize - offset < towrite ?
2909 sb->s_blocksize - offset : towrite;
2910 bh = ext3_bread(handle, inode, blk, 1, &err);
2913 if (journal_quota) {
2914 err = ext3_journal_get_write_access(handle, bh);
2921 memcpy(bh->b_data+offset, data, tocopy);
2922 flush_dcache_page(bh->b_page);
2925 err = ext3_journal_dirty_metadata(handle, bh);
2927 /* Always do at least ordered writes for quotas */
2928 err = ext3_journal_dirty_data(handle, bh);
2929 mark_buffer_dirty(bh);
2940 if (len == towrite) {
2941 mutex_unlock(&inode->i_mutex);
2944 if (inode->i_size < off+len-towrite) {
2945 i_size_write(inode, off+len-towrite);
2946 EXT3_I(inode)->i_disksize = inode->i_size;
2949 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2950 ext3_mark_inode_dirty(handle, inode);
2951 mutex_unlock(&inode->i_mutex);
2952 return len - towrite;
2957 static int ext3_get_sb(struct file_system_type *fs_type,
2958 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2960 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2963 static struct file_system_type ext3_fs_type = {
2964 .owner = THIS_MODULE,
2966 .get_sb = ext3_get_sb,
2967 .kill_sb = kill_block_super,
2968 .fs_flags = FS_REQUIRES_DEV,
2971 static int __init init_ext3_fs(void)
2973 int err = init_ext3_xattr();
2976 err = init_inodecache();
2979 err = register_filesystem(&ext3_fs_type);
2984 destroy_inodecache();
2990 static void __exit exit_ext3_fs(void)
2992 unregister_filesystem(&ext3_fs_type);
2993 destroy_inodecache();
2997 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2998 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2999 MODULE_LICENSE("GPL");
3000 module_init(init_ext3_fs)
3001 module_exit(exit_ext3_fs)