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/config.h>
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/jbd.h>
25 #include <linux/ext3_fs.h>
26 #include <linux/ext3_jbd.h>
27 #include <linux/slab.h>
28 #include <linux/init.h>
29 #include <linux/blkdev.h>
30 #include <linux/parser.h>
31 #include <linux/smp_lock.h>
32 #include <linux/buffer_head.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>
40 #include <asm/uaccess.h>
46 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
47 unsigned long journal_devnum);
48 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
50 static void ext3_commit_super (struct super_block * sb,
51 struct ext3_super_block * es,
53 static void ext3_mark_recovery_complete(struct super_block * sb,
54 struct ext3_super_block * es);
55 static void ext3_clear_journal_err(struct super_block * sb,
56 struct ext3_super_block * es);
57 static int ext3_sync_fs(struct super_block *sb, int wait);
58 static const char *ext3_decode_error(struct super_block * sb, int errno,
60 static int ext3_remount (struct super_block * sb, int * flags, char * data);
61 static int ext3_statfs (struct super_block * sb, struct kstatfs * buf);
62 static void ext3_unlockfs(struct super_block *sb);
63 static void ext3_write_super (struct super_block * sb);
64 static void ext3_write_super_lockfs(struct super_block *sb);
67 * Wrappers for journal_start/end.
69 * The only special thing we need to do here is to make sure that all
70 * journal_end calls result in the superblock being marked dirty, so
71 * that sync() will call the filesystem's write_super callback if
74 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
78 if (sb->s_flags & MS_RDONLY)
79 return ERR_PTR(-EROFS);
81 /* Special case here: if the journal has aborted behind our
82 * backs (eg. EIO in the commit thread), then we still need to
83 * take the FS itself readonly cleanly. */
84 journal = EXT3_SB(sb)->s_journal;
85 if (is_journal_aborted(journal)) {
86 ext3_abort(sb, __FUNCTION__,
87 "Detected aborted journal");
88 return ERR_PTR(-EROFS);
91 return journal_start(journal, nblocks);
95 * The only special thing we need to do here is to make sure that all
96 * journal_stop calls result in the superblock being marked dirty, so
97 * that sync() will call the filesystem's write_super callback if
100 int __ext3_journal_stop(const char *where, handle_t *handle)
102 struct super_block *sb;
106 sb = handle->h_transaction->t_journal->j_private;
108 rc = journal_stop(handle);
113 __ext3_std_error(sb, where, err);
117 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
118 struct buffer_head *bh, handle_t *handle, int err)
121 const char *errstr = ext3_decode_error(NULL, err, nbuf);
124 BUFFER_TRACE(bh, "abort");
129 if (is_handle_aborted(handle))
132 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
133 caller, errstr, err_fn);
135 journal_abort_handle(handle);
138 /* Deal with the reporting of failure conditions on a filesystem such as
139 * inconsistencies detected or read IO failures.
141 * On ext2, we can store the error state of the filesystem in the
142 * superblock. That is not possible on ext3, because we may have other
143 * write ordering constraints on the superblock which prevent us from
144 * writing it out straight away; and given that the journal is about to
145 * be aborted, we can't rely on the current, or future, transactions to
146 * write out the superblock safely.
148 * We'll just use the journal_abort() error code to record an error in
149 * the journal instead. On recovery, the journal will compain about
150 * that error until we've noted it down and cleared it.
153 static void ext3_handle_error(struct super_block *sb)
155 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
158 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160 if (sb->s_flags & MS_RDONLY)
163 if (test_opt (sb, ERRORS_RO)) {
164 printk (KERN_CRIT "Remounting filesystem read-only\n");
165 sb->s_flags |= MS_RDONLY;
167 journal_t *journal = EXT3_SB(sb)->s_journal;
169 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
171 journal_abort(journal, -EIO);
173 if (test_opt(sb, ERRORS_PANIC))
174 panic("EXT3-fs (device %s): panic forced after error\n",
176 ext3_commit_super(sb, es, 1);
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:%ld 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 init_MUTEX(&ei->truncate_sem);
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,
486 if (ext3_inode_cachep == NULL)
491 static void destroy_inodecache(void)
493 if (kmem_cache_destroy(ext3_inode_cachep))
494 printk(KERN_INFO "ext3_inode_cache: not all structures were freed\n");
497 static void ext3_clear_inode(struct inode *inode)
499 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
500 #ifdef CONFIG_EXT3_FS_POSIX_ACL
501 if (EXT3_I(inode)->i_acl &&
502 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
503 posix_acl_release(EXT3_I(inode)->i_acl);
504 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
506 if (EXT3_I(inode)->i_default_acl &&
507 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
508 posix_acl_release(EXT3_I(inode)->i_default_acl);
509 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
512 ext3_discard_reservation(inode);
513 EXT3_I(inode)->i_block_alloc_info = NULL;
517 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
519 #if defined(CONFIG_QUOTA)
520 struct ext3_sb_info *sbi = EXT3_SB(sb);
522 if (sbi->s_jquota_fmt)
523 seq_printf(seq, ",jqfmt=%s",
524 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
526 if (sbi->s_qf_names[USRQUOTA])
527 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
529 if (sbi->s_qf_names[GRPQUOTA])
530 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
532 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
533 seq_puts(seq, ",usrquota");
535 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
536 seq_puts(seq, ",grpquota");
540 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
542 struct super_block *sb = vfs->mnt_sb;
544 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
545 seq_puts(seq, ",data=journal");
546 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
547 seq_puts(seq, ",data=ordered");
548 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
549 seq_puts(seq, ",data=writeback");
551 ext3_show_quota_options(seq, sb);
557 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
558 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
560 static int ext3_dquot_initialize(struct inode *inode, int type);
561 static int ext3_dquot_drop(struct inode *inode);
562 static int ext3_write_dquot(struct dquot *dquot);
563 static int ext3_acquire_dquot(struct dquot *dquot);
564 static int ext3_release_dquot(struct dquot *dquot);
565 static int ext3_mark_dquot_dirty(struct dquot *dquot);
566 static int ext3_write_info(struct super_block *sb, int type);
567 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
568 static int ext3_quota_on_mount(struct super_block *sb, int type);
569 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
570 size_t len, loff_t off);
571 static ssize_t ext3_quota_write(struct super_block *sb, int type,
572 const char *data, size_t len, loff_t off);
574 static struct dquot_operations ext3_quota_operations = {
575 .initialize = ext3_dquot_initialize,
576 .drop = ext3_dquot_drop,
577 .alloc_space = dquot_alloc_space,
578 .alloc_inode = dquot_alloc_inode,
579 .free_space = dquot_free_space,
580 .free_inode = dquot_free_inode,
581 .transfer = dquot_transfer,
582 .write_dquot = ext3_write_dquot,
583 .acquire_dquot = ext3_acquire_dquot,
584 .release_dquot = ext3_release_dquot,
585 .mark_dirty = ext3_mark_dquot_dirty,
586 .write_info = ext3_write_info
589 static struct quotactl_ops ext3_qctl_operations = {
590 .quota_on = ext3_quota_on,
591 .quota_off = vfs_quota_off,
592 .quota_sync = vfs_quota_sync,
593 .get_info = vfs_get_dqinfo,
594 .set_info = vfs_set_dqinfo,
595 .get_dqblk = vfs_get_dqblk,
596 .set_dqblk = vfs_set_dqblk
600 static struct super_operations ext3_sops = {
601 .alloc_inode = ext3_alloc_inode,
602 .destroy_inode = ext3_destroy_inode,
603 .read_inode = ext3_read_inode,
604 .write_inode = ext3_write_inode,
605 .dirty_inode = ext3_dirty_inode,
606 .delete_inode = ext3_delete_inode,
607 .put_super = ext3_put_super,
608 .write_super = ext3_write_super,
609 .sync_fs = ext3_sync_fs,
610 .write_super_lockfs = ext3_write_super_lockfs,
611 .unlockfs = ext3_unlockfs,
612 .statfs = ext3_statfs,
613 .remount_fs = ext3_remount,
614 .clear_inode = ext3_clear_inode,
615 .show_options = ext3_show_options,
617 .quota_read = ext3_quota_read,
618 .quota_write = ext3_quota_write,
622 static struct dentry *ext3_get_dentry(struct super_block *sb, void *vobjp)
625 unsigned long ino = objp[0];
626 __u32 generation = objp[1];
628 struct dentry *result;
630 if (ino != EXT3_ROOT_INO && ino < EXT3_FIRST_INO(sb))
631 return ERR_PTR(-ESTALE);
632 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
633 return ERR_PTR(-ESTALE);
635 /* iget isn't really right if the inode is currently unallocated!!
636 * ext3_read_inode currently does appropriate checks, but
637 * it might be "neater" to call ext3_get_inode first and check
638 * if the inode is valid.....
640 inode = iget(sb, ino);
642 return ERR_PTR(-ENOMEM);
643 if (is_bad_inode(inode)
644 || (generation && inode->i_generation != generation)
646 /* we didn't find the right inode.. */
648 return ERR_PTR(-ESTALE);
650 /* now to find a dentry.
651 * If possible, get a well-connected one
653 result = d_alloc_anon(inode);
656 return ERR_PTR(-ENOMEM);
661 static struct export_operations ext3_export_ops = {
662 .get_parent = ext3_get_parent,
663 .get_dentry = ext3_get_dentry,
667 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
668 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
669 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
670 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
671 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh,
672 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
673 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
674 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
675 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
676 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
677 Opt_grpquota, Opt_tagxid
680 static match_table_t tokens = {
681 {Opt_bsd_df, "bsddf"},
682 {Opt_minix_df, "minixdf"},
683 {Opt_grpid, "grpid"},
684 {Opt_grpid, "bsdgroups"},
685 {Opt_nogrpid, "nogrpid"},
686 {Opt_nogrpid, "sysvgroups"},
687 {Opt_resgid, "resgid=%u"},
688 {Opt_resuid, "resuid=%u"},
690 {Opt_err_cont, "errors=continue"},
691 {Opt_err_panic, "errors=panic"},
692 {Opt_err_ro, "errors=remount-ro"},
693 {Opt_nouid32, "nouid32"},
694 {Opt_nocheck, "nocheck"},
695 {Opt_nocheck, "check=none"},
696 {Opt_debug, "debug"},
697 {Opt_oldalloc, "oldalloc"},
698 {Opt_orlov, "orlov"},
699 {Opt_user_xattr, "user_xattr"},
700 {Opt_nouser_xattr, "nouser_xattr"},
702 {Opt_noacl, "noacl"},
703 {Opt_reservation, "reservation"},
704 {Opt_noreservation, "noreservation"},
705 {Opt_noload, "noload"},
707 {Opt_commit, "commit=%u"},
708 {Opt_journal_update, "journal=update"},
709 {Opt_journal_inum, "journal=%u"},
710 {Opt_journal_dev, "journal_dev=%u"},
711 {Opt_abort, "abort"},
712 {Opt_data_journal, "data=journal"},
713 {Opt_data_ordered, "data=ordered"},
714 {Opt_data_writeback, "data=writeback"},
715 {Opt_offusrjquota, "usrjquota="},
716 {Opt_usrjquota, "usrjquota=%s"},
717 {Opt_offgrpjquota, "grpjquota="},
718 {Opt_grpjquota, "grpjquota=%s"},
719 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
720 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
721 {Opt_grpquota, "grpquota"},
722 {Opt_noquota, "noquota"},
723 {Opt_quota, "quota"},
724 {Opt_usrquota, "usrquota"},
725 {Opt_barrier, "barrier=%u"},
726 {Opt_tagxid, "tagxid"},
728 {Opt_resize, "resize"},
731 static unsigned long get_sb_block(void **data)
733 unsigned long sb_block;
734 char *options = (char *) *data;
736 if (!options || strncmp(options, "sb=", 3) != 0)
737 return 1; /* Default location */
739 sb_block = simple_strtoul(options, &options, 0);
740 if (*options && *options != ',') {
741 printk("EXT3-fs: Invalid sb specification: %s\n",
747 *data = (void *) options;
751 static int parse_options (char *options, struct super_block *sb,
752 unsigned long *inum, unsigned long *journal_devnum,
753 unsigned long *n_blocks_count, int is_remount)
755 struct ext3_sb_info *sbi = EXT3_SB(sb);
757 substring_t args[MAX_OPT_ARGS];
768 while ((p = strsep (&options, ",")) != NULL) {
773 token = match_token(p, tokens, args);
776 clear_opt (sbi->s_mount_opt, MINIX_DF);
779 set_opt (sbi->s_mount_opt, MINIX_DF);
782 set_opt (sbi->s_mount_opt, GRPID);
785 clear_opt (sbi->s_mount_opt, GRPID);
788 if (match_int(&args[0], &option))
790 sbi->s_resuid = option;
793 if (match_int(&args[0], &option))
795 sbi->s_resgid = option;
798 /* handled by get_sb_block() instead of here */
799 /* *sb_block = match_int(&args[0]); */
802 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
803 clear_opt (sbi->s_mount_opt, ERRORS_RO);
804 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
807 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
808 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
809 set_opt (sbi->s_mount_opt, ERRORS_RO);
812 clear_opt (sbi->s_mount_opt, ERRORS_RO);
813 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
814 set_opt (sbi->s_mount_opt, ERRORS_CONT);
817 set_opt (sbi->s_mount_opt, NO_UID32);
819 #ifndef CONFIG_INOXID_NONE
821 set_opt (sbi->s_mount_opt, TAGXID);
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);
1062 "EXT3-fs: Unrecognized mount option \"%s\" "
1063 "or missing value\n", p);
1068 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1069 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1070 sbi->s_qf_names[USRQUOTA])
1071 clear_opt(sbi->s_mount_opt, USRQUOTA);
1073 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1074 sbi->s_qf_names[GRPQUOTA])
1075 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1077 if ((sbi->s_qf_names[USRQUOTA] &&
1078 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1079 (sbi->s_qf_names[GRPQUOTA] &&
1080 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1081 printk(KERN_ERR "EXT3-fs: old and new quota "
1082 "format mixing.\n");
1086 if (!sbi->s_jquota_fmt) {
1087 printk(KERN_ERR "EXT3-fs: journalled quota format "
1088 "not specified.\n");
1092 if (sbi->s_jquota_fmt) {
1093 printk(KERN_ERR "EXT3-fs: journalled quota format "
1094 "specified with no journalling "
1103 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1106 struct ext3_sb_info *sbi = EXT3_SB(sb);
1109 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1110 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1111 "forcing read-only mode\n");
1116 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1117 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1118 "running e2fsck is recommended\n");
1119 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1120 printk (KERN_WARNING
1121 "EXT3-fs warning: mounting fs with errors, "
1122 "running e2fsck is recommended\n");
1123 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1124 le16_to_cpu(es->s_mnt_count) >=
1125 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1126 printk (KERN_WARNING
1127 "EXT3-fs warning: maximal mount count reached, "
1128 "running e2fsck is recommended\n");
1129 else if (le32_to_cpu(es->s_checkinterval) &&
1130 (le32_to_cpu(es->s_lastcheck) +
1131 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1132 printk (KERN_WARNING
1133 "EXT3-fs warning: checktime reached, "
1134 "running e2fsck is recommended\n");
1136 /* @@@ We _will_ want to clear the valid bit if we find
1137 inconsistencies, to force a fsck at reboot. But for
1138 a plain journaled filesystem we can keep it set as
1139 valid forever! :) */
1140 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
1142 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1143 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1144 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1145 es->s_mtime = cpu_to_le32(get_seconds());
1146 ext3_update_dynamic_rev(sb);
1147 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1149 ext3_commit_super(sb, es, 1);
1150 if (test_opt(sb, DEBUG))
1151 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1152 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1154 sbi->s_groups_count,
1155 EXT3_BLOCKS_PER_GROUP(sb),
1156 EXT3_INODES_PER_GROUP(sb),
1159 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1160 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1161 char b[BDEVNAME_SIZE];
1163 printk("external journal on %s\n",
1164 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1166 printk("internal journal\n");
1171 /* Called at mount-time, super-block is locked */
1172 static int ext3_check_descriptors (struct super_block * sb)
1174 struct ext3_sb_info *sbi = EXT3_SB(sb);
1175 unsigned long block = le32_to_cpu(sbi->s_es->s_first_data_block);
1176 struct ext3_group_desc * gdp = NULL;
1180 ext3_debug ("Checking group descriptors");
1182 for (i = 0; i < sbi->s_groups_count; i++)
1184 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1185 gdp = (struct ext3_group_desc *)
1186 sbi->s_group_desc[desc_block++]->b_data;
1187 if (le32_to_cpu(gdp->bg_block_bitmap) < block ||
1188 le32_to_cpu(gdp->bg_block_bitmap) >=
1189 block + EXT3_BLOCKS_PER_GROUP(sb))
1191 ext3_error (sb, "ext3_check_descriptors",
1192 "Block bitmap for group %d"
1193 " not in group (block %lu)!",
1195 le32_to_cpu(gdp->bg_block_bitmap));
1198 if (le32_to_cpu(gdp->bg_inode_bitmap) < block ||
1199 le32_to_cpu(gdp->bg_inode_bitmap) >=
1200 block + EXT3_BLOCKS_PER_GROUP(sb))
1202 ext3_error (sb, "ext3_check_descriptors",
1203 "Inode bitmap for group %d"
1204 " not in group (block %lu)!",
1206 le32_to_cpu(gdp->bg_inode_bitmap));
1209 if (le32_to_cpu(gdp->bg_inode_table) < block ||
1210 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >=
1211 block + EXT3_BLOCKS_PER_GROUP(sb))
1213 ext3_error (sb, "ext3_check_descriptors",
1214 "Inode table for group %d"
1215 " not in group (block %lu)!",
1217 le32_to_cpu(gdp->bg_inode_table));
1220 block += EXT3_BLOCKS_PER_GROUP(sb);
1224 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1225 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1230 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1231 * the superblock) which were deleted from all directories, but held open by
1232 * a process at the time of a crash. We walk the list and try to delete these
1233 * inodes at recovery time (only with a read-write filesystem).
1235 * In order to keep the orphan inode chain consistent during traversal (in
1236 * case of crash during recovery), we link each inode into the superblock
1237 * orphan list_head and handle it the same way as an inode deletion during
1238 * normal operation (which journals the operations for us).
1240 * We only do an iget() and an iput() on each inode, which is very safe if we
1241 * accidentally point at an in-use or already deleted inode. The worst that
1242 * can happen in this case is that we get a "bit already cleared" message from
1243 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1244 * e2fsck was run on this filesystem, and it must have already done the orphan
1245 * inode cleanup for us, so we can safely abort without any further action.
1247 static void ext3_orphan_cleanup (struct super_block * sb,
1248 struct ext3_super_block * es)
1250 unsigned int s_flags = sb->s_flags;
1251 int nr_orphans = 0, nr_truncates = 0;
1255 if (!es->s_last_orphan) {
1256 jbd_debug(4, "no orphan inodes to clean up\n");
1260 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1261 if (es->s_last_orphan)
1262 jbd_debug(1, "Errors on filesystem, "
1263 "clearing orphan list.\n");
1264 es->s_last_orphan = 0;
1265 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1269 if (s_flags & MS_RDONLY) {
1270 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1272 sb->s_flags &= ~MS_RDONLY;
1275 /* Needed for iput() to work correctly and not trash data */
1276 sb->s_flags |= MS_ACTIVE;
1277 /* Turn on quotas so that they are updated correctly */
1278 for (i = 0; i < MAXQUOTAS; i++) {
1279 if (EXT3_SB(sb)->s_qf_names[i]) {
1280 int ret = ext3_quota_on_mount(sb, i);
1283 "EXT3-fs: Cannot turn on journalled "
1284 "quota: error %d\n", ret);
1289 while (es->s_last_orphan) {
1290 struct inode *inode;
1293 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1294 es->s_last_orphan = 0;
1298 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1300 if (inode->i_nlink) {
1302 "%s: truncating inode %ld to %Ld bytes\n",
1303 __FUNCTION__, inode->i_ino, inode->i_size);
1304 jbd_debug(2, "truncating inode %ld to %Ld bytes\n",
1305 inode->i_ino, inode->i_size);
1306 ext3_truncate(inode);
1310 "%s: deleting unreferenced inode %ld\n",
1311 __FUNCTION__, inode->i_ino);
1312 jbd_debug(2, "deleting unreferenced inode %ld\n",
1316 iput(inode); /* The delete magic happens here! */
1319 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1322 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1323 sb->s_id, PLURAL(nr_orphans));
1325 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1326 sb->s_id, PLURAL(nr_truncates));
1328 /* Turn quotas off */
1329 for (i = 0; i < MAXQUOTAS; i++) {
1330 if (sb_dqopt(sb)->files[i])
1331 vfs_quota_off(sb, i);
1334 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1337 #define log2(n) ffz(~(n))
1340 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1341 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1342 * We need to be 1 filesystem block less than the 2^32 sector limit.
1344 static loff_t ext3_max_size(int bits)
1346 loff_t res = EXT3_NDIR_BLOCKS;
1347 /* This constant is calculated to be the largest file size for a
1348 * dense, 4k-blocksize file such that the total number of
1349 * sectors in the file, including data and all indirect blocks,
1350 * does not exceed 2^32. */
1351 const loff_t upper_limit = 0x1ff7fffd000LL;
1353 res += 1LL << (bits-2);
1354 res += 1LL << (2*(bits-2));
1355 res += 1LL << (3*(bits-2));
1357 if (res > upper_limit)
1362 static unsigned long descriptor_loc(struct super_block *sb,
1363 unsigned long logic_sb_block,
1366 struct ext3_sb_info *sbi = EXT3_SB(sb);
1367 unsigned long bg, first_data_block, first_meta_bg;
1370 first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1371 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1373 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1375 return (logic_sb_block + nr + 1);
1376 bg = sbi->s_desc_per_block * nr;
1377 if (ext3_bg_has_super(sb, bg))
1379 return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
1383 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1385 struct buffer_head * bh;
1386 struct ext3_super_block *es = NULL;
1387 struct ext3_sb_info *sbi;
1388 unsigned long block;
1389 unsigned long sb_block = get_sb_block(&data);
1390 unsigned long logic_sb_block;
1391 unsigned long offset = 0;
1392 unsigned long journal_inum = 0;
1393 unsigned long journal_devnum = 0;
1394 unsigned long def_mount_opts;
1403 sbi = kmalloc(sizeof(*sbi), GFP_KERNEL);
1406 sb->s_fs_info = sbi;
1407 memset(sbi, 0, sizeof(*sbi));
1408 sbi->s_mount_opt = 0;
1409 sbi->s_resuid = EXT3_DEF_RESUID;
1410 sbi->s_resgid = EXT3_DEF_RESGID;
1414 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1416 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1421 * The ext3 superblock will not be buffer aligned for other than 1kB
1422 * block sizes. We need to calculate the offset from buffer start.
1424 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1425 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1426 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1428 logic_sb_block = sb_block;
1431 if (!(bh = sb_bread(sb, logic_sb_block))) {
1432 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1436 * Note: s_es must be initialized as soon as possible because
1437 * some ext3 macro-instructions depend on its value
1439 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1441 sb->s_magic = le16_to_cpu(es->s_magic);
1442 if (sb->s_magic != EXT3_SUPER_MAGIC)
1445 /* Set defaults before we parse the mount options */
1446 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1447 if (def_mount_opts & EXT3_DEFM_DEBUG)
1448 set_opt(sbi->s_mount_opt, DEBUG);
1449 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1450 set_opt(sbi->s_mount_opt, GRPID);
1451 if (def_mount_opts & EXT3_DEFM_UID16)
1452 set_opt(sbi->s_mount_opt, NO_UID32);
1453 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1454 set_opt(sbi->s_mount_opt, XATTR_USER);
1455 if (def_mount_opts & EXT3_DEFM_ACL)
1456 set_opt(sbi->s_mount_opt, POSIX_ACL);
1457 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1458 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1459 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1460 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1461 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1462 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1464 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1465 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1466 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1467 set_opt(sbi->s_mount_opt, ERRORS_RO);
1469 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1470 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1472 set_opt(sbi->s_mount_opt, RESERVATION);
1474 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1478 if (EXT3_SB(sb)->s_mount_opt & EXT3_MOUNT_TAGXID)
1479 sb->s_flags |= MS_TAGXID;
1481 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1482 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1484 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1485 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1486 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1487 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1489 "EXT3-fs warning: feature flags set on rev 0 fs, "
1490 "running e2fsck is recommended\n");
1492 * Check feature flags regardless of the revision level, since we
1493 * previously didn't change the revision level when setting the flags,
1494 * so there is a chance incompat flags are set on a rev 0 filesystem.
1496 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1498 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1499 "unsupported optional features (%x).\n",
1500 sb->s_id, le32_to_cpu(features));
1503 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1504 if (!(sb->s_flags & MS_RDONLY) && features) {
1505 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1506 "unsupported optional features (%x).\n",
1507 sb->s_id, le32_to_cpu(features));
1510 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1512 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1513 blocksize > EXT3_MAX_BLOCK_SIZE) {
1515 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1516 blocksize, sb->s_id);
1520 hblock = bdev_hardsect_size(sb->s_bdev);
1521 if (sb->s_blocksize != blocksize) {
1523 * Make sure the blocksize for the filesystem is larger
1524 * than the hardware sectorsize for the machine.
1526 if (blocksize < hblock) {
1527 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1528 "device blocksize %d.\n", blocksize, hblock);
1533 sb_set_blocksize(sb, blocksize);
1534 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1535 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1536 bh = sb_bread(sb, logic_sb_block);
1539 "EXT3-fs: Can't read superblock on 2nd try.\n");
1542 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1544 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1546 "EXT3-fs: Magic mismatch, very weird !\n");
1551 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1553 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1554 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1555 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1557 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1558 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1559 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1560 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1561 (sbi->s_inode_size > blocksize)) {
1563 "EXT3-fs: unsupported inode size: %d\n",
1568 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1569 le32_to_cpu(es->s_log_frag_size);
1570 if (blocksize != sbi->s_frag_size) {
1572 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1573 sbi->s_frag_size, blocksize);
1576 sbi->s_frags_per_block = 1;
1577 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1578 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1579 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1580 if (EXT3_INODE_SIZE(sb) == 0)
1582 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1583 if (sbi->s_inodes_per_block == 0)
1585 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1586 sbi->s_inodes_per_block;
1587 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1589 sbi->s_mount_state = le16_to_cpu(es->s_state);
1590 sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb));
1591 sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb));
1592 for (i=0; i < 4; i++)
1593 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1594 sbi->s_def_hash_version = es->s_def_hash_version;
1596 if (sbi->s_blocks_per_group > blocksize * 8) {
1598 "EXT3-fs: #blocks per group too big: %lu\n",
1599 sbi->s_blocks_per_group);
1602 if (sbi->s_frags_per_group > blocksize * 8) {
1604 "EXT3-fs: #fragments per group too big: %lu\n",
1605 sbi->s_frags_per_group);
1608 if (sbi->s_inodes_per_group > blocksize * 8) {
1610 "EXT3-fs: #inodes per group too big: %lu\n",
1611 sbi->s_inodes_per_group);
1615 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1617 sbi->s_groups_count = (le32_to_cpu(es->s_blocks_count) -
1618 le32_to_cpu(es->s_first_data_block) +
1619 EXT3_BLOCKS_PER_GROUP(sb) - 1) /
1620 EXT3_BLOCKS_PER_GROUP(sb);
1621 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1622 EXT3_DESC_PER_BLOCK(sb);
1623 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1625 if (sbi->s_group_desc == NULL) {
1626 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1630 percpu_counter_init(&sbi->s_freeblocks_counter);
1631 percpu_counter_init(&sbi->s_freeinodes_counter);
1632 percpu_counter_init(&sbi->s_dirs_counter);
1633 bgl_lock_init(&sbi->s_blockgroup_lock);
1635 for (i = 0; i < db_count; i++) {
1636 block = descriptor_loc(sb, logic_sb_block, i);
1637 sbi->s_group_desc[i] = sb_bread(sb, block);
1638 if (!sbi->s_group_desc[i]) {
1639 printk (KERN_ERR "EXT3-fs: "
1640 "can't read group descriptor %d\n", i);
1645 if (!ext3_check_descriptors (sb)) {
1646 printk (KERN_ERR "EXT3-fs: group descriptors corrupted !\n");
1649 sbi->s_gdb_count = db_count;
1650 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1651 spin_lock_init(&sbi->s_next_gen_lock);
1652 /* per fileystem reservation list head & lock */
1653 spin_lock_init(&sbi->s_rsv_window_lock);
1654 sbi->s_rsv_window_root = RB_ROOT;
1655 /* Add a single, static dummy reservation to the start of the
1656 * reservation window list --- it gives us a placeholder for
1657 * append-at-start-of-list which makes the allocation logic
1658 * _much_ simpler. */
1659 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1660 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1661 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1662 sbi->s_rsv_window_head.rsv_goal_size = 0;
1663 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1666 * set up enough so that it can read an inode
1668 sb->s_op = &ext3_sops;
1669 sb->s_export_op = &ext3_export_ops;
1670 sb->s_xattr = ext3_xattr_handlers;
1672 sb->s_qcop = &ext3_qctl_operations;
1673 sb->dq_op = &ext3_quota_operations;
1675 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1679 needs_recovery = (es->s_last_orphan != 0 ||
1680 EXT3_HAS_INCOMPAT_FEATURE(sb,
1681 EXT3_FEATURE_INCOMPAT_RECOVER));
1684 * The first inode we look at is the journal inode. Don't try
1685 * root first: it may be modified in the journal!
1687 if (!test_opt(sb, NOLOAD) &&
1688 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1689 if (ext3_load_journal(sb, es, journal_devnum))
1691 } else if (journal_inum) {
1692 if (ext3_create_journal(sb, es, journal_inum))
1697 "ext3: No journal on filesystem on %s\n",
1702 /* We have now updated the journal if required, so we can
1703 * validate the data journaling mode. */
1704 switch (test_opt(sb, DATA_FLAGS)) {
1706 /* No mode set, assume a default based on the journal
1707 capabilities: ORDERED_DATA if the journal can
1708 cope, else JOURNAL_DATA */
1709 if (journal_check_available_features
1710 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1711 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1713 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1716 case EXT3_MOUNT_ORDERED_DATA:
1717 case EXT3_MOUNT_WRITEBACK_DATA:
1718 if (!journal_check_available_features
1719 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1720 printk(KERN_ERR "EXT3-fs: Journal does not support "
1721 "requested data journaling mode\n");
1728 if (test_opt(sb, NOBH)) {
1729 if (sb->s_blocksize_bits != PAGE_CACHE_SHIFT) {
1730 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option "
1731 "since filesystem blocksize doesn't match "
1733 clear_opt(sbi->s_mount_opt, NOBH);
1735 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1736 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1737 "its supported only with writeback mode\n");
1738 clear_opt(sbi->s_mount_opt, NOBH);
1742 * The journal_load will have done any necessary log recovery,
1743 * so we can safely mount the rest of the filesystem now.
1746 root = iget(sb, EXT3_ROOT_INO);
1747 sb->s_root = d_alloc_root(root);
1749 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1753 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1756 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1760 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1762 * akpm: core read_super() calls in here with the superblock locked.
1763 * That deadlocks, because orphan cleanup needs to lock the superblock
1764 * in numerous places. Here we just pop the lock - it's relatively
1765 * harmless, because we are now ready to accept write_super() requests,
1766 * and aviro says that's the only reason for hanging onto the
1769 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1770 ext3_orphan_cleanup(sb, es);
1771 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1773 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1774 ext3_mark_recovery_complete(sb, es);
1775 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1776 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1777 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1780 percpu_counter_mod(&sbi->s_freeblocks_counter,
1781 ext3_count_free_blocks(sb));
1782 percpu_counter_mod(&sbi->s_freeinodes_counter,
1783 ext3_count_free_inodes(sb));
1784 percpu_counter_mod(&sbi->s_dirs_counter,
1785 ext3_count_dirs(sb));
1792 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1797 journal_destroy(sbi->s_journal);
1799 for (i = 0; i < db_count; i++)
1800 brelse(sbi->s_group_desc[i]);
1801 kfree(sbi->s_group_desc);
1804 for (i = 0; i < MAXQUOTAS; i++)
1805 kfree(sbi->s_qf_names[i]);
1807 ext3_blkdev_remove(sbi);
1810 sb->s_fs_info = NULL;
1817 * Setup any per-fs journal parameters now. We'll do this both on
1818 * initial mount, once the journal has been initialised but before we've
1819 * done any recovery; and again on any subsequent remount.
1821 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1823 struct ext3_sb_info *sbi = EXT3_SB(sb);
1825 if (sbi->s_commit_interval)
1826 journal->j_commit_interval = sbi->s_commit_interval;
1827 /* We could also set up an ext3-specific default for the commit
1828 * interval here, but for now we'll just fall back to the jbd
1831 spin_lock(&journal->j_state_lock);
1832 if (test_opt(sb, BARRIER))
1833 journal->j_flags |= JFS_BARRIER;
1835 journal->j_flags &= ~JFS_BARRIER;
1836 spin_unlock(&journal->j_state_lock);
1839 static journal_t *ext3_get_journal(struct super_block *sb, int journal_inum)
1841 struct inode *journal_inode;
1844 /* First, test for the existence of a valid inode on disk. Bad
1845 * things happen if we iget() an unused inode, as the subsequent
1846 * iput() will try to delete it. */
1848 journal_inode = iget(sb, journal_inum);
1849 if (!journal_inode) {
1850 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1853 if (!journal_inode->i_nlink) {
1854 make_bad_inode(journal_inode);
1855 iput(journal_inode);
1856 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1860 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1861 journal_inode, journal_inode->i_size);
1862 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1863 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1864 iput(journal_inode);
1868 journal = journal_init_inode(journal_inode);
1870 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1871 iput(journal_inode);
1874 journal->j_private = sb;
1875 ext3_init_journal_params(sb, journal);
1879 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1882 struct buffer_head * bh;
1886 int hblock, blocksize;
1887 unsigned long sb_block;
1888 unsigned long offset;
1889 struct ext3_super_block * es;
1890 struct block_device *bdev;
1892 bdev = ext3_blkdev_get(j_dev);
1896 if (bd_claim(bdev, sb)) {
1898 "EXT3: failed to claim external journal device.\n");
1903 blocksize = sb->s_blocksize;
1904 hblock = bdev_hardsect_size(bdev);
1905 if (blocksize < hblock) {
1907 "EXT3-fs: blocksize too small for journal device.\n");
1911 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1912 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1913 set_blocksize(bdev, blocksize);
1914 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1915 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1916 "external journal\n");
1920 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1921 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1922 !(le32_to_cpu(es->s_feature_incompat) &
1923 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1924 printk(KERN_ERR "EXT3-fs: external journal has "
1925 "bad superblock\n");
1930 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1931 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1936 len = le32_to_cpu(es->s_blocks_count);
1937 start = sb_block + 1;
1938 brelse(bh); /* we're done with the superblock */
1940 journal = journal_init_dev(bdev, sb->s_bdev,
1941 start, len, blocksize);
1943 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1946 journal->j_private = sb;
1947 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1948 wait_on_buffer(journal->j_sb_buffer);
1949 if (!buffer_uptodate(journal->j_sb_buffer)) {
1950 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1953 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1954 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1955 "user (unsupported) - %d\n",
1956 be32_to_cpu(journal->j_superblock->s_nr_users));
1959 EXT3_SB(sb)->journal_bdev = bdev;
1960 ext3_init_journal_params(sb, journal);
1963 journal_destroy(journal);
1965 ext3_blkdev_put(bdev);
1969 static int ext3_load_journal(struct super_block *sb,
1970 struct ext3_super_block *es,
1971 unsigned long journal_devnum)
1974 int journal_inum = le32_to_cpu(es->s_journal_inum);
1977 int really_read_only;
1979 if (journal_devnum &&
1980 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
1981 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
1982 "numbers have changed\n");
1983 journal_dev = new_decode_dev(journal_devnum);
1985 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
1987 really_read_only = bdev_read_only(sb->s_bdev);
1990 * Are we loading a blank journal or performing recovery after a
1991 * crash? For recovery, we need to check in advance whether we
1992 * can get read-write access to the device.
1995 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
1996 if (sb->s_flags & MS_RDONLY) {
1997 printk(KERN_INFO "EXT3-fs: INFO: recovery "
1998 "required on readonly filesystem.\n");
1999 if (really_read_only) {
2000 printk(KERN_ERR "EXT3-fs: write access "
2001 "unavailable, cannot proceed.\n");
2004 printk (KERN_INFO "EXT3-fs: write access will "
2005 "be enabled during recovery.\n");
2009 if (journal_inum && journal_dev) {
2010 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2011 "and inode journals!\n");
2016 if (!(journal = ext3_get_journal(sb, journal_inum)))
2019 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2023 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2024 err = journal_update_format(journal);
2026 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2027 journal_destroy(journal);
2032 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2033 err = journal_wipe(journal, !really_read_only);
2035 err = journal_load(journal);
2038 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2039 journal_destroy(journal);
2043 EXT3_SB(sb)->s_journal = journal;
2044 ext3_clear_journal_err(sb, es);
2046 if (journal_devnum &&
2047 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2048 es->s_journal_dev = cpu_to_le32(journal_devnum);
2051 /* Make sure we flush the recovery flag to disk. */
2052 ext3_commit_super(sb, es, 1);
2058 static int ext3_create_journal(struct super_block * sb,
2059 struct ext3_super_block * es,
2064 if (sb->s_flags & MS_RDONLY) {
2065 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2066 "create journal.\n");
2070 if (!(journal = ext3_get_journal(sb, journal_inum)))
2073 printk(KERN_INFO "EXT3-fs: creating new journal on inode %d\n",
2076 if (journal_create(journal)) {
2077 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2078 journal_destroy(journal);
2082 EXT3_SB(sb)->s_journal = journal;
2084 ext3_update_dynamic_rev(sb);
2085 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2086 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2088 es->s_journal_inum = cpu_to_le32(journal_inum);
2091 /* Make sure we flush the recovery flag to disk. */
2092 ext3_commit_super(sb, es, 1);
2097 static void ext3_commit_super (struct super_block * sb,
2098 struct ext3_super_block * es,
2101 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2105 es->s_wtime = cpu_to_le32(get_seconds());
2106 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2107 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2108 BUFFER_TRACE(sbh, "marking dirty");
2109 mark_buffer_dirty(sbh);
2111 sync_dirty_buffer(sbh);
2116 * Have we just finished recovery? If so, and if we are mounting (or
2117 * remounting) the filesystem readonly, then we will end up with a
2118 * consistent fs on disk. Record that fact.
2120 static void ext3_mark_recovery_complete(struct super_block * sb,
2121 struct ext3_super_block * es)
2123 journal_t *journal = EXT3_SB(sb)->s_journal;
2125 journal_lock_updates(journal);
2126 journal_flush(journal);
2127 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2128 sb->s_flags & MS_RDONLY) {
2129 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2131 ext3_commit_super(sb, es, 1);
2133 journal_unlock_updates(journal);
2137 * If we are mounting (or read-write remounting) a filesystem whose journal
2138 * has recorded an error from a previous lifetime, move that error to the
2139 * main filesystem now.
2141 static void ext3_clear_journal_err(struct super_block * sb,
2142 struct ext3_super_block * es)
2148 journal = EXT3_SB(sb)->s_journal;
2151 * Now check for any error status which may have been recorded in the
2152 * journal by a prior ext3_error() or ext3_abort()
2155 j_errno = journal_errno(journal);
2159 errstr = ext3_decode_error(sb, j_errno, nbuf);
2160 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
2161 "from previous mount: %s", errstr);
2162 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
2163 "filesystem check.");
2165 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2166 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2167 ext3_commit_super (sb, es, 1);
2169 journal_clear_err(journal);
2174 * Force the running and committing transactions to commit,
2175 * and wait on the commit.
2177 int ext3_force_commit(struct super_block *sb)
2182 if (sb->s_flags & MS_RDONLY)
2185 journal = EXT3_SB(sb)->s_journal;
2187 ret = ext3_journal_force_commit(journal);
2192 * Ext3 always journals updates to the superblock itself, so we don't
2193 * have to propagate any other updates to the superblock on disk at this
2194 * point. Just start an async writeback to get the buffers on their way
2197 * This implicitly triggers the writebehind on sync().
2200 static void ext3_write_super (struct super_block * sb)
2202 if (mutex_trylock(&sb->s_lock) != 0)
2207 static int ext3_sync_fs(struct super_block *sb, int wait)
2212 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2214 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2220 * LVM calls this function before a (read-only) snapshot is created. This
2221 * gives us a chance to flush the journal completely and mark the fs clean.
2223 static void ext3_write_super_lockfs(struct super_block *sb)
2227 if (!(sb->s_flags & MS_RDONLY)) {
2228 journal_t *journal = EXT3_SB(sb)->s_journal;
2230 /* Now we set up the journal barrier. */
2231 journal_lock_updates(journal);
2232 journal_flush(journal);
2234 /* Journal blocked and flushed, clear needs_recovery flag. */
2235 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2236 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2241 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2242 * flag here, even though the filesystem is not technically dirty yet.
2244 static void ext3_unlockfs(struct super_block *sb)
2246 if (!(sb->s_flags & MS_RDONLY)) {
2248 /* Reser the needs_recovery flag before the fs is unlocked. */
2249 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2250 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2252 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2256 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2258 struct ext3_super_block * es;
2259 struct ext3_sb_info *sbi = EXT3_SB(sb);
2260 unsigned long n_blocks_count = 0;
2261 unsigned long old_sb_flags;
2262 struct ext3_mount_options old_opts;
2268 /* Store the original options */
2269 old_sb_flags = sb->s_flags;
2270 old_opts.s_mount_opt = sbi->s_mount_opt;
2271 old_opts.s_resuid = sbi->s_resuid;
2272 old_opts.s_resgid = sbi->s_resgid;
2273 old_opts.s_commit_interval = sbi->s_commit_interval;
2275 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2276 for (i = 0; i < MAXQUOTAS; i++)
2277 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2281 * Allow the "check" option to be passed as a remount option.
2283 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2288 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2289 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2290 if ((sbi->s_mount_opt & EXT3_MOUNT_TAGXID) &&
2291 !(sb->s_flags & MS_TAGXID)) {
2292 printk("EXT3-fs: %s: tagxid not permitted on remount.\n",
2297 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2298 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2302 ext3_init_journal_params(sb, sbi->s_journal);
2304 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2305 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2306 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2311 if (*flags & MS_RDONLY) {
2313 * First of all, the unconditional stuff we have to do
2314 * to disable replay of the journal when we next remount
2316 sb->s_flags |= MS_RDONLY;
2319 * OK, test if we are remounting a valid rw partition
2320 * readonly, and if so set the rdonly flag and then
2321 * mark the partition as valid again.
2323 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2324 (sbi->s_mount_state & EXT3_VALID_FS))
2325 es->s_state = cpu_to_le16(sbi->s_mount_state);
2327 ext3_mark_recovery_complete(sb, es);
2330 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2331 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2332 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2333 "remount RDWR because of unsupported "
2334 "optional features (%x).\n",
2335 sb->s_id, le32_to_cpu(ret));
2340 * Mounting a RDONLY partition read-write, so reread
2341 * and store the current valid flag. (It may have
2342 * been changed by e2fsck since we originally mounted
2345 ext3_clear_journal_err(sb, es);
2346 sbi->s_mount_state = le16_to_cpu(es->s_state);
2347 if ((ret = ext3_group_extend(sb, es, n_blocks_count))) {
2351 if (!ext3_setup_super (sb, es, 0))
2352 sb->s_flags &= ~MS_RDONLY;
2356 /* Release old quota file names */
2357 for (i = 0; i < MAXQUOTAS; i++)
2358 if (old_opts.s_qf_names[i] &&
2359 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2360 kfree(old_opts.s_qf_names[i]);
2364 sb->s_flags = old_sb_flags;
2365 sbi->s_mount_opt = old_opts.s_mount_opt;
2366 sbi->s_resuid = old_opts.s_resuid;
2367 sbi->s_resgid = old_opts.s_resgid;
2368 sbi->s_commit_interval = old_opts.s_commit_interval;
2370 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2371 for (i = 0; i < MAXQUOTAS; i++) {
2372 if (sbi->s_qf_names[i] &&
2373 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2374 kfree(sbi->s_qf_names[i]);
2375 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2381 static int ext3_statfs (struct super_block * sb, struct kstatfs * buf)
2383 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
2384 unsigned long overhead;
2387 if (test_opt (sb, MINIX_DF))
2390 unsigned long ngroups;
2391 ngroups = EXT3_SB(sb)->s_groups_count;
2395 * Compute the overhead (FS structures)
2399 * All of the blocks before first_data_block are
2402 overhead = le32_to_cpu(es->s_first_data_block);
2405 * Add the overhead attributed to the superblock and
2406 * block group descriptors. If the sparse superblocks
2407 * feature is turned on, then not all groups have this.
2409 for (i = 0; i < ngroups; i++) {
2410 overhead += ext3_bg_has_super(sb, i) +
2411 ext3_bg_num_gdb(sb, i);
2416 * Every block group has an inode bitmap, a block
2417 * bitmap, and an inode table.
2419 overhead += (ngroups * (2 + EXT3_SB(sb)->s_itb_per_group));
2422 buf->f_type = EXT3_SUPER_MAGIC;
2423 buf->f_bsize = sb->s_blocksize;
2424 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
2425 buf->f_bfree = ext3_count_free_blocks (sb);
2426 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2427 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2429 buf->f_files = le32_to_cpu(es->s_inodes_count);
2430 buf->f_ffree = ext3_count_free_inodes (sb);
2431 buf->f_namelen = EXT3_NAME_LEN;
2435 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2436 * is locked for write. Otherwise the are possible deadlocks:
2437 * Process 1 Process 2
2438 * ext3_create() quota_sync()
2439 * journal_start() write_dquot()
2440 * DQUOT_INIT() down(dqio_sem)
2441 * down(dqio_sem) journal_start()
2447 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2449 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2452 static int ext3_dquot_initialize(struct inode *inode, int type)
2457 /* We may create quota structure so we need to reserve enough blocks */
2458 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2460 return PTR_ERR(handle);
2461 ret = dquot_initialize(inode, type);
2462 err = ext3_journal_stop(handle);
2468 static int ext3_dquot_drop(struct inode *inode)
2473 /* We may delete quota structure so we need to reserve enough blocks */
2474 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2476 return PTR_ERR(handle);
2477 ret = dquot_drop(inode);
2478 err = ext3_journal_stop(handle);
2484 static int ext3_write_dquot(struct dquot *dquot)
2488 struct inode *inode;
2490 inode = dquot_to_inode(dquot);
2491 handle = ext3_journal_start(inode,
2492 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2494 return PTR_ERR(handle);
2495 ret = dquot_commit(dquot);
2496 err = ext3_journal_stop(handle);
2502 static int ext3_acquire_dquot(struct dquot *dquot)
2507 handle = ext3_journal_start(dquot_to_inode(dquot),
2508 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2510 return PTR_ERR(handle);
2511 ret = dquot_acquire(dquot);
2512 err = ext3_journal_stop(handle);
2518 static int ext3_release_dquot(struct dquot *dquot)
2523 handle = ext3_journal_start(dquot_to_inode(dquot),
2524 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2526 return PTR_ERR(handle);
2527 ret = dquot_release(dquot);
2528 err = ext3_journal_stop(handle);
2534 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2536 /* Are we journalling quotas? */
2537 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2538 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2539 dquot_mark_dquot_dirty(dquot);
2540 return ext3_write_dquot(dquot);
2542 return dquot_mark_dquot_dirty(dquot);
2546 static int ext3_write_info(struct super_block *sb, int type)
2551 /* Data block + inode block */
2552 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2554 return PTR_ERR(handle);
2555 ret = dquot_commit_info(sb, type);
2556 err = ext3_journal_stop(handle);
2563 * Turn on quotas during mount time - we need to find
2564 * the quota file and such...
2566 static int ext3_quota_on_mount(struct super_block *sb, int type)
2568 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2569 EXT3_SB(sb)->s_jquota_fmt, type);
2573 * Standard function to be called on quota_on
2575 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2579 struct nameidata nd;
2581 if (!test_opt(sb, QUOTA))
2583 /* Not journalling quota? */
2584 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2585 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2586 return vfs_quota_on(sb, type, format_id, path);
2587 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2590 /* Quotafile not on the same filesystem? */
2591 if (nd.mnt->mnt_sb != sb) {
2595 /* Quotafile not of fs root? */
2596 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2598 "EXT3-fs: Quota file not on filesystem root. "
2599 "Journalled quota will not work.\n");
2601 return vfs_quota_on(sb, type, format_id, path);
2604 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2605 * acquiring the locks... As quota files are never truncated and quota code
2606 * itself serializes the operations (and noone else should touch the files)
2607 * we don't have to be afraid of races */
2608 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2609 size_t len, loff_t off)
2611 struct inode *inode = sb_dqopt(sb)->files[type];
2612 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2614 int offset = off & (sb->s_blocksize - 1);
2617 struct buffer_head *bh;
2618 loff_t i_size = i_size_read(inode);
2622 if (off+len > i_size)
2625 while (toread > 0) {
2626 tocopy = sb->s_blocksize - offset < toread ?
2627 sb->s_blocksize - offset : toread;
2628 bh = ext3_bread(NULL, inode, blk, 0, &err);
2631 if (!bh) /* A hole? */
2632 memset(data, 0, tocopy);
2634 memcpy(data, bh->b_data+offset, tocopy);
2644 /* Write to quotafile (we know the transaction is already started and has
2645 * enough credits) */
2646 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2647 const char *data, size_t len, loff_t off)
2649 struct inode *inode = sb_dqopt(sb)->files[type];
2650 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2652 int offset = off & (sb->s_blocksize - 1);
2654 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2655 size_t towrite = len;
2656 struct buffer_head *bh;
2657 handle_t *handle = journal_current_handle();
2659 mutex_lock(&inode->i_mutex);
2660 while (towrite > 0) {
2661 tocopy = sb->s_blocksize - offset < towrite ?
2662 sb->s_blocksize - offset : towrite;
2663 bh = ext3_bread(handle, inode, blk, 1, &err);
2666 if (journal_quota) {
2667 err = ext3_journal_get_write_access(handle, bh);
2674 memcpy(bh->b_data+offset, data, tocopy);
2675 flush_dcache_page(bh->b_page);
2678 err = ext3_journal_dirty_metadata(handle, bh);
2680 /* Always do at least ordered writes for quotas */
2681 err = ext3_journal_dirty_data(handle, bh);
2682 mark_buffer_dirty(bh);
2695 if (inode->i_size < off+len-towrite) {
2696 i_size_write(inode, off+len-towrite);
2697 EXT3_I(inode)->i_disksize = inode->i_size;
2700 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2701 ext3_mark_inode_dirty(handle, inode);
2702 mutex_unlock(&inode->i_mutex);
2703 return len - towrite;
2708 static struct super_block *ext3_get_sb(struct file_system_type *fs_type,
2709 int flags, const char *dev_name, void *data)
2711 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
2714 static struct file_system_type ext3_fs_type = {
2715 .owner = THIS_MODULE,
2717 .get_sb = ext3_get_sb,
2718 .kill_sb = kill_block_super,
2719 .fs_flags = FS_REQUIRES_DEV,
2722 static int __init init_ext3_fs(void)
2724 int err = init_ext3_xattr();
2727 err = init_inodecache();
2730 err = register_filesystem(&ext3_fs_type);
2735 destroy_inodecache();
2741 static void __exit exit_ext3_fs(void)
2743 unregister_filesystem(&ext3_fs_type);
2744 destroy_inodecache();
2748 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2749 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2750 MODULE_LICENSE("GPL");
2751 module_init(init_ext3_fs)
2752 module_exit(exit_ext3_fs)