2 * linux/fs/ext4/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/jbd2.h>
24 #include <linux/ext4_fs.h>
25 #include <linux/ext4_jbd2.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
39 #include <asm/uaccess.h>
45 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
46 unsigned long journal_devnum);
47 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
49 static void ext4_commit_super (struct super_block * sb,
50 struct ext4_super_block * es,
52 static void ext4_mark_recovery_complete(struct super_block * sb,
53 struct ext4_super_block * es);
54 static void ext4_clear_journal_err(struct super_block * sb,
55 struct ext4_super_block * es);
56 static int ext4_sync_fs(struct super_block *sb, int wait);
57 static const char *ext4_decode_error(struct super_block * sb, int errno,
59 static int ext4_remount (struct super_block * sb, int * flags, char * data);
60 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
61 static void ext4_unlockfs(struct super_block *sb);
62 static void ext4_write_super (struct super_block * sb);
63 static void ext4_write_super_lockfs(struct super_block *sb);
66 * Wrappers for jbd2_journal_start/end.
68 * The only special thing we need to do here is to make sure that all
69 * journal_end calls result in the superblock being marked dirty, so
70 * that sync() will call the filesystem's write_super callback if
73 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
77 if (sb->s_flags & MS_RDONLY)
78 return ERR_PTR(-EROFS);
80 /* Special case here: if the journal has aborted behind our
81 * backs (eg. EIO in the commit thread), then we still need to
82 * take the FS itself readonly cleanly. */
83 journal = EXT4_SB(sb)->s_journal;
84 if (is_journal_aborted(journal)) {
85 ext4_abort(sb, __FUNCTION__,
86 "Detected aborted journal");
87 return ERR_PTR(-EROFS);
90 return jbd2_journal_start(journal, nblocks);
94 * The only special thing we need to do here is to make sure that all
95 * jbd2_journal_stop calls result in the superblock being marked dirty, so
96 * that sync() will call the filesystem's write_super callback if
99 int __ext4_journal_stop(const char *where, handle_t *handle)
101 struct super_block *sb;
105 sb = handle->h_transaction->t_journal->j_private;
107 rc = jbd2_journal_stop(handle);
112 __ext4_std_error(sb, where, err);
116 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
117 struct buffer_head *bh, handle_t *handle, int err)
120 const char *errstr = ext4_decode_error(NULL, err, nbuf);
123 BUFFER_TRACE(bh, "abort");
128 if (is_handle_aborted(handle))
131 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
132 caller, errstr, err_fn);
134 jbd2_journal_abort_handle(handle);
137 /* Deal with the reporting of failure conditions on a filesystem such as
138 * inconsistencies detected or read IO failures.
140 * On ext2, we can store the error state of the filesystem in the
141 * superblock. That is not possible on ext4, because we may have other
142 * write ordering constraints on the superblock which prevent us from
143 * writing it out straight away; and given that the journal is about to
144 * be aborted, we can't rely on the current, or future, transactions to
145 * write out the superblock safely.
147 * We'll just use the jbd2_journal_abort() error code to record an error in
148 * the journal instead. On recovery, the journal will compain about
149 * that error until we've noted it down and cleared it.
152 static void ext4_handle_error(struct super_block *sb)
154 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
156 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
157 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
159 if (sb->s_flags & MS_RDONLY)
162 if (!test_opt (sb, ERRORS_CONT)) {
163 journal_t *journal = EXT4_SB(sb)->s_journal;
165 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
167 jbd2_journal_abort(journal, -EIO);
169 if (test_opt (sb, ERRORS_RO)) {
170 printk (KERN_CRIT "Remounting filesystem read-only\n");
171 sb->s_flags |= MS_RDONLY;
173 ext4_commit_super(sb, es, 1);
174 if (test_opt(sb, ERRORS_PANIC))
175 panic("EXT4-fs (device %s): panic forced after error\n",
179 void ext4_error (struct super_block * sb, const char * function,
180 const char * fmt, ...)
185 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
190 ext4_handle_error(sb);
193 static const char *ext4_decode_error(struct super_block * sb, int errno,
200 errstr = "IO failure";
203 errstr = "Out of memory";
206 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_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 /* __ext4_std_error decodes expected errors from journaling functions
227 * automatically and invokes the appropriate error response. */
229 void __ext4_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 = ext4_decode_error(sb, errno, nbuf);
243 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
244 sb->s_id, function, errstr);
246 ext4_handle_error(sb);
250 * ext4_abort is a much stronger failure handler than ext4_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 ext4_abort (struct super_block * sb, const char * function,
260 const char * fmt, ...)
264 printk (KERN_CRIT "ext4_abort called.\n");
267 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
272 if (test_opt(sb, ERRORS_PANIC))
273 panic("EXT4-fs panic from previous error\n");
275 if (sb->s_flags & MS_RDONLY)
278 printk(KERN_CRIT "Remounting filesystem read-only\n");
279 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
280 sb->s_flags |= MS_RDONLY;
281 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
282 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
285 void ext4_warning (struct super_block * sb, const char * function,
286 const char * fmt, ...)
291 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
298 void ext4_update_dynamic_rev(struct super_block *sb)
300 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
302 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
305 ext4_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(EXT4_GOOD_OLD_FIRST_INO);
311 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
312 es->s_rev_level = cpu_to_le32(EXT4_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 *ext4_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 "EXT4: failed to open journal device %s: %ld\n",
338 __bdevname(dev, b), PTR_ERR(bdev));
343 * Release the journal device
345 static int ext4_blkdev_put(struct block_device *bdev)
348 return blkdev_put(bdev);
351 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
353 struct block_device *bdev;
356 bdev = sbi->journal_bdev;
358 ret = ext4_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 ext4_inode_info, i_orphan)->vfs_inode;
369 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
373 printk(KERN_ERR "sb orphan head is %d\n",
374 le32_to_cpu(sbi->s_es->s_last_orphan));
376 printk(KERN_ERR "sb_info orphan list:\n");
377 list_for_each(l, &sbi->s_orphan) {
378 struct inode *inode = orphan_list_entry(l);
380 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
381 inode->i_sb->s_id, inode->i_ino, inode,
382 inode->i_mode, inode->i_nlink,
387 static void ext4_put_super (struct super_block * sb)
389 struct ext4_sb_info *sbi = EXT4_SB(sb);
390 struct ext4_super_block *es = sbi->s_es;
393 ext4_ext_release(sb);
394 ext4_xattr_put_super(sb);
395 jbd2_journal_destroy(sbi->s_journal);
396 if (!(sb->s_flags & MS_RDONLY)) {
397 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
398 es->s_state = cpu_to_le16(sbi->s_mount_state);
399 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
400 mark_buffer_dirty(sbi->s_sbh);
401 ext4_commit_super(sb, es, 1);
404 for (i = 0; i < sbi->s_gdb_count; i++)
405 brelse(sbi->s_group_desc[i]);
406 kfree(sbi->s_group_desc);
407 percpu_counter_destroy(&sbi->s_freeblocks_counter);
408 percpu_counter_destroy(&sbi->s_freeinodes_counter);
409 percpu_counter_destroy(&sbi->s_dirs_counter);
412 for (i = 0; i < MAXQUOTAS; i++)
413 kfree(sbi->s_qf_names[i]);
416 /* Debugging code just in case the in-memory inode orphan list
417 * isn't empty. The on-disk one can be non-empty if we've
418 * detected an error and taken the fs readonly, but the
419 * in-memory list had better be clean by this point. */
420 if (!list_empty(&sbi->s_orphan))
421 dump_orphan_list(sb, sbi);
422 J_ASSERT(list_empty(&sbi->s_orphan));
424 invalidate_bdev(sb->s_bdev, 0);
425 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
427 * Invalidate the journal device's buffers. We don't want them
428 * floating about in memory - the physical journal device may
429 * hotswapped, and it breaks the `ro-after' testing code.
431 sync_blockdev(sbi->journal_bdev);
432 invalidate_bdev(sbi->journal_bdev, 0);
433 ext4_blkdev_remove(sbi);
435 sb->s_fs_info = NULL;
440 static kmem_cache_t *ext4_inode_cachep;
443 * Called inside transaction, so use GFP_NOFS
445 static struct inode *ext4_alloc_inode(struct super_block *sb)
447 struct ext4_inode_info *ei;
449 ei = kmem_cache_alloc(ext4_inode_cachep, SLAB_NOFS);
452 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
453 ei->i_acl = EXT4_ACL_NOT_CACHED;
454 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
456 ei->i_block_alloc_info = NULL;
457 ei->vfs_inode.i_version = 1;
458 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
459 return &ei->vfs_inode;
462 static void ext4_destroy_inode(struct inode *inode)
464 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
467 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
469 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
471 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
472 SLAB_CTOR_CONSTRUCTOR) {
473 INIT_LIST_HEAD(&ei->i_orphan);
474 #ifdef CONFIG_EXT4DEV_FS_XATTR
475 init_rwsem(&ei->xattr_sem);
477 mutex_init(&ei->truncate_mutex);
478 inode_init_once(&ei->vfs_inode);
482 static int init_inodecache(void)
484 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
485 sizeof(struct ext4_inode_info),
486 0, (SLAB_RECLAIM_ACCOUNT|
489 if (ext4_inode_cachep == NULL)
494 static void destroy_inodecache(void)
496 kmem_cache_destroy(ext4_inode_cachep);
499 static void ext4_clear_inode(struct inode *inode)
501 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
502 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
503 if (EXT4_I(inode)->i_acl &&
504 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
505 posix_acl_release(EXT4_I(inode)->i_acl);
506 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
508 if (EXT4_I(inode)->i_default_acl &&
509 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
510 posix_acl_release(EXT4_I(inode)->i_default_acl);
511 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
514 ext4_discard_reservation(inode);
515 EXT4_I(inode)->i_block_alloc_info = NULL;
520 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
522 #if defined(CONFIG_QUOTA)
523 struct ext4_sb_info *sbi = EXT4_SB(sb);
525 if (sbi->s_jquota_fmt)
526 seq_printf(seq, ",jqfmt=%s",
527 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
529 if (sbi->s_qf_names[USRQUOTA])
530 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
532 if (sbi->s_qf_names[GRPQUOTA])
533 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
535 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
536 seq_puts(seq, ",usrquota");
538 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
539 seq_puts(seq, ",grpquota");
543 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
545 struct super_block *sb = vfs->mnt_sb;
547 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
548 seq_puts(seq, ",data=journal");
549 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
550 seq_puts(seq, ",data=ordered");
551 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
552 seq_puts(seq, ",data=writeback");
554 ext4_show_quota_options(seq, sb);
560 static struct dentry *ext4_get_dentry(struct super_block *sb, void *vobjp)
563 unsigned long ino = objp[0];
564 __u32 generation = objp[1];
566 struct dentry *result;
568 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
569 return ERR_PTR(-ESTALE);
570 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
571 return ERR_PTR(-ESTALE);
573 /* iget isn't really right if the inode is currently unallocated!!
575 * ext4_read_inode will return a bad_inode if the inode had been
576 * deleted, so we should be safe.
578 * Currently we don't know the generation for parent directory, so
579 * a generation of 0 means "accept any"
581 inode = iget(sb, ino);
583 return ERR_PTR(-ENOMEM);
584 if (is_bad_inode(inode) ||
585 (generation && inode->i_generation != generation)) {
587 return ERR_PTR(-ESTALE);
589 /* now to find a dentry.
590 * If possible, get a well-connected one
592 result = d_alloc_anon(inode);
595 return ERR_PTR(-ENOMEM);
601 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
602 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
604 static int ext4_dquot_initialize(struct inode *inode, int type);
605 static int ext4_dquot_drop(struct inode *inode);
606 static int ext4_write_dquot(struct dquot *dquot);
607 static int ext4_acquire_dquot(struct dquot *dquot);
608 static int ext4_release_dquot(struct dquot *dquot);
609 static int ext4_mark_dquot_dirty(struct dquot *dquot);
610 static int ext4_write_info(struct super_block *sb, int type);
611 static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
612 static int ext4_quota_on_mount(struct super_block *sb, int type);
613 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
614 size_t len, loff_t off);
615 static ssize_t ext4_quota_write(struct super_block *sb, int type,
616 const char *data, size_t len, loff_t off);
618 static struct dquot_operations ext4_quota_operations = {
619 .initialize = ext4_dquot_initialize,
620 .drop = ext4_dquot_drop,
621 .alloc_space = dquot_alloc_space,
622 .alloc_inode = dquot_alloc_inode,
623 .free_space = dquot_free_space,
624 .free_inode = dquot_free_inode,
625 .transfer = dquot_transfer,
626 .write_dquot = ext4_write_dquot,
627 .acquire_dquot = ext4_acquire_dquot,
628 .release_dquot = ext4_release_dquot,
629 .mark_dirty = ext4_mark_dquot_dirty,
630 .write_info = ext4_write_info
633 static struct quotactl_ops ext4_qctl_operations = {
634 .quota_on = ext4_quota_on,
635 .quota_off = vfs_quota_off,
636 .quota_sync = vfs_quota_sync,
637 .get_info = vfs_get_dqinfo,
638 .set_info = vfs_set_dqinfo,
639 .get_dqblk = vfs_get_dqblk,
640 .set_dqblk = vfs_set_dqblk
644 static struct super_operations ext4_sops = {
645 .alloc_inode = ext4_alloc_inode,
646 .destroy_inode = ext4_destroy_inode,
647 .read_inode = ext4_read_inode,
648 .write_inode = ext4_write_inode,
649 .dirty_inode = ext4_dirty_inode,
650 .delete_inode = ext4_delete_inode,
651 .put_super = ext4_put_super,
652 .write_super = ext4_write_super,
653 .sync_fs = ext4_sync_fs,
654 .write_super_lockfs = ext4_write_super_lockfs,
655 .unlockfs = ext4_unlockfs,
656 .statfs = ext4_statfs,
657 .remount_fs = ext4_remount,
658 .clear_inode = ext4_clear_inode,
659 .show_options = ext4_show_options,
661 .quota_read = ext4_quota_read,
662 .quota_write = ext4_quota_write,
666 static struct export_operations ext4_export_ops = {
667 .get_parent = ext4_get_parent,
668 .get_dentry = ext4_get_dentry,
672 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
673 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
674 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
675 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
676 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
677 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
678 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
679 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
680 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
681 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
682 Opt_grpquota, Opt_extents,
685 static match_table_t tokens = {
686 {Opt_bsd_df, "bsddf"},
687 {Opt_minix_df, "minixdf"},
688 {Opt_grpid, "grpid"},
689 {Opt_grpid, "bsdgroups"},
690 {Opt_nogrpid, "nogrpid"},
691 {Opt_nogrpid, "sysvgroups"},
692 {Opt_resgid, "resgid=%u"},
693 {Opt_resuid, "resuid=%u"},
695 {Opt_err_cont, "errors=continue"},
696 {Opt_err_panic, "errors=panic"},
697 {Opt_err_ro, "errors=remount-ro"},
698 {Opt_nouid32, "nouid32"},
699 {Opt_nocheck, "nocheck"},
700 {Opt_nocheck, "check=none"},
701 {Opt_debug, "debug"},
702 {Opt_oldalloc, "oldalloc"},
703 {Opt_orlov, "orlov"},
704 {Opt_user_xattr, "user_xattr"},
705 {Opt_nouser_xattr, "nouser_xattr"},
707 {Opt_noacl, "noacl"},
708 {Opt_reservation, "reservation"},
709 {Opt_noreservation, "noreservation"},
710 {Opt_noload, "noload"},
713 {Opt_commit, "commit=%u"},
714 {Opt_journal_update, "journal=update"},
715 {Opt_journal_inum, "journal=%u"},
716 {Opt_journal_dev, "journal_dev=%u"},
717 {Opt_abort, "abort"},
718 {Opt_data_journal, "data=journal"},
719 {Opt_data_ordered, "data=ordered"},
720 {Opt_data_writeback, "data=writeback"},
721 {Opt_offusrjquota, "usrjquota="},
722 {Opt_usrjquota, "usrjquota=%s"},
723 {Opt_offgrpjquota, "grpjquota="},
724 {Opt_grpjquota, "grpjquota=%s"},
725 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
726 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
727 {Opt_grpquota, "grpquota"},
728 {Opt_noquota, "noquota"},
729 {Opt_quota, "quota"},
730 {Opt_usrquota, "usrquota"},
731 {Opt_barrier, "barrier=%u"},
732 {Opt_extents, "extents"},
734 {Opt_resize, "resize"},
737 static ext4_fsblk_t get_sb_block(void **data)
739 ext4_fsblk_t sb_block;
740 char *options = (char *) *data;
742 if (!options || strncmp(options, "sb=", 3) != 0)
743 return 1; /* Default location */
745 /*todo: use simple_strtoll with >32bit ext4 */
746 sb_block = simple_strtoul(options, &options, 0);
747 if (*options && *options != ',') {
748 printk("EXT4-fs: Invalid sb specification: %s\n",
754 *data = (void *) options;
758 static int parse_options (char *options, struct super_block *sb,
759 unsigned int *inum, unsigned long *journal_devnum,
760 ext4_fsblk_t *n_blocks_count, int is_remount)
762 struct ext4_sb_info *sbi = EXT4_SB(sb);
764 substring_t args[MAX_OPT_ARGS];
775 while ((p = strsep (&options, ",")) != NULL) {
780 token = match_token(p, tokens, args);
783 clear_opt (sbi->s_mount_opt, MINIX_DF);
786 set_opt (sbi->s_mount_opt, MINIX_DF);
789 set_opt (sbi->s_mount_opt, GRPID);
792 clear_opt (sbi->s_mount_opt, GRPID);
795 if (match_int(&args[0], &option))
797 sbi->s_resuid = option;
800 if (match_int(&args[0], &option))
802 sbi->s_resgid = option;
805 /* handled by get_sb_block() instead of here */
806 /* *sb_block = match_int(&args[0]); */
809 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
810 clear_opt (sbi->s_mount_opt, ERRORS_RO);
811 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
814 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
815 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
816 set_opt (sbi->s_mount_opt, ERRORS_RO);
819 clear_opt (sbi->s_mount_opt, ERRORS_RO);
820 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
821 set_opt (sbi->s_mount_opt, ERRORS_CONT);
824 set_opt (sbi->s_mount_opt, NO_UID32);
827 clear_opt (sbi->s_mount_opt, CHECK);
830 set_opt (sbi->s_mount_opt, DEBUG);
833 set_opt (sbi->s_mount_opt, OLDALLOC);
836 clear_opt (sbi->s_mount_opt, OLDALLOC);
838 #ifdef CONFIG_EXT4DEV_FS_XATTR
840 set_opt (sbi->s_mount_opt, XATTR_USER);
842 case Opt_nouser_xattr:
843 clear_opt (sbi->s_mount_opt, XATTR_USER);
847 case Opt_nouser_xattr:
848 printk("EXT4 (no)user_xattr options not supported\n");
851 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
853 set_opt(sbi->s_mount_opt, POSIX_ACL);
856 clear_opt(sbi->s_mount_opt, POSIX_ACL);
861 printk("EXT4 (no)acl options not supported\n");
864 case Opt_reservation:
865 set_opt(sbi->s_mount_opt, RESERVATION);
867 case Opt_noreservation:
868 clear_opt(sbi->s_mount_opt, RESERVATION);
870 case Opt_journal_update:
872 /* Eventually we will want to be able to create
873 a journal file here. For now, only allow the
874 user to specify an existing inode to be the
877 printk(KERN_ERR "EXT4-fs: cannot specify "
878 "journal on remount\n");
881 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
883 case Opt_journal_inum:
885 printk(KERN_ERR "EXT4-fs: cannot specify "
886 "journal on remount\n");
889 if (match_int(&args[0], &option))
893 case Opt_journal_dev:
895 printk(KERN_ERR "EXT4-fs: cannot specify "
896 "journal on remount\n");
899 if (match_int(&args[0], &option))
901 *journal_devnum = option;
904 set_opt (sbi->s_mount_opt, NOLOAD);
907 if (match_int(&args[0], &option))
912 option = JBD_DEFAULT_MAX_COMMIT_AGE;
913 sbi->s_commit_interval = HZ * option;
915 case Opt_data_journal:
916 data_opt = EXT4_MOUNT_JOURNAL_DATA;
918 case Opt_data_ordered:
919 data_opt = EXT4_MOUNT_ORDERED_DATA;
921 case Opt_data_writeback:
922 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
925 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
928 "EXT4-fs: cannot change data "
929 "mode on remount\n");
933 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
934 sbi->s_mount_opt |= data_opt;
944 if (sb_any_quota_enabled(sb)) {
946 "EXT4-fs: Cannot change journalled "
947 "quota options when quota turned on.\n");
950 qname = match_strdup(&args[0]);
953 "EXT4-fs: not enough memory for "
954 "storing quotafile name.\n");
957 if (sbi->s_qf_names[qtype] &&
958 strcmp(sbi->s_qf_names[qtype], qname)) {
960 "EXT4-fs: %s quota file already "
961 "specified.\n", QTYPE2NAME(qtype));
965 sbi->s_qf_names[qtype] = qname;
966 if (strchr(sbi->s_qf_names[qtype], '/')) {
968 "EXT4-fs: quotafile must be on "
969 "filesystem root.\n");
970 kfree(sbi->s_qf_names[qtype]);
971 sbi->s_qf_names[qtype] = NULL;
974 set_opt(sbi->s_mount_opt, QUOTA);
976 case Opt_offusrjquota:
979 case Opt_offgrpjquota:
982 if (sb_any_quota_enabled(sb)) {
983 printk(KERN_ERR "EXT4-fs: Cannot change "
984 "journalled quota options when "
985 "quota turned on.\n");
989 * The space will be released later when all options
990 * are confirmed to be correct
992 sbi->s_qf_names[qtype] = NULL;
994 case Opt_jqfmt_vfsold:
995 sbi->s_jquota_fmt = QFMT_VFS_OLD;
997 case Opt_jqfmt_vfsv0:
998 sbi->s_jquota_fmt = QFMT_VFS_V0;
1002 set_opt(sbi->s_mount_opt, QUOTA);
1003 set_opt(sbi->s_mount_opt, USRQUOTA);
1006 set_opt(sbi->s_mount_opt, QUOTA);
1007 set_opt(sbi->s_mount_opt, GRPQUOTA);
1010 if (sb_any_quota_enabled(sb)) {
1011 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1012 "options when quota turned on.\n");
1015 clear_opt(sbi->s_mount_opt, QUOTA);
1016 clear_opt(sbi->s_mount_opt, USRQUOTA);
1017 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1025 case Opt_offusrjquota:
1026 case Opt_offgrpjquota:
1027 case Opt_jqfmt_vfsold:
1028 case Opt_jqfmt_vfsv0:
1030 "EXT4-fs: journalled quota options not "
1037 set_opt(sbi->s_mount_opt, ABORT);
1040 if (match_int(&args[0], &option))
1043 set_opt(sbi->s_mount_opt, BARRIER);
1045 clear_opt(sbi->s_mount_opt, BARRIER);
1051 printk("EXT4-fs: resize option only available "
1055 if (match_int(&args[0], &option) != 0)
1057 *n_blocks_count = option;
1060 set_opt(sbi->s_mount_opt, NOBH);
1063 clear_opt(sbi->s_mount_opt, NOBH);
1066 set_opt (sbi->s_mount_opt, EXTENTS);
1070 "EXT4-fs: Unrecognized mount option \"%s\" "
1071 "or missing value\n", p);
1076 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1077 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1078 sbi->s_qf_names[USRQUOTA])
1079 clear_opt(sbi->s_mount_opt, USRQUOTA);
1081 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1082 sbi->s_qf_names[GRPQUOTA])
1083 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1085 if ((sbi->s_qf_names[USRQUOTA] &&
1086 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1087 (sbi->s_qf_names[GRPQUOTA] &&
1088 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1089 printk(KERN_ERR "EXT4-fs: old and new quota "
1090 "format mixing.\n");
1094 if (!sbi->s_jquota_fmt) {
1095 printk(KERN_ERR "EXT4-fs: journalled quota format "
1096 "not specified.\n");
1100 if (sbi->s_jquota_fmt) {
1101 printk(KERN_ERR "EXT4-fs: journalled quota format "
1102 "specified with no journalling "
1111 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1114 struct ext4_sb_info *sbi = EXT4_SB(sb);
1117 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1118 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1119 "forcing read-only mode\n");
1124 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1125 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1126 "running e2fsck is recommended\n");
1127 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1128 printk (KERN_WARNING
1129 "EXT4-fs warning: mounting fs with errors, "
1130 "running e2fsck is recommended\n");
1131 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1132 le16_to_cpu(es->s_mnt_count) >=
1133 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1134 printk (KERN_WARNING
1135 "EXT4-fs warning: maximal mount count reached, "
1136 "running e2fsck is recommended\n");
1137 else if (le32_to_cpu(es->s_checkinterval) &&
1138 (le32_to_cpu(es->s_lastcheck) +
1139 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1140 printk (KERN_WARNING
1141 "EXT4-fs warning: checktime reached, "
1142 "running e2fsck is recommended\n");
1144 /* @@@ We _will_ want to clear the valid bit if we find
1145 inconsistencies, to force a fsck at reboot. But for
1146 a plain journaled filesystem we can keep it set as
1147 valid forever! :) */
1148 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1150 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1151 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1152 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1153 es->s_mtime = cpu_to_le32(get_seconds());
1154 ext4_update_dynamic_rev(sb);
1155 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1157 ext4_commit_super(sb, es, 1);
1158 if (test_opt(sb, DEBUG))
1159 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1160 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1162 sbi->s_groups_count,
1163 EXT4_BLOCKS_PER_GROUP(sb),
1164 EXT4_INODES_PER_GROUP(sb),
1167 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1168 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1169 char b[BDEVNAME_SIZE];
1171 printk("external journal on %s\n",
1172 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1174 printk("internal journal\n");
1179 /* Called at mount-time, super-block is locked */
1180 static int ext4_check_descriptors (struct super_block * sb)
1182 struct ext4_sb_info *sbi = EXT4_SB(sb);
1183 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1184 ext4_fsblk_t last_block;
1185 struct ext4_group_desc * gdp = NULL;
1189 ext4_debug ("Checking group descriptors");
1191 for (i = 0; i < sbi->s_groups_count; i++)
1193 if (i == sbi->s_groups_count - 1)
1194 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1196 last_block = first_block +
1197 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1199 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1200 gdp = (struct ext4_group_desc *)
1201 sbi->s_group_desc[desc_block++]->b_data;
1202 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1203 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1205 ext4_error (sb, "ext4_check_descriptors",
1206 "Block bitmap for group %d"
1207 " not in group (block %lu)!",
1209 le32_to_cpu(gdp->bg_block_bitmap));
1212 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1213 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1215 ext4_error (sb, "ext4_check_descriptors",
1216 "Inode bitmap for group %d"
1217 " not in group (block %lu)!",
1219 le32_to_cpu(gdp->bg_inode_bitmap));
1222 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1223 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >
1226 ext4_error (sb, "ext4_check_descriptors",
1227 "Inode table for group %d"
1228 " not in group (block %lu)!",
1230 le32_to_cpu(gdp->bg_inode_table));
1233 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1237 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext4_count_free_blocks(sb));
1238 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1243 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1244 * the superblock) which were deleted from all directories, but held open by
1245 * a process at the time of a crash. We walk the list and try to delete these
1246 * inodes at recovery time (only with a read-write filesystem).
1248 * In order to keep the orphan inode chain consistent during traversal (in
1249 * case of crash during recovery), we link each inode into the superblock
1250 * orphan list_head and handle it the same way as an inode deletion during
1251 * normal operation (which journals the operations for us).
1253 * We only do an iget() and an iput() on each inode, which is very safe if we
1254 * accidentally point at an in-use or already deleted inode. The worst that
1255 * can happen in this case is that we get a "bit already cleared" message from
1256 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1257 * e2fsck was run on this filesystem, and it must have already done the orphan
1258 * inode cleanup for us, so we can safely abort without any further action.
1260 static void ext4_orphan_cleanup (struct super_block * sb,
1261 struct ext4_super_block * es)
1263 unsigned int s_flags = sb->s_flags;
1264 int nr_orphans = 0, nr_truncates = 0;
1268 if (!es->s_last_orphan) {
1269 jbd_debug(4, "no orphan inodes to clean up\n");
1273 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1274 if (es->s_last_orphan)
1275 jbd_debug(1, "Errors on filesystem, "
1276 "clearing orphan list.\n");
1277 es->s_last_orphan = 0;
1278 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1282 if (s_flags & MS_RDONLY) {
1283 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1285 sb->s_flags &= ~MS_RDONLY;
1288 /* Needed for iput() to work correctly and not trash data */
1289 sb->s_flags |= MS_ACTIVE;
1290 /* Turn on quotas so that they are updated correctly */
1291 for (i = 0; i < MAXQUOTAS; i++) {
1292 if (EXT4_SB(sb)->s_qf_names[i]) {
1293 int ret = ext4_quota_on_mount(sb, i);
1296 "EXT4-fs: Cannot turn on journalled "
1297 "quota: error %d\n", ret);
1302 while (es->s_last_orphan) {
1303 struct inode *inode;
1306 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1307 es->s_last_orphan = 0;
1311 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1313 if (inode->i_nlink) {
1315 "%s: truncating inode %lu to %Ld bytes\n",
1316 __FUNCTION__, inode->i_ino, inode->i_size);
1317 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1318 inode->i_ino, inode->i_size);
1319 ext4_truncate(inode);
1323 "%s: deleting unreferenced inode %lu\n",
1324 __FUNCTION__, inode->i_ino);
1325 jbd_debug(2, "deleting unreferenced inode %lu\n",
1329 iput(inode); /* The delete magic happens here! */
1332 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1335 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1336 sb->s_id, PLURAL(nr_orphans));
1338 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1339 sb->s_id, PLURAL(nr_truncates));
1341 /* Turn quotas off */
1342 for (i = 0; i < MAXQUOTAS; i++) {
1343 if (sb_dqopt(sb)->files[i])
1344 vfs_quota_off(sb, i);
1347 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1350 #define log2(n) ffz(~(n))
1353 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1354 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1355 * We need to be 1 filesystem block less than the 2^32 sector limit.
1357 static loff_t ext4_max_size(int bits)
1359 loff_t res = EXT4_NDIR_BLOCKS;
1360 /* This constant is calculated to be the largest file size for a
1361 * dense, 4k-blocksize file such that the total number of
1362 * sectors in the file, including data and all indirect blocks,
1363 * does not exceed 2^32. */
1364 const loff_t upper_limit = 0x1ff7fffd000LL;
1366 res += 1LL << (bits-2);
1367 res += 1LL << (2*(bits-2));
1368 res += 1LL << (3*(bits-2));
1370 if (res > upper_limit)
1375 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1376 ext4_fsblk_t logic_sb_block,
1379 struct ext4_sb_info *sbi = EXT4_SB(sb);
1380 unsigned long bg, first_meta_bg;
1383 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1385 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1387 return (logic_sb_block + nr + 1);
1388 bg = sbi->s_desc_per_block * nr;
1389 if (ext4_bg_has_super(sb, bg))
1391 return (has_super + ext4_group_first_block_no(sb, bg));
1395 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1397 struct buffer_head * bh;
1398 struct ext4_super_block *es = NULL;
1399 struct ext4_sb_info *sbi;
1401 ext4_fsblk_t sb_block = get_sb_block(&data);
1402 ext4_fsblk_t logic_sb_block;
1403 unsigned long offset = 0;
1404 unsigned int journal_inum = 0;
1405 unsigned long journal_devnum = 0;
1406 unsigned long def_mount_opts;
1415 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1418 sb->s_fs_info = sbi;
1419 sbi->s_mount_opt = 0;
1420 sbi->s_resuid = EXT4_DEF_RESUID;
1421 sbi->s_resgid = EXT4_DEF_RESGID;
1425 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1427 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1432 * The ext4 superblock will not be buffer aligned for other than 1kB
1433 * block sizes. We need to calculate the offset from buffer start.
1435 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1436 logic_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1437 offset = sector_div(logic_sb_block, blocksize);
1439 logic_sb_block = sb_block;
1442 if (!(bh = sb_bread(sb, logic_sb_block))) {
1443 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1447 * Note: s_es must be initialized as soon as possible because
1448 * some ext4 macro-instructions depend on its value
1450 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1452 sb->s_magic = le16_to_cpu(es->s_magic);
1453 if (sb->s_magic != EXT4_SUPER_MAGIC)
1456 /* Set defaults before we parse the mount options */
1457 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1458 if (def_mount_opts & EXT4_DEFM_DEBUG)
1459 set_opt(sbi->s_mount_opt, DEBUG);
1460 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1461 set_opt(sbi->s_mount_opt, GRPID);
1462 if (def_mount_opts & EXT4_DEFM_UID16)
1463 set_opt(sbi->s_mount_opt, NO_UID32);
1464 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1465 set_opt(sbi->s_mount_opt, XATTR_USER);
1466 if (def_mount_opts & EXT4_DEFM_ACL)
1467 set_opt(sbi->s_mount_opt, POSIX_ACL);
1468 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1469 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1470 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1471 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1472 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1473 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1475 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1476 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1477 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
1478 set_opt(sbi->s_mount_opt, ERRORS_RO);
1480 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1481 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1483 set_opt(sbi->s_mount_opt, RESERVATION);
1485 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1489 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1490 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1492 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1493 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1494 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1495 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1497 "EXT4-fs warning: feature flags set on rev 0 fs, "
1498 "running e2fsck is recommended\n");
1500 * Check feature flags regardless of the revision level, since we
1501 * previously didn't change the revision level when setting the flags,
1502 * so there is a chance incompat flags are set on a rev 0 filesystem.
1504 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1506 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1507 "unsupported optional features (%x).\n",
1508 sb->s_id, le32_to_cpu(features));
1511 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1512 if (!(sb->s_flags & MS_RDONLY) && features) {
1513 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1514 "unsupported optional features (%x).\n",
1515 sb->s_id, le32_to_cpu(features));
1518 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1520 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1521 blocksize > EXT4_MAX_BLOCK_SIZE) {
1523 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1524 blocksize, sb->s_id);
1528 hblock = bdev_hardsect_size(sb->s_bdev);
1529 if (sb->s_blocksize != blocksize) {
1531 * Make sure the blocksize for the filesystem is larger
1532 * than the hardware sectorsize for the machine.
1534 if (blocksize < hblock) {
1535 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
1536 "device blocksize %d.\n", blocksize, hblock);
1541 sb_set_blocksize(sb, blocksize);
1542 logic_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1543 offset = sector_div(logic_sb_block, blocksize);
1544 bh = sb_bread(sb, logic_sb_block);
1547 "EXT4-fs: Can't read superblock on 2nd try.\n");
1550 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1552 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1554 "EXT4-fs: Magic mismatch, very weird !\n");
1559 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1561 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1562 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1563 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1565 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1566 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1567 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1568 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1569 (sbi->s_inode_size > blocksize)) {
1571 "EXT4-fs: unsupported inode size: %d\n",
1576 sbi->s_frag_size = EXT4_MIN_FRAG_SIZE <<
1577 le32_to_cpu(es->s_log_frag_size);
1578 if (blocksize != sbi->s_frag_size) {
1580 "EXT4-fs: fragsize %lu != blocksize %u (unsupported)\n",
1581 sbi->s_frag_size, blocksize);
1584 sbi->s_frags_per_block = 1;
1585 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1586 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1587 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1588 if (EXT4_INODE_SIZE(sb) == 0)
1590 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
1591 if (sbi->s_inodes_per_block == 0)
1593 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1594 sbi->s_inodes_per_block;
1595 sbi->s_desc_per_block = blocksize / sizeof(struct ext4_group_desc);
1597 sbi->s_mount_state = le16_to_cpu(es->s_state);
1598 sbi->s_addr_per_block_bits = log2(EXT4_ADDR_PER_BLOCK(sb));
1599 sbi->s_desc_per_block_bits = log2(EXT4_DESC_PER_BLOCK(sb));
1600 for (i=0; i < 4; i++)
1601 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1602 sbi->s_def_hash_version = es->s_def_hash_version;
1604 if (sbi->s_blocks_per_group > blocksize * 8) {
1606 "EXT4-fs: #blocks per group too big: %lu\n",
1607 sbi->s_blocks_per_group);
1610 if (sbi->s_frags_per_group > blocksize * 8) {
1612 "EXT4-fs: #fragments per group too big: %lu\n",
1613 sbi->s_frags_per_group);
1616 if (sbi->s_inodes_per_group > blocksize * 8) {
1618 "EXT4-fs: #inodes per group too big: %lu\n",
1619 sbi->s_inodes_per_group);
1623 if (le32_to_cpu(es->s_blocks_count) >
1624 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1625 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
1626 " too large to mount safely\n", sb->s_id);
1627 if (sizeof(sector_t) < 8)
1628 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
1633 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1635 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1636 le32_to_cpu(es->s_first_data_block) - 1)
1637 / EXT4_BLOCKS_PER_GROUP(sb)) + 1;
1638 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1639 EXT4_DESC_PER_BLOCK(sb);
1640 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1642 if (sbi->s_group_desc == NULL) {
1643 printk (KERN_ERR "EXT4-fs: not enough memory\n");
1647 bgl_lock_init(&sbi->s_blockgroup_lock);
1649 for (i = 0; i < db_count; i++) {
1650 block = descriptor_loc(sb, logic_sb_block, i);
1651 sbi->s_group_desc[i] = sb_bread(sb, block);
1652 if (!sbi->s_group_desc[i]) {
1653 printk (KERN_ERR "EXT4-fs: "
1654 "can't read group descriptor %d\n", i);
1659 if (!ext4_check_descriptors (sb)) {
1660 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
1663 sbi->s_gdb_count = db_count;
1664 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1665 spin_lock_init(&sbi->s_next_gen_lock);
1667 percpu_counter_init(&sbi->s_freeblocks_counter,
1668 ext4_count_free_blocks(sb));
1669 percpu_counter_init(&sbi->s_freeinodes_counter,
1670 ext4_count_free_inodes(sb));
1671 percpu_counter_init(&sbi->s_dirs_counter,
1672 ext4_count_dirs(sb));
1674 /* per fileystem reservation list head & lock */
1675 spin_lock_init(&sbi->s_rsv_window_lock);
1676 sbi->s_rsv_window_root = RB_ROOT;
1677 /* Add a single, static dummy reservation to the start of the
1678 * reservation window list --- it gives us a placeholder for
1679 * append-at-start-of-list which makes the allocation logic
1680 * _much_ simpler. */
1681 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1682 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1683 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1684 sbi->s_rsv_window_head.rsv_goal_size = 0;
1685 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
1688 * set up enough so that it can read an inode
1690 sb->s_op = &ext4_sops;
1691 sb->s_export_op = &ext4_export_ops;
1692 sb->s_xattr = ext4_xattr_handlers;
1694 sb->s_qcop = &ext4_qctl_operations;
1695 sb->dq_op = &ext4_quota_operations;
1697 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1701 needs_recovery = (es->s_last_orphan != 0 ||
1702 EXT4_HAS_INCOMPAT_FEATURE(sb,
1703 EXT4_FEATURE_INCOMPAT_RECOVER));
1706 * The first inode we look at is the journal inode. Don't try
1707 * root first: it may be modified in the journal!
1709 if (!test_opt(sb, NOLOAD) &&
1710 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1711 if (ext4_load_journal(sb, es, journal_devnum))
1713 } else if (journal_inum) {
1714 if (ext4_create_journal(sb, es, journal_inum))
1719 "ext4: No journal on filesystem on %s\n",
1724 /* We have now updated the journal if required, so we can
1725 * validate the data journaling mode. */
1726 switch (test_opt(sb, DATA_FLAGS)) {
1728 /* No mode set, assume a default based on the journal
1729 capabilities: ORDERED_DATA if the journal can
1730 cope, else JOURNAL_DATA */
1731 if (jbd2_journal_check_available_features
1732 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
1733 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1735 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1738 case EXT4_MOUNT_ORDERED_DATA:
1739 case EXT4_MOUNT_WRITEBACK_DATA:
1740 if (!jbd2_journal_check_available_features
1741 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
1742 printk(KERN_ERR "EXT4-fs: Journal does not support "
1743 "requested data journaling mode\n");
1750 if (test_opt(sb, NOBH)) {
1751 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1752 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
1753 "its supported only with writeback mode\n");
1754 clear_opt(sbi->s_mount_opt, NOBH);
1758 * The jbd2_journal_load will have done any necessary log recovery,
1759 * so we can safely mount the rest of the filesystem now.
1762 root = iget(sb, EXT4_ROOT_INO);
1763 sb->s_root = d_alloc_root(root);
1765 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1769 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1772 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
1776 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1778 * akpm: core read_super() calls in here with the superblock locked.
1779 * That deadlocks, because orphan cleanup needs to lock the superblock
1780 * in numerous places. Here we just pop the lock - it's relatively
1781 * harmless, because we are now ready to accept write_super() requests,
1782 * and aviro says that's the only reason for hanging onto the
1785 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
1786 ext4_orphan_cleanup(sb, es);
1787 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
1789 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
1790 ext4_mark_recovery_complete(sb, es);
1791 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
1792 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
1793 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
1803 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
1808 jbd2_journal_destroy(sbi->s_journal);
1810 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1811 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1812 percpu_counter_destroy(&sbi->s_dirs_counter);
1814 for (i = 0; i < db_count; i++)
1815 brelse(sbi->s_group_desc[i]);
1816 kfree(sbi->s_group_desc);
1819 for (i = 0; i < MAXQUOTAS; i++)
1820 kfree(sbi->s_qf_names[i]);
1822 ext4_blkdev_remove(sbi);
1825 sb->s_fs_info = NULL;
1832 * Setup any per-fs journal parameters now. We'll do this both on
1833 * initial mount, once the journal has been initialised but before we've
1834 * done any recovery; and again on any subsequent remount.
1836 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
1838 struct ext4_sb_info *sbi = EXT4_SB(sb);
1840 if (sbi->s_commit_interval)
1841 journal->j_commit_interval = sbi->s_commit_interval;
1842 /* We could also set up an ext4-specific default for the commit
1843 * interval here, but for now we'll just fall back to the jbd
1846 spin_lock(&journal->j_state_lock);
1847 if (test_opt(sb, BARRIER))
1848 journal->j_flags |= JBD2_BARRIER;
1850 journal->j_flags &= ~JBD2_BARRIER;
1851 spin_unlock(&journal->j_state_lock);
1854 static journal_t *ext4_get_journal(struct super_block *sb,
1855 unsigned int journal_inum)
1857 struct inode *journal_inode;
1860 /* First, test for the existence of a valid inode on disk. Bad
1861 * things happen if we iget() an unused inode, as the subsequent
1862 * iput() will try to delete it. */
1864 journal_inode = iget(sb, journal_inum);
1865 if (!journal_inode) {
1866 printk(KERN_ERR "EXT4-fs: no journal found.\n");
1869 if (!journal_inode->i_nlink) {
1870 make_bad_inode(journal_inode);
1871 iput(journal_inode);
1872 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
1876 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1877 journal_inode, journal_inode->i_size);
1878 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1879 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
1880 iput(journal_inode);
1884 journal = jbd2_journal_init_inode(journal_inode);
1886 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
1887 iput(journal_inode);
1890 journal->j_private = sb;
1891 ext4_init_journal_params(sb, journal);
1895 static journal_t *ext4_get_dev_journal(struct super_block *sb,
1898 struct buffer_head * bh;
1902 int hblock, blocksize;
1903 ext4_fsblk_t sb_block;
1904 unsigned long offset;
1905 struct ext4_super_block * es;
1906 struct block_device *bdev;
1908 bdev = ext4_blkdev_get(j_dev);
1912 if (bd_claim(bdev, sb)) {
1914 "EXT4: failed to claim external journal device.\n");
1919 blocksize = sb->s_blocksize;
1920 hblock = bdev_hardsect_size(bdev);
1921 if (blocksize < hblock) {
1923 "EXT4-fs: blocksize too small for journal device.\n");
1927 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
1928 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
1929 set_blocksize(bdev, blocksize);
1930 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1931 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
1932 "external journal\n");
1936 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1937 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
1938 !(le32_to_cpu(es->s_feature_incompat) &
1939 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1940 printk(KERN_ERR "EXT4-fs: external journal has "
1941 "bad superblock\n");
1946 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1947 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
1952 len = le32_to_cpu(es->s_blocks_count);
1953 start = sb_block + 1;
1954 brelse(bh); /* we're done with the superblock */
1956 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
1957 start, len, blocksize);
1959 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
1962 journal->j_private = sb;
1963 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1964 wait_on_buffer(journal->j_sb_buffer);
1965 if (!buffer_uptodate(journal->j_sb_buffer)) {
1966 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
1969 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
1970 printk(KERN_ERR "EXT4-fs: External journal has more than one "
1971 "user (unsupported) - %d\n",
1972 be32_to_cpu(journal->j_superblock->s_nr_users));
1975 EXT4_SB(sb)->journal_bdev = bdev;
1976 ext4_init_journal_params(sb, journal);
1979 jbd2_journal_destroy(journal);
1981 ext4_blkdev_put(bdev);
1985 static int ext4_load_journal(struct super_block *sb,
1986 struct ext4_super_block *es,
1987 unsigned long journal_devnum)
1990 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
1993 int really_read_only;
1995 if (journal_devnum &&
1996 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
1997 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
1998 "numbers have changed\n");
1999 journal_dev = new_decode_dev(journal_devnum);
2001 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2003 really_read_only = bdev_read_only(sb->s_bdev);
2006 * Are we loading a blank journal or performing recovery after a
2007 * crash? For recovery, we need to check in advance whether we
2008 * can get read-write access to the device.
2011 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2012 if (sb->s_flags & MS_RDONLY) {
2013 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2014 "required on readonly filesystem.\n");
2015 if (really_read_only) {
2016 printk(KERN_ERR "EXT4-fs: write access "
2017 "unavailable, cannot proceed.\n");
2020 printk (KERN_INFO "EXT4-fs: write access will "
2021 "be enabled during recovery.\n");
2025 if (journal_inum && journal_dev) {
2026 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2027 "and inode journals!\n");
2032 if (!(journal = ext4_get_journal(sb, journal_inum)))
2035 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2039 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2040 err = jbd2_journal_update_format(journal);
2042 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2043 jbd2_journal_destroy(journal);
2048 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2049 err = jbd2_journal_wipe(journal, !really_read_only);
2051 err = jbd2_journal_load(journal);
2054 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2055 jbd2_journal_destroy(journal);
2059 EXT4_SB(sb)->s_journal = journal;
2060 ext4_clear_journal_err(sb, es);
2062 if (journal_devnum &&
2063 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2064 es->s_journal_dev = cpu_to_le32(journal_devnum);
2067 /* Make sure we flush the recovery flag to disk. */
2068 ext4_commit_super(sb, es, 1);
2074 static int ext4_create_journal(struct super_block * sb,
2075 struct ext4_super_block * es,
2076 unsigned int journal_inum)
2080 if (sb->s_flags & MS_RDONLY) {
2081 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2082 "create journal.\n");
2086 if (!(journal = ext4_get_journal(sb, journal_inum)))
2089 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2092 if (jbd2_journal_create(journal)) {
2093 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2094 jbd2_journal_destroy(journal);
2098 EXT4_SB(sb)->s_journal = journal;
2100 ext4_update_dynamic_rev(sb);
2101 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2102 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2104 es->s_journal_inum = cpu_to_le32(journal_inum);
2107 /* Make sure we flush the recovery flag to disk. */
2108 ext4_commit_super(sb, es, 1);
2113 static void ext4_commit_super (struct super_block * sb,
2114 struct ext4_super_block * es,
2117 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2121 es->s_wtime = cpu_to_le32(get_seconds());
2122 es->s_free_blocks_count = cpu_to_le32(ext4_count_free_blocks(sb));
2123 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2124 BUFFER_TRACE(sbh, "marking dirty");
2125 mark_buffer_dirty(sbh);
2127 sync_dirty_buffer(sbh);
2132 * Have we just finished recovery? If so, and if we are mounting (or
2133 * remounting) the filesystem readonly, then we will end up with a
2134 * consistent fs on disk. Record that fact.
2136 static void ext4_mark_recovery_complete(struct super_block * sb,
2137 struct ext4_super_block * es)
2139 journal_t *journal = EXT4_SB(sb)->s_journal;
2141 jbd2_journal_lock_updates(journal);
2142 jbd2_journal_flush(journal);
2143 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2144 sb->s_flags & MS_RDONLY) {
2145 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2147 ext4_commit_super(sb, es, 1);
2149 jbd2_journal_unlock_updates(journal);
2153 * If we are mounting (or read-write remounting) a filesystem whose journal
2154 * has recorded an error from a previous lifetime, move that error to the
2155 * main filesystem now.
2157 static void ext4_clear_journal_err(struct super_block * sb,
2158 struct ext4_super_block * es)
2164 journal = EXT4_SB(sb)->s_journal;
2167 * Now check for any error status which may have been recorded in the
2168 * journal by a prior ext4_error() or ext4_abort()
2171 j_errno = jbd2_journal_errno(journal);
2175 errstr = ext4_decode_error(sb, j_errno, nbuf);
2176 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2177 "from previous mount: %s", errstr);
2178 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2179 "filesystem check.");
2181 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2182 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2183 ext4_commit_super (sb, es, 1);
2185 jbd2_journal_clear_err(journal);
2190 * Force the running and committing transactions to commit,
2191 * and wait on the commit.
2193 int ext4_force_commit(struct super_block *sb)
2198 if (sb->s_flags & MS_RDONLY)
2201 journal = EXT4_SB(sb)->s_journal;
2203 ret = ext4_journal_force_commit(journal);
2208 * Ext4 always journals updates to the superblock itself, so we don't
2209 * have to propagate any other updates to the superblock on disk at this
2210 * point. Just start an async writeback to get the buffers on their way
2213 * This implicitly triggers the writebehind on sync().
2216 static void ext4_write_super (struct super_block * sb)
2218 if (mutex_trylock(&sb->s_lock) != 0)
2223 static int ext4_sync_fs(struct super_block *sb, int wait)
2228 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2230 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2236 * LVM calls this function before a (read-only) snapshot is created. This
2237 * gives us a chance to flush the journal completely and mark the fs clean.
2239 static void ext4_write_super_lockfs(struct super_block *sb)
2243 if (!(sb->s_flags & MS_RDONLY)) {
2244 journal_t *journal = EXT4_SB(sb)->s_journal;
2246 /* Now we set up the journal barrier. */
2247 jbd2_journal_lock_updates(journal);
2248 jbd2_journal_flush(journal);
2250 /* Journal blocked and flushed, clear needs_recovery flag. */
2251 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2252 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2257 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2258 * flag here, even though the filesystem is not technically dirty yet.
2260 static void ext4_unlockfs(struct super_block *sb)
2262 if (!(sb->s_flags & MS_RDONLY)) {
2264 /* Reser the needs_recovery flag before the fs is unlocked. */
2265 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2266 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2268 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2272 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2274 struct ext4_super_block * es;
2275 struct ext4_sb_info *sbi = EXT4_SB(sb);
2276 ext4_fsblk_t n_blocks_count = 0;
2277 unsigned long old_sb_flags;
2278 struct ext4_mount_options old_opts;
2284 /* Store the original options */
2285 old_sb_flags = sb->s_flags;
2286 old_opts.s_mount_opt = sbi->s_mount_opt;
2287 old_opts.s_resuid = sbi->s_resuid;
2288 old_opts.s_resgid = sbi->s_resgid;
2289 old_opts.s_commit_interval = sbi->s_commit_interval;
2291 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2292 for (i = 0; i < MAXQUOTAS; i++)
2293 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2297 * Allow the "check" option to be passed as a remount option.
2299 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2304 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2305 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2307 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2308 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2312 ext4_init_journal_params(sb, sbi->s_journal);
2314 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2315 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2316 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2321 if (*flags & MS_RDONLY) {
2323 * First of all, the unconditional stuff we have to do
2324 * to disable replay of the journal when we next remount
2326 sb->s_flags |= MS_RDONLY;
2329 * OK, test if we are remounting a valid rw partition
2330 * readonly, and if so set the rdonly flag and then
2331 * mark the partition as valid again.
2333 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2334 (sbi->s_mount_state & EXT4_VALID_FS))
2335 es->s_state = cpu_to_le16(sbi->s_mount_state);
2337 ext4_mark_recovery_complete(sb, es);
2340 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2341 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2342 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2343 "remount RDWR because of unsupported "
2344 "optional features (%x).\n",
2345 sb->s_id, le32_to_cpu(ret));
2350 * Mounting a RDONLY partition read-write, so reread
2351 * and store the current valid flag. (It may have
2352 * been changed by e2fsck since we originally mounted
2355 ext4_clear_journal_err(sb, es);
2356 sbi->s_mount_state = le16_to_cpu(es->s_state);
2357 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2359 if (!ext4_setup_super (sb, es, 0))
2360 sb->s_flags &= ~MS_RDONLY;
2364 /* Release old quota file names */
2365 for (i = 0; i < MAXQUOTAS; i++)
2366 if (old_opts.s_qf_names[i] &&
2367 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2368 kfree(old_opts.s_qf_names[i]);
2372 sb->s_flags = old_sb_flags;
2373 sbi->s_mount_opt = old_opts.s_mount_opt;
2374 sbi->s_resuid = old_opts.s_resuid;
2375 sbi->s_resgid = old_opts.s_resgid;
2376 sbi->s_commit_interval = old_opts.s_commit_interval;
2378 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2379 for (i = 0; i < MAXQUOTAS; i++) {
2380 if (sbi->s_qf_names[i] &&
2381 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2382 kfree(sbi->s_qf_names[i]);
2383 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2389 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2391 struct super_block *sb = dentry->d_sb;
2392 struct ext4_sb_info *sbi = EXT4_SB(sb);
2393 struct ext4_super_block *es = sbi->s_es;
2394 ext4_fsblk_t overhead;
2397 if (test_opt (sb, MINIX_DF))
2400 unsigned long ngroups;
2401 ngroups = EXT4_SB(sb)->s_groups_count;
2405 * Compute the overhead (FS structures)
2409 * All of the blocks before first_data_block are
2412 overhead = le32_to_cpu(es->s_first_data_block);
2415 * Add the overhead attributed to the superblock and
2416 * block group descriptors. If the sparse superblocks
2417 * feature is turned on, then not all groups have this.
2419 for (i = 0; i < ngroups; i++) {
2420 overhead += ext4_bg_has_super(sb, i) +
2421 ext4_bg_num_gdb(sb, i);
2426 * Every block group has an inode bitmap, a block
2427 * bitmap, and an inode table.
2429 overhead += (ngroups * (2 + EXT4_SB(sb)->s_itb_per_group));
2432 buf->f_type = EXT4_SUPER_MAGIC;
2433 buf->f_bsize = sb->s_blocksize;
2434 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
2435 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
2436 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2437 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2439 buf->f_files = le32_to_cpu(es->s_inodes_count);
2440 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
2441 buf->f_namelen = EXT4_NAME_LEN;
2445 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2446 * is locked for write. Otherwise the are possible deadlocks:
2447 * Process 1 Process 2
2448 * ext4_create() quota_sync()
2449 * jbd2_journal_start() write_dquot()
2450 * DQUOT_INIT() down(dqio_mutex)
2451 * down(dqio_mutex) jbd2_journal_start()
2457 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2459 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2462 static int ext4_dquot_initialize(struct inode *inode, int type)
2467 /* We may create quota structure so we need to reserve enough blocks */
2468 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
2470 return PTR_ERR(handle);
2471 ret = dquot_initialize(inode, type);
2472 err = ext4_journal_stop(handle);
2478 static int ext4_dquot_drop(struct inode *inode)
2483 /* We may delete quota structure so we need to reserve enough blocks */
2484 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2486 return PTR_ERR(handle);
2487 ret = dquot_drop(inode);
2488 err = ext4_journal_stop(handle);
2494 static int ext4_write_dquot(struct dquot *dquot)
2498 struct inode *inode;
2500 inode = dquot_to_inode(dquot);
2501 handle = ext4_journal_start(inode,
2502 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2504 return PTR_ERR(handle);
2505 ret = dquot_commit(dquot);
2506 err = ext4_journal_stop(handle);
2512 static int ext4_acquire_dquot(struct dquot *dquot)
2517 handle = ext4_journal_start(dquot_to_inode(dquot),
2518 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2520 return PTR_ERR(handle);
2521 ret = dquot_acquire(dquot);
2522 err = ext4_journal_stop(handle);
2528 static int ext4_release_dquot(struct dquot *dquot)
2533 handle = ext4_journal_start(dquot_to_inode(dquot),
2534 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2536 return PTR_ERR(handle);
2537 ret = dquot_release(dquot);
2538 err = ext4_journal_stop(handle);
2544 static int ext4_mark_dquot_dirty(struct dquot *dquot)
2546 /* Are we journalling quotas? */
2547 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2548 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2549 dquot_mark_dquot_dirty(dquot);
2550 return ext4_write_dquot(dquot);
2552 return dquot_mark_dquot_dirty(dquot);
2556 static int ext4_write_info(struct super_block *sb, int type)
2561 /* Data block + inode block */
2562 handle = ext4_journal_start(sb->s_root->d_inode, 2);
2564 return PTR_ERR(handle);
2565 ret = dquot_commit_info(sb, type);
2566 err = ext4_journal_stop(handle);
2573 * Turn on quotas during mount time - we need to find
2574 * the quota file and such...
2576 static int ext4_quota_on_mount(struct super_block *sb, int type)
2578 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2579 EXT4_SB(sb)->s_jquota_fmt, type);
2583 * Standard function to be called on quota_on
2585 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
2589 struct nameidata nd;
2591 if (!test_opt(sb, QUOTA))
2593 /* Not journalling quota? */
2594 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2595 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
2596 return vfs_quota_on(sb, type, format_id, path);
2597 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2600 /* Quotafile not on the same filesystem? */
2601 if (nd.mnt->mnt_sb != sb) {
2605 /* Quotafile not of fs root? */
2606 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2608 "EXT4-fs: Quota file not on filesystem root. "
2609 "Journalled quota will not work.\n");
2611 return vfs_quota_on(sb, type, format_id, path);
2614 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2615 * acquiring the locks... As quota files are never truncated and quota code
2616 * itself serializes the operations (and noone else should touch the files)
2617 * we don't have to be afraid of races */
2618 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
2619 size_t len, loff_t off)
2621 struct inode *inode = sb_dqopt(sb)->files[type];
2622 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2624 int offset = off & (sb->s_blocksize - 1);
2627 struct buffer_head *bh;
2628 loff_t i_size = i_size_read(inode);
2632 if (off+len > i_size)
2635 while (toread > 0) {
2636 tocopy = sb->s_blocksize - offset < toread ?
2637 sb->s_blocksize - offset : toread;
2638 bh = ext4_bread(NULL, inode, blk, 0, &err);
2641 if (!bh) /* A hole? */
2642 memset(data, 0, tocopy);
2644 memcpy(data, bh->b_data+offset, tocopy);
2654 /* Write to quotafile (we know the transaction is already started and has
2655 * enough credits) */
2656 static ssize_t ext4_quota_write(struct super_block *sb, int type,
2657 const char *data, size_t len, loff_t off)
2659 struct inode *inode = sb_dqopt(sb)->files[type];
2660 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2662 int offset = off & (sb->s_blocksize - 1);
2664 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
2665 size_t towrite = len;
2666 struct buffer_head *bh;
2667 handle_t *handle = journal_current_handle();
2669 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2670 while (towrite > 0) {
2671 tocopy = sb->s_blocksize - offset < towrite ?
2672 sb->s_blocksize - offset : towrite;
2673 bh = ext4_bread(handle, inode, blk, 1, &err);
2676 if (journal_quota) {
2677 err = ext4_journal_get_write_access(handle, bh);
2684 memcpy(bh->b_data+offset, data, tocopy);
2685 flush_dcache_page(bh->b_page);
2688 err = ext4_journal_dirty_metadata(handle, bh);
2690 /* Always do at least ordered writes for quotas */
2691 err = ext4_journal_dirty_data(handle, bh);
2692 mark_buffer_dirty(bh);
2705 if (inode->i_size < off+len-towrite) {
2706 i_size_write(inode, off+len-towrite);
2707 EXT4_I(inode)->i_disksize = inode->i_size;
2710 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2711 ext4_mark_inode_dirty(handle, inode);
2712 mutex_unlock(&inode->i_mutex);
2713 return len - towrite;
2718 static int ext4_get_sb(struct file_system_type *fs_type,
2719 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2721 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
2724 static struct file_system_type ext4dev_fs_type = {
2725 .owner = THIS_MODULE,
2727 .get_sb = ext4_get_sb,
2728 .kill_sb = kill_block_super,
2729 .fs_flags = FS_REQUIRES_DEV,
2732 static int __init init_ext4_fs(void)
2734 int err = init_ext4_xattr();
2737 err = init_inodecache();
2740 err = register_filesystem(&ext4dev_fs_type);
2745 destroy_inodecache();
2751 static void __exit exit_ext4_fs(void)
2753 unregister_filesystem(&ext4dev_fs_type);
2754 destroy_inodecache();
2758 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2759 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
2760 MODULE_LICENSE("GPL");
2761 module_init(init_ext4_fs)
2762 module_exit(exit_ext4_fs)