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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <asm/uaccess.h>
43 #include "ext4_jbd2.h"
49 struct proc_dir_entry *ext4_proc_root;
51 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
52 unsigned long journal_devnum);
53 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
55 static void ext4_commit_super(struct super_block *sb,
56 struct ext4_super_block *es, int sync);
57 static void ext4_mark_recovery_complete(struct super_block *sb,
58 struct ext4_super_block *es);
59 static void ext4_clear_journal_err(struct super_block *sb,
60 struct ext4_super_block *es);
61 static int ext4_sync_fs(struct super_block *sb, int wait);
62 static const char *ext4_decode_error(struct super_block *sb, int errno,
64 static int ext4_remount(struct super_block *sb, int *flags, char *data);
65 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
66 static void ext4_unlockfs(struct super_block *sb);
67 static void ext4_write_super(struct super_block *sb);
68 static void ext4_write_super_lockfs(struct super_block *sb);
71 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
72 struct ext4_group_desc *bg)
74 return le32_to_cpu(bg->bg_block_bitmap_lo) |
75 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
76 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
79 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
80 struct ext4_group_desc *bg)
82 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
83 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
84 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
87 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
88 struct ext4_group_desc *bg)
90 return le32_to_cpu(bg->bg_inode_table_lo) |
91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
92 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
95 void ext4_block_bitmap_set(struct super_block *sb,
96 struct ext4_group_desc *bg, ext4_fsblk_t blk)
98 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
99 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
100 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
103 void ext4_inode_bitmap_set(struct super_block *sb,
104 struct ext4_group_desc *bg, ext4_fsblk_t blk)
106 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
107 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
108 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
111 void ext4_inode_table_set(struct super_block *sb,
112 struct ext4_group_desc *bg, ext4_fsblk_t blk)
114 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
115 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
116 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
120 * Wrappers for jbd2_journal_start/end.
122 * The only special thing we need to do here is to make sure that all
123 * journal_end calls result in the superblock being marked dirty, so
124 * that sync() will call the filesystem's write_super callback if
127 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
131 if (sb->s_flags & MS_RDONLY)
132 return ERR_PTR(-EROFS);
134 /* Special case here: if the journal has aborted behind our
135 * backs (eg. EIO in the commit thread), then we still need to
136 * take the FS itself readonly cleanly. */
137 journal = EXT4_SB(sb)->s_journal;
138 if (is_journal_aborted(journal)) {
139 ext4_abort(sb, __func__,
140 "Detected aborted journal");
141 return ERR_PTR(-EROFS);
144 return jbd2_journal_start(journal, nblocks);
148 * The only special thing we need to do here is to make sure that all
149 * jbd2_journal_stop calls result in the superblock being marked dirty, so
150 * that sync() will call the filesystem's write_super callback if
153 int __ext4_journal_stop(const char *where, handle_t *handle)
155 struct super_block *sb;
159 sb = handle->h_transaction->t_journal->j_private;
161 rc = jbd2_journal_stop(handle);
166 __ext4_std_error(sb, where, err);
170 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
171 struct buffer_head *bh, handle_t *handle, int err)
174 const char *errstr = ext4_decode_error(NULL, err, nbuf);
177 BUFFER_TRACE(bh, "abort");
182 if (is_handle_aborted(handle))
185 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
186 caller, errstr, err_fn);
188 jbd2_journal_abort_handle(handle);
191 /* Deal with the reporting of failure conditions on a filesystem such as
192 * inconsistencies detected or read IO failures.
194 * On ext2, we can store the error state of the filesystem in the
195 * superblock. That is not possible on ext4, because we may have other
196 * write ordering constraints on the superblock which prevent us from
197 * writing it out straight away; and given that the journal is about to
198 * be aborted, we can't rely on the current, or future, transactions to
199 * write out the superblock safely.
201 * We'll just use the jbd2_journal_abort() error code to record an error in
202 * the journal instead. On recovery, the journal will compain about
203 * that error until we've noted it down and cleared it.
206 static void ext4_handle_error(struct super_block *sb)
208 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
210 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
211 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
213 if (sb->s_flags & MS_RDONLY)
216 if (!test_opt(sb, ERRORS_CONT)) {
217 journal_t *journal = EXT4_SB(sb)->s_journal;
219 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
221 jbd2_journal_abort(journal, -EIO);
223 if (test_opt(sb, ERRORS_RO)) {
224 printk(KERN_CRIT "Remounting filesystem read-only\n");
225 sb->s_flags |= MS_RDONLY;
227 ext4_commit_super(sb, es, 1);
228 if (test_opt(sb, ERRORS_PANIC))
229 panic("EXT4-fs (device %s): panic forced after error\n",
233 void ext4_error(struct super_block *sb, const char *function,
234 const char *fmt, ...)
239 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
244 ext4_handle_error(sb);
247 static const char *ext4_decode_error(struct super_block *sb, int errno,
254 errstr = "IO failure";
257 errstr = "Out of memory";
260 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
261 errstr = "Journal has aborted";
263 errstr = "Readonly filesystem";
266 /* If the caller passed in an extra buffer for unknown
267 * errors, textualise them now. Else we just return
270 /* Check for truncated error codes... */
271 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
280 /* __ext4_std_error decodes expected errors from journaling functions
281 * automatically and invokes the appropriate error response. */
283 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
288 /* Special case: if the error is EROFS, and we're not already
289 * inside a transaction, then there's really no point in logging
291 if (errno == -EROFS && journal_current_handle() == NULL &&
292 (sb->s_flags & MS_RDONLY))
295 errstr = ext4_decode_error(sb, errno, nbuf);
296 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
297 sb->s_id, function, errstr);
299 ext4_handle_error(sb);
303 * ext4_abort is a much stronger failure handler than ext4_error. The
304 * abort function may be used to deal with unrecoverable failures such
305 * as journal IO errors or ENOMEM at a critical moment in log management.
307 * We unconditionally force the filesystem into an ABORT|READONLY state,
308 * unless the error response on the fs has been set to panic in which
309 * case we take the easy way out and panic immediately.
312 void ext4_abort(struct super_block *sb, const char *function,
313 const char *fmt, ...)
317 printk(KERN_CRIT "ext4_abort called.\n");
320 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
325 if (test_opt(sb, ERRORS_PANIC))
326 panic("EXT4-fs panic from previous error\n");
328 if (sb->s_flags & MS_RDONLY)
331 printk(KERN_CRIT "Remounting filesystem read-only\n");
332 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
333 sb->s_flags |= MS_RDONLY;
334 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
335 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
338 void ext4_warning(struct super_block *sb, const char *function,
339 const char *fmt, ...)
344 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
351 void ext4_update_dynamic_rev(struct super_block *sb)
353 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
355 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
358 ext4_warning(sb, __func__,
359 "updating to rev %d because of new feature flag, "
360 "running e2fsck is recommended",
363 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
364 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
365 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
366 /* leave es->s_feature_*compat flags alone */
367 /* es->s_uuid will be set by e2fsck if empty */
370 * The rest of the superblock fields should be zero, and if not it
371 * means they are likely already in use, so leave them alone. We
372 * can leave it up to e2fsck to clean up any inconsistencies there.
376 int ext4_update_compat_feature(handle_t *handle,
377 struct super_block *sb, __u32 compat)
380 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
381 err = ext4_journal_get_write_access(handle,
385 EXT4_SET_COMPAT_FEATURE(sb, compat);
388 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
389 "call ext4_journal_dirty_met adata");
390 err = ext4_journal_dirty_metadata(handle,
396 int ext4_update_rocompat_feature(handle_t *handle,
397 struct super_block *sb, __u32 rocompat)
400 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
401 err = ext4_journal_get_write_access(handle,
405 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
408 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
409 "call ext4_journal_dirty_met adata");
410 err = ext4_journal_dirty_metadata(handle,
416 int ext4_update_incompat_feature(handle_t *handle,
417 struct super_block *sb, __u32 incompat)
420 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
421 err = ext4_journal_get_write_access(handle,
425 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
428 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
429 "call ext4_journal_dirty_met adata");
430 err = ext4_journal_dirty_metadata(handle,
437 * Open the external journal device
439 static struct block_device *ext4_blkdev_get(dev_t dev)
441 struct block_device *bdev;
442 char b[BDEVNAME_SIZE];
444 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
450 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
451 __bdevname(dev, b), PTR_ERR(bdev));
456 * Release the journal device
458 static int ext4_blkdev_put(struct block_device *bdev)
461 return blkdev_put(bdev);
464 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
466 struct block_device *bdev;
469 bdev = sbi->journal_bdev;
471 ret = ext4_blkdev_put(bdev);
472 sbi->journal_bdev = NULL;
477 static inline struct inode *orphan_list_entry(struct list_head *l)
479 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
482 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
486 printk(KERN_ERR "sb orphan head is %d\n",
487 le32_to_cpu(sbi->s_es->s_last_orphan));
489 printk(KERN_ERR "sb_info orphan list:\n");
490 list_for_each(l, &sbi->s_orphan) {
491 struct inode *inode = orphan_list_entry(l);
493 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
494 inode->i_sb->s_id, inode->i_ino, inode,
495 inode->i_mode, inode->i_nlink,
500 static void ext4_put_super(struct super_block *sb)
502 struct ext4_sb_info *sbi = EXT4_SB(sb);
503 struct ext4_super_block *es = sbi->s_es;
507 ext4_ext_release(sb);
508 ext4_xattr_put_super(sb);
509 jbd2_journal_destroy(sbi->s_journal);
510 sbi->s_journal = NULL;
511 if (!(sb->s_flags & MS_RDONLY)) {
512 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
513 es->s_state = cpu_to_le16(sbi->s_mount_state);
514 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
515 mark_buffer_dirty(sbi->s_sbh);
516 ext4_commit_super(sb, es, 1);
519 remove_proc_entry(sb->s_id, ext4_proc_root);
521 for (i = 0; i < sbi->s_gdb_count; i++)
522 brelse(sbi->s_group_desc[i]);
523 kfree(sbi->s_group_desc);
524 kfree(sbi->s_flex_groups);
525 percpu_counter_destroy(&sbi->s_freeblocks_counter);
526 percpu_counter_destroy(&sbi->s_freeinodes_counter);
527 percpu_counter_destroy(&sbi->s_dirs_counter);
528 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
531 for (i = 0; i < MAXQUOTAS; i++)
532 kfree(sbi->s_qf_names[i]);
535 /* Debugging code just in case the in-memory inode orphan list
536 * isn't empty. The on-disk one can be non-empty if we've
537 * detected an error and taken the fs readonly, but the
538 * in-memory list had better be clean by this point. */
539 if (!list_empty(&sbi->s_orphan))
540 dump_orphan_list(sb, sbi);
541 J_ASSERT(list_empty(&sbi->s_orphan));
543 invalidate_bdev(sb->s_bdev);
544 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
546 * Invalidate the journal device's buffers. We don't want them
547 * floating about in memory - the physical journal device may
548 * hotswapped, and it breaks the `ro-after' testing code.
550 sync_blockdev(sbi->journal_bdev);
551 invalidate_bdev(sbi->journal_bdev);
552 ext4_blkdev_remove(sbi);
554 sb->s_fs_info = NULL;
559 static struct kmem_cache *ext4_inode_cachep;
562 * Called inside transaction, so use GFP_NOFS
564 static struct inode *ext4_alloc_inode(struct super_block *sb)
566 struct ext4_inode_info *ei;
568 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
571 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
572 ei->i_acl = EXT4_ACL_NOT_CACHED;
573 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
575 ei->i_block_alloc_info = NULL;
576 ei->vfs_inode.i_version = 1;
577 ei->vfs_inode.i_data.writeback_index = 0;
578 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
579 INIT_LIST_HEAD(&ei->i_prealloc_list);
580 spin_lock_init(&ei->i_prealloc_lock);
581 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
582 ei->i_reserved_data_blocks = 0;
583 ei->i_reserved_meta_blocks = 0;
584 ei->i_allocated_meta_blocks = 0;
585 ei->i_delalloc_reserved_flag = 0;
586 spin_lock_init(&(ei->i_block_reservation_lock));
587 return &ei->vfs_inode;
590 static void ext4_destroy_inode(struct inode *inode)
592 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
593 printk("EXT4 Inode %p: orphan list check failed!\n",
595 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
596 EXT4_I(inode), sizeof(struct ext4_inode_info),
600 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
603 static void init_once(void *foo)
605 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
607 INIT_LIST_HEAD(&ei->i_orphan);
608 #ifdef CONFIG_EXT4DEV_FS_XATTR
609 init_rwsem(&ei->xattr_sem);
611 init_rwsem(&ei->i_data_sem);
612 inode_init_once(&ei->vfs_inode);
615 static int init_inodecache(void)
617 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
618 sizeof(struct ext4_inode_info),
619 0, (SLAB_RECLAIM_ACCOUNT|
622 if (ext4_inode_cachep == NULL)
627 static void destroy_inodecache(void)
629 kmem_cache_destroy(ext4_inode_cachep);
632 static void ext4_clear_inode(struct inode *inode)
634 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
635 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
636 if (EXT4_I(inode)->i_acl &&
637 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
638 posix_acl_release(EXT4_I(inode)->i_acl);
639 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
641 if (EXT4_I(inode)->i_default_acl &&
642 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
643 posix_acl_release(EXT4_I(inode)->i_default_acl);
644 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
647 ext4_discard_reservation(inode);
648 EXT4_I(inode)->i_block_alloc_info = NULL;
651 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
652 &EXT4_I(inode)->jinode);
655 static inline void ext4_show_quota_options(struct seq_file *seq,
656 struct super_block *sb)
658 #if defined(CONFIG_QUOTA)
659 struct ext4_sb_info *sbi = EXT4_SB(sb);
661 if (sbi->s_jquota_fmt)
662 seq_printf(seq, ",jqfmt=%s",
663 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
665 if (sbi->s_qf_names[USRQUOTA])
666 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
668 if (sbi->s_qf_names[GRPQUOTA])
669 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
671 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
672 seq_puts(seq, ",usrquota");
674 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
675 seq_puts(seq, ",grpquota");
681 * - it's set to a non-default value OR
682 * - if the per-sb default is different from the global default
684 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
687 unsigned long def_mount_opts;
688 struct super_block *sb = vfs->mnt_sb;
689 struct ext4_sb_info *sbi = EXT4_SB(sb);
690 struct ext4_super_block *es = sbi->s_es;
692 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
693 def_errors = le16_to_cpu(es->s_errors);
695 if (sbi->s_sb_block != 1)
696 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
697 if (test_opt(sb, MINIX_DF))
698 seq_puts(seq, ",minixdf");
699 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
700 seq_puts(seq, ",grpid");
701 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
702 seq_puts(seq, ",nogrpid");
703 if (sbi->s_resuid != EXT4_DEF_RESUID ||
704 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
705 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
707 if (sbi->s_resgid != EXT4_DEF_RESGID ||
708 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
709 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
711 if (test_opt(sb, ERRORS_RO)) {
712 if (def_errors == EXT4_ERRORS_PANIC ||
713 def_errors == EXT4_ERRORS_CONTINUE) {
714 seq_puts(seq, ",errors=remount-ro");
717 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
718 seq_puts(seq, ",errors=continue");
719 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
720 seq_puts(seq, ",errors=panic");
721 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
722 seq_puts(seq, ",nouid32");
723 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
724 seq_puts(seq, ",debug");
725 if (test_opt(sb, OLDALLOC))
726 seq_puts(seq, ",oldalloc");
727 #ifdef CONFIG_EXT4DEV_FS_XATTR
728 if (test_opt(sb, XATTR_USER) &&
729 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
730 seq_puts(seq, ",user_xattr");
731 if (!test_opt(sb, XATTR_USER) &&
732 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
733 seq_puts(seq, ",nouser_xattr");
736 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
737 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
738 seq_puts(seq, ",acl");
739 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
740 seq_puts(seq, ",noacl");
742 if (!test_opt(sb, RESERVATION))
743 seq_puts(seq, ",noreservation");
744 if (sbi->s_commit_interval) {
745 seq_printf(seq, ",commit=%u",
746 (unsigned) (sbi->s_commit_interval / HZ));
749 * We're changing the default of barrier mount option, so
750 * let's always display its mount state so it's clear what its
753 seq_puts(seq, ",barrier=");
754 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
755 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
756 seq_puts(seq, ",journal_async_commit");
757 if (test_opt(sb, NOBH))
758 seq_puts(seq, ",nobh");
759 if (!test_opt(sb, EXTENTS))
760 seq_puts(seq, ",noextents");
761 if (!test_opt(sb, MBALLOC))
762 seq_puts(seq, ",nomballoc");
763 if (test_opt(sb, I_VERSION))
764 seq_puts(seq, ",i_version");
765 if (!test_opt(sb, DELALLOC))
766 seq_puts(seq, ",nodelalloc");
770 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
772 * journal mode get enabled in different ways
773 * So just print the value even if we didn't specify it
775 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
776 seq_puts(seq, ",data=journal");
777 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
778 seq_puts(seq, ",data=ordered");
779 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
780 seq_puts(seq, ",data=writeback");
782 ext4_show_quota_options(seq, sb);
787 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
788 u64 ino, u32 generation)
792 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
793 return ERR_PTR(-ESTALE);
794 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
795 return ERR_PTR(-ESTALE);
797 /* iget isn't really right if the inode is currently unallocated!!
799 * ext4_read_inode will return a bad_inode if the inode had been
800 * deleted, so we should be safe.
802 * Currently we don't know the generation for parent directory, so
803 * a generation of 0 means "accept any"
805 inode = ext4_iget(sb, ino);
807 return ERR_CAST(inode);
808 if (generation && inode->i_generation != generation) {
810 return ERR_PTR(-ESTALE);
816 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
817 int fh_len, int fh_type)
819 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
823 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
824 int fh_len, int fh_type)
826 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
831 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
832 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
834 static int ext4_dquot_initialize(struct inode *inode, int type);
835 static int ext4_dquot_drop(struct inode *inode);
836 static int ext4_write_dquot(struct dquot *dquot);
837 static int ext4_acquire_dquot(struct dquot *dquot);
838 static int ext4_release_dquot(struct dquot *dquot);
839 static int ext4_mark_dquot_dirty(struct dquot *dquot);
840 static int ext4_write_info(struct super_block *sb, int type);
841 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
842 char *path, int remount);
843 static int ext4_quota_on_mount(struct super_block *sb, int type);
844 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
845 size_t len, loff_t off);
846 static ssize_t ext4_quota_write(struct super_block *sb, int type,
847 const char *data, size_t len, loff_t off);
849 static struct dquot_operations ext4_quota_operations = {
850 .initialize = ext4_dquot_initialize,
851 .drop = ext4_dquot_drop,
852 .alloc_space = dquot_alloc_space,
853 .alloc_inode = dquot_alloc_inode,
854 .free_space = dquot_free_space,
855 .free_inode = dquot_free_inode,
856 .transfer = dquot_transfer,
857 .write_dquot = ext4_write_dquot,
858 .acquire_dquot = ext4_acquire_dquot,
859 .release_dquot = ext4_release_dquot,
860 .mark_dirty = ext4_mark_dquot_dirty,
861 .write_info = ext4_write_info
864 static struct quotactl_ops ext4_qctl_operations = {
865 .quota_on = ext4_quota_on,
866 .quota_off = vfs_quota_off,
867 .quota_sync = vfs_quota_sync,
868 .get_info = vfs_get_dqinfo,
869 .set_info = vfs_set_dqinfo,
870 .get_dqblk = vfs_get_dqblk,
871 .set_dqblk = vfs_set_dqblk
875 static const struct super_operations ext4_sops = {
876 .alloc_inode = ext4_alloc_inode,
877 .destroy_inode = ext4_destroy_inode,
878 .write_inode = ext4_write_inode,
879 .dirty_inode = ext4_dirty_inode,
880 .delete_inode = ext4_delete_inode,
881 .put_super = ext4_put_super,
882 .write_super = ext4_write_super,
883 .sync_fs = ext4_sync_fs,
884 .write_super_lockfs = ext4_write_super_lockfs,
885 .unlockfs = ext4_unlockfs,
886 .statfs = ext4_statfs,
887 .remount_fs = ext4_remount,
888 .clear_inode = ext4_clear_inode,
889 .show_options = ext4_show_options,
891 .quota_read = ext4_quota_read,
892 .quota_write = ext4_quota_write,
896 static const struct export_operations ext4_export_ops = {
897 .fh_to_dentry = ext4_fh_to_dentry,
898 .fh_to_parent = ext4_fh_to_parent,
899 .get_parent = ext4_get_parent,
903 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
904 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
905 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
906 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
907 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
908 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
909 Opt_journal_checksum, Opt_journal_async_commit,
910 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
911 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
912 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
913 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
914 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
915 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc, Opt_nodelalloc,
918 static match_table_t tokens = {
919 {Opt_bsd_df, "bsddf"},
920 {Opt_minix_df, "minixdf"},
921 {Opt_grpid, "grpid"},
922 {Opt_grpid, "bsdgroups"},
923 {Opt_nogrpid, "nogrpid"},
924 {Opt_nogrpid, "sysvgroups"},
925 {Opt_resgid, "resgid=%u"},
926 {Opt_resuid, "resuid=%u"},
928 {Opt_err_cont, "errors=continue"},
929 {Opt_err_panic, "errors=panic"},
930 {Opt_err_ro, "errors=remount-ro"},
931 {Opt_nouid32, "nouid32"},
932 {Opt_nocheck, "nocheck"},
933 {Opt_nocheck, "check=none"},
934 {Opt_debug, "debug"},
935 {Opt_oldalloc, "oldalloc"},
936 {Opt_orlov, "orlov"},
937 {Opt_user_xattr, "user_xattr"},
938 {Opt_nouser_xattr, "nouser_xattr"},
940 {Opt_noacl, "noacl"},
941 {Opt_reservation, "reservation"},
942 {Opt_noreservation, "noreservation"},
943 {Opt_noload, "noload"},
946 {Opt_commit, "commit=%u"},
947 {Opt_journal_update, "journal=update"},
948 {Opt_journal_inum, "journal=%u"},
949 {Opt_journal_dev, "journal_dev=%u"},
950 {Opt_journal_checksum, "journal_checksum"},
951 {Opt_journal_async_commit, "journal_async_commit"},
952 {Opt_abort, "abort"},
953 {Opt_data_journal, "data=journal"},
954 {Opt_data_ordered, "data=ordered"},
955 {Opt_data_writeback, "data=writeback"},
956 {Opt_offusrjquota, "usrjquota="},
957 {Opt_usrjquota, "usrjquota=%s"},
958 {Opt_offgrpjquota, "grpjquota="},
959 {Opt_grpjquota, "grpjquota=%s"},
960 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
961 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
962 {Opt_grpquota, "grpquota"},
963 {Opt_noquota, "noquota"},
964 {Opt_quota, "quota"},
965 {Opt_usrquota, "usrquota"},
966 {Opt_barrier, "barrier=%u"},
967 {Opt_extents, "extents"},
968 {Opt_noextents, "noextents"},
969 {Opt_i_version, "i_version"},
970 {Opt_mballoc, "mballoc"},
971 {Opt_nomballoc, "nomballoc"},
972 {Opt_stripe, "stripe=%u"},
973 {Opt_resize, "resize"},
974 {Opt_delalloc, "delalloc"},
975 {Opt_nodelalloc, "nodelalloc"},
979 static ext4_fsblk_t get_sb_block(void **data)
981 ext4_fsblk_t sb_block;
982 char *options = (char *) *data;
984 if (!options || strncmp(options, "sb=", 3) != 0)
985 return 1; /* Default location */
987 /*todo: use simple_strtoll with >32bit ext4 */
988 sb_block = simple_strtoul(options, &options, 0);
989 if (*options && *options != ',') {
990 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
996 *data = (void *) options;
1000 static int parse_options(char *options, struct super_block *sb,
1001 unsigned int *inum, unsigned long *journal_devnum,
1002 ext4_fsblk_t *n_blocks_count, int is_remount)
1004 struct ext4_sb_info *sbi = EXT4_SB(sb);
1006 substring_t args[MAX_OPT_ARGS];
1013 ext4_fsblk_t last_block;
1018 while ((p = strsep(&options, ",")) != NULL) {
1023 token = match_token(p, tokens, args);
1026 clear_opt(sbi->s_mount_opt, MINIX_DF);
1029 set_opt(sbi->s_mount_opt, MINIX_DF);
1032 set_opt(sbi->s_mount_opt, GRPID);
1035 clear_opt(sbi->s_mount_opt, GRPID);
1038 if (match_int(&args[0], &option))
1040 sbi->s_resuid = option;
1043 if (match_int(&args[0], &option))
1045 sbi->s_resgid = option;
1048 /* handled by get_sb_block() instead of here */
1049 /* *sb_block = match_int(&args[0]); */
1052 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1053 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1054 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1057 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1058 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1059 set_opt(sbi->s_mount_opt, ERRORS_RO);
1062 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1063 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1064 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1067 set_opt(sbi->s_mount_opt, NO_UID32);
1070 clear_opt(sbi->s_mount_opt, CHECK);
1073 set_opt(sbi->s_mount_opt, DEBUG);
1076 set_opt(sbi->s_mount_opt, OLDALLOC);
1079 clear_opt(sbi->s_mount_opt, OLDALLOC);
1081 #ifdef CONFIG_EXT4DEV_FS_XATTR
1082 case Opt_user_xattr:
1083 set_opt(sbi->s_mount_opt, XATTR_USER);
1085 case Opt_nouser_xattr:
1086 clear_opt(sbi->s_mount_opt, XATTR_USER);
1089 case Opt_user_xattr:
1090 case Opt_nouser_xattr:
1091 printk(KERN_ERR "EXT4 (no)user_xattr options "
1095 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1097 set_opt(sbi->s_mount_opt, POSIX_ACL);
1100 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1105 printk(KERN_ERR "EXT4 (no)acl options "
1109 case Opt_reservation:
1110 set_opt(sbi->s_mount_opt, RESERVATION);
1112 case Opt_noreservation:
1113 clear_opt(sbi->s_mount_opt, RESERVATION);
1115 case Opt_journal_update:
1117 /* Eventually we will want to be able to create
1118 a journal file here. For now, only allow the
1119 user to specify an existing inode to be the
1122 printk(KERN_ERR "EXT4-fs: cannot specify "
1123 "journal on remount\n");
1126 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1128 case Opt_journal_inum:
1130 printk(KERN_ERR "EXT4-fs: cannot specify "
1131 "journal on remount\n");
1134 if (match_int(&args[0], &option))
1138 case Opt_journal_dev:
1140 printk(KERN_ERR "EXT4-fs: cannot specify "
1141 "journal on remount\n");
1144 if (match_int(&args[0], &option))
1146 *journal_devnum = option;
1148 case Opt_journal_checksum:
1149 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1151 case Opt_journal_async_commit:
1152 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1153 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1156 set_opt(sbi->s_mount_opt, NOLOAD);
1159 if (match_int(&args[0], &option))
1164 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1165 sbi->s_commit_interval = HZ * option;
1167 case Opt_data_journal:
1168 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1170 case Opt_data_ordered:
1171 data_opt = EXT4_MOUNT_ORDERED_DATA;
1173 case Opt_data_writeback:
1174 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1177 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1180 "EXT4-fs: cannot change data "
1181 "mode on remount\n");
1185 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1186 sbi->s_mount_opt |= data_opt;
1196 if ((sb_any_quota_enabled(sb) ||
1197 sb_any_quota_suspended(sb)) &&
1198 !sbi->s_qf_names[qtype]) {
1200 "EXT4-fs: Cannot change journaled "
1201 "quota options when quota turned on.\n");
1204 qname = match_strdup(&args[0]);
1207 "EXT4-fs: not enough memory for "
1208 "storing quotafile name.\n");
1211 if (sbi->s_qf_names[qtype] &&
1212 strcmp(sbi->s_qf_names[qtype], qname)) {
1214 "EXT4-fs: %s quota file already "
1215 "specified.\n", QTYPE2NAME(qtype));
1219 sbi->s_qf_names[qtype] = qname;
1220 if (strchr(sbi->s_qf_names[qtype], '/')) {
1222 "EXT4-fs: quotafile must be on "
1223 "filesystem root.\n");
1224 kfree(sbi->s_qf_names[qtype]);
1225 sbi->s_qf_names[qtype] = NULL;
1228 set_opt(sbi->s_mount_opt, QUOTA);
1230 case Opt_offusrjquota:
1233 case Opt_offgrpjquota:
1236 if ((sb_any_quota_enabled(sb) ||
1237 sb_any_quota_suspended(sb)) &&
1238 sbi->s_qf_names[qtype]) {
1239 printk(KERN_ERR "EXT4-fs: Cannot change "
1240 "journaled quota options when "
1241 "quota turned on.\n");
1245 * The space will be released later when all options
1246 * are confirmed to be correct
1248 sbi->s_qf_names[qtype] = NULL;
1250 case Opt_jqfmt_vfsold:
1251 qfmt = QFMT_VFS_OLD;
1253 case Opt_jqfmt_vfsv0:
1256 if ((sb_any_quota_enabled(sb) ||
1257 sb_any_quota_suspended(sb)) &&
1258 sbi->s_jquota_fmt != qfmt) {
1259 printk(KERN_ERR "EXT4-fs: Cannot change "
1260 "journaled quota options when "
1261 "quota turned on.\n");
1264 sbi->s_jquota_fmt = qfmt;
1268 set_opt(sbi->s_mount_opt, QUOTA);
1269 set_opt(sbi->s_mount_opt, USRQUOTA);
1272 set_opt(sbi->s_mount_opt, QUOTA);
1273 set_opt(sbi->s_mount_opt, GRPQUOTA);
1276 if (sb_any_quota_enabled(sb)) {
1277 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1278 "options when quota turned on.\n");
1281 clear_opt(sbi->s_mount_opt, QUOTA);
1282 clear_opt(sbi->s_mount_opt, USRQUOTA);
1283 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1290 "EXT4-fs: quota options not supported.\n");
1294 case Opt_offusrjquota:
1295 case Opt_offgrpjquota:
1296 case Opt_jqfmt_vfsold:
1297 case Opt_jqfmt_vfsv0:
1299 "EXT4-fs: journaled quota options not "
1306 set_opt(sbi->s_mount_opt, ABORT);
1309 if (match_int(&args[0], &option))
1312 set_opt(sbi->s_mount_opt, BARRIER);
1314 clear_opt(sbi->s_mount_opt, BARRIER);
1320 printk("EXT4-fs: resize option only available "
1324 if (match_int(&args[0], &option) != 0)
1326 *n_blocks_count = option;
1329 set_opt(sbi->s_mount_opt, NOBH);
1332 clear_opt(sbi->s_mount_opt, NOBH);
1335 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1336 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1337 ext4_warning(sb, __func__,
1338 "extents feature not enabled "
1339 "on this filesystem, use tune2fs\n");
1342 set_opt(sbi->s_mount_opt, EXTENTS);
1346 * When e2fsprogs support resizing an already existing
1347 * ext3 file system to greater than 2**32 we need to
1348 * add support to block allocator to handle growing
1349 * already existing block mapped inode so that blocks
1350 * allocated for them fall within 2**32
1352 last_block = ext4_blocks_count(sbi->s_es) - 1;
1353 if (last_block > 0xffffffffULL) {
1354 printk(KERN_ERR "EXT4-fs: Filesystem too "
1355 "large to mount with "
1356 "-o noextents options\n");
1359 clear_opt(sbi->s_mount_opt, EXTENTS);
1362 set_opt(sbi->s_mount_opt, I_VERSION);
1363 sb->s_flags |= MS_I_VERSION;
1365 case Opt_nodelalloc:
1366 clear_opt(sbi->s_mount_opt, DELALLOC);
1369 set_opt(sbi->s_mount_opt, MBALLOC);
1372 clear_opt(sbi->s_mount_opt, MBALLOC);
1375 if (match_int(&args[0], &option))
1379 sbi->s_stripe = option;
1382 set_opt(sbi->s_mount_opt, DELALLOC);
1386 "EXT4-fs: Unrecognized mount option \"%s\" "
1387 "or missing value\n", p);
1392 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1393 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1394 sbi->s_qf_names[USRQUOTA])
1395 clear_opt(sbi->s_mount_opt, USRQUOTA);
1397 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1398 sbi->s_qf_names[GRPQUOTA])
1399 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1401 if ((sbi->s_qf_names[USRQUOTA] &&
1402 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1403 (sbi->s_qf_names[GRPQUOTA] &&
1404 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1405 printk(KERN_ERR "EXT4-fs: old and new quota "
1406 "format mixing.\n");
1410 if (!sbi->s_jquota_fmt) {
1411 printk(KERN_ERR "EXT4-fs: journaled quota format "
1412 "not specified.\n");
1416 if (sbi->s_jquota_fmt) {
1417 printk(KERN_ERR "EXT4-fs: journaled quota format "
1418 "specified with no journaling "
1427 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1430 struct ext4_sb_info *sbi = EXT4_SB(sb);
1433 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1434 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1435 "forcing read-only mode\n");
1440 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1441 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1442 "running e2fsck is recommended\n");
1443 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1445 "EXT4-fs warning: mounting fs with errors, "
1446 "running e2fsck is recommended\n");
1447 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1448 le16_to_cpu(es->s_mnt_count) >=
1449 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1451 "EXT4-fs warning: maximal mount count reached, "
1452 "running e2fsck is recommended\n");
1453 else if (le32_to_cpu(es->s_checkinterval) &&
1454 (le32_to_cpu(es->s_lastcheck) +
1455 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1457 "EXT4-fs warning: checktime reached, "
1458 "running e2fsck is recommended\n");
1460 /* @@@ We _will_ want to clear the valid bit if we find
1461 * inconsistencies, to force a fsck at reboot. But for
1462 * a plain journaled filesystem we can keep it set as
1465 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1467 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1468 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1469 le16_add_cpu(&es->s_mnt_count, 1);
1470 es->s_mtime = cpu_to_le32(get_seconds());
1471 ext4_update_dynamic_rev(sb);
1472 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1474 ext4_commit_super(sb, es, 1);
1475 if (test_opt(sb, DEBUG))
1476 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1477 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1479 sbi->s_groups_count,
1480 EXT4_BLOCKS_PER_GROUP(sb),
1481 EXT4_INODES_PER_GROUP(sb),
1484 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1485 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1486 "external", EXT4_SB(sb)->s_journal->j_devname);
1490 static int ext4_fill_flex_info(struct super_block *sb)
1492 struct ext4_sb_info *sbi = EXT4_SB(sb);
1493 struct ext4_group_desc *gdp = NULL;
1494 struct buffer_head *bh;
1495 ext4_group_t flex_group_count;
1496 ext4_group_t flex_group;
1497 int groups_per_flex = 0;
1498 __u64 block_bitmap = 0;
1501 if (!sbi->s_es->s_log_groups_per_flex) {
1502 sbi->s_log_groups_per_flex = 0;
1506 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1507 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1509 /* We allocate both existing and potentially added groups */
1510 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1511 ((sbi->s_es->s_reserved_gdt_blocks +1 ) <<
1512 EXT4_DESC_PER_BLOCK_BITS(sb))) /
1514 sbi->s_flex_groups = kzalloc(flex_group_count *
1515 sizeof(struct flex_groups), GFP_KERNEL);
1516 if (sbi->s_flex_groups == NULL) {
1517 printk(KERN_ERR "EXT4-fs: not enough memory for "
1518 "%lu flex groups\n", flex_group_count);
1522 gdp = ext4_get_group_desc(sb, 1, &bh);
1523 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1525 for (i = 0; i < sbi->s_groups_count; i++) {
1526 gdp = ext4_get_group_desc(sb, i, &bh);
1528 flex_group = ext4_flex_group(sbi, i);
1529 sbi->s_flex_groups[flex_group].free_inodes +=
1530 le16_to_cpu(gdp->bg_free_inodes_count);
1531 sbi->s_flex_groups[flex_group].free_blocks +=
1532 le16_to_cpu(gdp->bg_free_blocks_count);
1540 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1541 struct ext4_group_desc *gdp)
1545 if (sbi->s_es->s_feature_ro_compat &
1546 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1547 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1548 __le32 le_group = cpu_to_le32(block_group);
1550 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1551 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1552 crc = crc16(crc, (__u8 *)gdp, offset);
1553 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1554 /* for checksum of struct ext4_group_desc do the rest...*/
1555 if ((sbi->s_es->s_feature_incompat &
1556 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1557 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1558 crc = crc16(crc, (__u8 *)gdp + offset,
1559 le16_to_cpu(sbi->s_es->s_desc_size) -
1563 return cpu_to_le16(crc);
1566 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1567 struct ext4_group_desc *gdp)
1569 if ((sbi->s_es->s_feature_ro_compat &
1570 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1571 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1577 /* Called at mount-time, super-block is locked */
1578 static int ext4_check_descriptors(struct super_block *sb)
1580 struct ext4_sb_info *sbi = EXT4_SB(sb);
1581 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1582 ext4_fsblk_t last_block;
1583 ext4_fsblk_t block_bitmap;
1584 ext4_fsblk_t inode_bitmap;
1585 ext4_fsblk_t inode_table;
1586 int flexbg_flag = 0;
1589 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1592 ext4_debug("Checking group descriptors");
1594 for (i = 0; i < sbi->s_groups_count; i++) {
1595 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1597 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1598 last_block = ext4_blocks_count(sbi->s_es) - 1;
1600 last_block = first_block +
1601 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1603 block_bitmap = ext4_block_bitmap(sb, gdp);
1604 if (block_bitmap < first_block || block_bitmap > last_block) {
1605 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1606 "Block bitmap for group %lu not in group "
1607 "(block %llu)!", i, block_bitmap);
1610 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1611 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1612 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1613 "Inode bitmap for group %lu not in group "
1614 "(block %llu)!", i, inode_bitmap);
1617 inode_table = ext4_inode_table(sb, gdp);
1618 if (inode_table < first_block ||
1619 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1620 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1621 "Inode table for group %lu not in group "
1622 "(block %llu)!", i, inode_table);
1625 spin_lock(sb_bgl_lock(sbi, i));
1626 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1627 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1628 "Checksum for group %lu failed (%u!=%u)\n",
1629 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1630 gdp)), le16_to_cpu(gdp->bg_checksum));
1631 if (!(sb->s_flags & MS_RDONLY)) {
1632 spin_unlock(sb_bgl_lock(sbi, i));
1636 spin_unlock(sb_bgl_lock(sbi, i));
1638 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1641 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1642 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1646 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1647 * the superblock) which were deleted from all directories, but held open by
1648 * a process at the time of a crash. We walk the list and try to delete these
1649 * inodes at recovery time (only with a read-write filesystem).
1651 * In order to keep the orphan inode chain consistent during traversal (in
1652 * case of crash during recovery), we link each inode into the superblock
1653 * orphan list_head and handle it the same way as an inode deletion during
1654 * normal operation (which journals the operations for us).
1656 * We only do an iget() and an iput() on each inode, which is very safe if we
1657 * accidentally point at an in-use or already deleted inode. The worst that
1658 * can happen in this case is that we get a "bit already cleared" message from
1659 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1660 * e2fsck was run on this filesystem, and it must have already done the orphan
1661 * inode cleanup for us, so we can safely abort without any further action.
1663 static void ext4_orphan_cleanup(struct super_block *sb,
1664 struct ext4_super_block *es)
1666 unsigned int s_flags = sb->s_flags;
1667 int nr_orphans = 0, nr_truncates = 0;
1671 if (!es->s_last_orphan) {
1672 jbd_debug(4, "no orphan inodes to clean up\n");
1676 if (bdev_read_only(sb->s_bdev)) {
1677 printk(KERN_ERR "EXT4-fs: write access "
1678 "unavailable, skipping orphan cleanup.\n");
1682 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1683 if (es->s_last_orphan)
1684 jbd_debug(1, "Errors on filesystem, "
1685 "clearing orphan list.\n");
1686 es->s_last_orphan = 0;
1687 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1691 if (s_flags & MS_RDONLY) {
1692 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1694 sb->s_flags &= ~MS_RDONLY;
1697 /* Needed for iput() to work correctly and not trash data */
1698 sb->s_flags |= MS_ACTIVE;
1699 /* Turn on quotas so that they are updated correctly */
1700 for (i = 0; i < MAXQUOTAS; i++) {
1701 if (EXT4_SB(sb)->s_qf_names[i]) {
1702 int ret = ext4_quota_on_mount(sb, i);
1705 "EXT4-fs: Cannot turn on journaled "
1706 "quota: error %d\n", ret);
1711 while (es->s_last_orphan) {
1712 struct inode *inode;
1714 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1715 if (IS_ERR(inode)) {
1716 es->s_last_orphan = 0;
1720 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1722 if (inode->i_nlink) {
1724 "%s: truncating inode %lu to %lld bytes\n",
1725 __func__, inode->i_ino, inode->i_size);
1726 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1727 inode->i_ino, inode->i_size);
1728 ext4_truncate(inode);
1732 "%s: deleting unreferenced inode %lu\n",
1733 __func__, inode->i_ino);
1734 jbd_debug(2, "deleting unreferenced inode %lu\n",
1738 iput(inode); /* The delete magic happens here! */
1741 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1744 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1745 sb->s_id, PLURAL(nr_orphans));
1747 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1748 sb->s_id, PLURAL(nr_truncates));
1750 /* Turn quotas off */
1751 for (i = 0; i < MAXQUOTAS; i++) {
1752 if (sb_dqopt(sb)->files[i])
1753 vfs_quota_off(sb, i, 0);
1756 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1759 * Maximal extent format file size.
1760 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1761 * extent format containers, within a sector_t, and within i_blocks
1762 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1763 * so that won't be a limiting factor.
1765 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1767 static loff_t ext4_max_size(int blkbits)
1770 loff_t upper_limit = MAX_LFS_FILESIZE;
1772 /* small i_blocks in vfs inode? */
1773 if (sizeof(blkcnt_t) < sizeof(u64)) {
1775 * CONFIG_LSF is not enabled implies the inode
1776 * i_block represent total blocks in 512 bytes
1777 * 32 == size of vfs inode i_blocks * 8
1779 upper_limit = (1LL << 32) - 1;
1781 /* total blocks in file system block size */
1782 upper_limit >>= (blkbits - 9);
1783 upper_limit <<= blkbits;
1786 /* 32-bit extent-start container, ee_block */
1791 /* Sanity check against vm- & vfs- imposed limits */
1792 if (res > upper_limit)
1799 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1800 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1801 * We need to be 1 filesystem block less than the 2^48 sector limit.
1803 static loff_t ext4_max_bitmap_size(int bits)
1805 loff_t res = EXT4_NDIR_BLOCKS;
1808 /* This is calculated to be the largest file size for a
1809 * dense, bitmapped file such that the total number of
1810 * sectors in the file, including data and all indirect blocks,
1811 * does not exceed 2^48 -1
1812 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1813 * total number of 512 bytes blocks of the file
1816 if (sizeof(blkcnt_t) < sizeof(u64)) {
1818 * CONFIG_LSF is not enabled implies the inode
1819 * i_block represent total blocks in 512 bytes
1820 * 32 == size of vfs inode i_blocks * 8
1822 upper_limit = (1LL << 32) - 1;
1824 /* total blocks in file system block size */
1825 upper_limit >>= (bits - 9);
1829 * We use 48 bit ext4_inode i_blocks
1830 * With EXT4_HUGE_FILE_FL set the i_blocks
1831 * represent total number of blocks in
1832 * file system block size
1834 upper_limit = (1LL << 48) - 1;
1838 /* indirect blocks */
1840 /* double indirect blocks */
1841 meta_blocks += 1 + (1LL << (bits-2));
1842 /* tripple indirect blocks */
1843 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1845 upper_limit -= meta_blocks;
1846 upper_limit <<= bits;
1848 res += 1LL << (bits-2);
1849 res += 1LL << (2*(bits-2));
1850 res += 1LL << (3*(bits-2));
1852 if (res > upper_limit)
1855 if (res > MAX_LFS_FILESIZE)
1856 res = MAX_LFS_FILESIZE;
1861 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1862 ext4_fsblk_t logical_sb_block, int nr)
1864 struct ext4_sb_info *sbi = EXT4_SB(sb);
1865 ext4_group_t bg, first_meta_bg;
1868 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1870 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1872 return logical_sb_block + nr + 1;
1873 bg = sbi->s_desc_per_block * nr;
1874 if (ext4_bg_has_super(sb, bg))
1876 return (has_super + ext4_group_first_block_no(sb, bg));
1880 * ext4_get_stripe_size: Get the stripe size.
1881 * @sbi: In memory super block info
1883 * If we have specified it via mount option, then
1884 * use the mount option value. If the value specified at mount time is
1885 * greater than the blocks per group use the super block value.
1886 * If the super block value is greater than blocks per group return 0.
1887 * Allocator needs it be less than blocks per group.
1890 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1892 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1893 unsigned long stripe_width =
1894 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1896 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1897 return sbi->s_stripe;
1899 if (stripe_width <= sbi->s_blocks_per_group)
1900 return stripe_width;
1902 if (stride <= sbi->s_blocks_per_group)
1908 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1909 __releases(kernel_lock)
1910 __acquires(kernel_lock)
1913 struct buffer_head *bh;
1914 struct ext4_super_block *es = NULL;
1915 struct ext4_sb_info *sbi;
1917 ext4_fsblk_t sb_block = get_sb_block(&data);
1918 ext4_fsblk_t logical_sb_block;
1919 unsigned long offset = 0;
1920 unsigned int journal_inum = 0;
1921 unsigned long journal_devnum = 0;
1922 unsigned long def_mount_opts;
1934 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1937 sb->s_fs_info = sbi;
1938 sbi->s_mount_opt = 0;
1939 sbi->s_resuid = EXT4_DEF_RESUID;
1940 sbi->s_resgid = EXT4_DEF_RESGID;
1941 sbi->s_sb_block = sb_block;
1945 /* Cleanup superblock name */
1946 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1949 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1951 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1956 * The ext4 superblock will not be buffer aligned for other than 1kB
1957 * block sizes. We need to calculate the offset from buffer start.
1959 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1960 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1961 offset = do_div(logical_sb_block, blocksize);
1963 logical_sb_block = sb_block;
1966 if (!(bh = sb_bread(sb, logical_sb_block))) {
1967 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
1971 * Note: s_es must be initialized as soon as possible because
1972 * some ext4 macro-instructions depend on its value
1974 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1976 sb->s_magic = le16_to_cpu(es->s_magic);
1977 if (sb->s_magic != EXT4_SUPER_MAGIC)
1980 /* Set defaults before we parse the mount options */
1981 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1982 if (def_mount_opts & EXT4_DEFM_DEBUG)
1983 set_opt(sbi->s_mount_opt, DEBUG);
1984 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1985 set_opt(sbi->s_mount_opt, GRPID);
1986 if (def_mount_opts & EXT4_DEFM_UID16)
1987 set_opt(sbi->s_mount_opt, NO_UID32);
1988 #ifdef CONFIG_EXT4DEV_FS_XATTR
1989 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1990 set_opt(sbi->s_mount_opt, XATTR_USER);
1992 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1993 if (def_mount_opts & EXT4_DEFM_ACL)
1994 set_opt(sbi->s_mount_opt, POSIX_ACL);
1996 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1997 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1998 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1999 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2000 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2001 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2003 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2004 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2005 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2006 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2008 set_opt(sbi->s_mount_opt, ERRORS_RO);
2010 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2011 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2013 set_opt(sbi->s_mount_opt, RESERVATION);
2014 set_opt(sbi->s_mount_opt, BARRIER);
2017 * turn on extents feature by default in ext4 filesystem
2018 * only if feature flag already set by mkfs or tune2fs.
2019 * Use -o noextents to turn it off
2021 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2022 set_opt(sbi->s_mount_opt, EXTENTS);
2024 ext4_warning(sb, __func__,
2025 "extents feature not enabled on this filesystem, "
2028 * turn on mballoc code by default in ext4 filesystem
2029 * Use -o nomballoc to turn it off
2031 set_opt(sbi->s_mount_opt, MBALLOC);
2034 * enable delayed allocation by default
2035 * Use -o nodelalloc to turn it off
2037 set_opt(sbi->s_mount_opt, DELALLOC);
2040 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2044 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2045 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2047 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2048 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2049 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2050 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2052 "EXT4-fs warning: feature flags set on rev 0 fs, "
2053 "running e2fsck is recommended\n");
2056 * Since ext4 is still considered development code, we require
2057 * that the TEST_FILESYS flag in s->flags be set.
2059 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2060 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2061 "OK to use with test code.\n", sb->s_id);
2066 * Check feature flags regardless of the revision level, since we
2067 * previously didn't change the revision level when setting the flags,
2068 * so there is a chance incompat flags are set on a rev 0 filesystem.
2070 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2072 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2073 "unsupported optional features (%x).\n",
2074 sb->s_id, le32_to_cpu(features));
2077 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2078 if (!(sb->s_flags & MS_RDONLY) && features) {
2079 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2080 "unsupported optional features (%x).\n",
2081 sb->s_id, le32_to_cpu(features));
2084 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2086 * Large file size enabled file system can only be
2087 * mount if kernel is build with CONFIG_LSF
2089 if (sizeof(root->i_blocks) < sizeof(u64) &&
2090 !(sb->s_flags & MS_RDONLY)) {
2091 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2092 "files cannot be mounted read-write "
2093 "without CONFIG_LSF.\n", sb->s_id);
2097 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2099 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2100 blocksize > EXT4_MAX_BLOCK_SIZE) {
2102 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2103 blocksize, sb->s_id);
2107 if (sb->s_blocksize != blocksize) {
2109 /* Validate the filesystem blocksize */
2110 if (!sb_set_blocksize(sb, blocksize)) {
2111 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2117 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2118 offset = do_div(logical_sb_block, blocksize);
2119 bh = sb_bread(sb, logical_sb_block);
2122 "EXT4-fs: Can't read superblock on 2nd try.\n");
2125 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2127 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2129 "EXT4-fs: Magic mismatch, very weird !\n");
2134 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2135 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2137 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2138 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2139 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2141 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2142 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2143 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2144 (!is_power_of_2(sbi->s_inode_size)) ||
2145 (sbi->s_inode_size > blocksize)) {
2147 "EXT4-fs: unsupported inode size: %d\n",
2151 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2152 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2154 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2155 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2156 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2157 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2158 !is_power_of_2(sbi->s_desc_size)) {
2160 "EXT4-fs: unsupported descriptor size %lu\n",
2165 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2166 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2167 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2168 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2170 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2171 if (sbi->s_inodes_per_block == 0)
2173 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2174 sbi->s_inodes_per_block;
2175 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2177 sbi->s_mount_state = le16_to_cpu(es->s_state);
2178 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2179 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2180 for (i = 0; i < 4; i++)
2181 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2182 sbi->s_def_hash_version = es->s_def_hash_version;
2184 if (sbi->s_blocks_per_group > blocksize * 8) {
2186 "EXT4-fs: #blocks per group too big: %lu\n",
2187 sbi->s_blocks_per_group);
2190 if (sbi->s_inodes_per_group > blocksize * 8) {
2192 "EXT4-fs: #inodes per group too big: %lu\n",
2193 sbi->s_inodes_per_group);
2197 if (ext4_blocks_count(es) >
2198 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2199 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2200 " too large to mount safely\n", sb->s_id);
2201 if (sizeof(sector_t) < 8)
2202 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2207 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2210 /* ensure blocks_count calculation below doesn't sign-extend */
2211 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2212 le32_to_cpu(es->s_first_data_block) + 1) {
2213 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2214 "first data block %u, blocks per group %lu\n",
2215 ext4_blocks_count(es),
2216 le32_to_cpu(es->s_first_data_block),
2217 EXT4_BLOCKS_PER_GROUP(sb));
2220 blocks_count = (ext4_blocks_count(es) -
2221 le32_to_cpu(es->s_first_data_block) +
2222 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2223 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2224 sbi->s_groups_count = blocks_count;
2225 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2226 EXT4_DESC_PER_BLOCK(sb);
2227 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2229 if (sbi->s_group_desc == NULL) {
2230 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2235 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2237 bgl_lock_init(&sbi->s_blockgroup_lock);
2239 for (i = 0; i < db_count; i++) {
2240 block = descriptor_loc(sb, logical_sb_block, i);
2241 sbi->s_group_desc[i] = sb_bread(sb, block);
2242 if (!sbi->s_group_desc[i]) {
2243 printk(KERN_ERR "EXT4-fs: "
2244 "can't read group descriptor %d\n", i);
2249 if (!ext4_check_descriptors(sb)) {
2250 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2253 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2254 if (!ext4_fill_flex_info(sb)) {
2256 "EXT4-fs: unable to initialize "
2257 "flex_bg meta info!\n");
2261 sbi->s_gdb_count = db_count;
2262 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2263 spin_lock_init(&sbi->s_next_gen_lock);
2265 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2266 ext4_count_free_blocks(sb));
2268 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2269 ext4_count_free_inodes(sb));
2272 err = percpu_counter_init(&sbi->s_dirs_counter,
2273 ext4_count_dirs(sb));
2276 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2279 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2283 /* per fileystem reservation list head & lock */
2284 spin_lock_init(&sbi->s_rsv_window_lock);
2285 sbi->s_rsv_window_root = RB_ROOT;
2286 /* Add a single, static dummy reservation to the start of the
2287 * reservation window list --- it gives us a placeholder for
2288 * append-at-start-of-list which makes the allocation logic
2289 * _much_ simpler. */
2290 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2291 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2292 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2293 sbi->s_rsv_window_head.rsv_goal_size = 0;
2294 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2296 sbi->s_stripe = ext4_get_stripe_size(sbi);
2299 * set up enough so that it can read an inode
2301 sb->s_op = &ext4_sops;
2302 sb->s_export_op = &ext4_export_ops;
2303 sb->s_xattr = ext4_xattr_handlers;
2305 sb->s_qcop = &ext4_qctl_operations;
2306 sb->dq_op = &ext4_quota_operations;
2308 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2312 needs_recovery = (es->s_last_orphan != 0 ||
2313 EXT4_HAS_INCOMPAT_FEATURE(sb,
2314 EXT4_FEATURE_INCOMPAT_RECOVER));
2317 * The first inode we look at is the journal inode. Don't try
2318 * root first: it may be modified in the journal!
2320 if (!test_opt(sb, NOLOAD) &&
2321 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2322 if (ext4_load_journal(sb, es, journal_devnum))
2324 if (!(sb->s_flags & MS_RDONLY) &&
2325 EXT4_SB(sb)->s_journal->j_failed_commit) {
2326 printk(KERN_CRIT "EXT4-fs error (device %s): "
2327 "ext4_fill_super: Journal transaction "
2328 "%u is corrupt\n", sb->s_id,
2329 EXT4_SB(sb)->s_journal->j_failed_commit);
2330 if (test_opt(sb, ERRORS_RO)) {
2332 "Mounting filesystem read-only\n");
2333 sb->s_flags |= MS_RDONLY;
2334 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2335 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2337 if (test_opt(sb, ERRORS_PANIC)) {
2338 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2339 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2340 ext4_commit_super(sb, es, 1);
2342 "EXT4-fs (device %s): mount failed\n",
2347 } else if (journal_inum) {
2348 if (ext4_create_journal(sb, es, journal_inum))
2353 "ext4: No journal on filesystem on %s\n",
2358 if (ext4_blocks_count(es) > 0xffffffffULL &&
2359 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2360 JBD2_FEATURE_INCOMPAT_64BIT)) {
2361 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2365 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2366 jbd2_journal_set_features(sbi->s_journal,
2367 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2368 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2369 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2370 jbd2_journal_set_features(sbi->s_journal,
2371 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2372 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2373 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2375 jbd2_journal_clear_features(sbi->s_journal,
2376 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2377 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2380 /* We have now updated the journal if required, so we can
2381 * validate the data journaling mode. */
2382 switch (test_opt(sb, DATA_FLAGS)) {
2384 /* No mode set, assume a default based on the journal
2385 * capabilities: ORDERED_DATA if the journal can
2386 * cope, else JOURNAL_DATA
2388 if (jbd2_journal_check_available_features
2389 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2390 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2392 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2395 case EXT4_MOUNT_ORDERED_DATA:
2396 case EXT4_MOUNT_WRITEBACK_DATA:
2397 if (!jbd2_journal_check_available_features
2398 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2399 printk(KERN_ERR "EXT4-fs: Journal does not support "
2400 "requested data journaling mode\n");
2407 if (test_opt(sb, NOBH)) {
2408 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2409 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2410 "its supported only with writeback mode\n");
2411 clear_opt(sbi->s_mount_opt, NOBH);
2415 * The jbd2_journal_load will have done any necessary log recovery,
2416 * so we can safely mount the rest of the filesystem now.
2419 root = ext4_iget(sb, EXT4_ROOT_INO);
2421 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2422 ret = PTR_ERR(root);
2425 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2427 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2430 sb->s_root = d_alloc_root(root);
2432 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2438 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2440 /* determine the minimum size of new large inodes, if present */
2441 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2442 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2443 EXT4_GOOD_OLD_INODE_SIZE;
2444 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2445 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2446 if (sbi->s_want_extra_isize <
2447 le16_to_cpu(es->s_want_extra_isize))
2448 sbi->s_want_extra_isize =
2449 le16_to_cpu(es->s_want_extra_isize);
2450 if (sbi->s_want_extra_isize <
2451 le16_to_cpu(es->s_min_extra_isize))
2452 sbi->s_want_extra_isize =
2453 le16_to_cpu(es->s_min_extra_isize);
2456 /* Check if enough inode space is available */
2457 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2458 sbi->s_inode_size) {
2459 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2460 EXT4_GOOD_OLD_INODE_SIZE;
2461 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2466 * akpm: core read_super() calls in here with the superblock locked.
2467 * That deadlocks, because orphan cleanup needs to lock the superblock
2468 * in numerous places. Here we just pop the lock - it's relatively
2469 * harmless, because we are now ready to accept write_super() requests,
2470 * and aviro says that's the only reason for hanging onto the
2473 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2474 ext4_orphan_cleanup(sb, es);
2475 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2477 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2478 ext4_mark_recovery_complete(sb, es);
2479 printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2480 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2481 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2484 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2485 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2486 "requested data journaling mode\n");
2487 clear_opt(sbi->s_mount_opt, DELALLOC);
2488 } else if (test_opt(sb, DELALLOC))
2489 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2492 ext4_mb_init(sb, needs_recovery);
2499 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2504 jbd2_journal_destroy(sbi->s_journal);
2505 sbi->s_journal = NULL;
2507 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2508 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2509 percpu_counter_destroy(&sbi->s_dirs_counter);
2510 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2512 for (i = 0; i < db_count; i++)
2513 brelse(sbi->s_group_desc[i]);
2514 kfree(sbi->s_group_desc);
2517 remove_proc_entry(sb->s_id, ext4_proc_root);
2519 for (i = 0; i < MAXQUOTAS; i++)
2520 kfree(sbi->s_qf_names[i]);
2522 ext4_blkdev_remove(sbi);
2525 sb->s_fs_info = NULL;
2532 * Setup any per-fs journal parameters now. We'll do this both on
2533 * initial mount, once the journal has been initialised but before we've
2534 * done any recovery; and again on any subsequent remount.
2536 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2538 struct ext4_sb_info *sbi = EXT4_SB(sb);
2540 if (sbi->s_commit_interval)
2541 journal->j_commit_interval = sbi->s_commit_interval;
2542 /* We could also set up an ext4-specific default for the commit
2543 * interval here, but for now we'll just fall back to the jbd
2546 spin_lock(&journal->j_state_lock);
2547 if (test_opt(sb, BARRIER))
2548 journal->j_flags |= JBD2_BARRIER;
2550 journal->j_flags &= ~JBD2_BARRIER;
2551 spin_unlock(&journal->j_state_lock);
2554 static journal_t *ext4_get_journal(struct super_block *sb,
2555 unsigned int journal_inum)
2557 struct inode *journal_inode;
2560 /* First, test for the existence of a valid inode on disk. Bad
2561 * things happen if we iget() an unused inode, as the subsequent
2562 * iput() will try to delete it. */
2564 journal_inode = ext4_iget(sb, journal_inum);
2565 if (IS_ERR(journal_inode)) {
2566 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2569 if (!journal_inode->i_nlink) {
2570 make_bad_inode(journal_inode);
2571 iput(journal_inode);
2572 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2576 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2577 journal_inode, journal_inode->i_size);
2578 if (!S_ISREG(journal_inode->i_mode)) {
2579 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2580 iput(journal_inode);
2584 journal = jbd2_journal_init_inode(journal_inode);
2586 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2587 iput(journal_inode);
2590 journal->j_private = sb;
2591 ext4_init_journal_params(sb, journal);
2595 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2598 struct buffer_head *bh;
2602 int hblock, blocksize;
2603 ext4_fsblk_t sb_block;
2604 unsigned long offset;
2605 struct ext4_super_block *es;
2606 struct block_device *bdev;
2608 bdev = ext4_blkdev_get(j_dev);
2612 if (bd_claim(bdev, sb)) {
2614 "EXT4: failed to claim external journal device.\n");
2619 blocksize = sb->s_blocksize;
2620 hblock = bdev_hardsect_size(bdev);
2621 if (blocksize < hblock) {
2623 "EXT4-fs: blocksize too small for journal device.\n");
2627 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2628 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2629 set_blocksize(bdev, blocksize);
2630 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2631 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2632 "external journal\n");
2636 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2637 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2638 !(le32_to_cpu(es->s_feature_incompat) &
2639 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2640 printk(KERN_ERR "EXT4-fs: external journal has "
2641 "bad superblock\n");
2646 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2647 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2652 len = ext4_blocks_count(es);
2653 start = sb_block + 1;
2654 brelse(bh); /* we're done with the superblock */
2656 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2657 start, len, blocksize);
2659 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2662 journal->j_private = sb;
2663 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2664 wait_on_buffer(journal->j_sb_buffer);
2665 if (!buffer_uptodate(journal->j_sb_buffer)) {
2666 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2669 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2670 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2671 "user (unsupported) - %d\n",
2672 be32_to_cpu(journal->j_superblock->s_nr_users));
2675 EXT4_SB(sb)->journal_bdev = bdev;
2676 ext4_init_journal_params(sb, journal);
2679 jbd2_journal_destroy(journal);
2681 ext4_blkdev_put(bdev);
2685 static int ext4_load_journal(struct super_block *sb,
2686 struct ext4_super_block *es,
2687 unsigned long journal_devnum)
2690 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2693 int really_read_only;
2695 if (journal_devnum &&
2696 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2697 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2698 "numbers have changed\n");
2699 journal_dev = new_decode_dev(journal_devnum);
2701 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2703 really_read_only = bdev_read_only(sb->s_bdev);
2706 * Are we loading a blank journal or performing recovery after a
2707 * crash? For recovery, we need to check in advance whether we
2708 * can get read-write access to the device.
2711 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2712 if (sb->s_flags & MS_RDONLY) {
2713 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2714 "required on readonly filesystem.\n");
2715 if (really_read_only) {
2716 printk(KERN_ERR "EXT4-fs: write access "
2717 "unavailable, cannot proceed.\n");
2720 printk(KERN_INFO "EXT4-fs: write access will "
2721 "be enabled during recovery.\n");
2725 if (journal_inum && journal_dev) {
2726 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2727 "and inode journals!\n");
2732 if (!(journal = ext4_get_journal(sb, journal_inum)))
2735 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2739 if (journal->j_flags & JBD2_BARRIER)
2740 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2742 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2744 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2745 err = jbd2_journal_update_format(journal);
2747 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2748 jbd2_journal_destroy(journal);
2753 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2754 err = jbd2_journal_wipe(journal, !really_read_only);
2756 err = jbd2_journal_load(journal);
2759 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2760 jbd2_journal_destroy(journal);
2764 EXT4_SB(sb)->s_journal = journal;
2765 ext4_clear_journal_err(sb, es);
2767 if (journal_devnum &&
2768 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2769 es->s_journal_dev = cpu_to_le32(journal_devnum);
2772 /* Make sure we flush the recovery flag to disk. */
2773 ext4_commit_super(sb, es, 1);
2779 static int ext4_create_journal(struct super_block *sb,
2780 struct ext4_super_block *es,
2781 unsigned int journal_inum)
2786 if (sb->s_flags & MS_RDONLY) {
2787 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2788 "create journal.\n");
2792 journal = ext4_get_journal(sb, journal_inum);
2796 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2799 err = jbd2_journal_create(journal);
2801 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2802 jbd2_journal_destroy(journal);
2806 EXT4_SB(sb)->s_journal = journal;
2808 ext4_update_dynamic_rev(sb);
2809 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2810 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2812 es->s_journal_inum = cpu_to_le32(journal_inum);
2815 /* Make sure we flush the recovery flag to disk. */
2816 ext4_commit_super(sb, es, 1);
2821 static void ext4_commit_super(struct super_block *sb,
2822 struct ext4_super_block *es, int sync)
2824 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2828 if (buffer_write_io_error(sbh)) {
2830 * Oh, dear. A previous attempt to write the
2831 * superblock failed. This could happen because the
2832 * USB device was yanked out. Or it could happen to
2833 * be a transient write error and maybe the block will
2834 * be remapped. Nothing we can do but to retry the
2835 * write and hope for the best.
2837 printk(KERN_ERR "ext4: previous I/O error to "
2838 "superblock detected for %s.\n", sb->s_id);
2839 clear_buffer_write_io_error(sbh);
2840 set_buffer_uptodate(sbh);
2842 es->s_wtime = cpu_to_le32(get_seconds());
2843 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2844 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2845 BUFFER_TRACE(sbh, "marking dirty");
2846 mark_buffer_dirty(sbh);
2848 sync_dirty_buffer(sbh);
2849 if (buffer_write_io_error(sbh)) {
2850 printk(KERN_ERR "ext4: I/O error while writing "
2851 "superblock for %s.\n", sb->s_id);
2852 clear_buffer_write_io_error(sbh);
2853 set_buffer_uptodate(sbh);
2860 * Have we just finished recovery? If so, and if we are mounting (or
2861 * remounting) the filesystem readonly, then we will end up with a
2862 * consistent fs on disk. Record that fact.
2864 static void ext4_mark_recovery_complete(struct super_block *sb,
2865 struct ext4_super_block *es)
2867 journal_t *journal = EXT4_SB(sb)->s_journal;
2869 jbd2_journal_lock_updates(journal);
2870 jbd2_journal_flush(journal);
2872 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2873 sb->s_flags & MS_RDONLY) {
2874 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2876 ext4_commit_super(sb, es, 1);
2879 jbd2_journal_unlock_updates(journal);
2883 * If we are mounting (or read-write remounting) a filesystem whose journal
2884 * has recorded an error from a previous lifetime, move that error to the
2885 * main filesystem now.
2887 static void ext4_clear_journal_err(struct super_block *sb,
2888 struct ext4_super_block *es)
2894 journal = EXT4_SB(sb)->s_journal;
2897 * Now check for any error status which may have been recorded in the
2898 * journal by a prior ext4_error() or ext4_abort()
2901 j_errno = jbd2_journal_errno(journal);
2905 errstr = ext4_decode_error(sb, j_errno, nbuf);
2906 ext4_warning(sb, __func__, "Filesystem error recorded "
2907 "from previous mount: %s", errstr);
2908 ext4_warning(sb, __func__, "Marking fs in need of "
2909 "filesystem check.");
2911 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2912 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2913 ext4_commit_super(sb, es, 1);
2915 jbd2_journal_clear_err(journal);
2920 * Force the running and committing transactions to commit,
2921 * and wait on the commit.
2923 int ext4_force_commit(struct super_block *sb)
2928 if (sb->s_flags & MS_RDONLY)
2931 journal = EXT4_SB(sb)->s_journal;
2933 ret = ext4_journal_force_commit(journal);
2938 * Ext4 always journals updates to the superblock itself, so we don't
2939 * have to propagate any other updates to the superblock on disk at this
2940 * point. Just start an async writeback to get the buffers on their way
2943 * This implicitly triggers the writebehind on sync().
2946 static void ext4_write_super(struct super_block *sb)
2948 if (mutex_trylock(&sb->s_lock) != 0)
2953 static int ext4_sync_fs(struct super_block *sb, int wait)
2958 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2960 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2966 * LVM calls this function before a (read-only) snapshot is created. This
2967 * gives us a chance to flush the journal completely and mark the fs clean.
2969 static void ext4_write_super_lockfs(struct super_block *sb)
2973 if (!(sb->s_flags & MS_RDONLY)) {
2974 journal_t *journal = EXT4_SB(sb)->s_journal;
2976 /* Now we set up the journal barrier. */
2977 jbd2_journal_lock_updates(journal);
2978 jbd2_journal_flush(journal);
2980 /* Journal blocked and flushed, clear needs_recovery flag. */
2981 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2982 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2987 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2988 * flag here, even though the filesystem is not technically dirty yet.
2990 static void ext4_unlockfs(struct super_block *sb)
2992 if (!(sb->s_flags & MS_RDONLY)) {
2994 /* Reser the needs_recovery flag before the fs is unlocked. */
2995 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2996 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2998 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3002 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3004 struct ext4_super_block *es;
3005 struct ext4_sb_info *sbi = EXT4_SB(sb);
3006 ext4_fsblk_t n_blocks_count = 0;
3007 unsigned long old_sb_flags;
3008 struct ext4_mount_options old_opts;
3015 /* Store the original options */
3016 old_sb_flags = sb->s_flags;
3017 old_opts.s_mount_opt = sbi->s_mount_opt;
3018 old_opts.s_resuid = sbi->s_resuid;
3019 old_opts.s_resgid = sbi->s_resgid;
3020 old_opts.s_commit_interval = sbi->s_commit_interval;
3022 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3023 for (i = 0; i < MAXQUOTAS; i++)
3024 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3028 * Allow the "check" option to be passed as a remount option.
3030 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
3035 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3036 ext4_abort(sb, __func__, "Abort forced by user");
3038 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3039 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3043 ext4_init_journal_params(sb, sbi->s_journal);
3045 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3046 n_blocks_count > ext4_blocks_count(es)) {
3047 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3052 if (*flags & MS_RDONLY) {
3054 * First of all, the unconditional stuff we have to do
3055 * to disable replay of the journal when we next remount
3057 sb->s_flags |= MS_RDONLY;
3060 * OK, test if we are remounting a valid rw partition
3061 * readonly, and if so set the rdonly flag and then
3062 * mark the partition as valid again.
3064 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3065 (sbi->s_mount_state & EXT4_VALID_FS))
3066 es->s_state = cpu_to_le16(sbi->s_mount_state);
3069 * We have to unlock super so that we can wait for
3073 ext4_mark_recovery_complete(sb, es);
3077 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3078 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3079 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3080 "remount RDWR because of unsupported "
3081 "optional features (%x).\n",
3082 sb->s_id, le32_to_cpu(ret));
3088 * Make sure the group descriptor checksums
3089 * are sane. If they aren't, refuse to
3092 for (g = 0; g < sbi->s_groups_count; g++) {
3093 struct ext4_group_desc *gdp =
3094 ext4_get_group_desc(sb, g, NULL);
3096 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3098 "EXT4-fs: ext4_remount: "
3099 "Checksum for group %lu failed (%u!=%u)\n",
3100 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3101 le16_to_cpu(gdp->bg_checksum));
3108 * If we have an unprocessed orphan list hanging
3109 * around from a previously readonly bdev mount,
3110 * require a full umount/remount for now.
3112 if (es->s_last_orphan) {
3113 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3114 "remount RDWR because of unprocessed "
3115 "orphan inode list. Please "
3116 "umount/remount instead.\n",
3123 * Mounting a RDONLY partition read-write, so reread
3124 * and store the current valid flag. (It may have
3125 * been changed by e2fsck since we originally mounted
3128 ext4_clear_journal_err(sb, es);
3129 sbi->s_mount_state = le16_to_cpu(es->s_state);
3130 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3132 if (!ext4_setup_super(sb, es, 0))
3133 sb->s_flags &= ~MS_RDONLY;
3137 /* Release old quota file names */
3138 for (i = 0; i < MAXQUOTAS; i++)
3139 if (old_opts.s_qf_names[i] &&
3140 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3141 kfree(old_opts.s_qf_names[i]);
3145 sb->s_flags = old_sb_flags;
3146 sbi->s_mount_opt = old_opts.s_mount_opt;
3147 sbi->s_resuid = old_opts.s_resuid;
3148 sbi->s_resgid = old_opts.s_resgid;
3149 sbi->s_commit_interval = old_opts.s_commit_interval;
3151 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3152 for (i = 0; i < MAXQUOTAS; i++) {
3153 if (sbi->s_qf_names[i] &&
3154 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3155 kfree(sbi->s_qf_names[i]);
3156 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3162 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3164 struct super_block *sb = dentry->d_sb;
3165 struct ext4_sb_info *sbi = EXT4_SB(sb);
3166 struct ext4_super_block *es = sbi->s_es;
3169 if (test_opt(sb, MINIX_DF)) {
3170 sbi->s_overhead_last = 0;
3171 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3172 ext4_group_t ngroups = sbi->s_groups_count, i;
3173 ext4_fsblk_t overhead = 0;
3177 * Compute the overhead (FS structures). This is constant
3178 * for a given filesystem unless the number of block groups
3179 * changes so we cache the previous value until it does.
3183 * All of the blocks before first_data_block are
3186 overhead = le32_to_cpu(es->s_first_data_block);
3189 * Add the overhead attributed to the superblock and
3190 * block group descriptors. If the sparse superblocks
3191 * feature is turned on, then not all groups have this.
3193 for (i = 0; i < ngroups; i++) {
3194 overhead += ext4_bg_has_super(sb, i) +
3195 ext4_bg_num_gdb(sb, i);
3200 * Every block group has an inode bitmap, a block
3201 * bitmap, and an inode table.
3203 overhead += ngroups * (2 + sbi->s_itb_per_group);
3204 sbi->s_overhead_last = overhead;
3206 sbi->s_blocks_last = ext4_blocks_count(es);
3209 buf->f_type = EXT4_SUPER_MAGIC;
3210 buf->f_bsize = sb->s_blocksize;
3211 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3212 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3213 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3214 ext4_free_blocks_count_set(es, buf->f_bfree);
3215 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3216 if (buf->f_bfree < ext4_r_blocks_count(es))
3218 buf->f_files = le32_to_cpu(es->s_inodes_count);
3219 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3220 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3221 buf->f_namelen = EXT4_NAME_LEN;
3222 fsid = le64_to_cpup((void *)es->s_uuid) ^
3223 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3224 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3225 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3229 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3230 * is locked for write. Otherwise the are possible deadlocks:
3231 * Process 1 Process 2
3232 * ext4_create() quota_sync()
3233 * jbd2_journal_start() write_dquot()
3234 * DQUOT_INIT() down(dqio_mutex)
3235 * down(dqio_mutex) jbd2_journal_start()
3241 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3243 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3246 static int ext4_dquot_initialize(struct inode *inode, int type)
3251 /* We may create quota structure so we need to reserve enough blocks */
3252 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3254 return PTR_ERR(handle);
3255 ret = dquot_initialize(inode, type);
3256 err = ext4_journal_stop(handle);
3262 static int ext4_dquot_drop(struct inode *inode)
3267 /* We may delete quota structure so we need to reserve enough blocks */
3268 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3269 if (IS_ERR(handle)) {
3271 * We call dquot_drop() anyway to at least release references
3272 * to quota structures so that umount does not hang.
3275 return PTR_ERR(handle);
3277 ret = dquot_drop(inode);
3278 err = ext4_journal_stop(handle);
3284 static int ext4_write_dquot(struct dquot *dquot)
3288 struct inode *inode;
3290 inode = dquot_to_inode(dquot);
3291 handle = ext4_journal_start(inode,
3292 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3294 return PTR_ERR(handle);
3295 ret = dquot_commit(dquot);
3296 err = ext4_journal_stop(handle);
3302 static int ext4_acquire_dquot(struct dquot *dquot)
3307 handle = ext4_journal_start(dquot_to_inode(dquot),
3308 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3310 return PTR_ERR(handle);
3311 ret = dquot_acquire(dquot);
3312 err = ext4_journal_stop(handle);
3318 static int ext4_release_dquot(struct dquot *dquot)
3323 handle = ext4_journal_start(dquot_to_inode(dquot),
3324 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3325 if (IS_ERR(handle)) {
3326 /* Release dquot anyway to avoid endless cycle in dqput() */
3327 dquot_release(dquot);
3328 return PTR_ERR(handle);
3330 ret = dquot_release(dquot);
3331 err = ext4_journal_stop(handle);
3337 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3339 /* Are we journaling quotas? */
3340 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3341 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3342 dquot_mark_dquot_dirty(dquot);
3343 return ext4_write_dquot(dquot);
3345 return dquot_mark_dquot_dirty(dquot);
3349 static int ext4_write_info(struct super_block *sb, int type)
3354 /* Data block + inode block */
3355 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3357 return PTR_ERR(handle);
3358 ret = dquot_commit_info(sb, type);
3359 err = ext4_journal_stop(handle);
3366 * Turn on quotas during mount time - we need to find
3367 * the quota file and such...
3369 static int ext4_quota_on_mount(struct super_block *sb, int type)
3371 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3372 EXT4_SB(sb)->s_jquota_fmt, type);
3376 * Standard function to be called on quota_on
3378 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3379 char *path, int remount)
3382 struct nameidata nd;
3384 if (!test_opt(sb, QUOTA))
3386 /* When remounting, no checks are needed and in fact, path is NULL */
3388 return vfs_quota_on(sb, type, format_id, path, remount);
3390 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3394 /* Quotafile not on the same filesystem? */
3395 if (nd.path.mnt->mnt_sb != sb) {
3399 /* Journaling quota? */
3400 if (EXT4_SB(sb)->s_qf_names[type]) {
3401 /* Quotafile not in fs root? */
3402 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3404 "EXT4-fs: Quota file not on filesystem root. "
3405 "Journaled quota will not work.\n");
3409 * When we journal data on quota file, we have to flush journal to see
3410 * all updates to the file when we bypass pagecache...
3412 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3414 * We don't need to lock updates but journal_flush() could
3415 * otherwise be livelocked...
3417 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3418 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3419 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3422 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
3427 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3428 * acquiring the locks... As quota files are never truncated and quota code
3429 * itself serializes the operations (and noone else should touch the files)
3430 * we don't have to be afraid of races */
3431 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3432 size_t len, loff_t off)
3434 struct inode *inode = sb_dqopt(sb)->files[type];
3435 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3437 int offset = off & (sb->s_blocksize - 1);
3440 struct buffer_head *bh;
3441 loff_t i_size = i_size_read(inode);
3445 if (off+len > i_size)
3448 while (toread > 0) {
3449 tocopy = sb->s_blocksize - offset < toread ?
3450 sb->s_blocksize - offset : toread;
3451 bh = ext4_bread(NULL, inode, blk, 0, &err);
3454 if (!bh) /* A hole? */
3455 memset(data, 0, tocopy);
3457 memcpy(data, bh->b_data+offset, tocopy);
3467 /* Write to quotafile (we know the transaction is already started and has
3468 * enough credits) */
3469 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3470 const char *data, size_t len, loff_t off)
3472 struct inode *inode = sb_dqopt(sb)->files[type];
3473 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3475 int offset = off & (sb->s_blocksize - 1);
3477 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3478 size_t towrite = len;
3479 struct buffer_head *bh;
3480 handle_t *handle = journal_current_handle();
3483 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3484 " cancelled because transaction is not started.\n",
3485 (unsigned long long)off, (unsigned long long)len);
3488 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3489 while (towrite > 0) {
3490 tocopy = sb->s_blocksize - offset < towrite ?
3491 sb->s_blocksize - offset : towrite;
3492 bh = ext4_bread(handle, inode, blk, 1, &err);
3495 if (journal_quota) {
3496 err = ext4_journal_get_write_access(handle, bh);
3503 memcpy(bh->b_data+offset, data, tocopy);
3504 flush_dcache_page(bh->b_page);
3507 err = ext4_journal_dirty_metadata(handle, bh);
3509 /* Always do at least ordered writes for quotas */
3510 err = ext4_jbd2_file_inode(handle, inode);
3511 mark_buffer_dirty(bh);
3522 if (len == towrite) {
3523 mutex_unlock(&inode->i_mutex);
3526 if (inode->i_size < off+len-towrite) {
3527 i_size_write(inode, off+len-towrite);
3528 EXT4_I(inode)->i_disksize = inode->i_size;
3530 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3531 ext4_mark_inode_dirty(handle, inode);
3532 mutex_unlock(&inode->i_mutex);
3533 return len - towrite;
3538 static int ext4_get_sb(struct file_system_type *fs_type,
3539 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3541 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3544 #ifdef CONFIG_PROC_FS
3545 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3547 unsigned int *p = m->private;
3549 seq_printf(m, "%u\n", *p);
3553 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3555 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3558 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3559 size_t cnt, loff_t *ppos)
3561 unsigned int *p = PDE(file->f_path.dentry->d_inode)->data;
3563 unsigned long value;
3565 if (cnt >= sizeof(str))
3567 if (copy_from_user(str, buf, cnt))
3569 value = simple_strtol(str, NULL, 0);
3576 const struct file_operations ext4_ui_proc_fops = {
3577 .owner = THIS_MODULE,
3578 .open = ext4_ui_proc_open,
3580 .llseek = seq_lseek,
3581 .release = single_release,
3582 .write = ext4_ui_proc_write,
3586 static struct file_system_type ext4dev_fs_type = {
3587 .owner = THIS_MODULE,
3589 .get_sb = ext4_get_sb,
3590 .kill_sb = kill_block_super,
3591 .fs_flags = FS_REQUIRES_DEV,
3594 static int __init init_ext4_fs(void)
3598 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3599 err = init_ext4_mballoc();
3603 err = init_ext4_xattr();
3606 err = init_inodecache();
3609 err = register_filesystem(&ext4dev_fs_type);
3614 destroy_inodecache();
3618 exit_ext4_mballoc();
3622 static void __exit exit_ext4_fs(void)
3624 unregister_filesystem(&ext4dev_fs_type);
3625 destroy_inodecache();
3627 exit_ext4_mballoc();
3628 remove_proc_entry("fs/ext4", NULL);
3631 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3632 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3633 MODULE_LICENSE("GPL");
3634 module_init(init_ext4_fs)
3635 module_exit(exit_ext4_fs)