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/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
71 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
72 const char *dev_name, void *data, struct vfsmount *mnt);
74 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75 static struct file_system_type ext3_fs_type = {
78 .get_sb = ext4_get_sb,
79 .kill_sb = kill_block_super,
80 .fs_flags = FS_REQUIRES_DEV,
82 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
84 #define IS_EXT3_SB(sb) (0)
87 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
90 return le32_to_cpu(bg->bg_block_bitmap_lo) |
91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
92 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
95 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
98 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
103 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
104 struct ext4_group_desc *bg)
106 return le32_to_cpu(bg->bg_inode_table_lo) |
107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
111 __u32 ext4_free_blks_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
114 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
116 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
119 __u32 ext4_free_inodes_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
122 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
124 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
127 __u32 ext4_used_dirs_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
130 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
132 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
135 __u32 ext4_itable_unused_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
138 return le16_to_cpu(bg->bg_itable_unused_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
140 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
143 void ext4_block_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
151 void ext4_inode_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
154 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
159 void ext4_inode_table_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
162 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
167 void ext4_free_blks_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
170 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
175 void ext4_free_inodes_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
178 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
183 void ext4_used_dirs_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
186 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
191 void ext4_itable_unused_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
194 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
200 /* Just increment the non-pointer handle value */
201 static handle_t *ext4_get_nojournal(void)
203 handle_t *handle = current->journal_info;
204 unsigned long ref_cnt = (unsigned long)handle;
206 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
209 handle = (handle_t *)ref_cnt;
211 current->journal_info = handle;
216 /* Decrement the non-pointer handle value */
217 static void ext4_put_nojournal(handle_t *handle)
219 unsigned long ref_cnt = (unsigned long)handle;
221 BUG_ON(ref_cnt == 0);
224 handle = (handle_t *)ref_cnt;
226 current->journal_info = handle;
230 * Wrappers for jbd2_journal_start/end.
232 * The only special thing we need to do here is to make sure that all
233 * journal_end calls result in the superblock being marked dirty, so
234 * that sync() will call the filesystem's write_super callback if
237 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
241 if (sb->s_flags & MS_RDONLY)
242 return ERR_PTR(-EROFS);
244 vfs_check_frozen(sb, SB_FREEZE_WRITE);
245 /* Special case here: if the journal has aborted behind our
246 * backs (eg. EIO in the commit thread), then we still need to
247 * take the FS itself readonly cleanly. */
248 journal = EXT4_SB(sb)->s_journal;
250 if (is_journal_aborted(journal)) {
251 ext4_abort(sb, "Detected aborted journal");
252 return ERR_PTR(-EROFS);
254 return jbd2_journal_start(journal, nblocks);
256 return ext4_get_nojournal();
260 * The only special thing we need to do here is to make sure that all
261 * jbd2_journal_stop calls result in the superblock being marked dirty, so
262 * that sync() will call the filesystem's write_super callback if
265 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
267 struct super_block *sb;
271 if (!ext4_handle_valid(handle)) {
272 ext4_put_nojournal(handle);
275 sb = handle->h_transaction->t_journal->j_private;
277 rc = jbd2_journal_stop(handle);
282 __ext4_std_error(sb, where, line, err);
286 void ext4_journal_abort_handle(const char *caller, unsigned int line,
287 const char *err_fn, struct buffer_head *bh,
288 handle_t *handle, int err)
291 const char *errstr = ext4_decode_error(NULL, err, nbuf);
293 BUG_ON(!ext4_handle_valid(handle));
296 BUFFER_TRACE(bh, "abort");
301 if (is_handle_aborted(handle))
304 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
305 caller, line, errstr, err_fn);
307 jbd2_journal_abort_handle(handle);
310 /* Deal with the reporting of failure conditions on a filesystem such as
311 * inconsistencies detected or read IO failures.
313 * On ext2, we can store the error state of the filesystem in the
314 * superblock. That is not possible on ext4, because we may have other
315 * write ordering constraints on the superblock which prevent us from
316 * writing it out straight away; and given that the journal is about to
317 * be aborted, we can't rely on the current, or future, transactions to
318 * write out the superblock safely.
320 * We'll just use the jbd2_journal_abort() error code to record an error in
321 * the journal instead. On recovery, the journal will complain about
322 * that error until we've noted it down and cleared it.
325 static void ext4_handle_error(struct super_block *sb)
327 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
329 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
330 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
332 if (sb->s_flags & MS_RDONLY)
335 if (!test_opt(sb, ERRORS_CONT)) {
336 journal_t *journal = EXT4_SB(sb)->s_journal;
338 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
340 jbd2_journal_abort(journal, -EIO);
342 if (test_opt(sb, ERRORS_RO)) {
343 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
344 sb->s_flags |= MS_RDONLY;
346 ext4_commit_super(sb, 1);
347 if (test_opt(sb, ERRORS_PANIC))
348 panic("EXT4-fs (device %s): panic forced after error\n",
352 void __ext4_error(struct super_block *sb, const char *function,
353 unsigned int line, const char *fmt, ...)
358 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
359 sb->s_id, function, line, current->comm);
364 ext4_handle_error(sb);
367 void ext4_error_inode(struct inode *inode, const char *function,
368 unsigned int line, ext4_fsblk_t block,
369 const char *fmt, ...)
374 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
375 inode->i_sb->s_id, function, line, inode->i_ino);
377 printk("block %llu: ", block);
378 printk("comm %s: ", current->comm);
383 ext4_handle_error(inode->i_sb);
386 void ext4_error_file(struct file *file, const char *function,
387 unsigned int line, const char *fmt, ...)
390 struct inode *inode = file->f_dentry->d_inode;
391 char pathname[80], *path;
394 path = d_path(&(file->f_path), pathname, sizeof(pathname));
398 "EXT4-fs error (device %s): %s:%d: inode #%lu "
399 "(comm %s path %s): ",
400 inode->i_sb->s_id, function, line, inode->i_ino,
401 current->comm, path);
406 ext4_handle_error(inode->i_sb);
409 static const char *ext4_decode_error(struct super_block *sb, int errno,
416 errstr = "IO failure";
419 errstr = "Out of memory";
422 if (!sb || (EXT4_SB(sb)->s_journal &&
423 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
424 errstr = "Journal has aborted";
426 errstr = "Readonly filesystem";
429 /* If the caller passed in an extra buffer for unknown
430 * errors, textualise them now. Else we just return
433 /* Check for truncated error codes... */
434 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
443 /* __ext4_std_error decodes expected errors from journaling functions
444 * automatically and invokes the appropriate error response. */
446 void __ext4_std_error(struct super_block *sb, const char *function,
447 unsigned int line, int errno)
452 /* Special case: if the error is EROFS, and we're not already
453 * inside a transaction, then there's really no point in logging
455 if (errno == -EROFS && journal_current_handle() == NULL &&
456 (sb->s_flags & MS_RDONLY))
459 errstr = ext4_decode_error(sb, errno, nbuf);
460 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
461 sb->s_id, function, line, errstr);
463 ext4_handle_error(sb);
467 * ext4_abort is a much stronger failure handler than ext4_error. The
468 * abort function may be used to deal with unrecoverable failures such
469 * as journal IO errors or ENOMEM at a critical moment in log management.
471 * We unconditionally force the filesystem into an ABORT|READONLY state,
472 * unless the error response on the fs has been set to panic in which
473 * case we take the easy way out and panic immediately.
476 void __ext4_abort(struct super_block *sb, const char *function,
477 unsigned int line, const char *fmt, ...)
482 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
488 if (test_opt(sb, ERRORS_PANIC))
489 panic("EXT4-fs panic from previous error\n");
491 if (sb->s_flags & MS_RDONLY)
494 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
495 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
496 sb->s_flags |= MS_RDONLY;
497 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
498 if (EXT4_SB(sb)->s_journal)
499 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
502 void ext4_msg (struct super_block * sb, const char *prefix,
503 const char *fmt, ...)
508 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
514 void __ext4_warning(struct super_block *sb, const char *function,
515 unsigned int line, const char *fmt, ...)
520 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
521 sb->s_id, function, line);
527 void __ext4_grp_locked_error(const char *function, unsigned int line,
528 struct super_block *sb, ext4_group_t grp,
529 unsigned long ino, ext4_fsblk_t block,
530 const char *fmt, ...)
535 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
538 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
539 sb->s_id, function, line, grp);
541 printk("inode %lu: ", ino);
543 printk("block %llu:", (unsigned long long) block);
548 if (test_opt(sb, ERRORS_CONT)) {
549 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
550 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
551 ext4_commit_super(sb, 0);
554 ext4_unlock_group(sb, grp);
555 ext4_handle_error(sb);
557 * We only get here in the ERRORS_RO case; relocking the group
558 * may be dangerous, but nothing bad will happen since the
559 * filesystem will have already been marked read/only and the
560 * journal has been aborted. We return 1 as a hint to callers
561 * who might what to use the return value from
562 * ext4_grp_locked_error() to distinguish beween the
563 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
564 * aggressively from the ext4 function in question, with a
565 * more appropriate error code.
567 ext4_lock_group(sb, grp);
571 void ext4_update_dynamic_rev(struct super_block *sb)
573 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
575 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
579 "updating to rev %d because of new feature flag, "
580 "running e2fsck is recommended",
583 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
584 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
585 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
586 /* leave es->s_feature_*compat flags alone */
587 /* es->s_uuid will be set by e2fsck if empty */
590 * The rest of the superblock fields should be zero, and if not it
591 * means they are likely already in use, so leave them alone. We
592 * can leave it up to e2fsck to clean up any inconsistencies there.
597 * Open the external journal device
599 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
601 struct block_device *bdev;
602 char b[BDEVNAME_SIZE];
604 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
610 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
611 __bdevname(dev, b), PTR_ERR(bdev));
616 * Release the journal device
618 static int ext4_blkdev_put(struct block_device *bdev)
621 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
624 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
626 struct block_device *bdev;
629 bdev = sbi->journal_bdev;
631 ret = ext4_blkdev_put(bdev);
632 sbi->journal_bdev = NULL;
637 static inline struct inode *orphan_list_entry(struct list_head *l)
639 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
642 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
646 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
647 le32_to_cpu(sbi->s_es->s_last_orphan));
649 printk(KERN_ERR "sb_info orphan list:\n");
650 list_for_each(l, &sbi->s_orphan) {
651 struct inode *inode = orphan_list_entry(l);
653 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
654 inode->i_sb->s_id, inode->i_ino, inode,
655 inode->i_mode, inode->i_nlink,
660 static void ext4_put_super(struct super_block *sb)
662 struct ext4_sb_info *sbi = EXT4_SB(sb);
663 struct ext4_super_block *es = sbi->s_es;
666 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
668 flush_workqueue(sbi->dio_unwritten_wq);
669 destroy_workqueue(sbi->dio_unwritten_wq);
674 ext4_commit_super(sb, 1);
676 if (sbi->s_journal) {
677 err = jbd2_journal_destroy(sbi->s_journal);
678 sbi->s_journal = NULL;
680 ext4_abort(sb, "Couldn't clean up the journal");
683 ext4_release_system_zone(sb);
685 ext4_ext_release(sb);
686 ext4_xattr_put_super(sb);
688 if (!(sb->s_flags & MS_RDONLY)) {
689 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
690 es->s_state = cpu_to_le16(sbi->s_mount_state);
691 ext4_commit_super(sb, 1);
694 remove_proc_entry(sb->s_id, ext4_proc_root);
696 kobject_del(&sbi->s_kobj);
698 for (i = 0; i < sbi->s_gdb_count; i++)
699 brelse(sbi->s_group_desc[i]);
700 kfree(sbi->s_group_desc);
701 if (is_vmalloc_addr(sbi->s_flex_groups))
702 vfree(sbi->s_flex_groups);
704 kfree(sbi->s_flex_groups);
705 percpu_counter_destroy(&sbi->s_freeblocks_counter);
706 percpu_counter_destroy(&sbi->s_freeinodes_counter);
707 percpu_counter_destroy(&sbi->s_dirs_counter);
708 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
711 for (i = 0; i < MAXQUOTAS; i++)
712 kfree(sbi->s_qf_names[i]);
715 /* Debugging code just in case the in-memory inode orphan list
716 * isn't empty. The on-disk one can be non-empty if we've
717 * detected an error and taken the fs readonly, but the
718 * in-memory list had better be clean by this point. */
719 if (!list_empty(&sbi->s_orphan))
720 dump_orphan_list(sb, sbi);
721 J_ASSERT(list_empty(&sbi->s_orphan));
723 invalidate_bdev(sb->s_bdev);
724 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
726 * Invalidate the journal device's buffers. We don't want them
727 * floating about in memory - the physical journal device may
728 * hotswapped, and it breaks the `ro-after' testing code.
730 sync_blockdev(sbi->journal_bdev);
731 invalidate_bdev(sbi->journal_bdev);
732 ext4_blkdev_remove(sbi);
734 sb->s_fs_info = NULL;
736 * Now that we are completely done shutting down the
737 * superblock, we need to actually destroy the kobject.
741 kobject_put(&sbi->s_kobj);
742 wait_for_completion(&sbi->s_kobj_unregister);
743 kfree(sbi->s_blockgroup_lock);
747 static struct kmem_cache *ext4_inode_cachep;
750 * Called inside transaction, so use GFP_NOFS
752 static struct inode *ext4_alloc_inode(struct super_block *sb)
754 struct ext4_inode_info *ei;
756 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
760 ei->vfs_inode.i_version = 1;
761 ei->vfs_inode.i_data.writeback_index = 0;
762 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
763 INIT_LIST_HEAD(&ei->i_prealloc_list);
764 spin_lock_init(&ei->i_prealloc_lock);
766 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
767 * therefore it can be null here. Don't check it, just initialize
770 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
771 ei->i_reserved_data_blocks = 0;
772 ei->i_reserved_meta_blocks = 0;
773 ei->i_allocated_meta_blocks = 0;
774 ei->i_da_metadata_calc_len = 0;
775 ei->i_delalloc_reserved_flag = 0;
776 spin_lock_init(&(ei->i_block_reservation_lock));
778 ei->i_reserved_quota = 0;
780 INIT_LIST_HEAD(&ei->i_completed_io_list);
781 spin_lock_init(&ei->i_completed_io_lock);
782 ei->cur_aio_dio = NULL;
784 ei->i_datasync_tid = 0;
786 return &ei->vfs_inode;
789 static void ext4_destroy_inode(struct inode *inode)
791 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
792 ext4_msg(inode->i_sb, KERN_ERR,
793 "Inode %lu (%p): orphan list check failed!",
794 inode->i_ino, EXT4_I(inode));
795 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
796 EXT4_I(inode), sizeof(struct ext4_inode_info),
800 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
803 static void init_once(void *foo)
805 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
807 INIT_LIST_HEAD(&ei->i_orphan);
808 #ifdef CONFIG_EXT4_FS_XATTR
809 init_rwsem(&ei->xattr_sem);
811 init_rwsem(&ei->i_data_sem);
812 inode_init_once(&ei->vfs_inode);
815 static int init_inodecache(void)
817 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
818 sizeof(struct ext4_inode_info),
819 0, (SLAB_RECLAIM_ACCOUNT|
822 if (ext4_inode_cachep == NULL)
827 static void destroy_inodecache(void)
829 kmem_cache_destroy(ext4_inode_cachep);
832 static void ext4_clear_inode(struct inode *inode)
835 ext4_discard_preallocations(inode);
836 if (EXT4_JOURNAL(inode))
837 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
838 &EXT4_I(inode)->jinode);
841 static inline void ext4_show_quota_options(struct seq_file *seq,
842 struct super_block *sb)
844 #if defined(CONFIG_QUOTA)
845 struct ext4_sb_info *sbi = EXT4_SB(sb);
847 if (sbi->s_jquota_fmt) {
850 switch (sbi->s_jquota_fmt) {
861 seq_printf(seq, ",jqfmt=%s", fmtname);
864 if (sbi->s_qf_names[USRQUOTA])
865 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
867 if (sbi->s_qf_names[GRPQUOTA])
868 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
870 if (test_opt(sb, USRQUOTA))
871 seq_puts(seq, ",usrquota");
873 if (test_opt(sb, GRPQUOTA))
874 seq_puts(seq, ",grpquota");
880 * - it's set to a non-default value OR
881 * - if the per-sb default is different from the global default
883 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
886 unsigned long def_mount_opts;
887 struct super_block *sb = vfs->mnt_sb;
888 struct ext4_sb_info *sbi = EXT4_SB(sb);
889 struct ext4_super_block *es = sbi->s_es;
891 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
892 def_errors = le16_to_cpu(es->s_errors);
894 if (sbi->s_sb_block != 1)
895 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
896 if (test_opt(sb, MINIX_DF))
897 seq_puts(seq, ",minixdf");
898 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
899 seq_puts(seq, ",grpid");
900 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
901 seq_puts(seq, ",nogrpid");
902 if (sbi->s_resuid != EXT4_DEF_RESUID ||
903 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
904 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
906 if (sbi->s_resgid != EXT4_DEF_RESGID ||
907 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
908 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
910 if (test_opt(sb, ERRORS_RO)) {
911 if (def_errors == EXT4_ERRORS_PANIC ||
912 def_errors == EXT4_ERRORS_CONTINUE) {
913 seq_puts(seq, ",errors=remount-ro");
916 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
917 seq_puts(seq, ",errors=continue");
918 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
919 seq_puts(seq, ",errors=panic");
920 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
921 seq_puts(seq, ",nouid32");
922 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
923 seq_puts(seq, ",debug");
924 if (test_opt(sb, OLDALLOC))
925 seq_puts(seq, ",oldalloc");
926 #ifdef CONFIG_EXT4_FS_XATTR
927 if (test_opt(sb, XATTR_USER) &&
928 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
929 seq_puts(seq, ",user_xattr");
930 if (!test_opt(sb, XATTR_USER) &&
931 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
932 seq_puts(seq, ",nouser_xattr");
935 #ifdef CONFIG_EXT4_FS_POSIX_ACL
936 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
937 seq_puts(seq, ",acl");
938 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
939 seq_puts(seq, ",noacl");
941 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
942 seq_printf(seq, ",commit=%u",
943 (unsigned) (sbi->s_commit_interval / HZ));
945 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
946 seq_printf(seq, ",min_batch_time=%u",
947 (unsigned) sbi->s_min_batch_time);
949 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
950 seq_printf(seq, ",max_batch_time=%u",
951 (unsigned) sbi->s_min_batch_time);
955 * We're changing the default of barrier mount option, so
956 * let's always display its mount state so it's clear what its
959 seq_puts(seq, ",barrier=");
960 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
961 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
962 seq_puts(seq, ",journal_async_commit");
963 else if (test_opt(sb, JOURNAL_CHECKSUM))
964 seq_puts(seq, ",journal_checksum");
965 if (test_opt(sb, I_VERSION))
966 seq_puts(seq, ",i_version");
967 if (!test_opt(sb, DELALLOC))
968 seq_puts(seq, ",nodelalloc");
972 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
974 * journal mode get enabled in different ways
975 * So just print the value even if we didn't specify it
977 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
978 seq_puts(seq, ",data=journal");
979 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
980 seq_puts(seq, ",data=ordered");
981 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
982 seq_puts(seq, ",data=writeback");
984 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
985 seq_printf(seq, ",inode_readahead_blks=%u",
986 sbi->s_inode_readahead_blks);
988 if (test_opt(sb, DATA_ERR_ABORT))
989 seq_puts(seq, ",data_err=abort");
991 if (test_opt(sb, NO_AUTO_DA_ALLOC))
992 seq_puts(seq, ",noauto_da_alloc");
994 if (test_opt(sb, DISCARD))
995 seq_puts(seq, ",discard");
997 if (test_opt(sb, NOLOAD))
998 seq_puts(seq, ",norecovery");
1000 if (test_opt(sb, DIOREAD_NOLOCK))
1001 seq_puts(seq, ",dioread_nolock");
1003 ext4_show_quota_options(seq, sb);
1008 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1009 u64 ino, u32 generation)
1011 struct inode *inode;
1013 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1014 return ERR_PTR(-ESTALE);
1015 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1016 return ERR_PTR(-ESTALE);
1018 /* iget isn't really right if the inode is currently unallocated!!
1020 * ext4_read_inode will return a bad_inode if the inode had been
1021 * deleted, so we should be safe.
1023 * Currently we don't know the generation for parent directory, so
1024 * a generation of 0 means "accept any"
1026 inode = ext4_iget(sb, ino);
1028 return ERR_CAST(inode);
1029 if (generation && inode->i_generation != generation) {
1031 return ERR_PTR(-ESTALE);
1037 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1038 int fh_len, int fh_type)
1040 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1041 ext4_nfs_get_inode);
1044 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1045 int fh_len, int fh_type)
1047 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1048 ext4_nfs_get_inode);
1052 * Try to release metadata pages (indirect blocks, directories) which are
1053 * mapped via the block device. Since these pages could have journal heads
1054 * which would prevent try_to_free_buffers() from freeing them, we must use
1055 * jbd2 layer's try_to_free_buffers() function to release them.
1057 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1060 journal_t *journal = EXT4_SB(sb)->s_journal;
1062 WARN_ON(PageChecked(page));
1063 if (!page_has_buffers(page))
1066 return jbd2_journal_try_to_free_buffers(journal, page,
1067 wait & ~__GFP_WAIT);
1068 return try_to_free_buffers(page);
1072 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1073 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1075 static int ext4_write_dquot(struct dquot *dquot);
1076 static int ext4_acquire_dquot(struct dquot *dquot);
1077 static int ext4_release_dquot(struct dquot *dquot);
1078 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1079 static int ext4_write_info(struct super_block *sb, int type);
1080 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1082 static int ext4_quota_on_mount(struct super_block *sb, int type);
1083 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1084 size_t len, loff_t off);
1085 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1086 const char *data, size_t len, loff_t off);
1088 static const struct dquot_operations ext4_quota_operations = {
1090 .get_reserved_space = ext4_get_reserved_space,
1092 .write_dquot = ext4_write_dquot,
1093 .acquire_dquot = ext4_acquire_dquot,
1094 .release_dquot = ext4_release_dquot,
1095 .mark_dirty = ext4_mark_dquot_dirty,
1096 .write_info = ext4_write_info,
1097 .alloc_dquot = dquot_alloc,
1098 .destroy_dquot = dquot_destroy,
1101 static const struct quotactl_ops ext4_qctl_operations = {
1102 .quota_on = ext4_quota_on,
1103 .quota_off = dquot_quota_off,
1104 .quota_sync = dquot_quota_sync,
1105 .get_info = dquot_get_dqinfo,
1106 .set_info = dquot_set_dqinfo,
1107 .get_dqblk = dquot_get_dqblk,
1108 .set_dqblk = dquot_set_dqblk
1112 static const struct super_operations ext4_sops = {
1113 .alloc_inode = ext4_alloc_inode,
1114 .destroy_inode = ext4_destroy_inode,
1115 .write_inode = ext4_write_inode,
1116 .dirty_inode = ext4_dirty_inode,
1117 .delete_inode = ext4_delete_inode,
1118 .put_super = ext4_put_super,
1119 .sync_fs = ext4_sync_fs,
1120 .freeze_fs = ext4_freeze,
1121 .unfreeze_fs = ext4_unfreeze,
1122 .statfs = ext4_statfs,
1123 .remount_fs = ext4_remount,
1124 .clear_inode = ext4_clear_inode,
1125 .show_options = ext4_show_options,
1127 .quota_read = ext4_quota_read,
1128 .quota_write = ext4_quota_write,
1130 .bdev_try_to_free_page = bdev_try_to_free_page,
1133 static const struct super_operations ext4_nojournal_sops = {
1134 .alloc_inode = ext4_alloc_inode,
1135 .destroy_inode = ext4_destroy_inode,
1136 .write_inode = ext4_write_inode,
1137 .dirty_inode = ext4_dirty_inode,
1138 .delete_inode = ext4_delete_inode,
1139 .write_super = ext4_write_super,
1140 .put_super = ext4_put_super,
1141 .statfs = ext4_statfs,
1142 .remount_fs = ext4_remount,
1143 .clear_inode = ext4_clear_inode,
1144 .show_options = ext4_show_options,
1146 .quota_read = ext4_quota_read,
1147 .quota_write = ext4_quota_write,
1149 .bdev_try_to_free_page = bdev_try_to_free_page,
1152 static const struct export_operations ext4_export_ops = {
1153 .fh_to_dentry = ext4_fh_to_dentry,
1154 .fh_to_parent = ext4_fh_to_parent,
1155 .get_parent = ext4_get_parent,
1159 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1160 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1161 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1162 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1163 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1164 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1165 Opt_journal_update, Opt_journal_dev,
1166 Opt_journal_checksum, Opt_journal_async_commit,
1167 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1168 Opt_data_err_abort, Opt_data_err_ignore,
1169 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1170 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1171 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1172 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1173 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1174 Opt_block_validity, Opt_noblock_validity,
1175 Opt_inode_readahead_blks, Opt_journal_ioprio,
1176 Opt_dioread_nolock, Opt_dioread_lock,
1177 Opt_discard, Opt_nodiscard,
1180 static const match_table_t tokens = {
1181 {Opt_bsd_df, "bsddf"},
1182 {Opt_minix_df, "minixdf"},
1183 {Opt_grpid, "grpid"},
1184 {Opt_grpid, "bsdgroups"},
1185 {Opt_nogrpid, "nogrpid"},
1186 {Opt_nogrpid, "sysvgroups"},
1187 {Opt_resgid, "resgid=%u"},
1188 {Opt_resuid, "resuid=%u"},
1190 {Opt_err_cont, "errors=continue"},
1191 {Opt_err_panic, "errors=panic"},
1192 {Opt_err_ro, "errors=remount-ro"},
1193 {Opt_nouid32, "nouid32"},
1194 {Opt_debug, "debug"},
1195 {Opt_oldalloc, "oldalloc"},
1196 {Opt_orlov, "orlov"},
1197 {Opt_user_xattr, "user_xattr"},
1198 {Opt_nouser_xattr, "nouser_xattr"},
1200 {Opt_noacl, "noacl"},
1201 {Opt_noload, "noload"},
1202 {Opt_noload, "norecovery"},
1205 {Opt_commit, "commit=%u"},
1206 {Opt_min_batch_time, "min_batch_time=%u"},
1207 {Opt_max_batch_time, "max_batch_time=%u"},
1208 {Opt_journal_update, "journal=update"},
1209 {Opt_journal_dev, "journal_dev=%u"},
1210 {Opt_journal_checksum, "journal_checksum"},
1211 {Opt_journal_async_commit, "journal_async_commit"},
1212 {Opt_abort, "abort"},
1213 {Opt_data_journal, "data=journal"},
1214 {Opt_data_ordered, "data=ordered"},
1215 {Opt_data_writeback, "data=writeback"},
1216 {Opt_data_err_abort, "data_err=abort"},
1217 {Opt_data_err_ignore, "data_err=ignore"},
1218 {Opt_offusrjquota, "usrjquota="},
1219 {Opt_usrjquota, "usrjquota=%s"},
1220 {Opt_offgrpjquota, "grpjquota="},
1221 {Opt_grpjquota, "grpjquota=%s"},
1222 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1223 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1224 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1225 {Opt_grpquota, "grpquota"},
1226 {Opt_noquota, "noquota"},
1227 {Opt_quota, "quota"},
1228 {Opt_usrquota, "usrquota"},
1229 {Opt_barrier, "barrier=%u"},
1230 {Opt_barrier, "barrier"},
1231 {Opt_nobarrier, "nobarrier"},
1232 {Opt_i_version, "i_version"},
1233 {Opt_stripe, "stripe=%u"},
1234 {Opt_resize, "resize"},
1235 {Opt_delalloc, "delalloc"},
1236 {Opt_nodelalloc, "nodelalloc"},
1237 {Opt_block_validity, "block_validity"},
1238 {Opt_noblock_validity, "noblock_validity"},
1239 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1240 {Opt_journal_ioprio, "journal_ioprio=%u"},
1241 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1242 {Opt_auto_da_alloc, "auto_da_alloc"},
1243 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1244 {Opt_dioread_nolock, "dioread_nolock"},
1245 {Opt_dioread_lock, "dioread_lock"},
1246 {Opt_discard, "discard"},
1247 {Opt_nodiscard, "nodiscard"},
1251 static ext4_fsblk_t get_sb_block(void **data)
1253 ext4_fsblk_t sb_block;
1254 char *options = (char *) *data;
1256 if (!options || strncmp(options, "sb=", 3) != 0)
1257 return 1; /* Default location */
1260 /* TODO: use simple_strtoll with >32bit ext4 */
1261 sb_block = simple_strtoul(options, &options, 0);
1262 if (*options && *options != ',') {
1263 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1267 if (*options == ',')
1269 *data = (void *) options;
1274 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1275 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1276 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1279 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1281 struct ext4_sb_info *sbi = EXT4_SB(sb);
1284 if (sb_any_quota_loaded(sb) &&
1285 !sbi->s_qf_names[qtype]) {
1286 ext4_msg(sb, KERN_ERR,
1287 "Cannot change journaled "
1288 "quota options when quota turned on");
1291 qname = match_strdup(args);
1293 ext4_msg(sb, KERN_ERR,
1294 "Not enough memory for storing quotafile name");
1297 if (sbi->s_qf_names[qtype] &&
1298 strcmp(sbi->s_qf_names[qtype], qname)) {
1299 ext4_msg(sb, KERN_ERR,
1300 "%s quota file already specified", QTYPE2NAME(qtype));
1304 sbi->s_qf_names[qtype] = qname;
1305 if (strchr(sbi->s_qf_names[qtype], '/')) {
1306 ext4_msg(sb, KERN_ERR,
1307 "quotafile must be on filesystem root");
1308 kfree(sbi->s_qf_names[qtype]);
1309 sbi->s_qf_names[qtype] = NULL;
1312 set_opt(sbi->s_mount_opt, QUOTA);
1316 static int clear_qf_name(struct super_block *sb, int qtype)
1319 struct ext4_sb_info *sbi = EXT4_SB(sb);
1321 if (sb_any_quota_loaded(sb) &&
1322 sbi->s_qf_names[qtype]) {
1323 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1324 " when quota turned on");
1328 * The space will be released later when all options are confirmed
1331 sbi->s_qf_names[qtype] = NULL;
1336 static int parse_options(char *options, struct super_block *sb,
1337 unsigned long *journal_devnum,
1338 unsigned int *journal_ioprio,
1339 ext4_fsblk_t *n_blocks_count, int is_remount)
1341 struct ext4_sb_info *sbi = EXT4_SB(sb);
1343 substring_t args[MAX_OPT_ARGS];
1353 while ((p = strsep(&options, ",")) != NULL) {
1359 * Initialize args struct so we know whether arg was
1360 * found; some options take optional arguments.
1362 args[0].to = args[0].from = 0;
1363 token = match_token(p, tokens, args);
1366 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1367 clear_opt(sbi->s_mount_opt, MINIX_DF);
1370 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1371 set_opt(sbi->s_mount_opt, MINIX_DF);
1375 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1376 set_opt(sbi->s_mount_opt, GRPID);
1380 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1381 clear_opt(sbi->s_mount_opt, GRPID);
1385 if (match_int(&args[0], &option))
1387 sbi->s_resuid = option;
1390 if (match_int(&args[0], &option))
1392 sbi->s_resgid = option;
1395 /* handled by get_sb_block() instead of here */
1396 /* *sb_block = match_int(&args[0]); */
1399 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1400 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1401 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1404 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1405 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1406 set_opt(sbi->s_mount_opt, ERRORS_RO);
1409 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1410 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1411 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1414 set_opt(sbi->s_mount_opt, NO_UID32);
1417 set_opt(sbi->s_mount_opt, DEBUG);
1420 set_opt(sbi->s_mount_opt, OLDALLOC);
1423 clear_opt(sbi->s_mount_opt, OLDALLOC);
1425 #ifdef CONFIG_EXT4_FS_XATTR
1426 case Opt_user_xattr:
1427 set_opt(sbi->s_mount_opt, XATTR_USER);
1429 case Opt_nouser_xattr:
1430 clear_opt(sbi->s_mount_opt, XATTR_USER);
1433 case Opt_user_xattr:
1434 case Opt_nouser_xattr:
1435 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1438 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1440 set_opt(sbi->s_mount_opt, POSIX_ACL);
1443 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1448 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1451 case Opt_journal_update:
1453 /* Eventually we will want to be able to create
1454 a journal file here. For now, only allow the
1455 user to specify an existing inode to be the
1458 ext4_msg(sb, KERN_ERR,
1459 "Cannot specify journal on remount");
1462 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1464 case Opt_journal_dev:
1466 ext4_msg(sb, KERN_ERR,
1467 "Cannot specify journal on remount");
1470 if (match_int(&args[0], &option))
1472 *journal_devnum = option;
1474 case Opt_journal_checksum:
1475 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1477 case Opt_journal_async_commit:
1478 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1479 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1482 set_opt(sbi->s_mount_opt, NOLOAD);
1485 if (match_int(&args[0], &option))
1490 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1491 sbi->s_commit_interval = HZ * option;
1493 case Opt_max_batch_time:
1494 if (match_int(&args[0], &option))
1499 option = EXT4_DEF_MAX_BATCH_TIME;
1500 sbi->s_max_batch_time = option;
1502 case Opt_min_batch_time:
1503 if (match_int(&args[0], &option))
1507 sbi->s_min_batch_time = option;
1509 case Opt_data_journal:
1510 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1512 case Opt_data_ordered:
1513 data_opt = EXT4_MOUNT_ORDERED_DATA;
1515 case Opt_data_writeback:
1516 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1519 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1520 ext4_msg(sb, KERN_ERR,
1521 "Cannot change data mode on remount");
1525 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1526 sbi->s_mount_opt |= data_opt;
1529 case Opt_data_err_abort:
1530 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1532 case Opt_data_err_ignore:
1533 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1537 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1541 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1544 case Opt_offusrjquota:
1545 if (!clear_qf_name(sb, USRQUOTA))
1548 case Opt_offgrpjquota:
1549 if (!clear_qf_name(sb, GRPQUOTA))
1553 case Opt_jqfmt_vfsold:
1554 qfmt = QFMT_VFS_OLD;
1556 case Opt_jqfmt_vfsv0:
1559 case Opt_jqfmt_vfsv1:
1562 if (sb_any_quota_loaded(sb) &&
1563 sbi->s_jquota_fmt != qfmt) {
1564 ext4_msg(sb, KERN_ERR, "Cannot change "
1565 "journaled quota options when "
1569 sbi->s_jquota_fmt = qfmt;
1573 set_opt(sbi->s_mount_opt, QUOTA);
1574 set_opt(sbi->s_mount_opt, USRQUOTA);
1577 set_opt(sbi->s_mount_opt, QUOTA);
1578 set_opt(sbi->s_mount_opt, GRPQUOTA);
1581 if (sb_any_quota_loaded(sb)) {
1582 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1583 "options when quota turned on");
1586 clear_opt(sbi->s_mount_opt, QUOTA);
1587 clear_opt(sbi->s_mount_opt, USRQUOTA);
1588 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1594 ext4_msg(sb, KERN_ERR,
1595 "quota options not supported");
1599 case Opt_offusrjquota:
1600 case Opt_offgrpjquota:
1601 case Opt_jqfmt_vfsold:
1602 case Opt_jqfmt_vfsv0:
1603 case Opt_jqfmt_vfsv1:
1604 ext4_msg(sb, KERN_ERR,
1605 "journaled quota options not supported");
1611 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1614 clear_opt(sbi->s_mount_opt, BARRIER);
1618 if (match_int(&args[0], &option))
1621 option = 1; /* No argument, default to 1 */
1623 set_opt(sbi->s_mount_opt, BARRIER);
1625 clear_opt(sbi->s_mount_opt, BARRIER);
1631 ext4_msg(sb, KERN_ERR,
1632 "resize option only available "
1636 if (match_int(&args[0], &option) != 0)
1638 *n_blocks_count = option;
1641 ext4_msg(sb, KERN_WARNING,
1642 "Ignoring deprecated nobh option");
1645 ext4_msg(sb, KERN_WARNING,
1646 "Ignoring deprecated bh option");
1649 set_opt(sbi->s_mount_opt, I_VERSION);
1650 sb->s_flags |= MS_I_VERSION;
1652 case Opt_nodelalloc:
1653 clear_opt(sbi->s_mount_opt, DELALLOC);
1656 if (match_int(&args[0], &option))
1660 sbi->s_stripe = option;
1663 set_opt(sbi->s_mount_opt, DELALLOC);
1665 case Opt_block_validity:
1666 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1668 case Opt_noblock_validity:
1669 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1671 case Opt_inode_readahead_blks:
1672 if (match_int(&args[0], &option))
1674 if (option < 0 || option > (1 << 30))
1676 if (!is_power_of_2(option)) {
1677 ext4_msg(sb, KERN_ERR,
1678 "EXT4-fs: inode_readahead_blks"
1679 " must be a power of 2");
1682 sbi->s_inode_readahead_blks = option;
1684 case Opt_journal_ioprio:
1685 if (match_int(&args[0], &option))
1687 if (option < 0 || option > 7)
1689 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1692 case Opt_noauto_da_alloc:
1693 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1695 case Opt_auto_da_alloc:
1697 if (match_int(&args[0], &option))
1700 option = 1; /* No argument, default to 1 */
1702 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1704 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1707 set_opt(sbi->s_mount_opt, DISCARD);
1710 clear_opt(sbi->s_mount_opt, DISCARD);
1712 case Opt_dioread_nolock:
1713 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1715 case Opt_dioread_lock:
1716 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1719 ext4_msg(sb, KERN_ERR,
1720 "Unrecognized mount option \"%s\" "
1721 "or missing value", p);
1726 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1727 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1728 clear_opt(sbi->s_mount_opt, USRQUOTA);
1730 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1731 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1733 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1734 ext4_msg(sb, KERN_ERR, "old and new quota "
1739 if (!sbi->s_jquota_fmt) {
1740 ext4_msg(sb, KERN_ERR, "journaled quota format "
1745 if (sbi->s_jquota_fmt) {
1746 ext4_msg(sb, KERN_ERR, "journaled quota format "
1747 "specified with no journaling "
1756 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1759 struct ext4_sb_info *sbi = EXT4_SB(sb);
1762 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1763 ext4_msg(sb, KERN_ERR, "revision level too high, "
1764 "forcing read-only mode");
1769 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1770 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1771 "running e2fsck is recommended");
1772 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1773 ext4_msg(sb, KERN_WARNING,
1774 "warning: mounting fs with errors, "
1775 "running e2fsck is recommended");
1776 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1777 le16_to_cpu(es->s_mnt_count) >=
1778 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1779 ext4_msg(sb, KERN_WARNING,
1780 "warning: maximal mount count reached, "
1781 "running e2fsck is recommended");
1782 else if (le32_to_cpu(es->s_checkinterval) &&
1783 (le32_to_cpu(es->s_lastcheck) +
1784 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1785 ext4_msg(sb, KERN_WARNING,
1786 "warning: checktime reached, "
1787 "running e2fsck is recommended");
1788 if (!sbi->s_journal)
1789 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1790 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1791 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1792 le16_add_cpu(&es->s_mnt_count, 1);
1793 es->s_mtime = cpu_to_le32(get_seconds());
1794 ext4_update_dynamic_rev(sb);
1796 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1798 ext4_commit_super(sb, 1);
1799 if (test_opt(sb, DEBUG))
1800 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1801 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1803 sbi->s_groups_count,
1804 EXT4_BLOCKS_PER_GROUP(sb),
1805 EXT4_INODES_PER_GROUP(sb),
1811 static int ext4_fill_flex_info(struct super_block *sb)
1813 struct ext4_sb_info *sbi = EXT4_SB(sb);
1814 struct ext4_group_desc *gdp = NULL;
1815 ext4_group_t flex_group_count;
1816 ext4_group_t flex_group;
1817 int groups_per_flex = 0;
1821 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1822 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1824 if (groups_per_flex < 2) {
1825 sbi->s_log_groups_per_flex = 0;
1829 /* We allocate both existing and potentially added groups */
1830 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1831 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1832 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1833 size = flex_group_count * sizeof(struct flex_groups);
1834 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1835 if (sbi->s_flex_groups == NULL) {
1836 sbi->s_flex_groups = vmalloc(size);
1837 if (sbi->s_flex_groups)
1838 memset(sbi->s_flex_groups, 0, size);
1840 if (sbi->s_flex_groups == NULL) {
1841 ext4_msg(sb, KERN_ERR, "not enough memory for "
1842 "%u flex groups", flex_group_count);
1846 for (i = 0; i < sbi->s_groups_count; i++) {
1847 gdp = ext4_get_group_desc(sb, i, NULL);
1849 flex_group = ext4_flex_group(sbi, i);
1850 atomic_add(ext4_free_inodes_count(sb, gdp),
1851 &sbi->s_flex_groups[flex_group].free_inodes);
1852 atomic_add(ext4_free_blks_count(sb, gdp),
1853 &sbi->s_flex_groups[flex_group].free_blocks);
1854 atomic_add(ext4_used_dirs_count(sb, gdp),
1855 &sbi->s_flex_groups[flex_group].used_dirs);
1863 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1864 struct ext4_group_desc *gdp)
1868 if (sbi->s_es->s_feature_ro_compat &
1869 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1870 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1871 __le32 le_group = cpu_to_le32(block_group);
1873 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1874 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1875 crc = crc16(crc, (__u8 *)gdp, offset);
1876 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1877 /* for checksum of struct ext4_group_desc do the rest...*/
1878 if ((sbi->s_es->s_feature_incompat &
1879 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1880 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1881 crc = crc16(crc, (__u8 *)gdp + offset,
1882 le16_to_cpu(sbi->s_es->s_desc_size) -
1886 return cpu_to_le16(crc);
1889 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1890 struct ext4_group_desc *gdp)
1892 if ((sbi->s_es->s_feature_ro_compat &
1893 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1894 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1900 /* Called at mount-time, super-block is locked */
1901 static int ext4_check_descriptors(struct super_block *sb)
1903 struct ext4_sb_info *sbi = EXT4_SB(sb);
1904 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1905 ext4_fsblk_t last_block;
1906 ext4_fsblk_t block_bitmap;
1907 ext4_fsblk_t inode_bitmap;
1908 ext4_fsblk_t inode_table;
1909 int flexbg_flag = 0;
1912 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1915 ext4_debug("Checking group descriptors");
1917 for (i = 0; i < sbi->s_groups_count; i++) {
1918 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1920 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1921 last_block = ext4_blocks_count(sbi->s_es) - 1;
1923 last_block = first_block +
1924 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1926 block_bitmap = ext4_block_bitmap(sb, gdp);
1927 if (block_bitmap < first_block || block_bitmap > last_block) {
1928 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1929 "Block bitmap for group %u not in group "
1930 "(block %llu)!", i, block_bitmap);
1933 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1934 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1935 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1936 "Inode bitmap for group %u not in group "
1937 "(block %llu)!", i, inode_bitmap);
1940 inode_table = ext4_inode_table(sb, gdp);
1941 if (inode_table < first_block ||
1942 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1943 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1944 "Inode table for group %u not in group "
1945 "(block %llu)!", i, inode_table);
1948 ext4_lock_group(sb, i);
1949 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1950 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1951 "Checksum for group %u failed (%u!=%u)",
1952 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1953 gdp)), le16_to_cpu(gdp->bg_checksum));
1954 if (!(sb->s_flags & MS_RDONLY)) {
1955 ext4_unlock_group(sb, i);
1959 ext4_unlock_group(sb, i);
1961 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1964 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1965 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1969 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1970 * the superblock) which were deleted from all directories, but held open by
1971 * a process at the time of a crash. We walk the list and try to delete these
1972 * inodes at recovery time (only with a read-write filesystem).
1974 * In order to keep the orphan inode chain consistent during traversal (in
1975 * case of crash during recovery), we link each inode into the superblock
1976 * orphan list_head and handle it the same way as an inode deletion during
1977 * normal operation (which journals the operations for us).
1979 * We only do an iget() and an iput() on each inode, which is very safe if we
1980 * accidentally point at an in-use or already deleted inode. The worst that
1981 * can happen in this case is that we get a "bit already cleared" message from
1982 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1983 * e2fsck was run on this filesystem, and it must have already done the orphan
1984 * inode cleanup for us, so we can safely abort without any further action.
1986 static void ext4_orphan_cleanup(struct super_block *sb,
1987 struct ext4_super_block *es)
1989 unsigned int s_flags = sb->s_flags;
1990 int nr_orphans = 0, nr_truncates = 0;
1994 if (!es->s_last_orphan) {
1995 jbd_debug(4, "no orphan inodes to clean up\n");
1999 if (bdev_read_only(sb->s_bdev)) {
2000 ext4_msg(sb, KERN_ERR, "write access "
2001 "unavailable, skipping orphan cleanup");
2005 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2006 if (es->s_last_orphan)
2007 jbd_debug(1, "Errors on filesystem, "
2008 "clearing orphan list.\n");
2009 es->s_last_orphan = 0;
2010 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2014 if (s_flags & MS_RDONLY) {
2015 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2016 sb->s_flags &= ~MS_RDONLY;
2019 /* Needed for iput() to work correctly and not trash data */
2020 sb->s_flags |= MS_ACTIVE;
2021 /* Turn on quotas so that they are updated correctly */
2022 for (i = 0; i < MAXQUOTAS; i++) {
2023 if (EXT4_SB(sb)->s_qf_names[i]) {
2024 int ret = ext4_quota_on_mount(sb, i);
2026 ext4_msg(sb, KERN_ERR,
2027 "Cannot turn on journaled "
2028 "quota: error %d", ret);
2033 while (es->s_last_orphan) {
2034 struct inode *inode;
2036 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2037 if (IS_ERR(inode)) {
2038 es->s_last_orphan = 0;
2042 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2043 dquot_initialize(inode);
2044 if (inode->i_nlink) {
2045 ext4_msg(sb, KERN_DEBUG,
2046 "%s: truncating inode %lu to %lld bytes",
2047 __func__, inode->i_ino, inode->i_size);
2048 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2049 inode->i_ino, inode->i_size);
2050 ext4_truncate(inode);
2053 ext4_msg(sb, KERN_DEBUG,
2054 "%s: deleting unreferenced inode %lu",
2055 __func__, inode->i_ino);
2056 jbd_debug(2, "deleting unreferenced inode %lu\n",
2060 iput(inode); /* The delete magic happens here! */
2063 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2066 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2067 PLURAL(nr_orphans));
2069 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2070 PLURAL(nr_truncates));
2072 /* Turn quotas off */
2073 for (i = 0; i < MAXQUOTAS; i++) {
2074 if (sb_dqopt(sb)->files[i])
2075 dquot_quota_off(sb, i);
2078 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2082 * Maximal extent format file size.
2083 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2084 * extent format containers, within a sector_t, and within i_blocks
2085 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2086 * so that won't be a limiting factor.
2088 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2090 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2093 loff_t upper_limit = MAX_LFS_FILESIZE;
2095 /* small i_blocks in vfs inode? */
2096 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2098 * CONFIG_LBDAF is not enabled implies the inode
2099 * i_block represent total blocks in 512 bytes
2100 * 32 == size of vfs inode i_blocks * 8
2102 upper_limit = (1LL << 32) - 1;
2104 /* total blocks in file system block size */
2105 upper_limit >>= (blkbits - 9);
2106 upper_limit <<= blkbits;
2109 /* 32-bit extent-start container, ee_block */
2114 /* Sanity check against vm- & vfs- imposed limits */
2115 if (res > upper_limit)
2122 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2123 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2124 * We need to be 1 filesystem block less than the 2^48 sector limit.
2126 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2128 loff_t res = EXT4_NDIR_BLOCKS;
2131 /* This is calculated to be the largest file size for a dense, block
2132 * mapped file such that the file's total number of 512-byte sectors,
2133 * including data and all indirect blocks, does not exceed (2^48 - 1).
2135 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2136 * number of 512-byte sectors of the file.
2139 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2141 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2142 * the inode i_block field represents total file blocks in
2143 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2145 upper_limit = (1LL << 32) - 1;
2147 /* total blocks in file system block size */
2148 upper_limit >>= (bits - 9);
2152 * We use 48 bit ext4_inode i_blocks
2153 * With EXT4_HUGE_FILE_FL set the i_blocks
2154 * represent total number of blocks in
2155 * file system block size
2157 upper_limit = (1LL << 48) - 1;
2161 /* indirect blocks */
2163 /* double indirect blocks */
2164 meta_blocks += 1 + (1LL << (bits-2));
2165 /* tripple indirect blocks */
2166 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2168 upper_limit -= meta_blocks;
2169 upper_limit <<= bits;
2171 res += 1LL << (bits-2);
2172 res += 1LL << (2*(bits-2));
2173 res += 1LL << (3*(bits-2));
2175 if (res > upper_limit)
2178 if (res > MAX_LFS_FILESIZE)
2179 res = MAX_LFS_FILESIZE;
2184 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2185 ext4_fsblk_t logical_sb_block, int nr)
2187 struct ext4_sb_info *sbi = EXT4_SB(sb);
2188 ext4_group_t bg, first_meta_bg;
2191 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2193 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2195 return logical_sb_block + nr + 1;
2196 bg = sbi->s_desc_per_block * nr;
2197 if (ext4_bg_has_super(sb, bg))
2200 return (has_super + ext4_group_first_block_no(sb, bg));
2204 * ext4_get_stripe_size: Get the stripe size.
2205 * @sbi: In memory super block info
2207 * If we have specified it via mount option, then
2208 * use the mount option value. If the value specified at mount time is
2209 * greater than the blocks per group use the super block value.
2210 * If the super block value is greater than blocks per group return 0.
2211 * Allocator needs it be less than blocks per group.
2214 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2216 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2217 unsigned long stripe_width =
2218 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2220 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2221 return sbi->s_stripe;
2223 if (stripe_width <= sbi->s_blocks_per_group)
2224 return stripe_width;
2226 if (stride <= sbi->s_blocks_per_group)
2235 struct attribute attr;
2236 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2237 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2238 const char *, size_t);
2242 static int parse_strtoul(const char *buf,
2243 unsigned long max, unsigned long *value)
2247 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2248 endp = skip_spaces(endp);
2249 if (*endp || *value > max)
2255 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2256 struct ext4_sb_info *sbi,
2259 return snprintf(buf, PAGE_SIZE, "%llu\n",
2260 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2263 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2264 struct ext4_sb_info *sbi, char *buf)
2266 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2268 return snprintf(buf, PAGE_SIZE, "%lu\n",
2269 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2270 sbi->s_sectors_written_start) >> 1);
2273 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2274 struct ext4_sb_info *sbi, char *buf)
2276 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2278 return snprintf(buf, PAGE_SIZE, "%llu\n",
2279 (unsigned long long)(sbi->s_kbytes_written +
2280 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2281 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2284 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2285 struct ext4_sb_info *sbi,
2286 const char *buf, size_t count)
2290 if (parse_strtoul(buf, 0x40000000, &t))
2293 if (!is_power_of_2(t))
2296 sbi->s_inode_readahead_blks = t;
2300 static ssize_t sbi_ui_show(struct ext4_attr *a,
2301 struct ext4_sb_info *sbi, char *buf)
2303 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2305 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2308 static ssize_t sbi_ui_store(struct ext4_attr *a,
2309 struct ext4_sb_info *sbi,
2310 const char *buf, size_t count)
2312 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2315 if (parse_strtoul(buf, 0xffffffff, &t))
2321 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2322 static struct ext4_attr ext4_attr_##_name = { \
2323 .attr = {.name = __stringify(_name), .mode = _mode }, \
2326 .offset = offsetof(struct ext4_sb_info, _elname), \
2328 #define EXT4_ATTR(name, mode, show, store) \
2329 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2331 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2332 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2333 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2334 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2335 #define ATTR_LIST(name) &ext4_attr_##name.attr
2337 EXT4_RO_ATTR(delayed_allocation_blocks);
2338 EXT4_RO_ATTR(session_write_kbytes);
2339 EXT4_RO_ATTR(lifetime_write_kbytes);
2340 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2341 inode_readahead_blks_store, s_inode_readahead_blks);
2342 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2343 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2344 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2345 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2346 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2347 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2348 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2349 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2351 static struct attribute *ext4_attrs[] = {
2352 ATTR_LIST(delayed_allocation_blocks),
2353 ATTR_LIST(session_write_kbytes),
2354 ATTR_LIST(lifetime_write_kbytes),
2355 ATTR_LIST(inode_readahead_blks),
2356 ATTR_LIST(inode_goal),
2357 ATTR_LIST(mb_stats),
2358 ATTR_LIST(mb_max_to_scan),
2359 ATTR_LIST(mb_min_to_scan),
2360 ATTR_LIST(mb_order2_req),
2361 ATTR_LIST(mb_stream_req),
2362 ATTR_LIST(mb_group_prealloc),
2363 ATTR_LIST(max_writeback_mb_bump),
2367 static ssize_t ext4_attr_show(struct kobject *kobj,
2368 struct attribute *attr, char *buf)
2370 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2372 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2374 return a->show ? a->show(a, sbi, buf) : 0;
2377 static ssize_t ext4_attr_store(struct kobject *kobj,
2378 struct attribute *attr,
2379 const char *buf, size_t len)
2381 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2383 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2385 return a->store ? a->store(a, sbi, buf, len) : 0;
2388 static void ext4_sb_release(struct kobject *kobj)
2390 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2392 complete(&sbi->s_kobj_unregister);
2396 static const struct sysfs_ops ext4_attr_ops = {
2397 .show = ext4_attr_show,
2398 .store = ext4_attr_store,
2401 static struct kobj_type ext4_ktype = {
2402 .default_attrs = ext4_attrs,
2403 .sysfs_ops = &ext4_attr_ops,
2404 .release = ext4_sb_release,
2408 * Check whether this filesystem can be mounted based on
2409 * the features present and the RDONLY/RDWR mount requested.
2410 * Returns 1 if this filesystem can be mounted as requested,
2411 * 0 if it cannot be.
2413 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2415 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2416 ext4_msg(sb, KERN_ERR,
2417 "Couldn't mount because of "
2418 "unsupported optional features (%x)",
2419 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2420 ~EXT4_FEATURE_INCOMPAT_SUPP));
2427 /* Check that feature set is OK for a read-write mount */
2428 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2429 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2430 "unsupported optional features (%x)",
2431 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2432 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2436 * Large file size enabled file system can only be mounted
2437 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2439 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2440 if (sizeof(blkcnt_t) < sizeof(u64)) {
2441 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2442 "cannot be mounted RDWR without "
2450 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2451 __releases(kernel_lock)
2452 __acquires(kernel_lock)
2454 char *orig_data = kstrdup(data, GFP_KERNEL);
2455 struct buffer_head *bh;
2456 struct ext4_super_block *es = NULL;
2457 struct ext4_sb_info *sbi;
2459 ext4_fsblk_t sb_block = get_sb_block(&data);
2460 ext4_fsblk_t logical_sb_block;
2461 unsigned long offset = 0;
2462 unsigned long journal_devnum = 0;
2463 unsigned long def_mount_opts;
2469 unsigned int db_count;
2471 int needs_recovery, has_huge_files;
2474 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2476 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2480 sbi->s_blockgroup_lock =
2481 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2482 if (!sbi->s_blockgroup_lock) {
2486 sb->s_fs_info = sbi;
2487 sbi->s_mount_opt = 0;
2488 sbi->s_resuid = EXT4_DEF_RESUID;
2489 sbi->s_resgid = EXT4_DEF_RESGID;
2490 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2491 sbi->s_sb_block = sb_block;
2492 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2497 /* Cleanup superblock name */
2498 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2501 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2503 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2508 * The ext4 superblock will not be buffer aligned for other than 1kB
2509 * block sizes. We need to calculate the offset from buffer start.
2511 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2512 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2513 offset = do_div(logical_sb_block, blocksize);
2515 logical_sb_block = sb_block;
2518 if (!(bh = sb_bread(sb, logical_sb_block))) {
2519 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2523 * Note: s_es must be initialized as soon as possible because
2524 * some ext4 macro-instructions depend on its value
2526 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2528 sb->s_magic = le16_to_cpu(es->s_magic);
2529 if (sb->s_magic != EXT4_SUPER_MAGIC)
2531 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2533 /* Set defaults before we parse the mount options */
2534 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2535 if (def_mount_opts & EXT4_DEFM_DEBUG)
2536 set_opt(sbi->s_mount_opt, DEBUG);
2537 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2538 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2540 set_opt(sbi->s_mount_opt, GRPID);
2542 if (def_mount_opts & EXT4_DEFM_UID16)
2543 set_opt(sbi->s_mount_opt, NO_UID32);
2544 #ifdef CONFIG_EXT4_FS_XATTR
2545 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2546 set_opt(sbi->s_mount_opt, XATTR_USER);
2548 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2549 if (def_mount_opts & EXT4_DEFM_ACL)
2550 set_opt(sbi->s_mount_opt, POSIX_ACL);
2552 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2553 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2554 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2555 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2556 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2557 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2559 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2560 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2561 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2562 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2564 set_opt(sbi->s_mount_opt, ERRORS_RO);
2566 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2567 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2568 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2569 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2570 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2572 set_opt(sbi->s_mount_opt, BARRIER);
2575 * enable delayed allocation by default
2576 * Use -o nodelalloc to turn it off
2578 if (!IS_EXT3_SB(sb))
2579 set_opt(sbi->s_mount_opt, DELALLOC);
2581 if (!parse_options((char *) data, sb, &journal_devnum,
2582 &journal_ioprio, NULL, 0))
2585 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2586 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2588 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2589 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2590 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2591 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2592 ext4_msg(sb, KERN_WARNING,
2593 "feature flags set on rev 0 fs, "
2594 "running e2fsck is recommended");
2597 * Check feature flags regardless of the revision level, since we
2598 * previously didn't change the revision level when setting the flags,
2599 * so there is a chance incompat flags are set on a rev 0 filesystem.
2601 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2604 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2606 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2607 blocksize > EXT4_MAX_BLOCK_SIZE) {
2608 ext4_msg(sb, KERN_ERR,
2609 "Unsupported filesystem blocksize %d", blocksize);
2613 if (sb->s_blocksize != blocksize) {
2614 /* Validate the filesystem blocksize */
2615 if (!sb_set_blocksize(sb, blocksize)) {
2616 ext4_msg(sb, KERN_ERR, "bad block size %d",
2622 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2623 offset = do_div(logical_sb_block, blocksize);
2624 bh = sb_bread(sb, logical_sb_block);
2626 ext4_msg(sb, KERN_ERR,
2627 "Can't read superblock on 2nd try");
2630 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2632 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2633 ext4_msg(sb, KERN_ERR,
2634 "Magic mismatch, very weird!");
2639 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2640 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2641 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2643 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2645 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2646 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2647 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2649 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2650 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2651 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2652 (!is_power_of_2(sbi->s_inode_size)) ||
2653 (sbi->s_inode_size > blocksize)) {
2654 ext4_msg(sb, KERN_ERR,
2655 "unsupported inode size: %d",
2659 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2660 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2663 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2664 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2665 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2666 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2667 !is_power_of_2(sbi->s_desc_size)) {
2668 ext4_msg(sb, KERN_ERR,
2669 "unsupported descriptor size %lu",
2674 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2676 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2677 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2678 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2681 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2682 if (sbi->s_inodes_per_block == 0)
2684 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2685 sbi->s_inodes_per_block;
2686 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2688 sbi->s_mount_state = le16_to_cpu(es->s_state);
2689 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2690 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2692 for (i = 0; i < 4; i++)
2693 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2694 sbi->s_def_hash_version = es->s_def_hash_version;
2695 i = le32_to_cpu(es->s_flags);
2696 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2697 sbi->s_hash_unsigned = 3;
2698 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2699 #ifdef __CHAR_UNSIGNED__
2700 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2701 sbi->s_hash_unsigned = 3;
2703 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2708 if (sbi->s_blocks_per_group > blocksize * 8) {
2709 ext4_msg(sb, KERN_ERR,
2710 "#blocks per group too big: %lu",
2711 sbi->s_blocks_per_group);
2714 if (sbi->s_inodes_per_group > blocksize * 8) {
2715 ext4_msg(sb, KERN_ERR,
2716 "#inodes per group too big: %lu",
2717 sbi->s_inodes_per_group);
2722 * Test whether we have more sectors than will fit in sector_t,
2723 * and whether the max offset is addressable by the page cache.
2725 if ((ext4_blocks_count(es) >
2726 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2727 (ext4_blocks_count(es) >
2728 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2729 ext4_msg(sb, KERN_ERR, "filesystem"
2730 " too large to mount safely on this system");
2731 if (sizeof(sector_t) < 8)
2732 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2737 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2740 /* check blocks count against device size */
2741 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2742 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2743 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2744 "exceeds size of device (%llu blocks)",
2745 ext4_blocks_count(es), blocks_count);
2750 * It makes no sense for the first data block to be beyond the end
2751 * of the filesystem.
2753 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2754 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2755 "block %u is beyond end of filesystem (%llu)",
2756 le32_to_cpu(es->s_first_data_block),
2757 ext4_blocks_count(es));
2760 blocks_count = (ext4_blocks_count(es) -
2761 le32_to_cpu(es->s_first_data_block) +
2762 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2763 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2764 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2765 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2766 "(block count %llu, first data block %u, "
2767 "blocks per group %lu)", sbi->s_groups_count,
2768 ext4_blocks_count(es),
2769 le32_to_cpu(es->s_first_data_block),
2770 EXT4_BLOCKS_PER_GROUP(sb));
2773 sbi->s_groups_count = blocks_count;
2774 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2775 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2776 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2777 EXT4_DESC_PER_BLOCK(sb);
2778 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2780 if (sbi->s_group_desc == NULL) {
2781 ext4_msg(sb, KERN_ERR, "not enough memory");
2785 #ifdef CONFIG_PROC_FS
2787 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2790 bgl_lock_init(sbi->s_blockgroup_lock);
2792 for (i = 0; i < db_count; i++) {
2793 block = descriptor_loc(sb, logical_sb_block, i);
2794 sbi->s_group_desc[i] = sb_bread(sb, block);
2795 if (!sbi->s_group_desc[i]) {
2796 ext4_msg(sb, KERN_ERR,
2797 "can't read group descriptor %d", i);
2802 if (!ext4_check_descriptors(sb)) {
2803 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2806 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2807 if (!ext4_fill_flex_info(sb)) {
2808 ext4_msg(sb, KERN_ERR,
2809 "unable to initialize "
2810 "flex_bg meta info!");
2814 sbi->s_gdb_count = db_count;
2815 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2816 spin_lock_init(&sbi->s_next_gen_lock);
2818 sbi->s_stripe = ext4_get_stripe_size(sbi);
2819 sbi->s_max_writeback_mb_bump = 128;
2822 * set up enough so that it can read an inode
2824 if (!test_opt(sb, NOLOAD) &&
2825 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2826 sb->s_op = &ext4_sops;
2828 sb->s_op = &ext4_nojournal_sops;
2829 sb->s_export_op = &ext4_export_ops;
2830 sb->s_xattr = ext4_xattr_handlers;
2832 sb->s_qcop = &ext4_qctl_operations;
2833 sb->dq_op = &ext4_quota_operations;
2835 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2836 mutex_init(&sbi->s_orphan_lock);
2837 mutex_init(&sbi->s_resize_lock);
2841 needs_recovery = (es->s_last_orphan != 0 ||
2842 EXT4_HAS_INCOMPAT_FEATURE(sb,
2843 EXT4_FEATURE_INCOMPAT_RECOVER));
2846 * The first inode we look at is the journal inode. Don't try
2847 * root first: it may be modified in the journal!
2849 if (!test_opt(sb, NOLOAD) &&
2850 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2851 if (ext4_load_journal(sb, es, journal_devnum))
2853 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2854 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2855 ext4_msg(sb, KERN_ERR, "required journal recovery "
2856 "suppressed and not mounted read-only");
2857 goto failed_mount_wq;
2859 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2860 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2861 sbi->s_journal = NULL;
2866 if (ext4_blocks_count(es) > 0xffffffffULL &&
2867 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2868 JBD2_FEATURE_INCOMPAT_64BIT)) {
2869 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2870 goto failed_mount_wq;
2873 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2874 jbd2_journal_set_features(sbi->s_journal,
2875 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2876 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2877 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2878 jbd2_journal_set_features(sbi->s_journal,
2879 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2880 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2881 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2883 jbd2_journal_clear_features(sbi->s_journal,
2884 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2885 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2888 /* We have now updated the journal if required, so we can
2889 * validate the data journaling mode. */
2890 switch (test_opt(sb, DATA_FLAGS)) {
2892 /* No mode set, assume a default based on the journal
2893 * capabilities: ORDERED_DATA if the journal can
2894 * cope, else JOURNAL_DATA
2896 if (jbd2_journal_check_available_features
2897 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2898 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2900 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2903 case EXT4_MOUNT_ORDERED_DATA:
2904 case EXT4_MOUNT_WRITEBACK_DATA:
2905 if (!jbd2_journal_check_available_features
2906 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2907 ext4_msg(sb, KERN_ERR, "Journal does not support "
2908 "requested data journaling mode");
2909 goto failed_mount_wq;
2914 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2917 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2918 ext4_count_free_blocks(sb));
2920 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2921 ext4_count_free_inodes(sb));
2923 err = percpu_counter_init(&sbi->s_dirs_counter,
2924 ext4_count_dirs(sb));
2926 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2928 ext4_msg(sb, KERN_ERR, "insufficient memory");
2929 goto failed_mount_wq;
2932 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2933 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2934 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2935 goto failed_mount_wq;
2939 * The jbd2_journal_load will have done any necessary log recovery,
2940 * so we can safely mount the rest of the filesystem now.
2943 root = ext4_iget(sb, EXT4_ROOT_INO);
2945 ext4_msg(sb, KERN_ERR, "get root inode failed");
2946 ret = PTR_ERR(root);
2949 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2951 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2954 sb->s_root = d_alloc_root(root);
2956 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2962 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2964 /* determine the minimum size of new large inodes, if present */
2965 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2966 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2967 EXT4_GOOD_OLD_INODE_SIZE;
2968 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2969 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2970 if (sbi->s_want_extra_isize <
2971 le16_to_cpu(es->s_want_extra_isize))
2972 sbi->s_want_extra_isize =
2973 le16_to_cpu(es->s_want_extra_isize);
2974 if (sbi->s_want_extra_isize <
2975 le16_to_cpu(es->s_min_extra_isize))
2976 sbi->s_want_extra_isize =
2977 le16_to_cpu(es->s_min_extra_isize);
2980 /* Check if enough inode space is available */
2981 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2982 sbi->s_inode_size) {
2983 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2984 EXT4_GOOD_OLD_INODE_SIZE;
2985 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2989 if (test_opt(sb, DELALLOC) &&
2990 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2991 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2992 "requested data journaling mode");
2993 clear_opt(sbi->s_mount_opt, DELALLOC);
2995 if (test_opt(sb, DIOREAD_NOLOCK)) {
2996 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2997 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2998 "option - requested data journaling mode");
2999 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3001 if (sb->s_blocksize < PAGE_SIZE) {
3002 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3003 "option - block size is too small");
3004 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3008 err = ext4_setup_system_zone(sb);
3010 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3016 err = ext4_mb_init(sb, needs_recovery);
3018 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3023 sbi->s_kobj.kset = ext4_kset;
3024 init_completion(&sbi->s_kobj_unregister);
3025 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3028 ext4_mb_release(sb);
3029 ext4_ext_release(sb);
3033 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3034 ext4_orphan_cleanup(sb, es);
3035 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3036 if (needs_recovery) {
3037 ext4_msg(sb, KERN_INFO, "recovery complete");
3038 ext4_mark_recovery_complete(sb, es);
3040 if (EXT4_SB(sb)->s_journal) {
3041 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3042 descr = " journalled data mode";
3043 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3044 descr = " ordered data mode";
3046 descr = " writeback data mode";
3048 descr = "out journal";
3050 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3051 "Opts: %s", descr, orig_data);
3059 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3063 ext4_msg(sb, KERN_ERR, "mount failed");
3064 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3066 ext4_release_system_zone(sb);
3067 if (sbi->s_journal) {
3068 jbd2_journal_destroy(sbi->s_journal);
3069 sbi->s_journal = NULL;
3071 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3072 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3073 percpu_counter_destroy(&sbi->s_dirs_counter);
3074 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3076 if (sbi->s_flex_groups) {
3077 if (is_vmalloc_addr(sbi->s_flex_groups))
3078 vfree(sbi->s_flex_groups);
3080 kfree(sbi->s_flex_groups);
3083 for (i = 0; i < db_count; i++)
3084 brelse(sbi->s_group_desc[i]);
3085 kfree(sbi->s_group_desc);
3088 remove_proc_entry(sb->s_id, ext4_proc_root);
3091 for (i = 0; i < MAXQUOTAS; i++)
3092 kfree(sbi->s_qf_names[i]);
3094 ext4_blkdev_remove(sbi);
3097 sb->s_fs_info = NULL;
3098 kfree(sbi->s_blockgroup_lock);
3106 * Setup any per-fs journal parameters now. We'll do this both on
3107 * initial mount, once the journal has been initialised but before we've
3108 * done any recovery; and again on any subsequent remount.
3110 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3112 struct ext4_sb_info *sbi = EXT4_SB(sb);
3114 journal->j_commit_interval = sbi->s_commit_interval;
3115 journal->j_min_batch_time = sbi->s_min_batch_time;
3116 journal->j_max_batch_time = sbi->s_max_batch_time;
3118 spin_lock(&journal->j_state_lock);
3119 if (test_opt(sb, BARRIER))
3120 journal->j_flags |= JBD2_BARRIER;
3122 journal->j_flags &= ~JBD2_BARRIER;
3123 if (test_opt(sb, DATA_ERR_ABORT))
3124 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3126 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3127 spin_unlock(&journal->j_state_lock);
3130 static journal_t *ext4_get_journal(struct super_block *sb,
3131 unsigned int journal_inum)
3133 struct inode *journal_inode;
3136 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3138 /* First, test for the existence of a valid inode on disk. Bad
3139 * things happen if we iget() an unused inode, as the subsequent
3140 * iput() will try to delete it. */
3142 journal_inode = ext4_iget(sb, journal_inum);
3143 if (IS_ERR(journal_inode)) {
3144 ext4_msg(sb, KERN_ERR, "no journal found");
3147 if (!journal_inode->i_nlink) {
3148 make_bad_inode(journal_inode);
3149 iput(journal_inode);
3150 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3154 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3155 journal_inode, journal_inode->i_size);
3156 if (!S_ISREG(journal_inode->i_mode)) {
3157 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3158 iput(journal_inode);
3162 journal = jbd2_journal_init_inode(journal_inode);
3164 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3165 iput(journal_inode);
3168 journal->j_private = sb;
3169 ext4_init_journal_params(sb, journal);
3173 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3176 struct buffer_head *bh;
3180 int hblock, blocksize;
3181 ext4_fsblk_t sb_block;
3182 unsigned long offset;
3183 struct ext4_super_block *es;
3184 struct block_device *bdev;
3186 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3188 bdev = ext4_blkdev_get(j_dev, sb);
3192 if (bd_claim(bdev, sb)) {
3193 ext4_msg(sb, KERN_ERR,
3194 "failed to claim external journal device");
3195 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3199 blocksize = sb->s_blocksize;
3200 hblock = bdev_logical_block_size(bdev);
3201 if (blocksize < hblock) {
3202 ext4_msg(sb, KERN_ERR,
3203 "blocksize too small for journal device");
3207 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3208 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3209 set_blocksize(bdev, blocksize);
3210 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3211 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3212 "external journal");
3216 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3217 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3218 !(le32_to_cpu(es->s_feature_incompat) &
3219 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3220 ext4_msg(sb, KERN_ERR, "external journal has "
3226 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3227 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3232 len = ext4_blocks_count(es);
3233 start = sb_block + 1;
3234 brelse(bh); /* we're done with the superblock */
3236 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3237 start, len, blocksize);
3239 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3242 journal->j_private = sb;
3243 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3244 wait_on_buffer(journal->j_sb_buffer);
3245 if (!buffer_uptodate(journal->j_sb_buffer)) {
3246 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3249 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3250 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3251 "user (unsupported) - %d",
3252 be32_to_cpu(journal->j_superblock->s_nr_users));
3255 EXT4_SB(sb)->journal_bdev = bdev;
3256 ext4_init_journal_params(sb, journal);
3260 jbd2_journal_destroy(journal);
3262 ext4_blkdev_put(bdev);
3266 static int ext4_load_journal(struct super_block *sb,
3267 struct ext4_super_block *es,
3268 unsigned long journal_devnum)
3271 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3274 int really_read_only;
3276 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3278 if (journal_devnum &&
3279 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3280 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3281 "numbers have changed");
3282 journal_dev = new_decode_dev(journal_devnum);
3284 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3286 really_read_only = bdev_read_only(sb->s_bdev);
3289 * Are we loading a blank journal or performing recovery after a
3290 * crash? For recovery, we need to check in advance whether we
3291 * can get read-write access to the device.
3293 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3294 if (sb->s_flags & MS_RDONLY) {
3295 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3296 "required on readonly filesystem");
3297 if (really_read_only) {
3298 ext4_msg(sb, KERN_ERR, "write access "
3299 "unavailable, cannot proceed");
3302 ext4_msg(sb, KERN_INFO, "write access will "
3303 "be enabled during recovery");
3307 if (journal_inum && journal_dev) {
3308 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3309 "and inode journals!");
3314 if (!(journal = ext4_get_journal(sb, journal_inum)))
3317 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3321 if (!(journal->j_flags & JBD2_BARRIER))
3322 ext4_msg(sb, KERN_INFO, "barriers disabled");
3324 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3325 err = jbd2_journal_update_format(journal);
3327 ext4_msg(sb, KERN_ERR, "error updating journal");
3328 jbd2_journal_destroy(journal);
3333 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3334 err = jbd2_journal_wipe(journal, !really_read_only);
3336 err = jbd2_journal_load(journal);
3339 ext4_msg(sb, KERN_ERR, "error loading journal");
3340 jbd2_journal_destroy(journal);
3344 EXT4_SB(sb)->s_journal = journal;
3345 ext4_clear_journal_err(sb, es);
3347 if (journal_devnum &&
3348 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3349 es->s_journal_dev = cpu_to_le32(journal_devnum);
3351 /* Make sure we flush the recovery flag to disk. */
3352 ext4_commit_super(sb, 1);
3358 static int ext4_commit_super(struct super_block *sb, int sync)
3360 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3361 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3366 if (buffer_write_io_error(sbh)) {
3368 * Oh, dear. A previous attempt to write the
3369 * superblock failed. This could happen because the
3370 * USB device was yanked out. Or it could happen to
3371 * be a transient write error and maybe the block will
3372 * be remapped. Nothing we can do but to retry the
3373 * write and hope for the best.
3375 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3376 "superblock detected");
3377 clear_buffer_write_io_error(sbh);
3378 set_buffer_uptodate(sbh);
3381 * If the file system is mounted read-only, don't update the
3382 * superblock write time. This avoids updating the superblock
3383 * write time when we are mounting the root file system
3384 * read/only but we need to replay the journal; at that point,
3385 * for people who are east of GMT and who make their clock
3386 * tick in localtime for Windows bug-for-bug compatibility,
3387 * the clock is set in the future, and this will cause e2fsck
3388 * to complain and force a full file system check.
3390 if (!(sb->s_flags & MS_RDONLY))
3391 es->s_wtime = cpu_to_le32(get_seconds());
3392 es->s_kbytes_written =
3393 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3394 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3395 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3396 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3397 &EXT4_SB(sb)->s_freeblocks_counter));
3398 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3399 &EXT4_SB(sb)->s_freeinodes_counter));
3401 BUFFER_TRACE(sbh, "marking dirty");
3402 mark_buffer_dirty(sbh);
3404 error = sync_dirty_buffer(sbh);
3408 error = buffer_write_io_error(sbh);
3410 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3412 clear_buffer_write_io_error(sbh);
3413 set_buffer_uptodate(sbh);
3420 * Have we just finished recovery? If so, and if we are mounting (or
3421 * remounting) the filesystem readonly, then we will end up with a
3422 * consistent fs on disk. Record that fact.
3424 static void ext4_mark_recovery_complete(struct super_block *sb,
3425 struct ext4_super_block *es)
3427 journal_t *journal = EXT4_SB(sb)->s_journal;
3429 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3430 BUG_ON(journal != NULL);
3433 jbd2_journal_lock_updates(journal);
3434 if (jbd2_journal_flush(journal) < 0)
3437 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3438 sb->s_flags & MS_RDONLY) {
3439 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3440 ext4_commit_super(sb, 1);
3444 jbd2_journal_unlock_updates(journal);
3448 * If we are mounting (or read-write remounting) a filesystem whose journal
3449 * has recorded an error from a previous lifetime, move that error to the
3450 * main filesystem now.
3452 static void ext4_clear_journal_err(struct super_block *sb,
3453 struct ext4_super_block *es)
3459 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3461 journal = EXT4_SB(sb)->s_journal;
3464 * Now check for any error status which may have been recorded in the
3465 * journal by a prior ext4_error() or ext4_abort()
3468 j_errno = jbd2_journal_errno(journal);
3472 errstr = ext4_decode_error(sb, j_errno, nbuf);
3473 ext4_warning(sb, "Filesystem error recorded "
3474 "from previous mount: %s", errstr);
3475 ext4_warning(sb, "Marking fs in need of filesystem check.");
3477 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3478 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3479 ext4_commit_super(sb, 1);
3481 jbd2_journal_clear_err(journal);
3486 * Force the running and committing transactions to commit,
3487 * and wait on the commit.
3489 int ext4_force_commit(struct super_block *sb)
3494 if (sb->s_flags & MS_RDONLY)
3497 journal = EXT4_SB(sb)->s_journal;
3499 vfs_check_frozen(sb, SB_FREEZE_WRITE);
3500 ret = ext4_journal_force_commit(journal);
3506 static void ext4_write_super(struct super_block *sb)
3509 ext4_commit_super(sb, 1);
3513 static int ext4_sync_fs(struct super_block *sb, int wait)
3517 struct ext4_sb_info *sbi = EXT4_SB(sb);
3519 trace_ext4_sync_fs(sb, wait);
3520 flush_workqueue(sbi->dio_unwritten_wq);
3521 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3523 jbd2_log_wait_commit(sbi->s_journal, target);
3529 * LVM calls this function before a (read-only) snapshot is created. This
3530 * gives us a chance to flush the journal completely and mark the fs clean.
3532 static int ext4_freeze(struct super_block *sb)
3537 if (sb->s_flags & MS_RDONLY)
3540 journal = EXT4_SB(sb)->s_journal;
3542 /* Now we set up the journal barrier. */
3543 jbd2_journal_lock_updates(journal);
3546 * Don't clear the needs_recovery flag if we failed to flush
3549 error = jbd2_journal_flush(journal);
3553 /* Journal blocked and flushed, clear needs_recovery flag. */
3554 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3555 error = ext4_commit_super(sb, 1);
3557 /* we rely on s_frozen to stop further updates */
3558 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3563 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3564 * flag here, even though the filesystem is not technically dirty yet.
3566 static int ext4_unfreeze(struct super_block *sb)
3568 if (sb->s_flags & MS_RDONLY)
3572 /* Reset the needs_recovery flag before the fs is unlocked. */
3573 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3574 ext4_commit_super(sb, 1);
3579 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3581 struct ext4_super_block *es;
3582 struct ext4_sb_info *sbi = EXT4_SB(sb);
3583 ext4_fsblk_t n_blocks_count = 0;
3584 unsigned long old_sb_flags;
3585 struct ext4_mount_options old_opts;
3586 int enable_quota = 0;
3588 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3593 char *orig_data = kstrdup(data, GFP_KERNEL);
3597 /* Store the original options */
3599 old_sb_flags = sb->s_flags;
3600 old_opts.s_mount_opt = sbi->s_mount_opt;
3601 old_opts.s_resuid = sbi->s_resuid;
3602 old_opts.s_resgid = sbi->s_resgid;
3603 old_opts.s_commit_interval = sbi->s_commit_interval;
3604 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3605 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3607 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3608 for (i = 0; i < MAXQUOTAS; i++)
3609 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3611 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3612 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3615 * Allow the "check" option to be passed as a remount option.
3617 if (!parse_options(data, sb, NULL, &journal_ioprio,
3618 &n_blocks_count, 1)) {
3623 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3624 ext4_abort(sb, "Abort forced by user");
3626 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3627 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3631 if (sbi->s_journal) {
3632 ext4_init_journal_params(sb, sbi->s_journal);
3633 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3636 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3637 n_blocks_count > ext4_blocks_count(es)) {
3638 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3643 if (*flags & MS_RDONLY) {
3644 err = dquot_suspend(sb, -1);
3649 * First of all, the unconditional stuff we have to do
3650 * to disable replay of the journal when we next remount
3652 sb->s_flags |= MS_RDONLY;
3655 * OK, test if we are remounting a valid rw partition
3656 * readonly, and if so set the rdonly flag and then
3657 * mark the partition as valid again.
3659 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3660 (sbi->s_mount_state & EXT4_VALID_FS))
3661 es->s_state = cpu_to_le16(sbi->s_mount_state);
3664 ext4_mark_recovery_complete(sb, es);
3666 /* Make sure we can mount this feature set readwrite */
3667 if (!ext4_feature_set_ok(sb, 0)) {
3672 * Make sure the group descriptor checksums
3673 * are sane. If they aren't, refuse to remount r/w.
3675 for (g = 0; g < sbi->s_groups_count; g++) {
3676 struct ext4_group_desc *gdp =
3677 ext4_get_group_desc(sb, g, NULL);
3679 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3680 ext4_msg(sb, KERN_ERR,
3681 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3682 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3683 le16_to_cpu(gdp->bg_checksum));
3690 * If we have an unprocessed orphan list hanging
3691 * around from a previously readonly bdev mount,
3692 * require a full umount/remount for now.
3694 if (es->s_last_orphan) {
3695 ext4_msg(sb, KERN_WARNING, "Couldn't "
3696 "remount RDWR because of unprocessed "
3697 "orphan inode list. Please "
3698 "umount/remount instead");
3704 * Mounting a RDONLY partition read-write, so reread
3705 * and store the current valid flag. (It may have
3706 * been changed by e2fsck since we originally mounted
3710 ext4_clear_journal_err(sb, es);
3711 sbi->s_mount_state = le16_to_cpu(es->s_state);
3712 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3714 if (!ext4_setup_super(sb, es, 0))
3715 sb->s_flags &= ~MS_RDONLY;
3719 ext4_setup_system_zone(sb);
3720 if (sbi->s_journal == NULL)
3721 ext4_commit_super(sb, 1);
3724 /* Release old quota file names */
3725 for (i = 0; i < MAXQUOTAS; i++)
3726 if (old_opts.s_qf_names[i] &&
3727 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3728 kfree(old_opts.s_qf_names[i]);
3733 dquot_resume(sb, -1);
3735 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3740 sb->s_flags = old_sb_flags;
3741 sbi->s_mount_opt = old_opts.s_mount_opt;
3742 sbi->s_resuid = old_opts.s_resuid;
3743 sbi->s_resgid = old_opts.s_resgid;
3744 sbi->s_commit_interval = old_opts.s_commit_interval;
3745 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3746 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3748 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3749 for (i = 0; i < MAXQUOTAS; i++) {
3750 if (sbi->s_qf_names[i] &&
3751 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3752 kfree(sbi->s_qf_names[i]);
3753 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3762 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3764 struct super_block *sb = dentry->d_sb;
3765 struct ext4_sb_info *sbi = EXT4_SB(sb);
3766 struct ext4_super_block *es = sbi->s_es;
3769 if (test_opt(sb, MINIX_DF)) {
3770 sbi->s_overhead_last = 0;
3771 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3772 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3773 ext4_fsblk_t overhead = 0;
3776 * Compute the overhead (FS structures). This is constant
3777 * for a given filesystem unless the number of block groups
3778 * changes so we cache the previous value until it does.
3782 * All of the blocks before first_data_block are
3785 overhead = le32_to_cpu(es->s_first_data_block);
3788 * Add the overhead attributed to the superblock and
3789 * block group descriptors. If the sparse superblocks
3790 * feature is turned on, then not all groups have this.
3792 for (i = 0; i < ngroups; i++) {
3793 overhead += ext4_bg_has_super(sb, i) +
3794 ext4_bg_num_gdb(sb, i);
3799 * Every block group has an inode bitmap, a block
3800 * bitmap, and an inode table.
3802 overhead += ngroups * (2 + sbi->s_itb_per_group);
3803 sbi->s_overhead_last = overhead;
3805 sbi->s_blocks_last = ext4_blocks_count(es);
3808 buf->f_type = EXT4_SUPER_MAGIC;
3809 buf->f_bsize = sb->s_blocksize;
3810 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3811 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3812 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3813 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3814 if (buf->f_bfree < ext4_r_blocks_count(es))
3816 buf->f_files = le32_to_cpu(es->s_inodes_count);
3817 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3818 buf->f_namelen = EXT4_NAME_LEN;
3819 fsid = le64_to_cpup((void *)es->s_uuid) ^
3820 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3821 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3822 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3827 /* Helper function for writing quotas on sync - we need to start transaction
3828 * before quota file is locked for write. Otherwise the are possible deadlocks:
3829 * Process 1 Process 2
3830 * ext4_create() quota_sync()
3831 * jbd2_journal_start() write_dquot()
3832 * dquot_initialize() down(dqio_mutex)
3833 * down(dqio_mutex) jbd2_journal_start()
3839 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3841 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3844 static int ext4_write_dquot(struct dquot *dquot)
3848 struct inode *inode;
3850 inode = dquot_to_inode(dquot);
3851 handle = ext4_journal_start(inode,
3852 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3854 return PTR_ERR(handle);
3855 ret = dquot_commit(dquot);
3856 err = ext4_journal_stop(handle);
3862 static int ext4_acquire_dquot(struct dquot *dquot)
3867 handle = ext4_journal_start(dquot_to_inode(dquot),
3868 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3870 return PTR_ERR(handle);
3871 ret = dquot_acquire(dquot);
3872 err = ext4_journal_stop(handle);
3878 static int ext4_release_dquot(struct dquot *dquot)
3883 handle = ext4_journal_start(dquot_to_inode(dquot),
3884 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3885 if (IS_ERR(handle)) {
3886 /* Release dquot anyway to avoid endless cycle in dqput() */
3887 dquot_release(dquot);
3888 return PTR_ERR(handle);
3890 ret = dquot_release(dquot);
3891 err = ext4_journal_stop(handle);
3897 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3899 /* Are we journaling quotas? */
3900 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3901 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3902 dquot_mark_dquot_dirty(dquot);
3903 return ext4_write_dquot(dquot);
3905 return dquot_mark_dquot_dirty(dquot);
3909 static int ext4_write_info(struct super_block *sb, int type)
3914 /* Data block + inode block */
3915 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3917 return PTR_ERR(handle);
3918 ret = dquot_commit_info(sb, type);
3919 err = ext4_journal_stop(handle);
3926 * Turn on quotas during mount time - we need to find
3927 * the quota file and such...
3929 static int ext4_quota_on_mount(struct super_block *sb, int type)
3931 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3932 EXT4_SB(sb)->s_jquota_fmt, type);
3936 * Standard function to be called on quota_on
3938 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3944 if (!test_opt(sb, QUOTA))
3947 err = kern_path(name, LOOKUP_FOLLOW, &path);
3951 /* Quotafile not on the same filesystem? */
3952 if (path.mnt->mnt_sb != sb) {
3956 /* Journaling quota? */
3957 if (EXT4_SB(sb)->s_qf_names[type]) {
3958 /* Quotafile not in fs root? */
3959 if (path.dentry->d_parent != sb->s_root)
3960 ext4_msg(sb, KERN_WARNING,
3961 "Quota file not on filesystem root. "
3962 "Journaled quota will not work");
3966 * When we journal data on quota file, we have to flush journal to see
3967 * all updates to the file when we bypass pagecache...
3969 if (EXT4_SB(sb)->s_journal &&
3970 ext4_should_journal_data(path.dentry->d_inode)) {
3972 * We don't need to lock updates but journal_flush() could
3973 * otherwise be livelocked...
3975 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3976 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3977 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3984 err = dquot_quota_on_path(sb, type, format_id, &path);
3989 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3990 * acquiring the locks... As quota files are never truncated and quota code
3991 * itself serializes the operations (and noone else should touch the files)
3992 * we don't have to be afraid of races */
3993 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3994 size_t len, loff_t off)
3996 struct inode *inode = sb_dqopt(sb)->files[type];
3997 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3999 int offset = off & (sb->s_blocksize - 1);
4002 struct buffer_head *bh;
4003 loff_t i_size = i_size_read(inode);
4007 if (off+len > i_size)
4010 while (toread > 0) {
4011 tocopy = sb->s_blocksize - offset < toread ?
4012 sb->s_blocksize - offset : toread;
4013 bh = ext4_bread(NULL, inode, blk, 0, &err);
4016 if (!bh) /* A hole? */
4017 memset(data, 0, tocopy);
4019 memcpy(data, bh->b_data+offset, tocopy);
4029 /* Write to quotafile (we know the transaction is already started and has
4030 * enough credits) */
4031 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4032 const char *data, size_t len, loff_t off)
4034 struct inode *inode = sb_dqopt(sb)->files[type];
4035 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4037 int offset = off & (sb->s_blocksize - 1);
4038 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
4039 struct buffer_head *bh;
4040 handle_t *handle = journal_current_handle();
4042 if (EXT4_SB(sb)->s_journal && !handle) {
4043 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4044 " cancelled because transaction is not started",
4045 (unsigned long long)off, (unsigned long long)len);
4049 * Since we account only one data block in transaction credits,
4050 * then it is impossible to cross a block boundary.
4052 if (sb->s_blocksize - offset < len) {
4053 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4054 " cancelled because not block aligned",
4055 (unsigned long long)off, (unsigned long long)len);
4059 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4060 bh = ext4_bread(handle, inode, blk, 1, &err);
4063 if (journal_quota) {
4064 err = ext4_journal_get_write_access(handle, bh);
4071 memcpy(bh->b_data+offset, data, len);
4072 flush_dcache_page(bh->b_page);
4075 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4077 /* Always do at least ordered writes for quotas */
4078 err = ext4_jbd2_file_inode(handle, inode);
4079 mark_buffer_dirty(bh);
4084 mutex_unlock(&inode->i_mutex);
4087 if (inode->i_size < off + len) {
4088 i_size_write(inode, off + len);
4089 EXT4_I(inode)->i_disksize = inode->i_size;
4091 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4092 ext4_mark_inode_dirty(handle, inode);
4093 mutex_unlock(&inode->i_mutex);
4099 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4100 const char *dev_name, void *data, struct vfsmount *mnt)
4102 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4105 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4106 static struct file_system_type ext2_fs_type = {
4107 .owner = THIS_MODULE,
4109 .get_sb = ext4_get_sb,
4110 .kill_sb = kill_block_super,
4111 .fs_flags = FS_REQUIRES_DEV,
4114 static inline void register_as_ext2(void)
4116 int err = register_filesystem(&ext2_fs_type);
4119 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4122 static inline void unregister_as_ext2(void)
4124 unregister_filesystem(&ext2_fs_type);
4126 MODULE_ALIAS("ext2");
4128 static inline void register_as_ext2(void) { }
4129 static inline void unregister_as_ext2(void) { }
4132 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4133 static inline void register_as_ext3(void)
4135 int err = register_filesystem(&ext3_fs_type);
4138 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4141 static inline void unregister_as_ext3(void)
4143 unregister_filesystem(&ext3_fs_type);
4145 MODULE_ALIAS("ext3");
4147 static inline void register_as_ext3(void) { }
4148 static inline void unregister_as_ext3(void) { }
4151 static struct file_system_type ext4_fs_type = {
4152 .owner = THIS_MODULE,
4154 .get_sb = ext4_get_sb,
4155 .kill_sb = kill_block_super,
4156 .fs_flags = FS_REQUIRES_DEV,
4159 static int __init init_ext4_fs(void)
4163 ext4_check_flag_values();
4164 err = init_ext4_system_zone();
4167 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4170 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4171 err = init_ext4_mballoc();
4175 err = init_ext4_xattr();
4178 err = init_inodecache();
4183 err = register_filesystem(&ext4_fs_type);
4188 unregister_as_ext2();
4189 unregister_as_ext3();
4190 destroy_inodecache();
4194 exit_ext4_mballoc();
4196 remove_proc_entry("fs/ext4", NULL);
4197 kset_unregister(ext4_kset);
4199 exit_ext4_system_zone();
4203 static void __exit exit_ext4_fs(void)
4205 unregister_as_ext2();
4206 unregister_as_ext3();
4207 unregister_filesystem(&ext4_fs_type);
4208 destroy_inodecache();
4210 exit_ext4_mballoc();
4211 remove_proc_entry("fs/ext4", NULL);
4212 kset_unregister(ext4_kset);
4213 exit_ext4_system_zone();
4216 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4217 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4218 MODULE_LICENSE("GPL");
4219 module_init(init_ext4_fs)
4220 module_exit(exit_ext4_fs)