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
, 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
, err
);
286 void ext4_journal_abort_handle(const char *caller
, const char *err_fn
,
287 struct buffer_head
*bh
, handle_t
*handle
, int err
)
290 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
292 BUG_ON(!ext4_handle_valid(handle
));
295 BUFFER_TRACE(bh
, "abort");
300 if (is_handle_aborted(handle
))
303 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
304 caller
, errstr
, err_fn
);
306 jbd2_journal_abort_handle(handle
);
309 /* Deal with the reporting of failure conditions on a filesystem such as
310 * inconsistencies detected or read IO failures.
312 * On ext2, we can store the error state of the filesystem in the
313 * superblock. That is not possible on ext4, because we may have other
314 * write ordering constraints on the superblock which prevent us from
315 * writing it out straight away; and given that the journal is about to
316 * be aborted, we can't rely on the current, or future, transactions to
317 * write out the superblock safely.
319 * We'll just use the jbd2_journal_abort() error code to record an error in
320 * the journal instead. On recovery, the journal will complain about
321 * that error until we've noted it down and cleared it.
324 static void ext4_handle_error(struct super_block
*sb
)
326 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
328 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
329 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
331 if (sb
->s_flags
& MS_RDONLY
)
334 if (!test_opt(sb
, ERRORS_CONT
)) {
335 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
337 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
339 jbd2_journal_abort(journal
, -EIO
);
341 if (test_opt(sb
, ERRORS_RO
)) {
342 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
343 sb
->s_flags
|= MS_RDONLY
;
345 ext4_commit_super(sb
, 1);
346 if (test_opt(sb
, ERRORS_PANIC
))
347 panic("EXT4-fs (device %s): panic forced after error\n",
351 void __ext4_error(struct super_block
*sb
, const char *function
,
352 const char *fmt
, ...)
357 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
362 ext4_handle_error(sb
);
365 void ext4_error_inode(const char *function
, struct inode
*inode
,
366 const char *fmt
, ...)
371 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: inode #%lu: (comm %s) ",
372 inode
->i_sb
->s_id
, function
, inode
->i_ino
, current
->comm
);
377 ext4_handle_error(inode
->i_sb
);
380 void ext4_error_file(const char *function
, struct file
*file
,
381 const char *fmt
, ...)
384 struct inode
*inode
= file
->f_dentry
->d_inode
;
385 char pathname
[80], *path
;
388 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
392 "EXT4-fs error (device %s): %s: inode #%lu (comm %s path %s): ",
393 inode
->i_sb
->s_id
, function
, inode
->i_ino
, current
->comm
, path
);
398 ext4_handle_error(inode
->i_sb
);
401 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
408 errstr
= "IO failure";
411 errstr
= "Out of memory";
414 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
415 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
416 errstr
= "Journal has aborted";
418 errstr
= "Readonly filesystem";
421 /* If the caller passed in an extra buffer for unknown
422 * errors, textualise them now. Else we just return
425 /* Check for truncated error codes... */
426 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
435 /* __ext4_std_error decodes expected errors from journaling functions
436 * automatically and invokes the appropriate error response. */
438 void __ext4_std_error(struct super_block
*sb
, const char *function
, int errno
)
443 /* Special case: if the error is EROFS, and we're not already
444 * inside a transaction, then there's really no point in logging
446 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
447 (sb
->s_flags
& MS_RDONLY
))
450 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
451 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s: %s\n",
452 sb
->s_id
, function
, errstr
);
454 ext4_handle_error(sb
);
458 * ext4_abort is a much stronger failure handler than ext4_error. The
459 * abort function may be used to deal with unrecoverable failures such
460 * as journal IO errors or ENOMEM at a critical moment in log management.
462 * We unconditionally force the filesystem into an ABORT|READONLY state,
463 * unless the error response on the fs has been set to panic in which
464 * case we take the easy way out and panic immediately.
467 void __ext4_abort(struct super_block
*sb
, const char *function
,
468 const char *fmt
, ...)
473 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
478 if (test_opt(sb
, ERRORS_PANIC
))
479 panic("EXT4-fs panic from previous error\n");
481 if (sb
->s_flags
& MS_RDONLY
)
484 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
485 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
486 sb
->s_flags
|= MS_RDONLY
;
487 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
488 if (EXT4_SB(sb
)->s_journal
)
489 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
492 void ext4_msg (struct super_block
* sb
, const char *prefix
,
493 const char *fmt
, ...)
498 printk("%sEXT4-fs (%s): ", prefix
, sb
->s_id
);
504 void __ext4_warning(struct super_block
*sb
, const char *function
,
505 const char *fmt
, ...)
510 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s: ",
517 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
518 struct super_block
*sb
, ext4_group_t grp
,
519 unsigned long ino
, ext4_fsblk_t block
,
520 const char *fmt
, ...)
525 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
528 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u",
529 sb
->s_id
, function
, line
, grp
);
531 printk("inode %lu: ", ino
);
533 printk("block %llu:", (unsigned long long) block
);
538 if (test_opt(sb
, ERRORS_CONT
)) {
539 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
540 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
541 ext4_commit_super(sb
, 0);
544 ext4_unlock_group(sb
, grp
);
545 ext4_handle_error(sb
);
547 * We only get here in the ERRORS_RO case; relocking the group
548 * may be dangerous, but nothing bad will happen since the
549 * filesystem will have already been marked read/only and the
550 * journal has been aborted. We return 1 as a hint to callers
551 * who might what to use the return value from
552 * ext4_grp_locked_error() to distinguish beween the
553 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
554 * aggressively from the ext4 function in question, with a
555 * more appropriate error code.
557 ext4_lock_group(sb
, grp
);
561 void ext4_update_dynamic_rev(struct super_block
*sb
)
563 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
565 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
569 "updating to rev %d because of new feature flag, "
570 "running e2fsck is recommended",
573 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
574 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
575 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
576 /* leave es->s_feature_*compat flags alone */
577 /* es->s_uuid will be set by e2fsck if empty */
580 * The rest of the superblock fields should be zero, and if not it
581 * means they are likely already in use, so leave them alone. We
582 * can leave it up to e2fsck to clean up any inconsistencies there.
587 * Open the external journal device
589 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
591 struct block_device
*bdev
;
592 char b
[BDEVNAME_SIZE
];
594 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
600 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
601 __bdevname(dev
, b
), PTR_ERR(bdev
));
606 * Release the journal device
608 static int ext4_blkdev_put(struct block_device
*bdev
)
611 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
614 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
616 struct block_device
*bdev
;
619 bdev
= sbi
->journal_bdev
;
621 ret
= ext4_blkdev_put(bdev
);
622 sbi
->journal_bdev
= NULL
;
627 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
629 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
632 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
636 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
637 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
639 printk(KERN_ERR
"sb_info orphan list:\n");
640 list_for_each(l
, &sbi
->s_orphan
) {
641 struct inode
*inode
= orphan_list_entry(l
);
643 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
644 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
645 inode
->i_mode
, inode
->i_nlink
,
650 static void ext4_put_super(struct super_block
*sb
)
652 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
653 struct ext4_super_block
*es
= sbi
->s_es
;
656 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
658 flush_workqueue(sbi
->dio_unwritten_wq
);
659 destroy_workqueue(sbi
->dio_unwritten_wq
);
664 ext4_commit_super(sb
, 1);
666 if (sbi
->s_journal
) {
667 err
= jbd2_journal_destroy(sbi
->s_journal
);
668 sbi
->s_journal
= NULL
;
670 ext4_abort(sb
, "Couldn't clean up the journal");
673 ext4_release_system_zone(sb
);
675 ext4_ext_release(sb
);
676 ext4_xattr_put_super(sb
);
678 if (!(sb
->s_flags
& MS_RDONLY
)) {
679 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
680 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
681 ext4_commit_super(sb
, 1);
684 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
686 kobject_del(&sbi
->s_kobj
);
688 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
689 brelse(sbi
->s_group_desc
[i
]);
690 kfree(sbi
->s_group_desc
);
691 if (is_vmalloc_addr(sbi
->s_flex_groups
))
692 vfree(sbi
->s_flex_groups
);
694 kfree(sbi
->s_flex_groups
);
695 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
696 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
697 percpu_counter_destroy(&sbi
->s_dirs_counter
);
698 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
701 for (i
= 0; i
< MAXQUOTAS
; i
++)
702 kfree(sbi
->s_qf_names
[i
]);
705 /* Debugging code just in case the in-memory inode orphan list
706 * isn't empty. The on-disk one can be non-empty if we've
707 * detected an error and taken the fs readonly, but the
708 * in-memory list had better be clean by this point. */
709 if (!list_empty(&sbi
->s_orphan
))
710 dump_orphan_list(sb
, sbi
);
711 J_ASSERT(list_empty(&sbi
->s_orphan
));
713 invalidate_bdev(sb
->s_bdev
);
714 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
716 * Invalidate the journal device's buffers. We don't want them
717 * floating about in memory - the physical journal device may
718 * hotswapped, and it breaks the `ro-after' testing code.
720 sync_blockdev(sbi
->journal_bdev
);
721 invalidate_bdev(sbi
->journal_bdev
);
722 ext4_blkdev_remove(sbi
);
724 sb
->s_fs_info
= NULL
;
726 * Now that we are completely done shutting down the
727 * superblock, we need to actually destroy the kobject.
731 kobject_put(&sbi
->s_kobj
);
732 wait_for_completion(&sbi
->s_kobj_unregister
);
733 kfree(sbi
->s_blockgroup_lock
);
737 static struct kmem_cache
*ext4_inode_cachep
;
740 * Called inside transaction, so use GFP_NOFS
742 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
744 struct ext4_inode_info
*ei
;
746 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
750 ei
->vfs_inode
.i_version
= 1;
751 ei
->vfs_inode
.i_data
.writeback_index
= 0;
752 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
753 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
754 spin_lock_init(&ei
->i_prealloc_lock
);
756 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
757 * therefore it can be null here. Don't check it, just initialize
760 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
761 ei
->i_reserved_data_blocks
= 0;
762 ei
->i_reserved_meta_blocks
= 0;
763 ei
->i_allocated_meta_blocks
= 0;
764 ei
->i_da_metadata_calc_len
= 0;
765 ei
->i_delalloc_reserved_flag
= 0;
766 spin_lock_init(&(ei
->i_block_reservation_lock
));
768 ei
->i_reserved_quota
= 0;
770 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
771 spin_lock_init(&ei
->i_completed_io_lock
);
772 ei
->cur_aio_dio
= NULL
;
774 ei
->i_datasync_tid
= 0;
776 return &ei
->vfs_inode
;
779 static void ext4_destroy_inode(struct inode
*inode
)
781 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
782 ext4_msg(inode
->i_sb
, KERN_ERR
,
783 "Inode %lu (%p): orphan list check failed!",
784 inode
->i_ino
, EXT4_I(inode
));
785 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
786 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
790 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
793 static void init_once(void *foo
)
795 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
797 INIT_LIST_HEAD(&ei
->i_orphan
);
798 #ifdef CONFIG_EXT4_FS_XATTR
799 init_rwsem(&ei
->xattr_sem
);
801 init_rwsem(&ei
->i_data_sem
);
802 inode_init_once(&ei
->vfs_inode
);
805 static int init_inodecache(void)
807 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
808 sizeof(struct ext4_inode_info
),
809 0, (SLAB_RECLAIM_ACCOUNT
|
812 if (ext4_inode_cachep
== NULL
)
817 static void destroy_inodecache(void)
819 kmem_cache_destroy(ext4_inode_cachep
);
822 static void ext4_clear_inode(struct inode
*inode
)
825 ext4_discard_preallocations(inode
);
826 if (EXT4_JOURNAL(inode
))
827 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
828 &EXT4_I(inode
)->jinode
);
831 static inline void ext4_show_quota_options(struct seq_file
*seq
,
832 struct super_block
*sb
)
834 #if defined(CONFIG_QUOTA)
835 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
837 if (sbi
->s_jquota_fmt
) {
840 switch (sbi
->s_jquota_fmt
) {
851 seq_printf(seq
, ",jqfmt=%s", fmtname
);
854 if (sbi
->s_qf_names
[USRQUOTA
])
855 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
857 if (sbi
->s_qf_names
[GRPQUOTA
])
858 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
860 if (test_opt(sb
, USRQUOTA
))
861 seq_puts(seq
, ",usrquota");
863 if (test_opt(sb
, GRPQUOTA
))
864 seq_puts(seq
, ",grpquota");
870 * - it's set to a non-default value OR
871 * - if the per-sb default is different from the global default
873 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
876 unsigned long def_mount_opts
;
877 struct super_block
*sb
= vfs
->mnt_sb
;
878 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
879 struct ext4_super_block
*es
= sbi
->s_es
;
881 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
882 def_errors
= le16_to_cpu(es
->s_errors
);
884 if (sbi
->s_sb_block
!= 1)
885 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
886 if (test_opt(sb
, MINIX_DF
))
887 seq_puts(seq
, ",minixdf");
888 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
889 seq_puts(seq
, ",grpid");
890 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
891 seq_puts(seq
, ",nogrpid");
892 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
893 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
894 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
896 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
897 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
898 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
900 if (test_opt(sb
, ERRORS_RO
)) {
901 if (def_errors
== EXT4_ERRORS_PANIC
||
902 def_errors
== EXT4_ERRORS_CONTINUE
) {
903 seq_puts(seq
, ",errors=remount-ro");
906 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
907 seq_puts(seq
, ",errors=continue");
908 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
909 seq_puts(seq
, ",errors=panic");
910 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
911 seq_puts(seq
, ",nouid32");
912 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
913 seq_puts(seq
, ",debug");
914 if (test_opt(sb
, OLDALLOC
))
915 seq_puts(seq
, ",oldalloc");
916 #ifdef CONFIG_EXT4_FS_XATTR
917 if (test_opt(sb
, XATTR_USER
) &&
918 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
919 seq_puts(seq
, ",user_xattr");
920 if (!test_opt(sb
, XATTR_USER
) &&
921 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
922 seq_puts(seq
, ",nouser_xattr");
925 #ifdef CONFIG_EXT4_FS_POSIX_ACL
926 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
927 seq_puts(seq
, ",acl");
928 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
929 seq_puts(seq
, ",noacl");
931 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
932 seq_printf(seq
, ",commit=%u",
933 (unsigned) (sbi
->s_commit_interval
/ HZ
));
935 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
936 seq_printf(seq
, ",min_batch_time=%u",
937 (unsigned) sbi
->s_min_batch_time
);
939 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
940 seq_printf(seq
, ",max_batch_time=%u",
941 (unsigned) sbi
->s_min_batch_time
);
945 * We're changing the default of barrier mount option, so
946 * let's always display its mount state so it's clear what its
949 seq_puts(seq
, ",barrier=");
950 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
951 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
952 seq_puts(seq
, ",journal_async_commit");
953 else if (test_opt(sb
, JOURNAL_CHECKSUM
))
954 seq_puts(seq
, ",journal_checksum");
955 if (test_opt(sb
, I_VERSION
))
956 seq_puts(seq
, ",i_version");
957 if (!test_opt(sb
, DELALLOC
))
958 seq_puts(seq
, ",nodelalloc");
962 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
964 * journal mode get enabled in different ways
965 * So just print the value even if we didn't specify it
967 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
968 seq_puts(seq
, ",data=journal");
969 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
970 seq_puts(seq
, ",data=ordered");
971 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
972 seq_puts(seq
, ",data=writeback");
974 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
975 seq_printf(seq
, ",inode_readahead_blks=%u",
976 sbi
->s_inode_readahead_blks
);
978 if (test_opt(sb
, DATA_ERR_ABORT
))
979 seq_puts(seq
, ",data_err=abort");
981 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
982 seq_puts(seq
, ",noauto_da_alloc");
984 if (test_opt(sb
, DISCARD
))
985 seq_puts(seq
, ",discard");
987 if (test_opt(sb
, NOLOAD
))
988 seq_puts(seq
, ",norecovery");
990 if (test_opt(sb
, DIOREAD_NOLOCK
))
991 seq_puts(seq
, ",dioread_nolock");
993 ext4_show_quota_options(seq
, sb
);
998 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
999 u64 ino
, u32 generation
)
1001 struct inode
*inode
;
1003 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1004 return ERR_PTR(-ESTALE
);
1005 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1006 return ERR_PTR(-ESTALE
);
1008 /* iget isn't really right if the inode is currently unallocated!!
1010 * ext4_read_inode will return a bad_inode if the inode had been
1011 * deleted, so we should be safe.
1013 * Currently we don't know the generation for parent directory, so
1014 * a generation of 0 means "accept any"
1016 inode
= ext4_iget(sb
, ino
);
1018 return ERR_CAST(inode
);
1019 if (generation
&& inode
->i_generation
!= generation
) {
1021 return ERR_PTR(-ESTALE
);
1027 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1028 int fh_len
, int fh_type
)
1030 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1031 ext4_nfs_get_inode
);
1034 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1035 int fh_len
, int fh_type
)
1037 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1038 ext4_nfs_get_inode
);
1042 * Try to release metadata pages (indirect blocks, directories) which are
1043 * mapped via the block device. Since these pages could have journal heads
1044 * which would prevent try_to_free_buffers() from freeing them, we must use
1045 * jbd2 layer's try_to_free_buffers() function to release them.
1047 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1050 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1052 WARN_ON(PageChecked(page
));
1053 if (!page_has_buffers(page
))
1056 return jbd2_journal_try_to_free_buffers(journal
, page
,
1057 wait
& ~__GFP_WAIT
);
1058 return try_to_free_buffers(page
);
1062 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1063 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1065 static int ext4_write_dquot(struct dquot
*dquot
);
1066 static int ext4_acquire_dquot(struct dquot
*dquot
);
1067 static int ext4_release_dquot(struct dquot
*dquot
);
1068 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1069 static int ext4_write_info(struct super_block
*sb
, int type
);
1070 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1072 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1073 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1074 size_t len
, loff_t off
);
1075 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1076 const char *data
, size_t len
, loff_t off
);
1078 static const struct dquot_operations ext4_quota_operations
= {
1080 .get_reserved_space
= ext4_get_reserved_space
,
1082 .write_dquot
= ext4_write_dquot
,
1083 .acquire_dquot
= ext4_acquire_dquot
,
1084 .release_dquot
= ext4_release_dquot
,
1085 .mark_dirty
= ext4_mark_dquot_dirty
,
1086 .write_info
= ext4_write_info
,
1087 .alloc_dquot
= dquot_alloc
,
1088 .destroy_dquot
= dquot_destroy
,
1091 static const struct quotactl_ops ext4_qctl_operations
= {
1092 .quota_on
= ext4_quota_on
,
1093 .quota_off
= dquot_quota_off
,
1094 .quota_sync
= dquot_quota_sync
,
1095 .get_info
= dquot_get_dqinfo
,
1096 .set_info
= dquot_set_dqinfo
,
1097 .get_dqblk
= dquot_get_dqblk
,
1098 .set_dqblk
= dquot_set_dqblk
1102 static const struct super_operations ext4_sops
= {
1103 .alloc_inode
= ext4_alloc_inode
,
1104 .destroy_inode
= ext4_destroy_inode
,
1105 .write_inode
= ext4_write_inode
,
1106 .dirty_inode
= ext4_dirty_inode
,
1107 .delete_inode
= ext4_delete_inode
,
1108 .put_super
= ext4_put_super
,
1109 .sync_fs
= ext4_sync_fs
,
1110 .freeze_fs
= ext4_freeze
,
1111 .unfreeze_fs
= ext4_unfreeze
,
1112 .statfs
= ext4_statfs
,
1113 .remount_fs
= ext4_remount
,
1114 .clear_inode
= ext4_clear_inode
,
1115 .show_options
= ext4_show_options
,
1117 .quota_read
= ext4_quota_read
,
1118 .quota_write
= ext4_quota_write
,
1120 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1123 static const struct super_operations ext4_nojournal_sops
= {
1124 .alloc_inode
= ext4_alloc_inode
,
1125 .destroy_inode
= ext4_destroy_inode
,
1126 .write_inode
= ext4_write_inode
,
1127 .dirty_inode
= ext4_dirty_inode
,
1128 .delete_inode
= ext4_delete_inode
,
1129 .write_super
= ext4_write_super
,
1130 .put_super
= ext4_put_super
,
1131 .statfs
= ext4_statfs
,
1132 .remount_fs
= ext4_remount
,
1133 .clear_inode
= ext4_clear_inode
,
1134 .show_options
= ext4_show_options
,
1136 .quota_read
= ext4_quota_read
,
1137 .quota_write
= ext4_quota_write
,
1139 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1142 static const struct export_operations ext4_export_ops
= {
1143 .fh_to_dentry
= ext4_fh_to_dentry
,
1144 .fh_to_parent
= ext4_fh_to_parent
,
1145 .get_parent
= ext4_get_parent
,
1149 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1150 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1151 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1152 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1153 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1154 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1155 Opt_journal_update
, Opt_journal_dev
,
1156 Opt_journal_checksum
, Opt_journal_async_commit
,
1157 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1158 Opt_data_err_abort
, Opt_data_err_ignore
,
1159 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1160 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1161 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1162 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1163 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
1164 Opt_block_validity
, Opt_noblock_validity
,
1165 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1166 Opt_dioread_nolock
, Opt_dioread_lock
,
1167 Opt_discard
, Opt_nodiscard
,
1170 static const match_table_t tokens
= {
1171 {Opt_bsd_df
, "bsddf"},
1172 {Opt_minix_df
, "minixdf"},
1173 {Opt_grpid
, "grpid"},
1174 {Opt_grpid
, "bsdgroups"},
1175 {Opt_nogrpid
, "nogrpid"},
1176 {Opt_nogrpid
, "sysvgroups"},
1177 {Opt_resgid
, "resgid=%u"},
1178 {Opt_resuid
, "resuid=%u"},
1180 {Opt_err_cont
, "errors=continue"},
1181 {Opt_err_panic
, "errors=panic"},
1182 {Opt_err_ro
, "errors=remount-ro"},
1183 {Opt_nouid32
, "nouid32"},
1184 {Opt_debug
, "debug"},
1185 {Opt_oldalloc
, "oldalloc"},
1186 {Opt_orlov
, "orlov"},
1187 {Opt_user_xattr
, "user_xattr"},
1188 {Opt_nouser_xattr
, "nouser_xattr"},
1190 {Opt_noacl
, "noacl"},
1191 {Opt_noload
, "noload"},
1192 {Opt_noload
, "norecovery"},
1195 {Opt_commit
, "commit=%u"},
1196 {Opt_min_batch_time
, "min_batch_time=%u"},
1197 {Opt_max_batch_time
, "max_batch_time=%u"},
1198 {Opt_journal_update
, "journal=update"},
1199 {Opt_journal_dev
, "journal_dev=%u"},
1200 {Opt_journal_checksum
, "journal_checksum"},
1201 {Opt_journal_async_commit
, "journal_async_commit"},
1202 {Opt_abort
, "abort"},
1203 {Opt_data_journal
, "data=journal"},
1204 {Opt_data_ordered
, "data=ordered"},
1205 {Opt_data_writeback
, "data=writeback"},
1206 {Opt_data_err_abort
, "data_err=abort"},
1207 {Opt_data_err_ignore
, "data_err=ignore"},
1208 {Opt_offusrjquota
, "usrjquota="},
1209 {Opt_usrjquota
, "usrjquota=%s"},
1210 {Opt_offgrpjquota
, "grpjquota="},
1211 {Opt_grpjquota
, "grpjquota=%s"},
1212 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1213 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1214 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1215 {Opt_grpquota
, "grpquota"},
1216 {Opt_noquota
, "noquota"},
1217 {Opt_quota
, "quota"},
1218 {Opt_usrquota
, "usrquota"},
1219 {Opt_barrier
, "barrier=%u"},
1220 {Opt_barrier
, "barrier"},
1221 {Opt_nobarrier
, "nobarrier"},
1222 {Opt_i_version
, "i_version"},
1223 {Opt_stripe
, "stripe=%u"},
1224 {Opt_resize
, "resize"},
1225 {Opt_delalloc
, "delalloc"},
1226 {Opt_nodelalloc
, "nodelalloc"},
1227 {Opt_block_validity
, "block_validity"},
1228 {Opt_noblock_validity
, "noblock_validity"},
1229 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1230 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1231 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1232 {Opt_auto_da_alloc
, "auto_da_alloc"},
1233 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1234 {Opt_dioread_nolock
, "dioread_nolock"},
1235 {Opt_dioread_lock
, "dioread_lock"},
1236 {Opt_discard
, "discard"},
1237 {Opt_nodiscard
, "nodiscard"},
1241 static ext4_fsblk_t
get_sb_block(void **data
)
1243 ext4_fsblk_t sb_block
;
1244 char *options
= (char *) *data
;
1246 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1247 return 1; /* Default location */
1250 /* TODO: use simple_strtoll with >32bit ext4 */
1251 sb_block
= simple_strtoul(options
, &options
, 0);
1252 if (*options
&& *options
!= ',') {
1253 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1257 if (*options
== ',')
1259 *data
= (void *) options
;
1264 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1265 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1266 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1269 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1271 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1274 if (sb_any_quota_loaded(sb
) &&
1275 !sbi
->s_qf_names
[qtype
]) {
1276 ext4_msg(sb
, KERN_ERR
,
1277 "Cannot change journaled "
1278 "quota options when quota turned on");
1281 qname
= match_strdup(args
);
1283 ext4_msg(sb
, KERN_ERR
,
1284 "Not enough memory for storing quotafile name");
1287 if (sbi
->s_qf_names
[qtype
] &&
1288 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1289 ext4_msg(sb
, KERN_ERR
,
1290 "%s quota file already specified", QTYPE2NAME(qtype
));
1294 sbi
->s_qf_names
[qtype
] = qname
;
1295 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1296 ext4_msg(sb
, KERN_ERR
,
1297 "quotafile must be on filesystem root");
1298 kfree(sbi
->s_qf_names
[qtype
]);
1299 sbi
->s_qf_names
[qtype
] = NULL
;
1302 set_opt(sbi
->s_mount_opt
, QUOTA
);
1306 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1309 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1311 if (sb_any_quota_loaded(sb
) &&
1312 sbi
->s_qf_names
[qtype
]) {
1313 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1314 " when quota turned on");
1318 * The space will be released later when all options are confirmed
1321 sbi
->s_qf_names
[qtype
] = NULL
;
1326 static int parse_options(char *options
, struct super_block
*sb
,
1327 unsigned long *journal_devnum
,
1328 unsigned int *journal_ioprio
,
1329 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1331 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1333 substring_t args
[MAX_OPT_ARGS
];
1343 while ((p
= strsep(&options
, ",")) != NULL
) {
1349 * Initialize args struct so we know whether arg was
1350 * found; some options take optional arguments.
1352 args
[0].to
= args
[0].from
= 0;
1353 token
= match_token(p
, tokens
, args
);
1356 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1357 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1360 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1361 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1365 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1366 set_opt(sbi
->s_mount_opt
, GRPID
);
1370 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1371 clear_opt(sbi
->s_mount_opt
, GRPID
);
1375 if (match_int(&args
[0], &option
))
1377 sbi
->s_resuid
= option
;
1380 if (match_int(&args
[0], &option
))
1382 sbi
->s_resgid
= option
;
1385 /* handled by get_sb_block() instead of here */
1386 /* *sb_block = match_int(&args[0]); */
1389 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1390 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1391 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1394 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1395 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1396 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1399 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1400 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1401 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1404 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1407 set_opt(sbi
->s_mount_opt
, DEBUG
);
1410 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1413 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1415 #ifdef CONFIG_EXT4_FS_XATTR
1416 case Opt_user_xattr
:
1417 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1419 case Opt_nouser_xattr
:
1420 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1423 case Opt_user_xattr
:
1424 case Opt_nouser_xattr
:
1425 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1428 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1430 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1433 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1438 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1441 case Opt_journal_update
:
1443 /* Eventually we will want to be able to create
1444 a journal file here. For now, only allow the
1445 user to specify an existing inode to be the
1448 ext4_msg(sb
, KERN_ERR
,
1449 "Cannot specify journal on remount");
1452 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1454 case Opt_journal_dev
:
1456 ext4_msg(sb
, KERN_ERR
,
1457 "Cannot specify journal on remount");
1460 if (match_int(&args
[0], &option
))
1462 *journal_devnum
= option
;
1464 case Opt_journal_checksum
:
1465 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1467 case Opt_journal_async_commit
:
1468 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1469 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1472 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1475 if (match_int(&args
[0], &option
))
1480 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1481 sbi
->s_commit_interval
= HZ
* option
;
1483 case Opt_max_batch_time
:
1484 if (match_int(&args
[0], &option
))
1489 option
= EXT4_DEF_MAX_BATCH_TIME
;
1490 sbi
->s_max_batch_time
= option
;
1492 case Opt_min_batch_time
:
1493 if (match_int(&args
[0], &option
))
1497 sbi
->s_min_batch_time
= option
;
1499 case Opt_data_journal
:
1500 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1502 case Opt_data_ordered
:
1503 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1505 case Opt_data_writeback
:
1506 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1509 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1510 ext4_msg(sb
, KERN_ERR
,
1511 "Cannot change data mode on remount");
1515 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
1516 sbi
->s_mount_opt
|= data_opt
;
1519 case Opt_data_err_abort
:
1520 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1522 case Opt_data_err_ignore
:
1523 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1527 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1531 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1534 case Opt_offusrjquota
:
1535 if (!clear_qf_name(sb
, USRQUOTA
))
1538 case Opt_offgrpjquota
:
1539 if (!clear_qf_name(sb
, GRPQUOTA
))
1543 case Opt_jqfmt_vfsold
:
1544 qfmt
= QFMT_VFS_OLD
;
1546 case Opt_jqfmt_vfsv0
:
1549 case Opt_jqfmt_vfsv1
:
1552 if (sb_any_quota_loaded(sb
) &&
1553 sbi
->s_jquota_fmt
!= qfmt
) {
1554 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1555 "journaled quota options when "
1559 sbi
->s_jquota_fmt
= qfmt
;
1563 set_opt(sbi
->s_mount_opt
, QUOTA
);
1564 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1567 set_opt(sbi
->s_mount_opt
, QUOTA
);
1568 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1571 if (sb_any_quota_loaded(sb
)) {
1572 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1573 "options when quota turned on");
1576 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1577 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1578 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1584 ext4_msg(sb
, KERN_ERR
,
1585 "quota options not supported");
1589 case Opt_offusrjquota
:
1590 case Opt_offgrpjquota
:
1591 case Opt_jqfmt_vfsold
:
1592 case Opt_jqfmt_vfsv0
:
1593 case Opt_jqfmt_vfsv1
:
1594 ext4_msg(sb
, KERN_ERR
,
1595 "journaled quota options not supported");
1601 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1604 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1608 if (match_int(&args
[0], &option
))
1611 option
= 1; /* No argument, default to 1 */
1613 set_opt(sbi
->s_mount_opt
, BARRIER
);
1615 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1621 ext4_msg(sb
, KERN_ERR
,
1622 "resize option only available "
1626 if (match_int(&args
[0], &option
) != 0)
1628 *n_blocks_count
= option
;
1631 ext4_msg(sb
, KERN_WARNING
,
1632 "Ignoring deprecated nobh option");
1635 ext4_msg(sb
, KERN_WARNING
,
1636 "Ignoring deprecated bh option");
1639 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1640 sb
->s_flags
|= MS_I_VERSION
;
1642 case Opt_nodelalloc
:
1643 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1646 if (match_int(&args
[0], &option
))
1650 sbi
->s_stripe
= option
;
1653 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1655 case Opt_block_validity
:
1656 set_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1658 case Opt_noblock_validity
:
1659 clear_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1661 case Opt_inode_readahead_blks
:
1662 if (match_int(&args
[0], &option
))
1664 if (option
< 0 || option
> (1 << 30))
1666 if (!is_power_of_2(option
)) {
1667 ext4_msg(sb
, KERN_ERR
,
1668 "EXT4-fs: inode_readahead_blks"
1669 " must be a power of 2");
1672 sbi
->s_inode_readahead_blks
= option
;
1674 case Opt_journal_ioprio
:
1675 if (match_int(&args
[0], &option
))
1677 if (option
< 0 || option
> 7)
1679 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1682 case Opt_noauto_da_alloc
:
1683 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1685 case Opt_auto_da_alloc
:
1687 if (match_int(&args
[0], &option
))
1690 option
= 1; /* No argument, default to 1 */
1692 clear_opt(sbi
->s_mount_opt
, NO_AUTO_DA_ALLOC
);
1694 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1697 set_opt(sbi
->s_mount_opt
, DISCARD
);
1700 clear_opt(sbi
->s_mount_opt
, DISCARD
);
1702 case Opt_dioread_nolock
:
1703 set_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
1705 case Opt_dioread_lock
:
1706 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
1709 ext4_msg(sb
, KERN_ERR
,
1710 "Unrecognized mount option \"%s\" "
1711 "or missing value", p
);
1716 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1717 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1718 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1720 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1721 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1723 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1724 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1729 if (!sbi
->s_jquota_fmt
) {
1730 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1735 if (sbi
->s_jquota_fmt
) {
1736 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1737 "specified with no journaling "
1746 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1749 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1752 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1753 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1754 "forcing read-only mode");
1759 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1760 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1761 "running e2fsck is recommended");
1762 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1763 ext4_msg(sb
, KERN_WARNING
,
1764 "warning: mounting fs with errors, "
1765 "running e2fsck is recommended");
1766 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1767 le16_to_cpu(es
->s_mnt_count
) >=
1768 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1769 ext4_msg(sb
, KERN_WARNING
,
1770 "warning: maximal mount count reached, "
1771 "running e2fsck is recommended");
1772 else if (le32_to_cpu(es
->s_checkinterval
) &&
1773 (le32_to_cpu(es
->s_lastcheck
) +
1774 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1775 ext4_msg(sb
, KERN_WARNING
,
1776 "warning: checktime reached, "
1777 "running e2fsck is recommended");
1778 if (!sbi
->s_journal
)
1779 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1780 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1781 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1782 le16_add_cpu(&es
->s_mnt_count
, 1);
1783 es
->s_mtime
= cpu_to_le32(get_seconds());
1784 ext4_update_dynamic_rev(sb
);
1786 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1788 ext4_commit_super(sb
, 1);
1789 if (test_opt(sb
, DEBUG
))
1790 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1791 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1793 sbi
->s_groups_count
,
1794 EXT4_BLOCKS_PER_GROUP(sb
),
1795 EXT4_INODES_PER_GROUP(sb
),
1801 static int ext4_fill_flex_info(struct super_block
*sb
)
1803 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1804 struct ext4_group_desc
*gdp
= NULL
;
1805 ext4_group_t flex_group_count
;
1806 ext4_group_t flex_group
;
1807 int groups_per_flex
= 0;
1811 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1812 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1814 if (groups_per_flex
< 2) {
1815 sbi
->s_log_groups_per_flex
= 0;
1819 /* We allocate both existing and potentially added groups */
1820 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1821 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1822 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1823 size
= flex_group_count
* sizeof(struct flex_groups
);
1824 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1825 if (sbi
->s_flex_groups
== NULL
) {
1826 sbi
->s_flex_groups
= vmalloc(size
);
1827 if (sbi
->s_flex_groups
)
1828 memset(sbi
->s_flex_groups
, 0, size
);
1830 if (sbi
->s_flex_groups
== NULL
) {
1831 ext4_msg(sb
, KERN_ERR
, "not enough memory for "
1832 "%u flex groups", flex_group_count
);
1836 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1837 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1839 flex_group
= ext4_flex_group(sbi
, i
);
1840 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1841 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1842 atomic_add(ext4_free_blks_count(sb
, gdp
),
1843 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1844 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1845 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1853 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1854 struct ext4_group_desc
*gdp
)
1858 if (sbi
->s_es
->s_feature_ro_compat
&
1859 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1860 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1861 __le32 le_group
= cpu_to_le32(block_group
);
1863 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1864 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1865 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1866 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1867 /* for checksum of struct ext4_group_desc do the rest...*/
1868 if ((sbi
->s_es
->s_feature_incompat
&
1869 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1870 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1871 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1872 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1876 return cpu_to_le16(crc
);
1879 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1880 struct ext4_group_desc
*gdp
)
1882 if ((sbi
->s_es
->s_feature_ro_compat
&
1883 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1884 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1890 /* Called at mount-time, super-block is locked */
1891 static int ext4_check_descriptors(struct super_block
*sb
)
1893 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1894 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1895 ext4_fsblk_t last_block
;
1896 ext4_fsblk_t block_bitmap
;
1897 ext4_fsblk_t inode_bitmap
;
1898 ext4_fsblk_t inode_table
;
1899 int flexbg_flag
= 0;
1902 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1905 ext4_debug("Checking group descriptors");
1907 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1908 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1910 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1911 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1913 last_block
= first_block
+
1914 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1916 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1917 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1918 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1919 "Block bitmap for group %u not in group "
1920 "(block %llu)!", i
, block_bitmap
);
1923 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1924 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1925 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1926 "Inode bitmap for group %u not in group "
1927 "(block %llu)!", i
, inode_bitmap
);
1930 inode_table
= ext4_inode_table(sb
, gdp
);
1931 if (inode_table
< first_block
||
1932 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1933 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1934 "Inode table for group %u not in group "
1935 "(block %llu)!", i
, inode_table
);
1938 ext4_lock_group(sb
, i
);
1939 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1940 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1941 "Checksum for group %u failed (%u!=%u)",
1942 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1943 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1944 if (!(sb
->s_flags
& MS_RDONLY
)) {
1945 ext4_unlock_group(sb
, i
);
1949 ext4_unlock_group(sb
, i
);
1951 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1954 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1955 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
1959 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1960 * the superblock) which were deleted from all directories, but held open by
1961 * a process at the time of a crash. We walk the list and try to delete these
1962 * inodes at recovery time (only with a read-write filesystem).
1964 * In order to keep the orphan inode chain consistent during traversal (in
1965 * case of crash during recovery), we link each inode into the superblock
1966 * orphan list_head and handle it the same way as an inode deletion during
1967 * normal operation (which journals the operations for us).
1969 * We only do an iget() and an iput() on each inode, which is very safe if we
1970 * accidentally point at an in-use or already deleted inode. The worst that
1971 * can happen in this case is that we get a "bit already cleared" message from
1972 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1973 * e2fsck was run on this filesystem, and it must have already done the orphan
1974 * inode cleanup for us, so we can safely abort without any further action.
1976 static void ext4_orphan_cleanup(struct super_block
*sb
,
1977 struct ext4_super_block
*es
)
1979 unsigned int s_flags
= sb
->s_flags
;
1980 int nr_orphans
= 0, nr_truncates
= 0;
1984 if (!es
->s_last_orphan
) {
1985 jbd_debug(4, "no orphan inodes to clean up\n");
1989 if (bdev_read_only(sb
->s_bdev
)) {
1990 ext4_msg(sb
, KERN_ERR
, "write access "
1991 "unavailable, skipping orphan cleanup");
1995 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1996 if (es
->s_last_orphan
)
1997 jbd_debug(1, "Errors on filesystem, "
1998 "clearing orphan list.\n");
1999 es
->s_last_orphan
= 0;
2000 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2004 if (s_flags
& MS_RDONLY
) {
2005 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2006 sb
->s_flags
&= ~MS_RDONLY
;
2009 /* Needed for iput() to work correctly and not trash data */
2010 sb
->s_flags
|= MS_ACTIVE
;
2011 /* Turn on quotas so that they are updated correctly */
2012 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2013 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2014 int ret
= ext4_quota_on_mount(sb
, i
);
2016 ext4_msg(sb
, KERN_ERR
,
2017 "Cannot turn on journaled "
2018 "quota: error %d", ret
);
2023 while (es
->s_last_orphan
) {
2024 struct inode
*inode
;
2026 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2027 if (IS_ERR(inode
)) {
2028 es
->s_last_orphan
= 0;
2032 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2033 dquot_initialize(inode
);
2034 if (inode
->i_nlink
) {
2035 ext4_msg(sb
, KERN_DEBUG
,
2036 "%s: truncating inode %lu to %lld bytes",
2037 __func__
, inode
->i_ino
, inode
->i_size
);
2038 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2039 inode
->i_ino
, inode
->i_size
);
2040 ext4_truncate(inode
);
2043 ext4_msg(sb
, KERN_DEBUG
,
2044 "%s: deleting unreferenced inode %lu",
2045 __func__
, inode
->i_ino
);
2046 jbd_debug(2, "deleting unreferenced inode %lu\n",
2050 iput(inode
); /* The delete magic happens here! */
2053 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2056 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2057 PLURAL(nr_orphans
));
2059 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2060 PLURAL(nr_truncates
));
2062 /* Turn quotas off */
2063 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2064 if (sb_dqopt(sb
)->files
[i
])
2065 dquot_quota_off(sb
, i
);
2068 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2072 * Maximal extent format file size.
2073 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2074 * extent format containers, within a sector_t, and within i_blocks
2075 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2076 * so that won't be a limiting factor.
2078 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2080 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2083 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2085 /* small i_blocks in vfs inode? */
2086 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2088 * CONFIG_LBDAF is not enabled implies the inode
2089 * i_block represent total blocks in 512 bytes
2090 * 32 == size of vfs inode i_blocks * 8
2092 upper_limit
= (1LL << 32) - 1;
2094 /* total blocks in file system block size */
2095 upper_limit
>>= (blkbits
- 9);
2096 upper_limit
<<= blkbits
;
2099 /* 32-bit extent-start container, ee_block */
2104 /* Sanity check against vm- & vfs- imposed limits */
2105 if (res
> upper_limit
)
2112 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2113 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2114 * We need to be 1 filesystem block less than the 2^48 sector limit.
2116 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2118 loff_t res
= EXT4_NDIR_BLOCKS
;
2121 /* This is calculated to be the largest file size for a dense, block
2122 * mapped file such that the file's total number of 512-byte sectors,
2123 * including data and all indirect blocks, does not exceed (2^48 - 1).
2125 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2126 * number of 512-byte sectors of the file.
2129 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2131 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2132 * the inode i_block field represents total file blocks in
2133 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2135 upper_limit
= (1LL << 32) - 1;
2137 /* total blocks in file system block size */
2138 upper_limit
>>= (bits
- 9);
2142 * We use 48 bit ext4_inode i_blocks
2143 * With EXT4_HUGE_FILE_FL set the i_blocks
2144 * represent total number of blocks in
2145 * file system block size
2147 upper_limit
= (1LL << 48) - 1;
2151 /* indirect blocks */
2153 /* double indirect blocks */
2154 meta_blocks
+= 1 + (1LL << (bits
-2));
2155 /* tripple indirect blocks */
2156 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2158 upper_limit
-= meta_blocks
;
2159 upper_limit
<<= bits
;
2161 res
+= 1LL << (bits
-2);
2162 res
+= 1LL << (2*(bits
-2));
2163 res
+= 1LL << (3*(bits
-2));
2165 if (res
> upper_limit
)
2168 if (res
> MAX_LFS_FILESIZE
)
2169 res
= MAX_LFS_FILESIZE
;
2174 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2175 ext4_fsblk_t logical_sb_block
, int nr
)
2177 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2178 ext4_group_t bg
, first_meta_bg
;
2181 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2183 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2185 return logical_sb_block
+ nr
+ 1;
2186 bg
= sbi
->s_desc_per_block
* nr
;
2187 if (ext4_bg_has_super(sb
, bg
))
2190 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2194 * ext4_get_stripe_size: Get the stripe size.
2195 * @sbi: In memory super block info
2197 * If we have specified it via mount option, then
2198 * use the mount option value. If the value specified at mount time is
2199 * greater than the blocks per group use the super block value.
2200 * If the super block value is greater than blocks per group return 0.
2201 * Allocator needs it be less than blocks per group.
2204 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2206 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2207 unsigned long stripe_width
=
2208 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2210 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2211 return sbi
->s_stripe
;
2213 if (stripe_width
<= sbi
->s_blocks_per_group
)
2214 return stripe_width
;
2216 if (stride
<= sbi
->s_blocks_per_group
)
2225 struct attribute attr
;
2226 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2227 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2228 const char *, size_t);
2232 static int parse_strtoul(const char *buf
,
2233 unsigned long max
, unsigned long *value
)
2237 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2238 endp
= skip_spaces(endp
);
2239 if (*endp
|| *value
> max
)
2245 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2246 struct ext4_sb_info
*sbi
,
2249 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2250 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2253 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2254 struct ext4_sb_info
*sbi
, char *buf
)
2256 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2258 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2259 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2260 sbi
->s_sectors_written_start
) >> 1);
2263 static ssize_t
lifetime_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
, "%llu\n",
2269 (unsigned long long)(sbi
->s_kbytes_written
+
2270 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2271 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2274 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2275 struct ext4_sb_info
*sbi
,
2276 const char *buf
, size_t count
)
2280 if (parse_strtoul(buf
, 0x40000000, &t
))
2283 if (!is_power_of_2(t
))
2286 sbi
->s_inode_readahead_blks
= t
;
2290 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2291 struct ext4_sb_info
*sbi
, char *buf
)
2293 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2295 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2298 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2299 struct ext4_sb_info
*sbi
,
2300 const char *buf
, size_t count
)
2302 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2305 if (parse_strtoul(buf
, 0xffffffff, &t
))
2311 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2312 static struct ext4_attr ext4_attr_##_name = { \
2313 .attr = {.name = __stringify(_name), .mode = _mode }, \
2316 .offset = offsetof(struct ext4_sb_info, _elname), \
2318 #define EXT4_ATTR(name, mode, show, store) \
2319 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2321 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2322 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2323 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2324 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2325 #define ATTR_LIST(name) &ext4_attr_##name.attr
2327 EXT4_RO_ATTR(delayed_allocation_blocks
);
2328 EXT4_RO_ATTR(session_write_kbytes
);
2329 EXT4_RO_ATTR(lifetime_write_kbytes
);
2330 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2331 inode_readahead_blks_store
, s_inode_readahead_blks
);
2332 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2333 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2334 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2335 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2336 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2337 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2338 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2339 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2341 static struct attribute
*ext4_attrs
[] = {
2342 ATTR_LIST(delayed_allocation_blocks
),
2343 ATTR_LIST(session_write_kbytes
),
2344 ATTR_LIST(lifetime_write_kbytes
),
2345 ATTR_LIST(inode_readahead_blks
),
2346 ATTR_LIST(inode_goal
),
2347 ATTR_LIST(mb_stats
),
2348 ATTR_LIST(mb_max_to_scan
),
2349 ATTR_LIST(mb_min_to_scan
),
2350 ATTR_LIST(mb_order2_req
),
2351 ATTR_LIST(mb_stream_req
),
2352 ATTR_LIST(mb_group_prealloc
),
2353 ATTR_LIST(max_writeback_mb_bump
),
2357 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2358 struct attribute
*attr
, char *buf
)
2360 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2362 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2364 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2367 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2368 struct attribute
*attr
,
2369 const char *buf
, size_t len
)
2371 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2373 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2375 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2378 static void ext4_sb_release(struct kobject
*kobj
)
2380 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2382 complete(&sbi
->s_kobj_unregister
);
2386 static const struct sysfs_ops ext4_attr_ops
= {
2387 .show
= ext4_attr_show
,
2388 .store
= ext4_attr_store
,
2391 static struct kobj_type ext4_ktype
= {
2392 .default_attrs
= ext4_attrs
,
2393 .sysfs_ops
= &ext4_attr_ops
,
2394 .release
= ext4_sb_release
,
2398 * Check whether this filesystem can be mounted based on
2399 * the features present and the RDONLY/RDWR mount requested.
2400 * Returns 1 if this filesystem can be mounted as requested,
2401 * 0 if it cannot be.
2403 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2405 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2406 ext4_msg(sb
, KERN_ERR
,
2407 "Couldn't mount because of "
2408 "unsupported optional features (%x)",
2409 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2410 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2417 /* Check that feature set is OK for a read-write mount */
2418 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2419 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2420 "unsupported optional features (%x)",
2421 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2422 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2426 * Large file size enabled file system can only be mounted
2427 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2429 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2430 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2431 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2432 "cannot be mounted RDWR without "
2440 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
2441 __releases(kernel_lock
)
2442 __acquires(kernel_lock
)
2444 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
2445 struct buffer_head
*bh
;
2446 struct ext4_super_block
*es
= NULL
;
2447 struct ext4_sb_info
*sbi
;
2449 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2450 ext4_fsblk_t logical_sb_block
;
2451 unsigned long offset
= 0;
2452 unsigned long journal_devnum
= 0;
2453 unsigned long def_mount_opts
;
2459 unsigned int db_count
;
2461 int needs_recovery
, has_huge_files
;
2464 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2466 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
2470 sbi
->s_blockgroup_lock
=
2471 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
2472 if (!sbi
->s_blockgroup_lock
) {
2476 sb
->s_fs_info
= sbi
;
2477 sbi
->s_mount_opt
= 0;
2478 sbi
->s_resuid
= EXT4_DEF_RESUID
;
2479 sbi
->s_resgid
= EXT4_DEF_RESGID
;
2480 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
2481 sbi
->s_sb_block
= sb_block
;
2482 sbi
->s_sectors_written_start
= part_stat_read(sb
->s_bdev
->bd_part
,
2487 /* Cleanup superblock name */
2488 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
2491 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
2493 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
2498 * The ext4 superblock will not be buffer aligned for other than 1kB
2499 * block sizes. We need to calculate the offset from buffer start.
2501 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
2502 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2503 offset
= do_div(logical_sb_block
, blocksize
);
2505 logical_sb_block
= sb_block
;
2508 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
2509 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
2513 * Note: s_es must be initialized as soon as possible because
2514 * some ext4 macro-instructions depend on its value
2516 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2518 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
2519 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
2521 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
2523 /* Set defaults before we parse the mount options */
2524 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
2525 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
2526 set_opt(sbi
->s_mount_opt
, DEBUG
);
2527 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
2528 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
2530 set_opt(sbi
->s_mount_opt
, GRPID
);
2532 if (def_mount_opts
& EXT4_DEFM_UID16
)
2533 set_opt(sbi
->s_mount_opt
, NO_UID32
);
2534 #ifdef CONFIG_EXT4_FS_XATTR
2535 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
2536 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
2538 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2539 if (def_mount_opts
& EXT4_DEFM_ACL
)
2540 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
2542 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
2543 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2544 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
2545 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2546 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
2547 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2549 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2550 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2551 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2552 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2554 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2556 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2557 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2558 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
2559 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
2560 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
2562 set_opt(sbi
->s_mount_opt
, BARRIER
);
2565 * enable delayed allocation by default
2566 * Use -o nodelalloc to turn it off
2568 if (!IS_EXT3_SB(sb
))
2569 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2571 if (!parse_options((char *) data
, sb
, &journal_devnum
,
2572 &journal_ioprio
, NULL
, 0))
2575 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2576 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
2578 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2579 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2580 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2581 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2582 ext4_msg(sb
, KERN_WARNING
,
2583 "feature flags set on rev 0 fs, "
2584 "running e2fsck is recommended");
2587 * Check feature flags regardless of the revision level, since we
2588 * previously didn't change the revision level when setting the flags,
2589 * so there is a chance incompat flags are set on a rev 0 filesystem.
2591 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
2594 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2596 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2597 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2598 ext4_msg(sb
, KERN_ERR
,
2599 "Unsupported filesystem blocksize %d", blocksize
);
2603 if (sb
->s_blocksize
!= blocksize
) {
2604 /* Validate the filesystem blocksize */
2605 if (!sb_set_blocksize(sb
, blocksize
)) {
2606 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
2612 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2613 offset
= do_div(logical_sb_block
, blocksize
);
2614 bh
= sb_bread(sb
, logical_sb_block
);
2616 ext4_msg(sb
, KERN_ERR
,
2617 "Can't read superblock on 2nd try");
2620 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2622 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2623 ext4_msg(sb
, KERN_ERR
,
2624 "Magic mismatch, very weird!");
2629 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2630 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
2631 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
2633 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
2635 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2636 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2637 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2639 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2640 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2641 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2642 (!is_power_of_2(sbi
->s_inode_size
)) ||
2643 (sbi
->s_inode_size
> blocksize
)) {
2644 ext4_msg(sb
, KERN_ERR
,
2645 "unsupported inode size: %d",
2649 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2650 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2653 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2654 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2655 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2656 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2657 !is_power_of_2(sbi
->s_desc_size
)) {
2658 ext4_msg(sb
, KERN_ERR
,
2659 "unsupported descriptor size %lu",
2664 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2666 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2667 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2668 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2671 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2672 if (sbi
->s_inodes_per_block
== 0)
2674 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2675 sbi
->s_inodes_per_block
;
2676 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2678 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2679 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2680 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2682 for (i
= 0; i
< 4; i
++)
2683 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2684 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2685 i
= le32_to_cpu(es
->s_flags
);
2686 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
2687 sbi
->s_hash_unsigned
= 3;
2688 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
2689 #ifdef __CHAR_UNSIGNED__
2690 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
2691 sbi
->s_hash_unsigned
= 3;
2693 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
2698 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2699 ext4_msg(sb
, KERN_ERR
,
2700 "#blocks per group too big: %lu",
2701 sbi
->s_blocks_per_group
);
2704 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2705 ext4_msg(sb
, KERN_ERR
,
2706 "#inodes per group too big: %lu",
2707 sbi
->s_inodes_per_group
);
2712 * Test whether we have more sectors than will fit in sector_t,
2713 * and whether the max offset is addressable by the page cache.
2715 if ((ext4_blocks_count(es
) >
2716 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) ||
2717 (ext4_blocks_count(es
) >
2718 (pgoff_t
)(~0ULL) >> (PAGE_CACHE_SHIFT
- sb
->s_blocksize_bits
))) {
2719 ext4_msg(sb
, KERN_ERR
, "filesystem"
2720 " too large to mount safely on this system");
2721 if (sizeof(sector_t
) < 8)
2722 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
2727 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2730 /* check blocks count against device size */
2731 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
2732 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
2733 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
2734 "exceeds size of device (%llu blocks)",
2735 ext4_blocks_count(es
), blocks_count
);
2740 * It makes no sense for the first data block to be beyond the end
2741 * of the filesystem.
2743 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
2744 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
2745 "block %u is beyond end of filesystem (%llu)",
2746 le32_to_cpu(es
->s_first_data_block
),
2747 ext4_blocks_count(es
));
2750 blocks_count
= (ext4_blocks_count(es
) -
2751 le32_to_cpu(es
->s_first_data_block
) +
2752 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2753 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2754 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
2755 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
2756 "(block count %llu, first data block %u, "
2757 "blocks per group %lu)", sbi
->s_groups_count
,
2758 ext4_blocks_count(es
),
2759 le32_to_cpu(es
->s_first_data_block
),
2760 EXT4_BLOCKS_PER_GROUP(sb
));
2763 sbi
->s_groups_count
= blocks_count
;
2764 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
2765 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
2766 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2767 EXT4_DESC_PER_BLOCK(sb
);
2768 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2770 if (sbi
->s_group_desc
== NULL
) {
2771 ext4_msg(sb
, KERN_ERR
, "not enough memory");
2775 #ifdef CONFIG_PROC_FS
2777 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2780 bgl_lock_init(sbi
->s_blockgroup_lock
);
2782 for (i
= 0; i
< db_count
; i
++) {
2783 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2784 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2785 if (!sbi
->s_group_desc
[i
]) {
2786 ext4_msg(sb
, KERN_ERR
,
2787 "can't read group descriptor %d", i
);
2792 if (!ext4_check_descriptors(sb
)) {
2793 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
2796 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2797 if (!ext4_fill_flex_info(sb
)) {
2798 ext4_msg(sb
, KERN_ERR
,
2799 "unable to initialize "
2800 "flex_bg meta info!");
2804 sbi
->s_gdb_count
= db_count
;
2805 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2806 spin_lock_init(&sbi
->s_next_gen_lock
);
2808 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2809 sbi
->s_max_writeback_mb_bump
= 128;
2812 * set up enough so that it can read an inode
2814 if (!test_opt(sb
, NOLOAD
) &&
2815 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
2816 sb
->s_op
= &ext4_sops
;
2818 sb
->s_op
= &ext4_nojournal_sops
;
2819 sb
->s_export_op
= &ext4_export_ops
;
2820 sb
->s_xattr
= ext4_xattr_handlers
;
2822 sb
->s_qcop
= &ext4_qctl_operations
;
2823 sb
->dq_op
= &ext4_quota_operations
;
2825 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2826 mutex_init(&sbi
->s_orphan_lock
);
2827 mutex_init(&sbi
->s_resize_lock
);
2831 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2832 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2833 EXT4_FEATURE_INCOMPAT_RECOVER
));
2836 * The first inode we look at is the journal inode. Don't try
2837 * root first: it may be modified in the journal!
2839 if (!test_opt(sb
, NOLOAD
) &&
2840 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2841 if (ext4_load_journal(sb
, es
, journal_devnum
))
2843 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
2844 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2845 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
2846 "suppressed and not mounted read-only");
2847 goto failed_mount_wq
;
2849 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
2850 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2851 sbi
->s_journal
= NULL
;
2856 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2857 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2858 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2859 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
2860 goto failed_mount_wq
;
2863 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2864 jbd2_journal_set_features(sbi
->s_journal
,
2865 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2866 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2867 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2868 jbd2_journal_set_features(sbi
->s_journal
,
2869 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2870 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2871 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2873 jbd2_journal_clear_features(sbi
->s_journal
,
2874 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2875 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2878 /* We have now updated the journal if required, so we can
2879 * validate the data journaling mode. */
2880 switch (test_opt(sb
, DATA_FLAGS
)) {
2882 /* No mode set, assume a default based on the journal
2883 * capabilities: ORDERED_DATA if the journal can
2884 * cope, else JOURNAL_DATA
2886 if (jbd2_journal_check_available_features
2887 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2888 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2890 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2893 case EXT4_MOUNT_ORDERED_DATA
:
2894 case EXT4_MOUNT_WRITEBACK_DATA
:
2895 if (!jbd2_journal_check_available_features
2896 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2897 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
2898 "requested data journaling mode");
2899 goto failed_mount_wq
;
2904 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
2907 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2908 ext4_count_free_blocks(sb
));
2910 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2911 ext4_count_free_inodes(sb
));
2913 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2914 ext4_count_dirs(sb
));
2916 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
2918 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
2919 goto failed_mount_wq
;
2922 EXT4_SB(sb
)->dio_unwritten_wq
= create_workqueue("ext4-dio-unwritten");
2923 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
2924 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
2925 goto failed_mount_wq
;
2929 * The jbd2_journal_load will have done any necessary log recovery,
2930 * so we can safely mount the rest of the filesystem now.
2933 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2935 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
2936 ret
= PTR_ERR(root
);
2939 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2941 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
2944 sb
->s_root
= d_alloc_root(root
);
2946 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
2952 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
2954 /* determine the minimum size of new large inodes, if present */
2955 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2956 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2957 EXT4_GOOD_OLD_INODE_SIZE
;
2958 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2959 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2960 if (sbi
->s_want_extra_isize
<
2961 le16_to_cpu(es
->s_want_extra_isize
))
2962 sbi
->s_want_extra_isize
=
2963 le16_to_cpu(es
->s_want_extra_isize
);
2964 if (sbi
->s_want_extra_isize
<
2965 le16_to_cpu(es
->s_min_extra_isize
))
2966 sbi
->s_want_extra_isize
=
2967 le16_to_cpu(es
->s_min_extra_isize
);
2970 /* Check if enough inode space is available */
2971 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2972 sbi
->s_inode_size
) {
2973 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2974 EXT4_GOOD_OLD_INODE_SIZE
;
2975 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
2979 if (test_opt(sb
, DELALLOC
) &&
2980 (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)) {
2981 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
2982 "requested data journaling mode");
2983 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2985 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
2986 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
2987 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
2988 "option - requested data journaling mode");
2989 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
2991 if (sb
->s_blocksize
< PAGE_SIZE
) {
2992 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
2993 "option - block size is too small");
2994 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
2998 err
= ext4_setup_system_zone(sb
);
3000 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3006 err
= ext4_mb_init(sb
, needs_recovery
);
3008 ext4_msg(sb
, KERN_ERR
, "failed to initalize mballoc (%d)",
3013 sbi
->s_kobj
.kset
= ext4_kset
;
3014 init_completion(&sbi
->s_kobj_unregister
);
3015 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3018 ext4_mb_release(sb
);
3019 ext4_ext_release(sb
);
3023 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3024 ext4_orphan_cleanup(sb
, es
);
3025 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3026 if (needs_recovery
) {
3027 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3028 ext4_mark_recovery_complete(sb
, es
);
3030 if (EXT4_SB(sb
)->s_journal
) {
3031 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3032 descr
= " journalled data mode";
3033 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3034 descr
= " ordered data mode";
3036 descr
= " writeback data mode";
3038 descr
= "out journal";
3040 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3041 "Opts: %s", descr
, orig_data
);
3049 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3053 ext4_msg(sb
, KERN_ERR
, "mount failed");
3054 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3056 ext4_release_system_zone(sb
);
3057 if (sbi
->s_journal
) {
3058 jbd2_journal_destroy(sbi
->s_journal
);
3059 sbi
->s_journal
= NULL
;
3061 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
3062 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3063 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3064 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
3066 if (sbi
->s_flex_groups
) {
3067 if (is_vmalloc_addr(sbi
->s_flex_groups
))
3068 vfree(sbi
->s_flex_groups
);
3070 kfree(sbi
->s_flex_groups
);
3073 for (i
= 0; i
< db_count
; i
++)
3074 brelse(sbi
->s_group_desc
[i
]);
3075 kfree(sbi
->s_group_desc
);
3078 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3081 for (i
= 0; i
< MAXQUOTAS
; i
++)
3082 kfree(sbi
->s_qf_names
[i
]);
3084 ext4_blkdev_remove(sbi
);
3087 sb
->s_fs_info
= NULL
;
3088 kfree(sbi
->s_blockgroup_lock
);
3096 * Setup any per-fs journal parameters now. We'll do this both on
3097 * initial mount, once the journal has been initialised but before we've
3098 * done any recovery; and again on any subsequent remount.
3100 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3102 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3104 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3105 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3106 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3108 spin_lock(&journal
->j_state_lock
);
3109 if (test_opt(sb
, BARRIER
))
3110 journal
->j_flags
|= JBD2_BARRIER
;
3112 journal
->j_flags
&= ~JBD2_BARRIER
;
3113 if (test_opt(sb
, DATA_ERR_ABORT
))
3114 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3116 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3117 spin_unlock(&journal
->j_state_lock
);
3120 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3121 unsigned int journal_inum
)
3123 struct inode
*journal_inode
;
3126 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3128 /* First, test for the existence of a valid inode on disk. Bad
3129 * things happen if we iget() an unused inode, as the subsequent
3130 * iput() will try to delete it. */
3132 journal_inode
= ext4_iget(sb
, journal_inum
);
3133 if (IS_ERR(journal_inode
)) {
3134 ext4_msg(sb
, KERN_ERR
, "no journal found");
3137 if (!journal_inode
->i_nlink
) {
3138 make_bad_inode(journal_inode
);
3139 iput(journal_inode
);
3140 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3144 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3145 journal_inode
, journal_inode
->i_size
);
3146 if (!S_ISREG(journal_inode
->i_mode
)) {
3147 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3148 iput(journal_inode
);
3152 journal
= jbd2_journal_init_inode(journal_inode
);
3154 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3155 iput(journal_inode
);
3158 journal
->j_private
= sb
;
3159 ext4_init_journal_params(sb
, journal
);
3163 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3166 struct buffer_head
*bh
;
3170 int hblock
, blocksize
;
3171 ext4_fsblk_t sb_block
;
3172 unsigned long offset
;
3173 struct ext4_super_block
*es
;
3174 struct block_device
*bdev
;
3176 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3178 bdev
= ext4_blkdev_get(j_dev
, sb
);
3182 if (bd_claim(bdev
, sb
)) {
3183 ext4_msg(sb
, KERN_ERR
,
3184 "failed to claim external journal device");
3185 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
3189 blocksize
= sb
->s_blocksize
;
3190 hblock
= bdev_logical_block_size(bdev
);
3191 if (blocksize
< hblock
) {
3192 ext4_msg(sb
, KERN_ERR
,
3193 "blocksize too small for journal device");
3197 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3198 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3199 set_blocksize(bdev
, blocksize
);
3200 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3201 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3202 "external journal");
3206 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3207 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3208 !(le32_to_cpu(es
->s_feature_incompat
) &
3209 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3210 ext4_msg(sb
, KERN_ERR
, "external journal has "
3216 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3217 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3222 len
= ext4_blocks_count(es
);
3223 start
= sb_block
+ 1;
3224 brelse(bh
); /* we're done with the superblock */
3226 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3227 start
, len
, blocksize
);
3229 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3232 journal
->j_private
= sb
;
3233 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3234 wait_on_buffer(journal
->j_sb_buffer
);
3235 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3236 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3239 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3240 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3241 "user (unsupported) - %d",
3242 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3245 EXT4_SB(sb
)->journal_bdev
= bdev
;
3246 ext4_init_journal_params(sb
, journal
);
3250 jbd2_journal_destroy(journal
);
3252 ext4_blkdev_put(bdev
);
3256 static int ext4_load_journal(struct super_block
*sb
,
3257 struct ext4_super_block
*es
,
3258 unsigned long journal_devnum
)
3261 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3264 int really_read_only
;
3266 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3268 if (journal_devnum
&&
3269 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3270 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3271 "numbers have changed");
3272 journal_dev
= new_decode_dev(journal_devnum
);
3274 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3276 really_read_only
= bdev_read_only(sb
->s_bdev
);
3279 * Are we loading a blank journal or performing recovery after a
3280 * crash? For recovery, we need to check in advance whether we
3281 * can get read-write access to the device.
3283 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3284 if (sb
->s_flags
& MS_RDONLY
) {
3285 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3286 "required on readonly filesystem");
3287 if (really_read_only
) {
3288 ext4_msg(sb
, KERN_ERR
, "write access "
3289 "unavailable, cannot proceed");
3292 ext4_msg(sb
, KERN_INFO
, "write access will "
3293 "be enabled during recovery");
3297 if (journal_inum
&& journal_dev
) {
3298 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3299 "and inode journals!");
3304 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3307 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3311 if (!(journal
->j_flags
& JBD2_BARRIER
))
3312 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3314 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
3315 err
= jbd2_journal_update_format(journal
);
3317 ext4_msg(sb
, KERN_ERR
, "error updating journal");
3318 jbd2_journal_destroy(journal
);
3323 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
3324 err
= jbd2_journal_wipe(journal
, !really_read_only
);
3326 err
= jbd2_journal_load(journal
);
3329 ext4_msg(sb
, KERN_ERR
, "error loading journal");
3330 jbd2_journal_destroy(journal
);
3334 EXT4_SB(sb
)->s_journal
= journal
;
3335 ext4_clear_journal_err(sb
, es
);
3337 if (journal_devnum
&&
3338 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3339 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
3341 /* Make sure we flush the recovery flag to disk. */
3342 ext4_commit_super(sb
, 1);
3348 static int ext4_commit_super(struct super_block
*sb
, int sync
)
3350 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
3351 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
3356 if (buffer_write_io_error(sbh
)) {
3358 * Oh, dear. A previous attempt to write the
3359 * superblock failed. This could happen because the
3360 * USB device was yanked out. Or it could happen to
3361 * be a transient write error and maybe the block will
3362 * be remapped. Nothing we can do but to retry the
3363 * write and hope for the best.
3365 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
3366 "superblock detected");
3367 clear_buffer_write_io_error(sbh
);
3368 set_buffer_uptodate(sbh
);
3371 * If the file system is mounted read-only, don't update the
3372 * superblock write time. This avoids updating the superblock
3373 * write time when we are mounting the root file system
3374 * read/only but we need to replay the journal; at that point,
3375 * for people who are east of GMT and who make their clock
3376 * tick in localtime for Windows bug-for-bug compatibility,
3377 * the clock is set in the future, and this will cause e2fsck
3378 * to complain and force a full file system check.
3380 if (!(sb
->s_flags
& MS_RDONLY
))
3381 es
->s_wtime
= cpu_to_le32(get_seconds());
3382 es
->s_kbytes_written
=
3383 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
3384 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
3385 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
3386 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
3387 &EXT4_SB(sb
)->s_freeblocks_counter
));
3388 es
->s_free_inodes_count
= cpu_to_le32(percpu_counter_sum_positive(
3389 &EXT4_SB(sb
)->s_freeinodes_counter
));
3391 BUFFER_TRACE(sbh
, "marking dirty");
3392 mark_buffer_dirty(sbh
);
3394 error
= sync_dirty_buffer(sbh
);
3398 error
= buffer_write_io_error(sbh
);
3400 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
3402 clear_buffer_write_io_error(sbh
);
3403 set_buffer_uptodate(sbh
);
3410 * Have we just finished recovery? If so, and if we are mounting (or
3411 * remounting) the filesystem readonly, then we will end up with a
3412 * consistent fs on disk. Record that fact.
3414 static void ext4_mark_recovery_complete(struct super_block
*sb
,
3415 struct ext4_super_block
*es
)
3417 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
3419 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3420 BUG_ON(journal
!= NULL
);
3423 jbd2_journal_lock_updates(journal
);
3424 if (jbd2_journal_flush(journal
) < 0)
3427 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
3428 sb
->s_flags
& MS_RDONLY
) {
3429 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3430 ext4_commit_super(sb
, 1);
3434 jbd2_journal_unlock_updates(journal
);
3438 * If we are mounting (or read-write remounting) a filesystem whose journal
3439 * has recorded an error from a previous lifetime, move that error to the
3440 * main filesystem now.
3442 static void ext4_clear_journal_err(struct super_block
*sb
,
3443 struct ext4_super_block
*es
)
3449 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3451 journal
= EXT4_SB(sb
)->s_journal
;
3454 * Now check for any error status which may have been recorded in the
3455 * journal by a prior ext4_error() or ext4_abort()
3458 j_errno
= jbd2_journal_errno(journal
);
3462 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
3463 ext4_warning(sb
, "Filesystem error recorded "
3464 "from previous mount: %s", errstr
);
3465 ext4_warning(sb
, "Marking fs in need of filesystem check.");
3467 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
3468 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
3469 ext4_commit_super(sb
, 1);
3471 jbd2_journal_clear_err(journal
);
3476 * Force the running and committing transactions to commit,
3477 * and wait on the commit.
3479 int ext4_force_commit(struct super_block
*sb
)
3484 if (sb
->s_flags
& MS_RDONLY
)
3487 journal
= EXT4_SB(sb
)->s_journal
;
3489 vfs_check_frozen(sb
, SB_FREEZE_WRITE
);
3490 ret
= ext4_journal_force_commit(journal
);
3496 static void ext4_write_super(struct super_block
*sb
)
3499 ext4_commit_super(sb
, 1);
3503 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
3507 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3509 trace_ext4_sync_fs(sb
, wait
);
3510 flush_workqueue(sbi
->dio_unwritten_wq
);
3511 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
3513 jbd2_log_wait_commit(sbi
->s_journal
, target
);
3519 * LVM calls this function before a (read-only) snapshot is created. This
3520 * gives us a chance to flush the journal completely and mark the fs clean.
3522 static int ext4_freeze(struct super_block
*sb
)
3527 if (sb
->s_flags
& MS_RDONLY
)
3530 journal
= EXT4_SB(sb
)->s_journal
;
3532 /* Now we set up the journal barrier. */
3533 jbd2_journal_lock_updates(journal
);
3536 * Don't clear the needs_recovery flag if we failed to flush
3539 error
= jbd2_journal_flush(journal
);
3543 /* Journal blocked and flushed, clear needs_recovery flag. */
3544 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3545 error
= ext4_commit_super(sb
, 1);
3547 /* we rely on s_frozen to stop further updates */
3548 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3553 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3554 * flag here, even though the filesystem is not technically dirty yet.
3556 static int ext4_unfreeze(struct super_block
*sb
)
3558 if (sb
->s_flags
& MS_RDONLY
)
3562 /* Reset the needs_recovery flag before the fs is unlocked. */
3563 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3564 ext4_commit_super(sb
, 1);
3569 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3571 struct ext4_super_block
*es
;
3572 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3573 ext4_fsblk_t n_blocks_count
= 0;
3574 unsigned long old_sb_flags
;
3575 struct ext4_mount_options old_opts
;
3576 int enable_quota
= 0;
3578 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3583 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3587 /* Store the original options */
3589 old_sb_flags
= sb
->s_flags
;
3590 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3591 old_opts
.s_resuid
= sbi
->s_resuid
;
3592 old_opts
.s_resgid
= sbi
->s_resgid
;
3593 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3594 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
3595 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
3597 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3598 for (i
= 0; i
< MAXQUOTAS
; i
++)
3599 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3601 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
3602 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
3605 * Allow the "check" option to be passed as a remount option.
3607 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
3608 &n_blocks_count
, 1)) {
3613 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
3614 ext4_abort(sb
, "Abort forced by user");
3616 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3617 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3621 if (sbi
->s_journal
) {
3622 ext4_init_journal_params(sb
, sbi
->s_journal
);
3623 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3626 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3627 n_blocks_count
> ext4_blocks_count(es
)) {
3628 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
3633 if (*flags
& MS_RDONLY
) {
3634 err
= dquot_suspend(sb
, -1);
3639 * First of all, the unconditional stuff we have to do
3640 * to disable replay of the journal when we next remount
3642 sb
->s_flags
|= MS_RDONLY
;
3645 * OK, test if we are remounting a valid rw partition
3646 * readonly, and if so set the rdonly flag and then
3647 * mark the partition as valid again.
3649 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3650 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3651 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3654 ext4_mark_recovery_complete(sb
, es
);
3656 /* Make sure we can mount this feature set readwrite */
3657 if (!ext4_feature_set_ok(sb
, 0)) {
3662 * Make sure the group descriptor checksums
3663 * are sane. If they aren't, refuse to remount r/w.
3665 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3666 struct ext4_group_desc
*gdp
=
3667 ext4_get_group_desc(sb
, g
, NULL
);
3669 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3670 ext4_msg(sb
, KERN_ERR
,
3671 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3672 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3673 le16_to_cpu(gdp
->bg_checksum
));
3680 * If we have an unprocessed orphan list hanging
3681 * around from a previously readonly bdev mount,
3682 * require a full umount/remount for now.
3684 if (es
->s_last_orphan
) {
3685 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
3686 "remount RDWR because of unprocessed "
3687 "orphan inode list. Please "
3688 "umount/remount instead");
3694 * Mounting a RDONLY partition read-write, so reread
3695 * and store the current valid flag. (It may have
3696 * been changed by e2fsck since we originally mounted
3700 ext4_clear_journal_err(sb
, es
);
3701 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3702 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3704 if (!ext4_setup_super(sb
, es
, 0))
3705 sb
->s_flags
&= ~MS_RDONLY
;
3709 ext4_setup_system_zone(sb
);
3710 if (sbi
->s_journal
== NULL
)
3711 ext4_commit_super(sb
, 1);
3714 /* Release old quota file names */
3715 for (i
= 0; i
< MAXQUOTAS
; i
++)
3716 if (old_opts
.s_qf_names
[i
] &&
3717 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3718 kfree(old_opts
.s_qf_names
[i
]);
3723 dquot_resume(sb
, -1);
3725 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
3730 sb
->s_flags
= old_sb_flags
;
3731 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3732 sbi
->s_resuid
= old_opts
.s_resuid
;
3733 sbi
->s_resgid
= old_opts
.s_resgid
;
3734 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3735 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
3736 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
3738 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3739 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3740 if (sbi
->s_qf_names
[i
] &&
3741 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3742 kfree(sbi
->s_qf_names
[i
]);
3743 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3752 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3754 struct super_block
*sb
= dentry
->d_sb
;
3755 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3756 struct ext4_super_block
*es
= sbi
->s_es
;
3759 if (test_opt(sb
, MINIX_DF
)) {
3760 sbi
->s_overhead_last
= 0;
3761 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3762 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3763 ext4_fsblk_t overhead
= 0;
3766 * Compute the overhead (FS structures). This is constant
3767 * for a given filesystem unless the number of block groups
3768 * changes so we cache the previous value until it does.
3772 * All of the blocks before first_data_block are
3775 overhead
= le32_to_cpu(es
->s_first_data_block
);
3778 * Add the overhead attributed to the superblock and
3779 * block group descriptors. If the sparse superblocks
3780 * feature is turned on, then not all groups have this.
3782 for (i
= 0; i
< ngroups
; i
++) {
3783 overhead
+= ext4_bg_has_super(sb
, i
) +
3784 ext4_bg_num_gdb(sb
, i
);
3789 * Every block group has an inode bitmap, a block
3790 * bitmap, and an inode table.
3792 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3793 sbi
->s_overhead_last
= overhead
;
3795 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3798 buf
->f_type
= EXT4_SUPER_MAGIC
;
3799 buf
->f_bsize
= sb
->s_blocksize
;
3800 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3801 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3802 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3803 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3804 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3806 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3807 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3808 buf
->f_namelen
= EXT4_NAME_LEN
;
3809 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3810 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3811 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3812 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3817 /* Helper function for writing quotas on sync - we need to start transaction
3818 * before quota file is locked for write. Otherwise the are possible deadlocks:
3819 * Process 1 Process 2
3820 * ext4_create() quota_sync()
3821 * jbd2_journal_start() write_dquot()
3822 * dquot_initialize() down(dqio_mutex)
3823 * down(dqio_mutex) jbd2_journal_start()
3829 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3831 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3834 static int ext4_write_dquot(struct dquot
*dquot
)
3838 struct inode
*inode
;
3840 inode
= dquot_to_inode(dquot
);
3841 handle
= ext4_journal_start(inode
,
3842 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3844 return PTR_ERR(handle
);
3845 ret
= dquot_commit(dquot
);
3846 err
= ext4_journal_stop(handle
);
3852 static int ext4_acquire_dquot(struct dquot
*dquot
)
3857 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3858 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3860 return PTR_ERR(handle
);
3861 ret
= dquot_acquire(dquot
);
3862 err
= ext4_journal_stop(handle
);
3868 static int ext4_release_dquot(struct dquot
*dquot
)
3873 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3874 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3875 if (IS_ERR(handle
)) {
3876 /* Release dquot anyway to avoid endless cycle in dqput() */
3877 dquot_release(dquot
);
3878 return PTR_ERR(handle
);
3880 ret
= dquot_release(dquot
);
3881 err
= ext4_journal_stop(handle
);
3887 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3889 /* Are we journaling quotas? */
3890 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3891 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3892 dquot_mark_dquot_dirty(dquot
);
3893 return ext4_write_dquot(dquot
);
3895 return dquot_mark_dquot_dirty(dquot
);
3899 static int ext4_write_info(struct super_block
*sb
, int type
)
3904 /* Data block + inode block */
3905 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3907 return PTR_ERR(handle
);
3908 ret
= dquot_commit_info(sb
, type
);
3909 err
= ext4_journal_stop(handle
);
3916 * Turn on quotas during mount time - we need to find
3917 * the quota file and such...
3919 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3921 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3922 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3926 * Standard function to be called on quota_on
3928 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3934 if (!test_opt(sb
, QUOTA
))
3937 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
3941 /* Quotafile not on the same filesystem? */
3942 if (path
.mnt
->mnt_sb
!= sb
) {
3946 /* Journaling quota? */
3947 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3948 /* Quotafile not in fs root? */
3949 if (path
.dentry
->d_parent
!= sb
->s_root
)
3950 ext4_msg(sb
, KERN_WARNING
,
3951 "Quota file not on filesystem root. "
3952 "Journaled quota will not work");
3956 * When we journal data on quota file, we have to flush journal to see
3957 * all updates to the file when we bypass pagecache...
3959 if (EXT4_SB(sb
)->s_journal
&&
3960 ext4_should_journal_data(path
.dentry
->d_inode
)) {
3962 * We don't need to lock updates but journal_flush() could
3963 * otherwise be livelocked...
3965 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3966 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3967 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3974 err
= dquot_quota_on_path(sb
, type
, format_id
, &path
);
3979 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3980 * acquiring the locks... As quota files are never truncated and quota code
3981 * itself serializes the operations (and noone else should touch the files)
3982 * we don't have to be afraid of races */
3983 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3984 size_t len
, loff_t off
)
3986 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3987 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3989 int offset
= off
& (sb
->s_blocksize
- 1);
3992 struct buffer_head
*bh
;
3993 loff_t i_size
= i_size_read(inode
);
3997 if (off
+len
> i_size
)
4000 while (toread
> 0) {
4001 tocopy
= sb
->s_blocksize
- offset
< toread
?
4002 sb
->s_blocksize
- offset
: toread
;
4003 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
4006 if (!bh
) /* A hole? */
4007 memset(data
, 0, tocopy
);
4009 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4019 /* Write to quotafile (we know the transaction is already started and has
4020 * enough credits) */
4021 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4022 const char *data
, size_t len
, loff_t off
)
4024 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4025 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4027 int offset
= off
& (sb
->s_blocksize
- 1);
4028 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
4029 struct buffer_head
*bh
;
4030 handle_t
*handle
= journal_current_handle();
4032 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4033 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4034 " cancelled because transaction is not started",
4035 (unsigned long long)off
, (unsigned long long)len
);
4039 * Since we account only one data block in transaction credits,
4040 * then it is impossible to cross a block boundary.
4042 if (sb
->s_blocksize
- offset
< len
) {
4043 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4044 " cancelled because not block aligned",
4045 (unsigned long long)off
, (unsigned long long)len
);
4049 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4050 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4053 if (journal_quota
) {
4054 err
= ext4_journal_get_write_access(handle
, bh
);
4061 memcpy(bh
->b_data
+offset
, data
, len
);
4062 flush_dcache_page(bh
->b_page
);
4065 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4067 /* Always do at least ordered writes for quotas */
4068 err
= ext4_jbd2_file_inode(handle
, inode
);
4069 mark_buffer_dirty(bh
);
4074 mutex_unlock(&inode
->i_mutex
);
4077 if (inode
->i_size
< off
+ len
) {
4078 i_size_write(inode
, off
+ len
);
4079 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4081 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4082 ext4_mark_inode_dirty(handle
, inode
);
4083 mutex_unlock(&inode
->i_mutex
);
4089 static int ext4_get_sb(struct file_system_type
*fs_type
, int flags
,
4090 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
4092 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
,mnt
);
4095 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4096 static struct file_system_type ext2_fs_type
= {
4097 .owner
= THIS_MODULE
,
4099 .get_sb
= ext4_get_sb
,
4100 .kill_sb
= kill_block_super
,
4101 .fs_flags
= FS_REQUIRES_DEV
,
4104 static inline void register_as_ext2(void)
4106 int err
= register_filesystem(&ext2_fs_type
);
4109 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4112 static inline void unregister_as_ext2(void)
4114 unregister_filesystem(&ext2_fs_type
);
4116 MODULE_ALIAS("ext2");
4118 static inline void register_as_ext2(void) { }
4119 static inline void unregister_as_ext2(void) { }
4122 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4123 static inline void register_as_ext3(void)
4125 int err
= register_filesystem(&ext3_fs_type
);
4128 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4131 static inline void unregister_as_ext3(void)
4133 unregister_filesystem(&ext3_fs_type
);
4135 MODULE_ALIAS("ext3");
4137 static inline void register_as_ext3(void) { }
4138 static inline void unregister_as_ext3(void) { }
4141 static struct file_system_type ext4_fs_type
= {
4142 .owner
= THIS_MODULE
,
4144 .get_sb
= ext4_get_sb
,
4145 .kill_sb
= kill_block_super
,
4146 .fs_flags
= FS_REQUIRES_DEV
,
4149 static int __init
init_ext4_fs(void)
4153 ext4_check_flag_values();
4154 err
= init_ext4_system_zone();
4157 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4160 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4161 err
= init_ext4_mballoc();
4165 err
= init_ext4_xattr();
4168 err
= init_inodecache();
4173 err
= register_filesystem(&ext4_fs_type
);
4178 unregister_as_ext2();
4179 unregister_as_ext3();
4180 destroy_inodecache();
4184 exit_ext4_mballoc();
4186 remove_proc_entry("fs/ext4", NULL
);
4187 kset_unregister(ext4_kset
);
4189 exit_ext4_system_zone();
4193 static void __exit
exit_ext4_fs(void)
4195 unregister_as_ext2();
4196 unregister_as_ext3();
4197 unregister_filesystem(&ext4_fs_type
);
4198 destroy_inodecache();
4200 exit_ext4_mballoc();
4201 remove_proc_entry("fs/ext4", NULL
);
4202 kset_unregister(ext4_kset
);
4203 exit_ext4_system_zone();
4206 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4207 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4208 MODULE_LICENSE("GPL");
4209 module_init(init_ext4_fs
)
4210 module_exit(exit_ext4_fs
)