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_TRANS
);
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
, __func__
, "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(struct super_block
*sb
, ext4_group_t grp
,
518 const char *function
, const char *fmt
, ...)
523 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
526 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
531 if (test_opt(sb
, ERRORS_CONT
)) {
532 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
533 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
534 ext4_commit_super(sb
, 0);
537 ext4_unlock_group(sb
, grp
);
538 ext4_handle_error(sb
);
540 * We only get here in the ERRORS_RO case; relocking the group
541 * may be dangerous, but nothing bad will happen since the
542 * filesystem will have already been marked read/only and the
543 * journal has been aborted. We return 1 as a hint to callers
544 * who might what to use the return value from
545 * ext4_grp_locked_error() to distinguish beween the
546 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
547 * aggressively from the ext4 function in question, with a
548 * more appropriate error code.
550 ext4_lock_group(sb
, grp
);
554 void ext4_update_dynamic_rev(struct super_block
*sb
)
556 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
558 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
562 "updating to rev %d because of new feature flag, "
563 "running e2fsck is recommended",
566 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
567 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
568 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
569 /* leave es->s_feature_*compat flags alone */
570 /* es->s_uuid will be set by e2fsck if empty */
573 * The rest of the superblock fields should be zero, and if not it
574 * means they are likely already in use, so leave them alone. We
575 * can leave it up to e2fsck to clean up any inconsistencies there.
580 * Open the external journal device
582 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
584 struct block_device
*bdev
;
585 char b
[BDEVNAME_SIZE
];
587 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
593 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
594 __bdevname(dev
, b
), PTR_ERR(bdev
));
599 * Release the journal device
601 static int ext4_blkdev_put(struct block_device
*bdev
)
604 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
607 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
609 struct block_device
*bdev
;
612 bdev
= sbi
->journal_bdev
;
614 ret
= ext4_blkdev_put(bdev
);
615 sbi
->journal_bdev
= NULL
;
620 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
622 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
625 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
629 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
630 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
632 printk(KERN_ERR
"sb_info orphan list:\n");
633 list_for_each(l
, &sbi
->s_orphan
) {
634 struct inode
*inode
= orphan_list_entry(l
);
636 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
637 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
638 inode
->i_mode
, inode
->i_nlink
,
643 static void ext4_put_super(struct super_block
*sb
)
645 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
646 struct ext4_super_block
*es
= sbi
->s_es
;
649 flush_workqueue(sbi
->dio_unwritten_wq
);
650 destroy_workqueue(sbi
->dio_unwritten_wq
);
655 ext4_commit_super(sb
, 1);
657 if (sbi
->s_journal
) {
658 err
= jbd2_journal_destroy(sbi
->s_journal
);
659 sbi
->s_journal
= NULL
;
661 ext4_abort(sb
, __func__
,
662 "Couldn't clean up the journal");
665 ext4_release_system_zone(sb
);
667 ext4_ext_release(sb
);
668 ext4_xattr_put_super(sb
);
670 if (!(sb
->s_flags
& MS_RDONLY
)) {
671 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
672 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
673 ext4_commit_super(sb
, 1);
676 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
678 kobject_del(&sbi
->s_kobj
);
680 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
681 brelse(sbi
->s_group_desc
[i
]);
682 kfree(sbi
->s_group_desc
);
683 if (is_vmalloc_addr(sbi
->s_flex_groups
))
684 vfree(sbi
->s_flex_groups
);
686 kfree(sbi
->s_flex_groups
);
687 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
688 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
689 percpu_counter_destroy(&sbi
->s_dirs_counter
);
690 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
693 for (i
= 0; i
< MAXQUOTAS
; i
++)
694 kfree(sbi
->s_qf_names
[i
]);
697 /* Debugging code just in case the in-memory inode orphan list
698 * isn't empty. The on-disk one can be non-empty if we've
699 * detected an error and taken the fs readonly, but the
700 * in-memory list had better be clean by this point. */
701 if (!list_empty(&sbi
->s_orphan
))
702 dump_orphan_list(sb
, sbi
);
703 J_ASSERT(list_empty(&sbi
->s_orphan
));
705 invalidate_bdev(sb
->s_bdev
);
706 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
708 * Invalidate the journal device's buffers. We don't want them
709 * floating about in memory - the physical journal device may
710 * hotswapped, and it breaks the `ro-after' testing code.
712 sync_blockdev(sbi
->journal_bdev
);
713 invalidate_bdev(sbi
->journal_bdev
);
714 ext4_blkdev_remove(sbi
);
716 sb
->s_fs_info
= NULL
;
718 * Now that we are completely done shutting down the
719 * superblock, we need to actually destroy the kobject.
723 kobject_put(&sbi
->s_kobj
);
724 wait_for_completion(&sbi
->s_kobj_unregister
);
725 kfree(sbi
->s_blockgroup_lock
);
729 static struct kmem_cache
*ext4_inode_cachep
;
732 * Called inside transaction, so use GFP_NOFS
734 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
736 struct ext4_inode_info
*ei
;
738 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
742 ei
->vfs_inode
.i_version
= 1;
743 ei
->vfs_inode
.i_data
.writeback_index
= 0;
744 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
745 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
746 spin_lock_init(&ei
->i_prealloc_lock
);
748 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
749 * therefore it can be null here. Don't check it, just initialize
752 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
753 ei
->i_reserved_data_blocks
= 0;
754 ei
->i_reserved_meta_blocks
= 0;
755 ei
->i_allocated_meta_blocks
= 0;
756 ei
->i_da_metadata_calc_len
= 0;
757 ei
->i_delalloc_reserved_flag
= 0;
758 spin_lock_init(&(ei
->i_block_reservation_lock
));
760 ei
->i_reserved_quota
= 0;
762 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
763 spin_lock_init(&ei
->i_completed_io_lock
);
764 ei
->cur_aio_dio
= NULL
;
766 ei
->i_datasync_tid
= 0;
768 return &ei
->vfs_inode
;
771 static void ext4_destroy_inode(struct inode
*inode
)
773 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
774 ext4_msg(inode
->i_sb
, KERN_ERR
,
775 "Inode %lu (%p): orphan list check failed!",
776 inode
->i_ino
, EXT4_I(inode
));
777 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
778 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
782 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
785 static void init_once(void *foo
)
787 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
789 INIT_LIST_HEAD(&ei
->i_orphan
);
790 #ifdef CONFIG_EXT4_FS_XATTR
791 init_rwsem(&ei
->xattr_sem
);
793 init_rwsem(&ei
->i_data_sem
);
794 inode_init_once(&ei
->vfs_inode
);
797 static int init_inodecache(void)
799 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
800 sizeof(struct ext4_inode_info
),
801 0, (SLAB_RECLAIM_ACCOUNT
|
804 if (ext4_inode_cachep
== NULL
)
809 static void destroy_inodecache(void)
811 kmem_cache_destroy(ext4_inode_cachep
);
814 static void ext4_clear_inode(struct inode
*inode
)
817 ext4_discard_preallocations(inode
);
818 if (EXT4_JOURNAL(inode
))
819 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
820 &EXT4_I(inode
)->jinode
);
823 static inline void ext4_show_quota_options(struct seq_file
*seq
,
824 struct super_block
*sb
)
826 #if defined(CONFIG_QUOTA)
827 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
829 if (sbi
->s_jquota_fmt
) {
832 switch (sbi
->s_jquota_fmt
) {
843 seq_printf(seq
, ",jqfmt=%s", fmtname
);
846 if (sbi
->s_qf_names
[USRQUOTA
])
847 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
849 if (sbi
->s_qf_names
[GRPQUOTA
])
850 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
852 if (test_opt(sb
, USRQUOTA
))
853 seq_puts(seq
, ",usrquota");
855 if (test_opt(sb
, GRPQUOTA
))
856 seq_puts(seq
, ",grpquota");
862 * - it's set to a non-default value OR
863 * - if the per-sb default is different from the global default
865 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
868 unsigned long def_mount_opts
;
869 struct super_block
*sb
= vfs
->mnt_sb
;
870 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
871 struct ext4_super_block
*es
= sbi
->s_es
;
873 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
874 def_errors
= le16_to_cpu(es
->s_errors
);
876 if (sbi
->s_sb_block
!= 1)
877 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
878 if (test_opt(sb
, MINIX_DF
))
879 seq_puts(seq
, ",minixdf");
880 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
881 seq_puts(seq
, ",grpid");
882 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
883 seq_puts(seq
, ",nogrpid");
884 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
885 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
886 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
888 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
889 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
890 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
892 if (test_opt(sb
, ERRORS_RO
)) {
893 if (def_errors
== EXT4_ERRORS_PANIC
||
894 def_errors
== EXT4_ERRORS_CONTINUE
) {
895 seq_puts(seq
, ",errors=remount-ro");
898 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
899 seq_puts(seq
, ",errors=continue");
900 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
901 seq_puts(seq
, ",errors=panic");
902 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
903 seq_puts(seq
, ",nouid32");
904 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
905 seq_puts(seq
, ",debug");
906 if (test_opt(sb
, OLDALLOC
))
907 seq_puts(seq
, ",oldalloc");
908 #ifdef CONFIG_EXT4_FS_XATTR
909 if (test_opt(sb
, XATTR_USER
) &&
910 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
911 seq_puts(seq
, ",user_xattr");
912 if (!test_opt(sb
, XATTR_USER
) &&
913 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
914 seq_puts(seq
, ",nouser_xattr");
917 #ifdef CONFIG_EXT4_FS_POSIX_ACL
918 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
919 seq_puts(seq
, ",acl");
920 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
921 seq_puts(seq
, ",noacl");
923 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
924 seq_printf(seq
, ",commit=%u",
925 (unsigned) (sbi
->s_commit_interval
/ HZ
));
927 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
928 seq_printf(seq
, ",min_batch_time=%u",
929 (unsigned) sbi
->s_min_batch_time
);
931 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
932 seq_printf(seq
, ",max_batch_time=%u",
933 (unsigned) sbi
->s_min_batch_time
);
937 * We're changing the default of barrier mount option, so
938 * let's always display its mount state so it's clear what its
941 seq_puts(seq
, ",barrier=");
942 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
943 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
944 seq_puts(seq
, ",journal_async_commit");
945 else if (test_opt(sb
, JOURNAL_CHECKSUM
))
946 seq_puts(seq
, ",journal_checksum");
947 if (test_opt(sb
, NOBH
))
948 seq_puts(seq
, ",nobh");
949 if (test_opt(sb
, I_VERSION
))
950 seq_puts(seq
, ",i_version");
951 if (!test_opt(sb
, DELALLOC
))
952 seq_puts(seq
, ",nodelalloc");
956 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
958 * journal mode get enabled in different ways
959 * So just print the value even if we didn't specify it
961 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
962 seq_puts(seq
, ",data=journal");
963 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
964 seq_puts(seq
, ",data=ordered");
965 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
966 seq_puts(seq
, ",data=writeback");
968 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
969 seq_printf(seq
, ",inode_readahead_blks=%u",
970 sbi
->s_inode_readahead_blks
);
972 if (test_opt(sb
, DATA_ERR_ABORT
))
973 seq_puts(seq
, ",data_err=abort");
975 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
976 seq_puts(seq
, ",noauto_da_alloc");
978 if (test_opt(sb
, DISCARD
))
979 seq_puts(seq
, ",discard");
981 if (test_opt(sb
, NOLOAD
))
982 seq_puts(seq
, ",norecovery");
984 if (test_opt(sb
, DIOREAD_NOLOCK
))
985 seq_puts(seq
, ",dioread_nolock");
987 ext4_show_quota_options(seq
, sb
);
992 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
993 u64 ino
, u32 generation
)
997 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
998 return ERR_PTR(-ESTALE
);
999 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1000 return ERR_PTR(-ESTALE
);
1002 /* iget isn't really right if the inode is currently unallocated!!
1004 * ext4_read_inode will return a bad_inode if the inode had been
1005 * deleted, so we should be safe.
1007 * Currently we don't know the generation for parent directory, so
1008 * a generation of 0 means "accept any"
1010 inode
= ext4_iget(sb
, ino
);
1012 return ERR_CAST(inode
);
1013 if (generation
&& inode
->i_generation
!= generation
) {
1015 return ERR_PTR(-ESTALE
);
1021 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1022 int fh_len
, int fh_type
)
1024 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1025 ext4_nfs_get_inode
);
1028 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1029 int fh_len
, int fh_type
)
1031 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1032 ext4_nfs_get_inode
);
1036 * Try to release metadata pages (indirect blocks, directories) which are
1037 * mapped via the block device. Since these pages could have journal heads
1038 * which would prevent try_to_free_buffers() from freeing them, we must use
1039 * jbd2 layer's try_to_free_buffers() function to release them.
1041 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1044 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1046 WARN_ON(PageChecked(page
));
1047 if (!page_has_buffers(page
))
1050 return jbd2_journal_try_to_free_buffers(journal
, page
,
1051 wait
& ~__GFP_WAIT
);
1052 return try_to_free_buffers(page
);
1056 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1057 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1059 static int ext4_write_dquot(struct dquot
*dquot
);
1060 static int ext4_acquire_dquot(struct dquot
*dquot
);
1061 static int ext4_release_dquot(struct dquot
*dquot
);
1062 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1063 static int ext4_write_info(struct super_block
*sb
, int type
);
1064 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1065 char *path
, int remount
);
1066 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1067 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1068 size_t len
, loff_t off
);
1069 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1070 const char *data
, size_t len
, loff_t off
);
1072 static const struct dquot_operations ext4_quota_operations
= {
1074 .get_reserved_space
= ext4_get_reserved_space
,
1076 .write_dquot
= ext4_write_dquot
,
1077 .acquire_dquot
= ext4_acquire_dquot
,
1078 .release_dquot
= ext4_release_dquot
,
1079 .mark_dirty
= ext4_mark_dquot_dirty
,
1080 .write_info
= ext4_write_info
,
1081 .alloc_dquot
= dquot_alloc
,
1082 .destroy_dquot
= dquot_destroy
,
1085 static const struct quotactl_ops ext4_qctl_operations
= {
1086 .quota_on
= ext4_quota_on
,
1087 .quota_off
= vfs_quota_off
,
1088 .quota_sync
= vfs_quota_sync
,
1089 .get_info
= vfs_get_dqinfo
,
1090 .set_info
= vfs_set_dqinfo
,
1091 .get_dqblk
= vfs_get_dqblk
,
1092 .set_dqblk
= vfs_set_dqblk
1096 static const struct super_operations ext4_sops
= {
1097 .alloc_inode
= ext4_alloc_inode
,
1098 .destroy_inode
= ext4_destroy_inode
,
1099 .write_inode
= ext4_write_inode
,
1100 .dirty_inode
= ext4_dirty_inode
,
1101 .delete_inode
= ext4_delete_inode
,
1102 .put_super
= ext4_put_super
,
1103 .sync_fs
= ext4_sync_fs
,
1104 .freeze_fs
= ext4_freeze
,
1105 .unfreeze_fs
= ext4_unfreeze
,
1106 .statfs
= ext4_statfs
,
1107 .remount_fs
= ext4_remount
,
1108 .clear_inode
= ext4_clear_inode
,
1109 .show_options
= ext4_show_options
,
1111 .quota_read
= ext4_quota_read
,
1112 .quota_write
= ext4_quota_write
,
1114 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1117 static const struct super_operations ext4_nojournal_sops
= {
1118 .alloc_inode
= ext4_alloc_inode
,
1119 .destroy_inode
= ext4_destroy_inode
,
1120 .write_inode
= ext4_write_inode
,
1121 .dirty_inode
= ext4_dirty_inode
,
1122 .delete_inode
= ext4_delete_inode
,
1123 .write_super
= ext4_write_super
,
1124 .put_super
= ext4_put_super
,
1125 .statfs
= ext4_statfs
,
1126 .remount_fs
= ext4_remount
,
1127 .clear_inode
= ext4_clear_inode
,
1128 .show_options
= ext4_show_options
,
1130 .quota_read
= ext4_quota_read
,
1131 .quota_write
= ext4_quota_write
,
1133 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1136 static const struct export_operations ext4_export_ops
= {
1137 .fh_to_dentry
= ext4_fh_to_dentry
,
1138 .fh_to_parent
= ext4_fh_to_parent
,
1139 .get_parent
= ext4_get_parent
,
1143 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1144 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1145 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1146 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1147 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1148 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1149 Opt_journal_update
, Opt_journal_dev
,
1150 Opt_journal_checksum
, Opt_journal_async_commit
,
1151 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1152 Opt_data_err_abort
, Opt_data_err_ignore
,
1153 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1154 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1155 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1156 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1157 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
1158 Opt_block_validity
, Opt_noblock_validity
,
1159 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1160 Opt_dioread_nolock
, Opt_dioread_lock
,
1161 Opt_discard
, Opt_nodiscard
,
1164 static const match_table_t tokens
= {
1165 {Opt_bsd_df
, "bsddf"},
1166 {Opt_minix_df
, "minixdf"},
1167 {Opt_grpid
, "grpid"},
1168 {Opt_grpid
, "bsdgroups"},
1169 {Opt_nogrpid
, "nogrpid"},
1170 {Opt_nogrpid
, "sysvgroups"},
1171 {Opt_resgid
, "resgid=%u"},
1172 {Opt_resuid
, "resuid=%u"},
1174 {Opt_err_cont
, "errors=continue"},
1175 {Opt_err_panic
, "errors=panic"},
1176 {Opt_err_ro
, "errors=remount-ro"},
1177 {Opt_nouid32
, "nouid32"},
1178 {Opt_debug
, "debug"},
1179 {Opt_oldalloc
, "oldalloc"},
1180 {Opt_orlov
, "orlov"},
1181 {Opt_user_xattr
, "user_xattr"},
1182 {Opt_nouser_xattr
, "nouser_xattr"},
1184 {Opt_noacl
, "noacl"},
1185 {Opt_noload
, "noload"},
1186 {Opt_noload
, "norecovery"},
1189 {Opt_commit
, "commit=%u"},
1190 {Opt_min_batch_time
, "min_batch_time=%u"},
1191 {Opt_max_batch_time
, "max_batch_time=%u"},
1192 {Opt_journal_update
, "journal=update"},
1193 {Opt_journal_dev
, "journal_dev=%u"},
1194 {Opt_journal_checksum
, "journal_checksum"},
1195 {Opt_journal_async_commit
, "journal_async_commit"},
1196 {Opt_abort
, "abort"},
1197 {Opt_data_journal
, "data=journal"},
1198 {Opt_data_ordered
, "data=ordered"},
1199 {Opt_data_writeback
, "data=writeback"},
1200 {Opt_data_err_abort
, "data_err=abort"},
1201 {Opt_data_err_ignore
, "data_err=ignore"},
1202 {Opt_offusrjquota
, "usrjquota="},
1203 {Opt_usrjquota
, "usrjquota=%s"},
1204 {Opt_offgrpjquota
, "grpjquota="},
1205 {Opt_grpjquota
, "grpjquota=%s"},
1206 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1207 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1208 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1209 {Opt_grpquota
, "grpquota"},
1210 {Opt_noquota
, "noquota"},
1211 {Opt_quota
, "quota"},
1212 {Opt_usrquota
, "usrquota"},
1213 {Opt_barrier
, "barrier=%u"},
1214 {Opt_barrier
, "barrier"},
1215 {Opt_nobarrier
, "nobarrier"},
1216 {Opt_i_version
, "i_version"},
1217 {Opt_stripe
, "stripe=%u"},
1218 {Opt_resize
, "resize"},
1219 {Opt_delalloc
, "delalloc"},
1220 {Opt_nodelalloc
, "nodelalloc"},
1221 {Opt_block_validity
, "block_validity"},
1222 {Opt_noblock_validity
, "noblock_validity"},
1223 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1224 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1225 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1226 {Opt_auto_da_alloc
, "auto_da_alloc"},
1227 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1228 {Opt_dioread_nolock
, "dioread_nolock"},
1229 {Opt_dioread_lock
, "dioread_lock"},
1230 {Opt_discard
, "discard"},
1231 {Opt_nodiscard
, "nodiscard"},
1235 static ext4_fsblk_t
get_sb_block(void **data
)
1237 ext4_fsblk_t sb_block
;
1238 char *options
= (char *) *data
;
1240 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1241 return 1; /* Default location */
1244 /* TODO: use simple_strtoll with >32bit ext4 */
1245 sb_block
= simple_strtoul(options
, &options
, 0);
1246 if (*options
&& *options
!= ',') {
1247 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1251 if (*options
== ',')
1253 *data
= (void *) options
;
1258 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1259 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1260 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1263 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1265 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1268 if (sb_any_quota_loaded(sb
) &&
1269 !sbi
->s_qf_names
[qtype
]) {
1270 ext4_msg(sb
, KERN_ERR
,
1271 "Cannot change journaled "
1272 "quota options when quota turned on");
1275 qname
= match_strdup(args
);
1277 ext4_msg(sb
, KERN_ERR
,
1278 "Not enough memory for storing quotafile name");
1281 if (sbi
->s_qf_names
[qtype
] &&
1282 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1283 ext4_msg(sb
, KERN_ERR
,
1284 "%s quota file already specified", QTYPE2NAME(qtype
));
1288 sbi
->s_qf_names
[qtype
] = qname
;
1289 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1290 ext4_msg(sb
, KERN_ERR
,
1291 "quotafile must be on filesystem root");
1292 kfree(sbi
->s_qf_names
[qtype
]);
1293 sbi
->s_qf_names
[qtype
] = NULL
;
1296 set_opt(sbi
->s_mount_opt
, QUOTA
);
1300 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1303 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1305 if (sb_any_quota_loaded(sb
) &&
1306 sbi
->s_qf_names
[qtype
]) {
1307 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1308 " when quota turned on");
1312 * The space will be released later when all options are confirmed
1315 sbi
->s_qf_names
[qtype
] = NULL
;
1320 static int parse_options(char *options
, struct super_block
*sb
,
1321 unsigned long *journal_devnum
,
1322 unsigned int *journal_ioprio
,
1323 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1325 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1327 substring_t args
[MAX_OPT_ARGS
];
1337 while ((p
= strsep(&options
, ",")) != NULL
) {
1343 * Initialize args struct so we know whether arg was
1344 * found; some options take optional arguments.
1346 args
[0].to
= args
[0].from
= 0;
1347 token
= match_token(p
, tokens
, args
);
1350 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1351 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1354 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1355 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1359 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1360 set_opt(sbi
->s_mount_opt
, GRPID
);
1364 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1365 clear_opt(sbi
->s_mount_opt
, GRPID
);
1369 if (match_int(&args
[0], &option
))
1371 sbi
->s_resuid
= option
;
1374 if (match_int(&args
[0], &option
))
1376 sbi
->s_resgid
= option
;
1379 /* handled by get_sb_block() instead of here */
1380 /* *sb_block = match_int(&args[0]); */
1383 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1384 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1385 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1388 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1389 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1390 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1393 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1394 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1395 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1398 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1401 set_opt(sbi
->s_mount_opt
, DEBUG
);
1404 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1407 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1409 #ifdef CONFIG_EXT4_FS_XATTR
1410 case Opt_user_xattr
:
1411 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1413 case Opt_nouser_xattr
:
1414 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1417 case Opt_user_xattr
:
1418 case Opt_nouser_xattr
:
1419 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1422 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1424 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1427 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1432 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1435 case Opt_journal_update
:
1437 /* Eventually we will want to be able to create
1438 a journal file here. For now, only allow the
1439 user to specify an existing inode to be the
1442 ext4_msg(sb
, KERN_ERR
,
1443 "Cannot specify journal on remount");
1446 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1448 case Opt_journal_dev
:
1450 ext4_msg(sb
, KERN_ERR
,
1451 "Cannot specify journal on remount");
1454 if (match_int(&args
[0], &option
))
1456 *journal_devnum
= option
;
1458 case Opt_journal_checksum
:
1459 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1461 case Opt_journal_async_commit
:
1462 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1463 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1466 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1469 if (match_int(&args
[0], &option
))
1474 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1475 sbi
->s_commit_interval
= HZ
* option
;
1477 case Opt_max_batch_time
:
1478 if (match_int(&args
[0], &option
))
1483 option
= EXT4_DEF_MAX_BATCH_TIME
;
1484 sbi
->s_max_batch_time
= option
;
1486 case Opt_min_batch_time
:
1487 if (match_int(&args
[0], &option
))
1491 sbi
->s_min_batch_time
= option
;
1493 case Opt_data_journal
:
1494 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1496 case Opt_data_ordered
:
1497 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1499 case Opt_data_writeback
:
1500 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1503 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1504 ext4_msg(sb
, KERN_ERR
,
1505 "Cannot change data mode on remount");
1509 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
1510 sbi
->s_mount_opt
|= data_opt
;
1513 case Opt_data_err_abort
:
1514 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1516 case Opt_data_err_ignore
:
1517 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1521 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1525 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1528 case Opt_offusrjquota
:
1529 if (!clear_qf_name(sb
, USRQUOTA
))
1532 case Opt_offgrpjquota
:
1533 if (!clear_qf_name(sb
, GRPQUOTA
))
1537 case Opt_jqfmt_vfsold
:
1538 qfmt
= QFMT_VFS_OLD
;
1540 case Opt_jqfmt_vfsv0
:
1543 case Opt_jqfmt_vfsv1
:
1546 if (sb_any_quota_loaded(sb
) &&
1547 sbi
->s_jquota_fmt
!= qfmt
) {
1548 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1549 "journaled quota options when "
1553 sbi
->s_jquota_fmt
= qfmt
;
1557 set_opt(sbi
->s_mount_opt
, QUOTA
);
1558 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1561 set_opt(sbi
->s_mount_opt
, QUOTA
);
1562 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1565 if (sb_any_quota_loaded(sb
)) {
1566 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1567 "options when quota turned on");
1570 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1571 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1572 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1578 ext4_msg(sb
, KERN_ERR
,
1579 "quota options not supported");
1583 case Opt_offusrjquota
:
1584 case Opt_offgrpjquota
:
1585 case Opt_jqfmt_vfsold
:
1586 case Opt_jqfmt_vfsv0
:
1587 case Opt_jqfmt_vfsv1
:
1588 ext4_msg(sb
, KERN_ERR
,
1589 "journaled quota options not supported");
1595 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1598 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1602 if (match_int(&args
[0], &option
))
1605 option
= 1; /* No argument, default to 1 */
1607 set_opt(sbi
->s_mount_opt
, BARRIER
);
1609 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1615 ext4_msg(sb
, KERN_ERR
,
1616 "resize option only available "
1620 if (match_int(&args
[0], &option
) != 0)
1622 *n_blocks_count
= option
;
1625 set_opt(sbi
->s_mount_opt
, NOBH
);
1628 clear_opt(sbi
->s_mount_opt
, NOBH
);
1631 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1632 sb
->s_flags
|= MS_I_VERSION
;
1634 case Opt_nodelalloc
:
1635 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1638 if (match_int(&args
[0], &option
))
1642 sbi
->s_stripe
= option
;
1645 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1647 case Opt_block_validity
:
1648 set_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1650 case Opt_noblock_validity
:
1651 clear_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1653 case Opt_inode_readahead_blks
:
1654 if (match_int(&args
[0], &option
))
1656 if (option
< 0 || option
> (1 << 30))
1658 if (!is_power_of_2(option
)) {
1659 ext4_msg(sb
, KERN_ERR
,
1660 "EXT4-fs: inode_readahead_blks"
1661 " must be a power of 2");
1664 sbi
->s_inode_readahead_blks
= option
;
1666 case Opt_journal_ioprio
:
1667 if (match_int(&args
[0], &option
))
1669 if (option
< 0 || option
> 7)
1671 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1674 case Opt_noauto_da_alloc
:
1675 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1677 case Opt_auto_da_alloc
:
1679 if (match_int(&args
[0], &option
))
1682 option
= 1; /* No argument, default to 1 */
1684 clear_opt(sbi
->s_mount_opt
, NO_AUTO_DA_ALLOC
);
1686 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1689 set_opt(sbi
->s_mount_opt
, DISCARD
);
1692 clear_opt(sbi
->s_mount_opt
, DISCARD
);
1694 case Opt_dioread_nolock
:
1695 set_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
1697 case Opt_dioread_lock
:
1698 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
1701 ext4_msg(sb
, KERN_ERR
,
1702 "Unrecognized mount option \"%s\" "
1703 "or missing value", p
);
1708 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1709 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1710 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1712 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1713 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1715 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1716 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1721 if (!sbi
->s_jquota_fmt
) {
1722 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1727 if (sbi
->s_jquota_fmt
) {
1728 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1729 "specified with no journaling "
1738 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1741 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1744 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1745 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1746 "forcing read-only mode");
1751 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1752 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1753 "running e2fsck is recommended");
1754 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1755 ext4_msg(sb
, KERN_WARNING
,
1756 "warning: mounting fs with errors, "
1757 "running e2fsck is recommended");
1758 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1759 le16_to_cpu(es
->s_mnt_count
) >=
1760 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1761 ext4_msg(sb
, KERN_WARNING
,
1762 "warning: maximal mount count reached, "
1763 "running e2fsck is recommended");
1764 else if (le32_to_cpu(es
->s_checkinterval
) &&
1765 (le32_to_cpu(es
->s_lastcheck
) +
1766 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1767 ext4_msg(sb
, KERN_WARNING
,
1768 "warning: checktime reached, "
1769 "running e2fsck is recommended");
1770 if (!sbi
->s_journal
)
1771 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1772 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1773 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1774 le16_add_cpu(&es
->s_mnt_count
, 1);
1775 es
->s_mtime
= cpu_to_le32(get_seconds());
1776 ext4_update_dynamic_rev(sb
);
1778 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1780 ext4_commit_super(sb
, 1);
1781 if (test_opt(sb
, DEBUG
))
1782 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1783 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1785 sbi
->s_groups_count
,
1786 EXT4_BLOCKS_PER_GROUP(sb
),
1787 EXT4_INODES_PER_GROUP(sb
),
1793 static int ext4_fill_flex_info(struct super_block
*sb
)
1795 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1796 struct ext4_group_desc
*gdp
= NULL
;
1797 ext4_group_t flex_group_count
;
1798 ext4_group_t flex_group
;
1799 unsigned int groups_per_flex
= 0;
1803 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1804 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
1805 sbi
->s_log_groups_per_flex
= 0;
1808 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1810 /* We allocate both existing and potentially added groups */
1811 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1812 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1813 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1814 size
= flex_group_count
* sizeof(struct flex_groups
);
1815 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1816 if (sbi
->s_flex_groups
== NULL
) {
1817 sbi
->s_flex_groups
= vmalloc(size
);
1818 if (sbi
->s_flex_groups
)
1819 memset(sbi
->s_flex_groups
, 0, size
);
1821 if (sbi
->s_flex_groups
== NULL
) {
1822 ext4_msg(sb
, KERN_ERR
, "not enough memory for "
1823 "%u flex groups", flex_group_count
);
1827 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1828 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1830 flex_group
= ext4_flex_group(sbi
, i
);
1831 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1832 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1833 atomic_add(ext4_free_blks_count(sb
, gdp
),
1834 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1835 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1836 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1844 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1845 struct ext4_group_desc
*gdp
)
1849 if (sbi
->s_es
->s_feature_ro_compat
&
1850 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1851 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1852 __le32 le_group
= cpu_to_le32(block_group
);
1854 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1855 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1856 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1857 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1858 /* for checksum of struct ext4_group_desc do the rest...*/
1859 if ((sbi
->s_es
->s_feature_incompat
&
1860 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1861 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1862 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1863 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1867 return cpu_to_le16(crc
);
1870 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1871 struct ext4_group_desc
*gdp
)
1873 if ((sbi
->s_es
->s_feature_ro_compat
&
1874 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1875 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1881 /* Called at mount-time, super-block is locked */
1882 static int ext4_check_descriptors(struct super_block
*sb
)
1884 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1885 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1886 ext4_fsblk_t last_block
;
1887 ext4_fsblk_t block_bitmap
;
1888 ext4_fsblk_t inode_bitmap
;
1889 ext4_fsblk_t inode_table
;
1890 int flexbg_flag
= 0;
1893 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1896 ext4_debug("Checking group descriptors");
1898 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1899 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1901 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1902 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1904 last_block
= first_block
+
1905 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1907 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1908 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1909 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1910 "Block bitmap for group %u not in group "
1911 "(block %llu)!", i
, block_bitmap
);
1914 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1915 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1916 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1917 "Inode bitmap for group %u not in group "
1918 "(block %llu)!", i
, inode_bitmap
);
1921 inode_table
= ext4_inode_table(sb
, gdp
);
1922 if (inode_table
< first_block
||
1923 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1924 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1925 "Inode table for group %u not in group "
1926 "(block %llu)!", i
, inode_table
);
1929 ext4_lock_group(sb
, i
);
1930 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1931 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1932 "Checksum for group %u failed (%u!=%u)",
1933 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1934 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1935 if (!(sb
->s_flags
& MS_RDONLY
)) {
1936 ext4_unlock_group(sb
, i
);
1940 ext4_unlock_group(sb
, i
);
1942 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1945 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1946 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
1950 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1951 * the superblock) which were deleted from all directories, but held open by
1952 * a process at the time of a crash. We walk the list and try to delete these
1953 * inodes at recovery time (only with a read-write filesystem).
1955 * In order to keep the orphan inode chain consistent during traversal (in
1956 * case of crash during recovery), we link each inode into the superblock
1957 * orphan list_head and handle it the same way as an inode deletion during
1958 * normal operation (which journals the operations for us).
1960 * We only do an iget() and an iput() on each inode, which is very safe if we
1961 * accidentally point at an in-use or already deleted inode. The worst that
1962 * can happen in this case is that we get a "bit already cleared" message from
1963 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1964 * e2fsck was run on this filesystem, and it must have already done the orphan
1965 * inode cleanup for us, so we can safely abort without any further action.
1967 static void ext4_orphan_cleanup(struct super_block
*sb
,
1968 struct ext4_super_block
*es
)
1970 unsigned int s_flags
= sb
->s_flags
;
1971 int nr_orphans
= 0, nr_truncates
= 0;
1975 if (!es
->s_last_orphan
) {
1976 jbd_debug(4, "no orphan inodes to clean up\n");
1980 if (bdev_read_only(sb
->s_bdev
)) {
1981 ext4_msg(sb
, KERN_ERR
, "write access "
1982 "unavailable, skipping orphan cleanup");
1986 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1987 if (es
->s_last_orphan
)
1988 jbd_debug(1, "Errors on filesystem, "
1989 "clearing orphan list.\n");
1990 es
->s_last_orphan
= 0;
1991 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1995 if (s_flags
& MS_RDONLY
) {
1996 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
1997 sb
->s_flags
&= ~MS_RDONLY
;
2000 /* Needed for iput() to work correctly and not trash data */
2001 sb
->s_flags
|= MS_ACTIVE
;
2002 /* Turn on quotas so that they are updated correctly */
2003 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2004 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2005 int ret
= ext4_quota_on_mount(sb
, i
);
2007 ext4_msg(sb
, KERN_ERR
,
2008 "Cannot turn on journaled "
2009 "quota: error %d", ret
);
2014 while (es
->s_last_orphan
) {
2015 struct inode
*inode
;
2017 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2018 if (IS_ERR(inode
)) {
2019 es
->s_last_orphan
= 0;
2023 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2024 dquot_initialize(inode
);
2025 if (inode
->i_nlink
) {
2026 ext4_msg(sb
, KERN_DEBUG
,
2027 "%s: truncating inode %lu to %lld bytes",
2028 __func__
, inode
->i_ino
, inode
->i_size
);
2029 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2030 inode
->i_ino
, inode
->i_size
);
2031 ext4_truncate(inode
);
2034 ext4_msg(sb
, KERN_DEBUG
,
2035 "%s: deleting unreferenced inode %lu",
2036 __func__
, inode
->i_ino
);
2037 jbd_debug(2, "deleting unreferenced inode %lu\n",
2041 iput(inode
); /* The delete magic happens here! */
2044 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2047 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2048 PLURAL(nr_orphans
));
2050 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2051 PLURAL(nr_truncates
));
2053 /* Turn quotas off */
2054 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2055 if (sb_dqopt(sb
)->files
[i
])
2056 vfs_quota_off(sb
, i
, 0);
2059 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2063 * Maximal extent format file size.
2064 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2065 * extent format containers, within a sector_t, and within i_blocks
2066 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2067 * so that won't be a limiting factor.
2069 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2071 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2074 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2076 /* small i_blocks in vfs inode? */
2077 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2079 * CONFIG_LBDAF is not enabled implies the inode
2080 * i_block represent total blocks in 512 bytes
2081 * 32 == size of vfs inode i_blocks * 8
2083 upper_limit
= (1LL << 32) - 1;
2085 /* total blocks in file system block size */
2086 upper_limit
>>= (blkbits
- 9);
2087 upper_limit
<<= blkbits
;
2090 /* 32-bit extent-start container, ee_block */
2095 /* Sanity check against vm- & vfs- imposed limits */
2096 if (res
> upper_limit
)
2103 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2104 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2105 * We need to be 1 filesystem block less than the 2^48 sector limit.
2107 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2109 loff_t res
= EXT4_NDIR_BLOCKS
;
2112 /* This is calculated to be the largest file size for a dense, block
2113 * mapped file such that the file's total number of 512-byte sectors,
2114 * including data and all indirect blocks, does not exceed (2^48 - 1).
2116 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2117 * number of 512-byte sectors of the file.
2120 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2122 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2123 * the inode i_block field represents total file blocks in
2124 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2126 upper_limit
= (1LL << 32) - 1;
2128 /* total blocks in file system block size */
2129 upper_limit
>>= (bits
- 9);
2133 * We use 48 bit ext4_inode i_blocks
2134 * With EXT4_HUGE_FILE_FL set the i_blocks
2135 * represent total number of blocks in
2136 * file system block size
2138 upper_limit
= (1LL << 48) - 1;
2142 /* indirect blocks */
2144 /* double indirect blocks */
2145 meta_blocks
+= 1 + (1LL << (bits
-2));
2146 /* tripple indirect blocks */
2147 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2149 upper_limit
-= meta_blocks
;
2150 upper_limit
<<= bits
;
2152 res
+= 1LL << (bits
-2);
2153 res
+= 1LL << (2*(bits
-2));
2154 res
+= 1LL << (3*(bits
-2));
2156 if (res
> upper_limit
)
2159 if (res
> MAX_LFS_FILESIZE
)
2160 res
= MAX_LFS_FILESIZE
;
2165 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2166 ext4_fsblk_t logical_sb_block
, int nr
)
2168 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2169 ext4_group_t bg
, first_meta_bg
;
2172 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2174 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2176 return logical_sb_block
+ nr
+ 1;
2177 bg
= sbi
->s_desc_per_block
* nr
;
2178 if (ext4_bg_has_super(sb
, bg
))
2181 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2185 * ext4_get_stripe_size: Get the stripe size.
2186 * @sbi: In memory super block info
2188 * If we have specified it via mount option, then
2189 * use the mount option value. If the value specified at mount time is
2190 * greater than the blocks per group use the super block value.
2191 * If the super block value is greater than blocks per group return 0.
2192 * Allocator needs it be less than blocks per group.
2195 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2197 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2198 unsigned long stripe_width
=
2199 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2201 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2202 return sbi
->s_stripe
;
2204 if (stripe_width
<= sbi
->s_blocks_per_group
)
2205 return stripe_width
;
2207 if (stride
<= sbi
->s_blocks_per_group
)
2216 struct attribute attr
;
2217 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2218 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2219 const char *, size_t);
2223 static int parse_strtoul(const char *buf
,
2224 unsigned long max
, unsigned long *value
)
2228 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2229 endp
= skip_spaces(endp
);
2230 if (*endp
|| *value
> max
)
2236 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2237 struct ext4_sb_info
*sbi
,
2240 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2241 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2244 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2245 struct ext4_sb_info
*sbi
, char *buf
)
2247 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2249 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2250 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2251 sbi
->s_sectors_written_start
) >> 1);
2254 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2255 struct ext4_sb_info
*sbi
, char *buf
)
2257 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2259 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2260 (unsigned long long)(sbi
->s_kbytes_written
+
2261 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2262 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2265 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2266 struct ext4_sb_info
*sbi
,
2267 const char *buf
, size_t count
)
2271 if (parse_strtoul(buf
, 0x40000000, &t
))
2274 if (!is_power_of_2(t
))
2277 sbi
->s_inode_readahead_blks
= t
;
2281 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2282 struct ext4_sb_info
*sbi
, char *buf
)
2284 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2286 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2289 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2290 struct ext4_sb_info
*sbi
,
2291 const char *buf
, size_t count
)
2293 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2296 if (parse_strtoul(buf
, 0xffffffff, &t
))
2302 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2303 static struct ext4_attr ext4_attr_##_name = { \
2304 .attr = {.name = __stringify(_name), .mode = _mode }, \
2307 .offset = offsetof(struct ext4_sb_info, _elname), \
2309 #define EXT4_ATTR(name, mode, show, store) \
2310 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2312 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2313 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2314 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2315 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2316 #define ATTR_LIST(name) &ext4_attr_##name.attr
2318 EXT4_RO_ATTR(delayed_allocation_blocks
);
2319 EXT4_RO_ATTR(session_write_kbytes
);
2320 EXT4_RO_ATTR(lifetime_write_kbytes
);
2321 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2322 inode_readahead_blks_store
, s_inode_readahead_blks
);
2323 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2324 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2325 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2326 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2327 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2328 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2329 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2330 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2332 static struct attribute
*ext4_attrs
[] = {
2333 ATTR_LIST(delayed_allocation_blocks
),
2334 ATTR_LIST(session_write_kbytes
),
2335 ATTR_LIST(lifetime_write_kbytes
),
2336 ATTR_LIST(inode_readahead_blks
),
2337 ATTR_LIST(inode_goal
),
2338 ATTR_LIST(mb_stats
),
2339 ATTR_LIST(mb_max_to_scan
),
2340 ATTR_LIST(mb_min_to_scan
),
2341 ATTR_LIST(mb_order2_req
),
2342 ATTR_LIST(mb_stream_req
),
2343 ATTR_LIST(mb_group_prealloc
),
2344 ATTR_LIST(max_writeback_mb_bump
),
2348 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2349 struct attribute
*attr
, char *buf
)
2351 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2353 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2355 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2358 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2359 struct attribute
*attr
,
2360 const char *buf
, size_t len
)
2362 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2364 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2366 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2369 static void ext4_sb_release(struct kobject
*kobj
)
2371 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2373 complete(&sbi
->s_kobj_unregister
);
2377 static const struct sysfs_ops ext4_attr_ops
= {
2378 .show
= ext4_attr_show
,
2379 .store
= ext4_attr_store
,
2382 static struct kobj_type ext4_ktype
= {
2383 .default_attrs
= ext4_attrs
,
2384 .sysfs_ops
= &ext4_attr_ops
,
2385 .release
= ext4_sb_release
,
2389 * Check whether this filesystem can be mounted based on
2390 * the features present and the RDONLY/RDWR mount requested.
2391 * Returns 1 if this filesystem can be mounted as requested,
2392 * 0 if it cannot be.
2394 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2396 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2397 ext4_msg(sb
, KERN_ERR
,
2398 "Couldn't mount because of "
2399 "unsupported optional features (%x)",
2400 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2401 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2408 /* Check that feature set is OK for a read-write mount */
2409 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2410 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2411 "unsupported optional features (%x)",
2412 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2413 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2417 * Large file size enabled file system can only be mounted
2418 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2420 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2421 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2422 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2423 "cannot be mounted RDWR without "
2431 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
2432 __releases(kernel_lock
)
2433 __acquires(kernel_lock
)
2435 struct buffer_head
*bh
;
2436 struct ext4_super_block
*es
= NULL
;
2437 struct ext4_sb_info
*sbi
;
2439 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2440 ext4_fsblk_t logical_sb_block
;
2441 unsigned long offset
= 0;
2442 unsigned long journal_devnum
= 0;
2443 unsigned long def_mount_opts
;
2449 unsigned int db_count
;
2451 int needs_recovery
, has_huge_files
;
2454 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2456 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
2460 sbi
->s_blockgroup_lock
=
2461 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
2462 if (!sbi
->s_blockgroup_lock
) {
2466 sb
->s_fs_info
= sbi
;
2467 sbi
->s_mount_opt
= 0;
2468 sbi
->s_resuid
= EXT4_DEF_RESUID
;
2469 sbi
->s_resgid
= EXT4_DEF_RESGID
;
2470 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
2471 sbi
->s_sb_block
= sb_block
;
2472 sbi
->s_sectors_written_start
= part_stat_read(sb
->s_bdev
->bd_part
,
2477 /* Cleanup superblock name */
2478 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
2481 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
2483 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
2488 * The ext4 superblock will not be buffer aligned for other than 1kB
2489 * block sizes. We need to calculate the offset from buffer start.
2491 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
2492 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2493 offset
= do_div(logical_sb_block
, blocksize
);
2495 logical_sb_block
= sb_block
;
2498 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
2499 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
2503 * Note: s_es must be initialized as soon as possible because
2504 * some ext4 macro-instructions depend on its value
2506 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2508 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
2509 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
2511 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
2513 /* Set defaults before we parse the mount options */
2514 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
2515 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
2516 set_opt(sbi
->s_mount_opt
, DEBUG
);
2517 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
2518 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
2520 set_opt(sbi
->s_mount_opt
, GRPID
);
2522 if (def_mount_opts
& EXT4_DEFM_UID16
)
2523 set_opt(sbi
->s_mount_opt
, NO_UID32
);
2524 #ifdef CONFIG_EXT4_FS_XATTR
2525 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
2526 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
2528 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2529 if (def_mount_opts
& EXT4_DEFM_ACL
)
2530 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
2532 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
2533 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2534 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
2535 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2536 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
2537 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2539 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2540 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2541 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2542 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2544 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2546 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2547 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2548 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
2549 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
2550 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
2552 set_opt(sbi
->s_mount_opt
, BARRIER
);
2555 * enable delayed allocation by default
2556 * Use -o nodelalloc to turn it off
2558 if (!IS_EXT3_SB(sb
))
2559 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2561 if (!parse_options((char *) data
, sb
, &journal_devnum
,
2562 &journal_ioprio
, NULL
, 0))
2565 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2566 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
2568 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2569 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2570 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2571 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2572 ext4_msg(sb
, KERN_WARNING
,
2573 "feature flags set on rev 0 fs, "
2574 "running e2fsck is recommended");
2577 * Check feature flags regardless of the revision level, since we
2578 * previously didn't change the revision level when setting the flags,
2579 * so there is a chance incompat flags are set on a rev 0 filesystem.
2581 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
2584 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2586 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2587 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2588 ext4_msg(sb
, KERN_ERR
,
2589 "Unsupported filesystem blocksize %d", blocksize
);
2593 if (sb
->s_blocksize
!= blocksize
) {
2594 /* Validate the filesystem blocksize */
2595 if (!sb_set_blocksize(sb
, blocksize
)) {
2596 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
2602 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2603 offset
= do_div(logical_sb_block
, blocksize
);
2604 bh
= sb_bread(sb
, logical_sb_block
);
2606 ext4_msg(sb
, KERN_ERR
,
2607 "Can't read superblock on 2nd try");
2610 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2612 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2613 ext4_msg(sb
, KERN_ERR
,
2614 "Magic mismatch, very weird!");
2619 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2620 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
2621 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
2623 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
2625 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2626 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2627 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2629 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2630 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2631 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2632 (!is_power_of_2(sbi
->s_inode_size
)) ||
2633 (sbi
->s_inode_size
> blocksize
)) {
2634 ext4_msg(sb
, KERN_ERR
,
2635 "unsupported inode size: %d",
2639 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2640 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2643 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2644 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2645 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2646 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2647 !is_power_of_2(sbi
->s_desc_size
)) {
2648 ext4_msg(sb
, KERN_ERR
,
2649 "unsupported descriptor size %lu",
2654 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2656 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2657 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2658 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2661 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2662 if (sbi
->s_inodes_per_block
== 0)
2664 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2665 sbi
->s_inodes_per_block
;
2666 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2668 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2669 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2670 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2672 for (i
= 0; i
< 4; i
++)
2673 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2674 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2675 i
= le32_to_cpu(es
->s_flags
);
2676 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
2677 sbi
->s_hash_unsigned
= 3;
2678 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
2679 #ifdef __CHAR_UNSIGNED__
2680 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
2681 sbi
->s_hash_unsigned
= 3;
2683 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
2688 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2689 ext4_msg(sb
, KERN_ERR
,
2690 "#blocks per group too big: %lu",
2691 sbi
->s_blocks_per_group
);
2694 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2695 ext4_msg(sb
, KERN_ERR
,
2696 "#inodes per group too big: %lu",
2697 sbi
->s_inodes_per_group
);
2702 * Test whether we have more sectors than will fit in sector_t,
2703 * and whether the max offset is addressable by the page cache.
2705 if ((ext4_blocks_count(es
) >
2706 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) ||
2707 (ext4_blocks_count(es
) >
2708 (pgoff_t
)(~0ULL) >> (PAGE_CACHE_SHIFT
- sb
->s_blocksize_bits
))) {
2709 ext4_msg(sb
, KERN_ERR
, "filesystem"
2710 " too large to mount safely on this system");
2711 if (sizeof(sector_t
) < 8)
2712 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
2717 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2720 /* check blocks count against device size */
2721 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
2722 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
2723 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
2724 "exceeds size of device (%llu blocks)",
2725 ext4_blocks_count(es
), blocks_count
);
2730 * It makes no sense for the first data block to be beyond the end
2731 * of the filesystem.
2733 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
2734 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
2735 "block %u is beyond end of filesystem (%llu)",
2736 le32_to_cpu(es
->s_first_data_block
),
2737 ext4_blocks_count(es
));
2740 blocks_count
= (ext4_blocks_count(es
) -
2741 le32_to_cpu(es
->s_first_data_block
) +
2742 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2743 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2744 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
2745 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
2746 "(block count %llu, first data block %u, "
2747 "blocks per group %lu)", sbi
->s_groups_count
,
2748 ext4_blocks_count(es
),
2749 le32_to_cpu(es
->s_first_data_block
),
2750 EXT4_BLOCKS_PER_GROUP(sb
));
2753 sbi
->s_groups_count
= blocks_count
;
2754 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
2755 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
2756 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2757 EXT4_DESC_PER_BLOCK(sb
);
2758 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2760 if (sbi
->s_group_desc
== NULL
) {
2761 ext4_msg(sb
, KERN_ERR
, "not enough memory");
2765 #ifdef CONFIG_PROC_FS
2767 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2770 bgl_lock_init(sbi
->s_blockgroup_lock
);
2772 for (i
= 0; i
< db_count
; i
++) {
2773 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2774 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2775 if (!sbi
->s_group_desc
[i
]) {
2776 ext4_msg(sb
, KERN_ERR
,
2777 "can't read group descriptor %d", i
);
2782 if (!ext4_check_descriptors(sb
)) {
2783 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
2786 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2787 if (!ext4_fill_flex_info(sb
)) {
2788 ext4_msg(sb
, KERN_ERR
,
2789 "unable to initialize "
2790 "flex_bg meta info!");
2794 sbi
->s_gdb_count
= db_count
;
2795 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2796 spin_lock_init(&sbi
->s_next_gen_lock
);
2798 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2799 sbi
->s_max_writeback_mb_bump
= 128;
2802 * set up enough so that it can read an inode
2804 if (!test_opt(sb
, NOLOAD
) &&
2805 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
2806 sb
->s_op
= &ext4_sops
;
2808 sb
->s_op
= &ext4_nojournal_sops
;
2809 sb
->s_export_op
= &ext4_export_ops
;
2810 sb
->s_xattr
= ext4_xattr_handlers
;
2812 sb
->s_qcop
= &ext4_qctl_operations
;
2813 sb
->dq_op
= &ext4_quota_operations
;
2815 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2816 mutex_init(&sbi
->s_orphan_lock
);
2817 mutex_init(&sbi
->s_resize_lock
);
2821 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2822 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2823 EXT4_FEATURE_INCOMPAT_RECOVER
));
2826 * The first inode we look at is the journal inode. Don't try
2827 * root first: it may be modified in the journal!
2829 if (!test_opt(sb
, NOLOAD
) &&
2830 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2831 if (ext4_load_journal(sb
, es
, journal_devnum
))
2833 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
2834 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2835 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
2836 "suppressed and not mounted read-only");
2837 goto failed_mount_wq
;
2839 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
2840 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2841 sbi
->s_journal
= NULL
;
2846 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2847 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2848 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2849 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
2850 goto failed_mount_wq
;
2853 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2854 jbd2_journal_set_features(sbi
->s_journal
,
2855 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2856 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2857 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2858 jbd2_journal_set_features(sbi
->s_journal
,
2859 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2860 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2861 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2863 jbd2_journal_clear_features(sbi
->s_journal
,
2864 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2865 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2868 /* We have now updated the journal if required, so we can
2869 * validate the data journaling mode. */
2870 switch (test_opt(sb
, DATA_FLAGS
)) {
2872 /* No mode set, assume a default based on the journal
2873 * capabilities: ORDERED_DATA if the journal can
2874 * cope, else JOURNAL_DATA
2876 if (jbd2_journal_check_available_features
2877 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2878 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2880 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2883 case EXT4_MOUNT_ORDERED_DATA
:
2884 case EXT4_MOUNT_WRITEBACK_DATA
:
2885 if (!jbd2_journal_check_available_features
2886 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2887 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
2888 "requested data journaling mode");
2889 goto failed_mount_wq
;
2894 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
2897 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2898 ext4_count_free_blocks(sb
));
2900 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2901 ext4_count_free_inodes(sb
));
2903 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2904 ext4_count_dirs(sb
));
2906 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
2908 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
2909 goto failed_mount_wq
;
2911 if (test_opt(sb
, NOBH
)) {
2912 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2913 ext4_msg(sb
, KERN_WARNING
, "Ignoring nobh option - "
2914 "its supported only with writeback mode");
2915 clear_opt(sbi
->s_mount_opt
, NOBH
);
2917 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
2918 ext4_msg(sb
, KERN_WARNING
, "dioread_nolock option is "
2919 "not supported with nobh mode");
2920 goto failed_mount_wq
;
2923 EXT4_SB(sb
)->dio_unwritten_wq
= create_workqueue("ext4-dio-unwritten");
2924 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
2925 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
2926 goto failed_mount_wq
;
2930 * The jbd2_journal_load will have done any necessary log recovery,
2931 * so we can safely mount the rest of the filesystem now.
2934 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2936 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
2937 ret
= PTR_ERR(root
);
2940 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2942 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
2945 sb
->s_root
= d_alloc_root(root
);
2947 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
2953 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
2955 /* determine the minimum size of new large inodes, if present */
2956 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2957 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2958 EXT4_GOOD_OLD_INODE_SIZE
;
2959 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2960 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2961 if (sbi
->s_want_extra_isize
<
2962 le16_to_cpu(es
->s_want_extra_isize
))
2963 sbi
->s_want_extra_isize
=
2964 le16_to_cpu(es
->s_want_extra_isize
);
2965 if (sbi
->s_want_extra_isize
<
2966 le16_to_cpu(es
->s_min_extra_isize
))
2967 sbi
->s_want_extra_isize
=
2968 le16_to_cpu(es
->s_min_extra_isize
);
2971 /* Check if enough inode space is available */
2972 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2973 sbi
->s_inode_size
) {
2974 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2975 EXT4_GOOD_OLD_INODE_SIZE
;
2976 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
2980 if (test_opt(sb
, DELALLOC
) &&
2981 (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)) {
2982 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
2983 "requested data journaling mode");
2984 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2986 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
2987 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
2988 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
2989 "option - requested data journaling mode");
2990 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
2992 if (sb
->s_blocksize
< PAGE_SIZE
) {
2993 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
2994 "option - block size is too small");
2995 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
2999 err
= ext4_setup_system_zone(sb
);
3001 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3007 err
= ext4_mb_init(sb
, needs_recovery
);
3009 ext4_msg(sb
, KERN_ERR
, "failed to initalize mballoc (%d)",
3014 sbi
->s_kobj
.kset
= ext4_kset
;
3015 init_completion(&sbi
->s_kobj_unregister
);
3016 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3019 ext4_mb_release(sb
);
3020 ext4_ext_release(sb
);
3024 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3025 ext4_orphan_cleanup(sb
, es
);
3026 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3027 if (needs_recovery
) {
3028 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3029 ext4_mark_recovery_complete(sb
, es
);
3031 if (EXT4_SB(sb
)->s_journal
) {
3032 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3033 descr
= " journalled data mode";
3034 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3035 descr
= " ordered data mode";
3037 descr
= " writeback data mode";
3039 descr
= "out journal";
3041 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s", descr
);
3048 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3052 ext4_msg(sb
, KERN_ERR
, "mount failed");
3053 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3055 ext4_release_system_zone(sb
);
3056 if (sbi
->s_journal
) {
3057 jbd2_journal_destroy(sbi
->s_journal
);
3058 sbi
->s_journal
= NULL
;
3060 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
3061 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3062 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3063 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
3065 if (sbi
->s_flex_groups
) {
3066 if (is_vmalloc_addr(sbi
->s_flex_groups
))
3067 vfree(sbi
->s_flex_groups
);
3069 kfree(sbi
->s_flex_groups
);
3072 for (i
= 0; i
< db_count
; i
++)
3073 brelse(sbi
->s_group_desc
[i
]);
3074 kfree(sbi
->s_group_desc
);
3077 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3080 for (i
= 0; i
< MAXQUOTAS
; i
++)
3081 kfree(sbi
->s_qf_names
[i
]);
3083 ext4_blkdev_remove(sbi
);
3086 sb
->s_fs_info
= NULL
;
3087 kfree(sbi
->s_blockgroup_lock
);
3094 * Setup any per-fs journal parameters now. We'll do this both on
3095 * initial mount, once the journal has been initialised but before we've
3096 * done any recovery; and again on any subsequent remount.
3098 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3100 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3102 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3103 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3104 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3106 spin_lock(&journal
->j_state_lock
);
3107 if (test_opt(sb
, BARRIER
))
3108 journal
->j_flags
|= JBD2_BARRIER
;
3110 journal
->j_flags
&= ~JBD2_BARRIER
;
3111 if (test_opt(sb
, DATA_ERR_ABORT
))
3112 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3114 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3115 spin_unlock(&journal
->j_state_lock
);
3118 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3119 unsigned int journal_inum
)
3121 struct inode
*journal_inode
;
3124 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3126 /* First, test for the existence of a valid inode on disk. Bad
3127 * things happen if we iget() an unused inode, as the subsequent
3128 * iput() will try to delete it. */
3130 journal_inode
= ext4_iget(sb
, journal_inum
);
3131 if (IS_ERR(journal_inode
)) {
3132 ext4_msg(sb
, KERN_ERR
, "no journal found");
3135 if (!journal_inode
->i_nlink
) {
3136 make_bad_inode(journal_inode
);
3137 iput(journal_inode
);
3138 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3142 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3143 journal_inode
, journal_inode
->i_size
);
3144 if (!S_ISREG(journal_inode
->i_mode
)) {
3145 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3146 iput(journal_inode
);
3150 journal
= jbd2_journal_init_inode(journal_inode
);
3152 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3153 iput(journal_inode
);
3156 journal
->j_private
= sb
;
3157 ext4_init_journal_params(sb
, journal
);
3161 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3164 struct buffer_head
*bh
;
3168 int hblock
, blocksize
;
3169 ext4_fsblk_t sb_block
;
3170 unsigned long offset
;
3171 struct ext4_super_block
*es
;
3172 struct block_device
*bdev
;
3174 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3176 bdev
= ext4_blkdev_get(j_dev
, sb
);
3180 if (bd_claim(bdev
, sb
)) {
3181 ext4_msg(sb
, KERN_ERR
,
3182 "failed to claim external journal device");
3183 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
3187 blocksize
= sb
->s_blocksize
;
3188 hblock
= bdev_logical_block_size(bdev
);
3189 if (blocksize
< hblock
) {
3190 ext4_msg(sb
, KERN_ERR
,
3191 "blocksize too small for journal device");
3195 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3196 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3197 set_blocksize(bdev
, blocksize
);
3198 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3199 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3200 "external journal");
3204 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3205 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3206 !(le32_to_cpu(es
->s_feature_incompat
) &
3207 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3208 ext4_msg(sb
, KERN_ERR
, "external journal has "
3214 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3215 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3220 len
= ext4_blocks_count(es
);
3221 start
= sb_block
+ 1;
3222 brelse(bh
); /* we're done with the superblock */
3224 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3225 start
, len
, blocksize
);
3227 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3230 journal
->j_private
= sb
;
3231 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3232 wait_on_buffer(journal
->j_sb_buffer
);
3233 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3234 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3237 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3238 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3239 "user (unsupported) - %d",
3240 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3243 EXT4_SB(sb
)->journal_bdev
= bdev
;
3244 ext4_init_journal_params(sb
, journal
);
3248 jbd2_journal_destroy(journal
);
3250 ext4_blkdev_put(bdev
);
3254 static int ext4_load_journal(struct super_block
*sb
,
3255 struct ext4_super_block
*es
,
3256 unsigned long journal_devnum
)
3259 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3262 int really_read_only
;
3264 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3266 if (journal_devnum
&&
3267 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3268 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3269 "numbers have changed");
3270 journal_dev
= new_decode_dev(journal_devnum
);
3272 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3274 really_read_only
= bdev_read_only(sb
->s_bdev
);
3277 * Are we loading a blank journal or performing recovery after a
3278 * crash? For recovery, we need to check in advance whether we
3279 * can get read-write access to the device.
3281 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3282 if (sb
->s_flags
& MS_RDONLY
) {
3283 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3284 "required on readonly filesystem");
3285 if (really_read_only
) {
3286 ext4_msg(sb
, KERN_ERR
, "write access "
3287 "unavailable, cannot proceed");
3290 ext4_msg(sb
, KERN_INFO
, "write access will "
3291 "be enabled during recovery");
3295 if (journal_inum
&& journal_dev
) {
3296 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3297 "and inode journals!");
3302 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3305 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3309 if (!(journal
->j_flags
& JBD2_BARRIER
))
3310 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3312 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
3313 err
= jbd2_journal_update_format(journal
);
3315 ext4_msg(sb
, KERN_ERR
, "error updating journal");
3316 jbd2_journal_destroy(journal
);
3321 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
3322 err
= jbd2_journal_wipe(journal
, !really_read_only
);
3324 err
= jbd2_journal_load(journal
);
3327 ext4_msg(sb
, KERN_ERR
, "error loading journal");
3328 jbd2_journal_destroy(journal
);
3332 EXT4_SB(sb
)->s_journal
= journal
;
3333 ext4_clear_journal_err(sb
, es
);
3335 if (journal_devnum
&&
3336 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3337 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
3339 /* Make sure we flush the recovery flag to disk. */
3340 ext4_commit_super(sb
, 1);
3346 static int ext4_commit_super(struct super_block
*sb
, int sync
)
3348 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
3349 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
3354 if (buffer_write_io_error(sbh
)) {
3356 * Oh, dear. A previous attempt to write the
3357 * superblock failed. This could happen because the
3358 * USB device was yanked out. Or it could happen to
3359 * be a transient write error and maybe the block will
3360 * be remapped. Nothing we can do but to retry the
3361 * write and hope for the best.
3363 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
3364 "superblock detected");
3365 clear_buffer_write_io_error(sbh
);
3366 set_buffer_uptodate(sbh
);
3369 * If the file system is mounted read-only, don't update the
3370 * superblock write time. This avoids updating the superblock
3371 * write time when we are mounting the root file system
3372 * read/only but we need to replay the journal; at that point,
3373 * for people who are east of GMT and who make their clock
3374 * tick in localtime for Windows bug-for-bug compatibility,
3375 * the clock is set in the future, and this will cause e2fsck
3376 * to complain and force a full file system check.
3378 if (!(sb
->s_flags
& MS_RDONLY
))
3379 es
->s_wtime
= cpu_to_le32(get_seconds());
3380 es
->s_kbytes_written
=
3381 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
3382 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
3383 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
3384 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
3385 &EXT4_SB(sb
)->s_freeblocks_counter
));
3386 es
->s_free_inodes_count
= cpu_to_le32(percpu_counter_sum_positive(
3387 &EXT4_SB(sb
)->s_freeinodes_counter
));
3389 BUFFER_TRACE(sbh
, "marking dirty");
3390 mark_buffer_dirty(sbh
);
3392 error
= sync_dirty_buffer(sbh
);
3396 error
= buffer_write_io_error(sbh
);
3398 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
3400 clear_buffer_write_io_error(sbh
);
3401 set_buffer_uptodate(sbh
);
3408 * Have we just finished recovery? If so, and if we are mounting (or
3409 * remounting) the filesystem readonly, then we will end up with a
3410 * consistent fs on disk. Record that fact.
3412 static void ext4_mark_recovery_complete(struct super_block
*sb
,
3413 struct ext4_super_block
*es
)
3415 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
3417 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3418 BUG_ON(journal
!= NULL
);
3421 jbd2_journal_lock_updates(journal
);
3422 if (jbd2_journal_flush(journal
) < 0)
3425 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
3426 sb
->s_flags
& MS_RDONLY
) {
3427 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3428 ext4_commit_super(sb
, 1);
3432 jbd2_journal_unlock_updates(journal
);
3436 * If we are mounting (or read-write remounting) a filesystem whose journal
3437 * has recorded an error from a previous lifetime, move that error to the
3438 * main filesystem now.
3440 static void ext4_clear_journal_err(struct super_block
*sb
,
3441 struct ext4_super_block
*es
)
3447 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3449 journal
= EXT4_SB(sb
)->s_journal
;
3452 * Now check for any error status which may have been recorded in the
3453 * journal by a prior ext4_error() or ext4_abort()
3456 j_errno
= jbd2_journal_errno(journal
);
3460 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
3461 ext4_warning(sb
, "Filesystem error recorded "
3462 "from previous mount: %s", errstr
);
3463 ext4_warning(sb
, "Marking fs in need of filesystem check.");
3465 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
3466 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
3467 ext4_commit_super(sb
, 1);
3469 jbd2_journal_clear_err(journal
);
3474 * Force the running and committing transactions to commit,
3475 * and wait on the commit.
3477 int ext4_force_commit(struct super_block
*sb
)
3482 if (sb
->s_flags
& MS_RDONLY
)
3485 journal
= EXT4_SB(sb
)->s_journal
;
3487 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
3488 ret
= ext4_journal_force_commit(journal
);
3494 static void ext4_write_super(struct super_block
*sb
)
3497 ext4_commit_super(sb
, 1);
3501 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
3505 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3507 trace_ext4_sync_fs(sb
, wait
);
3508 flush_workqueue(sbi
->dio_unwritten_wq
);
3509 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
3511 jbd2_log_wait_commit(sbi
->s_journal
, target
);
3517 * LVM calls this function before a (read-only) snapshot is created. This
3518 * gives us a chance to flush the journal completely and mark the fs clean.
3520 static int ext4_freeze(struct super_block
*sb
)
3525 if (sb
->s_flags
& MS_RDONLY
)
3528 journal
= EXT4_SB(sb
)->s_journal
;
3530 /* Now we set up the journal barrier. */
3531 jbd2_journal_lock_updates(journal
);
3534 * Don't clear the needs_recovery flag if we failed to flush
3537 error
= jbd2_journal_flush(journal
);
3541 /* Journal blocked and flushed, clear needs_recovery flag. */
3542 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3543 error
= ext4_commit_super(sb
, 1);
3545 /* we rely on s_frozen to stop further updates */
3546 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3551 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3552 * flag here, even though the filesystem is not technically dirty yet.
3554 static int ext4_unfreeze(struct super_block
*sb
)
3556 if (sb
->s_flags
& MS_RDONLY
)
3560 /* Reset the needs_recovery flag before the fs is unlocked. */
3561 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3562 ext4_commit_super(sb
, 1);
3567 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3569 struct ext4_super_block
*es
;
3570 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3571 ext4_fsblk_t n_blocks_count
= 0;
3572 unsigned long old_sb_flags
;
3573 struct ext4_mount_options old_opts
;
3575 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3583 /* Store the original options */
3585 old_sb_flags
= sb
->s_flags
;
3586 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3587 old_opts
.s_resuid
= sbi
->s_resuid
;
3588 old_opts
.s_resgid
= sbi
->s_resgid
;
3589 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3590 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
3591 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
3593 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3594 for (i
= 0; i
< MAXQUOTAS
; i
++)
3595 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3597 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
3598 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
3601 * Allow the "check" option to be passed as a remount option.
3603 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
3604 &n_blocks_count
, 1)) {
3609 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
3610 ext4_abort(sb
, __func__
, "Abort forced by user");
3612 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3613 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3617 if (sbi
->s_journal
) {
3618 ext4_init_journal_params(sb
, sbi
->s_journal
);
3619 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3622 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3623 n_blocks_count
> ext4_blocks_count(es
)) {
3624 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
3629 if (*flags
& MS_RDONLY
) {
3631 * First of all, the unconditional stuff we have to do
3632 * to disable replay of the journal when we next remount
3634 sb
->s_flags
|= MS_RDONLY
;
3637 * OK, test if we are remounting a valid rw partition
3638 * readonly, and if so set the rdonly flag and then
3639 * mark the partition as valid again.
3641 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3642 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3643 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3646 ext4_mark_recovery_complete(sb
, es
);
3648 /* Make sure we can mount this feature set readwrite */
3649 if (!ext4_feature_set_ok(sb
, 0)) {
3654 * Make sure the group descriptor checksums
3655 * are sane. If they aren't, refuse to remount r/w.
3657 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3658 struct ext4_group_desc
*gdp
=
3659 ext4_get_group_desc(sb
, g
, NULL
);
3661 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3662 ext4_msg(sb
, KERN_ERR
,
3663 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3664 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3665 le16_to_cpu(gdp
->bg_checksum
));
3672 * If we have an unprocessed orphan list hanging
3673 * around from a previously readonly bdev mount,
3674 * require a full umount/remount for now.
3676 if (es
->s_last_orphan
) {
3677 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
3678 "remount RDWR because of unprocessed "
3679 "orphan inode list. Please "
3680 "umount/remount instead");
3686 * Mounting a RDONLY partition read-write, so reread
3687 * and store the current valid flag. (It may have
3688 * been changed by e2fsck since we originally mounted
3692 ext4_clear_journal_err(sb
, es
);
3693 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3694 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3696 if (!ext4_setup_super(sb
, es
, 0))
3697 sb
->s_flags
&= ~MS_RDONLY
;
3700 ext4_setup_system_zone(sb
);
3701 if (sbi
->s_journal
== NULL
)
3702 ext4_commit_super(sb
, 1);
3705 /* Release old quota file names */
3706 for (i
= 0; i
< MAXQUOTAS
; i
++)
3707 if (old_opts
.s_qf_names
[i
] &&
3708 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3709 kfree(old_opts
.s_qf_names
[i
]);
3716 sb
->s_flags
= old_sb_flags
;
3717 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3718 sbi
->s_resuid
= old_opts
.s_resuid
;
3719 sbi
->s_resgid
= old_opts
.s_resgid
;
3720 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3721 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
3722 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
3724 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3725 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3726 if (sbi
->s_qf_names
[i
] &&
3727 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3728 kfree(sbi
->s_qf_names
[i
]);
3729 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3737 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3739 struct super_block
*sb
= dentry
->d_sb
;
3740 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3741 struct ext4_super_block
*es
= sbi
->s_es
;
3744 if (test_opt(sb
, MINIX_DF
)) {
3745 sbi
->s_overhead_last
= 0;
3746 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3747 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3748 ext4_fsblk_t overhead
= 0;
3751 * Compute the overhead (FS structures). This is constant
3752 * for a given filesystem unless the number of block groups
3753 * changes so we cache the previous value until it does.
3757 * All of the blocks before first_data_block are
3760 overhead
= le32_to_cpu(es
->s_first_data_block
);
3763 * Add the overhead attributed to the superblock and
3764 * block group descriptors. If the sparse superblocks
3765 * feature is turned on, then not all groups have this.
3767 for (i
= 0; i
< ngroups
; i
++) {
3768 overhead
+= ext4_bg_has_super(sb
, i
) +
3769 ext4_bg_num_gdb(sb
, i
);
3774 * Every block group has an inode bitmap, a block
3775 * bitmap, and an inode table.
3777 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3778 sbi
->s_overhead_last
= overhead
;
3780 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3783 buf
->f_type
= EXT4_SUPER_MAGIC
;
3784 buf
->f_bsize
= sb
->s_blocksize
;
3785 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3786 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3787 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3788 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3789 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3791 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3792 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3793 buf
->f_namelen
= EXT4_NAME_LEN
;
3794 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3795 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3796 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3797 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3802 /* Helper function for writing quotas on sync - we need to start transaction
3803 * before quota file is locked for write. Otherwise the are possible deadlocks:
3804 * Process 1 Process 2
3805 * ext4_create() quota_sync()
3806 * jbd2_journal_start() write_dquot()
3807 * dquot_initialize() down(dqio_mutex)
3808 * down(dqio_mutex) jbd2_journal_start()
3814 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3816 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3819 static int ext4_write_dquot(struct dquot
*dquot
)
3823 struct inode
*inode
;
3825 inode
= dquot_to_inode(dquot
);
3826 handle
= ext4_journal_start(inode
,
3827 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3829 return PTR_ERR(handle
);
3830 ret
= dquot_commit(dquot
);
3831 err
= ext4_journal_stop(handle
);
3837 static int ext4_acquire_dquot(struct dquot
*dquot
)
3842 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3843 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3845 return PTR_ERR(handle
);
3846 ret
= dquot_acquire(dquot
);
3847 err
= ext4_journal_stop(handle
);
3853 static int ext4_release_dquot(struct dquot
*dquot
)
3858 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3859 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3860 if (IS_ERR(handle
)) {
3861 /* Release dquot anyway to avoid endless cycle in dqput() */
3862 dquot_release(dquot
);
3863 return PTR_ERR(handle
);
3865 ret
= dquot_release(dquot
);
3866 err
= ext4_journal_stop(handle
);
3872 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3874 /* Are we journaling quotas? */
3875 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3876 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3877 dquot_mark_dquot_dirty(dquot
);
3878 return ext4_write_dquot(dquot
);
3880 return dquot_mark_dquot_dirty(dquot
);
3884 static int ext4_write_info(struct super_block
*sb
, int type
)
3889 /* Data block + inode block */
3890 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3892 return PTR_ERR(handle
);
3893 ret
= dquot_commit_info(sb
, type
);
3894 err
= ext4_journal_stop(handle
);
3901 * Turn on quotas during mount time - we need to find
3902 * the quota file and such...
3904 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3906 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3907 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3911 * Standard function to be called on quota_on
3913 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3914 char *name
, int remount
)
3919 if (!test_opt(sb
, QUOTA
))
3921 /* When remounting, no checks are needed and in fact, name is NULL */
3923 return vfs_quota_on(sb
, type
, format_id
, name
, remount
);
3925 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
3929 /* Quotafile not on the same filesystem? */
3930 if (path
.mnt
->mnt_sb
!= sb
) {
3934 /* Journaling quota? */
3935 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3936 /* Quotafile not in fs root? */
3937 if (path
.dentry
->d_parent
!= sb
->s_root
)
3938 ext4_msg(sb
, KERN_WARNING
,
3939 "Quota file not on filesystem root. "
3940 "Journaled quota will not work");
3944 * When we journal data on quota file, we have to flush journal to see
3945 * all updates to the file when we bypass pagecache...
3947 if (EXT4_SB(sb
)->s_journal
&&
3948 ext4_should_journal_data(path
.dentry
->d_inode
)) {
3950 * We don't need to lock updates but journal_flush() could
3951 * otherwise be livelocked...
3953 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3954 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3955 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3962 err
= vfs_quota_on_path(sb
, type
, format_id
, &path
);
3967 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3968 * acquiring the locks... As quota files are never truncated and quota code
3969 * itself serializes the operations (and noone else should touch the files)
3970 * we don't have to be afraid of races */
3971 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3972 size_t len
, loff_t off
)
3974 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3975 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3977 int offset
= off
& (sb
->s_blocksize
- 1);
3980 struct buffer_head
*bh
;
3981 loff_t i_size
= i_size_read(inode
);
3985 if (off
+len
> i_size
)
3988 while (toread
> 0) {
3989 tocopy
= sb
->s_blocksize
- offset
< toread
?
3990 sb
->s_blocksize
- offset
: toread
;
3991 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3994 if (!bh
) /* A hole? */
3995 memset(data
, 0, tocopy
);
3997 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4007 /* Write to quotafile (we know the transaction is already started and has
4008 * enough credits) */
4009 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4010 const char *data
, size_t len
, loff_t off
)
4012 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4013 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4015 int offset
= off
& (sb
->s_blocksize
- 1);
4016 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
4017 struct buffer_head
*bh
;
4018 handle_t
*handle
= journal_current_handle();
4020 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4021 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4022 " cancelled because transaction is not started",
4023 (unsigned long long)off
, (unsigned long long)len
);
4027 * Since we account only one data block in transaction credits,
4028 * then it is impossible to cross a block boundary.
4030 if (sb
->s_blocksize
- offset
< len
) {
4031 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4032 " cancelled because not block aligned",
4033 (unsigned long long)off
, (unsigned long long)len
);
4037 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4038 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4041 if (journal_quota
) {
4042 err
= ext4_journal_get_write_access(handle
, bh
);
4049 memcpy(bh
->b_data
+offset
, data
, len
);
4050 flush_dcache_page(bh
->b_page
);
4053 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4055 /* Always do at least ordered writes for quotas */
4056 err
= ext4_jbd2_file_inode(handle
, inode
);
4057 mark_buffer_dirty(bh
);
4062 mutex_unlock(&inode
->i_mutex
);
4065 if (inode
->i_size
< off
+ len
) {
4066 i_size_write(inode
, off
+ len
);
4067 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4069 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4070 ext4_mark_inode_dirty(handle
, inode
);
4071 mutex_unlock(&inode
->i_mutex
);
4077 static int ext4_get_sb(struct file_system_type
*fs_type
, int flags
,
4078 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
4080 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
,mnt
);
4083 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4084 static struct file_system_type ext2_fs_type
= {
4085 .owner
= THIS_MODULE
,
4087 .get_sb
= ext4_get_sb
,
4088 .kill_sb
= kill_block_super
,
4089 .fs_flags
= FS_REQUIRES_DEV
,
4092 static inline void register_as_ext2(void)
4094 int err
= register_filesystem(&ext2_fs_type
);
4097 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4100 static inline void unregister_as_ext2(void)
4102 unregister_filesystem(&ext2_fs_type
);
4104 MODULE_ALIAS("ext2");
4106 static inline void register_as_ext2(void) { }
4107 static inline void unregister_as_ext2(void) { }
4110 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4111 static inline void register_as_ext3(void)
4113 int err
= register_filesystem(&ext3_fs_type
);
4116 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4119 static inline void unregister_as_ext3(void)
4121 unregister_filesystem(&ext3_fs_type
);
4123 MODULE_ALIAS("ext3");
4125 static inline void register_as_ext3(void) { }
4126 static inline void unregister_as_ext3(void) { }
4129 static struct file_system_type ext4_fs_type
= {
4130 .owner
= THIS_MODULE
,
4132 .get_sb
= ext4_get_sb
,
4133 .kill_sb
= kill_block_super
,
4134 .fs_flags
= FS_REQUIRES_DEV
,
4137 static int __init
init_ext4_fs(void)
4141 ext4_check_flag_values();
4142 err
= init_ext4_system_zone();
4145 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4148 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4149 err
= init_ext4_mballoc();
4153 err
= init_ext4_xattr();
4156 err
= init_inodecache();
4161 err
= register_filesystem(&ext4_fs_type
);
4166 unregister_as_ext2();
4167 unregister_as_ext3();
4168 destroy_inodecache();
4172 exit_ext4_mballoc();
4174 remove_proc_entry("fs/ext4", NULL
);
4175 kset_unregister(ext4_kset
);
4177 exit_ext4_system_zone();
4181 static void __exit
exit_ext4_fs(void)
4183 unregister_as_ext2();
4184 unregister_as_ext3();
4185 unregister_filesystem(&ext4_fs_type
);
4186 destroy_inodecache();
4188 exit_ext4_mballoc();
4189 remove_proc_entry("fs/ext4", NULL
);
4190 kset_unregister(ext4_kset
);
4191 exit_ext4_system_zone();
4194 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4195 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4196 MODULE_LICENSE("GPL");
4197 module_init(init_ext4_fs
)
4198 module_exit(exit_ext4_fs
)