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 static int default_mb_history_length
= 1000;
55 module_param_named(default_mb_history_length
, default_mb_history_length
,
57 MODULE_PARM_DESC(default_mb_history_length
,
58 "Default number of entries saved for mb_history");
60 struct proc_dir_entry
*ext4_proc_root
;
61 static struct kset
*ext4_kset
;
63 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
64 unsigned long journal_devnum
);
65 static int ext4_commit_super(struct super_block
*sb
, int sync
);
66 static void ext4_mark_recovery_complete(struct super_block
*sb
,
67 struct ext4_super_block
*es
);
68 static void ext4_clear_journal_err(struct super_block
*sb
,
69 struct ext4_super_block
*es
);
70 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
71 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
73 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
74 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
75 static int ext4_unfreeze(struct super_block
*sb
);
76 static void ext4_write_super(struct super_block
*sb
);
77 static int ext4_freeze(struct super_block
*sb
);
80 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
81 struct ext4_group_desc
*bg
)
83 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
84 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
85 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
88 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
89 struct ext4_group_desc
*bg
)
91 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
92 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
93 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
96 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
97 struct ext4_group_desc
*bg
)
99 return le32_to_cpu(bg
->bg_inode_table_lo
) |
100 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
101 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
104 __u32
ext4_free_blks_count(struct super_block
*sb
,
105 struct ext4_group_desc
*bg
)
107 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
108 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
109 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
112 __u32
ext4_free_inodes_count(struct super_block
*sb
,
113 struct ext4_group_desc
*bg
)
115 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
116 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
117 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
120 __u32
ext4_used_dirs_count(struct super_block
*sb
,
121 struct ext4_group_desc
*bg
)
123 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
124 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
125 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
128 __u32
ext4_itable_unused_count(struct super_block
*sb
,
129 struct ext4_group_desc
*bg
)
131 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
132 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
133 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
136 void ext4_block_bitmap_set(struct super_block
*sb
,
137 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
139 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
140 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
141 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
144 void ext4_inode_bitmap_set(struct super_block
*sb
,
145 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
147 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
148 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
149 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
152 void ext4_inode_table_set(struct super_block
*sb
,
153 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
155 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
156 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
157 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
160 void ext4_free_blks_set(struct super_block
*sb
,
161 struct ext4_group_desc
*bg
, __u32 count
)
163 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
164 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
165 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
168 void ext4_free_inodes_set(struct super_block
*sb
,
169 struct ext4_group_desc
*bg
, __u32 count
)
171 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
172 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
173 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
176 void ext4_used_dirs_set(struct super_block
*sb
,
177 struct ext4_group_desc
*bg
, __u32 count
)
179 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
180 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
181 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
184 void ext4_itable_unused_set(struct super_block
*sb
,
185 struct ext4_group_desc
*bg
, __u32 count
)
187 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
188 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
189 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
193 /* Just increment the non-pointer handle value */
194 static handle_t
*ext4_get_nojournal(void)
196 handle_t
*handle
= current
->journal_info
;
197 unsigned long ref_cnt
= (unsigned long)handle
;
199 BUG_ON(ref_cnt
>= EXT4_NOJOURNAL_MAX_REF_COUNT
);
202 handle
= (handle_t
*)ref_cnt
;
204 current
->journal_info
= handle
;
209 /* Decrement the non-pointer handle value */
210 static void ext4_put_nojournal(handle_t
*handle
)
212 unsigned long ref_cnt
= (unsigned long)handle
;
214 BUG_ON(ref_cnt
== 0);
217 handle
= (handle_t
*)ref_cnt
;
219 current
->journal_info
= handle
;
223 * Wrappers for jbd2_journal_start/end.
225 * The only special thing we need to do here is to make sure that all
226 * journal_end calls result in the superblock being marked dirty, so
227 * that sync() will call the filesystem's write_super callback if
230 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
234 if (sb
->s_flags
& MS_RDONLY
)
235 return ERR_PTR(-EROFS
);
237 /* Special case here: if the journal has aborted behind our
238 * backs (eg. EIO in the commit thread), then we still need to
239 * take the FS itself readonly cleanly. */
240 journal
= EXT4_SB(sb
)->s_journal
;
242 if (is_journal_aborted(journal
)) {
243 ext4_abort(sb
, __func__
, "Detected aborted journal");
244 return ERR_PTR(-EROFS
);
246 return jbd2_journal_start(journal
, nblocks
);
248 return ext4_get_nojournal();
252 * The only special thing we need to do here is to make sure that all
253 * jbd2_journal_stop calls result in the superblock being marked dirty, so
254 * that sync() will call the filesystem's write_super callback if
257 int __ext4_journal_stop(const char *where
, handle_t
*handle
)
259 struct super_block
*sb
;
263 if (!ext4_handle_valid(handle
)) {
264 ext4_put_nojournal(handle
);
267 sb
= handle
->h_transaction
->t_journal
->j_private
;
269 rc
= jbd2_journal_stop(handle
);
274 __ext4_std_error(sb
, where
, err
);
278 void ext4_journal_abort_handle(const char *caller
, const char *err_fn
,
279 struct buffer_head
*bh
, handle_t
*handle
, int err
)
282 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
284 BUG_ON(!ext4_handle_valid(handle
));
287 BUFFER_TRACE(bh
, "abort");
292 if (is_handle_aborted(handle
))
295 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
296 caller
, errstr
, err_fn
);
298 jbd2_journal_abort_handle(handle
);
301 /* Deal with the reporting of failure conditions on a filesystem such as
302 * inconsistencies detected or read IO failures.
304 * On ext2, we can store the error state of the filesystem in the
305 * superblock. That is not possible on ext4, because we may have other
306 * write ordering constraints on the superblock which prevent us from
307 * writing it out straight away; and given that the journal is about to
308 * be aborted, we can't rely on the current, or future, transactions to
309 * write out the superblock safely.
311 * We'll just use the jbd2_journal_abort() error code to record an error in
312 * the journal instead. On recovery, the journal will compain about
313 * that error until we've noted it down and cleared it.
316 static void ext4_handle_error(struct super_block
*sb
)
318 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
320 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
321 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
323 if (sb
->s_flags
& MS_RDONLY
)
326 if (!test_opt(sb
, ERRORS_CONT
)) {
327 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
329 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
331 jbd2_journal_abort(journal
, -EIO
);
333 if (test_opt(sb
, ERRORS_RO
)) {
334 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
335 sb
->s_flags
|= MS_RDONLY
;
337 ext4_commit_super(sb
, 1);
338 if (test_opt(sb
, ERRORS_PANIC
))
339 panic("EXT4-fs (device %s): panic forced after error\n",
343 void ext4_error(struct super_block
*sb
, const char *function
,
344 const char *fmt
, ...)
349 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
354 ext4_handle_error(sb
);
357 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
364 errstr
= "IO failure";
367 errstr
= "Out of memory";
370 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
371 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
372 errstr
= "Journal has aborted";
374 errstr
= "Readonly filesystem";
377 /* If the caller passed in an extra buffer for unknown
378 * errors, textualise them now. Else we just return
381 /* Check for truncated error codes... */
382 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
391 /* __ext4_std_error decodes expected errors from journaling functions
392 * automatically and invokes the appropriate error response. */
394 void __ext4_std_error(struct super_block
*sb
, const char *function
, int errno
)
399 /* Special case: if the error is EROFS, and we're not already
400 * inside a transaction, then there's really no point in logging
402 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
403 (sb
->s_flags
& MS_RDONLY
))
406 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
407 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s: %s\n",
408 sb
->s_id
, function
, errstr
);
410 ext4_handle_error(sb
);
414 * ext4_abort is a much stronger failure handler than ext4_error. The
415 * abort function may be used to deal with unrecoverable failures such
416 * as journal IO errors or ENOMEM at a critical moment in log management.
418 * We unconditionally force the filesystem into an ABORT|READONLY state,
419 * unless the error response on the fs has been set to panic in which
420 * case we take the easy way out and panic immediately.
423 void ext4_abort(struct super_block
*sb
, const char *function
,
424 const char *fmt
, ...)
429 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
434 if (test_opt(sb
, ERRORS_PANIC
))
435 panic("EXT4-fs panic from previous error\n");
437 if (sb
->s_flags
& MS_RDONLY
)
440 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
441 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
442 sb
->s_flags
|= MS_RDONLY
;
443 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
444 if (EXT4_SB(sb
)->s_journal
)
445 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
448 void ext4_msg (struct super_block
* sb
, const char *prefix
,
449 const char *fmt
, ...)
454 printk("%sEXT4-fs (%s): ", prefix
, sb
->s_id
);
460 void ext4_warning(struct super_block
*sb
, const char *function
,
461 const char *fmt
, ...)
466 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s: ",
473 void ext4_grp_locked_error(struct super_block
*sb
, ext4_group_t grp
,
474 const char *function
, const char *fmt
, ...)
479 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
482 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
487 if (test_opt(sb
, ERRORS_CONT
)) {
488 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
489 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
490 ext4_commit_super(sb
, 0);
493 ext4_unlock_group(sb
, grp
);
494 ext4_handle_error(sb
);
496 * We only get here in the ERRORS_RO case; relocking the group
497 * may be dangerous, but nothing bad will happen since the
498 * filesystem will have already been marked read/only and the
499 * journal has been aborted. We return 1 as a hint to callers
500 * who might what to use the return value from
501 * ext4_grp_locked_error() to distinguish beween the
502 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
503 * aggressively from the ext4 function in question, with a
504 * more appropriate error code.
506 ext4_lock_group(sb
, grp
);
510 void ext4_update_dynamic_rev(struct super_block
*sb
)
512 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
514 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
517 ext4_warning(sb
, __func__
,
518 "updating to rev %d because of new feature flag, "
519 "running e2fsck is recommended",
522 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
523 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
524 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
525 /* leave es->s_feature_*compat flags alone */
526 /* es->s_uuid will be set by e2fsck if empty */
529 * The rest of the superblock fields should be zero, and if not it
530 * means they are likely already in use, so leave them alone. We
531 * can leave it up to e2fsck to clean up any inconsistencies there.
536 * Open the external journal device
538 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
540 struct block_device
*bdev
;
541 char b
[BDEVNAME_SIZE
];
543 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
549 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
550 __bdevname(dev
, b
), PTR_ERR(bdev
));
555 * Release the journal device
557 static int ext4_blkdev_put(struct block_device
*bdev
)
560 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
563 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
565 struct block_device
*bdev
;
568 bdev
= sbi
->journal_bdev
;
570 ret
= ext4_blkdev_put(bdev
);
571 sbi
->journal_bdev
= NULL
;
576 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
578 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
581 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
585 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
586 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
588 printk(KERN_ERR
"sb_info orphan list:\n");
589 list_for_each(l
, &sbi
->s_orphan
) {
590 struct inode
*inode
= orphan_list_entry(l
);
592 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
593 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
594 inode
->i_mode
, inode
->i_nlink
,
599 static void ext4_put_super(struct super_block
*sb
)
601 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
602 struct ext4_super_block
*es
= sbi
->s_es
;
605 flush_workqueue(sbi
->dio_unwritten_wq
);
606 destroy_workqueue(sbi
->dio_unwritten_wq
);
611 ext4_commit_super(sb
, 1);
613 ext4_release_system_zone(sb
);
615 ext4_ext_release(sb
);
616 ext4_xattr_put_super(sb
);
617 if (sbi
->s_journal
) {
618 err
= jbd2_journal_destroy(sbi
->s_journal
);
619 sbi
->s_journal
= NULL
;
621 ext4_abort(sb
, __func__
,
622 "Couldn't clean up the journal");
624 if (!(sb
->s_flags
& MS_RDONLY
)) {
625 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
626 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
627 ext4_commit_super(sb
, 1);
630 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
632 kobject_del(&sbi
->s_kobj
);
634 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
635 brelse(sbi
->s_group_desc
[i
]);
636 kfree(sbi
->s_group_desc
);
637 if (is_vmalloc_addr(sbi
->s_flex_groups
))
638 vfree(sbi
->s_flex_groups
);
640 kfree(sbi
->s_flex_groups
);
641 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
642 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
643 percpu_counter_destroy(&sbi
->s_dirs_counter
);
644 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
647 for (i
= 0; i
< MAXQUOTAS
; i
++)
648 kfree(sbi
->s_qf_names
[i
]);
651 /* Debugging code just in case the in-memory inode orphan list
652 * isn't empty. The on-disk one can be non-empty if we've
653 * detected an error and taken the fs readonly, but the
654 * in-memory list had better be clean by this point. */
655 if (!list_empty(&sbi
->s_orphan
))
656 dump_orphan_list(sb
, sbi
);
657 J_ASSERT(list_empty(&sbi
->s_orphan
));
659 invalidate_bdev(sb
->s_bdev
);
660 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
662 * Invalidate the journal device's buffers. We don't want them
663 * floating about in memory - the physical journal device may
664 * hotswapped, and it breaks the `ro-after' testing code.
666 sync_blockdev(sbi
->journal_bdev
);
667 invalidate_bdev(sbi
->journal_bdev
);
668 ext4_blkdev_remove(sbi
);
670 sb
->s_fs_info
= NULL
;
672 * Now that we are completely done shutting down the
673 * superblock, we need to actually destroy the kobject.
677 kobject_put(&sbi
->s_kobj
);
678 wait_for_completion(&sbi
->s_kobj_unregister
);
679 kfree(sbi
->s_blockgroup_lock
);
683 static struct kmem_cache
*ext4_inode_cachep
;
686 * Called inside transaction, so use GFP_NOFS
688 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
690 struct ext4_inode_info
*ei
;
692 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
696 ei
->vfs_inode
.i_version
= 1;
697 ei
->vfs_inode
.i_data
.writeback_index
= 0;
698 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
699 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
700 spin_lock_init(&ei
->i_prealloc_lock
);
702 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
703 * therefore it can be null here. Don't check it, just initialize
706 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
707 ei
->i_reserved_data_blocks
= 0;
708 ei
->i_reserved_meta_blocks
= 0;
709 ei
->i_allocated_meta_blocks
= 0;
710 ei
->i_delalloc_reserved_flag
= 0;
711 spin_lock_init(&(ei
->i_block_reservation_lock
));
712 INIT_LIST_HEAD(&ei
->i_aio_dio_complete_list
);
713 ei
->cur_aio_dio
= NULL
;
715 return &ei
->vfs_inode
;
718 static void ext4_destroy_inode(struct inode
*inode
)
720 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
721 ext4_msg(inode
->i_sb
, KERN_ERR
,
722 "Inode %lu (%p): orphan list check failed!",
723 inode
->i_ino
, EXT4_I(inode
));
724 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
725 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
729 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
732 static void init_once(void *foo
)
734 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
736 INIT_LIST_HEAD(&ei
->i_orphan
);
737 #ifdef CONFIG_EXT4_FS_XATTR
738 init_rwsem(&ei
->xattr_sem
);
740 init_rwsem(&ei
->i_data_sem
);
741 inode_init_once(&ei
->vfs_inode
);
744 static int init_inodecache(void)
746 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
747 sizeof(struct ext4_inode_info
),
748 0, (SLAB_RECLAIM_ACCOUNT
|
751 if (ext4_inode_cachep
== NULL
)
756 static void destroy_inodecache(void)
758 kmem_cache_destroy(ext4_inode_cachep
);
761 static void ext4_clear_inode(struct inode
*inode
)
763 ext4_discard_preallocations(inode
);
764 if (EXT4_JOURNAL(inode
))
765 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
766 &EXT4_I(inode
)->jinode
);
769 static inline void ext4_show_quota_options(struct seq_file
*seq
,
770 struct super_block
*sb
)
772 #if defined(CONFIG_QUOTA)
773 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
775 if (sbi
->s_jquota_fmt
)
776 seq_printf(seq
, ",jqfmt=%s",
777 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold" : "vfsv0");
779 if (sbi
->s_qf_names
[USRQUOTA
])
780 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
782 if (sbi
->s_qf_names
[GRPQUOTA
])
783 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
785 if (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
)
786 seq_puts(seq
, ",usrquota");
788 if (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)
789 seq_puts(seq
, ",grpquota");
795 * - it's set to a non-default value OR
796 * - if the per-sb default is different from the global default
798 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
801 unsigned long def_mount_opts
;
802 struct super_block
*sb
= vfs
->mnt_sb
;
803 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
804 struct ext4_super_block
*es
= sbi
->s_es
;
806 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
807 def_errors
= le16_to_cpu(es
->s_errors
);
809 if (sbi
->s_sb_block
!= 1)
810 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
811 if (test_opt(sb
, MINIX_DF
))
812 seq_puts(seq
, ",minixdf");
813 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
814 seq_puts(seq
, ",grpid");
815 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
816 seq_puts(seq
, ",nogrpid");
817 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
818 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
819 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
821 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
822 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
823 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
825 if (test_opt(sb
, ERRORS_RO
)) {
826 if (def_errors
== EXT4_ERRORS_PANIC
||
827 def_errors
== EXT4_ERRORS_CONTINUE
) {
828 seq_puts(seq
, ",errors=remount-ro");
831 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
832 seq_puts(seq
, ",errors=continue");
833 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
834 seq_puts(seq
, ",errors=panic");
835 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
836 seq_puts(seq
, ",nouid32");
837 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
838 seq_puts(seq
, ",debug");
839 if (test_opt(sb
, OLDALLOC
))
840 seq_puts(seq
, ",oldalloc");
841 #ifdef CONFIG_EXT4_FS_XATTR
842 if (test_opt(sb
, XATTR_USER
) &&
843 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
844 seq_puts(seq
, ",user_xattr");
845 if (!test_opt(sb
, XATTR_USER
) &&
846 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
847 seq_puts(seq
, ",nouser_xattr");
850 #ifdef CONFIG_EXT4_FS_POSIX_ACL
851 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
852 seq_puts(seq
, ",acl");
853 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
854 seq_puts(seq
, ",noacl");
856 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
857 seq_printf(seq
, ",commit=%u",
858 (unsigned) (sbi
->s_commit_interval
/ HZ
));
860 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
861 seq_printf(seq
, ",min_batch_time=%u",
862 (unsigned) sbi
->s_min_batch_time
);
864 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
865 seq_printf(seq
, ",max_batch_time=%u",
866 (unsigned) sbi
->s_min_batch_time
);
870 * We're changing the default of barrier mount option, so
871 * let's always display its mount state so it's clear what its
874 seq_puts(seq
, ",barrier=");
875 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
876 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
877 seq_puts(seq
, ",journal_async_commit");
878 if (test_opt(sb
, NOBH
))
879 seq_puts(seq
, ",nobh");
880 if (test_opt(sb
, I_VERSION
))
881 seq_puts(seq
, ",i_version");
882 if (!test_opt(sb
, DELALLOC
))
883 seq_puts(seq
, ",nodelalloc");
887 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
889 * journal mode get enabled in different ways
890 * So just print the value even if we didn't specify it
892 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
893 seq_puts(seq
, ",data=journal");
894 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
895 seq_puts(seq
, ",data=ordered");
896 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
897 seq_puts(seq
, ",data=writeback");
899 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
900 seq_printf(seq
, ",inode_readahead_blks=%u",
901 sbi
->s_inode_readahead_blks
);
903 if (test_opt(sb
, DATA_ERR_ABORT
))
904 seq_puts(seq
, ",data_err=abort");
906 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
907 seq_puts(seq
, ",noauto_da_alloc");
909 ext4_show_quota_options(seq
, sb
);
914 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
915 u64 ino
, u32 generation
)
919 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
920 return ERR_PTR(-ESTALE
);
921 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
922 return ERR_PTR(-ESTALE
);
924 /* iget isn't really right if the inode is currently unallocated!!
926 * ext4_read_inode will return a bad_inode if the inode had been
927 * deleted, so we should be safe.
929 * Currently we don't know the generation for parent directory, so
930 * a generation of 0 means "accept any"
932 inode
= ext4_iget(sb
, ino
);
934 return ERR_CAST(inode
);
935 if (generation
&& inode
->i_generation
!= generation
) {
937 return ERR_PTR(-ESTALE
);
943 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
944 int fh_len
, int fh_type
)
946 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
950 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
951 int fh_len
, int fh_type
)
953 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
958 * Try to release metadata pages (indirect blocks, directories) which are
959 * mapped via the block device. Since these pages could have journal heads
960 * which would prevent try_to_free_buffers() from freeing them, we must use
961 * jbd2 layer's try_to_free_buffers() function to release them.
963 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
966 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
968 WARN_ON(PageChecked(page
));
969 if (!page_has_buffers(page
))
972 return jbd2_journal_try_to_free_buffers(journal
, page
,
974 return try_to_free_buffers(page
);
978 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
979 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
981 static int ext4_write_dquot(struct dquot
*dquot
);
982 static int ext4_acquire_dquot(struct dquot
*dquot
);
983 static int ext4_release_dquot(struct dquot
*dquot
);
984 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
985 static int ext4_write_info(struct super_block
*sb
, int type
);
986 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
987 char *path
, int remount
);
988 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
989 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
990 size_t len
, loff_t off
);
991 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
992 const char *data
, size_t len
, loff_t off
);
994 static struct dquot_operations ext4_quota_operations
= {
995 .initialize
= dquot_initialize
,
997 .alloc_space
= dquot_alloc_space
,
998 .reserve_space
= dquot_reserve_space
,
999 .claim_space
= dquot_claim_space
,
1000 .release_rsv
= dquot_release_reserved_space
,
1001 .get_reserved_space
= ext4_get_reserved_space
,
1002 .alloc_inode
= dquot_alloc_inode
,
1003 .free_space
= dquot_free_space
,
1004 .free_inode
= dquot_free_inode
,
1005 .transfer
= dquot_transfer
,
1006 .write_dquot
= ext4_write_dquot
,
1007 .acquire_dquot
= ext4_acquire_dquot
,
1008 .release_dquot
= ext4_release_dquot
,
1009 .mark_dirty
= ext4_mark_dquot_dirty
,
1010 .write_info
= ext4_write_info
,
1011 .alloc_dquot
= dquot_alloc
,
1012 .destroy_dquot
= dquot_destroy
,
1015 static struct quotactl_ops ext4_qctl_operations
= {
1016 .quota_on
= ext4_quota_on
,
1017 .quota_off
= vfs_quota_off
,
1018 .quota_sync
= vfs_quota_sync
,
1019 .get_info
= vfs_get_dqinfo
,
1020 .set_info
= vfs_set_dqinfo
,
1021 .get_dqblk
= vfs_get_dqblk
,
1022 .set_dqblk
= vfs_set_dqblk
1026 static const struct super_operations ext4_sops
= {
1027 .alloc_inode
= ext4_alloc_inode
,
1028 .destroy_inode
= ext4_destroy_inode
,
1029 .write_inode
= ext4_write_inode
,
1030 .dirty_inode
= ext4_dirty_inode
,
1031 .delete_inode
= ext4_delete_inode
,
1032 .put_super
= ext4_put_super
,
1033 .sync_fs
= ext4_sync_fs
,
1034 .freeze_fs
= ext4_freeze
,
1035 .unfreeze_fs
= ext4_unfreeze
,
1036 .statfs
= ext4_statfs
,
1037 .remount_fs
= ext4_remount
,
1038 .clear_inode
= ext4_clear_inode
,
1039 .show_options
= ext4_show_options
,
1041 .quota_read
= ext4_quota_read
,
1042 .quota_write
= ext4_quota_write
,
1044 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1047 static const struct super_operations ext4_nojournal_sops
= {
1048 .alloc_inode
= ext4_alloc_inode
,
1049 .destroy_inode
= ext4_destroy_inode
,
1050 .write_inode
= ext4_write_inode
,
1051 .dirty_inode
= ext4_dirty_inode
,
1052 .delete_inode
= ext4_delete_inode
,
1053 .write_super
= ext4_write_super
,
1054 .put_super
= ext4_put_super
,
1055 .statfs
= ext4_statfs
,
1056 .remount_fs
= ext4_remount
,
1057 .clear_inode
= ext4_clear_inode
,
1058 .show_options
= ext4_show_options
,
1060 .quota_read
= ext4_quota_read
,
1061 .quota_write
= ext4_quota_write
,
1063 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1066 static const struct export_operations ext4_export_ops
= {
1067 .fh_to_dentry
= ext4_fh_to_dentry
,
1068 .fh_to_parent
= ext4_fh_to_parent
,
1069 .get_parent
= ext4_get_parent
,
1073 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1074 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1075 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1076 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1077 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1078 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1079 Opt_journal_update
, Opt_journal_dev
,
1080 Opt_journal_checksum
, Opt_journal_async_commit
,
1081 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1082 Opt_data_err_abort
, Opt_data_err_ignore
, Opt_mb_history_length
,
1083 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1084 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
1085 Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
, Opt_resize
,
1086 Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1087 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
1088 Opt_block_validity
, Opt_noblock_validity
,
1089 Opt_inode_readahead_blks
, Opt_journal_ioprio
1092 static const match_table_t tokens
= {
1093 {Opt_bsd_df
, "bsddf"},
1094 {Opt_minix_df
, "minixdf"},
1095 {Opt_grpid
, "grpid"},
1096 {Opt_grpid
, "bsdgroups"},
1097 {Opt_nogrpid
, "nogrpid"},
1098 {Opt_nogrpid
, "sysvgroups"},
1099 {Opt_resgid
, "resgid=%u"},
1100 {Opt_resuid
, "resuid=%u"},
1102 {Opt_err_cont
, "errors=continue"},
1103 {Opt_err_panic
, "errors=panic"},
1104 {Opt_err_ro
, "errors=remount-ro"},
1105 {Opt_nouid32
, "nouid32"},
1106 {Opt_debug
, "debug"},
1107 {Opt_oldalloc
, "oldalloc"},
1108 {Opt_orlov
, "orlov"},
1109 {Opt_user_xattr
, "user_xattr"},
1110 {Opt_nouser_xattr
, "nouser_xattr"},
1112 {Opt_noacl
, "noacl"},
1113 {Opt_noload
, "noload"},
1116 {Opt_commit
, "commit=%u"},
1117 {Opt_min_batch_time
, "min_batch_time=%u"},
1118 {Opt_max_batch_time
, "max_batch_time=%u"},
1119 {Opt_journal_update
, "journal=update"},
1120 {Opt_journal_dev
, "journal_dev=%u"},
1121 {Opt_journal_checksum
, "journal_checksum"},
1122 {Opt_journal_async_commit
, "journal_async_commit"},
1123 {Opt_abort
, "abort"},
1124 {Opt_data_journal
, "data=journal"},
1125 {Opt_data_ordered
, "data=ordered"},
1126 {Opt_data_writeback
, "data=writeback"},
1127 {Opt_data_err_abort
, "data_err=abort"},
1128 {Opt_data_err_ignore
, "data_err=ignore"},
1129 {Opt_mb_history_length
, "mb_history_length=%u"},
1130 {Opt_offusrjquota
, "usrjquota="},
1131 {Opt_usrjquota
, "usrjquota=%s"},
1132 {Opt_offgrpjquota
, "grpjquota="},
1133 {Opt_grpjquota
, "grpjquota=%s"},
1134 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1135 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1136 {Opt_grpquota
, "grpquota"},
1137 {Opt_noquota
, "noquota"},
1138 {Opt_quota
, "quota"},
1139 {Opt_usrquota
, "usrquota"},
1140 {Opt_barrier
, "barrier=%u"},
1141 {Opt_barrier
, "barrier"},
1142 {Opt_nobarrier
, "nobarrier"},
1143 {Opt_i_version
, "i_version"},
1144 {Opt_stripe
, "stripe=%u"},
1145 {Opt_resize
, "resize"},
1146 {Opt_delalloc
, "delalloc"},
1147 {Opt_nodelalloc
, "nodelalloc"},
1148 {Opt_block_validity
, "block_validity"},
1149 {Opt_noblock_validity
, "noblock_validity"},
1150 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1151 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1152 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1153 {Opt_auto_da_alloc
, "auto_da_alloc"},
1154 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1158 static ext4_fsblk_t
get_sb_block(void **data
)
1160 ext4_fsblk_t sb_block
;
1161 char *options
= (char *) *data
;
1163 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1164 return 1; /* Default location */
1167 /* TODO: use simple_strtoll with >32bit ext4 */
1168 sb_block
= simple_strtoul(options
, &options
, 0);
1169 if (*options
&& *options
!= ',') {
1170 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1174 if (*options
== ',')
1176 *data
= (void *) options
;
1181 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1183 static int parse_options(char *options
, struct super_block
*sb
,
1184 unsigned long *journal_devnum
,
1185 unsigned int *journal_ioprio
,
1186 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1188 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1190 substring_t args
[MAX_OPT_ARGS
];
1201 while ((p
= strsep(&options
, ",")) != NULL
) {
1206 token
= match_token(p
, tokens
, args
);
1209 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1212 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1215 set_opt(sbi
->s_mount_opt
, GRPID
);
1218 clear_opt(sbi
->s_mount_opt
, GRPID
);
1221 if (match_int(&args
[0], &option
))
1223 sbi
->s_resuid
= option
;
1226 if (match_int(&args
[0], &option
))
1228 sbi
->s_resgid
= option
;
1231 /* handled by get_sb_block() instead of here */
1232 /* *sb_block = match_int(&args[0]); */
1235 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1236 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1237 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1240 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1241 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1242 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1245 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1246 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1247 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1250 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1253 set_opt(sbi
->s_mount_opt
, DEBUG
);
1256 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1259 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1261 #ifdef CONFIG_EXT4_FS_XATTR
1262 case Opt_user_xattr
:
1263 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1265 case Opt_nouser_xattr
:
1266 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1269 case Opt_user_xattr
:
1270 case Opt_nouser_xattr
:
1271 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1274 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1276 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1279 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1284 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1287 case Opt_journal_update
:
1289 /* Eventually we will want to be able to create
1290 a journal file here. For now, only allow the
1291 user to specify an existing inode to be the
1294 ext4_msg(sb
, KERN_ERR
,
1295 "Cannot specify journal on remount");
1298 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1300 case Opt_journal_dev
:
1302 ext4_msg(sb
, KERN_ERR
,
1303 "Cannot specify journal on remount");
1306 if (match_int(&args
[0], &option
))
1308 *journal_devnum
= option
;
1310 case Opt_journal_checksum
:
1311 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1313 case Opt_journal_async_commit
:
1314 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1315 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1318 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1321 if (match_int(&args
[0], &option
))
1326 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1327 sbi
->s_commit_interval
= HZ
* option
;
1329 case Opt_max_batch_time
:
1330 if (match_int(&args
[0], &option
))
1335 option
= EXT4_DEF_MAX_BATCH_TIME
;
1336 sbi
->s_max_batch_time
= option
;
1338 case Opt_min_batch_time
:
1339 if (match_int(&args
[0], &option
))
1343 sbi
->s_min_batch_time
= option
;
1345 case Opt_data_journal
:
1346 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1348 case Opt_data_ordered
:
1349 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1351 case Opt_data_writeback
:
1352 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1355 if ((sbi
->s_mount_opt
& EXT4_MOUNT_DATA_FLAGS
)
1357 ext4_msg(sb
, KERN_ERR
,
1358 "Cannot change data mode on remount");
1362 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DATA_FLAGS
;
1363 sbi
->s_mount_opt
|= data_opt
;
1366 case Opt_data_err_abort
:
1367 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1369 case Opt_data_err_ignore
:
1370 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1372 case Opt_mb_history_length
:
1373 if (match_int(&args
[0], &option
))
1377 sbi
->s_mb_history_max
= option
;
1386 if (sb_any_quota_loaded(sb
) &&
1387 !sbi
->s_qf_names
[qtype
]) {
1388 ext4_msg(sb
, KERN_ERR
,
1389 "Cannot change journaled "
1390 "quota options when quota turned on");
1393 qname
= match_strdup(&args
[0]);
1395 ext4_msg(sb
, KERN_ERR
,
1396 "Not enough memory for "
1397 "storing quotafile name");
1400 if (sbi
->s_qf_names
[qtype
] &&
1401 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1402 ext4_msg(sb
, KERN_ERR
,
1403 "%s quota file already "
1404 "specified", QTYPE2NAME(qtype
));
1408 sbi
->s_qf_names
[qtype
] = qname
;
1409 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1410 ext4_msg(sb
, KERN_ERR
,
1411 "quotafile must be on "
1413 kfree(sbi
->s_qf_names
[qtype
]);
1414 sbi
->s_qf_names
[qtype
] = NULL
;
1417 set_opt(sbi
->s_mount_opt
, QUOTA
);
1419 case Opt_offusrjquota
:
1422 case Opt_offgrpjquota
:
1425 if (sb_any_quota_loaded(sb
) &&
1426 sbi
->s_qf_names
[qtype
]) {
1427 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1428 "journaled quota options when "
1433 * The space will be released later when all options
1434 * are confirmed to be correct
1436 sbi
->s_qf_names
[qtype
] = NULL
;
1438 case Opt_jqfmt_vfsold
:
1439 qfmt
= QFMT_VFS_OLD
;
1441 case Opt_jqfmt_vfsv0
:
1444 if (sb_any_quota_loaded(sb
) &&
1445 sbi
->s_jquota_fmt
!= qfmt
) {
1446 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1447 "journaled quota options when "
1451 sbi
->s_jquota_fmt
= qfmt
;
1455 set_opt(sbi
->s_mount_opt
, QUOTA
);
1456 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1459 set_opt(sbi
->s_mount_opt
, QUOTA
);
1460 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1463 if (sb_any_quota_loaded(sb
)) {
1464 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1465 "options when quota turned on");
1468 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1469 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1470 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1476 ext4_msg(sb
, KERN_ERR
,
1477 "quota options not supported");
1481 case Opt_offusrjquota
:
1482 case Opt_offgrpjquota
:
1483 case Opt_jqfmt_vfsold
:
1484 case Opt_jqfmt_vfsv0
:
1485 ext4_msg(sb
, KERN_ERR
,
1486 "journaled quota options not supported");
1492 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1495 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1498 if (match_int(&args
[0], &option
)) {
1499 set_opt(sbi
->s_mount_opt
, BARRIER
);
1503 set_opt(sbi
->s_mount_opt
, BARRIER
);
1505 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1511 ext4_msg(sb
, KERN_ERR
,
1512 "resize option only available "
1516 if (match_int(&args
[0], &option
) != 0)
1518 *n_blocks_count
= option
;
1521 set_opt(sbi
->s_mount_opt
, NOBH
);
1524 clear_opt(sbi
->s_mount_opt
, NOBH
);
1527 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1528 sb
->s_flags
|= MS_I_VERSION
;
1530 case Opt_nodelalloc
:
1531 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1534 if (match_int(&args
[0], &option
))
1538 sbi
->s_stripe
= option
;
1541 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1543 case Opt_block_validity
:
1544 set_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1546 case Opt_noblock_validity
:
1547 clear_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1549 case Opt_inode_readahead_blks
:
1550 if (match_int(&args
[0], &option
))
1552 if (option
< 0 || option
> (1 << 30))
1554 if (!is_power_of_2(option
)) {
1555 ext4_msg(sb
, KERN_ERR
,
1556 "EXT4-fs: inode_readahead_blks"
1557 " must be a power of 2");
1560 sbi
->s_inode_readahead_blks
= option
;
1562 case Opt_journal_ioprio
:
1563 if (match_int(&args
[0], &option
))
1565 if (option
< 0 || option
> 7)
1567 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1570 case Opt_noauto_da_alloc
:
1571 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1573 case Opt_auto_da_alloc
:
1574 if (match_int(&args
[0], &option
)) {
1575 clear_opt(sbi
->s_mount_opt
, NO_AUTO_DA_ALLOC
);
1579 clear_opt(sbi
->s_mount_opt
, NO_AUTO_DA_ALLOC
);
1581 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1584 ext4_msg(sb
, KERN_ERR
,
1585 "Unrecognized mount option \"%s\" "
1586 "or missing value", p
);
1591 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1592 if ((sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
) &&
1593 sbi
->s_qf_names
[USRQUOTA
])
1594 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1596 if ((sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
) &&
1597 sbi
->s_qf_names
[GRPQUOTA
])
1598 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1600 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1601 (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)) ||
1602 (sbi
->s_qf_names
[GRPQUOTA
] &&
1603 (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
))) {
1604 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1609 if (!sbi
->s_jquota_fmt
) {
1610 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1615 if (sbi
->s_jquota_fmt
) {
1616 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1617 "specified with no journaling "
1626 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1629 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1632 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1633 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1634 "forcing read-only mode");
1639 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1640 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1641 "running e2fsck is recommended");
1642 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1643 ext4_msg(sb
, KERN_WARNING
,
1644 "warning: mounting fs with errors, "
1645 "running e2fsck is recommended");
1646 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1647 le16_to_cpu(es
->s_mnt_count
) >=
1648 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1649 ext4_msg(sb
, KERN_WARNING
,
1650 "warning: maximal mount count reached, "
1651 "running e2fsck is recommended");
1652 else if (le32_to_cpu(es
->s_checkinterval
) &&
1653 (le32_to_cpu(es
->s_lastcheck
) +
1654 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1655 ext4_msg(sb
, KERN_WARNING
,
1656 "warning: checktime reached, "
1657 "running e2fsck is recommended");
1658 if (!sbi
->s_journal
)
1659 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1660 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1661 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1662 le16_add_cpu(&es
->s_mnt_count
, 1);
1663 es
->s_mtime
= cpu_to_le32(get_seconds());
1664 ext4_update_dynamic_rev(sb
);
1666 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1668 ext4_commit_super(sb
, 1);
1669 if (test_opt(sb
, DEBUG
))
1670 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1671 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1673 sbi
->s_groups_count
,
1674 EXT4_BLOCKS_PER_GROUP(sb
),
1675 EXT4_INODES_PER_GROUP(sb
),
1678 if (EXT4_SB(sb
)->s_journal
) {
1679 ext4_msg(sb
, KERN_INFO
, "%s journal on %s",
1680 EXT4_SB(sb
)->s_journal
->j_inode
? "internal" :
1681 "external", EXT4_SB(sb
)->s_journal
->j_devname
);
1683 ext4_msg(sb
, KERN_INFO
, "no journal");
1688 static int ext4_fill_flex_info(struct super_block
*sb
)
1690 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1691 struct ext4_group_desc
*gdp
= NULL
;
1692 ext4_group_t flex_group_count
;
1693 ext4_group_t flex_group
;
1694 int groups_per_flex
= 0;
1698 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1699 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1701 if (groups_per_flex
< 2) {
1702 sbi
->s_log_groups_per_flex
= 0;
1706 /* We allocate both existing and potentially added groups */
1707 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1708 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1709 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1710 size
= flex_group_count
* sizeof(struct flex_groups
);
1711 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1712 if (sbi
->s_flex_groups
== NULL
) {
1713 sbi
->s_flex_groups
= vmalloc(size
);
1714 if (sbi
->s_flex_groups
)
1715 memset(sbi
->s_flex_groups
, 0, size
);
1717 if (sbi
->s_flex_groups
== NULL
) {
1718 ext4_msg(sb
, KERN_ERR
, "not enough memory for "
1719 "%u flex groups", flex_group_count
);
1723 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1724 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1726 flex_group
= ext4_flex_group(sbi
, i
);
1727 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1728 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1729 atomic_add(ext4_free_blks_count(sb
, gdp
),
1730 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1731 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1732 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1740 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1741 struct ext4_group_desc
*gdp
)
1745 if (sbi
->s_es
->s_feature_ro_compat
&
1746 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1747 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1748 __le32 le_group
= cpu_to_le32(block_group
);
1750 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1751 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1752 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1753 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1754 /* for checksum of struct ext4_group_desc do the rest...*/
1755 if ((sbi
->s_es
->s_feature_incompat
&
1756 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1757 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1758 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1759 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1763 return cpu_to_le16(crc
);
1766 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1767 struct ext4_group_desc
*gdp
)
1769 if ((sbi
->s_es
->s_feature_ro_compat
&
1770 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1771 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1777 /* Called at mount-time, super-block is locked */
1778 static int ext4_check_descriptors(struct super_block
*sb
)
1780 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1781 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1782 ext4_fsblk_t last_block
;
1783 ext4_fsblk_t block_bitmap
;
1784 ext4_fsblk_t inode_bitmap
;
1785 ext4_fsblk_t inode_table
;
1786 int flexbg_flag
= 0;
1789 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1792 ext4_debug("Checking group descriptors");
1794 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1795 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1797 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1798 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1800 last_block
= first_block
+
1801 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1803 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1804 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1805 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1806 "Block bitmap for group %u not in group "
1807 "(block %llu)!", i
, block_bitmap
);
1810 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1811 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1812 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1813 "Inode bitmap for group %u not in group "
1814 "(block %llu)!", i
, inode_bitmap
);
1817 inode_table
= ext4_inode_table(sb
, gdp
);
1818 if (inode_table
< first_block
||
1819 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1820 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1821 "Inode table for group %u not in group "
1822 "(block %llu)!", i
, inode_table
);
1825 ext4_lock_group(sb
, i
);
1826 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1827 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1828 "Checksum for group %u failed (%u!=%u)",
1829 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1830 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1831 if (!(sb
->s_flags
& MS_RDONLY
)) {
1832 ext4_unlock_group(sb
, i
);
1836 ext4_unlock_group(sb
, i
);
1838 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1841 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1842 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
1846 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1847 * the superblock) which were deleted from all directories, but held open by
1848 * a process at the time of a crash. We walk the list and try to delete these
1849 * inodes at recovery time (only with a read-write filesystem).
1851 * In order to keep the orphan inode chain consistent during traversal (in
1852 * case of crash during recovery), we link each inode into the superblock
1853 * orphan list_head and handle it the same way as an inode deletion during
1854 * normal operation (which journals the operations for us).
1856 * We only do an iget() and an iput() on each inode, which is very safe if we
1857 * accidentally point at an in-use or already deleted inode. The worst that
1858 * can happen in this case is that we get a "bit already cleared" message from
1859 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1860 * e2fsck was run on this filesystem, and it must have already done the orphan
1861 * inode cleanup for us, so we can safely abort without any further action.
1863 static void ext4_orphan_cleanup(struct super_block
*sb
,
1864 struct ext4_super_block
*es
)
1866 unsigned int s_flags
= sb
->s_flags
;
1867 int nr_orphans
= 0, nr_truncates
= 0;
1871 if (!es
->s_last_orphan
) {
1872 jbd_debug(4, "no orphan inodes to clean up\n");
1876 if (bdev_read_only(sb
->s_bdev
)) {
1877 ext4_msg(sb
, KERN_ERR
, "write access "
1878 "unavailable, skipping orphan cleanup");
1882 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1883 if (es
->s_last_orphan
)
1884 jbd_debug(1, "Errors on filesystem, "
1885 "clearing orphan list.\n");
1886 es
->s_last_orphan
= 0;
1887 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1891 if (s_flags
& MS_RDONLY
) {
1892 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
1893 sb
->s_flags
&= ~MS_RDONLY
;
1896 /* Needed for iput() to work correctly and not trash data */
1897 sb
->s_flags
|= MS_ACTIVE
;
1898 /* Turn on quotas so that they are updated correctly */
1899 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1900 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
1901 int ret
= ext4_quota_on_mount(sb
, i
);
1903 ext4_msg(sb
, KERN_ERR
,
1904 "Cannot turn on journaled "
1905 "quota: error %d", ret
);
1910 while (es
->s_last_orphan
) {
1911 struct inode
*inode
;
1913 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1914 if (IS_ERR(inode
)) {
1915 es
->s_last_orphan
= 0;
1919 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
1921 if (inode
->i_nlink
) {
1922 ext4_msg(sb
, KERN_DEBUG
,
1923 "%s: truncating inode %lu to %lld bytes",
1924 __func__
, inode
->i_ino
, inode
->i_size
);
1925 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1926 inode
->i_ino
, inode
->i_size
);
1927 ext4_truncate(inode
);
1930 ext4_msg(sb
, KERN_DEBUG
,
1931 "%s: deleting unreferenced inode %lu",
1932 __func__
, inode
->i_ino
);
1933 jbd_debug(2, "deleting unreferenced inode %lu\n",
1937 iput(inode
); /* The delete magic happens here! */
1940 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1943 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
1944 PLURAL(nr_orphans
));
1946 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
1947 PLURAL(nr_truncates
));
1949 /* Turn quotas off */
1950 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1951 if (sb_dqopt(sb
)->files
[i
])
1952 vfs_quota_off(sb
, i
, 0);
1955 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1959 * Maximal extent format file size.
1960 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1961 * extent format containers, within a sector_t, and within i_blocks
1962 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1963 * so that won't be a limiting factor.
1965 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1967 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
1970 loff_t upper_limit
= MAX_LFS_FILESIZE
;
1972 /* small i_blocks in vfs inode? */
1973 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
1975 * CONFIG_LBDAF is not enabled implies the inode
1976 * i_block represent total blocks in 512 bytes
1977 * 32 == size of vfs inode i_blocks * 8
1979 upper_limit
= (1LL << 32) - 1;
1981 /* total blocks in file system block size */
1982 upper_limit
>>= (blkbits
- 9);
1983 upper_limit
<<= blkbits
;
1986 /* 32-bit extent-start container, ee_block */
1991 /* Sanity check against vm- & vfs- imposed limits */
1992 if (res
> upper_limit
)
1999 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2000 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2001 * We need to be 1 filesystem block less than the 2^48 sector limit.
2003 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2005 loff_t res
= EXT4_NDIR_BLOCKS
;
2008 /* This is calculated to be the largest file size for a dense, block
2009 * mapped file such that the file's total number of 512-byte sectors,
2010 * including data and all indirect blocks, does not exceed (2^48 - 1).
2012 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2013 * number of 512-byte sectors of the file.
2016 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2018 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2019 * the inode i_block field represents total file blocks in
2020 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2022 upper_limit
= (1LL << 32) - 1;
2024 /* total blocks in file system block size */
2025 upper_limit
>>= (bits
- 9);
2029 * We use 48 bit ext4_inode i_blocks
2030 * With EXT4_HUGE_FILE_FL set the i_blocks
2031 * represent total number of blocks in
2032 * file system block size
2034 upper_limit
= (1LL << 48) - 1;
2038 /* indirect blocks */
2040 /* double indirect blocks */
2041 meta_blocks
+= 1 + (1LL << (bits
-2));
2042 /* tripple indirect blocks */
2043 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2045 upper_limit
-= meta_blocks
;
2046 upper_limit
<<= bits
;
2048 res
+= 1LL << (bits
-2);
2049 res
+= 1LL << (2*(bits
-2));
2050 res
+= 1LL << (3*(bits
-2));
2052 if (res
> upper_limit
)
2055 if (res
> MAX_LFS_FILESIZE
)
2056 res
= MAX_LFS_FILESIZE
;
2061 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2062 ext4_fsblk_t logical_sb_block
, int nr
)
2064 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2065 ext4_group_t bg
, first_meta_bg
;
2068 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2070 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2072 return logical_sb_block
+ nr
+ 1;
2073 bg
= sbi
->s_desc_per_block
* nr
;
2074 if (ext4_bg_has_super(sb
, bg
))
2077 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2081 * ext4_get_stripe_size: Get the stripe size.
2082 * @sbi: In memory super block info
2084 * If we have specified it via mount option, then
2085 * use the mount option value. If the value specified at mount time is
2086 * greater than the blocks per group use the super block value.
2087 * If the super block value is greater than blocks per group return 0.
2088 * Allocator needs it be less than blocks per group.
2091 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2093 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2094 unsigned long stripe_width
=
2095 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2097 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2098 return sbi
->s_stripe
;
2100 if (stripe_width
<= sbi
->s_blocks_per_group
)
2101 return stripe_width
;
2103 if (stride
<= sbi
->s_blocks_per_group
)
2112 struct attribute attr
;
2113 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2114 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2115 const char *, size_t);
2119 static int parse_strtoul(const char *buf
,
2120 unsigned long max
, unsigned long *value
)
2124 while (*buf
&& isspace(*buf
))
2126 *value
= simple_strtoul(buf
, &endp
, 0);
2127 while (*endp
&& isspace(*endp
))
2129 if (*endp
|| *value
> max
)
2135 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2136 struct ext4_sb_info
*sbi
,
2139 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2140 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2143 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2144 struct ext4_sb_info
*sbi
, char *buf
)
2146 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2148 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2149 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2150 sbi
->s_sectors_written_start
) >> 1);
2153 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2154 struct ext4_sb_info
*sbi
, char *buf
)
2156 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2158 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2159 sbi
->s_kbytes_written
+
2160 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2161 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
2164 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2165 struct ext4_sb_info
*sbi
,
2166 const char *buf
, size_t count
)
2170 if (parse_strtoul(buf
, 0x40000000, &t
))
2173 if (!is_power_of_2(t
))
2176 sbi
->s_inode_readahead_blks
= t
;
2180 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2181 struct ext4_sb_info
*sbi
, char *buf
)
2183 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2185 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2188 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2189 struct ext4_sb_info
*sbi
,
2190 const char *buf
, size_t count
)
2192 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2195 if (parse_strtoul(buf
, 0xffffffff, &t
))
2201 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2202 static struct ext4_attr ext4_attr_##_name = { \
2203 .attr = {.name = __stringify(_name), .mode = _mode }, \
2206 .offset = offsetof(struct ext4_sb_info, _elname), \
2208 #define EXT4_ATTR(name, mode, show, store) \
2209 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2211 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2212 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2213 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2214 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2215 #define ATTR_LIST(name) &ext4_attr_##name.attr
2217 EXT4_RO_ATTR(delayed_allocation_blocks
);
2218 EXT4_RO_ATTR(session_write_kbytes
);
2219 EXT4_RO_ATTR(lifetime_write_kbytes
);
2220 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2221 inode_readahead_blks_store
, s_inode_readahead_blks
);
2222 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2223 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2224 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2225 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2226 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2227 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2228 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2229 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2231 static struct attribute
*ext4_attrs
[] = {
2232 ATTR_LIST(delayed_allocation_blocks
),
2233 ATTR_LIST(session_write_kbytes
),
2234 ATTR_LIST(lifetime_write_kbytes
),
2235 ATTR_LIST(inode_readahead_blks
),
2236 ATTR_LIST(inode_goal
),
2237 ATTR_LIST(mb_stats
),
2238 ATTR_LIST(mb_max_to_scan
),
2239 ATTR_LIST(mb_min_to_scan
),
2240 ATTR_LIST(mb_order2_req
),
2241 ATTR_LIST(mb_stream_req
),
2242 ATTR_LIST(mb_group_prealloc
),
2243 ATTR_LIST(max_writeback_mb_bump
),
2247 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2248 struct attribute
*attr
, char *buf
)
2250 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2252 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2254 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2257 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2258 struct attribute
*attr
,
2259 const char *buf
, size_t len
)
2261 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2263 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2265 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2268 static void ext4_sb_release(struct kobject
*kobj
)
2270 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2272 complete(&sbi
->s_kobj_unregister
);
2276 static struct sysfs_ops ext4_attr_ops
= {
2277 .show
= ext4_attr_show
,
2278 .store
= ext4_attr_store
,
2281 static struct kobj_type ext4_ktype
= {
2282 .default_attrs
= ext4_attrs
,
2283 .sysfs_ops
= &ext4_attr_ops
,
2284 .release
= ext4_sb_release
,
2288 * Check whether this filesystem can be mounted based on
2289 * the features present and the RDONLY/RDWR mount requested.
2290 * Returns 1 if this filesystem can be mounted as requested,
2291 * 0 if it cannot be.
2293 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2295 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2296 ext4_msg(sb
, KERN_ERR
,
2297 "Couldn't mount because of "
2298 "unsupported optional features (%x)",
2299 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2300 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2307 /* Check that feature set is OK for a read-write mount */
2308 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2309 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2310 "unsupported optional features (%x)",
2311 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2312 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2316 * Large file size enabled file system can only be mounted
2317 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2319 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2320 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2321 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2322 "cannot be mounted RDWR without "
2330 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
2331 __releases(kernel_lock
)
2332 __acquires(kernel_lock
)
2334 struct buffer_head
*bh
;
2335 struct ext4_super_block
*es
= NULL
;
2336 struct ext4_sb_info
*sbi
;
2338 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2339 ext4_fsblk_t logical_sb_block
;
2340 unsigned long offset
= 0;
2341 unsigned long journal_devnum
= 0;
2342 unsigned long def_mount_opts
;
2348 unsigned int db_count
;
2350 int needs_recovery
, has_huge_files
;
2353 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2355 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
2359 sbi
->s_blockgroup_lock
=
2360 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
2361 if (!sbi
->s_blockgroup_lock
) {
2365 sb
->s_fs_info
= sbi
;
2366 sbi
->s_mount_opt
= 0;
2367 sbi
->s_resuid
= EXT4_DEF_RESUID
;
2368 sbi
->s_resgid
= EXT4_DEF_RESGID
;
2369 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
2370 sbi
->s_sb_block
= sb_block
;
2371 sbi
->s_sectors_written_start
= part_stat_read(sb
->s_bdev
->bd_part
,
2376 /* Cleanup superblock name */
2377 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
2380 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
2382 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
2387 * The ext4 superblock will not be buffer aligned for other than 1kB
2388 * block sizes. We need to calculate the offset from buffer start.
2390 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
2391 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2392 offset
= do_div(logical_sb_block
, blocksize
);
2394 logical_sb_block
= sb_block
;
2397 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
2398 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
2402 * Note: s_es must be initialized as soon as possible because
2403 * some ext4 macro-instructions depend on its value
2405 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2407 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
2408 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
2410 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
2412 /* Set defaults before we parse the mount options */
2413 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
2414 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
2415 set_opt(sbi
->s_mount_opt
, DEBUG
);
2416 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
2417 set_opt(sbi
->s_mount_opt
, GRPID
);
2418 if (def_mount_opts
& EXT4_DEFM_UID16
)
2419 set_opt(sbi
->s_mount_opt
, NO_UID32
);
2420 #ifdef CONFIG_EXT4_FS_XATTR
2421 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
2422 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
2424 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2425 if (def_mount_opts
& EXT4_DEFM_ACL
)
2426 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
2428 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
2429 sbi
->s_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
2430 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
2431 sbi
->s_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
2432 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
2433 sbi
->s_mount_opt
|= EXT4_MOUNT_WRITEBACK_DATA
;
2435 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2436 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2437 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2438 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2440 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2442 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2443 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2444 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
2445 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
2446 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
2447 sbi
->s_mb_history_max
= default_mb_history_length
;
2449 set_opt(sbi
->s_mount_opt
, BARRIER
);
2452 * enable delayed allocation by default
2453 * Use -o nodelalloc to turn it off
2455 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2457 if (!parse_options((char *) data
, sb
, &journal_devnum
,
2458 &journal_ioprio
, NULL
, 0))
2461 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2462 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2464 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2465 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2466 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2467 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2468 ext4_msg(sb
, KERN_WARNING
,
2469 "feature flags set on rev 0 fs, "
2470 "running e2fsck is recommended");
2473 * Check feature flags regardless of the revision level, since we
2474 * previously didn't change the revision level when setting the flags,
2475 * so there is a chance incompat flags are set on a rev 0 filesystem.
2477 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
2480 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2482 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2483 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2484 ext4_msg(sb
, KERN_ERR
,
2485 "Unsupported filesystem blocksize %d", blocksize
);
2489 if (sb
->s_blocksize
!= blocksize
) {
2490 /* Validate the filesystem blocksize */
2491 if (!sb_set_blocksize(sb
, blocksize
)) {
2492 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
2498 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2499 offset
= do_div(logical_sb_block
, blocksize
);
2500 bh
= sb_bread(sb
, logical_sb_block
);
2502 ext4_msg(sb
, KERN_ERR
,
2503 "Can't read superblock on 2nd try");
2506 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2508 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2509 ext4_msg(sb
, KERN_ERR
,
2510 "Magic mismatch, very weird!");
2515 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2516 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
2517 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
2519 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
2521 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2522 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2523 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2525 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2526 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2527 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2528 (!is_power_of_2(sbi
->s_inode_size
)) ||
2529 (sbi
->s_inode_size
> blocksize
)) {
2530 ext4_msg(sb
, KERN_ERR
,
2531 "unsupported inode size: %d",
2535 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2536 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2539 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2540 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2541 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2542 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2543 !is_power_of_2(sbi
->s_desc_size
)) {
2544 ext4_msg(sb
, KERN_ERR
,
2545 "unsupported descriptor size %lu",
2550 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2552 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2553 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2554 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2557 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2558 if (sbi
->s_inodes_per_block
== 0)
2560 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2561 sbi
->s_inodes_per_block
;
2562 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2564 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2565 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2566 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2568 for (i
= 0; i
< 4; i
++)
2569 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2570 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2571 i
= le32_to_cpu(es
->s_flags
);
2572 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
2573 sbi
->s_hash_unsigned
= 3;
2574 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
2575 #ifdef __CHAR_UNSIGNED__
2576 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
2577 sbi
->s_hash_unsigned
= 3;
2579 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
2584 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2585 ext4_msg(sb
, KERN_ERR
,
2586 "#blocks per group too big: %lu",
2587 sbi
->s_blocks_per_group
);
2590 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2591 ext4_msg(sb
, KERN_ERR
,
2592 "#inodes per group too big: %lu",
2593 sbi
->s_inodes_per_group
);
2598 * Test whether we have more sectors than will fit in sector_t,
2599 * and whether the max offset is addressable by the page cache.
2601 if ((ext4_blocks_count(es
) >
2602 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) ||
2603 (ext4_blocks_count(es
) >
2604 (pgoff_t
)(~0ULL) >> (PAGE_CACHE_SHIFT
- sb
->s_blocksize_bits
))) {
2605 ext4_msg(sb
, KERN_ERR
, "filesystem"
2606 " too large to mount safely on this system");
2607 if (sizeof(sector_t
) < 8)
2608 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
2613 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2616 /* check blocks count against device size */
2617 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
2618 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
2619 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
2620 "exceeds size of device (%llu blocks)",
2621 ext4_blocks_count(es
), blocks_count
);
2626 * It makes no sense for the first data block to be beyond the end
2627 * of the filesystem.
2629 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
2630 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
2631 "block %u is beyond end of filesystem (%llu)",
2632 le32_to_cpu(es
->s_first_data_block
),
2633 ext4_blocks_count(es
));
2636 blocks_count
= (ext4_blocks_count(es
) -
2637 le32_to_cpu(es
->s_first_data_block
) +
2638 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2639 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2640 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
2641 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
2642 "(block count %llu, first data block %u, "
2643 "blocks per group %lu)", sbi
->s_groups_count
,
2644 ext4_blocks_count(es
),
2645 le32_to_cpu(es
->s_first_data_block
),
2646 EXT4_BLOCKS_PER_GROUP(sb
));
2649 sbi
->s_groups_count
= blocks_count
;
2650 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
2651 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
2652 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2653 EXT4_DESC_PER_BLOCK(sb
);
2654 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2656 if (sbi
->s_group_desc
== NULL
) {
2657 ext4_msg(sb
, KERN_ERR
, "not enough memory");
2661 #ifdef CONFIG_PROC_FS
2663 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2666 bgl_lock_init(sbi
->s_blockgroup_lock
);
2668 for (i
= 0; i
< db_count
; i
++) {
2669 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2670 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2671 if (!sbi
->s_group_desc
[i
]) {
2672 ext4_msg(sb
, KERN_ERR
,
2673 "can't read group descriptor %d", i
);
2678 if (!ext4_check_descriptors(sb
)) {
2679 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
2682 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2683 if (!ext4_fill_flex_info(sb
)) {
2684 ext4_msg(sb
, KERN_ERR
,
2685 "unable to initialize "
2686 "flex_bg meta info!");
2690 sbi
->s_gdb_count
= db_count
;
2691 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2692 spin_lock_init(&sbi
->s_next_gen_lock
);
2694 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2695 ext4_count_free_blocks(sb
));
2697 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2698 ext4_count_free_inodes(sb
));
2701 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2702 ext4_count_dirs(sb
));
2705 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
2708 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
2712 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2713 sbi
->s_max_writeback_mb_bump
= 128;
2716 * set up enough so that it can read an inode
2718 if (!test_opt(sb
, NOLOAD
) &&
2719 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
2720 sb
->s_op
= &ext4_sops
;
2722 sb
->s_op
= &ext4_nojournal_sops
;
2723 sb
->s_export_op
= &ext4_export_ops
;
2724 sb
->s_xattr
= ext4_xattr_handlers
;
2726 sb
->s_qcop
= &ext4_qctl_operations
;
2727 sb
->dq_op
= &ext4_quota_operations
;
2729 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2730 mutex_init(&sbi
->s_orphan_lock
);
2731 mutex_init(&sbi
->s_resize_lock
);
2735 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2736 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2737 EXT4_FEATURE_INCOMPAT_RECOVER
));
2740 * The first inode we look at is the journal inode. Don't try
2741 * root first: it may be modified in the journal!
2743 if (!test_opt(sb
, NOLOAD
) &&
2744 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2745 if (ext4_load_journal(sb
, es
, journal_devnum
))
2747 if (!(sb
->s_flags
& MS_RDONLY
) &&
2748 EXT4_SB(sb
)->s_journal
->j_failed_commit
) {
2749 ext4_msg(sb
, KERN_CRIT
, "error: "
2750 "ext4_fill_super: Journal transaction "
2752 EXT4_SB(sb
)->s_journal
->j_failed_commit
);
2753 if (test_opt(sb
, ERRORS_RO
)) {
2754 ext4_msg(sb
, KERN_CRIT
,
2755 "Mounting filesystem read-only");
2756 sb
->s_flags
|= MS_RDONLY
;
2757 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2758 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2760 if (test_opt(sb
, ERRORS_PANIC
)) {
2761 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2762 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2763 ext4_commit_super(sb
, 1);
2767 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
2768 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2769 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
2770 "suppressed and not mounted read-only");
2773 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
2774 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2775 sbi
->s_journal
= NULL
;
2780 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2781 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2782 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2783 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
2787 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2788 jbd2_journal_set_features(sbi
->s_journal
,
2789 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2790 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2791 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2792 jbd2_journal_set_features(sbi
->s_journal
,
2793 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2794 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2795 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2797 jbd2_journal_clear_features(sbi
->s_journal
,
2798 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2799 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2802 /* We have now updated the journal if required, so we can
2803 * validate the data journaling mode. */
2804 switch (test_opt(sb
, DATA_FLAGS
)) {
2806 /* No mode set, assume a default based on the journal
2807 * capabilities: ORDERED_DATA if the journal can
2808 * cope, else JOURNAL_DATA
2810 if (jbd2_journal_check_available_features
2811 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2812 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2814 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2817 case EXT4_MOUNT_ORDERED_DATA
:
2818 case EXT4_MOUNT_WRITEBACK_DATA
:
2819 if (!jbd2_journal_check_available_features
2820 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2821 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
2822 "requested data journaling mode");
2828 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
2832 if (test_opt(sb
, NOBH
)) {
2833 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2834 ext4_msg(sb
, KERN_WARNING
, "Ignoring nobh option - "
2835 "its supported only with writeback mode");
2836 clear_opt(sbi
->s_mount_opt
, NOBH
);
2839 EXT4_SB(sb
)->dio_unwritten_wq
= create_workqueue("ext4-dio-unwritten");
2840 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
2841 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
2842 goto failed_mount_wq
;
2846 * The jbd2_journal_load will have done any necessary log recovery,
2847 * so we can safely mount the rest of the filesystem now.
2850 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2852 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
2853 ret
= PTR_ERR(root
);
2856 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2858 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
2861 sb
->s_root
= d_alloc_root(root
);
2863 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
2869 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
2871 /* determine the minimum size of new large inodes, if present */
2872 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2873 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2874 EXT4_GOOD_OLD_INODE_SIZE
;
2875 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2876 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2877 if (sbi
->s_want_extra_isize
<
2878 le16_to_cpu(es
->s_want_extra_isize
))
2879 sbi
->s_want_extra_isize
=
2880 le16_to_cpu(es
->s_want_extra_isize
);
2881 if (sbi
->s_want_extra_isize
<
2882 le16_to_cpu(es
->s_min_extra_isize
))
2883 sbi
->s_want_extra_isize
=
2884 le16_to_cpu(es
->s_min_extra_isize
);
2887 /* Check if enough inode space is available */
2888 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2889 sbi
->s_inode_size
) {
2890 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2891 EXT4_GOOD_OLD_INODE_SIZE
;
2892 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
2896 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
2897 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
2898 "requested data journaling mode");
2899 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2900 } else if (test_opt(sb
, DELALLOC
))
2901 ext4_msg(sb
, KERN_INFO
, "delayed allocation enabled");
2903 err
= ext4_setup_system_zone(sb
);
2905 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
2906 "zone (%d)\n", err
);
2911 err
= ext4_mb_init(sb
, needs_recovery
);
2913 ext4_msg(sb
, KERN_ERR
, "failed to initalize mballoc (%d)",
2918 sbi
->s_kobj
.kset
= ext4_kset
;
2919 init_completion(&sbi
->s_kobj_unregister
);
2920 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
2923 ext4_mb_release(sb
);
2924 ext4_ext_release(sb
);
2928 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
2929 ext4_orphan_cleanup(sb
, es
);
2930 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
2931 if (needs_recovery
) {
2932 ext4_msg(sb
, KERN_INFO
, "recovery complete");
2933 ext4_mark_recovery_complete(sb
, es
);
2935 if (EXT4_SB(sb
)->s_journal
) {
2936 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2937 descr
= " journalled data mode";
2938 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2939 descr
= " ordered data mode";
2941 descr
= " writeback data mode";
2943 descr
= "out journal";
2945 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s", descr
);
2952 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
2956 ext4_msg(sb
, KERN_ERR
, "mount failed");
2957 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
2959 ext4_release_system_zone(sb
);
2960 if (sbi
->s_journal
) {
2961 jbd2_journal_destroy(sbi
->s_journal
);
2962 sbi
->s_journal
= NULL
;
2965 if (sbi
->s_flex_groups
) {
2966 if (is_vmalloc_addr(sbi
->s_flex_groups
))
2967 vfree(sbi
->s_flex_groups
);
2969 kfree(sbi
->s_flex_groups
);
2971 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2972 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2973 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2974 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
2976 for (i
= 0; i
< db_count
; i
++)
2977 brelse(sbi
->s_group_desc
[i
]);
2978 kfree(sbi
->s_group_desc
);
2981 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
2984 for (i
= 0; i
< MAXQUOTAS
; i
++)
2985 kfree(sbi
->s_qf_names
[i
]);
2987 ext4_blkdev_remove(sbi
);
2990 sb
->s_fs_info
= NULL
;
2991 kfree(sbi
->s_blockgroup_lock
);
2998 * Setup any per-fs journal parameters now. We'll do this both on
2999 * initial mount, once the journal has been initialised but before we've
3000 * done any recovery; and again on any subsequent remount.
3002 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3004 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3006 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3007 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3008 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3010 spin_lock(&journal
->j_state_lock
);
3011 if (test_opt(sb
, BARRIER
))
3012 journal
->j_flags
|= JBD2_BARRIER
;
3014 journal
->j_flags
&= ~JBD2_BARRIER
;
3015 if (test_opt(sb
, DATA_ERR_ABORT
))
3016 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3018 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3019 spin_unlock(&journal
->j_state_lock
);
3022 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3023 unsigned int journal_inum
)
3025 struct inode
*journal_inode
;
3028 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3030 /* First, test for the existence of a valid inode on disk. Bad
3031 * things happen if we iget() an unused inode, as the subsequent
3032 * iput() will try to delete it. */
3034 journal_inode
= ext4_iget(sb
, journal_inum
);
3035 if (IS_ERR(journal_inode
)) {
3036 ext4_msg(sb
, KERN_ERR
, "no journal found");
3039 if (!journal_inode
->i_nlink
) {
3040 make_bad_inode(journal_inode
);
3041 iput(journal_inode
);
3042 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3046 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3047 journal_inode
, journal_inode
->i_size
);
3048 if (!S_ISREG(journal_inode
->i_mode
)) {
3049 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3050 iput(journal_inode
);
3054 journal
= jbd2_journal_init_inode(journal_inode
);
3056 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3057 iput(journal_inode
);
3060 journal
->j_private
= sb
;
3061 ext4_init_journal_params(sb
, journal
);
3065 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3068 struct buffer_head
*bh
;
3072 int hblock
, blocksize
;
3073 ext4_fsblk_t sb_block
;
3074 unsigned long offset
;
3075 struct ext4_super_block
*es
;
3076 struct block_device
*bdev
;
3078 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3080 bdev
= ext4_blkdev_get(j_dev
, sb
);
3084 if (bd_claim(bdev
, sb
)) {
3085 ext4_msg(sb
, KERN_ERR
,
3086 "failed to claim external journal device");
3087 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
3091 blocksize
= sb
->s_blocksize
;
3092 hblock
= bdev_logical_block_size(bdev
);
3093 if (blocksize
< hblock
) {
3094 ext4_msg(sb
, KERN_ERR
,
3095 "blocksize too small for journal device");
3099 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3100 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3101 set_blocksize(bdev
, blocksize
);
3102 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3103 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3104 "external journal");
3108 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3109 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3110 !(le32_to_cpu(es
->s_feature_incompat
) &
3111 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3112 ext4_msg(sb
, KERN_ERR
, "external journal has "
3118 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3119 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3124 len
= ext4_blocks_count(es
);
3125 start
= sb_block
+ 1;
3126 brelse(bh
); /* we're done with the superblock */
3128 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3129 start
, len
, blocksize
);
3131 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3134 journal
->j_private
= sb
;
3135 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3136 wait_on_buffer(journal
->j_sb_buffer
);
3137 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3138 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3141 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3142 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3143 "user (unsupported) - %d",
3144 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3147 EXT4_SB(sb
)->journal_bdev
= bdev
;
3148 ext4_init_journal_params(sb
, journal
);
3152 jbd2_journal_destroy(journal
);
3154 ext4_blkdev_put(bdev
);
3158 static int ext4_load_journal(struct super_block
*sb
,
3159 struct ext4_super_block
*es
,
3160 unsigned long journal_devnum
)
3163 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3166 int really_read_only
;
3168 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3170 if (journal_devnum
&&
3171 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3172 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3173 "numbers have changed");
3174 journal_dev
= new_decode_dev(journal_devnum
);
3176 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3178 really_read_only
= bdev_read_only(sb
->s_bdev
);
3181 * Are we loading a blank journal or performing recovery after a
3182 * crash? For recovery, we need to check in advance whether we
3183 * can get read-write access to the device.
3185 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3186 if (sb
->s_flags
& MS_RDONLY
) {
3187 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3188 "required on readonly filesystem");
3189 if (really_read_only
) {
3190 ext4_msg(sb
, KERN_ERR
, "write access "
3191 "unavailable, cannot proceed");
3194 ext4_msg(sb
, KERN_INFO
, "write access will "
3195 "be enabled during recovery");
3199 if (journal_inum
&& journal_dev
) {
3200 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3201 "and inode journals!");
3206 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3209 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3213 if (journal
->j_flags
& JBD2_BARRIER
)
3214 ext4_msg(sb
, KERN_INFO
, "barriers enabled");
3216 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3218 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
3219 err
= jbd2_journal_update_format(journal
);
3221 ext4_msg(sb
, KERN_ERR
, "error updating journal");
3222 jbd2_journal_destroy(journal
);
3227 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
3228 err
= jbd2_journal_wipe(journal
, !really_read_only
);
3230 err
= jbd2_journal_load(journal
);
3233 ext4_msg(sb
, KERN_ERR
, "error loading journal");
3234 jbd2_journal_destroy(journal
);
3238 EXT4_SB(sb
)->s_journal
= journal
;
3239 ext4_clear_journal_err(sb
, es
);
3241 if (journal_devnum
&&
3242 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3243 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
3245 /* Make sure we flush the recovery flag to disk. */
3246 ext4_commit_super(sb
, 1);
3252 static int ext4_commit_super(struct super_block
*sb
, int sync
)
3254 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
3255 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
3260 if (buffer_write_io_error(sbh
)) {
3262 * Oh, dear. A previous attempt to write the
3263 * superblock failed. This could happen because the
3264 * USB device was yanked out. Or it could happen to
3265 * be a transient write error and maybe the block will
3266 * be remapped. Nothing we can do but to retry the
3267 * write and hope for the best.
3269 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
3270 "superblock detected");
3271 clear_buffer_write_io_error(sbh
);
3272 set_buffer_uptodate(sbh
);
3275 * If the file system is mounted read-only, don't update the
3276 * superblock write time. This avoids updating the superblock
3277 * write time when we are mounting the root file system
3278 * read/only but we need to replay the journal; at that point,
3279 * for people who are east of GMT and who make their clock
3280 * tick in localtime for Windows bug-for-bug compatibility,
3281 * the clock is set in the future, and this will cause e2fsck
3282 * to complain and force a full file system check.
3284 if (!(sb
->s_flags
& MS_RDONLY
))
3285 es
->s_wtime
= cpu_to_le32(get_seconds());
3286 es
->s_kbytes_written
=
3287 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
3288 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
3289 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
3290 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
3291 &EXT4_SB(sb
)->s_freeblocks_counter
));
3292 es
->s_free_inodes_count
= cpu_to_le32(percpu_counter_sum_positive(
3293 &EXT4_SB(sb
)->s_freeinodes_counter
));
3295 BUFFER_TRACE(sbh
, "marking dirty");
3296 mark_buffer_dirty(sbh
);
3298 error
= sync_dirty_buffer(sbh
);
3302 error
= buffer_write_io_error(sbh
);
3304 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
3306 clear_buffer_write_io_error(sbh
);
3307 set_buffer_uptodate(sbh
);
3314 * Have we just finished recovery? If so, and if we are mounting (or
3315 * remounting) the filesystem readonly, then we will end up with a
3316 * consistent fs on disk. Record that fact.
3318 static void ext4_mark_recovery_complete(struct super_block
*sb
,
3319 struct ext4_super_block
*es
)
3321 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
3323 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3324 BUG_ON(journal
!= NULL
);
3327 jbd2_journal_lock_updates(journal
);
3328 if (jbd2_journal_flush(journal
) < 0)
3331 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
3332 sb
->s_flags
& MS_RDONLY
) {
3333 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3334 ext4_commit_super(sb
, 1);
3338 jbd2_journal_unlock_updates(journal
);
3342 * If we are mounting (or read-write remounting) a filesystem whose journal
3343 * has recorded an error from a previous lifetime, move that error to the
3344 * main filesystem now.
3346 static void ext4_clear_journal_err(struct super_block
*sb
,
3347 struct ext4_super_block
*es
)
3353 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3355 journal
= EXT4_SB(sb
)->s_journal
;
3358 * Now check for any error status which may have been recorded in the
3359 * journal by a prior ext4_error() or ext4_abort()
3362 j_errno
= jbd2_journal_errno(journal
);
3366 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
3367 ext4_warning(sb
, __func__
, "Filesystem error recorded "
3368 "from previous mount: %s", errstr
);
3369 ext4_warning(sb
, __func__
, "Marking fs in need of "
3370 "filesystem check.");
3372 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
3373 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
3374 ext4_commit_super(sb
, 1);
3376 jbd2_journal_clear_err(journal
);
3381 * Force the running and committing transactions to commit,
3382 * and wait on the commit.
3384 int ext4_force_commit(struct super_block
*sb
)
3389 if (sb
->s_flags
& MS_RDONLY
)
3392 journal
= EXT4_SB(sb
)->s_journal
;
3394 ret
= ext4_journal_force_commit(journal
);
3399 static void ext4_write_super(struct super_block
*sb
)
3402 ext4_commit_super(sb
, 1);
3406 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
3410 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3412 trace_ext4_sync_fs(sb
, wait
);
3413 flush_workqueue(sbi
->dio_unwritten_wq
);
3414 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
3416 jbd2_log_wait_commit(sbi
->s_journal
, target
);
3422 * LVM calls this function before a (read-only) snapshot is created. This
3423 * gives us a chance to flush the journal completely and mark the fs clean.
3425 static int ext4_freeze(struct super_block
*sb
)
3430 if (sb
->s_flags
& MS_RDONLY
)
3433 journal
= EXT4_SB(sb
)->s_journal
;
3435 /* Now we set up the journal barrier. */
3436 jbd2_journal_lock_updates(journal
);
3439 * Don't clear the needs_recovery flag if we failed to flush
3442 error
= jbd2_journal_flush(journal
);
3445 jbd2_journal_unlock_updates(journal
);
3449 /* Journal blocked and flushed, clear needs_recovery flag. */
3450 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3451 error
= ext4_commit_super(sb
, 1);
3458 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3459 * flag here, even though the filesystem is not technically dirty yet.
3461 static int ext4_unfreeze(struct super_block
*sb
)
3463 if (sb
->s_flags
& MS_RDONLY
)
3467 /* Reset the needs_recovery flag before the fs is unlocked. */
3468 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3469 ext4_commit_super(sb
, 1);
3471 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3475 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3477 struct ext4_super_block
*es
;
3478 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3479 ext4_fsblk_t n_blocks_count
= 0;
3480 unsigned long old_sb_flags
;
3481 struct ext4_mount_options old_opts
;
3483 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3491 /* Store the original options */
3493 old_sb_flags
= sb
->s_flags
;
3494 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3495 old_opts
.s_resuid
= sbi
->s_resuid
;
3496 old_opts
.s_resgid
= sbi
->s_resgid
;
3497 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3498 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
3499 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
3501 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3502 for (i
= 0; i
< MAXQUOTAS
; i
++)
3503 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3505 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
3506 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
3509 * Allow the "check" option to be passed as a remount option.
3511 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
3512 &n_blocks_count
, 1)) {
3517 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
3518 ext4_abort(sb
, __func__
, "Abort forced by user");
3520 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3521 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
3525 if (sbi
->s_journal
) {
3526 ext4_init_journal_params(sb
, sbi
->s_journal
);
3527 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3530 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3531 n_blocks_count
> ext4_blocks_count(es
)) {
3532 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
3537 if (*flags
& MS_RDONLY
) {
3539 * First of all, the unconditional stuff we have to do
3540 * to disable replay of the journal when we next remount
3542 sb
->s_flags
|= MS_RDONLY
;
3545 * OK, test if we are remounting a valid rw partition
3546 * readonly, and if so set the rdonly flag and then
3547 * mark the partition as valid again.
3549 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3550 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3551 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3554 ext4_mark_recovery_complete(sb
, es
);
3556 /* Make sure we can mount this feature set readwrite */
3557 if (!ext4_feature_set_ok(sb
, 0)) {
3562 * Make sure the group descriptor checksums
3563 * are sane. If they aren't, refuse to remount r/w.
3565 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3566 struct ext4_group_desc
*gdp
=
3567 ext4_get_group_desc(sb
, g
, NULL
);
3569 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3570 ext4_msg(sb
, KERN_ERR
,
3571 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3572 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3573 le16_to_cpu(gdp
->bg_checksum
));
3580 * If we have an unprocessed orphan list hanging
3581 * around from a previously readonly bdev mount,
3582 * require a full umount/remount for now.
3584 if (es
->s_last_orphan
) {
3585 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
3586 "remount RDWR because of unprocessed "
3587 "orphan inode list. Please "
3588 "umount/remount instead");
3594 * Mounting a RDONLY partition read-write, so reread
3595 * and store the current valid flag. (It may have
3596 * been changed by e2fsck since we originally mounted
3600 ext4_clear_journal_err(sb
, es
);
3601 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3602 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3604 if (!ext4_setup_super(sb
, es
, 0))
3605 sb
->s_flags
&= ~MS_RDONLY
;
3608 ext4_setup_system_zone(sb
);
3609 if (sbi
->s_journal
== NULL
)
3610 ext4_commit_super(sb
, 1);
3613 /* Release old quota file names */
3614 for (i
= 0; i
< MAXQUOTAS
; i
++)
3615 if (old_opts
.s_qf_names
[i
] &&
3616 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3617 kfree(old_opts
.s_qf_names
[i
]);
3624 sb
->s_flags
= old_sb_flags
;
3625 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3626 sbi
->s_resuid
= old_opts
.s_resuid
;
3627 sbi
->s_resgid
= old_opts
.s_resgid
;
3628 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3629 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
3630 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
3632 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3633 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3634 if (sbi
->s_qf_names
[i
] &&
3635 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3636 kfree(sbi
->s_qf_names
[i
]);
3637 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3645 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3647 struct super_block
*sb
= dentry
->d_sb
;
3648 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3649 struct ext4_super_block
*es
= sbi
->s_es
;
3652 if (test_opt(sb
, MINIX_DF
)) {
3653 sbi
->s_overhead_last
= 0;
3654 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3655 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3656 ext4_fsblk_t overhead
= 0;
3659 * Compute the overhead (FS structures). This is constant
3660 * for a given filesystem unless the number of block groups
3661 * changes so we cache the previous value until it does.
3665 * All of the blocks before first_data_block are
3668 overhead
= le32_to_cpu(es
->s_first_data_block
);
3671 * Add the overhead attributed to the superblock and
3672 * block group descriptors. If the sparse superblocks
3673 * feature is turned on, then not all groups have this.
3675 for (i
= 0; i
< ngroups
; i
++) {
3676 overhead
+= ext4_bg_has_super(sb
, i
) +
3677 ext4_bg_num_gdb(sb
, i
);
3682 * Every block group has an inode bitmap, a block
3683 * bitmap, and an inode table.
3685 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3686 sbi
->s_overhead_last
= overhead
;
3688 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3691 buf
->f_type
= EXT4_SUPER_MAGIC
;
3692 buf
->f_bsize
= sb
->s_blocksize
;
3693 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3694 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3695 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3696 ext4_free_blocks_count_set(es
, buf
->f_bfree
);
3697 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3698 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3700 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3701 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3702 es
->s_free_inodes_count
= cpu_to_le32(buf
->f_ffree
);
3703 buf
->f_namelen
= EXT4_NAME_LEN
;
3704 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3705 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3706 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3707 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3712 /* Helper function for writing quotas on sync - we need to start transaction
3713 * before quota file is locked for write. Otherwise the are possible deadlocks:
3714 * Process 1 Process 2
3715 * ext4_create() quota_sync()
3716 * jbd2_journal_start() write_dquot()
3717 * vfs_dq_init() down(dqio_mutex)
3718 * down(dqio_mutex) jbd2_journal_start()
3724 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3726 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3729 static int ext4_write_dquot(struct dquot
*dquot
)
3733 struct inode
*inode
;
3735 inode
= dquot_to_inode(dquot
);
3736 handle
= ext4_journal_start(inode
,
3737 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3739 return PTR_ERR(handle
);
3740 ret
= dquot_commit(dquot
);
3741 err
= ext4_journal_stop(handle
);
3747 static int ext4_acquire_dquot(struct dquot
*dquot
)
3752 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3753 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3755 return PTR_ERR(handle
);
3756 ret
= dquot_acquire(dquot
);
3757 err
= ext4_journal_stop(handle
);
3763 static int ext4_release_dquot(struct dquot
*dquot
)
3768 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3769 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3770 if (IS_ERR(handle
)) {
3771 /* Release dquot anyway to avoid endless cycle in dqput() */
3772 dquot_release(dquot
);
3773 return PTR_ERR(handle
);
3775 ret
= dquot_release(dquot
);
3776 err
= ext4_journal_stop(handle
);
3782 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3784 /* Are we journaling quotas? */
3785 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3786 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3787 dquot_mark_dquot_dirty(dquot
);
3788 return ext4_write_dquot(dquot
);
3790 return dquot_mark_dquot_dirty(dquot
);
3794 static int ext4_write_info(struct super_block
*sb
, int type
)
3799 /* Data block + inode block */
3800 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3802 return PTR_ERR(handle
);
3803 ret
= dquot_commit_info(sb
, type
);
3804 err
= ext4_journal_stop(handle
);
3811 * Turn on quotas during mount time - we need to find
3812 * the quota file and such...
3814 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3816 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3817 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3821 * Standard function to be called on quota_on
3823 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3824 char *name
, int remount
)
3829 if (!test_opt(sb
, QUOTA
))
3831 /* When remounting, no checks are needed and in fact, name is NULL */
3833 return vfs_quota_on(sb
, type
, format_id
, name
, remount
);
3835 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
3839 /* Quotafile not on the same filesystem? */
3840 if (path
.mnt
->mnt_sb
!= sb
) {
3844 /* Journaling quota? */
3845 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3846 /* Quotafile not in fs root? */
3847 if (path
.dentry
->d_parent
!= sb
->s_root
)
3848 ext4_msg(sb
, KERN_WARNING
,
3849 "Quota file not on filesystem root. "
3850 "Journaled quota will not work");
3854 * When we journal data on quota file, we have to flush journal to see
3855 * all updates to the file when we bypass pagecache...
3857 if (EXT4_SB(sb
)->s_journal
&&
3858 ext4_should_journal_data(path
.dentry
->d_inode
)) {
3860 * We don't need to lock updates but journal_flush() could
3861 * otherwise be livelocked...
3863 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3864 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3865 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3872 err
= vfs_quota_on_path(sb
, type
, format_id
, &path
);
3877 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3878 * acquiring the locks... As quota files are never truncated and quota code
3879 * itself serializes the operations (and noone else should touch the files)
3880 * we don't have to be afraid of races */
3881 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3882 size_t len
, loff_t off
)
3884 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3885 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3887 int offset
= off
& (sb
->s_blocksize
- 1);
3890 struct buffer_head
*bh
;
3891 loff_t i_size
= i_size_read(inode
);
3895 if (off
+len
> i_size
)
3898 while (toread
> 0) {
3899 tocopy
= sb
->s_blocksize
- offset
< toread
?
3900 sb
->s_blocksize
- offset
: toread
;
3901 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3904 if (!bh
) /* A hole? */
3905 memset(data
, 0, tocopy
);
3907 memcpy(data
, bh
->b_data
+offset
, tocopy
);
3917 /* Write to quotafile (we know the transaction is already started and has
3918 * enough credits) */
3919 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
3920 const char *data
, size_t len
, loff_t off
)
3922 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3923 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3925 int offset
= off
& (sb
->s_blocksize
- 1);
3927 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
3928 size_t towrite
= len
;
3929 struct buffer_head
*bh
;
3930 handle_t
*handle
= journal_current_handle();
3932 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
3933 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
3934 " cancelled because transaction is not started",
3935 (unsigned long long)off
, (unsigned long long)len
);
3938 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
3939 while (towrite
> 0) {
3940 tocopy
= sb
->s_blocksize
- offset
< towrite
?
3941 sb
->s_blocksize
- offset
: towrite
;
3942 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
3945 if (journal_quota
) {
3946 err
= ext4_journal_get_write_access(handle
, bh
);
3953 memcpy(bh
->b_data
+offset
, data
, tocopy
);
3954 flush_dcache_page(bh
->b_page
);
3957 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
3959 /* Always do at least ordered writes for quotas */
3960 err
= ext4_jbd2_file_inode(handle
, inode
);
3961 mark_buffer_dirty(bh
);
3972 if (len
== towrite
) {
3973 mutex_unlock(&inode
->i_mutex
);
3976 if (inode
->i_size
< off
+len
-towrite
) {
3977 i_size_write(inode
, off
+len
-towrite
);
3978 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
3980 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3981 ext4_mark_inode_dirty(handle
, inode
);
3982 mutex_unlock(&inode
->i_mutex
);
3983 return len
- towrite
;
3988 static int ext4_get_sb(struct file_system_type
*fs_type
, int flags
,
3989 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3991 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
,mnt
);
3994 static struct file_system_type ext4_fs_type
= {
3995 .owner
= THIS_MODULE
,
3997 .get_sb
= ext4_get_sb
,
3998 .kill_sb
= kill_block_super
,
3999 .fs_flags
= FS_REQUIRES_DEV
,
4002 #ifdef CONFIG_EXT4DEV_COMPAT
4003 static int ext4dev_get_sb(struct file_system_type
*fs_type
, int flags
,
4004 const char *dev_name
, void *data
,struct vfsmount
*mnt
)
4006 printk(KERN_WARNING
"EXT4-fs (%s): Update your userspace programs "
4007 "to mount using ext4\n", dev_name
);
4008 printk(KERN_WARNING
"EXT4-fs (%s): ext4dev backwards compatibility "
4009 "will go away by 2.6.31\n", dev_name
);
4010 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
,mnt
);
4013 static struct file_system_type ext4dev_fs_type
= {
4014 .owner
= THIS_MODULE
,
4016 .get_sb
= ext4dev_get_sb
,
4017 .kill_sb
= kill_block_super
,
4018 .fs_flags
= FS_REQUIRES_DEV
,
4020 MODULE_ALIAS("ext4dev");
4023 static int __init
init_ext4_fs(void)
4027 err
= init_ext4_system_zone();
4030 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4033 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4034 err
= init_ext4_mballoc();
4038 err
= init_ext4_xattr();
4041 err
= init_inodecache();
4044 err
= register_filesystem(&ext4_fs_type
);
4047 #ifdef CONFIG_EXT4DEV_COMPAT
4048 err
= register_filesystem(&ext4dev_fs_type
);
4050 unregister_filesystem(&ext4_fs_type
);
4056 destroy_inodecache();
4060 exit_ext4_mballoc();
4062 remove_proc_entry("fs/ext4", NULL
);
4063 kset_unregister(ext4_kset
);
4065 exit_ext4_system_zone();
4069 static void __exit
exit_ext4_fs(void)
4071 unregister_filesystem(&ext4_fs_type
);
4072 #ifdef CONFIG_EXT4DEV_COMPAT
4073 unregister_filesystem(&ext4dev_fs_type
);
4075 destroy_inodecache();
4077 exit_ext4_mballoc();
4078 remove_proc_entry("fs/ext4", NULL
);
4079 kset_unregister(ext4_kset
);
4080 exit_ext4_system_zone();
4083 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4084 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4085 MODULE_LICENSE("GPL");
4086 module_init(init_ext4_fs
)
4087 module_exit(exit_ext4_fs
)