2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
53 struct proc_dir_entry
*ext4_proc_root
;
54 static struct kset
*ext4_kset
;
56 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
57 unsigned long journal_devnum
);
58 static int ext4_commit_super(struct super_block
*sb
, int sync
);
59 static void ext4_mark_recovery_complete(struct super_block
*sb
,
60 struct ext4_super_block
*es
);
61 static void ext4_clear_journal_err(struct super_block
*sb
,
62 struct ext4_super_block
*es
);
63 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
64 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
66 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
67 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
68 static int ext4_unfreeze(struct super_block
*sb
);
69 static void ext4_write_super(struct super_block
*sb
);
70 static int ext4_freeze(struct super_block
*sb
);
73 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
74 struct ext4_group_desc
*bg
)
76 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
77 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
78 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
81 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
82 struct ext4_group_desc
*bg
)
84 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
85 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
86 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
89 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
90 struct ext4_group_desc
*bg
)
92 return le32_to_cpu(bg
->bg_inode_table_lo
) |
93 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
94 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
97 __u32
ext4_free_blks_count(struct super_block
*sb
,
98 struct ext4_group_desc
*bg
)
100 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
101 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
102 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
105 __u32
ext4_free_inodes_count(struct super_block
*sb
,
106 struct ext4_group_desc
*bg
)
108 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
109 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
110 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
113 __u32
ext4_used_dirs_count(struct super_block
*sb
,
114 struct ext4_group_desc
*bg
)
116 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
117 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
118 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
121 __u32
ext4_itable_unused_count(struct super_block
*sb
,
122 struct ext4_group_desc
*bg
)
124 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
125 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
126 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
129 void ext4_block_bitmap_set(struct super_block
*sb
,
130 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
132 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
133 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
134 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
137 void ext4_inode_bitmap_set(struct super_block
*sb
,
138 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
140 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
141 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
142 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
145 void ext4_inode_table_set(struct super_block
*sb
,
146 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
148 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
149 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
150 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
153 void ext4_free_blks_set(struct super_block
*sb
,
154 struct ext4_group_desc
*bg
, __u32 count
)
156 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
157 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
158 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
161 void ext4_free_inodes_set(struct super_block
*sb
,
162 struct ext4_group_desc
*bg
, __u32 count
)
164 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
165 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
166 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
169 void ext4_used_dirs_set(struct super_block
*sb
,
170 struct ext4_group_desc
*bg
, __u32 count
)
172 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
173 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
174 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
177 void ext4_itable_unused_set(struct super_block
*sb
,
178 struct ext4_group_desc
*bg
, __u32 count
)
180 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
181 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
182 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
186 /* Just increment the non-pointer handle value */
187 static handle_t
*ext4_get_nojournal(void)
189 handle_t
*handle
= current
->journal_info
;
190 unsigned long ref_cnt
= (unsigned long)handle
;
192 BUG_ON(ref_cnt
>= EXT4_NOJOURNAL_MAX_REF_COUNT
);
195 handle
= (handle_t
*)ref_cnt
;
197 current
->journal_info
= handle
;
202 /* Decrement the non-pointer handle value */
203 static void ext4_put_nojournal(handle_t
*handle
)
205 unsigned long ref_cnt
= (unsigned long)handle
;
207 BUG_ON(ref_cnt
== 0);
210 handle
= (handle_t
*)ref_cnt
;
212 current
->journal_info
= handle
;
216 * Wrappers for jbd2_journal_start/end.
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
223 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
227 if (sb
->s_flags
& MS_RDONLY
)
228 return ERR_PTR(-EROFS
);
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
233 journal
= EXT4_SB(sb
)->s_journal
;
235 if (is_journal_aborted(journal
)) {
236 ext4_abort(sb
, __func__
, "Detected aborted journal");
237 return ERR_PTR(-EROFS
);
239 return jbd2_journal_start(journal
, nblocks
);
241 return ext4_get_nojournal();
245 * The only special thing we need to do here is to make sure that all
246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
247 * that sync() will call the filesystem's write_super callback if
250 int __ext4_journal_stop(const char *where
, handle_t
*handle
)
252 struct super_block
*sb
;
256 if (!ext4_handle_valid(handle
)) {
257 ext4_put_nojournal(handle
);
260 sb
= handle
->h_transaction
->t_journal
->j_private
;
262 rc
= jbd2_journal_stop(handle
);
267 __ext4_std_error(sb
, where
, err
);
271 void ext4_journal_abort_handle(const char *caller
, const char *err_fn
,
272 struct buffer_head
*bh
, handle_t
*handle
, int err
)
275 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
277 BUG_ON(!ext4_handle_valid(handle
));
280 BUFFER_TRACE(bh
, "abort");
285 if (is_handle_aborted(handle
))
288 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
289 caller
, errstr
, err_fn
);
291 jbd2_journal_abort_handle(handle
);
294 /* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
297 * On ext2, we can store the error state of the filesystem in the
298 * superblock. That is not possible on ext4, because we may have other
299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
304 * We'll just use the jbd2_journal_abort() error code to record an error in
305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
309 static void ext4_handle_error(struct super_block
*sb
)
311 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
313 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
314 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
316 if (sb
->s_flags
& MS_RDONLY
)
319 if (!test_opt(sb
, ERRORS_CONT
)) {
320 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
322 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
324 jbd2_journal_abort(journal
, -EIO
);
326 if (test_opt(sb
, ERRORS_RO
)) {
327 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
328 sb
->s_flags
|= MS_RDONLY
;
330 ext4_commit_super(sb
, 1);
331 if (test_opt(sb
, ERRORS_PANIC
))
332 panic("EXT4-fs (device %s): panic forced after error\n",
336 void ext4_error(struct super_block
*sb
, const char *function
,
337 const char *fmt
, ...)
342 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
347 ext4_handle_error(sb
);
350 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
357 errstr
= "IO failure";
360 errstr
= "Out of memory";
363 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
364 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
365 errstr
= "Journal has aborted";
367 errstr
= "Readonly filesystem";
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
384 /* __ext4_std_error decodes expected errors from journaling functions
385 * automatically and invokes the appropriate error response. */
387 void __ext4_std_error(struct super_block
*sb
, const char *function
, int errno
)
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
395 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
396 (sb
->s_flags
& MS_RDONLY
))
399 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
400 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s: %s\n",
401 sb
->s_id
, function
, errstr
);
403 ext4_handle_error(sb
);
407 * ext4_abort is a much stronger failure handler than ext4_error. The
408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
416 void ext4_abort(struct super_block
*sb
, const char *function
,
417 const char *fmt
, ...)
422 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
427 if (test_opt(sb
, ERRORS_PANIC
))
428 panic("EXT4-fs panic from previous error\n");
430 if (sb
->s_flags
& MS_RDONLY
)
433 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
434 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
435 sb
->s_flags
|= MS_RDONLY
;
436 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
437 if (EXT4_SB(sb
)->s_journal
)
438 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
441 void ext4_msg (struct super_block
* sb
, const char *prefix
,
442 const char *fmt
, ...)
447 printk("%sEXT4-fs (%s): ", prefix
, sb
->s_id
);
453 void ext4_warning(struct super_block
*sb
, const char *function
,
454 const char *fmt
, ...)
459 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s: ",
466 void ext4_grp_locked_error(struct super_block
*sb
, ext4_group_t grp
,
467 const char *function
, const char *fmt
, ...)
472 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
475 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
480 if (test_opt(sb
, ERRORS_CONT
)) {
481 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
482 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
483 ext4_commit_super(sb
, 0);
486 ext4_unlock_group(sb
, grp
);
487 ext4_handle_error(sb
);
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
499 ext4_lock_group(sb
, grp
);
503 void ext4_update_dynamic_rev(struct super_block
*sb
)
505 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
507 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
510 ext4_warning(sb
, __func__
,
511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
515 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
516 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
517 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
529 * Open the external journal device
531 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
533 struct block_device
*bdev
;
534 char b
[BDEVNAME_SIZE
];
536 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
542 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
543 __bdevname(dev
, b
), PTR_ERR(bdev
));
548 * Release the journal device
550 static int ext4_blkdev_put(struct block_device
*bdev
)
553 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
556 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
558 struct block_device
*bdev
;
561 bdev
= sbi
->journal_bdev
;
563 ret
= ext4_blkdev_put(bdev
);
564 sbi
->journal_bdev
= NULL
;
569 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
571 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
574 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
578 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
579 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
581 printk(KERN_ERR
"sb_info orphan list:\n");
582 list_for_each(l
, &sbi
->s_orphan
) {
583 struct inode
*inode
= orphan_list_entry(l
);
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
587 inode
->i_mode
, inode
->i_nlink
,
592 static void ext4_put_super(struct super_block
*sb
)
594 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
595 struct ext4_super_block
*es
= sbi
->s_es
;
598 flush_workqueue(sbi
->dio_unwritten_wq
);
599 destroy_workqueue(sbi
->dio_unwritten_wq
);
604 ext4_commit_super(sb
, 1);
606 ext4_release_system_zone(sb
);
608 ext4_ext_release(sb
);
609 ext4_xattr_put_super(sb
);
610 if (sbi
->s_journal
) {
611 err
= jbd2_journal_destroy(sbi
->s_journal
);
612 sbi
->s_journal
= NULL
;
614 ext4_abort(sb
, __func__
,
615 "Couldn't clean up the journal");
617 if (!(sb
->s_flags
& MS_RDONLY
)) {
618 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
619 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
620 ext4_commit_super(sb
, 1);
623 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
625 kobject_del(&sbi
->s_kobj
);
627 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
628 brelse(sbi
->s_group_desc
[i
]);
629 kfree(sbi
->s_group_desc
);
630 if (is_vmalloc_addr(sbi
->s_flex_groups
))
631 vfree(sbi
->s_flex_groups
);
633 kfree(sbi
->s_flex_groups
);
634 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
635 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
636 percpu_counter_destroy(&sbi
->s_dirs_counter
);
637 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
640 for (i
= 0; i
< MAXQUOTAS
; i
++)
641 kfree(sbi
->s_qf_names
[i
]);
644 /* Debugging code just in case the in-memory inode orphan list
645 * isn't empty. The on-disk one can be non-empty if we've
646 * detected an error and taken the fs readonly, but the
647 * in-memory list had better be clean by this point. */
648 if (!list_empty(&sbi
->s_orphan
))
649 dump_orphan_list(sb
, sbi
);
650 J_ASSERT(list_empty(&sbi
->s_orphan
));
652 invalidate_bdev(sb
->s_bdev
);
653 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
655 * Invalidate the journal device's buffers. We don't want them
656 * floating about in memory - the physical journal device may
657 * hotswapped, and it breaks the `ro-after' testing code.
659 sync_blockdev(sbi
->journal_bdev
);
660 invalidate_bdev(sbi
->journal_bdev
);
661 ext4_blkdev_remove(sbi
);
663 sb
->s_fs_info
= NULL
;
665 * Now that we are completely done shutting down the
666 * superblock, we need to actually destroy the kobject.
670 kobject_put(&sbi
->s_kobj
);
671 wait_for_completion(&sbi
->s_kobj_unregister
);
672 kfree(sbi
->s_blockgroup_lock
);
676 static struct kmem_cache
*ext4_inode_cachep
;
679 * Called inside transaction, so use GFP_NOFS
681 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
683 struct ext4_inode_info
*ei
;
685 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
689 ei
->vfs_inode
.i_version
= 1;
690 ei
->vfs_inode
.i_data
.writeback_index
= 0;
691 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
692 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
693 spin_lock_init(&ei
->i_prealloc_lock
);
695 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
696 * therefore it can be null here. Don't check it, just initialize
699 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
700 ei
->i_reserved_data_blocks
= 0;
701 ei
->i_reserved_meta_blocks
= 0;
702 ei
->i_allocated_meta_blocks
= 0;
703 ei
->i_delalloc_reserved_flag
= 0;
704 spin_lock_init(&(ei
->i_block_reservation_lock
));
705 INIT_LIST_HEAD(&ei
->i_aio_dio_complete_list
);
706 ei
->cur_aio_dio
= NULL
;
708 return &ei
->vfs_inode
;
711 static void ext4_destroy_inode(struct inode
*inode
)
713 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
714 ext4_msg(inode
->i_sb
, KERN_ERR
,
715 "Inode %lu (%p): orphan list check failed!",
716 inode
->i_ino
, EXT4_I(inode
));
717 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
718 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
722 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
725 static void init_once(void *foo
)
727 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
729 INIT_LIST_HEAD(&ei
->i_orphan
);
730 #ifdef CONFIG_EXT4_FS_XATTR
731 init_rwsem(&ei
->xattr_sem
);
733 init_rwsem(&ei
->i_data_sem
);
734 inode_init_once(&ei
->vfs_inode
);
737 static int init_inodecache(void)
739 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
740 sizeof(struct ext4_inode_info
),
741 0, (SLAB_RECLAIM_ACCOUNT
|
744 if (ext4_inode_cachep
== NULL
)
749 static void destroy_inodecache(void)
751 kmem_cache_destroy(ext4_inode_cachep
);
754 static void ext4_clear_inode(struct inode
*inode
)
756 ext4_discard_preallocations(inode
);
757 if (EXT4_JOURNAL(inode
))
758 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
759 &EXT4_I(inode
)->jinode
);
762 static inline void ext4_show_quota_options(struct seq_file
*seq
,
763 struct super_block
*sb
)
765 #if defined(CONFIG_QUOTA)
766 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
768 if (sbi
->s_jquota_fmt
)
769 seq_printf(seq
, ",jqfmt=%s",
770 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold" : "vfsv0");
772 if (sbi
->s_qf_names
[USRQUOTA
])
773 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
775 if (sbi
->s_qf_names
[GRPQUOTA
])
776 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
778 if (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
)
779 seq_puts(seq
, ",usrquota");
781 if (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)
782 seq_puts(seq
, ",grpquota");
788 * - it's set to a non-default value OR
789 * - if the per-sb default is different from the global default
791 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
794 unsigned long def_mount_opts
;
795 struct super_block
*sb
= vfs
->mnt_sb
;
796 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
797 struct ext4_super_block
*es
= sbi
->s_es
;
799 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
800 def_errors
= le16_to_cpu(es
->s_errors
);
802 if (sbi
->s_sb_block
!= 1)
803 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
804 if (test_opt(sb
, MINIX_DF
))
805 seq_puts(seq
, ",minixdf");
806 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
807 seq_puts(seq
, ",grpid");
808 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
809 seq_puts(seq
, ",nogrpid");
810 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
811 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
812 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
814 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
815 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
816 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
818 if (test_opt(sb
, ERRORS_RO
)) {
819 if (def_errors
== EXT4_ERRORS_PANIC
||
820 def_errors
== EXT4_ERRORS_CONTINUE
) {
821 seq_puts(seq
, ",errors=remount-ro");
824 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
825 seq_puts(seq
, ",errors=continue");
826 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
827 seq_puts(seq
, ",errors=panic");
828 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
829 seq_puts(seq
, ",nouid32");
830 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
831 seq_puts(seq
, ",debug");
832 if (test_opt(sb
, OLDALLOC
))
833 seq_puts(seq
, ",oldalloc");
834 #ifdef CONFIG_EXT4_FS_XATTR
835 if (test_opt(sb
, XATTR_USER
) &&
836 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
837 seq_puts(seq
, ",user_xattr");
838 if (!test_opt(sb
, XATTR_USER
) &&
839 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
840 seq_puts(seq
, ",nouser_xattr");
843 #ifdef CONFIG_EXT4_FS_POSIX_ACL
844 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
845 seq_puts(seq
, ",acl");
846 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
847 seq_puts(seq
, ",noacl");
849 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
850 seq_printf(seq
, ",commit=%u",
851 (unsigned) (sbi
->s_commit_interval
/ HZ
));
853 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
854 seq_printf(seq
, ",min_batch_time=%u",
855 (unsigned) sbi
->s_min_batch_time
);
857 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
858 seq_printf(seq
, ",max_batch_time=%u",
859 (unsigned) sbi
->s_min_batch_time
);
863 * We're changing the default of barrier mount option, so
864 * let's always display its mount state so it's clear what its
867 seq_puts(seq
, ",barrier=");
868 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
869 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
870 seq_puts(seq
, ",journal_async_commit");
871 if (test_opt(sb
, NOBH
))
872 seq_puts(seq
, ",nobh");
873 if (test_opt(sb
, I_VERSION
))
874 seq_puts(seq
, ",i_version");
875 if (!test_opt(sb
, DELALLOC
))
876 seq_puts(seq
, ",nodelalloc");
880 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
882 * journal mode get enabled in different ways
883 * So just print the value even if we didn't specify it
885 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
886 seq_puts(seq
, ",data=journal");
887 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
888 seq_puts(seq
, ",data=ordered");
889 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
890 seq_puts(seq
, ",data=writeback");
892 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
893 seq_printf(seq
, ",inode_readahead_blks=%u",
894 sbi
->s_inode_readahead_blks
);
896 if (test_opt(sb
, DATA_ERR_ABORT
))
897 seq_puts(seq
, ",data_err=abort");
899 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
900 seq_puts(seq
, ",noauto_da_alloc");
902 if (test_opt(sb
, DISCARD
))
903 seq_puts(seq
, ",discard");
905 if (test_opt(sb
, NOLOAD
))
906 seq_puts(seq
, ",norecovery");
908 ext4_show_quota_options(seq
, sb
);
913 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
914 u64 ino
, u32 generation
)
918 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
919 return ERR_PTR(-ESTALE
);
920 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
921 return ERR_PTR(-ESTALE
);
923 /* iget isn't really right if the inode is currently unallocated!!
925 * ext4_read_inode will return a bad_inode if the inode had been
926 * deleted, so we should be safe.
928 * Currently we don't know the generation for parent directory, so
929 * a generation of 0 means "accept any"
931 inode
= ext4_iget(sb
, ino
);
933 return ERR_CAST(inode
);
934 if (generation
&& inode
->i_generation
!= generation
) {
936 return ERR_PTR(-ESTALE
);
942 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
943 int fh_len
, int fh_type
)
945 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
949 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
950 int fh_len
, int fh_type
)
952 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
957 * Try to release metadata pages (indirect blocks, directories) which are
958 * mapped via the block device. Since these pages could have journal heads
959 * which would prevent try_to_free_buffers() from freeing them, we must use
960 * jbd2 layer's try_to_free_buffers() function to release them.
962 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
965 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
967 WARN_ON(PageChecked(page
));
968 if (!page_has_buffers(page
))
971 return jbd2_journal_try_to_free_buffers(journal
, page
,
973 return try_to_free_buffers(page
);
977 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
978 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
980 static int ext4_write_dquot(struct dquot
*dquot
);
981 static int ext4_acquire_dquot(struct dquot
*dquot
);
982 static int ext4_release_dquot(struct dquot
*dquot
);
983 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
984 static int ext4_write_info(struct super_block
*sb
, int type
);
985 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
986 char *path
, int remount
);
987 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
988 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
989 size_t len
, loff_t off
);
990 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
991 const char *data
, size_t len
, loff_t off
);
993 static const struct dquot_operations ext4_quota_operations
= {
994 .initialize
= dquot_initialize
,
996 .alloc_space
= dquot_alloc_space
,
997 .reserve_space
= dquot_reserve_space
,
998 .claim_space
= dquot_claim_space
,
999 .release_rsv
= dquot_release_reserved_space
,
1000 .get_reserved_space
= ext4_get_reserved_space
,
1001 .alloc_inode
= dquot_alloc_inode
,
1002 .free_space
= dquot_free_space
,
1003 .free_inode
= dquot_free_inode
,
1004 .transfer
= dquot_transfer
,
1005 .write_dquot
= ext4_write_dquot
,
1006 .acquire_dquot
= ext4_acquire_dquot
,
1007 .release_dquot
= ext4_release_dquot
,
1008 .mark_dirty
= ext4_mark_dquot_dirty
,
1009 .write_info
= ext4_write_info
,
1010 .alloc_dquot
= dquot_alloc
,
1011 .destroy_dquot
= dquot_destroy
,
1014 static const struct quotactl_ops ext4_qctl_operations
= {
1015 .quota_on
= ext4_quota_on
,
1016 .quota_off
= vfs_quota_off
,
1017 .quota_sync
= vfs_quota_sync
,
1018 .get_info
= vfs_get_dqinfo
,
1019 .set_info
= vfs_set_dqinfo
,
1020 .get_dqblk
= vfs_get_dqblk
,
1021 .set_dqblk
= vfs_set_dqblk
1025 static const struct super_operations ext4_sops
= {
1026 .alloc_inode
= ext4_alloc_inode
,
1027 .destroy_inode
= ext4_destroy_inode
,
1028 .write_inode
= ext4_write_inode
,
1029 .dirty_inode
= ext4_dirty_inode
,
1030 .delete_inode
= ext4_delete_inode
,
1031 .put_super
= ext4_put_super
,
1032 .sync_fs
= ext4_sync_fs
,
1033 .freeze_fs
= ext4_freeze
,
1034 .unfreeze_fs
= ext4_unfreeze
,
1035 .statfs
= ext4_statfs
,
1036 .remount_fs
= ext4_remount
,
1037 .clear_inode
= ext4_clear_inode
,
1038 .show_options
= ext4_show_options
,
1040 .quota_read
= ext4_quota_read
,
1041 .quota_write
= ext4_quota_write
,
1043 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1046 static const struct super_operations ext4_nojournal_sops
= {
1047 .alloc_inode
= ext4_alloc_inode
,
1048 .destroy_inode
= ext4_destroy_inode
,
1049 .write_inode
= ext4_write_inode
,
1050 .dirty_inode
= ext4_dirty_inode
,
1051 .delete_inode
= ext4_delete_inode
,
1052 .write_super
= ext4_write_super
,
1053 .put_super
= ext4_put_super
,
1054 .statfs
= ext4_statfs
,
1055 .remount_fs
= ext4_remount
,
1056 .clear_inode
= ext4_clear_inode
,
1057 .show_options
= ext4_show_options
,
1059 .quota_read
= ext4_quota_read
,
1060 .quota_write
= ext4_quota_write
,
1062 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1065 static const struct export_operations ext4_export_ops
= {
1066 .fh_to_dentry
= ext4_fh_to_dentry
,
1067 .fh_to_parent
= ext4_fh_to_parent
,
1068 .get_parent
= ext4_get_parent
,
1072 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1073 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1074 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1075 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1076 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1077 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1078 Opt_journal_update
, Opt_journal_dev
,
1079 Opt_journal_checksum
, Opt_journal_async_commit
,
1080 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1081 Opt_data_err_abort
, Opt_data_err_ignore
,
1082 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1083 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
1084 Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
, Opt_resize
,
1085 Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1086 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
1087 Opt_block_validity
, Opt_noblock_validity
,
1088 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1089 Opt_discard
, Opt_nodiscard
,
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"},
1114 {Opt_noload
, "norecovery"},
1117 {Opt_commit
, "commit=%u"},
1118 {Opt_min_batch_time
, "min_batch_time=%u"},
1119 {Opt_max_batch_time
, "max_batch_time=%u"},
1120 {Opt_journal_update
, "journal=update"},
1121 {Opt_journal_dev
, "journal_dev=%u"},
1122 {Opt_journal_checksum
, "journal_checksum"},
1123 {Opt_journal_async_commit
, "journal_async_commit"},
1124 {Opt_abort
, "abort"},
1125 {Opt_data_journal
, "data=journal"},
1126 {Opt_data_ordered
, "data=ordered"},
1127 {Opt_data_writeback
, "data=writeback"},
1128 {Opt_data_err_abort
, "data_err=abort"},
1129 {Opt_data_err_ignore
, "data_err=ignore"},
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"},
1155 {Opt_discard
, "discard"},
1156 {Opt_nodiscard
, "nodiscard"},
1160 static ext4_fsblk_t
get_sb_block(void **data
)
1162 ext4_fsblk_t sb_block
;
1163 char *options
= (char *) *data
;
1165 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1166 return 1; /* Default location */
1169 /* TODO: use simple_strtoll with >32bit ext4 */
1170 sb_block
= simple_strtoul(options
, &options
, 0);
1171 if (*options
&& *options
!= ',') {
1172 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1176 if (*options
== ',')
1178 *data
= (void *) options
;
1183 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1185 static int parse_options(char *options
, struct super_block
*sb
,
1186 unsigned long *journal_devnum
,
1187 unsigned int *journal_ioprio
,
1188 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1190 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1192 substring_t args
[MAX_OPT_ARGS
];
1203 while ((p
= strsep(&options
, ",")) != NULL
) {
1208 token
= match_token(p
, tokens
, args
);
1211 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1214 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1217 set_opt(sbi
->s_mount_opt
, GRPID
);
1220 clear_opt(sbi
->s_mount_opt
, GRPID
);
1223 if (match_int(&args
[0], &option
))
1225 sbi
->s_resuid
= option
;
1228 if (match_int(&args
[0], &option
))
1230 sbi
->s_resgid
= option
;
1233 /* handled by get_sb_block() instead of here */
1234 /* *sb_block = match_int(&args[0]); */
1237 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1238 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1239 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1242 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1243 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1244 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1247 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1248 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1249 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1252 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1255 set_opt(sbi
->s_mount_opt
, DEBUG
);
1258 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1261 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1263 #ifdef CONFIG_EXT4_FS_XATTR
1264 case Opt_user_xattr
:
1265 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1267 case Opt_nouser_xattr
:
1268 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1271 case Opt_user_xattr
:
1272 case Opt_nouser_xattr
:
1273 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1276 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1278 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1281 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1286 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1289 case Opt_journal_update
:
1291 /* Eventually we will want to be able to create
1292 a journal file here. For now, only allow the
1293 user to specify an existing inode to be the
1296 ext4_msg(sb
, KERN_ERR
,
1297 "Cannot specify journal on remount");
1300 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1302 case Opt_journal_dev
:
1304 ext4_msg(sb
, KERN_ERR
,
1305 "Cannot specify journal on remount");
1308 if (match_int(&args
[0], &option
))
1310 *journal_devnum
= option
;
1312 case Opt_journal_checksum
:
1313 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1315 case Opt_journal_async_commit
:
1316 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1317 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1320 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1323 if (match_int(&args
[0], &option
))
1328 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1329 sbi
->s_commit_interval
= HZ
* option
;
1331 case Opt_max_batch_time
:
1332 if (match_int(&args
[0], &option
))
1337 option
= EXT4_DEF_MAX_BATCH_TIME
;
1338 sbi
->s_max_batch_time
= option
;
1340 case Opt_min_batch_time
:
1341 if (match_int(&args
[0], &option
))
1345 sbi
->s_min_batch_time
= option
;
1347 case Opt_data_journal
:
1348 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1350 case Opt_data_ordered
:
1351 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1353 case Opt_data_writeback
:
1354 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1357 if ((sbi
->s_mount_opt
& EXT4_MOUNT_DATA_FLAGS
)
1359 ext4_msg(sb
, KERN_ERR
,
1360 "Cannot change data mode on remount");
1364 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DATA_FLAGS
;
1365 sbi
->s_mount_opt
|= data_opt
;
1368 case Opt_data_err_abort
:
1369 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1371 case Opt_data_err_ignore
:
1372 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1381 if (sb_any_quota_loaded(sb
) &&
1382 !sbi
->s_qf_names
[qtype
]) {
1383 ext4_msg(sb
, KERN_ERR
,
1384 "Cannot change journaled "
1385 "quota options when quota turned on");
1388 qname
= match_strdup(&args
[0]);
1390 ext4_msg(sb
, KERN_ERR
,
1391 "Not enough memory for "
1392 "storing quotafile name");
1395 if (sbi
->s_qf_names
[qtype
] &&
1396 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1397 ext4_msg(sb
, KERN_ERR
,
1398 "%s quota file already "
1399 "specified", QTYPE2NAME(qtype
));
1403 sbi
->s_qf_names
[qtype
] = qname
;
1404 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1405 ext4_msg(sb
, KERN_ERR
,
1406 "quotafile must be on "
1408 kfree(sbi
->s_qf_names
[qtype
]);
1409 sbi
->s_qf_names
[qtype
] = NULL
;
1412 set_opt(sbi
->s_mount_opt
, QUOTA
);
1414 case Opt_offusrjquota
:
1417 case Opt_offgrpjquota
:
1420 if (sb_any_quota_loaded(sb
) &&
1421 sbi
->s_qf_names
[qtype
]) {
1422 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1423 "journaled quota options when "
1428 * The space will be released later when all options
1429 * are confirmed to be correct
1431 sbi
->s_qf_names
[qtype
] = NULL
;
1433 case Opt_jqfmt_vfsold
:
1434 qfmt
= QFMT_VFS_OLD
;
1436 case Opt_jqfmt_vfsv0
:
1439 if (sb_any_quota_loaded(sb
) &&
1440 sbi
->s_jquota_fmt
!= qfmt
) {
1441 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1442 "journaled quota options when "
1446 sbi
->s_jquota_fmt
= qfmt
;
1450 set_opt(sbi
->s_mount_opt
, QUOTA
);
1451 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1454 set_opt(sbi
->s_mount_opt
, QUOTA
);
1455 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1458 if (sb_any_quota_loaded(sb
)) {
1459 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1460 "options when quota turned on");
1463 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1464 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1465 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1471 ext4_msg(sb
, KERN_ERR
,
1472 "quota options not supported");
1476 case Opt_offusrjquota
:
1477 case Opt_offgrpjquota
:
1478 case Opt_jqfmt_vfsold
:
1479 case Opt_jqfmt_vfsv0
:
1480 ext4_msg(sb
, KERN_ERR
,
1481 "journaled quota options not supported");
1487 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1490 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1493 if (match_int(&args
[0], &option
)) {
1494 set_opt(sbi
->s_mount_opt
, BARRIER
);
1498 set_opt(sbi
->s_mount_opt
, BARRIER
);
1500 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1506 ext4_msg(sb
, KERN_ERR
,
1507 "resize option only available "
1511 if (match_int(&args
[0], &option
) != 0)
1513 *n_blocks_count
= option
;
1516 set_opt(sbi
->s_mount_opt
, NOBH
);
1519 clear_opt(sbi
->s_mount_opt
, NOBH
);
1522 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1523 sb
->s_flags
|= MS_I_VERSION
;
1525 case Opt_nodelalloc
:
1526 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1529 if (match_int(&args
[0], &option
))
1533 sbi
->s_stripe
= option
;
1536 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1538 case Opt_block_validity
:
1539 set_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1541 case Opt_noblock_validity
:
1542 clear_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1544 case Opt_inode_readahead_blks
:
1545 if (match_int(&args
[0], &option
))
1547 if (option
< 0 || option
> (1 << 30))
1549 if (!is_power_of_2(option
)) {
1550 ext4_msg(sb
, KERN_ERR
,
1551 "EXT4-fs: inode_readahead_blks"
1552 " must be a power of 2");
1555 sbi
->s_inode_readahead_blks
= option
;
1557 case Opt_journal_ioprio
:
1558 if (match_int(&args
[0], &option
))
1560 if (option
< 0 || option
> 7)
1562 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1565 case Opt_noauto_da_alloc
:
1566 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1568 case Opt_auto_da_alloc
:
1569 if (match_int(&args
[0], &option
)) {
1570 clear_opt(sbi
->s_mount_opt
, NO_AUTO_DA_ALLOC
);
1574 clear_opt(sbi
->s_mount_opt
, NO_AUTO_DA_ALLOC
);
1576 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1579 set_opt(sbi
->s_mount_opt
, DISCARD
);
1582 clear_opt(sbi
->s_mount_opt
, DISCARD
);
1585 ext4_msg(sb
, KERN_ERR
,
1586 "Unrecognized mount option \"%s\" "
1587 "or missing value", p
);
1592 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1593 if ((sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
) &&
1594 sbi
->s_qf_names
[USRQUOTA
])
1595 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1597 if ((sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
) &&
1598 sbi
->s_qf_names
[GRPQUOTA
])
1599 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1601 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1602 (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)) ||
1603 (sbi
->s_qf_names
[GRPQUOTA
] &&
1604 (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
))) {
1605 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1610 if (!sbi
->s_jquota_fmt
) {
1611 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1616 if (sbi
->s_jquota_fmt
) {
1617 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1618 "specified with no journaling "
1627 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1630 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1633 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1634 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1635 "forcing read-only mode");
1640 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1641 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1642 "running e2fsck is recommended");
1643 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1644 ext4_msg(sb
, KERN_WARNING
,
1645 "warning: mounting fs with errors, "
1646 "running e2fsck is recommended");
1647 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1648 le16_to_cpu(es
->s_mnt_count
) >=
1649 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1650 ext4_msg(sb
, KERN_WARNING
,
1651 "warning: maximal mount count reached, "
1652 "running e2fsck is recommended");
1653 else if (le32_to_cpu(es
->s_checkinterval
) &&
1654 (le32_to_cpu(es
->s_lastcheck
) +
1655 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1656 ext4_msg(sb
, KERN_WARNING
,
1657 "warning: checktime reached, "
1658 "running e2fsck is recommended");
1659 if (!sbi
->s_journal
)
1660 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1661 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1662 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1663 le16_add_cpu(&es
->s_mnt_count
, 1);
1664 es
->s_mtime
= cpu_to_le32(get_seconds());
1665 ext4_update_dynamic_rev(sb
);
1667 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1669 ext4_commit_super(sb
, 1);
1670 if (test_opt(sb
, DEBUG
))
1671 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1672 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1674 sbi
->s_groups_count
,
1675 EXT4_BLOCKS_PER_GROUP(sb
),
1676 EXT4_INODES_PER_GROUP(sb
),
1682 static int ext4_fill_flex_info(struct super_block
*sb
)
1684 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1685 struct ext4_group_desc
*gdp
= NULL
;
1686 ext4_group_t flex_group_count
;
1687 ext4_group_t flex_group
;
1688 int groups_per_flex
= 0;
1692 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1693 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1695 if (groups_per_flex
< 2) {
1696 sbi
->s_log_groups_per_flex
= 0;
1700 /* We allocate both existing and potentially added groups */
1701 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1702 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1703 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1704 size
= flex_group_count
* sizeof(struct flex_groups
);
1705 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1706 if (sbi
->s_flex_groups
== NULL
) {
1707 sbi
->s_flex_groups
= vmalloc(size
);
1708 if (sbi
->s_flex_groups
)
1709 memset(sbi
->s_flex_groups
, 0, size
);
1711 if (sbi
->s_flex_groups
== NULL
) {
1712 ext4_msg(sb
, KERN_ERR
, "not enough memory for "
1713 "%u flex groups", flex_group_count
);
1717 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1718 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1720 flex_group
= ext4_flex_group(sbi
, i
);
1721 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1722 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1723 atomic_add(ext4_free_blks_count(sb
, gdp
),
1724 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1725 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1726 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1734 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1735 struct ext4_group_desc
*gdp
)
1739 if (sbi
->s_es
->s_feature_ro_compat
&
1740 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1741 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1742 __le32 le_group
= cpu_to_le32(block_group
);
1744 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1745 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1746 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1747 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1748 /* for checksum of struct ext4_group_desc do the rest...*/
1749 if ((sbi
->s_es
->s_feature_incompat
&
1750 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1751 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1752 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1753 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1757 return cpu_to_le16(crc
);
1760 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1761 struct ext4_group_desc
*gdp
)
1763 if ((sbi
->s_es
->s_feature_ro_compat
&
1764 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1765 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1771 /* Called at mount-time, super-block is locked */
1772 static int ext4_check_descriptors(struct super_block
*sb
)
1774 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1775 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1776 ext4_fsblk_t last_block
;
1777 ext4_fsblk_t block_bitmap
;
1778 ext4_fsblk_t inode_bitmap
;
1779 ext4_fsblk_t inode_table
;
1780 int flexbg_flag
= 0;
1783 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1786 ext4_debug("Checking group descriptors");
1788 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1789 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1791 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1792 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1794 last_block
= first_block
+
1795 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1797 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1798 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1799 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1800 "Block bitmap for group %u not in group "
1801 "(block %llu)!", i
, block_bitmap
);
1804 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1805 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1806 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1807 "Inode bitmap for group %u not in group "
1808 "(block %llu)!", i
, inode_bitmap
);
1811 inode_table
= ext4_inode_table(sb
, gdp
);
1812 if (inode_table
< first_block
||
1813 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1814 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1815 "Inode table for group %u not in group "
1816 "(block %llu)!", i
, inode_table
);
1819 ext4_lock_group(sb
, i
);
1820 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1821 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1822 "Checksum for group %u failed (%u!=%u)",
1823 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1824 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1825 if (!(sb
->s_flags
& MS_RDONLY
)) {
1826 ext4_unlock_group(sb
, i
);
1830 ext4_unlock_group(sb
, i
);
1832 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1835 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1836 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
1840 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1841 * the superblock) which were deleted from all directories, but held open by
1842 * a process at the time of a crash. We walk the list and try to delete these
1843 * inodes at recovery time (only with a read-write filesystem).
1845 * In order to keep the orphan inode chain consistent during traversal (in
1846 * case of crash during recovery), we link each inode into the superblock
1847 * orphan list_head and handle it the same way as an inode deletion during
1848 * normal operation (which journals the operations for us).
1850 * We only do an iget() and an iput() on each inode, which is very safe if we
1851 * accidentally point at an in-use or already deleted inode. The worst that
1852 * can happen in this case is that we get a "bit already cleared" message from
1853 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1854 * e2fsck was run on this filesystem, and it must have already done the orphan
1855 * inode cleanup for us, so we can safely abort without any further action.
1857 static void ext4_orphan_cleanup(struct super_block
*sb
,
1858 struct ext4_super_block
*es
)
1860 unsigned int s_flags
= sb
->s_flags
;
1861 int nr_orphans
= 0, nr_truncates
= 0;
1865 if (!es
->s_last_orphan
) {
1866 jbd_debug(4, "no orphan inodes to clean up\n");
1870 if (bdev_read_only(sb
->s_bdev
)) {
1871 ext4_msg(sb
, KERN_ERR
, "write access "
1872 "unavailable, skipping orphan cleanup");
1876 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1877 if (es
->s_last_orphan
)
1878 jbd_debug(1, "Errors on filesystem, "
1879 "clearing orphan list.\n");
1880 es
->s_last_orphan
= 0;
1881 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1885 if (s_flags
& MS_RDONLY
) {
1886 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
1887 sb
->s_flags
&= ~MS_RDONLY
;
1890 /* Needed for iput() to work correctly and not trash data */
1891 sb
->s_flags
|= MS_ACTIVE
;
1892 /* Turn on quotas so that they are updated correctly */
1893 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1894 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
1895 int ret
= ext4_quota_on_mount(sb
, i
);
1897 ext4_msg(sb
, KERN_ERR
,
1898 "Cannot turn on journaled "
1899 "quota: error %d", ret
);
1904 while (es
->s_last_orphan
) {
1905 struct inode
*inode
;
1907 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1908 if (IS_ERR(inode
)) {
1909 es
->s_last_orphan
= 0;
1913 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
1915 if (inode
->i_nlink
) {
1916 ext4_msg(sb
, KERN_DEBUG
,
1917 "%s: truncating inode %lu to %lld bytes",
1918 __func__
, inode
->i_ino
, inode
->i_size
);
1919 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1920 inode
->i_ino
, inode
->i_size
);
1921 ext4_truncate(inode
);
1924 ext4_msg(sb
, KERN_DEBUG
,
1925 "%s: deleting unreferenced inode %lu",
1926 __func__
, inode
->i_ino
);
1927 jbd_debug(2, "deleting unreferenced inode %lu\n",
1931 iput(inode
); /* The delete magic happens here! */
1934 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1937 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
1938 PLURAL(nr_orphans
));
1940 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
1941 PLURAL(nr_truncates
));
1943 /* Turn quotas off */
1944 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1945 if (sb_dqopt(sb
)->files
[i
])
1946 vfs_quota_off(sb
, i
, 0);
1949 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1953 * Maximal extent format file size.
1954 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1955 * extent format containers, within a sector_t, and within i_blocks
1956 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1957 * so that won't be a limiting factor.
1959 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1961 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
1964 loff_t upper_limit
= MAX_LFS_FILESIZE
;
1966 /* small i_blocks in vfs inode? */
1967 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
1969 * CONFIG_LBDAF is not enabled implies the inode
1970 * i_block represent total blocks in 512 bytes
1971 * 32 == size of vfs inode i_blocks * 8
1973 upper_limit
= (1LL << 32) - 1;
1975 /* total blocks in file system block size */
1976 upper_limit
>>= (blkbits
- 9);
1977 upper_limit
<<= blkbits
;
1980 /* 32-bit extent-start container, ee_block */
1985 /* Sanity check against vm- & vfs- imposed limits */
1986 if (res
> upper_limit
)
1993 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1994 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1995 * We need to be 1 filesystem block less than the 2^48 sector limit.
1997 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
1999 loff_t res
= EXT4_NDIR_BLOCKS
;
2002 /* This is calculated to be the largest file size for a dense, block
2003 * mapped file such that the file's total number of 512-byte sectors,
2004 * including data and all indirect blocks, does not exceed (2^48 - 1).
2006 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2007 * number of 512-byte sectors of the file.
2010 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2012 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2013 * the inode i_block field represents total file blocks in
2014 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2016 upper_limit
= (1LL << 32) - 1;
2018 /* total blocks in file system block size */
2019 upper_limit
>>= (bits
- 9);
2023 * We use 48 bit ext4_inode i_blocks
2024 * With EXT4_HUGE_FILE_FL set the i_blocks
2025 * represent total number of blocks in
2026 * file system block size
2028 upper_limit
= (1LL << 48) - 1;
2032 /* indirect blocks */
2034 /* double indirect blocks */
2035 meta_blocks
+= 1 + (1LL << (bits
-2));
2036 /* tripple indirect blocks */
2037 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2039 upper_limit
-= meta_blocks
;
2040 upper_limit
<<= bits
;
2042 res
+= 1LL << (bits
-2);
2043 res
+= 1LL << (2*(bits
-2));
2044 res
+= 1LL << (3*(bits
-2));
2046 if (res
> upper_limit
)
2049 if (res
> MAX_LFS_FILESIZE
)
2050 res
= MAX_LFS_FILESIZE
;
2055 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2056 ext4_fsblk_t logical_sb_block
, int nr
)
2058 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2059 ext4_group_t bg
, first_meta_bg
;
2062 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2064 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2066 return logical_sb_block
+ nr
+ 1;
2067 bg
= sbi
->s_desc_per_block
* nr
;
2068 if (ext4_bg_has_super(sb
, bg
))
2071 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2075 * ext4_get_stripe_size: Get the stripe size.
2076 * @sbi: In memory super block info
2078 * If we have specified it via mount option, then
2079 * use the mount option value. If the value specified at mount time is
2080 * greater than the blocks per group use the super block value.
2081 * If the super block value is greater than blocks per group return 0.
2082 * Allocator needs it be less than blocks per group.
2085 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2087 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2088 unsigned long stripe_width
=
2089 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2091 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2092 return sbi
->s_stripe
;
2094 if (stripe_width
<= sbi
->s_blocks_per_group
)
2095 return stripe_width
;
2097 if (stride
<= sbi
->s_blocks_per_group
)
2106 struct attribute attr
;
2107 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2108 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2109 const char *, size_t);
2113 static int parse_strtoul(const char *buf
,
2114 unsigned long max
, unsigned long *value
)
2118 while (*buf
&& isspace(*buf
))
2120 *value
= simple_strtoul(buf
, &endp
, 0);
2121 while (*endp
&& isspace(*endp
))
2123 if (*endp
|| *value
> max
)
2129 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2130 struct ext4_sb_info
*sbi
,
2133 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2134 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2137 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2138 struct ext4_sb_info
*sbi
, char *buf
)
2140 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2142 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2143 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2144 sbi
->s_sectors_written_start
) >> 1);
2147 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2148 struct ext4_sb_info
*sbi
, char *buf
)
2150 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2152 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2153 sbi
->s_kbytes_written
+
2154 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2155 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
2158 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2159 struct ext4_sb_info
*sbi
,
2160 const char *buf
, size_t count
)
2164 if (parse_strtoul(buf
, 0x40000000, &t
))
2167 if (!is_power_of_2(t
))
2170 sbi
->s_inode_readahead_blks
= t
;
2174 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2175 struct ext4_sb_info
*sbi
, char *buf
)
2177 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2179 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2182 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2183 struct ext4_sb_info
*sbi
,
2184 const char *buf
, size_t count
)
2186 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2189 if (parse_strtoul(buf
, 0xffffffff, &t
))
2195 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2196 static struct ext4_attr ext4_attr_##_name = { \
2197 .attr = {.name = __stringify(_name), .mode = _mode }, \
2200 .offset = offsetof(struct ext4_sb_info, _elname), \
2202 #define EXT4_ATTR(name, mode, show, store) \
2203 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2205 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2206 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2207 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2208 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2209 #define ATTR_LIST(name) &ext4_attr_##name.attr
2211 EXT4_RO_ATTR(delayed_allocation_blocks
);
2212 EXT4_RO_ATTR(session_write_kbytes
);
2213 EXT4_RO_ATTR(lifetime_write_kbytes
);
2214 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2215 inode_readahead_blks_store
, s_inode_readahead_blks
);
2216 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2217 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2218 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2219 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2220 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2221 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2222 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2223 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2225 static struct attribute
*ext4_attrs
[] = {
2226 ATTR_LIST(delayed_allocation_blocks
),
2227 ATTR_LIST(session_write_kbytes
),
2228 ATTR_LIST(lifetime_write_kbytes
),
2229 ATTR_LIST(inode_readahead_blks
),
2230 ATTR_LIST(inode_goal
),
2231 ATTR_LIST(mb_stats
),
2232 ATTR_LIST(mb_max_to_scan
),
2233 ATTR_LIST(mb_min_to_scan
),
2234 ATTR_LIST(mb_order2_req
),
2235 ATTR_LIST(mb_stream_req
),
2236 ATTR_LIST(mb_group_prealloc
),
2237 ATTR_LIST(max_writeback_mb_bump
),
2241 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2242 struct attribute
*attr
, char *buf
)
2244 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2246 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2248 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2251 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2252 struct attribute
*attr
,
2253 const char *buf
, size_t len
)
2255 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2257 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2259 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2262 static void ext4_sb_release(struct kobject
*kobj
)
2264 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2266 complete(&sbi
->s_kobj_unregister
);
2270 static struct sysfs_ops ext4_attr_ops
= {
2271 .show
= ext4_attr_show
,
2272 .store
= ext4_attr_store
,
2275 static struct kobj_type ext4_ktype
= {
2276 .default_attrs
= ext4_attrs
,
2277 .sysfs_ops
= &ext4_attr_ops
,
2278 .release
= ext4_sb_release
,
2282 * Check whether this filesystem can be mounted based on
2283 * the features present and the RDONLY/RDWR mount requested.
2284 * Returns 1 if this filesystem can be mounted as requested,
2285 * 0 if it cannot be.
2287 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2289 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2290 ext4_msg(sb
, KERN_ERR
,
2291 "Couldn't mount because of "
2292 "unsupported optional features (%x)",
2293 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2294 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2301 /* Check that feature set is OK for a read-write mount */
2302 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2303 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2304 "unsupported optional features (%x)",
2305 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2306 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2310 * Large file size enabled file system can only be mounted
2311 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2313 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2314 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2315 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2316 "cannot be mounted RDWR without "
2324 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
2325 __releases(kernel_lock
)
2326 __acquires(kernel_lock
)
2328 struct buffer_head
*bh
;
2329 struct ext4_super_block
*es
= NULL
;
2330 struct ext4_sb_info
*sbi
;
2332 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2333 ext4_fsblk_t logical_sb_block
;
2334 unsigned long offset
= 0;
2335 unsigned long journal_devnum
= 0;
2336 unsigned long def_mount_opts
;
2342 unsigned int db_count
;
2344 int needs_recovery
, has_huge_files
;
2347 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2349 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
2353 sbi
->s_blockgroup_lock
=
2354 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
2355 if (!sbi
->s_blockgroup_lock
) {
2359 sb
->s_fs_info
= sbi
;
2360 sbi
->s_mount_opt
= 0;
2361 sbi
->s_resuid
= EXT4_DEF_RESUID
;
2362 sbi
->s_resgid
= EXT4_DEF_RESGID
;
2363 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
2364 sbi
->s_sb_block
= sb_block
;
2365 sbi
->s_sectors_written_start
= part_stat_read(sb
->s_bdev
->bd_part
,
2370 /* Cleanup superblock name */
2371 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
2374 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
2376 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
2381 * The ext4 superblock will not be buffer aligned for other than 1kB
2382 * block sizes. We need to calculate the offset from buffer start.
2384 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
2385 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2386 offset
= do_div(logical_sb_block
, blocksize
);
2388 logical_sb_block
= sb_block
;
2391 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
2392 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
2396 * Note: s_es must be initialized as soon as possible because
2397 * some ext4 macro-instructions depend on its value
2399 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2401 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
2402 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
2404 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
2406 /* Set defaults before we parse the mount options */
2407 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
2408 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
2409 set_opt(sbi
->s_mount_opt
, DEBUG
);
2410 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
2411 set_opt(sbi
->s_mount_opt
, GRPID
);
2412 if (def_mount_opts
& EXT4_DEFM_UID16
)
2413 set_opt(sbi
->s_mount_opt
, NO_UID32
);
2414 #ifdef CONFIG_EXT4_FS_XATTR
2415 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
2416 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
2418 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2419 if (def_mount_opts
& EXT4_DEFM_ACL
)
2420 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
2422 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
2423 sbi
->s_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
2424 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
2425 sbi
->s_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
2426 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
2427 sbi
->s_mount_opt
|= EXT4_MOUNT_WRITEBACK_DATA
;
2429 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2430 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2431 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2432 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2434 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2436 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2437 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2438 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
2439 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
2440 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
2442 set_opt(sbi
->s_mount_opt
, BARRIER
);
2445 * enable delayed allocation by default
2446 * Use -o nodelalloc to turn it off
2448 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2450 if (!parse_options((char *) data
, sb
, &journal_devnum
,
2451 &journal_ioprio
, NULL
, 0))
2454 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2455 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2457 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2458 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2459 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2460 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2461 ext4_msg(sb
, KERN_WARNING
,
2462 "feature flags set on rev 0 fs, "
2463 "running e2fsck is recommended");
2466 * Check feature flags regardless of the revision level, since we
2467 * previously didn't change the revision level when setting the flags,
2468 * so there is a chance incompat flags are set on a rev 0 filesystem.
2470 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
2473 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2475 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2476 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2477 ext4_msg(sb
, KERN_ERR
,
2478 "Unsupported filesystem blocksize %d", blocksize
);
2482 if (sb
->s_blocksize
!= blocksize
) {
2483 /* Validate the filesystem blocksize */
2484 if (!sb_set_blocksize(sb
, blocksize
)) {
2485 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
2491 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2492 offset
= do_div(logical_sb_block
, blocksize
);
2493 bh
= sb_bread(sb
, logical_sb_block
);
2495 ext4_msg(sb
, KERN_ERR
,
2496 "Can't read superblock on 2nd try");
2499 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2501 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2502 ext4_msg(sb
, KERN_ERR
,
2503 "Magic mismatch, very weird!");
2508 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2509 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
2510 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
2512 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
2514 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2515 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2516 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2518 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2519 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2520 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2521 (!is_power_of_2(sbi
->s_inode_size
)) ||
2522 (sbi
->s_inode_size
> blocksize
)) {
2523 ext4_msg(sb
, KERN_ERR
,
2524 "unsupported inode size: %d",
2528 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2529 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2532 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2533 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2534 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2535 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2536 !is_power_of_2(sbi
->s_desc_size
)) {
2537 ext4_msg(sb
, KERN_ERR
,
2538 "unsupported descriptor size %lu",
2543 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2545 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2546 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2547 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2550 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2551 if (sbi
->s_inodes_per_block
== 0)
2553 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2554 sbi
->s_inodes_per_block
;
2555 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2557 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2558 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2559 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2561 for (i
= 0; i
< 4; i
++)
2562 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2563 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2564 i
= le32_to_cpu(es
->s_flags
);
2565 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
2566 sbi
->s_hash_unsigned
= 3;
2567 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
2568 #ifdef __CHAR_UNSIGNED__
2569 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
2570 sbi
->s_hash_unsigned
= 3;
2572 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
2577 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2578 ext4_msg(sb
, KERN_ERR
,
2579 "#blocks per group too big: %lu",
2580 sbi
->s_blocks_per_group
);
2583 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2584 ext4_msg(sb
, KERN_ERR
,
2585 "#inodes per group too big: %lu",
2586 sbi
->s_inodes_per_group
);
2591 * Test whether we have more sectors than will fit in sector_t,
2592 * and whether the max offset is addressable by the page cache.
2594 if ((ext4_blocks_count(es
) >
2595 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) ||
2596 (ext4_blocks_count(es
) >
2597 (pgoff_t
)(~0ULL) >> (PAGE_CACHE_SHIFT
- sb
->s_blocksize_bits
))) {
2598 ext4_msg(sb
, KERN_ERR
, "filesystem"
2599 " too large to mount safely on this system");
2600 if (sizeof(sector_t
) < 8)
2601 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
2606 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2609 /* check blocks count against device size */
2610 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
2611 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
2612 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
2613 "exceeds size of device (%llu blocks)",
2614 ext4_blocks_count(es
), blocks_count
);
2619 * It makes no sense for the first data block to be beyond the end
2620 * of the filesystem.
2622 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
2623 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
2624 "block %u is beyond end of filesystem (%llu)",
2625 le32_to_cpu(es
->s_first_data_block
),
2626 ext4_blocks_count(es
));
2629 blocks_count
= (ext4_blocks_count(es
) -
2630 le32_to_cpu(es
->s_first_data_block
) +
2631 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2632 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2633 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
2634 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
2635 "(block count %llu, first data block %u, "
2636 "blocks per group %lu)", sbi
->s_groups_count
,
2637 ext4_blocks_count(es
),
2638 le32_to_cpu(es
->s_first_data_block
),
2639 EXT4_BLOCKS_PER_GROUP(sb
));
2642 sbi
->s_groups_count
= blocks_count
;
2643 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
2644 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
2645 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2646 EXT4_DESC_PER_BLOCK(sb
);
2647 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2649 if (sbi
->s_group_desc
== NULL
) {
2650 ext4_msg(sb
, KERN_ERR
, "not enough memory");
2654 #ifdef CONFIG_PROC_FS
2656 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2659 bgl_lock_init(sbi
->s_blockgroup_lock
);
2661 for (i
= 0; i
< db_count
; i
++) {
2662 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2663 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2664 if (!sbi
->s_group_desc
[i
]) {
2665 ext4_msg(sb
, KERN_ERR
,
2666 "can't read group descriptor %d", i
);
2671 if (!ext4_check_descriptors(sb
)) {
2672 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
2675 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2676 if (!ext4_fill_flex_info(sb
)) {
2677 ext4_msg(sb
, KERN_ERR
,
2678 "unable to initialize "
2679 "flex_bg meta info!");
2683 sbi
->s_gdb_count
= db_count
;
2684 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2685 spin_lock_init(&sbi
->s_next_gen_lock
);
2687 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2688 ext4_count_free_blocks(sb
));
2690 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2691 ext4_count_free_inodes(sb
));
2694 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2695 ext4_count_dirs(sb
));
2698 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
2701 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
2705 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2706 sbi
->s_max_writeback_mb_bump
= 128;
2709 * set up enough so that it can read an inode
2711 if (!test_opt(sb
, NOLOAD
) &&
2712 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
2713 sb
->s_op
= &ext4_sops
;
2715 sb
->s_op
= &ext4_nojournal_sops
;
2716 sb
->s_export_op
= &ext4_export_ops
;
2717 sb
->s_xattr
= ext4_xattr_handlers
;
2719 sb
->s_qcop
= &ext4_qctl_operations
;
2720 sb
->dq_op
= &ext4_quota_operations
;
2722 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2723 mutex_init(&sbi
->s_orphan_lock
);
2724 mutex_init(&sbi
->s_resize_lock
);
2728 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2729 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2730 EXT4_FEATURE_INCOMPAT_RECOVER
));
2733 * The first inode we look at is the journal inode. Don't try
2734 * root first: it may be modified in the journal!
2736 if (!test_opt(sb
, NOLOAD
) &&
2737 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2738 if (ext4_load_journal(sb
, es
, journal_devnum
))
2740 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
2741 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2742 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
2743 "suppressed and not mounted read-only");
2746 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
2747 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2748 sbi
->s_journal
= NULL
;
2753 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2754 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2755 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2756 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
2760 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2761 jbd2_journal_set_features(sbi
->s_journal
,
2762 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2763 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2764 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2765 jbd2_journal_set_features(sbi
->s_journal
,
2766 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2767 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2768 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2770 jbd2_journal_clear_features(sbi
->s_journal
,
2771 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2772 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2775 /* We have now updated the journal if required, so we can
2776 * validate the data journaling mode. */
2777 switch (test_opt(sb
, DATA_FLAGS
)) {
2779 /* No mode set, assume a default based on the journal
2780 * capabilities: ORDERED_DATA if the journal can
2781 * cope, else JOURNAL_DATA
2783 if (jbd2_journal_check_available_features
2784 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2785 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2787 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2790 case EXT4_MOUNT_ORDERED_DATA
:
2791 case EXT4_MOUNT_WRITEBACK_DATA
:
2792 if (!jbd2_journal_check_available_features
2793 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2794 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
2795 "requested data journaling mode");
2801 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
2805 if (test_opt(sb
, NOBH
)) {
2806 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2807 ext4_msg(sb
, KERN_WARNING
, "Ignoring nobh option - "
2808 "its supported only with writeback mode");
2809 clear_opt(sbi
->s_mount_opt
, NOBH
);
2812 EXT4_SB(sb
)->dio_unwritten_wq
= create_workqueue("ext4-dio-unwritten");
2813 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
2814 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
2815 goto failed_mount_wq
;
2819 * The jbd2_journal_load will have done any necessary log recovery,
2820 * so we can safely mount the rest of the filesystem now.
2823 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2825 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
2826 ret
= PTR_ERR(root
);
2829 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2831 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
2834 sb
->s_root
= d_alloc_root(root
);
2836 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
2842 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
2844 /* determine the minimum size of new large inodes, if present */
2845 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2846 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2847 EXT4_GOOD_OLD_INODE_SIZE
;
2848 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2849 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2850 if (sbi
->s_want_extra_isize
<
2851 le16_to_cpu(es
->s_want_extra_isize
))
2852 sbi
->s_want_extra_isize
=
2853 le16_to_cpu(es
->s_want_extra_isize
);
2854 if (sbi
->s_want_extra_isize
<
2855 le16_to_cpu(es
->s_min_extra_isize
))
2856 sbi
->s_want_extra_isize
=
2857 le16_to_cpu(es
->s_min_extra_isize
);
2860 /* Check if enough inode space is available */
2861 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2862 sbi
->s_inode_size
) {
2863 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2864 EXT4_GOOD_OLD_INODE_SIZE
;
2865 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
2869 if (test_opt(sb
, DELALLOC
) &&
2870 (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)) {
2871 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
2872 "requested data journaling mode");
2873 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2876 err
= ext4_setup_system_zone(sb
);
2878 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
2879 "zone (%d)\n", err
);
2884 err
= ext4_mb_init(sb
, needs_recovery
);
2886 ext4_msg(sb
, KERN_ERR
, "failed to initalize mballoc (%d)",
2891 sbi
->s_kobj
.kset
= ext4_kset
;
2892 init_completion(&sbi
->s_kobj_unregister
);
2893 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
2896 ext4_mb_release(sb
);
2897 ext4_ext_release(sb
);
2901 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
2902 ext4_orphan_cleanup(sb
, es
);
2903 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
2904 if (needs_recovery
) {
2905 ext4_msg(sb
, KERN_INFO
, "recovery complete");
2906 ext4_mark_recovery_complete(sb
, es
);
2908 if (EXT4_SB(sb
)->s_journal
) {
2909 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2910 descr
= " journalled data mode";
2911 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2912 descr
= " ordered data mode";
2914 descr
= " writeback data mode";
2916 descr
= "out journal";
2918 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s", descr
);
2925 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
2929 ext4_msg(sb
, KERN_ERR
, "mount failed");
2930 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
2932 ext4_release_system_zone(sb
);
2933 if (sbi
->s_journal
) {
2934 jbd2_journal_destroy(sbi
->s_journal
);
2935 sbi
->s_journal
= NULL
;
2938 if (sbi
->s_flex_groups
) {
2939 if (is_vmalloc_addr(sbi
->s_flex_groups
))
2940 vfree(sbi
->s_flex_groups
);
2942 kfree(sbi
->s_flex_groups
);
2944 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2945 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2946 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2947 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
2949 for (i
= 0; i
< db_count
; i
++)
2950 brelse(sbi
->s_group_desc
[i
]);
2951 kfree(sbi
->s_group_desc
);
2954 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
2957 for (i
= 0; i
< MAXQUOTAS
; i
++)
2958 kfree(sbi
->s_qf_names
[i
]);
2960 ext4_blkdev_remove(sbi
);
2963 sb
->s_fs_info
= NULL
;
2964 kfree(sbi
->s_blockgroup_lock
);
2971 * Setup any per-fs journal parameters now. We'll do this both on
2972 * initial mount, once the journal has been initialised but before we've
2973 * done any recovery; and again on any subsequent remount.
2975 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
2977 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2979 journal
->j_commit_interval
= sbi
->s_commit_interval
;
2980 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
2981 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
2983 spin_lock(&journal
->j_state_lock
);
2984 if (test_opt(sb
, BARRIER
))
2985 journal
->j_flags
|= JBD2_BARRIER
;
2987 journal
->j_flags
&= ~JBD2_BARRIER
;
2988 if (test_opt(sb
, DATA_ERR_ABORT
))
2989 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
2991 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
2992 spin_unlock(&journal
->j_state_lock
);
2995 static journal_t
*ext4_get_journal(struct super_block
*sb
,
2996 unsigned int journal_inum
)
2998 struct inode
*journal_inode
;
3001 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3003 /* First, test for the existence of a valid inode on disk. Bad
3004 * things happen if we iget() an unused inode, as the subsequent
3005 * iput() will try to delete it. */
3007 journal_inode
= ext4_iget(sb
, journal_inum
);
3008 if (IS_ERR(journal_inode
)) {
3009 ext4_msg(sb
, KERN_ERR
, "no journal found");
3012 if (!journal_inode
->i_nlink
) {
3013 make_bad_inode(journal_inode
);
3014 iput(journal_inode
);
3015 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3019 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3020 journal_inode
, journal_inode
->i_size
);
3021 if (!S_ISREG(journal_inode
->i_mode
)) {
3022 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3023 iput(journal_inode
);
3027 journal
= jbd2_journal_init_inode(journal_inode
);
3029 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3030 iput(journal_inode
);
3033 journal
->j_private
= sb
;
3034 ext4_init_journal_params(sb
, journal
);
3038 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3041 struct buffer_head
*bh
;
3045 int hblock
, blocksize
;
3046 ext4_fsblk_t sb_block
;
3047 unsigned long offset
;
3048 struct ext4_super_block
*es
;
3049 struct block_device
*bdev
;
3051 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3053 bdev
= ext4_blkdev_get(j_dev
, sb
);
3057 if (bd_claim(bdev
, sb
)) {
3058 ext4_msg(sb
, KERN_ERR
,
3059 "failed to claim external journal device");
3060 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
3064 blocksize
= sb
->s_blocksize
;
3065 hblock
= bdev_logical_block_size(bdev
);
3066 if (blocksize
< hblock
) {
3067 ext4_msg(sb
, KERN_ERR
,
3068 "blocksize too small for journal device");
3072 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3073 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3074 set_blocksize(bdev
, blocksize
);
3075 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3076 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3077 "external journal");
3081 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3082 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3083 !(le32_to_cpu(es
->s_feature_incompat
) &
3084 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3085 ext4_msg(sb
, KERN_ERR
, "external journal has "
3091 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3092 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3097 len
= ext4_blocks_count(es
);
3098 start
= sb_block
+ 1;
3099 brelse(bh
); /* we're done with the superblock */
3101 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3102 start
, len
, blocksize
);
3104 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3107 journal
->j_private
= sb
;
3108 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3109 wait_on_buffer(journal
->j_sb_buffer
);
3110 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3111 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3114 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3115 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3116 "user (unsupported) - %d",
3117 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3120 EXT4_SB(sb
)->journal_bdev
= bdev
;
3121 ext4_init_journal_params(sb
, journal
);
3125 jbd2_journal_destroy(journal
);
3127 ext4_blkdev_put(bdev
);
3131 static int ext4_load_journal(struct super_block
*sb
,
3132 struct ext4_super_block
*es
,
3133 unsigned long journal_devnum
)
3136 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3139 int really_read_only
;
3141 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3143 if (journal_devnum
&&
3144 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3145 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3146 "numbers have changed");
3147 journal_dev
= new_decode_dev(journal_devnum
);
3149 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3151 really_read_only
= bdev_read_only(sb
->s_bdev
);
3154 * Are we loading a blank journal or performing recovery after a
3155 * crash? For recovery, we need to check in advance whether we
3156 * can get read-write access to the device.
3158 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3159 if (sb
->s_flags
& MS_RDONLY
) {
3160 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3161 "required on readonly filesystem");
3162 if (really_read_only
) {
3163 ext4_msg(sb
, KERN_ERR
, "write access "
3164 "unavailable, cannot proceed");
3167 ext4_msg(sb
, KERN_INFO
, "write access will "
3168 "be enabled during recovery");
3172 if (journal_inum
&& journal_dev
) {
3173 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3174 "and inode journals!");
3179 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3182 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3186 if (!(journal
->j_flags
& JBD2_BARRIER
))
3187 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3189 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
3190 err
= jbd2_journal_update_format(journal
);
3192 ext4_msg(sb
, KERN_ERR
, "error updating journal");
3193 jbd2_journal_destroy(journal
);
3198 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
3199 err
= jbd2_journal_wipe(journal
, !really_read_only
);
3201 err
= jbd2_journal_load(journal
);
3204 ext4_msg(sb
, KERN_ERR
, "error loading journal");
3205 jbd2_journal_destroy(journal
);
3209 EXT4_SB(sb
)->s_journal
= journal
;
3210 ext4_clear_journal_err(sb
, es
);
3212 if (journal_devnum
&&
3213 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3214 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
3216 /* Make sure we flush the recovery flag to disk. */
3217 ext4_commit_super(sb
, 1);
3223 static int ext4_commit_super(struct super_block
*sb
, int sync
)
3225 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
3226 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
3231 if (buffer_write_io_error(sbh
)) {
3233 * Oh, dear. A previous attempt to write the
3234 * superblock failed. This could happen because the
3235 * USB device was yanked out. Or it could happen to
3236 * be a transient write error and maybe the block will
3237 * be remapped. Nothing we can do but to retry the
3238 * write and hope for the best.
3240 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
3241 "superblock detected");
3242 clear_buffer_write_io_error(sbh
);
3243 set_buffer_uptodate(sbh
);
3246 * If the file system is mounted read-only, don't update the
3247 * superblock write time. This avoids updating the superblock
3248 * write time when we are mounting the root file system
3249 * read/only but we need to replay the journal; at that point,
3250 * for people who are east of GMT and who make their clock
3251 * tick in localtime for Windows bug-for-bug compatibility,
3252 * the clock is set in the future, and this will cause e2fsck
3253 * to complain and force a full file system check.
3255 if (!(sb
->s_flags
& MS_RDONLY
))
3256 es
->s_wtime
= cpu_to_le32(get_seconds());
3257 es
->s_kbytes_written
=
3258 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
3259 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
3260 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
3261 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
3262 &EXT4_SB(sb
)->s_freeblocks_counter
));
3263 es
->s_free_inodes_count
= cpu_to_le32(percpu_counter_sum_positive(
3264 &EXT4_SB(sb
)->s_freeinodes_counter
));
3266 BUFFER_TRACE(sbh
, "marking dirty");
3267 mark_buffer_dirty(sbh
);
3269 error
= sync_dirty_buffer(sbh
);
3273 error
= buffer_write_io_error(sbh
);
3275 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
3277 clear_buffer_write_io_error(sbh
);
3278 set_buffer_uptodate(sbh
);
3285 * Have we just finished recovery? If so, and if we are mounting (or
3286 * remounting) the filesystem readonly, then we will end up with a
3287 * consistent fs on disk. Record that fact.
3289 static void ext4_mark_recovery_complete(struct super_block
*sb
,
3290 struct ext4_super_block
*es
)
3292 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
3294 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3295 BUG_ON(journal
!= NULL
);
3298 jbd2_journal_lock_updates(journal
);
3299 if (jbd2_journal_flush(journal
) < 0)
3302 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
3303 sb
->s_flags
& MS_RDONLY
) {
3304 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3305 ext4_commit_super(sb
, 1);
3309 jbd2_journal_unlock_updates(journal
);
3313 * If we are mounting (or read-write remounting) a filesystem whose journal
3314 * has recorded an error from a previous lifetime, move that error to the
3315 * main filesystem now.
3317 static void ext4_clear_journal_err(struct super_block
*sb
,
3318 struct ext4_super_block
*es
)
3324 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3326 journal
= EXT4_SB(sb
)->s_journal
;
3329 * Now check for any error status which may have been recorded in the
3330 * journal by a prior ext4_error() or ext4_abort()
3333 j_errno
= jbd2_journal_errno(journal
);
3337 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
3338 ext4_warning(sb
, __func__
, "Filesystem error recorded "
3339 "from previous mount: %s", errstr
);
3340 ext4_warning(sb
, __func__
, "Marking fs in need of "
3341 "filesystem check.");
3343 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
3344 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
3345 ext4_commit_super(sb
, 1);
3347 jbd2_journal_clear_err(journal
);
3352 * Force the running and committing transactions to commit,
3353 * and wait on the commit.
3355 int ext4_force_commit(struct super_block
*sb
)
3360 if (sb
->s_flags
& MS_RDONLY
)
3363 journal
= EXT4_SB(sb
)->s_journal
;
3365 ret
= ext4_journal_force_commit(journal
);
3370 static void ext4_write_super(struct super_block
*sb
)
3373 ext4_commit_super(sb
, 1);
3377 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
3381 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3383 trace_ext4_sync_fs(sb
, wait
);
3384 flush_workqueue(sbi
->dio_unwritten_wq
);
3385 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
3387 jbd2_log_wait_commit(sbi
->s_journal
, target
);
3393 * LVM calls this function before a (read-only) snapshot is created. This
3394 * gives us a chance to flush the journal completely and mark the fs clean.
3396 static int ext4_freeze(struct super_block
*sb
)
3401 if (sb
->s_flags
& MS_RDONLY
)
3404 journal
= EXT4_SB(sb
)->s_journal
;
3406 /* Now we set up the journal barrier. */
3407 jbd2_journal_lock_updates(journal
);
3410 * Don't clear the needs_recovery flag if we failed to flush
3413 error
= jbd2_journal_flush(journal
);
3416 jbd2_journal_unlock_updates(journal
);
3420 /* Journal blocked and flushed, clear needs_recovery flag. */
3421 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3422 error
= ext4_commit_super(sb
, 1);
3429 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3430 * flag here, even though the filesystem is not technically dirty yet.
3432 static int ext4_unfreeze(struct super_block
*sb
)
3434 if (sb
->s_flags
& MS_RDONLY
)
3438 /* Reset the needs_recovery flag before the fs is unlocked. */
3439 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3440 ext4_commit_super(sb
, 1);
3442 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3446 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3448 struct ext4_super_block
*es
;
3449 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3450 ext4_fsblk_t n_blocks_count
= 0;
3451 unsigned long old_sb_flags
;
3452 struct ext4_mount_options old_opts
;
3454 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3462 /* Store the original options */
3464 old_sb_flags
= sb
->s_flags
;
3465 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3466 old_opts
.s_resuid
= sbi
->s_resuid
;
3467 old_opts
.s_resgid
= sbi
->s_resgid
;
3468 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3469 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
3470 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
3472 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3473 for (i
= 0; i
< MAXQUOTAS
; i
++)
3474 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3476 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
3477 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
3480 * Allow the "check" option to be passed as a remount option.
3482 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
3483 &n_blocks_count
, 1)) {
3488 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
3489 ext4_abort(sb
, __func__
, "Abort forced by user");
3491 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3492 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
3496 if (sbi
->s_journal
) {
3497 ext4_init_journal_params(sb
, sbi
->s_journal
);
3498 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3501 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3502 n_blocks_count
> ext4_blocks_count(es
)) {
3503 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
3508 if (*flags
& MS_RDONLY
) {
3510 * First of all, the unconditional stuff we have to do
3511 * to disable replay of the journal when we next remount
3513 sb
->s_flags
|= MS_RDONLY
;
3516 * OK, test if we are remounting a valid rw partition
3517 * readonly, and if so set the rdonly flag and then
3518 * mark the partition as valid again.
3520 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3521 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3522 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3525 ext4_mark_recovery_complete(sb
, es
);
3527 /* Make sure we can mount this feature set readwrite */
3528 if (!ext4_feature_set_ok(sb
, 0)) {
3533 * Make sure the group descriptor checksums
3534 * are sane. If they aren't, refuse to remount r/w.
3536 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3537 struct ext4_group_desc
*gdp
=
3538 ext4_get_group_desc(sb
, g
, NULL
);
3540 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3541 ext4_msg(sb
, KERN_ERR
,
3542 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3543 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3544 le16_to_cpu(gdp
->bg_checksum
));
3551 * If we have an unprocessed orphan list hanging
3552 * around from a previously readonly bdev mount,
3553 * require a full umount/remount for now.
3555 if (es
->s_last_orphan
) {
3556 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
3557 "remount RDWR because of unprocessed "
3558 "orphan inode list. Please "
3559 "umount/remount instead");
3565 * Mounting a RDONLY partition read-write, so reread
3566 * and store the current valid flag. (It may have
3567 * been changed by e2fsck since we originally mounted
3571 ext4_clear_journal_err(sb
, es
);
3572 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3573 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3575 if (!ext4_setup_super(sb
, es
, 0))
3576 sb
->s_flags
&= ~MS_RDONLY
;
3579 ext4_setup_system_zone(sb
);
3580 if (sbi
->s_journal
== NULL
)
3581 ext4_commit_super(sb
, 1);
3584 /* Release old quota file names */
3585 for (i
= 0; i
< MAXQUOTAS
; i
++)
3586 if (old_opts
.s_qf_names
[i
] &&
3587 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3588 kfree(old_opts
.s_qf_names
[i
]);
3595 sb
->s_flags
= old_sb_flags
;
3596 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3597 sbi
->s_resuid
= old_opts
.s_resuid
;
3598 sbi
->s_resgid
= old_opts
.s_resgid
;
3599 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3600 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
3601 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
3603 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3604 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3605 if (sbi
->s_qf_names
[i
] &&
3606 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3607 kfree(sbi
->s_qf_names
[i
]);
3608 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3616 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3618 struct super_block
*sb
= dentry
->d_sb
;
3619 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3620 struct ext4_super_block
*es
= sbi
->s_es
;
3623 if (test_opt(sb
, MINIX_DF
)) {
3624 sbi
->s_overhead_last
= 0;
3625 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3626 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3627 ext4_fsblk_t overhead
= 0;
3630 * Compute the overhead (FS structures). This is constant
3631 * for a given filesystem unless the number of block groups
3632 * changes so we cache the previous value until it does.
3636 * All of the blocks before first_data_block are
3639 overhead
= le32_to_cpu(es
->s_first_data_block
);
3642 * Add the overhead attributed to the superblock and
3643 * block group descriptors. If the sparse superblocks
3644 * feature is turned on, then not all groups have this.
3646 for (i
= 0; i
< ngroups
; i
++) {
3647 overhead
+= ext4_bg_has_super(sb
, i
) +
3648 ext4_bg_num_gdb(sb
, i
);
3653 * Every block group has an inode bitmap, a block
3654 * bitmap, and an inode table.
3656 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3657 sbi
->s_overhead_last
= overhead
;
3659 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3662 buf
->f_type
= EXT4_SUPER_MAGIC
;
3663 buf
->f_bsize
= sb
->s_blocksize
;
3664 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3665 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3666 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3667 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3668 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3670 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3671 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3672 buf
->f_namelen
= EXT4_NAME_LEN
;
3673 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3674 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3675 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3676 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3681 /* Helper function for writing quotas on sync - we need to start transaction
3682 * before quota file is locked for write. Otherwise the are possible deadlocks:
3683 * Process 1 Process 2
3684 * ext4_create() quota_sync()
3685 * jbd2_journal_start() write_dquot()
3686 * vfs_dq_init() down(dqio_mutex)
3687 * down(dqio_mutex) jbd2_journal_start()
3693 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3695 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3698 static int ext4_write_dquot(struct dquot
*dquot
)
3702 struct inode
*inode
;
3704 inode
= dquot_to_inode(dquot
);
3705 handle
= ext4_journal_start(inode
,
3706 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3708 return PTR_ERR(handle
);
3709 ret
= dquot_commit(dquot
);
3710 err
= ext4_journal_stop(handle
);
3716 static int ext4_acquire_dquot(struct dquot
*dquot
)
3721 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3722 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3724 return PTR_ERR(handle
);
3725 ret
= dquot_acquire(dquot
);
3726 err
= ext4_journal_stop(handle
);
3732 static int ext4_release_dquot(struct dquot
*dquot
)
3737 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3738 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3739 if (IS_ERR(handle
)) {
3740 /* Release dquot anyway to avoid endless cycle in dqput() */
3741 dquot_release(dquot
);
3742 return PTR_ERR(handle
);
3744 ret
= dquot_release(dquot
);
3745 err
= ext4_journal_stop(handle
);
3751 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3753 /* Are we journaling quotas? */
3754 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3755 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3756 dquot_mark_dquot_dirty(dquot
);
3757 return ext4_write_dquot(dquot
);
3759 return dquot_mark_dquot_dirty(dquot
);
3763 static int ext4_write_info(struct super_block
*sb
, int type
)
3768 /* Data block + inode block */
3769 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3771 return PTR_ERR(handle
);
3772 ret
= dquot_commit_info(sb
, type
);
3773 err
= ext4_journal_stop(handle
);
3780 * Turn on quotas during mount time - we need to find
3781 * the quota file and such...
3783 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3785 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3786 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3790 * Standard function to be called on quota_on
3792 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3793 char *name
, int remount
)
3798 if (!test_opt(sb
, QUOTA
))
3800 /* When remounting, no checks are needed and in fact, name is NULL */
3802 return vfs_quota_on(sb
, type
, format_id
, name
, remount
);
3804 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
3808 /* Quotafile not on the same filesystem? */
3809 if (path
.mnt
->mnt_sb
!= sb
) {
3813 /* Journaling quota? */
3814 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3815 /* Quotafile not in fs root? */
3816 if (path
.dentry
->d_parent
!= sb
->s_root
)
3817 ext4_msg(sb
, KERN_WARNING
,
3818 "Quota file not on filesystem root. "
3819 "Journaled quota will not work");
3823 * When we journal data on quota file, we have to flush journal to see
3824 * all updates to the file when we bypass pagecache...
3826 if (EXT4_SB(sb
)->s_journal
&&
3827 ext4_should_journal_data(path
.dentry
->d_inode
)) {
3829 * We don't need to lock updates but journal_flush() could
3830 * otherwise be livelocked...
3832 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3833 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3834 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3841 err
= vfs_quota_on_path(sb
, type
, format_id
, &path
);
3846 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3847 * acquiring the locks... As quota files are never truncated and quota code
3848 * itself serializes the operations (and noone else should touch the files)
3849 * we don't have to be afraid of races */
3850 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3851 size_t len
, loff_t off
)
3853 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3854 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3856 int offset
= off
& (sb
->s_blocksize
- 1);
3859 struct buffer_head
*bh
;
3860 loff_t i_size
= i_size_read(inode
);
3864 if (off
+len
> i_size
)
3867 while (toread
> 0) {
3868 tocopy
= sb
->s_blocksize
- offset
< toread
?
3869 sb
->s_blocksize
- offset
: toread
;
3870 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3873 if (!bh
) /* A hole? */
3874 memset(data
, 0, tocopy
);
3876 memcpy(data
, bh
->b_data
+offset
, tocopy
);
3886 /* Write to quotafile (we know the transaction is already started and has
3887 * enough credits) */
3888 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
3889 const char *data
, size_t len
, loff_t off
)
3891 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3892 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3894 int offset
= off
& (sb
->s_blocksize
- 1);
3896 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
3897 size_t towrite
= len
;
3898 struct buffer_head
*bh
;
3899 handle_t
*handle
= journal_current_handle();
3901 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
3902 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
3903 " cancelled because transaction is not started",
3904 (unsigned long long)off
, (unsigned long long)len
);
3907 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
3908 while (towrite
> 0) {
3909 tocopy
= sb
->s_blocksize
- offset
< towrite
?
3910 sb
->s_blocksize
- offset
: towrite
;
3911 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
3914 if (journal_quota
) {
3915 err
= ext4_journal_get_write_access(handle
, bh
);
3922 memcpy(bh
->b_data
+offset
, data
, tocopy
);
3923 flush_dcache_page(bh
->b_page
);
3926 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
3928 /* Always do at least ordered writes for quotas */
3929 err
= ext4_jbd2_file_inode(handle
, inode
);
3930 mark_buffer_dirty(bh
);
3941 if (len
== towrite
) {
3942 mutex_unlock(&inode
->i_mutex
);
3945 if (inode
->i_size
< off
+len
-towrite
) {
3946 i_size_write(inode
, off
+len
-towrite
);
3947 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
3949 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3950 ext4_mark_inode_dirty(handle
, inode
);
3951 mutex_unlock(&inode
->i_mutex
);
3952 return len
- towrite
;
3957 static int ext4_get_sb(struct file_system_type
*fs_type
, int flags
,
3958 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3960 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
,mnt
);
3963 #if !defined(CONTIG_EXT2_FS) && defined(CONFIG_EXT4_USE_FOR_EXT23)
3964 static struct file_system_type ext2_fs_type
= {
3965 .owner
= THIS_MODULE
,
3967 .get_sb
= ext4_get_sb
,
3968 .kill_sb
= kill_block_super
,
3969 .fs_flags
= FS_REQUIRES_DEV
,
3972 static inline void register_as_ext2(void)
3974 int err
= register_filesystem(&ext2_fs_type
);
3977 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
3980 static inline void unregister_as_ext2(void)
3982 unregister_filesystem(&ext2_fs_type
);
3985 static inline void register_as_ext2(void) { }
3986 static inline void unregister_as_ext2(void) { }
3989 #if !defined(CONTIG_EXT3_FS) && defined(CONFIG_EXT4_USE_FOR_EXT23)
3990 static struct file_system_type ext3_fs_type
= {
3991 .owner
= THIS_MODULE
,
3993 .get_sb
= ext4_get_sb
,
3994 .kill_sb
= kill_block_super
,
3995 .fs_flags
= FS_REQUIRES_DEV
,
3998 static inline void register_as_ext3(void)
4000 int err
= register_filesystem(&ext3_fs_type
);
4003 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4006 static inline void unregister_as_ext3(void)
4008 unregister_filesystem(&ext3_fs_type
);
4011 static inline void register_as_ext3(void) { }
4012 static inline void unregister_as_ext3(void) { }
4015 static struct file_system_type ext4_fs_type
= {
4016 .owner
= THIS_MODULE
,
4018 .get_sb
= ext4_get_sb
,
4019 .kill_sb
= kill_block_super
,
4020 .fs_flags
= FS_REQUIRES_DEV
,
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();
4046 err
= register_filesystem(&ext4_fs_type
);
4051 unregister_as_ext2();
4052 unregister_as_ext3();
4053 destroy_inodecache();
4057 exit_ext4_mballoc();
4059 remove_proc_entry("fs/ext4", NULL
);
4060 kset_unregister(ext4_kset
);
4062 exit_ext4_system_zone();
4066 static void __exit
exit_ext4_fs(void)
4068 unregister_as_ext2();
4069 unregister_as_ext3();
4070 unregister_filesystem(&ext4_fs_type
);
4071 destroy_inodecache();
4073 exit_ext4_mballoc();
4074 remove_proc_entry("fs/ext4", NULL
);
4075 kset_unregister(ext4_kset
);
4076 exit_ext4_system_zone();
4079 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4080 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4081 MODULE_LICENSE("GPL");
4082 module_init(init_ext4_fs
)
4083 module_exit(exit_ext4_fs
)