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/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
44 #include "ext4_jbd2.h"
50 struct proc_dir_entry
*ext4_proc_root
;
52 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
53 unsigned long journal_devnum
);
54 static void ext4_commit_super(struct super_block
*sb
,
55 struct ext4_super_block
*es
, int sync
);
56 static void ext4_mark_recovery_complete(struct super_block
*sb
,
57 struct ext4_super_block
*es
);
58 static void ext4_clear_journal_err(struct super_block
*sb
,
59 struct ext4_super_block
*es
);
60 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
61 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
63 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
64 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
65 static void ext4_unlockfs(struct super_block
*sb
);
66 static void ext4_write_super(struct super_block
*sb
);
67 static void ext4_write_super_lockfs(struct super_block
*sb
);
70 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
71 struct ext4_group_desc
*bg
)
73 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
74 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
75 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
78 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
79 struct ext4_group_desc
*bg
)
81 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
82 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
83 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
86 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
87 struct ext4_group_desc
*bg
)
89 return le32_to_cpu(bg
->bg_inode_table_lo
) |
90 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
91 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
94 __u32
ext4_free_blks_count(struct super_block
*sb
,
95 struct ext4_group_desc
*bg
)
97 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
98 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
99 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
102 __u32
ext4_free_inodes_count(struct super_block
*sb
,
103 struct ext4_group_desc
*bg
)
105 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
106 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
107 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
110 __u32
ext4_used_dirs_count(struct super_block
*sb
,
111 struct ext4_group_desc
*bg
)
113 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
114 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
115 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
118 __u32
ext4_itable_unused_count(struct super_block
*sb
,
119 struct ext4_group_desc
*bg
)
121 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
122 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
123 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
126 void ext4_block_bitmap_set(struct super_block
*sb
,
127 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
129 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
130 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
131 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
134 void ext4_inode_bitmap_set(struct super_block
*sb
,
135 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
137 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
138 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
139 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
142 void ext4_inode_table_set(struct super_block
*sb
,
143 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
145 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
146 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
147 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
150 void ext4_free_blks_set(struct super_block
*sb
,
151 struct ext4_group_desc
*bg
, __u32 count
)
153 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
154 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
155 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
158 void ext4_free_inodes_set(struct super_block
*sb
,
159 struct ext4_group_desc
*bg
, __u32 count
)
161 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
162 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
163 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
166 void ext4_used_dirs_set(struct super_block
*sb
,
167 struct ext4_group_desc
*bg
, __u32 count
)
169 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
170 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
171 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
174 void ext4_itable_unused_set(struct super_block
*sb
,
175 struct ext4_group_desc
*bg
, __u32 count
)
177 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
178 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
179 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
183 * Wrappers for jbd2_journal_start/end.
185 * The only special thing we need to do here is to make sure that all
186 * journal_end calls result in the superblock being marked dirty, so
187 * that sync() will call the filesystem's write_super callback if
190 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
194 if (sb
->s_flags
& MS_RDONLY
)
195 return ERR_PTR(-EROFS
);
197 /* Special case here: if the journal has aborted behind our
198 * backs (eg. EIO in the commit thread), then we still need to
199 * take the FS itself readonly cleanly. */
200 journal
= EXT4_SB(sb
)->s_journal
;
202 if (is_journal_aborted(journal
)) {
203 ext4_abort(sb
, __func__
,
204 "Detected aborted journal");
205 return ERR_PTR(-EROFS
);
207 return jbd2_journal_start(journal
, nblocks
);
210 * We're not journaling, return the appropriate indication.
212 current
->journal_info
= EXT4_NOJOURNAL_HANDLE
;
213 return current
->journal_info
;
217 * The only special thing we need to do here is to make sure that all
218 * jbd2_journal_stop calls result in the superblock being marked dirty, so
219 * that sync() will call the filesystem's write_super callback if
222 int __ext4_journal_stop(const char *where
, handle_t
*handle
)
224 struct super_block
*sb
;
228 if (!ext4_handle_valid(handle
)) {
230 * Do this here since we don't call jbd2_journal_stop() in
233 current
->journal_info
= NULL
;
236 sb
= handle
->h_transaction
->t_journal
->j_private
;
238 rc
= jbd2_journal_stop(handle
);
243 __ext4_std_error(sb
, where
, err
);
247 void ext4_journal_abort_handle(const char *caller
, const char *err_fn
,
248 struct buffer_head
*bh
, handle_t
*handle
, int err
)
251 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
253 BUG_ON(!ext4_handle_valid(handle
));
256 BUFFER_TRACE(bh
, "abort");
261 if (is_handle_aborted(handle
))
264 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
265 caller
, errstr
, err_fn
);
267 jbd2_journal_abort_handle(handle
);
270 /* Deal with the reporting of failure conditions on a filesystem such as
271 * inconsistencies detected or read IO failures.
273 * On ext2, we can store the error state of the filesystem in the
274 * superblock. That is not possible on ext4, because we may have other
275 * write ordering constraints on the superblock which prevent us from
276 * writing it out straight away; and given that the journal is about to
277 * be aborted, we can't rely on the current, or future, transactions to
278 * write out the superblock safely.
280 * We'll just use the jbd2_journal_abort() error code to record an error in
281 * the journal instead. On recovery, the journal will compain about
282 * that error until we've noted it down and cleared it.
285 static void ext4_handle_error(struct super_block
*sb
)
287 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
289 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
290 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
292 if (sb
->s_flags
& MS_RDONLY
)
295 if (!test_opt(sb
, ERRORS_CONT
)) {
296 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
298 EXT4_SB(sb
)->s_mount_opt
|= EXT4_MOUNT_ABORT
;
300 jbd2_journal_abort(journal
, -EIO
);
302 if (test_opt(sb
, ERRORS_RO
)) {
303 printk(KERN_CRIT
"Remounting filesystem read-only\n");
304 sb
->s_flags
|= MS_RDONLY
;
306 ext4_commit_super(sb
, es
, 1);
307 if (test_opt(sb
, ERRORS_PANIC
))
308 panic("EXT4-fs (device %s): panic forced after error\n",
312 void ext4_error(struct super_block
*sb
, const char *function
,
313 const char *fmt
, ...)
318 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
323 ext4_handle_error(sb
);
326 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
333 errstr
= "IO failure";
336 errstr
= "Out of memory";
339 if (!sb
|| EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
)
340 errstr
= "Journal has aborted";
342 errstr
= "Readonly filesystem";
345 /* If the caller passed in an extra buffer for unknown
346 * errors, textualise them now. Else we just return
349 /* Check for truncated error codes... */
350 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
359 /* __ext4_std_error decodes expected errors from journaling functions
360 * automatically and invokes the appropriate error response. */
362 void __ext4_std_error(struct super_block
*sb
, const char *function
, int errno
)
367 /* Special case: if the error is EROFS, and we're not already
368 * inside a transaction, then there's really no point in logging
370 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
371 (sb
->s_flags
& MS_RDONLY
))
374 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
375 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s: %s\n",
376 sb
->s_id
, function
, errstr
);
378 ext4_handle_error(sb
);
382 * ext4_abort is a much stronger failure handler than ext4_error. The
383 * abort function may be used to deal with unrecoverable failures such
384 * as journal IO errors or ENOMEM at a critical moment in log management.
386 * We unconditionally force the filesystem into an ABORT|READONLY state,
387 * unless the error response on the fs has been set to panic in which
388 * case we take the easy way out and panic immediately.
391 void ext4_abort(struct super_block
*sb
, const char *function
,
392 const char *fmt
, ...)
396 printk(KERN_CRIT
"ext4_abort called.\n");
399 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
404 if (test_opt(sb
, ERRORS_PANIC
))
405 panic("EXT4-fs panic from previous error\n");
407 if (sb
->s_flags
& MS_RDONLY
)
410 printk(KERN_CRIT
"Remounting filesystem read-only\n");
411 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
412 sb
->s_flags
|= MS_RDONLY
;
413 EXT4_SB(sb
)->s_mount_opt
|= EXT4_MOUNT_ABORT
;
414 if (EXT4_SB(sb
)->s_journal
)
415 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
418 void ext4_warning(struct super_block
*sb
, const char *function
,
419 const char *fmt
, ...)
424 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s: ",
431 void ext4_grp_locked_error(struct super_block
*sb
, ext4_group_t grp
,
432 const char *function
, const char *fmt
, ...)
437 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
440 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
445 if (test_opt(sb
, ERRORS_CONT
)) {
446 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
447 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
448 ext4_commit_super(sb
, es
, 0);
451 ext4_unlock_group(sb
, grp
);
452 ext4_handle_error(sb
);
454 * We only get here in the ERRORS_RO case; relocking the group
455 * may be dangerous, but nothing bad will happen since the
456 * filesystem will have already been marked read/only and the
457 * journal has been aborted. We return 1 as a hint to callers
458 * who might what to use the return value from
459 * ext4_grp_locked_error() to distinguish beween the
460 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
461 * aggressively from the ext4 function in question, with a
462 * more appropriate error code.
464 ext4_lock_group(sb
, grp
);
469 void ext4_update_dynamic_rev(struct super_block
*sb
)
471 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
473 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
476 ext4_warning(sb
, __func__
,
477 "updating to rev %d because of new feature flag, "
478 "running e2fsck is recommended",
481 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
482 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
483 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
484 /* leave es->s_feature_*compat flags alone */
485 /* es->s_uuid will be set by e2fsck if empty */
488 * The rest of the superblock fields should be zero, and if not it
489 * means they are likely already in use, so leave them alone. We
490 * can leave it up to e2fsck to clean up any inconsistencies there.
495 * Open the external journal device
497 static struct block_device
*ext4_blkdev_get(dev_t dev
)
499 struct block_device
*bdev
;
500 char b
[BDEVNAME_SIZE
];
502 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
508 printk(KERN_ERR
"EXT4-fs: failed to open journal device %s: %ld\n",
509 __bdevname(dev
, b
), PTR_ERR(bdev
));
514 * Release the journal device
516 static int ext4_blkdev_put(struct block_device
*bdev
)
519 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
522 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
524 struct block_device
*bdev
;
527 bdev
= sbi
->journal_bdev
;
529 ret
= ext4_blkdev_put(bdev
);
530 sbi
->journal_bdev
= NULL
;
535 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
537 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
540 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
544 printk(KERN_ERR
"sb orphan head is %d\n",
545 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
547 printk(KERN_ERR
"sb_info orphan list:\n");
548 list_for_each(l
, &sbi
->s_orphan
) {
549 struct inode
*inode
= orphan_list_entry(l
);
551 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
552 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
553 inode
->i_mode
, inode
->i_nlink
,
558 static void ext4_put_super(struct super_block
*sb
)
560 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
561 struct ext4_super_block
*es
= sbi
->s_es
;
565 ext4_ext_release(sb
);
566 ext4_xattr_put_super(sb
);
567 if (sbi
->s_journal
) {
568 err
= jbd2_journal_destroy(sbi
->s_journal
);
569 sbi
->s_journal
= NULL
;
571 ext4_abort(sb
, __func__
,
572 "Couldn't clean up the journal");
574 if (!(sb
->s_flags
& MS_RDONLY
)) {
575 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
576 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
577 ext4_commit_super(sb
, es
, 1);
580 remove_proc_entry("inode_readahead_blks", sbi
->s_proc
);
581 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
584 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
585 brelse(sbi
->s_group_desc
[i
]);
586 kfree(sbi
->s_group_desc
);
587 kfree(sbi
->s_flex_groups
);
588 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
589 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
590 percpu_counter_destroy(&sbi
->s_dirs_counter
);
591 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
594 for (i
= 0; i
< MAXQUOTAS
; i
++)
595 kfree(sbi
->s_qf_names
[i
]);
598 /* Debugging code just in case the in-memory inode orphan list
599 * isn't empty. The on-disk one can be non-empty if we've
600 * detected an error and taken the fs readonly, but the
601 * in-memory list had better be clean by this point. */
602 if (!list_empty(&sbi
->s_orphan
))
603 dump_orphan_list(sb
, sbi
);
604 J_ASSERT(list_empty(&sbi
->s_orphan
));
606 invalidate_bdev(sb
->s_bdev
);
607 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
609 * Invalidate the journal device's buffers. We don't want them
610 * floating about in memory - the physical journal device may
611 * hotswapped, and it breaks the `ro-after' testing code.
613 sync_blockdev(sbi
->journal_bdev
);
614 invalidate_bdev(sbi
->journal_bdev
);
615 ext4_blkdev_remove(sbi
);
617 sb
->s_fs_info
= NULL
;
622 static struct kmem_cache
*ext4_inode_cachep
;
625 * Called inside transaction, so use GFP_NOFS
627 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
629 struct ext4_inode_info
*ei
;
631 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
634 #ifdef CONFIG_EXT4_FS_POSIX_ACL
635 ei
->i_acl
= EXT4_ACL_NOT_CACHED
;
636 ei
->i_default_acl
= EXT4_ACL_NOT_CACHED
;
638 ei
->vfs_inode
.i_version
= 1;
639 ei
->vfs_inode
.i_data
.writeback_index
= 0;
640 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
641 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
642 spin_lock_init(&ei
->i_prealloc_lock
);
644 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
645 * therefore it can be null here. Don't check it, just initialize
648 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
649 ei
->i_reserved_data_blocks
= 0;
650 ei
->i_reserved_meta_blocks
= 0;
651 ei
->i_allocated_meta_blocks
= 0;
652 ei
->i_delalloc_reserved_flag
= 0;
653 spin_lock_init(&(ei
->i_block_reservation_lock
));
654 return &ei
->vfs_inode
;
657 static void ext4_destroy_inode(struct inode
*inode
)
659 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
660 printk("EXT4 Inode %p: orphan list check failed!\n",
662 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
663 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
667 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
670 static void init_once(void *foo
)
672 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
674 INIT_LIST_HEAD(&ei
->i_orphan
);
675 #ifdef CONFIG_EXT4_FS_XATTR
676 init_rwsem(&ei
->xattr_sem
);
678 init_rwsem(&ei
->i_data_sem
);
679 inode_init_once(&ei
->vfs_inode
);
682 static int init_inodecache(void)
684 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
685 sizeof(struct ext4_inode_info
),
686 0, (SLAB_RECLAIM_ACCOUNT
|
689 if (ext4_inode_cachep
== NULL
)
694 static void destroy_inodecache(void)
696 kmem_cache_destroy(ext4_inode_cachep
);
699 static void ext4_clear_inode(struct inode
*inode
)
701 #ifdef CONFIG_EXT4_FS_POSIX_ACL
702 if (EXT4_I(inode
)->i_acl
&&
703 EXT4_I(inode
)->i_acl
!= EXT4_ACL_NOT_CACHED
) {
704 posix_acl_release(EXT4_I(inode
)->i_acl
);
705 EXT4_I(inode
)->i_acl
= EXT4_ACL_NOT_CACHED
;
707 if (EXT4_I(inode
)->i_default_acl
&&
708 EXT4_I(inode
)->i_default_acl
!= EXT4_ACL_NOT_CACHED
) {
709 posix_acl_release(EXT4_I(inode
)->i_default_acl
);
710 EXT4_I(inode
)->i_default_acl
= EXT4_ACL_NOT_CACHED
;
713 ext4_discard_preallocations(inode
);
714 if (EXT4_JOURNAL(inode
))
715 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
716 &EXT4_I(inode
)->jinode
);
719 static inline void ext4_show_quota_options(struct seq_file
*seq
,
720 struct super_block
*sb
)
722 #if defined(CONFIG_QUOTA)
723 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
725 if (sbi
->s_jquota_fmt
)
726 seq_printf(seq
, ",jqfmt=%s",
727 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold" : "vfsv0");
729 if (sbi
->s_qf_names
[USRQUOTA
])
730 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
732 if (sbi
->s_qf_names
[GRPQUOTA
])
733 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
735 if (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
)
736 seq_puts(seq
, ",usrquota");
738 if (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)
739 seq_puts(seq
, ",grpquota");
745 * - it's set to a non-default value OR
746 * - if the per-sb default is different from the global default
748 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
751 unsigned long def_mount_opts
;
752 struct super_block
*sb
= vfs
->mnt_sb
;
753 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
754 struct ext4_super_block
*es
= sbi
->s_es
;
756 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
757 def_errors
= le16_to_cpu(es
->s_errors
);
759 if (sbi
->s_sb_block
!= 1)
760 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
761 if (test_opt(sb
, MINIX_DF
))
762 seq_puts(seq
, ",minixdf");
763 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
764 seq_puts(seq
, ",grpid");
765 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
766 seq_puts(seq
, ",nogrpid");
767 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
768 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
769 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
771 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
772 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
773 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
775 if (test_opt(sb
, ERRORS_RO
)) {
776 if (def_errors
== EXT4_ERRORS_PANIC
||
777 def_errors
== EXT4_ERRORS_CONTINUE
) {
778 seq_puts(seq
, ",errors=remount-ro");
781 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
782 seq_puts(seq
, ",errors=continue");
783 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
784 seq_puts(seq
, ",errors=panic");
785 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
786 seq_puts(seq
, ",nouid32");
787 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
788 seq_puts(seq
, ",debug");
789 if (test_opt(sb
, OLDALLOC
))
790 seq_puts(seq
, ",oldalloc");
791 #ifdef CONFIG_EXT4_FS_XATTR
792 if (test_opt(sb
, XATTR_USER
) &&
793 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
794 seq_puts(seq
, ",user_xattr");
795 if (!test_opt(sb
, XATTR_USER
) &&
796 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
797 seq_puts(seq
, ",nouser_xattr");
800 #ifdef CONFIG_EXT4_FS_POSIX_ACL
801 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
802 seq_puts(seq
, ",acl");
803 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
804 seq_puts(seq
, ",noacl");
806 if (!test_opt(sb
, RESERVATION
))
807 seq_puts(seq
, ",noreservation");
808 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
809 seq_printf(seq
, ",commit=%u",
810 (unsigned) (sbi
->s_commit_interval
/ HZ
));
812 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
813 seq_printf(seq
, ",min_batch_time=%u",
814 (unsigned) sbi
->s_min_batch_time
);
816 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
817 seq_printf(seq
, ",max_batch_time=%u",
818 (unsigned) sbi
->s_min_batch_time
);
822 * We're changing the default of barrier mount option, so
823 * let's always display its mount state so it's clear what its
826 seq_puts(seq
, ",barrier=");
827 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
828 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
829 seq_puts(seq
, ",journal_async_commit");
830 if (test_opt(sb
, NOBH
))
831 seq_puts(seq
, ",nobh");
832 if (test_opt(sb
, I_VERSION
))
833 seq_puts(seq
, ",i_version");
834 if (!test_opt(sb
, DELALLOC
))
835 seq_puts(seq
, ",nodelalloc");
839 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
841 * journal mode get enabled in different ways
842 * So just print the value even if we didn't specify it
844 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
845 seq_puts(seq
, ",data=journal");
846 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
847 seq_puts(seq
, ",data=ordered");
848 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
849 seq_puts(seq
, ",data=writeback");
851 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
852 seq_printf(seq
, ",inode_readahead_blks=%u",
853 sbi
->s_inode_readahead_blks
);
855 if (test_opt(sb
, DATA_ERR_ABORT
))
856 seq_puts(seq
, ",data_err=abort");
858 ext4_show_quota_options(seq
, sb
);
863 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
864 u64 ino
, u32 generation
)
868 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
869 return ERR_PTR(-ESTALE
);
870 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
871 return ERR_PTR(-ESTALE
);
873 /* iget isn't really right if the inode is currently unallocated!!
875 * ext4_read_inode will return a bad_inode if the inode had been
876 * deleted, so we should be safe.
878 * Currently we don't know the generation for parent directory, so
879 * a generation of 0 means "accept any"
881 inode
= ext4_iget(sb
, ino
);
883 return ERR_CAST(inode
);
884 if (generation
&& inode
->i_generation
!= generation
) {
886 return ERR_PTR(-ESTALE
);
892 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
893 int fh_len
, int fh_type
)
895 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
899 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
900 int fh_len
, int fh_type
)
902 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
907 * Try to release metadata pages (indirect blocks, directories) which are
908 * mapped via the block device. Since these pages could have journal heads
909 * which would prevent try_to_free_buffers() from freeing them, we must use
910 * jbd2 layer's try_to_free_buffers() function to release them.
912 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
, gfp_t wait
)
914 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
916 WARN_ON(PageChecked(page
));
917 if (!page_has_buffers(page
))
920 return jbd2_journal_try_to_free_buffers(journal
, page
,
922 return try_to_free_buffers(page
);
926 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
927 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
929 static int ext4_dquot_initialize(struct inode
*inode
, int type
);
930 static int ext4_dquot_drop(struct inode
*inode
);
931 static int ext4_write_dquot(struct dquot
*dquot
);
932 static int ext4_acquire_dquot(struct dquot
*dquot
);
933 static int ext4_release_dquot(struct dquot
*dquot
);
934 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
935 static int ext4_write_info(struct super_block
*sb
, int type
);
936 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
937 char *path
, int remount
);
938 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
939 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
940 size_t len
, loff_t off
);
941 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
942 const char *data
, size_t len
, loff_t off
);
944 static struct dquot_operations ext4_quota_operations
= {
945 .initialize
= ext4_dquot_initialize
,
946 .drop
= ext4_dquot_drop
,
947 .alloc_space
= dquot_alloc_space
,
948 .alloc_inode
= dquot_alloc_inode
,
949 .free_space
= dquot_free_space
,
950 .free_inode
= dquot_free_inode
,
951 .transfer
= dquot_transfer
,
952 .write_dquot
= ext4_write_dquot
,
953 .acquire_dquot
= ext4_acquire_dquot
,
954 .release_dquot
= ext4_release_dquot
,
955 .mark_dirty
= ext4_mark_dquot_dirty
,
956 .write_info
= ext4_write_info
,
957 .alloc_dquot
= dquot_alloc
,
958 .destroy_dquot
= dquot_destroy
,
961 static struct quotactl_ops ext4_qctl_operations
= {
962 .quota_on
= ext4_quota_on
,
963 .quota_off
= vfs_quota_off
,
964 .quota_sync
= vfs_quota_sync
,
965 .get_info
= vfs_get_dqinfo
,
966 .set_info
= vfs_set_dqinfo
,
967 .get_dqblk
= vfs_get_dqblk
,
968 .set_dqblk
= vfs_set_dqblk
972 static const struct super_operations ext4_sops
= {
973 .alloc_inode
= ext4_alloc_inode
,
974 .destroy_inode
= ext4_destroy_inode
,
975 .write_inode
= ext4_write_inode
,
976 .dirty_inode
= ext4_dirty_inode
,
977 .delete_inode
= ext4_delete_inode
,
978 .put_super
= ext4_put_super
,
979 .write_super
= ext4_write_super
,
980 .sync_fs
= ext4_sync_fs
,
981 .write_super_lockfs
= ext4_write_super_lockfs
,
982 .unlockfs
= ext4_unlockfs
,
983 .statfs
= ext4_statfs
,
984 .remount_fs
= ext4_remount
,
985 .clear_inode
= ext4_clear_inode
,
986 .show_options
= ext4_show_options
,
988 .quota_read
= ext4_quota_read
,
989 .quota_write
= ext4_quota_write
,
991 .bdev_try_to_free_page
= bdev_try_to_free_page
,
994 static const struct export_operations ext4_export_ops
= {
995 .fh_to_dentry
= ext4_fh_to_dentry
,
996 .fh_to_parent
= ext4_fh_to_parent
,
997 .get_parent
= ext4_get_parent
,
1001 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1002 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1003 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1004 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1005 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
1006 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1007 Opt_journal_update
, Opt_journal_dev
,
1008 Opt_journal_checksum
, Opt_journal_async_commit
,
1009 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1010 Opt_data_err_abort
, Opt_data_err_ignore
,
1011 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1012 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
1013 Opt_ignore
, Opt_barrier
, Opt_err
, Opt_resize
, Opt_usrquota
,
1014 Opt_grpquota
, Opt_i_version
,
1015 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
1016 Opt_inode_readahead_blks
, Opt_journal_ioprio
1019 static const match_table_t tokens
= {
1020 {Opt_bsd_df
, "bsddf"},
1021 {Opt_minix_df
, "minixdf"},
1022 {Opt_grpid
, "grpid"},
1023 {Opt_grpid
, "bsdgroups"},
1024 {Opt_nogrpid
, "nogrpid"},
1025 {Opt_nogrpid
, "sysvgroups"},
1026 {Opt_resgid
, "resgid=%u"},
1027 {Opt_resuid
, "resuid=%u"},
1029 {Opt_err_cont
, "errors=continue"},
1030 {Opt_err_panic
, "errors=panic"},
1031 {Opt_err_ro
, "errors=remount-ro"},
1032 {Opt_nouid32
, "nouid32"},
1033 {Opt_debug
, "debug"},
1034 {Opt_oldalloc
, "oldalloc"},
1035 {Opt_orlov
, "orlov"},
1036 {Opt_user_xattr
, "user_xattr"},
1037 {Opt_nouser_xattr
, "nouser_xattr"},
1039 {Opt_noacl
, "noacl"},
1040 {Opt_reservation
, "reservation"},
1041 {Opt_noreservation
, "noreservation"},
1042 {Opt_noload
, "noload"},
1045 {Opt_commit
, "commit=%u"},
1046 {Opt_min_batch_time
, "min_batch_time=%u"},
1047 {Opt_max_batch_time
, "max_batch_time=%u"},
1048 {Opt_journal_update
, "journal=update"},
1049 {Opt_journal_dev
, "journal_dev=%u"},
1050 {Opt_journal_checksum
, "journal_checksum"},
1051 {Opt_journal_async_commit
, "journal_async_commit"},
1052 {Opt_abort
, "abort"},
1053 {Opt_data_journal
, "data=journal"},
1054 {Opt_data_ordered
, "data=ordered"},
1055 {Opt_data_writeback
, "data=writeback"},
1056 {Opt_data_err_abort
, "data_err=abort"},
1057 {Opt_data_err_ignore
, "data_err=ignore"},
1058 {Opt_offusrjquota
, "usrjquota="},
1059 {Opt_usrjquota
, "usrjquota=%s"},
1060 {Opt_offgrpjquota
, "grpjquota="},
1061 {Opt_grpjquota
, "grpjquota=%s"},
1062 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1063 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1064 {Opt_grpquota
, "grpquota"},
1065 {Opt_noquota
, "noquota"},
1066 {Opt_quota
, "quota"},
1067 {Opt_usrquota
, "usrquota"},
1068 {Opt_barrier
, "barrier=%u"},
1069 {Opt_i_version
, "i_version"},
1070 {Opt_stripe
, "stripe=%u"},
1071 {Opt_resize
, "resize"},
1072 {Opt_delalloc
, "delalloc"},
1073 {Opt_nodelalloc
, "nodelalloc"},
1074 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1075 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1079 static ext4_fsblk_t
get_sb_block(void **data
)
1081 ext4_fsblk_t sb_block
;
1082 char *options
= (char *) *data
;
1084 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1085 return 1; /* Default location */
1087 /*todo: use simple_strtoll with >32bit ext4 */
1088 sb_block
= simple_strtoul(options
, &options
, 0);
1089 if (*options
&& *options
!= ',') {
1090 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1094 if (*options
== ',')
1096 *data
= (void *) options
;
1100 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1102 static int parse_options(char *options
, struct super_block
*sb
,
1103 unsigned long *journal_devnum
,
1104 unsigned int *journal_ioprio
,
1105 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1107 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1109 substring_t args
[MAX_OPT_ARGS
];
1120 while ((p
= strsep(&options
, ",")) != NULL
) {
1125 token
= match_token(p
, tokens
, args
);
1128 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1131 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1134 set_opt(sbi
->s_mount_opt
, GRPID
);
1137 clear_opt(sbi
->s_mount_opt
, GRPID
);
1140 if (match_int(&args
[0], &option
))
1142 sbi
->s_resuid
= option
;
1145 if (match_int(&args
[0], &option
))
1147 sbi
->s_resgid
= option
;
1150 /* handled by get_sb_block() instead of here */
1151 /* *sb_block = match_int(&args[0]); */
1154 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1155 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1156 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1159 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1160 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1161 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1164 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1165 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1166 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1169 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1172 set_opt(sbi
->s_mount_opt
, DEBUG
);
1175 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1178 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1180 #ifdef CONFIG_EXT4_FS_XATTR
1181 case Opt_user_xattr
:
1182 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1184 case Opt_nouser_xattr
:
1185 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1188 case Opt_user_xattr
:
1189 case Opt_nouser_xattr
:
1190 printk(KERN_ERR
"EXT4 (no)user_xattr options "
1194 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1196 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1199 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1204 printk(KERN_ERR
"EXT4 (no)acl options "
1208 case Opt_reservation
:
1209 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1211 case Opt_noreservation
:
1212 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
1214 case Opt_journal_update
:
1216 /* Eventually we will want to be able to create
1217 a journal file here. For now, only allow the
1218 user to specify an existing inode to be the
1221 printk(KERN_ERR
"EXT4-fs: cannot specify "
1222 "journal on remount\n");
1225 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1227 case Opt_journal_dev
:
1229 printk(KERN_ERR
"EXT4-fs: cannot specify "
1230 "journal on remount\n");
1233 if (match_int(&args
[0], &option
))
1235 *journal_devnum
= option
;
1237 case Opt_journal_checksum
:
1238 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1240 case Opt_journal_async_commit
:
1241 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1242 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1245 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1248 if (match_int(&args
[0], &option
))
1253 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1254 sbi
->s_commit_interval
= HZ
* option
;
1256 case Opt_max_batch_time
:
1257 if (match_int(&args
[0], &option
))
1262 option
= EXT4_DEF_MAX_BATCH_TIME
;
1263 sbi
->s_max_batch_time
= option
;
1265 case Opt_min_batch_time
:
1266 if (match_int(&args
[0], &option
))
1270 sbi
->s_min_batch_time
= option
;
1272 case Opt_data_journal
:
1273 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1275 case Opt_data_ordered
:
1276 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1278 case Opt_data_writeback
:
1279 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1282 if ((sbi
->s_mount_opt
& EXT4_MOUNT_DATA_FLAGS
)
1285 "EXT4-fs: cannot change data "
1286 "mode on remount\n");
1290 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DATA_FLAGS
;
1291 sbi
->s_mount_opt
|= data_opt
;
1294 case Opt_data_err_abort
:
1295 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1297 case Opt_data_err_ignore
:
1298 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1307 if (sb_any_quota_loaded(sb
) &&
1308 !sbi
->s_qf_names
[qtype
]) {
1310 "EXT4-fs: Cannot change journaled "
1311 "quota options when quota turned on.\n");
1314 qname
= match_strdup(&args
[0]);
1317 "EXT4-fs: not enough memory for "
1318 "storing quotafile name.\n");
1321 if (sbi
->s_qf_names
[qtype
] &&
1322 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1324 "EXT4-fs: %s quota file already "
1325 "specified.\n", QTYPE2NAME(qtype
));
1329 sbi
->s_qf_names
[qtype
] = qname
;
1330 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1332 "EXT4-fs: quotafile must be on "
1333 "filesystem root.\n");
1334 kfree(sbi
->s_qf_names
[qtype
]);
1335 sbi
->s_qf_names
[qtype
] = NULL
;
1338 set_opt(sbi
->s_mount_opt
, QUOTA
);
1340 case Opt_offusrjquota
:
1343 case Opt_offgrpjquota
:
1346 if (sb_any_quota_loaded(sb
) &&
1347 sbi
->s_qf_names
[qtype
]) {
1348 printk(KERN_ERR
"EXT4-fs: Cannot change "
1349 "journaled quota options when "
1350 "quota turned on.\n");
1354 * The space will be released later when all options
1355 * are confirmed to be correct
1357 sbi
->s_qf_names
[qtype
] = NULL
;
1359 case Opt_jqfmt_vfsold
:
1360 qfmt
= QFMT_VFS_OLD
;
1362 case Opt_jqfmt_vfsv0
:
1365 if (sb_any_quota_loaded(sb
) &&
1366 sbi
->s_jquota_fmt
!= qfmt
) {
1367 printk(KERN_ERR
"EXT4-fs: Cannot change "
1368 "journaled quota options when "
1369 "quota turned on.\n");
1372 sbi
->s_jquota_fmt
= qfmt
;
1376 set_opt(sbi
->s_mount_opt
, QUOTA
);
1377 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1380 set_opt(sbi
->s_mount_opt
, QUOTA
);
1381 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1384 if (sb_any_quota_loaded(sb
)) {
1385 printk(KERN_ERR
"EXT4-fs: Cannot change quota "
1386 "options when quota turned on.\n");
1389 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1390 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1391 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1398 "EXT4-fs: quota options not supported.\n");
1402 case Opt_offusrjquota
:
1403 case Opt_offgrpjquota
:
1404 case Opt_jqfmt_vfsold
:
1405 case Opt_jqfmt_vfsv0
:
1407 "EXT4-fs: journaled quota options not "
1414 set_opt(sbi
->s_mount_opt
, ABORT
);
1417 if (match_int(&args
[0], &option
))
1420 set_opt(sbi
->s_mount_opt
, BARRIER
);
1422 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1428 printk("EXT4-fs: resize option only available "
1432 if (match_int(&args
[0], &option
) != 0)
1434 *n_blocks_count
= option
;
1437 set_opt(sbi
->s_mount_opt
, NOBH
);
1440 clear_opt(sbi
->s_mount_opt
, NOBH
);
1443 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1444 sb
->s_flags
|= MS_I_VERSION
;
1446 case Opt_nodelalloc
:
1447 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1450 if (match_int(&args
[0], &option
))
1454 sbi
->s_stripe
= option
;
1457 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1459 case Opt_inode_readahead_blks
:
1460 if (match_int(&args
[0], &option
))
1462 if (option
< 0 || option
> (1 << 30))
1464 sbi
->s_inode_readahead_blks
= option
;
1466 case Opt_journal_ioprio
:
1467 if (match_int(&args
[0], &option
))
1469 if (option
< 0 || option
> 7)
1471 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1476 "EXT4-fs: Unrecognized mount option \"%s\" "
1477 "or missing value\n", p
);
1482 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1483 if ((sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
) &&
1484 sbi
->s_qf_names
[USRQUOTA
])
1485 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1487 if ((sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
) &&
1488 sbi
->s_qf_names
[GRPQUOTA
])
1489 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1491 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1492 (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)) ||
1493 (sbi
->s_qf_names
[GRPQUOTA
] &&
1494 (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
))) {
1495 printk(KERN_ERR
"EXT4-fs: old and new quota "
1496 "format mixing.\n");
1500 if (!sbi
->s_jquota_fmt
) {
1501 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1502 "not specified.\n");
1506 if (sbi
->s_jquota_fmt
) {
1507 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1508 "specified with no journaling "
1517 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1520 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1523 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1524 printk(KERN_ERR
"EXT4-fs warning: revision level too high, "
1525 "forcing read-only mode\n");
1530 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1531 printk(KERN_WARNING
"EXT4-fs warning: mounting unchecked fs, "
1532 "running e2fsck is recommended\n");
1533 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1535 "EXT4-fs warning: mounting fs with errors, "
1536 "running e2fsck is recommended\n");
1537 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1538 le16_to_cpu(es
->s_mnt_count
) >=
1539 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1541 "EXT4-fs warning: maximal mount count reached, "
1542 "running e2fsck is recommended\n");
1543 else if (le32_to_cpu(es
->s_checkinterval
) &&
1544 (le32_to_cpu(es
->s_lastcheck
) +
1545 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1547 "EXT4-fs warning: checktime reached, "
1548 "running e2fsck is recommended\n");
1549 if (!sbi
->s_journal
)
1550 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1551 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1552 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1553 le16_add_cpu(&es
->s_mnt_count
, 1);
1554 es
->s_mtime
= cpu_to_le32(get_seconds());
1555 ext4_update_dynamic_rev(sb
);
1557 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1559 ext4_commit_super(sb
, es
, 1);
1560 if (test_opt(sb
, DEBUG
))
1561 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1562 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1564 sbi
->s_groups_count
,
1565 EXT4_BLOCKS_PER_GROUP(sb
),
1566 EXT4_INODES_PER_GROUP(sb
),
1569 if (EXT4_SB(sb
)->s_journal
) {
1570 printk(KERN_INFO
"EXT4 FS on %s, %s journal on %s\n",
1571 sb
->s_id
, EXT4_SB(sb
)->s_journal
->j_inode
? "internal" :
1572 "external", EXT4_SB(sb
)->s_journal
->j_devname
);
1574 printk(KERN_INFO
"EXT4 FS on %s, no journal\n", sb
->s_id
);
1579 static int ext4_fill_flex_info(struct super_block
*sb
)
1581 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1582 struct ext4_group_desc
*gdp
= NULL
;
1583 struct buffer_head
*bh
;
1584 ext4_group_t flex_group_count
;
1585 ext4_group_t flex_group
;
1586 int groups_per_flex
= 0;
1589 if (!sbi
->s_es
->s_log_groups_per_flex
) {
1590 sbi
->s_log_groups_per_flex
= 0;
1594 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1595 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1597 /* We allocate both existing and potentially added groups */
1598 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1599 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1600 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1601 sbi
->s_flex_groups
= kzalloc(flex_group_count
*
1602 sizeof(struct flex_groups
), GFP_KERNEL
);
1603 if (sbi
->s_flex_groups
== NULL
) {
1604 printk(KERN_ERR
"EXT4-fs: not enough memory for "
1605 "%u flex groups\n", flex_group_count
);
1609 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1610 gdp
= ext4_get_group_desc(sb
, i
, &bh
);
1612 flex_group
= ext4_flex_group(sbi
, i
);
1613 sbi
->s_flex_groups
[flex_group
].free_inodes
+=
1614 ext4_free_inodes_count(sb
, gdp
);
1615 sbi
->s_flex_groups
[flex_group
].free_blocks
+=
1616 ext4_free_blks_count(sb
, gdp
);
1624 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1625 struct ext4_group_desc
*gdp
)
1629 if (sbi
->s_es
->s_feature_ro_compat
&
1630 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1631 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1632 __le32 le_group
= cpu_to_le32(block_group
);
1634 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1635 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1636 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1637 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1638 /* for checksum of struct ext4_group_desc do the rest...*/
1639 if ((sbi
->s_es
->s_feature_incompat
&
1640 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1641 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1642 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1643 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1647 return cpu_to_le16(crc
);
1650 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1651 struct ext4_group_desc
*gdp
)
1653 if ((sbi
->s_es
->s_feature_ro_compat
&
1654 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1655 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1661 /* Called at mount-time, super-block is locked */
1662 static int ext4_check_descriptors(struct super_block
*sb
)
1664 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1665 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1666 ext4_fsblk_t last_block
;
1667 ext4_fsblk_t block_bitmap
;
1668 ext4_fsblk_t inode_bitmap
;
1669 ext4_fsblk_t inode_table
;
1670 int flexbg_flag
= 0;
1673 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1676 ext4_debug("Checking group descriptors");
1678 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1679 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1681 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1682 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1684 last_block
= first_block
+
1685 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1687 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1688 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1689 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1690 "Block bitmap for group %u not in group "
1691 "(block %llu)!\n", i
, block_bitmap
);
1694 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1695 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1696 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1697 "Inode bitmap for group %u not in group "
1698 "(block %llu)!\n", i
, inode_bitmap
);
1701 inode_table
= ext4_inode_table(sb
, gdp
);
1702 if (inode_table
< first_block
||
1703 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1704 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1705 "Inode table for group %u not in group "
1706 "(block %llu)!\n", i
, inode_table
);
1709 spin_lock(sb_bgl_lock(sbi
, i
));
1710 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1711 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1712 "Checksum for group %u failed (%u!=%u)\n",
1713 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1714 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1715 if (!(sb
->s_flags
& MS_RDONLY
)) {
1716 spin_unlock(sb_bgl_lock(sbi
, i
));
1720 spin_unlock(sb_bgl_lock(sbi
, i
));
1722 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1725 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1726 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(ext4_count_free_inodes(sb
));
1730 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1731 * the superblock) which were deleted from all directories, but held open by
1732 * a process at the time of a crash. We walk the list and try to delete these
1733 * inodes at recovery time (only with a read-write filesystem).
1735 * In order to keep the orphan inode chain consistent during traversal (in
1736 * case of crash during recovery), we link each inode into the superblock
1737 * orphan list_head and handle it the same way as an inode deletion during
1738 * normal operation (which journals the operations for us).
1740 * We only do an iget() and an iput() on each inode, which is very safe if we
1741 * accidentally point at an in-use or already deleted inode. The worst that
1742 * can happen in this case is that we get a "bit already cleared" message from
1743 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1744 * e2fsck was run on this filesystem, and it must have already done the orphan
1745 * inode cleanup for us, so we can safely abort without any further action.
1747 static void ext4_orphan_cleanup(struct super_block
*sb
,
1748 struct ext4_super_block
*es
)
1750 unsigned int s_flags
= sb
->s_flags
;
1751 int nr_orphans
= 0, nr_truncates
= 0;
1755 if (!es
->s_last_orphan
) {
1756 jbd_debug(4, "no orphan inodes to clean up\n");
1760 if (bdev_read_only(sb
->s_bdev
)) {
1761 printk(KERN_ERR
"EXT4-fs: write access "
1762 "unavailable, skipping orphan cleanup.\n");
1766 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1767 if (es
->s_last_orphan
)
1768 jbd_debug(1, "Errors on filesystem, "
1769 "clearing orphan list.\n");
1770 es
->s_last_orphan
= 0;
1771 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1775 if (s_flags
& MS_RDONLY
) {
1776 printk(KERN_INFO
"EXT4-fs: %s: orphan cleanup on readonly fs\n",
1778 sb
->s_flags
&= ~MS_RDONLY
;
1781 /* Needed for iput() to work correctly and not trash data */
1782 sb
->s_flags
|= MS_ACTIVE
;
1783 /* Turn on quotas so that they are updated correctly */
1784 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1785 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
1786 int ret
= ext4_quota_on_mount(sb
, i
);
1789 "EXT4-fs: Cannot turn on journaled "
1790 "quota: error %d\n", ret
);
1795 while (es
->s_last_orphan
) {
1796 struct inode
*inode
;
1798 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1799 if (IS_ERR(inode
)) {
1800 es
->s_last_orphan
= 0;
1804 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
1806 if (inode
->i_nlink
) {
1808 "%s: truncating inode %lu to %lld bytes\n",
1809 __func__
, inode
->i_ino
, inode
->i_size
);
1810 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1811 inode
->i_ino
, inode
->i_size
);
1812 ext4_truncate(inode
);
1816 "%s: deleting unreferenced inode %lu\n",
1817 __func__
, inode
->i_ino
);
1818 jbd_debug(2, "deleting unreferenced inode %lu\n",
1822 iput(inode
); /* The delete magic happens here! */
1825 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1828 printk(KERN_INFO
"EXT4-fs: %s: %d orphan inode%s deleted\n",
1829 sb
->s_id
, PLURAL(nr_orphans
));
1831 printk(KERN_INFO
"EXT4-fs: %s: %d truncate%s cleaned up\n",
1832 sb
->s_id
, PLURAL(nr_truncates
));
1834 /* Turn quotas off */
1835 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1836 if (sb_dqopt(sb
)->files
[i
])
1837 vfs_quota_off(sb
, i
, 0);
1840 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1843 * Maximal extent format file size.
1844 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1845 * extent format containers, within a sector_t, and within i_blocks
1846 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1847 * so that won't be a limiting factor.
1849 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1851 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
1854 loff_t upper_limit
= MAX_LFS_FILESIZE
;
1856 /* small i_blocks in vfs inode? */
1857 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
1859 * CONFIG_LBD is not enabled implies the inode
1860 * i_block represent total blocks in 512 bytes
1861 * 32 == size of vfs inode i_blocks * 8
1863 upper_limit
= (1LL << 32) - 1;
1865 /* total blocks in file system block size */
1866 upper_limit
>>= (blkbits
- 9);
1867 upper_limit
<<= blkbits
;
1870 /* 32-bit extent-start container, ee_block */
1875 /* Sanity check against vm- & vfs- imposed limits */
1876 if (res
> upper_limit
)
1883 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1884 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1885 * We need to be 1 filesystem block less than the 2^48 sector limit.
1887 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
1889 loff_t res
= EXT4_NDIR_BLOCKS
;
1892 /* This is calculated to be the largest file size for a
1893 * dense, bitmapped file such that the total number of
1894 * sectors in the file, including data and all indirect blocks,
1895 * does not exceed 2^48 -1
1896 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1897 * total number of 512 bytes blocks of the file
1900 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
1902 * !has_huge_files or CONFIG_LBD is not enabled
1903 * implies the inode i_block represent total blocks in
1904 * 512 bytes 32 == size of vfs inode i_blocks * 8
1906 upper_limit
= (1LL << 32) - 1;
1908 /* total blocks in file system block size */
1909 upper_limit
>>= (bits
- 9);
1913 * We use 48 bit ext4_inode i_blocks
1914 * With EXT4_HUGE_FILE_FL set the i_blocks
1915 * represent total number of blocks in
1916 * file system block size
1918 upper_limit
= (1LL << 48) - 1;
1922 /* indirect blocks */
1924 /* double indirect blocks */
1925 meta_blocks
+= 1 + (1LL << (bits
-2));
1926 /* tripple indirect blocks */
1927 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
1929 upper_limit
-= meta_blocks
;
1930 upper_limit
<<= bits
;
1932 res
+= 1LL << (bits
-2);
1933 res
+= 1LL << (2*(bits
-2));
1934 res
+= 1LL << (3*(bits
-2));
1936 if (res
> upper_limit
)
1939 if (res
> MAX_LFS_FILESIZE
)
1940 res
= MAX_LFS_FILESIZE
;
1945 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
1946 ext4_fsblk_t logical_sb_block
, int nr
)
1948 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1949 ext4_group_t bg
, first_meta_bg
;
1952 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1954 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
1956 return logical_sb_block
+ nr
+ 1;
1957 bg
= sbi
->s_desc_per_block
* nr
;
1958 if (ext4_bg_has_super(sb
, bg
))
1960 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
1964 * ext4_get_stripe_size: Get the stripe size.
1965 * @sbi: In memory super block info
1967 * If we have specified it via mount option, then
1968 * use the mount option value. If the value specified at mount time is
1969 * greater than the blocks per group use the super block value.
1970 * If the super block value is greater than blocks per group return 0.
1971 * Allocator needs it be less than blocks per group.
1974 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
1976 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
1977 unsigned long stripe_width
=
1978 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
1980 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
1981 return sbi
->s_stripe
;
1983 if (stripe_width
<= sbi
->s_blocks_per_group
)
1984 return stripe_width
;
1986 if (stride
<= sbi
->s_blocks_per_group
)
1992 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
1993 __releases(kernel_lock
)
1994 __acquires(kernel_lock
)
1997 struct buffer_head
*bh
;
1998 struct ext4_super_block
*es
= NULL
;
1999 struct ext4_sb_info
*sbi
;
2001 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2002 ext4_fsblk_t logical_sb_block
;
2003 unsigned long offset
= 0;
2004 unsigned long journal_devnum
= 0;
2005 unsigned long def_mount_opts
;
2011 unsigned int db_count
;
2013 int needs_recovery
, has_huge_files
;
2017 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2019 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
2022 sb
->s_fs_info
= sbi
;
2023 sbi
->s_mount_opt
= 0;
2024 sbi
->s_resuid
= EXT4_DEF_RESUID
;
2025 sbi
->s_resgid
= EXT4_DEF_RESGID
;
2026 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
2027 sbi
->s_sb_block
= sb_block
;
2031 /* Cleanup superblock name */
2032 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
2035 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
2037 printk(KERN_ERR
"EXT4-fs: unable to set blocksize\n");
2042 * The ext4 superblock will not be buffer aligned for other than 1kB
2043 * block sizes. We need to calculate the offset from buffer start.
2045 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
2046 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2047 offset
= do_div(logical_sb_block
, blocksize
);
2049 logical_sb_block
= sb_block
;
2052 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
2053 printk(KERN_ERR
"EXT4-fs: unable to read superblock\n");
2057 * Note: s_es must be initialized as soon as possible because
2058 * some ext4 macro-instructions depend on its value
2060 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2062 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
2063 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
2066 /* Set defaults before we parse the mount options */
2067 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
2068 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
2069 set_opt(sbi
->s_mount_opt
, DEBUG
);
2070 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
2071 set_opt(sbi
->s_mount_opt
, GRPID
);
2072 if (def_mount_opts
& EXT4_DEFM_UID16
)
2073 set_opt(sbi
->s_mount_opt
, NO_UID32
);
2074 #ifdef CONFIG_EXT4_FS_XATTR
2075 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
2076 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
2078 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2079 if (def_mount_opts
& EXT4_DEFM_ACL
)
2080 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
2082 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
2083 sbi
->s_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
2084 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
2085 sbi
->s_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
2086 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
2087 sbi
->s_mount_opt
|= EXT4_MOUNT_WRITEBACK_DATA
;
2089 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2090 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2091 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2092 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2094 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2096 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2097 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2098 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
2099 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
2100 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
2102 set_opt(sbi
->s_mount_opt
, RESERVATION
);
2103 set_opt(sbi
->s_mount_opt
, BARRIER
);
2106 * enable delayed allocation by default
2107 * Use -o nodelalloc to turn it off
2109 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2112 if (!parse_options((char *) data
, sb
, &journal_devnum
,
2113 &journal_ioprio
, NULL
, 0))
2116 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2117 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2119 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2120 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2121 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2122 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2124 "EXT4-fs warning: feature flags set on rev 0 fs, "
2125 "running e2fsck is recommended\n");
2128 * Check feature flags regardless of the revision level, since we
2129 * previously didn't change the revision level when setting the flags,
2130 * so there is a chance incompat flags are set on a rev 0 filesystem.
2132 features
= EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
);
2134 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount because of "
2135 "unsupported optional features (%x).\n", sb
->s_id
,
2136 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2137 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2140 features
= EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
);
2141 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
2142 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount RDWR because of "
2143 "unsupported optional features (%x).\n", sb
->s_id
,
2144 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2145 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2148 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2149 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
2150 if (has_huge_files
) {
2152 * Large file size enabled file system can only be
2153 * mount if kernel is build with CONFIG_LBD
2155 if (sizeof(root
->i_blocks
) < sizeof(u64
) &&
2156 !(sb
->s_flags
& MS_RDONLY
)) {
2157 printk(KERN_ERR
"EXT4-fs: %s: Filesystem with huge "
2158 "files cannot be mounted read-write "
2159 "without CONFIG_LBD.\n", sb
->s_id
);
2163 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2165 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2166 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2168 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2169 blocksize
, sb
->s_id
);
2173 if (sb
->s_blocksize
!= blocksize
) {
2175 /* Validate the filesystem blocksize */
2176 if (!sb_set_blocksize(sb
, blocksize
)) {
2177 printk(KERN_ERR
"EXT4-fs: bad block size %d.\n",
2183 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2184 offset
= do_div(logical_sb_block
, blocksize
);
2185 bh
= sb_bread(sb
, logical_sb_block
);
2188 "EXT4-fs: Can't read superblock on 2nd try.\n");
2191 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2193 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2195 "EXT4-fs: Magic mismatch, very weird !\n");
2200 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
2202 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
2204 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2205 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2206 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2208 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2209 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2210 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2211 (!is_power_of_2(sbi
->s_inode_size
)) ||
2212 (sbi
->s_inode_size
> blocksize
)) {
2214 "EXT4-fs: unsupported inode size: %d\n",
2218 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2219 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2221 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2222 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2223 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2224 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2225 !is_power_of_2(sbi
->s_desc_size
)) {
2227 "EXT4-fs: unsupported descriptor size %lu\n",
2232 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2233 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2234 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2235 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2237 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2238 if (sbi
->s_inodes_per_block
== 0)
2240 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2241 sbi
->s_inodes_per_block
;
2242 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2244 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2245 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2246 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2247 for (i
= 0; i
< 4; i
++)
2248 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2249 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2250 i
= le32_to_cpu(es
->s_flags
);
2251 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
2252 sbi
->s_hash_unsigned
= 3;
2253 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
2254 #ifdef __CHAR_UNSIGNED__
2255 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
2256 sbi
->s_hash_unsigned
= 3;
2258 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
2263 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2265 "EXT4-fs: #blocks per group too big: %lu\n",
2266 sbi
->s_blocks_per_group
);
2269 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2271 "EXT4-fs: #inodes per group too big: %lu\n",
2272 sbi
->s_inodes_per_group
);
2276 if (ext4_blocks_count(es
) >
2277 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) {
2278 printk(KERN_ERR
"EXT4-fs: filesystem on %s:"
2279 " too large to mount safely\n", sb
->s_id
);
2280 if (sizeof(sector_t
) < 8)
2281 printk(KERN_WARNING
"EXT4-fs: CONFIG_LBD not "
2286 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2290 * It makes no sense for the first data block to be beyond the end
2291 * of the filesystem.
2293 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
2294 printk(KERN_WARNING
"EXT4-fs: bad geometry: first data"
2295 "block %u is beyond end of filesystem (%llu)\n",
2296 le32_to_cpu(es
->s_first_data_block
),
2297 ext4_blocks_count(es
));
2300 blocks_count
= (ext4_blocks_count(es
) -
2301 le32_to_cpu(es
->s_first_data_block
) +
2302 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2303 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2304 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
2305 printk(KERN_WARNING
"EXT4-fs: groups count too large: %u "
2306 "(block count %llu, first data block %u, "
2307 "blocks per group %lu)\n", sbi
->s_groups_count
,
2308 ext4_blocks_count(es
),
2309 le32_to_cpu(es
->s_first_data_block
),
2310 EXT4_BLOCKS_PER_GROUP(sb
));
2313 sbi
->s_groups_count
= blocks_count
;
2314 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2315 EXT4_DESC_PER_BLOCK(sb
);
2316 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2318 if (sbi
->s_group_desc
== NULL
) {
2319 printk(KERN_ERR
"EXT4-fs: not enough memory\n");
2323 #ifdef CONFIG_PROC_FS
2325 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2328 proc_create_data("inode_readahead_blks", 0644, sbi
->s_proc
,
2330 &sbi
->s_inode_readahead_blks
);
2333 bgl_lock_init(&sbi
->s_blockgroup_lock
);
2335 for (i
= 0; i
< db_count
; i
++) {
2336 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2337 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2338 if (!sbi
->s_group_desc
[i
]) {
2339 printk(KERN_ERR
"EXT4-fs: "
2340 "can't read group descriptor %d\n", i
);
2345 if (!ext4_check_descriptors(sb
)) {
2346 printk(KERN_ERR
"EXT4-fs: group descriptors corrupted!\n");
2349 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2350 if (!ext4_fill_flex_info(sb
)) {
2352 "EXT4-fs: unable to initialize "
2353 "flex_bg meta info!\n");
2357 sbi
->s_gdb_count
= db_count
;
2358 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2359 spin_lock_init(&sbi
->s_next_gen_lock
);
2361 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2362 ext4_count_free_blocks(sb
));
2364 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2365 ext4_count_free_inodes(sb
));
2368 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2369 ext4_count_dirs(sb
));
2372 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
2375 printk(KERN_ERR
"EXT4-fs: insufficient memory\n");
2379 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2382 * set up enough so that it can read an inode
2384 sb
->s_op
= &ext4_sops
;
2385 sb
->s_export_op
= &ext4_export_ops
;
2386 sb
->s_xattr
= ext4_xattr_handlers
;
2388 sb
->s_qcop
= &ext4_qctl_operations
;
2389 sb
->dq_op
= &ext4_quota_operations
;
2391 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2395 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2396 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2397 EXT4_FEATURE_INCOMPAT_RECOVER
));
2400 * The first inode we look at is the journal inode. Don't try
2401 * root first: it may be modified in the journal!
2403 if (!test_opt(sb
, NOLOAD
) &&
2404 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2405 if (ext4_load_journal(sb
, es
, journal_devnum
))
2407 if (!(sb
->s_flags
& MS_RDONLY
) &&
2408 EXT4_SB(sb
)->s_journal
->j_failed_commit
) {
2409 printk(KERN_CRIT
"EXT4-fs error (device %s): "
2410 "ext4_fill_super: Journal transaction "
2411 "%u is corrupt\n", sb
->s_id
,
2412 EXT4_SB(sb
)->s_journal
->j_failed_commit
);
2413 if (test_opt(sb
, ERRORS_RO
)) {
2415 "Mounting filesystem read-only\n");
2416 sb
->s_flags
|= MS_RDONLY
;
2417 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2418 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2420 if (test_opt(sb
, ERRORS_PANIC
)) {
2421 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2422 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2423 ext4_commit_super(sb
, es
, 1);
2427 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
2428 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2429 printk(KERN_ERR
"EXT4-fs: required journal recovery "
2430 "suppressed and not mounted read-only\n");
2433 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
2434 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2435 sbi
->s_journal
= NULL
;
2440 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2441 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2442 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2443 printk(KERN_ERR
"EXT4-fs: Failed to set 64-bit journal feature\n");
2447 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2448 jbd2_journal_set_features(sbi
->s_journal
,
2449 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2450 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2451 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2452 jbd2_journal_set_features(sbi
->s_journal
,
2453 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2454 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2455 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2457 jbd2_journal_clear_features(sbi
->s_journal
,
2458 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2459 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2462 /* We have now updated the journal if required, so we can
2463 * validate the data journaling mode. */
2464 switch (test_opt(sb
, DATA_FLAGS
)) {
2466 /* No mode set, assume a default based on the journal
2467 * capabilities: ORDERED_DATA if the journal can
2468 * cope, else JOURNAL_DATA
2470 if (jbd2_journal_check_available_features
2471 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2472 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2474 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2477 case EXT4_MOUNT_ORDERED_DATA
:
2478 case EXT4_MOUNT_WRITEBACK_DATA
:
2479 if (!jbd2_journal_check_available_features
2480 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2481 printk(KERN_ERR
"EXT4-fs: Journal does not support "
2482 "requested data journaling mode\n");
2488 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
2492 if (test_opt(sb
, NOBH
)) {
2493 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2494 printk(KERN_WARNING
"EXT4-fs: Ignoring nobh option - "
2495 "its supported only with writeback mode\n");
2496 clear_opt(sbi
->s_mount_opt
, NOBH
);
2500 * The jbd2_journal_load will have done any necessary log recovery,
2501 * so we can safely mount the rest of the filesystem now.
2504 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2506 printk(KERN_ERR
"EXT4-fs: get root inode failed\n");
2507 ret
= PTR_ERR(root
);
2510 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2512 printk(KERN_ERR
"EXT4-fs: corrupt root inode, run e2fsck\n");
2515 sb
->s_root
= d_alloc_root(root
);
2517 printk(KERN_ERR
"EXT4-fs: get root dentry failed\n");
2523 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
2525 /* determine the minimum size of new large inodes, if present */
2526 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2527 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2528 EXT4_GOOD_OLD_INODE_SIZE
;
2529 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2530 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2531 if (sbi
->s_want_extra_isize
<
2532 le16_to_cpu(es
->s_want_extra_isize
))
2533 sbi
->s_want_extra_isize
=
2534 le16_to_cpu(es
->s_want_extra_isize
);
2535 if (sbi
->s_want_extra_isize
<
2536 le16_to_cpu(es
->s_min_extra_isize
))
2537 sbi
->s_want_extra_isize
=
2538 le16_to_cpu(es
->s_min_extra_isize
);
2541 /* Check if enough inode space is available */
2542 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2543 sbi
->s_inode_size
) {
2544 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2545 EXT4_GOOD_OLD_INODE_SIZE
;
2546 printk(KERN_INFO
"EXT4-fs: required extra inode space not"
2550 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
2551 printk(KERN_WARNING
"EXT4-fs: Ignoring delalloc option - "
2552 "requested data journaling mode\n");
2553 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2554 } else if (test_opt(sb
, DELALLOC
))
2555 printk(KERN_INFO
"EXT4-fs: delayed allocation enabled\n");
2558 err
= ext4_mb_init(sb
, needs_recovery
);
2560 printk(KERN_ERR
"EXT4-fs: failed to initalize mballoc (%d)\n",
2566 * akpm: core read_super() calls in here with the superblock locked.
2567 * That deadlocks, because orphan cleanup needs to lock the superblock
2568 * in numerous places. Here we just pop the lock - it's relatively
2569 * harmless, because we are now ready to accept write_super() requests,
2570 * and aviro says that's the only reason for hanging onto the
2573 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
2574 ext4_orphan_cleanup(sb
, es
);
2575 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
2576 if (needs_recovery
) {
2577 printk(KERN_INFO
"EXT4-fs: recovery complete.\n");
2578 ext4_mark_recovery_complete(sb
, es
);
2580 if (EXT4_SB(sb
)->s_journal
) {
2581 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2582 descr
= " journalled data mode";
2583 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2584 descr
= " ordered data mode";
2586 descr
= " writeback data mode";
2588 descr
= "out journal";
2590 printk(KERN_INFO
"EXT4-fs: mounted filesystem %s with%s\n",
2598 printk(KERN_ERR
"VFS: Can't find ext4 filesystem on dev %s.\n",
2603 printk(KERN_ERR
"EXT4-fs (device %s): mount failed\n", sb
->s_id
);
2604 if (sbi
->s_journal
) {
2605 jbd2_journal_destroy(sbi
->s_journal
);
2606 sbi
->s_journal
= NULL
;
2609 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2610 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2611 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2612 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
2614 for (i
= 0; i
< db_count
; i
++)
2615 brelse(sbi
->s_group_desc
[i
]);
2616 kfree(sbi
->s_group_desc
);
2619 remove_proc_entry("inode_readahead_blks", sbi
->s_proc
);
2620 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
2623 for (i
= 0; i
< MAXQUOTAS
; i
++)
2624 kfree(sbi
->s_qf_names
[i
]);
2626 ext4_blkdev_remove(sbi
);
2629 sb
->s_fs_info
= NULL
;
2636 * Setup any per-fs journal parameters now. We'll do this both on
2637 * initial mount, once the journal has been initialised but before we've
2638 * done any recovery; and again on any subsequent remount.
2640 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
2642 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2644 journal
->j_commit_interval
= sbi
->s_commit_interval
;
2645 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
2646 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
2648 spin_lock(&journal
->j_state_lock
);
2649 if (test_opt(sb
, BARRIER
))
2650 journal
->j_flags
|= JBD2_BARRIER
;
2652 journal
->j_flags
&= ~JBD2_BARRIER
;
2653 if (test_opt(sb
, DATA_ERR_ABORT
))
2654 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
2656 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
2657 spin_unlock(&journal
->j_state_lock
);
2660 static journal_t
*ext4_get_journal(struct super_block
*sb
,
2661 unsigned int journal_inum
)
2663 struct inode
*journal_inode
;
2666 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2668 /* First, test for the existence of a valid inode on disk. Bad
2669 * things happen if we iget() an unused inode, as the subsequent
2670 * iput() will try to delete it. */
2672 journal_inode
= ext4_iget(sb
, journal_inum
);
2673 if (IS_ERR(journal_inode
)) {
2674 printk(KERN_ERR
"EXT4-fs: no journal found.\n");
2677 if (!journal_inode
->i_nlink
) {
2678 make_bad_inode(journal_inode
);
2679 iput(journal_inode
);
2680 printk(KERN_ERR
"EXT4-fs: journal inode is deleted.\n");
2684 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2685 journal_inode
, journal_inode
->i_size
);
2686 if (!S_ISREG(journal_inode
->i_mode
)) {
2687 printk(KERN_ERR
"EXT4-fs: invalid journal inode.\n");
2688 iput(journal_inode
);
2692 journal
= jbd2_journal_init_inode(journal_inode
);
2694 printk(KERN_ERR
"EXT4-fs: Could not load journal inode\n");
2695 iput(journal_inode
);
2698 journal
->j_private
= sb
;
2699 ext4_init_journal_params(sb
, journal
);
2703 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
2706 struct buffer_head
*bh
;
2710 int hblock
, blocksize
;
2711 ext4_fsblk_t sb_block
;
2712 unsigned long offset
;
2713 struct ext4_super_block
*es
;
2714 struct block_device
*bdev
;
2716 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2718 bdev
= ext4_blkdev_get(j_dev
);
2722 if (bd_claim(bdev
, sb
)) {
2724 "EXT4-fs: failed to claim external journal device.\n");
2725 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
2729 blocksize
= sb
->s_blocksize
;
2730 hblock
= bdev_hardsect_size(bdev
);
2731 if (blocksize
< hblock
) {
2733 "EXT4-fs: blocksize too small for journal device.\n");
2737 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
2738 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
2739 set_blocksize(bdev
, blocksize
);
2740 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
2741 printk(KERN_ERR
"EXT4-fs: couldn't read superblock of "
2742 "external journal\n");
2746 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2747 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
2748 !(le32_to_cpu(es
->s_feature_incompat
) &
2749 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
2750 printk(KERN_ERR
"EXT4-fs: external journal has "
2751 "bad superblock\n");
2756 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
2757 printk(KERN_ERR
"EXT4-fs: journal UUID does not match\n");
2762 len
= ext4_blocks_count(es
);
2763 start
= sb_block
+ 1;
2764 brelse(bh
); /* we're done with the superblock */
2766 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
2767 start
, len
, blocksize
);
2769 printk(KERN_ERR
"EXT4-fs: failed to create device journal\n");
2772 journal
->j_private
= sb
;
2773 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
2774 wait_on_buffer(journal
->j_sb_buffer
);
2775 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
2776 printk(KERN_ERR
"EXT4-fs: I/O error on journal device\n");
2779 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
2780 printk(KERN_ERR
"EXT4-fs: External journal has more than one "
2781 "user (unsupported) - %d\n",
2782 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
2785 EXT4_SB(sb
)->journal_bdev
= bdev
;
2786 ext4_init_journal_params(sb
, journal
);
2789 jbd2_journal_destroy(journal
);
2791 ext4_blkdev_put(bdev
);
2795 static int ext4_load_journal(struct super_block
*sb
,
2796 struct ext4_super_block
*es
,
2797 unsigned long journal_devnum
)
2800 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
2803 int really_read_only
;
2805 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2807 if (journal_devnum
&&
2808 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2809 printk(KERN_INFO
"EXT4-fs: external journal device major/minor "
2810 "numbers have changed\n");
2811 journal_dev
= new_decode_dev(journal_devnum
);
2813 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
2815 really_read_only
= bdev_read_only(sb
->s_bdev
);
2818 * Are we loading a blank journal or performing recovery after a
2819 * crash? For recovery, we need to check in advance whether we
2820 * can get read-write access to the device.
2823 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2824 if (sb
->s_flags
& MS_RDONLY
) {
2825 printk(KERN_INFO
"EXT4-fs: INFO: recovery "
2826 "required on readonly filesystem.\n");
2827 if (really_read_only
) {
2828 printk(KERN_ERR
"EXT4-fs: write access "
2829 "unavailable, cannot proceed.\n");
2832 printk(KERN_INFO
"EXT4-fs: write access will "
2833 "be enabled during recovery.\n");
2837 if (journal_inum
&& journal_dev
) {
2838 printk(KERN_ERR
"EXT4-fs: filesystem has both journal "
2839 "and inode journals!\n");
2844 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
2847 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
2851 if (journal
->j_flags
& JBD2_BARRIER
)
2852 printk(KERN_INFO
"EXT4-fs: barriers enabled\n");
2854 printk(KERN_INFO
"EXT4-fs: barriers disabled\n");
2856 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
2857 err
= jbd2_journal_update_format(journal
);
2859 printk(KERN_ERR
"EXT4-fs: error updating journal.\n");
2860 jbd2_journal_destroy(journal
);
2865 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
2866 err
= jbd2_journal_wipe(journal
, !really_read_only
);
2868 err
= jbd2_journal_load(journal
);
2871 printk(KERN_ERR
"EXT4-fs: error loading journal.\n");
2872 jbd2_journal_destroy(journal
);
2876 EXT4_SB(sb
)->s_journal
= journal
;
2877 ext4_clear_journal_err(sb
, es
);
2879 if (journal_devnum
&&
2880 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2881 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2884 /* Make sure we flush the recovery flag to disk. */
2885 ext4_commit_super(sb
, es
, 1);
2891 static void ext4_commit_super(struct super_block
*sb
,
2892 struct ext4_super_block
*es
, int sync
)
2894 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
2898 if (buffer_write_io_error(sbh
)) {
2900 * Oh, dear. A previous attempt to write the
2901 * superblock failed. This could happen because the
2902 * USB device was yanked out. Or it could happen to
2903 * be a transient write error and maybe the block will
2904 * be remapped. Nothing we can do but to retry the
2905 * write and hope for the best.
2907 printk(KERN_ERR
"EXT4-fs: previous I/O error to "
2908 "superblock detected for %s.\n", sb
->s_id
);
2909 clear_buffer_write_io_error(sbh
);
2910 set_buffer_uptodate(sbh
);
2912 es
->s_wtime
= cpu_to_le32(get_seconds());
2913 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
2914 &EXT4_SB(sb
)->s_freeblocks_counter
));
2915 es
->s_free_inodes_count
= cpu_to_le32(percpu_counter_sum_positive(
2916 &EXT4_SB(sb
)->s_freeinodes_counter
));
2918 BUFFER_TRACE(sbh
, "marking dirty");
2919 mark_buffer_dirty(sbh
);
2921 sync_dirty_buffer(sbh
);
2922 if (buffer_write_io_error(sbh
)) {
2923 printk(KERN_ERR
"EXT4-fs: I/O error while writing "
2924 "superblock for %s.\n", sb
->s_id
);
2925 clear_buffer_write_io_error(sbh
);
2926 set_buffer_uptodate(sbh
);
2933 * Have we just finished recovery? If so, and if we are mounting (or
2934 * remounting) the filesystem readonly, then we will end up with a
2935 * consistent fs on disk. Record that fact.
2937 static void ext4_mark_recovery_complete(struct super_block
*sb
,
2938 struct ext4_super_block
*es
)
2940 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
2942 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2943 BUG_ON(journal
!= NULL
);
2946 jbd2_journal_lock_updates(journal
);
2947 if (jbd2_journal_flush(journal
) < 0)
2951 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
2952 sb
->s_flags
& MS_RDONLY
) {
2953 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2955 ext4_commit_super(sb
, es
, 1);
2960 jbd2_journal_unlock_updates(journal
);
2964 * If we are mounting (or read-write remounting) a filesystem whose journal
2965 * has recorded an error from a previous lifetime, move that error to the
2966 * main filesystem now.
2968 static void ext4_clear_journal_err(struct super_block
*sb
,
2969 struct ext4_super_block
*es
)
2975 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2977 journal
= EXT4_SB(sb
)->s_journal
;
2980 * Now check for any error status which may have been recorded in the
2981 * journal by a prior ext4_error() or ext4_abort()
2984 j_errno
= jbd2_journal_errno(journal
);
2988 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
2989 ext4_warning(sb
, __func__
, "Filesystem error recorded "
2990 "from previous mount: %s", errstr
);
2991 ext4_warning(sb
, __func__
, "Marking fs in need of "
2992 "filesystem check.");
2994 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2995 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2996 ext4_commit_super(sb
, es
, 1);
2998 jbd2_journal_clear_err(journal
);
3003 * Force the running and committing transactions to commit,
3004 * and wait on the commit.
3006 int ext4_force_commit(struct super_block
*sb
)
3011 if (sb
->s_flags
& MS_RDONLY
)
3014 journal
= EXT4_SB(sb
)->s_journal
;
3017 ret
= ext4_journal_force_commit(journal
);
3024 * Ext4 always journals updates to the superblock itself, so we don't
3025 * have to propagate any other updates to the superblock on disk at this
3026 * point. (We can probably nuke this function altogether, and remove
3027 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3029 static void ext4_write_super(struct super_block
*sb
)
3031 if (EXT4_SB(sb
)->s_journal
) {
3032 if (mutex_trylock(&sb
->s_lock
) != 0)
3036 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
3040 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
3044 trace_mark(ext4_sync_fs
, "dev %s wait %d", sb
->s_id
, wait
);
3046 if (EXT4_SB(sb
)->s_journal
) {
3048 ret
= ext4_force_commit(sb
);
3050 jbd2_journal_start_commit(EXT4_SB(sb
)->s_journal
, NULL
);
3052 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, wait
);
3058 * LVM calls this function before a (read-only) snapshot is created. This
3059 * gives us a chance to flush the journal completely and mark the fs clean.
3061 static void ext4_write_super_lockfs(struct super_block
*sb
)
3065 if (!(sb
->s_flags
& MS_RDONLY
)) {
3066 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
3069 /* Now we set up the journal barrier. */
3070 jbd2_journal_lock_updates(journal
);
3073 * We don't want to clear needs_recovery flag when we
3074 * failed to flush the journal.
3076 if (jbd2_journal_flush(journal
) < 0)
3080 /* Journal blocked and flushed, clear needs_recovery flag. */
3081 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3082 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
3087 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3088 * flag here, even though the filesystem is not technically dirty yet.
3090 static void ext4_unlockfs(struct super_block
*sb
)
3092 if (EXT4_SB(sb
)->s_journal
&& !(sb
->s_flags
& MS_RDONLY
)) {
3094 /* Reser the needs_recovery flag before the fs is unlocked. */
3095 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3096 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
3098 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3102 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3104 struct ext4_super_block
*es
;
3105 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3106 ext4_fsblk_t n_blocks_count
= 0;
3107 unsigned long old_sb_flags
;
3108 struct ext4_mount_options old_opts
;
3110 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3116 /* Store the original options */
3117 old_sb_flags
= sb
->s_flags
;
3118 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3119 old_opts
.s_resuid
= sbi
->s_resuid
;
3120 old_opts
.s_resgid
= sbi
->s_resgid
;
3121 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3122 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
3123 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
3125 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3126 for (i
= 0; i
< MAXQUOTAS
; i
++)
3127 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3129 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
3130 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
3133 * Allow the "check" option to be passed as a remount option.
3135 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
3136 &n_blocks_count
, 1)) {
3141 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
)
3142 ext4_abort(sb
, __func__
, "Abort forced by user");
3144 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3145 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
3149 if (sbi
->s_journal
) {
3150 ext4_init_journal_params(sb
, sbi
->s_journal
);
3151 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3154 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3155 n_blocks_count
> ext4_blocks_count(es
)) {
3156 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
) {
3161 if (*flags
& MS_RDONLY
) {
3163 * First of all, the unconditional stuff we have to do
3164 * to disable replay of the journal when we next remount
3166 sb
->s_flags
|= MS_RDONLY
;
3169 * OK, test if we are remounting a valid rw partition
3170 * readonly, and if so set the rdonly flag and then
3171 * mark the partition as valid again.
3173 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3174 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3175 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3178 * We have to unlock super so that we can wait for
3181 if (sbi
->s_journal
) {
3183 ext4_mark_recovery_complete(sb
, es
);
3188 if ((ret
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3189 ~EXT4_FEATURE_RO_COMPAT_SUPP
))) {
3190 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3191 "remount RDWR because of unsupported "
3192 "optional features (%x).\n", sb
->s_id
,
3193 (le32_to_cpu(sbi
->s_es
->s_feature_ro_compat
) &
3194 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
3200 * Make sure the group descriptor checksums
3201 * are sane. If they aren't, refuse to
3204 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3205 struct ext4_group_desc
*gdp
=
3206 ext4_get_group_desc(sb
, g
, NULL
);
3208 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3210 "EXT4-fs: ext4_remount: "
3211 "Checksum for group %u failed (%u!=%u)\n",
3212 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3213 le16_to_cpu(gdp
->bg_checksum
));
3220 * If we have an unprocessed orphan list hanging
3221 * around from a previously readonly bdev mount,
3222 * require a full umount/remount for now.
3224 if (es
->s_last_orphan
) {
3225 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3226 "remount RDWR because of unprocessed "
3227 "orphan inode list. Please "
3228 "umount/remount instead.\n",
3235 * Mounting a RDONLY partition read-write, so reread
3236 * and store the current valid flag. (It may have
3237 * been changed by e2fsck since we originally mounted
3241 ext4_clear_journal_err(sb
, es
);
3242 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3243 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3245 if (!ext4_setup_super(sb
, es
, 0))
3246 sb
->s_flags
&= ~MS_RDONLY
;
3249 if (sbi
->s_journal
== NULL
)
3250 ext4_commit_super(sb
, es
, 1);
3253 /* Release old quota file names */
3254 for (i
= 0; i
< MAXQUOTAS
; i
++)
3255 if (old_opts
.s_qf_names
[i
] &&
3256 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3257 kfree(old_opts
.s_qf_names
[i
]);
3261 sb
->s_flags
= old_sb_flags
;
3262 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3263 sbi
->s_resuid
= old_opts
.s_resuid
;
3264 sbi
->s_resgid
= old_opts
.s_resgid
;
3265 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3266 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
3267 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
3269 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3270 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3271 if (sbi
->s_qf_names
[i
] &&
3272 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3273 kfree(sbi
->s_qf_names
[i
]);
3274 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3280 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3282 struct super_block
*sb
= dentry
->d_sb
;
3283 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3284 struct ext4_super_block
*es
= sbi
->s_es
;
3287 if (test_opt(sb
, MINIX_DF
)) {
3288 sbi
->s_overhead_last
= 0;
3289 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3290 ext4_group_t ngroups
= sbi
->s_groups_count
, i
;
3291 ext4_fsblk_t overhead
= 0;
3295 * Compute the overhead (FS structures). This is constant
3296 * for a given filesystem unless the number of block groups
3297 * changes so we cache the previous value until it does.
3301 * All of the blocks before first_data_block are
3304 overhead
= le32_to_cpu(es
->s_first_data_block
);
3307 * Add the overhead attributed to the superblock and
3308 * block group descriptors. If the sparse superblocks
3309 * feature is turned on, then not all groups have this.
3311 for (i
= 0; i
< ngroups
; i
++) {
3312 overhead
+= ext4_bg_has_super(sb
, i
) +
3313 ext4_bg_num_gdb(sb
, i
);
3318 * Every block group has an inode bitmap, a block
3319 * bitmap, and an inode table.
3321 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3322 sbi
->s_overhead_last
= overhead
;
3324 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3327 buf
->f_type
= EXT4_SUPER_MAGIC
;
3328 buf
->f_bsize
= sb
->s_blocksize
;
3329 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3330 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3331 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3332 ext4_free_blocks_count_set(es
, buf
->f_bfree
);
3333 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3334 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3336 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3337 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3338 es
->s_free_inodes_count
= cpu_to_le32(buf
->f_ffree
);
3339 buf
->f_namelen
= EXT4_NAME_LEN
;
3340 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3341 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3342 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3343 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3347 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3348 * is locked for write. Otherwise the are possible deadlocks:
3349 * Process 1 Process 2
3350 * ext4_create() quota_sync()
3351 * jbd2_journal_start() write_dquot()
3352 * DQUOT_INIT() down(dqio_mutex)
3353 * down(dqio_mutex) jbd2_journal_start()
3359 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3361 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3364 static int ext4_dquot_initialize(struct inode
*inode
, int type
)
3369 /* We may create quota structure so we need to reserve enough blocks */
3370 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_INIT_BLOCKS(inode
->i_sb
));
3372 return PTR_ERR(handle
);
3373 ret
= dquot_initialize(inode
, type
);
3374 err
= ext4_journal_stop(handle
);
3380 static int ext4_dquot_drop(struct inode
*inode
)
3385 /* We may delete quota structure so we need to reserve enough blocks */
3386 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_DEL_BLOCKS(inode
->i_sb
));
3387 if (IS_ERR(handle
)) {
3389 * We call dquot_drop() anyway to at least release references
3390 * to quota structures so that umount does not hang.
3393 return PTR_ERR(handle
);
3395 ret
= dquot_drop(inode
);
3396 err
= ext4_journal_stop(handle
);
3402 static int ext4_write_dquot(struct dquot
*dquot
)
3406 struct inode
*inode
;
3408 inode
= dquot_to_inode(dquot
);
3409 handle
= ext4_journal_start(inode
,
3410 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3412 return PTR_ERR(handle
);
3413 ret
= dquot_commit(dquot
);
3414 err
= ext4_journal_stop(handle
);
3420 static int ext4_acquire_dquot(struct dquot
*dquot
)
3425 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3426 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3428 return PTR_ERR(handle
);
3429 ret
= dquot_acquire(dquot
);
3430 err
= ext4_journal_stop(handle
);
3436 static int ext4_release_dquot(struct dquot
*dquot
)
3441 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3442 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3443 if (IS_ERR(handle
)) {
3444 /* Release dquot anyway to avoid endless cycle in dqput() */
3445 dquot_release(dquot
);
3446 return PTR_ERR(handle
);
3448 ret
= dquot_release(dquot
);
3449 err
= ext4_journal_stop(handle
);
3455 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3457 /* Are we journaling quotas? */
3458 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3459 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3460 dquot_mark_dquot_dirty(dquot
);
3461 return ext4_write_dquot(dquot
);
3463 return dquot_mark_dquot_dirty(dquot
);
3467 static int ext4_write_info(struct super_block
*sb
, int type
)
3472 /* Data block + inode block */
3473 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3475 return PTR_ERR(handle
);
3476 ret
= dquot_commit_info(sb
, type
);
3477 err
= ext4_journal_stop(handle
);
3484 * Turn on quotas during mount time - we need to find
3485 * the quota file and such...
3487 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3489 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3490 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3494 * Standard function to be called on quota_on
3496 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3497 char *name
, int remount
)
3502 if (!test_opt(sb
, QUOTA
))
3504 /* When remounting, no checks are needed and in fact, name is NULL */
3506 return vfs_quota_on(sb
, type
, format_id
, name
, remount
);
3508 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
3512 /* Quotafile not on the same filesystem? */
3513 if (path
.mnt
->mnt_sb
!= sb
) {
3517 /* Journaling quota? */
3518 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3519 /* Quotafile not in fs root? */
3520 if (path
.dentry
->d_parent
!= sb
->s_root
)
3522 "EXT4-fs: Quota file not on filesystem root. "
3523 "Journaled quota will not work.\n");
3527 * When we journal data on quota file, we have to flush journal to see
3528 * all updates to the file when we bypass pagecache...
3530 if (EXT4_SB(sb
)->s_journal
&&
3531 ext4_should_journal_data(path
.dentry
->d_inode
)) {
3533 * We don't need to lock updates but journal_flush() could
3534 * otherwise be livelocked...
3536 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3537 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3538 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3545 err
= vfs_quota_on_path(sb
, type
, format_id
, &path
);
3550 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3551 * acquiring the locks... As quota files are never truncated and quota code
3552 * itself serializes the operations (and noone else should touch the files)
3553 * we don't have to be afraid of races */
3554 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3555 size_t len
, loff_t off
)
3557 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3558 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3560 int offset
= off
& (sb
->s_blocksize
- 1);
3563 struct buffer_head
*bh
;
3564 loff_t i_size
= i_size_read(inode
);
3568 if (off
+len
> i_size
)
3571 while (toread
> 0) {
3572 tocopy
= sb
->s_blocksize
- offset
< toread
?
3573 sb
->s_blocksize
- offset
: toread
;
3574 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3577 if (!bh
) /* A hole? */
3578 memset(data
, 0, tocopy
);
3580 memcpy(data
, bh
->b_data
+offset
, tocopy
);
3590 /* Write to quotafile (we know the transaction is already started and has
3591 * enough credits) */
3592 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
3593 const char *data
, size_t len
, loff_t off
)
3595 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3596 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3598 int offset
= off
& (sb
->s_blocksize
- 1);
3600 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
3601 size_t towrite
= len
;
3602 struct buffer_head
*bh
;
3603 handle_t
*handle
= journal_current_handle();
3605 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
3606 printk(KERN_WARNING
"EXT4-fs: Quota write (off=%llu, len=%llu)"
3607 " cancelled because transaction is not started.\n",
3608 (unsigned long long)off
, (unsigned long long)len
);
3611 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
3612 while (towrite
> 0) {
3613 tocopy
= sb
->s_blocksize
- offset
< towrite
?
3614 sb
->s_blocksize
- offset
: towrite
;
3615 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
3618 if (journal_quota
) {
3619 err
= ext4_journal_get_write_access(handle
, bh
);
3626 memcpy(bh
->b_data
+offset
, data
, tocopy
);
3627 flush_dcache_page(bh
->b_page
);
3630 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
3632 /* Always do at least ordered writes for quotas */
3633 err
= ext4_jbd2_file_inode(handle
, inode
);
3634 mark_buffer_dirty(bh
);
3645 if (len
== towrite
) {
3646 mutex_unlock(&inode
->i_mutex
);
3649 if (inode
->i_size
< off
+len
-towrite
) {
3650 i_size_write(inode
, off
+len
-towrite
);
3651 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
3653 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3654 ext4_mark_inode_dirty(handle
, inode
);
3655 mutex_unlock(&inode
->i_mutex
);
3656 return len
- towrite
;
3661 static int ext4_get_sb(struct file_system_type
*fs_type
,
3662 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3664 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3667 #ifdef CONFIG_PROC_FS
3668 static int ext4_ui_proc_show(struct seq_file
*m
, void *v
)
3670 unsigned int *p
= m
->private;
3672 seq_printf(m
, "%u\n", *p
);
3676 static int ext4_ui_proc_open(struct inode
*inode
, struct file
*file
)
3678 return single_open(file
, ext4_ui_proc_show
, PDE(inode
)->data
);
3681 static ssize_t
ext4_ui_proc_write(struct file
*file
, const char __user
*buf
,
3682 size_t cnt
, loff_t
*ppos
)
3684 unsigned long *p
= PDE(file
->f_path
.dentry
->d_inode
)->data
;
3687 if (cnt
>= sizeof(str
))
3689 if (copy_from_user(str
, buf
, cnt
))
3692 *p
= simple_strtoul(str
, NULL
, 0);
3696 const struct file_operations ext4_ui_proc_fops
= {
3697 .owner
= THIS_MODULE
,
3698 .open
= ext4_ui_proc_open
,
3700 .llseek
= seq_lseek
,
3701 .release
= single_release
,
3702 .write
= ext4_ui_proc_write
,
3706 static struct file_system_type ext4_fs_type
= {
3707 .owner
= THIS_MODULE
,
3709 .get_sb
= ext4_get_sb
,
3710 .kill_sb
= kill_block_super
,
3711 .fs_flags
= FS_REQUIRES_DEV
,
3714 #ifdef CONFIG_EXT4DEV_COMPAT
3715 static int ext4dev_get_sb(struct file_system_type
*fs_type
,
3716 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3718 printk(KERN_WARNING
"EXT4-fs: Update your userspace programs "
3719 "to mount using ext4\n");
3720 printk(KERN_WARNING
"EXT4-fs: ext4dev backwards compatibility "
3721 "will go away by 2.6.31\n");
3722 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3725 static struct file_system_type ext4dev_fs_type
= {
3726 .owner
= THIS_MODULE
,
3728 .get_sb
= ext4dev_get_sb
,
3729 .kill_sb
= kill_block_super
,
3730 .fs_flags
= FS_REQUIRES_DEV
,
3732 MODULE_ALIAS("ext4dev");
3735 static int __init
init_ext4_fs(void)
3739 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
3740 err
= init_ext4_mballoc();
3744 err
= init_ext4_xattr();
3747 err
= init_inodecache();
3750 err
= register_filesystem(&ext4_fs_type
);
3753 #ifdef CONFIG_EXT4DEV_COMPAT
3754 err
= register_filesystem(&ext4dev_fs_type
);
3756 unregister_filesystem(&ext4_fs_type
);
3762 destroy_inodecache();
3766 exit_ext4_mballoc();
3770 static void __exit
exit_ext4_fs(void)
3772 unregister_filesystem(&ext4_fs_type
);
3773 #ifdef CONFIG_EXT4DEV_COMPAT
3774 unregister_filesystem(&ext4dev_fs_type
);
3776 destroy_inodecache();
3778 exit_ext4_mballoc();
3779 remove_proc_entry("fs/ext4", NULL
);
3782 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3783 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3784 MODULE_LICENSE("GPL");
3785 module_init(init_ext4_fs
)
3786 module_exit(exit_ext4_fs
)