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 int 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 int ext4_unfreeze(struct super_block
*sb
);
66 static void ext4_write_super(struct super_block
*sb
);
67 static int ext4_freeze(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
;
618 kfree(sbi
->s_blockgroup_lock
);
623 static struct kmem_cache
*ext4_inode_cachep
;
626 * Called inside transaction, so use GFP_NOFS
628 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
630 struct ext4_inode_info
*ei
;
632 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
635 #ifdef CONFIG_EXT4_FS_POSIX_ACL
636 ei
->i_acl
= EXT4_ACL_NOT_CACHED
;
637 ei
->i_default_acl
= EXT4_ACL_NOT_CACHED
;
639 ei
->vfs_inode
.i_version
= 1;
640 ei
->vfs_inode
.i_data
.writeback_index
= 0;
641 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
642 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
643 spin_lock_init(&ei
->i_prealloc_lock
);
645 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
646 * therefore it can be null here. Don't check it, just initialize
649 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
650 ei
->i_reserved_data_blocks
= 0;
651 ei
->i_reserved_meta_blocks
= 0;
652 ei
->i_allocated_meta_blocks
= 0;
653 ei
->i_delalloc_reserved_flag
= 0;
654 spin_lock_init(&(ei
->i_block_reservation_lock
));
655 return &ei
->vfs_inode
;
658 static void ext4_destroy_inode(struct inode
*inode
)
660 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
661 printk("EXT4 Inode %p: orphan list check failed!\n",
663 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
664 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
668 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
671 static void init_once(void *foo
)
673 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
675 INIT_LIST_HEAD(&ei
->i_orphan
);
676 #ifdef CONFIG_EXT4_FS_XATTR
677 init_rwsem(&ei
->xattr_sem
);
679 init_rwsem(&ei
->i_data_sem
);
680 inode_init_once(&ei
->vfs_inode
);
683 static int init_inodecache(void)
685 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
686 sizeof(struct ext4_inode_info
),
687 0, (SLAB_RECLAIM_ACCOUNT
|
690 if (ext4_inode_cachep
== NULL
)
695 static void destroy_inodecache(void)
697 kmem_cache_destroy(ext4_inode_cachep
);
700 static void ext4_clear_inode(struct inode
*inode
)
702 #ifdef CONFIG_EXT4_FS_POSIX_ACL
703 if (EXT4_I(inode
)->i_acl
&&
704 EXT4_I(inode
)->i_acl
!= EXT4_ACL_NOT_CACHED
) {
705 posix_acl_release(EXT4_I(inode
)->i_acl
);
706 EXT4_I(inode
)->i_acl
= EXT4_ACL_NOT_CACHED
;
708 if (EXT4_I(inode
)->i_default_acl
&&
709 EXT4_I(inode
)->i_default_acl
!= EXT4_ACL_NOT_CACHED
) {
710 posix_acl_release(EXT4_I(inode
)->i_default_acl
);
711 EXT4_I(inode
)->i_default_acl
= EXT4_ACL_NOT_CACHED
;
714 ext4_discard_preallocations(inode
);
715 if (EXT4_JOURNAL(inode
))
716 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
717 &EXT4_I(inode
)->jinode
);
720 static inline void ext4_show_quota_options(struct seq_file
*seq
,
721 struct super_block
*sb
)
723 #if defined(CONFIG_QUOTA)
724 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
726 if (sbi
->s_jquota_fmt
)
727 seq_printf(seq
, ",jqfmt=%s",
728 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold" : "vfsv0");
730 if (sbi
->s_qf_names
[USRQUOTA
])
731 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
733 if (sbi
->s_qf_names
[GRPQUOTA
])
734 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
736 if (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
)
737 seq_puts(seq
, ",usrquota");
739 if (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)
740 seq_puts(seq
, ",grpquota");
746 * - it's set to a non-default value OR
747 * - if the per-sb default is different from the global default
749 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
752 unsigned long def_mount_opts
;
753 struct super_block
*sb
= vfs
->mnt_sb
;
754 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
755 struct ext4_super_block
*es
= sbi
->s_es
;
757 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
758 def_errors
= le16_to_cpu(es
->s_errors
);
760 if (sbi
->s_sb_block
!= 1)
761 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
762 if (test_opt(sb
, MINIX_DF
))
763 seq_puts(seq
, ",minixdf");
764 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
765 seq_puts(seq
, ",grpid");
766 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
767 seq_puts(seq
, ",nogrpid");
768 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
769 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
770 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
772 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
773 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
774 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
776 if (test_opt(sb
, ERRORS_RO
)) {
777 if (def_errors
== EXT4_ERRORS_PANIC
||
778 def_errors
== EXT4_ERRORS_CONTINUE
) {
779 seq_puts(seq
, ",errors=remount-ro");
782 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
783 seq_puts(seq
, ",errors=continue");
784 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
785 seq_puts(seq
, ",errors=panic");
786 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
787 seq_puts(seq
, ",nouid32");
788 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
789 seq_puts(seq
, ",debug");
790 if (test_opt(sb
, OLDALLOC
))
791 seq_puts(seq
, ",oldalloc");
792 #ifdef CONFIG_EXT4_FS_XATTR
793 if (test_opt(sb
, XATTR_USER
) &&
794 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
795 seq_puts(seq
, ",user_xattr");
796 if (!test_opt(sb
, XATTR_USER
) &&
797 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
798 seq_puts(seq
, ",nouser_xattr");
801 #ifdef CONFIG_EXT4_FS_POSIX_ACL
802 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
803 seq_puts(seq
, ",acl");
804 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
805 seq_puts(seq
, ",noacl");
807 if (!test_opt(sb
, RESERVATION
))
808 seq_puts(seq
, ",noreservation");
809 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
810 seq_printf(seq
, ",commit=%u",
811 (unsigned) (sbi
->s_commit_interval
/ HZ
));
813 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
814 seq_printf(seq
, ",min_batch_time=%u",
815 (unsigned) sbi
->s_min_batch_time
);
817 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
818 seq_printf(seq
, ",max_batch_time=%u",
819 (unsigned) sbi
->s_min_batch_time
);
823 * We're changing the default of barrier mount option, so
824 * let's always display its mount state so it's clear what its
827 seq_puts(seq
, ",barrier=");
828 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
829 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
830 seq_puts(seq
, ",journal_async_commit");
831 if (test_opt(sb
, NOBH
))
832 seq_puts(seq
, ",nobh");
833 if (test_opt(sb
, I_VERSION
))
834 seq_puts(seq
, ",i_version");
835 if (!test_opt(sb
, DELALLOC
))
836 seq_puts(seq
, ",nodelalloc");
840 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
842 * journal mode get enabled in different ways
843 * So just print the value even if we didn't specify it
845 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
846 seq_puts(seq
, ",data=journal");
847 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
848 seq_puts(seq
, ",data=ordered");
849 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
850 seq_puts(seq
, ",data=writeback");
852 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
853 seq_printf(seq
, ",inode_readahead_blks=%u",
854 sbi
->s_inode_readahead_blks
);
856 if (test_opt(sb
, DATA_ERR_ABORT
))
857 seq_puts(seq
, ",data_err=abort");
859 ext4_show_quota_options(seq
, sb
);
864 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
865 u64 ino
, u32 generation
)
869 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
870 return ERR_PTR(-ESTALE
);
871 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
872 return ERR_PTR(-ESTALE
);
874 /* iget isn't really right if the inode is currently unallocated!!
876 * ext4_read_inode will return a bad_inode if the inode had been
877 * deleted, so we should be safe.
879 * Currently we don't know the generation for parent directory, so
880 * a generation of 0 means "accept any"
882 inode
= ext4_iget(sb
, ino
);
884 return ERR_CAST(inode
);
885 if (generation
&& inode
->i_generation
!= generation
) {
887 return ERR_PTR(-ESTALE
);
893 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
894 int fh_len
, int fh_type
)
896 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
900 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
901 int fh_len
, int fh_type
)
903 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
908 * Try to release metadata pages (indirect blocks, directories) which are
909 * mapped via the block device. Since these pages could have journal heads
910 * which would prevent try_to_free_buffers() from freeing them, we must use
911 * jbd2 layer's try_to_free_buffers() function to release them.
913 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
, gfp_t wait
)
915 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
917 WARN_ON(PageChecked(page
));
918 if (!page_has_buffers(page
))
921 return jbd2_journal_try_to_free_buffers(journal
, page
,
923 return try_to_free_buffers(page
);
927 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
928 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
930 static int ext4_write_dquot(struct dquot
*dquot
);
931 static int ext4_acquire_dquot(struct dquot
*dquot
);
932 static int ext4_release_dquot(struct dquot
*dquot
);
933 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
934 static int ext4_write_info(struct super_block
*sb
, int type
);
935 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
936 char *path
, int remount
);
937 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
938 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
939 size_t len
, loff_t off
);
940 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
941 const char *data
, size_t len
, loff_t off
);
943 static struct dquot_operations ext4_quota_operations
= {
944 .initialize
= dquot_initialize
,
946 .alloc_space
= dquot_alloc_space
,
947 .reserve_space
= dquot_reserve_space
,
948 .claim_space
= dquot_claim_space
,
949 .release_rsv
= dquot_release_reserved_space
,
950 .get_reserved_space
= ext4_get_reserved_space
,
951 .alloc_inode
= dquot_alloc_inode
,
952 .free_space
= dquot_free_space
,
953 .free_inode
= dquot_free_inode
,
954 .transfer
= dquot_transfer
,
955 .write_dquot
= ext4_write_dquot
,
956 .acquire_dquot
= ext4_acquire_dquot
,
957 .release_dquot
= ext4_release_dquot
,
958 .mark_dirty
= ext4_mark_dquot_dirty
,
959 .write_info
= ext4_write_info
,
960 .alloc_dquot
= dquot_alloc
,
961 .destroy_dquot
= dquot_destroy
,
964 static struct quotactl_ops ext4_qctl_operations
= {
965 .quota_on
= ext4_quota_on
,
966 .quota_off
= vfs_quota_off
,
967 .quota_sync
= vfs_quota_sync
,
968 .get_info
= vfs_get_dqinfo
,
969 .set_info
= vfs_set_dqinfo
,
970 .get_dqblk
= vfs_get_dqblk
,
971 .set_dqblk
= vfs_set_dqblk
975 static const struct super_operations ext4_sops
= {
976 .alloc_inode
= ext4_alloc_inode
,
977 .destroy_inode
= ext4_destroy_inode
,
978 .write_inode
= ext4_write_inode
,
979 .dirty_inode
= ext4_dirty_inode
,
980 .delete_inode
= ext4_delete_inode
,
981 .put_super
= ext4_put_super
,
982 .write_super
= ext4_write_super
,
983 .sync_fs
= ext4_sync_fs
,
984 .freeze_fs
= ext4_freeze
,
985 .unfreeze_fs
= ext4_unfreeze
,
986 .statfs
= ext4_statfs
,
987 .remount_fs
= ext4_remount
,
988 .clear_inode
= ext4_clear_inode
,
989 .show_options
= ext4_show_options
,
991 .quota_read
= ext4_quota_read
,
992 .quota_write
= ext4_quota_write
,
994 .bdev_try_to_free_page
= bdev_try_to_free_page
,
997 static const struct export_operations ext4_export_ops
= {
998 .fh_to_dentry
= ext4_fh_to_dentry
,
999 .fh_to_parent
= ext4_fh_to_parent
,
1000 .get_parent
= ext4_get_parent
,
1004 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1005 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1006 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1007 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1008 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
1009 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1010 Opt_journal_update
, Opt_journal_dev
,
1011 Opt_journal_checksum
, Opt_journal_async_commit
,
1012 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1013 Opt_data_err_abort
, Opt_data_err_ignore
,
1014 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1015 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
1016 Opt_ignore
, Opt_barrier
, Opt_err
, Opt_resize
, Opt_usrquota
,
1017 Opt_grpquota
, Opt_i_version
,
1018 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
1019 Opt_inode_readahead_blks
, Opt_journal_ioprio
1022 static const match_table_t tokens
= {
1023 {Opt_bsd_df
, "bsddf"},
1024 {Opt_minix_df
, "minixdf"},
1025 {Opt_grpid
, "grpid"},
1026 {Opt_grpid
, "bsdgroups"},
1027 {Opt_nogrpid
, "nogrpid"},
1028 {Opt_nogrpid
, "sysvgroups"},
1029 {Opt_resgid
, "resgid=%u"},
1030 {Opt_resuid
, "resuid=%u"},
1032 {Opt_err_cont
, "errors=continue"},
1033 {Opt_err_panic
, "errors=panic"},
1034 {Opt_err_ro
, "errors=remount-ro"},
1035 {Opt_nouid32
, "nouid32"},
1036 {Opt_debug
, "debug"},
1037 {Opt_oldalloc
, "oldalloc"},
1038 {Opt_orlov
, "orlov"},
1039 {Opt_user_xattr
, "user_xattr"},
1040 {Opt_nouser_xattr
, "nouser_xattr"},
1042 {Opt_noacl
, "noacl"},
1043 {Opt_reservation
, "reservation"},
1044 {Opt_noreservation
, "noreservation"},
1045 {Opt_noload
, "noload"},
1048 {Opt_commit
, "commit=%u"},
1049 {Opt_min_batch_time
, "min_batch_time=%u"},
1050 {Opt_max_batch_time
, "max_batch_time=%u"},
1051 {Opt_journal_update
, "journal=update"},
1052 {Opt_journal_dev
, "journal_dev=%u"},
1053 {Opt_journal_checksum
, "journal_checksum"},
1054 {Opt_journal_async_commit
, "journal_async_commit"},
1055 {Opt_abort
, "abort"},
1056 {Opt_data_journal
, "data=journal"},
1057 {Opt_data_ordered
, "data=ordered"},
1058 {Opt_data_writeback
, "data=writeback"},
1059 {Opt_data_err_abort
, "data_err=abort"},
1060 {Opt_data_err_ignore
, "data_err=ignore"},
1061 {Opt_offusrjquota
, "usrjquota="},
1062 {Opt_usrjquota
, "usrjquota=%s"},
1063 {Opt_offgrpjquota
, "grpjquota="},
1064 {Opt_grpjquota
, "grpjquota=%s"},
1065 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1066 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1067 {Opt_grpquota
, "grpquota"},
1068 {Opt_noquota
, "noquota"},
1069 {Opt_quota
, "quota"},
1070 {Opt_usrquota
, "usrquota"},
1071 {Opt_barrier
, "barrier=%u"},
1072 {Opt_i_version
, "i_version"},
1073 {Opt_stripe
, "stripe=%u"},
1074 {Opt_resize
, "resize"},
1075 {Opt_delalloc
, "delalloc"},
1076 {Opt_nodelalloc
, "nodelalloc"},
1077 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1078 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1082 static ext4_fsblk_t
get_sb_block(void **data
)
1084 ext4_fsblk_t sb_block
;
1085 char *options
= (char *) *data
;
1087 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1088 return 1; /* Default location */
1090 /*todo: use simple_strtoll with >32bit ext4 */
1091 sb_block
= simple_strtoul(options
, &options
, 0);
1092 if (*options
&& *options
!= ',') {
1093 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1097 if (*options
== ',')
1099 *data
= (void *) options
;
1103 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1105 static int parse_options(char *options
, struct super_block
*sb
,
1106 unsigned long *journal_devnum
,
1107 unsigned int *journal_ioprio
,
1108 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1110 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1112 substring_t args
[MAX_OPT_ARGS
];
1123 while ((p
= strsep(&options
, ",")) != NULL
) {
1128 token
= match_token(p
, tokens
, args
);
1131 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1134 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1137 set_opt(sbi
->s_mount_opt
, GRPID
);
1140 clear_opt(sbi
->s_mount_opt
, GRPID
);
1143 if (match_int(&args
[0], &option
))
1145 sbi
->s_resuid
= option
;
1148 if (match_int(&args
[0], &option
))
1150 sbi
->s_resgid
= option
;
1153 /* handled by get_sb_block() instead of here */
1154 /* *sb_block = match_int(&args[0]); */
1157 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1158 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1159 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1162 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1163 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1164 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1167 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1168 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1169 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1172 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1175 set_opt(sbi
->s_mount_opt
, DEBUG
);
1178 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1181 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1183 #ifdef CONFIG_EXT4_FS_XATTR
1184 case Opt_user_xattr
:
1185 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1187 case Opt_nouser_xattr
:
1188 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1191 case Opt_user_xattr
:
1192 case Opt_nouser_xattr
:
1193 printk(KERN_ERR
"EXT4 (no)user_xattr options "
1197 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1199 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1202 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1207 printk(KERN_ERR
"EXT4 (no)acl options "
1211 case Opt_reservation
:
1212 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1214 case Opt_noreservation
:
1215 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
1217 case Opt_journal_update
:
1219 /* Eventually we will want to be able to create
1220 a journal file here. For now, only allow the
1221 user to specify an existing inode to be the
1224 printk(KERN_ERR
"EXT4-fs: cannot specify "
1225 "journal on remount\n");
1228 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1230 case Opt_journal_dev
:
1232 printk(KERN_ERR
"EXT4-fs: cannot specify "
1233 "journal on remount\n");
1236 if (match_int(&args
[0], &option
))
1238 *journal_devnum
= option
;
1240 case Opt_journal_checksum
:
1241 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1243 case Opt_journal_async_commit
:
1244 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1245 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1248 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1251 if (match_int(&args
[0], &option
))
1256 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1257 sbi
->s_commit_interval
= HZ
* option
;
1259 case Opt_max_batch_time
:
1260 if (match_int(&args
[0], &option
))
1265 option
= EXT4_DEF_MAX_BATCH_TIME
;
1266 sbi
->s_max_batch_time
= option
;
1268 case Opt_min_batch_time
:
1269 if (match_int(&args
[0], &option
))
1273 sbi
->s_min_batch_time
= option
;
1275 case Opt_data_journal
:
1276 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1278 case Opt_data_ordered
:
1279 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1281 case Opt_data_writeback
:
1282 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1285 if ((sbi
->s_mount_opt
& EXT4_MOUNT_DATA_FLAGS
)
1288 "EXT4-fs: cannot change data "
1289 "mode on remount\n");
1293 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DATA_FLAGS
;
1294 sbi
->s_mount_opt
|= data_opt
;
1297 case Opt_data_err_abort
:
1298 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1300 case Opt_data_err_ignore
:
1301 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1310 if (sb_any_quota_loaded(sb
) &&
1311 !sbi
->s_qf_names
[qtype
]) {
1313 "EXT4-fs: Cannot change journaled "
1314 "quota options when quota turned on.\n");
1317 qname
= match_strdup(&args
[0]);
1320 "EXT4-fs: not enough memory for "
1321 "storing quotafile name.\n");
1324 if (sbi
->s_qf_names
[qtype
] &&
1325 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1327 "EXT4-fs: %s quota file already "
1328 "specified.\n", QTYPE2NAME(qtype
));
1332 sbi
->s_qf_names
[qtype
] = qname
;
1333 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1335 "EXT4-fs: quotafile must be on "
1336 "filesystem root.\n");
1337 kfree(sbi
->s_qf_names
[qtype
]);
1338 sbi
->s_qf_names
[qtype
] = NULL
;
1341 set_opt(sbi
->s_mount_opt
, QUOTA
);
1343 case Opt_offusrjquota
:
1346 case Opt_offgrpjquota
:
1349 if (sb_any_quota_loaded(sb
) &&
1350 sbi
->s_qf_names
[qtype
]) {
1351 printk(KERN_ERR
"EXT4-fs: Cannot change "
1352 "journaled quota options when "
1353 "quota turned on.\n");
1357 * The space will be released later when all options
1358 * are confirmed to be correct
1360 sbi
->s_qf_names
[qtype
] = NULL
;
1362 case Opt_jqfmt_vfsold
:
1363 qfmt
= QFMT_VFS_OLD
;
1365 case Opt_jqfmt_vfsv0
:
1368 if (sb_any_quota_loaded(sb
) &&
1369 sbi
->s_jquota_fmt
!= qfmt
) {
1370 printk(KERN_ERR
"EXT4-fs: Cannot change "
1371 "journaled quota options when "
1372 "quota turned on.\n");
1375 sbi
->s_jquota_fmt
= qfmt
;
1379 set_opt(sbi
->s_mount_opt
, QUOTA
);
1380 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1383 set_opt(sbi
->s_mount_opt
, QUOTA
);
1384 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1387 if (sb_any_quota_loaded(sb
)) {
1388 printk(KERN_ERR
"EXT4-fs: Cannot change quota "
1389 "options when quota turned on.\n");
1392 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1393 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1394 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1401 "EXT4-fs: quota options not supported.\n");
1405 case Opt_offusrjquota
:
1406 case Opt_offgrpjquota
:
1407 case Opt_jqfmt_vfsold
:
1408 case Opt_jqfmt_vfsv0
:
1410 "EXT4-fs: journaled quota options not "
1417 set_opt(sbi
->s_mount_opt
, ABORT
);
1420 if (match_int(&args
[0], &option
))
1423 set_opt(sbi
->s_mount_opt
, BARRIER
);
1425 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1431 printk("EXT4-fs: resize option only available "
1435 if (match_int(&args
[0], &option
) != 0)
1437 *n_blocks_count
= option
;
1440 set_opt(sbi
->s_mount_opt
, NOBH
);
1443 clear_opt(sbi
->s_mount_opt
, NOBH
);
1446 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1447 sb
->s_flags
|= MS_I_VERSION
;
1449 case Opt_nodelalloc
:
1450 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1453 if (match_int(&args
[0], &option
))
1457 sbi
->s_stripe
= option
;
1460 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1462 case Opt_inode_readahead_blks
:
1463 if (match_int(&args
[0], &option
))
1465 if (option
< 0 || option
> (1 << 30))
1467 sbi
->s_inode_readahead_blks
= option
;
1469 case Opt_journal_ioprio
:
1470 if (match_int(&args
[0], &option
))
1472 if (option
< 0 || option
> 7)
1474 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1479 "EXT4-fs: Unrecognized mount option \"%s\" "
1480 "or missing value\n", p
);
1485 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1486 if ((sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
) &&
1487 sbi
->s_qf_names
[USRQUOTA
])
1488 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1490 if ((sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
) &&
1491 sbi
->s_qf_names
[GRPQUOTA
])
1492 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1494 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1495 (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)) ||
1496 (sbi
->s_qf_names
[GRPQUOTA
] &&
1497 (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
))) {
1498 printk(KERN_ERR
"EXT4-fs: old and new quota "
1499 "format mixing.\n");
1503 if (!sbi
->s_jquota_fmt
) {
1504 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1505 "not specified.\n");
1509 if (sbi
->s_jquota_fmt
) {
1510 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1511 "specified with no journaling "
1520 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1523 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1526 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1527 printk(KERN_ERR
"EXT4-fs warning: revision level too high, "
1528 "forcing read-only mode\n");
1533 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1534 printk(KERN_WARNING
"EXT4-fs warning: mounting unchecked fs, "
1535 "running e2fsck is recommended\n");
1536 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1538 "EXT4-fs warning: mounting fs with errors, "
1539 "running e2fsck is recommended\n");
1540 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1541 le16_to_cpu(es
->s_mnt_count
) >=
1542 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1544 "EXT4-fs warning: maximal mount count reached, "
1545 "running e2fsck is recommended\n");
1546 else if (le32_to_cpu(es
->s_checkinterval
) &&
1547 (le32_to_cpu(es
->s_lastcheck
) +
1548 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1550 "EXT4-fs warning: checktime reached, "
1551 "running e2fsck is recommended\n");
1552 if (!sbi
->s_journal
)
1553 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1554 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1555 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1556 le16_add_cpu(&es
->s_mnt_count
, 1);
1557 es
->s_mtime
= cpu_to_le32(get_seconds());
1558 ext4_update_dynamic_rev(sb
);
1560 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1562 ext4_commit_super(sb
, es
, 1);
1563 if (test_opt(sb
, DEBUG
))
1564 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1565 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1567 sbi
->s_groups_count
,
1568 EXT4_BLOCKS_PER_GROUP(sb
),
1569 EXT4_INODES_PER_GROUP(sb
),
1572 if (EXT4_SB(sb
)->s_journal
) {
1573 printk(KERN_INFO
"EXT4 FS on %s, %s journal on %s\n",
1574 sb
->s_id
, EXT4_SB(sb
)->s_journal
->j_inode
? "internal" :
1575 "external", EXT4_SB(sb
)->s_journal
->j_devname
);
1577 printk(KERN_INFO
"EXT4 FS on %s, no journal\n", sb
->s_id
);
1582 static int ext4_fill_flex_info(struct super_block
*sb
)
1584 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1585 struct ext4_group_desc
*gdp
= NULL
;
1586 struct buffer_head
*bh
;
1587 ext4_group_t flex_group_count
;
1588 ext4_group_t flex_group
;
1589 int groups_per_flex
= 0;
1592 if (!sbi
->s_es
->s_log_groups_per_flex
) {
1593 sbi
->s_log_groups_per_flex
= 0;
1597 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1598 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1600 /* We allocate both existing and potentially added groups */
1601 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1602 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1603 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1604 sbi
->s_flex_groups
= kzalloc(flex_group_count
*
1605 sizeof(struct flex_groups
), GFP_KERNEL
);
1606 if (sbi
->s_flex_groups
== NULL
) {
1607 printk(KERN_ERR
"EXT4-fs: not enough memory for "
1608 "%u flex groups\n", flex_group_count
);
1612 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1613 gdp
= ext4_get_group_desc(sb
, i
, &bh
);
1615 flex_group
= ext4_flex_group(sbi
, i
);
1616 sbi
->s_flex_groups
[flex_group
].free_inodes
+=
1617 ext4_free_inodes_count(sb
, gdp
);
1618 sbi
->s_flex_groups
[flex_group
].free_blocks
+=
1619 ext4_free_blks_count(sb
, gdp
);
1627 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1628 struct ext4_group_desc
*gdp
)
1632 if (sbi
->s_es
->s_feature_ro_compat
&
1633 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1634 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1635 __le32 le_group
= cpu_to_le32(block_group
);
1637 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1638 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1639 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1640 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1641 /* for checksum of struct ext4_group_desc do the rest...*/
1642 if ((sbi
->s_es
->s_feature_incompat
&
1643 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1644 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1645 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1646 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1650 return cpu_to_le16(crc
);
1653 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1654 struct ext4_group_desc
*gdp
)
1656 if ((sbi
->s_es
->s_feature_ro_compat
&
1657 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1658 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1664 /* Called at mount-time, super-block is locked */
1665 static int ext4_check_descriptors(struct super_block
*sb
)
1667 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1668 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1669 ext4_fsblk_t last_block
;
1670 ext4_fsblk_t block_bitmap
;
1671 ext4_fsblk_t inode_bitmap
;
1672 ext4_fsblk_t inode_table
;
1673 int flexbg_flag
= 0;
1676 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1679 ext4_debug("Checking group descriptors");
1681 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1682 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1684 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1685 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1687 last_block
= first_block
+
1688 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1690 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1691 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1692 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1693 "Block bitmap for group %u not in group "
1694 "(block %llu)!\n", i
, block_bitmap
);
1697 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1698 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1699 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1700 "Inode bitmap for group %u not in group "
1701 "(block %llu)!\n", i
, inode_bitmap
);
1704 inode_table
= ext4_inode_table(sb
, gdp
);
1705 if (inode_table
< first_block
||
1706 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1707 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1708 "Inode table for group %u not in group "
1709 "(block %llu)!\n", i
, inode_table
);
1712 spin_lock(sb_bgl_lock(sbi
, i
));
1713 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1714 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1715 "Checksum for group %u failed (%u!=%u)\n",
1716 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1717 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1718 if (!(sb
->s_flags
& MS_RDONLY
)) {
1719 spin_unlock(sb_bgl_lock(sbi
, i
));
1723 spin_unlock(sb_bgl_lock(sbi
, i
));
1725 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1728 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1729 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(ext4_count_free_inodes(sb
));
1733 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1734 * the superblock) which were deleted from all directories, but held open by
1735 * a process at the time of a crash. We walk the list and try to delete these
1736 * inodes at recovery time (only with a read-write filesystem).
1738 * In order to keep the orphan inode chain consistent during traversal (in
1739 * case of crash during recovery), we link each inode into the superblock
1740 * orphan list_head and handle it the same way as an inode deletion during
1741 * normal operation (which journals the operations for us).
1743 * We only do an iget() and an iput() on each inode, which is very safe if we
1744 * accidentally point at an in-use or already deleted inode. The worst that
1745 * can happen in this case is that we get a "bit already cleared" message from
1746 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1747 * e2fsck was run on this filesystem, and it must have already done the orphan
1748 * inode cleanup for us, so we can safely abort without any further action.
1750 static void ext4_orphan_cleanup(struct super_block
*sb
,
1751 struct ext4_super_block
*es
)
1753 unsigned int s_flags
= sb
->s_flags
;
1754 int nr_orphans
= 0, nr_truncates
= 0;
1758 if (!es
->s_last_orphan
) {
1759 jbd_debug(4, "no orphan inodes to clean up\n");
1763 if (bdev_read_only(sb
->s_bdev
)) {
1764 printk(KERN_ERR
"EXT4-fs: write access "
1765 "unavailable, skipping orphan cleanup.\n");
1769 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1770 if (es
->s_last_orphan
)
1771 jbd_debug(1, "Errors on filesystem, "
1772 "clearing orphan list.\n");
1773 es
->s_last_orphan
= 0;
1774 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1778 if (s_flags
& MS_RDONLY
) {
1779 printk(KERN_INFO
"EXT4-fs: %s: orphan cleanup on readonly fs\n",
1781 sb
->s_flags
&= ~MS_RDONLY
;
1784 /* Needed for iput() to work correctly and not trash data */
1785 sb
->s_flags
|= MS_ACTIVE
;
1786 /* Turn on quotas so that they are updated correctly */
1787 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1788 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
1789 int ret
= ext4_quota_on_mount(sb
, i
);
1792 "EXT4-fs: Cannot turn on journaled "
1793 "quota: error %d\n", ret
);
1798 while (es
->s_last_orphan
) {
1799 struct inode
*inode
;
1801 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1802 if (IS_ERR(inode
)) {
1803 es
->s_last_orphan
= 0;
1807 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
1809 if (inode
->i_nlink
) {
1811 "%s: truncating inode %lu to %lld bytes\n",
1812 __func__
, inode
->i_ino
, inode
->i_size
);
1813 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1814 inode
->i_ino
, inode
->i_size
);
1815 ext4_truncate(inode
);
1819 "%s: deleting unreferenced inode %lu\n",
1820 __func__
, inode
->i_ino
);
1821 jbd_debug(2, "deleting unreferenced inode %lu\n",
1825 iput(inode
); /* The delete magic happens here! */
1828 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1831 printk(KERN_INFO
"EXT4-fs: %s: %d orphan inode%s deleted\n",
1832 sb
->s_id
, PLURAL(nr_orphans
));
1834 printk(KERN_INFO
"EXT4-fs: %s: %d truncate%s cleaned up\n",
1835 sb
->s_id
, PLURAL(nr_truncates
));
1837 /* Turn quotas off */
1838 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1839 if (sb_dqopt(sb
)->files
[i
])
1840 vfs_quota_off(sb
, i
, 0);
1843 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1846 * Maximal extent format file size.
1847 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1848 * extent format containers, within a sector_t, and within i_blocks
1849 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1850 * so that won't be a limiting factor.
1852 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1854 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
1857 loff_t upper_limit
= MAX_LFS_FILESIZE
;
1859 /* small i_blocks in vfs inode? */
1860 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
1862 * CONFIG_LBD is not enabled implies the inode
1863 * i_block represent total blocks in 512 bytes
1864 * 32 == size of vfs inode i_blocks * 8
1866 upper_limit
= (1LL << 32) - 1;
1868 /* total blocks in file system block size */
1869 upper_limit
>>= (blkbits
- 9);
1870 upper_limit
<<= blkbits
;
1873 /* 32-bit extent-start container, ee_block */
1878 /* Sanity check against vm- & vfs- imposed limits */
1879 if (res
> upper_limit
)
1886 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1887 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1888 * We need to be 1 filesystem block less than the 2^48 sector limit.
1890 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
1892 loff_t res
= EXT4_NDIR_BLOCKS
;
1895 /* This is calculated to be the largest file size for a
1896 * dense, bitmapped file such that the total number of
1897 * sectors in the file, including data and all indirect blocks,
1898 * does not exceed 2^48 -1
1899 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1900 * total number of 512 bytes blocks of the file
1903 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
1905 * !has_huge_files or CONFIG_LBD is not enabled
1906 * implies the inode i_block represent total blocks in
1907 * 512 bytes 32 == size of vfs inode i_blocks * 8
1909 upper_limit
= (1LL << 32) - 1;
1911 /* total blocks in file system block size */
1912 upper_limit
>>= (bits
- 9);
1916 * We use 48 bit ext4_inode i_blocks
1917 * With EXT4_HUGE_FILE_FL set the i_blocks
1918 * represent total number of blocks in
1919 * file system block size
1921 upper_limit
= (1LL << 48) - 1;
1925 /* indirect blocks */
1927 /* double indirect blocks */
1928 meta_blocks
+= 1 + (1LL << (bits
-2));
1929 /* tripple indirect blocks */
1930 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
1932 upper_limit
-= meta_blocks
;
1933 upper_limit
<<= bits
;
1935 res
+= 1LL << (bits
-2);
1936 res
+= 1LL << (2*(bits
-2));
1937 res
+= 1LL << (3*(bits
-2));
1939 if (res
> upper_limit
)
1942 if (res
> MAX_LFS_FILESIZE
)
1943 res
= MAX_LFS_FILESIZE
;
1948 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
1949 ext4_fsblk_t logical_sb_block
, int nr
)
1951 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1952 ext4_group_t bg
, first_meta_bg
;
1955 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1957 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
1959 return logical_sb_block
+ nr
+ 1;
1960 bg
= sbi
->s_desc_per_block
* nr
;
1961 if (ext4_bg_has_super(sb
, bg
))
1963 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
1967 * ext4_get_stripe_size: Get the stripe size.
1968 * @sbi: In memory super block info
1970 * If we have specified it via mount option, then
1971 * use the mount option value. If the value specified at mount time is
1972 * greater than the blocks per group use the super block value.
1973 * If the super block value is greater than blocks per group return 0.
1974 * Allocator needs it be less than blocks per group.
1977 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
1979 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
1980 unsigned long stripe_width
=
1981 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
1983 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
1984 return sbi
->s_stripe
;
1986 if (stripe_width
<= sbi
->s_blocks_per_group
)
1987 return stripe_width
;
1989 if (stride
<= sbi
->s_blocks_per_group
)
1995 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
1996 __releases(kernel_lock
)
1997 __acquires(kernel_lock
)
2000 struct buffer_head
*bh
;
2001 struct ext4_super_block
*es
= NULL
;
2002 struct ext4_sb_info
*sbi
;
2004 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2005 ext4_fsblk_t logical_sb_block
;
2006 unsigned long offset
= 0;
2007 unsigned long journal_devnum
= 0;
2008 unsigned long def_mount_opts
;
2014 unsigned int db_count
;
2016 int needs_recovery
, has_huge_files
;
2020 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2022 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
2026 sbi
->s_blockgroup_lock
=
2027 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
2028 if (!sbi
->s_blockgroup_lock
) {
2032 sb
->s_fs_info
= sbi
;
2033 sbi
->s_mount_opt
= 0;
2034 sbi
->s_resuid
= EXT4_DEF_RESUID
;
2035 sbi
->s_resgid
= EXT4_DEF_RESGID
;
2036 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
2037 sbi
->s_sb_block
= sb_block
;
2041 /* Cleanup superblock name */
2042 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
2045 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
2047 printk(KERN_ERR
"EXT4-fs: unable to set blocksize\n");
2052 * The ext4 superblock will not be buffer aligned for other than 1kB
2053 * block sizes. We need to calculate the offset from buffer start.
2055 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
2056 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2057 offset
= do_div(logical_sb_block
, blocksize
);
2059 logical_sb_block
= sb_block
;
2062 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
2063 printk(KERN_ERR
"EXT4-fs: unable to read superblock\n");
2067 * Note: s_es must be initialized as soon as possible because
2068 * some ext4 macro-instructions depend on its value
2070 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2072 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
2073 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
2076 /* Set defaults before we parse the mount options */
2077 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
2078 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
2079 set_opt(sbi
->s_mount_opt
, DEBUG
);
2080 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
2081 set_opt(sbi
->s_mount_opt
, GRPID
);
2082 if (def_mount_opts
& EXT4_DEFM_UID16
)
2083 set_opt(sbi
->s_mount_opt
, NO_UID32
);
2084 #ifdef CONFIG_EXT4_FS_XATTR
2085 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
2086 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
2088 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2089 if (def_mount_opts
& EXT4_DEFM_ACL
)
2090 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
2092 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
2093 sbi
->s_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
2094 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
2095 sbi
->s_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
2096 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
2097 sbi
->s_mount_opt
|= EXT4_MOUNT_WRITEBACK_DATA
;
2099 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2100 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2101 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2102 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2104 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2106 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2107 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2108 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
2109 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
2110 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
2112 set_opt(sbi
->s_mount_opt
, RESERVATION
);
2113 set_opt(sbi
->s_mount_opt
, BARRIER
);
2116 * enable delayed allocation by default
2117 * Use -o nodelalloc to turn it off
2119 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2122 if (!parse_options((char *) data
, sb
, &journal_devnum
,
2123 &journal_ioprio
, NULL
, 0))
2126 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2127 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2129 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2130 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2131 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2132 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2134 "EXT4-fs warning: feature flags set on rev 0 fs, "
2135 "running e2fsck is recommended\n");
2138 * Check feature flags regardless of the revision level, since we
2139 * previously didn't change the revision level when setting the flags,
2140 * so there is a chance incompat flags are set on a rev 0 filesystem.
2142 features
= EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
);
2144 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount because of "
2145 "unsupported optional features (%x).\n", sb
->s_id
,
2146 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2147 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2150 features
= EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
);
2151 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
2152 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount RDWR because of "
2153 "unsupported optional features (%x).\n", sb
->s_id
,
2154 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2155 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2158 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2159 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
2160 if (has_huge_files
) {
2162 * Large file size enabled file system can only be
2163 * mount if kernel is build with CONFIG_LBD
2165 if (sizeof(root
->i_blocks
) < sizeof(u64
) &&
2166 !(sb
->s_flags
& MS_RDONLY
)) {
2167 printk(KERN_ERR
"EXT4-fs: %s: Filesystem with huge "
2168 "files cannot be mounted read-write "
2169 "without CONFIG_LBD.\n", sb
->s_id
);
2173 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2175 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2176 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2178 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2179 blocksize
, sb
->s_id
);
2183 if (sb
->s_blocksize
!= blocksize
) {
2185 /* Validate the filesystem blocksize */
2186 if (!sb_set_blocksize(sb
, blocksize
)) {
2187 printk(KERN_ERR
"EXT4-fs: bad block size %d.\n",
2193 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2194 offset
= do_div(logical_sb_block
, blocksize
);
2195 bh
= sb_bread(sb
, logical_sb_block
);
2198 "EXT4-fs: Can't read superblock on 2nd try.\n");
2201 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2203 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2205 "EXT4-fs: Magic mismatch, very weird !\n");
2210 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
2212 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
2214 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2215 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2216 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2218 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2219 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2220 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2221 (!is_power_of_2(sbi
->s_inode_size
)) ||
2222 (sbi
->s_inode_size
> blocksize
)) {
2224 "EXT4-fs: unsupported inode size: %d\n",
2228 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2229 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2231 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2232 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2233 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2234 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2235 !is_power_of_2(sbi
->s_desc_size
)) {
2237 "EXT4-fs: unsupported descriptor size %lu\n",
2242 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2243 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2244 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2245 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2247 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2248 if (sbi
->s_inodes_per_block
== 0)
2250 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2251 sbi
->s_inodes_per_block
;
2252 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2254 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2255 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2256 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2257 for (i
= 0; i
< 4; i
++)
2258 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2259 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2260 i
= le32_to_cpu(es
->s_flags
);
2261 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
2262 sbi
->s_hash_unsigned
= 3;
2263 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
2264 #ifdef __CHAR_UNSIGNED__
2265 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
2266 sbi
->s_hash_unsigned
= 3;
2268 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
2273 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2275 "EXT4-fs: #blocks per group too big: %lu\n",
2276 sbi
->s_blocks_per_group
);
2279 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2281 "EXT4-fs: #inodes per group too big: %lu\n",
2282 sbi
->s_inodes_per_group
);
2286 if (ext4_blocks_count(es
) >
2287 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) {
2288 printk(KERN_ERR
"EXT4-fs: filesystem on %s:"
2289 " too large to mount safely\n", sb
->s_id
);
2290 if (sizeof(sector_t
) < 8)
2291 printk(KERN_WARNING
"EXT4-fs: CONFIG_LBD not "
2296 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2300 * It makes no sense for the first data block to be beyond the end
2301 * of the filesystem.
2303 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
2304 printk(KERN_WARNING
"EXT4-fs: bad geometry: first data"
2305 "block %u is beyond end of filesystem (%llu)\n",
2306 le32_to_cpu(es
->s_first_data_block
),
2307 ext4_blocks_count(es
));
2310 blocks_count
= (ext4_blocks_count(es
) -
2311 le32_to_cpu(es
->s_first_data_block
) +
2312 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2313 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2314 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
2315 printk(KERN_WARNING
"EXT4-fs: groups count too large: %u "
2316 "(block count %llu, first data block %u, "
2317 "blocks per group %lu)\n", sbi
->s_groups_count
,
2318 ext4_blocks_count(es
),
2319 le32_to_cpu(es
->s_first_data_block
),
2320 EXT4_BLOCKS_PER_GROUP(sb
));
2323 sbi
->s_groups_count
= blocks_count
;
2324 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2325 EXT4_DESC_PER_BLOCK(sb
);
2326 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2328 if (sbi
->s_group_desc
== NULL
) {
2329 printk(KERN_ERR
"EXT4-fs: not enough memory\n");
2333 #ifdef CONFIG_PROC_FS
2335 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2338 proc_create_data("inode_readahead_blks", 0644, sbi
->s_proc
,
2340 &sbi
->s_inode_readahead_blks
);
2343 bgl_lock_init(sbi
->s_blockgroup_lock
);
2345 for (i
= 0; i
< db_count
; i
++) {
2346 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2347 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2348 if (!sbi
->s_group_desc
[i
]) {
2349 printk(KERN_ERR
"EXT4-fs: "
2350 "can't read group descriptor %d\n", i
);
2355 if (!ext4_check_descriptors(sb
)) {
2356 printk(KERN_ERR
"EXT4-fs: group descriptors corrupted!\n");
2359 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2360 if (!ext4_fill_flex_info(sb
)) {
2362 "EXT4-fs: unable to initialize "
2363 "flex_bg meta info!\n");
2367 sbi
->s_gdb_count
= db_count
;
2368 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2369 spin_lock_init(&sbi
->s_next_gen_lock
);
2371 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2372 ext4_count_free_blocks(sb
));
2374 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2375 ext4_count_free_inodes(sb
));
2378 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2379 ext4_count_dirs(sb
));
2382 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
2385 printk(KERN_ERR
"EXT4-fs: insufficient memory\n");
2389 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2392 * set up enough so that it can read an inode
2394 sb
->s_op
= &ext4_sops
;
2395 sb
->s_export_op
= &ext4_export_ops
;
2396 sb
->s_xattr
= ext4_xattr_handlers
;
2398 sb
->s_qcop
= &ext4_qctl_operations
;
2399 sb
->dq_op
= &ext4_quota_operations
;
2401 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2405 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2406 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2407 EXT4_FEATURE_INCOMPAT_RECOVER
));
2410 * The first inode we look at is the journal inode. Don't try
2411 * root first: it may be modified in the journal!
2413 if (!test_opt(sb
, NOLOAD
) &&
2414 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2415 if (ext4_load_journal(sb
, es
, journal_devnum
))
2417 if (!(sb
->s_flags
& MS_RDONLY
) &&
2418 EXT4_SB(sb
)->s_journal
->j_failed_commit
) {
2419 printk(KERN_CRIT
"EXT4-fs error (device %s): "
2420 "ext4_fill_super: Journal transaction "
2421 "%u is corrupt\n", sb
->s_id
,
2422 EXT4_SB(sb
)->s_journal
->j_failed_commit
);
2423 if (test_opt(sb
, ERRORS_RO
)) {
2425 "Mounting filesystem read-only\n");
2426 sb
->s_flags
|= MS_RDONLY
;
2427 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2428 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2430 if (test_opt(sb
, ERRORS_PANIC
)) {
2431 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2432 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2433 ext4_commit_super(sb
, es
, 1);
2437 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
2438 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2439 printk(KERN_ERR
"EXT4-fs: required journal recovery "
2440 "suppressed and not mounted read-only\n");
2443 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
2444 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2445 sbi
->s_journal
= NULL
;
2450 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2451 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2452 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2453 printk(KERN_ERR
"EXT4-fs: Failed to set 64-bit journal feature\n");
2457 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2458 jbd2_journal_set_features(sbi
->s_journal
,
2459 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2460 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2461 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2462 jbd2_journal_set_features(sbi
->s_journal
,
2463 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2464 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2465 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2467 jbd2_journal_clear_features(sbi
->s_journal
,
2468 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2469 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2472 /* We have now updated the journal if required, so we can
2473 * validate the data journaling mode. */
2474 switch (test_opt(sb
, DATA_FLAGS
)) {
2476 /* No mode set, assume a default based on the journal
2477 * capabilities: ORDERED_DATA if the journal can
2478 * cope, else JOURNAL_DATA
2480 if (jbd2_journal_check_available_features
2481 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2482 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2484 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2487 case EXT4_MOUNT_ORDERED_DATA
:
2488 case EXT4_MOUNT_WRITEBACK_DATA
:
2489 if (!jbd2_journal_check_available_features
2490 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2491 printk(KERN_ERR
"EXT4-fs: Journal does not support "
2492 "requested data journaling mode\n");
2498 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
2502 if (test_opt(sb
, NOBH
)) {
2503 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2504 printk(KERN_WARNING
"EXT4-fs: Ignoring nobh option - "
2505 "its supported only with writeback mode\n");
2506 clear_opt(sbi
->s_mount_opt
, NOBH
);
2510 * The jbd2_journal_load will have done any necessary log recovery,
2511 * so we can safely mount the rest of the filesystem now.
2514 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2516 printk(KERN_ERR
"EXT4-fs: get root inode failed\n");
2517 ret
= PTR_ERR(root
);
2520 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2522 printk(KERN_ERR
"EXT4-fs: corrupt root inode, run e2fsck\n");
2525 sb
->s_root
= d_alloc_root(root
);
2527 printk(KERN_ERR
"EXT4-fs: get root dentry failed\n");
2533 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
2535 /* determine the minimum size of new large inodes, if present */
2536 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2537 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2538 EXT4_GOOD_OLD_INODE_SIZE
;
2539 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2540 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2541 if (sbi
->s_want_extra_isize
<
2542 le16_to_cpu(es
->s_want_extra_isize
))
2543 sbi
->s_want_extra_isize
=
2544 le16_to_cpu(es
->s_want_extra_isize
);
2545 if (sbi
->s_want_extra_isize
<
2546 le16_to_cpu(es
->s_min_extra_isize
))
2547 sbi
->s_want_extra_isize
=
2548 le16_to_cpu(es
->s_min_extra_isize
);
2551 /* Check if enough inode space is available */
2552 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2553 sbi
->s_inode_size
) {
2554 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2555 EXT4_GOOD_OLD_INODE_SIZE
;
2556 printk(KERN_INFO
"EXT4-fs: required extra inode space not"
2560 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
2561 printk(KERN_WARNING
"EXT4-fs: Ignoring delalloc option - "
2562 "requested data journaling mode\n");
2563 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2564 } else if (test_opt(sb
, DELALLOC
))
2565 printk(KERN_INFO
"EXT4-fs: delayed allocation enabled\n");
2568 err
= ext4_mb_init(sb
, needs_recovery
);
2570 printk(KERN_ERR
"EXT4-fs: failed to initalize mballoc (%d)\n",
2576 * akpm: core read_super() calls in here with the superblock locked.
2577 * That deadlocks, because orphan cleanup needs to lock the superblock
2578 * in numerous places. Here we just pop the lock - it's relatively
2579 * harmless, because we are now ready to accept write_super() requests,
2580 * and aviro says that's the only reason for hanging onto the
2583 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
2584 ext4_orphan_cleanup(sb
, es
);
2585 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
2586 if (needs_recovery
) {
2587 printk(KERN_INFO
"EXT4-fs: recovery complete.\n");
2588 ext4_mark_recovery_complete(sb
, es
);
2590 if (EXT4_SB(sb
)->s_journal
) {
2591 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2592 descr
= " journalled data mode";
2593 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2594 descr
= " ordered data mode";
2596 descr
= " writeback data mode";
2598 descr
= "out journal";
2600 printk(KERN_INFO
"EXT4-fs: mounted filesystem %s with%s\n",
2608 printk(KERN_ERR
"VFS: Can't find ext4 filesystem on dev %s.\n",
2613 printk(KERN_ERR
"EXT4-fs (device %s): mount failed\n", sb
->s_id
);
2614 if (sbi
->s_journal
) {
2615 jbd2_journal_destroy(sbi
->s_journal
);
2616 sbi
->s_journal
= NULL
;
2619 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2620 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2621 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2622 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
2624 for (i
= 0; i
< db_count
; i
++)
2625 brelse(sbi
->s_group_desc
[i
]);
2626 kfree(sbi
->s_group_desc
);
2629 remove_proc_entry("inode_readahead_blks", sbi
->s_proc
);
2630 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
2633 for (i
= 0; i
< MAXQUOTAS
; i
++)
2634 kfree(sbi
->s_qf_names
[i
]);
2636 ext4_blkdev_remove(sbi
);
2639 sb
->s_fs_info
= NULL
;
2646 * Setup any per-fs journal parameters now. We'll do this both on
2647 * initial mount, once the journal has been initialised but before we've
2648 * done any recovery; and again on any subsequent remount.
2650 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
2652 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2654 journal
->j_commit_interval
= sbi
->s_commit_interval
;
2655 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
2656 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
2658 spin_lock(&journal
->j_state_lock
);
2659 if (test_opt(sb
, BARRIER
))
2660 journal
->j_flags
|= JBD2_BARRIER
;
2662 journal
->j_flags
&= ~JBD2_BARRIER
;
2663 if (test_opt(sb
, DATA_ERR_ABORT
))
2664 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
2666 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
2667 spin_unlock(&journal
->j_state_lock
);
2670 static journal_t
*ext4_get_journal(struct super_block
*sb
,
2671 unsigned int journal_inum
)
2673 struct inode
*journal_inode
;
2676 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2678 /* First, test for the existence of a valid inode on disk. Bad
2679 * things happen if we iget() an unused inode, as the subsequent
2680 * iput() will try to delete it. */
2682 journal_inode
= ext4_iget(sb
, journal_inum
);
2683 if (IS_ERR(journal_inode
)) {
2684 printk(KERN_ERR
"EXT4-fs: no journal found.\n");
2687 if (!journal_inode
->i_nlink
) {
2688 make_bad_inode(journal_inode
);
2689 iput(journal_inode
);
2690 printk(KERN_ERR
"EXT4-fs: journal inode is deleted.\n");
2694 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2695 journal_inode
, journal_inode
->i_size
);
2696 if (!S_ISREG(journal_inode
->i_mode
)) {
2697 printk(KERN_ERR
"EXT4-fs: invalid journal inode.\n");
2698 iput(journal_inode
);
2702 journal
= jbd2_journal_init_inode(journal_inode
);
2704 printk(KERN_ERR
"EXT4-fs: Could not load journal inode\n");
2705 iput(journal_inode
);
2708 journal
->j_private
= sb
;
2709 ext4_init_journal_params(sb
, journal
);
2713 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
2716 struct buffer_head
*bh
;
2720 int hblock
, blocksize
;
2721 ext4_fsblk_t sb_block
;
2722 unsigned long offset
;
2723 struct ext4_super_block
*es
;
2724 struct block_device
*bdev
;
2726 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2728 bdev
= ext4_blkdev_get(j_dev
);
2732 if (bd_claim(bdev
, sb
)) {
2734 "EXT4-fs: failed to claim external journal device.\n");
2735 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
2739 blocksize
= sb
->s_blocksize
;
2740 hblock
= bdev_hardsect_size(bdev
);
2741 if (blocksize
< hblock
) {
2743 "EXT4-fs: blocksize too small for journal device.\n");
2747 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
2748 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
2749 set_blocksize(bdev
, blocksize
);
2750 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
2751 printk(KERN_ERR
"EXT4-fs: couldn't read superblock of "
2752 "external journal\n");
2756 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2757 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
2758 !(le32_to_cpu(es
->s_feature_incompat
) &
2759 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
2760 printk(KERN_ERR
"EXT4-fs: external journal has "
2761 "bad superblock\n");
2766 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
2767 printk(KERN_ERR
"EXT4-fs: journal UUID does not match\n");
2772 len
= ext4_blocks_count(es
);
2773 start
= sb_block
+ 1;
2774 brelse(bh
); /* we're done with the superblock */
2776 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
2777 start
, len
, blocksize
);
2779 printk(KERN_ERR
"EXT4-fs: failed to create device journal\n");
2782 journal
->j_private
= sb
;
2783 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
2784 wait_on_buffer(journal
->j_sb_buffer
);
2785 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
2786 printk(KERN_ERR
"EXT4-fs: I/O error on journal device\n");
2789 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
2790 printk(KERN_ERR
"EXT4-fs: External journal has more than one "
2791 "user (unsupported) - %d\n",
2792 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
2795 EXT4_SB(sb
)->journal_bdev
= bdev
;
2796 ext4_init_journal_params(sb
, journal
);
2799 jbd2_journal_destroy(journal
);
2801 ext4_blkdev_put(bdev
);
2805 static int ext4_load_journal(struct super_block
*sb
,
2806 struct ext4_super_block
*es
,
2807 unsigned long journal_devnum
)
2810 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
2813 int really_read_only
;
2815 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2817 if (journal_devnum
&&
2818 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2819 printk(KERN_INFO
"EXT4-fs: external journal device major/minor "
2820 "numbers have changed\n");
2821 journal_dev
= new_decode_dev(journal_devnum
);
2823 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
2825 really_read_only
= bdev_read_only(sb
->s_bdev
);
2828 * Are we loading a blank journal or performing recovery after a
2829 * crash? For recovery, we need to check in advance whether we
2830 * can get read-write access to the device.
2833 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2834 if (sb
->s_flags
& MS_RDONLY
) {
2835 printk(KERN_INFO
"EXT4-fs: INFO: recovery "
2836 "required on readonly filesystem.\n");
2837 if (really_read_only
) {
2838 printk(KERN_ERR
"EXT4-fs: write access "
2839 "unavailable, cannot proceed.\n");
2842 printk(KERN_INFO
"EXT4-fs: write access will "
2843 "be enabled during recovery.\n");
2847 if (journal_inum
&& journal_dev
) {
2848 printk(KERN_ERR
"EXT4-fs: filesystem has both journal "
2849 "and inode journals!\n");
2854 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
2857 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
2861 if (journal
->j_flags
& JBD2_BARRIER
)
2862 printk(KERN_INFO
"EXT4-fs: barriers enabled\n");
2864 printk(KERN_INFO
"EXT4-fs: barriers disabled\n");
2866 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
2867 err
= jbd2_journal_update_format(journal
);
2869 printk(KERN_ERR
"EXT4-fs: error updating journal.\n");
2870 jbd2_journal_destroy(journal
);
2875 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
2876 err
= jbd2_journal_wipe(journal
, !really_read_only
);
2878 err
= jbd2_journal_load(journal
);
2881 printk(KERN_ERR
"EXT4-fs: error loading journal.\n");
2882 jbd2_journal_destroy(journal
);
2886 EXT4_SB(sb
)->s_journal
= journal
;
2887 ext4_clear_journal_err(sb
, es
);
2889 if (journal_devnum
&&
2890 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2891 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2894 /* Make sure we flush the recovery flag to disk. */
2895 ext4_commit_super(sb
, es
, 1);
2901 static int ext4_commit_super(struct super_block
*sb
,
2902 struct ext4_super_block
*es
, int sync
)
2904 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
2909 if (buffer_write_io_error(sbh
)) {
2911 * Oh, dear. A previous attempt to write the
2912 * superblock failed. This could happen because the
2913 * USB device was yanked out. Or it could happen to
2914 * be a transient write error and maybe the block will
2915 * be remapped. Nothing we can do but to retry the
2916 * write and hope for the best.
2918 printk(KERN_ERR
"EXT4-fs: previous I/O error to "
2919 "superblock detected for %s.\n", sb
->s_id
);
2920 clear_buffer_write_io_error(sbh
);
2921 set_buffer_uptodate(sbh
);
2923 es
->s_wtime
= cpu_to_le32(get_seconds());
2924 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
2925 &EXT4_SB(sb
)->s_freeblocks_counter
));
2926 es
->s_free_inodes_count
= cpu_to_le32(percpu_counter_sum_positive(
2927 &EXT4_SB(sb
)->s_freeinodes_counter
));
2929 BUFFER_TRACE(sbh
, "marking dirty");
2930 mark_buffer_dirty(sbh
);
2932 error
= sync_dirty_buffer(sbh
);
2936 error
= buffer_write_io_error(sbh
);
2938 printk(KERN_ERR
"EXT4-fs: I/O error while writing "
2939 "superblock for %s.\n", sb
->s_id
);
2940 clear_buffer_write_io_error(sbh
);
2941 set_buffer_uptodate(sbh
);
2949 * Have we just finished recovery? If so, and if we are mounting (or
2950 * remounting) the filesystem readonly, then we will end up with a
2951 * consistent fs on disk. Record that fact.
2953 static void ext4_mark_recovery_complete(struct super_block
*sb
,
2954 struct ext4_super_block
*es
)
2956 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
2958 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2959 BUG_ON(journal
!= NULL
);
2962 jbd2_journal_lock_updates(journal
);
2963 if (jbd2_journal_flush(journal
) < 0)
2967 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
2968 sb
->s_flags
& MS_RDONLY
) {
2969 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2971 ext4_commit_super(sb
, es
, 1);
2976 jbd2_journal_unlock_updates(journal
);
2980 * If we are mounting (or read-write remounting) a filesystem whose journal
2981 * has recorded an error from a previous lifetime, move that error to the
2982 * main filesystem now.
2984 static void ext4_clear_journal_err(struct super_block
*sb
,
2985 struct ext4_super_block
*es
)
2991 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2993 journal
= EXT4_SB(sb
)->s_journal
;
2996 * Now check for any error status which may have been recorded in the
2997 * journal by a prior ext4_error() or ext4_abort()
3000 j_errno
= jbd2_journal_errno(journal
);
3004 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
3005 ext4_warning(sb
, __func__
, "Filesystem error recorded "
3006 "from previous mount: %s", errstr
);
3007 ext4_warning(sb
, __func__
, "Marking fs in need of "
3008 "filesystem check.");
3010 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
3011 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
3012 ext4_commit_super(sb
, es
, 1);
3014 jbd2_journal_clear_err(journal
);
3019 * Force the running and committing transactions to commit,
3020 * and wait on the commit.
3022 int ext4_force_commit(struct super_block
*sb
)
3027 if (sb
->s_flags
& MS_RDONLY
)
3030 journal
= EXT4_SB(sb
)->s_journal
;
3033 ret
= ext4_journal_force_commit(journal
);
3040 * Ext4 always journals updates to the superblock itself, so we don't
3041 * have to propagate any other updates to the superblock on disk at this
3042 * point. (We can probably nuke this function altogether, and remove
3043 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3045 static void ext4_write_super(struct super_block
*sb
)
3047 if (EXT4_SB(sb
)->s_journal
) {
3048 if (mutex_trylock(&sb
->s_lock
) != 0)
3052 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
3056 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
3061 trace_mark(ext4_sync_fs
, "dev %s wait %d", sb
->s_id
, wait
);
3063 if (EXT4_SB(sb
)->s_journal
) {
3064 if (jbd2_journal_start_commit(EXT4_SB(sb
)->s_journal
,
3067 jbd2_log_wait_commit(EXT4_SB(sb
)->s_journal
,
3071 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, wait
);
3077 * LVM calls this function before a (read-only) snapshot is created. This
3078 * gives us a chance to flush the journal completely and mark the fs clean.
3080 static int ext4_freeze(struct super_block
*sb
)
3086 if (!(sb
->s_flags
& MS_RDONLY
)) {
3087 journal
= EXT4_SB(sb
)->s_journal
;
3090 /* Now we set up the journal barrier. */
3091 jbd2_journal_lock_updates(journal
);
3094 * We don't want to clear needs_recovery flag when we
3095 * failed to flush the journal.
3097 error
= jbd2_journal_flush(journal
);
3102 /* Journal blocked and flushed, clear needs_recovery flag. */
3103 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3104 error
= ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
3110 jbd2_journal_unlock_updates(journal
);
3115 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3116 * flag here, even though the filesystem is not technically dirty yet.
3118 static int ext4_unfreeze(struct super_block
*sb
)
3120 if (EXT4_SB(sb
)->s_journal
&& !(sb
->s_flags
& MS_RDONLY
)) {
3122 /* Reser the needs_recovery flag before the fs is unlocked. */
3123 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3124 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
3126 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3131 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3133 struct ext4_super_block
*es
;
3134 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3135 ext4_fsblk_t n_blocks_count
= 0;
3136 unsigned long old_sb_flags
;
3137 struct ext4_mount_options old_opts
;
3139 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3145 /* Store the original options */
3146 old_sb_flags
= sb
->s_flags
;
3147 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3148 old_opts
.s_resuid
= sbi
->s_resuid
;
3149 old_opts
.s_resgid
= sbi
->s_resgid
;
3150 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3151 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
3152 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
3154 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3155 for (i
= 0; i
< MAXQUOTAS
; i
++)
3156 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3158 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
3159 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
3162 * Allow the "check" option to be passed as a remount option.
3164 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
3165 &n_blocks_count
, 1)) {
3170 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
)
3171 ext4_abort(sb
, __func__
, "Abort forced by user");
3173 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3174 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
3178 if (sbi
->s_journal
) {
3179 ext4_init_journal_params(sb
, sbi
->s_journal
);
3180 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3183 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3184 n_blocks_count
> ext4_blocks_count(es
)) {
3185 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
) {
3190 if (*flags
& MS_RDONLY
) {
3192 * First of all, the unconditional stuff we have to do
3193 * to disable replay of the journal when we next remount
3195 sb
->s_flags
|= MS_RDONLY
;
3198 * OK, test if we are remounting a valid rw partition
3199 * readonly, and if so set the rdonly flag and then
3200 * mark the partition as valid again.
3202 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3203 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3204 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3207 * We have to unlock super so that we can wait for
3210 if (sbi
->s_journal
) {
3212 ext4_mark_recovery_complete(sb
, es
);
3217 if ((ret
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3218 ~EXT4_FEATURE_RO_COMPAT_SUPP
))) {
3219 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3220 "remount RDWR because of unsupported "
3221 "optional features (%x).\n", sb
->s_id
,
3222 (le32_to_cpu(sbi
->s_es
->s_feature_ro_compat
) &
3223 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
3229 * Make sure the group descriptor checksums
3230 * are sane. If they aren't, refuse to
3233 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3234 struct ext4_group_desc
*gdp
=
3235 ext4_get_group_desc(sb
, g
, NULL
);
3237 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3239 "EXT4-fs: ext4_remount: "
3240 "Checksum for group %u failed (%u!=%u)\n",
3241 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3242 le16_to_cpu(gdp
->bg_checksum
));
3249 * If we have an unprocessed orphan list hanging
3250 * around from a previously readonly bdev mount,
3251 * require a full umount/remount for now.
3253 if (es
->s_last_orphan
) {
3254 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3255 "remount RDWR because of unprocessed "
3256 "orphan inode list. Please "
3257 "umount/remount instead.\n",
3264 * Mounting a RDONLY partition read-write, so reread
3265 * and store the current valid flag. (It may have
3266 * been changed by e2fsck since we originally mounted
3270 ext4_clear_journal_err(sb
, es
);
3271 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3272 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3274 if (!ext4_setup_super(sb
, es
, 0))
3275 sb
->s_flags
&= ~MS_RDONLY
;
3278 if (sbi
->s_journal
== NULL
)
3279 ext4_commit_super(sb
, es
, 1);
3282 /* Release old quota file names */
3283 for (i
= 0; i
< MAXQUOTAS
; i
++)
3284 if (old_opts
.s_qf_names
[i
] &&
3285 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3286 kfree(old_opts
.s_qf_names
[i
]);
3290 sb
->s_flags
= old_sb_flags
;
3291 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3292 sbi
->s_resuid
= old_opts
.s_resuid
;
3293 sbi
->s_resgid
= old_opts
.s_resgid
;
3294 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3295 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
3296 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
3298 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3299 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3300 if (sbi
->s_qf_names
[i
] &&
3301 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3302 kfree(sbi
->s_qf_names
[i
]);
3303 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3309 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3311 struct super_block
*sb
= dentry
->d_sb
;
3312 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3313 struct ext4_super_block
*es
= sbi
->s_es
;
3316 if (test_opt(sb
, MINIX_DF
)) {
3317 sbi
->s_overhead_last
= 0;
3318 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3319 ext4_group_t ngroups
= sbi
->s_groups_count
, i
;
3320 ext4_fsblk_t overhead
= 0;
3324 * Compute the overhead (FS structures). This is constant
3325 * for a given filesystem unless the number of block groups
3326 * changes so we cache the previous value until it does.
3330 * All of the blocks before first_data_block are
3333 overhead
= le32_to_cpu(es
->s_first_data_block
);
3336 * Add the overhead attributed to the superblock and
3337 * block group descriptors. If the sparse superblocks
3338 * feature is turned on, then not all groups have this.
3340 for (i
= 0; i
< ngroups
; i
++) {
3341 overhead
+= ext4_bg_has_super(sb
, i
) +
3342 ext4_bg_num_gdb(sb
, i
);
3347 * Every block group has an inode bitmap, a block
3348 * bitmap, and an inode table.
3350 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3351 sbi
->s_overhead_last
= overhead
;
3353 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3356 buf
->f_type
= EXT4_SUPER_MAGIC
;
3357 buf
->f_bsize
= sb
->s_blocksize
;
3358 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3359 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3360 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3361 ext4_free_blocks_count_set(es
, buf
->f_bfree
);
3362 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3363 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3365 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3366 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3367 es
->s_free_inodes_count
= cpu_to_le32(buf
->f_ffree
);
3368 buf
->f_namelen
= EXT4_NAME_LEN
;
3369 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3370 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3371 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3372 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3376 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3377 * is locked for write. Otherwise the are possible deadlocks:
3378 * Process 1 Process 2
3379 * ext4_create() quota_sync()
3380 * jbd2_journal_start() write_dquot()
3381 * vfs_dq_init() down(dqio_mutex)
3382 * down(dqio_mutex) jbd2_journal_start()
3388 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3390 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3393 static int ext4_write_dquot(struct dquot
*dquot
)
3397 struct inode
*inode
;
3399 inode
= dquot_to_inode(dquot
);
3400 handle
= ext4_journal_start(inode
,
3401 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3403 return PTR_ERR(handle
);
3404 ret
= dquot_commit(dquot
);
3405 err
= ext4_journal_stop(handle
);
3411 static int ext4_acquire_dquot(struct dquot
*dquot
)
3416 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3417 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3419 return PTR_ERR(handle
);
3420 ret
= dquot_acquire(dquot
);
3421 err
= ext4_journal_stop(handle
);
3427 static int ext4_release_dquot(struct dquot
*dquot
)
3432 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3433 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3434 if (IS_ERR(handle
)) {
3435 /* Release dquot anyway to avoid endless cycle in dqput() */
3436 dquot_release(dquot
);
3437 return PTR_ERR(handle
);
3439 ret
= dquot_release(dquot
);
3440 err
= ext4_journal_stop(handle
);
3446 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3448 /* Are we journaling quotas? */
3449 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3450 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3451 dquot_mark_dquot_dirty(dquot
);
3452 return ext4_write_dquot(dquot
);
3454 return dquot_mark_dquot_dirty(dquot
);
3458 static int ext4_write_info(struct super_block
*sb
, int type
)
3463 /* Data block + inode block */
3464 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3466 return PTR_ERR(handle
);
3467 ret
= dquot_commit_info(sb
, type
);
3468 err
= ext4_journal_stop(handle
);
3475 * Turn on quotas during mount time - we need to find
3476 * the quota file and such...
3478 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3480 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3481 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3485 * Standard function to be called on quota_on
3487 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3488 char *name
, int remount
)
3493 if (!test_opt(sb
, QUOTA
))
3495 /* When remounting, no checks are needed and in fact, name is NULL */
3497 return vfs_quota_on(sb
, type
, format_id
, name
, remount
);
3499 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
3503 /* Quotafile not on the same filesystem? */
3504 if (path
.mnt
->mnt_sb
!= sb
) {
3508 /* Journaling quota? */
3509 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3510 /* Quotafile not in fs root? */
3511 if (path
.dentry
->d_parent
!= sb
->s_root
)
3513 "EXT4-fs: Quota file not on filesystem root. "
3514 "Journaled quota will not work.\n");
3518 * When we journal data on quota file, we have to flush journal to see
3519 * all updates to the file when we bypass pagecache...
3521 if (EXT4_SB(sb
)->s_journal
&&
3522 ext4_should_journal_data(path
.dentry
->d_inode
)) {
3524 * We don't need to lock updates but journal_flush() could
3525 * otherwise be livelocked...
3527 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3528 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3529 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3536 err
= vfs_quota_on_path(sb
, type
, format_id
, &path
);
3541 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3542 * acquiring the locks... As quota files are never truncated and quota code
3543 * itself serializes the operations (and noone else should touch the files)
3544 * we don't have to be afraid of races */
3545 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3546 size_t len
, loff_t off
)
3548 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3549 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3551 int offset
= off
& (sb
->s_blocksize
- 1);
3554 struct buffer_head
*bh
;
3555 loff_t i_size
= i_size_read(inode
);
3559 if (off
+len
> i_size
)
3562 while (toread
> 0) {
3563 tocopy
= sb
->s_blocksize
- offset
< toread
?
3564 sb
->s_blocksize
- offset
: toread
;
3565 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3568 if (!bh
) /* A hole? */
3569 memset(data
, 0, tocopy
);
3571 memcpy(data
, bh
->b_data
+offset
, tocopy
);
3581 /* Write to quotafile (we know the transaction is already started and has
3582 * enough credits) */
3583 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
3584 const char *data
, size_t len
, loff_t off
)
3586 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3587 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3589 int offset
= off
& (sb
->s_blocksize
- 1);
3591 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
3592 size_t towrite
= len
;
3593 struct buffer_head
*bh
;
3594 handle_t
*handle
= journal_current_handle();
3596 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
3597 printk(KERN_WARNING
"EXT4-fs: Quota write (off=%llu, len=%llu)"
3598 " cancelled because transaction is not started.\n",
3599 (unsigned long long)off
, (unsigned long long)len
);
3602 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
3603 while (towrite
> 0) {
3604 tocopy
= sb
->s_blocksize
- offset
< towrite
?
3605 sb
->s_blocksize
- offset
: towrite
;
3606 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
3609 if (journal_quota
) {
3610 err
= ext4_journal_get_write_access(handle
, bh
);
3617 memcpy(bh
->b_data
+offset
, data
, tocopy
);
3618 flush_dcache_page(bh
->b_page
);
3621 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
3623 /* Always do at least ordered writes for quotas */
3624 err
= ext4_jbd2_file_inode(handle
, inode
);
3625 mark_buffer_dirty(bh
);
3636 if (len
== towrite
) {
3637 mutex_unlock(&inode
->i_mutex
);
3640 if (inode
->i_size
< off
+len
-towrite
) {
3641 i_size_write(inode
, off
+len
-towrite
);
3642 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
3644 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3645 ext4_mark_inode_dirty(handle
, inode
);
3646 mutex_unlock(&inode
->i_mutex
);
3647 return len
- towrite
;
3652 static int ext4_get_sb(struct file_system_type
*fs_type
,
3653 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3655 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3658 #ifdef CONFIG_PROC_FS
3659 static int ext4_ui_proc_show(struct seq_file
*m
, void *v
)
3661 unsigned int *p
= m
->private;
3663 seq_printf(m
, "%u\n", *p
);
3667 static int ext4_ui_proc_open(struct inode
*inode
, struct file
*file
)
3669 return single_open(file
, ext4_ui_proc_show
, PDE(inode
)->data
);
3672 static ssize_t
ext4_ui_proc_write(struct file
*file
, const char __user
*buf
,
3673 size_t cnt
, loff_t
*ppos
)
3675 unsigned long *p
= PDE(file
->f_path
.dentry
->d_inode
)->data
;
3678 if (cnt
>= sizeof(str
))
3680 if (copy_from_user(str
, buf
, cnt
))
3683 *p
= simple_strtoul(str
, NULL
, 0);
3687 const struct file_operations ext4_ui_proc_fops
= {
3688 .owner
= THIS_MODULE
,
3689 .open
= ext4_ui_proc_open
,
3691 .llseek
= seq_lseek
,
3692 .release
= single_release
,
3693 .write
= ext4_ui_proc_write
,
3697 static struct file_system_type ext4_fs_type
= {
3698 .owner
= THIS_MODULE
,
3700 .get_sb
= ext4_get_sb
,
3701 .kill_sb
= kill_block_super
,
3702 .fs_flags
= FS_REQUIRES_DEV
,
3705 #ifdef CONFIG_EXT4DEV_COMPAT
3706 static int ext4dev_get_sb(struct file_system_type
*fs_type
,
3707 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3709 printk(KERN_WARNING
"EXT4-fs: Update your userspace programs "
3710 "to mount using ext4\n");
3711 printk(KERN_WARNING
"EXT4-fs: ext4dev backwards compatibility "
3712 "will go away by 2.6.31\n");
3713 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3716 static struct file_system_type ext4dev_fs_type
= {
3717 .owner
= THIS_MODULE
,
3719 .get_sb
= ext4dev_get_sb
,
3720 .kill_sb
= kill_block_super
,
3721 .fs_flags
= FS_REQUIRES_DEV
,
3723 MODULE_ALIAS("ext4dev");
3726 static int __init
init_ext4_fs(void)
3730 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
3731 err
= init_ext4_mballoc();
3735 err
= init_ext4_xattr();
3738 err
= init_inodecache();
3741 err
= register_filesystem(&ext4_fs_type
);
3744 #ifdef CONFIG_EXT4DEV_COMPAT
3745 err
= register_filesystem(&ext4dev_fs_type
);
3747 unregister_filesystem(&ext4_fs_type
);
3753 destroy_inodecache();
3757 exit_ext4_mballoc();
3761 static void __exit
exit_ext4_fs(void)
3763 unregister_filesystem(&ext4_fs_type
);
3764 #ifdef CONFIG_EXT4DEV_COMPAT
3765 unregister_filesystem(&ext4dev_fs_type
);
3767 destroy_inodecache();
3769 exit_ext4_mballoc();
3770 remove_proc_entry("fs/ext4", NULL
);
3773 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3774 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3775 MODULE_LICENSE("GPL");
3776 module_init(init_ext4_fs
)
3777 module_exit(exit_ext4_fs
)