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/log2.h>
38 #include <linux/crc16.h>
39 #include <asm/uaccess.h>
42 #include "ext4_jbd2.h"
48 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
49 unsigned long journal_devnum
);
50 static int ext4_create_journal(struct super_block
*, struct ext4_super_block
*,
52 static void ext4_commit_super (struct super_block
* sb
,
53 struct ext4_super_block
* es
,
55 static void ext4_mark_recovery_complete(struct super_block
* sb
,
56 struct ext4_super_block
* es
);
57 static void ext4_clear_journal_err(struct super_block
* sb
,
58 struct ext4_super_block
* es
);
59 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
60 static const char *ext4_decode_error(struct super_block
* sb
, int errno
,
62 static int ext4_remount (struct super_block
* sb
, int * flags
, char * data
);
63 static int ext4_statfs (struct dentry
* dentry
, struct kstatfs
* buf
);
64 static void ext4_unlockfs(struct super_block
*sb
);
65 static void ext4_write_super (struct super_block
* sb
);
66 static void ext4_write_super_lockfs(struct super_block
*sb
);
69 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
70 struct ext4_group_desc
*bg
)
72 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
73 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
74 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
77 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
78 struct ext4_group_desc
*bg
)
80 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
81 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
82 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
85 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
86 struct ext4_group_desc
*bg
)
88 return le32_to_cpu(bg
->bg_inode_table_lo
) |
89 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
90 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
93 void ext4_block_bitmap_set(struct super_block
*sb
,
94 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
96 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
97 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
98 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
101 void ext4_inode_bitmap_set(struct super_block
*sb
,
102 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
104 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
105 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
106 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
109 void ext4_inode_table_set(struct super_block
*sb
,
110 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
112 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
113 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
114 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
118 * Wrappers for jbd2_journal_start/end.
120 * The only special thing we need to do here is to make sure that all
121 * journal_end calls result in the superblock being marked dirty, so
122 * that sync() will call the filesystem's write_super callback if
125 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
129 if (sb
->s_flags
& MS_RDONLY
)
130 return ERR_PTR(-EROFS
);
132 /* Special case here: if the journal has aborted behind our
133 * backs (eg. EIO in the commit thread), then we still need to
134 * take the FS itself readonly cleanly. */
135 journal
= EXT4_SB(sb
)->s_journal
;
136 if (is_journal_aborted(journal
)) {
137 ext4_abort(sb
, __func__
,
138 "Detected aborted journal");
139 return ERR_PTR(-EROFS
);
142 return jbd2_journal_start(journal
, nblocks
);
146 * The only special thing we need to do here is to make sure that all
147 * jbd2_journal_stop calls result in the superblock being marked dirty, so
148 * that sync() will call the filesystem's write_super callback if
151 int __ext4_journal_stop(const char *where
, handle_t
*handle
)
153 struct super_block
*sb
;
157 sb
= handle
->h_transaction
->t_journal
->j_private
;
159 rc
= jbd2_journal_stop(handle
);
164 __ext4_std_error(sb
, where
, err
);
168 void ext4_journal_abort_handle(const char *caller
, const char *err_fn
,
169 struct buffer_head
*bh
, handle_t
*handle
, int err
)
172 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
175 BUFFER_TRACE(bh
, "abort");
180 if (is_handle_aborted(handle
))
183 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
184 caller
, errstr
, err_fn
);
186 jbd2_journal_abort_handle(handle
);
189 /* Deal with the reporting of failure conditions on a filesystem such as
190 * inconsistencies detected or read IO failures.
192 * On ext2, we can store the error state of the filesystem in the
193 * superblock. That is not possible on ext4, because we may have other
194 * write ordering constraints on the superblock which prevent us from
195 * writing it out straight away; and given that the journal is about to
196 * be aborted, we can't rely on the current, or future, transactions to
197 * write out the superblock safely.
199 * We'll just use the jbd2_journal_abort() error code to record an error in
200 * the journal instead. On recovery, the journal will compain about
201 * that error until we've noted it down and cleared it.
204 static void ext4_handle_error(struct super_block
*sb
)
206 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
208 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
209 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
211 if (sb
->s_flags
& MS_RDONLY
)
214 if (!test_opt (sb
, ERRORS_CONT
)) {
215 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
217 EXT4_SB(sb
)->s_mount_opt
|= EXT4_MOUNT_ABORT
;
219 jbd2_journal_abort(journal
, -EIO
);
221 if (test_opt (sb
, ERRORS_RO
)) {
222 printk (KERN_CRIT
"Remounting filesystem read-only\n");
223 sb
->s_flags
|= MS_RDONLY
;
225 ext4_commit_super(sb
, es
, 1);
226 if (test_opt(sb
, ERRORS_PANIC
))
227 panic("EXT4-fs (device %s): panic forced after error\n",
231 void ext4_error (struct super_block
* sb
, const char * function
,
232 const char * fmt
, ...)
237 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ",sb
->s_id
, function
);
242 ext4_handle_error(sb
);
245 static const char *ext4_decode_error(struct super_block
* sb
, int errno
,
252 errstr
= "IO failure";
255 errstr
= "Out of memory";
258 if (!sb
|| EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
)
259 errstr
= "Journal has aborted";
261 errstr
= "Readonly filesystem";
264 /* If the caller passed in an extra buffer for unknown
265 * errors, textualise them now. Else we just return
268 /* Check for truncated error codes... */
269 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
278 /* __ext4_std_error decodes expected errors from journaling functions
279 * automatically and invokes the appropriate error response. */
281 void __ext4_std_error (struct super_block
* sb
, const char * function
,
287 /* Special case: if the error is EROFS, and we're not already
288 * inside a transaction, then there's really no point in logging
290 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
291 (sb
->s_flags
& MS_RDONLY
))
294 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
295 printk (KERN_CRIT
"EXT4-fs error (device %s) in %s: %s\n",
296 sb
->s_id
, function
, errstr
);
298 ext4_handle_error(sb
);
302 * ext4_abort is a much stronger failure handler than ext4_error. The
303 * abort function may be used to deal with unrecoverable failures such
304 * as journal IO errors or ENOMEM at a critical moment in log management.
306 * We unconditionally force the filesystem into an ABORT|READONLY state,
307 * unless the error response on the fs has been set to panic in which
308 * case we take the easy way out and panic immediately.
311 void ext4_abort (struct super_block
* sb
, const char * function
,
312 const char * fmt
, ...)
316 printk (KERN_CRIT
"ext4_abort called.\n");
319 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ",sb
->s_id
, function
);
324 if (test_opt(sb
, ERRORS_PANIC
))
325 panic("EXT4-fs panic from previous error\n");
327 if (sb
->s_flags
& MS_RDONLY
)
330 printk(KERN_CRIT
"Remounting filesystem read-only\n");
331 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
332 sb
->s_flags
|= MS_RDONLY
;
333 EXT4_SB(sb
)->s_mount_opt
|= EXT4_MOUNT_ABORT
;
334 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
337 void ext4_warning (struct super_block
* sb
, const char * function
,
338 const char * fmt
, ...)
343 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s: ",
350 void ext4_update_dynamic_rev(struct super_block
*sb
)
352 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
354 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
357 ext4_warning(sb
, __func__
,
358 "updating to rev %d because of new feature flag, "
359 "running e2fsck is recommended",
362 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
363 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
364 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
365 /* leave es->s_feature_*compat flags alone */
366 /* es->s_uuid will be set by e2fsck if empty */
369 * The rest of the superblock fields should be zero, and if not it
370 * means they are likely already in use, so leave them alone. We
371 * can leave it up to e2fsck to clean up any inconsistencies there.
375 int ext4_update_compat_feature(handle_t
*handle
,
376 struct super_block
*sb
, __u32 compat
)
379 if (!EXT4_HAS_COMPAT_FEATURE(sb
, compat
)) {
380 err
= ext4_journal_get_write_access(handle
,
384 EXT4_SET_COMPAT_FEATURE(sb
, compat
);
387 BUFFER_TRACE(EXT4_SB(sb
)->s_sbh
,
388 "call ext4_journal_dirty_met adata");
389 err
= ext4_journal_dirty_metadata(handle
,
395 int ext4_update_rocompat_feature(handle_t
*handle
,
396 struct super_block
*sb
, __u32 rocompat
)
399 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, rocompat
)) {
400 err
= ext4_journal_get_write_access(handle
,
404 EXT4_SET_RO_COMPAT_FEATURE(sb
, rocompat
);
407 BUFFER_TRACE(EXT4_SB(sb
)->s_sbh
,
408 "call ext4_journal_dirty_met adata");
409 err
= ext4_journal_dirty_metadata(handle
,
415 int ext4_update_incompat_feature(handle_t
*handle
,
416 struct super_block
*sb
, __u32 incompat
)
419 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, incompat
)) {
420 err
= ext4_journal_get_write_access(handle
,
424 EXT4_SET_INCOMPAT_FEATURE(sb
, incompat
);
427 BUFFER_TRACE(EXT4_SB(sb
)->s_sbh
,
428 "call ext4_journal_dirty_met adata");
429 err
= ext4_journal_dirty_metadata(handle
,
436 * Open the external journal device
438 static struct block_device
*ext4_blkdev_get(dev_t dev
)
440 struct block_device
*bdev
;
441 char b
[BDEVNAME_SIZE
];
443 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
449 printk(KERN_ERR
"EXT4: failed to open journal device %s: %ld\n",
450 __bdevname(dev
, b
), PTR_ERR(bdev
));
455 * Release the journal device
457 static int ext4_blkdev_put(struct block_device
*bdev
)
460 return blkdev_put(bdev
);
463 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
465 struct block_device
*bdev
;
468 bdev
= sbi
->journal_bdev
;
470 ret
= ext4_blkdev_put(bdev
);
471 sbi
->journal_bdev
= NULL
;
476 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
478 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
481 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
485 printk(KERN_ERR
"sb orphan head is %d\n",
486 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
488 printk(KERN_ERR
"sb_info orphan list:\n");
489 list_for_each(l
, &sbi
->s_orphan
) {
490 struct inode
*inode
= orphan_list_entry(l
);
492 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
493 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
494 inode
->i_mode
, inode
->i_nlink
,
499 static void ext4_put_super (struct super_block
* sb
)
501 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
502 struct ext4_super_block
*es
= sbi
->s_es
;
506 ext4_ext_release(sb
);
507 ext4_xattr_put_super(sb
);
508 jbd2_journal_destroy(sbi
->s_journal
);
509 if (!(sb
->s_flags
& MS_RDONLY
)) {
510 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
511 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
512 BUFFER_TRACE(sbi
->s_sbh
, "marking dirty");
513 mark_buffer_dirty(sbi
->s_sbh
);
514 ext4_commit_super(sb
, es
, 1);
517 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
518 brelse(sbi
->s_group_desc
[i
]);
519 kfree(sbi
->s_group_desc
);
520 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
521 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
522 percpu_counter_destroy(&sbi
->s_dirs_counter
);
525 for (i
= 0; i
< MAXQUOTAS
; i
++)
526 kfree(sbi
->s_qf_names
[i
]);
529 /* Debugging code just in case the in-memory inode orphan list
530 * isn't empty. The on-disk one can be non-empty if we've
531 * detected an error and taken the fs readonly, but the
532 * in-memory list had better be clean by this point. */
533 if (!list_empty(&sbi
->s_orphan
))
534 dump_orphan_list(sb
, sbi
);
535 J_ASSERT(list_empty(&sbi
->s_orphan
));
537 invalidate_bdev(sb
->s_bdev
);
538 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
540 * Invalidate the journal device's buffers. We don't want them
541 * floating about in memory - the physical journal device may
542 * hotswapped, and it breaks the `ro-after' testing code.
544 sync_blockdev(sbi
->journal_bdev
);
545 invalidate_bdev(sbi
->journal_bdev
);
546 ext4_blkdev_remove(sbi
);
548 sb
->s_fs_info
= NULL
;
553 static struct kmem_cache
*ext4_inode_cachep
;
556 * Called inside transaction, so use GFP_NOFS
558 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
560 struct ext4_inode_info
*ei
;
562 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
565 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 ei
->i_acl
= EXT4_ACL_NOT_CACHED
;
567 ei
->i_default_acl
= EXT4_ACL_NOT_CACHED
;
569 ei
->i_block_alloc_info
= NULL
;
570 ei
->vfs_inode
.i_version
= 1;
571 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
572 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
573 spin_lock_init(&ei
->i_prealloc_lock
);
574 return &ei
->vfs_inode
;
577 static void ext4_destroy_inode(struct inode
*inode
)
579 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
580 printk("EXT4 Inode %p: orphan list check failed!\n",
582 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
583 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
587 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
590 static void init_once(struct kmem_cache
*cachep
, void *foo
)
592 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
594 INIT_LIST_HEAD(&ei
->i_orphan
);
595 #ifdef CONFIG_EXT4DEV_FS_XATTR
596 init_rwsem(&ei
->xattr_sem
);
598 init_rwsem(&ei
->i_data_sem
);
599 inode_init_once(&ei
->vfs_inode
);
602 static int init_inodecache(void)
604 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
605 sizeof(struct ext4_inode_info
),
606 0, (SLAB_RECLAIM_ACCOUNT
|
609 if (ext4_inode_cachep
== NULL
)
614 static void destroy_inodecache(void)
616 kmem_cache_destroy(ext4_inode_cachep
);
619 static void ext4_clear_inode(struct inode
*inode
)
621 struct ext4_block_alloc_info
*rsv
= EXT4_I(inode
)->i_block_alloc_info
;
622 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
623 if (EXT4_I(inode
)->i_acl
&&
624 EXT4_I(inode
)->i_acl
!= EXT4_ACL_NOT_CACHED
) {
625 posix_acl_release(EXT4_I(inode
)->i_acl
);
626 EXT4_I(inode
)->i_acl
= EXT4_ACL_NOT_CACHED
;
628 if (EXT4_I(inode
)->i_default_acl
&&
629 EXT4_I(inode
)->i_default_acl
!= EXT4_ACL_NOT_CACHED
) {
630 posix_acl_release(EXT4_I(inode
)->i_default_acl
);
631 EXT4_I(inode
)->i_default_acl
= EXT4_ACL_NOT_CACHED
;
634 ext4_discard_reservation(inode
);
635 EXT4_I(inode
)->i_block_alloc_info
= NULL
;
640 static inline void ext4_show_quota_options(struct seq_file
*seq
, struct super_block
*sb
)
642 #if defined(CONFIG_QUOTA)
643 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
645 if (sbi
->s_jquota_fmt
)
646 seq_printf(seq
, ",jqfmt=%s",
647 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold": "vfsv0");
649 if (sbi
->s_qf_names
[USRQUOTA
])
650 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
652 if (sbi
->s_qf_names
[GRPQUOTA
])
653 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
655 if (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
)
656 seq_puts(seq
, ",usrquota");
658 if (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)
659 seq_puts(seq
, ",grpquota");
665 * - it's set to a non-default value OR
666 * - if the per-sb default is different from the global default
668 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
671 unsigned long def_mount_opts
;
672 struct super_block
*sb
= vfs
->mnt_sb
;
673 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
674 struct ext4_super_block
*es
= sbi
->s_es
;
676 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
677 def_errors
= le16_to_cpu(es
->s_errors
);
679 if (sbi
->s_sb_block
!= 1)
680 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
681 if (test_opt(sb
, MINIX_DF
))
682 seq_puts(seq
, ",minixdf");
683 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
684 seq_puts(seq
, ",grpid");
685 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
686 seq_puts(seq
, ",nogrpid");
687 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
688 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
689 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
691 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
692 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
693 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
695 if (test_opt(sb
, ERRORS_RO
)) {
696 if (def_errors
== EXT4_ERRORS_PANIC
||
697 def_errors
== EXT4_ERRORS_CONTINUE
) {
698 seq_puts(seq
, ",errors=remount-ro");
701 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
702 seq_puts(seq
, ",errors=continue");
703 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
704 seq_puts(seq
, ",errors=panic");
705 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
706 seq_puts(seq
, ",nouid32");
707 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
708 seq_puts(seq
, ",debug");
709 if (test_opt(sb
, OLDALLOC
))
710 seq_puts(seq
, ",oldalloc");
711 #ifdef CONFIG_EXT4DEV_FS_XATTR
712 if (test_opt(sb
, XATTR_USER
) &&
713 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
714 seq_puts(seq
, ",user_xattr");
715 if (!test_opt(sb
, XATTR_USER
) &&
716 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
717 seq_puts(seq
, ",nouser_xattr");
720 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
721 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
722 seq_puts(seq
, ",acl");
723 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
724 seq_puts(seq
, ",noacl");
726 if (!test_opt(sb
, RESERVATION
))
727 seq_puts(seq
, ",noreservation");
728 if (sbi
->s_commit_interval
) {
729 seq_printf(seq
, ",commit=%u",
730 (unsigned) (sbi
->s_commit_interval
/ HZ
));
732 if (test_opt(sb
, BARRIER
))
733 seq_puts(seq
, ",barrier=1");
734 if (test_opt(sb
, NOBH
))
735 seq_puts(seq
, ",nobh");
736 if (!test_opt(sb
, EXTENTS
))
737 seq_puts(seq
, ",noextents");
738 if (!test_opt(sb
, MBALLOC
))
739 seq_puts(seq
, ",nomballoc");
740 if (test_opt(sb
, I_VERSION
))
741 seq_puts(seq
, ",i_version");
744 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
746 * journal mode get enabled in different ways
747 * So just print the value even if we didn't specify it
749 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
750 seq_puts(seq
, ",data=journal");
751 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
752 seq_puts(seq
, ",data=ordered");
753 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
754 seq_puts(seq
, ",data=writeback");
756 ext4_show_quota_options(seq
, sb
);
761 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
762 u64 ino
, u32 generation
)
766 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
767 return ERR_PTR(-ESTALE
);
768 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
769 return ERR_PTR(-ESTALE
);
771 /* iget isn't really right if the inode is currently unallocated!!
773 * ext4_read_inode will return a bad_inode if the inode had been
774 * deleted, so we should be safe.
776 * Currently we don't know the generation for parent directory, so
777 * a generation of 0 means "accept any"
779 inode
= ext4_iget(sb
, ino
);
781 return ERR_CAST(inode
);
782 if (generation
&& inode
->i_generation
!= generation
) {
784 return ERR_PTR(-ESTALE
);
790 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
791 int fh_len
, int fh_type
)
793 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
797 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
798 int fh_len
, int fh_type
)
800 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
805 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
806 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
808 static int ext4_dquot_initialize(struct inode
*inode
, int type
);
809 static int ext4_dquot_drop(struct inode
*inode
);
810 static int ext4_write_dquot(struct dquot
*dquot
);
811 static int ext4_acquire_dquot(struct dquot
*dquot
);
812 static int ext4_release_dquot(struct dquot
*dquot
);
813 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
814 static int ext4_write_info(struct super_block
*sb
, int type
);
815 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
816 char *path
, int remount
);
817 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
818 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
819 size_t len
, loff_t off
);
820 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
821 const char *data
, size_t len
, loff_t off
);
823 static struct dquot_operations ext4_quota_operations
= {
824 .initialize
= ext4_dquot_initialize
,
825 .drop
= ext4_dquot_drop
,
826 .alloc_space
= dquot_alloc_space
,
827 .alloc_inode
= dquot_alloc_inode
,
828 .free_space
= dquot_free_space
,
829 .free_inode
= dquot_free_inode
,
830 .transfer
= dquot_transfer
,
831 .write_dquot
= ext4_write_dquot
,
832 .acquire_dquot
= ext4_acquire_dquot
,
833 .release_dquot
= ext4_release_dquot
,
834 .mark_dirty
= ext4_mark_dquot_dirty
,
835 .write_info
= ext4_write_info
838 static struct quotactl_ops ext4_qctl_operations
= {
839 .quota_on
= ext4_quota_on
,
840 .quota_off
= vfs_quota_off
,
841 .quota_sync
= vfs_quota_sync
,
842 .get_info
= vfs_get_dqinfo
,
843 .set_info
= vfs_set_dqinfo
,
844 .get_dqblk
= vfs_get_dqblk
,
845 .set_dqblk
= vfs_set_dqblk
849 static const struct super_operations ext4_sops
= {
850 .alloc_inode
= ext4_alloc_inode
,
851 .destroy_inode
= ext4_destroy_inode
,
852 .write_inode
= ext4_write_inode
,
853 .dirty_inode
= ext4_dirty_inode
,
854 .delete_inode
= ext4_delete_inode
,
855 .put_super
= ext4_put_super
,
856 .write_super
= ext4_write_super
,
857 .sync_fs
= ext4_sync_fs
,
858 .write_super_lockfs
= ext4_write_super_lockfs
,
859 .unlockfs
= ext4_unlockfs
,
860 .statfs
= ext4_statfs
,
861 .remount_fs
= ext4_remount
,
862 .clear_inode
= ext4_clear_inode
,
863 .show_options
= ext4_show_options
,
865 .quota_read
= ext4_quota_read
,
866 .quota_write
= ext4_quota_write
,
870 static const struct export_operations ext4_export_ops
= {
871 .fh_to_dentry
= ext4_fh_to_dentry
,
872 .fh_to_parent
= ext4_fh_to_parent
,
873 .get_parent
= ext4_get_parent
,
877 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
878 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
879 Opt_nouid32
, Opt_nocheck
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
880 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
881 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
882 Opt_commit
, Opt_journal_update
, Opt_journal_inum
, Opt_journal_dev
,
883 Opt_journal_checksum
, Opt_journal_async_commit
,
884 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
885 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
886 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
887 Opt_ignore
, Opt_barrier
, Opt_err
, Opt_resize
, Opt_usrquota
,
888 Opt_grpquota
, Opt_extents
, Opt_noextents
, Opt_i_version
,
889 Opt_mballoc
, Opt_nomballoc
, Opt_stripe
,
892 static match_table_t tokens
= {
893 {Opt_bsd_df
, "bsddf"},
894 {Opt_minix_df
, "minixdf"},
895 {Opt_grpid
, "grpid"},
896 {Opt_grpid
, "bsdgroups"},
897 {Opt_nogrpid
, "nogrpid"},
898 {Opt_nogrpid
, "sysvgroups"},
899 {Opt_resgid
, "resgid=%u"},
900 {Opt_resuid
, "resuid=%u"},
902 {Opt_err_cont
, "errors=continue"},
903 {Opt_err_panic
, "errors=panic"},
904 {Opt_err_ro
, "errors=remount-ro"},
905 {Opt_nouid32
, "nouid32"},
906 {Opt_nocheck
, "nocheck"},
907 {Opt_nocheck
, "check=none"},
908 {Opt_debug
, "debug"},
909 {Opt_oldalloc
, "oldalloc"},
910 {Opt_orlov
, "orlov"},
911 {Opt_user_xattr
, "user_xattr"},
912 {Opt_nouser_xattr
, "nouser_xattr"},
914 {Opt_noacl
, "noacl"},
915 {Opt_reservation
, "reservation"},
916 {Opt_noreservation
, "noreservation"},
917 {Opt_noload
, "noload"},
920 {Opt_commit
, "commit=%u"},
921 {Opt_journal_update
, "journal=update"},
922 {Opt_journal_inum
, "journal=%u"},
923 {Opt_journal_dev
, "journal_dev=%u"},
924 {Opt_journal_checksum
, "journal_checksum"},
925 {Opt_journal_async_commit
, "journal_async_commit"},
926 {Opt_abort
, "abort"},
927 {Opt_data_journal
, "data=journal"},
928 {Opt_data_ordered
, "data=ordered"},
929 {Opt_data_writeback
, "data=writeback"},
930 {Opt_offusrjquota
, "usrjquota="},
931 {Opt_usrjquota
, "usrjquota=%s"},
932 {Opt_offgrpjquota
, "grpjquota="},
933 {Opt_grpjquota
, "grpjquota=%s"},
934 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
935 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
936 {Opt_grpquota
, "grpquota"},
937 {Opt_noquota
, "noquota"},
938 {Opt_quota
, "quota"},
939 {Opt_usrquota
, "usrquota"},
940 {Opt_barrier
, "barrier=%u"},
941 {Opt_extents
, "extents"},
942 {Opt_noextents
, "noextents"},
943 {Opt_i_version
, "i_version"},
944 {Opt_mballoc
, "mballoc"},
945 {Opt_nomballoc
, "nomballoc"},
946 {Opt_stripe
, "stripe=%u"},
947 {Opt_resize
, "resize"},
951 static ext4_fsblk_t
get_sb_block(void **data
)
953 ext4_fsblk_t sb_block
;
954 char *options
= (char *) *data
;
956 if (!options
|| strncmp(options
, "sb=", 3) != 0)
957 return 1; /* Default location */
959 /*todo: use simple_strtoll with >32bit ext4 */
960 sb_block
= simple_strtoul(options
, &options
, 0);
961 if (*options
&& *options
!= ',') {
962 printk("EXT4-fs: Invalid sb specification: %s\n",
968 *data
= (void *) options
;
972 static int parse_options (char *options
, struct super_block
*sb
,
973 unsigned int *inum
, unsigned long *journal_devnum
,
974 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
976 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
978 substring_t args
[MAX_OPT_ARGS
];
989 while ((p
= strsep (&options
, ",")) != NULL
) {
994 token
= match_token(p
, tokens
, args
);
997 clear_opt (sbi
->s_mount_opt
, MINIX_DF
);
1000 set_opt (sbi
->s_mount_opt
, MINIX_DF
);
1003 set_opt (sbi
->s_mount_opt
, GRPID
);
1006 clear_opt (sbi
->s_mount_opt
, GRPID
);
1009 if (match_int(&args
[0], &option
))
1011 sbi
->s_resuid
= option
;
1014 if (match_int(&args
[0], &option
))
1016 sbi
->s_resgid
= option
;
1019 /* handled by get_sb_block() instead of here */
1020 /* *sb_block = match_int(&args[0]); */
1023 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1024 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1025 set_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1028 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1029 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1030 set_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1033 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1034 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1035 set_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1038 set_opt (sbi
->s_mount_opt
, NO_UID32
);
1041 clear_opt (sbi
->s_mount_opt
, CHECK
);
1044 set_opt (sbi
->s_mount_opt
, DEBUG
);
1047 set_opt (sbi
->s_mount_opt
, OLDALLOC
);
1050 clear_opt (sbi
->s_mount_opt
, OLDALLOC
);
1052 #ifdef CONFIG_EXT4DEV_FS_XATTR
1053 case Opt_user_xattr
:
1054 set_opt (sbi
->s_mount_opt
, XATTR_USER
);
1056 case Opt_nouser_xattr
:
1057 clear_opt (sbi
->s_mount_opt
, XATTR_USER
);
1060 case Opt_user_xattr
:
1061 case Opt_nouser_xattr
:
1062 printk("EXT4 (no)user_xattr options not supported\n");
1065 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1067 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1070 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1075 printk("EXT4 (no)acl options not supported\n");
1078 case Opt_reservation
:
1079 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1081 case Opt_noreservation
:
1082 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
1084 case Opt_journal_update
:
1086 /* Eventually we will want to be able to create
1087 a journal file here. For now, only allow the
1088 user to specify an existing inode to be the
1091 printk(KERN_ERR
"EXT4-fs: cannot specify "
1092 "journal on remount\n");
1095 set_opt (sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1097 case Opt_journal_inum
:
1099 printk(KERN_ERR
"EXT4-fs: cannot specify "
1100 "journal on remount\n");
1103 if (match_int(&args
[0], &option
))
1107 case Opt_journal_dev
:
1109 printk(KERN_ERR
"EXT4-fs: cannot specify "
1110 "journal on remount\n");
1113 if (match_int(&args
[0], &option
))
1115 *journal_devnum
= option
;
1117 case Opt_journal_checksum
:
1118 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1120 case Opt_journal_async_commit
:
1121 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1122 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1125 set_opt (sbi
->s_mount_opt
, NOLOAD
);
1128 if (match_int(&args
[0], &option
))
1133 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1134 sbi
->s_commit_interval
= HZ
* option
;
1136 case Opt_data_journal
:
1137 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1139 case Opt_data_ordered
:
1140 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1142 case Opt_data_writeback
:
1143 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1146 if ((sbi
->s_mount_opt
& EXT4_MOUNT_DATA_FLAGS
)
1149 "EXT4-fs: cannot change data "
1150 "mode on remount\n");
1154 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DATA_FLAGS
;
1155 sbi
->s_mount_opt
|= data_opt
;
1165 if ((sb_any_quota_enabled(sb
) ||
1166 sb_any_quota_suspended(sb
)) &&
1167 !sbi
->s_qf_names
[qtype
]) {
1169 "EXT4-fs: Cannot change journaled "
1170 "quota options when quota turned on.\n");
1173 qname
= match_strdup(&args
[0]);
1176 "EXT4-fs: not enough memory for "
1177 "storing quotafile name.\n");
1180 if (sbi
->s_qf_names
[qtype
] &&
1181 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1183 "EXT4-fs: %s quota file already "
1184 "specified.\n", QTYPE2NAME(qtype
));
1188 sbi
->s_qf_names
[qtype
] = qname
;
1189 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1191 "EXT4-fs: quotafile must be on "
1192 "filesystem root.\n");
1193 kfree(sbi
->s_qf_names
[qtype
]);
1194 sbi
->s_qf_names
[qtype
] = NULL
;
1197 set_opt(sbi
->s_mount_opt
, QUOTA
);
1199 case Opt_offusrjquota
:
1202 case Opt_offgrpjquota
:
1205 if ((sb_any_quota_enabled(sb
) ||
1206 sb_any_quota_suspended(sb
)) &&
1207 sbi
->s_qf_names
[qtype
]) {
1208 printk(KERN_ERR
"EXT4-fs: Cannot change "
1209 "journaled quota options when "
1210 "quota turned on.\n");
1214 * The space will be released later when all options
1215 * are confirmed to be correct
1217 sbi
->s_qf_names
[qtype
] = NULL
;
1219 case Opt_jqfmt_vfsold
:
1220 qfmt
= QFMT_VFS_OLD
;
1222 case Opt_jqfmt_vfsv0
:
1225 if ((sb_any_quota_enabled(sb
) ||
1226 sb_any_quota_suspended(sb
)) &&
1227 sbi
->s_jquota_fmt
!= qfmt
) {
1228 printk(KERN_ERR
"EXT4-fs: Cannot change "
1229 "journaled quota options when "
1230 "quota turned on.\n");
1233 sbi
->s_jquota_fmt
= qfmt
;
1237 set_opt(sbi
->s_mount_opt
, QUOTA
);
1238 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1241 set_opt(sbi
->s_mount_opt
, QUOTA
);
1242 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1245 if (sb_any_quota_enabled(sb
)) {
1246 printk(KERN_ERR
"EXT4-fs: Cannot change quota "
1247 "options when quota turned on.\n");
1250 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1251 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1252 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1259 "EXT4-fs: quota options not supported.\n");
1263 case Opt_offusrjquota
:
1264 case Opt_offgrpjquota
:
1265 case Opt_jqfmt_vfsold
:
1266 case Opt_jqfmt_vfsv0
:
1268 "EXT4-fs: journaled quota options not "
1275 set_opt(sbi
->s_mount_opt
, ABORT
);
1278 if (match_int(&args
[0], &option
))
1281 set_opt(sbi
->s_mount_opt
, BARRIER
);
1283 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1289 printk("EXT4-fs: resize option only available "
1293 if (match_int(&args
[0], &option
) != 0)
1295 *n_blocks_count
= option
;
1298 set_opt(sbi
->s_mount_opt
, NOBH
);
1301 clear_opt(sbi
->s_mount_opt
, NOBH
);
1304 set_opt (sbi
->s_mount_opt
, EXTENTS
);
1307 clear_opt (sbi
->s_mount_opt
, EXTENTS
);
1310 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1311 sb
->s_flags
|= MS_I_VERSION
;
1314 set_opt(sbi
->s_mount_opt
, MBALLOC
);
1317 clear_opt(sbi
->s_mount_opt
, MBALLOC
);
1320 if (match_int(&args
[0], &option
))
1324 sbi
->s_stripe
= option
;
1328 "EXT4-fs: Unrecognized mount option \"%s\" "
1329 "or missing value\n", p
);
1334 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1335 if ((sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
) &&
1336 sbi
->s_qf_names
[USRQUOTA
])
1337 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1339 if ((sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
) &&
1340 sbi
->s_qf_names
[GRPQUOTA
])
1341 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1343 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1344 (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)) ||
1345 (sbi
->s_qf_names
[GRPQUOTA
] &&
1346 (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
))) {
1347 printk(KERN_ERR
"EXT4-fs: old and new quota "
1348 "format mixing.\n");
1352 if (!sbi
->s_jquota_fmt
) {
1353 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1354 "not specified.\n");
1358 if (sbi
->s_jquota_fmt
) {
1359 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1360 "specified with no journaling "
1369 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1372 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1375 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1376 printk (KERN_ERR
"EXT4-fs warning: revision level too high, "
1377 "forcing read-only mode\n");
1382 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1383 printk (KERN_WARNING
"EXT4-fs warning: mounting unchecked fs, "
1384 "running e2fsck is recommended\n");
1385 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1386 printk (KERN_WARNING
1387 "EXT4-fs warning: mounting fs with errors, "
1388 "running e2fsck is recommended\n");
1389 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1390 le16_to_cpu(es
->s_mnt_count
) >=
1391 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1392 printk (KERN_WARNING
1393 "EXT4-fs warning: maximal mount count reached, "
1394 "running e2fsck is recommended\n");
1395 else if (le32_to_cpu(es
->s_checkinterval
) &&
1396 (le32_to_cpu(es
->s_lastcheck
) +
1397 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1398 printk (KERN_WARNING
1399 "EXT4-fs warning: checktime reached, "
1400 "running e2fsck is recommended\n");
1402 /* @@@ We _will_ want to clear the valid bit if we find
1403 * inconsistencies, to force a fsck at reboot. But for
1404 * a plain journaled filesystem we can keep it set as
1407 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1409 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1410 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1411 le16_add_cpu(&es
->s_mnt_count
, 1);
1412 es
->s_mtime
= cpu_to_le32(get_seconds());
1413 ext4_update_dynamic_rev(sb
);
1414 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1416 ext4_commit_super(sb
, es
, 1);
1417 if (test_opt(sb
, DEBUG
))
1418 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%lu, "
1419 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1421 sbi
->s_groups_count
,
1422 EXT4_BLOCKS_PER_GROUP(sb
),
1423 EXT4_INODES_PER_GROUP(sb
),
1426 printk(KERN_INFO
"EXT4 FS on %s, ", sb
->s_id
);
1427 if (EXT4_SB(sb
)->s_journal
->j_inode
== NULL
) {
1428 char b
[BDEVNAME_SIZE
];
1430 printk("external journal on %s\n",
1431 bdevname(EXT4_SB(sb
)->s_journal
->j_dev
, b
));
1433 printk("internal journal\n");
1438 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1439 struct ext4_group_desc
*gdp
)
1443 if (sbi
->s_es
->s_feature_ro_compat
&
1444 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1445 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1446 __le32 le_group
= cpu_to_le32(block_group
);
1448 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1449 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1450 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1451 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1452 /* for checksum of struct ext4_group_desc do the rest...*/
1453 if ((sbi
->s_es
->s_feature_incompat
&
1454 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1455 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1456 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1457 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1461 return cpu_to_le16(crc
);
1464 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1465 struct ext4_group_desc
*gdp
)
1467 if ((sbi
->s_es
->s_feature_ro_compat
&
1468 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1469 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1475 /* Called at mount-time, super-block is locked */
1476 static int ext4_check_descriptors(struct super_block
*sb
)
1478 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1479 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1480 ext4_fsblk_t last_block
;
1481 ext4_fsblk_t block_bitmap
;
1482 ext4_fsblk_t inode_bitmap
;
1483 ext4_fsblk_t inode_table
;
1484 int flexbg_flag
= 0;
1487 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1490 ext4_debug ("Checking group descriptors");
1492 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1493 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1495 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1496 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1498 last_block
= first_block
+
1499 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1501 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1502 if (block_bitmap
< first_block
|| block_bitmap
> last_block
)
1504 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1505 "Block bitmap for group %lu not in group "
1506 "(block %llu)!", i
, block_bitmap
);
1509 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1510 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
)
1512 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1513 "Inode bitmap for group %lu not in group "
1514 "(block %llu)!", i
, inode_bitmap
);
1517 inode_table
= ext4_inode_table(sb
, gdp
);
1518 if (inode_table
< first_block
||
1519 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
)
1521 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1522 "Inode table for group %lu not in group "
1523 "(block %llu)!", i
, inode_table
);
1526 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1527 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1528 "Checksum for group %lu failed (%u!=%u)\n",
1529 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1530 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1534 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1537 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1538 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
1542 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1543 * the superblock) which were deleted from all directories, but held open by
1544 * a process at the time of a crash. We walk the list and try to delete these
1545 * inodes at recovery time (only with a read-write filesystem).
1547 * In order to keep the orphan inode chain consistent during traversal (in
1548 * case of crash during recovery), we link each inode into the superblock
1549 * orphan list_head and handle it the same way as an inode deletion during
1550 * normal operation (which journals the operations for us).
1552 * We only do an iget() and an iput() on each inode, which is very safe if we
1553 * accidentally point at an in-use or already deleted inode. The worst that
1554 * can happen in this case is that we get a "bit already cleared" message from
1555 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1556 * e2fsck was run on this filesystem, and it must have already done the orphan
1557 * inode cleanup for us, so we can safely abort without any further action.
1559 static void ext4_orphan_cleanup (struct super_block
* sb
,
1560 struct ext4_super_block
* es
)
1562 unsigned int s_flags
= sb
->s_flags
;
1563 int nr_orphans
= 0, nr_truncates
= 0;
1567 if (!es
->s_last_orphan
) {
1568 jbd_debug(4, "no orphan inodes to clean up\n");
1572 if (bdev_read_only(sb
->s_bdev
)) {
1573 printk(KERN_ERR
"EXT4-fs: write access "
1574 "unavailable, skipping orphan cleanup.\n");
1578 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1579 if (es
->s_last_orphan
)
1580 jbd_debug(1, "Errors on filesystem, "
1581 "clearing orphan list.\n");
1582 es
->s_last_orphan
= 0;
1583 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1587 if (s_flags
& MS_RDONLY
) {
1588 printk(KERN_INFO
"EXT4-fs: %s: orphan cleanup on readonly fs\n",
1590 sb
->s_flags
&= ~MS_RDONLY
;
1593 /* Needed for iput() to work correctly and not trash data */
1594 sb
->s_flags
|= MS_ACTIVE
;
1595 /* Turn on quotas so that they are updated correctly */
1596 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1597 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
1598 int ret
= ext4_quota_on_mount(sb
, i
);
1601 "EXT4-fs: Cannot turn on journaled "
1602 "quota: error %d\n", ret
);
1607 while (es
->s_last_orphan
) {
1608 struct inode
*inode
;
1610 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1611 if (IS_ERR(inode
)) {
1612 es
->s_last_orphan
= 0;
1616 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
1618 if (inode
->i_nlink
) {
1620 "%s: truncating inode %lu to %Ld bytes\n",
1621 __func__
, inode
->i_ino
, inode
->i_size
);
1622 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1623 inode
->i_ino
, inode
->i_size
);
1624 ext4_truncate(inode
);
1628 "%s: deleting unreferenced inode %lu\n",
1629 __func__
, inode
->i_ino
);
1630 jbd_debug(2, "deleting unreferenced inode %lu\n",
1634 iput(inode
); /* The delete magic happens here! */
1637 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1640 printk(KERN_INFO
"EXT4-fs: %s: %d orphan inode%s deleted\n",
1641 sb
->s_id
, PLURAL(nr_orphans
));
1643 printk(KERN_INFO
"EXT4-fs: %s: %d truncate%s cleaned up\n",
1644 sb
->s_id
, PLURAL(nr_truncates
));
1646 /* Turn quotas off */
1647 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1648 if (sb_dqopt(sb
)->files
[i
])
1649 vfs_quota_off(sb
, i
, 0);
1652 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1655 * Maximal extent format file size.
1656 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1657 * extent format containers, within a sector_t, and within i_blocks
1658 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1659 * so that won't be a limiting factor.
1661 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1663 static loff_t
ext4_max_size(int blkbits
)
1666 loff_t upper_limit
= MAX_LFS_FILESIZE
;
1668 /* small i_blocks in vfs inode? */
1669 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
1671 * CONFIG_LSF is not enabled implies the inode
1672 * i_block represent total blocks in 512 bytes
1673 * 32 == size of vfs inode i_blocks * 8
1675 upper_limit
= (1LL << 32) - 1;
1677 /* total blocks in file system block size */
1678 upper_limit
>>= (blkbits
- 9);
1679 upper_limit
<<= blkbits
;
1682 /* 32-bit extent-start container, ee_block */
1687 /* Sanity check against vm- & vfs- imposed limits */
1688 if (res
> upper_limit
)
1695 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1696 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1697 * We need to be 1 filesystem block less than the 2^48 sector limit.
1699 static loff_t
ext4_max_bitmap_size(int bits
)
1701 loff_t res
= EXT4_NDIR_BLOCKS
;
1704 /* This is calculated to be the largest file size for a
1705 * dense, bitmapped file such that the total number of
1706 * sectors in the file, including data and all indirect blocks,
1707 * does not exceed 2^48 -1
1708 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1709 * total number of 512 bytes blocks of the file
1712 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
1714 * CONFIG_LSF is not enabled implies the inode
1715 * i_block represent total blocks in 512 bytes
1716 * 32 == size of vfs inode i_blocks * 8
1718 upper_limit
= (1LL << 32) - 1;
1720 /* total blocks in file system block size */
1721 upper_limit
>>= (bits
- 9);
1725 * We use 48 bit ext4_inode i_blocks
1726 * With EXT4_HUGE_FILE_FL set the i_blocks
1727 * represent total number of blocks in
1728 * file system block size
1730 upper_limit
= (1LL << 48) - 1;
1734 /* indirect blocks */
1736 /* double indirect blocks */
1737 meta_blocks
+= 1 + (1LL << (bits
-2));
1738 /* tripple indirect blocks */
1739 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
1741 upper_limit
-= meta_blocks
;
1742 upper_limit
<<= bits
;
1744 res
+= 1LL << (bits
-2);
1745 res
+= 1LL << (2*(bits
-2));
1746 res
+= 1LL << (3*(bits
-2));
1748 if (res
> upper_limit
)
1751 if (res
> MAX_LFS_FILESIZE
)
1752 res
= MAX_LFS_FILESIZE
;
1757 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
1758 ext4_fsblk_t logical_sb_block
, int nr
)
1760 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1761 ext4_group_t bg
, first_meta_bg
;
1764 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1766 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
1768 return logical_sb_block
+ nr
+ 1;
1769 bg
= sbi
->s_desc_per_block
* nr
;
1770 if (ext4_bg_has_super(sb
, bg
))
1772 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
1776 * ext4_get_stripe_size: Get the stripe size.
1777 * @sbi: In memory super block info
1779 * If we have specified it via mount option, then
1780 * use the mount option value. If the value specified at mount time is
1781 * greater than the blocks per group use the super block value.
1782 * If the super block value is greater than blocks per group return 0.
1783 * Allocator needs it be less than blocks per group.
1786 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
1788 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
1789 unsigned long stripe_width
=
1790 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
1792 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
1793 return sbi
->s_stripe
;
1795 if (stripe_width
<= sbi
->s_blocks_per_group
)
1796 return stripe_width
;
1798 if (stride
<= sbi
->s_blocks_per_group
)
1804 static int ext4_fill_super (struct super_block
*sb
, void *data
, int silent
)
1805 __releases(kernel_sem
)
1806 __acquires(kernel_sem
)
1809 struct buffer_head
* bh
;
1810 struct ext4_super_block
*es
= NULL
;
1811 struct ext4_sb_info
*sbi
;
1813 ext4_fsblk_t sb_block
= get_sb_block(&data
);
1814 ext4_fsblk_t logical_sb_block
;
1815 unsigned long offset
= 0;
1816 unsigned int journal_inum
= 0;
1817 unsigned long journal_devnum
= 0;
1818 unsigned long def_mount_opts
;
1829 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
1832 sb
->s_fs_info
= sbi
;
1833 sbi
->s_mount_opt
= 0;
1834 sbi
->s_resuid
= EXT4_DEF_RESUID
;
1835 sbi
->s_resgid
= EXT4_DEF_RESGID
;
1836 sbi
->s_sb_block
= sb_block
;
1840 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
1842 printk(KERN_ERR
"EXT4-fs: unable to set blocksize\n");
1846 if (!sb_set_blocksize(sb
, blocksize
)) {
1847 printk(KERN_ERR
"EXT4-fs: bad blocksize %d.\n", blocksize
);
1852 * The ext4 superblock will not be buffer aligned for other than 1kB
1853 * block sizes. We need to calculate the offset from buffer start.
1855 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
1856 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
1857 offset
= do_div(logical_sb_block
, blocksize
);
1859 logical_sb_block
= sb_block
;
1862 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
1863 printk (KERN_ERR
"EXT4-fs: unable to read superblock\n");
1867 * Note: s_es must be initialized as soon as possible because
1868 * some ext4 macro-instructions depend on its value
1870 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
1872 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1873 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
1876 /* Set defaults before we parse the mount options */
1877 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1878 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
1879 set_opt(sbi
->s_mount_opt
, DEBUG
);
1880 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
1881 set_opt(sbi
->s_mount_opt
, GRPID
);
1882 if (def_mount_opts
& EXT4_DEFM_UID16
)
1883 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1884 #ifdef CONFIG_EXT4DEV_FS_XATTR
1885 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
1886 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1888 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1889 if (def_mount_opts
& EXT4_DEFM_ACL
)
1890 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1892 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
1893 sbi
->s_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
1894 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
1895 sbi
->s_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
1896 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
1897 sbi
->s_mount_opt
|= EXT4_MOUNT_WRITEBACK_DATA
;
1899 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
1900 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1901 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
1902 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1904 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1906 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
1907 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
1909 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1912 * turn on extents feature by default in ext4 filesystem
1913 * User -o noextents to turn it off
1915 set_opt(sbi
->s_mount_opt
, EXTENTS
);
1917 * turn on mballoc feature by default in ext4 filesystem
1918 * User -o nomballoc to turn it off
1920 set_opt(sbi
->s_mount_opt
, MBALLOC
);
1922 if (!parse_options ((char *) data
, sb
, &journal_inum
, &journal_devnum
,
1926 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1927 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1929 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
1930 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
1931 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
1932 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
1934 "EXT4-fs warning: feature flags set on rev 0 fs, "
1935 "running e2fsck is recommended\n");
1938 * Since ext4 is still considered development code, we require
1939 * that the TEST_FILESYS flag in s->flags be set.
1941 if (!(le32_to_cpu(es
->s_flags
) & EXT2_FLAGS_TEST_FILESYS
)) {
1942 printk(KERN_WARNING
"EXT4-fs: %s: not marked "
1943 "OK to use with test code.\n", sb
->s_id
);
1948 * Check feature flags regardless of the revision level, since we
1949 * previously didn't change the revision level when setting the flags,
1950 * so there is a chance incompat flags are set on a rev 0 filesystem.
1952 features
= EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
);
1954 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount because of "
1955 "unsupported optional features (%x).\n",
1956 sb
->s_id
, le32_to_cpu(features
));
1959 features
= EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
);
1960 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1961 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount RDWR because of "
1962 "unsupported optional features (%x).\n",
1963 sb
->s_id
, le32_to_cpu(features
));
1966 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
1968 * Large file size enabled file system can only be
1969 * mount if kernel is build with CONFIG_LSF
1971 if (sizeof(root
->i_blocks
) < sizeof(u64
) &&
1972 !(sb
->s_flags
& MS_RDONLY
)) {
1973 printk(KERN_ERR
"EXT4-fs: %s: Filesystem with huge "
1974 "files cannot be mounted read-write "
1975 "without CONFIG_LSF.\n", sb
->s_id
);
1979 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
1981 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
1982 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
1984 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1985 blocksize
, sb
->s_id
);
1989 if (sb
->s_blocksize
!= blocksize
) {
1991 /* Validate the filesystem blocksize */
1992 if (!sb_set_blocksize(sb
, blocksize
)) {
1993 printk(KERN_ERR
"EXT4-fs: bad block size %d.\n",
1999 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2000 offset
= do_div(logical_sb_block
, blocksize
);
2001 bh
= sb_bread(sb
, logical_sb_block
);
2004 "EXT4-fs: Can't read superblock on 2nd try.\n");
2007 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2009 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2011 "EXT4-fs: Magic mismatch, very weird !\n");
2016 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
);
2017 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
);
2019 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2020 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2021 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2023 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2024 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2025 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2026 (!is_power_of_2(sbi
->s_inode_size
)) ||
2027 (sbi
->s_inode_size
> blocksize
)) {
2029 "EXT4-fs: unsupported inode size: %d\n",
2033 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2034 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2036 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2037 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2038 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2039 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2040 !is_power_of_2(sbi
->s_desc_size
)) {
2042 "EXT4-fs: unsupported descriptor size %lu\n",
2047 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2048 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2049 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2050 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2052 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2053 if (sbi
->s_inodes_per_block
== 0)
2055 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2056 sbi
->s_inodes_per_block
;
2057 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2059 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2060 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2061 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2062 for (i
=0; i
< 4; i
++)
2063 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2064 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2066 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2068 "EXT4-fs: #blocks per group too big: %lu\n",
2069 sbi
->s_blocks_per_group
);
2072 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2074 "EXT4-fs: #inodes per group too big: %lu\n",
2075 sbi
->s_inodes_per_group
);
2079 if (ext4_blocks_count(es
) >
2080 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) {
2081 printk(KERN_ERR
"EXT4-fs: filesystem on %s:"
2082 " too large to mount safely\n", sb
->s_id
);
2083 if (sizeof(sector_t
) < 8)
2084 printk(KERN_WARNING
"EXT4-fs: CONFIG_LBD not "
2089 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2092 /* ensure blocks_count calculation below doesn't sign-extend */
2093 if (ext4_blocks_count(es
) + EXT4_BLOCKS_PER_GROUP(sb
) <
2094 le32_to_cpu(es
->s_first_data_block
) + 1) {
2095 printk(KERN_WARNING
"EXT4-fs: bad geometry: block count %llu, "
2096 "first data block %u, blocks per group %lu\n",
2097 ext4_blocks_count(es
),
2098 le32_to_cpu(es
->s_first_data_block
),
2099 EXT4_BLOCKS_PER_GROUP(sb
));
2102 blocks_count
= (ext4_blocks_count(es
) -
2103 le32_to_cpu(es
->s_first_data_block
) +
2104 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2105 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2106 sbi
->s_groups_count
= blocks_count
;
2107 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2108 EXT4_DESC_PER_BLOCK(sb
);
2109 sbi
->s_group_desc
= kmalloc(db_count
* sizeof (struct buffer_head
*),
2111 if (sbi
->s_group_desc
== NULL
) {
2112 printk (KERN_ERR
"EXT4-fs: not enough memory\n");
2116 bgl_lock_init(&sbi
->s_blockgroup_lock
);
2118 for (i
= 0; i
< db_count
; i
++) {
2119 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2120 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2121 if (!sbi
->s_group_desc
[i
]) {
2122 printk (KERN_ERR
"EXT4-fs: "
2123 "can't read group descriptor %d\n", i
);
2128 if (!ext4_check_descriptors (sb
)) {
2129 printk(KERN_ERR
"EXT4-fs: group descriptors corrupted!\n");
2132 sbi
->s_gdb_count
= db_count
;
2133 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2134 spin_lock_init(&sbi
->s_next_gen_lock
);
2136 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2137 ext4_count_free_blocks(sb
));
2139 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2140 ext4_count_free_inodes(sb
));
2143 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2144 ext4_count_dirs(sb
));
2147 printk(KERN_ERR
"EXT4-fs: insufficient memory\n");
2151 /* per fileystem reservation list head & lock */
2152 spin_lock_init(&sbi
->s_rsv_window_lock
);
2153 sbi
->s_rsv_window_root
= RB_ROOT
;
2154 /* Add a single, static dummy reservation to the start of the
2155 * reservation window list --- it gives us a placeholder for
2156 * append-at-start-of-list which makes the allocation logic
2157 * _much_ simpler. */
2158 sbi
->s_rsv_window_head
.rsv_start
= EXT4_RESERVE_WINDOW_NOT_ALLOCATED
;
2159 sbi
->s_rsv_window_head
.rsv_end
= EXT4_RESERVE_WINDOW_NOT_ALLOCATED
;
2160 sbi
->s_rsv_window_head
.rsv_alloc_hit
= 0;
2161 sbi
->s_rsv_window_head
.rsv_goal_size
= 0;
2162 ext4_rsv_window_add(sb
, &sbi
->s_rsv_window_head
);
2164 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2167 * set up enough so that it can read an inode
2169 sb
->s_op
= &ext4_sops
;
2170 sb
->s_export_op
= &ext4_export_ops
;
2171 sb
->s_xattr
= ext4_xattr_handlers
;
2173 sb
->s_qcop
= &ext4_qctl_operations
;
2174 sb
->dq_op
= &ext4_quota_operations
;
2176 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2180 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2181 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2182 EXT4_FEATURE_INCOMPAT_RECOVER
));
2185 * The first inode we look at is the journal inode. Don't try
2186 * root first: it may be modified in the journal!
2188 if (!test_opt(sb
, NOLOAD
) &&
2189 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2190 if (ext4_load_journal(sb
, es
, journal_devnum
))
2192 } else if (journal_inum
) {
2193 if (ext4_create_journal(sb
, es
, journal_inum
))
2198 "ext4: No journal on filesystem on %s\n",
2203 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2204 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2205 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2206 printk(KERN_ERR
"ext4: Failed to set 64-bit journal feature\n");
2210 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2211 jbd2_journal_set_features(sbi
->s_journal
,
2212 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2213 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2214 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2215 jbd2_journal_set_features(sbi
->s_journal
,
2216 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2217 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2218 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2220 jbd2_journal_clear_features(sbi
->s_journal
,
2221 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2222 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2225 /* We have now updated the journal if required, so we can
2226 * validate the data journaling mode. */
2227 switch (test_opt(sb
, DATA_FLAGS
)) {
2229 /* No mode set, assume a default based on the journal
2230 * capabilities: ORDERED_DATA if the journal can
2231 * cope, else JOURNAL_DATA
2233 if (jbd2_journal_check_available_features
2234 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2235 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2237 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2240 case EXT4_MOUNT_ORDERED_DATA
:
2241 case EXT4_MOUNT_WRITEBACK_DATA
:
2242 if (!jbd2_journal_check_available_features
2243 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2244 printk(KERN_ERR
"EXT4-fs: Journal does not support "
2245 "requested data journaling mode\n");
2252 if (test_opt(sb
, NOBH
)) {
2253 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2254 printk(KERN_WARNING
"EXT4-fs: Ignoring nobh option - "
2255 "its supported only with writeback mode\n");
2256 clear_opt(sbi
->s_mount_opt
, NOBH
);
2260 * The jbd2_journal_load will have done any necessary log recovery,
2261 * so we can safely mount the rest of the filesystem now.
2264 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2266 printk(KERN_ERR
"EXT4-fs: get root inode failed\n");
2267 ret
= PTR_ERR(root
);
2270 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2272 printk(KERN_ERR
"EXT4-fs: corrupt root inode, run e2fsck\n");
2275 sb
->s_root
= d_alloc_root(root
);
2277 printk(KERN_ERR
"EXT4-fs: get root dentry failed\n");
2283 ext4_setup_super (sb
, es
, sb
->s_flags
& MS_RDONLY
);
2285 /* determine the minimum size of new large inodes, if present */
2286 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2287 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2288 EXT4_GOOD_OLD_INODE_SIZE
;
2289 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2290 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2291 if (sbi
->s_want_extra_isize
<
2292 le16_to_cpu(es
->s_want_extra_isize
))
2293 sbi
->s_want_extra_isize
=
2294 le16_to_cpu(es
->s_want_extra_isize
);
2295 if (sbi
->s_want_extra_isize
<
2296 le16_to_cpu(es
->s_min_extra_isize
))
2297 sbi
->s_want_extra_isize
=
2298 le16_to_cpu(es
->s_min_extra_isize
);
2301 /* Check if enough inode space is available */
2302 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2303 sbi
->s_inode_size
) {
2304 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2305 EXT4_GOOD_OLD_INODE_SIZE
;
2306 printk(KERN_INFO
"EXT4-fs: required extra inode space not"
2311 * akpm: core read_super() calls in here with the superblock locked.
2312 * That deadlocks, because orphan cleanup needs to lock the superblock
2313 * in numerous places. Here we just pop the lock - it's relatively
2314 * harmless, because we are now ready to accept write_super() requests,
2315 * and aviro says that's the only reason for hanging onto the
2318 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
2319 ext4_orphan_cleanup(sb
, es
);
2320 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
2322 printk (KERN_INFO
"EXT4-fs: recovery complete.\n");
2323 ext4_mark_recovery_complete(sb
, es
);
2324 printk (KERN_INFO
"EXT4-fs: mounted filesystem with %s data mode.\n",
2325 test_opt(sb
,DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
? "journal":
2326 test_opt(sb
,DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
? "ordered":
2330 ext4_mb_init(sb
, needs_recovery
);
2337 printk(KERN_ERR
"VFS: Can't find ext4 filesystem on dev %s.\n",
2342 jbd2_journal_destroy(sbi
->s_journal
);
2344 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2345 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2346 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2348 for (i
= 0; i
< db_count
; i
++)
2349 brelse(sbi
->s_group_desc
[i
]);
2350 kfree(sbi
->s_group_desc
);
2353 for (i
= 0; i
< MAXQUOTAS
; i
++)
2354 kfree(sbi
->s_qf_names
[i
]);
2356 ext4_blkdev_remove(sbi
);
2359 sb
->s_fs_info
= NULL
;
2366 * Setup any per-fs journal parameters now. We'll do this both on
2367 * initial mount, once the journal has been initialised but before we've
2368 * done any recovery; and again on any subsequent remount.
2370 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
2372 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2374 if (sbi
->s_commit_interval
)
2375 journal
->j_commit_interval
= sbi
->s_commit_interval
;
2376 /* We could also set up an ext4-specific default for the commit
2377 * interval here, but for now we'll just fall back to the jbd
2380 spin_lock(&journal
->j_state_lock
);
2381 if (test_opt(sb
, BARRIER
))
2382 journal
->j_flags
|= JBD2_BARRIER
;
2384 journal
->j_flags
&= ~JBD2_BARRIER
;
2385 spin_unlock(&journal
->j_state_lock
);
2388 static journal_t
*ext4_get_journal(struct super_block
*sb
,
2389 unsigned int journal_inum
)
2391 struct inode
*journal_inode
;
2394 /* First, test for the existence of a valid inode on disk. Bad
2395 * things happen if we iget() an unused inode, as the subsequent
2396 * iput() will try to delete it. */
2398 journal_inode
= ext4_iget(sb
, journal_inum
);
2399 if (IS_ERR(journal_inode
)) {
2400 printk(KERN_ERR
"EXT4-fs: no journal found.\n");
2403 if (!journal_inode
->i_nlink
) {
2404 make_bad_inode(journal_inode
);
2405 iput(journal_inode
);
2406 printk(KERN_ERR
"EXT4-fs: journal inode is deleted.\n");
2410 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2411 journal_inode
, journal_inode
->i_size
);
2412 if (!S_ISREG(journal_inode
->i_mode
)) {
2413 printk(KERN_ERR
"EXT4-fs: invalid journal inode.\n");
2414 iput(journal_inode
);
2418 journal
= jbd2_journal_init_inode(journal_inode
);
2420 printk(KERN_ERR
"EXT4-fs: Could not load journal inode\n");
2421 iput(journal_inode
);
2424 journal
->j_private
= sb
;
2425 ext4_init_journal_params(sb
, journal
);
2429 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
2432 struct buffer_head
* bh
;
2436 int hblock
, blocksize
;
2437 ext4_fsblk_t sb_block
;
2438 unsigned long offset
;
2439 struct ext4_super_block
* es
;
2440 struct block_device
*bdev
;
2442 bdev
= ext4_blkdev_get(j_dev
);
2446 if (bd_claim(bdev
, sb
)) {
2448 "EXT4: failed to claim external journal device.\n");
2453 blocksize
= sb
->s_blocksize
;
2454 hblock
= bdev_hardsect_size(bdev
);
2455 if (blocksize
< hblock
) {
2457 "EXT4-fs: blocksize too small for journal device.\n");
2461 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
2462 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
2463 set_blocksize(bdev
, blocksize
);
2464 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
2465 printk(KERN_ERR
"EXT4-fs: couldn't read superblock of "
2466 "external journal\n");
2470 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2471 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
2472 !(le32_to_cpu(es
->s_feature_incompat
) &
2473 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
2474 printk(KERN_ERR
"EXT4-fs: external journal has "
2475 "bad superblock\n");
2480 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
2481 printk(KERN_ERR
"EXT4-fs: journal UUID does not match\n");
2486 len
= ext4_blocks_count(es
);
2487 start
= sb_block
+ 1;
2488 brelse(bh
); /* we're done with the superblock */
2490 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
2491 start
, len
, blocksize
);
2493 printk(KERN_ERR
"EXT4-fs: failed to create device journal\n");
2496 journal
->j_private
= sb
;
2497 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
2498 wait_on_buffer(journal
->j_sb_buffer
);
2499 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
2500 printk(KERN_ERR
"EXT4-fs: I/O error on journal device\n");
2503 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
2504 printk(KERN_ERR
"EXT4-fs: External journal has more than one "
2505 "user (unsupported) - %d\n",
2506 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
2509 EXT4_SB(sb
)->journal_bdev
= bdev
;
2510 ext4_init_journal_params(sb
, journal
);
2513 jbd2_journal_destroy(journal
);
2515 ext4_blkdev_put(bdev
);
2519 static int ext4_load_journal(struct super_block
*sb
,
2520 struct ext4_super_block
*es
,
2521 unsigned long journal_devnum
)
2524 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
2527 int really_read_only
;
2529 if (journal_devnum
&&
2530 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2531 printk(KERN_INFO
"EXT4-fs: external journal device major/minor "
2532 "numbers have changed\n");
2533 journal_dev
= new_decode_dev(journal_devnum
);
2535 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
2537 really_read_only
= bdev_read_only(sb
->s_bdev
);
2540 * Are we loading a blank journal or performing recovery after a
2541 * crash? For recovery, we need to check in advance whether we
2542 * can get read-write access to the device.
2545 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2546 if (sb
->s_flags
& MS_RDONLY
) {
2547 printk(KERN_INFO
"EXT4-fs: INFO: recovery "
2548 "required on readonly filesystem.\n");
2549 if (really_read_only
) {
2550 printk(KERN_ERR
"EXT4-fs: write access "
2551 "unavailable, cannot proceed.\n");
2554 printk (KERN_INFO
"EXT4-fs: write access will "
2555 "be enabled during recovery.\n");
2559 if (journal_inum
&& journal_dev
) {
2560 printk(KERN_ERR
"EXT4-fs: filesystem has both journal "
2561 "and inode journals!\n");
2566 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
2569 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
2573 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
2574 err
= jbd2_journal_update_format(journal
);
2576 printk(KERN_ERR
"EXT4-fs: error updating journal.\n");
2577 jbd2_journal_destroy(journal
);
2582 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
2583 err
= jbd2_journal_wipe(journal
, !really_read_only
);
2585 err
= jbd2_journal_load(journal
);
2588 printk(KERN_ERR
"EXT4-fs: error loading journal.\n");
2589 jbd2_journal_destroy(journal
);
2593 EXT4_SB(sb
)->s_journal
= journal
;
2594 ext4_clear_journal_err(sb
, es
);
2596 if (journal_devnum
&&
2597 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2598 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2601 /* Make sure we flush the recovery flag to disk. */
2602 ext4_commit_super(sb
, es
, 1);
2608 static int ext4_create_journal(struct super_block
* sb
,
2609 struct ext4_super_block
* es
,
2610 unsigned int journal_inum
)
2615 if (sb
->s_flags
& MS_RDONLY
) {
2616 printk(KERN_ERR
"EXT4-fs: readonly filesystem when trying to "
2617 "create journal.\n");
2621 journal
= ext4_get_journal(sb
, journal_inum
);
2625 printk(KERN_INFO
"EXT4-fs: creating new journal on inode %u\n",
2628 err
= jbd2_journal_create(journal
);
2630 printk(KERN_ERR
"EXT4-fs: error creating journal.\n");
2631 jbd2_journal_destroy(journal
);
2635 EXT4_SB(sb
)->s_journal
= journal
;
2637 ext4_update_dynamic_rev(sb
);
2638 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2639 EXT4_SET_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
);
2641 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
2644 /* Make sure we flush the recovery flag to disk. */
2645 ext4_commit_super(sb
, es
, 1);
2650 static void ext4_commit_super (struct super_block
* sb
,
2651 struct ext4_super_block
* es
,
2654 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
2658 es
->s_wtime
= cpu_to_le32(get_seconds());
2659 ext4_free_blocks_count_set(es
, ext4_count_free_blocks(sb
));
2660 es
->s_free_inodes_count
= cpu_to_le32(ext4_count_free_inodes(sb
));
2661 BUFFER_TRACE(sbh
, "marking dirty");
2662 mark_buffer_dirty(sbh
);
2664 sync_dirty_buffer(sbh
);
2669 * Have we just finished recovery? If so, and if we are mounting (or
2670 * remounting) the filesystem readonly, then we will end up with a
2671 * consistent fs on disk. Record that fact.
2673 static void ext4_mark_recovery_complete(struct super_block
* sb
,
2674 struct ext4_super_block
* es
)
2676 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
2678 jbd2_journal_lock_updates(journal
);
2679 jbd2_journal_flush(journal
);
2681 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
2682 sb
->s_flags
& MS_RDONLY
) {
2683 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2685 ext4_commit_super(sb
, es
, 1);
2688 jbd2_journal_unlock_updates(journal
);
2692 * If we are mounting (or read-write remounting) a filesystem whose journal
2693 * has recorded an error from a previous lifetime, move that error to the
2694 * main filesystem now.
2696 static void ext4_clear_journal_err(struct super_block
* sb
,
2697 struct ext4_super_block
* es
)
2703 journal
= EXT4_SB(sb
)->s_journal
;
2706 * Now check for any error status which may have been recorded in the
2707 * journal by a prior ext4_error() or ext4_abort()
2710 j_errno
= jbd2_journal_errno(journal
);
2714 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
2715 ext4_warning(sb
, __func__
, "Filesystem error recorded "
2716 "from previous mount: %s", errstr
);
2717 ext4_warning(sb
, __func__
, "Marking fs in need of "
2718 "filesystem check.");
2720 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2721 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2722 ext4_commit_super (sb
, es
, 1);
2724 jbd2_journal_clear_err(journal
);
2729 * Force the running and committing transactions to commit,
2730 * and wait on the commit.
2732 int ext4_force_commit(struct super_block
*sb
)
2737 if (sb
->s_flags
& MS_RDONLY
)
2740 journal
= EXT4_SB(sb
)->s_journal
;
2742 ret
= ext4_journal_force_commit(journal
);
2747 * Ext4 always journals updates to the superblock itself, so we don't
2748 * have to propagate any other updates to the superblock on disk at this
2749 * point. Just start an async writeback to get the buffers on their way
2752 * This implicitly triggers the writebehind on sync().
2755 static void ext4_write_super (struct super_block
* sb
)
2757 if (mutex_trylock(&sb
->s_lock
) != 0)
2762 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
2767 if (jbd2_journal_start_commit(EXT4_SB(sb
)->s_journal
, &target
)) {
2769 jbd2_log_wait_commit(EXT4_SB(sb
)->s_journal
, target
);
2775 * LVM calls this function before a (read-only) snapshot is created. This
2776 * gives us a chance to flush the journal completely and mark the fs clean.
2778 static void ext4_write_super_lockfs(struct super_block
*sb
)
2782 if (!(sb
->s_flags
& MS_RDONLY
)) {
2783 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
2785 /* Now we set up the journal barrier. */
2786 jbd2_journal_lock_updates(journal
);
2787 jbd2_journal_flush(journal
);
2789 /* Journal blocked and flushed, clear needs_recovery flag. */
2790 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2791 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
2796 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2797 * flag here, even though the filesystem is not technically dirty yet.
2799 static void ext4_unlockfs(struct super_block
*sb
)
2801 if (!(sb
->s_flags
& MS_RDONLY
)) {
2803 /* Reser the needs_recovery flag before the fs is unlocked. */
2804 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2805 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
2807 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
2811 static int ext4_remount (struct super_block
* sb
, int * flags
, char * data
)
2813 struct ext4_super_block
* es
;
2814 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2815 ext4_fsblk_t n_blocks_count
= 0;
2816 unsigned long old_sb_flags
;
2817 struct ext4_mount_options old_opts
;
2823 /* Store the original options */
2824 old_sb_flags
= sb
->s_flags
;
2825 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
2826 old_opts
.s_resuid
= sbi
->s_resuid
;
2827 old_opts
.s_resgid
= sbi
->s_resgid
;
2828 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
2830 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
2831 for (i
= 0; i
< MAXQUOTAS
; i
++)
2832 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
2836 * Allow the "check" option to be passed as a remount option.
2838 if (!parse_options(data
, sb
, NULL
, NULL
, &n_blocks_count
, 1)) {
2843 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
)
2844 ext4_abort(sb
, __func__
, "Abort forced by user");
2846 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2847 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2851 ext4_init_journal_params(sb
, sbi
->s_journal
);
2853 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
2854 n_blocks_count
> ext4_blocks_count(es
)) {
2855 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
) {
2860 if (*flags
& MS_RDONLY
) {
2862 * First of all, the unconditional stuff we have to do
2863 * to disable replay of the journal when we next remount
2865 sb
->s_flags
|= MS_RDONLY
;
2868 * OK, test if we are remounting a valid rw partition
2869 * readonly, and if so set the rdonly flag and then
2870 * mark the partition as valid again.
2872 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
2873 (sbi
->s_mount_state
& EXT4_VALID_FS
))
2874 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
2877 * We have to unlock super so that we can wait for
2881 ext4_mark_recovery_complete(sb
, es
);
2885 if ((ret
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2886 ~EXT4_FEATURE_RO_COMPAT_SUPP
))) {
2887 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
2888 "remount RDWR because of unsupported "
2889 "optional features (%x).\n",
2890 sb
->s_id
, le32_to_cpu(ret
));
2896 * If we have an unprocessed orphan list hanging
2897 * around from a previously readonly bdev mount,
2898 * require a full umount/remount for now.
2900 if (es
->s_last_orphan
) {
2901 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
2902 "remount RDWR because of unprocessed "
2903 "orphan inode list. Please "
2904 "umount/remount instead.\n",
2911 * Mounting a RDONLY partition read-write, so reread
2912 * and store the current valid flag. (It may have
2913 * been changed by e2fsck since we originally mounted
2916 ext4_clear_journal_err(sb
, es
);
2917 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2918 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
2920 if (!ext4_setup_super (sb
, es
, 0))
2921 sb
->s_flags
&= ~MS_RDONLY
;
2925 /* Release old quota file names */
2926 for (i
= 0; i
< MAXQUOTAS
; i
++)
2927 if (old_opts
.s_qf_names
[i
] &&
2928 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2929 kfree(old_opts
.s_qf_names
[i
]);
2933 sb
->s_flags
= old_sb_flags
;
2934 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
2935 sbi
->s_resuid
= old_opts
.s_resuid
;
2936 sbi
->s_resgid
= old_opts
.s_resgid
;
2937 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
2939 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
2940 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2941 if (sbi
->s_qf_names
[i
] &&
2942 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2943 kfree(sbi
->s_qf_names
[i
]);
2944 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
2950 static int ext4_statfs (struct dentry
* dentry
, struct kstatfs
* buf
)
2952 struct super_block
*sb
= dentry
->d_sb
;
2953 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2954 struct ext4_super_block
*es
= sbi
->s_es
;
2957 if (test_opt(sb
, MINIX_DF
)) {
2958 sbi
->s_overhead_last
= 0;
2959 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
2960 ext4_group_t ngroups
= sbi
->s_groups_count
, i
;
2961 ext4_fsblk_t overhead
= 0;
2965 * Compute the overhead (FS structures). This is constant
2966 * for a given filesystem unless the number of block groups
2967 * changes so we cache the previous value until it does.
2971 * All of the blocks before first_data_block are
2974 overhead
= le32_to_cpu(es
->s_first_data_block
);
2977 * Add the overhead attributed to the superblock and
2978 * block group descriptors. If the sparse superblocks
2979 * feature is turned on, then not all groups have this.
2981 for (i
= 0; i
< ngroups
; i
++) {
2982 overhead
+= ext4_bg_has_super(sb
, i
) +
2983 ext4_bg_num_gdb(sb
, i
);
2988 * Every block group has an inode bitmap, a block
2989 * bitmap, and an inode table.
2991 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
2992 sbi
->s_overhead_last
= overhead
;
2994 sbi
->s_blocks_last
= ext4_blocks_count(es
);
2997 buf
->f_type
= EXT4_SUPER_MAGIC
;
2998 buf
->f_bsize
= sb
->s_blocksize
;
2999 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3000 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
);
3001 ext4_free_blocks_count_set(es
, buf
->f_bfree
);
3002 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3003 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3005 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3006 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3007 es
->s_free_inodes_count
= cpu_to_le32(buf
->f_ffree
);
3008 buf
->f_namelen
= EXT4_NAME_LEN
;
3009 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3010 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3011 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3012 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3016 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3017 * is locked for write. Otherwise the are possible deadlocks:
3018 * Process 1 Process 2
3019 * ext4_create() quota_sync()
3020 * jbd2_journal_start() write_dquot()
3021 * DQUOT_INIT() down(dqio_mutex)
3022 * down(dqio_mutex) jbd2_journal_start()
3028 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3030 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3033 static int ext4_dquot_initialize(struct inode
*inode
, int type
)
3038 /* We may create quota structure so we need to reserve enough blocks */
3039 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_INIT_BLOCKS(inode
->i_sb
));
3041 return PTR_ERR(handle
);
3042 ret
= dquot_initialize(inode
, type
);
3043 err
= ext4_journal_stop(handle
);
3049 static int ext4_dquot_drop(struct inode
*inode
)
3054 /* We may delete quota structure so we need to reserve enough blocks */
3055 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_DEL_BLOCKS(inode
->i_sb
));
3056 if (IS_ERR(handle
)) {
3058 * We call dquot_drop() anyway to at least release references
3059 * to quota structures so that umount does not hang.
3062 return PTR_ERR(handle
);
3064 ret
= dquot_drop(inode
);
3065 err
= ext4_journal_stop(handle
);
3071 static int ext4_write_dquot(struct dquot
*dquot
)
3075 struct inode
*inode
;
3077 inode
= dquot_to_inode(dquot
);
3078 handle
= ext4_journal_start(inode
,
3079 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3081 return PTR_ERR(handle
);
3082 ret
= dquot_commit(dquot
);
3083 err
= ext4_journal_stop(handle
);
3089 static int ext4_acquire_dquot(struct dquot
*dquot
)
3094 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3095 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3097 return PTR_ERR(handle
);
3098 ret
= dquot_acquire(dquot
);
3099 err
= ext4_journal_stop(handle
);
3105 static int ext4_release_dquot(struct dquot
*dquot
)
3110 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3111 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3112 if (IS_ERR(handle
)) {
3113 /* Release dquot anyway to avoid endless cycle in dqput() */
3114 dquot_release(dquot
);
3115 return PTR_ERR(handle
);
3117 ret
= dquot_release(dquot
);
3118 err
= ext4_journal_stop(handle
);
3124 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3126 /* Are we journaling quotas? */
3127 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3128 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3129 dquot_mark_dquot_dirty(dquot
);
3130 return ext4_write_dquot(dquot
);
3132 return dquot_mark_dquot_dirty(dquot
);
3136 static int ext4_write_info(struct super_block
*sb
, int type
)
3141 /* Data block + inode block */
3142 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3144 return PTR_ERR(handle
);
3145 ret
= dquot_commit_info(sb
, type
);
3146 err
= ext4_journal_stop(handle
);
3153 * Turn on quotas during mount time - we need to find
3154 * the quota file and such...
3156 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3158 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3159 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3163 * Standard function to be called on quota_on
3165 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3166 char *path
, int remount
)
3169 struct nameidata nd
;
3171 if (!test_opt(sb
, QUOTA
))
3173 /* When remounting, no checks are needed and in fact, path is NULL */
3175 return vfs_quota_on(sb
, type
, format_id
, path
, remount
);
3177 err
= path_lookup(path
, LOOKUP_FOLLOW
, &nd
);
3181 /* Quotafile not on the same filesystem? */
3182 if (nd
.path
.mnt
->mnt_sb
!= sb
) {
3186 /* Journaling quota? */
3187 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3188 /* Quotafile not of fs root? */
3189 if (nd
.path
.dentry
->d_parent
->d_inode
!= sb
->s_root
->d_inode
)
3191 "EXT4-fs: Quota file not on filesystem root. "
3192 "Journaled quota will not work.\n");
3196 * When we journal data on quota file, we have to flush journal to see
3197 * all updates to the file when we bypass pagecache...
3199 if (ext4_should_journal_data(nd
.path
.dentry
->d_inode
)) {
3201 * We don't need to lock updates but journal_flush() could
3202 * otherwise be livelocked...
3204 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3205 jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3206 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3210 return vfs_quota_on(sb
, type
, format_id
, path
, remount
);
3213 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3214 * acquiring the locks... As quota files are never truncated and quota code
3215 * itself serializes the operations (and noone else should touch the files)
3216 * we don't have to be afraid of races */
3217 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3218 size_t len
, loff_t off
)
3220 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3221 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3223 int offset
= off
& (sb
->s_blocksize
- 1);
3226 struct buffer_head
*bh
;
3227 loff_t i_size
= i_size_read(inode
);
3231 if (off
+len
> i_size
)
3234 while (toread
> 0) {
3235 tocopy
= sb
->s_blocksize
- offset
< toread
?
3236 sb
->s_blocksize
- offset
: toread
;
3237 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3240 if (!bh
) /* A hole? */
3241 memset(data
, 0, tocopy
);
3243 memcpy(data
, bh
->b_data
+offset
, tocopy
);
3253 /* Write to quotafile (we know the transaction is already started and has
3254 * enough credits) */
3255 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
3256 const char *data
, size_t len
, loff_t off
)
3258 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3259 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3261 int offset
= off
& (sb
->s_blocksize
- 1);
3263 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
3264 size_t towrite
= len
;
3265 struct buffer_head
*bh
;
3266 handle_t
*handle
= journal_current_handle();
3269 printk(KERN_WARNING
"EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3270 " cancelled because transaction is not started.\n",
3271 (unsigned long long)off
, (unsigned long long)len
);
3274 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
3275 while (towrite
> 0) {
3276 tocopy
= sb
->s_blocksize
- offset
< towrite
?
3277 sb
->s_blocksize
- offset
: towrite
;
3278 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
3281 if (journal_quota
) {
3282 err
= ext4_journal_get_write_access(handle
, bh
);
3289 memcpy(bh
->b_data
+offset
, data
, tocopy
);
3290 flush_dcache_page(bh
->b_page
);
3293 err
= ext4_journal_dirty_metadata(handle
, bh
);
3295 /* Always do at least ordered writes for quotas */
3296 err
= ext4_journal_dirty_data(handle
, bh
);
3297 mark_buffer_dirty(bh
);
3310 if (inode
->i_size
< off
+len
-towrite
) {
3311 i_size_write(inode
, off
+len
-towrite
);
3312 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
3314 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3315 ext4_mark_inode_dirty(handle
, inode
);
3316 mutex_unlock(&inode
->i_mutex
);
3317 return len
- towrite
;
3322 static int ext4_get_sb(struct file_system_type
*fs_type
,
3323 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3325 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3328 static struct file_system_type ext4dev_fs_type
= {
3329 .owner
= THIS_MODULE
,
3331 .get_sb
= ext4_get_sb
,
3332 .kill_sb
= kill_block_super
,
3333 .fs_flags
= FS_REQUIRES_DEV
,
3336 static int __init
init_ext4_fs(void)
3340 err
= init_ext4_mballoc();
3344 err
= init_ext4_xattr();
3347 err
= init_inodecache();
3350 err
= register_filesystem(&ext4dev_fs_type
);
3355 destroy_inodecache();
3359 exit_ext4_mballoc();
3363 static void __exit
exit_ext4_fs(void)
3365 unregister_filesystem(&ext4dev_fs_type
);
3366 destroy_inodecache();
3368 exit_ext4_mballoc();
3371 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3372 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3373 MODULE_LICENSE("GPL");
3374 module_init(init_ext4_fs
)
3375 module_exit(exit_ext4_fs
)