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 sbi
->s_journal
= NULL
;
510 if (!(sb
->s_flags
& MS_RDONLY
)) {
511 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
512 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
513 BUFFER_TRACE(sbi
->s_sbh
, "marking dirty");
514 mark_buffer_dirty(sbi
->s_sbh
);
515 ext4_commit_super(sb
, es
, 1);
518 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
519 brelse(sbi
->s_group_desc
[i
]);
520 kfree(sbi
->s_group_desc
);
521 kfree(sbi
->s_flex_groups
);
522 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
523 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
524 percpu_counter_destroy(&sbi
->s_dirs_counter
);
527 for (i
= 0; i
< MAXQUOTAS
; i
++)
528 kfree(sbi
->s_qf_names
[i
]);
531 /* Debugging code just in case the in-memory inode orphan list
532 * isn't empty. The on-disk one can be non-empty if we've
533 * detected an error and taken the fs readonly, but the
534 * in-memory list had better be clean by this point. */
535 if (!list_empty(&sbi
->s_orphan
))
536 dump_orphan_list(sb
, sbi
);
537 J_ASSERT(list_empty(&sbi
->s_orphan
));
539 invalidate_bdev(sb
->s_bdev
);
540 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
542 * Invalidate the journal device's buffers. We don't want them
543 * floating about in memory - the physical journal device may
544 * hotswapped, and it breaks the `ro-after' testing code.
546 sync_blockdev(sbi
->journal_bdev
);
547 invalidate_bdev(sbi
->journal_bdev
);
548 ext4_blkdev_remove(sbi
);
550 sb
->s_fs_info
= NULL
;
555 static struct kmem_cache
*ext4_inode_cachep
;
558 * Called inside transaction, so use GFP_NOFS
560 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
562 struct ext4_inode_info
*ei
;
564 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
567 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
568 ei
->i_acl
= EXT4_ACL_NOT_CACHED
;
569 ei
->i_default_acl
= EXT4_ACL_NOT_CACHED
;
571 ei
->i_block_alloc_info
= NULL
;
572 ei
->vfs_inode
.i_version
= 1;
573 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
574 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
575 spin_lock_init(&ei
->i_prealloc_lock
);
576 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
577 ei
->i_reserved_data_blocks
= 0;
578 ei
->i_reserved_meta_blocks
= 0;
579 ei
->i_allocated_meta_blocks
= 0;
580 ei
->i_delalloc_reserved_flag
= 0;
581 spin_lock_init(&(ei
->i_block_reservation_lock
));
582 return &ei
->vfs_inode
;
585 static void ext4_destroy_inode(struct inode
*inode
)
587 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
588 printk("EXT4 Inode %p: orphan list check failed!\n",
590 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
591 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
595 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
598 static void init_once(struct kmem_cache
*cachep
, void *foo
)
600 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
602 INIT_LIST_HEAD(&ei
->i_orphan
);
603 #ifdef CONFIG_EXT4DEV_FS_XATTR
604 init_rwsem(&ei
->xattr_sem
);
606 init_rwsem(&ei
->i_data_sem
);
607 inode_init_once(&ei
->vfs_inode
);
610 static int init_inodecache(void)
612 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
613 sizeof(struct ext4_inode_info
),
614 0, (SLAB_RECLAIM_ACCOUNT
|
617 if (ext4_inode_cachep
== NULL
)
622 static void destroy_inodecache(void)
624 kmem_cache_destroy(ext4_inode_cachep
);
627 static void ext4_clear_inode(struct inode
*inode
)
629 struct ext4_block_alloc_info
*rsv
= EXT4_I(inode
)->i_block_alloc_info
;
630 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
631 if (EXT4_I(inode
)->i_acl
&&
632 EXT4_I(inode
)->i_acl
!= EXT4_ACL_NOT_CACHED
) {
633 posix_acl_release(EXT4_I(inode
)->i_acl
);
634 EXT4_I(inode
)->i_acl
= EXT4_ACL_NOT_CACHED
;
636 if (EXT4_I(inode
)->i_default_acl
&&
637 EXT4_I(inode
)->i_default_acl
!= EXT4_ACL_NOT_CACHED
) {
638 posix_acl_release(EXT4_I(inode
)->i_default_acl
);
639 EXT4_I(inode
)->i_default_acl
= EXT4_ACL_NOT_CACHED
;
642 ext4_discard_reservation(inode
);
643 EXT4_I(inode
)->i_block_alloc_info
= NULL
;
646 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
647 &EXT4_I(inode
)->jinode
);
650 static inline void ext4_show_quota_options(struct seq_file
*seq
, struct super_block
*sb
)
652 #if defined(CONFIG_QUOTA)
653 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
655 if (sbi
->s_jquota_fmt
)
656 seq_printf(seq
, ",jqfmt=%s",
657 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold": "vfsv0");
659 if (sbi
->s_qf_names
[USRQUOTA
])
660 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
662 if (sbi
->s_qf_names
[GRPQUOTA
])
663 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
665 if (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
)
666 seq_puts(seq
, ",usrquota");
668 if (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)
669 seq_puts(seq
, ",grpquota");
675 * - it's set to a non-default value OR
676 * - if the per-sb default is different from the global default
678 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
681 unsigned long def_mount_opts
;
682 struct super_block
*sb
= vfs
->mnt_sb
;
683 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
684 struct ext4_super_block
*es
= sbi
->s_es
;
686 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
687 def_errors
= le16_to_cpu(es
->s_errors
);
689 if (sbi
->s_sb_block
!= 1)
690 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
691 if (test_opt(sb
, MINIX_DF
))
692 seq_puts(seq
, ",minixdf");
693 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
694 seq_puts(seq
, ",grpid");
695 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
696 seq_puts(seq
, ",nogrpid");
697 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
698 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
699 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
701 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
702 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
703 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
705 if (test_opt(sb
, ERRORS_RO
)) {
706 if (def_errors
== EXT4_ERRORS_PANIC
||
707 def_errors
== EXT4_ERRORS_CONTINUE
) {
708 seq_puts(seq
, ",errors=remount-ro");
711 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
712 seq_puts(seq
, ",errors=continue");
713 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
714 seq_puts(seq
, ",errors=panic");
715 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
716 seq_puts(seq
, ",nouid32");
717 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
718 seq_puts(seq
, ",debug");
719 if (test_opt(sb
, OLDALLOC
))
720 seq_puts(seq
, ",oldalloc");
721 #ifdef CONFIG_EXT4DEV_FS_XATTR
722 if (test_opt(sb
, XATTR_USER
) &&
723 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
724 seq_puts(seq
, ",user_xattr");
725 if (!test_opt(sb
, XATTR_USER
) &&
726 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
727 seq_puts(seq
, ",nouser_xattr");
730 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
731 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
732 seq_puts(seq
, ",acl");
733 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
734 seq_puts(seq
, ",noacl");
736 if (!test_opt(sb
, RESERVATION
))
737 seq_puts(seq
, ",noreservation");
738 if (sbi
->s_commit_interval
) {
739 seq_printf(seq
, ",commit=%u",
740 (unsigned) (sbi
->s_commit_interval
/ HZ
));
743 * We're changing the default of barrier mount option, so
744 * let's always display its mount state so it's clear what its
747 seq_puts(seq
, ",barrier=");
748 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
749 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
750 seq_puts(seq
, ",journal_async_commit");
751 if (test_opt(sb
, NOBH
))
752 seq_puts(seq
, ",nobh");
753 if (!test_opt(sb
, EXTENTS
))
754 seq_puts(seq
, ",noextents");
755 if (!test_opt(sb
, MBALLOC
))
756 seq_puts(seq
, ",nomballoc");
757 if (test_opt(sb
, I_VERSION
))
758 seq_puts(seq
, ",i_version");
759 if (!test_opt(sb
, DELALLOC
))
760 seq_puts(seq
, ",nodelalloc");
764 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
766 * journal mode get enabled in different ways
767 * So just print the value even if we didn't specify it
769 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
770 seq_puts(seq
, ",data=journal");
771 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
772 seq_puts(seq
, ",data=ordered");
773 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
774 seq_puts(seq
, ",data=writeback");
776 ext4_show_quota_options(seq
, sb
);
781 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
782 u64 ino
, u32 generation
)
786 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
787 return ERR_PTR(-ESTALE
);
788 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
789 return ERR_PTR(-ESTALE
);
791 /* iget isn't really right if the inode is currently unallocated!!
793 * ext4_read_inode will return a bad_inode if the inode had been
794 * deleted, so we should be safe.
796 * Currently we don't know the generation for parent directory, so
797 * a generation of 0 means "accept any"
799 inode
= ext4_iget(sb
, ino
);
801 return ERR_CAST(inode
);
802 if (generation
&& inode
->i_generation
!= generation
) {
804 return ERR_PTR(-ESTALE
);
810 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
811 int fh_len
, int fh_type
)
813 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
817 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
818 int fh_len
, int fh_type
)
820 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
825 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
826 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
828 static int ext4_dquot_initialize(struct inode
*inode
, int type
);
829 static int ext4_dquot_drop(struct inode
*inode
);
830 static int ext4_write_dquot(struct dquot
*dquot
);
831 static int ext4_acquire_dquot(struct dquot
*dquot
);
832 static int ext4_release_dquot(struct dquot
*dquot
);
833 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
834 static int ext4_write_info(struct super_block
*sb
, int type
);
835 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
836 char *path
, int remount
);
837 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
838 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
839 size_t len
, loff_t off
);
840 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
841 const char *data
, size_t len
, loff_t off
);
843 static struct dquot_operations ext4_quota_operations
= {
844 .initialize
= ext4_dquot_initialize
,
845 .drop
= ext4_dquot_drop
,
846 .alloc_space
= dquot_alloc_space
,
847 .alloc_inode
= dquot_alloc_inode
,
848 .free_space
= dquot_free_space
,
849 .free_inode
= dquot_free_inode
,
850 .transfer
= dquot_transfer
,
851 .write_dquot
= ext4_write_dquot
,
852 .acquire_dquot
= ext4_acquire_dquot
,
853 .release_dquot
= ext4_release_dquot
,
854 .mark_dirty
= ext4_mark_dquot_dirty
,
855 .write_info
= ext4_write_info
858 static struct quotactl_ops ext4_qctl_operations
= {
859 .quota_on
= ext4_quota_on
,
860 .quota_off
= vfs_quota_off
,
861 .quota_sync
= vfs_quota_sync
,
862 .get_info
= vfs_get_dqinfo
,
863 .set_info
= vfs_set_dqinfo
,
864 .get_dqblk
= vfs_get_dqblk
,
865 .set_dqblk
= vfs_set_dqblk
869 static const struct super_operations ext4_sops
= {
870 .alloc_inode
= ext4_alloc_inode
,
871 .destroy_inode
= ext4_destroy_inode
,
872 .write_inode
= ext4_write_inode
,
873 .dirty_inode
= ext4_dirty_inode
,
874 .delete_inode
= ext4_delete_inode
,
875 .put_super
= ext4_put_super
,
876 .write_super
= ext4_write_super
,
877 .sync_fs
= ext4_sync_fs
,
878 .write_super_lockfs
= ext4_write_super_lockfs
,
879 .unlockfs
= ext4_unlockfs
,
880 .statfs
= ext4_statfs
,
881 .remount_fs
= ext4_remount
,
882 .clear_inode
= ext4_clear_inode
,
883 .show_options
= ext4_show_options
,
885 .quota_read
= ext4_quota_read
,
886 .quota_write
= ext4_quota_write
,
890 static const struct export_operations ext4_export_ops
= {
891 .fh_to_dentry
= ext4_fh_to_dentry
,
892 .fh_to_parent
= ext4_fh_to_parent
,
893 .get_parent
= ext4_get_parent
,
897 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
898 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
899 Opt_nouid32
, Opt_nocheck
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
900 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
901 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
902 Opt_commit
, Opt_journal_update
, Opt_journal_inum
, Opt_journal_dev
,
903 Opt_journal_checksum
, Opt_journal_async_commit
,
904 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
905 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
906 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
907 Opt_ignore
, Opt_barrier
, Opt_err
, Opt_resize
, Opt_usrquota
,
908 Opt_grpquota
, Opt_extents
, Opt_noextents
, Opt_i_version
,
909 Opt_mballoc
, Opt_nomballoc
, Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
912 static match_table_t tokens
= {
913 {Opt_bsd_df
, "bsddf"},
914 {Opt_minix_df
, "minixdf"},
915 {Opt_grpid
, "grpid"},
916 {Opt_grpid
, "bsdgroups"},
917 {Opt_nogrpid
, "nogrpid"},
918 {Opt_nogrpid
, "sysvgroups"},
919 {Opt_resgid
, "resgid=%u"},
920 {Opt_resuid
, "resuid=%u"},
922 {Opt_err_cont
, "errors=continue"},
923 {Opt_err_panic
, "errors=panic"},
924 {Opt_err_ro
, "errors=remount-ro"},
925 {Opt_nouid32
, "nouid32"},
926 {Opt_nocheck
, "nocheck"},
927 {Opt_nocheck
, "check=none"},
928 {Opt_debug
, "debug"},
929 {Opt_oldalloc
, "oldalloc"},
930 {Opt_orlov
, "orlov"},
931 {Opt_user_xattr
, "user_xattr"},
932 {Opt_nouser_xattr
, "nouser_xattr"},
934 {Opt_noacl
, "noacl"},
935 {Opt_reservation
, "reservation"},
936 {Opt_noreservation
, "noreservation"},
937 {Opt_noload
, "noload"},
940 {Opt_commit
, "commit=%u"},
941 {Opt_journal_update
, "journal=update"},
942 {Opt_journal_inum
, "journal=%u"},
943 {Opt_journal_dev
, "journal_dev=%u"},
944 {Opt_journal_checksum
, "journal_checksum"},
945 {Opt_journal_async_commit
, "journal_async_commit"},
946 {Opt_abort
, "abort"},
947 {Opt_data_journal
, "data=journal"},
948 {Opt_data_ordered
, "data=ordered"},
949 {Opt_data_writeback
, "data=writeback"},
950 {Opt_offusrjquota
, "usrjquota="},
951 {Opt_usrjquota
, "usrjquota=%s"},
952 {Opt_offgrpjquota
, "grpjquota="},
953 {Opt_grpjquota
, "grpjquota=%s"},
954 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
955 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
956 {Opt_grpquota
, "grpquota"},
957 {Opt_noquota
, "noquota"},
958 {Opt_quota
, "quota"},
959 {Opt_usrquota
, "usrquota"},
960 {Opt_barrier
, "barrier=%u"},
961 {Opt_extents
, "extents"},
962 {Opt_noextents
, "noextents"},
963 {Opt_i_version
, "i_version"},
964 {Opt_mballoc
, "mballoc"},
965 {Opt_nomballoc
, "nomballoc"},
966 {Opt_stripe
, "stripe=%u"},
967 {Opt_resize
, "resize"},
968 {Opt_delalloc
, "delalloc"},
969 {Opt_nodelalloc
, "nodelalloc"},
973 static ext4_fsblk_t
get_sb_block(void **data
)
975 ext4_fsblk_t sb_block
;
976 char *options
= (char *) *data
;
978 if (!options
|| strncmp(options
, "sb=", 3) != 0)
979 return 1; /* Default location */
981 /*todo: use simple_strtoll with >32bit ext4 */
982 sb_block
= simple_strtoul(options
, &options
, 0);
983 if (*options
&& *options
!= ',') {
984 printk("EXT4-fs: Invalid sb specification: %s\n",
990 *data
= (void *) options
;
994 static int parse_options (char *options
, struct super_block
*sb
,
995 unsigned int *inum
, unsigned long *journal_devnum
,
996 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
998 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1000 substring_t args
[MAX_OPT_ARGS
];
1011 while ((p
= strsep (&options
, ",")) != NULL
) {
1016 token
= match_token(p
, tokens
, args
);
1019 clear_opt (sbi
->s_mount_opt
, MINIX_DF
);
1022 set_opt (sbi
->s_mount_opt
, MINIX_DF
);
1025 set_opt (sbi
->s_mount_opt
, GRPID
);
1028 clear_opt (sbi
->s_mount_opt
, GRPID
);
1031 if (match_int(&args
[0], &option
))
1033 sbi
->s_resuid
= option
;
1036 if (match_int(&args
[0], &option
))
1038 sbi
->s_resgid
= option
;
1041 /* handled by get_sb_block() instead of here */
1042 /* *sb_block = match_int(&args[0]); */
1045 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1046 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1047 set_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1050 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1051 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1052 set_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1055 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1056 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1057 set_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1060 set_opt (sbi
->s_mount_opt
, NO_UID32
);
1063 clear_opt (sbi
->s_mount_opt
, CHECK
);
1066 set_opt (sbi
->s_mount_opt
, DEBUG
);
1069 set_opt (sbi
->s_mount_opt
, OLDALLOC
);
1072 clear_opt (sbi
->s_mount_opt
, OLDALLOC
);
1074 #ifdef CONFIG_EXT4DEV_FS_XATTR
1075 case Opt_user_xattr
:
1076 set_opt (sbi
->s_mount_opt
, XATTR_USER
);
1078 case Opt_nouser_xattr
:
1079 clear_opt (sbi
->s_mount_opt
, XATTR_USER
);
1082 case Opt_user_xattr
:
1083 case Opt_nouser_xattr
:
1084 printk("EXT4 (no)user_xattr options not supported\n");
1087 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1089 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1092 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1097 printk("EXT4 (no)acl options not supported\n");
1100 case Opt_reservation
:
1101 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1103 case Opt_noreservation
:
1104 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
1106 case Opt_journal_update
:
1108 /* Eventually we will want to be able to create
1109 a journal file here. For now, only allow the
1110 user to specify an existing inode to be the
1113 printk(KERN_ERR
"EXT4-fs: cannot specify "
1114 "journal on remount\n");
1117 set_opt (sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1119 case Opt_journal_inum
:
1121 printk(KERN_ERR
"EXT4-fs: cannot specify "
1122 "journal on remount\n");
1125 if (match_int(&args
[0], &option
))
1129 case Opt_journal_dev
:
1131 printk(KERN_ERR
"EXT4-fs: cannot specify "
1132 "journal on remount\n");
1135 if (match_int(&args
[0], &option
))
1137 *journal_devnum
= option
;
1139 case Opt_journal_checksum
:
1140 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1142 case Opt_journal_async_commit
:
1143 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1144 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1147 set_opt (sbi
->s_mount_opt
, NOLOAD
);
1150 if (match_int(&args
[0], &option
))
1155 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1156 sbi
->s_commit_interval
= HZ
* option
;
1158 case Opt_data_journal
:
1159 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1161 case Opt_data_ordered
:
1162 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1164 case Opt_data_writeback
:
1165 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1168 if ((sbi
->s_mount_opt
& EXT4_MOUNT_DATA_FLAGS
)
1171 "EXT4-fs: cannot change data "
1172 "mode on remount\n");
1176 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DATA_FLAGS
;
1177 sbi
->s_mount_opt
|= data_opt
;
1187 if ((sb_any_quota_enabled(sb
) ||
1188 sb_any_quota_suspended(sb
)) &&
1189 !sbi
->s_qf_names
[qtype
]) {
1191 "EXT4-fs: Cannot change journaled "
1192 "quota options when quota turned on.\n");
1195 qname
= match_strdup(&args
[0]);
1198 "EXT4-fs: not enough memory for "
1199 "storing quotafile name.\n");
1202 if (sbi
->s_qf_names
[qtype
] &&
1203 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1205 "EXT4-fs: %s quota file already "
1206 "specified.\n", QTYPE2NAME(qtype
));
1210 sbi
->s_qf_names
[qtype
] = qname
;
1211 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1213 "EXT4-fs: quotafile must be on "
1214 "filesystem root.\n");
1215 kfree(sbi
->s_qf_names
[qtype
]);
1216 sbi
->s_qf_names
[qtype
] = NULL
;
1219 set_opt(sbi
->s_mount_opt
, QUOTA
);
1221 case Opt_offusrjquota
:
1224 case Opt_offgrpjquota
:
1227 if ((sb_any_quota_enabled(sb
) ||
1228 sb_any_quota_suspended(sb
)) &&
1229 sbi
->s_qf_names
[qtype
]) {
1230 printk(KERN_ERR
"EXT4-fs: Cannot change "
1231 "journaled quota options when "
1232 "quota turned on.\n");
1236 * The space will be released later when all options
1237 * are confirmed to be correct
1239 sbi
->s_qf_names
[qtype
] = NULL
;
1241 case Opt_jqfmt_vfsold
:
1242 qfmt
= QFMT_VFS_OLD
;
1244 case Opt_jqfmt_vfsv0
:
1247 if ((sb_any_quota_enabled(sb
) ||
1248 sb_any_quota_suspended(sb
)) &&
1249 sbi
->s_jquota_fmt
!= qfmt
) {
1250 printk(KERN_ERR
"EXT4-fs: Cannot change "
1251 "journaled quota options when "
1252 "quota turned on.\n");
1255 sbi
->s_jquota_fmt
= qfmt
;
1259 set_opt(sbi
->s_mount_opt
, QUOTA
);
1260 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1263 set_opt(sbi
->s_mount_opt
, QUOTA
);
1264 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1267 if (sb_any_quota_enabled(sb
)) {
1268 printk(KERN_ERR
"EXT4-fs: Cannot change quota "
1269 "options when quota turned on.\n");
1272 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1273 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1274 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1281 "EXT4-fs: quota options not supported.\n");
1285 case Opt_offusrjquota
:
1286 case Opt_offgrpjquota
:
1287 case Opt_jqfmt_vfsold
:
1288 case Opt_jqfmt_vfsv0
:
1290 "EXT4-fs: journaled quota options not "
1297 set_opt(sbi
->s_mount_opt
, ABORT
);
1300 if (match_int(&args
[0], &option
))
1303 set_opt(sbi
->s_mount_opt
, BARRIER
);
1305 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1311 printk("EXT4-fs: resize option only available "
1315 if (match_int(&args
[0], &option
) != 0)
1317 *n_blocks_count
= option
;
1320 set_opt(sbi
->s_mount_opt
, NOBH
);
1323 clear_opt(sbi
->s_mount_opt
, NOBH
);
1326 set_opt (sbi
->s_mount_opt
, EXTENTS
);
1329 clear_opt (sbi
->s_mount_opt
, EXTENTS
);
1332 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1333 sb
->s_flags
|= MS_I_VERSION
;
1335 case Opt_nodelalloc
:
1336 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1339 set_opt(sbi
->s_mount_opt
, MBALLOC
);
1342 clear_opt(sbi
->s_mount_opt
, MBALLOC
);
1345 if (match_int(&args
[0], &option
))
1349 sbi
->s_stripe
= option
;
1352 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1356 "EXT4-fs: Unrecognized mount option \"%s\" "
1357 "or missing value\n", p
);
1362 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1363 if ((sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
) &&
1364 sbi
->s_qf_names
[USRQUOTA
])
1365 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1367 if ((sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
) &&
1368 sbi
->s_qf_names
[GRPQUOTA
])
1369 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1371 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1372 (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)) ||
1373 (sbi
->s_qf_names
[GRPQUOTA
] &&
1374 (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
))) {
1375 printk(KERN_ERR
"EXT4-fs: old and new quota "
1376 "format mixing.\n");
1380 if (!sbi
->s_jquota_fmt
) {
1381 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1382 "not specified.\n");
1386 if (sbi
->s_jquota_fmt
) {
1387 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1388 "specified with no journaling "
1397 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1400 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1403 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1404 printk (KERN_ERR
"EXT4-fs warning: revision level too high, "
1405 "forcing read-only mode\n");
1410 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1411 printk (KERN_WARNING
"EXT4-fs warning: mounting unchecked fs, "
1412 "running e2fsck is recommended\n");
1413 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1414 printk (KERN_WARNING
1415 "EXT4-fs warning: mounting fs with errors, "
1416 "running e2fsck is recommended\n");
1417 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1418 le16_to_cpu(es
->s_mnt_count
) >=
1419 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1420 printk (KERN_WARNING
1421 "EXT4-fs warning: maximal mount count reached, "
1422 "running e2fsck is recommended\n");
1423 else if (le32_to_cpu(es
->s_checkinterval
) &&
1424 (le32_to_cpu(es
->s_lastcheck
) +
1425 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1426 printk (KERN_WARNING
1427 "EXT4-fs warning: checktime reached, "
1428 "running e2fsck is recommended\n");
1430 /* @@@ We _will_ want to clear the valid bit if we find
1431 * inconsistencies, to force a fsck at reboot. But for
1432 * a plain journaled filesystem we can keep it set as
1435 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1437 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1438 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1439 le16_add_cpu(&es
->s_mnt_count
, 1);
1440 es
->s_mtime
= cpu_to_le32(get_seconds());
1441 ext4_update_dynamic_rev(sb
);
1442 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1444 ext4_commit_super(sb
, es
, 1);
1445 if (test_opt(sb
, DEBUG
))
1446 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%lu, "
1447 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1449 sbi
->s_groups_count
,
1450 EXT4_BLOCKS_PER_GROUP(sb
),
1451 EXT4_INODES_PER_GROUP(sb
),
1454 printk(KERN_INFO
"EXT4 FS on %s, ", sb
->s_id
);
1455 if (EXT4_SB(sb
)->s_journal
->j_inode
== NULL
) {
1456 char b
[BDEVNAME_SIZE
];
1458 printk("external journal on %s\n",
1459 bdevname(EXT4_SB(sb
)->s_journal
->j_dev
, b
));
1461 printk("internal journal\n");
1466 static int ext4_fill_flex_info(struct super_block
*sb
)
1468 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1469 struct ext4_group_desc
*gdp
= NULL
;
1470 struct buffer_head
*bh
;
1471 ext4_group_t flex_group_count
;
1472 ext4_group_t flex_group
;
1473 int groups_per_flex
= 0;
1474 __u64 block_bitmap
= 0;
1477 if (!sbi
->s_es
->s_log_groups_per_flex
) {
1478 sbi
->s_log_groups_per_flex
= 0;
1482 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1483 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1485 flex_group_count
= (sbi
->s_groups_count
+ groups_per_flex
- 1) /
1487 sbi
->s_flex_groups
= kmalloc(flex_group_count
*
1488 sizeof(struct flex_groups
), GFP_KERNEL
);
1489 if (sbi
->s_flex_groups
== NULL
) {
1490 printk(KERN_ERR
"EXT4-fs: not enough memory\n");
1493 memset(sbi
->s_flex_groups
, 0, flex_group_count
*
1494 sizeof(struct flex_groups
));
1496 gdp
= ext4_get_group_desc(sb
, 1, &bh
);
1497 block_bitmap
= ext4_block_bitmap(sb
, gdp
) - 1;
1499 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1500 gdp
= ext4_get_group_desc(sb
, i
, &bh
);
1502 flex_group
= ext4_flex_group(sbi
, i
);
1503 sbi
->s_flex_groups
[flex_group
].free_inodes
+=
1504 le16_to_cpu(gdp
->bg_free_inodes_count
);
1505 sbi
->s_flex_groups
[flex_group
].free_blocks
+=
1506 le16_to_cpu(gdp
->bg_free_blocks_count
);
1514 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1515 struct ext4_group_desc
*gdp
)
1519 if (sbi
->s_es
->s_feature_ro_compat
&
1520 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1521 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1522 __le32 le_group
= cpu_to_le32(block_group
);
1524 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1525 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1526 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1527 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1528 /* for checksum of struct ext4_group_desc do the rest...*/
1529 if ((sbi
->s_es
->s_feature_incompat
&
1530 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1531 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1532 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1533 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1537 return cpu_to_le16(crc
);
1540 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1541 struct ext4_group_desc
*gdp
)
1543 if ((sbi
->s_es
->s_feature_ro_compat
&
1544 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1545 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1551 /* Called at mount-time, super-block is locked */
1552 static int ext4_check_descriptors(struct super_block
*sb
)
1554 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1555 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1556 ext4_fsblk_t last_block
;
1557 ext4_fsblk_t block_bitmap
;
1558 ext4_fsblk_t inode_bitmap
;
1559 ext4_fsblk_t inode_table
;
1560 int flexbg_flag
= 0;
1563 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1566 ext4_debug ("Checking group descriptors");
1568 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1569 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1571 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1572 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1574 last_block
= first_block
+
1575 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1577 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1578 if (block_bitmap
< first_block
|| block_bitmap
> last_block
)
1580 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1581 "Block bitmap for group %lu not in group "
1582 "(block %llu)!", i
, block_bitmap
);
1585 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1586 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
)
1588 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1589 "Inode bitmap for group %lu not in group "
1590 "(block %llu)!", i
, inode_bitmap
);
1593 inode_table
= ext4_inode_table(sb
, gdp
);
1594 if (inode_table
< first_block
||
1595 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
)
1597 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1598 "Inode table for group %lu not in group "
1599 "(block %llu)!", i
, inode_table
);
1602 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1603 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1604 "Checksum for group %lu failed (%u!=%u)\n",
1605 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1606 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1610 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1613 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1614 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
1618 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1619 * the superblock) which were deleted from all directories, but held open by
1620 * a process at the time of a crash. We walk the list and try to delete these
1621 * inodes at recovery time (only with a read-write filesystem).
1623 * In order to keep the orphan inode chain consistent during traversal (in
1624 * case of crash during recovery), we link each inode into the superblock
1625 * orphan list_head and handle it the same way as an inode deletion during
1626 * normal operation (which journals the operations for us).
1628 * We only do an iget() and an iput() on each inode, which is very safe if we
1629 * accidentally point at an in-use or already deleted inode. The worst that
1630 * can happen in this case is that we get a "bit already cleared" message from
1631 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1632 * e2fsck was run on this filesystem, and it must have already done the orphan
1633 * inode cleanup for us, so we can safely abort without any further action.
1635 static void ext4_orphan_cleanup (struct super_block
* sb
,
1636 struct ext4_super_block
* es
)
1638 unsigned int s_flags
= sb
->s_flags
;
1639 int nr_orphans
= 0, nr_truncates
= 0;
1643 if (!es
->s_last_orphan
) {
1644 jbd_debug(4, "no orphan inodes to clean up\n");
1648 if (bdev_read_only(sb
->s_bdev
)) {
1649 printk(KERN_ERR
"EXT4-fs: write access "
1650 "unavailable, skipping orphan cleanup.\n");
1654 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1655 if (es
->s_last_orphan
)
1656 jbd_debug(1, "Errors on filesystem, "
1657 "clearing orphan list.\n");
1658 es
->s_last_orphan
= 0;
1659 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1663 if (s_flags
& MS_RDONLY
) {
1664 printk(KERN_INFO
"EXT4-fs: %s: orphan cleanup on readonly fs\n",
1666 sb
->s_flags
&= ~MS_RDONLY
;
1669 /* Needed for iput() to work correctly and not trash data */
1670 sb
->s_flags
|= MS_ACTIVE
;
1671 /* Turn on quotas so that they are updated correctly */
1672 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1673 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
1674 int ret
= ext4_quota_on_mount(sb
, i
);
1677 "EXT4-fs: Cannot turn on journaled "
1678 "quota: error %d\n", ret
);
1683 while (es
->s_last_orphan
) {
1684 struct inode
*inode
;
1686 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1687 if (IS_ERR(inode
)) {
1688 es
->s_last_orphan
= 0;
1692 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
1694 if (inode
->i_nlink
) {
1696 "%s: truncating inode %lu to %Ld bytes\n",
1697 __func__
, inode
->i_ino
, inode
->i_size
);
1698 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1699 inode
->i_ino
, inode
->i_size
);
1700 ext4_truncate(inode
);
1704 "%s: deleting unreferenced inode %lu\n",
1705 __func__
, inode
->i_ino
);
1706 jbd_debug(2, "deleting unreferenced inode %lu\n",
1710 iput(inode
); /* The delete magic happens here! */
1713 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1716 printk(KERN_INFO
"EXT4-fs: %s: %d orphan inode%s deleted\n",
1717 sb
->s_id
, PLURAL(nr_orphans
));
1719 printk(KERN_INFO
"EXT4-fs: %s: %d truncate%s cleaned up\n",
1720 sb
->s_id
, PLURAL(nr_truncates
));
1722 /* Turn quotas off */
1723 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1724 if (sb_dqopt(sb
)->files
[i
])
1725 vfs_quota_off(sb
, i
, 0);
1728 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1731 * Maximal extent format file size.
1732 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1733 * extent format containers, within a sector_t, and within i_blocks
1734 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1735 * so that won't be a limiting factor.
1737 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1739 static loff_t
ext4_max_size(int blkbits
)
1742 loff_t upper_limit
= MAX_LFS_FILESIZE
;
1744 /* small i_blocks in vfs inode? */
1745 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
1747 * CONFIG_LSF is not enabled implies the inode
1748 * i_block represent total blocks in 512 bytes
1749 * 32 == size of vfs inode i_blocks * 8
1751 upper_limit
= (1LL << 32) - 1;
1753 /* total blocks in file system block size */
1754 upper_limit
>>= (blkbits
- 9);
1755 upper_limit
<<= blkbits
;
1758 /* 32-bit extent-start container, ee_block */
1763 /* Sanity check against vm- & vfs- imposed limits */
1764 if (res
> upper_limit
)
1771 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1772 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1773 * We need to be 1 filesystem block less than the 2^48 sector limit.
1775 static loff_t
ext4_max_bitmap_size(int bits
)
1777 loff_t res
= EXT4_NDIR_BLOCKS
;
1780 /* This is calculated to be the largest file size for a
1781 * dense, bitmapped file such that the total number of
1782 * sectors in the file, including data and all indirect blocks,
1783 * does not exceed 2^48 -1
1784 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1785 * total number of 512 bytes blocks of the file
1788 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
1790 * CONFIG_LSF is not enabled implies the inode
1791 * i_block represent total blocks in 512 bytes
1792 * 32 == size of vfs inode i_blocks * 8
1794 upper_limit
= (1LL << 32) - 1;
1796 /* total blocks in file system block size */
1797 upper_limit
>>= (bits
- 9);
1801 * We use 48 bit ext4_inode i_blocks
1802 * With EXT4_HUGE_FILE_FL set the i_blocks
1803 * represent total number of blocks in
1804 * file system block size
1806 upper_limit
= (1LL << 48) - 1;
1810 /* indirect blocks */
1812 /* double indirect blocks */
1813 meta_blocks
+= 1 + (1LL << (bits
-2));
1814 /* tripple indirect blocks */
1815 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
1817 upper_limit
-= meta_blocks
;
1818 upper_limit
<<= bits
;
1820 res
+= 1LL << (bits
-2);
1821 res
+= 1LL << (2*(bits
-2));
1822 res
+= 1LL << (3*(bits
-2));
1824 if (res
> upper_limit
)
1827 if (res
> MAX_LFS_FILESIZE
)
1828 res
= MAX_LFS_FILESIZE
;
1833 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
1834 ext4_fsblk_t logical_sb_block
, int nr
)
1836 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1837 ext4_group_t bg
, first_meta_bg
;
1840 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1842 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
1844 return logical_sb_block
+ nr
+ 1;
1845 bg
= sbi
->s_desc_per_block
* nr
;
1846 if (ext4_bg_has_super(sb
, bg
))
1848 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
1852 * ext4_get_stripe_size: Get the stripe size.
1853 * @sbi: In memory super block info
1855 * If we have specified it via mount option, then
1856 * use the mount option value. If the value specified at mount time is
1857 * greater than the blocks per group use the super block value.
1858 * If the super block value is greater than blocks per group return 0.
1859 * Allocator needs it be less than blocks per group.
1862 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
1864 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
1865 unsigned long stripe_width
=
1866 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
1868 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
1869 return sbi
->s_stripe
;
1871 if (stripe_width
<= sbi
->s_blocks_per_group
)
1872 return stripe_width
;
1874 if (stride
<= sbi
->s_blocks_per_group
)
1880 static int ext4_fill_super (struct super_block
*sb
, void *data
, int silent
)
1881 __releases(kernel_lock
)
1882 __acquires(kernel_lock
)
1885 struct buffer_head
* bh
;
1886 struct ext4_super_block
*es
= NULL
;
1887 struct ext4_sb_info
*sbi
;
1889 ext4_fsblk_t sb_block
= get_sb_block(&data
);
1890 ext4_fsblk_t logical_sb_block
;
1891 unsigned long offset
= 0;
1892 unsigned int journal_inum
= 0;
1893 unsigned long journal_devnum
= 0;
1894 unsigned long def_mount_opts
;
1905 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
1908 sb
->s_fs_info
= sbi
;
1909 sbi
->s_mount_opt
= 0;
1910 sbi
->s_resuid
= EXT4_DEF_RESUID
;
1911 sbi
->s_resgid
= EXT4_DEF_RESGID
;
1912 sbi
->s_sb_block
= sb_block
;
1916 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
1918 printk(KERN_ERR
"EXT4-fs: unable to set blocksize\n");
1923 * The ext4 superblock will not be buffer aligned for other than 1kB
1924 * block sizes. We need to calculate the offset from buffer start.
1926 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
1927 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
1928 offset
= do_div(logical_sb_block
, blocksize
);
1930 logical_sb_block
= sb_block
;
1933 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
1934 printk (KERN_ERR
"EXT4-fs: unable to read superblock\n");
1938 * Note: s_es must be initialized as soon as possible because
1939 * some ext4 macro-instructions depend on its value
1941 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
1943 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1944 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
1947 /* Set defaults before we parse the mount options */
1948 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1949 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
1950 set_opt(sbi
->s_mount_opt
, DEBUG
);
1951 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
1952 set_opt(sbi
->s_mount_opt
, GRPID
);
1953 if (def_mount_opts
& EXT4_DEFM_UID16
)
1954 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1955 #ifdef CONFIG_EXT4DEV_FS_XATTR
1956 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
1957 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1959 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1960 if (def_mount_opts
& EXT4_DEFM_ACL
)
1961 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1963 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
1964 sbi
->s_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
1965 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
1966 sbi
->s_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
1967 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
1968 sbi
->s_mount_opt
|= EXT4_MOUNT_WRITEBACK_DATA
;
1970 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
1971 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1972 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
1973 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1975 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1977 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
1978 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
1980 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1981 set_opt(sbi
->s_mount_opt
, BARRIER
);
1984 * turn on extents feature by default in ext4 filesystem
1985 * User -o noextents to turn it off
1987 set_opt(sbi
->s_mount_opt
, EXTENTS
);
1989 * turn on mballoc feature by default in ext4 filesystem
1990 * User -o nomballoc to turn it off
1992 set_opt(sbi
->s_mount_opt
, MBALLOC
);
1995 * enable delayed allocation by default
1996 * Use -o nodelalloc to turn it off
1998 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2001 if (!parse_options ((char *) data
, sb
, &journal_inum
, &journal_devnum
,
2005 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2006 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2008 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2009 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2010 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2011 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2013 "EXT4-fs warning: feature flags set on rev 0 fs, "
2014 "running e2fsck is recommended\n");
2017 * Since ext4 is still considered development code, we require
2018 * that the TEST_FILESYS flag in s->flags be set.
2020 if (!(le32_to_cpu(es
->s_flags
) & EXT2_FLAGS_TEST_FILESYS
)) {
2021 printk(KERN_WARNING
"EXT4-fs: %s: not marked "
2022 "OK to use with test code.\n", sb
->s_id
);
2027 * Check feature flags regardless of the revision level, since we
2028 * previously didn't change the revision level when setting the flags,
2029 * so there is a chance incompat flags are set on a rev 0 filesystem.
2031 features
= EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
);
2033 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount because of "
2034 "unsupported optional features (%x).\n",
2035 sb
->s_id
, le32_to_cpu(features
));
2038 features
= EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
);
2039 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
2040 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount RDWR because of "
2041 "unsupported optional features (%x).\n",
2042 sb
->s_id
, le32_to_cpu(features
));
2045 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2047 * Large file size enabled file system can only be
2048 * mount if kernel is build with CONFIG_LSF
2050 if (sizeof(root
->i_blocks
) < sizeof(u64
) &&
2051 !(sb
->s_flags
& MS_RDONLY
)) {
2052 printk(KERN_ERR
"EXT4-fs: %s: Filesystem with huge "
2053 "files cannot be mounted read-write "
2054 "without CONFIG_LSF.\n", sb
->s_id
);
2058 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2060 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2061 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2063 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2064 blocksize
, sb
->s_id
);
2068 if (sb
->s_blocksize
!= blocksize
) {
2070 /* Validate the filesystem blocksize */
2071 if (!sb_set_blocksize(sb
, blocksize
)) {
2072 printk(KERN_ERR
"EXT4-fs: bad block size %d.\n",
2078 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2079 offset
= do_div(logical_sb_block
, blocksize
);
2080 bh
= sb_bread(sb
, logical_sb_block
);
2083 "EXT4-fs: Can't read superblock on 2nd try.\n");
2086 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2088 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2090 "EXT4-fs: Magic mismatch, very weird !\n");
2095 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
);
2096 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
);
2098 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2099 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2100 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2102 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2103 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2104 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2105 (!is_power_of_2(sbi
->s_inode_size
)) ||
2106 (sbi
->s_inode_size
> blocksize
)) {
2108 "EXT4-fs: unsupported inode size: %d\n",
2112 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2113 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2115 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2116 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2117 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2118 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2119 !is_power_of_2(sbi
->s_desc_size
)) {
2121 "EXT4-fs: unsupported descriptor size %lu\n",
2126 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2127 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2128 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2129 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2131 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2132 if (sbi
->s_inodes_per_block
== 0)
2134 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2135 sbi
->s_inodes_per_block
;
2136 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2138 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2139 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2140 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2141 for (i
=0; i
< 4; i
++)
2142 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2143 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2145 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2147 "EXT4-fs: #blocks per group too big: %lu\n",
2148 sbi
->s_blocks_per_group
);
2151 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2153 "EXT4-fs: #inodes per group too big: %lu\n",
2154 sbi
->s_inodes_per_group
);
2158 if (ext4_blocks_count(es
) >
2159 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) {
2160 printk(KERN_ERR
"EXT4-fs: filesystem on %s:"
2161 " too large to mount safely\n", sb
->s_id
);
2162 if (sizeof(sector_t
) < 8)
2163 printk(KERN_WARNING
"EXT4-fs: CONFIG_LBD not "
2168 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2171 /* ensure blocks_count calculation below doesn't sign-extend */
2172 if (ext4_blocks_count(es
) + EXT4_BLOCKS_PER_GROUP(sb
) <
2173 le32_to_cpu(es
->s_first_data_block
) + 1) {
2174 printk(KERN_WARNING
"EXT4-fs: bad geometry: block count %llu, "
2175 "first data block %u, blocks per group %lu\n",
2176 ext4_blocks_count(es
),
2177 le32_to_cpu(es
->s_first_data_block
),
2178 EXT4_BLOCKS_PER_GROUP(sb
));
2181 blocks_count
= (ext4_blocks_count(es
) -
2182 le32_to_cpu(es
->s_first_data_block
) +
2183 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2184 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2185 sbi
->s_groups_count
= blocks_count
;
2186 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2187 EXT4_DESC_PER_BLOCK(sb
);
2188 sbi
->s_group_desc
= kmalloc(db_count
* sizeof (struct buffer_head
*),
2190 if (sbi
->s_group_desc
== NULL
) {
2191 printk (KERN_ERR
"EXT4-fs: not enough memory\n");
2195 bgl_lock_init(&sbi
->s_blockgroup_lock
);
2197 for (i
= 0; i
< db_count
; i
++) {
2198 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2199 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2200 if (!sbi
->s_group_desc
[i
]) {
2201 printk (KERN_ERR
"EXT4-fs: "
2202 "can't read group descriptor %d\n", i
);
2207 if (!ext4_check_descriptors (sb
)) {
2208 printk(KERN_ERR
"EXT4-fs: group descriptors corrupted!\n");
2211 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2212 if (!ext4_fill_flex_info(sb
)) {
2214 "EXT4-fs: unable to initialize "
2215 "flex_bg meta info!\n");
2219 sbi
->s_gdb_count
= db_count
;
2220 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2221 spin_lock_init(&sbi
->s_next_gen_lock
);
2223 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2224 ext4_count_free_blocks(sb
));
2226 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2227 ext4_count_free_inodes(sb
));
2230 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2231 ext4_count_dirs(sb
));
2234 printk(KERN_ERR
"EXT4-fs: insufficient memory\n");
2238 /* per fileystem reservation list head & lock */
2239 spin_lock_init(&sbi
->s_rsv_window_lock
);
2240 sbi
->s_rsv_window_root
= RB_ROOT
;
2241 /* Add a single, static dummy reservation to the start of the
2242 * reservation window list --- it gives us a placeholder for
2243 * append-at-start-of-list which makes the allocation logic
2244 * _much_ simpler. */
2245 sbi
->s_rsv_window_head
.rsv_start
= EXT4_RESERVE_WINDOW_NOT_ALLOCATED
;
2246 sbi
->s_rsv_window_head
.rsv_end
= EXT4_RESERVE_WINDOW_NOT_ALLOCATED
;
2247 sbi
->s_rsv_window_head
.rsv_alloc_hit
= 0;
2248 sbi
->s_rsv_window_head
.rsv_goal_size
= 0;
2249 ext4_rsv_window_add(sb
, &sbi
->s_rsv_window_head
);
2251 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2254 * set up enough so that it can read an inode
2256 sb
->s_op
= &ext4_sops
;
2257 sb
->s_export_op
= &ext4_export_ops
;
2258 sb
->s_xattr
= ext4_xattr_handlers
;
2260 sb
->s_qcop
= &ext4_qctl_operations
;
2261 sb
->dq_op
= &ext4_quota_operations
;
2263 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2267 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2268 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2269 EXT4_FEATURE_INCOMPAT_RECOVER
));
2272 * The first inode we look at is the journal inode. Don't try
2273 * root first: it may be modified in the journal!
2275 if (!test_opt(sb
, NOLOAD
) &&
2276 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2277 if (ext4_load_journal(sb
, es
, journal_devnum
))
2279 if (!(sb
->s_flags
& MS_RDONLY
) &&
2280 EXT4_SB(sb
)->s_journal
->j_failed_commit
) {
2281 printk(KERN_CRIT
"EXT4-fs error (device %s): "
2282 "ext4_fill_super: Journal transaction "
2283 "%u is corrupt\n", sb
->s_id
,
2284 EXT4_SB(sb
)->s_journal
->j_failed_commit
);
2285 if (test_opt (sb
, ERRORS_RO
)) {
2287 "Mounting filesystem read-only\n");
2288 sb
->s_flags
|= MS_RDONLY
;
2289 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2290 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2292 if (test_opt(sb
, ERRORS_PANIC
)) {
2293 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2294 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2295 ext4_commit_super(sb
, es
, 1);
2297 "EXT4-fs (device %s): mount failed\n",
2302 } else if (journal_inum
) {
2303 if (ext4_create_journal(sb
, es
, journal_inum
))
2308 "ext4: No journal on filesystem on %s\n",
2313 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2314 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2315 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2316 printk(KERN_ERR
"ext4: Failed to set 64-bit journal feature\n");
2320 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2321 jbd2_journal_set_features(sbi
->s_journal
,
2322 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2323 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2324 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2325 jbd2_journal_set_features(sbi
->s_journal
,
2326 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2327 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2328 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2330 jbd2_journal_clear_features(sbi
->s_journal
,
2331 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2332 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2335 /* We have now updated the journal if required, so we can
2336 * validate the data journaling mode. */
2337 switch (test_opt(sb
, DATA_FLAGS
)) {
2339 /* No mode set, assume a default based on the journal
2340 * capabilities: ORDERED_DATA if the journal can
2341 * cope, else JOURNAL_DATA
2343 if (jbd2_journal_check_available_features
2344 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2345 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2347 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2350 case EXT4_MOUNT_ORDERED_DATA
:
2351 case EXT4_MOUNT_WRITEBACK_DATA
:
2352 if (!jbd2_journal_check_available_features
2353 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2354 printk(KERN_ERR
"EXT4-fs: Journal does not support "
2355 "requested data journaling mode\n");
2362 if (test_opt(sb
, NOBH
)) {
2363 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2364 printk(KERN_WARNING
"EXT4-fs: Ignoring nobh option - "
2365 "its supported only with writeback mode\n");
2366 clear_opt(sbi
->s_mount_opt
, NOBH
);
2370 * The jbd2_journal_load will have done any necessary log recovery,
2371 * so we can safely mount the rest of the filesystem now.
2374 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2376 printk(KERN_ERR
"EXT4-fs: get root inode failed\n");
2377 ret
= PTR_ERR(root
);
2380 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2382 printk(KERN_ERR
"EXT4-fs: corrupt root inode, run e2fsck\n");
2385 sb
->s_root
= d_alloc_root(root
);
2387 printk(KERN_ERR
"EXT4-fs: get root dentry failed\n");
2393 ext4_setup_super (sb
, es
, sb
->s_flags
& MS_RDONLY
);
2395 /* determine the minimum size of new large inodes, if present */
2396 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2397 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2398 EXT4_GOOD_OLD_INODE_SIZE
;
2399 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2400 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2401 if (sbi
->s_want_extra_isize
<
2402 le16_to_cpu(es
->s_want_extra_isize
))
2403 sbi
->s_want_extra_isize
=
2404 le16_to_cpu(es
->s_want_extra_isize
);
2405 if (sbi
->s_want_extra_isize
<
2406 le16_to_cpu(es
->s_min_extra_isize
))
2407 sbi
->s_want_extra_isize
=
2408 le16_to_cpu(es
->s_min_extra_isize
);
2411 /* Check if enough inode space is available */
2412 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2413 sbi
->s_inode_size
) {
2414 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2415 EXT4_GOOD_OLD_INODE_SIZE
;
2416 printk(KERN_INFO
"EXT4-fs: required extra inode space not"
2421 * akpm: core read_super() calls in here with the superblock locked.
2422 * That deadlocks, because orphan cleanup needs to lock the superblock
2423 * in numerous places. Here we just pop the lock - it's relatively
2424 * harmless, because we are now ready to accept write_super() requests,
2425 * and aviro says that's the only reason for hanging onto the
2428 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
2429 ext4_orphan_cleanup(sb
, es
);
2430 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
2432 printk (KERN_INFO
"EXT4-fs: recovery complete.\n");
2433 ext4_mark_recovery_complete(sb
, es
);
2434 printk (KERN_INFO
"EXT4-fs: mounted filesystem with %s data mode.\n",
2435 test_opt(sb
,DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
? "journal":
2436 test_opt(sb
,DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
? "ordered":
2439 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
2440 printk(KERN_WARNING
"EXT4-fs: Ignoring delalloc option - "
2441 "requested data journaling mode\n");
2442 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2443 } else if (test_opt(sb
, DELALLOC
))
2444 printk(KERN_INFO
"EXT4-fs: delayed allocation enabled\n");
2447 ext4_mb_init(sb
, needs_recovery
);
2454 printk(KERN_ERR
"VFS: Can't find ext4 filesystem on dev %s.\n",
2459 jbd2_journal_destroy(sbi
->s_journal
);
2460 sbi
->s_journal
= NULL
;
2462 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2463 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2464 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2466 for (i
= 0; i
< db_count
; i
++)
2467 brelse(sbi
->s_group_desc
[i
]);
2468 kfree(sbi
->s_group_desc
);
2471 for (i
= 0; i
< MAXQUOTAS
; i
++)
2472 kfree(sbi
->s_qf_names
[i
]);
2474 ext4_blkdev_remove(sbi
);
2477 sb
->s_fs_info
= NULL
;
2484 * Setup any per-fs journal parameters now. We'll do this both on
2485 * initial mount, once the journal has been initialised but before we've
2486 * done any recovery; and again on any subsequent remount.
2488 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
2490 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2492 if (sbi
->s_commit_interval
)
2493 journal
->j_commit_interval
= sbi
->s_commit_interval
;
2494 /* We could also set up an ext4-specific default for the commit
2495 * interval here, but for now we'll just fall back to the jbd
2498 spin_lock(&journal
->j_state_lock
);
2499 if (test_opt(sb
, BARRIER
))
2500 journal
->j_flags
|= JBD2_BARRIER
;
2502 journal
->j_flags
&= ~JBD2_BARRIER
;
2503 spin_unlock(&journal
->j_state_lock
);
2506 static journal_t
*ext4_get_journal(struct super_block
*sb
,
2507 unsigned int journal_inum
)
2509 struct inode
*journal_inode
;
2512 /* First, test for the existence of a valid inode on disk. Bad
2513 * things happen if we iget() an unused inode, as the subsequent
2514 * iput() will try to delete it. */
2516 journal_inode
= ext4_iget(sb
, journal_inum
);
2517 if (IS_ERR(journal_inode
)) {
2518 printk(KERN_ERR
"EXT4-fs: no journal found.\n");
2521 if (!journal_inode
->i_nlink
) {
2522 make_bad_inode(journal_inode
);
2523 iput(journal_inode
);
2524 printk(KERN_ERR
"EXT4-fs: journal inode is deleted.\n");
2528 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2529 journal_inode
, journal_inode
->i_size
);
2530 if (!S_ISREG(journal_inode
->i_mode
)) {
2531 printk(KERN_ERR
"EXT4-fs: invalid journal inode.\n");
2532 iput(journal_inode
);
2536 journal
= jbd2_journal_init_inode(journal_inode
);
2538 printk(KERN_ERR
"EXT4-fs: Could not load journal inode\n");
2539 iput(journal_inode
);
2542 journal
->j_private
= sb
;
2543 ext4_init_journal_params(sb
, journal
);
2547 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
2550 struct buffer_head
* bh
;
2554 int hblock
, blocksize
;
2555 ext4_fsblk_t sb_block
;
2556 unsigned long offset
;
2557 struct ext4_super_block
* es
;
2558 struct block_device
*bdev
;
2560 bdev
= ext4_blkdev_get(j_dev
);
2564 if (bd_claim(bdev
, sb
)) {
2566 "EXT4: failed to claim external journal device.\n");
2571 blocksize
= sb
->s_blocksize
;
2572 hblock
= bdev_hardsect_size(bdev
);
2573 if (blocksize
< hblock
) {
2575 "EXT4-fs: blocksize too small for journal device.\n");
2579 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
2580 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
2581 set_blocksize(bdev
, blocksize
);
2582 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
2583 printk(KERN_ERR
"EXT4-fs: couldn't read superblock of "
2584 "external journal\n");
2588 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2589 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
2590 !(le32_to_cpu(es
->s_feature_incompat
) &
2591 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
2592 printk(KERN_ERR
"EXT4-fs: external journal has "
2593 "bad superblock\n");
2598 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
2599 printk(KERN_ERR
"EXT4-fs: journal UUID does not match\n");
2604 len
= ext4_blocks_count(es
);
2605 start
= sb_block
+ 1;
2606 brelse(bh
); /* we're done with the superblock */
2608 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
2609 start
, len
, blocksize
);
2611 printk(KERN_ERR
"EXT4-fs: failed to create device journal\n");
2614 journal
->j_private
= sb
;
2615 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
2616 wait_on_buffer(journal
->j_sb_buffer
);
2617 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
2618 printk(KERN_ERR
"EXT4-fs: I/O error on journal device\n");
2621 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
2622 printk(KERN_ERR
"EXT4-fs: External journal has more than one "
2623 "user (unsupported) - %d\n",
2624 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
2627 EXT4_SB(sb
)->journal_bdev
= bdev
;
2628 ext4_init_journal_params(sb
, journal
);
2631 jbd2_journal_destroy(journal
);
2633 ext4_blkdev_put(bdev
);
2637 static int ext4_load_journal(struct super_block
*sb
,
2638 struct ext4_super_block
*es
,
2639 unsigned long journal_devnum
)
2642 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
2645 int really_read_only
;
2647 if (journal_devnum
&&
2648 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2649 printk(KERN_INFO
"EXT4-fs: external journal device major/minor "
2650 "numbers have changed\n");
2651 journal_dev
= new_decode_dev(journal_devnum
);
2653 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
2655 really_read_only
= bdev_read_only(sb
->s_bdev
);
2658 * Are we loading a blank journal or performing recovery after a
2659 * crash? For recovery, we need to check in advance whether we
2660 * can get read-write access to the device.
2663 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2664 if (sb
->s_flags
& MS_RDONLY
) {
2665 printk(KERN_INFO
"EXT4-fs: INFO: recovery "
2666 "required on readonly filesystem.\n");
2667 if (really_read_only
) {
2668 printk(KERN_ERR
"EXT4-fs: write access "
2669 "unavailable, cannot proceed.\n");
2672 printk (KERN_INFO
"EXT4-fs: write access will "
2673 "be enabled during recovery.\n");
2677 if (journal_inum
&& journal_dev
) {
2678 printk(KERN_ERR
"EXT4-fs: filesystem has both journal "
2679 "and inode journals!\n");
2684 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
2687 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
2691 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
2692 err
= jbd2_journal_update_format(journal
);
2694 printk(KERN_ERR
"EXT4-fs: error updating journal.\n");
2695 jbd2_journal_destroy(journal
);
2700 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
2701 err
= jbd2_journal_wipe(journal
, !really_read_only
);
2703 err
= jbd2_journal_load(journal
);
2706 printk(KERN_ERR
"EXT4-fs: error loading journal.\n");
2707 jbd2_journal_destroy(journal
);
2711 EXT4_SB(sb
)->s_journal
= journal
;
2712 ext4_clear_journal_err(sb
, es
);
2714 if (journal_devnum
&&
2715 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2716 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2719 /* Make sure we flush the recovery flag to disk. */
2720 ext4_commit_super(sb
, es
, 1);
2726 static int ext4_create_journal(struct super_block
* sb
,
2727 struct ext4_super_block
* es
,
2728 unsigned int journal_inum
)
2733 if (sb
->s_flags
& MS_RDONLY
) {
2734 printk(KERN_ERR
"EXT4-fs: readonly filesystem when trying to "
2735 "create journal.\n");
2739 journal
= ext4_get_journal(sb
, journal_inum
);
2743 printk(KERN_INFO
"EXT4-fs: creating new journal on inode %u\n",
2746 err
= jbd2_journal_create(journal
);
2748 printk(KERN_ERR
"EXT4-fs: error creating journal.\n");
2749 jbd2_journal_destroy(journal
);
2753 EXT4_SB(sb
)->s_journal
= journal
;
2755 ext4_update_dynamic_rev(sb
);
2756 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2757 EXT4_SET_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
);
2759 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
2762 /* Make sure we flush the recovery flag to disk. */
2763 ext4_commit_super(sb
, es
, 1);
2768 static void ext4_commit_super (struct super_block
* sb
,
2769 struct ext4_super_block
* es
,
2772 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
2776 es
->s_wtime
= cpu_to_le32(get_seconds());
2777 ext4_free_blocks_count_set(es
, ext4_count_free_blocks(sb
));
2778 es
->s_free_inodes_count
= cpu_to_le32(ext4_count_free_inodes(sb
));
2779 BUFFER_TRACE(sbh
, "marking dirty");
2780 mark_buffer_dirty(sbh
);
2782 sync_dirty_buffer(sbh
);
2787 * Have we just finished recovery? If so, and if we are mounting (or
2788 * remounting) the filesystem readonly, then we will end up with a
2789 * consistent fs on disk. Record that fact.
2791 static void ext4_mark_recovery_complete(struct super_block
* sb
,
2792 struct ext4_super_block
* es
)
2794 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
2796 jbd2_journal_lock_updates(journal
);
2797 jbd2_journal_flush(journal
);
2799 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
2800 sb
->s_flags
& MS_RDONLY
) {
2801 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2803 ext4_commit_super(sb
, es
, 1);
2806 jbd2_journal_unlock_updates(journal
);
2810 * If we are mounting (or read-write remounting) a filesystem whose journal
2811 * has recorded an error from a previous lifetime, move that error to the
2812 * main filesystem now.
2814 static void ext4_clear_journal_err(struct super_block
* sb
,
2815 struct ext4_super_block
* es
)
2821 journal
= EXT4_SB(sb
)->s_journal
;
2824 * Now check for any error status which may have been recorded in the
2825 * journal by a prior ext4_error() or ext4_abort()
2828 j_errno
= jbd2_journal_errno(journal
);
2832 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
2833 ext4_warning(sb
, __func__
, "Filesystem error recorded "
2834 "from previous mount: %s", errstr
);
2835 ext4_warning(sb
, __func__
, "Marking fs in need of "
2836 "filesystem check.");
2838 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2839 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2840 ext4_commit_super (sb
, es
, 1);
2842 jbd2_journal_clear_err(journal
);
2847 * Force the running and committing transactions to commit,
2848 * and wait on the commit.
2850 int ext4_force_commit(struct super_block
*sb
)
2855 if (sb
->s_flags
& MS_RDONLY
)
2858 journal
= EXT4_SB(sb
)->s_journal
;
2860 ret
= ext4_journal_force_commit(journal
);
2865 * Ext4 always journals updates to the superblock itself, so we don't
2866 * have to propagate any other updates to the superblock on disk at this
2867 * point. Just start an async writeback to get the buffers on their way
2870 * This implicitly triggers the writebehind on sync().
2873 static void ext4_write_super (struct super_block
* sb
)
2875 if (mutex_trylock(&sb
->s_lock
) != 0)
2880 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
2885 if (jbd2_journal_start_commit(EXT4_SB(sb
)->s_journal
, &target
)) {
2887 jbd2_log_wait_commit(EXT4_SB(sb
)->s_journal
, target
);
2893 * LVM calls this function before a (read-only) snapshot is created. This
2894 * gives us a chance to flush the journal completely and mark the fs clean.
2896 static void ext4_write_super_lockfs(struct super_block
*sb
)
2900 if (!(sb
->s_flags
& MS_RDONLY
)) {
2901 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
2903 /* Now we set up the journal barrier. */
2904 jbd2_journal_lock_updates(journal
);
2905 jbd2_journal_flush(journal
);
2907 /* Journal blocked and flushed, clear needs_recovery flag. */
2908 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2909 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
2914 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2915 * flag here, even though the filesystem is not technically dirty yet.
2917 static void ext4_unlockfs(struct super_block
*sb
)
2919 if (!(sb
->s_flags
& MS_RDONLY
)) {
2921 /* Reser the needs_recovery flag before the fs is unlocked. */
2922 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2923 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
2925 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
2929 static int ext4_remount (struct super_block
* sb
, int * flags
, char * data
)
2931 struct ext4_super_block
* es
;
2932 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2933 ext4_fsblk_t n_blocks_count
= 0;
2934 unsigned long old_sb_flags
;
2935 struct ext4_mount_options old_opts
;
2941 /* Store the original options */
2942 old_sb_flags
= sb
->s_flags
;
2943 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
2944 old_opts
.s_resuid
= sbi
->s_resuid
;
2945 old_opts
.s_resgid
= sbi
->s_resgid
;
2946 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
2948 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
2949 for (i
= 0; i
< MAXQUOTAS
; i
++)
2950 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
2954 * Allow the "check" option to be passed as a remount option.
2956 if (!parse_options(data
, sb
, NULL
, NULL
, &n_blocks_count
, 1)) {
2961 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
)
2962 ext4_abort(sb
, __func__
, "Abort forced by user");
2964 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2965 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2969 ext4_init_journal_params(sb
, sbi
->s_journal
);
2971 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
2972 n_blocks_count
> ext4_blocks_count(es
)) {
2973 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
) {
2978 if (*flags
& MS_RDONLY
) {
2980 * First of all, the unconditional stuff we have to do
2981 * to disable replay of the journal when we next remount
2983 sb
->s_flags
|= MS_RDONLY
;
2986 * OK, test if we are remounting a valid rw partition
2987 * readonly, and if so set the rdonly flag and then
2988 * mark the partition as valid again.
2990 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
2991 (sbi
->s_mount_state
& EXT4_VALID_FS
))
2992 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
2995 * We have to unlock super so that we can wait for
2999 ext4_mark_recovery_complete(sb
, es
);
3003 if ((ret
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3004 ~EXT4_FEATURE_RO_COMPAT_SUPP
))) {
3005 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3006 "remount RDWR because of unsupported "
3007 "optional features (%x).\n",
3008 sb
->s_id
, le32_to_cpu(ret
));
3014 * If we have an unprocessed orphan list hanging
3015 * around from a previously readonly bdev mount,
3016 * require a full umount/remount for now.
3018 if (es
->s_last_orphan
) {
3019 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3020 "remount RDWR because of unprocessed "
3021 "orphan inode list. Please "
3022 "umount/remount instead.\n",
3029 * Mounting a RDONLY partition read-write, so reread
3030 * and store the current valid flag. (It may have
3031 * been changed by e2fsck since we originally mounted
3034 ext4_clear_journal_err(sb
, es
);
3035 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3036 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3038 if (!ext4_setup_super (sb
, es
, 0))
3039 sb
->s_flags
&= ~MS_RDONLY
;
3043 /* Release old quota file names */
3044 for (i
= 0; i
< MAXQUOTAS
; i
++)
3045 if (old_opts
.s_qf_names
[i
] &&
3046 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3047 kfree(old_opts
.s_qf_names
[i
]);
3051 sb
->s_flags
= old_sb_flags
;
3052 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3053 sbi
->s_resuid
= old_opts
.s_resuid
;
3054 sbi
->s_resgid
= old_opts
.s_resgid
;
3055 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3057 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3058 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3059 if (sbi
->s_qf_names
[i
] &&
3060 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3061 kfree(sbi
->s_qf_names
[i
]);
3062 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3068 static int ext4_statfs (struct dentry
* dentry
, struct kstatfs
* buf
)
3070 struct super_block
*sb
= dentry
->d_sb
;
3071 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3072 struct ext4_super_block
*es
= sbi
->s_es
;
3075 if (test_opt(sb
, MINIX_DF
)) {
3076 sbi
->s_overhead_last
= 0;
3077 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3078 ext4_group_t ngroups
= sbi
->s_groups_count
, i
;
3079 ext4_fsblk_t overhead
= 0;
3083 * Compute the overhead (FS structures). This is constant
3084 * for a given filesystem unless the number of block groups
3085 * changes so we cache the previous value until it does.
3089 * All of the blocks before first_data_block are
3092 overhead
= le32_to_cpu(es
->s_first_data_block
);
3095 * Add the overhead attributed to the superblock and
3096 * block group descriptors. If the sparse superblocks
3097 * feature is turned on, then not all groups have this.
3099 for (i
= 0; i
< ngroups
; i
++) {
3100 overhead
+= ext4_bg_has_super(sb
, i
) +
3101 ext4_bg_num_gdb(sb
, i
);
3106 * Every block group has an inode bitmap, a block
3107 * bitmap, and an inode table.
3109 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3110 sbi
->s_overhead_last
= overhead
;
3112 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3115 buf
->f_type
= EXT4_SUPER_MAGIC
;
3116 buf
->f_bsize
= sb
->s_blocksize
;
3117 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3118 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
);
3119 ext4_free_blocks_count_set(es
, buf
->f_bfree
);
3120 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3121 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3123 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3124 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3125 es
->s_free_inodes_count
= cpu_to_le32(buf
->f_ffree
);
3126 buf
->f_namelen
= EXT4_NAME_LEN
;
3127 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3128 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3129 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3130 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3134 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3135 * is locked for write. Otherwise the are possible deadlocks:
3136 * Process 1 Process 2
3137 * ext4_create() quota_sync()
3138 * jbd2_journal_start() write_dquot()
3139 * DQUOT_INIT() down(dqio_mutex)
3140 * down(dqio_mutex) jbd2_journal_start()
3146 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3148 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3151 static int ext4_dquot_initialize(struct inode
*inode
, int type
)
3156 /* We may create quota structure so we need to reserve enough blocks */
3157 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_INIT_BLOCKS(inode
->i_sb
));
3159 return PTR_ERR(handle
);
3160 ret
= dquot_initialize(inode
, type
);
3161 err
= ext4_journal_stop(handle
);
3167 static int ext4_dquot_drop(struct inode
*inode
)
3172 /* We may delete quota structure so we need to reserve enough blocks */
3173 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_DEL_BLOCKS(inode
->i_sb
));
3174 if (IS_ERR(handle
)) {
3176 * We call dquot_drop() anyway to at least release references
3177 * to quota structures so that umount does not hang.
3180 return PTR_ERR(handle
);
3182 ret
= dquot_drop(inode
);
3183 err
= ext4_journal_stop(handle
);
3189 static int ext4_write_dquot(struct dquot
*dquot
)
3193 struct inode
*inode
;
3195 inode
= dquot_to_inode(dquot
);
3196 handle
= ext4_journal_start(inode
,
3197 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3199 return PTR_ERR(handle
);
3200 ret
= dquot_commit(dquot
);
3201 err
= ext4_journal_stop(handle
);
3207 static int ext4_acquire_dquot(struct dquot
*dquot
)
3212 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3213 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3215 return PTR_ERR(handle
);
3216 ret
= dquot_acquire(dquot
);
3217 err
= ext4_journal_stop(handle
);
3223 static int ext4_release_dquot(struct dquot
*dquot
)
3228 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3229 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3230 if (IS_ERR(handle
)) {
3231 /* Release dquot anyway to avoid endless cycle in dqput() */
3232 dquot_release(dquot
);
3233 return PTR_ERR(handle
);
3235 ret
= dquot_release(dquot
);
3236 err
= ext4_journal_stop(handle
);
3242 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3244 /* Are we journaling quotas? */
3245 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3246 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3247 dquot_mark_dquot_dirty(dquot
);
3248 return ext4_write_dquot(dquot
);
3250 return dquot_mark_dquot_dirty(dquot
);
3254 static int ext4_write_info(struct super_block
*sb
, int type
)
3259 /* Data block + inode block */
3260 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3262 return PTR_ERR(handle
);
3263 ret
= dquot_commit_info(sb
, type
);
3264 err
= ext4_journal_stop(handle
);
3271 * Turn on quotas during mount time - we need to find
3272 * the quota file and such...
3274 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3276 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3277 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3281 * Standard function to be called on quota_on
3283 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3284 char *path
, int remount
)
3287 struct nameidata nd
;
3289 if (!test_opt(sb
, QUOTA
))
3291 /* When remounting, no checks are needed and in fact, path is NULL */
3293 return vfs_quota_on(sb
, type
, format_id
, path
, remount
);
3295 err
= path_lookup(path
, LOOKUP_FOLLOW
, &nd
);
3299 /* Quotafile not on the same filesystem? */
3300 if (nd
.path
.mnt
->mnt_sb
!= sb
) {
3304 /* Journaling quota? */
3305 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3306 /* Quotafile not of fs root? */
3307 if (nd
.path
.dentry
->d_parent
->d_inode
!= sb
->s_root
->d_inode
)
3309 "EXT4-fs: Quota file not on filesystem root. "
3310 "Journaled quota will not work.\n");
3314 * When we journal data on quota file, we have to flush journal to see
3315 * all updates to the file when we bypass pagecache...
3317 if (ext4_should_journal_data(nd
.path
.dentry
->d_inode
)) {
3319 * We don't need to lock updates but journal_flush() could
3320 * otherwise be livelocked...
3322 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3323 jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3324 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3328 return vfs_quota_on(sb
, type
, format_id
, path
, remount
);
3331 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3332 * acquiring the locks... As quota files are never truncated and quota code
3333 * itself serializes the operations (and noone else should touch the files)
3334 * we don't have to be afraid of races */
3335 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3336 size_t len
, loff_t off
)
3338 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3339 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3341 int offset
= off
& (sb
->s_blocksize
- 1);
3344 struct buffer_head
*bh
;
3345 loff_t i_size
= i_size_read(inode
);
3349 if (off
+len
> i_size
)
3352 while (toread
> 0) {
3353 tocopy
= sb
->s_blocksize
- offset
< toread
?
3354 sb
->s_blocksize
- offset
: toread
;
3355 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3358 if (!bh
) /* A hole? */
3359 memset(data
, 0, tocopy
);
3361 memcpy(data
, bh
->b_data
+offset
, tocopy
);
3371 /* Write to quotafile (we know the transaction is already started and has
3372 * enough credits) */
3373 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
3374 const char *data
, size_t len
, loff_t off
)
3376 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3377 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3379 int offset
= off
& (sb
->s_blocksize
- 1);
3381 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
3382 size_t towrite
= len
;
3383 struct buffer_head
*bh
;
3384 handle_t
*handle
= journal_current_handle();
3387 printk(KERN_WARNING
"EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3388 " cancelled because transaction is not started.\n",
3389 (unsigned long long)off
, (unsigned long long)len
);
3392 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
3393 while (towrite
> 0) {
3394 tocopy
= sb
->s_blocksize
- offset
< towrite
?
3395 sb
->s_blocksize
- offset
: towrite
;
3396 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
3399 if (journal_quota
) {
3400 err
= ext4_journal_get_write_access(handle
, bh
);
3407 memcpy(bh
->b_data
+offset
, data
, tocopy
);
3408 flush_dcache_page(bh
->b_page
);
3411 err
= ext4_journal_dirty_metadata(handle
, bh
);
3413 /* Always do at least ordered writes for quotas */
3414 err
= ext4_jbd2_file_inode(handle
, inode
);
3415 mark_buffer_dirty(bh
);
3426 if (len
== towrite
) {
3427 mutex_unlock(&inode
->i_mutex
);
3430 if (inode
->i_size
< off
+len
-towrite
) {
3431 i_size_write(inode
, off
+len
-towrite
);
3432 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
3434 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3435 ext4_mark_inode_dirty(handle
, inode
);
3436 mutex_unlock(&inode
->i_mutex
);
3437 return len
- towrite
;
3442 static int ext4_get_sb(struct file_system_type
*fs_type
,
3443 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3445 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3448 static struct file_system_type ext4dev_fs_type
= {
3449 .owner
= THIS_MODULE
,
3451 .get_sb
= ext4_get_sb
,
3452 .kill_sb
= kill_block_super
,
3453 .fs_flags
= FS_REQUIRES_DEV
,
3456 static int __init
init_ext4_fs(void)
3460 err
= init_ext4_mballoc();
3464 err
= init_ext4_xattr();
3467 err
= init_inodecache();
3470 err
= register_filesystem(&ext4dev_fs_type
);
3475 destroy_inodecache();
3479 exit_ext4_mballoc();
3483 static void __exit
exit_ext4_fs(void)
3485 unregister_filesystem(&ext4dev_fs_type
);
3486 destroy_inodecache();
3488 exit_ext4_mballoc();
3491 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3492 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3493 MODULE_LICENSE("GPL");
3494 module_init(init_ext4_fs
)
3495 module_exit(exit_ext4_fs
)