2 * linux/fs/ext3/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/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/log2.h>
40 #include <asm/uaccess.h>
46 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
47 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
49 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
52 static int ext3_load_journal(struct super_block
*, struct ext3_super_block
*,
53 unsigned long journal_devnum
);
54 static int ext3_create_journal(struct super_block
*, struct ext3_super_block
*,
56 static int ext3_commit_super(struct super_block
*sb
,
57 struct ext3_super_block
*es
,
59 static void ext3_mark_recovery_complete(struct super_block
* sb
,
60 struct ext3_super_block
* es
);
61 static void ext3_clear_journal_err(struct super_block
* sb
,
62 struct ext3_super_block
* es
);
63 static int ext3_sync_fs(struct super_block
*sb
, int wait
);
64 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
66 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
);
67 static int ext3_statfs (struct dentry
* dentry
, struct kstatfs
* buf
);
68 static int ext3_unfreeze(struct super_block
*sb
);
69 static int ext3_freeze(struct super_block
*sb
);
72 * Wrappers for journal_start/end.
74 * The only special thing we need to do here is to make sure that all
75 * journal_end calls result in the superblock being marked dirty, so
76 * that sync() will call the filesystem's write_super callback if
79 handle_t
*ext3_journal_start_sb(struct super_block
*sb
, int nblocks
)
83 if (sb
->s_flags
& MS_RDONLY
)
84 return ERR_PTR(-EROFS
);
86 /* Special case here: if the journal has aborted behind our
87 * backs (eg. EIO in the commit thread), then we still need to
88 * take the FS itself readonly cleanly. */
89 journal
= EXT3_SB(sb
)->s_journal
;
90 if (is_journal_aborted(journal
)) {
91 ext3_abort(sb
, __func__
,
92 "Detected aborted journal");
93 return ERR_PTR(-EROFS
);
96 return journal_start(journal
, nblocks
);
100 * The only special thing we need to do here is to make sure that all
101 * journal_stop calls result in the superblock being marked dirty, so
102 * that sync() will call the filesystem's write_super callback if
105 int __ext3_journal_stop(const char *where
, handle_t
*handle
)
107 struct super_block
*sb
;
111 sb
= handle
->h_transaction
->t_journal
->j_private
;
113 rc
= journal_stop(handle
);
118 __ext3_std_error(sb
, where
, err
);
122 void ext3_journal_abort_handle(const char *caller
, const char *err_fn
,
123 struct buffer_head
*bh
, handle_t
*handle
, int err
)
126 const char *errstr
= ext3_decode_error(NULL
, err
, nbuf
);
129 BUFFER_TRACE(bh
, "abort");
134 if (is_handle_aborted(handle
))
137 printk(KERN_ERR
"EXT3-fs: %s: aborting transaction: %s in %s\n",
138 caller
, errstr
, err_fn
);
140 journal_abort_handle(handle
);
143 void ext3_msg(struct super_block
*sb
, const char *prefix
,
144 const char *fmt
, ...)
149 printk("%sEXT3-fs (%s): ", prefix
, sb
->s_id
);
155 /* Deal with the reporting of failure conditions on a filesystem such as
156 * inconsistencies detected or read IO failures.
158 * On ext2, we can store the error state of the filesystem in the
159 * superblock. That is not possible on ext3, because we may have other
160 * write ordering constraints on the superblock which prevent us from
161 * writing it out straight away; and given that the journal is about to
162 * be aborted, we can't rely on the current, or future, transactions to
163 * write out the superblock safely.
165 * We'll just use the journal_abort() error code to record an error in
166 * the journal instead. On recovery, the journal will complain about
167 * that error until we've noted it down and cleared it.
170 static void ext3_handle_error(struct super_block
*sb
)
172 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
174 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
175 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
177 if (sb
->s_flags
& MS_RDONLY
)
180 if (!test_opt (sb
, ERRORS_CONT
)) {
181 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
183 set_opt(EXT3_SB(sb
)->s_mount_opt
, ABORT
);
185 journal_abort(journal
, -EIO
);
187 if (test_opt (sb
, ERRORS_RO
)) {
188 ext3_msg(sb
, KERN_CRIT
,
189 "error: remounting filesystem read-only");
190 sb
->s_flags
|= MS_RDONLY
;
192 ext3_commit_super(sb
, es
, 1);
193 if (test_opt(sb
, ERRORS_PANIC
))
194 panic("EXT3-fs (%s): panic forced after error\n",
198 void ext3_error (struct super_block
* sb
, const char * function
,
199 const char * fmt
, ...)
204 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
209 ext3_handle_error(sb
);
212 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
219 errstr
= "IO failure";
222 errstr
= "Out of memory";
225 if (!sb
|| EXT3_SB(sb
)->s_journal
->j_flags
& JFS_ABORT
)
226 errstr
= "Journal has aborted";
228 errstr
= "Readonly filesystem";
231 /* If the caller passed in an extra buffer for unknown
232 * errors, textualise them now. Else we just return
235 /* Check for truncated error codes... */
236 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
245 /* __ext3_std_error decodes expected errors from journaling functions
246 * automatically and invokes the appropriate error response. */
248 void __ext3_std_error (struct super_block
* sb
, const char * function
,
254 /* Special case: if the error is EROFS, and we're not already
255 * inside a transaction, then there's really no point in logging
257 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
258 (sb
->s_flags
& MS_RDONLY
))
261 errstr
= ext3_decode_error(sb
, errno
, nbuf
);
262 ext3_msg(sb
, KERN_CRIT
, "error in %s: %s", function
, errstr
);
264 ext3_handle_error(sb
);
268 * ext3_abort is a much stronger failure handler than ext3_error. The
269 * abort function may be used to deal with unrecoverable failures such
270 * as journal IO errors or ENOMEM at a critical moment in log management.
272 * We unconditionally force the filesystem into an ABORT|READONLY state,
273 * unless the error response on the fs has been set to panic in which
274 * case we take the easy way out and panic immediately.
277 void ext3_abort (struct super_block
* sb
, const char * function
,
278 const char * fmt
, ...)
283 printk(KERN_CRIT
"EXT3-fs (%s): error: %s: ", sb
->s_id
, function
);
288 if (test_opt(sb
, ERRORS_PANIC
))
289 panic("EXT3-fs: panic from previous error\n");
291 if (sb
->s_flags
& MS_RDONLY
)
294 ext3_msg(sb
, KERN_CRIT
,
295 "error: remounting filesystem read-only");
296 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
297 sb
->s_flags
|= MS_RDONLY
;
298 set_opt(EXT3_SB(sb
)->s_mount_opt
, ABORT
);
299 if (EXT3_SB(sb
)->s_journal
)
300 journal_abort(EXT3_SB(sb
)->s_journal
, -EIO
);
303 void ext3_warning (struct super_block
* sb
, const char * function
,
304 const char * fmt
, ...)
309 printk(KERN_WARNING
"EXT3-fs (%s): warning: %s: ",
316 void ext3_update_dynamic_rev(struct super_block
*sb
)
318 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
320 if (le32_to_cpu(es
->s_rev_level
) > EXT3_GOOD_OLD_REV
)
323 ext3_msg(sb
, KERN_WARNING
,
324 "warning: updating to rev %d because of "
325 "new feature flag, running e2fsck is recommended",
328 es
->s_first_ino
= cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO
);
329 es
->s_inode_size
= cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE
);
330 es
->s_rev_level
= cpu_to_le32(EXT3_DYNAMIC_REV
);
331 /* leave es->s_feature_*compat flags alone */
332 /* es->s_uuid will be set by e2fsck if empty */
335 * The rest of the superblock fields should be zero, and if not it
336 * means they are likely already in use, so leave them alone. We
337 * can leave it up to e2fsck to clean up any inconsistencies there.
342 * Open the external journal device
344 static struct block_device
*ext3_blkdev_get(dev_t dev
, struct super_block
*sb
)
346 struct block_device
*bdev
;
347 char b
[BDEVNAME_SIZE
];
349 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
355 ext3_msg(sb
, "error: failed to open journal device %s: %ld",
356 __bdevname(dev
, b
), PTR_ERR(bdev
));
362 * Release the journal device
364 static int ext3_blkdev_put(struct block_device
*bdev
)
367 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
370 static int ext3_blkdev_remove(struct ext3_sb_info
*sbi
)
372 struct block_device
*bdev
;
375 bdev
= sbi
->journal_bdev
;
377 ret
= ext3_blkdev_put(bdev
);
378 sbi
->journal_bdev
= NULL
;
383 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
385 return &list_entry(l
, struct ext3_inode_info
, i_orphan
)->vfs_inode
;
388 static void dump_orphan_list(struct super_block
*sb
, struct ext3_sb_info
*sbi
)
392 ext3_msg(sb
, KERN_ERR
, "error: sb orphan head is %d",
393 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
395 ext3_msg(sb
, KERN_ERR
, "sb_info orphan list:");
396 list_for_each(l
, &sbi
->s_orphan
) {
397 struct inode
*inode
= orphan_list_entry(l
);
398 ext3_msg(sb
, KERN_ERR
, " "
399 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
400 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
401 inode
->i_mode
, inode
->i_nlink
,
406 static void ext3_put_super (struct super_block
* sb
)
408 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
409 struct ext3_super_block
*es
= sbi
->s_es
;
412 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
413 ext3_xattr_put_super(sb
);
414 err
= journal_destroy(sbi
->s_journal
);
415 sbi
->s_journal
= NULL
;
417 ext3_abort(sb
, __func__
, "Couldn't clean up the journal");
419 if (!(sb
->s_flags
& MS_RDONLY
)) {
420 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
421 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
422 BUFFER_TRACE(sbi
->s_sbh
, "marking dirty");
423 mark_buffer_dirty(sbi
->s_sbh
);
424 ext3_commit_super(sb
, es
, 1);
427 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
428 brelse(sbi
->s_group_desc
[i
]);
429 kfree(sbi
->s_group_desc
);
430 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
431 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
432 percpu_counter_destroy(&sbi
->s_dirs_counter
);
435 for (i
= 0; i
< MAXQUOTAS
; i
++)
436 kfree(sbi
->s_qf_names
[i
]);
439 /* Debugging code just in case the in-memory inode orphan list
440 * isn't empty. The on-disk one can be non-empty if we've
441 * detected an error and taken the fs readonly, but the
442 * in-memory list had better be clean by this point. */
443 if (!list_empty(&sbi
->s_orphan
))
444 dump_orphan_list(sb
, sbi
);
445 J_ASSERT(list_empty(&sbi
->s_orphan
));
447 invalidate_bdev(sb
->s_bdev
);
448 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
450 * Invalidate the journal device's buffers. We don't want them
451 * floating about in memory - the physical journal device may
452 * hotswapped, and it breaks the `ro-after' testing code.
454 sync_blockdev(sbi
->journal_bdev
);
455 invalidate_bdev(sbi
->journal_bdev
);
456 ext3_blkdev_remove(sbi
);
458 sb
->s_fs_info
= NULL
;
459 kfree(sbi
->s_blockgroup_lock
);
463 static struct kmem_cache
*ext3_inode_cachep
;
466 * Called inside transaction, so use GFP_NOFS
468 static struct inode
*ext3_alloc_inode(struct super_block
*sb
)
470 struct ext3_inode_info
*ei
;
472 ei
= kmem_cache_alloc(ext3_inode_cachep
, GFP_NOFS
);
475 ei
->i_block_alloc_info
= NULL
;
476 ei
->vfs_inode
.i_version
= 1;
477 atomic_set(&ei
->i_datasync_tid
, 0);
478 atomic_set(&ei
->i_sync_tid
, 0);
479 return &ei
->vfs_inode
;
482 static void ext3_i_callback(struct rcu_head
*head
)
484 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
485 INIT_LIST_HEAD(&inode
->i_dentry
);
486 kmem_cache_free(ext3_inode_cachep
, EXT3_I(inode
));
489 static void ext3_destroy_inode(struct inode
*inode
)
491 if (!list_empty(&(EXT3_I(inode
)->i_orphan
))) {
492 printk("EXT3 Inode %p: orphan list check failed!\n",
494 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
495 EXT3_I(inode
), sizeof(struct ext3_inode_info
),
499 call_rcu(&inode
->i_rcu
, ext3_i_callback
);
502 static void init_once(void *foo
)
504 struct ext3_inode_info
*ei
= (struct ext3_inode_info
*) foo
;
506 INIT_LIST_HEAD(&ei
->i_orphan
);
507 #ifdef CONFIG_EXT3_FS_XATTR
508 init_rwsem(&ei
->xattr_sem
);
510 mutex_init(&ei
->truncate_mutex
);
511 inode_init_once(&ei
->vfs_inode
);
514 static int init_inodecache(void)
516 ext3_inode_cachep
= kmem_cache_create("ext3_inode_cache",
517 sizeof(struct ext3_inode_info
),
518 0, (SLAB_RECLAIM_ACCOUNT
|
521 if (ext3_inode_cachep
== NULL
)
526 static void destroy_inodecache(void)
528 kmem_cache_destroy(ext3_inode_cachep
);
531 static inline void ext3_show_quota_options(struct seq_file
*seq
, struct super_block
*sb
)
533 #if defined(CONFIG_QUOTA)
534 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
536 if (sbi
->s_jquota_fmt
) {
539 switch (sbi
->s_jquota_fmt
) {
550 seq_printf(seq
, ",jqfmt=%s", fmtname
);
553 if (sbi
->s_qf_names
[USRQUOTA
])
554 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
556 if (sbi
->s_qf_names
[GRPQUOTA
])
557 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
559 if (test_opt(sb
, USRQUOTA
))
560 seq_puts(seq
, ",usrquota");
562 if (test_opt(sb
, GRPQUOTA
))
563 seq_puts(seq
, ",grpquota");
567 static char *data_mode_string(unsigned long mode
)
570 case EXT3_MOUNT_JOURNAL_DATA
:
572 case EXT3_MOUNT_ORDERED_DATA
:
574 case EXT3_MOUNT_WRITEBACK_DATA
:
582 * - it's set to a non-default value OR
583 * - if the per-sb default is different from the global default
585 static int ext3_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
587 struct super_block
*sb
= vfs
->mnt_sb
;
588 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
589 struct ext3_super_block
*es
= sbi
->s_es
;
590 unsigned long def_mount_opts
;
592 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
594 if (sbi
->s_sb_block
!= 1)
595 seq_printf(seq
, ",sb=%lu", sbi
->s_sb_block
);
596 if (test_opt(sb
, MINIX_DF
))
597 seq_puts(seq
, ",minixdf");
598 if (test_opt(sb
, GRPID
))
599 seq_puts(seq
, ",grpid");
600 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT3_DEFM_BSDGROUPS
))
601 seq_puts(seq
, ",nogrpid");
602 if (sbi
->s_resuid
!= EXT3_DEF_RESUID
||
603 le16_to_cpu(es
->s_def_resuid
) != EXT3_DEF_RESUID
) {
604 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
606 if (sbi
->s_resgid
!= EXT3_DEF_RESGID
||
607 le16_to_cpu(es
->s_def_resgid
) != EXT3_DEF_RESGID
) {
608 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
610 if (test_opt(sb
, ERRORS_RO
)) {
611 int def_errors
= le16_to_cpu(es
->s_errors
);
613 if (def_errors
== EXT3_ERRORS_PANIC
||
614 def_errors
== EXT3_ERRORS_CONTINUE
) {
615 seq_puts(seq
, ",errors=remount-ro");
618 if (test_opt(sb
, ERRORS_CONT
))
619 seq_puts(seq
, ",errors=continue");
620 if (test_opt(sb
, ERRORS_PANIC
))
621 seq_puts(seq
, ",errors=panic");
622 if (test_opt(sb
, NO_UID32
))
623 seq_puts(seq
, ",nouid32");
624 if (test_opt(sb
, DEBUG
))
625 seq_puts(seq
, ",debug");
626 if (test_opt(sb
, OLDALLOC
))
627 seq_puts(seq
, ",oldalloc");
628 #ifdef CONFIG_EXT3_FS_XATTR
629 if (test_opt(sb
, XATTR_USER
))
630 seq_puts(seq
, ",user_xattr");
631 if (!test_opt(sb
, XATTR_USER
) &&
632 (def_mount_opts
& EXT3_DEFM_XATTR_USER
)) {
633 seq_puts(seq
, ",nouser_xattr");
636 #ifdef CONFIG_EXT3_FS_POSIX_ACL
637 if (test_opt(sb
, POSIX_ACL
))
638 seq_puts(seq
, ",acl");
639 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT3_DEFM_ACL
))
640 seq_puts(seq
, ",noacl");
642 if (!test_opt(sb
, RESERVATION
))
643 seq_puts(seq
, ",noreservation");
644 if (sbi
->s_commit_interval
) {
645 seq_printf(seq
, ",commit=%u",
646 (unsigned) (sbi
->s_commit_interval
/ HZ
));
650 * Always display barrier state so it's clear what the status is.
652 seq_puts(seq
, ",barrier=");
653 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
654 seq_printf(seq
, ",data=%s", data_mode_string(test_opt(sb
, DATA_FLAGS
)));
655 if (test_opt(sb
, DATA_ERR_ABORT
))
656 seq_puts(seq
, ",data_err=abort");
658 if (test_opt(sb
, NOLOAD
))
659 seq_puts(seq
, ",norecovery");
661 ext3_show_quota_options(seq
, sb
);
667 static struct inode
*ext3_nfs_get_inode(struct super_block
*sb
,
668 u64 ino
, u32 generation
)
672 if (ino
< EXT3_FIRST_INO(sb
) && ino
!= EXT3_ROOT_INO
)
673 return ERR_PTR(-ESTALE
);
674 if (ino
> le32_to_cpu(EXT3_SB(sb
)->s_es
->s_inodes_count
))
675 return ERR_PTR(-ESTALE
);
677 /* iget isn't really right if the inode is currently unallocated!!
679 * ext3_read_inode will return a bad_inode if the inode had been
680 * deleted, so we should be safe.
682 * Currently we don't know the generation for parent directory, so
683 * a generation of 0 means "accept any"
685 inode
= ext3_iget(sb
, ino
);
687 return ERR_CAST(inode
);
688 if (generation
&& inode
->i_generation
!= generation
) {
690 return ERR_PTR(-ESTALE
);
696 static struct dentry
*ext3_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
697 int fh_len
, int fh_type
)
699 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
703 static struct dentry
*ext3_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
704 int fh_len
, int fh_type
)
706 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
711 * Try to release metadata pages (indirect blocks, directories) which are
712 * mapped via the block device. Since these pages could have journal heads
713 * which would prevent try_to_free_buffers() from freeing them, we must use
714 * jbd layer's try_to_free_buffers() function to release them.
716 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
719 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
721 WARN_ON(PageChecked(page
));
722 if (!page_has_buffers(page
))
725 return journal_try_to_free_buffers(journal
, page
,
727 return try_to_free_buffers(page
);
731 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
732 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
734 static int ext3_write_dquot(struct dquot
*dquot
);
735 static int ext3_acquire_dquot(struct dquot
*dquot
);
736 static int ext3_release_dquot(struct dquot
*dquot
);
737 static int ext3_mark_dquot_dirty(struct dquot
*dquot
);
738 static int ext3_write_info(struct super_block
*sb
, int type
);
739 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
,
741 static int ext3_quota_on_mount(struct super_block
*sb
, int type
);
742 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
743 size_t len
, loff_t off
);
744 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
745 const char *data
, size_t len
, loff_t off
);
747 static const struct dquot_operations ext3_quota_operations
= {
748 .write_dquot
= ext3_write_dquot
,
749 .acquire_dquot
= ext3_acquire_dquot
,
750 .release_dquot
= ext3_release_dquot
,
751 .mark_dirty
= ext3_mark_dquot_dirty
,
752 .write_info
= ext3_write_info
,
753 .alloc_dquot
= dquot_alloc
,
754 .destroy_dquot
= dquot_destroy
,
757 static const struct quotactl_ops ext3_qctl_operations
= {
758 .quota_on
= ext3_quota_on
,
759 .quota_off
= dquot_quota_off
,
760 .quota_sync
= dquot_quota_sync
,
761 .get_info
= dquot_get_dqinfo
,
762 .set_info
= dquot_set_dqinfo
,
763 .get_dqblk
= dquot_get_dqblk
,
764 .set_dqblk
= dquot_set_dqblk
768 static const struct super_operations ext3_sops
= {
769 .alloc_inode
= ext3_alloc_inode
,
770 .destroy_inode
= ext3_destroy_inode
,
771 .write_inode
= ext3_write_inode
,
772 .dirty_inode
= ext3_dirty_inode
,
773 .evict_inode
= ext3_evict_inode
,
774 .put_super
= ext3_put_super
,
775 .sync_fs
= ext3_sync_fs
,
776 .freeze_fs
= ext3_freeze
,
777 .unfreeze_fs
= ext3_unfreeze
,
778 .statfs
= ext3_statfs
,
779 .remount_fs
= ext3_remount
,
780 .show_options
= ext3_show_options
,
782 .quota_read
= ext3_quota_read
,
783 .quota_write
= ext3_quota_write
,
785 .bdev_try_to_free_page
= bdev_try_to_free_page
,
788 static const struct export_operations ext3_export_ops
= {
789 .fh_to_dentry
= ext3_fh_to_dentry
,
790 .fh_to_parent
= ext3_fh_to_parent
,
791 .get_parent
= ext3_get_parent
,
795 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
796 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
797 Opt_nouid32
, Opt_nocheck
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
798 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
799 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
800 Opt_commit
, Opt_journal_update
, Opt_journal_inum
, Opt_journal_dev
,
801 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
802 Opt_data_err_abort
, Opt_data_err_ignore
,
803 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
804 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
805 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
806 Opt_resize
, Opt_usrquota
, Opt_grpquota
809 static const match_table_t tokens
= {
810 {Opt_bsd_df
, "bsddf"},
811 {Opt_minix_df
, "minixdf"},
812 {Opt_grpid
, "grpid"},
813 {Opt_grpid
, "bsdgroups"},
814 {Opt_nogrpid
, "nogrpid"},
815 {Opt_nogrpid
, "sysvgroups"},
816 {Opt_resgid
, "resgid=%u"},
817 {Opt_resuid
, "resuid=%u"},
819 {Opt_err_cont
, "errors=continue"},
820 {Opt_err_panic
, "errors=panic"},
821 {Opt_err_ro
, "errors=remount-ro"},
822 {Opt_nouid32
, "nouid32"},
823 {Opt_nocheck
, "nocheck"},
824 {Opt_nocheck
, "check=none"},
825 {Opt_debug
, "debug"},
826 {Opt_oldalloc
, "oldalloc"},
827 {Opt_orlov
, "orlov"},
828 {Opt_user_xattr
, "user_xattr"},
829 {Opt_nouser_xattr
, "nouser_xattr"},
831 {Opt_noacl
, "noacl"},
832 {Opt_reservation
, "reservation"},
833 {Opt_noreservation
, "noreservation"},
834 {Opt_noload
, "noload"},
835 {Opt_noload
, "norecovery"},
838 {Opt_commit
, "commit=%u"},
839 {Opt_journal_update
, "journal=update"},
840 {Opt_journal_inum
, "journal=%u"},
841 {Opt_journal_dev
, "journal_dev=%u"},
842 {Opt_abort
, "abort"},
843 {Opt_data_journal
, "data=journal"},
844 {Opt_data_ordered
, "data=ordered"},
845 {Opt_data_writeback
, "data=writeback"},
846 {Opt_data_err_abort
, "data_err=abort"},
847 {Opt_data_err_ignore
, "data_err=ignore"},
848 {Opt_offusrjquota
, "usrjquota="},
849 {Opt_usrjquota
, "usrjquota=%s"},
850 {Opt_offgrpjquota
, "grpjquota="},
851 {Opt_grpjquota
, "grpjquota=%s"},
852 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
853 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
854 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
855 {Opt_grpquota
, "grpquota"},
856 {Opt_noquota
, "noquota"},
857 {Opt_quota
, "quota"},
858 {Opt_usrquota
, "usrquota"},
859 {Opt_barrier
, "barrier=%u"},
860 {Opt_barrier
, "barrier"},
861 {Opt_nobarrier
, "nobarrier"},
862 {Opt_resize
, "resize"},
866 static ext3_fsblk_t
get_sb_block(void **data
, struct super_block
*sb
)
868 ext3_fsblk_t sb_block
;
869 char *options
= (char *) *data
;
871 if (!options
|| strncmp(options
, "sb=", 3) != 0)
872 return 1; /* Default location */
874 /*todo: use simple_strtoll with >32bit ext3 */
875 sb_block
= simple_strtoul(options
, &options
, 0);
876 if (*options
&& *options
!= ',') {
877 ext3_msg(sb
, "error: invalid sb specification: %s",
883 *data
= (void *) options
;
888 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
890 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
893 if (sb_any_quota_loaded(sb
) &&
894 !sbi
->s_qf_names
[qtype
]) {
895 ext3_msg(sb
, KERN_ERR
,
896 "Cannot change journaled "
897 "quota options when quota turned on");
900 qname
= match_strdup(args
);
902 ext3_msg(sb
, KERN_ERR
,
903 "Not enough memory for storing quotafile name");
906 if (sbi
->s_qf_names
[qtype
] &&
907 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
908 ext3_msg(sb
, KERN_ERR
,
909 "%s quota file already specified", QTYPE2NAME(qtype
));
913 sbi
->s_qf_names
[qtype
] = qname
;
914 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
915 ext3_msg(sb
, KERN_ERR
,
916 "quotafile must be on filesystem root");
917 kfree(sbi
->s_qf_names
[qtype
]);
918 sbi
->s_qf_names
[qtype
] = NULL
;
921 set_opt(sbi
->s_mount_opt
, QUOTA
);
925 static int clear_qf_name(struct super_block
*sb
, int qtype
) {
927 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
929 if (sb_any_quota_loaded(sb
) &&
930 sbi
->s_qf_names
[qtype
]) {
931 ext3_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
932 " when quota turned on");
936 * The space will be released later when all options are confirmed
939 sbi
->s_qf_names
[qtype
] = NULL
;
944 static int parse_options (char *options
, struct super_block
*sb
,
945 unsigned int *inum
, unsigned long *journal_devnum
,
946 ext3_fsblk_t
*n_blocks_count
, int is_remount
)
948 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
950 substring_t args
[MAX_OPT_ARGS
];
960 while ((p
= strsep (&options
, ",")) != NULL
) {
965 * Initialize args struct so we know whether arg was
966 * found; some options take optional arguments.
968 args
[0].to
= args
[0].from
= 0;
969 token
= match_token(p
, tokens
, args
);
972 clear_opt (sbi
->s_mount_opt
, MINIX_DF
);
975 set_opt (sbi
->s_mount_opt
, MINIX_DF
);
978 set_opt (sbi
->s_mount_opt
, GRPID
);
981 clear_opt (sbi
->s_mount_opt
, GRPID
);
984 if (match_int(&args
[0], &option
))
986 sbi
->s_resuid
= option
;
989 if (match_int(&args
[0], &option
))
991 sbi
->s_resgid
= option
;
994 /* handled by get_sb_block() instead of here */
995 /* *sb_block = match_int(&args[0]); */
998 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
999 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1000 set_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1003 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1004 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1005 set_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1008 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1009 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1010 set_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1013 set_opt (sbi
->s_mount_opt
, NO_UID32
);
1016 clear_opt (sbi
->s_mount_opt
, CHECK
);
1019 set_opt (sbi
->s_mount_opt
, DEBUG
);
1022 set_opt (sbi
->s_mount_opt
, OLDALLOC
);
1025 clear_opt (sbi
->s_mount_opt
, OLDALLOC
);
1027 #ifdef CONFIG_EXT3_FS_XATTR
1028 case Opt_user_xattr
:
1029 set_opt (sbi
->s_mount_opt
, XATTR_USER
);
1031 case Opt_nouser_xattr
:
1032 clear_opt (sbi
->s_mount_opt
, XATTR_USER
);
1035 case Opt_user_xattr
:
1036 case Opt_nouser_xattr
:
1037 ext3_msg(sb
, KERN_INFO
,
1038 "(no)user_xattr options not supported");
1041 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1043 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1046 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1051 ext3_msg(sb
, KERN_INFO
,
1052 "(no)acl options not supported");
1055 case Opt_reservation
:
1056 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1058 case Opt_noreservation
:
1059 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
1061 case Opt_journal_update
:
1063 /* Eventually we will want to be able to create
1064 a journal file here. For now, only allow the
1065 user to specify an existing inode to be the
1068 ext3_msg(sb
, KERN_ERR
, "error: cannot specify "
1069 "journal on remount");
1072 set_opt (sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1074 case Opt_journal_inum
:
1076 ext3_msg(sb
, KERN_ERR
, "error: cannot specify "
1077 "journal on remount");
1080 if (match_int(&args
[0], &option
))
1084 case Opt_journal_dev
:
1086 ext3_msg(sb
, KERN_ERR
, "error: cannot specify "
1087 "journal on remount");
1090 if (match_int(&args
[0], &option
))
1092 *journal_devnum
= option
;
1095 set_opt (sbi
->s_mount_opt
, NOLOAD
);
1098 if (match_int(&args
[0], &option
))
1103 option
= JBD_DEFAULT_MAX_COMMIT_AGE
;
1104 sbi
->s_commit_interval
= HZ
* option
;
1106 case Opt_data_journal
:
1107 data_opt
= EXT3_MOUNT_JOURNAL_DATA
;
1109 case Opt_data_ordered
:
1110 data_opt
= EXT3_MOUNT_ORDERED_DATA
;
1112 case Opt_data_writeback
:
1113 data_opt
= EXT3_MOUNT_WRITEBACK_DATA
;
1116 if (test_opt(sb
, DATA_FLAGS
) == data_opt
)
1118 ext3_msg(sb
, KERN_ERR
,
1119 "error: cannot change "
1120 "data mode on remount. The filesystem "
1121 "is mounted in data=%s mode and you "
1122 "try to remount it in data=%s mode.",
1123 data_mode_string(test_opt(sb
,
1125 data_mode_string(data_opt
));
1128 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
1129 sbi
->s_mount_opt
|= data_opt
;
1132 case Opt_data_err_abort
:
1133 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1135 case Opt_data_err_ignore
:
1136 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1140 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1144 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1147 case Opt_offusrjquota
:
1148 if (!clear_qf_name(sb
, USRQUOTA
))
1151 case Opt_offgrpjquota
:
1152 if (!clear_qf_name(sb
, GRPQUOTA
))
1155 case Opt_jqfmt_vfsold
:
1156 qfmt
= QFMT_VFS_OLD
;
1158 case Opt_jqfmt_vfsv0
:
1161 case Opt_jqfmt_vfsv1
:
1164 if (sb_any_quota_loaded(sb
) &&
1165 sbi
->s_jquota_fmt
!= qfmt
) {
1166 ext3_msg(sb
, KERN_ERR
, "error: cannot change "
1167 "journaled quota options when "
1168 "quota turned on.");
1171 sbi
->s_jquota_fmt
= qfmt
;
1175 set_opt(sbi
->s_mount_opt
, QUOTA
);
1176 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1179 set_opt(sbi
->s_mount_opt
, QUOTA
);
1180 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1183 if (sb_any_quota_loaded(sb
)) {
1184 ext3_msg(sb
, KERN_ERR
, "error: cannot change "
1185 "quota options when quota turned on.");
1188 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1189 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1190 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1196 ext3_msg(sb
, KERN_ERR
,
1197 "error: quota options not supported.");
1201 case Opt_offusrjquota
:
1202 case Opt_offgrpjquota
:
1203 case Opt_jqfmt_vfsold
:
1204 case Opt_jqfmt_vfsv0
:
1205 case Opt_jqfmt_vfsv1
:
1206 ext3_msg(sb
, KERN_ERR
,
1207 "error: journaled quota options not "
1214 set_opt(sbi
->s_mount_opt
, ABORT
);
1217 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1221 if (match_int(&args
[0], &option
))
1224 option
= 1; /* No argument, default to 1 */
1226 set_opt(sbi
->s_mount_opt
, BARRIER
);
1228 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1234 ext3_msg(sb
, KERN_ERR
,
1235 "error: resize option only available "
1239 if (match_int(&args
[0], &option
) != 0)
1241 *n_blocks_count
= option
;
1244 ext3_msg(sb
, KERN_WARNING
,
1245 "warning: ignoring deprecated nobh option");
1248 ext3_msg(sb
, KERN_WARNING
,
1249 "warning: ignoring deprecated bh option");
1252 ext3_msg(sb
, KERN_ERR
,
1253 "error: unrecognized mount option \"%s\" "
1254 "or missing value", p
);
1259 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1260 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1261 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1262 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1263 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1265 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1266 ext3_msg(sb
, KERN_ERR
, "error: old and new quota "
1271 if (!sbi
->s_jquota_fmt
) {
1272 ext3_msg(sb
, KERN_ERR
, "error: journaled quota format "
1277 if (sbi
->s_jquota_fmt
) {
1278 ext3_msg(sb
, KERN_ERR
, "error: journaled quota format "
1279 "specified with no journaling "
1288 static int ext3_setup_super(struct super_block
*sb
, struct ext3_super_block
*es
,
1291 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1294 if (le32_to_cpu(es
->s_rev_level
) > EXT3_MAX_SUPP_REV
) {
1295 ext3_msg(sb
, KERN_ERR
,
1296 "error: revision level too high, "
1297 "forcing read-only mode");
1302 if (!(sbi
->s_mount_state
& EXT3_VALID_FS
))
1303 ext3_msg(sb
, KERN_WARNING
,
1304 "warning: mounting unchecked fs, "
1305 "running e2fsck is recommended");
1306 else if ((sbi
->s_mount_state
& EXT3_ERROR_FS
))
1307 ext3_msg(sb
, KERN_WARNING
,
1308 "warning: mounting fs with errors, "
1309 "running e2fsck is recommended");
1310 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1311 le16_to_cpu(es
->s_mnt_count
) >=
1312 le16_to_cpu(es
->s_max_mnt_count
))
1313 ext3_msg(sb
, KERN_WARNING
,
1314 "warning: maximal mount count reached, "
1315 "running e2fsck is recommended");
1316 else if (le32_to_cpu(es
->s_checkinterval
) &&
1317 (le32_to_cpu(es
->s_lastcheck
) +
1318 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1319 ext3_msg(sb
, KERN_WARNING
,
1320 "warning: checktime reached, "
1321 "running e2fsck is recommended");
1323 /* @@@ We _will_ want to clear the valid bit if we find
1324 inconsistencies, to force a fsck at reboot. But for
1325 a plain journaled filesystem we can keep it set as
1326 valid forever! :) */
1327 es
->s_state
&= cpu_to_le16(~EXT3_VALID_FS
);
1329 if (!le16_to_cpu(es
->s_max_mnt_count
))
1330 es
->s_max_mnt_count
= cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT
);
1331 le16_add_cpu(&es
->s_mnt_count
, 1);
1332 es
->s_mtime
= cpu_to_le32(get_seconds());
1333 ext3_update_dynamic_rev(sb
);
1334 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
1336 ext3_commit_super(sb
, es
, 1);
1337 if (test_opt(sb
, DEBUG
))
1338 ext3_msg(sb
, KERN_INFO
, "[bs=%lu, gc=%lu, "
1339 "bpg=%lu, ipg=%lu, mo=%04lx]",
1341 sbi
->s_groups_count
,
1342 EXT3_BLOCKS_PER_GROUP(sb
),
1343 EXT3_INODES_PER_GROUP(sb
),
1346 if (EXT3_SB(sb
)->s_journal
->j_inode
== NULL
) {
1347 char b
[BDEVNAME_SIZE
];
1348 ext3_msg(sb
, KERN_INFO
, "using external journal on %s",
1349 bdevname(EXT3_SB(sb
)->s_journal
->j_dev
, b
));
1351 ext3_msg(sb
, KERN_INFO
, "using internal journal");
1356 /* Called at mount-time, super-block is locked */
1357 static int ext3_check_descriptors(struct super_block
*sb
)
1359 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1362 ext3_debug ("Checking group descriptors");
1364 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1365 struct ext3_group_desc
*gdp
= ext3_get_group_desc(sb
, i
, NULL
);
1366 ext3_fsblk_t first_block
= ext3_group_first_block_no(sb
, i
);
1367 ext3_fsblk_t last_block
;
1369 if (i
== sbi
->s_groups_count
- 1)
1370 last_block
= le32_to_cpu(sbi
->s_es
->s_blocks_count
) - 1;
1372 last_block
= first_block
+
1373 (EXT3_BLOCKS_PER_GROUP(sb
) - 1);
1375 if (le32_to_cpu(gdp
->bg_block_bitmap
) < first_block
||
1376 le32_to_cpu(gdp
->bg_block_bitmap
) > last_block
)
1378 ext3_error (sb
, "ext3_check_descriptors",
1379 "Block bitmap for group %d"
1380 " not in group (block %lu)!",
1382 le32_to_cpu(gdp
->bg_block_bitmap
));
1385 if (le32_to_cpu(gdp
->bg_inode_bitmap
) < first_block
||
1386 le32_to_cpu(gdp
->bg_inode_bitmap
) > last_block
)
1388 ext3_error (sb
, "ext3_check_descriptors",
1389 "Inode bitmap for group %d"
1390 " not in group (block %lu)!",
1392 le32_to_cpu(gdp
->bg_inode_bitmap
));
1395 if (le32_to_cpu(gdp
->bg_inode_table
) < first_block
||
1396 le32_to_cpu(gdp
->bg_inode_table
) + sbi
->s_itb_per_group
- 1 >
1399 ext3_error (sb
, "ext3_check_descriptors",
1400 "Inode table for group %d"
1401 " not in group (block %lu)!",
1403 le32_to_cpu(gdp
->bg_inode_table
));
1408 sbi
->s_es
->s_free_blocks_count
=cpu_to_le32(ext3_count_free_blocks(sb
));
1409 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext3_count_free_inodes(sb
));
1414 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1415 * the superblock) which were deleted from all directories, but held open by
1416 * a process at the time of a crash. We walk the list and try to delete these
1417 * inodes at recovery time (only with a read-write filesystem).
1419 * In order to keep the orphan inode chain consistent during traversal (in
1420 * case of crash during recovery), we link each inode into the superblock
1421 * orphan list_head and handle it the same way as an inode deletion during
1422 * normal operation (which journals the operations for us).
1424 * We only do an iget() and an iput() on each inode, which is very safe if we
1425 * accidentally point at an in-use or already deleted inode. The worst that
1426 * can happen in this case is that we get a "bit already cleared" message from
1427 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1428 * e2fsck was run on this filesystem, and it must have already done the orphan
1429 * inode cleanup for us, so we can safely abort without any further action.
1431 static void ext3_orphan_cleanup (struct super_block
* sb
,
1432 struct ext3_super_block
* es
)
1434 unsigned int s_flags
= sb
->s_flags
;
1435 int nr_orphans
= 0, nr_truncates
= 0;
1439 if (!es
->s_last_orphan
) {
1440 jbd_debug(4, "no orphan inodes to clean up\n");
1444 if (bdev_read_only(sb
->s_bdev
)) {
1445 ext3_msg(sb
, KERN_ERR
, "error: write access "
1446 "unavailable, skipping orphan cleanup.");
1450 if (EXT3_SB(sb
)->s_mount_state
& EXT3_ERROR_FS
) {
1451 if (es
->s_last_orphan
)
1452 jbd_debug(1, "Errors on filesystem, "
1453 "clearing orphan list.\n");
1454 es
->s_last_orphan
= 0;
1455 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1459 if (s_flags
& MS_RDONLY
) {
1460 ext3_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
1461 sb
->s_flags
&= ~MS_RDONLY
;
1464 /* Needed for iput() to work correctly and not trash data */
1465 sb
->s_flags
|= MS_ACTIVE
;
1466 /* Turn on quotas so that they are updated correctly */
1467 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1468 if (EXT3_SB(sb
)->s_qf_names
[i
]) {
1469 int ret
= ext3_quota_on_mount(sb
, i
);
1471 ext3_msg(sb
, KERN_ERR
,
1472 "error: cannot turn on journaled "
1478 while (es
->s_last_orphan
) {
1479 struct inode
*inode
;
1481 inode
= ext3_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1482 if (IS_ERR(inode
)) {
1483 es
->s_last_orphan
= 0;
1487 list_add(&EXT3_I(inode
)->i_orphan
, &EXT3_SB(sb
)->s_orphan
);
1488 dquot_initialize(inode
);
1489 if (inode
->i_nlink
) {
1491 "%s: truncating inode %lu to %Ld bytes\n",
1492 __func__
, inode
->i_ino
, inode
->i_size
);
1493 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1494 inode
->i_ino
, inode
->i_size
);
1495 ext3_truncate(inode
);
1499 "%s: deleting unreferenced inode %lu\n",
1500 __func__
, inode
->i_ino
);
1501 jbd_debug(2, "deleting unreferenced inode %lu\n",
1505 iput(inode
); /* The delete magic happens here! */
1508 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1511 ext3_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
1512 PLURAL(nr_orphans
));
1514 ext3_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
1515 PLURAL(nr_truncates
));
1517 /* Turn quotas off */
1518 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1519 if (sb_dqopt(sb
)->files
[i
])
1520 dquot_quota_off(sb
, i
);
1523 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1527 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1528 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1529 * We need to be 1 filesystem block less than the 2^32 sector limit.
1531 static loff_t
ext3_max_size(int bits
)
1533 loff_t res
= EXT3_NDIR_BLOCKS
;
1537 /* This is calculated to be the largest file size for a
1538 * dense, file such that the total number of
1539 * sectors in the file, including data and all indirect blocks,
1540 * does not exceed 2^32 -1
1541 * __u32 i_blocks representing the total number of
1542 * 512 bytes blocks of the file
1544 upper_limit
= (1LL << 32) - 1;
1546 /* total blocks in file system block size */
1547 upper_limit
>>= (bits
- 9);
1550 /* indirect blocks */
1552 /* double indirect blocks */
1553 meta_blocks
+= 1 + (1LL << (bits
-2));
1554 /* tripple indirect blocks */
1555 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
1557 upper_limit
-= meta_blocks
;
1558 upper_limit
<<= bits
;
1560 res
+= 1LL << (bits
-2);
1561 res
+= 1LL << (2*(bits
-2));
1562 res
+= 1LL << (3*(bits
-2));
1564 if (res
> upper_limit
)
1567 if (res
> MAX_LFS_FILESIZE
)
1568 res
= MAX_LFS_FILESIZE
;
1573 static ext3_fsblk_t
descriptor_loc(struct super_block
*sb
,
1574 ext3_fsblk_t logic_sb_block
,
1577 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1578 unsigned long bg
, first_meta_bg
;
1581 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1583 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_META_BG
) ||
1585 return (logic_sb_block
+ nr
+ 1);
1586 bg
= sbi
->s_desc_per_block
* nr
;
1587 if (ext3_bg_has_super(sb
, bg
))
1589 return (has_super
+ ext3_group_first_block_no(sb
, bg
));
1593 static int ext3_fill_super (struct super_block
*sb
, void *data
, int silent
)
1595 struct buffer_head
* bh
;
1596 struct ext3_super_block
*es
= NULL
;
1597 struct ext3_sb_info
*sbi
;
1599 ext3_fsblk_t sb_block
= get_sb_block(&data
, sb
);
1600 ext3_fsblk_t logic_sb_block
;
1601 unsigned long offset
= 0;
1602 unsigned int journal_inum
= 0;
1603 unsigned long journal_devnum
= 0;
1604 unsigned long def_mount_opts
;
1615 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
1619 sbi
->s_blockgroup_lock
=
1620 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
1621 if (!sbi
->s_blockgroup_lock
) {
1625 sb
->s_fs_info
= sbi
;
1626 sbi
->s_mount_opt
= 0;
1627 sbi
->s_resuid
= EXT3_DEF_RESUID
;
1628 sbi
->s_resgid
= EXT3_DEF_RESGID
;
1629 sbi
->s_sb_block
= sb_block
;
1631 blocksize
= sb_min_blocksize(sb
, EXT3_MIN_BLOCK_SIZE
);
1633 ext3_msg(sb
, KERN_ERR
, "error: unable to set blocksize");
1638 * The ext3 superblock will not be buffer aligned for other than 1kB
1639 * block sizes. We need to calculate the offset from buffer start.
1641 if (blocksize
!= EXT3_MIN_BLOCK_SIZE
) {
1642 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1643 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1645 logic_sb_block
= sb_block
;
1648 if (!(bh
= sb_bread(sb
, logic_sb_block
))) {
1649 ext3_msg(sb
, KERN_ERR
, "error: unable to read superblock");
1653 * Note: s_es must be initialized as soon as possible because
1654 * some ext3 macro-instructions depend on its value
1656 es
= (struct ext3_super_block
*) (bh
->b_data
+ offset
);
1658 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1659 if (sb
->s_magic
!= EXT3_SUPER_MAGIC
)
1662 /* Set defaults before we parse the mount options */
1663 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1664 if (def_mount_opts
& EXT3_DEFM_DEBUG
)
1665 set_opt(sbi
->s_mount_opt
, DEBUG
);
1666 if (def_mount_opts
& EXT3_DEFM_BSDGROUPS
)
1667 set_opt(sbi
->s_mount_opt
, GRPID
);
1668 if (def_mount_opts
& EXT3_DEFM_UID16
)
1669 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1670 #ifdef CONFIG_EXT3_FS_XATTR
1671 if (def_mount_opts
& EXT3_DEFM_XATTR_USER
)
1672 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1674 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1675 if (def_mount_opts
& EXT3_DEFM_ACL
)
1676 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1678 if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_DATA
)
1679 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
1680 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_ORDERED
)
1681 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
1682 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_WBACK
)
1683 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
1685 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_PANIC
)
1686 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1687 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_CONTINUE
)
1688 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1690 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1692 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
1693 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
1695 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1697 if (!parse_options ((char *) data
, sb
, &journal_inum
, &journal_devnum
,
1701 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1702 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1704 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
&&
1705 (EXT3_HAS_COMPAT_FEATURE(sb
, ~0U) ||
1706 EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
1707 EXT3_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
1708 ext3_msg(sb
, KERN_WARNING
,
1709 "warning: feature flags set on rev 0 fs, "
1710 "running e2fsck is recommended");
1712 * Check feature flags regardless of the revision level, since we
1713 * previously didn't change the revision level when setting the flags,
1714 * so there is a chance incompat flags are set on a rev 0 filesystem.
1716 features
= EXT3_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
);
1718 ext3_msg(sb
, KERN_ERR
,
1719 "error: couldn't mount because of unsupported "
1720 "optional features (%x)", le32_to_cpu(features
));
1723 features
= EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
);
1724 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1725 ext3_msg(sb
, KERN_ERR
,
1726 "error: couldn't mount RDWR because of unsupported "
1727 "optional features (%x)", le32_to_cpu(features
));
1730 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
1732 if (blocksize
< EXT3_MIN_BLOCK_SIZE
||
1733 blocksize
> EXT3_MAX_BLOCK_SIZE
) {
1734 ext3_msg(sb
, KERN_ERR
,
1735 "error: couldn't mount because of unsupported "
1736 "filesystem blocksize %d", blocksize
);
1740 hblock
= bdev_logical_block_size(sb
->s_bdev
);
1741 if (sb
->s_blocksize
!= blocksize
) {
1743 * Make sure the blocksize for the filesystem is larger
1744 * than the hardware sectorsize for the machine.
1746 if (blocksize
< hblock
) {
1747 ext3_msg(sb
, KERN_ERR
,
1748 "error: fsblocksize %d too small for "
1749 "hardware sectorsize %d", blocksize
, hblock
);
1754 if (!sb_set_blocksize(sb
, blocksize
)) {
1755 ext3_msg(sb
, KERN_ERR
,
1756 "error: bad blocksize %d", blocksize
);
1759 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1760 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1761 bh
= sb_bread(sb
, logic_sb_block
);
1763 ext3_msg(sb
, KERN_ERR
,
1764 "error: can't read superblock on 2nd try");
1767 es
= (struct ext3_super_block
*)(bh
->b_data
+ offset
);
1769 if (es
->s_magic
!= cpu_to_le16(EXT3_SUPER_MAGIC
)) {
1770 ext3_msg(sb
, KERN_ERR
,
1771 "error: magic mismatch");
1776 sb
->s_maxbytes
= ext3_max_size(sb
->s_blocksize_bits
);
1778 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
) {
1779 sbi
->s_inode_size
= EXT3_GOOD_OLD_INODE_SIZE
;
1780 sbi
->s_first_ino
= EXT3_GOOD_OLD_FIRST_INO
;
1782 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
1783 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
1784 if ((sbi
->s_inode_size
< EXT3_GOOD_OLD_INODE_SIZE
) ||
1785 (!is_power_of_2(sbi
->s_inode_size
)) ||
1786 (sbi
->s_inode_size
> blocksize
)) {
1787 ext3_msg(sb
, KERN_ERR
,
1788 "error: unsupported inode size: %d",
1793 sbi
->s_frag_size
= EXT3_MIN_FRAG_SIZE
<<
1794 le32_to_cpu(es
->s_log_frag_size
);
1795 if (blocksize
!= sbi
->s_frag_size
) {
1796 ext3_msg(sb
, KERN_ERR
,
1797 "error: fragsize %lu != blocksize %u (unsupported)",
1798 sbi
->s_frag_size
, blocksize
);
1801 sbi
->s_frags_per_block
= 1;
1802 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
1803 sbi
->s_frags_per_group
= le32_to_cpu(es
->s_frags_per_group
);
1804 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
1805 if (EXT3_INODE_SIZE(sb
) == 0 || EXT3_INODES_PER_GROUP(sb
) == 0)
1807 sbi
->s_inodes_per_block
= blocksize
/ EXT3_INODE_SIZE(sb
);
1808 if (sbi
->s_inodes_per_block
== 0)
1810 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
1811 sbi
->s_inodes_per_block
;
1812 sbi
->s_desc_per_block
= blocksize
/ sizeof(struct ext3_group_desc
);
1814 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
1815 sbi
->s_addr_per_block_bits
= ilog2(EXT3_ADDR_PER_BLOCK(sb
));
1816 sbi
->s_desc_per_block_bits
= ilog2(EXT3_DESC_PER_BLOCK(sb
));
1817 for (i
=0; i
< 4; i
++)
1818 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
1819 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
1820 i
= le32_to_cpu(es
->s_flags
);
1821 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
1822 sbi
->s_hash_unsigned
= 3;
1823 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
1824 #ifdef __CHAR_UNSIGNED__
1825 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
1826 sbi
->s_hash_unsigned
= 3;
1828 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
1832 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
1833 ext3_msg(sb
, KERN_ERR
,
1834 "#blocks per group too big: %lu",
1835 sbi
->s_blocks_per_group
);
1838 if (sbi
->s_frags_per_group
> blocksize
* 8) {
1839 ext3_msg(sb
, KERN_ERR
,
1840 "error: #fragments per group too big: %lu",
1841 sbi
->s_frags_per_group
);
1844 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
1845 ext3_msg(sb
, KERN_ERR
,
1846 "error: #inodes per group too big: %lu",
1847 sbi
->s_inodes_per_group
);
1851 if (generic_check_addressable(sb
->s_blocksize_bits
,
1852 le32_to_cpu(es
->s_blocks_count
))) {
1853 ext3_msg(sb
, KERN_ERR
,
1854 "error: filesystem is too large to mount safely");
1855 if (sizeof(sector_t
) < 8)
1856 ext3_msg(sb
, KERN_ERR
,
1857 "error: CONFIG_LBDAF not enabled");
1861 if (EXT3_BLOCKS_PER_GROUP(sb
) == 0)
1863 sbi
->s_groups_count
= ((le32_to_cpu(es
->s_blocks_count
) -
1864 le32_to_cpu(es
->s_first_data_block
) - 1)
1865 / EXT3_BLOCKS_PER_GROUP(sb
)) + 1;
1866 db_count
= DIV_ROUND_UP(sbi
->s_groups_count
, EXT3_DESC_PER_BLOCK(sb
));
1867 sbi
->s_group_desc
= kmalloc(db_count
* sizeof (struct buffer_head
*),
1869 if (sbi
->s_group_desc
== NULL
) {
1870 ext3_msg(sb
, KERN_ERR
,
1871 "error: not enough memory");
1876 bgl_lock_init(sbi
->s_blockgroup_lock
);
1878 for (i
= 0; i
< db_count
; i
++) {
1879 block
= descriptor_loc(sb
, logic_sb_block
, i
);
1880 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
1881 if (!sbi
->s_group_desc
[i
]) {
1882 ext3_msg(sb
, KERN_ERR
,
1883 "error: can't read group descriptor %d", i
);
1888 if (!ext3_check_descriptors (sb
)) {
1889 ext3_msg(sb
, KERN_ERR
,
1890 "error: group descriptors corrupted");
1893 sbi
->s_gdb_count
= db_count
;
1894 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1895 spin_lock_init(&sbi
->s_next_gen_lock
);
1897 /* per fileystem reservation list head & lock */
1898 spin_lock_init(&sbi
->s_rsv_window_lock
);
1899 sbi
->s_rsv_window_root
= RB_ROOT
;
1900 /* Add a single, static dummy reservation to the start of the
1901 * reservation window list --- it gives us a placeholder for
1902 * append-at-start-of-list which makes the allocation logic
1903 * _much_ simpler. */
1904 sbi
->s_rsv_window_head
.rsv_start
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1905 sbi
->s_rsv_window_head
.rsv_end
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1906 sbi
->s_rsv_window_head
.rsv_alloc_hit
= 0;
1907 sbi
->s_rsv_window_head
.rsv_goal_size
= 0;
1908 ext3_rsv_window_add(sb
, &sbi
->s_rsv_window_head
);
1911 * set up enough so that it can read an inode
1913 sb
->s_op
= &ext3_sops
;
1914 sb
->s_export_op
= &ext3_export_ops
;
1915 sb
->s_xattr
= ext3_xattr_handlers
;
1917 sb
->s_qcop
= &ext3_qctl_operations
;
1918 sb
->dq_op
= &ext3_quota_operations
;
1920 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
1921 mutex_init(&sbi
->s_orphan_lock
);
1922 mutex_init(&sbi
->s_resize_lock
);
1926 needs_recovery
= (es
->s_last_orphan
!= 0 ||
1927 EXT3_HAS_INCOMPAT_FEATURE(sb
,
1928 EXT3_FEATURE_INCOMPAT_RECOVER
));
1931 * The first inode we look at is the journal inode. Don't try
1932 * root first: it may be modified in the journal!
1934 if (!test_opt(sb
, NOLOAD
) &&
1935 EXT3_HAS_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
)) {
1936 if (ext3_load_journal(sb
, es
, journal_devnum
))
1938 } else if (journal_inum
) {
1939 if (ext3_create_journal(sb
, es
, journal_inum
))
1943 ext3_msg(sb
, KERN_ERR
,
1944 "error: no journal found. "
1945 "mounting ext3 over ext2?");
1948 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
1949 ext3_count_free_blocks(sb
));
1951 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
1952 ext3_count_free_inodes(sb
));
1955 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
1956 ext3_count_dirs(sb
));
1959 ext3_msg(sb
, KERN_ERR
, "error: insufficient memory");
1964 /* We have now updated the journal if required, so we can
1965 * validate the data journaling mode. */
1966 switch (test_opt(sb
, DATA_FLAGS
)) {
1968 /* No mode set, assume a default based on the journal
1969 capabilities: ORDERED_DATA if the journal can
1970 cope, else JOURNAL_DATA */
1971 if (journal_check_available_features
1972 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
))
1973 set_opt(sbi
->s_mount_opt
, DEFAULT_DATA_MODE
);
1975 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
1978 case EXT3_MOUNT_ORDERED_DATA
:
1979 case EXT3_MOUNT_WRITEBACK_DATA
:
1980 if (!journal_check_available_features
1981 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
)) {
1982 ext3_msg(sb
, KERN_ERR
,
1983 "error: journal does not support "
1984 "requested data journaling mode");
1992 * The journal_load will have done any necessary log recovery,
1993 * so we can safely mount the rest of the filesystem now.
1996 root
= ext3_iget(sb
, EXT3_ROOT_INO
);
1998 ext3_msg(sb
, KERN_ERR
, "error: get root inode failed");
1999 ret
= PTR_ERR(root
);
2002 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2004 ext3_msg(sb
, KERN_ERR
, "error: corrupt root inode, run e2fsck");
2007 sb
->s_root
= d_alloc_root(root
);
2009 ext3_msg(sb
, KERN_ERR
, "error: get root dentry failed");
2015 ext3_setup_super (sb
, es
, sb
->s_flags
& MS_RDONLY
);
2017 EXT3_SB(sb
)->s_mount_state
|= EXT3_ORPHAN_FS
;
2018 ext3_orphan_cleanup(sb
, es
);
2019 EXT3_SB(sb
)->s_mount_state
&= ~EXT3_ORPHAN_FS
;
2021 ext3_msg(sb
, KERN_INFO
, "recovery complete");
2022 ext3_mark_recovery_complete(sb
, es
);
2023 ext3_msg(sb
, KERN_INFO
, "mounted filesystem with %s data mode",
2024 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_JOURNAL_DATA
? "journal":
2025 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_ORDERED_DATA
? "ordered":
2032 ext3_msg(sb
, KERN_INFO
,
2033 "error: can't find ext3 filesystem on dev %s.",
2038 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2039 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2040 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2041 journal_destroy(sbi
->s_journal
);
2043 for (i
= 0; i
< db_count
; i
++)
2044 brelse(sbi
->s_group_desc
[i
]);
2045 kfree(sbi
->s_group_desc
);
2048 for (i
= 0; i
< MAXQUOTAS
; i
++)
2049 kfree(sbi
->s_qf_names
[i
]);
2051 ext3_blkdev_remove(sbi
);
2054 sb
->s_fs_info
= NULL
;
2055 kfree(sbi
->s_blockgroup_lock
);
2061 * Setup any per-fs journal parameters now. We'll do this both on
2062 * initial mount, once the journal has been initialised but before we've
2063 * done any recovery; and again on any subsequent remount.
2065 static void ext3_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
2067 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2069 if (sbi
->s_commit_interval
)
2070 journal
->j_commit_interval
= sbi
->s_commit_interval
;
2071 /* We could also set up an ext3-specific default for the commit
2072 * interval here, but for now we'll just fall back to the jbd
2075 spin_lock(&journal
->j_state_lock
);
2076 if (test_opt(sb
, BARRIER
))
2077 journal
->j_flags
|= JFS_BARRIER
;
2079 journal
->j_flags
&= ~JFS_BARRIER
;
2080 if (test_opt(sb
, DATA_ERR_ABORT
))
2081 journal
->j_flags
|= JFS_ABORT_ON_SYNCDATA_ERR
;
2083 journal
->j_flags
&= ~JFS_ABORT_ON_SYNCDATA_ERR
;
2084 spin_unlock(&journal
->j_state_lock
);
2087 static journal_t
*ext3_get_journal(struct super_block
*sb
,
2088 unsigned int journal_inum
)
2090 struct inode
*journal_inode
;
2093 /* First, test for the existence of a valid inode on disk. Bad
2094 * things happen if we iget() an unused inode, as the subsequent
2095 * iput() will try to delete it. */
2097 journal_inode
= ext3_iget(sb
, journal_inum
);
2098 if (IS_ERR(journal_inode
)) {
2099 ext3_msg(sb
, KERN_ERR
, "error: no journal found");
2102 if (!journal_inode
->i_nlink
) {
2103 make_bad_inode(journal_inode
);
2104 iput(journal_inode
);
2105 ext3_msg(sb
, KERN_ERR
, "error: journal inode is deleted");
2109 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2110 journal_inode
, journal_inode
->i_size
);
2111 if (!S_ISREG(journal_inode
->i_mode
)) {
2112 ext3_msg(sb
, KERN_ERR
, "error: invalid journal inode");
2113 iput(journal_inode
);
2117 journal
= journal_init_inode(journal_inode
);
2119 ext3_msg(sb
, KERN_ERR
, "error: could not load journal inode");
2120 iput(journal_inode
);
2123 journal
->j_private
= sb
;
2124 ext3_init_journal_params(sb
, journal
);
2128 static journal_t
*ext3_get_dev_journal(struct super_block
*sb
,
2131 struct buffer_head
* bh
;
2135 int hblock
, blocksize
;
2136 ext3_fsblk_t sb_block
;
2137 unsigned long offset
;
2138 struct ext3_super_block
* es
;
2139 struct block_device
*bdev
;
2141 bdev
= ext3_blkdev_get(j_dev
, sb
);
2145 if (bd_claim(bdev
, sb
)) {
2146 ext3_msg(sb
, KERN_ERR
,
2147 "error: failed to claim external journal device");
2148 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
2152 blocksize
= sb
->s_blocksize
;
2153 hblock
= bdev_logical_block_size(bdev
);
2154 if (blocksize
< hblock
) {
2155 ext3_msg(sb
, KERN_ERR
,
2156 "error: blocksize too small for journal device");
2160 sb_block
= EXT3_MIN_BLOCK_SIZE
/ blocksize
;
2161 offset
= EXT3_MIN_BLOCK_SIZE
% blocksize
;
2162 set_blocksize(bdev
, blocksize
);
2163 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
2164 ext3_msg(sb
, KERN_ERR
, "error: couldn't read superblock of "
2165 "external journal");
2169 es
= (struct ext3_super_block
*) (bh
->b_data
+ offset
);
2170 if ((le16_to_cpu(es
->s_magic
) != EXT3_SUPER_MAGIC
) ||
2171 !(le32_to_cpu(es
->s_feature_incompat
) &
2172 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
2173 ext3_msg(sb
, KERN_ERR
, "error: external journal has "
2179 if (memcmp(EXT3_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
2180 ext3_msg(sb
, KERN_ERR
, "error: journal UUID does not match");
2185 len
= le32_to_cpu(es
->s_blocks_count
);
2186 start
= sb_block
+ 1;
2187 brelse(bh
); /* we're done with the superblock */
2189 journal
= journal_init_dev(bdev
, sb
->s_bdev
,
2190 start
, len
, blocksize
);
2192 ext3_msg(sb
, KERN_ERR
,
2193 "error: failed to create device journal");
2196 journal
->j_private
= sb
;
2197 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
2198 wait_on_buffer(journal
->j_sb_buffer
);
2199 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
2200 ext3_msg(sb
, KERN_ERR
, "I/O error on journal device");
2203 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
2204 ext3_msg(sb
, KERN_ERR
,
2205 "error: external journal has more than one "
2206 "user (unsupported) - %d",
2207 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
2210 EXT3_SB(sb
)->journal_bdev
= bdev
;
2211 ext3_init_journal_params(sb
, journal
);
2214 journal_destroy(journal
);
2216 ext3_blkdev_put(bdev
);
2220 static int ext3_load_journal(struct super_block
*sb
,
2221 struct ext3_super_block
*es
,
2222 unsigned long journal_devnum
)
2225 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
2228 int really_read_only
;
2230 if (journal_devnum
&&
2231 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2232 ext3_msg(sb
, KERN_INFO
, "external journal device major/minor "
2233 "numbers have changed");
2234 journal_dev
= new_decode_dev(journal_devnum
);
2236 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
2238 really_read_only
= bdev_read_only(sb
->s_bdev
);
2241 * Are we loading a blank journal or performing recovery after a
2242 * crash? For recovery, we need to check in advance whether we
2243 * can get read-write access to the device.
2246 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
)) {
2247 if (sb
->s_flags
& MS_RDONLY
) {
2248 ext3_msg(sb
, KERN_INFO
,
2249 "recovery required on readonly filesystem");
2250 if (really_read_only
) {
2251 ext3_msg(sb
, KERN_ERR
, "error: write access "
2252 "unavailable, cannot proceed");
2255 ext3_msg(sb
, KERN_INFO
,
2256 "write access will be enabled during recovery");
2260 if (journal_inum
&& journal_dev
) {
2261 ext3_msg(sb
, KERN_ERR
, "error: filesystem has both journal "
2262 "and inode journals");
2267 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
2270 if (!(journal
= ext3_get_dev_journal(sb
, journal_dev
)))
2274 if (!(journal
->j_flags
& JFS_BARRIER
))
2275 printk(KERN_INFO
"EXT3-fs: barriers not enabled\n");
2277 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
2278 err
= journal_update_format(journal
);
2280 ext3_msg(sb
, KERN_ERR
, "error updating journal");
2281 journal_destroy(journal
);
2286 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
))
2287 err
= journal_wipe(journal
, !really_read_only
);
2289 err
= journal_load(journal
);
2292 ext3_msg(sb
, KERN_ERR
, "error loading journal");
2293 journal_destroy(journal
);
2297 EXT3_SB(sb
)->s_journal
= journal
;
2298 ext3_clear_journal_err(sb
, es
);
2300 if (journal_devnum
&&
2301 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2302 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2304 /* Make sure we flush the recovery flag to disk. */
2305 ext3_commit_super(sb
, es
, 1);
2311 static int ext3_create_journal(struct super_block
*sb
,
2312 struct ext3_super_block
*es
,
2313 unsigned int journal_inum
)
2318 if (sb
->s_flags
& MS_RDONLY
) {
2319 ext3_msg(sb
, KERN_ERR
,
2320 "error: readonly filesystem when trying to "
2325 journal
= ext3_get_journal(sb
, journal_inum
);
2329 ext3_msg(sb
, KERN_INFO
, "creating new journal on inode %u",
2332 err
= journal_create(journal
);
2334 ext3_msg(sb
, KERN_ERR
, "error creating journal");
2335 journal_destroy(journal
);
2339 EXT3_SB(sb
)->s_journal
= journal
;
2341 ext3_update_dynamic_rev(sb
);
2342 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2343 EXT3_SET_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
);
2345 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
2347 /* Make sure we flush the recovery flag to disk. */
2348 ext3_commit_super(sb
, es
, 1);
2353 static int ext3_commit_super(struct super_block
*sb
,
2354 struct ext3_super_block
*es
,
2357 struct buffer_head
*sbh
= EXT3_SB(sb
)->s_sbh
;
2363 if (buffer_write_io_error(sbh
)) {
2365 * Oh, dear. A previous attempt to write the
2366 * superblock failed. This could happen because the
2367 * USB device was yanked out. Or it could happen to
2368 * be a transient write error and maybe the block will
2369 * be remapped. Nothing we can do but to retry the
2370 * write and hope for the best.
2372 ext3_msg(sb
, KERN_ERR
, "previous I/O error to "
2373 "superblock detected");
2374 clear_buffer_write_io_error(sbh
);
2375 set_buffer_uptodate(sbh
);
2378 * If the file system is mounted read-only, don't update the
2379 * superblock write time. This avoids updating the superblock
2380 * write time when we are mounting the root file system
2381 * read/only but we need to replay the journal; at that point,
2382 * for people who are east of GMT and who make their clock
2383 * tick in localtime for Windows bug-for-bug compatibility,
2384 * the clock is set in the future, and this will cause e2fsck
2385 * to complain and force a full file system check.
2387 if (!(sb
->s_flags
& MS_RDONLY
))
2388 es
->s_wtime
= cpu_to_le32(get_seconds());
2389 es
->s_free_blocks_count
= cpu_to_le32(ext3_count_free_blocks(sb
));
2390 es
->s_free_inodes_count
= cpu_to_le32(ext3_count_free_inodes(sb
));
2391 BUFFER_TRACE(sbh
, "marking dirty");
2392 mark_buffer_dirty(sbh
);
2394 error
= sync_dirty_buffer(sbh
);
2395 if (buffer_write_io_error(sbh
)) {
2396 ext3_msg(sb
, KERN_ERR
, "I/O error while writing "
2398 clear_buffer_write_io_error(sbh
);
2399 set_buffer_uptodate(sbh
);
2407 * Have we just finished recovery? If so, and if we are mounting (or
2408 * remounting) the filesystem readonly, then we will end up with a
2409 * consistent fs on disk. Record that fact.
2411 static void ext3_mark_recovery_complete(struct super_block
* sb
,
2412 struct ext3_super_block
* es
)
2414 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
2416 journal_lock_updates(journal
);
2417 if (journal_flush(journal
) < 0)
2420 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
) &&
2421 sb
->s_flags
& MS_RDONLY
) {
2422 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2423 ext3_commit_super(sb
, es
, 1);
2427 journal_unlock_updates(journal
);
2431 * If we are mounting (or read-write remounting) a filesystem whose journal
2432 * has recorded an error from a previous lifetime, move that error to the
2433 * main filesystem now.
2435 static void ext3_clear_journal_err(struct super_block
*sb
,
2436 struct ext3_super_block
*es
)
2442 journal
= EXT3_SB(sb
)->s_journal
;
2445 * Now check for any error status which may have been recorded in the
2446 * journal by a prior ext3_error() or ext3_abort()
2449 j_errno
= journal_errno(journal
);
2453 errstr
= ext3_decode_error(sb
, j_errno
, nbuf
);
2454 ext3_warning(sb
, __func__
, "Filesystem error recorded "
2455 "from previous mount: %s", errstr
);
2456 ext3_warning(sb
, __func__
, "Marking fs in need of "
2457 "filesystem check.");
2459 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
2460 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
2461 ext3_commit_super (sb
, es
, 1);
2463 journal_clear_err(journal
);
2468 * Force the running and committing transactions to commit,
2469 * and wait on the commit.
2471 int ext3_force_commit(struct super_block
*sb
)
2476 if (sb
->s_flags
& MS_RDONLY
)
2479 journal
= EXT3_SB(sb
)->s_journal
;
2480 ret
= ext3_journal_force_commit(journal
);
2484 static int ext3_sync_fs(struct super_block
*sb
, int wait
)
2488 if (journal_start_commit(EXT3_SB(sb
)->s_journal
, &target
)) {
2490 log_wait_commit(EXT3_SB(sb
)->s_journal
, target
);
2496 * LVM calls this function before a (read-only) snapshot is created. This
2497 * gives us a chance to flush the journal completely and mark the fs clean.
2499 static int ext3_freeze(struct super_block
*sb
)
2504 if (!(sb
->s_flags
& MS_RDONLY
)) {
2505 journal
= EXT3_SB(sb
)->s_journal
;
2507 /* Now we set up the journal barrier. */
2508 journal_lock_updates(journal
);
2511 * We don't want to clear needs_recovery flag when we failed
2512 * to flush the journal.
2514 error
= journal_flush(journal
);
2518 /* Journal blocked and flushed, clear needs_recovery flag. */
2519 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2520 error
= ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2527 journal_unlock_updates(journal
);
2532 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2533 * flag here, even though the filesystem is not technically dirty yet.
2535 static int ext3_unfreeze(struct super_block
*sb
)
2537 if (!(sb
->s_flags
& MS_RDONLY
)) {
2539 /* Reser the needs_recovery flag before the fs is unlocked. */
2540 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2541 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2543 journal_unlock_updates(EXT3_SB(sb
)->s_journal
);
2548 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
)
2550 struct ext3_super_block
* es
;
2551 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2552 ext3_fsblk_t n_blocks_count
= 0;
2553 unsigned long old_sb_flags
;
2554 struct ext3_mount_options old_opts
;
2555 int enable_quota
= 0;
2561 /* Store the original options */
2563 old_sb_flags
= sb
->s_flags
;
2564 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
2565 old_opts
.s_resuid
= sbi
->s_resuid
;
2566 old_opts
.s_resgid
= sbi
->s_resgid
;
2567 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
2569 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
2570 for (i
= 0; i
< MAXQUOTAS
; i
++)
2571 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
2575 * Allow the "check" option to be passed as a remount option.
2577 if (!parse_options(data
, sb
, NULL
, NULL
, &n_blocks_count
, 1)) {
2582 if (test_opt(sb
, ABORT
))
2583 ext3_abort(sb
, __func__
, "Abort forced by user");
2585 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2586 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
2590 ext3_init_journal_params(sb
, sbi
->s_journal
);
2592 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
2593 n_blocks_count
> le32_to_cpu(es
->s_blocks_count
)) {
2594 if (test_opt(sb
, ABORT
)) {
2599 if (*flags
& MS_RDONLY
) {
2600 err
= dquot_suspend(sb
, -1);
2605 * First of all, the unconditional stuff we have to do
2606 * to disable replay of the journal when we next remount
2608 sb
->s_flags
|= MS_RDONLY
;
2611 * OK, test if we are remounting a valid rw partition
2612 * readonly, and if so set the rdonly flag and then
2613 * mark the partition as valid again.
2615 if (!(es
->s_state
& cpu_to_le16(EXT3_VALID_FS
)) &&
2616 (sbi
->s_mount_state
& EXT3_VALID_FS
))
2617 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
2619 ext3_mark_recovery_complete(sb
, es
);
2622 if ((ret
= EXT3_HAS_RO_COMPAT_FEATURE(sb
,
2623 ~EXT3_FEATURE_RO_COMPAT_SUPP
))) {
2624 ext3_msg(sb
, KERN_WARNING
,
2625 "warning: couldn't remount RDWR "
2626 "because of unsupported optional "
2627 "features (%x)", le32_to_cpu(ret
));
2633 * If we have an unprocessed orphan list hanging
2634 * around from a previously readonly bdev mount,
2635 * require a full umount/remount for now.
2637 if (es
->s_last_orphan
) {
2638 ext3_msg(sb
, KERN_WARNING
, "warning: couldn't "
2639 "remount RDWR because of unprocessed "
2640 "orphan inode list. Please "
2641 "umount/remount instead.");
2647 * Mounting a RDONLY partition read-write, so reread
2648 * and store the current valid flag. (It may have
2649 * been changed by e2fsck since we originally mounted
2652 ext3_clear_journal_err(sb
, es
);
2653 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2654 if ((err
= ext3_group_extend(sb
, es
, n_blocks_count
)))
2656 if (!ext3_setup_super (sb
, es
, 0))
2657 sb
->s_flags
&= ~MS_RDONLY
;
2662 /* Release old quota file names */
2663 for (i
= 0; i
< MAXQUOTAS
; i
++)
2664 if (old_opts
.s_qf_names
[i
] &&
2665 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2666 kfree(old_opts
.s_qf_names
[i
]);
2671 dquot_resume(sb
, -1);
2674 sb
->s_flags
= old_sb_flags
;
2675 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
2676 sbi
->s_resuid
= old_opts
.s_resuid
;
2677 sbi
->s_resgid
= old_opts
.s_resgid
;
2678 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
2680 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
2681 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2682 if (sbi
->s_qf_names
[i
] &&
2683 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2684 kfree(sbi
->s_qf_names
[i
]);
2685 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
2692 static int ext3_statfs (struct dentry
* dentry
, struct kstatfs
* buf
)
2694 struct super_block
*sb
= dentry
->d_sb
;
2695 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2696 struct ext3_super_block
*es
= sbi
->s_es
;
2699 if (test_opt(sb
, MINIX_DF
)) {
2700 sbi
->s_overhead_last
= 0;
2701 } else if (sbi
->s_blocks_last
!= le32_to_cpu(es
->s_blocks_count
)) {
2702 unsigned long ngroups
= sbi
->s_groups_count
, i
;
2703 ext3_fsblk_t overhead
= 0;
2707 * Compute the overhead (FS structures). This is constant
2708 * for a given filesystem unless the number of block groups
2709 * changes so we cache the previous value until it does.
2713 * All of the blocks before first_data_block are
2716 overhead
= le32_to_cpu(es
->s_first_data_block
);
2719 * Add the overhead attributed to the superblock and
2720 * block group descriptors. If the sparse superblocks
2721 * feature is turned on, then not all groups have this.
2723 for (i
= 0; i
< ngroups
; i
++) {
2724 overhead
+= ext3_bg_has_super(sb
, i
) +
2725 ext3_bg_num_gdb(sb
, i
);
2730 * Every block group has an inode bitmap, a block
2731 * bitmap, and an inode table.
2733 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
2734 sbi
->s_overhead_last
= overhead
;
2736 sbi
->s_blocks_last
= le32_to_cpu(es
->s_blocks_count
);
2739 buf
->f_type
= EXT3_SUPER_MAGIC
;
2740 buf
->f_bsize
= sb
->s_blocksize
;
2741 buf
->f_blocks
= le32_to_cpu(es
->s_blocks_count
) - sbi
->s_overhead_last
;
2742 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
);
2743 buf
->f_bavail
= buf
->f_bfree
- le32_to_cpu(es
->s_r_blocks_count
);
2744 if (buf
->f_bfree
< le32_to_cpu(es
->s_r_blocks_count
))
2746 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
2747 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
2748 buf
->f_namelen
= EXT3_NAME_LEN
;
2749 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
2750 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
2751 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
2752 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
2756 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2757 * is locked for write. Otherwise the are possible deadlocks:
2758 * Process 1 Process 2
2759 * ext3_create() quota_sync()
2760 * journal_start() write_dquot()
2761 * dquot_initialize() down(dqio_mutex)
2762 * down(dqio_mutex) journal_start()
2768 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
2770 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
2773 static int ext3_write_dquot(struct dquot
*dquot
)
2777 struct inode
*inode
;
2779 inode
= dquot_to_inode(dquot
);
2780 handle
= ext3_journal_start(inode
,
2781 EXT3_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
2783 return PTR_ERR(handle
);
2784 ret
= dquot_commit(dquot
);
2785 err
= ext3_journal_stop(handle
);
2791 static int ext3_acquire_dquot(struct dquot
*dquot
)
2796 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2797 EXT3_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
2799 return PTR_ERR(handle
);
2800 ret
= dquot_acquire(dquot
);
2801 err
= ext3_journal_stop(handle
);
2807 static int ext3_release_dquot(struct dquot
*dquot
)
2812 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2813 EXT3_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
2814 if (IS_ERR(handle
)) {
2815 /* Release dquot anyway to avoid endless cycle in dqput() */
2816 dquot_release(dquot
);
2817 return PTR_ERR(handle
);
2819 ret
= dquot_release(dquot
);
2820 err
= ext3_journal_stop(handle
);
2826 static int ext3_mark_dquot_dirty(struct dquot
*dquot
)
2828 /* Are we journaling quotas? */
2829 if (EXT3_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
2830 EXT3_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
2831 dquot_mark_dquot_dirty(dquot
);
2832 return ext3_write_dquot(dquot
);
2834 return dquot_mark_dquot_dirty(dquot
);
2838 static int ext3_write_info(struct super_block
*sb
, int type
)
2843 /* Data block + inode block */
2844 handle
= ext3_journal_start(sb
->s_root
->d_inode
, 2);
2846 return PTR_ERR(handle
);
2847 ret
= dquot_commit_info(sb
, type
);
2848 err
= ext3_journal_stop(handle
);
2855 * Turn on quotas during mount time - we need to find
2856 * the quota file and such...
2858 static int ext3_quota_on_mount(struct super_block
*sb
, int type
)
2860 return dquot_quota_on_mount(sb
, EXT3_SB(sb
)->s_qf_names
[type
],
2861 EXT3_SB(sb
)->s_jquota_fmt
, type
);
2865 * Standard function to be called on quota_on
2867 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
,
2873 if (!test_opt(sb
, QUOTA
))
2876 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
2880 /* Quotafile not on the same filesystem? */
2881 if (path
.mnt
->mnt_sb
!= sb
) {
2885 /* Journaling quota? */
2886 if (EXT3_SB(sb
)->s_qf_names
[type
]) {
2887 /* Quotafile not of fs root? */
2888 if (path
.dentry
->d_parent
!= sb
->s_root
)
2889 ext3_msg(sb
, KERN_WARNING
,
2890 "warning: Quota file not on filesystem root. "
2891 "Journaled quota will not work.");
2895 * When we journal data on quota file, we have to flush journal to see
2896 * all updates to the file when we bypass pagecache...
2898 if (ext3_should_journal_data(path
.dentry
->d_inode
)) {
2900 * We don't need to lock updates but journal_flush() could
2901 * otherwise be livelocked...
2903 journal_lock_updates(EXT3_SB(sb
)->s_journal
);
2904 err
= journal_flush(EXT3_SB(sb
)->s_journal
);
2905 journal_unlock_updates(EXT3_SB(sb
)->s_journal
);
2912 err
= dquot_quota_on_path(sb
, type
, format_id
, &path
);
2917 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2918 * acquiring the locks... As quota files are never truncated and quota code
2919 * itself serializes the operations (and noone else should touch the files)
2920 * we don't have to be afraid of races */
2921 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
2922 size_t len
, loff_t off
)
2924 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2925 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2927 int offset
= off
& (sb
->s_blocksize
- 1);
2930 struct buffer_head
*bh
;
2931 loff_t i_size
= i_size_read(inode
);
2935 if (off
+len
> i_size
)
2938 while (toread
> 0) {
2939 tocopy
= sb
->s_blocksize
- offset
< toread
?
2940 sb
->s_blocksize
- offset
: toread
;
2941 bh
= ext3_bread(NULL
, inode
, blk
, 0, &err
);
2944 if (!bh
) /* A hole? */
2945 memset(data
, 0, tocopy
);
2947 memcpy(data
, bh
->b_data
+offset
, tocopy
);
2957 /* Write to quotafile (we know the transaction is already started and has
2958 * enough credits) */
2959 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
2960 const char *data
, size_t len
, loff_t off
)
2962 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2963 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2965 int offset
= off
& (sb
->s_blocksize
- 1);
2966 int journal_quota
= EXT3_SB(sb
)->s_qf_names
[type
] != NULL
;
2967 struct buffer_head
*bh
;
2968 handle_t
*handle
= journal_current_handle();
2971 ext3_msg(sb
, KERN_WARNING
,
2972 "warning: quota write (off=%llu, len=%llu)"
2973 " cancelled because transaction is not started.",
2974 (unsigned long long)off
, (unsigned long long)len
);
2979 * Since we account only one data block in transaction credits,
2980 * then it is impossible to cross a block boundary.
2982 if (sb
->s_blocksize
- offset
< len
) {
2983 ext3_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
2984 " cancelled because not block aligned",
2985 (unsigned long long)off
, (unsigned long long)len
);
2988 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
2989 bh
= ext3_bread(handle
, inode
, blk
, 1, &err
);
2992 if (journal_quota
) {
2993 err
= ext3_journal_get_write_access(handle
, bh
);
3000 memcpy(bh
->b_data
+offset
, data
, len
);
3001 flush_dcache_page(bh
->b_page
);
3004 err
= ext3_journal_dirty_metadata(handle
, bh
);
3006 /* Always do at least ordered writes for quotas */
3007 err
= ext3_journal_dirty_data(handle
, bh
);
3008 mark_buffer_dirty(bh
);
3013 mutex_unlock(&inode
->i_mutex
);
3016 if (inode
->i_size
< off
+ len
) {
3017 i_size_write(inode
, off
+ len
);
3018 EXT3_I(inode
)->i_disksize
= inode
->i_size
;
3021 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3022 ext3_mark_inode_dirty(handle
, inode
);
3023 mutex_unlock(&inode
->i_mutex
);
3029 static struct dentry
*ext3_mount(struct file_system_type
*fs_type
,
3030 int flags
, const char *dev_name
, void *data
)
3032 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext3_fill_super
);
3035 static struct file_system_type ext3_fs_type
= {
3036 .owner
= THIS_MODULE
,
3038 .mount
= ext3_mount
,
3039 .kill_sb
= kill_block_super
,
3040 .fs_flags
= FS_REQUIRES_DEV
,
3043 static int __init
init_ext3_fs(void)
3045 int err
= init_ext3_xattr();
3048 err
= init_inodecache();
3051 err
= register_filesystem(&ext3_fs_type
);
3056 destroy_inodecache();
3062 static void __exit
exit_ext3_fs(void)
3064 unregister_filesystem(&ext3_fs_type
);
3065 destroy_inodecache();
3069 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3070 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3071 MODULE_LICENSE("GPL");
3072 module_init(init_ext3_fs
)
3073 module_exit(exit_ext3_fs
)