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/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
41 #include <asm/uaccess.h>
47 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
48 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
50 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
53 static int ext3_load_journal(struct super_block
*, struct ext3_super_block
*,
54 unsigned long journal_devnum
);
55 static int ext3_create_journal(struct super_block
*, struct ext3_super_block
*,
57 static int ext3_commit_super(struct super_block
*sb
,
58 struct ext3_super_block
*es
,
60 static void ext3_mark_recovery_complete(struct super_block
* sb
,
61 struct ext3_super_block
* es
);
62 static void ext3_clear_journal_err(struct super_block
* sb
,
63 struct ext3_super_block
* es
);
64 static int ext3_sync_fs(struct super_block
*sb
, int wait
);
65 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
67 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
);
68 static int ext3_statfs (struct dentry
* dentry
, struct kstatfs
* buf
);
69 static int ext3_unfreeze(struct super_block
*sb
);
70 static int ext3_freeze(struct super_block
*sb
);
73 * Wrappers for journal_start/end.
75 * The only special thing we need to do here is to make sure that all
76 * journal_end calls result in the superblock being marked dirty, so
77 * that sync() will call the filesystem's write_super callback if
80 handle_t
*ext3_journal_start_sb(struct super_block
*sb
, int nblocks
)
84 if (sb
->s_flags
& MS_RDONLY
)
85 return ERR_PTR(-EROFS
);
87 /* Special case here: if the journal has aborted behind our
88 * backs (eg. EIO in the commit thread), then we still need to
89 * take the FS itself readonly cleanly. */
90 journal
= EXT3_SB(sb
)->s_journal
;
91 if (is_journal_aborted(journal
)) {
92 ext3_abort(sb
, __func__
,
93 "Detected aborted journal");
94 return ERR_PTR(-EROFS
);
97 return journal_start(journal
, nblocks
);
101 * The only special thing we need to do here is to make sure that all
102 * journal_stop calls result in the superblock being marked dirty, so
103 * that sync() will call the filesystem's write_super callback if
106 int __ext3_journal_stop(const char *where
, handle_t
*handle
)
108 struct super_block
*sb
;
112 sb
= handle
->h_transaction
->t_journal
->j_private
;
114 rc
= journal_stop(handle
);
119 __ext3_std_error(sb
, where
, err
);
123 void ext3_journal_abort_handle(const char *caller
, const char *err_fn
,
124 struct buffer_head
*bh
, handle_t
*handle
, int err
)
127 const char *errstr
= ext3_decode_error(NULL
, err
, nbuf
);
130 BUFFER_TRACE(bh
, "abort");
135 if (is_handle_aborted(handle
))
138 printk(KERN_ERR
"EXT3-fs: %s: aborting transaction: %s in %s\n",
139 caller
, errstr
, err_fn
);
141 journal_abort_handle(handle
);
144 void ext3_msg(struct super_block
*sb
, const char *prefix
,
145 const char *fmt
, ...)
150 printk("%sEXT3-fs (%s): ", prefix
, sb
->s_id
);
156 /* Deal with the reporting of failure conditions on a filesystem such as
157 * inconsistencies detected or read IO failures.
159 * On ext2, we can store the error state of the filesystem in the
160 * superblock. That is not possible on ext3, because we may have other
161 * write ordering constraints on the superblock which prevent us from
162 * writing it out straight away; and given that the journal is about to
163 * be aborted, we can't rely on the current, or future, transactions to
164 * write out the superblock safely.
166 * We'll just use the journal_abort() error code to record an error in
167 * the journal instead. On recovery, the journal will complain about
168 * that error until we've noted it down and cleared it.
171 static void ext3_handle_error(struct super_block
*sb
)
173 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
175 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
176 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
178 if (sb
->s_flags
& MS_RDONLY
)
181 if (!test_opt (sb
, ERRORS_CONT
)) {
182 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
184 set_opt(EXT3_SB(sb
)->s_mount_opt
, ABORT
);
186 journal_abort(journal
, -EIO
);
188 if (test_opt (sb
, ERRORS_RO
)) {
189 ext3_msg(sb
, KERN_CRIT
,
190 "error: remounting filesystem read-only");
191 sb
->s_flags
|= MS_RDONLY
;
193 ext3_commit_super(sb
, es
, 1);
194 if (test_opt(sb
, ERRORS_PANIC
))
195 panic("EXT3-fs (%s): panic forced after error\n",
199 void ext3_error (struct super_block
* sb
, const char * function
,
200 const char * fmt
, ...)
205 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
210 ext3_handle_error(sb
);
213 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
220 errstr
= "IO failure";
223 errstr
= "Out of memory";
226 if (!sb
|| EXT3_SB(sb
)->s_journal
->j_flags
& JFS_ABORT
)
227 errstr
= "Journal has aborted";
229 errstr
= "Readonly filesystem";
232 /* If the caller passed in an extra buffer for unknown
233 * errors, textualise them now. Else we just return
236 /* Check for truncated error codes... */
237 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
246 /* __ext3_std_error decodes expected errors from journaling functions
247 * automatically and invokes the appropriate error response. */
249 void __ext3_std_error (struct super_block
* sb
, const char * function
,
255 /* Special case: if the error is EROFS, and we're not already
256 * inside a transaction, then there's really no point in logging
258 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
259 (sb
->s_flags
& MS_RDONLY
))
262 errstr
= ext3_decode_error(sb
, errno
, nbuf
);
263 ext3_msg(sb
, KERN_CRIT
, "error in %s: %s", function
, errstr
);
265 ext3_handle_error(sb
);
269 * ext3_abort is a much stronger failure handler than ext3_error. The
270 * abort function may be used to deal with unrecoverable failures such
271 * as journal IO errors or ENOMEM at a critical moment in log management.
273 * We unconditionally force the filesystem into an ABORT|READONLY state,
274 * unless the error response on the fs has been set to panic in which
275 * case we take the easy way out and panic immediately.
278 void ext3_abort (struct super_block
* sb
, const char * function
,
279 const char * fmt
, ...)
284 printk(KERN_CRIT
"EXT3-fs (%s): error: %s: ", sb
->s_id
, function
);
289 if (test_opt(sb
, ERRORS_PANIC
))
290 panic("EXT3-fs: panic from previous error\n");
292 if (sb
->s_flags
& MS_RDONLY
)
295 ext3_msg(sb
, KERN_CRIT
,
296 "error: remounting filesystem read-only");
297 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
298 sb
->s_flags
|= MS_RDONLY
;
299 set_opt(EXT3_SB(sb
)->s_mount_opt
, ABORT
);
300 if (EXT3_SB(sb
)->s_journal
)
301 journal_abort(EXT3_SB(sb
)->s_journal
, -EIO
);
304 void ext3_warning (struct super_block
* sb
, const char * function
,
305 const char * fmt
, ...)
310 printk(KERN_WARNING
"EXT3-fs (%s): warning: %s: ",
317 void ext3_update_dynamic_rev(struct super_block
*sb
)
319 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
321 if (le32_to_cpu(es
->s_rev_level
) > EXT3_GOOD_OLD_REV
)
324 ext3_msg(sb
, KERN_WARNING
,
325 "warning: updating to rev %d because of "
326 "new feature flag, running e2fsck is recommended",
329 es
->s_first_ino
= cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO
);
330 es
->s_inode_size
= cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE
);
331 es
->s_rev_level
= cpu_to_le32(EXT3_DYNAMIC_REV
);
332 /* leave es->s_feature_*compat flags alone */
333 /* es->s_uuid will be set by e2fsck if empty */
336 * The rest of the superblock fields should be zero, and if not it
337 * means they are likely already in use, so leave them alone. We
338 * can leave it up to e2fsck to clean up any inconsistencies there.
343 * Open the external journal device
345 static struct block_device
*ext3_blkdev_get(dev_t dev
, struct super_block
*sb
)
347 struct block_device
*bdev
;
348 char b
[BDEVNAME_SIZE
];
350 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
356 ext3_msg(sb
, "error: failed to open journal device %s: %ld",
357 __bdevname(dev
, b
), PTR_ERR(bdev
));
363 * Release the journal device
365 static int ext3_blkdev_put(struct block_device
*bdev
)
368 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
371 static int ext3_blkdev_remove(struct ext3_sb_info
*sbi
)
373 struct block_device
*bdev
;
376 bdev
= sbi
->journal_bdev
;
378 ret
= ext3_blkdev_put(bdev
);
379 sbi
->journal_bdev
= NULL
;
384 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
386 return &list_entry(l
, struct ext3_inode_info
, i_orphan
)->vfs_inode
;
389 static void dump_orphan_list(struct super_block
*sb
, struct ext3_sb_info
*sbi
)
393 ext3_msg(sb
, KERN_ERR
, "error: sb orphan head is %d",
394 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
396 ext3_msg(sb
, KERN_ERR
, "sb_info orphan list:");
397 list_for_each(l
, &sbi
->s_orphan
) {
398 struct inode
*inode
= orphan_list_entry(l
);
399 ext3_msg(sb
, KERN_ERR
, " "
400 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
401 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
402 inode
->i_mode
, inode
->i_nlink
,
407 static void ext3_put_super (struct super_block
* sb
)
409 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
410 struct ext3_super_block
*es
= sbi
->s_es
;
413 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
417 ext3_xattr_put_super(sb
);
418 err
= journal_destroy(sbi
->s_journal
);
419 sbi
->s_journal
= NULL
;
421 ext3_abort(sb
, __func__
, "Couldn't clean up the journal");
423 if (!(sb
->s_flags
& MS_RDONLY
)) {
424 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
425 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
426 BUFFER_TRACE(sbi
->s_sbh
, "marking dirty");
427 mark_buffer_dirty(sbi
->s_sbh
);
428 ext3_commit_super(sb
, es
, 1);
431 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
432 brelse(sbi
->s_group_desc
[i
]);
433 kfree(sbi
->s_group_desc
);
434 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
435 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
436 percpu_counter_destroy(&sbi
->s_dirs_counter
);
439 for (i
= 0; i
< MAXQUOTAS
; i
++)
440 kfree(sbi
->s_qf_names
[i
]);
443 /* Debugging code just in case the in-memory inode orphan list
444 * isn't empty. The on-disk one can be non-empty if we've
445 * detected an error and taken the fs readonly, but the
446 * in-memory list had better be clean by this point. */
447 if (!list_empty(&sbi
->s_orphan
))
448 dump_orphan_list(sb
, sbi
);
449 J_ASSERT(list_empty(&sbi
->s_orphan
));
451 invalidate_bdev(sb
->s_bdev
);
452 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
454 * Invalidate the journal device's buffers. We don't want them
455 * floating about in memory - the physical journal device may
456 * hotswapped, and it breaks the `ro-after' testing code.
458 sync_blockdev(sbi
->journal_bdev
);
459 invalidate_bdev(sbi
->journal_bdev
);
460 ext3_blkdev_remove(sbi
);
462 sb
->s_fs_info
= NULL
;
463 kfree(sbi
->s_blockgroup_lock
);
469 static struct kmem_cache
*ext3_inode_cachep
;
472 * Called inside transaction, so use GFP_NOFS
474 static struct inode
*ext3_alloc_inode(struct super_block
*sb
)
476 struct ext3_inode_info
*ei
;
478 ei
= kmem_cache_alloc(ext3_inode_cachep
, GFP_NOFS
);
481 ei
->i_block_alloc_info
= NULL
;
482 ei
->vfs_inode
.i_version
= 1;
483 atomic_set(&ei
->i_datasync_tid
, 0);
484 atomic_set(&ei
->i_sync_tid
, 0);
485 return &ei
->vfs_inode
;
488 static void ext3_destroy_inode(struct inode
*inode
)
490 if (!list_empty(&(EXT3_I(inode
)->i_orphan
))) {
491 printk("EXT3 Inode %p: orphan list check failed!\n",
493 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
494 EXT3_I(inode
), sizeof(struct ext3_inode_info
),
498 kmem_cache_free(ext3_inode_cachep
, EXT3_I(inode
));
501 static void init_once(void *foo
)
503 struct ext3_inode_info
*ei
= (struct ext3_inode_info
*) foo
;
505 INIT_LIST_HEAD(&ei
->i_orphan
);
506 #ifdef CONFIG_EXT3_FS_XATTR
507 init_rwsem(&ei
->xattr_sem
);
509 mutex_init(&ei
->truncate_mutex
);
510 inode_init_once(&ei
->vfs_inode
);
513 static int init_inodecache(void)
515 ext3_inode_cachep
= kmem_cache_create("ext3_inode_cache",
516 sizeof(struct ext3_inode_info
),
517 0, (SLAB_RECLAIM_ACCOUNT
|
520 if (ext3_inode_cachep
== NULL
)
525 static void destroy_inodecache(void)
527 kmem_cache_destroy(ext3_inode_cachep
);
530 static inline void ext3_show_quota_options(struct seq_file
*seq
, struct super_block
*sb
)
532 #if defined(CONFIG_QUOTA)
533 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
535 if (sbi
->s_jquota_fmt
) {
538 switch (sbi
->s_jquota_fmt
) {
549 seq_printf(seq
, ",jqfmt=%s", fmtname
);
552 if (sbi
->s_qf_names
[USRQUOTA
])
553 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
555 if (sbi
->s_qf_names
[GRPQUOTA
])
556 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
558 if (test_opt(sb
, USRQUOTA
))
559 seq_puts(seq
, ",usrquota");
561 if (test_opt(sb
, GRPQUOTA
))
562 seq_puts(seq
, ",grpquota");
566 static char *data_mode_string(unsigned long mode
)
569 case EXT3_MOUNT_JOURNAL_DATA
:
571 case EXT3_MOUNT_ORDERED_DATA
:
573 case EXT3_MOUNT_WRITEBACK_DATA
:
581 * - it's set to a non-default value OR
582 * - if the per-sb default is different from the global default
584 static int ext3_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
586 struct super_block
*sb
= vfs
->mnt_sb
;
587 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
588 struct ext3_super_block
*es
= sbi
->s_es
;
589 unsigned long def_mount_opts
;
591 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
593 if (sbi
->s_sb_block
!= 1)
594 seq_printf(seq
, ",sb=%lu", sbi
->s_sb_block
);
595 if (test_opt(sb
, MINIX_DF
))
596 seq_puts(seq
, ",minixdf");
597 if (test_opt(sb
, GRPID
))
598 seq_puts(seq
, ",grpid");
599 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT3_DEFM_BSDGROUPS
))
600 seq_puts(seq
, ",nogrpid");
601 if (sbi
->s_resuid
!= EXT3_DEF_RESUID
||
602 le16_to_cpu(es
->s_def_resuid
) != EXT3_DEF_RESUID
) {
603 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
605 if (sbi
->s_resgid
!= EXT3_DEF_RESGID
||
606 le16_to_cpu(es
->s_def_resgid
) != EXT3_DEF_RESGID
) {
607 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
609 if (test_opt(sb
, ERRORS_RO
)) {
610 int def_errors
= le16_to_cpu(es
->s_errors
);
612 if (def_errors
== EXT3_ERRORS_PANIC
||
613 def_errors
== EXT3_ERRORS_CONTINUE
) {
614 seq_puts(seq
, ",errors=remount-ro");
617 if (test_opt(sb
, ERRORS_CONT
))
618 seq_puts(seq
, ",errors=continue");
619 if (test_opt(sb
, ERRORS_PANIC
))
620 seq_puts(seq
, ",errors=panic");
621 if (test_opt(sb
, NO_UID32
))
622 seq_puts(seq
, ",nouid32");
623 if (test_opt(sb
, DEBUG
))
624 seq_puts(seq
, ",debug");
625 if (test_opt(sb
, OLDALLOC
))
626 seq_puts(seq
, ",oldalloc");
627 #ifdef CONFIG_EXT3_FS_XATTR
628 if (test_opt(sb
, XATTR_USER
))
629 seq_puts(seq
, ",user_xattr");
630 if (!test_opt(sb
, XATTR_USER
) &&
631 (def_mount_opts
& EXT3_DEFM_XATTR_USER
)) {
632 seq_puts(seq
, ",nouser_xattr");
635 #ifdef CONFIG_EXT3_FS_POSIX_ACL
636 if (test_opt(sb
, POSIX_ACL
))
637 seq_puts(seq
, ",acl");
638 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT3_DEFM_ACL
))
639 seq_puts(seq
, ",noacl");
641 if (!test_opt(sb
, RESERVATION
))
642 seq_puts(seq
, ",noreservation");
643 if (sbi
->s_commit_interval
) {
644 seq_printf(seq
, ",commit=%u",
645 (unsigned) (sbi
->s_commit_interval
/ HZ
));
649 * Always display barrier state so it's clear what the status is.
651 seq_puts(seq
, ",barrier=");
652 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
653 seq_printf(seq
, ",data=%s", data_mode_string(test_opt(sb
, DATA_FLAGS
)));
654 if (test_opt(sb
, DATA_ERR_ABORT
))
655 seq_puts(seq
, ",data_err=abort");
657 if (test_opt(sb
, NOLOAD
))
658 seq_puts(seq
, ",norecovery");
660 ext3_show_quota_options(seq
, sb
);
666 static struct inode
*ext3_nfs_get_inode(struct super_block
*sb
,
667 u64 ino
, u32 generation
)
671 if (ino
< EXT3_FIRST_INO(sb
) && ino
!= EXT3_ROOT_INO
)
672 return ERR_PTR(-ESTALE
);
673 if (ino
> le32_to_cpu(EXT3_SB(sb
)->s_es
->s_inodes_count
))
674 return ERR_PTR(-ESTALE
);
676 /* iget isn't really right if the inode is currently unallocated!!
678 * ext3_read_inode will return a bad_inode if the inode had been
679 * deleted, so we should be safe.
681 * Currently we don't know the generation for parent directory, so
682 * a generation of 0 means "accept any"
684 inode
= ext3_iget(sb
, ino
);
686 return ERR_CAST(inode
);
687 if (generation
&& inode
->i_generation
!= generation
) {
689 return ERR_PTR(-ESTALE
);
695 static struct dentry
*ext3_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
696 int fh_len
, int fh_type
)
698 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
702 static struct dentry
*ext3_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
703 int fh_len
, int fh_type
)
705 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
710 * Try to release metadata pages (indirect blocks, directories) which are
711 * mapped via the block device. Since these pages could have journal heads
712 * which would prevent try_to_free_buffers() from freeing them, we must use
713 * jbd layer's try_to_free_buffers() function to release them.
715 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
718 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
720 WARN_ON(PageChecked(page
));
721 if (!page_has_buffers(page
))
724 return journal_try_to_free_buffers(journal
, page
,
726 return try_to_free_buffers(page
);
730 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
731 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
733 static int ext3_write_dquot(struct dquot
*dquot
);
734 static int ext3_acquire_dquot(struct dquot
*dquot
);
735 static int ext3_release_dquot(struct dquot
*dquot
);
736 static int ext3_mark_dquot_dirty(struct dquot
*dquot
);
737 static int ext3_write_info(struct super_block
*sb
, int type
);
738 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
,
740 static int ext3_quota_on_mount(struct super_block
*sb
, int type
);
741 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
742 size_t len
, loff_t off
);
743 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
744 const char *data
, size_t len
, loff_t off
);
746 static const struct dquot_operations ext3_quota_operations
= {
747 .write_dquot
= ext3_write_dquot
,
748 .acquire_dquot
= ext3_acquire_dquot
,
749 .release_dquot
= ext3_release_dquot
,
750 .mark_dirty
= ext3_mark_dquot_dirty
,
751 .write_info
= ext3_write_info
,
752 .alloc_dquot
= dquot_alloc
,
753 .destroy_dquot
= dquot_destroy
,
756 static const struct quotactl_ops ext3_qctl_operations
= {
757 .quota_on
= ext3_quota_on
,
758 .quota_off
= dquot_quota_off
,
759 .quota_sync
= dquot_quota_sync
,
760 .get_info
= dquot_get_dqinfo
,
761 .set_info
= dquot_set_dqinfo
,
762 .get_dqblk
= dquot_get_dqblk
,
763 .set_dqblk
= dquot_set_dqblk
767 static const struct super_operations ext3_sops
= {
768 .alloc_inode
= ext3_alloc_inode
,
769 .destroy_inode
= ext3_destroy_inode
,
770 .write_inode
= ext3_write_inode
,
771 .dirty_inode
= ext3_dirty_inode
,
772 .evict_inode
= ext3_evict_inode
,
773 .put_super
= ext3_put_super
,
774 .sync_fs
= ext3_sync_fs
,
775 .freeze_fs
= ext3_freeze
,
776 .unfreeze_fs
= ext3_unfreeze
,
777 .statfs
= ext3_statfs
,
778 .remount_fs
= ext3_remount
,
779 .show_options
= ext3_show_options
,
781 .quota_read
= ext3_quota_read
,
782 .quota_write
= ext3_quota_write
,
784 .bdev_try_to_free_page
= bdev_try_to_free_page
,
787 static const struct export_operations ext3_export_ops
= {
788 .fh_to_dentry
= ext3_fh_to_dentry
,
789 .fh_to_parent
= ext3_fh_to_parent
,
790 .get_parent
= ext3_get_parent
,
794 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
795 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
796 Opt_nouid32
, Opt_nocheck
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
797 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
798 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
799 Opt_commit
, Opt_journal_update
, Opt_journal_inum
, Opt_journal_dev
,
800 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
801 Opt_data_err_abort
, Opt_data_err_ignore
,
802 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
803 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
804 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
805 Opt_resize
, Opt_usrquota
, Opt_grpquota
808 static const match_table_t tokens
= {
809 {Opt_bsd_df
, "bsddf"},
810 {Opt_minix_df
, "minixdf"},
811 {Opt_grpid
, "grpid"},
812 {Opt_grpid
, "bsdgroups"},
813 {Opt_nogrpid
, "nogrpid"},
814 {Opt_nogrpid
, "sysvgroups"},
815 {Opt_resgid
, "resgid=%u"},
816 {Opt_resuid
, "resuid=%u"},
818 {Opt_err_cont
, "errors=continue"},
819 {Opt_err_panic
, "errors=panic"},
820 {Opt_err_ro
, "errors=remount-ro"},
821 {Opt_nouid32
, "nouid32"},
822 {Opt_nocheck
, "nocheck"},
823 {Opt_nocheck
, "check=none"},
824 {Opt_debug
, "debug"},
825 {Opt_oldalloc
, "oldalloc"},
826 {Opt_orlov
, "orlov"},
827 {Opt_user_xattr
, "user_xattr"},
828 {Opt_nouser_xattr
, "nouser_xattr"},
830 {Opt_noacl
, "noacl"},
831 {Opt_reservation
, "reservation"},
832 {Opt_noreservation
, "noreservation"},
833 {Opt_noload
, "noload"},
834 {Opt_noload
, "norecovery"},
837 {Opt_commit
, "commit=%u"},
838 {Opt_journal_update
, "journal=update"},
839 {Opt_journal_inum
, "journal=%u"},
840 {Opt_journal_dev
, "journal_dev=%u"},
841 {Opt_abort
, "abort"},
842 {Opt_data_journal
, "data=journal"},
843 {Opt_data_ordered
, "data=ordered"},
844 {Opt_data_writeback
, "data=writeback"},
845 {Opt_data_err_abort
, "data_err=abort"},
846 {Opt_data_err_ignore
, "data_err=ignore"},
847 {Opt_offusrjquota
, "usrjquota="},
848 {Opt_usrjquota
, "usrjquota=%s"},
849 {Opt_offgrpjquota
, "grpjquota="},
850 {Opt_grpjquota
, "grpjquota=%s"},
851 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
852 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
853 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
854 {Opt_grpquota
, "grpquota"},
855 {Opt_noquota
, "noquota"},
856 {Opt_quota
, "quota"},
857 {Opt_usrquota
, "usrquota"},
858 {Opt_barrier
, "barrier=%u"},
859 {Opt_barrier
, "barrier"},
860 {Opt_nobarrier
, "nobarrier"},
861 {Opt_resize
, "resize"},
865 static ext3_fsblk_t
get_sb_block(void **data
, struct super_block
*sb
)
867 ext3_fsblk_t sb_block
;
868 char *options
= (char *) *data
;
870 if (!options
|| strncmp(options
, "sb=", 3) != 0)
871 return 1; /* Default location */
873 /*todo: use simple_strtoll with >32bit ext3 */
874 sb_block
= simple_strtoul(options
, &options
, 0);
875 if (*options
&& *options
!= ',') {
876 ext3_msg(sb
, "error: invalid sb specification: %s",
882 *data
= (void *) options
;
887 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
889 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
892 if (sb_any_quota_loaded(sb
) &&
893 !sbi
->s_qf_names
[qtype
]) {
894 ext3_msg(sb
, KERN_ERR
,
895 "Cannot change journaled "
896 "quota options when quota turned on");
899 qname
= match_strdup(args
);
901 ext3_msg(sb
, KERN_ERR
,
902 "Not enough memory for storing quotafile name");
905 if (sbi
->s_qf_names
[qtype
] &&
906 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
907 ext3_msg(sb
, KERN_ERR
,
908 "%s quota file already specified", QTYPE2NAME(qtype
));
912 sbi
->s_qf_names
[qtype
] = qname
;
913 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
914 ext3_msg(sb
, KERN_ERR
,
915 "quotafile must be on filesystem root");
916 kfree(sbi
->s_qf_names
[qtype
]);
917 sbi
->s_qf_names
[qtype
] = NULL
;
920 set_opt(sbi
->s_mount_opt
, QUOTA
);
924 static int clear_qf_name(struct super_block
*sb
, int qtype
) {
926 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
928 if (sb_any_quota_loaded(sb
) &&
929 sbi
->s_qf_names
[qtype
]) {
930 ext3_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
931 " when quota turned on");
935 * The space will be released later when all options are confirmed
938 sbi
->s_qf_names
[qtype
] = NULL
;
943 static int parse_options (char *options
, struct super_block
*sb
,
944 unsigned int *inum
, unsigned long *journal_devnum
,
945 ext3_fsblk_t
*n_blocks_count
, int is_remount
)
947 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
949 substring_t args
[MAX_OPT_ARGS
];
959 while ((p
= strsep (&options
, ",")) != NULL
) {
964 * Initialize args struct so we know whether arg was
965 * found; some options take optional arguments.
967 args
[0].to
= args
[0].from
= 0;
968 token
= match_token(p
, tokens
, args
);
971 clear_opt (sbi
->s_mount_opt
, MINIX_DF
);
974 set_opt (sbi
->s_mount_opt
, MINIX_DF
);
977 set_opt (sbi
->s_mount_opt
, GRPID
);
980 clear_opt (sbi
->s_mount_opt
, GRPID
);
983 if (match_int(&args
[0], &option
))
985 sbi
->s_resuid
= option
;
988 if (match_int(&args
[0], &option
))
990 sbi
->s_resgid
= option
;
993 /* handled by get_sb_block() instead of here */
994 /* *sb_block = match_int(&args[0]); */
997 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
998 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
999 set_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1002 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1003 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1004 set_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1007 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
1008 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
1009 set_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
1012 set_opt (sbi
->s_mount_opt
, NO_UID32
);
1015 clear_opt (sbi
->s_mount_opt
, CHECK
);
1018 set_opt (sbi
->s_mount_opt
, DEBUG
);
1021 set_opt (sbi
->s_mount_opt
, OLDALLOC
);
1024 clear_opt (sbi
->s_mount_opt
, OLDALLOC
);
1026 #ifdef CONFIG_EXT3_FS_XATTR
1027 case Opt_user_xattr
:
1028 set_opt (sbi
->s_mount_opt
, XATTR_USER
);
1030 case Opt_nouser_xattr
:
1031 clear_opt (sbi
->s_mount_opt
, XATTR_USER
);
1034 case Opt_user_xattr
:
1035 case Opt_nouser_xattr
:
1036 ext3_msg(sb
, KERN_INFO
,
1037 "(no)user_xattr options not supported");
1040 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1042 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1045 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1050 ext3_msg(sb
, KERN_INFO
,
1051 "(no)acl options not supported");
1054 case Opt_reservation
:
1055 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1057 case Opt_noreservation
:
1058 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
1060 case Opt_journal_update
:
1062 /* Eventually we will want to be able to create
1063 a journal file here. For now, only allow the
1064 user to specify an existing inode to be the
1067 ext3_msg(sb
, KERN_ERR
, "error: cannot specify "
1068 "journal on remount");
1071 set_opt (sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1073 case Opt_journal_inum
:
1075 ext3_msg(sb
, KERN_ERR
, "error: cannot specify "
1076 "journal on remount");
1079 if (match_int(&args
[0], &option
))
1083 case Opt_journal_dev
:
1085 ext3_msg(sb
, KERN_ERR
, "error: cannot specify "
1086 "journal on remount");
1089 if (match_int(&args
[0], &option
))
1091 *journal_devnum
= option
;
1094 set_opt (sbi
->s_mount_opt
, NOLOAD
);
1097 if (match_int(&args
[0], &option
))
1102 option
= JBD_DEFAULT_MAX_COMMIT_AGE
;
1103 sbi
->s_commit_interval
= HZ
* option
;
1105 case Opt_data_journal
:
1106 data_opt
= EXT3_MOUNT_JOURNAL_DATA
;
1108 case Opt_data_ordered
:
1109 data_opt
= EXT3_MOUNT_ORDERED_DATA
;
1111 case Opt_data_writeback
:
1112 data_opt
= EXT3_MOUNT_WRITEBACK_DATA
;
1115 if (test_opt(sb
, DATA_FLAGS
) == data_opt
)
1117 ext3_msg(sb
, KERN_ERR
,
1118 "error: cannot change "
1119 "data mode on remount. The filesystem "
1120 "is mounted in data=%s mode and you "
1121 "try to remount it in data=%s mode.",
1122 data_mode_string(test_opt(sb
,
1124 data_mode_string(data_opt
));
1127 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
1128 sbi
->s_mount_opt
|= data_opt
;
1131 case Opt_data_err_abort
:
1132 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1134 case Opt_data_err_ignore
:
1135 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1139 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1143 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1146 case Opt_offusrjquota
:
1147 if (!clear_qf_name(sb
, USRQUOTA
))
1150 case Opt_offgrpjquota
:
1151 if (!clear_qf_name(sb
, GRPQUOTA
))
1154 case Opt_jqfmt_vfsold
:
1155 qfmt
= QFMT_VFS_OLD
;
1157 case Opt_jqfmt_vfsv0
:
1160 case Opt_jqfmt_vfsv1
:
1163 if (sb_any_quota_loaded(sb
) &&
1164 sbi
->s_jquota_fmt
!= qfmt
) {
1165 ext3_msg(sb
, KERN_ERR
, "error: cannot change "
1166 "journaled quota options when "
1167 "quota turned on.");
1170 sbi
->s_jquota_fmt
= qfmt
;
1174 set_opt(sbi
->s_mount_opt
, QUOTA
);
1175 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1178 set_opt(sbi
->s_mount_opt
, QUOTA
);
1179 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1182 if (sb_any_quota_loaded(sb
)) {
1183 ext3_msg(sb
, KERN_ERR
, "error: cannot change "
1184 "quota options when quota turned on.");
1187 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1188 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1189 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1195 ext3_msg(sb
, KERN_ERR
,
1196 "error: quota options not supported.");
1200 case Opt_offusrjquota
:
1201 case Opt_offgrpjquota
:
1202 case Opt_jqfmt_vfsold
:
1203 case Opt_jqfmt_vfsv0
:
1204 case Opt_jqfmt_vfsv1
:
1205 ext3_msg(sb
, KERN_ERR
,
1206 "error: journaled quota options not "
1213 set_opt(sbi
->s_mount_opt
, ABORT
);
1216 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1220 if (match_int(&args
[0], &option
))
1223 option
= 1; /* No argument, default to 1 */
1225 set_opt(sbi
->s_mount_opt
, BARRIER
);
1227 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1233 ext3_msg(sb
, KERN_ERR
,
1234 "error: resize option only available "
1238 if (match_int(&args
[0], &option
) != 0)
1240 *n_blocks_count
= option
;
1243 ext3_msg(sb
, KERN_WARNING
,
1244 "warning: ignoring deprecated nobh option");
1247 ext3_msg(sb
, KERN_WARNING
,
1248 "warning: ignoring deprecated bh option");
1251 ext3_msg(sb
, KERN_ERR
,
1252 "error: unrecognized mount option \"%s\" "
1253 "or missing value", p
);
1258 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1259 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1260 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1261 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1262 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1264 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1265 ext3_msg(sb
, KERN_ERR
, "error: old and new quota "
1270 if (!sbi
->s_jquota_fmt
) {
1271 ext3_msg(sb
, KERN_ERR
, "error: journaled quota format "
1276 if (sbi
->s_jquota_fmt
) {
1277 ext3_msg(sb
, KERN_ERR
, "error: journaled quota format "
1278 "specified with no journaling "
1287 static int ext3_setup_super(struct super_block
*sb
, struct ext3_super_block
*es
,
1290 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1293 if (le32_to_cpu(es
->s_rev_level
) > EXT3_MAX_SUPP_REV
) {
1294 ext3_msg(sb
, KERN_ERR
,
1295 "error: revision level too high, "
1296 "forcing read-only mode");
1301 if (!(sbi
->s_mount_state
& EXT3_VALID_FS
))
1302 ext3_msg(sb
, KERN_WARNING
,
1303 "warning: mounting unchecked fs, "
1304 "running e2fsck is recommended");
1305 else if ((sbi
->s_mount_state
& EXT3_ERROR_FS
))
1306 ext3_msg(sb
, KERN_WARNING
,
1307 "warning: mounting fs with errors, "
1308 "running e2fsck is recommended");
1309 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1310 le16_to_cpu(es
->s_mnt_count
) >=
1311 le16_to_cpu(es
->s_max_mnt_count
))
1312 ext3_msg(sb
, KERN_WARNING
,
1313 "warning: maximal mount count reached, "
1314 "running e2fsck is recommended");
1315 else if (le32_to_cpu(es
->s_checkinterval
) &&
1316 (le32_to_cpu(es
->s_lastcheck
) +
1317 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1318 ext3_msg(sb
, KERN_WARNING
,
1319 "warning: checktime reached, "
1320 "running e2fsck is recommended");
1322 /* @@@ We _will_ want to clear the valid bit if we find
1323 inconsistencies, to force a fsck at reboot. But for
1324 a plain journaled filesystem we can keep it set as
1325 valid forever! :) */
1326 es
->s_state
&= cpu_to_le16(~EXT3_VALID_FS
);
1328 if (!le16_to_cpu(es
->s_max_mnt_count
))
1329 es
->s_max_mnt_count
= cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT
);
1330 le16_add_cpu(&es
->s_mnt_count
, 1);
1331 es
->s_mtime
= cpu_to_le32(get_seconds());
1332 ext3_update_dynamic_rev(sb
);
1333 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
1335 ext3_commit_super(sb
, es
, 1);
1336 if (test_opt(sb
, DEBUG
))
1337 ext3_msg(sb
, KERN_INFO
, "[bs=%lu, gc=%lu, "
1338 "bpg=%lu, ipg=%lu, mo=%04lx]",
1340 sbi
->s_groups_count
,
1341 EXT3_BLOCKS_PER_GROUP(sb
),
1342 EXT3_INODES_PER_GROUP(sb
),
1345 if (EXT3_SB(sb
)->s_journal
->j_inode
== NULL
) {
1346 char b
[BDEVNAME_SIZE
];
1347 ext3_msg(sb
, KERN_INFO
, "using external journal on %s",
1348 bdevname(EXT3_SB(sb
)->s_journal
->j_dev
, b
));
1350 ext3_msg(sb
, KERN_INFO
, "using internal journal");
1355 /* Called at mount-time, super-block is locked */
1356 static int ext3_check_descriptors(struct super_block
*sb
)
1358 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1361 ext3_debug ("Checking group descriptors");
1363 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1364 struct ext3_group_desc
*gdp
= ext3_get_group_desc(sb
, i
, NULL
);
1365 ext3_fsblk_t first_block
= ext3_group_first_block_no(sb
, i
);
1366 ext3_fsblk_t last_block
;
1368 if (i
== sbi
->s_groups_count
- 1)
1369 last_block
= le32_to_cpu(sbi
->s_es
->s_blocks_count
) - 1;
1371 last_block
= first_block
+
1372 (EXT3_BLOCKS_PER_GROUP(sb
) - 1);
1374 if (le32_to_cpu(gdp
->bg_block_bitmap
) < first_block
||
1375 le32_to_cpu(gdp
->bg_block_bitmap
) > last_block
)
1377 ext3_error (sb
, "ext3_check_descriptors",
1378 "Block bitmap for group %d"
1379 " not in group (block %lu)!",
1381 le32_to_cpu(gdp
->bg_block_bitmap
));
1384 if (le32_to_cpu(gdp
->bg_inode_bitmap
) < first_block
||
1385 le32_to_cpu(gdp
->bg_inode_bitmap
) > last_block
)
1387 ext3_error (sb
, "ext3_check_descriptors",
1388 "Inode bitmap for group %d"
1389 " not in group (block %lu)!",
1391 le32_to_cpu(gdp
->bg_inode_bitmap
));
1394 if (le32_to_cpu(gdp
->bg_inode_table
) < first_block
||
1395 le32_to_cpu(gdp
->bg_inode_table
) + sbi
->s_itb_per_group
- 1 >
1398 ext3_error (sb
, "ext3_check_descriptors",
1399 "Inode table for group %d"
1400 " not in group (block %lu)!",
1402 le32_to_cpu(gdp
->bg_inode_table
));
1407 sbi
->s_es
->s_free_blocks_count
=cpu_to_le32(ext3_count_free_blocks(sb
));
1408 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext3_count_free_inodes(sb
));
1413 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1414 * the superblock) which were deleted from all directories, but held open by
1415 * a process at the time of a crash. We walk the list and try to delete these
1416 * inodes at recovery time (only with a read-write filesystem).
1418 * In order to keep the orphan inode chain consistent during traversal (in
1419 * case of crash during recovery), we link each inode into the superblock
1420 * orphan list_head and handle it the same way as an inode deletion during
1421 * normal operation (which journals the operations for us).
1423 * We only do an iget() and an iput() on each inode, which is very safe if we
1424 * accidentally point at an in-use or already deleted inode. The worst that
1425 * can happen in this case is that we get a "bit already cleared" message from
1426 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1427 * e2fsck was run on this filesystem, and it must have already done the orphan
1428 * inode cleanup for us, so we can safely abort without any further action.
1430 static void ext3_orphan_cleanup (struct super_block
* sb
,
1431 struct ext3_super_block
* es
)
1433 unsigned int s_flags
= sb
->s_flags
;
1434 int nr_orphans
= 0, nr_truncates
= 0;
1438 if (!es
->s_last_orphan
) {
1439 jbd_debug(4, "no orphan inodes to clean up\n");
1443 if (bdev_read_only(sb
->s_bdev
)) {
1444 ext3_msg(sb
, KERN_ERR
, "error: write access "
1445 "unavailable, skipping orphan cleanup.");
1449 if (EXT3_SB(sb
)->s_mount_state
& EXT3_ERROR_FS
) {
1450 if (es
->s_last_orphan
)
1451 jbd_debug(1, "Errors on filesystem, "
1452 "clearing orphan list.\n");
1453 es
->s_last_orphan
= 0;
1454 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1458 if (s_flags
& MS_RDONLY
) {
1459 ext3_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
1460 sb
->s_flags
&= ~MS_RDONLY
;
1463 /* Needed for iput() to work correctly and not trash data */
1464 sb
->s_flags
|= MS_ACTIVE
;
1465 /* Turn on quotas so that they are updated correctly */
1466 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1467 if (EXT3_SB(sb
)->s_qf_names
[i
]) {
1468 int ret
= ext3_quota_on_mount(sb
, i
);
1470 ext3_msg(sb
, KERN_ERR
,
1471 "error: cannot turn on journaled "
1477 while (es
->s_last_orphan
) {
1478 struct inode
*inode
;
1480 inode
= ext3_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1481 if (IS_ERR(inode
)) {
1482 es
->s_last_orphan
= 0;
1486 list_add(&EXT3_I(inode
)->i_orphan
, &EXT3_SB(sb
)->s_orphan
);
1487 dquot_initialize(inode
);
1488 if (inode
->i_nlink
) {
1490 "%s: truncating inode %lu to %Ld bytes\n",
1491 __func__
, inode
->i_ino
, inode
->i_size
);
1492 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1493 inode
->i_ino
, inode
->i_size
);
1494 ext3_truncate(inode
);
1498 "%s: deleting unreferenced inode %lu\n",
1499 __func__
, inode
->i_ino
);
1500 jbd_debug(2, "deleting unreferenced inode %lu\n",
1504 iput(inode
); /* The delete magic happens here! */
1507 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1510 ext3_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
1511 PLURAL(nr_orphans
));
1513 ext3_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
1514 PLURAL(nr_truncates
));
1516 /* Turn quotas off */
1517 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1518 if (sb_dqopt(sb
)->files
[i
])
1519 dquot_quota_off(sb
, i
);
1522 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1526 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1527 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1528 * We need to be 1 filesystem block less than the 2^32 sector limit.
1530 static loff_t
ext3_max_size(int bits
)
1532 loff_t res
= EXT3_NDIR_BLOCKS
;
1536 /* This is calculated to be the largest file size for a
1537 * dense, file such that the total number of
1538 * sectors in the file, including data and all indirect blocks,
1539 * does not exceed 2^32 -1
1540 * __u32 i_blocks representing the total number of
1541 * 512 bytes blocks of the file
1543 upper_limit
= (1LL << 32) - 1;
1545 /* total blocks in file system block size */
1546 upper_limit
>>= (bits
- 9);
1549 /* indirect blocks */
1551 /* double indirect blocks */
1552 meta_blocks
+= 1 + (1LL << (bits
-2));
1553 /* tripple indirect blocks */
1554 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
1556 upper_limit
-= meta_blocks
;
1557 upper_limit
<<= bits
;
1559 res
+= 1LL << (bits
-2);
1560 res
+= 1LL << (2*(bits
-2));
1561 res
+= 1LL << (3*(bits
-2));
1563 if (res
> upper_limit
)
1566 if (res
> MAX_LFS_FILESIZE
)
1567 res
= MAX_LFS_FILESIZE
;
1572 static ext3_fsblk_t
descriptor_loc(struct super_block
*sb
,
1573 ext3_fsblk_t logic_sb_block
,
1576 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1577 unsigned long bg
, first_meta_bg
;
1580 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1582 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_META_BG
) ||
1584 return (logic_sb_block
+ nr
+ 1);
1585 bg
= sbi
->s_desc_per_block
* nr
;
1586 if (ext3_bg_has_super(sb
, bg
))
1588 return (has_super
+ ext3_group_first_block_no(sb
, bg
));
1592 static int ext3_fill_super (struct super_block
*sb
, void *data
, int silent
)
1594 struct buffer_head
* bh
;
1595 struct ext3_super_block
*es
= NULL
;
1596 struct ext3_sb_info
*sbi
;
1598 ext3_fsblk_t sb_block
= get_sb_block(&data
, sb
);
1599 ext3_fsblk_t logic_sb_block
;
1600 unsigned long offset
= 0;
1601 unsigned int journal_inum
= 0;
1602 unsigned long journal_devnum
= 0;
1603 unsigned long def_mount_opts
;
1614 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
1618 sbi
->s_blockgroup_lock
=
1619 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
1620 if (!sbi
->s_blockgroup_lock
) {
1624 sb
->s_fs_info
= sbi
;
1625 sbi
->s_mount_opt
= 0;
1626 sbi
->s_resuid
= EXT3_DEF_RESUID
;
1627 sbi
->s_resgid
= EXT3_DEF_RESGID
;
1628 sbi
->s_sb_block
= sb_block
;
1632 blocksize
= sb_min_blocksize(sb
, EXT3_MIN_BLOCK_SIZE
);
1634 ext3_msg(sb
, KERN_ERR
, "error: unable to set blocksize");
1639 * The ext3 superblock will not be buffer aligned for other than 1kB
1640 * block sizes. We need to calculate the offset from buffer start.
1642 if (blocksize
!= EXT3_MIN_BLOCK_SIZE
) {
1643 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1644 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1646 logic_sb_block
= sb_block
;
1649 if (!(bh
= sb_bread(sb
, logic_sb_block
))) {
1650 ext3_msg(sb
, KERN_ERR
, "error: unable to read superblock");
1654 * Note: s_es must be initialized as soon as possible because
1655 * some ext3 macro-instructions depend on its value
1657 es
= (struct ext3_super_block
*) (bh
->b_data
+ offset
);
1659 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1660 if (sb
->s_magic
!= EXT3_SUPER_MAGIC
)
1663 /* Set defaults before we parse the mount options */
1664 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1665 if (def_mount_opts
& EXT3_DEFM_DEBUG
)
1666 set_opt(sbi
->s_mount_opt
, DEBUG
);
1667 if (def_mount_opts
& EXT3_DEFM_BSDGROUPS
)
1668 set_opt(sbi
->s_mount_opt
, GRPID
);
1669 if (def_mount_opts
& EXT3_DEFM_UID16
)
1670 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1671 #ifdef CONFIG_EXT3_FS_XATTR
1672 if (def_mount_opts
& EXT3_DEFM_XATTR_USER
)
1673 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1675 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1676 if (def_mount_opts
& EXT3_DEFM_ACL
)
1677 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1679 if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_DATA
)
1680 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
1681 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_ORDERED
)
1682 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
1683 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_WBACK
)
1684 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
1686 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_PANIC
)
1687 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1688 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_CONTINUE
)
1689 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1691 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1693 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
1694 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
1696 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1698 if (!parse_options ((char *) data
, sb
, &journal_inum
, &journal_devnum
,
1702 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1703 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1705 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
&&
1706 (EXT3_HAS_COMPAT_FEATURE(sb
, ~0U) ||
1707 EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
1708 EXT3_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
1709 ext3_msg(sb
, KERN_WARNING
,
1710 "warning: feature flags set on rev 0 fs, "
1711 "running e2fsck is recommended");
1713 * Check feature flags regardless of the revision level, since we
1714 * previously didn't change the revision level when setting the flags,
1715 * so there is a chance incompat flags are set on a rev 0 filesystem.
1717 features
= EXT3_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
);
1719 ext3_msg(sb
, KERN_ERR
,
1720 "error: couldn't mount because of unsupported "
1721 "optional features (%x)", le32_to_cpu(features
));
1724 features
= EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
);
1725 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1726 ext3_msg(sb
, KERN_ERR
,
1727 "error: couldn't mount RDWR because of unsupported "
1728 "optional features (%x)", le32_to_cpu(features
));
1731 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
1733 if (blocksize
< EXT3_MIN_BLOCK_SIZE
||
1734 blocksize
> EXT3_MAX_BLOCK_SIZE
) {
1735 ext3_msg(sb
, KERN_ERR
,
1736 "error: couldn't mount because of unsupported "
1737 "filesystem blocksize %d", blocksize
);
1741 hblock
= bdev_logical_block_size(sb
->s_bdev
);
1742 if (sb
->s_blocksize
!= blocksize
) {
1744 * Make sure the blocksize for the filesystem is larger
1745 * than the hardware sectorsize for the machine.
1747 if (blocksize
< hblock
) {
1748 ext3_msg(sb
, KERN_ERR
,
1749 "error: fsblocksize %d too small for "
1750 "hardware sectorsize %d", blocksize
, hblock
);
1755 if (!sb_set_blocksize(sb
, blocksize
)) {
1756 ext3_msg(sb
, KERN_ERR
,
1757 "error: bad blocksize %d", blocksize
);
1760 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1761 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1762 bh
= sb_bread(sb
, logic_sb_block
);
1764 ext3_msg(sb
, KERN_ERR
,
1765 "error: can't read superblock on 2nd try");
1768 es
= (struct ext3_super_block
*)(bh
->b_data
+ offset
);
1770 if (es
->s_magic
!= cpu_to_le16(EXT3_SUPER_MAGIC
)) {
1771 ext3_msg(sb
, KERN_ERR
,
1772 "error: magic mismatch");
1777 sb
->s_maxbytes
= ext3_max_size(sb
->s_blocksize_bits
);
1779 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
) {
1780 sbi
->s_inode_size
= EXT3_GOOD_OLD_INODE_SIZE
;
1781 sbi
->s_first_ino
= EXT3_GOOD_OLD_FIRST_INO
;
1783 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
1784 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
1785 if ((sbi
->s_inode_size
< EXT3_GOOD_OLD_INODE_SIZE
) ||
1786 (!is_power_of_2(sbi
->s_inode_size
)) ||
1787 (sbi
->s_inode_size
> blocksize
)) {
1788 ext3_msg(sb
, KERN_ERR
,
1789 "error: unsupported inode size: %d",
1794 sbi
->s_frag_size
= EXT3_MIN_FRAG_SIZE
<<
1795 le32_to_cpu(es
->s_log_frag_size
);
1796 if (blocksize
!= sbi
->s_frag_size
) {
1797 ext3_msg(sb
, KERN_ERR
,
1798 "error: fragsize %lu != blocksize %u (unsupported)",
1799 sbi
->s_frag_size
, blocksize
);
1802 sbi
->s_frags_per_block
= 1;
1803 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
1804 sbi
->s_frags_per_group
= le32_to_cpu(es
->s_frags_per_group
);
1805 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
1806 if (EXT3_INODE_SIZE(sb
) == 0 || EXT3_INODES_PER_GROUP(sb
) == 0)
1808 sbi
->s_inodes_per_block
= blocksize
/ EXT3_INODE_SIZE(sb
);
1809 if (sbi
->s_inodes_per_block
== 0)
1811 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
1812 sbi
->s_inodes_per_block
;
1813 sbi
->s_desc_per_block
= blocksize
/ sizeof(struct ext3_group_desc
);
1815 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
1816 sbi
->s_addr_per_block_bits
= ilog2(EXT3_ADDR_PER_BLOCK(sb
));
1817 sbi
->s_desc_per_block_bits
= ilog2(EXT3_DESC_PER_BLOCK(sb
));
1818 for (i
=0; i
< 4; i
++)
1819 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
1820 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
1821 i
= le32_to_cpu(es
->s_flags
);
1822 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
1823 sbi
->s_hash_unsigned
= 3;
1824 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
1825 #ifdef __CHAR_UNSIGNED__
1826 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
1827 sbi
->s_hash_unsigned
= 3;
1829 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
1833 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
1834 ext3_msg(sb
, KERN_ERR
,
1835 "#blocks per group too big: %lu",
1836 sbi
->s_blocks_per_group
);
1839 if (sbi
->s_frags_per_group
> blocksize
* 8) {
1840 ext3_msg(sb
, KERN_ERR
,
1841 "error: #fragments per group too big: %lu",
1842 sbi
->s_frags_per_group
);
1845 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
1846 ext3_msg(sb
, KERN_ERR
,
1847 "error: #inodes per group too big: %lu",
1848 sbi
->s_inodes_per_group
);
1852 if (le32_to_cpu(es
->s_blocks_count
) >
1853 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) {
1854 ext3_msg(sb
, KERN_ERR
,
1855 "error: filesystem is too large to mount safely");
1856 if (sizeof(sector_t
) < 8)
1857 ext3_msg(sb
, KERN_ERR
,
1858 "error: CONFIG_LBDAF not enabled");
1862 if (EXT3_BLOCKS_PER_GROUP(sb
) == 0)
1864 sbi
->s_groups_count
= ((le32_to_cpu(es
->s_blocks_count
) -
1865 le32_to_cpu(es
->s_first_data_block
) - 1)
1866 / EXT3_BLOCKS_PER_GROUP(sb
)) + 1;
1867 db_count
= (sbi
->s_groups_count
+ EXT3_DESC_PER_BLOCK(sb
) - 1) /
1868 EXT3_DESC_PER_BLOCK(sb
);
1869 sbi
->s_group_desc
= kmalloc(db_count
* sizeof (struct buffer_head
*),
1871 if (sbi
->s_group_desc
== NULL
) {
1872 ext3_msg(sb
, KERN_ERR
,
1873 "error: not enough memory");
1878 bgl_lock_init(sbi
->s_blockgroup_lock
);
1880 for (i
= 0; i
< db_count
; i
++) {
1881 block
= descriptor_loc(sb
, logic_sb_block
, i
);
1882 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
1883 if (!sbi
->s_group_desc
[i
]) {
1884 ext3_msg(sb
, KERN_ERR
,
1885 "error: can't read group descriptor %d", i
);
1890 if (!ext3_check_descriptors (sb
)) {
1891 ext3_msg(sb
, KERN_ERR
,
1892 "error: group descriptors corrupted");
1895 sbi
->s_gdb_count
= db_count
;
1896 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1897 spin_lock_init(&sbi
->s_next_gen_lock
);
1899 /* per fileystem reservation list head & lock */
1900 spin_lock_init(&sbi
->s_rsv_window_lock
);
1901 sbi
->s_rsv_window_root
= RB_ROOT
;
1902 /* Add a single, static dummy reservation to the start of the
1903 * reservation window list --- it gives us a placeholder for
1904 * append-at-start-of-list which makes the allocation logic
1905 * _much_ simpler. */
1906 sbi
->s_rsv_window_head
.rsv_start
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1907 sbi
->s_rsv_window_head
.rsv_end
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1908 sbi
->s_rsv_window_head
.rsv_alloc_hit
= 0;
1909 sbi
->s_rsv_window_head
.rsv_goal_size
= 0;
1910 ext3_rsv_window_add(sb
, &sbi
->s_rsv_window_head
);
1913 * set up enough so that it can read an inode
1915 sb
->s_op
= &ext3_sops
;
1916 sb
->s_export_op
= &ext3_export_ops
;
1917 sb
->s_xattr
= ext3_xattr_handlers
;
1919 sb
->s_qcop
= &ext3_qctl_operations
;
1920 sb
->dq_op
= &ext3_quota_operations
;
1922 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
1923 mutex_init(&sbi
->s_orphan_lock
);
1924 mutex_init(&sbi
->s_resize_lock
);
1928 needs_recovery
= (es
->s_last_orphan
!= 0 ||
1929 EXT3_HAS_INCOMPAT_FEATURE(sb
,
1930 EXT3_FEATURE_INCOMPAT_RECOVER
));
1933 * The first inode we look at is the journal inode. Don't try
1934 * root first: it may be modified in the journal!
1936 if (!test_opt(sb
, NOLOAD
) &&
1937 EXT3_HAS_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
)) {
1938 if (ext3_load_journal(sb
, es
, journal_devnum
))
1940 } else if (journal_inum
) {
1941 if (ext3_create_journal(sb
, es
, journal_inum
))
1945 ext3_msg(sb
, KERN_ERR
,
1946 "error: no journal found. "
1947 "mounting ext3 over ext2?");
1950 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
1951 ext3_count_free_blocks(sb
));
1953 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
1954 ext3_count_free_inodes(sb
));
1957 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
1958 ext3_count_dirs(sb
));
1961 ext3_msg(sb
, KERN_ERR
, "error: insufficient memory");
1966 /* We have now updated the journal if required, so we can
1967 * validate the data journaling mode. */
1968 switch (test_opt(sb
, DATA_FLAGS
)) {
1970 /* No mode set, assume a default based on the journal
1971 capabilities: ORDERED_DATA if the journal can
1972 cope, else JOURNAL_DATA */
1973 if (journal_check_available_features
1974 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
))
1975 set_opt(sbi
->s_mount_opt
, DEFAULT_DATA_MODE
);
1977 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
1980 case EXT3_MOUNT_ORDERED_DATA
:
1981 case EXT3_MOUNT_WRITEBACK_DATA
:
1982 if (!journal_check_available_features
1983 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
)) {
1984 ext3_msg(sb
, KERN_ERR
,
1985 "error: journal does not support "
1986 "requested data journaling mode");
1994 * The journal_load will have done any necessary log recovery,
1995 * so we can safely mount the rest of the filesystem now.
1998 root
= ext3_iget(sb
, EXT3_ROOT_INO
);
2000 ext3_msg(sb
, KERN_ERR
, "error: get root inode failed");
2001 ret
= PTR_ERR(root
);
2004 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2006 ext3_msg(sb
, KERN_ERR
, "error: corrupt root inode, run e2fsck");
2009 sb
->s_root
= d_alloc_root(root
);
2011 ext3_msg(sb
, KERN_ERR
, "error: get root dentry failed");
2017 ext3_setup_super (sb
, es
, sb
->s_flags
& MS_RDONLY
);
2019 EXT3_SB(sb
)->s_mount_state
|= EXT3_ORPHAN_FS
;
2020 ext3_orphan_cleanup(sb
, es
);
2021 EXT3_SB(sb
)->s_mount_state
&= ~EXT3_ORPHAN_FS
;
2023 ext3_msg(sb
, KERN_INFO
, "recovery complete");
2024 ext3_mark_recovery_complete(sb
, es
);
2025 ext3_msg(sb
, KERN_INFO
, "mounted filesystem with %s data mode",
2026 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_JOURNAL_DATA
? "journal":
2027 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_ORDERED_DATA
? "ordered":
2035 ext3_msg(sb
, KERN_INFO
,
2036 "error: can't find ext3 filesystem on dev %s.",
2041 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2042 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2043 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2044 journal_destroy(sbi
->s_journal
);
2046 for (i
= 0; i
< db_count
; i
++)
2047 brelse(sbi
->s_group_desc
[i
]);
2048 kfree(sbi
->s_group_desc
);
2051 for (i
= 0; i
< MAXQUOTAS
; i
++)
2052 kfree(sbi
->s_qf_names
[i
]);
2054 ext3_blkdev_remove(sbi
);
2057 sb
->s_fs_info
= NULL
;
2058 kfree(sbi
->s_blockgroup_lock
);
2065 * Setup any per-fs journal parameters now. We'll do this both on
2066 * initial mount, once the journal has been initialised but before we've
2067 * done any recovery; and again on any subsequent remount.
2069 static void ext3_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
2071 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2073 if (sbi
->s_commit_interval
)
2074 journal
->j_commit_interval
= sbi
->s_commit_interval
;
2075 /* We could also set up an ext3-specific default for the commit
2076 * interval here, but for now we'll just fall back to the jbd
2079 spin_lock(&journal
->j_state_lock
);
2080 if (test_opt(sb
, BARRIER
))
2081 journal
->j_flags
|= JFS_BARRIER
;
2083 journal
->j_flags
&= ~JFS_BARRIER
;
2084 if (test_opt(sb
, DATA_ERR_ABORT
))
2085 journal
->j_flags
|= JFS_ABORT_ON_SYNCDATA_ERR
;
2087 journal
->j_flags
&= ~JFS_ABORT_ON_SYNCDATA_ERR
;
2088 spin_unlock(&journal
->j_state_lock
);
2091 static journal_t
*ext3_get_journal(struct super_block
*sb
,
2092 unsigned int journal_inum
)
2094 struct inode
*journal_inode
;
2097 /* First, test for the existence of a valid inode on disk. Bad
2098 * things happen if we iget() an unused inode, as the subsequent
2099 * iput() will try to delete it. */
2101 journal_inode
= ext3_iget(sb
, journal_inum
);
2102 if (IS_ERR(journal_inode
)) {
2103 ext3_msg(sb
, KERN_ERR
, "error: no journal found");
2106 if (!journal_inode
->i_nlink
) {
2107 make_bad_inode(journal_inode
);
2108 iput(journal_inode
);
2109 ext3_msg(sb
, KERN_ERR
, "error: journal inode is deleted");
2113 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2114 journal_inode
, journal_inode
->i_size
);
2115 if (!S_ISREG(journal_inode
->i_mode
)) {
2116 ext3_msg(sb
, KERN_ERR
, "error: invalid journal inode");
2117 iput(journal_inode
);
2121 journal
= journal_init_inode(journal_inode
);
2123 ext3_msg(sb
, KERN_ERR
, "error: could not load journal inode");
2124 iput(journal_inode
);
2127 journal
->j_private
= sb
;
2128 ext3_init_journal_params(sb
, journal
);
2132 static journal_t
*ext3_get_dev_journal(struct super_block
*sb
,
2135 struct buffer_head
* bh
;
2139 int hblock
, blocksize
;
2140 ext3_fsblk_t sb_block
;
2141 unsigned long offset
;
2142 struct ext3_super_block
* es
;
2143 struct block_device
*bdev
;
2145 bdev
= ext3_blkdev_get(j_dev
, sb
);
2149 if (bd_claim(bdev
, sb
)) {
2150 ext3_msg(sb
, KERN_ERR
,
2151 "error: failed to claim external journal device");
2152 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
2156 blocksize
= sb
->s_blocksize
;
2157 hblock
= bdev_logical_block_size(bdev
);
2158 if (blocksize
< hblock
) {
2159 ext3_msg(sb
, KERN_ERR
,
2160 "error: blocksize too small for journal device");
2164 sb_block
= EXT3_MIN_BLOCK_SIZE
/ blocksize
;
2165 offset
= EXT3_MIN_BLOCK_SIZE
% blocksize
;
2166 set_blocksize(bdev
, blocksize
);
2167 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
2168 ext3_msg(sb
, KERN_ERR
, "error: couldn't read superblock of "
2169 "external journal");
2173 es
= (struct ext3_super_block
*) (bh
->b_data
+ offset
);
2174 if ((le16_to_cpu(es
->s_magic
) != EXT3_SUPER_MAGIC
) ||
2175 !(le32_to_cpu(es
->s_feature_incompat
) &
2176 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
2177 ext3_msg(sb
, KERN_ERR
, "error: external journal has "
2183 if (memcmp(EXT3_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
2184 ext3_msg(sb
, KERN_ERR
, "error: journal UUID does not match");
2189 len
= le32_to_cpu(es
->s_blocks_count
);
2190 start
= sb_block
+ 1;
2191 brelse(bh
); /* we're done with the superblock */
2193 journal
= journal_init_dev(bdev
, sb
->s_bdev
,
2194 start
, len
, blocksize
);
2196 ext3_msg(sb
, KERN_ERR
,
2197 "error: failed to create device journal");
2200 journal
->j_private
= sb
;
2201 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
2202 wait_on_buffer(journal
->j_sb_buffer
);
2203 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
2204 ext3_msg(sb
, KERN_ERR
, "I/O error on journal device");
2207 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
2208 ext3_msg(sb
, KERN_ERR
,
2209 "error: external journal has more than one "
2210 "user (unsupported) - %d",
2211 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
2214 EXT3_SB(sb
)->journal_bdev
= bdev
;
2215 ext3_init_journal_params(sb
, journal
);
2218 journal_destroy(journal
);
2220 ext3_blkdev_put(bdev
);
2224 static int ext3_load_journal(struct super_block
*sb
,
2225 struct ext3_super_block
*es
,
2226 unsigned long journal_devnum
)
2229 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
2232 int really_read_only
;
2234 if (journal_devnum
&&
2235 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2236 ext3_msg(sb
, KERN_INFO
, "external journal device major/minor "
2237 "numbers have changed");
2238 journal_dev
= new_decode_dev(journal_devnum
);
2240 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
2242 really_read_only
= bdev_read_only(sb
->s_bdev
);
2245 * Are we loading a blank journal or performing recovery after a
2246 * crash? For recovery, we need to check in advance whether we
2247 * can get read-write access to the device.
2250 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
)) {
2251 if (sb
->s_flags
& MS_RDONLY
) {
2252 ext3_msg(sb
, KERN_INFO
,
2253 "recovery required on readonly filesystem");
2254 if (really_read_only
) {
2255 ext3_msg(sb
, KERN_ERR
, "error: write access "
2256 "unavailable, cannot proceed");
2259 ext3_msg(sb
, KERN_INFO
,
2260 "write access will be enabled during recovery");
2264 if (journal_inum
&& journal_dev
) {
2265 ext3_msg(sb
, KERN_ERR
, "error: filesystem has both journal "
2266 "and inode journals");
2271 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
2274 if (!(journal
= ext3_get_dev_journal(sb
, journal_dev
)))
2278 if (!(journal
->j_flags
& JFS_BARRIER
))
2279 printk(KERN_INFO
"EXT3-fs: barriers not enabled\n");
2281 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
2282 err
= journal_update_format(journal
);
2284 ext3_msg(sb
, KERN_ERR
, "error updating journal");
2285 journal_destroy(journal
);
2290 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
))
2291 err
= journal_wipe(journal
, !really_read_only
);
2293 err
= journal_load(journal
);
2296 ext3_msg(sb
, KERN_ERR
, "error loading journal");
2297 journal_destroy(journal
);
2301 EXT3_SB(sb
)->s_journal
= journal
;
2302 ext3_clear_journal_err(sb
, es
);
2304 if (journal_devnum
&&
2305 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2306 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2308 /* Make sure we flush the recovery flag to disk. */
2309 ext3_commit_super(sb
, es
, 1);
2315 static int ext3_create_journal(struct super_block
*sb
,
2316 struct ext3_super_block
*es
,
2317 unsigned int journal_inum
)
2322 if (sb
->s_flags
& MS_RDONLY
) {
2323 ext3_msg(sb
, KERN_ERR
,
2324 "error: readonly filesystem when trying to "
2329 journal
= ext3_get_journal(sb
, journal_inum
);
2333 ext3_msg(sb
, KERN_INFO
, "creating new journal on inode %u",
2336 err
= journal_create(journal
);
2338 ext3_msg(sb
, KERN_ERR
, "error creating journal");
2339 journal_destroy(journal
);
2343 EXT3_SB(sb
)->s_journal
= journal
;
2345 ext3_update_dynamic_rev(sb
);
2346 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2347 EXT3_SET_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
);
2349 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
2351 /* Make sure we flush the recovery flag to disk. */
2352 ext3_commit_super(sb
, es
, 1);
2357 static int ext3_commit_super(struct super_block
*sb
,
2358 struct ext3_super_block
*es
,
2361 struct buffer_head
*sbh
= EXT3_SB(sb
)->s_sbh
;
2367 if (buffer_write_io_error(sbh
)) {
2369 * Oh, dear. A previous attempt to write the
2370 * superblock failed. This could happen because the
2371 * USB device was yanked out. Or it could happen to
2372 * be a transient write error and maybe the block will
2373 * be remapped. Nothing we can do but to retry the
2374 * write and hope for the best.
2376 ext3_msg(sb
, KERN_ERR
, "previous I/O error to "
2377 "superblock detected");
2378 clear_buffer_write_io_error(sbh
);
2379 set_buffer_uptodate(sbh
);
2382 * If the file system is mounted read-only, don't update the
2383 * superblock write time. This avoids updating the superblock
2384 * write time when we are mounting the root file system
2385 * read/only but we need to replay the journal; at that point,
2386 * for people who are east of GMT and who make their clock
2387 * tick in localtime for Windows bug-for-bug compatibility,
2388 * the clock is set in the future, and this will cause e2fsck
2389 * to complain and force a full file system check.
2391 if (!(sb
->s_flags
& MS_RDONLY
))
2392 es
->s_wtime
= cpu_to_le32(get_seconds());
2393 es
->s_free_blocks_count
= cpu_to_le32(ext3_count_free_blocks(sb
));
2394 es
->s_free_inodes_count
= cpu_to_le32(ext3_count_free_inodes(sb
));
2395 BUFFER_TRACE(sbh
, "marking dirty");
2396 mark_buffer_dirty(sbh
);
2398 error
= sync_dirty_buffer(sbh
);
2399 if (buffer_write_io_error(sbh
)) {
2400 ext3_msg(sb
, KERN_ERR
, "I/O error while writing "
2402 clear_buffer_write_io_error(sbh
);
2403 set_buffer_uptodate(sbh
);
2411 * Have we just finished recovery? If so, and if we are mounting (or
2412 * remounting) the filesystem readonly, then we will end up with a
2413 * consistent fs on disk. Record that fact.
2415 static void ext3_mark_recovery_complete(struct super_block
* sb
,
2416 struct ext3_super_block
* es
)
2418 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
2420 journal_lock_updates(journal
);
2421 if (journal_flush(journal
) < 0)
2424 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
) &&
2425 sb
->s_flags
& MS_RDONLY
) {
2426 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2427 ext3_commit_super(sb
, es
, 1);
2431 journal_unlock_updates(journal
);
2435 * If we are mounting (or read-write remounting) a filesystem whose journal
2436 * has recorded an error from a previous lifetime, move that error to the
2437 * main filesystem now.
2439 static void ext3_clear_journal_err(struct super_block
*sb
,
2440 struct ext3_super_block
*es
)
2446 journal
= EXT3_SB(sb
)->s_journal
;
2449 * Now check for any error status which may have been recorded in the
2450 * journal by a prior ext3_error() or ext3_abort()
2453 j_errno
= journal_errno(journal
);
2457 errstr
= ext3_decode_error(sb
, j_errno
, nbuf
);
2458 ext3_warning(sb
, __func__
, "Filesystem error recorded "
2459 "from previous mount: %s", errstr
);
2460 ext3_warning(sb
, __func__
, "Marking fs in need of "
2461 "filesystem check.");
2463 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
2464 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
2465 ext3_commit_super (sb
, es
, 1);
2467 journal_clear_err(journal
);
2472 * Force the running and committing transactions to commit,
2473 * and wait on the commit.
2475 int ext3_force_commit(struct super_block
*sb
)
2480 if (sb
->s_flags
& MS_RDONLY
)
2483 journal
= EXT3_SB(sb
)->s_journal
;
2484 ret
= ext3_journal_force_commit(journal
);
2488 static int ext3_sync_fs(struct super_block
*sb
, int wait
)
2492 if (journal_start_commit(EXT3_SB(sb
)->s_journal
, &target
)) {
2494 log_wait_commit(EXT3_SB(sb
)->s_journal
, target
);
2500 * LVM calls this function before a (read-only) snapshot is created. This
2501 * gives us a chance to flush the journal completely and mark the fs clean.
2503 static int ext3_freeze(struct super_block
*sb
)
2508 if (!(sb
->s_flags
& MS_RDONLY
)) {
2509 journal
= EXT3_SB(sb
)->s_journal
;
2511 /* Now we set up the journal barrier. */
2512 journal_lock_updates(journal
);
2515 * We don't want to clear needs_recovery flag when we failed
2516 * to flush the journal.
2518 error
= journal_flush(journal
);
2522 /* Journal blocked and flushed, clear needs_recovery flag. */
2523 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2524 error
= ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2531 journal_unlock_updates(journal
);
2536 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2537 * flag here, even though the filesystem is not technically dirty yet.
2539 static int ext3_unfreeze(struct super_block
*sb
)
2541 if (!(sb
->s_flags
& MS_RDONLY
)) {
2543 /* Reser the needs_recovery flag before the fs is unlocked. */
2544 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2545 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2547 journal_unlock_updates(EXT3_SB(sb
)->s_journal
);
2552 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
)
2554 struct ext3_super_block
* es
;
2555 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2556 ext3_fsblk_t n_blocks_count
= 0;
2557 unsigned long old_sb_flags
;
2558 struct ext3_mount_options old_opts
;
2559 int enable_quota
= 0;
2567 /* Store the original options */
2569 old_sb_flags
= sb
->s_flags
;
2570 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
2571 old_opts
.s_resuid
= sbi
->s_resuid
;
2572 old_opts
.s_resgid
= sbi
->s_resgid
;
2573 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
2575 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
2576 for (i
= 0; i
< MAXQUOTAS
; i
++)
2577 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
2581 * Allow the "check" option to be passed as a remount option.
2583 if (!parse_options(data
, sb
, NULL
, NULL
, &n_blocks_count
, 1)) {
2588 if (test_opt(sb
, ABORT
))
2589 ext3_abort(sb
, __func__
, "Abort forced by user");
2591 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2592 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
2596 ext3_init_journal_params(sb
, sbi
->s_journal
);
2598 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
2599 n_blocks_count
> le32_to_cpu(es
->s_blocks_count
)) {
2600 if (test_opt(sb
, ABORT
)) {
2605 if (*flags
& MS_RDONLY
) {
2606 err
= dquot_suspend(sb
, -1);
2611 * First of all, the unconditional stuff we have to do
2612 * to disable replay of the journal when we next remount
2614 sb
->s_flags
|= MS_RDONLY
;
2617 * OK, test if we are remounting a valid rw partition
2618 * readonly, and if so set the rdonly flag and then
2619 * mark the partition as valid again.
2621 if (!(es
->s_state
& cpu_to_le16(EXT3_VALID_FS
)) &&
2622 (sbi
->s_mount_state
& EXT3_VALID_FS
))
2623 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
2625 ext3_mark_recovery_complete(sb
, es
);
2628 if ((ret
= EXT3_HAS_RO_COMPAT_FEATURE(sb
,
2629 ~EXT3_FEATURE_RO_COMPAT_SUPP
))) {
2630 ext3_msg(sb
, KERN_WARNING
,
2631 "warning: couldn't remount RDWR "
2632 "because of unsupported optional "
2633 "features (%x)", le32_to_cpu(ret
));
2639 * If we have an unprocessed orphan list hanging
2640 * around from a previously readonly bdev mount,
2641 * require a full umount/remount for now.
2643 if (es
->s_last_orphan
) {
2644 ext3_msg(sb
, KERN_WARNING
, "warning: couldn't "
2645 "remount RDWR because of unprocessed "
2646 "orphan inode list. Please "
2647 "umount/remount instead.");
2653 * Mounting a RDONLY partition read-write, so reread
2654 * and store the current valid flag. (It may have
2655 * been changed by e2fsck since we originally mounted
2658 ext3_clear_journal_err(sb
, es
);
2659 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2660 if ((err
= ext3_group_extend(sb
, es
, n_blocks_count
)))
2662 if (!ext3_setup_super (sb
, es
, 0))
2663 sb
->s_flags
&= ~MS_RDONLY
;
2668 /* Release old quota file names */
2669 for (i
= 0; i
< MAXQUOTAS
; i
++)
2670 if (old_opts
.s_qf_names
[i
] &&
2671 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2672 kfree(old_opts
.s_qf_names
[i
]);
2678 dquot_resume(sb
, -1);
2681 sb
->s_flags
= old_sb_flags
;
2682 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
2683 sbi
->s_resuid
= old_opts
.s_resuid
;
2684 sbi
->s_resgid
= old_opts
.s_resgid
;
2685 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
2687 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
2688 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2689 if (sbi
->s_qf_names
[i
] &&
2690 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2691 kfree(sbi
->s_qf_names
[i
]);
2692 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
2700 static int ext3_statfs (struct dentry
* dentry
, struct kstatfs
* buf
)
2702 struct super_block
*sb
= dentry
->d_sb
;
2703 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2704 struct ext3_super_block
*es
= sbi
->s_es
;
2707 if (test_opt(sb
, MINIX_DF
)) {
2708 sbi
->s_overhead_last
= 0;
2709 } else if (sbi
->s_blocks_last
!= le32_to_cpu(es
->s_blocks_count
)) {
2710 unsigned long ngroups
= sbi
->s_groups_count
, i
;
2711 ext3_fsblk_t overhead
= 0;
2715 * Compute the overhead (FS structures). This is constant
2716 * for a given filesystem unless the number of block groups
2717 * changes so we cache the previous value until it does.
2721 * All of the blocks before first_data_block are
2724 overhead
= le32_to_cpu(es
->s_first_data_block
);
2727 * Add the overhead attributed to the superblock and
2728 * block group descriptors. If the sparse superblocks
2729 * feature is turned on, then not all groups have this.
2731 for (i
= 0; i
< ngroups
; i
++) {
2732 overhead
+= ext3_bg_has_super(sb
, i
) +
2733 ext3_bg_num_gdb(sb
, i
);
2738 * Every block group has an inode bitmap, a block
2739 * bitmap, and an inode table.
2741 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
2742 sbi
->s_overhead_last
= overhead
;
2744 sbi
->s_blocks_last
= le32_to_cpu(es
->s_blocks_count
);
2747 buf
->f_type
= EXT3_SUPER_MAGIC
;
2748 buf
->f_bsize
= sb
->s_blocksize
;
2749 buf
->f_blocks
= le32_to_cpu(es
->s_blocks_count
) - sbi
->s_overhead_last
;
2750 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
);
2751 buf
->f_bavail
= buf
->f_bfree
- le32_to_cpu(es
->s_r_blocks_count
);
2752 if (buf
->f_bfree
< le32_to_cpu(es
->s_r_blocks_count
))
2754 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
2755 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
2756 buf
->f_namelen
= EXT3_NAME_LEN
;
2757 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
2758 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
2759 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
2760 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
2764 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2765 * is locked for write. Otherwise the are possible deadlocks:
2766 * Process 1 Process 2
2767 * ext3_create() quota_sync()
2768 * journal_start() write_dquot()
2769 * dquot_initialize() down(dqio_mutex)
2770 * down(dqio_mutex) journal_start()
2776 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
2778 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
2781 static int ext3_write_dquot(struct dquot
*dquot
)
2785 struct inode
*inode
;
2787 inode
= dquot_to_inode(dquot
);
2788 handle
= ext3_journal_start(inode
,
2789 EXT3_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
2791 return PTR_ERR(handle
);
2792 ret
= dquot_commit(dquot
);
2793 err
= ext3_journal_stop(handle
);
2799 static int ext3_acquire_dquot(struct dquot
*dquot
)
2804 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2805 EXT3_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
2807 return PTR_ERR(handle
);
2808 ret
= dquot_acquire(dquot
);
2809 err
= ext3_journal_stop(handle
);
2815 static int ext3_release_dquot(struct dquot
*dquot
)
2820 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2821 EXT3_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
2822 if (IS_ERR(handle
)) {
2823 /* Release dquot anyway to avoid endless cycle in dqput() */
2824 dquot_release(dquot
);
2825 return PTR_ERR(handle
);
2827 ret
= dquot_release(dquot
);
2828 err
= ext3_journal_stop(handle
);
2834 static int ext3_mark_dquot_dirty(struct dquot
*dquot
)
2836 /* Are we journaling quotas? */
2837 if (EXT3_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
2838 EXT3_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
2839 dquot_mark_dquot_dirty(dquot
);
2840 return ext3_write_dquot(dquot
);
2842 return dquot_mark_dquot_dirty(dquot
);
2846 static int ext3_write_info(struct super_block
*sb
, int type
)
2851 /* Data block + inode block */
2852 handle
= ext3_journal_start(sb
->s_root
->d_inode
, 2);
2854 return PTR_ERR(handle
);
2855 ret
= dquot_commit_info(sb
, type
);
2856 err
= ext3_journal_stop(handle
);
2863 * Turn on quotas during mount time - we need to find
2864 * the quota file and such...
2866 static int ext3_quota_on_mount(struct super_block
*sb
, int type
)
2868 return dquot_quota_on_mount(sb
, EXT3_SB(sb
)->s_qf_names
[type
],
2869 EXT3_SB(sb
)->s_jquota_fmt
, type
);
2873 * Standard function to be called on quota_on
2875 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
,
2881 if (!test_opt(sb
, QUOTA
))
2884 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
2888 /* Quotafile not on the same filesystem? */
2889 if (path
.mnt
->mnt_sb
!= sb
) {
2893 /* Journaling quota? */
2894 if (EXT3_SB(sb
)->s_qf_names
[type
]) {
2895 /* Quotafile not of fs root? */
2896 if (path
.dentry
->d_parent
!= sb
->s_root
)
2897 ext3_msg(sb
, KERN_WARNING
,
2898 "warning: Quota file not on filesystem root. "
2899 "Journaled quota will not work.");
2903 * When we journal data on quota file, we have to flush journal to see
2904 * all updates to the file when we bypass pagecache...
2906 if (ext3_should_journal_data(path
.dentry
->d_inode
)) {
2908 * We don't need to lock updates but journal_flush() could
2909 * otherwise be livelocked...
2911 journal_lock_updates(EXT3_SB(sb
)->s_journal
);
2912 err
= journal_flush(EXT3_SB(sb
)->s_journal
);
2913 journal_unlock_updates(EXT3_SB(sb
)->s_journal
);
2920 err
= dquot_quota_on_path(sb
, type
, format_id
, &path
);
2925 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2926 * acquiring the locks... As quota files are never truncated and quota code
2927 * itself serializes the operations (and noone else should touch the files)
2928 * we don't have to be afraid of races */
2929 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
2930 size_t len
, loff_t off
)
2932 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2933 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2935 int offset
= off
& (sb
->s_blocksize
- 1);
2938 struct buffer_head
*bh
;
2939 loff_t i_size
= i_size_read(inode
);
2943 if (off
+len
> i_size
)
2946 while (toread
> 0) {
2947 tocopy
= sb
->s_blocksize
- offset
< toread
?
2948 sb
->s_blocksize
- offset
: toread
;
2949 bh
= ext3_bread(NULL
, inode
, blk
, 0, &err
);
2952 if (!bh
) /* A hole? */
2953 memset(data
, 0, tocopy
);
2955 memcpy(data
, bh
->b_data
+offset
, tocopy
);
2965 /* Write to quotafile (we know the transaction is already started and has
2966 * enough credits) */
2967 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
2968 const char *data
, size_t len
, loff_t off
)
2970 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2971 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2973 int offset
= off
& (sb
->s_blocksize
- 1);
2974 int journal_quota
= EXT3_SB(sb
)->s_qf_names
[type
] != NULL
;
2975 struct buffer_head
*bh
;
2976 handle_t
*handle
= journal_current_handle();
2979 ext3_msg(sb
, KERN_WARNING
,
2980 "warning: quota write (off=%llu, len=%llu)"
2981 " cancelled because transaction is not started.",
2982 (unsigned long long)off
, (unsigned long long)len
);
2987 * Since we account only one data block in transaction credits,
2988 * then it is impossible to cross a block boundary.
2990 if (sb
->s_blocksize
- offset
< len
) {
2991 ext3_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
2992 " cancelled because not block aligned",
2993 (unsigned long long)off
, (unsigned long long)len
);
2996 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
2997 bh
= ext3_bread(handle
, inode
, blk
, 1, &err
);
3000 if (journal_quota
) {
3001 err
= ext3_journal_get_write_access(handle
, bh
);
3008 memcpy(bh
->b_data
+offset
, data
, len
);
3009 flush_dcache_page(bh
->b_page
);
3012 err
= ext3_journal_dirty_metadata(handle
, bh
);
3014 /* Always do at least ordered writes for quotas */
3015 err
= ext3_journal_dirty_data(handle
, bh
);
3016 mark_buffer_dirty(bh
);
3021 mutex_unlock(&inode
->i_mutex
);
3024 if (inode
->i_size
< off
+ len
) {
3025 i_size_write(inode
, off
+ len
);
3026 EXT3_I(inode
)->i_disksize
= inode
->i_size
;
3029 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3030 ext3_mark_inode_dirty(handle
, inode
);
3031 mutex_unlock(&inode
->i_mutex
);
3037 static int ext3_get_sb(struct file_system_type
*fs_type
,
3038 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3040 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext3_fill_super
, mnt
);
3043 static struct file_system_type ext3_fs_type
= {
3044 .owner
= THIS_MODULE
,
3046 .get_sb
= ext3_get_sb
,
3047 .kill_sb
= kill_block_super
,
3048 .fs_flags
= FS_REQUIRES_DEV
,
3051 static int __init
init_ext3_fs(void)
3053 int err
= init_ext3_xattr();
3056 err
= init_inodecache();
3059 err
= register_filesystem(&ext3_fs_type
);
3064 destroy_inodecache();
3070 static void __exit
exit_ext3_fs(void)
3072 unregister_filesystem(&ext3_fs_type
);
3073 destroy_inodecache();
3077 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3078 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3079 MODULE_LICENSE("GPL");
3080 module_init(init_ext3_fs
)
3081 module_exit(exit_ext3_fs
)