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/config.h>
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/jbd.h>
25 #include <linux/ext3_fs.h>
26 #include <linux/ext3_jbd.h>
27 #include <linux/slab.h>
28 #include <linux/init.h>
29 #include <linux/blkdev.h>
30 #include <linux/parser.h>
31 #include <linux/smp_lock.h>
32 #include <linux/buffer_head.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 <asm/uaccess.h>
42 static int ext3_load_journal(struct super_block
*, struct ext3_super_block
*);
43 static int ext3_create_journal(struct super_block
*, struct ext3_super_block
*,
45 static void ext3_commit_super (struct super_block
* sb
,
46 struct ext3_super_block
* es
,
48 static void ext3_mark_recovery_complete(struct super_block
* sb
,
49 struct ext3_super_block
* es
);
50 static void ext3_clear_journal_err(struct super_block
* sb
,
51 struct ext3_super_block
* es
);
52 static int ext3_sync_fs(struct super_block
*sb
, int wait
);
55 * Wrappers for journal_start/end.
57 * The only special thing we need to do here is to make sure that all
58 * journal_end calls result in the superblock being marked dirty, so
59 * that sync() will call the filesystem's write_super callback if
62 handle_t
*ext3_journal_start(struct inode
*inode
, int nblocks
)
66 if (inode
->i_sb
->s_flags
& MS_RDONLY
)
67 return ERR_PTR(-EROFS
);
69 /* Special case here: if the journal has aborted behind our
70 * backs (eg. EIO in the commit thread), then we still need to
71 * take the FS itself readonly cleanly. */
72 journal
= EXT3_JOURNAL(inode
);
73 if (is_journal_aborted(journal
)) {
74 ext3_abort(inode
->i_sb
, __FUNCTION__
,
75 "Detected aborted journal");
76 return ERR_PTR(-EROFS
);
79 return journal_start(journal
, nblocks
);
83 * The only special thing we need to do here is to make sure that all
84 * journal_stop calls result in the superblock being marked dirty, so
85 * that sync() will call the filesystem's write_super callback if
88 int __ext3_journal_stop(const char *where
, handle_t
*handle
)
90 struct super_block
*sb
;
94 sb
= handle
->h_transaction
->t_journal
->j_private
;
96 rc
= journal_stop(handle
);
101 __ext3_std_error(sb
, where
, err
);
105 void ext3_journal_abort_handle(const char *caller
, const char *err_fn
,
106 struct buffer_head
*bh
, handle_t
*handle
, int err
)
109 const char *errstr
= ext3_decode_error(NULL
, err
, nbuf
);
111 printk(KERN_ERR
"%s: aborting transaction: %s in %s",
112 caller
, errstr
, err_fn
);
115 BUFFER_TRACE(bh
, "abort");
116 journal_abort_handle(handle
);
121 /* Deal with the reporting of failure conditions on a filesystem such as
122 * inconsistencies detected or read IO failures.
124 * On ext2, we can store the error state of the filesystem in the
125 * superblock. That is not possible on ext3, because we may have other
126 * write ordering constraints on the superblock which prevent us from
127 * writing it out straight away; and given that the journal is about to
128 * be aborted, we can't rely on the current, or future, transactions to
129 * write out the superblock safely.
131 * We'll just use the journal_abort() error code to record an error in
132 * the journal instead. On recovery, the journal will compain about
133 * that error until we've noted it down and cleared it.
136 static void ext3_handle_error(struct super_block
*sb
)
138 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
140 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
141 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
143 if (sb
->s_flags
& MS_RDONLY
)
146 if (test_opt (sb
, ERRORS_PANIC
))
147 panic ("EXT3-fs (device %s): panic forced after error\n",
149 if (test_opt (sb
, ERRORS_RO
)) {
150 printk (KERN_CRIT
"Remounting filesystem read-only\n");
151 sb
->s_flags
|= MS_RDONLY
;
153 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
155 EXT3_SB(sb
)->s_mount_opt
|= EXT3_MOUNT_ABORT
;
157 journal_abort(journal
, -EIO
);
159 ext3_commit_super(sb
, es
, 1);
162 void ext3_error (struct super_block
* sb
, const char * function
,
163 const char * fmt
, ...)
168 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
173 ext3_handle_error(sb
);
176 const char *ext3_decode_error(struct super_block
* sb
, int errno
, char nbuf
[16])
182 errstr
= "IO failure";
185 errstr
= "Out of memory";
188 if (!sb
|| EXT3_SB(sb
)->s_journal
->j_flags
& JFS_ABORT
)
189 errstr
= "Journal has aborted";
191 errstr
= "Readonly filesystem";
194 /* If the caller passed in an extra buffer for unknown
195 * errors, textualise them now. Else we just return
198 /* Check for truncated error codes... */
199 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
208 /* __ext3_std_error decodes expected errors from journaling functions
209 * automatically and invokes the appropriate error response. */
211 void __ext3_std_error (struct super_block
* sb
, const char * function
,
215 const char *errstr
= ext3_decode_error(sb
, errno
, nbuf
);
217 printk (KERN_CRIT
"EXT3-fs error (device %s) in %s: %s\n",
218 sb
->s_id
, function
, errstr
);
220 ext3_handle_error(sb
);
224 * ext3_abort is a much stronger failure handler than ext3_error. The
225 * abort function may be used to deal with unrecoverable failures such
226 * as journal IO errors or ENOMEM at a critical moment in log management.
228 * We unconditionally force the filesystem into an ABORT|READONLY state,
229 * unless the error response on the fs has been set to panic in which
230 * case we take the easy way out and panic immediately.
233 void ext3_abort (struct super_block
* sb
, const char * function
,
234 const char * fmt
, ...)
238 printk (KERN_CRIT
"ext3_abort called.\n");
241 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
246 if (test_opt(sb
, ERRORS_PANIC
))
247 panic("EXT3-fs panic from previous error\n");
249 if (sb
->s_flags
& MS_RDONLY
)
252 printk(KERN_CRIT
"Remounting filesystem read-only\n");
253 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
254 sb
->s_flags
|= MS_RDONLY
;
255 EXT3_SB(sb
)->s_mount_opt
|= EXT3_MOUNT_ABORT
;
256 journal_abort(EXT3_SB(sb
)->s_journal
, -EIO
);
259 /* Deal with the reporting of failure conditions while running, such as
260 * inconsistencies in operation or invalid system states.
262 * Use ext3_error() for cases of invalid filesystem states, as that will
263 * record an error on disk and force a filesystem check on the next boot.
265 NORET_TYPE
void ext3_panic (struct super_block
* sb
, const char * function
,
266 const char * fmt
, ...)
271 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
276 /* this is to prevent panic from syncing this filesystem */
277 /* AKPM: is this sufficient? */
278 sb
->s_flags
|= MS_RDONLY
;
279 panic ("EXT3-fs panic forced\n");
282 void ext3_warning (struct super_block
* sb
, const char * function
,
283 const char * fmt
, ...)
288 printk(KERN_WARNING
"EXT3-fs warning (device %s): %s: ",
295 void ext3_update_dynamic_rev(struct super_block
*sb
)
297 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
299 if (le32_to_cpu(es
->s_rev_level
) > EXT3_GOOD_OLD_REV
)
302 ext3_warning(sb
, __FUNCTION__
,
303 "updating to rev %d because of new feature flag, "
304 "running e2fsck is recommended",
307 es
->s_first_ino
= cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO
);
308 es
->s_inode_size
= cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE
);
309 es
->s_rev_level
= cpu_to_le32(EXT3_DYNAMIC_REV
);
310 /* leave es->s_feature_*compat flags alone */
311 /* es->s_uuid will be set by e2fsck if empty */
314 * The rest of the superblock fields should be zero, and if not it
315 * means they are likely already in use, so leave them alone. We
316 * can leave it up to e2fsck to clean up any inconsistencies there.
321 * Open the external journal device
323 static struct block_device
*ext3_blkdev_get(dev_t dev
)
325 struct block_device
*bdev
;
326 char b
[BDEVNAME_SIZE
];
328 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
334 printk(KERN_ERR
"EXT3: failed to open journal device %s: %ld\n",
335 __bdevname(dev
, b
), PTR_ERR(bdev
));
340 * Release the journal device
342 static int ext3_blkdev_put(struct block_device
*bdev
)
345 return blkdev_put(bdev
);
348 static int ext3_blkdev_remove(struct ext3_sb_info
*sbi
)
350 struct block_device
*bdev
;
353 bdev
= sbi
->journal_bdev
;
355 ret
= ext3_blkdev_put(bdev
);
356 sbi
->journal_bdev
= NULL
;
361 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
363 return &list_entry(l
, struct ext3_inode_info
, i_orphan
)->vfs_inode
;
366 static void dump_orphan_list(struct super_block
*sb
, struct ext3_sb_info
*sbi
)
370 printk(KERN_ERR
"sb orphan head is %d\n",
371 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
373 printk(KERN_ERR
"sb_info orphan list:\n");
374 list_for_each(l
, &sbi
->s_orphan
) {
375 struct inode
*inode
= orphan_list_entry(l
);
377 "inode %s:%ld at %p: mode %o, nlink %d, next %d\n",
378 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
379 inode
->i_mode
, inode
->i_nlink
,
384 void ext3_put_super (struct super_block
* sb
)
386 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
387 struct ext3_super_block
*es
= sbi
->s_es
;
390 ext3_xattr_put_super(sb
);
391 journal_destroy(sbi
->s_journal
);
392 if (!(sb
->s_flags
& MS_RDONLY
)) {
393 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
394 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
395 BUFFER_TRACE(sbi
->s_sbh
, "marking dirty");
396 mark_buffer_dirty(sbi
->s_sbh
);
397 ext3_commit_super(sb
, es
, 1);
400 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
401 brelse(sbi
->s_group_desc
[i
]);
402 kfree(sbi
->s_group_desc
);
406 for (i
= 0; i
< MAXQUOTAS
; i
++) {
407 if (sbi
->s_qf_names
[i
])
408 kfree(sbi
->s_qf_names
[i
]);
412 /* Debugging code just in case the in-memory inode orphan list
413 * isn't empty. The on-disk one can be non-empty if we've
414 * detected an error and taken the fs readonly, but the
415 * in-memory list had better be clean by this point. */
416 if (!list_empty(&sbi
->s_orphan
))
417 dump_orphan_list(sb
, sbi
);
418 J_ASSERT(list_empty(&sbi
->s_orphan
));
420 invalidate_bdev(sb
->s_bdev
, 0);
421 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
423 * Invalidate the journal device's buffers. We don't want them
424 * floating about in memory - the physical journal device may
425 * hotswapped, and it breaks the `ro-after' testing code.
427 sync_blockdev(sbi
->journal_bdev
);
428 invalidate_bdev(sbi
->journal_bdev
, 0);
429 ext3_blkdev_remove(sbi
);
431 sb
->s_fs_info
= NULL
;
436 static kmem_cache_t
*ext3_inode_cachep
;
439 * Called inside transaction, so use GFP_NOFS
441 static struct inode
*ext3_alloc_inode(struct super_block
*sb
)
443 struct ext3_inode_info
*ei
;
445 ei
= kmem_cache_alloc(ext3_inode_cachep
, SLAB_NOFS
);
448 #ifdef CONFIG_EXT3_FS_POSIX_ACL
449 ei
->i_acl
= EXT3_ACL_NOT_CACHED
;
450 ei
->i_default_acl
= EXT3_ACL_NOT_CACHED
;
452 ei
->vfs_inode
.i_version
= 1;
453 return &ei
->vfs_inode
;
456 static void ext3_destroy_inode(struct inode
*inode
)
458 kmem_cache_free(ext3_inode_cachep
, EXT3_I(inode
));
461 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
463 struct ext3_inode_info
*ei
= (struct ext3_inode_info
*) foo
;
465 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
466 SLAB_CTOR_CONSTRUCTOR
) {
467 INIT_LIST_HEAD(&ei
->i_orphan
);
468 #ifdef CONFIG_EXT3_FS_XATTR
469 init_rwsem(&ei
->xattr_sem
);
471 init_MUTEX(&ei
->truncate_sem
);
472 inode_init_once(&ei
->vfs_inode
);
476 static int init_inodecache(void)
478 ext3_inode_cachep
= kmem_cache_create("ext3_inode_cache",
479 sizeof(struct ext3_inode_info
),
480 0, SLAB_RECLAIM_ACCOUNT
,
482 if (ext3_inode_cachep
== NULL
)
487 static void destroy_inodecache(void)
489 if (kmem_cache_destroy(ext3_inode_cachep
))
490 printk(KERN_INFO
"ext3_inode_cache: not all structures were freed\n");
493 #ifdef CONFIG_EXT3_FS_POSIX_ACL
495 static void ext3_clear_inode(struct inode
*inode
)
497 if (EXT3_I(inode
)->i_acl
&&
498 EXT3_I(inode
)->i_acl
!= EXT3_ACL_NOT_CACHED
) {
499 posix_acl_release(EXT3_I(inode
)->i_acl
);
500 EXT3_I(inode
)->i_acl
= EXT3_ACL_NOT_CACHED
;
502 if (EXT3_I(inode
)->i_default_acl
&&
503 EXT3_I(inode
)->i_default_acl
!= EXT3_ACL_NOT_CACHED
) {
504 posix_acl_release(EXT3_I(inode
)->i_default_acl
);
505 EXT3_I(inode
)->i_default_acl
= EXT3_ACL_NOT_CACHED
;
510 # define ext3_clear_inode NULL
515 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
516 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
518 static int ext3_dquot_initialize(struct inode
*inode
, int type
);
519 static int ext3_dquot_drop(struct inode
*inode
);
520 static int ext3_write_dquot(struct dquot
*dquot
);
521 static int ext3_acquire_dquot(struct dquot
*dquot
);
522 static int ext3_release_dquot(struct dquot
*dquot
);
523 static int ext3_mark_dquot_dirty(struct dquot
*dquot
);
524 static int ext3_write_info(struct super_block
*sb
, int type
);
525 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
, char *path
);
526 static int ext3_quota_on_mount(struct super_block
*sb
, int type
);
527 static int ext3_quota_off_mount(struct super_block
*sb
, int type
);
529 static struct dquot_operations ext3_quota_operations
= {
530 .initialize
= ext3_dquot_initialize
,
531 .drop
= ext3_dquot_drop
,
532 .alloc_space
= dquot_alloc_space
,
533 .alloc_inode
= dquot_alloc_inode
,
534 .free_space
= dquot_free_space
,
535 .free_inode
= dquot_free_inode
,
536 .transfer
= dquot_transfer
,
537 .write_dquot
= ext3_write_dquot
,
538 .acquire_dquot
= ext3_acquire_dquot
,
539 .release_dquot
= ext3_release_dquot
,
540 .mark_dirty
= ext3_mark_dquot_dirty
,
541 .write_info
= ext3_write_info
544 static struct quotactl_ops ext3_qctl_operations
= {
545 .quota_on
= ext3_quota_on
,
546 .quota_off
= vfs_quota_off
,
547 .quota_sync
= vfs_quota_sync
,
548 .get_info
= vfs_get_dqinfo
,
549 .set_info
= vfs_set_dqinfo
,
550 .get_dqblk
= vfs_get_dqblk
,
551 .set_dqblk
= vfs_set_dqblk
555 static struct super_operations ext3_sops
= {
556 .alloc_inode
= ext3_alloc_inode
,
557 .destroy_inode
= ext3_destroy_inode
,
558 .read_inode
= ext3_read_inode
,
559 .write_inode
= ext3_write_inode
,
560 .dirty_inode
= ext3_dirty_inode
,
561 .put_inode
= ext3_put_inode
,
562 .delete_inode
= ext3_delete_inode
,
563 .put_super
= ext3_put_super
,
564 .write_super
= ext3_write_super
,
565 .sync_fs
= ext3_sync_fs
,
566 .write_super_lockfs
= ext3_write_super_lockfs
,
567 .unlockfs
= ext3_unlockfs
,
568 .statfs
= ext3_statfs
,
569 .remount_fs
= ext3_remount
,
570 .clear_inode
= ext3_clear_inode
,
573 struct dentry
*ext3_get_parent(struct dentry
*child
);
574 static struct export_operations ext3_export_ops
= {
575 .get_parent
= ext3_get_parent
,
579 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
580 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
581 Opt_nouid32
, Opt_check
, Opt_nocheck
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
582 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
, Opt_noload
,
583 Opt_commit
, Opt_journal_update
, Opt_journal_inum
,
584 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
585 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
586 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
,
587 Opt_ignore
, Opt_barrier
, Opt_err
,
590 static match_table_t tokens
= {
591 {Opt_bsd_df
, "bsddf"},
592 {Opt_minix_df
, "minixdf"},
593 {Opt_grpid
, "grpid"},
594 {Opt_grpid
, "bsdgroups"},
595 {Opt_nogrpid
, "nogrpid"},
596 {Opt_nogrpid
, "sysvgroups"},
597 {Opt_resgid
, "resgid=%u"},
598 {Opt_resuid
, "resuid=%u"},
600 {Opt_err_cont
, "errors=continue"},
601 {Opt_err_panic
, "errors=panic"},
602 {Opt_err_ro
, "errors=remount-ro"},
603 {Opt_nouid32
, "nouid32"},
604 {Opt_nocheck
, "nocheck"},
605 {Opt_nocheck
, "check=none"},
606 {Opt_check
, "check"},
607 {Opt_debug
, "debug"},
608 {Opt_oldalloc
, "oldalloc"},
609 {Opt_orlov
, "orlov"},
610 {Opt_user_xattr
, "user_xattr"},
611 {Opt_nouser_xattr
, "nouser_xattr"},
613 {Opt_noacl
, "noacl"},
614 {Opt_noload
, "noload"},
615 {Opt_commit
, "commit=%u"},
616 {Opt_journal_update
, "journal=update"},
617 {Opt_journal_inum
, "journal=%u"},
618 {Opt_abort
, "abort"},
619 {Opt_data_journal
, "data=journal"},
620 {Opt_data_ordered
, "data=ordered"},
621 {Opt_data_writeback
, "data=writeback"},
622 {Opt_offusrjquota
, "usrjquota="},
623 {Opt_usrjquota
, "usrjquota=%s"},
624 {Opt_offgrpjquota
, "grpjquota="},
625 {Opt_grpjquota
, "grpjquota=%s"},
626 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
627 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
628 {Opt_ignore
, "grpquota"},
629 {Opt_ignore
, "noquota"},
630 {Opt_ignore
, "quota"},
631 {Opt_ignore
, "usrquota"},
632 {Opt_barrier
, "barrier=%u"},
636 static unsigned long get_sb_block(void **data
)
638 unsigned long sb_block
;
639 char *options
= (char *) *data
;
641 if (!options
|| strncmp(options
, "sb=", 3) != 0)
642 return 1; /* Default location */
644 sb_block
= simple_strtoul(options
, &options
, 0);
645 if (*options
&& *options
!= ',') {
646 printk("EXT3-fs: Invalid sb specification: %s\n",
652 *data
= (void *) options
;
656 static int parse_options (char * options
, struct super_block
*sb
,
657 unsigned long * inum
, int is_remount
)
659 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
661 substring_t args
[MAX_OPT_ARGS
];
671 while ((p
= strsep (&options
, ",")) != NULL
) {
676 token
= match_token(p
, tokens
, args
);
679 clear_opt (sbi
->s_mount_opt
, MINIX_DF
);
682 set_opt (sbi
->s_mount_opt
, MINIX_DF
);
685 set_opt (sbi
->s_mount_opt
, GRPID
);
688 clear_opt (sbi
->s_mount_opt
, GRPID
);
691 if (match_int(&args
[0], &option
))
693 sbi
->s_resuid
= option
;
696 if (match_int(&args
[0], &option
))
698 sbi
->s_resgid
= option
;
701 /* handled by get_sb_block() instead of here */
702 /* *sb_block = match_int(&args[0]); */
705 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
706 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
707 set_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
710 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
711 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
712 set_opt (sbi
->s_mount_opt
, ERRORS_RO
);
715 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
716 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
717 set_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
720 set_opt (sbi
->s_mount_opt
, NO_UID32
);
723 #ifdef CONFIG_EXT3_CHECK
724 set_opt (sbi
->s_mount_opt
, CHECK
);
727 "EXT3 Check option not supported\n");
731 clear_opt (sbi
->s_mount_opt
, CHECK
);
734 set_opt (sbi
->s_mount_opt
, DEBUG
);
737 set_opt (sbi
->s_mount_opt
, OLDALLOC
);
740 clear_opt (sbi
->s_mount_opt
, OLDALLOC
);
742 #ifdef CONFIG_EXT3_FS_XATTR
744 set_opt (sbi
->s_mount_opt
, XATTR_USER
);
746 case Opt_nouser_xattr
:
747 clear_opt (sbi
->s_mount_opt
, XATTR_USER
);
751 case Opt_nouser_xattr
:
752 printk("EXT3 (no)user_xattr options not supported\n");
755 #ifdef CONFIG_EXT3_FS_POSIX_ACL
757 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
760 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
765 printk("EXT3 (no)acl options not supported\n");
768 case Opt_journal_update
:
770 /* Eventually we will want to be able to create
771 a journal file here. For now, only allow the
772 user to specify an existing inode to be the
775 printk(KERN_ERR
"EXT3-fs: cannot specify "
776 "journal on remount\n");
779 set_opt (sbi
->s_mount_opt
, UPDATE_JOURNAL
);
781 case Opt_journal_inum
:
783 printk(KERN_ERR
"EXT3-fs: cannot specify "
784 "journal on remount\n");
787 if (match_int(&args
[0], &option
))
792 set_opt (sbi
->s_mount_opt
, NOLOAD
);
795 if (match_int(&args
[0], &option
))
800 option
= JBD_DEFAULT_MAX_COMMIT_AGE
;
801 sbi
->s_commit_interval
= HZ
* option
;
803 case Opt_data_journal
:
804 data_opt
= EXT3_MOUNT_JOURNAL_DATA
;
806 case Opt_data_ordered
:
807 data_opt
= EXT3_MOUNT_ORDERED_DATA
;
809 case Opt_data_writeback
:
810 data_opt
= EXT3_MOUNT_WRITEBACK_DATA
;
813 if ((sbi
->s_mount_opt
& EXT3_MOUNT_DATA_FLAGS
)
816 "EXT3-fs: cannot change data "
817 "mode on remount\n");
821 sbi
->s_mount_opt
&= ~EXT3_MOUNT_DATA_FLAGS
;
822 sbi
->s_mount_opt
|= data_opt
;
832 if (sb_any_quota_enabled(sb
)) {
834 "EXT3-fs: Cannot change journalled "
835 "quota options when quota turned on.\n");
838 if (sbi
->s_qf_names
[qtype
]) {
840 "EXT3-fs: %s quota file already "
841 "specified.\n", QTYPE2NAME(qtype
));
844 sbi
->s_qf_names
[qtype
] = match_strdup(&args
[0]);
845 if (!sbi
->s_qf_names
[qtype
]) {
847 "EXT3-fs: not enough memory for "
848 "storing quotafile name.\n");
851 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
853 "EXT3-fs: quotafile must be on "
854 "filesystem root.\n");
855 kfree(sbi
->s_qf_names
[qtype
]);
856 sbi
->s_qf_names
[qtype
] = NULL
;
860 case Opt_offusrjquota
:
863 case Opt_offgrpjquota
:
866 if (sb_any_quota_enabled(sb
)) {
867 printk(KERN_ERR
"EXT3-fs: Cannot change "
868 "journalled quota options when "
869 "quota turned on.\n");
872 if (sbi
->s_qf_names
[qtype
]) {
873 kfree(sbi
->s_qf_names
[qtype
]);
874 sbi
->s_qf_names
[qtype
] = NULL
;
877 case Opt_jqfmt_vfsold
:
878 sbi
->s_jquota_fmt
= QFMT_VFS_OLD
;
880 case Opt_jqfmt_vfsv0
:
881 sbi
->s_jquota_fmt
= QFMT_VFS_V0
;
886 case Opt_offusrjquota
:
887 case Opt_offgrpjquota
:
888 case Opt_jqfmt_vfsold
:
889 case Opt_jqfmt_vfsv0
:
891 "EXT3-fs: journalled quota options not "
896 set_opt(sbi
->s_mount_opt
, ABORT
);
899 if (match_int(&args
[0], &option
))
902 set_opt(sbi
->s_mount_opt
, BARRIER
);
904 clear_opt(sbi
->s_mount_opt
, BARRIER
);
910 "EXT3-fs: Unrecognized mount option \"%s\" "
911 "or missing value\n", p
);
916 if (!sbi
->s_jquota_fmt
&& (sbi
->s_qf_names
[USRQUOTA
] ||
917 sbi
->s_qf_names
[GRPQUOTA
])) {
919 "EXT3-fs: journalled quota format not specified.\n");
927 static int ext3_setup_super(struct super_block
*sb
, struct ext3_super_block
*es
,
930 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
933 if (le32_to_cpu(es
->s_rev_level
) > EXT3_MAX_SUPP_REV
) {
934 printk (KERN_ERR
"EXT3-fs warning: revision level too high, "
935 "forcing read-only mode\n");
940 if (!(sbi
->s_mount_state
& EXT3_VALID_FS
))
941 printk (KERN_WARNING
"EXT3-fs warning: mounting unchecked fs, "
942 "running e2fsck is recommended\n");
943 else if ((sbi
->s_mount_state
& EXT3_ERROR_FS
))
945 "EXT3-fs warning: mounting fs with errors, "
946 "running e2fsck is recommended\n");
947 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
948 le16_to_cpu(es
->s_mnt_count
) >=
949 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
951 "EXT3-fs warning: maximal mount count reached, "
952 "running e2fsck is recommended\n");
953 else if (le32_to_cpu(es
->s_checkinterval
) &&
954 (le32_to_cpu(es
->s_lastcheck
) +
955 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
957 "EXT3-fs warning: checktime reached, "
958 "running e2fsck is recommended\n");
960 /* @@@ We _will_ want to clear the valid bit if we find
961 inconsistencies, to force a fsck at reboot. But for
962 a plain journaled filesystem we can keep it set as
964 es
->s_state
= cpu_to_le16(le16_to_cpu(es
->s_state
) & ~EXT3_VALID_FS
);
966 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
967 es
->s_max_mnt_count
= cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT
);
968 es
->s_mnt_count
=cpu_to_le16(le16_to_cpu(es
->s_mnt_count
) + 1);
969 es
->s_mtime
= cpu_to_le32(get_seconds());
970 ext3_update_dynamic_rev(sb
);
971 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
973 ext3_commit_super(sb
, es
, 1);
974 if (test_opt(sb
, DEBUG
))
975 printk(KERN_INFO
"[EXT3 FS bs=%lu, gc=%lu, "
976 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
979 EXT3_BLOCKS_PER_GROUP(sb
),
980 EXT3_INODES_PER_GROUP(sb
),
983 printk(KERN_INFO
"EXT3 FS on %s, ", sb
->s_id
);
984 if (EXT3_SB(sb
)->s_journal
->j_inode
== NULL
) {
985 char b
[BDEVNAME_SIZE
];
987 printk("external journal on %s\n",
988 bdevname(EXT3_SB(sb
)->s_journal
->j_dev
, b
));
990 printk("internal journal\n");
992 #ifdef CONFIG_EXT3_CHECK
993 if (test_opt (sb
, CHECK
)) {
994 ext3_check_blocks_bitmap (sb
);
995 ext3_check_inodes_bitmap (sb
);
1001 static int ext3_check_descriptors (struct super_block
* sb
)
1003 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1004 unsigned long block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1005 struct ext3_group_desc
* gdp
= NULL
;
1009 ext3_debug ("Checking group descriptors");
1011 for (i
= 0; i
< sbi
->s_groups_count
; i
++)
1013 if ((i
% EXT3_DESC_PER_BLOCK(sb
)) == 0)
1014 gdp
= (struct ext3_group_desc
*)
1015 sbi
->s_group_desc
[desc_block
++]->b_data
;
1016 if (le32_to_cpu(gdp
->bg_block_bitmap
) < block
||
1017 le32_to_cpu(gdp
->bg_block_bitmap
) >=
1018 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1020 ext3_error (sb
, "ext3_check_descriptors",
1021 "Block bitmap for group %d"
1022 " not in group (block %lu)!",
1024 le32_to_cpu(gdp
->bg_block_bitmap
));
1027 if (le32_to_cpu(gdp
->bg_inode_bitmap
) < block
||
1028 le32_to_cpu(gdp
->bg_inode_bitmap
) >=
1029 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1031 ext3_error (sb
, "ext3_check_descriptors",
1032 "Inode bitmap for group %d"
1033 " not in group (block %lu)!",
1035 le32_to_cpu(gdp
->bg_inode_bitmap
));
1038 if (le32_to_cpu(gdp
->bg_inode_table
) < block
||
1039 le32_to_cpu(gdp
->bg_inode_table
) + sbi
->s_itb_per_group
>=
1040 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1042 ext3_error (sb
, "ext3_check_descriptors",
1043 "Inode table for group %d"
1044 " not in group (block %lu)!",
1046 le32_to_cpu(gdp
->bg_inode_table
));
1049 block
+= EXT3_BLOCKS_PER_GROUP(sb
);
1053 sbi
->s_es
->s_free_blocks_count
=cpu_to_le32(ext3_count_free_blocks(sb
));
1054 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext3_count_free_inodes(sb
));
1059 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1060 * the superblock) which were deleted from all directories, but held open by
1061 * a process at the time of a crash. We walk the list and try to delete these
1062 * inodes at recovery time (only with a read-write filesystem).
1064 * In order to keep the orphan inode chain consistent during traversal (in
1065 * case of crash during recovery), we link each inode into the superblock
1066 * orphan list_head and handle it the same way as an inode deletion during
1067 * normal operation (which journals the operations for us).
1069 * We only do an iget() and an iput() on each inode, which is very safe if we
1070 * accidentally point at an in-use or already deleted inode. The worst that
1071 * can happen in this case is that we get a "bit already cleared" message from
1072 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1073 * e2fsck was run on this filesystem, and it must have already done the orphan
1074 * inode cleanup for us, so we can safely abort without any further action.
1076 static void ext3_orphan_cleanup (struct super_block
* sb
,
1077 struct ext3_super_block
* es
)
1079 unsigned int s_flags
= sb
->s_flags
;
1080 int nr_orphans
= 0, nr_truncates
= 0;
1084 if (!es
->s_last_orphan
) {
1085 jbd_debug(4, "no orphan inodes to clean up\n");
1089 if (EXT3_SB(sb
)->s_mount_state
& EXT3_ERROR_FS
) {
1090 if (es
->s_last_orphan
)
1091 jbd_debug(1, "Errors on filesystem, "
1092 "clearing orphan list.\n");
1093 es
->s_last_orphan
= 0;
1094 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1098 if (s_flags
& MS_RDONLY
) {
1099 printk(KERN_INFO
"EXT3-fs: %s: orphan cleanup on readonly fs\n",
1101 sb
->s_flags
&= ~MS_RDONLY
;
1104 /* Needed for iput() to work correctly and not trash data */
1105 sb
->s_flags
|= MS_ACTIVE
;
1106 /* Turn on quotas so that they are updated correctly */
1107 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1108 if (EXT3_SB(sb
)->s_qf_names
[i
]) {
1109 int ret
= ext3_quota_on_mount(sb
, i
);
1112 "EXT3-fs: Cannot turn on journalled "
1113 "quota: error %d\n", ret
);
1118 while (es
->s_last_orphan
) {
1119 struct inode
*inode
;
1122 ext3_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
)))) {
1123 es
->s_last_orphan
= 0;
1127 list_add(&EXT3_I(inode
)->i_orphan
, &EXT3_SB(sb
)->s_orphan
);
1129 if (inode
->i_nlink
) {
1131 "%s: truncating inode %ld to %Ld bytes\n",
1132 __FUNCTION__
, inode
->i_ino
, inode
->i_size
);
1133 jbd_debug(2, "truncating inode %ld to %Ld bytes\n",
1134 inode
->i_ino
, inode
->i_size
);
1135 ext3_truncate(inode
);
1139 "%s: deleting unreferenced inode %ld\n",
1140 __FUNCTION__
, inode
->i_ino
);
1141 jbd_debug(2, "deleting unreferenced inode %ld\n",
1145 iput(inode
); /* The delete magic happens here! */
1148 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1151 printk(KERN_INFO
"EXT3-fs: %s: %d orphan inode%s deleted\n",
1152 sb
->s_id
, PLURAL(nr_orphans
));
1154 printk(KERN_INFO
"EXT3-fs: %s: %d truncate%s cleaned up\n",
1155 sb
->s_id
, PLURAL(nr_truncates
));
1157 /* Turn quotas off */
1158 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1159 if (sb_dqopt(sb
)->files
[i
])
1160 ext3_quota_off_mount(sb
, i
);
1163 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1166 #define log2(n) ffz(~(n))
1169 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1170 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1171 * We need to be 1 filesystem block less than the 2^32 sector limit.
1173 static loff_t
ext3_max_size(int bits
)
1175 loff_t res
= EXT3_NDIR_BLOCKS
;
1176 res
+= 1LL << (bits
-2);
1177 res
+= 1LL << (2*(bits
-2));
1178 res
+= 1LL << (3*(bits
-2));
1180 if (res
> (512LL << 32) - (1 << bits
))
1181 res
= (512LL << 32) - (1 << bits
);
1185 static unsigned long descriptor_loc(struct super_block
*sb
,
1186 unsigned long logic_sb_block
,
1189 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1190 unsigned long bg
, first_data_block
, first_meta_bg
;
1193 first_data_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1194 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1196 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_META_BG
) ||
1198 return (logic_sb_block
+ nr
+ 1);
1199 bg
= sbi
->s_desc_per_block
* nr
;
1200 if (ext3_bg_has_super(sb
, bg
))
1202 return (first_data_block
+ has_super
+ (bg
* sbi
->s_blocks_per_group
));
1206 static int ext3_fill_super (struct super_block
*sb
, void *data
, int silent
)
1208 struct buffer_head
* bh
;
1209 struct ext3_super_block
*es
= NULL
;
1210 struct ext3_sb_info
*sbi
;
1211 unsigned long block
;
1212 unsigned long sb_block
= get_sb_block(&data
);
1213 unsigned long logic_sb_block
;
1214 unsigned long offset
= 0;
1215 unsigned long journal_inum
= 0;
1216 unsigned long def_mount_opts
;
1225 sbi
= kmalloc(sizeof(*sbi
), GFP_KERNEL
);
1228 sb
->s_fs_info
= sbi
;
1229 memset(sbi
, 0, sizeof(*sbi
));
1230 sbi
->s_mount_opt
= 0;
1231 sbi
->s_resuid
= EXT3_DEF_RESUID
;
1232 sbi
->s_resgid
= EXT3_DEF_RESGID
;
1234 blocksize
= sb_min_blocksize(sb
, EXT3_MIN_BLOCK_SIZE
);
1236 printk(KERN_ERR
"EXT3-fs: unable to set blocksize\n");
1241 * The ext3 superblock will not be buffer aligned for other than 1kB
1242 * block sizes. We need to calculate the offset from buffer start.
1244 if (blocksize
!= EXT3_MIN_BLOCK_SIZE
) {
1245 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1246 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1248 logic_sb_block
= sb_block
;
1251 if (!(bh
= sb_bread(sb
, logic_sb_block
))) {
1252 printk (KERN_ERR
"EXT3-fs: unable to read superblock\n");
1256 * Note: s_es must be initialized as soon as possible because
1257 * some ext3 macro-instructions depend on its value
1259 es
= (struct ext3_super_block
*) (((char *)bh
->b_data
) + offset
);
1261 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1262 if (sb
->s_magic
!= EXT3_SUPER_MAGIC
) {
1265 "VFS: Can't find ext3 filesystem on dev %s.\n",
1270 /* Set defaults before we parse the mount options */
1271 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1272 if (def_mount_opts
& EXT3_DEFM_DEBUG
)
1273 set_opt(sbi
->s_mount_opt
, DEBUG
);
1274 if (def_mount_opts
& EXT3_DEFM_BSDGROUPS
)
1275 set_opt(sbi
->s_mount_opt
, GRPID
);
1276 if (def_mount_opts
& EXT3_DEFM_UID16
)
1277 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1278 if (def_mount_opts
& EXT3_DEFM_XATTR_USER
)
1279 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1280 if (def_mount_opts
& EXT3_DEFM_ACL
)
1281 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1282 if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_DATA
)
1283 sbi
->s_mount_opt
|= EXT3_MOUNT_JOURNAL_DATA
;
1284 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_ORDERED
)
1285 sbi
->s_mount_opt
|= EXT3_MOUNT_ORDERED_DATA
;
1286 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_WBACK
)
1287 sbi
->s_mount_opt
|= EXT3_MOUNT_WRITEBACK_DATA
;
1289 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_PANIC
)
1290 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1291 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_RO
)
1292 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1294 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
1295 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
1297 if (!parse_options ((char *) data
, sb
, &journal_inum
, 0))
1300 sb
->s_flags
|= MS_ONE_SECOND
;
1301 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1302 ((sbi
->s_mount_opt
& EXT3_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1304 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
&&
1305 (EXT3_HAS_COMPAT_FEATURE(sb
, ~0U) ||
1306 EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
1307 EXT3_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
1309 "EXT3-fs warning: feature flags set on rev 0 fs, "
1310 "running e2fsck is recommended\n");
1312 * Check feature flags regardless of the revision level, since we
1313 * previously didn't change the revision level when setting the flags,
1314 * so there is a chance incompat flags are set on a rev 0 filesystem.
1316 features
= EXT3_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
);
1318 printk(KERN_ERR
"EXT3-fs: %s: couldn't mount because of "
1319 "unsupported optional features (%x).\n",
1320 sb
->s_id
, le32_to_cpu(features
));
1323 features
= EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
);
1324 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1325 printk(KERN_ERR
"EXT3-fs: %s: couldn't mount RDWR because of "
1326 "unsupported optional features (%x).\n",
1327 sb
->s_id
, le32_to_cpu(features
));
1330 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
1332 if (blocksize
< EXT3_MIN_BLOCK_SIZE
||
1333 blocksize
> EXT3_MAX_BLOCK_SIZE
) {
1335 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1336 blocksize
, sb
->s_id
);
1340 hblock
= bdev_hardsect_size(sb
->s_bdev
);
1341 if (sb
->s_blocksize
!= blocksize
) {
1343 * Make sure the blocksize for the filesystem is larger
1344 * than the hardware sectorsize for the machine.
1346 if (blocksize
< hblock
) {
1347 printk(KERN_ERR
"EXT3-fs: blocksize %d too small for "
1348 "device blocksize %d.\n", blocksize
, hblock
);
1353 sb_set_blocksize(sb
, blocksize
);
1354 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1355 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1356 bh
= sb_bread(sb
, logic_sb_block
);
1359 "EXT3-fs: Can't read superblock on 2nd try.\n");
1362 es
= (struct ext3_super_block
*)(((char *)bh
->b_data
) + offset
);
1364 if (es
->s_magic
!= cpu_to_le16(EXT3_SUPER_MAGIC
)) {
1366 "EXT3-fs: Magic mismatch, very weird !\n");
1371 sb
->s_maxbytes
= ext3_max_size(sb
->s_blocksize_bits
);
1373 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
) {
1374 sbi
->s_inode_size
= EXT3_GOOD_OLD_INODE_SIZE
;
1375 sbi
->s_first_ino
= EXT3_GOOD_OLD_FIRST_INO
;
1377 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
1378 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
1379 if ((sbi
->s_inode_size
< EXT3_GOOD_OLD_INODE_SIZE
) ||
1380 (sbi
->s_inode_size
& (sbi
->s_inode_size
- 1)) ||
1381 (sbi
->s_inode_size
> blocksize
)) {
1383 "EXT3-fs: unsupported inode size: %d\n",
1388 sbi
->s_frag_size
= EXT3_MIN_FRAG_SIZE
<<
1389 le32_to_cpu(es
->s_log_frag_size
);
1390 if (blocksize
!= sbi
->s_frag_size
) {
1392 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1393 sbi
->s_frag_size
, blocksize
);
1396 sbi
->s_frags_per_block
= 1;
1397 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
1398 sbi
->s_frags_per_group
= le32_to_cpu(es
->s_frags_per_group
);
1399 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
1400 sbi
->s_inodes_per_block
= blocksize
/ EXT3_INODE_SIZE(sb
);
1401 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/sbi
->s_inodes_per_block
;
1402 sbi
->s_desc_per_block
= blocksize
/ sizeof(struct ext3_group_desc
);
1404 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
1405 sbi
->s_addr_per_block_bits
= log2(EXT3_ADDR_PER_BLOCK(sb
));
1406 sbi
->s_desc_per_block_bits
= log2(EXT3_DESC_PER_BLOCK(sb
));
1407 for (i
=0; i
< 4; i
++)
1408 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
1409 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
1411 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
1413 "EXT3-fs: #blocks per group too big: %lu\n",
1414 sbi
->s_blocks_per_group
);
1417 if (sbi
->s_frags_per_group
> blocksize
* 8) {
1419 "EXT3-fs: #fragments per group too big: %lu\n",
1420 sbi
->s_frags_per_group
);
1423 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
1425 "EXT3-fs: #inodes per group too big: %lu\n",
1426 sbi
->s_inodes_per_group
);
1430 sbi
->s_groups_count
= (le32_to_cpu(es
->s_blocks_count
) -
1431 le32_to_cpu(es
->s_first_data_block
) +
1432 EXT3_BLOCKS_PER_GROUP(sb
) - 1) /
1433 EXT3_BLOCKS_PER_GROUP(sb
);
1434 db_count
= (sbi
->s_groups_count
+ EXT3_DESC_PER_BLOCK(sb
) - 1) /
1435 EXT3_DESC_PER_BLOCK(sb
);
1436 sbi
->s_group_desc
= kmalloc(db_count
* sizeof (struct buffer_head
*),
1438 if (sbi
->s_group_desc
== NULL
) {
1439 printk (KERN_ERR
"EXT3-fs: not enough memory\n");
1442 sbi
->s_debts
= kmalloc(sbi
->s_groups_count
* sizeof(u8
),
1444 if (!sbi
->s_debts
) {
1445 printk("EXT3-fs: not enough memory to allocate s_bgi\n");
1448 memset(sbi
->s_debts
, 0, sbi
->s_groups_count
* sizeof(u8
));
1450 percpu_counter_init(&sbi
->s_freeblocks_counter
);
1451 percpu_counter_init(&sbi
->s_freeinodes_counter
);
1452 percpu_counter_init(&sbi
->s_dirs_counter
);
1453 bgl_lock_init(&sbi
->s_blockgroup_lock
);
1455 for (i
= 0; i
< db_count
; i
++) {
1456 block
= descriptor_loc(sb
, logic_sb_block
, i
);
1457 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
1458 if (!sbi
->s_group_desc
[i
]) {
1459 printk (KERN_ERR
"EXT3-fs: "
1460 "can't read group descriptor %d\n", i
);
1465 if (!ext3_check_descriptors (sb
)) {
1466 printk (KERN_ERR
"EXT3-fs: group descriptors corrupted !\n");
1469 sbi
->s_gdb_count
= db_count
;
1470 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1471 spin_lock_init(&sbi
->s_next_gen_lock
);
1473 * set up enough so that it can read an inode
1475 sb
->s_op
= &ext3_sops
;
1476 sb
->s_export_op
= &ext3_export_ops
;
1478 sb
->s_qcop
= &ext3_qctl_operations
;
1479 sb
->dq_op
= &ext3_quota_operations
;
1481 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
1485 needs_recovery
= (es
->s_last_orphan
!= 0 ||
1486 EXT3_HAS_INCOMPAT_FEATURE(sb
,
1487 EXT3_FEATURE_INCOMPAT_RECOVER
));
1490 * The first inode we look at is the journal inode. Don't try
1491 * root first: it may be modified in the journal!
1493 if (!test_opt(sb
, NOLOAD
) &&
1494 EXT3_HAS_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
)) {
1495 if (ext3_load_journal(sb
, es
))
1497 } else if (journal_inum
) {
1498 if (ext3_create_journal(sb
, es
, journal_inum
))
1503 "ext3: No journal on filesystem on %s\n",
1508 /* We have now updated the journal if required, so we can
1509 * validate the data journaling mode. */
1510 switch (test_opt(sb
, DATA_FLAGS
)) {
1512 /* No mode set, assume a default based on the journal
1513 capabilities: ORDERED_DATA if the journal can
1514 cope, else JOURNAL_DATA */
1515 if (journal_check_available_features
1516 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
))
1517 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
1519 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
1522 case EXT3_MOUNT_ORDERED_DATA
:
1523 case EXT3_MOUNT_WRITEBACK_DATA
:
1524 if (!journal_check_available_features
1525 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
)) {
1526 printk(KERN_ERR
"EXT3-fs: Journal does not support "
1527 "requested data journaling mode\n");
1535 * The journal_load will have done any necessary log recovery,
1536 * so we can safely mount the rest of the filesystem now.
1539 root
= iget(sb
, EXT3_ROOT_INO
);
1540 sb
->s_root
= d_alloc_root(root
);
1542 printk(KERN_ERR
"EXT3-fs: get root inode failed\n");
1546 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1549 printk(KERN_ERR
"EXT3-fs: corrupt root inode, run e2fsck\n");
1553 ext3_setup_super (sb
, es
, sb
->s_flags
& MS_RDONLY
);
1555 * akpm: core read_super() calls in here with the superblock locked.
1556 * That deadlocks, because orphan cleanup needs to lock the superblock
1557 * in numerous places. Here we just pop the lock - it's relatively
1558 * harmless, because we are now ready to accept write_super() requests,
1559 * and aviro says that's the only reason for hanging onto the
1562 EXT3_SB(sb
)->s_mount_state
|= EXT3_ORPHAN_FS
;
1563 ext3_orphan_cleanup(sb
, es
);
1564 EXT3_SB(sb
)->s_mount_state
&= ~EXT3_ORPHAN_FS
;
1566 printk (KERN_INFO
"EXT3-fs: recovery complete.\n");
1567 ext3_mark_recovery_complete(sb
, es
);
1568 printk (KERN_INFO
"EXT3-fs: mounted filesystem with %s data mode.\n",
1569 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_JOURNAL_DATA
? "journal":
1570 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_ORDERED_DATA
? "ordered":
1573 percpu_counter_mod(&sbi
->s_freeblocks_counter
,
1574 ext3_count_free_blocks(sb
));
1575 percpu_counter_mod(&sbi
->s_freeinodes_counter
,
1576 ext3_count_free_inodes(sb
));
1577 percpu_counter_mod(&sbi
->s_dirs_counter
,
1578 ext3_count_dirs(sb
));
1583 journal_destroy(sbi
->s_journal
);
1585 kfree(sbi
->s_debts
);
1586 for (i
= 0; i
< db_count
; i
++)
1587 brelse(sbi
->s_group_desc
[i
]);
1588 kfree(sbi
->s_group_desc
);
1591 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1592 if (sbi
->s_qf_names
[i
])
1593 kfree(sbi
->s_qf_names
[i
]);
1596 ext3_blkdev_remove(sbi
);
1599 sb
->s_fs_info
= NULL
;
1605 * Setup any per-fs journal parameters now. We'll do this both on
1606 * initial mount, once the journal has been initialised but before we've
1607 * done any recovery; and again on any subsequent remount.
1609 static void ext3_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
1611 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1613 if (sbi
->s_commit_interval
)
1614 journal
->j_commit_interval
= sbi
->s_commit_interval
;
1615 /* We could also set up an ext3-specific default for the commit
1616 * interval here, but for now we'll just fall back to the jbd
1619 spin_lock(&journal
->j_state_lock
);
1620 if (test_opt(sb
, BARRIER
))
1621 journal
->j_flags
|= JFS_BARRIER
;
1623 journal
->j_flags
&= ~JFS_BARRIER
;
1624 spin_unlock(&journal
->j_state_lock
);
1627 static journal_t
*ext3_get_journal(struct super_block
*sb
, int journal_inum
)
1629 struct inode
*journal_inode
;
1632 /* First, test for the existence of a valid inode on disk. Bad
1633 * things happen if we iget() an unused inode, as the subsequent
1634 * iput() will try to delete it. */
1636 journal_inode
= iget(sb
, journal_inum
);
1637 if (!journal_inode
) {
1638 printk(KERN_ERR
"EXT3-fs: no journal found.\n");
1641 if (!journal_inode
->i_nlink
) {
1642 make_bad_inode(journal_inode
);
1643 iput(journal_inode
);
1644 printk(KERN_ERR
"EXT3-fs: journal inode is deleted.\n");
1648 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1649 journal_inode
, journal_inode
->i_size
);
1650 if (is_bad_inode(journal_inode
) || !S_ISREG(journal_inode
->i_mode
)) {
1651 printk(KERN_ERR
"EXT3-fs: invalid journal inode.\n");
1652 iput(journal_inode
);
1656 journal
= journal_init_inode(journal_inode
);
1658 printk(KERN_ERR
"EXT3-fs: Could not load journal inode\n");
1659 iput(journal_inode
);
1662 journal
->j_private
= sb
;
1663 ext3_init_journal_params(sb
, journal
);
1667 static journal_t
*ext3_get_dev_journal(struct super_block
*sb
,
1670 struct buffer_head
* bh
;
1674 int hblock
, blocksize
;
1675 unsigned long sb_block
;
1676 unsigned long offset
;
1677 struct ext3_super_block
* es
;
1678 struct block_device
*bdev
;
1680 bdev
= ext3_blkdev_get(j_dev
);
1684 if (bd_claim(bdev
, sb
)) {
1686 "EXT3: failed to claim external journal device.\n");
1691 blocksize
= sb
->s_blocksize
;
1692 hblock
= bdev_hardsect_size(bdev
);
1693 if (blocksize
< hblock
) {
1695 "EXT3-fs: blocksize too small for journal device.\n");
1699 sb_block
= EXT3_MIN_BLOCK_SIZE
/ blocksize
;
1700 offset
= EXT3_MIN_BLOCK_SIZE
% blocksize
;
1701 set_blocksize(bdev
, blocksize
);
1702 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
1703 printk(KERN_ERR
"EXT3-fs: couldn't read superblock of "
1704 "external journal\n");
1708 es
= (struct ext3_super_block
*) (((char *)bh
->b_data
) + offset
);
1709 if ((le16_to_cpu(es
->s_magic
) != EXT3_SUPER_MAGIC
) ||
1710 !(le32_to_cpu(es
->s_feature_incompat
) &
1711 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
1712 printk(KERN_ERR
"EXT3-fs: external journal has "
1713 "bad superblock\n");
1718 if (memcmp(EXT3_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
1719 printk(KERN_ERR
"EXT3-fs: journal UUID does not match\n");
1724 len
= le32_to_cpu(es
->s_blocks_count
);
1725 start
= sb_block
+ 1;
1726 brelse(bh
); /* we're done with the superblock */
1728 journal
= journal_init_dev(bdev
, sb
->s_bdev
,
1729 start
, len
, blocksize
);
1731 printk(KERN_ERR
"EXT3-fs: failed to create device journal\n");
1734 journal
->j_private
= sb
;
1735 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
1736 wait_on_buffer(journal
->j_sb_buffer
);
1737 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
1738 printk(KERN_ERR
"EXT3-fs: I/O error on journal device\n");
1741 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
1742 printk(KERN_ERR
"EXT3-fs: External journal has more than one "
1743 "user (unsupported) - %d\n",
1744 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
1747 EXT3_SB(sb
)->journal_bdev
= bdev
;
1748 ext3_init_journal_params(sb
, journal
);
1751 journal_destroy(journal
);
1753 ext3_blkdev_put(bdev
);
1757 static int ext3_load_journal(struct super_block
* sb
,
1758 struct ext3_super_block
* es
)
1761 int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
1762 dev_t journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
1764 int really_read_only
;
1766 really_read_only
= bdev_read_only(sb
->s_bdev
);
1769 * Are we loading a blank journal or performing recovery after a
1770 * crash? For recovery, we need to check in advance whether we
1771 * can get read-write access to the device.
1774 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
)) {
1775 if (sb
->s_flags
& MS_RDONLY
) {
1776 printk(KERN_INFO
"EXT3-fs: INFO: recovery "
1777 "required on readonly filesystem.\n");
1778 if (really_read_only
) {
1779 printk(KERN_ERR
"EXT3-fs: write access "
1780 "unavailable, cannot proceed.\n");
1783 printk (KERN_INFO
"EXT3-fs: write access will "
1784 "be enabled during recovery.\n");
1788 if (journal_inum
&& journal_dev
) {
1789 printk(KERN_ERR
"EXT3-fs: filesystem has both journal "
1790 "and inode journals!\n");
1795 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
1798 if (!(journal
= ext3_get_dev_journal(sb
, journal_dev
)))
1802 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
1803 err
= journal_update_format(journal
);
1805 printk(KERN_ERR
"EXT3-fs: error updating journal.\n");
1806 journal_destroy(journal
);
1811 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
))
1812 err
= journal_wipe(journal
, !really_read_only
);
1814 err
= journal_load(journal
);
1817 printk(KERN_ERR
"EXT3-fs: error loading journal.\n");
1818 journal_destroy(journal
);
1822 EXT3_SB(sb
)->s_journal
= journal
;
1823 ext3_clear_journal_err(sb
, es
);
1827 static int ext3_create_journal(struct super_block
* sb
,
1828 struct ext3_super_block
* es
,
1833 if (sb
->s_flags
& MS_RDONLY
) {
1834 printk(KERN_ERR
"EXT3-fs: readonly filesystem when trying to "
1835 "create journal.\n");
1839 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
1842 printk(KERN_INFO
"EXT3-fs: creating new journal on inode %d\n",
1845 if (journal_create(journal
)) {
1846 printk(KERN_ERR
"EXT3-fs: error creating journal.\n");
1847 journal_destroy(journal
);
1851 EXT3_SB(sb
)->s_journal
= journal
;
1853 ext3_update_dynamic_rev(sb
);
1854 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
1855 EXT3_SET_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
);
1857 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
1860 /* Make sure we flush the recovery flag to disk. */
1861 ext3_commit_super(sb
, es
, 1);
1866 static void ext3_commit_super (struct super_block
* sb
,
1867 struct ext3_super_block
* es
,
1870 struct buffer_head
*sbh
= EXT3_SB(sb
)->s_sbh
;
1874 es
->s_wtime
= cpu_to_le32(get_seconds());
1875 es
->s_free_blocks_count
= cpu_to_le32(ext3_count_free_blocks(sb
));
1876 es
->s_free_inodes_count
= cpu_to_le32(ext3_count_free_inodes(sb
));
1877 BUFFER_TRACE(sbh
, "marking dirty");
1878 mark_buffer_dirty(sbh
);
1880 sync_dirty_buffer(sbh
);
1885 * Have we just finished recovery? If so, and if we are mounting (or
1886 * remounting) the filesystem readonly, then we will end up with a
1887 * consistent fs on disk. Record that fact.
1889 static void ext3_mark_recovery_complete(struct super_block
* sb
,
1890 struct ext3_super_block
* es
)
1892 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
1894 journal_lock_updates(journal
);
1895 journal_flush(journal
);
1896 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
) &&
1897 sb
->s_flags
& MS_RDONLY
) {
1898 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
1900 ext3_commit_super(sb
, es
, 1);
1902 journal_unlock_updates(journal
);
1906 * If we are mounting (or read-write remounting) a filesystem whose journal
1907 * has recorded an error from a previous lifetime, move that error to the
1908 * main filesystem now.
1910 static void ext3_clear_journal_err(struct super_block
* sb
,
1911 struct ext3_super_block
* es
)
1917 journal
= EXT3_SB(sb
)->s_journal
;
1920 * Now check for any error status which may have been recorded in the
1921 * journal by a prior ext3_error() or ext3_abort()
1924 j_errno
= journal_errno(journal
);
1928 errstr
= ext3_decode_error(sb
, j_errno
, nbuf
);
1929 ext3_warning(sb
, __FUNCTION__
, "Filesystem error recorded "
1930 "from previous mount: %s", errstr
);
1931 ext3_warning(sb
, __FUNCTION__
, "Marking fs in need of "
1932 "filesystem check.");
1934 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
1935 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
1936 ext3_commit_super (sb
, es
, 1);
1938 journal_clear_err(journal
);
1943 * Force the running and committing transactions to commit,
1944 * and wait on the commit.
1946 int ext3_force_commit(struct super_block
*sb
)
1951 if (sb
->s_flags
& MS_RDONLY
)
1954 journal
= EXT3_SB(sb
)->s_journal
;
1956 ret
= ext3_journal_force_commit(journal
);
1961 * Ext3 always journals updates to the superblock itself, so we don't
1962 * have to propagate any other updates to the superblock on disk at this
1963 * point. Just start an async writeback to get the buffers on their way
1966 * This implicitly triggers the writebehind on sync().
1969 void ext3_write_super (struct super_block
* sb
)
1971 if (down_trylock(&sb
->s_lock
) == 0)
1976 static int ext3_sync_fs(struct super_block
*sb
, int wait
)
1981 if (journal_start_commit(EXT3_SB(sb
)->s_journal
, &target
)) {
1983 log_wait_commit(EXT3_SB(sb
)->s_journal
, target
);
1989 * LVM calls this function before a (read-only) snapshot is created. This
1990 * gives us a chance to flush the journal completely and mark the fs clean.
1992 void ext3_write_super_lockfs(struct super_block
*sb
)
1996 if (!(sb
->s_flags
& MS_RDONLY
)) {
1997 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
1999 /* Now we set up the journal barrier. */
2000 journal_lock_updates(journal
);
2001 journal_flush(journal
);
2003 /* Journal blocked and flushed, clear needs_recovery flag. */
2004 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2005 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2010 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2011 * flag here, even though the filesystem is not technically dirty yet.
2013 void ext3_unlockfs(struct super_block
*sb
)
2015 if (!(sb
->s_flags
& MS_RDONLY
)) {
2017 /* Reser the needs_recovery flag before the fs is unlocked. */
2018 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2019 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2021 journal_unlock_updates(EXT3_SB(sb
)->s_journal
);
2025 int ext3_remount (struct super_block
* sb
, int * flags
, char * data
)
2027 struct ext3_super_block
* es
;
2028 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2032 * Allow the "check" option to be passed as a remount option.
2034 if (!parse_options(data
, sb
, &tmp
, 1))
2037 if (sbi
->s_mount_opt
& EXT3_MOUNT_ABORT
)
2038 ext3_abort(sb
, __FUNCTION__
, "Abort forced by user");
2040 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2041 ((sbi
->s_mount_opt
& EXT3_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2045 ext3_init_journal_params(sb
, sbi
->s_journal
);
2047 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
2048 if (sbi
->s_mount_opt
& EXT3_MOUNT_ABORT
)
2051 if (*flags
& MS_RDONLY
) {
2053 * First of all, the unconditional stuff we have to do
2054 * to disable replay of the journal when we next remount
2056 sb
->s_flags
|= MS_RDONLY
;
2059 * OK, test if we are remounting a valid rw partition
2060 * readonly, and if so set the rdonly flag and then
2061 * mark the partition as valid again.
2063 if (!(es
->s_state
& cpu_to_le16(EXT3_VALID_FS
)) &&
2064 (sbi
->s_mount_state
& EXT3_VALID_FS
))
2065 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
2067 ext3_mark_recovery_complete(sb
, es
);
2070 if ((ret
= EXT3_HAS_RO_COMPAT_FEATURE(sb
,
2071 ~EXT3_FEATURE_RO_COMPAT_SUPP
))) {
2072 printk(KERN_WARNING
"EXT3-fs: %s: couldn't "
2073 "remount RDWR because of unsupported "
2074 "optional features (%x).\n",
2075 sb
->s_id
, le32_to_cpu(ret
));
2079 * Mounting a RDONLY partition read-write, so reread
2080 * and store the current valid flag. (It may have
2081 * been changed by e2fsck since we originally mounted
2084 ext3_clear_journal_err(sb
, es
);
2085 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2086 if (!ext3_setup_super (sb
, es
, 0))
2087 sb
->s_flags
&= ~MS_RDONLY
;
2093 int ext3_statfs (struct super_block
* sb
, struct kstatfs
* buf
)
2095 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
2096 unsigned long overhead
;
2099 if (test_opt (sb
, MINIX_DF
))
2103 * Compute the overhead (FS structures)
2107 * All of the blocks before first_data_block are
2110 overhead
= le32_to_cpu(es
->s_first_data_block
);
2113 * Add the overhead attributed to the superblock and
2114 * block group descriptors. If the sparse superblocks
2115 * feature is turned on, then not all groups have this.
2117 for (i
= 0; i
< EXT3_SB(sb
)->s_groups_count
; i
++)
2118 overhead
+= ext3_bg_has_super(sb
, i
) +
2119 ext3_bg_num_gdb(sb
, i
);
2122 * Every block group has an inode bitmap, a block
2123 * bitmap, and an inode table.
2125 overhead
+= (EXT3_SB(sb
)->s_groups_count
*
2126 (2 + EXT3_SB(sb
)->s_itb_per_group
));
2129 buf
->f_type
= EXT3_SUPER_MAGIC
;
2130 buf
->f_bsize
= sb
->s_blocksize
;
2131 buf
->f_blocks
= le32_to_cpu(es
->s_blocks_count
) - overhead
;
2132 buf
->f_bfree
= ext3_count_free_blocks (sb
);
2133 buf
->f_bavail
= buf
->f_bfree
- le32_to_cpu(es
->s_r_blocks_count
);
2134 if (buf
->f_bfree
< le32_to_cpu(es
->s_r_blocks_count
))
2136 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
2137 buf
->f_ffree
= ext3_count_free_inodes (sb
);
2138 buf
->f_namelen
= EXT3_NAME_LEN
;
2142 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2143 * is locked for write. Otherwise the are possible deadlocks:
2144 * Process 1 Process 2
2145 * ext3_create() quota_sync()
2146 * journal_start() write_dquot()
2147 * DQUOT_INIT() down(dqio_sem)
2148 * down(dqio_sem) journal_start()
2154 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
2156 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
]->f_dentry
->d_inode
;
2159 static int ext3_dquot_initialize(struct inode
*inode
, int type
)
2164 /* We may create quota structure so we need to reserve enough blocks */
2165 handle
= ext3_journal_start(inode
, 2*EXT3_QUOTA_INIT_BLOCKS
);
2167 return PTR_ERR(handle
);
2168 ret
= dquot_initialize(inode
, type
);
2169 err
= ext3_journal_stop(handle
);
2175 static int ext3_dquot_drop(struct inode
*inode
)
2180 /* We may delete quota structure so we need to reserve enough blocks */
2181 handle
= ext3_journal_start(inode
, 2*EXT3_QUOTA_INIT_BLOCKS
);
2183 return PTR_ERR(handle
);
2184 ret
= dquot_drop(inode
);
2185 err
= ext3_journal_stop(handle
);
2191 static int ext3_write_dquot(struct dquot
*dquot
)
2196 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2197 EXT3_QUOTA_TRANS_BLOCKS
);
2199 return PTR_ERR(handle
);
2200 ret
= dquot_commit(dquot
);
2201 err
= ext3_journal_stop(handle
);
2207 static int ext3_acquire_dquot(struct dquot
*dquot
)
2212 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2213 EXT3_QUOTA_INIT_BLOCKS
);
2215 return PTR_ERR(handle
);
2216 ret
= dquot_acquire(dquot
);
2217 err
= ext3_journal_stop(handle
);
2223 static int ext3_release_dquot(struct dquot
*dquot
)
2228 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2229 EXT3_QUOTA_INIT_BLOCKS
);
2231 return PTR_ERR(handle
);
2232 ret
= dquot_release(dquot
);
2233 err
= ext3_journal_stop(handle
);
2239 static int ext3_mark_dquot_dirty(struct dquot
*dquot
)
2241 /* Are we journalling quotas? */
2242 if (EXT3_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
2243 EXT3_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
2244 dquot_mark_dquot_dirty(dquot
);
2245 return ext3_write_dquot(dquot
);
2247 return dquot_mark_dquot_dirty(dquot
);
2251 static int ext3_write_info(struct super_block
*sb
, int type
)
2256 /* Data block + inode block */
2257 handle
= ext3_journal_start(sb
->s_root
->d_inode
, 2);
2259 return PTR_ERR(handle
);
2260 ret
= dquot_commit_info(sb
, type
);
2261 err
= ext3_journal_stop(handle
);
2268 * Turn on quotas during mount time - we need to find
2269 * the quota file and such...
2271 static int ext3_quota_on_mount(struct super_block
*sb
, int type
)
2274 struct dentry
*dentry
;
2275 struct qstr name
= { .name
= EXT3_SB(sb
)->s_qf_names
[type
],
2277 .len
= strlen(EXT3_SB(sb
)->s_qf_names
[type
])};
2279 dentry
= lookup_hash(&name
, sb
->s_root
);
2281 return PTR_ERR(dentry
);
2282 err
= vfs_quota_on_mount(type
, EXT3_SB(sb
)->s_jquota_fmt
, dentry
);
2285 /* We keep the dentry reference if everything went ok - we drop it
2286 * on quota_off time */
2290 /* Turn quotas off during mount time */
2291 static int ext3_quota_off_mount(struct super_block
*sb
, int type
)
2294 struct dentry
*dentry
;
2296 dentry
= sb_dqopt(sb
)->files
[type
]->f_dentry
;
2297 err
= vfs_quota_off_mount(sb
, type
);
2298 /* We invalidate dentry - it has at least wrong hash... */
2299 d_invalidate(dentry
);
2305 * Standard function to be called on quota_on
2307 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
,
2311 struct nameidata nd
;
2313 /* Not journalling quota? */
2314 if (!EXT3_SB(sb
)->s_qf_names
[USRQUOTA
] &&
2315 !EXT3_SB(sb
)->s_qf_names
[GRPQUOTA
])
2316 return vfs_quota_on(sb
, type
, format_id
, path
);
2317 err
= path_lookup(path
, LOOKUP_FOLLOW
, &nd
);
2320 /* Quotafile not on the same filesystem? */
2321 if (nd
.mnt
->mnt_sb
!= sb
)
2323 /* Quotafile not of fs root? */
2324 if (nd
.dentry
->d_parent
->d_inode
!= sb
->s_root
->d_inode
)
2326 "EXT3-fs: Quota file not on filesystem root. "
2327 "Journalled quota will not work.\n");
2328 if (!ext3_should_journal_data(nd
.dentry
->d_inode
))
2329 printk(KERN_WARNING
"EXT3-fs: Quota file does not have "
2330 "data-journalling. Journalled quota will not work.\n");
2332 return vfs_quota_on(sb
, type
, format_id
, path
);
2337 static struct super_block
*ext3_get_sb(struct file_system_type
*fs_type
,
2338 int flags
, const char *dev_name
, void *data
)
2340 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext3_fill_super
);
2343 static struct file_system_type ext3_fs_type
= {
2344 .owner
= THIS_MODULE
,
2346 .get_sb
= ext3_get_sb
,
2347 .kill_sb
= kill_block_super
,
2348 .fs_flags
= FS_REQUIRES_DEV
,
2351 static int __init
init_ext3_fs(void)
2353 int err
= init_ext3_xattr();
2356 err
= init_inodecache();
2359 err
= register_filesystem(&ext3_fs_type
);
2364 destroy_inodecache();
2370 static void __exit
exit_ext3_fs(void)
2372 unregister_filesystem(&ext3_fs_type
);
2373 destroy_inodecache();
2377 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2378 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2379 MODULE_LICENSE("GPL");
2380 module_init(init_ext3_fs
)
2381 module_exit(exit_ext3_fs
)