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
);
53 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
55 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
);
56 static int ext3_statfs (struct super_block
* sb
, struct kstatfs
* buf
);
57 static void ext3_unlockfs(struct super_block
*sb
);
58 static void ext3_write_super (struct super_block
* sb
);
59 static void ext3_write_super_lockfs(struct super_block
*sb
);
62 * Wrappers for journal_start/end.
64 * The only special thing we need to do here is to make sure that all
65 * journal_end calls result in the superblock being marked dirty, so
66 * that sync() will call the filesystem's write_super callback if
69 handle_t
*ext3_journal_start_sb(struct super_block
*sb
, int nblocks
)
73 if (sb
->s_flags
& MS_RDONLY
)
74 return ERR_PTR(-EROFS
);
76 /* Special case here: if the journal has aborted behind our
77 * backs (eg. EIO in the commit thread), then we still need to
78 * take the FS itself readonly cleanly. */
79 journal
= EXT3_SB(sb
)->s_journal
;
80 if (is_journal_aborted(journal
)) {
81 ext3_abort(sb
, __FUNCTION__
,
82 "Detected aborted journal");
83 return ERR_PTR(-EROFS
);
86 return journal_start(journal
, nblocks
);
90 * The only special thing we need to do here is to make sure that all
91 * journal_stop calls result in the superblock being marked dirty, so
92 * that sync() will call the filesystem's write_super callback if
95 int __ext3_journal_stop(const char *where
, handle_t
*handle
)
97 struct super_block
*sb
;
101 sb
= handle
->h_transaction
->t_journal
->j_private
;
103 rc
= journal_stop(handle
);
108 __ext3_std_error(sb
, where
, err
);
112 void ext3_journal_abort_handle(const char *caller
, const char *err_fn
,
113 struct buffer_head
*bh
, handle_t
*handle
, int err
)
116 const char *errstr
= ext3_decode_error(NULL
, err
, nbuf
);
119 BUFFER_TRACE(bh
, "abort");
124 if (is_handle_aborted(handle
))
127 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
128 caller
, errstr
, err_fn
);
130 journal_abort_handle(handle
);
133 /* Deal with the reporting of failure conditions on a filesystem such as
134 * inconsistencies detected or read IO failures.
136 * On ext2, we can store the error state of the filesystem in the
137 * superblock. That is not possible on ext3, because we may have other
138 * write ordering constraints on the superblock which prevent us from
139 * writing it out straight away; and given that the journal is about to
140 * be aborted, we can't rely on the current, or future, transactions to
141 * write out the superblock safely.
143 * We'll just use the journal_abort() error code to record an error in
144 * the journal instead. On recovery, the journal will compain about
145 * that error until we've noted it down and cleared it.
148 static void ext3_handle_error(struct super_block
*sb
)
150 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
152 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
153 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
155 if (sb
->s_flags
& MS_RDONLY
)
158 if (test_opt (sb
, ERRORS_RO
)) {
159 printk (KERN_CRIT
"Remounting filesystem read-only\n");
160 sb
->s_flags
|= MS_RDONLY
;
162 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
164 EXT3_SB(sb
)->s_mount_opt
|= EXT3_MOUNT_ABORT
;
166 journal_abort(journal
, -EIO
);
168 if (test_opt(sb
, ERRORS_PANIC
))
169 panic("EXT3-fs (device %s): panic forced after error\n",
171 ext3_commit_super(sb
, es
, 1);
174 void ext3_error (struct super_block
* sb
, const char * function
,
175 const char * fmt
, ...)
180 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
185 ext3_handle_error(sb
);
188 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
195 errstr
= "IO failure";
198 errstr
= "Out of memory";
201 if (!sb
|| EXT3_SB(sb
)->s_journal
->j_flags
& JFS_ABORT
)
202 errstr
= "Journal has aborted";
204 errstr
= "Readonly filesystem";
207 /* If the caller passed in an extra buffer for unknown
208 * errors, textualise them now. Else we just return
211 /* Check for truncated error codes... */
212 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
221 /* __ext3_std_error decodes expected errors from journaling functions
222 * automatically and invokes the appropriate error response. */
224 void __ext3_std_error (struct super_block
* sb
, const char * function
,
230 /* Special case: if the error is EROFS, and we're not already
231 * inside a transaction, then there's really no point in logging
233 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
234 (sb
->s_flags
& MS_RDONLY
))
237 errstr
= ext3_decode_error(sb
, errno
, nbuf
);
238 printk (KERN_CRIT
"EXT3-fs error (device %s) in %s: %s\n",
239 sb
->s_id
, function
, errstr
);
241 ext3_handle_error(sb
);
245 * ext3_abort is a much stronger failure handler than ext3_error. The
246 * abort function may be used to deal with unrecoverable failures such
247 * as journal IO errors or ENOMEM at a critical moment in log management.
249 * We unconditionally force the filesystem into an ABORT|READONLY state,
250 * unless the error response on the fs has been set to panic in which
251 * case we take the easy way out and panic immediately.
254 void ext3_abort (struct super_block
* sb
, const char * function
,
255 const char * fmt
, ...)
259 printk (KERN_CRIT
"ext3_abort called.\n");
262 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
267 if (test_opt(sb
, ERRORS_PANIC
))
268 panic("EXT3-fs panic from previous error\n");
270 if (sb
->s_flags
& MS_RDONLY
)
273 printk(KERN_CRIT
"Remounting filesystem read-only\n");
274 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
275 sb
->s_flags
|= MS_RDONLY
;
276 EXT3_SB(sb
)->s_mount_opt
|= EXT3_MOUNT_ABORT
;
277 journal_abort(EXT3_SB(sb
)->s_journal
, -EIO
);
280 void ext3_warning (struct super_block
* sb
, const char * function
,
281 const char * fmt
, ...)
286 printk(KERN_WARNING
"EXT3-fs warning (device %s): %s: ",
293 void ext3_update_dynamic_rev(struct super_block
*sb
)
295 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
297 if (le32_to_cpu(es
->s_rev_level
) > EXT3_GOOD_OLD_REV
)
300 ext3_warning(sb
, __FUNCTION__
,
301 "updating to rev %d because of new feature flag, "
302 "running e2fsck is recommended",
305 es
->s_first_ino
= cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO
);
306 es
->s_inode_size
= cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE
);
307 es
->s_rev_level
= cpu_to_le32(EXT3_DYNAMIC_REV
);
308 /* leave es->s_feature_*compat flags alone */
309 /* es->s_uuid will be set by e2fsck if empty */
312 * The rest of the superblock fields should be zero, and if not it
313 * means they are likely already in use, so leave them alone. We
314 * can leave it up to e2fsck to clean up any inconsistencies there.
319 * Open the external journal device
321 static struct block_device
*ext3_blkdev_get(dev_t dev
)
323 struct block_device
*bdev
;
324 char b
[BDEVNAME_SIZE
];
326 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
332 printk(KERN_ERR
"EXT3: failed to open journal device %s: %ld\n",
333 __bdevname(dev
, b
), PTR_ERR(bdev
));
338 * Release the journal device
340 static int ext3_blkdev_put(struct block_device
*bdev
)
343 return blkdev_put(bdev
);
346 static int ext3_blkdev_remove(struct ext3_sb_info
*sbi
)
348 struct block_device
*bdev
;
351 bdev
= sbi
->journal_bdev
;
353 ret
= ext3_blkdev_put(bdev
);
354 sbi
->journal_bdev
= NULL
;
359 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
361 return &list_entry(l
, struct ext3_inode_info
, i_orphan
)->vfs_inode
;
364 static void dump_orphan_list(struct super_block
*sb
, struct ext3_sb_info
*sbi
)
368 printk(KERN_ERR
"sb orphan head is %d\n",
369 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
371 printk(KERN_ERR
"sb_info orphan list:\n");
372 list_for_each(l
, &sbi
->s_orphan
) {
373 struct inode
*inode
= orphan_list_entry(l
);
375 "inode %s:%ld at %p: mode %o, nlink %d, next %d\n",
376 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
377 inode
->i_mode
, inode
->i_nlink
,
382 static void ext3_put_super (struct super_block
* sb
)
384 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
385 struct ext3_super_block
*es
= sbi
->s_es
;
388 ext3_xattr_put_super(sb
);
389 journal_destroy(sbi
->s_journal
);
390 if (!(sb
->s_flags
& MS_RDONLY
)) {
391 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
392 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
393 BUFFER_TRACE(sbi
->s_sbh
, "marking dirty");
394 mark_buffer_dirty(sbi
->s_sbh
);
395 ext3_commit_super(sb
, es
, 1);
398 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
399 brelse(sbi
->s_group_desc
[i
]);
400 kfree(sbi
->s_group_desc
);
401 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
402 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
403 percpu_counter_destroy(&sbi
->s_dirs_counter
);
406 for (i
= 0; i
< MAXQUOTAS
; i
++)
407 kfree(sbi
->s_qf_names
[i
]);
410 /* Debugging code just in case the in-memory inode orphan list
411 * isn't empty. The on-disk one can be non-empty if we've
412 * detected an error and taken the fs readonly, but the
413 * in-memory list had better be clean by this point. */
414 if (!list_empty(&sbi
->s_orphan
))
415 dump_orphan_list(sb
, sbi
);
416 J_ASSERT(list_empty(&sbi
->s_orphan
));
418 invalidate_bdev(sb
->s_bdev
, 0);
419 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
421 * Invalidate the journal device's buffers. We don't want them
422 * floating about in memory - the physical journal device may
423 * hotswapped, and it breaks the `ro-after' testing code.
425 sync_blockdev(sbi
->journal_bdev
);
426 invalidate_bdev(sbi
->journal_bdev
, 0);
427 ext3_blkdev_remove(sbi
);
429 sb
->s_fs_info
= NULL
;
434 static kmem_cache_t
*ext3_inode_cachep
;
437 * Called inside transaction, so use GFP_NOFS
439 static struct inode
*ext3_alloc_inode(struct super_block
*sb
)
441 struct ext3_inode_info
*ei
;
443 ei
= kmem_cache_alloc(ext3_inode_cachep
, SLAB_NOFS
);
446 #ifdef CONFIG_EXT3_FS_POSIX_ACL
447 ei
->i_acl
= EXT3_ACL_NOT_CACHED
;
448 ei
->i_default_acl
= EXT3_ACL_NOT_CACHED
;
450 ei
->i_block_alloc_info
= NULL
;
451 ei
->vfs_inode
.i_version
= 1;
452 return &ei
->vfs_inode
;
455 static void ext3_destroy_inode(struct inode
*inode
)
457 kmem_cache_free(ext3_inode_cachep
, EXT3_I(inode
));
460 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
462 struct ext3_inode_info
*ei
= (struct ext3_inode_info
*) foo
;
464 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
465 SLAB_CTOR_CONSTRUCTOR
) {
466 INIT_LIST_HEAD(&ei
->i_orphan
);
467 #ifdef CONFIG_EXT3_FS_XATTR
468 init_rwsem(&ei
->xattr_sem
);
470 init_MUTEX(&ei
->truncate_sem
);
471 inode_init_once(&ei
->vfs_inode
);
475 static int init_inodecache(void)
477 ext3_inode_cachep
= kmem_cache_create("ext3_inode_cache",
478 sizeof(struct ext3_inode_info
),
479 0, SLAB_RECLAIM_ACCOUNT
,
481 if (ext3_inode_cachep
== NULL
)
486 static void destroy_inodecache(void)
488 if (kmem_cache_destroy(ext3_inode_cachep
))
489 printk(KERN_INFO
"ext3_inode_cache: not all structures were freed\n");
492 static void ext3_clear_inode(struct inode
*inode
)
494 struct ext3_block_alloc_info
*rsv
= EXT3_I(inode
)->i_block_alloc_info
;
495 #ifdef CONFIG_EXT3_FS_POSIX_ACL
496 if (EXT3_I(inode
)->i_acl
&&
497 EXT3_I(inode
)->i_acl
!= EXT3_ACL_NOT_CACHED
) {
498 posix_acl_release(EXT3_I(inode
)->i_acl
);
499 EXT3_I(inode
)->i_acl
= EXT3_ACL_NOT_CACHED
;
501 if (EXT3_I(inode
)->i_default_acl
&&
502 EXT3_I(inode
)->i_default_acl
!= EXT3_ACL_NOT_CACHED
) {
503 posix_acl_release(EXT3_I(inode
)->i_default_acl
);
504 EXT3_I(inode
)->i_default_acl
= EXT3_ACL_NOT_CACHED
;
507 ext3_discard_reservation(inode
);
508 EXT3_I(inode
)->i_block_alloc_info
= NULL
;
514 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
515 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
517 static int ext3_dquot_initialize(struct inode
*inode
, int type
);
518 static int ext3_dquot_drop(struct inode
*inode
);
519 static int ext3_write_dquot(struct dquot
*dquot
);
520 static int ext3_acquire_dquot(struct dquot
*dquot
);
521 static int ext3_release_dquot(struct dquot
*dquot
);
522 static int ext3_mark_dquot_dirty(struct dquot
*dquot
);
523 static int ext3_write_info(struct super_block
*sb
, int type
);
524 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
, char *path
);
525 static int ext3_quota_on_mount(struct super_block
*sb
, int type
);
526 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
527 size_t len
, loff_t off
);
528 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
529 const char *data
, size_t len
, loff_t off
);
531 static struct dquot_operations ext3_quota_operations
= {
532 .initialize
= ext3_dquot_initialize
,
533 .drop
= ext3_dquot_drop
,
534 .alloc_space
= dquot_alloc_space
,
535 .alloc_inode
= dquot_alloc_inode
,
536 .free_space
= dquot_free_space
,
537 .free_inode
= dquot_free_inode
,
538 .transfer
= dquot_transfer
,
539 .write_dquot
= ext3_write_dquot
,
540 .acquire_dquot
= ext3_acquire_dquot
,
541 .release_dquot
= ext3_release_dquot
,
542 .mark_dirty
= ext3_mark_dquot_dirty
,
543 .write_info
= ext3_write_info
546 static struct quotactl_ops ext3_qctl_operations
= {
547 .quota_on
= ext3_quota_on
,
548 .quota_off
= vfs_quota_off
,
549 .quota_sync
= vfs_quota_sync
,
550 .get_info
= vfs_get_dqinfo
,
551 .set_info
= vfs_set_dqinfo
,
552 .get_dqblk
= vfs_get_dqblk
,
553 .set_dqblk
= vfs_set_dqblk
557 static struct super_operations ext3_sops
= {
558 .alloc_inode
= ext3_alloc_inode
,
559 .destroy_inode
= ext3_destroy_inode
,
560 .read_inode
= ext3_read_inode
,
561 .write_inode
= ext3_write_inode
,
562 .dirty_inode
= ext3_dirty_inode
,
563 .delete_inode
= ext3_delete_inode
,
564 .put_super
= ext3_put_super
,
565 .write_super
= ext3_write_super
,
566 .sync_fs
= ext3_sync_fs
,
567 .write_super_lockfs
= ext3_write_super_lockfs
,
568 .unlockfs
= ext3_unlockfs
,
569 .statfs
= ext3_statfs
,
570 .remount_fs
= ext3_remount
,
571 .clear_inode
= ext3_clear_inode
,
573 .quota_read
= ext3_quota_read
,
574 .quota_write
= ext3_quota_write
,
578 struct dentry
*ext3_get_parent(struct dentry
*child
);
579 static struct export_operations ext3_export_ops
= {
580 .get_parent
= ext3_get_parent
,
584 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
585 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
586 Opt_nouid32
, Opt_check
, Opt_nocheck
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
587 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
588 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
,
589 Opt_commit
, Opt_journal_update
, Opt_journal_inum
,
590 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
591 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
592 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
593 Opt_ignore
, Opt_barrier
, Opt_err
, Opt_resize
,
596 static match_table_t tokens
= {
597 {Opt_bsd_df
, "bsddf"},
598 {Opt_minix_df
, "minixdf"},
599 {Opt_grpid
, "grpid"},
600 {Opt_grpid
, "bsdgroups"},
601 {Opt_nogrpid
, "nogrpid"},
602 {Opt_nogrpid
, "sysvgroups"},
603 {Opt_resgid
, "resgid=%u"},
604 {Opt_resuid
, "resuid=%u"},
606 {Opt_err_cont
, "errors=continue"},
607 {Opt_err_panic
, "errors=panic"},
608 {Opt_err_ro
, "errors=remount-ro"},
609 {Opt_nouid32
, "nouid32"},
610 {Opt_nocheck
, "nocheck"},
611 {Opt_nocheck
, "check=none"},
612 {Opt_check
, "check"},
613 {Opt_debug
, "debug"},
614 {Opt_oldalloc
, "oldalloc"},
615 {Opt_orlov
, "orlov"},
616 {Opt_user_xattr
, "user_xattr"},
617 {Opt_nouser_xattr
, "nouser_xattr"},
619 {Opt_noacl
, "noacl"},
620 {Opt_reservation
, "reservation"},
621 {Opt_noreservation
, "noreservation"},
622 {Opt_noload
, "noload"},
624 {Opt_commit
, "commit=%u"},
625 {Opt_journal_update
, "journal=update"},
626 {Opt_journal_inum
, "journal=%u"},
627 {Opt_abort
, "abort"},
628 {Opt_data_journal
, "data=journal"},
629 {Opt_data_ordered
, "data=ordered"},
630 {Opt_data_writeback
, "data=writeback"},
631 {Opt_offusrjquota
, "usrjquota="},
632 {Opt_usrjquota
, "usrjquota=%s"},
633 {Opt_offgrpjquota
, "grpjquota="},
634 {Opt_grpjquota
, "grpjquota=%s"},
635 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
636 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
637 {Opt_quota
, "grpquota"},
638 {Opt_noquota
, "noquota"},
639 {Opt_quota
, "quota"},
640 {Opt_quota
, "usrquota"},
641 {Opt_barrier
, "barrier=%u"},
643 {Opt_resize
, "resize"},
646 static unsigned long get_sb_block(void **data
)
648 unsigned long sb_block
;
649 char *options
= (char *) *data
;
651 if (!options
|| strncmp(options
, "sb=", 3) != 0)
652 return 1; /* Default location */
654 sb_block
= simple_strtoul(options
, &options
, 0);
655 if (*options
&& *options
!= ',') {
656 printk("EXT3-fs: Invalid sb specification: %s\n",
662 *data
= (void *) options
;
666 static int parse_options (char * options
, struct super_block
*sb
,
667 unsigned long * inum
, unsigned long *n_blocks_count
, int is_remount
)
669 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
671 substring_t args
[MAX_OPT_ARGS
];
682 while ((p
= strsep (&options
, ",")) != NULL
) {
687 token
= match_token(p
, tokens
, args
);
690 clear_opt (sbi
->s_mount_opt
, MINIX_DF
);
693 set_opt (sbi
->s_mount_opt
, MINIX_DF
);
696 set_opt (sbi
->s_mount_opt
, GRPID
);
699 clear_opt (sbi
->s_mount_opt
, GRPID
);
702 if (match_int(&args
[0], &option
))
704 sbi
->s_resuid
= option
;
707 if (match_int(&args
[0], &option
))
709 sbi
->s_resgid
= option
;
712 /* handled by get_sb_block() instead of here */
713 /* *sb_block = match_int(&args[0]); */
716 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
717 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
718 set_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
721 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
722 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
723 set_opt (sbi
->s_mount_opt
, ERRORS_RO
);
726 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
727 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
728 set_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
731 set_opt (sbi
->s_mount_opt
, NO_UID32
);
734 #ifdef CONFIG_EXT3_CHECK
735 set_opt (sbi
->s_mount_opt
, CHECK
);
738 "EXT3 Check option not supported\n");
742 clear_opt (sbi
->s_mount_opt
, CHECK
);
745 set_opt (sbi
->s_mount_opt
, DEBUG
);
748 set_opt (sbi
->s_mount_opt
, OLDALLOC
);
751 clear_opt (sbi
->s_mount_opt
, OLDALLOC
);
753 #ifdef CONFIG_EXT3_FS_XATTR
755 set_opt (sbi
->s_mount_opt
, XATTR_USER
);
757 case Opt_nouser_xattr
:
758 clear_opt (sbi
->s_mount_opt
, XATTR_USER
);
762 case Opt_nouser_xattr
:
763 printk("EXT3 (no)user_xattr options not supported\n");
766 #ifdef CONFIG_EXT3_FS_POSIX_ACL
768 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
771 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
776 printk("EXT3 (no)acl options not supported\n");
779 case Opt_reservation
:
780 set_opt(sbi
->s_mount_opt
, RESERVATION
);
782 case Opt_noreservation
:
783 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
785 case Opt_journal_update
:
787 /* Eventually we will want to be able to create
788 a journal file here. For now, only allow the
789 user to specify an existing inode to be the
792 printk(KERN_ERR
"EXT3-fs: cannot specify "
793 "journal on remount\n");
796 set_opt (sbi
->s_mount_opt
, UPDATE_JOURNAL
);
798 case Opt_journal_inum
:
800 printk(KERN_ERR
"EXT3-fs: cannot specify "
801 "journal on remount\n");
804 if (match_int(&args
[0], &option
))
809 set_opt (sbi
->s_mount_opt
, NOLOAD
);
812 if (match_int(&args
[0], &option
))
817 option
= JBD_DEFAULT_MAX_COMMIT_AGE
;
818 sbi
->s_commit_interval
= HZ
* option
;
820 case Opt_data_journal
:
821 data_opt
= EXT3_MOUNT_JOURNAL_DATA
;
823 case Opt_data_ordered
:
824 data_opt
= EXT3_MOUNT_ORDERED_DATA
;
826 case Opt_data_writeback
:
827 data_opt
= EXT3_MOUNT_WRITEBACK_DATA
;
830 if ((sbi
->s_mount_opt
& EXT3_MOUNT_DATA_FLAGS
)
833 "EXT3-fs: cannot change data "
834 "mode on remount\n");
838 sbi
->s_mount_opt
&= ~EXT3_MOUNT_DATA_FLAGS
;
839 sbi
->s_mount_opt
|= data_opt
;
849 if (sb_any_quota_enabled(sb
)) {
851 "EXT3-fs: Cannot change journalled "
852 "quota options when quota turned on.\n");
855 qname
= match_strdup(&args
[0]);
858 "EXT3-fs: not enough memory for "
859 "storing quotafile name.\n");
862 if (sbi
->s_qf_names
[qtype
] &&
863 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
865 "EXT3-fs: %s quota file already "
866 "specified.\n", QTYPE2NAME(qtype
));
870 sbi
->s_qf_names
[qtype
] = qname
;
871 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
873 "EXT3-fs: quotafile must be on "
874 "filesystem root.\n");
875 kfree(sbi
->s_qf_names
[qtype
]);
876 sbi
->s_qf_names
[qtype
] = NULL
;
879 set_opt(sbi
->s_mount_opt
, QUOTA
);
881 case Opt_offusrjquota
:
884 case Opt_offgrpjquota
:
887 if (sb_any_quota_enabled(sb
)) {
888 printk(KERN_ERR
"EXT3-fs: Cannot change "
889 "journalled quota options when "
890 "quota turned on.\n");
893 kfree(sbi
->s_qf_names
[qtype
]);
894 sbi
->s_qf_names
[qtype
] = NULL
;
896 case Opt_jqfmt_vfsold
:
897 sbi
->s_jquota_fmt
= QFMT_VFS_OLD
;
899 case Opt_jqfmt_vfsv0
:
900 sbi
->s_jquota_fmt
= QFMT_VFS_V0
;
903 set_opt(sbi
->s_mount_opt
, QUOTA
);
906 if (sb_any_quota_enabled(sb
)) {
907 printk(KERN_ERR
"EXT3-fs: Cannot change quota "
908 "options when quota turned on.\n");
911 clear_opt(sbi
->s_mount_opt
, QUOTA
);
916 case Opt_offusrjquota
:
917 case Opt_offgrpjquota
:
918 case Opt_jqfmt_vfsold
:
919 case Opt_jqfmt_vfsv0
:
921 "EXT3-fs: journalled quota options not "
929 set_opt(sbi
->s_mount_opt
, ABORT
);
932 if (match_int(&args
[0], &option
))
935 set_opt(sbi
->s_mount_opt
, BARRIER
);
937 clear_opt(sbi
->s_mount_opt
, BARRIER
);
942 if (!n_blocks_count
) {
943 printk("EXT3-fs: resize option only available "
947 match_int(&args
[0], &option
);
948 *n_blocks_count
= option
;
951 set_opt(sbi
->s_mount_opt
, NOBH
);
955 "EXT3-fs: Unrecognized mount option \"%s\" "
956 "or missing value\n", p
);
961 if (!sbi
->s_jquota_fmt
&& (sbi
->s_qf_names
[USRQUOTA
] ||
962 sbi
->s_qf_names
[GRPQUOTA
])) {
964 "EXT3-fs: journalled quota format not specified.\n");
972 static int ext3_setup_super(struct super_block
*sb
, struct ext3_super_block
*es
,
975 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
978 if (le32_to_cpu(es
->s_rev_level
) > EXT3_MAX_SUPP_REV
) {
979 printk (KERN_ERR
"EXT3-fs warning: revision level too high, "
980 "forcing read-only mode\n");
985 if (!(sbi
->s_mount_state
& EXT3_VALID_FS
))
986 printk (KERN_WARNING
"EXT3-fs warning: mounting unchecked fs, "
987 "running e2fsck is recommended\n");
988 else if ((sbi
->s_mount_state
& EXT3_ERROR_FS
))
990 "EXT3-fs warning: mounting fs with errors, "
991 "running e2fsck is recommended\n");
992 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
993 le16_to_cpu(es
->s_mnt_count
) >=
994 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
996 "EXT3-fs warning: maximal mount count reached, "
997 "running e2fsck is recommended\n");
998 else if (le32_to_cpu(es
->s_checkinterval
) &&
999 (le32_to_cpu(es
->s_lastcheck
) +
1000 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1001 printk (KERN_WARNING
1002 "EXT3-fs warning: checktime reached, "
1003 "running e2fsck is recommended\n");
1005 /* @@@ We _will_ want to clear the valid bit if we find
1006 inconsistencies, to force a fsck at reboot. But for
1007 a plain journaled filesystem we can keep it set as
1008 valid forever! :) */
1009 es
->s_state
= cpu_to_le16(le16_to_cpu(es
->s_state
) & ~EXT3_VALID_FS
);
1011 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1012 es
->s_max_mnt_count
= cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT
);
1013 es
->s_mnt_count
=cpu_to_le16(le16_to_cpu(es
->s_mnt_count
) + 1);
1014 es
->s_mtime
= cpu_to_le32(get_seconds());
1015 ext3_update_dynamic_rev(sb
);
1016 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
1018 ext3_commit_super(sb
, es
, 1);
1019 if (test_opt(sb
, DEBUG
))
1020 printk(KERN_INFO
"[EXT3 FS bs=%lu, gc=%lu, "
1021 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1023 sbi
->s_groups_count
,
1024 EXT3_BLOCKS_PER_GROUP(sb
),
1025 EXT3_INODES_PER_GROUP(sb
),
1028 printk(KERN_INFO
"EXT3 FS on %s, ", sb
->s_id
);
1029 if (EXT3_SB(sb
)->s_journal
->j_inode
== NULL
) {
1030 char b
[BDEVNAME_SIZE
];
1032 printk("external journal on %s\n",
1033 bdevname(EXT3_SB(sb
)->s_journal
->j_dev
, b
));
1035 printk("internal journal\n");
1037 #ifdef CONFIG_EXT3_CHECK
1038 if (test_opt (sb
, CHECK
)) {
1039 ext3_check_blocks_bitmap (sb
);
1040 ext3_check_inodes_bitmap (sb
);
1046 /* Called at mount-time, super-block is locked */
1047 static int ext3_check_descriptors (struct super_block
* sb
)
1049 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1050 unsigned long block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1051 struct ext3_group_desc
* gdp
= NULL
;
1055 ext3_debug ("Checking group descriptors");
1057 for (i
= 0; i
< sbi
->s_groups_count
; i
++)
1059 if ((i
% EXT3_DESC_PER_BLOCK(sb
)) == 0)
1060 gdp
= (struct ext3_group_desc
*)
1061 sbi
->s_group_desc
[desc_block
++]->b_data
;
1062 if (le32_to_cpu(gdp
->bg_block_bitmap
) < block
||
1063 le32_to_cpu(gdp
->bg_block_bitmap
) >=
1064 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1066 ext3_error (sb
, "ext3_check_descriptors",
1067 "Block bitmap for group %d"
1068 " not in group (block %lu)!",
1070 le32_to_cpu(gdp
->bg_block_bitmap
));
1073 if (le32_to_cpu(gdp
->bg_inode_bitmap
) < block
||
1074 le32_to_cpu(gdp
->bg_inode_bitmap
) >=
1075 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1077 ext3_error (sb
, "ext3_check_descriptors",
1078 "Inode bitmap for group %d"
1079 " not in group (block %lu)!",
1081 le32_to_cpu(gdp
->bg_inode_bitmap
));
1084 if (le32_to_cpu(gdp
->bg_inode_table
) < block
||
1085 le32_to_cpu(gdp
->bg_inode_table
) + sbi
->s_itb_per_group
>=
1086 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1088 ext3_error (sb
, "ext3_check_descriptors",
1089 "Inode table for group %d"
1090 " not in group (block %lu)!",
1092 le32_to_cpu(gdp
->bg_inode_table
));
1095 block
+= EXT3_BLOCKS_PER_GROUP(sb
);
1099 sbi
->s_es
->s_free_blocks_count
=cpu_to_le32(ext3_count_free_blocks(sb
));
1100 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext3_count_free_inodes(sb
));
1105 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1106 * the superblock) which were deleted from all directories, but held open by
1107 * a process at the time of a crash. We walk the list and try to delete these
1108 * inodes at recovery time (only with a read-write filesystem).
1110 * In order to keep the orphan inode chain consistent during traversal (in
1111 * case of crash during recovery), we link each inode into the superblock
1112 * orphan list_head and handle it the same way as an inode deletion during
1113 * normal operation (which journals the operations for us).
1115 * We only do an iget() and an iput() on each inode, which is very safe if we
1116 * accidentally point at an in-use or already deleted inode. The worst that
1117 * can happen in this case is that we get a "bit already cleared" message from
1118 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1119 * e2fsck was run on this filesystem, and it must have already done the orphan
1120 * inode cleanup for us, so we can safely abort without any further action.
1122 static void ext3_orphan_cleanup (struct super_block
* sb
,
1123 struct ext3_super_block
* es
)
1125 unsigned int s_flags
= sb
->s_flags
;
1126 int nr_orphans
= 0, nr_truncates
= 0;
1130 if (!es
->s_last_orphan
) {
1131 jbd_debug(4, "no orphan inodes to clean up\n");
1135 if (EXT3_SB(sb
)->s_mount_state
& EXT3_ERROR_FS
) {
1136 if (es
->s_last_orphan
)
1137 jbd_debug(1, "Errors on filesystem, "
1138 "clearing orphan list.\n");
1139 es
->s_last_orphan
= 0;
1140 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1144 if (s_flags
& MS_RDONLY
) {
1145 printk(KERN_INFO
"EXT3-fs: %s: orphan cleanup on readonly fs\n",
1147 sb
->s_flags
&= ~MS_RDONLY
;
1150 /* Needed for iput() to work correctly and not trash data */
1151 sb
->s_flags
|= MS_ACTIVE
;
1152 /* Turn on quotas so that they are updated correctly */
1153 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1154 if (EXT3_SB(sb
)->s_qf_names
[i
]) {
1155 int ret
= ext3_quota_on_mount(sb
, i
);
1158 "EXT3-fs: Cannot turn on journalled "
1159 "quota: error %d\n", ret
);
1164 while (es
->s_last_orphan
) {
1165 struct inode
*inode
;
1168 ext3_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
)))) {
1169 es
->s_last_orphan
= 0;
1173 list_add(&EXT3_I(inode
)->i_orphan
, &EXT3_SB(sb
)->s_orphan
);
1175 if (inode
->i_nlink
) {
1177 "%s: truncating inode %ld to %Ld bytes\n",
1178 __FUNCTION__
, inode
->i_ino
, inode
->i_size
);
1179 jbd_debug(2, "truncating inode %ld to %Ld bytes\n",
1180 inode
->i_ino
, inode
->i_size
);
1181 ext3_truncate(inode
);
1185 "%s: deleting unreferenced inode %ld\n",
1186 __FUNCTION__
, inode
->i_ino
);
1187 jbd_debug(2, "deleting unreferenced inode %ld\n",
1191 iput(inode
); /* The delete magic happens here! */
1194 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1197 printk(KERN_INFO
"EXT3-fs: %s: %d orphan inode%s deleted\n",
1198 sb
->s_id
, PLURAL(nr_orphans
));
1200 printk(KERN_INFO
"EXT3-fs: %s: %d truncate%s cleaned up\n",
1201 sb
->s_id
, PLURAL(nr_truncates
));
1203 /* Turn quotas off */
1204 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1205 if (sb_dqopt(sb
)->files
[i
])
1206 vfs_quota_off(sb
, i
);
1209 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1212 #define log2(n) ffz(~(n))
1215 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1216 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1217 * We need to be 1 filesystem block less than the 2^32 sector limit.
1219 static loff_t
ext3_max_size(int bits
)
1221 loff_t res
= EXT3_NDIR_BLOCKS
;
1222 /* This constant is calculated to be the largest file size for a
1223 * dense, 4k-blocksize file such that the total number of
1224 * sectors in the file, including data and all indirect blocks,
1225 * does not exceed 2^32. */
1226 const loff_t upper_limit
= 0x1ff7fffd000LL
;
1228 res
+= 1LL << (bits
-2);
1229 res
+= 1LL << (2*(bits
-2));
1230 res
+= 1LL << (3*(bits
-2));
1232 if (res
> upper_limit
)
1237 static unsigned long descriptor_loc(struct super_block
*sb
,
1238 unsigned long logic_sb_block
,
1241 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1242 unsigned long bg
, first_data_block
, first_meta_bg
;
1245 first_data_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1246 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1248 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_META_BG
) ||
1250 return (logic_sb_block
+ nr
+ 1);
1251 bg
= sbi
->s_desc_per_block
* nr
;
1252 if (ext3_bg_has_super(sb
, bg
))
1254 return (first_data_block
+ has_super
+ (bg
* sbi
->s_blocks_per_group
));
1258 static int ext3_fill_super (struct super_block
*sb
, void *data
, int silent
)
1260 struct buffer_head
* bh
;
1261 struct ext3_super_block
*es
= NULL
;
1262 struct ext3_sb_info
*sbi
;
1263 unsigned long block
;
1264 unsigned long sb_block
= get_sb_block(&data
);
1265 unsigned long logic_sb_block
;
1266 unsigned long offset
= 0;
1267 unsigned long journal_inum
= 0;
1268 unsigned long def_mount_opts
;
1277 sbi
= kmalloc(sizeof(*sbi
), GFP_KERNEL
);
1280 sb
->s_fs_info
= sbi
;
1281 memset(sbi
, 0, sizeof(*sbi
));
1282 sbi
->s_mount_opt
= 0;
1283 sbi
->s_resuid
= EXT3_DEF_RESUID
;
1284 sbi
->s_resgid
= EXT3_DEF_RESGID
;
1288 blocksize
= sb_min_blocksize(sb
, EXT3_MIN_BLOCK_SIZE
);
1290 printk(KERN_ERR
"EXT3-fs: unable to set blocksize\n");
1295 * The ext3 superblock will not be buffer aligned for other than 1kB
1296 * block sizes. We need to calculate the offset from buffer start.
1298 if (blocksize
!= EXT3_MIN_BLOCK_SIZE
) {
1299 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1300 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1302 logic_sb_block
= sb_block
;
1305 if (!(bh
= sb_bread(sb
, logic_sb_block
))) {
1306 printk (KERN_ERR
"EXT3-fs: unable to read superblock\n");
1310 * Note: s_es must be initialized as soon as possible because
1311 * some ext3 macro-instructions depend on its value
1313 es
= (struct ext3_super_block
*) (((char *)bh
->b_data
) + offset
);
1315 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1316 if (sb
->s_magic
!= EXT3_SUPER_MAGIC
)
1319 /* Set defaults before we parse the mount options */
1320 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1321 if (def_mount_opts
& EXT3_DEFM_DEBUG
)
1322 set_opt(sbi
->s_mount_opt
, DEBUG
);
1323 if (def_mount_opts
& EXT3_DEFM_BSDGROUPS
)
1324 set_opt(sbi
->s_mount_opt
, GRPID
);
1325 if (def_mount_opts
& EXT3_DEFM_UID16
)
1326 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1327 if (def_mount_opts
& EXT3_DEFM_XATTR_USER
)
1328 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1329 if (def_mount_opts
& EXT3_DEFM_ACL
)
1330 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1331 if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_DATA
)
1332 sbi
->s_mount_opt
|= EXT3_MOUNT_JOURNAL_DATA
;
1333 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_ORDERED
)
1334 sbi
->s_mount_opt
|= EXT3_MOUNT_ORDERED_DATA
;
1335 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_WBACK
)
1336 sbi
->s_mount_opt
|= EXT3_MOUNT_WRITEBACK_DATA
;
1338 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_PANIC
)
1339 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1340 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_RO
)
1341 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1343 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
1344 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
1346 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1348 if (!parse_options ((char *) data
, sb
, &journal_inum
, NULL
, 0))
1351 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1352 ((sbi
->s_mount_opt
& EXT3_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1354 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
&&
1355 (EXT3_HAS_COMPAT_FEATURE(sb
, ~0U) ||
1356 EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
1357 EXT3_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
1359 "EXT3-fs warning: feature flags set on rev 0 fs, "
1360 "running e2fsck is recommended\n");
1362 * Check feature flags regardless of the revision level, since we
1363 * previously didn't change the revision level when setting the flags,
1364 * so there is a chance incompat flags are set on a rev 0 filesystem.
1366 features
= EXT3_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
);
1368 printk(KERN_ERR
"EXT3-fs: %s: couldn't mount because of "
1369 "unsupported optional features (%x).\n",
1370 sb
->s_id
, le32_to_cpu(features
));
1373 features
= EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
);
1374 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1375 printk(KERN_ERR
"EXT3-fs: %s: couldn't mount RDWR because of "
1376 "unsupported optional features (%x).\n",
1377 sb
->s_id
, le32_to_cpu(features
));
1380 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
1382 if (blocksize
< EXT3_MIN_BLOCK_SIZE
||
1383 blocksize
> EXT3_MAX_BLOCK_SIZE
) {
1385 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1386 blocksize
, sb
->s_id
);
1390 hblock
= bdev_hardsect_size(sb
->s_bdev
);
1391 if (sb
->s_blocksize
!= blocksize
) {
1393 * Make sure the blocksize for the filesystem is larger
1394 * than the hardware sectorsize for the machine.
1396 if (blocksize
< hblock
) {
1397 printk(KERN_ERR
"EXT3-fs: blocksize %d too small for "
1398 "device blocksize %d.\n", blocksize
, hblock
);
1403 sb_set_blocksize(sb
, blocksize
);
1404 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1405 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1406 bh
= sb_bread(sb
, logic_sb_block
);
1409 "EXT3-fs: Can't read superblock on 2nd try.\n");
1412 es
= (struct ext3_super_block
*)(((char *)bh
->b_data
) + offset
);
1414 if (es
->s_magic
!= cpu_to_le16(EXT3_SUPER_MAGIC
)) {
1416 "EXT3-fs: Magic mismatch, very weird !\n");
1421 sb
->s_maxbytes
= ext3_max_size(sb
->s_blocksize_bits
);
1423 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
) {
1424 sbi
->s_inode_size
= EXT3_GOOD_OLD_INODE_SIZE
;
1425 sbi
->s_first_ino
= EXT3_GOOD_OLD_FIRST_INO
;
1427 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
1428 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
1429 if ((sbi
->s_inode_size
< EXT3_GOOD_OLD_INODE_SIZE
) ||
1430 (sbi
->s_inode_size
& (sbi
->s_inode_size
- 1)) ||
1431 (sbi
->s_inode_size
> blocksize
)) {
1433 "EXT3-fs: unsupported inode size: %d\n",
1438 sbi
->s_frag_size
= EXT3_MIN_FRAG_SIZE
<<
1439 le32_to_cpu(es
->s_log_frag_size
);
1440 if (blocksize
!= sbi
->s_frag_size
) {
1442 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1443 sbi
->s_frag_size
, blocksize
);
1446 sbi
->s_frags_per_block
= 1;
1447 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
1448 sbi
->s_frags_per_group
= le32_to_cpu(es
->s_frags_per_group
);
1449 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
1450 if (EXT3_INODE_SIZE(sb
) == 0)
1452 sbi
->s_inodes_per_block
= blocksize
/ EXT3_INODE_SIZE(sb
);
1453 if (sbi
->s_inodes_per_block
== 0)
1455 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
1456 sbi
->s_inodes_per_block
;
1457 sbi
->s_desc_per_block
= blocksize
/ sizeof(struct ext3_group_desc
);
1459 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
1460 sbi
->s_addr_per_block_bits
= log2(EXT3_ADDR_PER_BLOCK(sb
));
1461 sbi
->s_desc_per_block_bits
= log2(EXT3_DESC_PER_BLOCK(sb
));
1462 for (i
=0; i
< 4; i
++)
1463 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
1464 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
1466 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
1468 "EXT3-fs: #blocks per group too big: %lu\n",
1469 sbi
->s_blocks_per_group
);
1472 if (sbi
->s_frags_per_group
> blocksize
* 8) {
1474 "EXT3-fs: #fragments per group too big: %lu\n",
1475 sbi
->s_frags_per_group
);
1478 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
1480 "EXT3-fs: #inodes per group too big: %lu\n",
1481 sbi
->s_inodes_per_group
);
1485 if (EXT3_BLOCKS_PER_GROUP(sb
) == 0)
1487 sbi
->s_groups_count
= (le32_to_cpu(es
->s_blocks_count
) -
1488 le32_to_cpu(es
->s_first_data_block
) +
1489 EXT3_BLOCKS_PER_GROUP(sb
) - 1) /
1490 EXT3_BLOCKS_PER_GROUP(sb
);
1491 db_count
= (sbi
->s_groups_count
+ EXT3_DESC_PER_BLOCK(sb
) - 1) /
1492 EXT3_DESC_PER_BLOCK(sb
);
1493 sbi
->s_group_desc
= kmalloc(db_count
* sizeof (struct buffer_head
*),
1495 if (sbi
->s_group_desc
== NULL
) {
1496 printk (KERN_ERR
"EXT3-fs: not enough memory\n");
1500 percpu_counter_init(&sbi
->s_freeblocks_counter
);
1501 percpu_counter_init(&sbi
->s_freeinodes_counter
);
1502 percpu_counter_init(&sbi
->s_dirs_counter
);
1503 bgl_lock_init(&sbi
->s_blockgroup_lock
);
1505 for (i
= 0; i
< db_count
; i
++) {
1506 block
= descriptor_loc(sb
, logic_sb_block
, i
);
1507 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
1508 if (!sbi
->s_group_desc
[i
]) {
1509 printk (KERN_ERR
"EXT3-fs: "
1510 "can't read group descriptor %d\n", i
);
1515 if (!ext3_check_descriptors (sb
)) {
1516 printk (KERN_ERR
"EXT3-fs: group descriptors corrupted !\n");
1519 sbi
->s_gdb_count
= db_count
;
1520 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1521 spin_lock_init(&sbi
->s_next_gen_lock
);
1522 /* per fileystem reservation list head & lock */
1523 spin_lock_init(&sbi
->s_rsv_window_lock
);
1524 sbi
->s_rsv_window_root
= RB_ROOT
;
1525 /* Add a single, static dummy reservation to the start of the
1526 * reservation window list --- it gives us a placeholder for
1527 * append-at-start-of-list which makes the allocation logic
1528 * _much_ simpler. */
1529 sbi
->s_rsv_window_head
.rsv_start
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1530 sbi
->s_rsv_window_head
.rsv_end
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1531 sbi
->s_rsv_window_head
.rsv_alloc_hit
= 0;
1532 sbi
->s_rsv_window_head
.rsv_goal_size
= 0;
1533 ext3_rsv_window_add(sb
, &sbi
->s_rsv_window_head
);
1536 * set up enough so that it can read an inode
1538 sb
->s_op
= &ext3_sops
;
1539 sb
->s_export_op
= &ext3_export_ops
;
1540 sb
->s_xattr
= ext3_xattr_handlers
;
1542 sb
->s_qcop
= &ext3_qctl_operations
;
1543 sb
->dq_op
= &ext3_quota_operations
;
1545 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
1549 needs_recovery
= (es
->s_last_orphan
!= 0 ||
1550 EXT3_HAS_INCOMPAT_FEATURE(sb
,
1551 EXT3_FEATURE_INCOMPAT_RECOVER
));
1554 * The first inode we look at is the journal inode. Don't try
1555 * root first: it may be modified in the journal!
1557 if (!test_opt(sb
, NOLOAD
) &&
1558 EXT3_HAS_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
)) {
1559 if (ext3_load_journal(sb
, es
))
1561 } else if (journal_inum
) {
1562 if (ext3_create_journal(sb
, es
, journal_inum
))
1567 "ext3: No journal on filesystem on %s\n",
1572 /* We have now updated the journal if required, so we can
1573 * validate the data journaling mode. */
1574 switch (test_opt(sb
, DATA_FLAGS
)) {
1576 /* No mode set, assume a default based on the journal
1577 capabilities: ORDERED_DATA if the journal can
1578 cope, else JOURNAL_DATA */
1579 if (journal_check_available_features
1580 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
))
1581 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
1583 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
1586 case EXT3_MOUNT_ORDERED_DATA
:
1587 case EXT3_MOUNT_WRITEBACK_DATA
:
1588 if (!journal_check_available_features
1589 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
)) {
1590 printk(KERN_ERR
"EXT3-fs: Journal does not support "
1591 "requested data journaling mode\n");
1598 if (test_opt(sb
, NOBH
)) {
1599 if (sb
->s_blocksize_bits
!= PAGE_CACHE_SHIFT
) {
1600 printk(KERN_WARNING
"EXT3-fs: Ignoring nobh option "
1601 "since filesystem blocksize doesn't match "
1603 clear_opt(sbi
->s_mount_opt
, NOBH
);
1605 if (!(test_opt(sb
, DATA_FLAGS
) == EXT3_MOUNT_WRITEBACK_DATA
)) {
1606 printk(KERN_WARNING
"EXT3-fs: Ignoring nobh option - "
1607 "its supported only with writeback mode\n");
1608 clear_opt(sbi
->s_mount_opt
, NOBH
);
1612 * The journal_load will have done any necessary log recovery,
1613 * so we can safely mount the rest of the filesystem now.
1616 root
= iget(sb
, EXT3_ROOT_INO
);
1617 sb
->s_root
= d_alloc_root(root
);
1619 printk(KERN_ERR
"EXT3-fs: get root inode failed\n");
1623 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1626 printk(KERN_ERR
"EXT3-fs: corrupt root inode, run e2fsck\n");
1630 ext3_setup_super (sb
, es
, sb
->s_flags
& MS_RDONLY
);
1632 * akpm: core read_super() calls in here with the superblock locked.
1633 * That deadlocks, because orphan cleanup needs to lock the superblock
1634 * in numerous places. Here we just pop the lock - it's relatively
1635 * harmless, because we are now ready to accept write_super() requests,
1636 * and aviro says that's the only reason for hanging onto the
1639 EXT3_SB(sb
)->s_mount_state
|= EXT3_ORPHAN_FS
;
1640 ext3_orphan_cleanup(sb
, es
);
1641 EXT3_SB(sb
)->s_mount_state
&= ~EXT3_ORPHAN_FS
;
1643 printk (KERN_INFO
"EXT3-fs: recovery complete.\n");
1644 ext3_mark_recovery_complete(sb
, es
);
1645 printk (KERN_INFO
"EXT3-fs: mounted filesystem with %s data mode.\n",
1646 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_JOURNAL_DATA
? "journal":
1647 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_ORDERED_DATA
? "ordered":
1650 percpu_counter_mod(&sbi
->s_freeblocks_counter
,
1651 ext3_count_free_blocks(sb
));
1652 percpu_counter_mod(&sbi
->s_freeinodes_counter
,
1653 ext3_count_free_inodes(sb
));
1654 percpu_counter_mod(&sbi
->s_dirs_counter
,
1655 ext3_count_dirs(sb
));
1662 printk(KERN_ERR
"VFS: Can't find ext3 filesystem on dev %s.\n",
1667 journal_destroy(sbi
->s_journal
);
1669 for (i
= 0; i
< db_count
; i
++)
1670 brelse(sbi
->s_group_desc
[i
]);
1671 kfree(sbi
->s_group_desc
);
1674 for (i
= 0; i
< MAXQUOTAS
; i
++)
1675 kfree(sbi
->s_qf_names
[i
]);
1677 ext3_blkdev_remove(sbi
);
1680 sb
->s_fs_info
= NULL
;
1687 * Setup any per-fs journal parameters now. We'll do this both on
1688 * initial mount, once the journal has been initialised but before we've
1689 * done any recovery; and again on any subsequent remount.
1691 static void ext3_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
1693 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1695 if (sbi
->s_commit_interval
)
1696 journal
->j_commit_interval
= sbi
->s_commit_interval
;
1697 /* We could also set up an ext3-specific default for the commit
1698 * interval here, but for now we'll just fall back to the jbd
1701 spin_lock(&journal
->j_state_lock
);
1702 if (test_opt(sb
, BARRIER
))
1703 journal
->j_flags
|= JFS_BARRIER
;
1705 journal
->j_flags
&= ~JFS_BARRIER
;
1706 spin_unlock(&journal
->j_state_lock
);
1709 static journal_t
*ext3_get_journal(struct super_block
*sb
, int journal_inum
)
1711 struct inode
*journal_inode
;
1714 /* First, test for the existence of a valid inode on disk. Bad
1715 * things happen if we iget() an unused inode, as the subsequent
1716 * iput() will try to delete it. */
1718 journal_inode
= iget(sb
, journal_inum
);
1719 if (!journal_inode
) {
1720 printk(KERN_ERR
"EXT3-fs: no journal found.\n");
1723 if (!journal_inode
->i_nlink
) {
1724 make_bad_inode(journal_inode
);
1725 iput(journal_inode
);
1726 printk(KERN_ERR
"EXT3-fs: journal inode is deleted.\n");
1730 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1731 journal_inode
, journal_inode
->i_size
);
1732 if (is_bad_inode(journal_inode
) || !S_ISREG(journal_inode
->i_mode
)) {
1733 printk(KERN_ERR
"EXT3-fs: invalid journal inode.\n");
1734 iput(journal_inode
);
1738 journal
= journal_init_inode(journal_inode
);
1740 printk(KERN_ERR
"EXT3-fs: Could not load journal inode\n");
1741 iput(journal_inode
);
1744 journal
->j_private
= sb
;
1745 ext3_init_journal_params(sb
, journal
);
1749 static journal_t
*ext3_get_dev_journal(struct super_block
*sb
,
1752 struct buffer_head
* bh
;
1756 int hblock
, blocksize
;
1757 unsigned long sb_block
;
1758 unsigned long offset
;
1759 struct ext3_super_block
* es
;
1760 struct block_device
*bdev
;
1762 bdev
= ext3_blkdev_get(j_dev
);
1766 if (bd_claim(bdev
, sb
)) {
1768 "EXT3: failed to claim external journal device.\n");
1773 blocksize
= sb
->s_blocksize
;
1774 hblock
= bdev_hardsect_size(bdev
);
1775 if (blocksize
< hblock
) {
1777 "EXT3-fs: blocksize too small for journal device.\n");
1781 sb_block
= EXT3_MIN_BLOCK_SIZE
/ blocksize
;
1782 offset
= EXT3_MIN_BLOCK_SIZE
% blocksize
;
1783 set_blocksize(bdev
, blocksize
);
1784 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
1785 printk(KERN_ERR
"EXT3-fs: couldn't read superblock of "
1786 "external journal\n");
1790 es
= (struct ext3_super_block
*) (((char *)bh
->b_data
) + offset
);
1791 if ((le16_to_cpu(es
->s_magic
) != EXT3_SUPER_MAGIC
) ||
1792 !(le32_to_cpu(es
->s_feature_incompat
) &
1793 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
1794 printk(KERN_ERR
"EXT3-fs: external journal has "
1795 "bad superblock\n");
1800 if (memcmp(EXT3_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
1801 printk(KERN_ERR
"EXT3-fs: journal UUID does not match\n");
1806 len
= le32_to_cpu(es
->s_blocks_count
);
1807 start
= sb_block
+ 1;
1808 brelse(bh
); /* we're done with the superblock */
1810 journal
= journal_init_dev(bdev
, sb
->s_bdev
,
1811 start
, len
, blocksize
);
1813 printk(KERN_ERR
"EXT3-fs: failed to create device journal\n");
1816 journal
->j_private
= sb
;
1817 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
1818 wait_on_buffer(journal
->j_sb_buffer
);
1819 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
1820 printk(KERN_ERR
"EXT3-fs: I/O error on journal device\n");
1823 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
1824 printk(KERN_ERR
"EXT3-fs: External journal has more than one "
1825 "user (unsupported) - %d\n",
1826 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
1829 EXT3_SB(sb
)->journal_bdev
= bdev
;
1830 ext3_init_journal_params(sb
, journal
);
1833 journal_destroy(journal
);
1835 ext3_blkdev_put(bdev
);
1839 static int ext3_load_journal(struct super_block
* sb
,
1840 struct ext3_super_block
* es
)
1843 int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
1844 dev_t journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
1846 int really_read_only
;
1848 really_read_only
= bdev_read_only(sb
->s_bdev
);
1851 * Are we loading a blank journal or performing recovery after a
1852 * crash? For recovery, we need to check in advance whether we
1853 * can get read-write access to the device.
1856 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
)) {
1857 if (sb
->s_flags
& MS_RDONLY
) {
1858 printk(KERN_INFO
"EXT3-fs: INFO: recovery "
1859 "required on readonly filesystem.\n");
1860 if (really_read_only
) {
1861 printk(KERN_ERR
"EXT3-fs: write access "
1862 "unavailable, cannot proceed.\n");
1865 printk (KERN_INFO
"EXT3-fs: write access will "
1866 "be enabled during recovery.\n");
1870 if (journal_inum
&& journal_dev
) {
1871 printk(KERN_ERR
"EXT3-fs: filesystem has both journal "
1872 "and inode journals!\n");
1877 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
1880 if (!(journal
= ext3_get_dev_journal(sb
, journal_dev
)))
1884 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
1885 err
= journal_update_format(journal
);
1887 printk(KERN_ERR
"EXT3-fs: error updating journal.\n");
1888 journal_destroy(journal
);
1893 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
))
1894 err
= journal_wipe(journal
, !really_read_only
);
1896 err
= journal_load(journal
);
1899 printk(KERN_ERR
"EXT3-fs: error loading journal.\n");
1900 journal_destroy(journal
);
1904 EXT3_SB(sb
)->s_journal
= journal
;
1905 ext3_clear_journal_err(sb
, es
);
1909 static int ext3_create_journal(struct super_block
* sb
,
1910 struct ext3_super_block
* es
,
1915 if (sb
->s_flags
& MS_RDONLY
) {
1916 printk(KERN_ERR
"EXT3-fs: readonly filesystem when trying to "
1917 "create journal.\n");
1921 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
1924 printk(KERN_INFO
"EXT3-fs: creating new journal on inode %d\n",
1927 if (journal_create(journal
)) {
1928 printk(KERN_ERR
"EXT3-fs: error creating journal.\n");
1929 journal_destroy(journal
);
1933 EXT3_SB(sb
)->s_journal
= journal
;
1935 ext3_update_dynamic_rev(sb
);
1936 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
1937 EXT3_SET_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
);
1939 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
1942 /* Make sure we flush the recovery flag to disk. */
1943 ext3_commit_super(sb
, es
, 1);
1948 static void ext3_commit_super (struct super_block
* sb
,
1949 struct ext3_super_block
* es
,
1952 struct buffer_head
*sbh
= EXT3_SB(sb
)->s_sbh
;
1956 es
->s_wtime
= cpu_to_le32(get_seconds());
1957 es
->s_free_blocks_count
= cpu_to_le32(ext3_count_free_blocks(sb
));
1958 es
->s_free_inodes_count
= cpu_to_le32(ext3_count_free_inodes(sb
));
1959 BUFFER_TRACE(sbh
, "marking dirty");
1960 mark_buffer_dirty(sbh
);
1962 sync_dirty_buffer(sbh
);
1967 * Have we just finished recovery? If so, and if we are mounting (or
1968 * remounting) the filesystem readonly, then we will end up with a
1969 * consistent fs on disk. Record that fact.
1971 static void ext3_mark_recovery_complete(struct super_block
* sb
,
1972 struct ext3_super_block
* es
)
1974 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
1976 journal_lock_updates(journal
);
1977 journal_flush(journal
);
1978 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
) &&
1979 sb
->s_flags
& MS_RDONLY
) {
1980 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
1982 ext3_commit_super(sb
, es
, 1);
1984 journal_unlock_updates(journal
);
1988 * If we are mounting (or read-write remounting) a filesystem whose journal
1989 * has recorded an error from a previous lifetime, move that error to the
1990 * main filesystem now.
1992 static void ext3_clear_journal_err(struct super_block
* sb
,
1993 struct ext3_super_block
* es
)
1999 journal
= EXT3_SB(sb
)->s_journal
;
2002 * Now check for any error status which may have been recorded in the
2003 * journal by a prior ext3_error() or ext3_abort()
2006 j_errno
= journal_errno(journal
);
2010 errstr
= ext3_decode_error(sb
, j_errno
, nbuf
);
2011 ext3_warning(sb
, __FUNCTION__
, "Filesystem error recorded "
2012 "from previous mount: %s", errstr
);
2013 ext3_warning(sb
, __FUNCTION__
, "Marking fs in need of "
2014 "filesystem check.");
2016 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
2017 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
2018 ext3_commit_super (sb
, es
, 1);
2020 journal_clear_err(journal
);
2025 * Force the running and committing transactions to commit,
2026 * and wait on the commit.
2028 int ext3_force_commit(struct super_block
*sb
)
2033 if (sb
->s_flags
& MS_RDONLY
)
2036 journal
= EXT3_SB(sb
)->s_journal
;
2038 ret
= ext3_journal_force_commit(journal
);
2043 * Ext3 always journals updates to the superblock itself, so we don't
2044 * have to propagate any other updates to the superblock on disk at this
2045 * point. Just start an async writeback to get the buffers on their way
2048 * This implicitly triggers the writebehind on sync().
2051 static void ext3_write_super (struct super_block
* sb
)
2053 if (down_trylock(&sb
->s_lock
) == 0)
2058 static int ext3_sync_fs(struct super_block
*sb
, int wait
)
2063 if (journal_start_commit(EXT3_SB(sb
)->s_journal
, &target
)) {
2065 log_wait_commit(EXT3_SB(sb
)->s_journal
, target
);
2071 * LVM calls this function before a (read-only) snapshot is created. This
2072 * gives us a chance to flush the journal completely and mark the fs clean.
2074 static void ext3_write_super_lockfs(struct super_block
*sb
)
2078 if (!(sb
->s_flags
& MS_RDONLY
)) {
2079 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
2081 /* Now we set up the journal barrier. */
2082 journal_lock_updates(journal
);
2083 journal_flush(journal
);
2085 /* Journal blocked and flushed, clear needs_recovery flag. */
2086 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2087 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2092 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2093 * flag here, even though the filesystem is not technically dirty yet.
2095 static void ext3_unlockfs(struct super_block
*sb
)
2097 if (!(sb
->s_flags
& MS_RDONLY
)) {
2099 /* Reser the needs_recovery flag before the fs is unlocked. */
2100 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2101 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2103 journal_unlock_updates(EXT3_SB(sb
)->s_journal
);
2107 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
)
2109 struct ext3_super_block
* es
;
2110 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2112 unsigned long n_blocks_count
= 0;
2115 * Allow the "check" option to be passed as a remount option.
2117 if (!parse_options(data
, sb
, &tmp
, &n_blocks_count
, 1))
2120 if (sbi
->s_mount_opt
& EXT3_MOUNT_ABORT
)
2121 ext3_abort(sb
, __FUNCTION__
, "Abort forced by user");
2123 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2124 ((sbi
->s_mount_opt
& EXT3_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2128 ext3_init_journal_params(sb
, sbi
->s_journal
);
2130 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
2131 n_blocks_count
> le32_to_cpu(es
->s_blocks_count
)) {
2132 if (sbi
->s_mount_opt
& EXT3_MOUNT_ABORT
)
2135 if (*flags
& MS_RDONLY
) {
2137 * First of all, the unconditional stuff we have to do
2138 * to disable replay of the journal when we next remount
2140 sb
->s_flags
|= MS_RDONLY
;
2143 * OK, test if we are remounting a valid rw partition
2144 * readonly, and if so set the rdonly flag and then
2145 * mark the partition as valid again.
2147 if (!(es
->s_state
& cpu_to_le16(EXT3_VALID_FS
)) &&
2148 (sbi
->s_mount_state
& EXT3_VALID_FS
))
2149 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
2151 ext3_mark_recovery_complete(sb
, es
);
2154 if ((ret
= EXT3_HAS_RO_COMPAT_FEATURE(sb
,
2155 ~EXT3_FEATURE_RO_COMPAT_SUPP
))) {
2156 printk(KERN_WARNING
"EXT3-fs: %s: couldn't "
2157 "remount RDWR because of unsupported "
2158 "optional features (%x).\n",
2159 sb
->s_id
, le32_to_cpu(ret
));
2163 * Mounting a RDONLY partition read-write, so reread
2164 * and store the current valid flag. (It may have
2165 * been changed by e2fsck since we originally mounted
2168 ext3_clear_journal_err(sb
, es
);
2169 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2170 if ((ret
= ext3_group_extend(sb
, es
, n_blocks_count
)))
2172 if (!ext3_setup_super (sb
, es
, 0))
2173 sb
->s_flags
&= ~MS_RDONLY
;
2179 static int ext3_statfs (struct super_block
* sb
, struct kstatfs
* buf
)
2181 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
2182 unsigned long overhead
;
2185 if (test_opt (sb
, MINIX_DF
))
2188 unsigned long ngroups
;
2189 ngroups
= EXT3_SB(sb
)->s_groups_count
;
2193 * Compute the overhead (FS structures)
2197 * All of the blocks before first_data_block are
2200 overhead
= le32_to_cpu(es
->s_first_data_block
);
2203 * Add the overhead attributed to the superblock and
2204 * block group descriptors. If the sparse superblocks
2205 * feature is turned on, then not all groups have this.
2207 for (i
= 0; i
< ngroups
; i
++) {
2208 overhead
+= ext3_bg_has_super(sb
, i
) +
2209 ext3_bg_num_gdb(sb
, i
);
2214 * Every block group has an inode bitmap, a block
2215 * bitmap, and an inode table.
2217 overhead
+= (ngroups
* (2 + EXT3_SB(sb
)->s_itb_per_group
));
2220 buf
->f_type
= EXT3_SUPER_MAGIC
;
2221 buf
->f_bsize
= sb
->s_blocksize
;
2222 buf
->f_blocks
= le32_to_cpu(es
->s_blocks_count
) - overhead
;
2223 buf
->f_bfree
= ext3_count_free_blocks (sb
);
2224 buf
->f_bavail
= buf
->f_bfree
- le32_to_cpu(es
->s_r_blocks_count
);
2225 if (buf
->f_bfree
< le32_to_cpu(es
->s_r_blocks_count
))
2227 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
2228 buf
->f_ffree
= ext3_count_free_inodes (sb
);
2229 buf
->f_namelen
= EXT3_NAME_LEN
;
2233 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2234 * is locked for write. Otherwise the are possible deadlocks:
2235 * Process 1 Process 2
2236 * ext3_create() quota_sync()
2237 * journal_start() write_dquot()
2238 * DQUOT_INIT() down(dqio_sem)
2239 * down(dqio_sem) journal_start()
2245 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
2247 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
2250 static int ext3_dquot_initialize(struct inode
*inode
, int type
)
2255 /* We may create quota structure so we need to reserve enough blocks */
2256 handle
= ext3_journal_start(inode
, 2*EXT3_QUOTA_INIT_BLOCKS(inode
->i_sb
));
2258 return PTR_ERR(handle
);
2259 ret
= dquot_initialize(inode
, type
);
2260 err
= ext3_journal_stop(handle
);
2266 static int ext3_dquot_drop(struct inode
*inode
)
2271 /* We may delete quota structure so we need to reserve enough blocks */
2272 handle
= ext3_journal_start(inode
, 2*EXT3_QUOTA_DEL_BLOCKS(inode
->i_sb
));
2274 return PTR_ERR(handle
);
2275 ret
= dquot_drop(inode
);
2276 err
= ext3_journal_stop(handle
);
2282 static int ext3_write_dquot(struct dquot
*dquot
)
2286 struct inode
*inode
;
2288 inode
= dquot_to_inode(dquot
);
2289 handle
= ext3_journal_start(inode
,
2290 EXT3_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
2292 return PTR_ERR(handle
);
2293 ret
= dquot_commit(dquot
);
2294 err
= ext3_journal_stop(handle
);
2300 static int ext3_acquire_dquot(struct dquot
*dquot
)
2305 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2306 EXT3_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
2308 return PTR_ERR(handle
);
2309 ret
= dquot_acquire(dquot
);
2310 err
= ext3_journal_stop(handle
);
2316 static int ext3_release_dquot(struct dquot
*dquot
)
2321 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2322 EXT3_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
2324 return PTR_ERR(handle
);
2325 ret
= dquot_release(dquot
);
2326 err
= ext3_journal_stop(handle
);
2332 static int ext3_mark_dquot_dirty(struct dquot
*dquot
)
2334 /* Are we journalling quotas? */
2335 if (EXT3_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
2336 EXT3_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
2337 dquot_mark_dquot_dirty(dquot
);
2338 return ext3_write_dquot(dquot
);
2340 return dquot_mark_dquot_dirty(dquot
);
2344 static int ext3_write_info(struct super_block
*sb
, int type
)
2349 /* Data block + inode block */
2350 handle
= ext3_journal_start(sb
->s_root
->d_inode
, 2);
2352 return PTR_ERR(handle
);
2353 ret
= dquot_commit_info(sb
, type
);
2354 err
= ext3_journal_stop(handle
);
2361 * Turn on quotas during mount time - we need to find
2362 * the quota file and such...
2364 static int ext3_quota_on_mount(struct super_block
*sb
, int type
)
2366 return vfs_quota_on_mount(sb
, EXT3_SB(sb
)->s_qf_names
[type
],
2367 EXT3_SB(sb
)->s_jquota_fmt
, type
);
2371 * Standard function to be called on quota_on
2373 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
,
2377 struct nameidata nd
;
2379 if (!test_opt(sb
, QUOTA
))
2381 /* Not journalling quota? */
2382 if (!EXT3_SB(sb
)->s_qf_names
[USRQUOTA
] &&
2383 !EXT3_SB(sb
)->s_qf_names
[GRPQUOTA
])
2384 return vfs_quota_on(sb
, type
, format_id
, path
);
2385 err
= path_lookup(path
, LOOKUP_FOLLOW
, &nd
);
2388 /* Quotafile not on the same filesystem? */
2389 if (nd
.mnt
->mnt_sb
!= sb
) {
2393 /* Quotafile not of fs root? */
2394 if (nd
.dentry
->d_parent
->d_inode
!= sb
->s_root
->d_inode
)
2396 "EXT3-fs: Quota file not on filesystem root. "
2397 "Journalled quota will not work.\n");
2399 return vfs_quota_on(sb
, type
, format_id
, path
);
2402 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2403 * acquiring the locks... As quota files are never truncated and quota code
2404 * itself serializes the operations (and noone else should touch the files)
2405 * we don't have to be afraid of races */
2406 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
2407 size_t len
, loff_t off
)
2409 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2410 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2412 int offset
= off
& (sb
->s_blocksize
- 1);
2415 struct buffer_head
*bh
;
2416 loff_t i_size
= i_size_read(inode
);
2420 if (off
+len
> i_size
)
2423 while (toread
> 0) {
2424 tocopy
= sb
->s_blocksize
- offset
< toread
?
2425 sb
->s_blocksize
- offset
: toread
;
2426 bh
= ext3_bread(NULL
, inode
, blk
, 0, &err
);
2429 if (!bh
) /* A hole? */
2430 memset(data
, 0, tocopy
);
2432 memcpy(data
, bh
->b_data
+offset
, tocopy
);
2442 /* Write to quotafile (we know the transaction is already started and has
2443 * enough credits) */
2444 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
2445 const char *data
, size_t len
, loff_t off
)
2447 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2448 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2450 int offset
= off
& (sb
->s_blocksize
- 1);
2452 int journal_quota
= EXT3_SB(sb
)->s_qf_names
[type
] != NULL
;
2453 size_t towrite
= len
;
2454 struct buffer_head
*bh
;
2455 handle_t
*handle
= journal_current_handle();
2457 down(&inode
->i_sem
);
2458 while (towrite
> 0) {
2459 tocopy
= sb
->s_blocksize
- offset
< towrite
?
2460 sb
->s_blocksize
- offset
: towrite
;
2461 bh
= ext3_bread(handle
, inode
, blk
, 1, &err
);
2464 if (journal_quota
) {
2465 err
= ext3_journal_get_write_access(handle
, bh
);
2472 memcpy(bh
->b_data
+offset
, data
, tocopy
);
2473 flush_dcache_page(bh
->b_page
);
2476 err
= ext3_journal_dirty_metadata(handle
, bh
);
2478 /* Always do at least ordered writes for quotas */
2479 err
= ext3_journal_dirty_data(handle
, bh
);
2480 mark_buffer_dirty(bh
);
2493 if (inode
->i_size
< off
+len
-towrite
) {
2494 i_size_write(inode
, off
+len
-towrite
);
2495 EXT3_I(inode
)->i_disksize
= inode
->i_size
;
2498 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2499 ext3_mark_inode_dirty(handle
, inode
);
2501 return len
- towrite
;
2506 static struct super_block
*ext3_get_sb(struct file_system_type
*fs_type
,
2507 int flags
, const char *dev_name
, void *data
)
2509 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext3_fill_super
);
2512 static struct file_system_type ext3_fs_type
= {
2513 .owner
= THIS_MODULE
,
2515 .get_sb
= ext3_get_sb
,
2516 .kill_sb
= kill_block_super
,
2517 .fs_flags
= FS_REQUIRES_DEV
,
2520 static int __init
init_ext3_fs(void)
2522 int err
= init_ext3_xattr();
2525 err
= init_inodecache();
2528 err
= register_filesystem(&ext3_fs_type
);
2533 destroy_inodecache();
2539 static void __exit
exit_ext3_fs(void)
2541 unregister_filesystem(&ext3_fs_type
);
2542 destroy_inodecache();
2546 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2547 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2548 MODULE_LICENSE("GPL");
2549 module_init(init_ext3_fs
)
2550 module_exit(exit_ext3_fs
)