2 * linux/fs/ext3/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
39 #include <asm/uaccess.h>
45 static int ext3_load_journal(struct super_block
*, struct ext3_super_block
*,
46 unsigned long journal_devnum
);
47 static int ext3_create_journal(struct super_block
*, struct ext3_super_block
*,
49 static void ext3_commit_super (struct super_block
* sb
,
50 struct ext3_super_block
* es
,
52 static void ext3_mark_recovery_complete(struct super_block
* sb
,
53 struct ext3_super_block
* es
);
54 static void ext3_clear_journal_err(struct super_block
* sb
,
55 struct ext3_super_block
* es
);
56 static int ext3_sync_fs(struct super_block
*sb
, int wait
);
57 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
59 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
);
60 static int ext3_statfs (struct dentry
* dentry
, struct kstatfs
* buf
);
61 static void ext3_unlockfs(struct super_block
*sb
);
62 static void ext3_write_super (struct super_block
* sb
);
63 static void ext3_write_super_lockfs(struct super_block
*sb
);
66 * Wrappers for journal_start/end.
68 * The only special thing we need to do here is to make sure that all
69 * journal_end calls result in the superblock being marked dirty, so
70 * that sync() will call the filesystem's write_super callback if
73 handle_t
*ext3_journal_start_sb(struct super_block
*sb
, int nblocks
)
77 if (sb
->s_flags
& MS_RDONLY
)
78 return ERR_PTR(-EROFS
);
80 /* Special case here: if the journal has aborted behind our
81 * backs (eg. EIO in the commit thread), then we still need to
82 * take the FS itself readonly cleanly. */
83 journal
= EXT3_SB(sb
)->s_journal
;
84 if (is_journal_aborted(journal
)) {
85 ext3_abort(sb
, __FUNCTION__
,
86 "Detected aborted journal");
87 return ERR_PTR(-EROFS
);
90 return journal_start(journal
, nblocks
);
94 * The only special thing we need to do here is to make sure that all
95 * journal_stop calls result in the superblock being marked dirty, so
96 * that sync() will call the filesystem's write_super callback if
99 int __ext3_journal_stop(const char *where
, handle_t
*handle
)
101 struct super_block
*sb
;
105 sb
= handle
->h_transaction
->t_journal
->j_private
;
107 rc
= journal_stop(handle
);
112 __ext3_std_error(sb
, where
, err
);
116 void ext3_journal_abort_handle(const char *caller
, const char *err_fn
,
117 struct buffer_head
*bh
, handle_t
*handle
, int err
)
120 const char *errstr
= ext3_decode_error(NULL
, err
, nbuf
);
123 BUFFER_TRACE(bh
, "abort");
128 if (is_handle_aborted(handle
))
131 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
132 caller
, errstr
, err_fn
);
134 journal_abort_handle(handle
);
137 /* Deal with the reporting of failure conditions on a filesystem such as
138 * inconsistencies detected or read IO failures.
140 * On ext2, we can store the error state of the filesystem in the
141 * superblock. That is not possible on ext3, because we may have other
142 * write ordering constraints on the superblock which prevent us from
143 * writing it out straight away; and given that the journal is about to
144 * be aborted, we can't rely on the current, or future, transactions to
145 * write out the superblock safely.
147 * We'll just use the journal_abort() error code to record an error in
148 * the journal instead. On recovery, the journal will compain about
149 * that error until we've noted it down and cleared it.
152 static void ext3_handle_error(struct super_block
*sb
)
154 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
156 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
157 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
159 if (sb
->s_flags
& MS_RDONLY
)
162 if (test_opt (sb
, ERRORS_RO
)) {
163 printk (KERN_CRIT
"Remounting filesystem read-only\n");
164 sb
->s_flags
|= MS_RDONLY
;
166 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
168 EXT3_SB(sb
)->s_mount_opt
|= EXT3_MOUNT_ABORT
;
170 journal_abort(journal
, -EIO
);
172 if (test_opt(sb
, ERRORS_PANIC
))
173 panic("EXT3-fs (device %s): panic forced after error\n",
175 ext3_commit_super(sb
, es
, 1);
178 void ext3_error (struct super_block
* sb
, const char * function
,
179 const char * fmt
, ...)
184 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
189 ext3_handle_error(sb
);
192 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
199 errstr
= "IO failure";
202 errstr
= "Out of memory";
205 if (!sb
|| EXT3_SB(sb
)->s_journal
->j_flags
& JFS_ABORT
)
206 errstr
= "Journal has aborted";
208 errstr
= "Readonly filesystem";
211 /* If the caller passed in an extra buffer for unknown
212 * errors, textualise them now. Else we just return
215 /* Check for truncated error codes... */
216 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
225 /* __ext3_std_error decodes expected errors from journaling functions
226 * automatically and invokes the appropriate error response. */
228 void __ext3_std_error (struct super_block
* sb
, const char * function
,
234 /* Special case: if the error is EROFS, and we're not already
235 * inside a transaction, then there's really no point in logging
237 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
238 (sb
->s_flags
& MS_RDONLY
))
241 errstr
= ext3_decode_error(sb
, errno
, nbuf
);
242 printk (KERN_CRIT
"EXT3-fs error (device %s) in %s: %s\n",
243 sb
->s_id
, function
, errstr
);
245 ext3_handle_error(sb
);
249 * ext3_abort is a much stronger failure handler than ext3_error. The
250 * abort function may be used to deal with unrecoverable failures such
251 * as journal IO errors or ENOMEM at a critical moment in log management.
253 * We unconditionally force the filesystem into an ABORT|READONLY state,
254 * unless the error response on the fs has been set to panic in which
255 * case we take the easy way out and panic immediately.
258 void ext3_abort (struct super_block
* sb
, const char * function
,
259 const char * fmt
, ...)
263 printk (KERN_CRIT
"ext3_abort called.\n");
266 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
271 if (test_opt(sb
, ERRORS_PANIC
))
272 panic("EXT3-fs panic from previous error\n");
274 if (sb
->s_flags
& MS_RDONLY
)
277 printk(KERN_CRIT
"Remounting filesystem read-only\n");
278 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
279 sb
->s_flags
|= MS_RDONLY
;
280 EXT3_SB(sb
)->s_mount_opt
|= EXT3_MOUNT_ABORT
;
281 journal_abort(EXT3_SB(sb
)->s_journal
, -EIO
);
284 void ext3_warning (struct super_block
* sb
, const char * function
,
285 const char * fmt
, ...)
290 printk(KERN_WARNING
"EXT3-fs warning (device %s): %s: ",
297 void ext3_update_dynamic_rev(struct super_block
*sb
)
299 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
301 if (le32_to_cpu(es
->s_rev_level
) > EXT3_GOOD_OLD_REV
)
304 ext3_warning(sb
, __FUNCTION__
,
305 "updating to rev %d because of new feature flag, "
306 "running e2fsck is recommended",
309 es
->s_first_ino
= cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO
);
310 es
->s_inode_size
= cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE
);
311 es
->s_rev_level
= cpu_to_le32(EXT3_DYNAMIC_REV
);
312 /* leave es->s_feature_*compat flags alone */
313 /* es->s_uuid will be set by e2fsck if empty */
316 * The rest of the superblock fields should be zero, and if not it
317 * means they are likely already in use, so leave them alone. We
318 * can leave it up to e2fsck to clean up any inconsistencies there.
323 * Open the external journal device
325 static struct block_device
*ext3_blkdev_get(dev_t dev
)
327 struct block_device
*bdev
;
328 char b
[BDEVNAME_SIZE
];
330 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
336 printk(KERN_ERR
"EXT3: failed to open journal device %s: %ld\n",
337 __bdevname(dev
, b
), PTR_ERR(bdev
));
342 * Release the journal device
344 static int ext3_blkdev_put(struct block_device
*bdev
)
347 return blkdev_put(bdev
);
350 static int ext3_blkdev_remove(struct ext3_sb_info
*sbi
)
352 struct block_device
*bdev
;
355 bdev
= sbi
->journal_bdev
;
357 ret
= ext3_blkdev_put(bdev
);
358 sbi
->journal_bdev
= NULL
;
363 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
365 return &list_entry(l
, struct ext3_inode_info
, i_orphan
)->vfs_inode
;
368 static void dump_orphan_list(struct super_block
*sb
, struct ext3_sb_info
*sbi
)
372 printk(KERN_ERR
"sb orphan head is %d\n",
373 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
375 printk(KERN_ERR
"sb_info orphan list:\n");
376 list_for_each(l
, &sbi
->s_orphan
) {
377 struct inode
*inode
= orphan_list_entry(l
);
379 "inode %s:%ld at %p: mode %o, nlink %d, next %d\n",
380 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
381 inode
->i_mode
, inode
->i_nlink
,
386 static void ext3_put_super (struct super_block
* sb
)
388 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
389 struct ext3_super_block
*es
= sbi
->s_es
;
392 ext3_xattr_put_super(sb
);
393 journal_destroy(sbi
->s_journal
);
394 if (!(sb
->s_flags
& MS_RDONLY
)) {
395 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
396 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
397 BUFFER_TRACE(sbi
->s_sbh
, "marking dirty");
398 mark_buffer_dirty(sbi
->s_sbh
);
399 ext3_commit_super(sb
, es
, 1);
402 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
403 brelse(sbi
->s_group_desc
[i
]);
404 kfree(sbi
->s_group_desc
);
405 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
406 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
407 percpu_counter_destroy(&sbi
->s_dirs_counter
);
410 for (i
= 0; i
< MAXQUOTAS
; i
++)
411 kfree(sbi
->s_qf_names
[i
]);
414 /* Debugging code just in case the in-memory inode orphan list
415 * isn't empty. The on-disk one can be non-empty if we've
416 * detected an error and taken the fs readonly, but the
417 * in-memory list had better be clean by this point. */
418 if (!list_empty(&sbi
->s_orphan
))
419 dump_orphan_list(sb
, sbi
);
420 J_ASSERT(list_empty(&sbi
->s_orphan
));
422 invalidate_bdev(sb
->s_bdev
, 0);
423 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
425 * Invalidate the journal device's buffers. We don't want them
426 * floating about in memory - the physical journal device may
427 * hotswapped, and it breaks the `ro-after' testing code.
429 sync_blockdev(sbi
->journal_bdev
);
430 invalidate_bdev(sbi
->journal_bdev
, 0);
431 ext3_blkdev_remove(sbi
);
433 sb
->s_fs_info
= NULL
;
438 static kmem_cache_t
*ext3_inode_cachep
;
441 * Called inside transaction, so use GFP_NOFS
443 static struct inode
*ext3_alloc_inode(struct super_block
*sb
)
445 struct ext3_inode_info
*ei
;
447 ei
= kmem_cache_alloc(ext3_inode_cachep
, SLAB_NOFS
);
450 #ifdef CONFIG_EXT3_FS_POSIX_ACL
451 ei
->i_acl
= EXT3_ACL_NOT_CACHED
;
452 ei
->i_default_acl
= EXT3_ACL_NOT_CACHED
;
454 ei
->i_block_alloc_info
= NULL
;
455 ei
->vfs_inode
.i_version
= 1;
456 return &ei
->vfs_inode
;
459 static void ext3_destroy_inode(struct inode
*inode
)
461 kmem_cache_free(ext3_inode_cachep
, EXT3_I(inode
));
464 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
466 struct ext3_inode_info
*ei
= (struct ext3_inode_info
*) foo
;
468 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
469 SLAB_CTOR_CONSTRUCTOR
) {
470 INIT_LIST_HEAD(&ei
->i_orphan
);
471 #ifdef CONFIG_EXT3_FS_XATTR
472 init_rwsem(&ei
->xattr_sem
);
474 mutex_init(&ei
->truncate_mutex
);
475 inode_init_once(&ei
->vfs_inode
);
479 static int init_inodecache(void)
481 ext3_inode_cachep
= kmem_cache_create("ext3_inode_cache",
482 sizeof(struct ext3_inode_info
),
483 0, (SLAB_RECLAIM_ACCOUNT
|
486 if (ext3_inode_cachep
== NULL
)
491 static void destroy_inodecache(void)
493 if (kmem_cache_destroy(ext3_inode_cachep
))
494 printk(KERN_INFO
"ext3_inode_cache: not all structures were freed\n");
497 static void ext3_clear_inode(struct inode
*inode
)
499 struct ext3_block_alloc_info
*rsv
= EXT3_I(inode
)->i_block_alloc_info
;
500 #ifdef CONFIG_EXT3_FS_POSIX_ACL
501 if (EXT3_I(inode
)->i_acl
&&
502 EXT3_I(inode
)->i_acl
!= EXT3_ACL_NOT_CACHED
) {
503 posix_acl_release(EXT3_I(inode
)->i_acl
);
504 EXT3_I(inode
)->i_acl
= EXT3_ACL_NOT_CACHED
;
506 if (EXT3_I(inode
)->i_default_acl
&&
507 EXT3_I(inode
)->i_default_acl
!= EXT3_ACL_NOT_CACHED
) {
508 posix_acl_release(EXT3_I(inode
)->i_default_acl
);
509 EXT3_I(inode
)->i_default_acl
= EXT3_ACL_NOT_CACHED
;
512 ext3_discard_reservation(inode
);
513 EXT3_I(inode
)->i_block_alloc_info
= NULL
;
518 static inline void ext3_show_quota_options(struct seq_file
*seq
, struct super_block
*sb
)
520 #if defined(CONFIG_QUOTA)
521 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
523 if (sbi
->s_jquota_fmt
)
524 seq_printf(seq
, ",jqfmt=%s",
525 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold": "vfsv0");
527 if (sbi
->s_qf_names
[USRQUOTA
])
528 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
530 if (sbi
->s_qf_names
[GRPQUOTA
])
531 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
533 if (sbi
->s_mount_opt
& EXT3_MOUNT_USRQUOTA
)
534 seq_puts(seq
, ",usrquota");
536 if (sbi
->s_mount_opt
& EXT3_MOUNT_GRPQUOTA
)
537 seq_puts(seq
, ",grpquota");
541 static int ext3_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
543 struct super_block
*sb
= vfs
->mnt_sb
;
545 if (test_opt(sb
, DATA_FLAGS
) == EXT3_MOUNT_JOURNAL_DATA
)
546 seq_puts(seq
, ",data=journal");
547 else if (test_opt(sb
, DATA_FLAGS
) == EXT3_MOUNT_ORDERED_DATA
)
548 seq_puts(seq
, ",data=ordered");
549 else if (test_opt(sb
, DATA_FLAGS
) == EXT3_MOUNT_WRITEBACK_DATA
)
550 seq_puts(seq
, ",data=writeback");
552 ext3_show_quota_options(seq
, sb
);
558 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
559 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
561 static int ext3_dquot_initialize(struct inode
*inode
, int type
);
562 static int ext3_dquot_drop(struct inode
*inode
);
563 static int ext3_write_dquot(struct dquot
*dquot
);
564 static int ext3_acquire_dquot(struct dquot
*dquot
);
565 static int ext3_release_dquot(struct dquot
*dquot
);
566 static int ext3_mark_dquot_dirty(struct dquot
*dquot
);
567 static int ext3_write_info(struct super_block
*sb
, int type
);
568 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
, char *path
);
569 static int ext3_quota_on_mount(struct super_block
*sb
, int type
);
570 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
571 size_t len
, loff_t off
);
572 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
573 const char *data
, size_t len
, loff_t off
);
575 static struct dquot_operations ext3_quota_operations
= {
576 .initialize
= ext3_dquot_initialize
,
577 .drop
= ext3_dquot_drop
,
578 .alloc_space
= dquot_alloc_space
,
579 .alloc_inode
= dquot_alloc_inode
,
580 .free_space
= dquot_free_space
,
581 .free_inode
= dquot_free_inode
,
582 .transfer
= dquot_transfer
,
583 .write_dquot
= ext3_write_dquot
,
584 .acquire_dquot
= ext3_acquire_dquot
,
585 .release_dquot
= ext3_release_dquot
,
586 .mark_dirty
= ext3_mark_dquot_dirty
,
587 .write_info
= ext3_write_info
590 static struct quotactl_ops ext3_qctl_operations
= {
591 .quota_on
= ext3_quota_on
,
592 .quota_off
= vfs_quota_off
,
593 .quota_sync
= vfs_quota_sync
,
594 .get_info
= vfs_get_dqinfo
,
595 .set_info
= vfs_set_dqinfo
,
596 .get_dqblk
= vfs_get_dqblk
,
597 .set_dqblk
= vfs_set_dqblk
601 static struct super_operations ext3_sops
= {
602 .alloc_inode
= ext3_alloc_inode
,
603 .destroy_inode
= ext3_destroy_inode
,
604 .read_inode
= ext3_read_inode
,
605 .write_inode
= ext3_write_inode
,
606 .dirty_inode
= ext3_dirty_inode
,
607 .delete_inode
= ext3_delete_inode
,
608 .put_super
= ext3_put_super
,
609 .write_super
= ext3_write_super
,
610 .sync_fs
= ext3_sync_fs
,
611 .write_super_lockfs
= ext3_write_super_lockfs
,
612 .unlockfs
= ext3_unlockfs
,
613 .statfs
= ext3_statfs
,
614 .remount_fs
= ext3_remount
,
615 .clear_inode
= ext3_clear_inode
,
616 .show_options
= ext3_show_options
,
618 .quota_read
= ext3_quota_read
,
619 .quota_write
= ext3_quota_write
,
623 static struct export_operations ext3_export_ops
= {
624 .get_parent
= ext3_get_parent
,
628 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
629 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
630 Opt_nouid32
, Opt_nocheck
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
631 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
632 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
633 Opt_commit
, Opt_journal_update
, Opt_journal_inum
, Opt_journal_dev
,
634 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
635 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
636 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
637 Opt_ignore
, Opt_barrier
, Opt_err
, Opt_resize
, Opt_usrquota
,
641 static match_table_t tokens
= {
642 {Opt_bsd_df
, "bsddf"},
643 {Opt_minix_df
, "minixdf"},
644 {Opt_grpid
, "grpid"},
645 {Opt_grpid
, "bsdgroups"},
646 {Opt_nogrpid
, "nogrpid"},
647 {Opt_nogrpid
, "sysvgroups"},
648 {Opt_resgid
, "resgid=%u"},
649 {Opt_resuid
, "resuid=%u"},
651 {Opt_err_cont
, "errors=continue"},
652 {Opt_err_panic
, "errors=panic"},
653 {Opt_err_ro
, "errors=remount-ro"},
654 {Opt_nouid32
, "nouid32"},
655 {Opt_nocheck
, "nocheck"},
656 {Opt_nocheck
, "check=none"},
657 {Opt_debug
, "debug"},
658 {Opt_oldalloc
, "oldalloc"},
659 {Opt_orlov
, "orlov"},
660 {Opt_user_xattr
, "user_xattr"},
661 {Opt_nouser_xattr
, "nouser_xattr"},
663 {Opt_noacl
, "noacl"},
664 {Opt_reservation
, "reservation"},
665 {Opt_noreservation
, "noreservation"},
666 {Opt_noload
, "noload"},
669 {Opt_commit
, "commit=%u"},
670 {Opt_journal_update
, "journal=update"},
671 {Opt_journal_inum
, "journal=%u"},
672 {Opt_journal_dev
, "journal_dev=%u"},
673 {Opt_abort
, "abort"},
674 {Opt_data_journal
, "data=journal"},
675 {Opt_data_ordered
, "data=ordered"},
676 {Opt_data_writeback
, "data=writeback"},
677 {Opt_offusrjquota
, "usrjquota="},
678 {Opt_usrjquota
, "usrjquota=%s"},
679 {Opt_offgrpjquota
, "grpjquota="},
680 {Opt_grpjquota
, "grpjquota=%s"},
681 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
682 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
683 {Opt_grpquota
, "grpquota"},
684 {Opt_noquota
, "noquota"},
685 {Opt_quota
, "quota"},
686 {Opt_usrquota
, "usrquota"},
687 {Opt_barrier
, "barrier=%u"},
689 {Opt_resize
, "resize"},
692 static ext3_fsblk_t
get_sb_block(void **data
)
694 ext3_fsblk_t sb_block
;
695 char *options
= (char *) *data
;
697 if (!options
|| strncmp(options
, "sb=", 3) != 0)
698 return 1; /* Default location */
700 /*todo: use simple_strtoll with >32bit ext3 */
701 sb_block
= simple_strtoul(options
, &options
, 0);
702 if (*options
&& *options
!= ',') {
703 printk("EXT3-fs: Invalid sb specification: %s\n",
709 *data
= (void *) options
;
713 static int parse_options (char *options
, struct super_block
*sb
,
714 unsigned long *inum
, unsigned long *journal_devnum
,
715 ext3_fsblk_t
*n_blocks_count
, int is_remount
)
717 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
719 substring_t args
[MAX_OPT_ARGS
];
730 while ((p
= strsep (&options
, ",")) != NULL
) {
735 token
= match_token(p
, tokens
, args
);
738 clear_opt (sbi
->s_mount_opt
, MINIX_DF
);
741 set_opt (sbi
->s_mount_opt
, MINIX_DF
);
744 set_opt (sbi
->s_mount_opt
, GRPID
);
747 clear_opt (sbi
->s_mount_opt
, GRPID
);
750 if (match_int(&args
[0], &option
))
752 sbi
->s_resuid
= option
;
755 if (match_int(&args
[0], &option
))
757 sbi
->s_resgid
= option
;
760 /* handled by get_sb_block() instead of here */
761 /* *sb_block = match_int(&args[0]); */
764 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
765 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
766 set_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
769 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
770 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
771 set_opt (sbi
->s_mount_opt
, ERRORS_RO
);
774 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
775 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
776 set_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
779 set_opt (sbi
->s_mount_opt
, NO_UID32
);
782 clear_opt (sbi
->s_mount_opt
, CHECK
);
785 set_opt (sbi
->s_mount_opt
, DEBUG
);
788 set_opt (sbi
->s_mount_opt
, OLDALLOC
);
791 clear_opt (sbi
->s_mount_opt
, OLDALLOC
);
793 #ifdef CONFIG_EXT3_FS_XATTR
795 set_opt (sbi
->s_mount_opt
, XATTR_USER
);
797 case Opt_nouser_xattr
:
798 clear_opt (sbi
->s_mount_opt
, XATTR_USER
);
802 case Opt_nouser_xattr
:
803 printk("EXT3 (no)user_xattr options not supported\n");
806 #ifdef CONFIG_EXT3_FS_POSIX_ACL
808 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
811 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
816 printk("EXT3 (no)acl options not supported\n");
819 case Opt_reservation
:
820 set_opt(sbi
->s_mount_opt
, RESERVATION
);
822 case Opt_noreservation
:
823 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
825 case Opt_journal_update
:
827 /* Eventually we will want to be able to create
828 a journal file here. For now, only allow the
829 user to specify an existing inode to be the
832 printk(KERN_ERR
"EXT3-fs: cannot specify "
833 "journal on remount\n");
836 set_opt (sbi
->s_mount_opt
, UPDATE_JOURNAL
);
838 case Opt_journal_inum
:
840 printk(KERN_ERR
"EXT3-fs: cannot specify "
841 "journal on remount\n");
844 if (match_int(&args
[0], &option
))
848 case Opt_journal_dev
:
850 printk(KERN_ERR
"EXT3-fs: cannot specify "
851 "journal on remount\n");
854 if (match_int(&args
[0], &option
))
856 *journal_devnum
= option
;
859 set_opt (sbi
->s_mount_opt
, NOLOAD
);
862 if (match_int(&args
[0], &option
))
867 option
= JBD_DEFAULT_MAX_COMMIT_AGE
;
868 sbi
->s_commit_interval
= HZ
* option
;
870 case Opt_data_journal
:
871 data_opt
= EXT3_MOUNT_JOURNAL_DATA
;
873 case Opt_data_ordered
:
874 data_opt
= EXT3_MOUNT_ORDERED_DATA
;
876 case Opt_data_writeback
:
877 data_opt
= EXT3_MOUNT_WRITEBACK_DATA
;
880 if ((sbi
->s_mount_opt
& EXT3_MOUNT_DATA_FLAGS
)
883 "EXT3-fs: cannot change data "
884 "mode on remount\n");
888 sbi
->s_mount_opt
&= ~EXT3_MOUNT_DATA_FLAGS
;
889 sbi
->s_mount_opt
|= data_opt
;
899 if (sb_any_quota_enabled(sb
)) {
901 "EXT3-fs: Cannot change journalled "
902 "quota options when quota turned on.\n");
905 qname
= match_strdup(&args
[0]);
908 "EXT3-fs: not enough memory for "
909 "storing quotafile name.\n");
912 if (sbi
->s_qf_names
[qtype
] &&
913 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
915 "EXT3-fs: %s quota file already "
916 "specified.\n", QTYPE2NAME(qtype
));
920 sbi
->s_qf_names
[qtype
] = qname
;
921 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
923 "EXT3-fs: quotafile must be on "
924 "filesystem root.\n");
925 kfree(sbi
->s_qf_names
[qtype
]);
926 sbi
->s_qf_names
[qtype
] = NULL
;
929 set_opt(sbi
->s_mount_opt
, QUOTA
);
931 case Opt_offusrjquota
:
934 case Opt_offgrpjquota
:
937 if (sb_any_quota_enabled(sb
)) {
938 printk(KERN_ERR
"EXT3-fs: Cannot change "
939 "journalled quota options when "
940 "quota turned on.\n");
944 * The space will be released later when all options
945 * are confirmed to be correct
947 sbi
->s_qf_names
[qtype
] = NULL
;
949 case Opt_jqfmt_vfsold
:
950 sbi
->s_jquota_fmt
= QFMT_VFS_OLD
;
952 case Opt_jqfmt_vfsv0
:
953 sbi
->s_jquota_fmt
= QFMT_VFS_V0
;
957 set_opt(sbi
->s_mount_opt
, QUOTA
);
958 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
961 set_opt(sbi
->s_mount_opt
, QUOTA
);
962 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
965 if (sb_any_quota_enabled(sb
)) {
966 printk(KERN_ERR
"EXT3-fs: Cannot change quota "
967 "options when quota turned on.\n");
970 clear_opt(sbi
->s_mount_opt
, QUOTA
);
971 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
972 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
980 case Opt_offusrjquota
:
981 case Opt_offgrpjquota
:
982 case Opt_jqfmt_vfsold
:
983 case Opt_jqfmt_vfsv0
:
985 "EXT3-fs: journalled quota options not "
992 set_opt(sbi
->s_mount_opt
, ABORT
);
995 if (match_int(&args
[0], &option
))
998 set_opt(sbi
->s_mount_opt
, BARRIER
);
1000 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1006 printk("EXT3-fs: resize option only available "
1010 if (match_int(&args
[0], &option
) != 0)
1012 *n_blocks_count
= option
;
1015 set_opt(sbi
->s_mount_opt
, NOBH
);
1018 clear_opt(sbi
->s_mount_opt
, NOBH
);
1022 "EXT3-fs: Unrecognized mount option \"%s\" "
1023 "or missing value\n", p
);
1028 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1029 if ((sbi
->s_mount_opt
& EXT3_MOUNT_USRQUOTA
) &&
1030 sbi
->s_qf_names
[USRQUOTA
])
1031 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1033 if ((sbi
->s_mount_opt
& EXT3_MOUNT_GRPQUOTA
) &&
1034 sbi
->s_qf_names
[GRPQUOTA
])
1035 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1037 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1038 (sbi
->s_mount_opt
& EXT3_MOUNT_GRPQUOTA
)) ||
1039 (sbi
->s_qf_names
[GRPQUOTA
] &&
1040 (sbi
->s_mount_opt
& EXT3_MOUNT_USRQUOTA
))) {
1041 printk(KERN_ERR
"EXT3-fs: old and new quota "
1042 "format mixing.\n");
1046 if (!sbi
->s_jquota_fmt
) {
1047 printk(KERN_ERR
"EXT3-fs: journalled quota format "
1048 "not specified.\n");
1052 if (sbi
->s_jquota_fmt
) {
1053 printk(KERN_ERR
"EXT3-fs: journalled quota format "
1054 "specified with no journalling "
1063 static int ext3_setup_super(struct super_block
*sb
, struct ext3_super_block
*es
,
1066 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1069 if (le32_to_cpu(es
->s_rev_level
) > EXT3_MAX_SUPP_REV
) {
1070 printk (KERN_ERR
"EXT3-fs warning: revision level too high, "
1071 "forcing read-only mode\n");
1076 if (!(sbi
->s_mount_state
& EXT3_VALID_FS
))
1077 printk (KERN_WARNING
"EXT3-fs warning: mounting unchecked fs, "
1078 "running e2fsck is recommended\n");
1079 else if ((sbi
->s_mount_state
& EXT3_ERROR_FS
))
1080 printk (KERN_WARNING
1081 "EXT3-fs warning: mounting fs with errors, "
1082 "running e2fsck is recommended\n");
1083 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1084 le16_to_cpu(es
->s_mnt_count
) >=
1085 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1086 printk (KERN_WARNING
1087 "EXT3-fs warning: maximal mount count reached, "
1088 "running e2fsck is recommended\n");
1089 else if (le32_to_cpu(es
->s_checkinterval
) &&
1090 (le32_to_cpu(es
->s_lastcheck
) +
1091 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1092 printk (KERN_WARNING
1093 "EXT3-fs warning: checktime reached, "
1094 "running e2fsck is recommended\n");
1096 /* @@@ We _will_ want to clear the valid bit if we find
1097 inconsistencies, to force a fsck at reboot. But for
1098 a plain journaled filesystem we can keep it set as
1099 valid forever! :) */
1100 es
->s_state
= cpu_to_le16(le16_to_cpu(es
->s_state
) & ~EXT3_VALID_FS
);
1102 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1103 es
->s_max_mnt_count
= cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT
);
1104 es
->s_mnt_count
=cpu_to_le16(le16_to_cpu(es
->s_mnt_count
) + 1);
1105 es
->s_mtime
= cpu_to_le32(get_seconds());
1106 ext3_update_dynamic_rev(sb
);
1107 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
1109 ext3_commit_super(sb
, es
, 1);
1110 if (test_opt(sb
, DEBUG
))
1111 printk(KERN_INFO
"[EXT3 FS bs=%lu, gc=%lu, "
1112 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1114 sbi
->s_groups_count
,
1115 EXT3_BLOCKS_PER_GROUP(sb
),
1116 EXT3_INODES_PER_GROUP(sb
),
1119 printk(KERN_INFO
"EXT3 FS on %s, ", sb
->s_id
);
1120 if (EXT3_SB(sb
)->s_journal
->j_inode
== NULL
) {
1121 char b
[BDEVNAME_SIZE
];
1123 printk("external journal on %s\n",
1124 bdevname(EXT3_SB(sb
)->s_journal
->j_dev
, b
));
1126 printk("internal journal\n");
1131 /* Called at mount-time, super-block is locked */
1132 static int ext3_check_descriptors (struct super_block
* sb
)
1134 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1135 ext3_fsblk_t block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1136 struct ext3_group_desc
* gdp
= NULL
;
1140 ext3_debug ("Checking group descriptors");
1142 for (i
= 0; i
< sbi
->s_groups_count
; i
++)
1144 if ((i
% EXT3_DESC_PER_BLOCK(sb
)) == 0)
1145 gdp
= (struct ext3_group_desc
*)
1146 sbi
->s_group_desc
[desc_block
++]->b_data
;
1147 if (le32_to_cpu(gdp
->bg_block_bitmap
) < block
||
1148 le32_to_cpu(gdp
->bg_block_bitmap
) >=
1149 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1151 ext3_error (sb
, "ext3_check_descriptors",
1152 "Block bitmap for group %d"
1153 " not in group (block %lu)!",
1155 le32_to_cpu(gdp
->bg_block_bitmap
));
1158 if (le32_to_cpu(gdp
->bg_inode_bitmap
) < block
||
1159 le32_to_cpu(gdp
->bg_inode_bitmap
) >=
1160 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1162 ext3_error (sb
, "ext3_check_descriptors",
1163 "Inode bitmap for group %d"
1164 " not in group (block %lu)!",
1166 le32_to_cpu(gdp
->bg_inode_bitmap
));
1169 if (le32_to_cpu(gdp
->bg_inode_table
) < block
||
1170 le32_to_cpu(gdp
->bg_inode_table
) + sbi
->s_itb_per_group
>=
1171 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1173 ext3_error (sb
, "ext3_check_descriptors",
1174 "Inode table for group %d"
1175 " not in group (block %lu)!",
1177 le32_to_cpu(gdp
->bg_inode_table
));
1180 block
+= EXT3_BLOCKS_PER_GROUP(sb
);
1184 sbi
->s_es
->s_free_blocks_count
=cpu_to_le32(ext3_count_free_blocks(sb
));
1185 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext3_count_free_inodes(sb
));
1190 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1191 * the superblock) which were deleted from all directories, but held open by
1192 * a process at the time of a crash. We walk the list and try to delete these
1193 * inodes at recovery time (only with a read-write filesystem).
1195 * In order to keep the orphan inode chain consistent during traversal (in
1196 * case of crash during recovery), we link each inode into the superblock
1197 * orphan list_head and handle it the same way as an inode deletion during
1198 * normal operation (which journals the operations for us).
1200 * We only do an iget() and an iput() on each inode, which is very safe if we
1201 * accidentally point at an in-use or already deleted inode. The worst that
1202 * can happen in this case is that we get a "bit already cleared" message from
1203 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1204 * e2fsck was run on this filesystem, and it must have already done the orphan
1205 * inode cleanup for us, so we can safely abort without any further action.
1207 static void ext3_orphan_cleanup (struct super_block
* sb
,
1208 struct ext3_super_block
* es
)
1210 unsigned int s_flags
= sb
->s_flags
;
1211 int nr_orphans
= 0, nr_truncates
= 0;
1215 if (!es
->s_last_orphan
) {
1216 jbd_debug(4, "no orphan inodes to clean up\n");
1220 if (EXT3_SB(sb
)->s_mount_state
& EXT3_ERROR_FS
) {
1221 if (es
->s_last_orphan
)
1222 jbd_debug(1, "Errors on filesystem, "
1223 "clearing orphan list.\n");
1224 es
->s_last_orphan
= 0;
1225 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1229 if (s_flags
& MS_RDONLY
) {
1230 printk(KERN_INFO
"EXT3-fs: %s: orphan cleanup on readonly fs\n",
1232 sb
->s_flags
&= ~MS_RDONLY
;
1235 /* Needed for iput() to work correctly and not trash data */
1236 sb
->s_flags
|= MS_ACTIVE
;
1237 /* Turn on quotas so that they are updated correctly */
1238 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1239 if (EXT3_SB(sb
)->s_qf_names
[i
]) {
1240 int ret
= ext3_quota_on_mount(sb
, i
);
1243 "EXT3-fs: Cannot turn on journalled "
1244 "quota: error %d\n", ret
);
1249 while (es
->s_last_orphan
) {
1250 struct inode
*inode
;
1253 ext3_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
)))) {
1254 es
->s_last_orphan
= 0;
1258 list_add(&EXT3_I(inode
)->i_orphan
, &EXT3_SB(sb
)->s_orphan
);
1260 if (inode
->i_nlink
) {
1262 "%s: truncating inode %ld to %Ld bytes\n",
1263 __FUNCTION__
, inode
->i_ino
, inode
->i_size
);
1264 jbd_debug(2, "truncating inode %ld to %Ld bytes\n",
1265 inode
->i_ino
, inode
->i_size
);
1266 ext3_truncate(inode
);
1270 "%s: deleting unreferenced inode %ld\n",
1271 __FUNCTION__
, inode
->i_ino
);
1272 jbd_debug(2, "deleting unreferenced inode %ld\n",
1276 iput(inode
); /* The delete magic happens here! */
1279 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1282 printk(KERN_INFO
"EXT3-fs: %s: %d orphan inode%s deleted\n",
1283 sb
->s_id
, PLURAL(nr_orphans
));
1285 printk(KERN_INFO
"EXT3-fs: %s: %d truncate%s cleaned up\n",
1286 sb
->s_id
, PLURAL(nr_truncates
));
1288 /* Turn quotas off */
1289 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1290 if (sb_dqopt(sb
)->files
[i
])
1291 vfs_quota_off(sb
, i
);
1294 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1297 #define log2(n) ffz(~(n))
1300 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1301 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1302 * We need to be 1 filesystem block less than the 2^32 sector limit.
1304 static loff_t
ext3_max_size(int bits
)
1306 loff_t res
= EXT3_NDIR_BLOCKS
;
1307 /* This constant is calculated to be the largest file size for a
1308 * dense, 4k-blocksize file such that the total number of
1309 * sectors in the file, including data and all indirect blocks,
1310 * does not exceed 2^32. */
1311 const loff_t upper_limit
= 0x1ff7fffd000LL
;
1313 res
+= 1LL << (bits
-2);
1314 res
+= 1LL << (2*(bits
-2));
1315 res
+= 1LL << (3*(bits
-2));
1317 if (res
> upper_limit
)
1322 static ext3_fsblk_t
descriptor_loc(struct super_block
*sb
,
1323 ext3_fsblk_t logic_sb_block
,
1326 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1327 unsigned long bg
, first_meta_bg
;
1330 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1332 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_META_BG
) ||
1334 return (logic_sb_block
+ nr
+ 1);
1335 bg
= sbi
->s_desc_per_block
* nr
;
1336 if (ext3_bg_has_super(sb
, bg
))
1338 return (has_super
+ ext3_group_first_block_no(sb
, bg
));
1342 static int ext3_fill_super (struct super_block
*sb
, void *data
, int silent
)
1344 struct buffer_head
* bh
;
1345 struct ext3_super_block
*es
= NULL
;
1346 struct ext3_sb_info
*sbi
;
1348 ext3_fsblk_t sb_block
= get_sb_block(&data
);
1349 ext3_fsblk_t logic_sb_block
;
1350 unsigned long offset
= 0;
1351 unsigned long journal_inum
= 0;
1352 unsigned long journal_devnum
= 0;
1353 unsigned long def_mount_opts
;
1362 sbi
= kmalloc(sizeof(*sbi
), GFP_KERNEL
);
1365 sb
->s_fs_info
= sbi
;
1366 memset(sbi
, 0, sizeof(*sbi
));
1367 sbi
->s_mount_opt
= 0;
1368 sbi
->s_resuid
= EXT3_DEF_RESUID
;
1369 sbi
->s_resgid
= EXT3_DEF_RESGID
;
1373 blocksize
= sb_min_blocksize(sb
, EXT3_MIN_BLOCK_SIZE
);
1375 printk(KERN_ERR
"EXT3-fs: unable to set blocksize\n");
1380 * The ext3 superblock will not be buffer aligned for other than 1kB
1381 * block sizes. We need to calculate the offset from buffer start.
1383 if (blocksize
!= EXT3_MIN_BLOCK_SIZE
) {
1384 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1385 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1387 logic_sb_block
= sb_block
;
1390 if (!(bh
= sb_bread(sb
, logic_sb_block
))) {
1391 printk (KERN_ERR
"EXT3-fs: unable to read superblock\n");
1395 * Note: s_es must be initialized as soon as possible because
1396 * some ext3 macro-instructions depend on its value
1398 es
= (struct ext3_super_block
*) (((char *)bh
->b_data
) + offset
);
1400 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1401 if (sb
->s_magic
!= EXT3_SUPER_MAGIC
)
1404 /* Set defaults before we parse the mount options */
1405 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1406 if (def_mount_opts
& EXT3_DEFM_DEBUG
)
1407 set_opt(sbi
->s_mount_opt
, DEBUG
);
1408 if (def_mount_opts
& EXT3_DEFM_BSDGROUPS
)
1409 set_opt(sbi
->s_mount_opt
, GRPID
);
1410 if (def_mount_opts
& EXT3_DEFM_UID16
)
1411 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1412 if (def_mount_opts
& EXT3_DEFM_XATTR_USER
)
1413 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1414 if (def_mount_opts
& EXT3_DEFM_ACL
)
1415 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1416 if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_DATA
)
1417 sbi
->s_mount_opt
|= EXT3_MOUNT_JOURNAL_DATA
;
1418 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_ORDERED
)
1419 sbi
->s_mount_opt
|= EXT3_MOUNT_ORDERED_DATA
;
1420 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_WBACK
)
1421 sbi
->s_mount_opt
|= EXT3_MOUNT_WRITEBACK_DATA
;
1423 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_PANIC
)
1424 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1425 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_RO
)
1426 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1428 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
1429 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
1431 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1433 if (!parse_options ((char *) data
, sb
, &journal_inum
, &journal_devnum
,
1437 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1438 ((sbi
->s_mount_opt
& EXT3_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1440 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
&&
1441 (EXT3_HAS_COMPAT_FEATURE(sb
, ~0U) ||
1442 EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
1443 EXT3_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
1445 "EXT3-fs warning: feature flags set on rev 0 fs, "
1446 "running e2fsck is recommended\n");
1448 * Check feature flags regardless of the revision level, since we
1449 * previously didn't change the revision level when setting the flags,
1450 * so there is a chance incompat flags are set on a rev 0 filesystem.
1452 features
= EXT3_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
);
1454 printk(KERN_ERR
"EXT3-fs: %s: couldn't mount because of "
1455 "unsupported optional features (%x).\n",
1456 sb
->s_id
, le32_to_cpu(features
));
1459 features
= EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
);
1460 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1461 printk(KERN_ERR
"EXT3-fs: %s: couldn't mount RDWR because of "
1462 "unsupported optional features (%x).\n",
1463 sb
->s_id
, le32_to_cpu(features
));
1466 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
1468 if (blocksize
< EXT3_MIN_BLOCK_SIZE
||
1469 blocksize
> EXT3_MAX_BLOCK_SIZE
) {
1471 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1472 blocksize
, sb
->s_id
);
1476 hblock
= bdev_hardsect_size(sb
->s_bdev
);
1477 if (sb
->s_blocksize
!= blocksize
) {
1479 * Make sure the blocksize for the filesystem is larger
1480 * than the hardware sectorsize for the machine.
1482 if (blocksize
< hblock
) {
1483 printk(KERN_ERR
"EXT3-fs: blocksize %d too small for "
1484 "device blocksize %d.\n", blocksize
, hblock
);
1489 sb_set_blocksize(sb
, blocksize
);
1490 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1491 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1492 bh
= sb_bread(sb
, logic_sb_block
);
1495 "EXT3-fs: Can't read superblock on 2nd try.\n");
1498 es
= (struct ext3_super_block
*)(((char *)bh
->b_data
) + offset
);
1500 if (es
->s_magic
!= cpu_to_le16(EXT3_SUPER_MAGIC
)) {
1502 "EXT3-fs: Magic mismatch, very weird !\n");
1507 sb
->s_maxbytes
= ext3_max_size(sb
->s_blocksize_bits
);
1509 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
) {
1510 sbi
->s_inode_size
= EXT3_GOOD_OLD_INODE_SIZE
;
1511 sbi
->s_first_ino
= EXT3_GOOD_OLD_FIRST_INO
;
1513 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
1514 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
1515 if ((sbi
->s_inode_size
< EXT3_GOOD_OLD_INODE_SIZE
) ||
1516 (sbi
->s_inode_size
& (sbi
->s_inode_size
- 1)) ||
1517 (sbi
->s_inode_size
> blocksize
)) {
1519 "EXT3-fs: unsupported inode size: %d\n",
1524 sbi
->s_frag_size
= EXT3_MIN_FRAG_SIZE
<<
1525 le32_to_cpu(es
->s_log_frag_size
);
1526 if (blocksize
!= sbi
->s_frag_size
) {
1528 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1529 sbi
->s_frag_size
, blocksize
);
1532 sbi
->s_frags_per_block
= 1;
1533 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
1534 sbi
->s_frags_per_group
= le32_to_cpu(es
->s_frags_per_group
);
1535 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
1536 if (EXT3_INODE_SIZE(sb
) == 0)
1538 sbi
->s_inodes_per_block
= blocksize
/ EXT3_INODE_SIZE(sb
);
1539 if (sbi
->s_inodes_per_block
== 0)
1541 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
1542 sbi
->s_inodes_per_block
;
1543 sbi
->s_desc_per_block
= blocksize
/ sizeof(struct ext3_group_desc
);
1545 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
1546 sbi
->s_addr_per_block_bits
= log2(EXT3_ADDR_PER_BLOCK(sb
));
1547 sbi
->s_desc_per_block_bits
= log2(EXT3_DESC_PER_BLOCK(sb
));
1548 for (i
=0; i
< 4; i
++)
1549 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
1550 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
1552 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
1554 "EXT3-fs: #blocks per group too big: %lu\n",
1555 sbi
->s_blocks_per_group
);
1558 if (sbi
->s_frags_per_group
> blocksize
* 8) {
1560 "EXT3-fs: #fragments per group too big: %lu\n",
1561 sbi
->s_frags_per_group
);
1564 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
1566 "EXT3-fs: #inodes per group too big: %lu\n",
1567 sbi
->s_inodes_per_group
);
1571 if (le32_to_cpu(es
->s_blocks_count
) >
1572 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) {
1573 printk(KERN_ERR
"EXT3-fs: filesystem on %s:"
1574 " too large to mount safely\n", sb
->s_id
);
1575 if (sizeof(sector_t
) < 8)
1576 printk(KERN_WARNING
"EXT3-fs: CONFIG_LBD not "
1581 if (EXT3_BLOCKS_PER_GROUP(sb
) == 0)
1583 sbi
->s_groups_count
= (le32_to_cpu(es
->s_blocks_count
) -
1584 le32_to_cpu(es
->s_first_data_block
) +
1585 EXT3_BLOCKS_PER_GROUP(sb
) - 1) /
1586 EXT3_BLOCKS_PER_GROUP(sb
);
1587 db_count
= (sbi
->s_groups_count
+ EXT3_DESC_PER_BLOCK(sb
) - 1) /
1588 EXT3_DESC_PER_BLOCK(sb
);
1589 sbi
->s_group_desc
= kmalloc(db_count
* sizeof (struct buffer_head
*),
1591 if (sbi
->s_group_desc
== NULL
) {
1592 printk (KERN_ERR
"EXT3-fs: not enough memory\n");
1596 bgl_lock_init(&sbi
->s_blockgroup_lock
);
1598 for (i
= 0; i
< db_count
; i
++) {
1599 block
= descriptor_loc(sb
, logic_sb_block
, i
);
1600 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
1601 if (!sbi
->s_group_desc
[i
]) {
1602 printk (KERN_ERR
"EXT3-fs: "
1603 "can't read group descriptor %d\n", i
);
1608 if (!ext3_check_descriptors (sb
)) {
1609 printk(KERN_ERR
"EXT3-fs: group descriptors corrupted!\n");
1612 sbi
->s_gdb_count
= db_count
;
1613 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1614 spin_lock_init(&sbi
->s_next_gen_lock
);
1616 percpu_counter_init(&sbi
->s_freeblocks_counter
,
1617 ext3_count_free_blocks(sb
));
1618 percpu_counter_init(&sbi
->s_freeinodes_counter
,
1619 ext3_count_free_inodes(sb
));
1620 percpu_counter_init(&sbi
->s_dirs_counter
,
1621 ext3_count_dirs(sb
));
1623 /* per fileystem reservation list head & lock */
1624 spin_lock_init(&sbi
->s_rsv_window_lock
);
1625 sbi
->s_rsv_window_root
= RB_ROOT
;
1626 /* Add a single, static dummy reservation to the start of the
1627 * reservation window list --- it gives us a placeholder for
1628 * append-at-start-of-list which makes the allocation logic
1629 * _much_ simpler. */
1630 sbi
->s_rsv_window_head
.rsv_start
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1631 sbi
->s_rsv_window_head
.rsv_end
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1632 sbi
->s_rsv_window_head
.rsv_alloc_hit
= 0;
1633 sbi
->s_rsv_window_head
.rsv_goal_size
= 0;
1634 ext3_rsv_window_add(sb
, &sbi
->s_rsv_window_head
);
1637 * set up enough so that it can read an inode
1639 sb
->s_op
= &ext3_sops
;
1640 sb
->s_export_op
= &ext3_export_ops
;
1641 sb
->s_xattr
= ext3_xattr_handlers
;
1643 sb
->s_qcop
= &ext3_qctl_operations
;
1644 sb
->dq_op
= &ext3_quota_operations
;
1646 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
1650 needs_recovery
= (es
->s_last_orphan
!= 0 ||
1651 EXT3_HAS_INCOMPAT_FEATURE(sb
,
1652 EXT3_FEATURE_INCOMPAT_RECOVER
));
1655 * The first inode we look at is the journal inode. Don't try
1656 * root first: it may be modified in the journal!
1658 if (!test_opt(sb
, NOLOAD
) &&
1659 EXT3_HAS_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
)) {
1660 if (ext3_load_journal(sb
, es
, journal_devnum
))
1662 } else if (journal_inum
) {
1663 if (ext3_create_journal(sb
, es
, journal_inum
))
1668 "ext3: No journal on filesystem on %s\n",
1673 /* We have now updated the journal if required, so we can
1674 * validate the data journaling mode. */
1675 switch (test_opt(sb
, DATA_FLAGS
)) {
1677 /* No mode set, assume a default based on the journal
1678 capabilities: ORDERED_DATA if the journal can
1679 cope, else JOURNAL_DATA */
1680 if (journal_check_available_features
1681 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
))
1682 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
1684 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
1687 case EXT3_MOUNT_ORDERED_DATA
:
1688 case EXT3_MOUNT_WRITEBACK_DATA
:
1689 if (!journal_check_available_features
1690 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
)) {
1691 printk(KERN_ERR
"EXT3-fs: Journal does not support "
1692 "requested data journaling mode\n");
1699 if (test_opt(sb
, NOBH
)) {
1700 if (!(test_opt(sb
, DATA_FLAGS
) == EXT3_MOUNT_WRITEBACK_DATA
)) {
1701 printk(KERN_WARNING
"EXT3-fs: Ignoring nobh option - "
1702 "its supported only with writeback mode\n");
1703 clear_opt(sbi
->s_mount_opt
, NOBH
);
1707 * The journal_load will have done any necessary log recovery,
1708 * so we can safely mount the rest of the filesystem now.
1711 root
= iget(sb
, EXT3_ROOT_INO
);
1712 sb
->s_root
= d_alloc_root(root
);
1714 printk(KERN_ERR
"EXT3-fs: get root inode failed\n");
1718 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1721 printk(KERN_ERR
"EXT3-fs: corrupt root inode, run e2fsck\n");
1725 ext3_setup_super (sb
, es
, sb
->s_flags
& MS_RDONLY
);
1727 * akpm: core read_super() calls in here with the superblock locked.
1728 * That deadlocks, because orphan cleanup needs to lock the superblock
1729 * in numerous places. Here we just pop the lock - it's relatively
1730 * harmless, because we are now ready to accept write_super() requests,
1731 * and aviro says that's the only reason for hanging onto the
1734 EXT3_SB(sb
)->s_mount_state
|= EXT3_ORPHAN_FS
;
1735 ext3_orphan_cleanup(sb
, es
);
1736 EXT3_SB(sb
)->s_mount_state
&= ~EXT3_ORPHAN_FS
;
1738 printk (KERN_INFO
"EXT3-fs: recovery complete.\n");
1739 ext3_mark_recovery_complete(sb
, es
);
1740 printk (KERN_INFO
"EXT3-fs: mounted filesystem with %s data mode.\n",
1741 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_JOURNAL_DATA
? "journal":
1742 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_ORDERED_DATA
? "ordered":
1750 printk(KERN_ERR
"VFS: Can't find ext3 filesystem on dev %s.\n",
1755 journal_destroy(sbi
->s_journal
);
1757 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
1758 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
1759 percpu_counter_destroy(&sbi
->s_dirs_counter
);
1761 for (i
= 0; i
< db_count
; i
++)
1762 brelse(sbi
->s_group_desc
[i
]);
1763 kfree(sbi
->s_group_desc
);
1766 for (i
= 0; i
< MAXQUOTAS
; i
++)
1767 kfree(sbi
->s_qf_names
[i
]);
1769 ext3_blkdev_remove(sbi
);
1772 sb
->s_fs_info
= NULL
;
1779 * Setup any per-fs journal parameters now. We'll do this both on
1780 * initial mount, once the journal has been initialised but before we've
1781 * done any recovery; and again on any subsequent remount.
1783 static void ext3_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
1785 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1787 if (sbi
->s_commit_interval
)
1788 journal
->j_commit_interval
= sbi
->s_commit_interval
;
1789 /* We could also set up an ext3-specific default for the commit
1790 * interval here, but for now we'll just fall back to the jbd
1793 spin_lock(&journal
->j_state_lock
);
1794 if (test_opt(sb
, BARRIER
))
1795 journal
->j_flags
|= JFS_BARRIER
;
1797 journal
->j_flags
&= ~JFS_BARRIER
;
1798 spin_unlock(&journal
->j_state_lock
);
1801 static journal_t
*ext3_get_journal(struct super_block
*sb
, int journal_inum
)
1803 struct inode
*journal_inode
;
1806 /* First, test for the existence of a valid inode on disk. Bad
1807 * things happen if we iget() an unused inode, as the subsequent
1808 * iput() will try to delete it. */
1810 journal_inode
= iget(sb
, journal_inum
);
1811 if (!journal_inode
) {
1812 printk(KERN_ERR
"EXT3-fs: no journal found.\n");
1815 if (!journal_inode
->i_nlink
) {
1816 make_bad_inode(journal_inode
);
1817 iput(journal_inode
);
1818 printk(KERN_ERR
"EXT3-fs: journal inode is deleted.\n");
1822 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1823 journal_inode
, journal_inode
->i_size
);
1824 if (is_bad_inode(journal_inode
) || !S_ISREG(journal_inode
->i_mode
)) {
1825 printk(KERN_ERR
"EXT3-fs: invalid journal inode.\n");
1826 iput(journal_inode
);
1830 journal
= journal_init_inode(journal_inode
);
1832 printk(KERN_ERR
"EXT3-fs: Could not load journal inode\n");
1833 iput(journal_inode
);
1836 journal
->j_private
= sb
;
1837 ext3_init_journal_params(sb
, journal
);
1841 static journal_t
*ext3_get_dev_journal(struct super_block
*sb
,
1844 struct buffer_head
* bh
;
1848 int hblock
, blocksize
;
1849 ext3_fsblk_t sb_block
;
1850 unsigned long offset
;
1851 struct ext3_super_block
* es
;
1852 struct block_device
*bdev
;
1854 bdev
= ext3_blkdev_get(j_dev
);
1858 if (bd_claim(bdev
, sb
)) {
1860 "EXT3: failed to claim external journal device.\n");
1865 blocksize
= sb
->s_blocksize
;
1866 hblock
= bdev_hardsect_size(bdev
);
1867 if (blocksize
< hblock
) {
1869 "EXT3-fs: blocksize too small for journal device.\n");
1873 sb_block
= EXT3_MIN_BLOCK_SIZE
/ blocksize
;
1874 offset
= EXT3_MIN_BLOCK_SIZE
% blocksize
;
1875 set_blocksize(bdev
, blocksize
);
1876 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
1877 printk(KERN_ERR
"EXT3-fs: couldn't read superblock of "
1878 "external journal\n");
1882 es
= (struct ext3_super_block
*) (((char *)bh
->b_data
) + offset
);
1883 if ((le16_to_cpu(es
->s_magic
) != EXT3_SUPER_MAGIC
) ||
1884 !(le32_to_cpu(es
->s_feature_incompat
) &
1885 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
1886 printk(KERN_ERR
"EXT3-fs: external journal has "
1887 "bad superblock\n");
1892 if (memcmp(EXT3_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
1893 printk(KERN_ERR
"EXT3-fs: journal UUID does not match\n");
1898 len
= le32_to_cpu(es
->s_blocks_count
);
1899 start
= sb_block
+ 1;
1900 brelse(bh
); /* we're done with the superblock */
1902 journal
= journal_init_dev(bdev
, sb
->s_bdev
,
1903 start
, len
, blocksize
);
1905 printk(KERN_ERR
"EXT3-fs: failed to create device journal\n");
1908 journal
->j_private
= sb
;
1909 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
1910 wait_on_buffer(journal
->j_sb_buffer
);
1911 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
1912 printk(KERN_ERR
"EXT3-fs: I/O error on journal device\n");
1915 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
1916 printk(KERN_ERR
"EXT3-fs: External journal has more than one "
1917 "user (unsupported) - %d\n",
1918 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
1921 EXT3_SB(sb
)->journal_bdev
= bdev
;
1922 ext3_init_journal_params(sb
, journal
);
1925 journal_destroy(journal
);
1927 ext3_blkdev_put(bdev
);
1931 static int ext3_load_journal(struct super_block
*sb
,
1932 struct ext3_super_block
*es
,
1933 unsigned long journal_devnum
)
1936 int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
1939 int really_read_only
;
1941 if (journal_devnum
&&
1942 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
1943 printk(KERN_INFO
"EXT3-fs: external journal device major/minor "
1944 "numbers have changed\n");
1945 journal_dev
= new_decode_dev(journal_devnum
);
1947 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
1949 really_read_only
= bdev_read_only(sb
->s_bdev
);
1952 * Are we loading a blank journal or performing recovery after a
1953 * crash? For recovery, we need to check in advance whether we
1954 * can get read-write access to the device.
1957 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
)) {
1958 if (sb
->s_flags
& MS_RDONLY
) {
1959 printk(KERN_INFO
"EXT3-fs: INFO: recovery "
1960 "required on readonly filesystem.\n");
1961 if (really_read_only
) {
1962 printk(KERN_ERR
"EXT3-fs: write access "
1963 "unavailable, cannot proceed.\n");
1966 printk (KERN_INFO
"EXT3-fs: write access will "
1967 "be enabled during recovery.\n");
1971 if (journal_inum
&& journal_dev
) {
1972 printk(KERN_ERR
"EXT3-fs: filesystem has both journal "
1973 "and inode journals!\n");
1978 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
1981 if (!(journal
= ext3_get_dev_journal(sb
, journal_dev
)))
1985 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
1986 err
= journal_update_format(journal
);
1988 printk(KERN_ERR
"EXT3-fs: error updating journal.\n");
1989 journal_destroy(journal
);
1994 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
))
1995 err
= journal_wipe(journal
, !really_read_only
);
1997 err
= journal_load(journal
);
2000 printk(KERN_ERR
"EXT3-fs: error loading journal.\n");
2001 journal_destroy(journal
);
2005 EXT3_SB(sb
)->s_journal
= journal
;
2006 ext3_clear_journal_err(sb
, es
);
2008 if (journal_devnum
&&
2009 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2010 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2013 /* Make sure we flush the recovery flag to disk. */
2014 ext3_commit_super(sb
, es
, 1);
2020 static int ext3_create_journal(struct super_block
* sb
,
2021 struct ext3_super_block
* es
,
2026 if (sb
->s_flags
& MS_RDONLY
) {
2027 printk(KERN_ERR
"EXT3-fs: readonly filesystem when trying to "
2028 "create journal.\n");
2032 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
2035 printk(KERN_INFO
"EXT3-fs: creating new journal on inode %d\n",
2038 if (journal_create(journal
)) {
2039 printk(KERN_ERR
"EXT3-fs: error creating journal.\n");
2040 journal_destroy(journal
);
2044 EXT3_SB(sb
)->s_journal
= journal
;
2046 ext3_update_dynamic_rev(sb
);
2047 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2048 EXT3_SET_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
);
2050 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
2053 /* Make sure we flush the recovery flag to disk. */
2054 ext3_commit_super(sb
, es
, 1);
2059 static void ext3_commit_super (struct super_block
* sb
,
2060 struct ext3_super_block
* es
,
2063 struct buffer_head
*sbh
= EXT3_SB(sb
)->s_sbh
;
2067 es
->s_wtime
= cpu_to_le32(get_seconds());
2068 es
->s_free_blocks_count
= cpu_to_le32(ext3_count_free_blocks(sb
));
2069 es
->s_free_inodes_count
= cpu_to_le32(ext3_count_free_inodes(sb
));
2070 BUFFER_TRACE(sbh
, "marking dirty");
2071 mark_buffer_dirty(sbh
);
2073 sync_dirty_buffer(sbh
);
2078 * Have we just finished recovery? If so, and if we are mounting (or
2079 * remounting) the filesystem readonly, then we will end up with a
2080 * consistent fs on disk. Record that fact.
2082 static void ext3_mark_recovery_complete(struct super_block
* sb
,
2083 struct ext3_super_block
* es
)
2085 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
2087 journal_lock_updates(journal
);
2088 journal_flush(journal
);
2089 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
) &&
2090 sb
->s_flags
& MS_RDONLY
) {
2091 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2093 ext3_commit_super(sb
, es
, 1);
2095 journal_unlock_updates(journal
);
2099 * If we are mounting (or read-write remounting) a filesystem whose journal
2100 * has recorded an error from a previous lifetime, move that error to the
2101 * main filesystem now.
2103 static void ext3_clear_journal_err(struct super_block
* sb
,
2104 struct ext3_super_block
* es
)
2110 journal
= EXT3_SB(sb
)->s_journal
;
2113 * Now check for any error status which may have been recorded in the
2114 * journal by a prior ext3_error() or ext3_abort()
2117 j_errno
= journal_errno(journal
);
2121 errstr
= ext3_decode_error(sb
, j_errno
, nbuf
);
2122 ext3_warning(sb
, __FUNCTION__
, "Filesystem error recorded "
2123 "from previous mount: %s", errstr
);
2124 ext3_warning(sb
, __FUNCTION__
, "Marking fs in need of "
2125 "filesystem check.");
2127 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
2128 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
2129 ext3_commit_super (sb
, es
, 1);
2131 journal_clear_err(journal
);
2136 * Force the running and committing transactions to commit,
2137 * and wait on the commit.
2139 int ext3_force_commit(struct super_block
*sb
)
2144 if (sb
->s_flags
& MS_RDONLY
)
2147 journal
= EXT3_SB(sb
)->s_journal
;
2149 ret
= ext3_journal_force_commit(journal
);
2154 * Ext3 always journals updates to the superblock itself, so we don't
2155 * have to propagate any other updates to the superblock on disk at this
2156 * point. Just start an async writeback to get the buffers on their way
2159 * This implicitly triggers the writebehind on sync().
2162 static void ext3_write_super (struct super_block
* sb
)
2164 if (mutex_trylock(&sb
->s_lock
) != 0)
2169 static int ext3_sync_fs(struct super_block
*sb
, int wait
)
2174 if (journal_start_commit(EXT3_SB(sb
)->s_journal
, &target
)) {
2176 log_wait_commit(EXT3_SB(sb
)->s_journal
, target
);
2182 * LVM calls this function before a (read-only) snapshot is created. This
2183 * gives us a chance to flush the journal completely and mark the fs clean.
2185 static void ext3_write_super_lockfs(struct super_block
*sb
)
2189 if (!(sb
->s_flags
& MS_RDONLY
)) {
2190 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
2192 /* Now we set up the journal barrier. */
2193 journal_lock_updates(journal
);
2194 journal_flush(journal
);
2196 /* Journal blocked and flushed, clear needs_recovery flag. */
2197 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2198 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2203 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2204 * flag here, even though the filesystem is not technically dirty yet.
2206 static void ext3_unlockfs(struct super_block
*sb
)
2208 if (!(sb
->s_flags
& MS_RDONLY
)) {
2210 /* Reser the needs_recovery flag before the fs is unlocked. */
2211 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2212 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2214 journal_unlock_updates(EXT3_SB(sb
)->s_journal
);
2218 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
)
2220 struct ext3_super_block
* es
;
2221 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2222 ext3_fsblk_t n_blocks_count
= 0;
2223 unsigned long old_sb_flags
;
2224 struct ext3_mount_options old_opts
;
2230 /* Store the original options */
2231 old_sb_flags
= sb
->s_flags
;
2232 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
2233 old_opts
.s_resuid
= sbi
->s_resuid
;
2234 old_opts
.s_resgid
= sbi
->s_resgid
;
2235 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
2237 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
2238 for (i
= 0; i
< MAXQUOTAS
; i
++)
2239 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
2243 * Allow the "check" option to be passed as a remount option.
2245 if (!parse_options(data
, sb
, NULL
, NULL
, &n_blocks_count
, 1)) {
2250 if (sbi
->s_mount_opt
& EXT3_MOUNT_ABORT
)
2251 ext3_abort(sb
, __FUNCTION__
, "Abort forced by user");
2253 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2254 ((sbi
->s_mount_opt
& EXT3_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2258 ext3_init_journal_params(sb
, sbi
->s_journal
);
2260 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
2261 n_blocks_count
> le32_to_cpu(es
->s_blocks_count
)) {
2262 if (sbi
->s_mount_opt
& EXT3_MOUNT_ABORT
) {
2267 if (*flags
& MS_RDONLY
) {
2269 * First of all, the unconditional stuff we have to do
2270 * to disable replay of the journal when we next remount
2272 sb
->s_flags
|= MS_RDONLY
;
2275 * OK, test if we are remounting a valid rw partition
2276 * readonly, and if so set the rdonly flag and then
2277 * mark the partition as valid again.
2279 if (!(es
->s_state
& cpu_to_le16(EXT3_VALID_FS
)) &&
2280 (sbi
->s_mount_state
& EXT3_VALID_FS
))
2281 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
2283 ext3_mark_recovery_complete(sb
, es
);
2286 if ((ret
= EXT3_HAS_RO_COMPAT_FEATURE(sb
,
2287 ~EXT3_FEATURE_RO_COMPAT_SUPP
))) {
2288 printk(KERN_WARNING
"EXT3-fs: %s: couldn't "
2289 "remount RDWR because of unsupported "
2290 "optional features (%x).\n",
2291 sb
->s_id
, le32_to_cpu(ret
));
2296 * Mounting a RDONLY partition read-write, so reread
2297 * and store the current valid flag. (It may have
2298 * been changed by e2fsck since we originally mounted
2301 ext3_clear_journal_err(sb
, es
);
2302 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2303 if ((ret
= ext3_group_extend(sb
, es
, n_blocks_count
))) {
2307 if (!ext3_setup_super (sb
, es
, 0))
2308 sb
->s_flags
&= ~MS_RDONLY
;
2312 /* Release old quota file names */
2313 for (i
= 0; i
< MAXQUOTAS
; i
++)
2314 if (old_opts
.s_qf_names
[i
] &&
2315 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2316 kfree(old_opts
.s_qf_names
[i
]);
2320 sb
->s_flags
= old_sb_flags
;
2321 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
2322 sbi
->s_resuid
= old_opts
.s_resuid
;
2323 sbi
->s_resgid
= old_opts
.s_resgid
;
2324 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
2326 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
2327 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2328 if (sbi
->s_qf_names
[i
] &&
2329 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2330 kfree(sbi
->s_qf_names
[i
]);
2331 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
2337 static int ext3_statfs (struct dentry
* dentry
, struct kstatfs
* buf
)
2339 struct super_block
*sb
= dentry
->d_sb
;
2340 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2341 struct ext3_super_block
*es
= sbi
->s_es
;
2342 ext3_fsblk_t overhead
;
2345 if (test_opt (sb
, MINIX_DF
))
2348 unsigned long ngroups
;
2349 ngroups
= EXT3_SB(sb
)->s_groups_count
;
2353 * Compute the overhead (FS structures)
2357 * All of the blocks before first_data_block are
2360 overhead
= le32_to_cpu(es
->s_first_data_block
);
2363 * Add the overhead attributed to the superblock and
2364 * block group descriptors. If the sparse superblocks
2365 * feature is turned on, then not all groups have this.
2367 for (i
= 0; i
< ngroups
; i
++) {
2368 overhead
+= ext3_bg_has_super(sb
, i
) +
2369 ext3_bg_num_gdb(sb
, i
);
2374 * Every block group has an inode bitmap, a block
2375 * bitmap, and an inode table.
2377 overhead
+= (ngroups
* (2 + EXT3_SB(sb
)->s_itb_per_group
));
2380 buf
->f_type
= EXT3_SUPER_MAGIC
;
2381 buf
->f_bsize
= sb
->s_blocksize
;
2382 buf
->f_blocks
= le32_to_cpu(es
->s_blocks_count
) - overhead
;
2383 buf
->f_bfree
= percpu_counter_sum(&sbi
->s_freeblocks_counter
);
2384 buf
->f_bavail
= buf
->f_bfree
- le32_to_cpu(es
->s_r_blocks_count
);
2385 if (buf
->f_bfree
< le32_to_cpu(es
->s_r_blocks_count
))
2387 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
2388 buf
->f_ffree
= percpu_counter_sum(&sbi
->s_freeinodes_counter
);
2389 buf
->f_namelen
= EXT3_NAME_LEN
;
2393 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2394 * is locked for write. Otherwise the are possible deadlocks:
2395 * Process 1 Process 2
2396 * ext3_create() quota_sync()
2397 * journal_start() write_dquot()
2398 * DQUOT_INIT() down(dqio_mutex)
2399 * down(dqio_mutex) journal_start()
2405 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
2407 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
2410 static int ext3_dquot_initialize(struct inode
*inode
, int type
)
2415 /* We may create quota structure so we need to reserve enough blocks */
2416 handle
= ext3_journal_start(inode
, 2*EXT3_QUOTA_INIT_BLOCKS(inode
->i_sb
));
2418 return PTR_ERR(handle
);
2419 ret
= dquot_initialize(inode
, type
);
2420 err
= ext3_journal_stop(handle
);
2426 static int ext3_dquot_drop(struct inode
*inode
)
2431 /* We may delete quota structure so we need to reserve enough blocks */
2432 handle
= ext3_journal_start(inode
, 2*EXT3_QUOTA_DEL_BLOCKS(inode
->i_sb
));
2434 return PTR_ERR(handle
);
2435 ret
= dquot_drop(inode
);
2436 err
= ext3_journal_stop(handle
);
2442 static int ext3_write_dquot(struct dquot
*dquot
)
2446 struct inode
*inode
;
2448 inode
= dquot_to_inode(dquot
);
2449 handle
= ext3_journal_start(inode
,
2450 EXT3_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
2452 return PTR_ERR(handle
);
2453 ret
= dquot_commit(dquot
);
2454 err
= ext3_journal_stop(handle
);
2460 static int ext3_acquire_dquot(struct dquot
*dquot
)
2465 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2466 EXT3_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
2468 return PTR_ERR(handle
);
2469 ret
= dquot_acquire(dquot
);
2470 err
= ext3_journal_stop(handle
);
2476 static int ext3_release_dquot(struct dquot
*dquot
)
2481 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2482 EXT3_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
2484 return PTR_ERR(handle
);
2485 ret
= dquot_release(dquot
);
2486 err
= ext3_journal_stop(handle
);
2492 static int ext3_mark_dquot_dirty(struct dquot
*dquot
)
2494 /* Are we journalling quotas? */
2495 if (EXT3_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
2496 EXT3_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
2497 dquot_mark_dquot_dirty(dquot
);
2498 return ext3_write_dquot(dquot
);
2500 return dquot_mark_dquot_dirty(dquot
);
2504 static int ext3_write_info(struct super_block
*sb
, int type
)
2509 /* Data block + inode block */
2510 handle
= ext3_journal_start(sb
->s_root
->d_inode
, 2);
2512 return PTR_ERR(handle
);
2513 ret
= dquot_commit_info(sb
, type
);
2514 err
= ext3_journal_stop(handle
);
2521 * Turn on quotas during mount time - we need to find
2522 * the quota file and such...
2524 static int ext3_quota_on_mount(struct super_block
*sb
, int type
)
2526 return vfs_quota_on_mount(sb
, EXT3_SB(sb
)->s_qf_names
[type
],
2527 EXT3_SB(sb
)->s_jquota_fmt
, type
);
2531 * Standard function to be called on quota_on
2533 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
,
2537 struct nameidata nd
;
2539 if (!test_opt(sb
, QUOTA
))
2541 /* Not journalling quota? */
2542 if (!EXT3_SB(sb
)->s_qf_names
[USRQUOTA
] &&
2543 !EXT3_SB(sb
)->s_qf_names
[GRPQUOTA
])
2544 return vfs_quota_on(sb
, type
, format_id
, path
);
2545 err
= path_lookup(path
, LOOKUP_FOLLOW
, &nd
);
2548 /* Quotafile not on the same filesystem? */
2549 if (nd
.mnt
->mnt_sb
!= sb
) {
2553 /* Quotafile not of fs root? */
2554 if (nd
.dentry
->d_parent
->d_inode
!= sb
->s_root
->d_inode
)
2556 "EXT3-fs: Quota file not on filesystem root. "
2557 "Journalled quota will not work.\n");
2559 return vfs_quota_on(sb
, type
, format_id
, path
);
2562 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2563 * acquiring the locks... As quota files are never truncated and quota code
2564 * itself serializes the operations (and noone else should touch the files)
2565 * we don't have to be afraid of races */
2566 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
2567 size_t len
, loff_t off
)
2569 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2570 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2572 int offset
= off
& (sb
->s_blocksize
- 1);
2575 struct buffer_head
*bh
;
2576 loff_t i_size
= i_size_read(inode
);
2580 if (off
+len
> i_size
)
2583 while (toread
> 0) {
2584 tocopy
= sb
->s_blocksize
- offset
< toread
?
2585 sb
->s_blocksize
- offset
: toread
;
2586 bh
= ext3_bread(NULL
, inode
, blk
, 0, &err
);
2589 if (!bh
) /* A hole? */
2590 memset(data
, 0, tocopy
);
2592 memcpy(data
, bh
->b_data
+offset
, tocopy
);
2602 /* Write to quotafile (we know the transaction is already started and has
2603 * enough credits) */
2604 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
2605 const char *data
, size_t len
, loff_t off
)
2607 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2608 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2610 int offset
= off
& (sb
->s_blocksize
- 1);
2612 int journal_quota
= EXT3_SB(sb
)->s_qf_names
[type
] != NULL
;
2613 size_t towrite
= len
;
2614 struct buffer_head
*bh
;
2615 handle_t
*handle
= journal_current_handle();
2617 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
2618 while (towrite
> 0) {
2619 tocopy
= sb
->s_blocksize
- offset
< towrite
?
2620 sb
->s_blocksize
- offset
: towrite
;
2621 bh
= ext3_bread(handle
, inode
, blk
, 1, &err
);
2624 if (journal_quota
) {
2625 err
= ext3_journal_get_write_access(handle
, bh
);
2632 memcpy(bh
->b_data
+offset
, data
, tocopy
);
2633 flush_dcache_page(bh
->b_page
);
2636 err
= ext3_journal_dirty_metadata(handle
, bh
);
2638 /* Always do at least ordered writes for quotas */
2639 err
= ext3_journal_dirty_data(handle
, bh
);
2640 mark_buffer_dirty(bh
);
2653 if (inode
->i_size
< off
+len
-towrite
) {
2654 i_size_write(inode
, off
+len
-towrite
);
2655 EXT3_I(inode
)->i_disksize
= inode
->i_size
;
2658 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2659 ext3_mark_inode_dirty(handle
, inode
);
2660 mutex_unlock(&inode
->i_mutex
);
2661 return len
- towrite
;
2666 static int ext3_get_sb(struct file_system_type
*fs_type
,
2667 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
2669 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext3_fill_super
, mnt
);
2672 static struct file_system_type ext3_fs_type
= {
2673 .owner
= THIS_MODULE
,
2675 .get_sb
= ext3_get_sb
,
2676 .kill_sb
= kill_block_super
,
2677 .fs_flags
= FS_REQUIRES_DEV
,
2680 static int __init
init_ext3_fs(void)
2682 int err
= init_ext3_xattr();
2685 err
= init_inodecache();
2688 err
= register_filesystem(&ext3_fs_type
);
2693 destroy_inodecache();
2699 static void __exit
exit_ext3_fs(void)
2701 unregister_filesystem(&ext3_fs_type
);
2702 destroy_inodecache();
2706 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2707 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2708 MODULE_LICENSE("GPL");
2709 module_init(init_ext3_fs
)
2710 module_exit(exit_ext3_fs
)