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 <linux/seq_file.h>
40 #include <asm/uaccess.h>
46 static int ext3_load_journal(struct super_block
*, struct ext3_super_block
*,
47 unsigned long journal_devnum
);
48 static int ext3_create_journal(struct super_block
*, struct ext3_super_block
*,
50 static void ext3_commit_super (struct super_block
* sb
,
51 struct ext3_super_block
* es
,
53 static void ext3_mark_recovery_complete(struct super_block
* sb
,
54 struct ext3_super_block
* es
);
55 static void ext3_clear_journal_err(struct super_block
* sb
,
56 struct ext3_super_block
* es
);
57 static int ext3_sync_fs(struct super_block
*sb
, int wait
);
58 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
60 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
);
61 static int ext3_statfs (struct dentry
* dentry
, struct kstatfs
* buf
);
62 static void ext3_unlockfs(struct super_block
*sb
);
63 static void ext3_write_super (struct super_block
* sb
);
64 static void ext3_write_super_lockfs(struct super_block
*sb
);
67 * Wrappers for journal_start/end.
69 * The only special thing we need to do here is to make sure that all
70 * journal_end calls result in the superblock being marked dirty, so
71 * that sync() will call the filesystem's write_super callback if
74 handle_t
*ext3_journal_start_sb(struct super_block
*sb
, int nblocks
)
78 if (sb
->s_flags
& MS_RDONLY
)
79 return ERR_PTR(-EROFS
);
81 /* Special case here: if the journal has aborted behind our
82 * backs (eg. EIO in the commit thread), then we still need to
83 * take the FS itself readonly cleanly. */
84 journal
= EXT3_SB(sb
)->s_journal
;
85 if (is_journal_aborted(journal
)) {
86 ext3_abort(sb
, __FUNCTION__
,
87 "Detected aborted journal");
88 return ERR_PTR(-EROFS
);
91 return journal_start(journal
, nblocks
);
95 * The only special thing we need to do here is to make sure that all
96 * journal_stop calls result in the superblock being marked dirty, so
97 * that sync() will call the filesystem's write_super callback if
100 int __ext3_journal_stop(const char *where
, handle_t
*handle
)
102 struct super_block
*sb
;
106 sb
= handle
->h_transaction
->t_journal
->j_private
;
108 rc
= journal_stop(handle
);
113 __ext3_std_error(sb
, where
, err
);
117 void ext3_journal_abort_handle(const char *caller
, const char *err_fn
,
118 struct buffer_head
*bh
, handle_t
*handle
, int err
)
121 const char *errstr
= ext3_decode_error(NULL
, err
, nbuf
);
124 BUFFER_TRACE(bh
, "abort");
129 if (is_handle_aborted(handle
))
132 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
133 caller
, errstr
, err_fn
);
135 journal_abort_handle(handle
);
138 /* Deal with the reporting of failure conditions on a filesystem such as
139 * inconsistencies detected or read IO failures.
141 * On ext2, we can store the error state of the filesystem in the
142 * superblock. That is not possible on ext3, because we may have other
143 * write ordering constraints on the superblock which prevent us from
144 * writing it out straight away; and given that the journal is about to
145 * be aborted, we can't rely on the current, or future, transactions to
146 * write out the superblock safely.
148 * We'll just use the journal_abort() error code to record an error in
149 * the journal instead. On recovery, the journal will compain about
150 * that error until we've noted it down and cleared it.
153 static void ext3_handle_error(struct super_block
*sb
)
155 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
157 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
158 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
160 if (sb
->s_flags
& MS_RDONLY
)
163 if (test_opt (sb
, ERRORS_RO
)) {
164 printk (KERN_CRIT
"Remounting filesystem read-only\n");
165 sb
->s_flags
|= MS_RDONLY
;
167 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
169 EXT3_SB(sb
)->s_mount_opt
|= EXT3_MOUNT_ABORT
;
171 journal_abort(journal
, -EIO
);
173 if (test_opt(sb
, ERRORS_PANIC
))
174 panic("EXT3-fs (device %s): panic forced after error\n",
176 ext3_commit_super(sb
, es
, 1);
179 void ext3_error (struct super_block
* sb
, const char * function
,
180 const char * fmt
, ...)
185 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
190 ext3_handle_error(sb
);
193 static const char *ext3_decode_error(struct super_block
* sb
, int errno
,
200 errstr
= "IO failure";
203 errstr
= "Out of memory";
206 if (!sb
|| EXT3_SB(sb
)->s_journal
->j_flags
& JFS_ABORT
)
207 errstr
= "Journal has aborted";
209 errstr
= "Readonly filesystem";
212 /* If the caller passed in an extra buffer for unknown
213 * errors, textualise them now. Else we just return
216 /* Check for truncated error codes... */
217 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
226 /* __ext3_std_error decodes expected errors from journaling functions
227 * automatically and invokes the appropriate error response. */
229 void __ext3_std_error (struct super_block
* sb
, const char * function
,
235 /* Special case: if the error is EROFS, and we're not already
236 * inside a transaction, then there's really no point in logging
238 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
239 (sb
->s_flags
& MS_RDONLY
))
242 errstr
= ext3_decode_error(sb
, errno
, nbuf
);
243 printk (KERN_CRIT
"EXT3-fs error (device %s) in %s: %s\n",
244 sb
->s_id
, function
, errstr
);
246 ext3_handle_error(sb
);
250 * ext3_abort is a much stronger failure handler than ext3_error. The
251 * abort function may be used to deal with unrecoverable failures such
252 * as journal IO errors or ENOMEM at a critical moment in log management.
254 * We unconditionally force the filesystem into an ABORT|READONLY state,
255 * unless the error response on the fs has been set to panic in which
256 * case we take the easy way out and panic immediately.
259 void ext3_abort (struct super_block
* sb
, const char * function
,
260 const char * fmt
, ...)
264 printk (KERN_CRIT
"ext3_abort called.\n");
267 printk(KERN_CRIT
"EXT3-fs error (device %s): %s: ",sb
->s_id
, function
);
272 if (test_opt(sb
, ERRORS_PANIC
))
273 panic("EXT3-fs panic from previous error\n");
275 if (sb
->s_flags
& MS_RDONLY
)
278 printk(KERN_CRIT
"Remounting filesystem read-only\n");
279 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
280 sb
->s_flags
|= MS_RDONLY
;
281 EXT3_SB(sb
)->s_mount_opt
|= EXT3_MOUNT_ABORT
;
282 journal_abort(EXT3_SB(sb
)->s_journal
, -EIO
);
285 void ext3_warning (struct super_block
* sb
, const char * function
,
286 const char * fmt
, ...)
291 printk(KERN_WARNING
"EXT3-fs warning (device %s): %s: ",
298 void ext3_update_dynamic_rev(struct super_block
*sb
)
300 struct ext3_super_block
*es
= EXT3_SB(sb
)->s_es
;
302 if (le32_to_cpu(es
->s_rev_level
) > EXT3_GOOD_OLD_REV
)
305 ext3_warning(sb
, __FUNCTION__
,
306 "updating to rev %d because of new feature flag, "
307 "running e2fsck is recommended",
310 es
->s_first_ino
= cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO
);
311 es
->s_inode_size
= cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE
);
312 es
->s_rev_level
= cpu_to_le32(EXT3_DYNAMIC_REV
);
313 /* leave es->s_feature_*compat flags alone */
314 /* es->s_uuid will be set by e2fsck if empty */
317 * The rest of the superblock fields should be zero, and if not it
318 * means they are likely already in use, so leave them alone. We
319 * can leave it up to e2fsck to clean up any inconsistencies there.
324 * Open the external journal device
326 static struct block_device
*ext3_blkdev_get(dev_t dev
)
328 struct block_device
*bdev
;
329 char b
[BDEVNAME_SIZE
];
331 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
337 printk(KERN_ERR
"EXT3: failed to open journal device %s: %ld\n",
338 __bdevname(dev
, b
), PTR_ERR(bdev
));
343 * Release the journal device
345 static int ext3_blkdev_put(struct block_device
*bdev
)
348 return blkdev_put(bdev
);
351 static int ext3_blkdev_remove(struct ext3_sb_info
*sbi
)
353 struct block_device
*bdev
;
356 bdev
= sbi
->journal_bdev
;
358 ret
= ext3_blkdev_put(bdev
);
359 sbi
->journal_bdev
= NULL
;
364 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
366 return &list_entry(l
, struct ext3_inode_info
, i_orphan
)->vfs_inode
;
369 static void dump_orphan_list(struct super_block
*sb
, struct ext3_sb_info
*sbi
)
373 printk(KERN_ERR
"sb orphan head is %d\n",
374 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
376 printk(KERN_ERR
"sb_info orphan list:\n");
377 list_for_each(l
, &sbi
->s_orphan
) {
378 struct inode
*inode
= orphan_list_entry(l
);
380 "inode %s:%ld at %p: mode %o, nlink %d, next %d\n",
381 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
382 inode
->i_mode
, inode
->i_nlink
,
387 static void ext3_put_super (struct super_block
* sb
)
389 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
390 struct ext3_super_block
*es
= sbi
->s_es
;
393 ext3_xattr_put_super(sb
);
394 journal_destroy(sbi
->s_journal
);
395 if (!(sb
->s_flags
& MS_RDONLY
)) {
396 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
397 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
398 BUFFER_TRACE(sbi
->s_sbh
, "marking dirty");
399 mark_buffer_dirty(sbi
->s_sbh
);
400 ext3_commit_super(sb
, es
, 1);
403 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
404 brelse(sbi
->s_group_desc
[i
]);
405 kfree(sbi
->s_group_desc
);
406 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
407 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
408 percpu_counter_destroy(&sbi
->s_dirs_counter
);
411 for (i
= 0; i
< MAXQUOTAS
; i
++)
412 kfree(sbi
->s_qf_names
[i
]);
415 /* Debugging code just in case the in-memory inode orphan list
416 * isn't empty. The on-disk one can be non-empty if we've
417 * detected an error and taken the fs readonly, but the
418 * in-memory list had better be clean by this point. */
419 if (!list_empty(&sbi
->s_orphan
))
420 dump_orphan_list(sb
, sbi
);
421 J_ASSERT(list_empty(&sbi
->s_orphan
));
423 invalidate_bdev(sb
->s_bdev
, 0);
424 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
426 * Invalidate the journal device's buffers. We don't want them
427 * floating about in memory - the physical journal device may
428 * hotswapped, and it breaks the `ro-after' testing code.
430 sync_blockdev(sbi
->journal_bdev
);
431 invalidate_bdev(sbi
->journal_bdev
, 0);
432 ext3_blkdev_remove(sbi
);
434 sb
->s_fs_info
= NULL
;
439 static kmem_cache_t
*ext3_inode_cachep
;
442 * Called inside transaction, so use GFP_NOFS
444 static struct inode
*ext3_alloc_inode(struct super_block
*sb
)
446 struct ext3_inode_info
*ei
;
448 ei
= kmem_cache_alloc(ext3_inode_cachep
, SLAB_NOFS
);
451 #ifdef CONFIG_EXT3_FS_POSIX_ACL
452 ei
->i_acl
= EXT3_ACL_NOT_CACHED
;
453 ei
->i_default_acl
= EXT3_ACL_NOT_CACHED
;
455 ei
->i_block_alloc_info
= NULL
;
456 ei
->vfs_inode
.i_version
= 1;
457 return &ei
->vfs_inode
;
460 static void ext3_destroy_inode(struct inode
*inode
)
462 kmem_cache_free(ext3_inode_cachep
, EXT3_I(inode
));
465 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
467 struct ext3_inode_info
*ei
= (struct ext3_inode_info
*) foo
;
469 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
470 SLAB_CTOR_CONSTRUCTOR
) {
471 INIT_LIST_HEAD(&ei
->i_orphan
);
472 #ifdef CONFIG_EXT3_FS_XATTR
473 init_rwsem(&ei
->xattr_sem
);
475 mutex_init(&ei
->truncate_mutex
);
476 inode_init_once(&ei
->vfs_inode
);
480 static int init_inodecache(void)
482 ext3_inode_cachep
= kmem_cache_create("ext3_inode_cache",
483 sizeof(struct ext3_inode_info
),
484 0, (SLAB_RECLAIM_ACCOUNT
|
487 if (ext3_inode_cachep
== NULL
)
492 static void destroy_inodecache(void)
494 if (kmem_cache_destroy(ext3_inode_cachep
))
495 printk(KERN_INFO
"ext3_inode_cache: not all structures were freed\n");
498 static void ext3_clear_inode(struct inode
*inode
)
500 struct ext3_block_alloc_info
*rsv
= EXT3_I(inode
)->i_block_alloc_info
;
501 #ifdef CONFIG_EXT3_FS_POSIX_ACL
502 if (EXT3_I(inode
)->i_acl
&&
503 EXT3_I(inode
)->i_acl
!= EXT3_ACL_NOT_CACHED
) {
504 posix_acl_release(EXT3_I(inode
)->i_acl
);
505 EXT3_I(inode
)->i_acl
= EXT3_ACL_NOT_CACHED
;
507 if (EXT3_I(inode
)->i_default_acl
&&
508 EXT3_I(inode
)->i_default_acl
!= EXT3_ACL_NOT_CACHED
) {
509 posix_acl_release(EXT3_I(inode
)->i_default_acl
);
510 EXT3_I(inode
)->i_default_acl
= EXT3_ACL_NOT_CACHED
;
513 ext3_discard_reservation(inode
);
514 EXT3_I(inode
)->i_block_alloc_info
= NULL
;
519 static inline void ext3_show_quota_options(struct seq_file
*seq
, struct super_block
*sb
)
521 #if defined(CONFIG_QUOTA)
522 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
524 if (sbi
->s_jquota_fmt
)
525 seq_printf(seq
, ",jqfmt=%s",
526 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold": "vfsv0");
528 if (sbi
->s_qf_names
[USRQUOTA
])
529 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
531 if (sbi
->s_qf_names
[GRPQUOTA
])
532 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
534 if (sbi
->s_mount_opt
& EXT3_MOUNT_USRQUOTA
)
535 seq_puts(seq
, ",usrquota");
537 if (sbi
->s_mount_opt
& EXT3_MOUNT_GRPQUOTA
)
538 seq_puts(seq
, ",grpquota");
542 static int ext3_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
544 struct super_block
*sb
= vfs
->mnt_sb
;
546 if (test_opt(sb
, DATA_FLAGS
) == EXT3_MOUNT_JOURNAL_DATA
)
547 seq_puts(seq
, ",data=journal");
548 else if (test_opt(sb
, DATA_FLAGS
) == EXT3_MOUNT_ORDERED_DATA
)
549 seq_puts(seq
, ",data=ordered");
550 else if (test_opt(sb
, DATA_FLAGS
) == EXT3_MOUNT_WRITEBACK_DATA
)
551 seq_puts(seq
, ",data=writeback");
553 ext3_show_quota_options(seq
, sb
);
559 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
560 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
562 static int ext3_dquot_initialize(struct inode
*inode
, int type
);
563 static int ext3_dquot_drop(struct inode
*inode
);
564 static int ext3_write_dquot(struct dquot
*dquot
);
565 static int ext3_acquire_dquot(struct dquot
*dquot
);
566 static int ext3_release_dquot(struct dquot
*dquot
);
567 static int ext3_mark_dquot_dirty(struct dquot
*dquot
);
568 static int ext3_write_info(struct super_block
*sb
, int type
);
569 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
, char *path
);
570 static int ext3_quota_on_mount(struct super_block
*sb
, int type
);
571 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
572 size_t len
, loff_t off
);
573 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
574 const char *data
, size_t len
, loff_t off
);
576 static struct dquot_operations ext3_quota_operations
= {
577 .initialize
= ext3_dquot_initialize
,
578 .drop
= ext3_dquot_drop
,
579 .alloc_space
= dquot_alloc_space
,
580 .alloc_inode
= dquot_alloc_inode
,
581 .free_space
= dquot_free_space
,
582 .free_inode
= dquot_free_inode
,
583 .transfer
= dquot_transfer
,
584 .write_dquot
= ext3_write_dquot
,
585 .acquire_dquot
= ext3_acquire_dquot
,
586 .release_dquot
= ext3_release_dquot
,
587 .mark_dirty
= ext3_mark_dquot_dirty
,
588 .write_info
= ext3_write_info
591 static struct quotactl_ops ext3_qctl_operations
= {
592 .quota_on
= ext3_quota_on
,
593 .quota_off
= vfs_quota_off
,
594 .quota_sync
= vfs_quota_sync
,
595 .get_info
= vfs_get_dqinfo
,
596 .set_info
= vfs_set_dqinfo
,
597 .get_dqblk
= vfs_get_dqblk
,
598 .set_dqblk
= vfs_set_dqblk
602 static struct super_operations ext3_sops
= {
603 .alloc_inode
= ext3_alloc_inode
,
604 .destroy_inode
= ext3_destroy_inode
,
605 .read_inode
= ext3_read_inode
,
606 .write_inode
= ext3_write_inode
,
607 .dirty_inode
= ext3_dirty_inode
,
608 .delete_inode
= ext3_delete_inode
,
609 .put_super
= ext3_put_super
,
610 .write_super
= ext3_write_super
,
611 .sync_fs
= ext3_sync_fs
,
612 .write_super_lockfs
= ext3_write_super_lockfs
,
613 .unlockfs
= ext3_unlockfs
,
614 .statfs
= ext3_statfs
,
615 .remount_fs
= ext3_remount
,
616 .clear_inode
= ext3_clear_inode
,
617 .show_options
= ext3_show_options
,
619 .quota_read
= ext3_quota_read
,
620 .quota_write
= ext3_quota_write
,
624 static struct export_operations ext3_export_ops
= {
625 .get_parent
= ext3_get_parent
,
629 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
630 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
631 Opt_nouid32
, Opt_nocheck
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
632 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
633 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
634 Opt_commit
, Opt_journal_update
, Opt_journal_inum
, Opt_journal_dev
,
635 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
636 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
637 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
638 Opt_ignore
, Opt_barrier
, Opt_err
, Opt_resize
, Opt_usrquota
,
642 static match_table_t tokens
= {
643 {Opt_bsd_df
, "bsddf"},
644 {Opt_minix_df
, "minixdf"},
645 {Opt_grpid
, "grpid"},
646 {Opt_grpid
, "bsdgroups"},
647 {Opt_nogrpid
, "nogrpid"},
648 {Opt_nogrpid
, "sysvgroups"},
649 {Opt_resgid
, "resgid=%u"},
650 {Opt_resuid
, "resuid=%u"},
652 {Opt_err_cont
, "errors=continue"},
653 {Opt_err_panic
, "errors=panic"},
654 {Opt_err_ro
, "errors=remount-ro"},
655 {Opt_nouid32
, "nouid32"},
656 {Opt_nocheck
, "nocheck"},
657 {Opt_nocheck
, "check=none"},
658 {Opt_debug
, "debug"},
659 {Opt_oldalloc
, "oldalloc"},
660 {Opt_orlov
, "orlov"},
661 {Opt_user_xattr
, "user_xattr"},
662 {Opt_nouser_xattr
, "nouser_xattr"},
664 {Opt_noacl
, "noacl"},
665 {Opt_reservation
, "reservation"},
666 {Opt_noreservation
, "noreservation"},
667 {Opt_noload
, "noload"},
670 {Opt_commit
, "commit=%u"},
671 {Opt_journal_update
, "journal=update"},
672 {Opt_journal_inum
, "journal=%u"},
673 {Opt_journal_dev
, "journal_dev=%u"},
674 {Opt_abort
, "abort"},
675 {Opt_data_journal
, "data=journal"},
676 {Opt_data_ordered
, "data=ordered"},
677 {Opt_data_writeback
, "data=writeback"},
678 {Opt_offusrjquota
, "usrjquota="},
679 {Opt_usrjquota
, "usrjquota=%s"},
680 {Opt_offgrpjquota
, "grpjquota="},
681 {Opt_grpjquota
, "grpjquota=%s"},
682 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
683 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
684 {Opt_grpquota
, "grpquota"},
685 {Opt_noquota
, "noquota"},
686 {Opt_quota
, "quota"},
687 {Opt_usrquota
, "usrquota"},
688 {Opt_barrier
, "barrier=%u"},
690 {Opt_resize
, "resize"},
693 static ext3_fsblk_t
get_sb_block(void **data
)
695 ext3_fsblk_t sb_block
;
696 char *options
= (char *) *data
;
698 if (!options
|| strncmp(options
, "sb=", 3) != 0)
699 return 1; /* Default location */
701 /*todo: use simple_strtoll with >32bit ext3 */
702 sb_block
= simple_strtoul(options
, &options
, 0);
703 if (*options
&& *options
!= ',') {
704 printk("EXT3-fs: Invalid sb specification: %s\n",
710 *data
= (void *) options
;
714 static int parse_options (char *options
, struct super_block
*sb
,
715 unsigned long *inum
, unsigned long *journal_devnum
,
716 ext3_fsblk_t
*n_blocks_count
, int is_remount
)
718 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
720 substring_t args
[MAX_OPT_ARGS
];
731 while ((p
= strsep (&options
, ",")) != NULL
) {
736 token
= match_token(p
, tokens
, args
);
739 clear_opt (sbi
->s_mount_opt
, MINIX_DF
);
742 set_opt (sbi
->s_mount_opt
, MINIX_DF
);
745 set_opt (sbi
->s_mount_opt
, GRPID
);
748 clear_opt (sbi
->s_mount_opt
, GRPID
);
751 if (match_int(&args
[0], &option
))
753 sbi
->s_resuid
= option
;
756 if (match_int(&args
[0], &option
))
758 sbi
->s_resgid
= option
;
761 /* handled by get_sb_block() instead of here */
762 /* *sb_block = match_int(&args[0]); */
765 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
766 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
767 set_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
770 clear_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
771 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
772 set_opt (sbi
->s_mount_opt
, ERRORS_RO
);
775 clear_opt (sbi
->s_mount_opt
, ERRORS_RO
);
776 clear_opt (sbi
->s_mount_opt
, ERRORS_PANIC
);
777 set_opt (sbi
->s_mount_opt
, ERRORS_CONT
);
780 set_opt (sbi
->s_mount_opt
, NO_UID32
);
783 clear_opt (sbi
->s_mount_opt
, CHECK
);
786 set_opt (sbi
->s_mount_opt
, DEBUG
);
789 set_opt (sbi
->s_mount_opt
, OLDALLOC
);
792 clear_opt (sbi
->s_mount_opt
, OLDALLOC
);
794 #ifdef CONFIG_EXT3_FS_XATTR
796 set_opt (sbi
->s_mount_opt
, XATTR_USER
);
798 case Opt_nouser_xattr
:
799 clear_opt (sbi
->s_mount_opt
, XATTR_USER
);
803 case Opt_nouser_xattr
:
804 printk("EXT3 (no)user_xattr options not supported\n");
807 #ifdef CONFIG_EXT3_FS_POSIX_ACL
809 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
812 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
817 printk("EXT3 (no)acl options not supported\n");
820 case Opt_reservation
:
821 set_opt(sbi
->s_mount_opt
, RESERVATION
);
823 case Opt_noreservation
:
824 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
826 case Opt_journal_update
:
828 /* Eventually we will want to be able to create
829 a journal file here. For now, only allow the
830 user to specify an existing inode to be the
833 printk(KERN_ERR
"EXT3-fs: cannot specify "
834 "journal on remount\n");
837 set_opt (sbi
->s_mount_opt
, UPDATE_JOURNAL
);
839 case Opt_journal_inum
:
841 printk(KERN_ERR
"EXT3-fs: cannot specify "
842 "journal on remount\n");
845 if (match_int(&args
[0], &option
))
849 case Opt_journal_dev
:
851 printk(KERN_ERR
"EXT3-fs: cannot specify "
852 "journal on remount\n");
855 if (match_int(&args
[0], &option
))
857 *journal_devnum
= option
;
860 set_opt (sbi
->s_mount_opt
, NOLOAD
);
863 if (match_int(&args
[0], &option
))
868 option
= JBD_DEFAULT_MAX_COMMIT_AGE
;
869 sbi
->s_commit_interval
= HZ
* option
;
871 case Opt_data_journal
:
872 data_opt
= EXT3_MOUNT_JOURNAL_DATA
;
874 case Opt_data_ordered
:
875 data_opt
= EXT3_MOUNT_ORDERED_DATA
;
877 case Opt_data_writeback
:
878 data_opt
= EXT3_MOUNT_WRITEBACK_DATA
;
881 if ((sbi
->s_mount_opt
& EXT3_MOUNT_DATA_FLAGS
)
884 "EXT3-fs: cannot change data "
885 "mode on remount\n");
889 sbi
->s_mount_opt
&= ~EXT3_MOUNT_DATA_FLAGS
;
890 sbi
->s_mount_opt
|= data_opt
;
900 if (sb_any_quota_enabled(sb
)) {
902 "EXT3-fs: Cannot change journalled "
903 "quota options when quota turned on.\n");
906 qname
= match_strdup(&args
[0]);
909 "EXT3-fs: not enough memory for "
910 "storing quotafile name.\n");
913 if (sbi
->s_qf_names
[qtype
] &&
914 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
916 "EXT3-fs: %s quota file already "
917 "specified.\n", QTYPE2NAME(qtype
));
921 sbi
->s_qf_names
[qtype
] = qname
;
922 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
924 "EXT3-fs: quotafile must be on "
925 "filesystem root.\n");
926 kfree(sbi
->s_qf_names
[qtype
]);
927 sbi
->s_qf_names
[qtype
] = NULL
;
930 set_opt(sbi
->s_mount_opt
, QUOTA
);
932 case Opt_offusrjquota
:
935 case Opt_offgrpjquota
:
938 if (sb_any_quota_enabled(sb
)) {
939 printk(KERN_ERR
"EXT3-fs: Cannot change "
940 "journalled quota options when "
941 "quota turned on.\n");
945 * The space will be released later when all options
946 * are confirmed to be correct
948 sbi
->s_qf_names
[qtype
] = NULL
;
950 case Opt_jqfmt_vfsold
:
951 sbi
->s_jquota_fmt
= QFMT_VFS_OLD
;
953 case Opt_jqfmt_vfsv0
:
954 sbi
->s_jquota_fmt
= QFMT_VFS_V0
;
958 set_opt(sbi
->s_mount_opt
, QUOTA
);
959 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
962 set_opt(sbi
->s_mount_opt
, QUOTA
);
963 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
966 if (sb_any_quota_enabled(sb
)) {
967 printk(KERN_ERR
"EXT3-fs: Cannot change quota "
968 "options when quota turned on.\n");
971 clear_opt(sbi
->s_mount_opt
, QUOTA
);
972 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
973 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
981 case Opt_offusrjquota
:
982 case Opt_offgrpjquota
:
983 case Opt_jqfmt_vfsold
:
984 case Opt_jqfmt_vfsv0
:
986 "EXT3-fs: journalled quota options not "
993 set_opt(sbi
->s_mount_opt
, ABORT
);
996 if (match_int(&args
[0], &option
))
999 set_opt(sbi
->s_mount_opt
, BARRIER
);
1001 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1007 printk("EXT3-fs: resize option only available "
1011 if (match_int(&args
[0], &option
) != 0)
1013 *n_blocks_count
= option
;
1016 set_opt(sbi
->s_mount_opt
, NOBH
);
1019 clear_opt(sbi
->s_mount_opt
, NOBH
);
1023 "EXT3-fs: Unrecognized mount option \"%s\" "
1024 "or missing value\n", p
);
1029 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1030 if ((sbi
->s_mount_opt
& EXT3_MOUNT_USRQUOTA
) &&
1031 sbi
->s_qf_names
[USRQUOTA
])
1032 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1034 if ((sbi
->s_mount_opt
& EXT3_MOUNT_GRPQUOTA
) &&
1035 sbi
->s_qf_names
[GRPQUOTA
])
1036 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1038 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1039 (sbi
->s_mount_opt
& EXT3_MOUNT_GRPQUOTA
)) ||
1040 (sbi
->s_qf_names
[GRPQUOTA
] &&
1041 (sbi
->s_mount_opt
& EXT3_MOUNT_USRQUOTA
))) {
1042 printk(KERN_ERR
"EXT3-fs: old and new quota "
1043 "format mixing.\n");
1047 if (!sbi
->s_jquota_fmt
) {
1048 printk(KERN_ERR
"EXT3-fs: journalled quota format "
1049 "not specified.\n");
1053 if (sbi
->s_jquota_fmt
) {
1054 printk(KERN_ERR
"EXT3-fs: journalled quota format "
1055 "specified with no journalling "
1064 static int ext3_setup_super(struct super_block
*sb
, struct ext3_super_block
*es
,
1067 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1070 if (le32_to_cpu(es
->s_rev_level
) > EXT3_MAX_SUPP_REV
) {
1071 printk (KERN_ERR
"EXT3-fs warning: revision level too high, "
1072 "forcing read-only mode\n");
1077 if (!(sbi
->s_mount_state
& EXT3_VALID_FS
))
1078 printk (KERN_WARNING
"EXT3-fs warning: mounting unchecked fs, "
1079 "running e2fsck is recommended\n");
1080 else if ((sbi
->s_mount_state
& EXT3_ERROR_FS
))
1081 printk (KERN_WARNING
1082 "EXT3-fs warning: mounting fs with errors, "
1083 "running e2fsck is recommended\n");
1084 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1085 le16_to_cpu(es
->s_mnt_count
) >=
1086 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1087 printk (KERN_WARNING
1088 "EXT3-fs warning: maximal mount count reached, "
1089 "running e2fsck is recommended\n");
1090 else if (le32_to_cpu(es
->s_checkinterval
) &&
1091 (le32_to_cpu(es
->s_lastcheck
) +
1092 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1093 printk (KERN_WARNING
1094 "EXT3-fs warning: checktime reached, "
1095 "running e2fsck is recommended\n");
1097 /* @@@ We _will_ want to clear the valid bit if we find
1098 inconsistencies, to force a fsck at reboot. But for
1099 a plain journaled filesystem we can keep it set as
1100 valid forever! :) */
1101 es
->s_state
= cpu_to_le16(le16_to_cpu(es
->s_state
) & ~EXT3_VALID_FS
);
1103 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1104 es
->s_max_mnt_count
= cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT
);
1105 es
->s_mnt_count
=cpu_to_le16(le16_to_cpu(es
->s_mnt_count
) + 1);
1106 es
->s_mtime
= cpu_to_le32(get_seconds());
1107 ext3_update_dynamic_rev(sb
);
1108 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
1110 ext3_commit_super(sb
, es
, 1);
1111 if (test_opt(sb
, DEBUG
))
1112 printk(KERN_INFO
"[EXT3 FS bs=%lu, gc=%lu, "
1113 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1115 sbi
->s_groups_count
,
1116 EXT3_BLOCKS_PER_GROUP(sb
),
1117 EXT3_INODES_PER_GROUP(sb
),
1120 printk(KERN_INFO
"EXT3 FS on %s, ", sb
->s_id
);
1121 if (EXT3_SB(sb
)->s_journal
->j_inode
== NULL
) {
1122 char b
[BDEVNAME_SIZE
];
1124 printk("external journal on %s\n",
1125 bdevname(EXT3_SB(sb
)->s_journal
->j_dev
, b
));
1127 printk("internal journal\n");
1132 /* Called at mount-time, super-block is locked */
1133 static int ext3_check_descriptors (struct super_block
* sb
)
1135 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1136 ext3_fsblk_t block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1137 struct ext3_group_desc
* gdp
= NULL
;
1141 ext3_debug ("Checking group descriptors");
1143 for (i
= 0; i
< sbi
->s_groups_count
; i
++)
1145 if ((i
% EXT3_DESC_PER_BLOCK(sb
)) == 0)
1146 gdp
= (struct ext3_group_desc
*)
1147 sbi
->s_group_desc
[desc_block
++]->b_data
;
1148 if (le32_to_cpu(gdp
->bg_block_bitmap
) < block
||
1149 le32_to_cpu(gdp
->bg_block_bitmap
) >=
1150 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1152 ext3_error (sb
, "ext3_check_descriptors",
1153 "Block bitmap for group %d"
1154 " not in group (block %lu)!",
1156 le32_to_cpu(gdp
->bg_block_bitmap
));
1159 if (le32_to_cpu(gdp
->bg_inode_bitmap
) < block
||
1160 le32_to_cpu(gdp
->bg_inode_bitmap
) >=
1161 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1163 ext3_error (sb
, "ext3_check_descriptors",
1164 "Inode bitmap for group %d"
1165 " not in group (block %lu)!",
1167 le32_to_cpu(gdp
->bg_inode_bitmap
));
1170 if (le32_to_cpu(gdp
->bg_inode_table
) < block
||
1171 le32_to_cpu(gdp
->bg_inode_table
) + sbi
->s_itb_per_group
>=
1172 block
+ EXT3_BLOCKS_PER_GROUP(sb
))
1174 ext3_error (sb
, "ext3_check_descriptors",
1175 "Inode table for group %d"
1176 " not in group (block %lu)!",
1178 le32_to_cpu(gdp
->bg_inode_table
));
1181 block
+= EXT3_BLOCKS_PER_GROUP(sb
);
1185 sbi
->s_es
->s_free_blocks_count
=cpu_to_le32(ext3_count_free_blocks(sb
));
1186 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext3_count_free_inodes(sb
));
1191 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1192 * the superblock) which were deleted from all directories, but held open by
1193 * a process at the time of a crash. We walk the list and try to delete these
1194 * inodes at recovery time (only with a read-write filesystem).
1196 * In order to keep the orphan inode chain consistent during traversal (in
1197 * case of crash during recovery), we link each inode into the superblock
1198 * orphan list_head and handle it the same way as an inode deletion during
1199 * normal operation (which journals the operations for us).
1201 * We only do an iget() and an iput() on each inode, which is very safe if we
1202 * accidentally point at an in-use or already deleted inode. The worst that
1203 * can happen in this case is that we get a "bit already cleared" message from
1204 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1205 * e2fsck was run on this filesystem, and it must have already done the orphan
1206 * inode cleanup for us, so we can safely abort without any further action.
1208 static void ext3_orphan_cleanup (struct super_block
* sb
,
1209 struct ext3_super_block
* es
)
1211 unsigned int s_flags
= sb
->s_flags
;
1212 int nr_orphans
= 0, nr_truncates
= 0;
1216 if (!es
->s_last_orphan
) {
1217 jbd_debug(4, "no orphan inodes to clean up\n");
1221 if (EXT3_SB(sb
)->s_mount_state
& EXT3_ERROR_FS
) {
1222 if (es
->s_last_orphan
)
1223 jbd_debug(1, "Errors on filesystem, "
1224 "clearing orphan list.\n");
1225 es
->s_last_orphan
= 0;
1226 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1230 if (s_flags
& MS_RDONLY
) {
1231 printk(KERN_INFO
"EXT3-fs: %s: orphan cleanup on readonly fs\n",
1233 sb
->s_flags
&= ~MS_RDONLY
;
1236 /* Needed for iput() to work correctly and not trash data */
1237 sb
->s_flags
|= MS_ACTIVE
;
1238 /* Turn on quotas so that they are updated correctly */
1239 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1240 if (EXT3_SB(sb
)->s_qf_names
[i
]) {
1241 int ret
= ext3_quota_on_mount(sb
, i
);
1244 "EXT3-fs: Cannot turn on journalled "
1245 "quota: error %d\n", ret
);
1250 while (es
->s_last_orphan
) {
1251 struct inode
*inode
;
1254 ext3_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
)))) {
1255 es
->s_last_orphan
= 0;
1259 list_add(&EXT3_I(inode
)->i_orphan
, &EXT3_SB(sb
)->s_orphan
);
1261 if (inode
->i_nlink
) {
1263 "%s: truncating inode %ld to %Ld bytes\n",
1264 __FUNCTION__
, inode
->i_ino
, inode
->i_size
);
1265 jbd_debug(2, "truncating inode %ld to %Ld bytes\n",
1266 inode
->i_ino
, inode
->i_size
);
1267 ext3_truncate(inode
);
1271 "%s: deleting unreferenced inode %ld\n",
1272 __FUNCTION__
, inode
->i_ino
);
1273 jbd_debug(2, "deleting unreferenced inode %ld\n",
1277 iput(inode
); /* The delete magic happens here! */
1280 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1283 printk(KERN_INFO
"EXT3-fs: %s: %d orphan inode%s deleted\n",
1284 sb
->s_id
, PLURAL(nr_orphans
));
1286 printk(KERN_INFO
"EXT3-fs: %s: %d truncate%s cleaned up\n",
1287 sb
->s_id
, PLURAL(nr_truncates
));
1289 /* Turn quotas off */
1290 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1291 if (sb_dqopt(sb
)->files
[i
])
1292 vfs_quota_off(sb
, i
);
1295 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1298 #define log2(n) ffz(~(n))
1301 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1302 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1303 * We need to be 1 filesystem block less than the 2^32 sector limit.
1305 static loff_t
ext3_max_size(int bits
)
1307 loff_t res
= EXT3_NDIR_BLOCKS
;
1308 /* This constant is calculated to be the largest file size for a
1309 * dense, 4k-blocksize file such that the total number of
1310 * sectors in the file, including data and all indirect blocks,
1311 * does not exceed 2^32. */
1312 const loff_t upper_limit
= 0x1ff7fffd000LL
;
1314 res
+= 1LL << (bits
-2);
1315 res
+= 1LL << (2*(bits
-2));
1316 res
+= 1LL << (3*(bits
-2));
1318 if (res
> upper_limit
)
1323 static ext3_fsblk_t
descriptor_loc(struct super_block
*sb
,
1324 ext3_fsblk_t logic_sb_block
,
1327 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1328 unsigned long bg
, first_meta_bg
;
1331 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1333 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_META_BG
) ||
1335 return (logic_sb_block
+ nr
+ 1);
1336 bg
= sbi
->s_desc_per_block
* nr
;
1337 if (ext3_bg_has_super(sb
, bg
))
1339 return (has_super
+ ext3_group_first_block_no(sb
, bg
));
1343 static int ext3_fill_super (struct super_block
*sb
, void *data
, int silent
)
1345 struct buffer_head
* bh
;
1346 struct ext3_super_block
*es
= NULL
;
1347 struct ext3_sb_info
*sbi
;
1349 ext3_fsblk_t sb_block
= get_sb_block(&data
);
1350 ext3_fsblk_t logic_sb_block
;
1351 unsigned long offset
= 0;
1352 unsigned long journal_inum
= 0;
1353 unsigned long journal_devnum
= 0;
1354 unsigned long def_mount_opts
;
1363 sbi
= kmalloc(sizeof(*sbi
), GFP_KERNEL
);
1366 sb
->s_fs_info
= sbi
;
1367 memset(sbi
, 0, sizeof(*sbi
));
1368 sbi
->s_mount_opt
= 0;
1369 sbi
->s_resuid
= EXT3_DEF_RESUID
;
1370 sbi
->s_resgid
= EXT3_DEF_RESGID
;
1374 blocksize
= sb_min_blocksize(sb
, EXT3_MIN_BLOCK_SIZE
);
1376 printk(KERN_ERR
"EXT3-fs: unable to set blocksize\n");
1381 * The ext3 superblock will not be buffer aligned for other than 1kB
1382 * block sizes. We need to calculate the offset from buffer start.
1384 if (blocksize
!= EXT3_MIN_BLOCK_SIZE
) {
1385 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1386 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1388 logic_sb_block
= sb_block
;
1391 if (!(bh
= sb_bread(sb
, logic_sb_block
))) {
1392 printk (KERN_ERR
"EXT3-fs: unable to read superblock\n");
1396 * Note: s_es must be initialized as soon as possible because
1397 * some ext3 macro-instructions depend on its value
1399 es
= (struct ext3_super_block
*) (((char *)bh
->b_data
) + offset
);
1401 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1402 if (sb
->s_magic
!= EXT3_SUPER_MAGIC
)
1405 /* Set defaults before we parse the mount options */
1406 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1407 if (def_mount_opts
& EXT3_DEFM_DEBUG
)
1408 set_opt(sbi
->s_mount_opt
, DEBUG
);
1409 if (def_mount_opts
& EXT3_DEFM_BSDGROUPS
)
1410 set_opt(sbi
->s_mount_opt
, GRPID
);
1411 if (def_mount_opts
& EXT3_DEFM_UID16
)
1412 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1413 if (def_mount_opts
& EXT3_DEFM_XATTR_USER
)
1414 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1415 if (def_mount_opts
& EXT3_DEFM_ACL
)
1416 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1417 if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_DATA
)
1418 sbi
->s_mount_opt
|= EXT3_MOUNT_JOURNAL_DATA
;
1419 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_ORDERED
)
1420 sbi
->s_mount_opt
|= EXT3_MOUNT_ORDERED_DATA
;
1421 else if ((def_mount_opts
& EXT3_DEFM_JMODE
) == EXT3_DEFM_JMODE_WBACK
)
1422 sbi
->s_mount_opt
|= EXT3_MOUNT_WRITEBACK_DATA
;
1424 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_PANIC
)
1425 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1426 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT3_ERRORS_RO
)
1427 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1429 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
1430 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
1432 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1434 if (!parse_options ((char *) data
, sb
, &journal_inum
, &journal_devnum
,
1438 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1439 ((sbi
->s_mount_opt
& EXT3_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
1441 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
&&
1442 (EXT3_HAS_COMPAT_FEATURE(sb
, ~0U) ||
1443 EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
1444 EXT3_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
1446 "EXT3-fs warning: feature flags set on rev 0 fs, "
1447 "running e2fsck is recommended\n");
1449 * Check feature flags regardless of the revision level, since we
1450 * previously didn't change the revision level when setting the flags,
1451 * so there is a chance incompat flags are set on a rev 0 filesystem.
1453 features
= EXT3_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
);
1455 printk(KERN_ERR
"EXT3-fs: %s: couldn't mount because of "
1456 "unsupported optional features (%x).\n",
1457 sb
->s_id
, le32_to_cpu(features
));
1460 features
= EXT3_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
);
1461 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1462 printk(KERN_ERR
"EXT3-fs: %s: couldn't mount RDWR because of "
1463 "unsupported optional features (%x).\n",
1464 sb
->s_id
, le32_to_cpu(features
));
1467 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
1469 if (blocksize
< EXT3_MIN_BLOCK_SIZE
||
1470 blocksize
> EXT3_MAX_BLOCK_SIZE
) {
1472 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1473 blocksize
, sb
->s_id
);
1477 hblock
= bdev_hardsect_size(sb
->s_bdev
);
1478 if (sb
->s_blocksize
!= blocksize
) {
1480 * Make sure the blocksize for the filesystem is larger
1481 * than the hardware sectorsize for the machine.
1483 if (blocksize
< hblock
) {
1484 printk(KERN_ERR
"EXT3-fs: blocksize %d too small for "
1485 "device blocksize %d.\n", blocksize
, hblock
);
1490 sb_set_blocksize(sb
, blocksize
);
1491 logic_sb_block
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) / blocksize
;
1492 offset
= (sb_block
* EXT3_MIN_BLOCK_SIZE
) % blocksize
;
1493 bh
= sb_bread(sb
, logic_sb_block
);
1496 "EXT3-fs: Can't read superblock on 2nd try.\n");
1499 es
= (struct ext3_super_block
*)(((char *)bh
->b_data
) + offset
);
1501 if (es
->s_magic
!= cpu_to_le16(EXT3_SUPER_MAGIC
)) {
1503 "EXT3-fs: Magic mismatch, very weird !\n");
1508 sb
->s_maxbytes
= ext3_max_size(sb
->s_blocksize_bits
);
1510 if (le32_to_cpu(es
->s_rev_level
) == EXT3_GOOD_OLD_REV
) {
1511 sbi
->s_inode_size
= EXT3_GOOD_OLD_INODE_SIZE
;
1512 sbi
->s_first_ino
= EXT3_GOOD_OLD_FIRST_INO
;
1514 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
1515 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
1516 if ((sbi
->s_inode_size
< EXT3_GOOD_OLD_INODE_SIZE
) ||
1517 (sbi
->s_inode_size
& (sbi
->s_inode_size
- 1)) ||
1518 (sbi
->s_inode_size
> blocksize
)) {
1520 "EXT3-fs: unsupported inode size: %d\n",
1525 sbi
->s_frag_size
= EXT3_MIN_FRAG_SIZE
<<
1526 le32_to_cpu(es
->s_log_frag_size
);
1527 if (blocksize
!= sbi
->s_frag_size
) {
1529 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1530 sbi
->s_frag_size
, blocksize
);
1533 sbi
->s_frags_per_block
= 1;
1534 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
1535 sbi
->s_frags_per_group
= le32_to_cpu(es
->s_frags_per_group
);
1536 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
1537 if (EXT3_INODE_SIZE(sb
) == 0)
1539 sbi
->s_inodes_per_block
= blocksize
/ EXT3_INODE_SIZE(sb
);
1540 if (sbi
->s_inodes_per_block
== 0)
1542 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
1543 sbi
->s_inodes_per_block
;
1544 sbi
->s_desc_per_block
= blocksize
/ sizeof(struct ext3_group_desc
);
1546 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
1547 sbi
->s_addr_per_block_bits
= log2(EXT3_ADDR_PER_BLOCK(sb
));
1548 sbi
->s_desc_per_block_bits
= log2(EXT3_DESC_PER_BLOCK(sb
));
1549 for (i
=0; i
< 4; i
++)
1550 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
1551 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
1553 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
1555 "EXT3-fs: #blocks per group too big: %lu\n",
1556 sbi
->s_blocks_per_group
);
1559 if (sbi
->s_frags_per_group
> blocksize
* 8) {
1561 "EXT3-fs: #fragments per group too big: %lu\n",
1562 sbi
->s_frags_per_group
);
1565 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
1567 "EXT3-fs: #inodes per group too big: %lu\n",
1568 sbi
->s_inodes_per_group
);
1572 if (le32_to_cpu(es
->s_blocks_count
) >
1573 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) {
1574 printk(KERN_ERR
"EXT3-fs: filesystem on %s:"
1575 " too large to mount safely\n", sb
->s_id
);
1576 if (sizeof(sector_t
) < 8)
1577 printk(KERN_WARNING
"EXT3-fs: CONFIG_LBD not "
1582 if (EXT3_BLOCKS_PER_GROUP(sb
) == 0)
1584 sbi
->s_groups_count
= (le32_to_cpu(es
->s_blocks_count
) -
1585 le32_to_cpu(es
->s_first_data_block
) +
1586 EXT3_BLOCKS_PER_GROUP(sb
) - 1) /
1587 EXT3_BLOCKS_PER_GROUP(sb
);
1588 db_count
= (sbi
->s_groups_count
+ EXT3_DESC_PER_BLOCK(sb
) - 1) /
1589 EXT3_DESC_PER_BLOCK(sb
);
1590 sbi
->s_group_desc
= kmalloc(db_count
* sizeof (struct buffer_head
*),
1592 if (sbi
->s_group_desc
== NULL
) {
1593 printk (KERN_ERR
"EXT3-fs: not enough memory\n");
1597 bgl_lock_init(&sbi
->s_blockgroup_lock
);
1599 for (i
= 0; i
< db_count
; i
++) {
1600 block
= descriptor_loc(sb
, logic_sb_block
, i
);
1601 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
1602 if (!sbi
->s_group_desc
[i
]) {
1603 printk (KERN_ERR
"EXT3-fs: "
1604 "can't read group descriptor %d\n", i
);
1609 if (!ext3_check_descriptors (sb
)) {
1610 printk(KERN_ERR
"EXT3-fs: group descriptors corrupted!\n");
1613 sbi
->s_gdb_count
= db_count
;
1614 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1615 spin_lock_init(&sbi
->s_next_gen_lock
);
1617 percpu_counter_init(&sbi
->s_freeblocks_counter
,
1618 ext3_count_free_blocks(sb
));
1619 percpu_counter_init(&sbi
->s_freeinodes_counter
,
1620 ext3_count_free_inodes(sb
));
1621 percpu_counter_init(&sbi
->s_dirs_counter
,
1622 ext3_count_dirs(sb
));
1624 /* per fileystem reservation list head & lock */
1625 spin_lock_init(&sbi
->s_rsv_window_lock
);
1626 sbi
->s_rsv_window_root
= RB_ROOT
;
1627 /* Add a single, static dummy reservation to the start of the
1628 * reservation window list --- it gives us a placeholder for
1629 * append-at-start-of-list which makes the allocation logic
1630 * _much_ simpler. */
1631 sbi
->s_rsv_window_head
.rsv_start
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1632 sbi
->s_rsv_window_head
.rsv_end
= EXT3_RESERVE_WINDOW_NOT_ALLOCATED
;
1633 sbi
->s_rsv_window_head
.rsv_alloc_hit
= 0;
1634 sbi
->s_rsv_window_head
.rsv_goal_size
= 0;
1635 ext3_rsv_window_add(sb
, &sbi
->s_rsv_window_head
);
1638 * set up enough so that it can read an inode
1640 sb
->s_op
= &ext3_sops
;
1641 sb
->s_export_op
= &ext3_export_ops
;
1642 sb
->s_xattr
= ext3_xattr_handlers
;
1644 sb
->s_qcop
= &ext3_qctl_operations
;
1645 sb
->dq_op
= &ext3_quota_operations
;
1647 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
1651 needs_recovery
= (es
->s_last_orphan
!= 0 ||
1652 EXT3_HAS_INCOMPAT_FEATURE(sb
,
1653 EXT3_FEATURE_INCOMPAT_RECOVER
));
1656 * The first inode we look at is the journal inode. Don't try
1657 * root first: it may be modified in the journal!
1659 if (!test_opt(sb
, NOLOAD
) &&
1660 EXT3_HAS_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
)) {
1661 if (ext3_load_journal(sb
, es
, journal_devnum
))
1663 } else if (journal_inum
) {
1664 if (ext3_create_journal(sb
, es
, journal_inum
))
1669 "ext3: No journal on filesystem on %s\n",
1674 /* We have now updated the journal if required, so we can
1675 * validate the data journaling mode. */
1676 switch (test_opt(sb
, DATA_FLAGS
)) {
1678 /* No mode set, assume a default based on the journal
1679 capabilities: ORDERED_DATA if the journal can
1680 cope, else JOURNAL_DATA */
1681 if (journal_check_available_features
1682 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
))
1683 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
1685 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
1688 case EXT3_MOUNT_ORDERED_DATA
:
1689 case EXT3_MOUNT_WRITEBACK_DATA
:
1690 if (!journal_check_available_features
1691 (sbi
->s_journal
, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE
)) {
1692 printk(KERN_ERR
"EXT3-fs: Journal does not support "
1693 "requested data journaling mode\n");
1700 if (test_opt(sb
, NOBH
)) {
1701 if (!(test_opt(sb
, DATA_FLAGS
) == EXT3_MOUNT_WRITEBACK_DATA
)) {
1702 printk(KERN_WARNING
"EXT3-fs: Ignoring nobh option - "
1703 "its supported only with writeback mode\n");
1704 clear_opt(sbi
->s_mount_opt
, NOBH
);
1708 * The journal_load will have done any necessary log recovery,
1709 * so we can safely mount the rest of the filesystem now.
1712 root
= iget(sb
, EXT3_ROOT_INO
);
1713 sb
->s_root
= d_alloc_root(root
);
1715 printk(KERN_ERR
"EXT3-fs: get root inode failed\n");
1719 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1722 printk(KERN_ERR
"EXT3-fs: corrupt root inode, run e2fsck\n");
1726 ext3_setup_super (sb
, es
, sb
->s_flags
& MS_RDONLY
);
1728 * akpm: core read_super() calls in here with the superblock locked.
1729 * That deadlocks, because orphan cleanup needs to lock the superblock
1730 * in numerous places. Here we just pop the lock - it's relatively
1731 * harmless, because we are now ready to accept write_super() requests,
1732 * and aviro says that's the only reason for hanging onto the
1735 EXT3_SB(sb
)->s_mount_state
|= EXT3_ORPHAN_FS
;
1736 ext3_orphan_cleanup(sb
, es
);
1737 EXT3_SB(sb
)->s_mount_state
&= ~EXT3_ORPHAN_FS
;
1739 printk (KERN_INFO
"EXT3-fs: recovery complete.\n");
1740 ext3_mark_recovery_complete(sb
, es
);
1741 printk (KERN_INFO
"EXT3-fs: mounted filesystem with %s data mode.\n",
1742 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_JOURNAL_DATA
? "journal":
1743 test_opt(sb
,DATA_FLAGS
) == EXT3_MOUNT_ORDERED_DATA
? "ordered":
1751 printk(KERN_ERR
"VFS: Can't find ext3 filesystem on dev %s.\n",
1756 journal_destroy(sbi
->s_journal
);
1758 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
1759 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
1760 percpu_counter_destroy(&sbi
->s_dirs_counter
);
1762 for (i
= 0; i
< db_count
; i
++)
1763 brelse(sbi
->s_group_desc
[i
]);
1764 kfree(sbi
->s_group_desc
);
1767 for (i
= 0; i
< MAXQUOTAS
; i
++)
1768 kfree(sbi
->s_qf_names
[i
]);
1770 ext3_blkdev_remove(sbi
);
1773 sb
->s_fs_info
= NULL
;
1780 * Setup any per-fs journal parameters now. We'll do this both on
1781 * initial mount, once the journal has been initialised but before we've
1782 * done any recovery; and again on any subsequent remount.
1784 static void ext3_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
1786 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
1788 if (sbi
->s_commit_interval
)
1789 journal
->j_commit_interval
= sbi
->s_commit_interval
;
1790 /* We could also set up an ext3-specific default for the commit
1791 * interval here, but for now we'll just fall back to the jbd
1794 spin_lock(&journal
->j_state_lock
);
1795 if (test_opt(sb
, BARRIER
))
1796 journal
->j_flags
|= JFS_BARRIER
;
1798 journal
->j_flags
&= ~JFS_BARRIER
;
1799 spin_unlock(&journal
->j_state_lock
);
1802 static journal_t
*ext3_get_journal(struct super_block
*sb
, int journal_inum
)
1804 struct inode
*journal_inode
;
1807 /* First, test for the existence of a valid inode on disk. Bad
1808 * things happen if we iget() an unused inode, as the subsequent
1809 * iput() will try to delete it. */
1811 journal_inode
= iget(sb
, journal_inum
);
1812 if (!journal_inode
) {
1813 printk(KERN_ERR
"EXT3-fs: no journal found.\n");
1816 if (!journal_inode
->i_nlink
) {
1817 make_bad_inode(journal_inode
);
1818 iput(journal_inode
);
1819 printk(KERN_ERR
"EXT3-fs: journal inode is deleted.\n");
1823 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1824 journal_inode
, journal_inode
->i_size
);
1825 if (is_bad_inode(journal_inode
) || !S_ISREG(journal_inode
->i_mode
)) {
1826 printk(KERN_ERR
"EXT3-fs: invalid journal inode.\n");
1827 iput(journal_inode
);
1831 journal
= journal_init_inode(journal_inode
);
1833 printk(KERN_ERR
"EXT3-fs: Could not load journal inode\n");
1834 iput(journal_inode
);
1837 journal
->j_private
= sb
;
1838 ext3_init_journal_params(sb
, journal
);
1842 static journal_t
*ext3_get_dev_journal(struct super_block
*sb
,
1845 struct buffer_head
* bh
;
1849 int hblock
, blocksize
;
1850 ext3_fsblk_t sb_block
;
1851 unsigned long offset
;
1852 struct ext3_super_block
* es
;
1853 struct block_device
*bdev
;
1855 bdev
= ext3_blkdev_get(j_dev
);
1859 if (bd_claim(bdev
, sb
)) {
1861 "EXT3: failed to claim external journal device.\n");
1866 blocksize
= sb
->s_blocksize
;
1867 hblock
= bdev_hardsect_size(bdev
);
1868 if (blocksize
< hblock
) {
1870 "EXT3-fs: blocksize too small for journal device.\n");
1874 sb_block
= EXT3_MIN_BLOCK_SIZE
/ blocksize
;
1875 offset
= EXT3_MIN_BLOCK_SIZE
% blocksize
;
1876 set_blocksize(bdev
, blocksize
);
1877 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
1878 printk(KERN_ERR
"EXT3-fs: couldn't read superblock of "
1879 "external journal\n");
1883 es
= (struct ext3_super_block
*) (((char *)bh
->b_data
) + offset
);
1884 if ((le16_to_cpu(es
->s_magic
) != EXT3_SUPER_MAGIC
) ||
1885 !(le32_to_cpu(es
->s_feature_incompat
) &
1886 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
1887 printk(KERN_ERR
"EXT3-fs: external journal has "
1888 "bad superblock\n");
1893 if (memcmp(EXT3_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
1894 printk(KERN_ERR
"EXT3-fs: journal UUID does not match\n");
1899 len
= le32_to_cpu(es
->s_blocks_count
);
1900 start
= sb_block
+ 1;
1901 brelse(bh
); /* we're done with the superblock */
1903 journal
= journal_init_dev(bdev
, sb
->s_bdev
,
1904 start
, len
, blocksize
);
1906 printk(KERN_ERR
"EXT3-fs: failed to create device journal\n");
1909 journal
->j_private
= sb
;
1910 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
1911 wait_on_buffer(journal
->j_sb_buffer
);
1912 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
1913 printk(KERN_ERR
"EXT3-fs: I/O error on journal device\n");
1916 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
1917 printk(KERN_ERR
"EXT3-fs: External journal has more than one "
1918 "user (unsupported) - %d\n",
1919 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
1922 EXT3_SB(sb
)->journal_bdev
= bdev
;
1923 ext3_init_journal_params(sb
, journal
);
1926 journal_destroy(journal
);
1928 ext3_blkdev_put(bdev
);
1932 static int ext3_load_journal(struct super_block
*sb
,
1933 struct ext3_super_block
*es
,
1934 unsigned long journal_devnum
)
1937 int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
1940 int really_read_only
;
1942 if (journal_devnum
&&
1943 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
1944 printk(KERN_INFO
"EXT3-fs: external journal device major/minor "
1945 "numbers have changed\n");
1946 journal_dev
= new_decode_dev(journal_devnum
);
1948 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
1950 really_read_only
= bdev_read_only(sb
->s_bdev
);
1953 * Are we loading a blank journal or performing recovery after a
1954 * crash? For recovery, we need to check in advance whether we
1955 * can get read-write access to the device.
1958 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
)) {
1959 if (sb
->s_flags
& MS_RDONLY
) {
1960 printk(KERN_INFO
"EXT3-fs: INFO: recovery "
1961 "required on readonly filesystem.\n");
1962 if (really_read_only
) {
1963 printk(KERN_ERR
"EXT3-fs: write access "
1964 "unavailable, cannot proceed.\n");
1967 printk (KERN_INFO
"EXT3-fs: write access will "
1968 "be enabled during recovery.\n");
1972 if (journal_inum
&& journal_dev
) {
1973 printk(KERN_ERR
"EXT3-fs: filesystem has both journal "
1974 "and inode journals!\n");
1979 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
1982 if (!(journal
= ext3_get_dev_journal(sb
, journal_dev
)))
1986 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
1987 err
= journal_update_format(journal
);
1989 printk(KERN_ERR
"EXT3-fs: error updating journal.\n");
1990 journal_destroy(journal
);
1995 if (!EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
))
1996 err
= journal_wipe(journal
, !really_read_only
);
1998 err
= journal_load(journal
);
2001 printk(KERN_ERR
"EXT3-fs: error loading journal.\n");
2002 journal_destroy(journal
);
2006 EXT3_SB(sb
)->s_journal
= journal
;
2007 ext3_clear_journal_err(sb
, es
);
2009 if (journal_devnum
&&
2010 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2011 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2014 /* Make sure we flush the recovery flag to disk. */
2015 ext3_commit_super(sb
, es
, 1);
2021 static int ext3_create_journal(struct super_block
* sb
,
2022 struct ext3_super_block
* es
,
2027 if (sb
->s_flags
& MS_RDONLY
) {
2028 printk(KERN_ERR
"EXT3-fs: readonly filesystem when trying to "
2029 "create journal.\n");
2033 if (!(journal
= ext3_get_journal(sb
, journal_inum
)))
2036 printk(KERN_INFO
"EXT3-fs: creating new journal on inode %d\n",
2039 if (journal_create(journal
)) {
2040 printk(KERN_ERR
"EXT3-fs: error creating journal.\n");
2041 journal_destroy(journal
);
2045 EXT3_SB(sb
)->s_journal
= journal
;
2047 ext3_update_dynamic_rev(sb
);
2048 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2049 EXT3_SET_COMPAT_FEATURE(sb
, EXT3_FEATURE_COMPAT_HAS_JOURNAL
);
2051 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
2054 /* Make sure we flush the recovery flag to disk. */
2055 ext3_commit_super(sb
, es
, 1);
2060 static void ext3_commit_super (struct super_block
* sb
,
2061 struct ext3_super_block
* es
,
2064 struct buffer_head
*sbh
= EXT3_SB(sb
)->s_sbh
;
2068 es
->s_wtime
= cpu_to_le32(get_seconds());
2069 es
->s_free_blocks_count
= cpu_to_le32(ext3_count_free_blocks(sb
));
2070 es
->s_free_inodes_count
= cpu_to_le32(ext3_count_free_inodes(sb
));
2071 BUFFER_TRACE(sbh
, "marking dirty");
2072 mark_buffer_dirty(sbh
);
2074 sync_dirty_buffer(sbh
);
2079 * Have we just finished recovery? If so, and if we are mounting (or
2080 * remounting) the filesystem readonly, then we will end up with a
2081 * consistent fs on disk. Record that fact.
2083 static void ext3_mark_recovery_complete(struct super_block
* sb
,
2084 struct ext3_super_block
* es
)
2086 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
2088 journal_lock_updates(journal
);
2089 journal_flush(journal
);
2090 if (EXT3_HAS_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
) &&
2091 sb
->s_flags
& MS_RDONLY
) {
2092 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2094 ext3_commit_super(sb
, es
, 1);
2096 journal_unlock_updates(journal
);
2100 * If we are mounting (or read-write remounting) a filesystem whose journal
2101 * has recorded an error from a previous lifetime, move that error to the
2102 * main filesystem now.
2104 static void ext3_clear_journal_err(struct super_block
* sb
,
2105 struct ext3_super_block
* es
)
2111 journal
= EXT3_SB(sb
)->s_journal
;
2114 * Now check for any error status which may have been recorded in the
2115 * journal by a prior ext3_error() or ext3_abort()
2118 j_errno
= journal_errno(journal
);
2122 errstr
= ext3_decode_error(sb
, j_errno
, nbuf
);
2123 ext3_warning(sb
, __FUNCTION__
, "Filesystem error recorded "
2124 "from previous mount: %s", errstr
);
2125 ext3_warning(sb
, __FUNCTION__
, "Marking fs in need of "
2126 "filesystem check.");
2128 EXT3_SB(sb
)->s_mount_state
|= EXT3_ERROR_FS
;
2129 es
->s_state
|= cpu_to_le16(EXT3_ERROR_FS
);
2130 ext3_commit_super (sb
, es
, 1);
2132 journal_clear_err(journal
);
2137 * Force the running and committing transactions to commit,
2138 * and wait on the commit.
2140 int ext3_force_commit(struct super_block
*sb
)
2145 if (sb
->s_flags
& MS_RDONLY
)
2148 journal
= EXT3_SB(sb
)->s_journal
;
2150 ret
= ext3_journal_force_commit(journal
);
2155 * Ext3 always journals updates to the superblock itself, so we don't
2156 * have to propagate any other updates to the superblock on disk at this
2157 * point. Just start an async writeback to get the buffers on their way
2160 * This implicitly triggers the writebehind on sync().
2163 static void ext3_write_super (struct super_block
* sb
)
2165 if (mutex_trylock(&sb
->s_lock
) != 0)
2170 static int ext3_sync_fs(struct super_block
*sb
, int wait
)
2175 if (journal_start_commit(EXT3_SB(sb
)->s_journal
, &target
)) {
2177 log_wait_commit(EXT3_SB(sb
)->s_journal
, target
);
2183 * LVM calls this function before a (read-only) snapshot is created. This
2184 * gives us a chance to flush the journal completely and mark the fs clean.
2186 static void ext3_write_super_lockfs(struct super_block
*sb
)
2190 if (!(sb
->s_flags
& MS_RDONLY
)) {
2191 journal_t
*journal
= EXT3_SB(sb
)->s_journal
;
2193 /* Now we set up the journal barrier. */
2194 journal_lock_updates(journal
);
2195 journal_flush(journal
);
2197 /* Journal blocked and flushed, clear needs_recovery flag. */
2198 EXT3_CLEAR_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2199 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2204 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2205 * flag here, even though the filesystem is not technically dirty yet.
2207 static void ext3_unlockfs(struct super_block
*sb
)
2209 if (!(sb
->s_flags
& MS_RDONLY
)) {
2211 /* Reser the needs_recovery flag before the fs is unlocked. */
2212 EXT3_SET_INCOMPAT_FEATURE(sb
, EXT3_FEATURE_INCOMPAT_RECOVER
);
2213 ext3_commit_super(sb
, EXT3_SB(sb
)->s_es
, 1);
2215 journal_unlock_updates(EXT3_SB(sb
)->s_journal
);
2219 static int ext3_remount (struct super_block
* sb
, int * flags
, char * data
)
2221 struct ext3_super_block
* es
;
2222 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2223 ext3_fsblk_t n_blocks_count
= 0;
2224 unsigned long old_sb_flags
;
2225 struct ext3_mount_options old_opts
;
2231 /* Store the original options */
2232 old_sb_flags
= sb
->s_flags
;
2233 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
2234 old_opts
.s_resuid
= sbi
->s_resuid
;
2235 old_opts
.s_resgid
= sbi
->s_resgid
;
2236 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
2238 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
2239 for (i
= 0; i
< MAXQUOTAS
; i
++)
2240 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
2244 * Allow the "check" option to be passed as a remount option.
2246 if (!parse_options(data
, sb
, NULL
, NULL
, &n_blocks_count
, 1)) {
2251 if (sbi
->s_mount_opt
& EXT3_MOUNT_ABORT
)
2252 ext3_abort(sb
, __FUNCTION__
, "Abort forced by user");
2254 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2255 ((sbi
->s_mount_opt
& EXT3_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2259 ext3_init_journal_params(sb
, sbi
->s_journal
);
2261 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
2262 n_blocks_count
> le32_to_cpu(es
->s_blocks_count
)) {
2263 if (sbi
->s_mount_opt
& EXT3_MOUNT_ABORT
) {
2268 if (*flags
& MS_RDONLY
) {
2270 * First of all, the unconditional stuff we have to do
2271 * to disable replay of the journal when we next remount
2273 sb
->s_flags
|= MS_RDONLY
;
2276 * OK, test if we are remounting a valid rw partition
2277 * readonly, and if so set the rdonly flag and then
2278 * mark the partition as valid again.
2280 if (!(es
->s_state
& cpu_to_le16(EXT3_VALID_FS
)) &&
2281 (sbi
->s_mount_state
& EXT3_VALID_FS
))
2282 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
2284 ext3_mark_recovery_complete(sb
, es
);
2287 if ((ret
= EXT3_HAS_RO_COMPAT_FEATURE(sb
,
2288 ~EXT3_FEATURE_RO_COMPAT_SUPP
))) {
2289 printk(KERN_WARNING
"EXT3-fs: %s: couldn't "
2290 "remount RDWR because of unsupported "
2291 "optional features (%x).\n",
2292 sb
->s_id
, le32_to_cpu(ret
));
2297 * Mounting a RDONLY partition read-write, so reread
2298 * and store the current valid flag. (It may have
2299 * been changed by e2fsck since we originally mounted
2302 ext3_clear_journal_err(sb
, es
);
2303 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2304 if ((ret
= ext3_group_extend(sb
, es
, n_blocks_count
))) {
2308 if (!ext3_setup_super (sb
, es
, 0))
2309 sb
->s_flags
&= ~MS_RDONLY
;
2313 /* Release old quota file names */
2314 for (i
= 0; i
< MAXQUOTAS
; i
++)
2315 if (old_opts
.s_qf_names
[i
] &&
2316 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2317 kfree(old_opts
.s_qf_names
[i
]);
2321 sb
->s_flags
= old_sb_flags
;
2322 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
2323 sbi
->s_resuid
= old_opts
.s_resuid
;
2324 sbi
->s_resgid
= old_opts
.s_resgid
;
2325 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
2327 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
2328 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2329 if (sbi
->s_qf_names
[i
] &&
2330 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
2331 kfree(sbi
->s_qf_names
[i
]);
2332 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
2338 static int ext3_statfs (struct dentry
* dentry
, struct kstatfs
* buf
)
2340 struct super_block
*sb
= dentry
->d_sb
;
2341 struct ext3_sb_info
*sbi
= EXT3_SB(sb
);
2342 struct ext3_super_block
*es
= sbi
->s_es
;
2343 ext3_fsblk_t overhead
;
2346 if (test_opt (sb
, MINIX_DF
))
2349 unsigned long ngroups
;
2350 ngroups
= EXT3_SB(sb
)->s_groups_count
;
2354 * Compute the overhead (FS structures)
2358 * All of the blocks before first_data_block are
2361 overhead
= le32_to_cpu(es
->s_first_data_block
);
2364 * Add the overhead attributed to the superblock and
2365 * block group descriptors. If the sparse superblocks
2366 * feature is turned on, then not all groups have this.
2368 for (i
= 0; i
< ngroups
; i
++) {
2369 overhead
+= ext3_bg_has_super(sb
, i
) +
2370 ext3_bg_num_gdb(sb
, i
);
2375 * Every block group has an inode bitmap, a block
2376 * bitmap, and an inode table.
2378 overhead
+= (ngroups
* (2 + EXT3_SB(sb
)->s_itb_per_group
));
2381 buf
->f_type
= EXT3_SUPER_MAGIC
;
2382 buf
->f_bsize
= sb
->s_blocksize
;
2383 buf
->f_blocks
= le32_to_cpu(es
->s_blocks_count
) - overhead
;
2384 buf
->f_bfree
= percpu_counter_sum(&sbi
->s_freeblocks_counter
);
2385 buf
->f_bavail
= buf
->f_bfree
- le32_to_cpu(es
->s_r_blocks_count
);
2386 if (buf
->f_bfree
< le32_to_cpu(es
->s_r_blocks_count
))
2388 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
2389 buf
->f_ffree
= percpu_counter_sum(&sbi
->s_freeinodes_counter
);
2390 buf
->f_namelen
= EXT3_NAME_LEN
;
2394 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2395 * is locked for write. Otherwise the are possible deadlocks:
2396 * Process 1 Process 2
2397 * ext3_create() quota_sync()
2398 * journal_start() write_dquot()
2399 * DQUOT_INIT() down(dqio_mutex)
2400 * down(dqio_mutex) journal_start()
2406 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
2408 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
2411 static int ext3_dquot_initialize(struct inode
*inode
, int type
)
2416 /* We may create quota structure so we need to reserve enough blocks */
2417 handle
= ext3_journal_start(inode
, 2*EXT3_QUOTA_INIT_BLOCKS(inode
->i_sb
));
2419 return PTR_ERR(handle
);
2420 ret
= dquot_initialize(inode
, type
);
2421 err
= ext3_journal_stop(handle
);
2427 static int ext3_dquot_drop(struct inode
*inode
)
2432 /* We may delete quota structure so we need to reserve enough blocks */
2433 handle
= ext3_journal_start(inode
, 2*EXT3_QUOTA_DEL_BLOCKS(inode
->i_sb
));
2435 return PTR_ERR(handle
);
2436 ret
= dquot_drop(inode
);
2437 err
= ext3_journal_stop(handle
);
2443 static int ext3_write_dquot(struct dquot
*dquot
)
2447 struct inode
*inode
;
2449 inode
= dquot_to_inode(dquot
);
2450 handle
= ext3_journal_start(inode
,
2451 EXT3_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
2453 return PTR_ERR(handle
);
2454 ret
= dquot_commit(dquot
);
2455 err
= ext3_journal_stop(handle
);
2461 static int ext3_acquire_dquot(struct dquot
*dquot
)
2466 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2467 EXT3_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
2469 return PTR_ERR(handle
);
2470 ret
= dquot_acquire(dquot
);
2471 err
= ext3_journal_stop(handle
);
2477 static int ext3_release_dquot(struct dquot
*dquot
)
2482 handle
= ext3_journal_start(dquot_to_inode(dquot
),
2483 EXT3_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
2485 return PTR_ERR(handle
);
2486 ret
= dquot_release(dquot
);
2487 err
= ext3_journal_stop(handle
);
2493 static int ext3_mark_dquot_dirty(struct dquot
*dquot
)
2495 /* Are we journalling quotas? */
2496 if (EXT3_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
2497 EXT3_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
2498 dquot_mark_dquot_dirty(dquot
);
2499 return ext3_write_dquot(dquot
);
2501 return dquot_mark_dquot_dirty(dquot
);
2505 static int ext3_write_info(struct super_block
*sb
, int type
)
2510 /* Data block + inode block */
2511 handle
= ext3_journal_start(sb
->s_root
->d_inode
, 2);
2513 return PTR_ERR(handle
);
2514 ret
= dquot_commit_info(sb
, type
);
2515 err
= ext3_journal_stop(handle
);
2522 * Turn on quotas during mount time - we need to find
2523 * the quota file and such...
2525 static int ext3_quota_on_mount(struct super_block
*sb
, int type
)
2527 return vfs_quota_on_mount(sb
, EXT3_SB(sb
)->s_qf_names
[type
],
2528 EXT3_SB(sb
)->s_jquota_fmt
, type
);
2532 * Standard function to be called on quota_on
2534 static int ext3_quota_on(struct super_block
*sb
, int type
, int format_id
,
2538 struct nameidata nd
;
2540 if (!test_opt(sb
, QUOTA
))
2542 /* Not journalling quota? */
2543 if (!EXT3_SB(sb
)->s_qf_names
[USRQUOTA
] &&
2544 !EXT3_SB(sb
)->s_qf_names
[GRPQUOTA
])
2545 return vfs_quota_on(sb
, type
, format_id
, path
);
2546 err
= path_lookup(path
, LOOKUP_FOLLOW
, &nd
);
2549 /* Quotafile not on the same filesystem? */
2550 if (nd
.mnt
->mnt_sb
!= sb
) {
2554 /* Quotafile not of fs root? */
2555 if (nd
.dentry
->d_parent
->d_inode
!= sb
->s_root
->d_inode
)
2557 "EXT3-fs: Quota file not on filesystem root. "
2558 "Journalled quota will not work.\n");
2560 return vfs_quota_on(sb
, type
, format_id
, path
);
2563 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2564 * acquiring the locks... As quota files are never truncated and quota code
2565 * itself serializes the operations (and noone else should touch the files)
2566 * we don't have to be afraid of races */
2567 static ssize_t
ext3_quota_read(struct super_block
*sb
, int type
, char *data
,
2568 size_t len
, loff_t off
)
2570 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2571 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2573 int offset
= off
& (sb
->s_blocksize
- 1);
2576 struct buffer_head
*bh
;
2577 loff_t i_size
= i_size_read(inode
);
2581 if (off
+len
> i_size
)
2584 while (toread
> 0) {
2585 tocopy
= sb
->s_blocksize
- offset
< toread
?
2586 sb
->s_blocksize
- offset
: toread
;
2587 bh
= ext3_bread(NULL
, inode
, blk
, 0, &err
);
2590 if (!bh
) /* A hole? */
2591 memset(data
, 0, tocopy
);
2593 memcpy(data
, bh
->b_data
+offset
, tocopy
);
2603 /* Write to quotafile (we know the transaction is already started and has
2604 * enough credits) */
2605 static ssize_t
ext3_quota_write(struct super_block
*sb
, int type
,
2606 const char *data
, size_t len
, loff_t off
)
2608 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
2609 sector_t blk
= off
>> EXT3_BLOCK_SIZE_BITS(sb
);
2611 int offset
= off
& (sb
->s_blocksize
- 1);
2613 int journal_quota
= EXT3_SB(sb
)->s_qf_names
[type
] != NULL
;
2614 size_t towrite
= len
;
2615 struct buffer_head
*bh
;
2616 handle_t
*handle
= journal_current_handle();
2618 mutex_lock(&inode
->i_mutex
);
2619 while (towrite
> 0) {
2620 tocopy
= sb
->s_blocksize
- offset
< towrite
?
2621 sb
->s_blocksize
- offset
: towrite
;
2622 bh
= ext3_bread(handle
, inode
, blk
, 1, &err
);
2625 if (journal_quota
) {
2626 err
= ext3_journal_get_write_access(handle
, bh
);
2633 memcpy(bh
->b_data
+offset
, data
, tocopy
);
2634 flush_dcache_page(bh
->b_page
);
2637 err
= ext3_journal_dirty_metadata(handle
, bh
);
2639 /* Always do at least ordered writes for quotas */
2640 err
= ext3_journal_dirty_data(handle
, bh
);
2641 mark_buffer_dirty(bh
);
2654 if (inode
->i_size
< off
+len
-towrite
) {
2655 i_size_write(inode
, off
+len
-towrite
);
2656 EXT3_I(inode
)->i_disksize
= inode
->i_size
;
2659 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2660 ext3_mark_inode_dirty(handle
, inode
);
2661 mutex_unlock(&inode
->i_mutex
);
2662 return len
- towrite
;
2667 static int ext3_get_sb(struct file_system_type
*fs_type
,
2668 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
2670 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext3_fill_super
, mnt
);
2673 static struct file_system_type ext3_fs_type
= {
2674 .owner
= THIS_MODULE
,
2676 .get_sb
= ext3_get_sb
,
2677 .kill_sb
= kill_block_super
,
2678 .fs_flags
= FS_REQUIRES_DEV
,
2681 static int __init
init_ext3_fs(void)
2683 int err
= init_ext3_xattr();
2686 err
= init_inodecache();
2689 err
= register_filesystem(&ext3_fs_type
);
2694 destroy_inodecache();
2700 static void __exit
exit_ext3_fs(void)
2702 unregister_filesystem(&ext3_fs_type
);
2703 destroy_inodecache();
2707 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2708 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2709 MODULE_LICENSE("GPL");
2710 module_init(init_ext3_fs
)
2711 module_exit(exit_ext3_fs
)