2 * linux/fs/ext4/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/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
44 #include "ext4_jbd2.h"
50 struct proc_dir_entry
*ext4_proc_root
;
52 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
53 unsigned long journal_devnum
);
54 static int ext4_create_journal(struct super_block
*, struct ext4_super_block
*,
56 static void ext4_commit_super(struct super_block
*sb
,
57 struct ext4_super_block
*es
, int sync
);
58 static void ext4_mark_recovery_complete(struct super_block
*sb
,
59 struct ext4_super_block
*es
);
60 static void ext4_clear_journal_err(struct super_block
*sb
,
61 struct ext4_super_block
*es
);
62 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
63 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
65 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
66 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
67 static void ext4_unlockfs(struct super_block
*sb
);
68 static void ext4_write_super(struct super_block
*sb
);
69 static void ext4_write_super_lockfs(struct super_block
*sb
);
72 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
73 struct ext4_group_desc
*bg
)
75 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
76 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
77 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
80 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
81 struct ext4_group_desc
*bg
)
83 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
84 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
85 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
88 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
89 struct ext4_group_desc
*bg
)
91 return le32_to_cpu(bg
->bg_inode_table_lo
) |
92 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
93 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
96 void ext4_block_bitmap_set(struct super_block
*sb
,
97 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
99 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
100 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
101 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
104 void ext4_inode_bitmap_set(struct super_block
*sb
,
105 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
107 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
108 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
109 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
112 void ext4_inode_table_set(struct super_block
*sb
,
113 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
115 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
116 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
117 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
121 * Wrappers for jbd2_journal_start/end.
123 * The only special thing we need to do here is to make sure that all
124 * journal_end calls result in the superblock being marked dirty, so
125 * that sync() will call the filesystem's write_super callback if
128 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
132 if (sb
->s_flags
& MS_RDONLY
)
133 return ERR_PTR(-EROFS
);
135 /* Special case here: if the journal has aborted behind our
136 * backs (eg. EIO in the commit thread), then we still need to
137 * take the FS itself readonly cleanly. */
138 journal
= EXT4_SB(sb
)->s_journal
;
140 if (is_journal_aborted(journal
)) {
141 ext4_abort(sb
, __func__
,
142 "Detected aborted journal");
143 return ERR_PTR(-EROFS
);
145 return jbd2_journal_start(journal
, nblocks
);
148 * We're not journaling, return the appropriate indication.
150 current
->journal_info
= EXT4_NOJOURNAL_HANDLE
;
151 return current
->journal_info
;
155 * The only special thing we need to do here is to make sure that all
156 * jbd2_journal_stop calls result in the superblock being marked dirty, so
157 * that sync() will call the filesystem's write_super callback if
160 int __ext4_journal_stop(const char *where
, handle_t
*handle
)
162 struct super_block
*sb
;
166 if (!ext4_handle_valid(handle
)) {
168 * Do this here since we don't call jbd2_journal_stop() in
171 current
->journal_info
= NULL
;
174 sb
= handle
->h_transaction
->t_journal
->j_private
;
176 rc
= jbd2_journal_stop(handle
);
181 __ext4_std_error(sb
, where
, err
);
185 void ext4_journal_abort_handle(const char *caller
, const char *err_fn
,
186 struct buffer_head
*bh
, handle_t
*handle
, int err
)
189 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
191 BUG_ON(!ext4_handle_valid(handle
));
194 BUFFER_TRACE(bh
, "abort");
199 if (is_handle_aborted(handle
))
202 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
203 caller
, errstr
, err_fn
);
205 jbd2_journal_abort_handle(handle
);
208 /* Deal with the reporting of failure conditions on a filesystem such as
209 * inconsistencies detected or read IO failures.
211 * On ext2, we can store the error state of the filesystem in the
212 * superblock. That is not possible on ext4, because we may have other
213 * write ordering constraints on the superblock which prevent us from
214 * writing it out straight away; and given that the journal is about to
215 * be aborted, we can't rely on the current, or future, transactions to
216 * write out the superblock safely.
218 * We'll just use the jbd2_journal_abort() error code to record an error in
219 * the journal instead. On recovery, the journal will compain about
220 * that error until we've noted it down and cleared it.
223 static void ext4_handle_error(struct super_block
*sb
)
225 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
227 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
228 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
230 if (sb
->s_flags
& MS_RDONLY
)
233 if (!test_opt(sb
, ERRORS_CONT
)) {
234 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
236 EXT4_SB(sb
)->s_mount_opt
|= EXT4_MOUNT_ABORT
;
238 jbd2_journal_abort(journal
, -EIO
);
240 if (test_opt(sb
, ERRORS_RO
)) {
241 printk(KERN_CRIT
"Remounting filesystem read-only\n");
242 sb
->s_flags
|= MS_RDONLY
;
244 ext4_commit_super(sb
, es
, 1);
245 if (test_opt(sb
, ERRORS_PANIC
))
246 panic("EXT4-fs (device %s): panic forced after error\n",
250 void ext4_error(struct super_block
*sb
, const char *function
,
251 const char *fmt
, ...)
256 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
261 ext4_handle_error(sb
);
264 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
271 errstr
= "IO failure";
274 errstr
= "Out of memory";
277 if (!sb
|| EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
)
278 errstr
= "Journal has aborted";
280 errstr
= "Readonly filesystem";
283 /* If the caller passed in an extra buffer for unknown
284 * errors, textualise them now. Else we just return
287 /* Check for truncated error codes... */
288 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
297 /* __ext4_std_error decodes expected errors from journaling functions
298 * automatically and invokes the appropriate error response. */
300 void __ext4_std_error(struct super_block
*sb
, const char *function
, int errno
)
305 /* Special case: if the error is EROFS, and we're not already
306 * inside a transaction, then there's really no point in logging
308 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
309 (sb
->s_flags
& MS_RDONLY
))
312 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
313 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s: %s\n",
314 sb
->s_id
, function
, errstr
);
316 ext4_handle_error(sb
);
320 * ext4_abort is a much stronger failure handler than ext4_error. The
321 * abort function may be used to deal with unrecoverable failures such
322 * as journal IO errors or ENOMEM at a critical moment in log management.
324 * We unconditionally force the filesystem into an ABORT|READONLY state,
325 * unless the error response on the fs has been set to panic in which
326 * case we take the easy way out and panic immediately.
329 void ext4_abort(struct super_block
*sb
, const char *function
,
330 const char *fmt
, ...)
334 printk(KERN_CRIT
"ext4_abort called.\n");
337 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
342 if (test_opt(sb
, ERRORS_PANIC
))
343 panic("EXT4-fs panic from previous error\n");
345 if (sb
->s_flags
& MS_RDONLY
)
348 printk(KERN_CRIT
"Remounting filesystem read-only\n");
349 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
350 sb
->s_flags
|= MS_RDONLY
;
351 EXT4_SB(sb
)->s_mount_opt
|= EXT4_MOUNT_ABORT
;
352 if (EXT4_SB(sb
)->s_journal
)
353 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
356 void ext4_warning(struct super_block
*sb
, const char *function
,
357 const char *fmt
, ...)
362 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s: ",
369 void ext4_update_dynamic_rev(struct super_block
*sb
)
371 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
373 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
376 ext4_warning(sb
, __func__
,
377 "updating to rev %d because of new feature flag, "
378 "running e2fsck is recommended",
381 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
382 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
383 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
384 /* leave es->s_feature_*compat flags alone */
385 /* es->s_uuid will be set by e2fsck if empty */
388 * The rest of the superblock fields should be zero, and if not it
389 * means they are likely already in use, so leave them alone. We
390 * can leave it up to e2fsck to clean up any inconsistencies there.
395 * Open the external journal device
397 static struct block_device
*ext4_blkdev_get(dev_t dev
)
399 struct block_device
*bdev
;
400 char b
[BDEVNAME_SIZE
];
402 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
408 printk(KERN_ERR
"EXT4: failed to open journal device %s: %ld\n",
409 __bdevname(dev
, b
), PTR_ERR(bdev
));
414 * Release the journal device
416 static int ext4_blkdev_put(struct block_device
*bdev
)
419 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
422 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
424 struct block_device
*bdev
;
427 bdev
= sbi
->journal_bdev
;
429 ret
= ext4_blkdev_put(bdev
);
430 sbi
->journal_bdev
= NULL
;
435 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
437 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
440 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
444 printk(KERN_ERR
"sb orphan head is %d\n",
445 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
447 printk(KERN_ERR
"sb_info orphan list:\n");
448 list_for_each(l
, &sbi
->s_orphan
) {
449 struct inode
*inode
= orphan_list_entry(l
);
451 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
452 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
453 inode
->i_mode
, inode
->i_nlink
,
458 static void ext4_put_super(struct super_block
*sb
)
460 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
461 struct ext4_super_block
*es
= sbi
->s_es
;
465 ext4_ext_release(sb
);
466 ext4_xattr_put_super(sb
);
467 if (sbi
->s_journal
) {
468 err
= jbd2_journal_destroy(sbi
->s_journal
);
469 sbi
->s_journal
= NULL
;
471 ext4_abort(sb
, __func__
,
472 "Couldn't clean up the journal");
474 if (!(sb
->s_flags
& MS_RDONLY
)) {
475 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
476 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
477 ext4_commit_super(sb
, es
, 1);
480 remove_proc_entry("inode_readahead_blks", sbi
->s_proc
);
481 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
484 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
485 brelse(sbi
->s_group_desc
[i
]);
486 kfree(sbi
->s_group_desc
);
487 kfree(sbi
->s_flex_groups
);
488 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
489 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
490 percpu_counter_destroy(&sbi
->s_dirs_counter
);
491 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
494 for (i
= 0; i
< MAXQUOTAS
; i
++)
495 kfree(sbi
->s_qf_names
[i
]);
498 /* Debugging code just in case the in-memory inode orphan list
499 * isn't empty. The on-disk one can be non-empty if we've
500 * detected an error and taken the fs readonly, but the
501 * in-memory list had better be clean by this point. */
502 if (!list_empty(&sbi
->s_orphan
))
503 dump_orphan_list(sb
, sbi
);
504 J_ASSERT(list_empty(&sbi
->s_orphan
));
506 invalidate_bdev(sb
->s_bdev
);
507 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
509 * Invalidate the journal device's buffers. We don't want them
510 * floating about in memory - the physical journal device may
511 * hotswapped, and it breaks the `ro-after' testing code.
513 sync_blockdev(sbi
->journal_bdev
);
514 invalidate_bdev(sbi
->journal_bdev
);
515 ext4_blkdev_remove(sbi
);
517 sb
->s_fs_info
= NULL
;
522 static struct kmem_cache
*ext4_inode_cachep
;
525 * Called inside transaction, so use GFP_NOFS
527 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
529 struct ext4_inode_info
*ei
;
531 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
534 #ifdef CONFIG_EXT4_FS_POSIX_ACL
535 ei
->i_acl
= EXT4_ACL_NOT_CACHED
;
536 ei
->i_default_acl
= EXT4_ACL_NOT_CACHED
;
538 ei
->vfs_inode
.i_version
= 1;
539 ei
->vfs_inode
.i_data
.writeback_index
= 0;
540 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
541 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
542 spin_lock_init(&ei
->i_prealloc_lock
);
544 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
545 * therefore it can be null here. Don't check it, just initialize
548 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
549 ei
->i_reserved_data_blocks
= 0;
550 ei
->i_reserved_meta_blocks
= 0;
551 ei
->i_allocated_meta_blocks
= 0;
552 ei
->i_delalloc_reserved_flag
= 0;
553 spin_lock_init(&(ei
->i_block_reservation_lock
));
554 return &ei
->vfs_inode
;
557 static void ext4_destroy_inode(struct inode
*inode
)
559 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
560 printk("EXT4 Inode %p: orphan list check failed!\n",
562 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
563 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
567 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
570 static void init_once(void *foo
)
572 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
574 INIT_LIST_HEAD(&ei
->i_orphan
);
575 #ifdef CONFIG_EXT4_FS_XATTR
576 init_rwsem(&ei
->xattr_sem
);
578 init_rwsem(&ei
->i_data_sem
);
579 inode_init_once(&ei
->vfs_inode
);
582 static int init_inodecache(void)
584 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
585 sizeof(struct ext4_inode_info
),
586 0, (SLAB_RECLAIM_ACCOUNT
|
589 if (ext4_inode_cachep
== NULL
)
594 static void destroy_inodecache(void)
596 kmem_cache_destroy(ext4_inode_cachep
);
599 static void ext4_clear_inode(struct inode
*inode
)
601 #ifdef CONFIG_EXT4_FS_POSIX_ACL
602 if (EXT4_I(inode
)->i_acl
&&
603 EXT4_I(inode
)->i_acl
!= EXT4_ACL_NOT_CACHED
) {
604 posix_acl_release(EXT4_I(inode
)->i_acl
);
605 EXT4_I(inode
)->i_acl
= EXT4_ACL_NOT_CACHED
;
607 if (EXT4_I(inode
)->i_default_acl
&&
608 EXT4_I(inode
)->i_default_acl
!= EXT4_ACL_NOT_CACHED
) {
609 posix_acl_release(EXT4_I(inode
)->i_default_acl
);
610 EXT4_I(inode
)->i_default_acl
= EXT4_ACL_NOT_CACHED
;
613 ext4_discard_preallocations(inode
);
614 if (EXT4_JOURNAL(inode
))
615 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
616 &EXT4_I(inode
)->jinode
);
619 static inline void ext4_show_quota_options(struct seq_file
*seq
,
620 struct super_block
*sb
)
622 #if defined(CONFIG_QUOTA)
623 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
625 if (sbi
->s_jquota_fmt
)
626 seq_printf(seq
, ",jqfmt=%s",
627 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold" : "vfsv0");
629 if (sbi
->s_qf_names
[USRQUOTA
])
630 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
632 if (sbi
->s_qf_names
[GRPQUOTA
])
633 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
635 if (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
)
636 seq_puts(seq
, ",usrquota");
638 if (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)
639 seq_puts(seq
, ",grpquota");
645 * - it's set to a non-default value OR
646 * - if the per-sb default is different from the global default
648 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
651 unsigned long def_mount_opts
;
652 struct super_block
*sb
= vfs
->mnt_sb
;
653 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
654 struct ext4_super_block
*es
= sbi
->s_es
;
656 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
657 def_errors
= le16_to_cpu(es
->s_errors
);
659 if (sbi
->s_sb_block
!= 1)
660 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
661 if (test_opt(sb
, MINIX_DF
))
662 seq_puts(seq
, ",minixdf");
663 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
664 seq_puts(seq
, ",grpid");
665 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
666 seq_puts(seq
, ",nogrpid");
667 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
668 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
669 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
671 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
672 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
673 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
675 if (test_opt(sb
, ERRORS_RO
)) {
676 if (def_errors
== EXT4_ERRORS_PANIC
||
677 def_errors
== EXT4_ERRORS_CONTINUE
) {
678 seq_puts(seq
, ",errors=remount-ro");
681 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
682 seq_puts(seq
, ",errors=continue");
683 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
684 seq_puts(seq
, ",errors=panic");
685 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
686 seq_puts(seq
, ",nouid32");
687 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
688 seq_puts(seq
, ",debug");
689 if (test_opt(sb
, OLDALLOC
))
690 seq_puts(seq
, ",oldalloc");
691 #ifdef CONFIG_EXT4_FS_XATTR
692 if (test_opt(sb
, XATTR_USER
) &&
693 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
694 seq_puts(seq
, ",user_xattr");
695 if (!test_opt(sb
, XATTR_USER
) &&
696 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
697 seq_puts(seq
, ",nouser_xattr");
700 #ifdef CONFIG_EXT4_FS_POSIX_ACL
701 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
702 seq_puts(seq
, ",acl");
703 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
704 seq_puts(seq
, ",noacl");
706 if (!test_opt(sb
, RESERVATION
))
707 seq_puts(seq
, ",noreservation");
708 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
709 seq_printf(seq
, ",commit=%u",
710 (unsigned) (sbi
->s_commit_interval
/ HZ
));
712 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
713 seq_printf(seq
, ",min_batch_time=%u",
714 (unsigned) sbi
->s_min_batch_time
);
716 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
717 seq_printf(seq
, ",max_batch_time=%u",
718 (unsigned) sbi
->s_min_batch_time
);
722 * We're changing the default of barrier mount option, so
723 * let's always display its mount state so it's clear what its
726 seq_puts(seq
, ",barrier=");
727 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
728 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
729 seq_puts(seq
, ",journal_async_commit");
730 if (test_opt(sb
, NOBH
))
731 seq_puts(seq
, ",nobh");
732 if (!test_opt(sb
, EXTENTS
))
733 seq_puts(seq
, ",noextents");
734 if (test_opt(sb
, I_VERSION
))
735 seq_puts(seq
, ",i_version");
736 if (!test_opt(sb
, DELALLOC
))
737 seq_puts(seq
, ",nodelalloc");
741 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
743 * journal mode get enabled in different ways
744 * So just print the value even if we didn't specify it
746 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
747 seq_puts(seq
, ",data=journal");
748 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
749 seq_puts(seq
, ",data=ordered");
750 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
751 seq_puts(seq
, ",data=writeback");
753 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
754 seq_printf(seq
, ",inode_readahead_blks=%u",
755 sbi
->s_inode_readahead_blks
);
757 if (test_opt(sb
, DATA_ERR_ABORT
))
758 seq_puts(seq
, ",data_err=abort");
760 ext4_show_quota_options(seq
, sb
);
765 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
766 u64 ino
, u32 generation
)
770 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
771 return ERR_PTR(-ESTALE
);
772 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
773 return ERR_PTR(-ESTALE
);
775 /* iget isn't really right if the inode is currently unallocated!!
777 * ext4_read_inode will return a bad_inode if the inode had been
778 * deleted, so we should be safe.
780 * Currently we don't know the generation for parent directory, so
781 * a generation of 0 means "accept any"
783 inode
= ext4_iget(sb
, ino
);
785 return ERR_CAST(inode
);
786 if (generation
&& inode
->i_generation
!= generation
) {
788 return ERR_PTR(-ESTALE
);
794 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
795 int fh_len
, int fh_type
)
797 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
801 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
802 int fh_len
, int fh_type
)
804 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
809 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
810 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
812 static int ext4_dquot_initialize(struct inode
*inode
, int type
);
813 static int ext4_dquot_drop(struct inode
*inode
);
814 static int ext4_write_dquot(struct dquot
*dquot
);
815 static int ext4_acquire_dquot(struct dquot
*dquot
);
816 static int ext4_release_dquot(struct dquot
*dquot
);
817 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
818 static int ext4_write_info(struct super_block
*sb
, int type
);
819 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
820 char *path
, int remount
);
821 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
822 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
823 size_t len
, loff_t off
);
824 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
825 const char *data
, size_t len
, loff_t off
);
827 static struct dquot_operations ext4_quota_operations
= {
828 .initialize
= ext4_dquot_initialize
,
829 .drop
= ext4_dquot_drop
,
830 .alloc_space
= dquot_alloc_space
,
831 .alloc_inode
= dquot_alloc_inode
,
832 .free_space
= dquot_free_space
,
833 .free_inode
= dquot_free_inode
,
834 .transfer
= dquot_transfer
,
835 .write_dquot
= ext4_write_dquot
,
836 .acquire_dquot
= ext4_acquire_dquot
,
837 .release_dquot
= ext4_release_dquot
,
838 .mark_dirty
= ext4_mark_dquot_dirty
,
839 .write_info
= ext4_write_info
842 static struct quotactl_ops ext4_qctl_operations
= {
843 .quota_on
= ext4_quota_on
,
844 .quota_off
= vfs_quota_off
,
845 .quota_sync
= vfs_quota_sync
,
846 .get_info
= vfs_get_dqinfo
,
847 .set_info
= vfs_set_dqinfo
,
848 .get_dqblk
= vfs_get_dqblk
,
849 .set_dqblk
= vfs_set_dqblk
853 static const struct super_operations ext4_sops
= {
854 .alloc_inode
= ext4_alloc_inode
,
855 .destroy_inode
= ext4_destroy_inode
,
856 .write_inode
= ext4_write_inode
,
857 .dirty_inode
= ext4_dirty_inode
,
858 .delete_inode
= ext4_delete_inode
,
859 .put_super
= ext4_put_super
,
860 .write_super
= ext4_write_super
,
861 .sync_fs
= ext4_sync_fs
,
862 .write_super_lockfs
= ext4_write_super_lockfs
,
863 .unlockfs
= ext4_unlockfs
,
864 .statfs
= ext4_statfs
,
865 .remount_fs
= ext4_remount
,
866 .clear_inode
= ext4_clear_inode
,
867 .show_options
= ext4_show_options
,
869 .quota_read
= ext4_quota_read
,
870 .quota_write
= ext4_quota_write
,
874 static const struct export_operations ext4_export_ops
= {
875 .fh_to_dentry
= ext4_fh_to_dentry
,
876 .fh_to_parent
= ext4_fh_to_parent
,
877 .get_parent
= ext4_get_parent
,
881 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
882 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
883 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
884 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
885 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
886 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
887 Opt_journal_update
, Opt_journal_inum
, Opt_journal_dev
,
888 Opt_journal_checksum
, Opt_journal_async_commit
,
889 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
890 Opt_data_err_abort
, Opt_data_err_ignore
,
891 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
892 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
893 Opt_ignore
, Opt_barrier
, Opt_err
, Opt_resize
, Opt_usrquota
,
894 Opt_grpquota
, Opt_extents
, Opt_noextents
, Opt_i_version
,
895 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
896 Opt_inode_readahead_blks
899 static const match_table_t tokens
= {
900 {Opt_bsd_df
, "bsddf"},
901 {Opt_minix_df
, "minixdf"},
902 {Opt_grpid
, "grpid"},
903 {Opt_grpid
, "bsdgroups"},
904 {Opt_nogrpid
, "nogrpid"},
905 {Opt_nogrpid
, "sysvgroups"},
906 {Opt_resgid
, "resgid=%u"},
907 {Opt_resuid
, "resuid=%u"},
909 {Opt_err_cont
, "errors=continue"},
910 {Opt_err_panic
, "errors=panic"},
911 {Opt_err_ro
, "errors=remount-ro"},
912 {Opt_nouid32
, "nouid32"},
913 {Opt_debug
, "debug"},
914 {Opt_oldalloc
, "oldalloc"},
915 {Opt_orlov
, "orlov"},
916 {Opt_user_xattr
, "user_xattr"},
917 {Opt_nouser_xattr
, "nouser_xattr"},
919 {Opt_noacl
, "noacl"},
920 {Opt_reservation
, "reservation"},
921 {Opt_noreservation
, "noreservation"},
922 {Opt_noload
, "noload"},
925 {Opt_commit
, "commit=%u"},
926 {Opt_min_batch_time
, "min_batch_time=%u"},
927 {Opt_max_batch_time
, "max_batch_time=%u"},
928 {Opt_journal_update
, "journal=update"},
929 {Opt_journal_inum
, "journal=%u"},
930 {Opt_journal_dev
, "journal_dev=%u"},
931 {Opt_journal_checksum
, "journal_checksum"},
932 {Opt_journal_async_commit
, "journal_async_commit"},
933 {Opt_abort
, "abort"},
934 {Opt_data_journal
, "data=journal"},
935 {Opt_data_ordered
, "data=ordered"},
936 {Opt_data_writeback
, "data=writeback"},
937 {Opt_data_err_abort
, "data_err=abort"},
938 {Opt_data_err_ignore
, "data_err=ignore"},
939 {Opt_offusrjquota
, "usrjquota="},
940 {Opt_usrjquota
, "usrjquota=%s"},
941 {Opt_offgrpjquota
, "grpjquota="},
942 {Opt_grpjquota
, "grpjquota=%s"},
943 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
944 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
945 {Opt_grpquota
, "grpquota"},
946 {Opt_noquota
, "noquota"},
947 {Opt_quota
, "quota"},
948 {Opt_usrquota
, "usrquota"},
949 {Opt_barrier
, "barrier=%u"},
950 {Opt_extents
, "extents"},
951 {Opt_noextents
, "noextents"},
952 {Opt_i_version
, "i_version"},
953 {Opt_stripe
, "stripe=%u"},
954 {Opt_resize
, "resize"},
955 {Opt_delalloc
, "delalloc"},
956 {Opt_nodelalloc
, "nodelalloc"},
957 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
961 static ext4_fsblk_t
get_sb_block(void **data
)
963 ext4_fsblk_t sb_block
;
964 char *options
= (char *) *data
;
966 if (!options
|| strncmp(options
, "sb=", 3) != 0)
967 return 1; /* Default location */
969 /*todo: use simple_strtoll with >32bit ext4 */
970 sb_block
= simple_strtoul(options
, &options
, 0);
971 if (*options
&& *options
!= ',') {
972 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
978 *data
= (void *) options
;
982 static int parse_options(char *options
, struct super_block
*sb
,
983 unsigned int *inum
, unsigned long *journal_devnum
,
984 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
986 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
988 substring_t args
[MAX_OPT_ARGS
];
995 ext4_fsblk_t last_block
;
1000 while ((p
= strsep(&options
, ",")) != NULL
) {
1005 token
= match_token(p
, tokens
, args
);
1008 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1011 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1014 set_opt(sbi
->s_mount_opt
, GRPID
);
1017 clear_opt(sbi
->s_mount_opt
, GRPID
);
1020 if (match_int(&args
[0], &option
))
1022 sbi
->s_resuid
= option
;
1025 if (match_int(&args
[0], &option
))
1027 sbi
->s_resgid
= option
;
1030 /* handled by get_sb_block() instead of here */
1031 /* *sb_block = match_int(&args[0]); */
1034 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1035 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1036 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1039 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1040 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1041 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1044 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1045 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1046 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1049 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1052 set_opt(sbi
->s_mount_opt
, DEBUG
);
1055 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1058 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1060 #ifdef CONFIG_EXT4_FS_XATTR
1061 case Opt_user_xattr
:
1062 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1064 case Opt_nouser_xattr
:
1065 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1068 case Opt_user_xattr
:
1069 case Opt_nouser_xattr
:
1070 printk(KERN_ERR
"EXT4 (no)user_xattr options "
1074 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1076 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1079 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1084 printk(KERN_ERR
"EXT4 (no)acl options "
1088 case Opt_reservation
:
1089 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1091 case Opt_noreservation
:
1092 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
1094 case Opt_journal_update
:
1096 /* Eventually we will want to be able to create
1097 a journal file here. For now, only allow the
1098 user to specify an existing inode to be the
1101 printk(KERN_ERR
"EXT4-fs: cannot specify "
1102 "journal on remount\n");
1105 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1107 case Opt_journal_inum
:
1109 printk(KERN_ERR
"EXT4-fs: cannot specify "
1110 "journal on remount\n");
1113 if (match_int(&args
[0], &option
))
1117 case Opt_journal_dev
:
1119 printk(KERN_ERR
"EXT4-fs: cannot specify "
1120 "journal on remount\n");
1123 if (match_int(&args
[0], &option
))
1125 *journal_devnum
= option
;
1127 case Opt_journal_checksum
:
1128 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1130 case Opt_journal_async_commit
:
1131 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1132 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1135 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1138 if (match_int(&args
[0], &option
))
1143 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1144 sbi
->s_commit_interval
= HZ
* option
;
1146 case Opt_max_batch_time
:
1147 if (match_int(&args
[0], &option
))
1152 option
= EXT4_DEF_MAX_BATCH_TIME
;
1153 sbi
->s_max_batch_time
= option
;
1155 case Opt_min_batch_time
:
1156 if (match_int(&args
[0], &option
))
1160 sbi
->s_min_batch_time
= option
;
1162 case Opt_data_journal
:
1163 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1165 case Opt_data_ordered
:
1166 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1168 case Opt_data_writeback
:
1169 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1172 if ((sbi
->s_mount_opt
& EXT4_MOUNT_DATA_FLAGS
)
1175 "EXT4-fs: cannot change data "
1176 "mode on remount\n");
1180 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DATA_FLAGS
;
1181 sbi
->s_mount_opt
|= data_opt
;
1184 case Opt_data_err_abort
:
1185 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1187 case Opt_data_err_ignore
:
1188 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1197 if ((sb_any_quota_enabled(sb
) ||
1198 sb_any_quota_suspended(sb
)) &&
1199 !sbi
->s_qf_names
[qtype
]) {
1201 "EXT4-fs: Cannot change journaled "
1202 "quota options when quota turned on.\n");
1205 qname
= match_strdup(&args
[0]);
1208 "EXT4-fs: not enough memory for "
1209 "storing quotafile name.\n");
1212 if (sbi
->s_qf_names
[qtype
] &&
1213 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1215 "EXT4-fs: %s quota file already "
1216 "specified.\n", QTYPE2NAME(qtype
));
1220 sbi
->s_qf_names
[qtype
] = qname
;
1221 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1223 "EXT4-fs: quotafile must be on "
1224 "filesystem root.\n");
1225 kfree(sbi
->s_qf_names
[qtype
]);
1226 sbi
->s_qf_names
[qtype
] = NULL
;
1229 set_opt(sbi
->s_mount_opt
, QUOTA
);
1231 case Opt_offusrjquota
:
1234 case Opt_offgrpjquota
:
1237 if ((sb_any_quota_enabled(sb
) ||
1238 sb_any_quota_suspended(sb
)) &&
1239 sbi
->s_qf_names
[qtype
]) {
1240 printk(KERN_ERR
"EXT4-fs: Cannot change "
1241 "journaled quota options when "
1242 "quota turned on.\n");
1246 * The space will be released later when all options
1247 * are confirmed to be correct
1249 sbi
->s_qf_names
[qtype
] = NULL
;
1251 case Opt_jqfmt_vfsold
:
1252 qfmt
= QFMT_VFS_OLD
;
1254 case Opt_jqfmt_vfsv0
:
1257 if ((sb_any_quota_enabled(sb
) ||
1258 sb_any_quota_suspended(sb
)) &&
1259 sbi
->s_jquota_fmt
!= qfmt
) {
1260 printk(KERN_ERR
"EXT4-fs: Cannot change "
1261 "journaled quota options when "
1262 "quota turned on.\n");
1265 sbi
->s_jquota_fmt
= qfmt
;
1269 set_opt(sbi
->s_mount_opt
, QUOTA
);
1270 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1273 set_opt(sbi
->s_mount_opt
, QUOTA
);
1274 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1277 if (sb_any_quota_enabled(sb
)) {
1278 printk(KERN_ERR
"EXT4-fs: Cannot change quota "
1279 "options when quota turned on.\n");
1282 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1283 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1284 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1291 "EXT4-fs: quota options not supported.\n");
1295 case Opt_offusrjquota
:
1296 case Opt_offgrpjquota
:
1297 case Opt_jqfmt_vfsold
:
1298 case Opt_jqfmt_vfsv0
:
1300 "EXT4-fs: journaled quota options not "
1307 set_opt(sbi
->s_mount_opt
, ABORT
);
1310 if (match_int(&args
[0], &option
))
1313 set_opt(sbi
->s_mount_opt
, BARRIER
);
1315 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1321 printk("EXT4-fs: resize option only available "
1325 if (match_int(&args
[0], &option
) != 0)
1327 *n_blocks_count
= option
;
1330 set_opt(sbi
->s_mount_opt
, NOBH
);
1333 clear_opt(sbi
->s_mount_opt
, NOBH
);
1336 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
,
1337 EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
1338 ext4_warning(sb
, __func__
,
1339 "extents feature not enabled "
1340 "on this filesystem, use tune2fs");
1343 set_opt(sbi
->s_mount_opt
, EXTENTS
);
1347 * When e2fsprogs support resizing an already existing
1348 * ext3 file system to greater than 2**32 we need to
1349 * add support to block allocator to handle growing
1350 * already existing block mapped inode so that blocks
1351 * allocated for them fall within 2**32
1353 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1354 if (last_block
> 0xffffffffULL
) {
1355 printk(KERN_ERR
"EXT4-fs: Filesystem too "
1356 "large to mount with "
1357 "-o noextents options\n");
1360 clear_opt(sbi
->s_mount_opt
, EXTENTS
);
1363 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1364 sb
->s_flags
|= MS_I_VERSION
;
1366 case Opt_nodelalloc
:
1367 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1370 if (match_int(&args
[0], &option
))
1374 sbi
->s_stripe
= option
;
1377 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1379 case Opt_inode_readahead_blks
:
1380 if (match_int(&args
[0], &option
))
1382 if (option
< 0 || option
> (1 << 30))
1384 sbi
->s_inode_readahead_blks
= option
;
1388 "EXT4-fs: Unrecognized mount option \"%s\" "
1389 "or missing value\n", p
);
1394 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1395 if ((sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
) &&
1396 sbi
->s_qf_names
[USRQUOTA
])
1397 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1399 if ((sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
) &&
1400 sbi
->s_qf_names
[GRPQUOTA
])
1401 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1403 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1404 (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)) ||
1405 (sbi
->s_qf_names
[GRPQUOTA
] &&
1406 (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
))) {
1407 printk(KERN_ERR
"EXT4-fs: old and new quota "
1408 "format mixing.\n");
1412 if (!sbi
->s_jquota_fmt
) {
1413 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1414 "not specified.\n");
1418 if (sbi
->s_jquota_fmt
) {
1419 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1420 "specified with no journaling "
1429 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1432 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1435 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1436 printk(KERN_ERR
"EXT4-fs warning: revision level too high, "
1437 "forcing read-only mode\n");
1442 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1443 printk(KERN_WARNING
"EXT4-fs warning: mounting unchecked fs, "
1444 "running e2fsck is recommended\n");
1445 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1447 "EXT4-fs warning: mounting fs with errors, "
1448 "running e2fsck is recommended\n");
1449 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1450 le16_to_cpu(es
->s_mnt_count
) >=
1451 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1453 "EXT4-fs warning: maximal mount count reached, "
1454 "running e2fsck is recommended\n");
1455 else if (le32_to_cpu(es
->s_checkinterval
) &&
1456 (le32_to_cpu(es
->s_lastcheck
) +
1457 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1459 "EXT4-fs warning: checktime reached, "
1460 "running e2fsck is recommended\n");
1461 if (!sbi
->s_journal
)
1462 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1463 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1464 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1465 le16_add_cpu(&es
->s_mnt_count
, 1);
1466 es
->s_mtime
= cpu_to_le32(get_seconds());
1467 ext4_update_dynamic_rev(sb
);
1469 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1471 ext4_commit_super(sb
, es
, 1);
1472 if (test_opt(sb
, DEBUG
))
1473 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1474 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1476 sbi
->s_groups_count
,
1477 EXT4_BLOCKS_PER_GROUP(sb
),
1478 EXT4_INODES_PER_GROUP(sb
),
1481 if (EXT4_SB(sb
)->s_journal
) {
1482 printk(KERN_INFO
"EXT4 FS on %s, %s journal on %s\n",
1483 sb
->s_id
, EXT4_SB(sb
)->s_journal
->j_inode
? "internal" :
1484 "external", EXT4_SB(sb
)->s_journal
->j_devname
);
1486 printk(KERN_INFO
"EXT4 FS on %s, no journal\n", sb
->s_id
);
1491 static int ext4_fill_flex_info(struct super_block
*sb
)
1493 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1494 struct ext4_group_desc
*gdp
= NULL
;
1495 struct buffer_head
*bh
;
1496 ext4_group_t flex_group_count
;
1497 ext4_group_t flex_group
;
1498 int groups_per_flex
= 0;
1501 if (!sbi
->s_es
->s_log_groups_per_flex
) {
1502 sbi
->s_log_groups_per_flex
= 0;
1506 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1507 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1509 /* We allocate both existing and potentially added groups */
1510 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1511 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1512 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1513 sbi
->s_flex_groups
= kzalloc(flex_group_count
*
1514 sizeof(struct flex_groups
), GFP_KERNEL
);
1515 if (sbi
->s_flex_groups
== NULL
) {
1516 printk(KERN_ERR
"EXT4-fs: not enough memory for "
1517 "%u flex groups\n", flex_group_count
);
1521 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1522 gdp
= ext4_get_group_desc(sb
, i
, &bh
);
1524 flex_group
= ext4_flex_group(sbi
, i
);
1525 sbi
->s_flex_groups
[flex_group
].free_inodes
+=
1526 le16_to_cpu(gdp
->bg_free_inodes_count
);
1527 sbi
->s_flex_groups
[flex_group
].free_blocks
+=
1528 le16_to_cpu(gdp
->bg_free_blocks_count
);
1536 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1537 struct ext4_group_desc
*gdp
)
1541 if (sbi
->s_es
->s_feature_ro_compat
&
1542 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1543 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1544 __le32 le_group
= cpu_to_le32(block_group
);
1546 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1547 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1548 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1549 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1550 /* for checksum of struct ext4_group_desc do the rest...*/
1551 if ((sbi
->s_es
->s_feature_incompat
&
1552 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1553 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1554 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1555 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1559 return cpu_to_le16(crc
);
1562 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1563 struct ext4_group_desc
*gdp
)
1565 if ((sbi
->s_es
->s_feature_ro_compat
&
1566 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1567 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1573 /* Called at mount-time, super-block is locked */
1574 static int ext4_check_descriptors(struct super_block
*sb
)
1576 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1577 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1578 ext4_fsblk_t last_block
;
1579 ext4_fsblk_t block_bitmap
;
1580 ext4_fsblk_t inode_bitmap
;
1581 ext4_fsblk_t inode_table
;
1582 int flexbg_flag
= 0;
1585 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1588 ext4_debug("Checking group descriptors");
1590 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1591 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1593 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1594 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1596 last_block
= first_block
+
1597 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1599 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1600 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1601 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1602 "Block bitmap for group %u not in group "
1603 "(block %llu)!\n", i
, block_bitmap
);
1606 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1607 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1608 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1609 "Inode bitmap for group %u not in group "
1610 "(block %llu)!\n", i
, inode_bitmap
);
1613 inode_table
= ext4_inode_table(sb
, gdp
);
1614 if (inode_table
< first_block
||
1615 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1616 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1617 "Inode table for group %u not in group "
1618 "(block %llu)!\n", i
, inode_table
);
1621 spin_lock(sb_bgl_lock(sbi
, i
));
1622 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1623 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1624 "Checksum for group %u failed (%u!=%u)\n",
1625 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1626 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1627 if (!(sb
->s_flags
& MS_RDONLY
)) {
1628 spin_unlock(sb_bgl_lock(sbi
, i
));
1632 spin_unlock(sb_bgl_lock(sbi
, i
));
1634 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1637 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1638 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(ext4_count_free_inodes(sb
));
1642 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1643 * the superblock) which were deleted from all directories, but held open by
1644 * a process at the time of a crash. We walk the list and try to delete these
1645 * inodes at recovery time (only with a read-write filesystem).
1647 * In order to keep the orphan inode chain consistent during traversal (in
1648 * case of crash during recovery), we link each inode into the superblock
1649 * orphan list_head and handle it the same way as an inode deletion during
1650 * normal operation (which journals the operations for us).
1652 * We only do an iget() and an iput() on each inode, which is very safe if we
1653 * accidentally point at an in-use or already deleted inode. The worst that
1654 * can happen in this case is that we get a "bit already cleared" message from
1655 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1656 * e2fsck was run on this filesystem, and it must have already done the orphan
1657 * inode cleanup for us, so we can safely abort without any further action.
1659 static void ext4_orphan_cleanup(struct super_block
*sb
,
1660 struct ext4_super_block
*es
)
1662 unsigned int s_flags
= sb
->s_flags
;
1663 int nr_orphans
= 0, nr_truncates
= 0;
1667 if (!es
->s_last_orphan
) {
1668 jbd_debug(4, "no orphan inodes to clean up\n");
1672 if (bdev_read_only(sb
->s_bdev
)) {
1673 printk(KERN_ERR
"EXT4-fs: write access "
1674 "unavailable, skipping orphan cleanup.\n");
1678 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1679 if (es
->s_last_orphan
)
1680 jbd_debug(1, "Errors on filesystem, "
1681 "clearing orphan list.\n");
1682 es
->s_last_orphan
= 0;
1683 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1687 if (s_flags
& MS_RDONLY
) {
1688 printk(KERN_INFO
"EXT4-fs: %s: orphan cleanup on readonly fs\n",
1690 sb
->s_flags
&= ~MS_RDONLY
;
1693 /* Needed for iput() to work correctly and not trash data */
1694 sb
->s_flags
|= MS_ACTIVE
;
1695 /* Turn on quotas so that they are updated correctly */
1696 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1697 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
1698 int ret
= ext4_quota_on_mount(sb
, i
);
1701 "EXT4-fs: Cannot turn on journaled "
1702 "quota: error %d\n", ret
);
1707 while (es
->s_last_orphan
) {
1708 struct inode
*inode
;
1710 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1711 if (IS_ERR(inode
)) {
1712 es
->s_last_orphan
= 0;
1716 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
1718 if (inode
->i_nlink
) {
1720 "%s: truncating inode %lu to %lld bytes\n",
1721 __func__
, inode
->i_ino
, inode
->i_size
);
1722 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1723 inode
->i_ino
, inode
->i_size
);
1724 ext4_truncate(inode
);
1728 "%s: deleting unreferenced inode %lu\n",
1729 __func__
, inode
->i_ino
);
1730 jbd_debug(2, "deleting unreferenced inode %lu\n",
1734 iput(inode
); /* The delete magic happens here! */
1737 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1740 printk(KERN_INFO
"EXT4-fs: %s: %d orphan inode%s deleted\n",
1741 sb
->s_id
, PLURAL(nr_orphans
));
1743 printk(KERN_INFO
"EXT4-fs: %s: %d truncate%s cleaned up\n",
1744 sb
->s_id
, PLURAL(nr_truncates
));
1746 /* Turn quotas off */
1747 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1748 if (sb_dqopt(sb
)->files
[i
])
1749 vfs_quota_off(sb
, i
, 0);
1752 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1755 * Maximal extent format file size.
1756 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1757 * extent format containers, within a sector_t, and within i_blocks
1758 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1759 * so that won't be a limiting factor.
1761 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1763 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
1766 loff_t upper_limit
= MAX_LFS_FILESIZE
;
1768 /* small i_blocks in vfs inode? */
1769 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
1771 * CONFIG_LBD is not enabled implies the inode
1772 * i_block represent total blocks in 512 bytes
1773 * 32 == size of vfs inode i_blocks * 8
1775 upper_limit
= (1LL << 32) - 1;
1777 /* total blocks in file system block size */
1778 upper_limit
>>= (blkbits
- 9);
1779 upper_limit
<<= blkbits
;
1782 /* 32-bit extent-start container, ee_block */
1787 /* Sanity check against vm- & vfs- imposed limits */
1788 if (res
> upper_limit
)
1795 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1796 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1797 * We need to be 1 filesystem block less than the 2^48 sector limit.
1799 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
1801 loff_t res
= EXT4_NDIR_BLOCKS
;
1804 /* This is calculated to be the largest file size for a
1805 * dense, bitmapped file such that the total number of
1806 * sectors in the file, including data and all indirect blocks,
1807 * does not exceed 2^48 -1
1808 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1809 * total number of 512 bytes blocks of the file
1812 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
1814 * !has_huge_files or CONFIG_LBD is not enabled
1815 * implies the inode i_block represent total blocks in
1816 * 512 bytes 32 == size of vfs inode i_blocks * 8
1818 upper_limit
= (1LL << 32) - 1;
1820 /* total blocks in file system block size */
1821 upper_limit
>>= (bits
- 9);
1825 * We use 48 bit ext4_inode i_blocks
1826 * With EXT4_HUGE_FILE_FL set the i_blocks
1827 * represent total number of blocks in
1828 * file system block size
1830 upper_limit
= (1LL << 48) - 1;
1834 /* indirect blocks */
1836 /* double indirect blocks */
1837 meta_blocks
+= 1 + (1LL << (bits
-2));
1838 /* tripple indirect blocks */
1839 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
1841 upper_limit
-= meta_blocks
;
1842 upper_limit
<<= bits
;
1844 res
+= 1LL << (bits
-2);
1845 res
+= 1LL << (2*(bits
-2));
1846 res
+= 1LL << (3*(bits
-2));
1848 if (res
> upper_limit
)
1851 if (res
> MAX_LFS_FILESIZE
)
1852 res
= MAX_LFS_FILESIZE
;
1857 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
1858 ext4_fsblk_t logical_sb_block
, int nr
)
1860 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1861 ext4_group_t bg
, first_meta_bg
;
1864 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1866 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
1868 return logical_sb_block
+ nr
+ 1;
1869 bg
= sbi
->s_desc_per_block
* nr
;
1870 if (ext4_bg_has_super(sb
, bg
))
1872 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
1876 * ext4_get_stripe_size: Get the stripe size.
1877 * @sbi: In memory super block info
1879 * If we have specified it via mount option, then
1880 * use the mount option value. If the value specified at mount time is
1881 * greater than the blocks per group use the super block value.
1882 * If the super block value is greater than blocks per group return 0.
1883 * Allocator needs it be less than blocks per group.
1886 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
1888 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
1889 unsigned long stripe_width
=
1890 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
1892 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
1893 return sbi
->s_stripe
;
1895 if (stripe_width
<= sbi
->s_blocks_per_group
)
1896 return stripe_width
;
1898 if (stride
<= sbi
->s_blocks_per_group
)
1904 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
1905 __releases(kernel_lock
)
1906 __acquires(kernel_lock
)
1909 struct buffer_head
*bh
;
1910 struct ext4_super_block
*es
= NULL
;
1911 struct ext4_sb_info
*sbi
;
1913 ext4_fsblk_t sb_block
= get_sb_block(&data
);
1914 ext4_fsblk_t logical_sb_block
;
1915 unsigned long offset
= 0;
1916 unsigned int journal_inum
= 0;
1917 unsigned long journal_devnum
= 0;
1918 unsigned long def_mount_opts
;
1926 int needs_recovery
, has_huge_files
;
1931 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
1934 sb
->s_fs_info
= sbi
;
1935 sbi
->s_mount_opt
= 0;
1936 sbi
->s_resuid
= EXT4_DEF_RESUID
;
1937 sbi
->s_resgid
= EXT4_DEF_RESGID
;
1938 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
1939 sbi
->s_sb_block
= sb_block
;
1943 /* Cleanup superblock name */
1944 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
1947 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
1949 printk(KERN_ERR
"EXT4-fs: unable to set blocksize\n");
1954 * The ext4 superblock will not be buffer aligned for other than 1kB
1955 * block sizes. We need to calculate the offset from buffer start.
1957 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
1958 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
1959 offset
= do_div(logical_sb_block
, blocksize
);
1961 logical_sb_block
= sb_block
;
1964 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
1965 printk(KERN_ERR
"EXT4-fs: unable to read superblock\n");
1969 * Note: s_es must be initialized as soon as possible because
1970 * some ext4 macro-instructions depend on its value
1972 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
1974 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1975 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
1978 /* Set defaults before we parse the mount options */
1979 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1980 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
1981 set_opt(sbi
->s_mount_opt
, DEBUG
);
1982 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
1983 set_opt(sbi
->s_mount_opt
, GRPID
);
1984 if (def_mount_opts
& EXT4_DEFM_UID16
)
1985 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1986 #ifdef CONFIG_EXT4_FS_XATTR
1987 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
1988 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1990 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1991 if (def_mount_opts
& EXT4_DEFM_ACL
)
1992 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1994 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
1995 sbi
->s_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
1996 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
1997 sbi
->s_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
1998 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
1999 sbi
->s_mount_opt
|= EXT4_MOUNT_WRITEBACK_DATA
;
2001 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2002 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2003 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2004 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2006 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2008 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2009 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2010 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
2011 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
2012 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
2014 set_opt(sbi
->s_mount_opt
, RESERVATION
);
2015 set_opt(sbi
->s_mount_opt
, BARRIER
);
2018 * turn on extents feature by default in ext4 filesystem
2019 * only if feature flag already set by mkfs or tune2fs.
2020 * Use -o noextents to turn it off
2022 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
))
2023 set_opt(sbi
->s_mount_opt
, EXTENTS
);
2025 ext4_warning(sb
, __func__
,
2026 "extents feature not enabled on this filesystem, "
2030 * enable delayed allocation by default
2031 * Use -o nodelalloc to turn it off
2033 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2036 if (!parse_options((char *) data
, sb
, &journal_inum
, &journal_devnum
,
2040 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2041 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2043 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2044 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2045 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2046 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2048 "EXT4-fs warning: feature flags set on rev 0 fs, "
2049 "running e2fsck is recommended\n");
2052 * Check feature flags regardless of the revision level, since we
2053 * previously didn't change the revision level when setting the flags,
2054 * so there is a chance incompat flags are set on a rev 0 filesystem.
2056 features
= EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
);
2058 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount because of "
2059 "unsupported optional features (%x).\n",
2060 sb
->s_id
, le32_to_cpu(features
));
2063 features
= EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
);
2064 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
2065 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount RDWR because of "
2066 "unsupported optional features (%x).\n",
2067 sb
->s_id
, le32_to_cpu(features
));
2070 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2071 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
2072 if (has_huge_files
) {
2074 * Large file size enabled file system can only be
2075 * mount if kernel is build with CONFIG_LBD
2077 if (sizeof(root
->i_blocks
) < sizeof(u64
) &&
2078 !(sb
->s_flags
& MS_RDONLY
)) {
2079 printk(KERN_ERR
"EXT4-fs: %s: Filesystem with huge "
2080 "files cannot be mounted read-write "
2081 "without CONFIG_LBD.\n", sb
->s_id
);
2085 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2087 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2088 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2090 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2091 blocksize
, sb
->s_id
);
2095 if (sb
->s_blocksize
!= blocksize
) {
2097 /* Validate the filesystem blocksize */
2098 if (!sb_set_blocksize(sb
, blocksize
)) {
2099 printk(KERN_ERR
"EXT4-fs: bad block size %d.\n",
2105 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2106 offset
= do_div(logical_sb_block
, blocksize
);
2107 bh
= sb_bread(sb
, logical_sb_block
);
2110 "EXT4-fs: Can't read superblock on 2nd try.\n");
2113 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2115 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2117 "EXT4-fs: Magic mismatch, very weird !\n");
2122 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
2124 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
2126 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2127 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2128 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2130 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2131 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2132 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2133 (!is_power_of_2(sbi
->s_inode_size
)) ||
2134 (sbi
->s_inode_size
> blocksize
)) {
2136 "EXT4-fs: unsupported inode size: %d\n",
2140 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2141 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2143 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2144 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2145 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2146 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2147 !is_power_of_2(sbi
->s_desc_size
)) {
2149 "EXT4-fs: unsupported descriptor size %lu\n",
2154 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2155 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2156 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2157 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2159 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2160 if (sbi
->s_inodes_per_block
== 0)
2162 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2163 sbi
->s_inodes_per_block
;
2164 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2166 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2167 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2168 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2169 for (i
= 0; i
< 4; i
++)
2170 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2171 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2172 i
= le32_to_cpu(es
->s_flags
);
2173 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
2174 sbi
->s_hash_unsigned
= 3;
2175 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
2176 #ifdef __CHAR_UNSIGNED__
2177 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
2178 sbi
->s_hash_unsigned
= 3;
2180 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
2185 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2187 "EXT4-fs: #blocks per group too big: %lu\n",
2188 sbi
->s_blocks_per_group
);
2191 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2193 "EXT4-fs: #inodes per group too big: %lu\n",
2194 sbi
->s_inodes_per_group
);
2198 if (ext4_blocks_count(es
) >
2199 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) {
2200 printk(KERN_ERR
"EXT4-fs: filesystem on %s:"
2201 " too large to mount safely\n", sb
->s_id
);
2202 if (sizeof(sector_t
) < 8)
2203 printk(KERN_WARNING
"EXT4-fs: CONFIG_LBD not "
2208 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2211 /* ensure blocks_count calculation below doesn't sign-extend */
2212 if (ext4_blocks_count(es
) + EXT4_BLOCKS_PER_GROUP(sb
) <
2213 le32_to_cpu(es
->s_first_data_block
) + 1) {
2214 printk(KERN_WARNING
"EXT4-fs: bad geometry: block count %llu, "
2215 "first data block %u, blocks per group %lu\n",
2216 ext4_blocks_count(es
),
2217 le32_to_cpu(es
->s_first_data_block
),
2218 EXT4_BLOCKS_PER_GROUP(sb
));
2221 blocks_count
= (ext4_blocks_count(es
) -
2222 le32_to_cpu(es
->s_first_data_block
) +
2223 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2224 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2225 sbi
->s_groups_count
= blocks_count
;
2226 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2227 EXT4_DESC_PER_BLOCK(sb
);
2228 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2230 if (sbi
->s_group_desc
== NULL
) {
2231 printk(KERN_ERR
"EXT4-fs: not enough memory\n");
2235 #ifdef CONFIG_PROC_FS
2237 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2240 proc_create_data("inode_readahead_blks", 0644, sbi
->s_proc
,
2242 &sbi
->s_inode_readahead_blks
);
2245 bgl_lock_init(&sbi
->s_blockgroup_lock
);
2247 for (i
= 0; i
< db_count
; i
++) {
2248 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2249 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2250 if (!sbi
->s_group_desc
[i
]) {
2251 printk(KERN_ERR
"EXT4-fs: "
2252 "can't read group descriptor %d\n", i
);
2257 if (!ext4_check_descriptors(sb
)) {
2258 printk(KERN_ERR
"EXT4-fs: group descriptors corrupted!\n");
2261 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2262 if (!ext4_fill_flex_info(sb
)) {
2264 "EXT4-fs: unable to initialize "
2265 "flex_bg meta info!\n");
2269 sbi
->s_gdb_count
= db_count
;
2270 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2271 spin_lock_init(&sbi
->s_next_gen_lock
);
2273 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2274 ext4_count_free_blocks(sb
));
2276 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2277 ext4_count_free_inodes(sb
));
2280 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2281 ext4_count_dirs(sb
));
2284 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
2287 printk(KERN_ERR
"EXT4-fs: insufficient memory\n");
2291 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2294 * set up enough so that it can read an inode
2296 sb
->s_op
= &ext4_sops
;
2297 sb
->s_export_op
= &ext4_export_ops
;
2298 sb
->s_xattr
= ext4_xattr_handlers
;
2300 sb
->s_qcop
= &ext4_qctl_operations
;
2301 sb
->dq_op
= &ext4_quota_operations
;
2303 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2307 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2308 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2309 EXT4_FEATURE_INCOMPAT_RECOVER
));
2312 * The first inode we look at is the journal inode. Don't try
2313 * root first: it may be modified in the journal!
2315 if (!test_opt(sb
, NOLOAD
) &&
2316 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2317 if (ext4_load_journal(sb
, es
, journal_devnum
))
2319 if (!(sb
->s_flags
& MS_RDONLY
) &&
2320 EXT4_SB(sb
)->s_journal
->j_failed_commit
) {
2321 printk(KERN_CRIT
"EXT4-fs error (device %s): "
2322 "ext4_fill_super: Journal transaction "
2323 "%u is corrupt\n", sb
->s_id
,
2324 EXT4_SB(sb
)->s_journal
->j_failed_commit
);
2325 if (test_opt(sb
, ERRORS_RO
)) {
2327 "Mounting filesystem read-only\n");
2328 sb
->s_flags
|= MS_RDONLY
;
2329 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2330 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2332 if (test_opt(sb
, ERRORS_PANIC
)) {
2333 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2334 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2335 ext4_commit_super(sb
, es
, 1);
2339 } else if (journal_inum
) {
2340 if (ext4_create_journal(sb
, es
, journal_inum
))
2342 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
2343 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2344 printk(KERN_ERR
"EXT4-fs: required journal recovery "
2345 "suppressed and not mounted read-only\n");
2348 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
2349 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2350 sbi
->s_journal
= NULL
;
2355 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2356 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2357 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2358 printk(KERN_ERR
"ext4: Failed to set 64-bit journal feature\n");
2362 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2363 jbd2_journal_set_features(sbi
->s_journal
,
2364 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2365 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2366 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2367 jbd2_journal_set_features(sbi
->s_journal
,
2368 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2369 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2370 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2372 jbd2_journal_clear_features(sbi
->s_journal
,
2373 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2374 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2377 /* We have now updated the journal if required, so we can
2378 * validate the data journaling mode. */
2379 switch (test_opt(sb
, DATA_FLAGS
)) {
2381 /* No mode set, assume a default based on the journal
2382 * capabilities: ORDERED_DATA if the journal can
2383 * cope, else JOURNAL_DATA
2385 if (jbd2_journal_check_available_features
2386 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2387 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2389 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2392 case EXT4_MOUNT_ORDERED_DATA
:
2393 case EXT4_MOUNT_WRITEBACK_DATA
:
2394 if (!jbd2_journal_check_available_features
2395 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2396 printk(KERN_ERR
"EXT4-fs: Journal does not support "
2397 "requested data journaling mode\n");
2406 if (test_opt(sb
, NOBH
)) {
2407 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2408 printk(KERN_WARNING
"EXT4-fs: Ignoring nobh option - "
2409 "its supported only with writeback mode\n");
2410 clear_opt(sbi
->s_mount_opt
, NOBH
);
2414 * The jbd2_journal_load will have done any necessary log recovery,
2415 * so we can safely mount the rest of the filesystem now.
2418 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2420 printk(KERN_ERR
"EXT4-fs: get root inode failed\n");
2421 ret
= PTR_ERR(root
);
2424 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2426 printk(KERN_ERR
"EXT4-fs: corrupt root inode, run e2fsck\n");
2429 sb
->s_root
= d_alloc_root(root
);
2431 printk(KERN_ERR
"EXT4-fs: get root dentry failed\n");
2437 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
2439 /* determine the minimum size of new large inodes, if present */
2440 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2441 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2442 EXT4_GOOD_OLD_INODE_SIZE
;
2443 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2444 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2445 if (sbi
->s_want_extra_isize
<
2446 le16_to_cpu(es
->s_want_extra_isize
))
2447 sbi
->s_want_extra_isize
=
2448 le16_to_cpu(es
->s_want_extra_isize
);
2449 if (sbi
->s_want_extra_isize
<
2450 le16_to_cpu(es
->s_min_extra_isize
))
2451 sbi
->s_want_extra_isize
=
2452 le16_to_cpu(es
->s_min_extra_isize
);
2455 /* Check if enough inode space is available */
2456 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2457 sbi
->s_inode_size
) {
2458 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2459 EXT4_GOOD_OLD_INODE_SIZE
;
2460 printk(KERN_INFO
"EXT4-fs: required extra inode space not"
2464 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
2465 printk(KERN_WARNING
"EXT4-fs: Ignoring delalloc option - "
2466 "requested data journaling mode\n");
2467 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2468 } else if (test_opt(sb
, DELALLOC
))
2469 printk(KERN_INFO
"EXT4-fs: delayed allocation enabled\n");
2472 err
= ext4_mb_init(sb
, needs_recovery
);
2474 printk(KERN_ERR
"EXT4-fs: failed to initalize mballoc (%d)\n",
2480 * akpm: core read_super() calls in here with the superblock locked.
2481 * That deadlocks, because orphan cleanup needs to lock the superblock
2482 * in numerous places. Here we just pop the lock - it's relatively
2483 * harmless, because we are now ready to accept write_super() requests,
2484 * and aviro says that's the only reason for hanging onto the
2487 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
2488 ext4_orphan_cleanup(sb
, es
);
2489 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
2490 if (needs_recovery
) {
2491 printk(KERN_INFO
"EXT4-fs: recovery complete.\n");
2492 ext4_mark_recovery_complete(sb
, es
);
2494 if (EXT4_SB(sb
)->s_journal
) {
2495 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2496 descr
= " journalled data mode";
2497 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2498 descr
= " ordered data mode";
2500 descr
= " writeback data mode";
2502 descr
= "out journal";
2504 printk(KERN_INFO
"EXT4-fs: mounted filesystem %s with%s\n",
2512 printk(KERN_ERR
"VFS: Can't find ext4 filesystem on dev %s.\n",
2517 printk(KERN_ERR
"EXT4-fs (device %s): mount failed\n", sb
->s_id
);
2518 if (sbi
->s_journal
) {
2519 jbd2_journal_destroy(sbi
->s_journal
);
2520 sbi
->s_journal
= NULL
;
2523 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2524 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2525 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2526 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
2528 for (i
= 0; i
< db_count
; i
++)
2529 brelse(sbi
->s_group_desc
[i
]);
2530 kfree(sbi
->s_group_desc
);
2533 remove_proc_entry("inode_readahead_blks", sbi
->s_proc
);
2534 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
2537 for (i
= 0; i
< MAXQUOTAS
; i
++)
2538 kfree(sbi
->s_qf_names
[i
]);
2540 ext4_blkdev_remove(sbi
);
2543 sb
->s_fs_info
= NULL
;
2550 * Setup any per-fs journal parameters now. We'll do this both on
2551 * initial mount, once the journal has been initialised but before we've
2552 * done any recovery; and again on any subsequent remount.
2554 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
2556 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2558 journal
->j_commit_interval
= sbi
->s_commit_interval
;
2559 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
2560 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
2562 spin_lock(&journal
->j_state_lock
);
2563 if (test_opt(sb
, BARRIER
))
2564 journal
->j_flags
|= JBD2_BARRIER
;
2566 journal
->j_flags
&= ~JBD2_BARRIER
;
2567 if (test_opt(sb
, DATA_ERR_ABORT
))
2568 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
2570 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
2571 spin_unlock(&journal
->j_state_lock
);
2574 static journal_t
*ext4_get_journal(struct super_block
*sb
,
2575 unsigned int journal_inum
)
2577 struct inode
*journal_inode
;
2580 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2582 /* First, test for the existence of a valid inode on disk. Bad
2583 * things happen if we iget() an unused inode, as the subsequent
2584 * iput() will try to delete it. */
2586 journal_inode
= ext4_iget(sb
, journal_inum
);
2587 if (IS_ERR(journal_inode
)) {
2588 printk(KERN_ERR
"EXT4-fs: no journal found.\n");
2591 if (!journal_inode
->i_nlink
) {
2592 make_bad_inode(journal_inode
);
2593 iput(journal_inode
);
2594 printk(KERN_ERR
"EXT4-fs: journal inode is deleted.\n");
2598 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2599 journal_inode
, journal_inode
->i_size
);
2600 if (!S_ISREG(journal_inode
->i_mode
)) {
2601 printk(KERN_ERR
"EXT4-fs: invalid journal inode.\n");
2602 iput(journal_inode
);
2606 journal
= jbd2_journal_init_inode(journal_inode
);
2608 printk(KERN_ERR
"EXT4-fs: Could not load journal inode\n");
2609 iput(journal_inode
);
2612 journal
->j_private
= sb
;
2613 ext4_init_journal_params(sb
, journal
);
2617 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
2620 struct buffer_head
*bh
;
2624 int hblock
, blocksize
;
2625 ext4_fsblk_t sb_block
;
2626 unsigned long offset
;
2627 struct ext4_super_block
*es
;
2628 struct block_device
*bdev
;
2630 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2632 bdev
= ext4_blkdev_get(j_dev
);
2636 if (bd_claim(bdev
, sb
)) {
2638 "EXT4: failed to claim external journal device.\n");
2639 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
2643 blocksize
= sb
->s_blocksize
;
2644 hblock
= bdev_hardsect_size(bdev
);
2645 if (blocksize
< hblock
) {
2647 "EXT4-fs: blocksize too small for journal device.\n");
2651 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
2652 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
2653 set_blocksize(bdev
, blocksize
);
2654 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
2655 printk(KERN_ERR
"EXT4-fs: couldn't read superblock of "
2656 "external journal\n");
2660 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2661 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
2662 !(le32_to_cpu(es
->s_feature_incompat
) &
2663 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
2664 printk(KERN_ERR
"EXT4-fs: external journal has "
2665 "bad superblock\n");
2670 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
2671 printk(KERN_ERR
"EXT4-fs: journal UUID does not match\n");
2676 len
= ext4_blocks_count(es
);
2677 start
= sb_block
+ 1;
2678 brelse(bh
); /* we're done with the superblock */
2680 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
2681 start
, len
, blocksize
);
2683 printk(KERN_ERR
"EXT4-fs: failed to create device journal\n");
2686 journal
->j_private
= sb
;
2687 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
2688 wait_on_buffer(journal
->j_sb_buffer
);
2689 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
2690 printk(KERN_ERR
"EXT4-fs: I/O error on journal device\n");
2693 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
2694 printk(KERN_ERR
"EXT4-fs: External journal has more than one "
2695 "user (unsupported) - %d\n",
2696 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
2699 EXT4_SB(sb
)->journal_bdev
= bdev
;
2700 ext4_init_journal_params(sb
, journal
);
2703 jbd2_journal_destroy(journal
);
2705 ext4_blkdev_put(bdev
);
2709 static int ext4_load_journal(struct super_block
*sb
,
2710 struct ext4_super_block
*es
,
2711 unsigned long journal_devnum
)
2714 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
2717 int really_read_only
;
2719 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2721 if (journal_devnum
&&
2722 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2723 printk(KERN_INFO
"EXT4-fs: external journal device major/minor "
2724 "numbers have changed\n");
2725 journal_dev
= new_decode_dev(journal_devnum
);
2727 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
2729 really_read_only
= bdev_read_only(sb
->s_bdev
);
2732 * Are we loading a blank journal or performing recovery after a
2733 * crash? For recovery, we need to check in advance whether we
2734 * can get read-write access to the device.
2737 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2738 if (sb
->s_flags
& MS_RDONLY
) {
2739 printk(KERN_INFO
"EXT4-fs: INFO: recovery "
2740 "required on readonly filesystem.\n");
2741 if (really_read_only
) {
2742 printk(KERN_ERR
"EXT4-fs: write access "
2743 "unavailable, cannot proceed.\n");
2746 printk(KERN_INFO
"EXT4-fs: write access will "
2747 "be enabled during recovery.\n");
2751 if (journal_inum
&& journal_dev
) {
2752 printk(KERN_ERR
"EXT4-fs: filesystem has both journal "
2753 "and inode journals!\n");
2758 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
2761 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
2765 if (journal
->j_flags
& JBD2_BARRIER
)
2766 printk(KERN_INFO
"EXT4-fs: barriers enabled\n");
2768 printk(KERN_INFO
"EXT4-fs: barriers disabled\n");
2770 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
2771 err
= jbd2_journal_update_format(journal
);
2773 printk(KERN_ERR
"EXT4-fs: error updating journal.\n");
2774 jbd2_journal_destroy(journal
);
2779 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
2780 err
= jbd2_journal_wipe(journal
, !really_read_only
);
2782 err
= jbd2_journal_load(journal
);
2785 printk(KERN_ERR
"EXT4-fs: error loading journal.\n");
2786 jbd2_journal_destroy(journal
);
2790 EXT4_SB(sb
)->s_journal
= journal
;
2791 ext4_clear_journal_err(sb
, es
);
2793 if (journal_devnum
&&
2794 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2795 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2798 /* Make sure we flush the recovery flag to disk. */
2799 ext4_commit_super(sb
, es
, 1);
2805 static int ext4_create_journal(struct super_block
*sb
,
2806 struct ext4_super_block
*es
,
2807 unsigned int journal_inum
)
2812 if (sb
->s_flags
& MS_RDONLY
) {
2813 printk(KERN_ERR
"EXT4-fs: readonly filesystem when trying to "
2814 "create journal.\n");
2818 journal
= ext4_get_journal(sb
, journal_inum
);
2822 printk(KERN_INFO
"EXT4-fs: creating new journal on inode %u\n",
2825 err
= jbd2_journal_create(journal
);
2827 printk(KERN_ERR
"EXT4-fs: error creating journal.\n");
2828 jbd2_journal_destroy(journal
);
2832 EXT4_SB(sb
)->s_journal
= journal
;
2834 ext4_update_dynamic_rev(sb
);
2835 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2836 EXT4_SET_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
);
2838 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
2841 /* Make sure we flush the recovery flag to disk. */
2842 ext4_commit_super(sb
, es
, 1);
2847 static void ext4_commit_super(struct super_block
*sb
,
2848 struct ext4_super_block
*es
, int sync
)
2850 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
2854 if (buffer_write_io_error(sbh
)) {
2856 * Oh, dear. A previous attempt to write the
2857 * superblock failed. This could happen because the
2858 * USB device was yanked out. Or it could happen to
2859 * be a transient write error and maybe the block will
2860 * be remapped. Nothing we can do but to retry the
2861 * write and hope for the best.
2863 printk(KERN_ERR
"ext4: previous I/O error to "
2864 "superblock detected for %s.\n", sb
->s_id
);
2865 clear_buffer_write_io_error(sbh
);
2866 set_buffer_uptodate(sbh
);
2868 es
->s_wtime
= cpu_to_le32(get_seconds());
2869 ext4_free_blocks_count_set(es
, ext4_count_free_blocks(sb
));
2870 es
->s_free_inodes_count
= cpu_to_le32(ext4_count_free_inodes(sb
));
2871 BUFFER_TRACE(sbh
, "marking dirty");
2872 mark_buffer_dirty(sbh
);
2874 sync_dirty_buffer(sbh
);
2875 if (buffer_write_io_error(sbh
)) {
2876 printk(KERN_ERR
"ext4: I/O error while writing "
2877 "superblock for %s.\n", sb
->s_id
);
2878 clear_buffer_write_io_error(sbh
);
2879 set_buffer_uptodate(sbh
);
2886 * Have we just finished recovery? If so, and if we are mounting (or
2887 * remounting) the filesystem readonly, then we will end up with a
2888 * consistent fs on disk. Record that fact.
2890 static void ext4_mark_recovery_complete(struct super_block
*sb
,
2891 struct ext4_super_block
*es
)
2893 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
2895 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2896 BUG_ON(journal
!= NULL
);
2899 jbd2_journal_lock_updates(journal
);
2900 if (jbd2_journal_flush(journal
) < 0)
2904 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
2905 sb
->s_flags
& MS_RDONLY
) {
2906 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2908 ext4_commit_super(sb
, es
, 1);
2913 jbd2_journal_unlock_updates(journal
);
2917 * If we are mounting (or read-write remounting) a filesystem whose journal
2918 * has recorded an error from a previous lifetime, move that error to the
2919 * main filesystem now.
2921 static void ext4_clear_journal_err(struct super_block
*sb
,
2922 struct ext4_super_block
*es
)
2928 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
2930 journal
= EXT4_SB(sb
)->s_journal
;
2933 * Now check for any error status which may have been recorded in the
2934 * journal by a prior ext4_error() or ext4_abort()
2937 j_errno
= jbd2_journal_errno(journal
);
2941 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
2942 ext4_warning(sb
, __func__
, "Filesystem error recorded "
2943 "from previous mount: %s", errstr
);
2944 ext4_warning(sb
, __func__
, "Marking fs in need of "
2945 "filesystem check.");
2947 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2948 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2949 ext4_commit_super(sb
, es
, 1);
2951 jbd2_journal_clear_err(journal
);
2956 * Force the running and committing transactions to commit,
2957 * and wait on the commit.
2959 int ext4_force_commit(struct super_block
*sb
)
2964 if (sb
->s_flags
& MS_RDONLY
)
2967 journal
= EXT4_SB(sb
)->s_journal
;
2970 ret
= ext4_journal_force_commit(journal
);
2977 * Ext4 always journals updates to the superblock itself, so we don't
2978 * have to propagate any other updates to the superblock on disk at this
2979 * point. (We can probably nuke this function altogether, and remove
2980 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
2982 static void ext4_write_super(struct super_block
*sb
)
2984 if (EXT4_SB(sb
)->s_journal
) {
2985 if (mutex_trylock(&sb
->s_lock
) != 0)
2989 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
2993 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
2997 trace_mark(ext4_sync_fs
, "dev %s wait %d", sb
->s_id
, wait
);
2999 if (EXT4_SB(sb
)->s_journal
) {
3001 ret
= ext4_force_commit(sb
);
3003 jbd2_journal_start_commit(EXT4_SB(sb
)->s_journal
, NULL
);
3005 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, wait
);
3011 * LVM calls this function before a (read-only) snapshot is created. This
3012 * gives us a chance to flush the journal completely and mark the fs clean.
3014 static void ext4_write_super_lockfs(struct super_block
*sb
)
3018 if (!(sb
->s_flags
& MS_RDONLY
)) {
3019 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
3022 /* Now we set up the journal barrier. */
3023 jbd2_journal_lock_updates(journal
);
3026 * We don't want to clear needs_recovery flag when we
3027 * failed to flush the journal.
3029 if (jbd2_journal_flush(journal
) < 0)
3033 /* Journal blocked and flushed, clear needs_recovery flag. */
3034 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3035 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
3040 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3041 * flag here, even though the filesystem is not technically dirty yet.
3043 static void ext4_unlockfs(struct super_block
*sb
)
3045 if (EXT4_SB(sb
)->s_journal
&& !(sb
->s_flags
& MS_RDONLY
)) {
3047 /* Reser the needs_recovery flag before the fs is unlocked. */
3048 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3049 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
3051 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3055 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3057 struct ext4_super_block
*es
;
3058 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3059 ext4_fsblk_t n_blocks_count
= 0;
3060 unsigned long old_sb_flags
;
3061 struct ext4_mount_options old_opts
;
3068 /* Store the original options */
3069 old_sb_flags
= sb
->s_flags
;
3070 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3071 old_opts
.s_resuid
= sbi
->s_resuid
;
3072 old_opts
.s_resgid
= sbi
->s_resgid
;
3073 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3074 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
3075 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
3077 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3078 for (i
= 0; i
< MAXQUOTAS
; i
++)
3079 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3083 * Allow the "check" option to be passed as a remount option.
3085 if (!parse_options(data
, sb
, NULL
, NULL
, &n_blocks_count
, 1)) {
3090 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
)
3091 ext4_abort(sb
, __func__
, "Abort forced by user");
3093 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3094 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
3099 ext4_init_journal_params(sb
, sbi
->s_journal
);
3101 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3102 n_blocks_count
> ext4_blocks_count(es
)) {
3103 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
) {
3108 if (*flags
& MS_RDONLY
) {
3110 * First of all, the unconditional stuff we have to do
3111 * to disable replay of the journal when we next remount
3113 sb
->s_flags
|= MS_RDONLY
;
3116 * OK, test if we are remounting a valid rw partition
3117 * readonly, and if so set the rdonly flag and then
3118 * mark the partition as valid again.
3120 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3121 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3122 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3125 * We have to unlock super so that we can wait for
3128 if (sbi
->s_journal
) {
3130 ext4_mark_recovery_complete(sb
, es
);
3135 if ((ret
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3136 ~EXT4_FEATURE_RO_COMPAT_SUPP
))) {
3137 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3138 "remount RDWR because of unsupported "
3139 "optional features (%x).\n",
3140 sb
->s_id
, le32_to_cpu(ret
));
3146 * Make sure the group descriptor checksums
3147 * are sane. If they aren't, refuse to
3150 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3151 struct ext4_group_desc
*gdp
=
3152 ext4_get_group_desc(sb
, g
, NULL
);
3154 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3156 "EXT4-fs: ext4_remount: "
3157 "Checksum for group %u failed (%u!=%u)\n",
3158 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3159 le16_to_cpu(gdp
->bg_checksum
));
3166 * If we have an unprocessed orphan list hanging
3167 * around from a previously readonly bdev mount,
3168 * require a full umount/remount for now.
3170 if (es
->s_last_orphan
) {
3171 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3172 "remount RDWR because of unprocessed "
3173 "orphan inode list. Please "
3174 "umount/remount instead.\n",
3181 * Mounting a RDONLY partition read-write, so reread
3182 * and store the current valid flag. (It may have
3183 * been changed by e2fsck since we originally mounted
3187 ext4_clear_journal_err(sb
, es
);
3188 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3189 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3191 if (!ext4_setup_super(sb
, es
, 0))
3192 sb
->s_flags
&= ~MS_RDONLY
;
3195 if (sbi
->s_journal
== NULL
)
3196 ext4_commit_super(sb
, es
, 1);
3199 /* Release old quota file names */
3200 for (i
= 0; i
< MAXQUOTAS
; i
++)
3201 if (old_opts
.s_qf_names
[i
] &&
3202 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3203 kfree(old_opts
.s_qf_names
[i
]);
3207 sb
->s_flags
= old_sb_flags
;
3208 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3209 sbi
->s_resuid
= old_opts
.s_resuid
;
3210 sbi
->s_resgid
= old_opts
.s_resgid
;
3211 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3212 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
3213 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
3215 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3216 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3217 if (sbi
->s_qf_names
[i
] &&
3218 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3219 kfree(sbi
->s_qf_names
[i
]);
3220 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3226 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3228 struct super_block
*sb
= dentry
->d_sb
;
3229 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3230 struct ext4_super_block
*es
= sbi
->s_es
;
3233 if (test_opt(sb
, MINIX_DF
)) {
3234 sbi
->s_overhead_last
= 0;
3235 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3236 ext4_group_t ngroups
= sbi
->s_groups_count
, i
;
3237 ext4_fsblk_t overhead
= 0;
3241 * Compute the overhead (FS structures). This is constant
3242 * for a given filesystem unless the number of block groups
3243 * changes so we cache the previous value until it does.
3247 * All of the blocks before first_data_block are
3250 overhead
= le32_to_cpu(es
->s_first_data_block
);
3253 * Add the overhead attributed to the superblock and
3254 * block group descriptors. If the sparse superblocks
3255 * feature is turned on, then not all groups have this.
3257 for (i
= 0; i
< ngroups
; i
++) {
3258 overhead
+= ext4_bg_has_super(sb
, i
) +
3259 ext4_bg_num_gdb(sb
, i
);
3264 * Every block group has an inode bitmap, a block
3265 * bitmap, and an inode table.
3267 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3268 sbi
->s_overhead_last
= overhead
;
3270 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3273 buf
->f_type
= EXT4_SUPER_MAGIC
;
3274 buf
->f_bsize
= sb
->s_blocksize
;
3275 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3276 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3277 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3278 ext4_free_blocks_count_set(es
, buf
->f_bfree
);
3279 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3280 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3282 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3283 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3284 es
->s_free_inodes_count
= cpu_to_le32(buf
->f_ffree
);
3285 buf
->f_namelen
= EXT4_NAME_LEN
;
3286 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3287 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3288 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3289 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3293 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3294 * is locked for write. Otherwise the are possible deadlocks:
3295 * Process 1 Process 2
3296 * ext4_create() quota_sync()
3297 * jbd2_journal_start() write_dquot()
3298 * DQUOT_INIT() down(dqio_mutex)
3299 * down(dqio_mutex) jbd2_journal_start()
3305 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3307 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3310 static int ext4_dquot_initialize(struct inode
*inode
, int type
)
3315 /* We may create quota structure so we need to reserve enough blocks */
3316 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_INIT_BLOCKS(inode
->i_sb
));
3318 return PTR_ERR(handle
);
3319 ret
= dquot_initialize(inode
, type
);
3320 err
= ext4_journal_stop(handle
);
3326 static int ext4_dquot_drop(struct inode
*inode
)
3331 /* We may delete quota structure so we need to reserve enough blocks */
3332 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_DEL_BLOCKS(inode
->i_sb
));
3333 if (IS_ERR(handle
)) {
3335 * We call dquot_drop() anyway to at least release references
3336 * to quota structures so that umount does not hang.
3339 return PTR_ERR(handle
);
3341 ret
= dquot_drop(inode
);
3342 err
= ext4_journal_stop(handle
);
3348 static int ext4_write_dquot(struct dquot
*dquot
)
3352 struct inode
*inode
;
3354 inode
= dquot_to_inode(dquot
);
3355 handle
= ext4_journal_start(inode
,
3356 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3358 return PTR_ERR(handle
);
3359 ret
= dquot_commit(dquot
);
3360 err
= ext4_journal_stop(handle
);
3366 static int ext4_acquire_dquot(struct dquot
*dquot
)
3371 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3372 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3374 return PTR_ERR(handle
);
3375 ret
= dquot_acquire(dquot
);
3376 err
= ext4_journal_stop(handle
);
3382 static int ext4_release_dquot(struct dquot
*dquot
)
3387 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3388 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3389 if (IS_ERR(handle
)) {
3390 /* Release dquot anyway to avoid endless cycle in dqput() */
3391 dquot_release(dquot
);
3392 return PTR_ERR(handle
);
3394 ret
= dquot_release(dquot
);
3395 err
= ext4_journal_stop(handle
);
3401 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3403 /* Are we journaling quotas? */
3404 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3405 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3406 dquot_mark_dquot_dirty(dquot
);
3407 return ext4_write_dquot(dquot
);
3409 return dquot_mark_dquot_dirty(dquot
);
3413 static int ext4_write_info(struct super_block
*sb
, int type
)
3418 /* Data block + inode block */
3419 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3421 return PTR_ERR(handle
);
3422 ret
= dquot_commit_info(sb
, type
);
3423 err
= ext4_journal_stop(handle
);
3430 * Turn on quotas during mount time - we need to find
3431 * the quota file and such...
3433 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3435 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3436 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3440 * Standard function to be called on quota_on
3442 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3443 char *name
, int remount
)
3448 if (!test_opt(sb
, QUOTA
))
3450 /* When remounting, no checks are needed and in fact, name is NULL */
3452 return vfs_quota_on(sb
, type
, format_id
, name
, remount
);
3454 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
3458 /* Quotafile not on the same filesystem? */
3459 if (path
.mnt
->mnt_sb
!= sb
) {
3463 /* Journaling quota? */
3464 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3465 /* Quotafile not in fs root? */
3466 if (path
.dentry
->d_parent
!= sb
->s_root
)
3468 "EXT4-fs: Quota file not on filesystem root. "
3469 "Journaled quota will not work.\n");
3473 * When we journal data on quota file, we have to flush journal to see
3474 * all updates to the file when we bypass pagecache...
3476 if (EXT4_SB(sb
)->s_journal
&&
3477 ext4_should_journal_data(path
.dentry
->d_inode
)) {
3479 * We don't need to lock updates but journal_flush() could
3480 * otherwise be livelocked...
3482 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3483 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3484 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3491 err
= vfs_quota_on_path(sb
, type
, format_id
, &path
);
3496 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3497 * acquiring the locks... As quota files are never truncated and quota code
3498 * itself serializes the operations (and noone else should touch the files)
3499 * we don't have to be afraid of races */
3500 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3501 size_t len
, loff_t off
)
3503 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3504 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3506 int offset
= off
& (sb
->s_blocksize
- 1);
3509 struct buffer_head
*bh
;
3510 loff_t i_size
= i_size_read(inode
);
3514 if (off
+len
> i_size
)
3517 while (toread
> 0) {
3518 tocopy
= sb
->s_blocksize
- offset
< toread
?
3519 sb
->s_blocksize
- offset
: toread
;
3520 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3523 if (!bh
) /* A hole? */
3524 memset(data
, 0, tocopy
);
3526 memcpy(data
, bh
->b_data
+offset
, tocopy
);
3536 /* Write to quotafile (we know the transaction is already started and has
3537 * enough credits) */
3538 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
3539 const char *data
, size_t len
, loff_t off
)
3541 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3542 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3544 int offset
= off
& (sb
->s_blocksize
- 1);
3546 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
3547 size_t towrite
= len
;
3548 struct buffer_head
*bh
;
3549 handle_t
*handle
= journal_current_handle();
3551 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
3552 printk(KERN_WARNING
"EXT4-fs: Quota write (off=%llu, len=%llu)"
3553 " cancelled because transaction is not started.\n",
3554 (unsigned long long)off
, (unsigned long long)len
);
3557 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
3558 while (towrite
> 0) {
3559 tocopy
= sb
->s_blocksize
- offset
< towrite
?
3560 sb
->s_blocksize
- offset
: towrite
;
3561 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
3564 if (journal_quota
) {
3565 err
= ext4_journal_get_write_access(handle
, bh
);
3572 memcpy(bh
->b_data
+offset
, data
, tocopy
);
3573 flush_dcache_page(bh
->b_page
);
3576 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
3578 /* Always do at least ordered writes for quotas */
3579 err
= ext4_jbd2_file_inode(handle
, inode
);
3580 mark_buffer_dirty(bh
);
3591 if (len
== towrite
) {
3592 mutex_unlock(&inode
->i_mutex
);
3595 if (inode
->i_size
< off
+len
-towrite
) {
3596 i_size_write(inode
, off
+len
-towrite
);
3597 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
3599 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3600 ext4_mark_inode_dirty(handle
, inode
);
3601 mutex_unlock(&inode
->i_mutex
);
3602 return len
- towrite
;
3607 static int ext4_get_sb(struct file_system_type
*fs_type
,
3608 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3610 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3613 #ifdef CONFIG_PROC_FS
3614 static int ext4_ui_proc_show(struct seq_file
*m
, void *v
)
3616 unsigned int *p
= m
->private;
3618 seq_printf(m
, "%u\n", *p
);
3622 static int ext4_ui_proc_open(struct inode
*inode
, struct file
*file
)
3624 return single_open(file
, ext4_ui_proc_show
, PDE(inode
)->data
);
3627 static ssize_t
ext4_ui_proc_write(struct file
*file
, const char __user
*buf
,
3628 size_t cnt
, loff_t
*ppos
)
3630 unsigned long *p
= PDE(file
->f_path
.dentry
->d_inode
)->data
;
3633 if (cnt
>= sizeof(str
))
3635 if (copy_from_user(str
, buf
, cnt
))
3638 *p
= simple_strtoul(str
, NULL
, 0);
3642 const struct file_operations ext4_ui_proc_fops
= {
3643 .owner
= THIS_MODULE
,
3644 .open
= ext4_ui_proc_open
,
3646 .llseek
= seq_lseek
,
3647 .release
= single_release
,
3648 .write
= ext4_ui_proc_write
,
3652 static struct file_system_type ext4_fs_type
= {
3653 .owner
= THIS_MODULE
,
3655 .get_sb
= ext4_get_sb
,
3656 .kill_sb
= kill_block_super
,
3657 .fs_flags
= FS_REQUIRES_DEV
,
3660 #ifdef CONFIG_EXT4DEV_COMPAT
3661 static int ext4dev_get_sb(struct file_system_type
*fs_type
,
3662 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3664 printk(KERN_WARNING
"EXT4-fs: Update your userspace programs "
3665 "to mount using ext4\n");
3666 printk(KERN_WARNING
"EXT4-fs: ext4dev backwards compatibility "
3667 "will go away by 2.6.31\n");
3668 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3671 static struct file_system_type ext4dev_fs_type
= {
3672 .owner
= THIS_MODULE
,
3674 .get_sb
= ext4dev_get_sb
,
3675 .kill_sb
= kill_block_super
,
3676 .fs_flags
= FS_REQUIRES_DEV
,
3678 MODULE_ALIAS("ext4dev");
3681 static int __init
init_ext4_fs(void)
3685 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
3686 err
= init_ext4_mballoc();
3690 err
= init_ext4_xattr();
3693 err
= init_inodecache();
3696 err
= register_filesystem(&ext4_fs_type
);
3699 #ifdef CONFIG_EXT4DEV_COMPAT
3700 err
= register_filesystem(&ext4dev_fs_type
);
3702 unregister_filesystem(&ext4_fs_type
);
3708 destroy_inodecache();
3712 exit_ext4_mballoc();
3716 static void __exit
exit_ext4_fs(void)
3718 unregister_filesystem(&ext4_fs_type
);
3719 #ifdef CONFIG_EXT4DEV_COMPAT
3720 unregister_filesystem(&ext4dev_fs_type
);
3722 destroy_inodecache();
3724 exit_ext4_mballoc();
3725 remove_proc_entry("fs/ext4", NULL
);
3728 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3729 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3730 MODULE_LICENSE("GPL");
3731 module_init(init_ext4_fs
)
3732 module_exit(exit_ext4_fs
)