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
;
139 if (is_journal_aborted(journal
)) {
140 ext4_abort(sb
, __func__
,
141 "Detected aborted journal");
142 return ERR_PTR(-EROFS
);
145 return jbd2_journal_start(journal
, nblocks
);
149 * The only special thing we need to do here is to make sure that all
150 * jbd2_journal_stop calls result in the superblock being marked dirty, so
151 * that sync() will call the filesystem's write_super callback if
154 int __ext4_journal_stop(const char *where
, handle_t
*handle
)
156 struct super_block
*sb
;
160 sb
= handle
->h_transaction
->t_journal
->j_private
;
162 rc
= jbd2_journal_stop(handle
);
167 __ext4_std_error(sb
, where
, err
);
171 void ext4_journal_abort_handle(const char *caller
, const char *err_fn
,
172 struct buffer_head
*bh
, handle_t
*handle
, int err
)
175 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
178 BUFFER_TRACE(bh
, "abort");
183 if (is_handle_aborted(handle
))
186 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
187 caller
, errstr
, err_fn
);
189 jbd2_journal_abort_handle(handle
);
192 /* Deal with the reporting of failure conditions on a filesystem such as
193 * inconsistencies detected or read IO failures.
195 * On ext2, we can store the error state of the filesystem in the
196 * superblock. That is not possible on ext4, because we may have other
197 * write ordering constraints on the superblock which prevent us from
198 * writing it out straight away; and given that the journal is about to
199 * be aborted, we can't rely on the current, or future, transactions to
200 * write out the superblock safely.
202 * We'll just use the jbd2_journal_abort() error code to record an error in
203 * the journal instead. On recovery, the journal will compain about
204 * that error until we've noted it down and cleared it.
207 static void ext4_handle_error(struct super_block
*sb
)
209 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
211 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
212 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
214 if (sb
->s_flags
& MS_RDONLY
)
217 if (!test_opt(sb
, ERRORS_CONT
)) {
218 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
220 EXT4_SB(sb
)->s_mount_opt
|= EXT4_MOUNT_ABORT
;
222 jbd2_journal_abort(journal
, -EIO
);
224 if (test_opt(sb
, ERRORS_RO
)) {
225 printk(KERN_CRIT
"Remounting filesystem read-only\n");
226 sb
->s_flags
|= MS_RDONLY
;
228 ext4_commit_super(sb
, es
, 1);
229 if (test_opt(sb
, ERRORS_PANIC
))
230 panic("EXT4-fs (device %s): panic forced after error\n",
234 void ext4_error(struct super_block
*sb
, const char *function
,
235 const char *fmt
, ...)
240 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
245 ext4_handle_error(sb
);
248 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
255 errstr
= "IO failure";
258 errstr
= "Out of memory";
261 if (!sb
|| EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
)
262 errstr
= "Journal has aborted";
264 errstr
= "Readonly filesystem";
267 /* If the caller passed in an extra buffer for unknown
268 * errors, textualise them now. Else we just return
271 /* Check for truncated error codes... */
272 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
281 /* __ext4_std_error decodes expected errors from journaling functions
282 * automatically and invokes the appropriate error response. */
284 void __ext4_std_error(struct super_block
*sb
, const char *function
, int errno
)
289 /* Special case: if the error is EROFS, and we're not already
290 * inside a transaction, then there's really no point in logging
292 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
293 (sb
->s_flags
& MS_RDONLY
))
296 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
297 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s: %s\n",
298 sb
->s_id
, function
, errstr
);
300 ext4_handle_error(sb
);
304 * ext4_abort is a much stronger failure handler than ext4_error. The
305 * abort function may be used to deal with unrecoverable failures such
306 * as journal IO errors or ENOMEM at a critical moment in log management.
308 * We unconditionally force the filesystem into an ABORT|READONLY state,
309 * unless the error response on the fs has been set to panic in which
310 * case we take the easy way out and panic immediately.
313 void ext4_abort(struct super_block
*sb
, const char *function
,
314 const char *fmt
, ...)
318 printk(KERN_CRIT
"ext4_abort called.\n");
321 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
326 if (test_opt(sb
, ERRORS_PANIC
))
327 panic("EXT4-fs panic from previous error\n");
329 if (sb
->s_flags
& MS_RDONLY
)
332 printk(KERN_CRIT
"Remounting filesystem read-only\n");
333 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
334 sb
->s_flags
|= MS_RDONLY
;
335 EXT4_SB(sb
)->s_mount_opt
|= EXT4_MOUNT_ABORT
;
336 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
339 void ext4_warning(struct super_block
*sb
, const char *function
,
340 const char *fmt
, ...)
345 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s: ",
352 void ext4_update_dynamic_rev(struct super_block
*sb
)
354 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
356 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
359 ext4_warning(sb
, __func__
,
360 "updating to rev %d because of new feature flag, "
361 "running e2fsck is recommended",
364 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
365 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
366 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
367 /* leave es->s_feature_*compat flags alone */
368 /* es->s_uuid will be set by e2fsck if empty */
371 * The rest of the superblock fields should be zero, and if not it
372 * means they are likely already in use, so leave them alone. We
373 * can leave it up to e2fsck to clean up any inconsistencies there.
377 int ext4_update_compat_feature(handle_t
*handle
,
378 struct super_block
*sb
, __u32 compat
)
381 if (!EXT4_HAS_COMPAT_FEATURE(sb
, compat
)) {
382 err
= ext4_journal_get_write_access(handle
,
386 EXT4_SET_COMPAT_FEATURE(sb
, compat
);
389 BUFFER_TRACE(EXT4_SB(sb
)->s_sbh
,
390 "call ext4_journal_dirty_met adata");
391 err
= ext4_journal_dirty_metadata(handle
,
397 int ext4_update_rocompat_feature(handle_t
*handle
,
398 struct super_block
*sb
, __u32 rocompat
)
401 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb
, rocompat
)) {
402 err
= ext4_journal_get_write_access(handle
,
406 EXT4_SET_RO_COMPAT_FEATURE(sb
, rocompat
);
409 BUFFER_TRACE(EXT4_SB(sb
)->s_sbh
,
410 "call ext4_journal_dirty_met adata");
411 err
= ext4_journal_dirty_metadata(handle
,
417 int ext4_update_incompat_feature(handle_t
*handle
,
418 struct super_block
*sb
, __u32 incompat
)
421 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, incompat
)) {
422 err
= ext4_journal_get_write_access(handle
,
426 EXT4_SET_INCOMPAT_FEATURE(sb
, incompat
);
429 BUFFER_TRACE(EXT4_SB(sb
)->s_sbh
,
430 "call ext4_journal_dirty_met adata");
431 err
= ext4_journal_dirty_metadata(handle
,
438 * Open the external journal device
440 static struct block_device
*ext4_blkdev_get(dev_t dev
)
442 struct block_device
*bdev
;
443 char b
[BDEVNAME_SIZE
];
445 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
451 printk(KERN_ERR
"EXT4: failed to open journal device %s: %ld\n",
452 __bdevname(dev
, b
), PTR_ERR(bdev
));
457 * Release the journal device
459 static int ext4_blkdev_put(struct block_device
*bdev
)
462 return blkdev_put(bdev
);
465 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
467 struct block_device
*bdev
;
470 bdev
= sbi
->journal_bdev
;
472 ret
= ext4_blkdev_put(bdev
);
473 sbi
->journal_bdev
= NULL
;
478 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
480 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
483 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
487 printk(KERN_ERR
"sb orphan head is %d\n",
488 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
490 printk(KERN_ERR
"sb_info orphan list:\n");
491 list_for_each(l
, &sbi
->s_orphan
) {
492 struct inode
*inode
= orphan_list_entry(l
);
494 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
495 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
496 inode
->i_mode
, inode
->i_nlink
,
501 static void ext4_put_super(struct super_block
*sb
)
503 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
504 struct ext4_super_block
*es
= sbi
->s_es
;
508 ext4_ext_release(sb
);
509 ext4_xattr_put_super(sb
);
510 if (jbd2_journal_destroy(sbi
->s_journal
) < 0)
511 ext4_abort(sb
, __func__
, "Couldn't clean up the journal");
512 sbi
->s_journal
= NULL
;
513 if (!(sb
->s_flags
& MS_RDONLY
)) {
514 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
515 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
516 ext4_commit_super(sb
, es
, 1);
519 remove_proc_entry("inode_readahead_blks", sbi
->s_proc
);
520 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
523 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
524 brelse(sbi
->s_group_desc
[i
]);
525 kfree(sbi
->s_group_desc
);
526 kfree(sbi
->s_flex_groups
);
527 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
528 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
529 percpu_counter_destroy(&sbi
->s_dirs_counter
);
530 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
533 for (i
= 0; i
< MAXQUOTAS
; i
++)
534 kfree(sbi
->s_qf_names
[i
]);
537 /* Debugging code just in case the in-memory inode orphan list
538 * isn't empty. The on-disk one can be non-empty if we've
539 * detected an error and taken the fs readonly, but the
540 * in-memory list had better be clean by this point. */
541 if (!list_empty(&sbi
->s_orphan
))
542 dump_orphan_list(sb
, sbi
);
543 J_ASSERT(list_empty(&sbi
->s_orphan
));
545 invalidate_bdev(sb
->s_bdev
);
546 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
548 * Invalidate the journal device's buffers. We don't want them
549 * floating about in memory - the physical journal device may
550 * hotswapped, and it breaks the `ro-after' testing code.
552 sync_blockdev(sbi
->journal_bdev
);
553 invalidate_bdev(sbi
->journal_bdev
);
554 ext4_blkdev_remove(sbi
);
556 sb
->s_fs_info
= NULL
;
561 static struct kmem_cache
*ext4_inode_cachep
;
564 * Called inside transaction, so use GFP_NOFS
566 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
568 struct ext4_inode_info
*ei
;
570 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
573 #ifdef CONFIG_EXT4_FS_POSIX_ACL
574 ei
->i_acl
= EXT4_ACL_NOT_CACHED
;
575 ei
->i_default_acl
= EXT4_ACL_NOT_CACHED
;
577 ei
->vfs_inode
.i_version
= 1;
578 ei
->vfs_inode
.i_data
.writeback_index
= 0;
579 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
580 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
581 spin_lock_init(&ei
->i_prealloc_lock
);
582 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
583 ei
->i_reserved_data_blocks
= 0;
584 ei
->i_reserved_meta_blocks
= 0;
585 ei
->i_allocated_meta_blocks
= 0;
586 ei
->i_delalloc_reserved_flag
= 0;
587 spin_lock_init(&(ei
->i_block_reservation_lock
));
588 return &ei
->vfs_inode
;
591 static void ext4_destroy_inode(struct inode
*inode
)
593 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
594 printk("EXT4 Inode %p: orphan list check failed!\n",
596 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
597 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
601 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
604 static void init_once(void *foo
)
606 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
608 INIT_LIST_HEAD(&ei
->i_orphan
);
609 #ifdef CONFIG_EXT4_FS_XATTR
610 init_rwsem(&ei
->xattr_sem
);
612 init_rwsem(&ei
->i_data_sem
);
613 inode_init_once(&ei
->vfs_inode
);
616 static int init_inodecache(void)
618 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
619 sizeof(struct ext4_inode_info
),
620 0, (SLAB_RECLAIM_ACCOUNT
|
623 if (ext4_inode_cachep
== NULL
)
628 static void destroy_inodecache(void)
630 kmem_cache_destroy(ext4_inode_cachep
);
633 static void ext4_clear_inode(struct inode
*inode
)
635 #ifdef CONFIG_EXT4_FS_POSIX_ACL
636 if (EXT4_I(inode
)->i_acl
&&
637 EXT4_I(inode
)->i_acl
!= EXT4_ACL_NOT_CACHED
) {
638 posix_acl_release(EXT4_I(inode
)->i_acl
);
639 EXT4_I(inode
)->i_acl
= EXT4_ACL_NOT_CACHED
;
641 if (EXT4_I(inode
)->i_default_acl
&&
642 EXT4_I(inode
)->i_default_acl
!= EXT4_ACL_NOT_CACHED
) {
643 posix_acl_release(EXT4_I(inode
)->i_default_acl
);
644 EXT4_I(inode
)->i_default_acl
= EXT4_ACL_NOT_CACHED
;
647 ext4_discard_preallocations(inode
);
648 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
649 &EXT4_I(inode
)->jinode
);
652 static inline void ext4_show_quota_options(struct seq_file
*seq
,
653 struct super_block
*sb
)
655 #if defined(CONFIG_QUOTA)
656 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
658 if (sbi
->s_jquota_fmt
)
659 seq_printf(seq
, ",jqfmt=%s",
660 (sbi
->s_jquota_fmt
== QFMT_VFS_OLD
) ? "vfsold" : "vfsv0");
662 if (sbi
->s_qf_names
[USRQUOTA
])
663 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
665 if (sbi
->s_qf_names
[GRPQUOTA
])
666 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
668 if (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
)
669 seq_puts(seq
, ",usrquota");
671 if (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)
672 seq_puts(seq
, ",grpquota");
678 * - it's set to a non-default value OR
679 * - if the per-sb default is different from the global default
681 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
684 unsigned long def_mount_opts
;
685 struct super_block
*sb
= vfs
->mnt_sb
;
686 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
687 struct ext4_super_block
*es
= sbi
->s_es
;
689 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
690 def_errors
= le16_to_cpu(es
->s_errors
);
692 if (sbi
->s_sb_block
!= 1)
693 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
694 if (test_opt(sb
, MINIX_DF
))
695 seq_puts(seq
, ",minixdf");
696 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
697 seq_puts(seq
, ",grpid");
698 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
699 seq_puts(seq
, ",nogrpid");
700 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
701 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
702 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
704 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
705 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
706 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
708 if (test_opt(sb
, ERRORS_RO
)) {
709 if (def_errors
== EXT4_ERRORS_PANIC
||
710 def_errors
== EXT4_ERRORS_CONTINUE
) {
711 seq_puts(seq
, ",errors=remount-ro");
714 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
715 seq_puts(seq
, ",errors=continue");
716 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
717 seq_puts(seq
, ",errors=panic");
718 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
719 seq_puts(seq
, ",nouid32");
720 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
721 seq_puts(seq
, ",debug");
722 if (test_opt(sb
, OLDALLOC
))
723 seq_puts(seq
, ",oldalloc");
724 #ifdef CONFIG_EXT4_FS_XATTR
725 if (test_opt(sb
, XATTR_USER
) &&
726 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
727 seq_puts(seq
, ",user_xattr");
728 if (!test_opt(sb
, XATTR_USER
) &&
729 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
730 seq_puts(seq
, ",nouser_xattr");
733 #ifdef CONFIG_EXT4_FS_POSIX_ACL
734 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
735 seq_puts(seq
, ",acl");
736 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
737 seq_puts(seq
, ",noacl");
739 if (!test_opt(sb
, RESERVATION
))
740 seq_puts(seq
, ",noreservation");
741 if (sbi
->s_commit_interval
) {
742 seq_printf(seq
, ",commit=%u",
743 (unsigned) (sbi
->s_commit_interval
/ HZ
));
746 * We're changing the default of barrier mount option, so
747 * let's always display its mount state so it's clear what its
750 seq_puts(seq
, ",barrier=");
751 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
752 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
753 seq_puts(seq
, ",journal_async_commit");
754 if (test_opt(sb
, NOBH
))
755 seq_puts(seq
, ",nobh");
756 if (!test_opt(sb
, EXTENTS
))
757 seq_puts(seq
, ",noextents");
758 if (test_opt(sb
, I_VERSION
))
759 seq_puts(seq
, ",i_version");
760 if (!test_opt(sb
, DELALLOC
))
761 seq_puts(seq
, ",nodelalloc");
765 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
767 * journal mode get enabled in different ways
768 * So just print the value even if we didn't specify it
770 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
771 seq_puts(seq
, ",data=journal");
772 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
773 seq_puts(seq
, ",data=ordered");
774 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
775 seq_puts(seq
, ",data=writeback");
777 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
778 seq_printf(seq
, ",inode_readahead_blks=%u",
779 sbi
->s_inode_readahead_blks
);
781 if (test_opt(sb
, DATA_ERR_ABORT
))
782 seq_puts(seq
, ",data_err=abort");
784 ext4_show_quota_options(seq
, sb
);
789 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
790 u64 ino
, u32 generation
)
794 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
795 return ERR_PTR(-ESTALE
);
796 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
797 return ERR_PTR(-ESTALE
);
799 /* iget isn't really right if the inode is currently unallocated!!
801 * ext4_read_inode will return a bad_inode if the inode had been
802 * deleted, so we should be safe.
804 * Currently we don't know the generation for parent directory, so
805 * a generation of 0 means "accept any"
807 inode
= ext4_iget(sb
, ino
);
809 return ERR_CAST(inode
);
810 if (generation
&& inode
->i_generation
!= generation
) {
812 return ERR_PTR(-ESTALE
);
818 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
819 int fh_len
, int fh_type
)
821 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
825 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
826 int fh_len
, int fh_type
)
828 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
833 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
834 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
836 static int ext4_dquot_initialize(struct inode
*inode
, int type
);
837 static int ext4_dquot_drop(struct inode
*inode
);
838 static int ext4_write_dquot(struct dquot
*dquot
);
839 static int ext4_acquire_dquot(struct dquot
*dquot
);
840 static int ext4_release_dquot(struct dquot
*dquot
);
841 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
842 static int ext4_write_info(struct super_block
*sb
, int type
);
843 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
844 char *path
, int remount
);
845 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
846 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
847 size_t len
, loff_t off
);
848 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
849 const char *data
, size_t len
, loff_t off
);
851 static struct dquot_operations ext4_quota_operations
= {
852 .initialize
= ext4_dquot_initialize
,
853 .drop
= ext4_dquot_drop
,
854 .alloc_space
= dquot_alloc_space
,
855 .alloc_inode
= dquot_alloc_inode
,
856 .free_space
= dquot_free_space
,
857 .free_inode
= dquot_free_inode
,
858 .transfer
= dquot_transfer
,
859 .write_dquot
= ext4_write_dquot
,
860 .acquire_dquot
= ext4_acquire_dquot
,
861 .release_dquot
= ext4_release_dquot
,
862 .mark_dirty
= ext4_mark_dquot_dirty
,
863 .write_info
= ext4_write_info
866 static struct quotactl_ops ext4_qctl_operations
= {
867 .quota_on
= ext4_quota_on
,
868 .quota_off
= vfs_quota_off
,
869 .quota_sync
= vfs_quota_sync
,
870 .get_info
= vfs_get_dqinfo
,
871 .set_info
= vfs_set_dqinfo
,
872 .get_dqblk
= vfs_get_dqblk
,
873 .set_dqblk
= vfs_set_dqblk
877 static const struct super_operations ext4_sops
= {
878 .alloc_inode
= ext4_alloc_inode
,
879 .destroy_inode
= ext4_destroy_inode
,
880 .write_inode
= ext4_write_inode
,
881 .dirty_inode
= ext4_dirty_inode
,
882 .delete_inode
= ext4_delete_inode
,
883 .put_super
= ext4_put_super
,
884 .write_super
= ext4_write_super
,
885 .sync_fs
= ext4_sync_fs
,
886 .write_super_lockfs
= ext4_write_super_lockfs
,
887 .unlockfs
= ext4_unlockfs
,
888 .statfs
= ext4_statfs
,
889 .remount_fs
= ext4_remount
,
890 .clear_inode
= ext4_clear_inode
,
891 .show_options
= ext4_show_options
,
893 .quota_read
= ext4_quota_read
,
894 .quota_write
= ext4_quota_write
,
898 static const struct export_operations ext4_export_ops
= {
899 .fh_to_dentry
= ext4_fh_to_dentry
,
900 .fh_to_parent
= ext4_fh_to_parent
,
901 .get_parent
= ext4_get_parent
,
905 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
906 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
907 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
908 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
909 Opt_reservation
, Opt_noreservation
, Opt_noload
, Opt_nobh
, Opt_bh
,
910 Opt_commit
, Opt_journal_update
, Opt_journal_inum
, Opt_journal_dev
,
911 Opt_journal_checksum
, Opt_journal_async_commit
,
912 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
913 Opt_data_err_abort
, Opt_data_err_ignore
,
914 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
915 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_quota
, Opt_noquota
,
916 Opt_ignore
, Opt_barrier
, Opt_err
, Opt_resize
, Opt_usrquota
,
917 Opt_grpquota
, Opt_extents
, Opt_noextents
, Opt_i_version
,
918 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
919 Opt_inode_readahead_blks
922 static const match_table_t tokens
= {
923 {Opt_bsd_df
, "bsddf"},
924 {Opt_minix_df
, "minixdf"},
925 {Opt_grpid
, "grpid"},
926 {Opt_grpid
, "bsdgroups"},
927 {Opt_nogrpid
, "nogrpid"},
928 {Opt_nogrpid
, "sysvgroups"},
929 {Opt_resgid
, "resgid=%u"},
930 {Opt_resuid
, "resuid=%u"},
932 {Opt_err_cont
, "errors=continue"},
933 {Opt_err_panic
, "errors=panic"},
934 {Opt_err_ro
, "errors=remount-ro"},
935 {Opt_nouid32
, "nouid32"},
936 {Opt_debug
, "debug"},
937 {Opt_oldalloc
, "oldalloc"},
938 {Opt_orlov
, "orlov"},
939 {Opt_user_xattr
, "user_xattr"},
940 {Opt_nouser_xattr
, "nouser_xattr"},
942 {Opt_noacl
, "noacl"},
943 {Opt_reservation
, "reservation"},
944 {Opt_noreservation
, "noreservation"},
945 {Opt_noload
, "noload"},
948 {Opt_commit
, "commit=%u"},
949 {Opt_journal_update
, "journal=update"},
950 {Opt_journal_inum
, "journal=%u"},
951 {Opt_journal_dev
, "journal_dev=%u"},
952 {Opt_journal_checksum
, "journal_checksum"},
953 {Opt_journal_async_commit
, "journal_async_commit"},
954 {Opt_abort
, "abort"},
955 {Opt_data_journal
, "data=journal"},
956 {Opt_data_ordered
, "data=ordered"},
957 {Opt_data_writeback
, "data=writeback"},
958 {Opt_data_err_abort
, "data_err=abort"},
959 {Opt_data_err_ignore
, "data_err=ignore"},
960 {Opt_offusrjquota
, "usrjquota="},
961 {Opt_usrjquota
, "usrjquota=%s"},
962 {Opt_offgrpjquota
, "grpjquota="},
963 {Opt_grpjquota
, "grpjquota=%s"},
964 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
965 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
966 {Opt_grpquota
, "grpquota"},
967 {Opt_noquota
, "noquota"},
968 {Opt_quota
, "quota"},
969 {Opt_usrquota
, "usrquota"},
970 {Opt_barrier
, "barrier=%u"},
971 {Opt_extents
, "extents"},
972 {Opt_noextents
, "noextents"},
973 {Opt_i_version
, "i_version"},
974 {Opt_stripe
, "stripe=%u"},
975 {Opt_resize
, "resize"},
976 {Opt_delalloc
, "delalloc"},
977 {Opt_nodelalloc
, "nodelalloc"},
978 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
982 static ext4_fsblk_t
get_sb_block(void **data
)
984 ext4_fsblk_t sb_block
;
985 char *options
= (char *) *data
;
987 if (!options
|| strncmp(options
, "sb=", 3) != 0)
988 return 1; /* Default location */
990 /*todo: use simple_strtoll with >32bit ext4 */
991 sb_block
= simple_strtoul(options
, &options
, 0);
992 if (*options
&& *options
!= ',') {
993 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
999 *data
= (void *) options
;
1003 static int parse_options(char *options
, struct super_block
*sb
,
1004 unsigned int *inum
, unsigned long *journal_devnum
,
1005 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1007 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1009 substring_t args
[MAX_OPT_ARGS
];
1016 ext4_fsblk_t last_block
;
1021 while ((p
= strsep(&options
, ",")) != NULL
) {
1026 token
= match_token(p
, tokens
, args
);
1029 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1032 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1035 set_opt(sbi
->s_mount_opt
, GRPID
);
1038 clear_opt(sbi
->s_mount_opt
, GRPID
);
1041 if (match_int(&args
[0], &option
))
1043 sbi
->s_resuid
= option
;
1046 if (match_int(&args
[0], &option
))
1048 sbi
->s_resgid
= option
;
1051 /* handled by get_sb_block() instead of here */
1052 /* *sb_block = match_int(&args[0]); */
1055 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1056 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1057 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1060 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1061 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1062 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1065 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1066 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1067 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1070 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1073 set_opt(sbi
->s_mount_opt
, DEBUG
);
1076 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1079 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1081 #ifdef CONFIG_EXT4_FS_XATTR
1082 case Opt_user_xattr
:
1083 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1085 case Opt_nouser_xattr
:
1086 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1089 case Opt_user_xattr
:
1090 case Opt_nouser_xattr
:
1091 printk(KERN_ERR
"EXT4 (no)user_xattr options "
1095 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1097 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1100 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1105 printk(KERN_ERR
"EXT4 (no)acl options "
1109 case Opt_reservation
:
1110 set_opt(sbi
->s_mount_opt
, RESERVATION
);
1112 case Opt_noreservation
:
1113 clear_opt(sbi
->s_mount_opt
, RESERVATION
);
1115 case Opt_journal_update
:
1117 /* Eventually we will want to be able to create
1118 a journal file here. For now, only allow the
1119 user to specify an existing inode to be the
1122 printk(KERN_ERR
"EXT4-fs: cannot specify "
1123 "journal on remount\n");
1126 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1128 case Opt_journal_inum
:
1130 printk(KERN_ERR
"EXT4-fs: cannot specify "
1131 "journal on remount\n");
1134 if (match_int(&args
[0], &option
))
1138 case Opt_journal_dev
:
1140 printk(KERN_ERR
"EXT4-fs: cannot specify "
1141 "journal on remount\n");
1144 if (match_int(&args
[0], &option
))
1146 *journal_devnum
= option
;
1148 case Opt_journal_checksum
:
1149 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1151 case Opt_journal_async_commit
:
1152 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1153 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1156 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1159 if (match_int(&args
[0], &option
))
1164 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1165 sbi
->s_commit_interval
= HZ
* option
;
1167 case Opt_data_journal
:
1168 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1170 case Opt_data_ordered
:
1171 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1173 case Opt_data_writeback
:
1174 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1177 if ((sbi
->s_mount_opt
& EXT4_MOUNT_DATA_FLAGS
)
1180 "EXT4-fs: cannot change data "
1181 "mode on remount\n");
1185 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DATA_FLAGS
;
1186 sbi
->s_mount_opt
|= data_opt
;
1189 case Opt_data_err_abort
:
1190 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1192 case Opt_data_err_ignore
:
1193 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1202 if ((sb_any_quota_enabled(sb
) ||
1203 sb_any_quota_suspended(sb
)) &&
1204 !sbi
->s_qf_names
[qtype
]) {
1206 "EXT4-fs: Cannot change journaled "
1207 "quota options when quota turned on.\n");
1210 qname
= match_strdup(&args
[0]);
1213 "EXT4-fs: not enough memory for "
1214 "storing quotafile name.\n");
1217 if (sbi
->s_qf_names
[qtype
] &&
1218 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1220 "EXT4-fs: %s quota file already "
1221 "specified.\n", QTYPE2NAME(qtype
));
1225 sbi
->s_qf_names
[qtype
] = qname
;
1226 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1228 "EXT4-fs: quotafile must be on "
1229 "filesystem root.\n");
1230 kfree(sbi
->s_qf_names
[qtype
]);
1231 sbi
->s_qf_names
[qtype
] = NULL
;
1234 set_opt(sbi
->s_mount_opt
, QUOTA
);
1236 case Opt_offusrjquota
:
1239 case Opt_offgrpjquota
:
1242 if ((sb_any_quota_enabled(sb
) ||
1243 sb_any_quota_suspended(sb
)) &&
1244 sbi
->s_qf_names
[qtype
]) {
1245 printk(KERN_ERR
"EXT4-fs: Cannot change "
1246 "journaled quota options when "
1247 "quota turned on.\n");
1251 * The space will be released later when all options
1252 * are confirmed to be correct
1254 sbi
->s_qf_names
[qtype
] = NULL
;
1256 case Opt_jqfmt_vfsold
:
1257 qfmt
= QFMT_VFS_OLD
;
1259 case Opt_jqfmt_vfsv0
:
1262 if ((sb_any_quota_enabled(sb
) ||
1263 sb_any_quota_suspended(sb
)) &&
1264 sbi
->s_jquota_fmt
!= qfmt
) {
1265 printk(KERN_ERR
"EXT4-fs: Cannot change "
1266 "journaled quota options when "
1267 "quota turned on.\n");
1270 sbi
->s_jquota_fmt
= qfmt
;
1274 set_opt(sbi
->s_mount_opt
, QUOTA
);
1275 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1278 set_opt(sbi
->s_mount_opt
, QUOTA
);
1279 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1282 if (sb_any_quota_enabled(sb
)) {
1283 printk(KERN_ERR
"EXT4-fs: Cannot change quota "
1284 "options when quota turned on.\n");
1287 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1288 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1289 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1296 "EXT4-fs: quota options not supported.\n");
1300 case Opt_offusrjquota
:
1301 case Opt_offgrpjquota
:
1302 case Opt_jqfmt_vfsold
:
1303 case Opt_jqfmt_vfsv0
:
1305 "EXT4-fs: journaled quota options not "
1312 set_opt(sbi
->s_mount_opt
, ABORT
);
1315 if (match_int(&args
[0], &option
))
1318 set_opt(sbi
->s_mount_opt
, BARRIER
);
1320 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1326 printk("EXT4-fs: resize option only available "
1330 if (match_int(&args
[0], &option
) != 0)
1332 *n_blocks_count
= option
;
1335 set_opt(sbi
->s_mount_opt
, NOBH
);
1338 clear_opt(sbi
->s_mount_opt
, NOBH
);
1341 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
,
1342 EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
1343 ext4_warning(sb
, __func__
,
1344 "extents feature not enabled "
1345 "on this filesystem, use tune2fs\n");
1348 set_opt(sbi
->s_mount_opt
, EXTENTS
);
1352 * When e2fsprogs support resizing an already existing
1353 * ext3 file system to greater than 2**32 we need to
1354 * add support to block allocator to handle growing
1355 * already existing block mapped inode so that blocks
1356 * allocated for them fall within 2**32
1358 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1359 if (last_block
> 0xffffffffULL
) {
1360 printk(KERN_ERR
"EXT4-fs: Filesystem too "
1361 "large to mount with "
1362 "-o noextents options\n");
1365 clear_opt(sbi
->s_mount_opt
, EXTENTS
);
1368 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1369 sb
->s_flags
|= MS_I_VERSION
;
1371 case Opt_nodelalloc
:
1372 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1375 if (match_int(&args
[0], &option
))
1379 sbi
->s_stripe
= option
;
1382 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1384 case Opt_inode_readahead_blks
:
1385 if (match_int(&args
[0], &option
))
1387 if (option
< 0 || option
> (1 << 30))
1389 sbi
->s_inode_readahead_blks
= option
;
1393 "EXT4-fs: Unrecognized mount option \"%s\" "
1394 "or missing value\n", p
);
1399 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1400 if ((sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
) &&
1401 sbi
->s_qf_names
[USRQUOTA
])
1402 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1404 if ((sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
) &&
1405 sbi
->s_qf_names
[GRPQUOTA
])
1406 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1408 if ((sbi
->s_qf_names
[USRQUOTA
] &&
1409 (sbi
->s_mount_opt
& EXT4_MOUNT_GRPQUOTA
)) ||
1410 (sbi
->s_qf_names
[GRPQUOTA
] &&
1411 (sbi
->s_mount_opt
& EXT4_MOUNT_USRQUOTA
))) {
1412 printk(KERN_ERR
"EXT4-fs: old and new quota "
1413 "format mixing.\n");
1417 if (!sbi
->s_jquota_fmt
) {
1418 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1419 "not specified.\n");
1423 if (sbi
->s_jquota_fmt
) {
1424 printk(KERN_ERR
"EXT4-fs: journaled quota format "
1425 "specified with no journaling "
1434 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1437 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1440 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1441 printk(KERN_ERR
"EXT4-fs warning: revision level too high, "
1442 "forcing read-only mode\n");
1447 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1448 printk(KERN_WARNING
"EXT4-fs warning: mounting unchecked fs, "
1449 "running e2fsck is recommended\n");
1450 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1452 "EXT4-fs warning: mounting fs with errors, "
1453 "running e2fsck is recommended\n");
1454 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1455 le16_to_cpu(es
->s_mnt_count
) >=
1456 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1458 "EXT4-fs warning: maximal mount count reached, "
1459 "running e2fsck is recommended\n");
1460 else if (le32_to_cpu(es
->s_checkinterval
) &&
1461 (le32_to_cpu(es
->s_lastcheck
) +
1462 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1464 "EXT4-fs warning: checktime reached, "
1465 "running e2fsck is recommended\n");
1467 /* @@@ We _will_ want to clear the valid bit if we find
1468 * inconsistencies, to force a fsck at reboot. But for
1469 * a plain journaled filesystem we can keep it set as
1472 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1474 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1475 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1476 le16_add_cpu(&es
->s_mnt_count
, 1);
1477 es
->s_mtime
= cpu_to_le32(get_seconds());
1478 ext4_update_dynamic_rev(sb
);
1479 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1481 ext4_commit_super(sb
, es
, 1);
1482 if (test_opt(sb
, DEBUG
))
1483 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%lu, "
1484 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1486 sbi
->s_groups_count
,
1487 EXT4_BLOCKS_PER_GROUP(sb
),
1488 EXT4_INODES_PER_GROUP(sb
),
1491 printk(KERN_INFO
"EXT4 FS on %s, %s journal on %s\n",
1492 sb
->s_id
, EXT4_SB(sb
)->s_journal
->j_inode
? "internal" :
1493 "external", EXT4_SB(sb
)->s_journal
->j_devname
);
1497 static int ext4_fill_flex_info(struct super_block
*sb
)
1499 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1500 struct ext4_group_desc
*gdp
= NULL
;
1501 struct buffer_head
*bh
;
1502 ext4_group_t flex_group_count
;
1503 ext4_group_t flex_group
;
1504 int groups_per_flex
= 0;
1505 __u64 block_bitmap
= 0;
1508 if (!sbi
->s_es
->s_log_groups_per_flex
) {
1509 sbi
->s_log_groups_per_flex
= 0;
1513 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1514 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1516 /* We allocate both existing and potentially added groups */
1517 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1518 ((sbi
->s_es
->s_reserved_gdt_blocks
+1 ) <<
1519 EXT4_DESC_PER_BLOCK_BITS(sb
))) /
1521 sbi
->s_flex_groups
= kzalloc(flex_group_count
*
1522 sizeof(struct flex_groups
), GFP_KERNEL
);
1523 if (sbi
->s_flex_groups
== NULL
) {
1524 printk(KERN_ERR
"EXT4-fs: not enough memory for "
1525 "%lu flex groups\n", flex_group_count
);
1529 gdp
= ext4_get_group_desc(sb
, 1, &bh
);
1530 block_bitmap
= ext4_block_bitmap(sb
, gdp
) - 1;
1532 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1533 gdp
= ext4_get_group_desc(sb
, i
, &bh
);
1535 flex_group
= ext4_flex_group(sbi
, i
);
1536 sbi
->s_flex_groups
[flex_group
].free_inodes
+=
1537 le16_to_cpu(gdp
->bg_free_inodes_count
);
1538 sbi
->s_flex_groups
[flex_group
].free_blocks
+=
1539 le16_to_cpu(gdp
->bg_free_blocks_count
);
1547 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1548 struct ext4_group_desc
*gdp
)
1552 if (sbi
->s_es
->s_feature_ro_compat
&
1553 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1554 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1555 __le32 le_group
= cpu_to_le32(block_group
);
1557 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1558 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1559 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1560 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1561 /* for checksum of struct ext4_group_desc do the rest...*/
1562 if ((sbi
->s_es
->s_feature_incompat
&
1563 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1564 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1565 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1566 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1570 return cpu_to_le16(crc
);
1573 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1574 struct ext4_group_desc
*gdp
)
1576 if ((sbi
->s_es
->s_feature_ro_compat
&
1577 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1578 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1584 /* Called at mount-time, super-block is locked */
1585 static int ext4_check_descriptors(struct super_block
*sb
)
1587 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1588 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1589 ext4_fsblk_t last_block
;
1590 ext4_fsblk_t block_bitmap
;
1591 ext4_fsblk_t inode_bitmap
;
1592 ext4_fsblk_t inode_table
;
1593 int flexbg_flag
= 0;
1596 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1599 ext4_debug("Checking group descriptors");
1601 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1602 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1604 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1605 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1607 last_block
= first_block
+
1608 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1610 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1611 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1612 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1613 "Block bitmap for group %lu not in group "
1614 "(block %llu)!\n", i
, block_bitmap
);
1617 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1618 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1619 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1620 "Inode bitmap for group %lu not in group "
1621 "(block %llu)!\n", i
, inode_bitmap
);
1624 inode_table
= ext4_inode_table(sb
, gdp
);
1625 if (inode_table
< first_block
||
1626 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1627 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1628 "Inode table for group %lu not in group "
1629 "(block %llu)!\n", i
, inode_table
);
1632 spin_lock(sb_bgl_lock(sbi
, i
));
1633 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1634 printk(KERN_ERR
"EXT4-fs: ext4_check_descriptors: "
1635 "Checksum for group %lu failed (%u!=%u)\n",
1636 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1637 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1638 if (!(sb
->s_flags
& MS_RDONLY
)) {
1639 spin_unlock(sb_bgl_lock(sbi
, i
));
1643 spin_unlock(sb_bgl_lock(sbi
, i
));
1645 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1648 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1649 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(ext4_count_free_inodes(sb
));
1653 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1654 * the superblock) which were deleted from all directories, but held open by
1655 * a process at the time of a crash. We walk the list and try to delete these
1656 * inodes at recovery time (only with a read-write filesystem).
1658 * In order to keep the orphan inode chain consistent during traversal (in
1659 * case of crash during recovery), we link each inode into the superblock
1660 * orphan list_head and handle it the same way as an inode deletion during
1661 * normal operation (which journals the operations for us).
1663 * We only do an iget() and an iput() on each inode, which is very safe if we
1664 * accidentally point at an in-use or already deleted inode. The worst that
1665 * can happen in this case is that we get a "bit already cleared" message from
1666 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1667 * e2fsck was run on this filesystem, and it must have already done the orphan
1668 * inode cleanup for us, so we can safely abort without any further action.
1670 static void ext4_orphan_cleanup(struct super_block
*sb
,
1671 struct ext4_super_block
*es
)
1673 unsigned int s_flags
= sb
->s_flags
;
1674 int nr_orphans
= 0, nr_truncates
= 0;
1678 if (!es
->s_last_orphan
) {
1679 jbd_debug(4, "no orphan inodes to clean up\n");
1683 if (bdev_read_only(sb
->s_bdev
)) {
1684 printk(KERN_ERR
"EXT4-fs: write access "
1685 "unavailable, skipping orphan cleanup.\n");
1689 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1690 if (es
->s_last_orphan
)
1691 jbd_debug(1, "Errors on filesystem, "
1692 "clearing orphan list.\n");
1693 es
->s_last_orphan
= 0;
1694 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1698 if (s_flags
& MS_RDONLY
) {
1699 printk(KERN_INFO
"EXT4-fs: %s: orphan cleanup on readonly fs\n",
1701 sb
->s_flags
&= ~MS_RDONLY
;
1704 /* Needed for iput() to work correctly and not trash data */
1705 sb
->s_flags
|= MS_ACTIVE
;
1706 /* Turn on quotas so that they are updated correctly */
1707 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1708 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
1709 int ret
= ext4_quota_on_mount(sb
, i
);
1712 "EXT4-fs: Cannot turn on journaled "
1713 "quota: error %d\n", ret
);
1718 while (es
->s_last_orphan
) {
1719 struct inode
*inode
;
1721 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1722 if (IS_ERR(inode
)) {
1723 es
->s_last_orphan
= 0;
1727 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
1729 if (inode
->i_nlink
) {
1731 "%s: truncating inode %lu to %lld bytes\n",
1732 __func__
, inode
->i_ino
, inode
->i_size
);
1733 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1734 inode
->i_ino
, inode
->i_size
);
1735 ext4_truncate(inode
);
1739 "%s: deleting unreferenced inode %lu\n",
1740 __func__
, inode
->i_ino
);
1741 jbd_debug(2, "deleting unreferenced inode %lu\n",
1745 iput(inode
); /* The delete magic happens here! */
1748 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1751 printk(KERN_INFO
"EXT4-fs: %s: %d orphan inode%s deleted\n",
1752 sb
->s_id
, PLURAL(nr_orphans
));
1754 printk(KERN_INFO
"EXT4-fs: %s: %d truncate%s cleaned up\n",
1755 sb
->s_id
, PLURAL(nr_truncates
));
1757 /* Turn quotas off */
1758 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1759 if (sb_dqopt(sb
)->files
[i
])
1760 vfs_quota_off(sb
, i
, 0);
1763 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
1766 * Maximal extent format file size.
1767 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1768 * extent format containers, within a sector_t, and within i_blocks
1769 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1770 * so that won't be a limiting factor.
1772 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1774 static loff_t
ext4_max_size(int blkbits
)
1777 loff_t upper_limit
= MAX_LFS_FILESIZE
;
1779 /* small i_blocks in vfs inode? */
1780 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
1782 * CONFIG_LSF is not enabled implies the inode
1783 * i_block represent total blocks in 512 bytes
1784 * 32 == size of vfs inode i_blocks * 8
1786 upper_limit
= (1LL << 32) - 1;
1788 /* total blocks in file system block size */
1789 upper_limit
>>= (blkbits
- 9);
1790 upper_limit
<<= blkbits
;
1793 /* 32-bit extent-start container, ee_block */
1798 /* Sanity check against vm- & vfs- imposed limits */
1799 if (res
> upper_limit
)
1806 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1807 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1808 * We need to be 1 filesystem block less than the 2^48 sector limit.
1810 static loff_t
ext4_max_bitmap_size(int bits
)
1812 loff_t res
= EXT4_NDIR_BLOCKS
;
1815 /* This is calculated to be the largest file size for a
1816 * dense, bitmapped file such that the total number of
1817 * sectors in the file, including data and all indirect blocks,
1818 * does not exceed 2^48 -1
1819 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1820 * total number of 512 bytes blocks of the file
1823 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
1825 * CONFIG_LSF is not enabled implies the inode
1826 * i_block represent total blocks in 512 bytes
1827 * 32 == size of vfs inode i_blocks * 8
1829 upper_limit
= (1LL << 32) - 1;
1831 /* total blocks in file system block size */
1832 upper_limit
>>= (bits
- 9);
1836 * We use 48 bit ext4_inode i_blocks
1837 * With EXT4_HUGE_FILE_FL set the i_blocks
1838 * represent total number of blocks in
1839 * file system block size
1841 upper_limit
= (1LL << 48) - 1;
1845 /* indirect blocks */
1847 /* double indirect blocks */
1848 meta_blocks
+= 1 + (1LL << (bits
-2));
1849 /* tripple indirect blocks */
1850 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
1852 upper_limit
-= meta_blocks
;
1853 upper_limit
<<= bits
;
1855 res
+= 1LL << (bits
-2);
1856 res
+= 1LL << (2*(bits
-2));
1857 res
+= 1LL << (3*(bits
-2));
1859 if (res
> upper_limit
)
1862 if (res
> MAX_LFS_FILESIZE
)
1863 res
= MAX_LFS_FILESIZE
;
1868 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
1869 ext4_fsblk_t logical_sb_block
, int nr
)
1871 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1872 ext4_group_t bg
, first_meta_bg
;
1875 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
1877 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
1879 return logical_sb_block
+ nr
+ 1;
1880 bg
= sbi
->s_desc_per_block
* nr
;
1881 if (ext4_bg_has_super(sb
, bg
))
1883 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
1887 * ext4_get_stripe_size: Get the stripe size.
1888 * @sbi: In memory super block info
1890 * If we have specified it via mount option, then
1891 * use the mount option value. If the value specified at mount time is
1892 * greater than the blocks per group use the super block value.
1893 * If the super block value is greater than blocks per group return 0.
1894 * Allocator needs it be less than blocks per group.
1897 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
1899 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
1900 unsigned long stripe_width
=
1901 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
1903 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
1904 return sbi
->s_stripe
;
1906 if (stripe_width
<= sbi
->s_blocks_per_group
)
1907 return stripe_width
;
1909 if (stride
<= sbi
->s_blocks_per_group
)
1915 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
1916 __releases(kernel_lock
)
1917 __acquires(kernel_lock
)
1920 struct buffer_head
*bh
;
1921 struct ext4_super_block
*es
= NULL
;
1922 struct ext4_sb_info
*sbi
;
1924 ext4_fsblk_t sb_block
= get_sb_block(&data
);
1925 ext4_fsblk_t logical_sb_block
;
1926 unsigned long offset
= 0;
1927 unsigned int journal_inum
= 0;
1928 unsigned long journal_devnum
= 0;
1929 unsigned long def_mount_opts
;
1941 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
1944 sb
->s_fs_info
= sbi
;
1945 sbi
->s_mount_opt
= 0;
1946 sbi
->s_resuid
= EXT4_DEF_RESUID
;
1947 sbi
->s_resgid
= EXT4_DEF_RESGID
;
1948 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
1949 sbi
->s_sb_block
= sb_block
;
1953 /* Cleanup superblock name */
1954 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
1957 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
1959 printk(KERN_ERR
"EXT4-fs: unable to set blocksize\n");
1964 * The ext4 superblock will not be buffer aligned for other than 1kB
1965 * block sizes. We need to calculate the offset from buffer start.
1967 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
1968 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
1969 offset
= do_div(logical_sb_block
, blocksize
);
1971 logical_sb_block
= sb_block
;
1974 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
1975 printk(KERN_ERR
"EXT4-fs: unable to read superblock\n");
1979 * Note: s_es must be initialized as soon as possible because
1980 * some ext4 macro-instructions depend on its value
1982 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
1984 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
1985 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
1988 /* Set defaults before we parse the mount options */
1989 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1990 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
1991 set_opt(sbi
->s_mount_opt
, DEBUG
);
1992 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
1993 set_opt(sbi
->s_mount_opt
, GRPID
);
1994 if (def_mount_opts
& EXT4_DEFM_UID16
)
1995 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1996 #ifdef CONFIG_EXT4_FS_XATTR
1997 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
1998 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
2000 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2001 if (def_mount_opts
& EXT4_DEFM_ACL
)
2002 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
2004 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
2005 sbi
->s_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
2006 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
2007 sbi
->s_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
2008 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
2009 sbi
->s_mount_opt
|= EXT4_MOUNT_WRITEBACK_DATA
;
2011 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2012 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2013 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2014 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2016 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2018 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2019 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2021 set_opt(sbi
->s_mount_opt
, RESERVATION
);
2022 set_opt(sbi
->s_mount_opt
, BARRIER
);
2025 * turn on extents feature by default in ext4 filesystem
2026 * only if feature flag already set by mkfs or tune2fs.
2027 * Use -o noextents to turn it off
2029 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
))
2030 set_opt(sbi
->s_mount_opt
, EXTENTS
);
2032 ext4_warning(sb
, __func__
,
2033 "extents feature not enabled on this filesystem, "
2037 * enable delayed allocation by default
2038 * Use -o nodelalloc to turn it off
2040 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2043 if (!parse_options((char *) data
, sb
, &journal_inum
, &journal_devnum
,
2047 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2048 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
2050 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2051 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2052 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2053 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2055 "EXT4-fs warning: feature flags set on rev 0 fs, "
2056 "running e2fsck is recommended\n");
2059 * Check feature flags regardless of the revision level, since we
2060 * previously didn't change the revision level when setting the flags,
2061 * so there is a chance incompat flags are set on a rev 0 filesystem.
2063 features
= EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
);
2065 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount because of "
2066 "unsupported optional features (%x).\n",
2067 sb
->s_id
, le32_to_cpu(features
));
2070 features
= EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
);
2071 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
2072 printk(KERN_ERR
"EXT4-fs: %s: couldn't mount RDWR because of "
2073 "unsupported optional features (%x).\n",
2074 sb
->s_id
, le32_to_cpu(features
));
2077 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2079 * Large file size enabled file system can only be
2080 * mount if kernel is build with CONFIG_LSF
2082 if (sizeof(root
->i_blocks
) < sizeof(u64
) &&
2083 !(sb
->s_flags
& MS_RDONLY
)) {
2084 printk(KERN_ERR
"EXT4-fs: %s: Filesystem with huge "
2085 "files cannot be mounted read-write "
2086 "without CONFIG_LSF.\n", sb
->s_id
);
2090 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2092 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2093 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2095 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2096 blocksize
, sb
->s_id
);
2100 if (sb
->s_blocksize
!= blocksize
) {
2102 /* Validate the filesystem blocksize */
2103 if (!sb_set_blocksize(sb
, blocksize
)) {
2104 printk(KERN_ERR
"EXT4-fs: bad block size %d.\n",
2110 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2111 offset
= do_div(logical_sb_block
, blocksize
);
2112 bh
= sb_bread(sb
, logical_sb_block
);
2115 "EXT4-fs: Can't read superblock on 2nd try.\n");
2118 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2120 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2122 "EXT4-fs: Magic mismatch, very weird !\n");
2127 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
);
2128 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
);
2130 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2131 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2132 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2134 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2135 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2136 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2137 (!is_power_of_2(sbi
->s_inode_size
)) ||
2138 (sbi
->s_inode_size
> blocksize
)) {
2140 "EXT4-fs: unsupported inode size: %d\n",
2144 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2145 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2147 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2148 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2149 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2150 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2151 !is_power_of_2(sbi
->s_desc_size
)) {
2153 "EXT4-fs: unsupported descriptor size %lu\n",
2158 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2159 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2160 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2161 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2163 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2164 if (sbi
->s_inodes_per_block
== 0)
2166 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2167 sbi
->s_inodes_per_block
;
2168 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2170 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2171 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2172 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2173 for (i
= 0; i
< 4; i
++)
2174 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2175 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2177 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2179 "EXT4-fs: #blocks per group too big: %lu\n",
2180 sbi
->s_blocks_per_group
);
2183 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2185 "EXT4-fs: #inodes per group too big: %lu\n",
2186 sbi
->s_inodes_per_group
);
2190 if (ext4_blocks_count(es
) >
2191 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) {
2192 printk(KERN_ERR
"EXT4-fs: filesystem on %s:"
2193 " too large to mount safely\n", sb
->s_id
);
2194 if (sizeof(sector_t
) < 8)
2195 printk(KERN_WARNING
"EXT4-fs: CONFIG_LBD not "
2200 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2203 /* ensure blocks_count calculation below doesn't sign-extend */
2204 if (ext4_blocks_count(es
) + EXT4_BLOCKS_PER_GROUP(sb
) <
2205 le32_to_cpu(es
->s_first_data_block
) + 1) {
2206 printk(KERN_WARNING
"EXT4-fs: bad geometry: block count %llu, "
2207 "first data block %u, blocks per group %lu\n",
2208 ext4_blocks_count(es
),
2209 le32_to_cpu(es
->s_first_data_block
),
2210 EXT4_BLOCKS_PER_GROUP(sb
));
2213 blocks_count
= (ext4_blocks_count(es
) -
2214 le32_to_cpu(es
->s_first_data_block
) +
2215 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2216 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2217 sbi
->s_groups_count
= blocks_count
;
2218 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2219 EXT4_DESC_PER_BLOCK(sb
);
2220 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2222 if (sbi
->s_group_desc
== NULL
) {
2223 printk(KERN_ERR
"EXT4-fs: not enough memory\n");
2227 #ifdef CONFIG_PROC_FS
2229 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2232 proc_create_data("inode_readahead_blks", 0644, sbi
->s_proc
,
2234 &sbi
->s_inode_readahead_blks
);
2237 bgl_lock_init(&sbi
->s_blockgroup_lock
);
2239 for (i
= 0; i
< db_count
; i
++) {
2240 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2241 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2242 if (!sbi
->s_group_desc
[i
]) {
2243 printk(KERN_ERR
"EXT4-fs: "
2244 "can't read group descriptor %d\n", i
);
2249 if (!ext4_check_descriptors(sb
)) {
2250 printk(KERN_ERR
"EXT4-fs: group descriptors corrupted!\n");
2253 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2254 if (!ext4_fill_flex_info(sb
)) {
2256 "EXT4-fs: unable to initialize "
2257 "flex_bg meta info!\n");
2261 sbi
->s_gdb_count
= db_count
;
2262 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2263 spin_lock_init(&sbi
->s_next_gen_lock
);
2265 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2266 ext4_count_free_blocks(sb
));
2268 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2269 ext4_count_free_inodes(sb
));
2272 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2273 ext4_count_dirs(sb
));
2276 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
2279 printk(KERN_ERR
"EXT4-fs: insufficient memory\n");
2283 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2286 * set up enough so that it can read an inode
2288 sb
->s_op
= &ext4_sops
;
2289 sb
->s_export_op
= &ext4_export_ops
;
2290 sb
->s_xattr
= ext4_xattr_handlers
;
2292 sb
->s_qcop
= &ext4_qctl_operations
;
2293 sb
->dq_op
= &ext4_quota_operations
;
2295 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2299 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2300 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2301 EXT4_FEATURE_INCOMPAT_RECOVER
));
2304 * The first inode we look at is the journal inode. Don't try
2305 * root first: it may be modified in the journal!
2307 if (!test_opt(sb
, NOLOAD
) &&
2308 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2309 if (ext4_load_journal(sb
, es
, journal_devnum
))
2311 if (!(sb
->s_flags
& MS_RDONLY
) &&
2312 EXT4_SB(sb
)->s_journal
->j_failed_commit
) {
2313 printk(KERN_CRIT
"EXT4-fs error (device %s): "
2314 "ext4_fill_super: Journal transaction "
2315 "%u is corrupt\n", sb
->s_id
,
2316 EXT4_SB(sb
)->s_journal
->j_failed_commit
);
2317 if (test_opt(sb
, ERRORS_RO
)) {
2319 "Mounting filesystem read-only\n");
2320 sb
->s_flags
|= MS_RDONLY
;
2321 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2322 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2324 if (test_opt(sb
, ERRORS_PANIC
)) {
2325 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2326 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2327 ext4_commit_super(sb
, es
, 1);
2329 "EXT4-fs (device %s): mount failed\n",
2334 } else if (journal_inum
) {
2335 if (ext4_create_journal(sb
, es
, journal_inum
))
2340 "ext4: No journal on filesystem on %s\n",
2345 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2346 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2347 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2348 printk(KERN_ERR
"ext4: Failed to set 64-bit journal feature\n");
2352 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2353 jbd2_journal_set_features(sbi
->s_journal
,
2354 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2355 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2356 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2357 jbd2_journal_set_features(sbi
->s_journal
,
2358 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2359 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2360 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2362 jbd2_journal_clear_features(sbi
->s_journal
,
2363 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2364 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2367 /* We have now updated the journal if required, so we can
2368 * validate the data journaling mode. */
2369 switch (test_opt(sb
, DATA_FLAGS
)) {
2371 /* No mode set, assume a default based on the journal
2372 * capabilities: ORDERED_DATA if the journal can
2373 * cope, else JOURNAL_DATA
2375 if (jbd2_journal_check_available_features
2376 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2377 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2379 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2382 case EXT4_MOUNT_ORDERED_DATA
:
2383 case EXT4_MOUNT_WRITEBACK_DATA
:
2384 if (!jbd2_journal_check_available_features
2385 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2386 printk(KERN_ERR
"EXT4-fs: Journal does not support "
2387 "requested data journaling mode\n");
2394 if (test_opt(sb
, NOBH
)) {
2395 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2396 printk(KERN_WARNING
"EXT4-fs: Ignoring nobh option - "
2397 "its supported only with writeback mode\n");
2398 clear_opt(sbi
->s_mount_opt
, NOBH
);
2402 * The jbd2_journal_load will have done any necessary log recovery,
2403 * so we can safely mount the rest of the filesystem now.
2406 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2408 printk(KERN_ERR
"EXT4-fs: get root inode failed\n");
2409 ret
= PTR_ERR(root
);
2412 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2414 printk(KERN_ERR
"EXT4-fs: corrupt root inode, run e2fsck\n");
2417 sb
->s_root
= d_alloc_root(root
);
2419 printk(KERN_ERR
"EXT4-fs: get root dentry failed\n");
2425 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
2427 /* determine the minimum size of new large inodes, if present */
2428 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2429 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2430 EXT4_GOOD_OLD_INODE_SIZE
;
2431 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2432 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2433 if (sbi
->s_want_extra_isize
<
2434 le16_to_cpu(es
->s_want_extra_isize
))
2435 sbi
->s_want_extra_isize
=
2436 le16_to_cpu(es
->s_want_extra_isize
);
2437 if (sbi
->s_want_extra_isize
<
2438 le16_to_cpu(es
->s_min_extra_isize
))
2439 sbi
->s_want_extra_isize
=
2440 le16_to_cpu(es
->s_min_extra_isize
);
2443 /* Check if enough inode space is available */
2444 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2445 sbi
->s_inode_size
) {
2446 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2447 EXT4_GOOD_OLD_INODE_SIZE
;
2448 printk(KERN_INFO
"EXT4-fs: required extra inode space not"
2452 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
2453 printk(KERN_WARNING
"EXT4-fs: Ignoring delalloc option - "
2454 "requested data journaling mode\n");
2455 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2456 } else if (test_opt(sb
, DELALLOC
))
2457 printk(KERN_INFO
"EXT4-fs: delayed allocation enabled\n");
2460 err
= ext4_mb_init(sb
, needs_recovery
);
2462 printk(KERN_ERR
"EXT4-fs: failed to initalize mballoc (%d)\n",
2468 * akpm: core read_super() calls in here with the superblock locked.
2469 * That deadlocks, because orphan cleanup needs to lock the superblock
2470 * in numerous places. Here we just pop the lock - it's relatively
2471 * harmless, because we are now ready to accept write_super() requests,
2472 * and aviro says that's the only reason for hanging onto the
2475 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
2476 ext4_orphan_cleanup(sb
, es
);
2477 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
2479 printk(KERN_INFO
"EXT4-fs: recovery complete.\n");
2480 ext4_mark_recovery_complete(sb
, es
);
2481 printk(KERN_INFO
"EXT4-fs: mounted filesystem with %s data mode.\n",
2482 test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
? "journal":
2483 test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
? "ordered":
2491 printk(KERN_ERR
"VFS: Can't find ext4 filesystem on dev %s.\n",
2496 jbd2_journal_destroy(sbi
->s_journal
);
2497 sbi
->s_journal
= NULL
;
2499 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2500 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
2501 percpu_counter_destroy(&sbi
->s_dirs_counter
);
2502 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
2504 for (i
= 0; i
< db_count
; i
++)
2505 brelse(sbi
->s_group_desc
[i
]);
2506 kfree(sbi
->s_group_desc
);
2509 remove_proc_entry("inode_readahead_blks", sbi
->s_proc
);
2510 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
2513 for (i
= 0; i
< MAXQUOTAS
; i
++)
2514 kfree(sbi
->s_qf_names
[i
]);
2516 ext4_blkdev_remove(sbi
);
2519 sb
->s_fs_info
= NULL
;
2526 * Setup any per-fs journal parameters now. We'll do this both on
2527 * initial mount, once the journal has been initialised but before we've
2528 * done any recovery; and again on any subsequent remount.
2530 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
2532 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2534 if (sbi
->s_commit_interval
)
2535 journal
->j_commit_interval
= sbi
->s_commit_interval
;
2536 /* We could also set up an ext4-specific default for the commit
2537 * interval here, but for now we'll just fall back to the jbd
2540 spin_lock(&journal
->j_state_lock
);
2541 if (test_opt(sb
, BARRIER
))
2542 journal
->j_flags
|= JBD2_BARRIER
;
2544 journal
->j_flags
&= ~JBD2_BARRIER
;
2545 if (test_opt(sb
, DATA_ERR_ABORT
))
2546 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
2548 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
2549 spin_unlock(&journal
->j_state_lock
);
2552 static journal_t
*ext4_get_journal(struct super_block
*sb
,
2553 unsigned int journal_inum
)
2555 struct inode
*journal_inode
;
2558 /* First, test for the existence of a valid inode on disk. Bad
2559 * things happen if we iget() an unused inode, as the subsequent
2560 * iput() will try to delete it. */
2562 journal_inode
= ext4_iget(sb
, journal_inum
);
2563 if (IS_ERR(journal_inode
)) {
2564 printk(KERN_ERR
"EXT4-fs: no journal found.\n");
2567 if (!journal_inode
->i_nlink
) {
2568 make_bad_inode(journal_inode
);
2569 iput(journal_inode
);
2570 printk(KERN_ERR
"EXT4-fs: journal inode is deleted.\n");
2574 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2575 journal_inode
, journal_inode
->i_size
);
2576 if (!S_ISREG(journal_inode
->i_mode
)) {
2577 printk(KERN_ERR
"EXT4-fs: invalid journal inode.\n");
2578 iput(journal_inode
);
2582 journal
= jbd2_journal_init_inode(journal_inode
);
2584 printk(KERN_ERR
"EXT4-fs: Could not load journal inode\n");
2585 iput(journal_inode
);
2588 journal
->j_private
= sb
;
2589 ext4_init_journal_params(sb
, journal
);
2593 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
2596 struct buffer_head
*bh
;
2600 int hblock
, blocksize
;
2601 ext4_fsblk_t sb_block
;
2602 unsigned long offset
;
2603 struct ext4_super_block
*es
;
2604 struct block_device
*bdev
;
2606 bdev
= ext4_blkdev_get(j_dev
);
2610 if (bd_claim(bdev
, sb
)) {
2612 "EXT4: failed to claim external journal device.\n");
2617 blocksize
= sb
->s_blocksize
;
2618 hblock
= bdev_hardsect_size(bdev
);
2619 if (blocksize
< hblock
) {
2621 "EXT4-fs: blocksize too small for journal device.\n");
2625 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
2626 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
2627 set_blocksize(bdev
, blocksize
);
2628 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
2629 printk(KERN_ERR
"EXT4-fs: couldn't read superblock of "
2630 "external journal\n");
2634 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2635 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
2636 !(le32_to_cpu(es
->s_feature_incompat
) &
2637 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
2638 printk(KERN_ERR
"EXT4-fs: external journal has "
2639 "bad superblock\n");
2644 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
2645 printk(KERN_ERR
"EXT4-fs: journal UUID does not match\n");
2650 len
= ext4_blocks_count(es
);
2651 start
= sb_block
+ 1;
2652 brelse(bh
); /* we're done with the superblock */
2654 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
2655 start
, len
, blocksize
);
2657 printk(KERN_ERR
"EXT4-fs: failed to create device journal\n");
2660 journal
->j_private
= sb
;
2661 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
2662 wait_on_buffer(journal
->j_sb_buffer
);
2663 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
2664 printk(KERN_ERR
"EXT4-fs: I/O error on journal device\n");
2667 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
2668 printk(KERN_ERR
"EXT4-fs: External journal has more than one "
2669 "user (unsupported) - %d\n",
2670 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
2673 EXT4_SB(sb
)->journal_bdev
= bdev
;
2674 ext4_init_journal_params(sb
, journal
);
2677 jbd2_journal_destroy(journal
);
2679 ext4_blkdev_put(bdev
);
2683 static int ext4_load_journal(struct super_block
*sb
,
2684 struct ext4_super_block
*es
,
2685 unsigned long journal_devnum
)
2688 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
2691 int really_read_only
;
2693 if (journal_devnum
&&
2694 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2695 printk(KERN_INFO
"EXT4-fs: external journal device major/minor "
2696 "numbers have changed\n");
2697 journal_dev
= new_decode_dev(journal_devnum
);
2699 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
2701 really_read_only
= bdev_read_only(sb
->s_bdev
);
2704 * Are we loading a blank journal or performing recovery after a
2705 * crash? For recovery, we need to check in advance whether we
2706 * can get read-write access to the device.
2709 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2710 if (sb
->s_flags
& MS_RDONLY
) {
2711 printk(KERN_INFO
"EXT4-fs: INFO: recovery "
2712 "required on readonly filesystem.\n");
2713 if (really_read_only
) {
2714 printk(KERN_ERR
"EXT4-fs: write access "
2715 "unavailable, cannot proceed.\n");
2718 printk(KERN_INFO
"EXT4-fs: write access will "
2719 "be enabled during recovery.\n");
2723 if (journal_inum
&& journal_dev
) {
2724 printk(KERN_ERR
"EXT4-fs: filesystem has both journal "
2725 "and inode journals!\n");
2730 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
2733 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
2737 if (journal
->j_flags
& JBD2_BARRIER
)
2738 printk(KERN_INFO
"EXT4-fs: barriers enabled\n");
2740 printk(KERN_INFO
"EXT4-fs: barriers disabled\n");
2742 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
2743 err
= jbd2_journal_update_format(journal
);
2745 printk(KERN_ERR
"EXT4-fs: error updating journal.\n");
2746 jbd2_journal_destroy(journal
);
2751 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
2752 err
= jbd2_journal_wipe(journal
, !really_read_only
);
2754 err
= jbd2_journal_load(journal
);
2757 printk(KERN_ERR
"EXT4-fs: error loading journal.\n");
2758 jbd2_journal_destroy(journal
);
2762 EXT4_SB(sb
)->s_journal
= journal
;
2763 ext4_clear_journal_err(sb
, es
);
2765 if (journal_devnum
&&
2766 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
2767 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
2770 /* Make sure we flush the recovery flag to disk. */
2771 ext4_commit_super(sb
, es
, 1);
2777 static int ext4_create_journal(struct super_block
*sb
,
2778 struct ext4_super_block
*es
,
2779 unsigned int journal_inum
)
2784 if (sb
->s_flags
& MS_RDONLY
) {
2785 printk(KERN_ERR
"EXT4-fs: readonly filesystem when trying to "
2786 "create journal.\n");
2790 journal
= ext4_get_journal(sb
, journal_inum
);
2794 printk(KERN_INFO
"EXT4-fs: creating new journal on inode %u\n",
2797 err
= jbd2_journal_create(journal
);
2799 printk(KERN_ERR
"EXT4-fs: error creating journal.\n");
2800 jbd2_journal_destroy(journal
);
2804 EXT4_SB(sb
)->s_journal
= journal
;
2806 ext4_update_dynamic_rev(sb
);
2807 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2808 EXT4_SET_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
);
2810 es
->s_journal_inum
= cpu_to_le32(journal_inum
);
2813 /* Make sure we flush the recovery flag to disk. */
2814 ext4_commit_super(sb
, es
, 1);
2819 static void ext4_commit_super(struct super_block
*sb
,
2820 struct ext4_super_block
*es
, int sync
)
2822 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
2826 if (buffer_write_io_error(sbh
)) {
2828 * Oh, dear. A previous attempt to write the
2829 * superblock failed. This could happen because the
2830 * USB device was yanked out. Or it could happen to
2831 * be a transient write error and maybe the block will
2832 * be remapped. Nothing we can do but to retry the
2833 * write and hope for the best.
2835 printk(KERN_ERR
"ext4: previous I/O error to "
2836 "superblock detected for %s.\n", sb
->s_id
);
2837 clear_buffer_write_io_error(sbh
);
2838 set_buffer_uptodate(sbh
);
2840 es
->s_wtime
= cpu_to_le32(get_seconds());
2841 ext4_free_blocks_count_set(es
, ext4_count_free_blocks(sb
));
2842 es
->s_free_inodes_count
= cpu_to_le32(ext4_count_free_inodes(sb
));
2843 BUFFER_TRACE(sbh
, "marking dirty");
2844 mark_buffer_dirty(sbh
);
2846 sync_dirty_buffer(sbh
);
2847 if (buffer_write_io_error(sbh
)) {
2848 printk(KERN_ERR
"ext4: I/O error while writing "
2849 "superblock for %s.\n", sb
->s_id
);
2850 clear_buffer_write_io_error(sbh
);
2851 set_buffer_uptodate(sbh
);
2858 * Have we just finished recovery? If so, and if we are mounting (or
2859 * remounting) the filesystem readonly, then we will end up with a
2860 * consistent fs on disk. Record that fact.
2862 static void ext4_mark_recovery_complete(struct super_block
*sb
,
2863 struct ext4_super_block
*es
)
2865 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
2867 jbd2_journal_lock_updates(journal
);
2868 if (jbd2_journal_flush(journal
) < 0)
2872 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
2873 sb
->s_flags
& MS_RDONLY
) {
2874 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2876 ext4_commit_super(sb
, es
, 1);
2881 jbd2_journal_unlock_updates(journal
);
2885 * If we are mounting (or read-write remounting) a filesystem whose journal
2886 * has recorded an error from a previous lifetime, move that error to the
2887 * main filesystem now.
2889 static void ext4_clear_journal_err(struct super_block
*sb
,
2890 struct ext4_super_block
*es
)
2896 journal
= EXT4_SB(sb
)->s_journal
;
2899 * Now check for any error status which may have been recorded in the
2900 * journal by a prior ext4_error() or ext4_abort()
2903 j_errno
= jbd2_journal_errno(journal
);
2907 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
2908 ext4_warning(sb
, __func__
, "Filesystem error recorded "
2909 "from previous mount: %s", errstr
);
2910 ext4_warning(sb
, __func__
, "Marking fs in need of "
2911 "filesystem check.");
2913 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
2914 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
2915 ext4_commit_super(sb
, es
, 1);
2917 jbd2_journal_clear_err(journal
);
2922 * Force the running and committing transactions to commit,
2923 * and wait on the commit.
2925 int ext4_force_commit(struct super_block
*sb
)
2930 if (sb
->s_flags
& MS_RDONLY
)
2933 journal
= EXT4_SB(sb
)->s_journal
;
2935 ret
= ext4_journal_force_commit(journal
);
2940 * Ext4 always journals updates to the superblock itself, so we don't
2941 * have to propagate any other updates to the superblock on disk at this
2942 * point. Just start an async writeback to get the buffers on their way
2945 * This implicitly triggers the writebehind on sync().
2948 static void ext4_write_super(struct super_block
*sb
)
2950 if (mutex_trylock(&sb
->s_lock
) != 0)
2955 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
2959 trace_mark(ext4_sync_fs
, "dev %s wait %d", sb
->s_id
, wait
);
2961 if (jbd2_journal_start_commit(EXT4_SB(sb
)->s_journal
, &target
)) {
2963 jbd2_log_wait_commit(EXT4_SB(sb
)->s_journal
, target
);
2969 * LVM calls this function before a (read-only) snapshot is created. This
2970 * gives us a chance to flush the journal completely and mark the fs clean.
2972 static void ext4_write_super_lockfs(struct super_block
*sb
)
2976 if (!(sb
->s_flags
& MS_RDONLY
)) {
2977 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
2979 /* Now we set up the journal barrier. */
2980 jbd2_journal_lock_updates(journal
);
2983 * We don't want to clear needs_recovery flag when we failed
2984 * to flush the journal.
2986 if (jbd2_journal_flush(journal
) < 0)
2989 /* Journal blocked and flushed, clear needs_recovery flag. */
2990 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
2991 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
2996 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2997 * flag here, even though the filesystem is not technically dirty yet.
2999 static void ext4_unlockfs(struct super_block
*sb
)
3001 if (!(sb
->s_flags
& MS_RDONLY
)) {
3003 /* Reser the needs_recovery flag before the fs is unlocked. */
3004 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3005 ext4_commit_super(sb
, EXT4_SB(sb
)->s_es
, 1);
3007 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3011 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3013 struct ext4_super_block
*es
;
3014 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3015 ext4_fsblk_t n_blocks_count
= 0;
3016 unsigned long old_sb_flags
;
3017 struct ext4_mount_options old_opts
;
3024 /* Store the original options */
3025 old_sb_flags
= sb
->s_flags
;
3026 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3027 old_opts
.s_resuid
= sbi
->s_resuid
;
3028 old_opts
.s_resgid
= sbi
->s_resgid
;
3029 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3031 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3032 for (i
= 0; i
< MAXQUOTAS
; i
++)
3033 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3037 * Allow the "check" option to be passed as a remount option.
3039 if (!parse_options(data
, sb
, NULL
, NULL
, &n_blocks_count
, 1)) {
3044 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
)
3045 ext4_abort(sb
, __func__
, "Abort forced by user");
3047 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3048 ((sbi
->s_mount_opt
& EXT4_MOUNT_POSIX_ACL
) ? MS_POSIXACL
: 0);
3052 ext4_init_journal_params(sb
, sbi
->s_journal
);
3054 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3055 n_blocks_count
> ext4_blocks_count(es
)) {
3056 if (sbi
->s_mount_opt
& EXT4_MOUNT_ABORT
) {
3061 if (*flags
& MS_RDONLY
) {
3063 * First of all, the unconditional stuff we have to do
3064 * to disable replay of the journal when we next remount
3066 sb
->s_flags
|= MS_RDONLY
;
3069 * OK, test if we are remounting a valid rw partition
3070 * readonly, and if so set the rdonly flag and then
3071 * mark the partition as valid again.
3073 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3074 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3075 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3078 * We have to unlock super so that we can wait for
3082 ext4_mark_recovery_complete(sb
, es
);
3086 if ((ret
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3087 ~EXT4_FEATURE_RO_COMPAT_SUPP
))) {
3088 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3089 "remount RDWR because of unsupported "
3090 "optional features (%x).\n",
3091 sb
->s_id
, le32_to_cpu(ret
));
3097 * Make sure the group descriptor checksums
3098 * are sane. If they aren't, refuse to
3101 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3102 struct ext4_group_desc
*gdp
=
3103 ext4_get_group_desc(sb
, g
, NULL
);
3105 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3107 "EXT4-fs: ext4_remount: "
3108 "Checksum for group %lu failed (%u!=%u)\n",
3109 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3110 le16_to_cpu(gdp
->bg_checksum
));
3117 * If we have an unprocessed orphan list hanging
3118 * around from a previously readonly bdev mount,
3119 * require a full umount/remount for now.
3121 if (es
->s_last_orphan
) {
3122 printk(KERN_WARNING
"EXT4-fs: %s: couldn't "
3123 "remount RDWR because of unprocessed "
3124 "orphan inode list. Please "
3125 "umount/remount instead.\n",
3132 * Mounting a RDONLY partition read-write, so reread
3133 * and store the current valid flag. (It may have
3134 * been changed by e2fsck since we originally mounted
3137 ext4_clear_journal_err(sb
, es
);
3138 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3139 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3141 if (!ext4_setup_super(sb
, es
, 0))
3142 sb
->s_flags
&= ~MS_RDONLY
;
3146 /* Release old quota file names */
3147 for (i
= 0; i
< MAXQUOTAS
; i
++)
3148 if (old_opts
.s_qf_names
[i
] &&
3149 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3150 kfree(old_opts
.s_qf_names
[i
]);
3154 sb
->s_flags
= old_sb_flags
;
3155 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3156 sbi
->s_resuid
= old_opts
.s_resuid
;
3157 sbi
->s_resgid
= old_opts
.s_resgid
;
3158 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3160 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3161 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3162 if (sbi
->s_qf_names
[i
] &&
3163 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3164 kfree(sbi
->s_qf_names
[i
]);
3165 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3171 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3173 struct super_block
*sb
= dentry
->d_sb
;
3174 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3175 struct ext4_super_block
*es
= sbi
->s_es
;
3178 if (test_opt(sb
, MINIX_DF
)) {
3179 sbi
->s_overhead_last
= 0;
3180 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3181 ext4_group_t ngroups
= sbi
->s_groups_count
, i
;
3182 ext4_fsblk_t overhead
= 0;
3186 * Compute the overhead (FS structures). This is constant
3187 * for a given filesystem unless the number of block groups
3188 * changes so we cache the previous value until it does.
3192 * All of the blocks before first_data_block are
3195 overhead
= le32_to_cpu(es
->s_first_data_block
);
3198 * Add the overhead attributed to the superblock and
3199 * block group descriptors. If the sparse superblocks
3200 * feature is turned on, then not all groups have this.
3202 for (i
= 0; i
< ngroups
; i
++) {
3203 overhead
+= ext4_bg_has_super(sb
, i
) +
3204 ext4_bg_num_gdb(sb
, i
);
3209 * Every block group has an inode bitmap, a block
3210 * bitmap, and an inode table.
3212 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3213 sbi
->s_overhead_last
= overhead
;
3215 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3218 buf
->f_type
= EXT4_SUPER_MAGIC
;
3219 buf
->f_bsize
= sb
->s_blocksize
;
3220 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3221 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3222 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3223 ext4_free_blocks_count_set(es
, buf
->f_bfree
);
3224 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3225 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3227 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3228 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3229 es
->s_free_inodes_count
= cpu_to_le32(buf
->f_ffree
);
3230 buf
->f_namelen
= EXT4_NAME_LEN
;
3231 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3232 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3233 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3234 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3238 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3239 * is locked for write. Otherwise the are possible deadlocks:
3240 * Process 1 Process 2
3241 * ext4_create() quota_sync()
3242 * jbd2_journal_start() write_dquot()
3243 * DQUOT_INIT() down(dqio_mutex)
3244 * down(dqio_mutex) jbd2_journal_start()
3250 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3252 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3255 static int ext4_dquot_initialize(struct inode
*inode
, int type
)
3260 /* We may create quota structure so we need to reserve enough blocks */
3261 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_INIT_BLOCKS(inode
->i_sb
));
3263 return PTR_ERR(handle
);
3264 ret
= dquot_initialize(inode
, type
);
3265 err
= ext4_journal_stop(handle
);
3271 static int ext4_dquot_drop(struct inode
*inode
)
3276 /* We may delete quota structure so we need to reserve enough blocks */
3277 handle
= ext4_journal_start(inode
, 2*EXT4_QUOTA_DEL_BLOCKS(inode
->i_sb
));
3278 if (IS_ERR(handle
)) {
3280 * We call dquot_drop() anyway to at least release references
3281 * to quota structures so that umount does not hang.
3284 return PTR_ERR(handle
);
3286 ret
= dquot_drop(inode
);
3287 err
= ext4_journal_stop(handle
);
3293 static int ext4_write_dquot(struct dquot
*dquot
)
3297 struct inode
*inode
;
3299 inode
= dquot_to_inode(dquot
);
3300 handle
= ext4_journal_start(inode
,
3301 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3303 return PTR_ERR(handle
);
3304 ret
= dquot_commit(dquot
);
3305 err
= ext4_journal_stop(handle
);
3311 static int ext4_acquire_dquot(struct dquot
*dquot
)
3316 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3317 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3319 return PTR_ERR(handle
);
3320 ret
= dquot_acquire(dquot
);
3321 err
= ext4_journal_stop(handle
);
3327 static int ext4_release_dquot(struct dquot
*dquot
)
3332 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3333 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3334 if (IS_ERR(handle
)) {
3335 /* Release dquot anyway to avoid endless cycle in dqput() */
3336 dquot_release(dquot
);
3337 return PTR_ERR(handle
);
3339 ret
= dquot_release(dquot
);
3340 err
= ext4_journal_stop(handle
);
3346 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3348 /* Are we journaling quotas? */
3349 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3350 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3351 dquot_mark_dquot_dirty(dquot
);
3352 return ext4_write_dquot(dquot
);
3354 return dquot_mark_dquot_dirty(dquot
);
3358 static int ext4_write_info(struct super_block
*sb
, int type
)
3363 /* Data block + inode block */
3364 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3366 return PTR_ERR(handle
);
3367 ret
= dquot_commit_info(sb
, type
);
3368 err
= ext4_journal_stop(handle
);
3375 * Turn on quotas during mount time - we need to find
3376 * the quota file and such...
3378 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3380 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3381 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3385 * Standard function to be called on quota_on
3387 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3388 char *path
, int remount
)
3391 struct nameidata nd
;
3393 if (!test_opt(sb
, QUOTA
))
3395 /* When remounting, no checks are needed and in fact, path is NULL */
3397 return vfs_quota_on(sb
, type
, format_id
, path
, remount
);
3399 err
= path_lookup(path
, LOOKUP_FOLLOW
, &nd
);
3403 /* Quotafile not on the same filesystem? */
3404 if (nd
.path
.mnt
->mnt_sb
!= sb
) {
3408 /* Journaling quota? */
3409 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3410 /* Quotafile not in fs root? */
3411 if (nd
.path
.dentry
->d_parent
->d_inode
!= sb
->s_root
->d_inode
)
3413 "EXT4-fs: Quota file not on filesystem root. "
3414 "Journaled quota will not work.\n");
3418 * When we journal data on quota file, we have to flush journal to see
3419 * all updates to the file when we bypass pagecache...
3421 if (ext4_should_journal_data(nd
.path
.dentry
->d_inode
)) {
3423 * We don't need to lock updates but journal_flush() could
3424 * otherwise be livelocked...
3426 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3427 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3428 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3435 err
= vfs_quota_on_path(sb
, type
, format_id
, &nd
.path
);
3440 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3441 * acquiring the locks... As quota files are never truncated and quota code
3442 * itself serializes the operations (and noone else should touch the files)
3443 * we don't have to be afraid of races */
3444 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3445 size_t len
, loff_t off
)
3447 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3448 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3450 int offset
= off
& (sb
->s_blocksize
- 1);
3453 struct buffer_head
*bh
;
3454 loff_t i_size
= i_size_read(inode
);
3458 if (off
+len
> i_size
)
3461 while (toread
> 0) {
3462 tocopy
= sb
->s_blocksize
- offset
< toread
?
3463 sb
->s_blocksize
- offset
: toread
;
3464 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3467 if (!bh
) /* A hole? */
3468 memset(data
, 0, tocopy
);
3470 memcpy(data
, bh
->b_data
+offset
, tocopy
);
3480 /* Write to quotafile (we know the transaction is already started and has
3481 * enough credits) */
3482 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
3483 const char *data
, size_t len
, loff_t off
)
3485 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3486 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3488 int offset
= off
& (sb
->s_blocksize
- 1);
3490 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
3491 size_t towrite
= len
;
3492 struct buffer_head
*bh
;
3493 handle_t
*handle
= journal_current_handle();
3496 printk(KERN_WARNING
"EXT4-fs: Quota write (off=%llu, len=%llu)"
3497 " cancelled because transaction is not started.\n",
3498 (unsigned long long)off
, (unsigned long long)len
);
3501 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
3502 while (towrite
> 0) {
3503 tocopy
= sb
->s_blocksize
- offset
< towrite
?
3504 sb
->s_blocksize
- offset
: towrite
;
3505 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
3508 if (journal_quota
) {
3509 err
= ext4_journal_get_write_access(handle
, bh
);
3516 memcpy(bh
->b_data
+offset
, data
, tocopy
);
3517 flush_dcache_page(bh
->b_page
);
3520 err
= ext4_journal_dirty_metadata(handle
, bh
);
3522 /* Always do at least ordered writes for quotas */
3523 err
= ext4_jbd2_file_inode(handle
, inode
);
3524 mark_buffer_dirty(bh
);
3535 if (len
== towrite
) {
3536 mutex_unlock(&inode
->i_mutex
);
3539 if (inode
->i_size
< off
+len
-towrite
) {
3540 i_size_write(inode
, off
+len
-towrite
);
3541 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
3543 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3544 ext4_mark_inode_dirty(handle
, inode
);
3545 mutex_unlock(&inode
->i_mutex
);
3546 return len
- towrite
;
3551 static int ext4_get_sb(struct file_system_type
*fs_type
,
3552 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3554 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3557 #ifdef CONFIG_PROC_FS
3558 static int ext4_ui_proc_show(struct seq_file
*m
, void *v
)
3560 unsigned int *p
= m
->private;
3562 seq_printf(m
, "%u\n", *p
);
3566 static int ext4_ui_proc_open(struct inode
*inode
, struct file
*file
)
3568 return single_open(file
, ext4_ui_proc_show
, PDE(inode
)->data
);
3571 static ssize_t
ext4_ui_proc_write(struct file
*file
, const char __user
*buf
,
3572 size_t cnt
, loff_t
*ppos
)
3574 unsigned int *p
= PDE(file
->f_path
.dentry
->d_inode
)->data
;
3576 unsigned long value
;
3578 if (cnt
>= sizeof(str
))
3580 if (copy_from_user(str
, buf
, cnt
))
3582 value
= simple_strtol(str
, NULL
, 0);
3589 const struct file_operations ext4_ui_proc_fops
= {
3590 .owner
= THIS_MODULE
,
3591 .open
= ext4_ui_proc_open
,
3593 .llseek
= seq_lseek
,
3594 .release
= single_release
,
3595 .write
= ext4_ui_proc_write
,
3599 static struct file_system_type ext4_fs_type
= {
3600 .owner
= THIS_MODULE
,
3602 .get_sb
= ext4_get_sb
,
3603 .kill_sb
= kill_block_super
,
3604 .fs_flags
= FS_REQUIRES_DEV
,
3607 #ifdef CONFIG_EXT4DEV_COMPAT
3608 static int ext4dev_get_sb(struct file_system_type
*fs_type
,
3609 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
3611 printk(KERN_WARNING
"EXT4-fs: Update your userspace programs "
3612 "to mount using ext4\n");
3613 printk(KERN_WARNING
"EXT4-fs: ext4dev backwards compatibility "
3614 "will go away by 2.6.31\n");
3615 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
, mnt
);
3618 static struct file_system_type ext4dev_fs_type
= {
3619 .owner
= THIS_MODULE
,
3621 .get_sb
= ext4dev_get_sb
,
3622 .kill_sb
= kill_block_super
,
3623 .fs_flags
= FS_REQUIRES_DEV
,
3625 MODULE_ALIAS("ext4dev");
3628 static int __init
init_ext4_fs(void)
3632 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
3633 err
= init_ext4_mballoc();
3637 err
= init_ext4_xattr();
3640 err
= init_inodecache();
3643 err
= register_filesystem(&ext4_fs_type
);
3646 #ifdef CONFIG_EXT4DEV_COMPAT
3647 err
= register_filesystem(&ext4dev_fs_type
);
3649 unregister_filesystem(&ext4_fs_type
);
3655 destroy_inodecache();
3659 exit_ext4_mballoc();
3663 static void __exit
exit_ext4_fs(void)
3665 unregister_filesystem(&ext4_fs_type
);
3666 #ifdef CONFIG_EXT4DEV_COMPAT
3667 unregister_filesystem(&ext4dev_fs_type
);
3669 destroy_inodecache();
3671 exit_ext4_mballoc();
3672 remove_proc_entry("fs/ext4", NULL
);
3675 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3676 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3677 MODULE_LICENSE("GPL");
3678 module_init(init_ext4_fs
)
3679 module_exit(exit_ext4_fs
)