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/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
53 struct proc_dir_entry
*ext4_proc_root
;
54 static struct kset
*ext4_kset
;
56 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
57 unsigned long journal_devnum
);
58 static int ext4_commit_super(struct super_block
*sb
, int sync
);
59 static void ext4_mark_recovery_complete(struct super_block
*sb
,
60 struct ext4_super_block
*es
);
61 static void ext4_clear_journal_err(struct super_block
*sb
,
62 struct ext4_super_block
*es
);
63 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
64 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
66 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
67 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
68 static int ext4_unfreeze(struct super_block
*sb
);
69 static void ext4_write_super(struct super_block
*sb
);
70 static int ext4_freeze(struct super_block
*sb
);
73 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
74 struct ext4_group_desc
*bg
)
76 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
77 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
78 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
81 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
82 struct ext4_group_desc
*bg
)
84 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
85 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
86 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
89 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
90 struct ext4_group_desc
*bg
)
92 return le32_to_cpu(bg
->bg_inode_table_lo
) |
93 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
94 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
97 __u32
ext4_free_blks_count(struct super_block
*sb
,
98 struct ext4_group_desc
*bg
)
100 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
101 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
102 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
105 __u32
ext4_free_inodes_count(struct super_block
*sb
,
106 struct ext4_group_desc
*bg
)
108 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
109 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
110 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
113 __u32
ext4_used_dirs_count(struct super_block
*sb
,
114 struct ext4_group_desc
*bg
)
116 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
117 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
118 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
121 __u32
ext4_itable_unused_count(struct super_block
*sb
,
122 struct ext4_group_desc
*bg
)
124 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
125 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
126 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
129 void ext4_block_bitmap_set(struct super_block
*sb
,
130 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
132 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
133 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
134 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
137 void ext4_inode_bitmap_set(struct super_block
*sb
,
138 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
140 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
141 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
142 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
145 void ext4_inode_table_set(struct super_block
*sb
,
146 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
148 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
149 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
150 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
153 void ext4_free_blks_set(struct super_block
*sb
,
154 struct ext4_group_desc
*bg
, __u32 count
)
156 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
157 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
158 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
161 void ext4_free_inodes_set(struct super_block
*sb
,
162 struct ext4_group_desc
*bg
, __u32 count
)
164 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
165 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
166 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
169 void ext4_used_dirs_set(struct super_block
*sb
,
170 struct ext4_group_desc
*bg
, __u32 count
)
172 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
173 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
174 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
177 void ext4_itable_unused_set(struct super_block
*sb
,
178 struct ext4_group_desc
*bg
, __u32 count
)
180 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
181 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
182 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
186 /* Just increment the non-pointer handle value */
187 static handle_t
*ext4_get_nojournal(void)
189 handle_t
*handle
= current
->journal_info
;
190 unsigned long ref_cnt
= (unsigned long)handle
;
192 BUG_ON(ref_cnt
>= EXT4_NOJOURNAL_MAX_REF_COUNT
);
195 handle
= (handle_t
*)ref_cnt
;
197 current
->journal_info
= handle
;
202 /* Decrement the non-pointer handle value */
203 static void ext4_put_nojournal(handle_t
*handle
)
205 unsigned long ref_cnt
= (unsigned long)handle
;
207 BUG_ON(ref_cnt
== 0);
210 handle
= (handle_t
*)ref_cnt
;
212 current
->journal_info
= handle
;
216 * Wrappers for jbd2_journal_start/end.
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
223 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
227 if (sb
->s_flags
& MS_RDONLY
)
228 return ERR_PTR(-EROFS
);
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
233 journal
= EXT4_SB(sb
)->s_journal
;
235 if (is_journal_aborted(journal
)) {
236 ext4_abort(sb
, __func__
, "Detected aborted journal");
237 return ERR_PTR(-EROFS
);
239 return jbd2_journal_start(journal
, nblocks
);
241 return ext4_get_nojournal();
245 * The only special thing we need to do here is to make sure that all
246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
247 * that sync() will call the filesystem's write_super callback if
250 int __ext4_journal_stop(const char *where
, handle_t
*handle
)
252 struct super_block
*sb
;
256 if (!ext4_handle_valid(handle
)) {
257 ext4_put_nojournal(handle
);
260 sb
= handle
->h_transaction
->t_journal
->j_private
;
262 rc
= jbd2_journal_stop(handle
);
267 __ext4_std_error(sb
, where
, err
);
271 void ext4_journal_abort_handle(const char *caller
, const char *err_fn
,
272 struct buffer_head
*bh
, handle_t
*handle
, int err
)
275 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
277 BUG_ON(!ext4_handle_valid(handle
));
280 BUFFER_TRACE(bh
, "abort");
285 if (is_handle_aborted(handle
))
288 printk(KERN_ERR
"%s: aborting transaction: %s in %s\n",
289 caller
, errstr
, err_fn
);
291 jbd2_journal_abort_handle(handle
);
294 /* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
297 * On ext2, we can store the error state of the filesystem in the
298 * superblock. That is not possible on ext4, because we may have other
299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
304 * We'll just use the jbd2_journal_abort() error code to record an error in
305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
309 static void ext4_handle_error(struct super_block
*sb
)
311 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
313 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
314 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
316 if (sb
->s_flags
& MS_RDONLY
)
319 if (!test_opt(sb
, ERRORS_CONT
)) {
320 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
322 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
324 jbd2_journal_abort(journal
, -EIO
);
326 if (test_opt(sb
, ERRORS_RO
)) {
327 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
328 sb
->s_flags
|= MS_RDONLY
;
330 ext4_commit_super(sb
, 1);
331 if (test_opt(sb
, ERRORS_PANIC
))
332 panic("EXT4-fs (device %s): panic forced after error\n",
336 void __ext4_error(struct super_block
*sb
, const char *function
,
337 const char *fmt
, ...)
342 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
347 ext4_handle_error(sb
);
350 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
357 errstr
= "IO failure";
360 errstr
= "Out of memory";
363 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
364 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
365 errstr
= "Journal has aborted";
367 errstr
= "Readonly filesystem";
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
384 /* __ext4_std_error decodes expected errors from journaling functions
385 * automatically and invokes the appropriate error response. */
387 void __ext4_std_error(struct super_block
*sb
, const char *function
, int errno
)
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
395 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
396 (sb
->s_flags
& MS_RDONLY
))
399 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
400 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s: %s\n",
401 sb
->s_id
, function
, errstr
);
403 ext4_handle_error(sb
);
407 * ext4_abort is a much stronger failure handler than ext4_error. The
408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
416 void ext4_abort(struct super_block
*sb
, const char *function
,
417 const char *fmt
, ...)
422 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
427 if (test_opt(sb
, ERRORS_PANIC
))
428 panic("EXT4-fs panic from previous error\n");
430 if (sb
->s_flags
& MS_RDONLY
)
433 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
434 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
435 sb
->s_flags
|= MS_RDONLY
;
436 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
437 if (EXT4_SB(sb
)->s_journal
)
438 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
441 void ext4_msg (struct super_block
* sb
, const char *prefix
,
442 const char *fmt
, ...)
447 printk("%sEXT4-fs (%s): ", prefix
, sb
->s_id
);
453 void __ext4_warning(struct super_block
*sb
, const char *function
,
454 const char *fmt
, ...)
459 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s: ",
466 void ext4_grp_locked_error(struct super_block
*sb
, ext4_group_t grp
,
467 const char *function
, const char *fmt
, ...)
472 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
475 printk(KERN_CRIT
"EXT4-fs error (device %s): %s: ", sb
->s_id
, function
);
480 if (test_opt(sb
, ERRORS_CONT
)) {
481 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
482 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
483 ext4_commit_super(sb
, 0);
486 ext4_unlock_group(sb
, grp
);
487 ext4_handle_error(sb
);
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
499 ext4_lock_group(sb
, grp
);
503 void ext4_update_dynamic_rev(struct super_block
*sb
)
505 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
507 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
515 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
516 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
517 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
529 * Open the external journal device
531 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
533 struct block_device
*bdev
;
534 char b
[BDEVNAME_SIZE
];
536 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
542 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
543 __bdevname(dev
, b
), PTR_ERR(bdev
));
548 * Release the journal device
550 static int ext4_blkdev_put(struct block_device
*bdev
)
553 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
556 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
558 struct block_device
*bdev
;
561 bdev
= sbi
->journal_bdev
;
563 ret
= ext4_blkdev_put(bdev
);
564 sbi
->journal_bdev
= NULL
;
569 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
571 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
574 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
578 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
579 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
581 printk(KERN_ERR
"sb_info orphan list:\n");
582 list_for_each(l
, &sbi
->s_orphan
) {
583 struct inode
*inode
= orphan_list_entry(l
);
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
587 inode
->i_mode
, inode
->i_nlink
,
592 static void ext4_put_super(struct super_block
*sb
)
594 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
595 struct ext4_super_block
*es
= sbi
->s_es
;
598 flush_workqueue(sbi
->dio_unwritten_wq
);
599 destroy_workqueue(sbi
->dio_unwritten_wq
);
604 ext4_commit_super(sb
, 1);
606 if (sbi
->s_journal
) {
607 err
= jbd2_journal_destroy(sbi
->s_journal
);
608 sbi
->s_journal
= NULL
;
610 ext4_abort(sb
, __func__
,
611 "Couldn't clean up the journal");
614 ext4_release_system_zone(sb
);
616 ext4_ext_release(sb
);
617 ext4_xattr_put_super(sb
);
619 if (!(sb
->s_flags
& MS_RDONLY
)) {
620 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
621 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
622 ext4_commit_super(sb
, 1);
625 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
627 kobject_del(&sbi
->s_kobj
);
629 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
630 brelse(sbi
->s_group_desc
[i
]);
631 kfree(sbi
->s_group_desc
);
632 if (is_vmalloc_addr(sbi
->s_flex_groups
))
633 vfree(sbi
->s_flex_groups
);
635 kfree(sbi
->s_flex_groups
);
636 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
637 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
638 percpu_counter_destroy(&sbi
->s_dirs_counter
);
639 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
642 for (i
= 0; i
< MAXQUOTAS
; i
++)
643 kfree(sbi
->s_qf_names
[i
]);
646 /* Debugging code just in case the in-memory inode orphan list
647 * isn't empty. The on-disk one can be non-empty if we've
648 * detected an error and taken the fs readonly, but the
649 * in-memory list had better be clean by this point. */
650 if (!list_empty(&sbi
->s_orphan
))
651 dump_orphan_list(sb
, sbi
);
652 J_ASSERT(list_empty(&sbi
->s_orphan
));
654 invalidate_bdev(sb
->s_bdev
);
655 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
657 * Invalidate the journal device's buffers. We don't want them
658 * floating about in memory - the physical journal device may
659 * hotswapped, and it breaks the `ro-after' testing code.
661 sync_blockdev(sbi
->journal_bdev
);
662 invalidate_bdev(sbi
->journal_bdev
);
663 ext4_blkdev_remove(sbi
);
665 sb
->s_fs_info
= NULL
;
667 * Now that we are completely done shutting down the
668 * superblock, we need to actually destroy the kobject.
672 kobject_put(&sbi
->s_kobj
);
673 wait_for_completion(&sbi
->s_kobj_unregister
);
674 kfree(sbi
->s_blockgroup_lock
);
678 static struct kmem_cache
*ext4_inode_cachep
;
681 * Called inside transaction, so use GFP_NOFS
683 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
685 struct ext4_inode_info
*ei
;
687 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
691 ei
->vfs_inode
.i_version
= 1;
692 ei
->vfs_inode
.i_data
.writeback_index
= 0;
693 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
694 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
695 spin_lock_init(&ei
->i_prealloc_lock
);
697 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
698 * therefore it can be null here. Don't check it, just initialize
701 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
702 ei
->i_reserved_data_blocks
= 0;
703 ei
->i_reserved_meta_blocks
= 0;
704 ei
->i_allocated_meta_blocks
= 0;
705 ei
->i_da_metadata_calc_len
= 0;
706 ei
->i_delalloc_reserved_flag
= 0;
707 spin_lock_init(&(ei
->i_block_reservation_lock
));
709 ei
->i_reserved_quota
= 0;
711 INIT_LIST_HEAD(&ei
->i_aio_dio_complete_list
);
712 ei
->cur_aio_dio
= NULL
;
714 ei
->i_datasync_tid
= 0;
716 return &ei
->vfs_inode
;
719 static void ext4_destroy_inode(struct inode
*inode
)
721 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
722 ext4_msg(inode
->i_sb
, KERN_ERR
,
723 "Inode %lu (%p): orphan list check failed!",
724 inode
->i_ino
, EXT4_I(inode
));
725 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
726 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
730 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
733 static void init_once(void *foo
)
735 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
737 INIT_LIST_HEAD(&ei
->i_orphan
);
738 #ifdef CONFIG_EXT4_FS_XATTR
739 init_rwsem(&ei
->xattr_sem
);
741 init_rwsem(&ei
->i_data_sem
);
742 inode_init_once(&ei
->vfs_inode
);
745 static int init_inodecache(void)
747 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
748 sizeof(struct ext4_inode_info
),
749 0, (SLAB_RECLAIM_ACCOUNT
|
752 if (ext4_inode_cachep
== NULL
)
757 static void destroy_inodecache(void)
759 kmem_cache_destroy(ext4_inode_cachep
);
762 static void ext4_clear_inode(struct inode
*inode
)
764 ext4_discard_preallocations(inode
);
765 if (EXT4_JOURNAL(inode
))
766 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
767 &EXT4_I(inode
)->jinode
);
770 static inline void ext4_show_quota_options(struct seq_file
*seq
,
771 struct super_block
*sb
)
773 #if defined(CONFIG_QUOTA)
774 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
776 if (sbi
->s_jquota_fmt
) {
779 switch (sbi
->s_jquota_fmt
) {
790 seq_printf(seq
, ",jqfmt=%s", fmtname
);
793 if (sbi
->s_qf_names
[USRQUOTA
])
794 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
796 if (sbi
->s_qf_names
[GRPQUOTA
])
797 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
799 if (test_opt(sb
, USRQUOTA
))
800 seq_puts(seq
, ",usrquota");
802 if (test_opt(sb
, GRPQUOTA
))
803 seq_puts(seq
, ",grpquota");
809 * - it's set to a non-default value OR
810 * - if the per-sb default is different from the global default
812 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
815 unsigned long def_mount_opts
;
816 struct super_block
*sb
= vfs
->mnt_sb
;
817 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
818 struct ext4_super_block
*es
= sbi
->s_es
;
820 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
821 def_errors
= le16_to_cpu(es
->s_errors
);
823 if (sbi
->s_sb_block
!= 1)
824 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
825 if (test_opt(sb
, MINIX_DF
))
826 seq_puts(seq
, ",minixdf");
827 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
828 seq_puts(seq
, ",grpid");
829 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
830 seq_puts(seq
, ",nogrpid");
831 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
832 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
833 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
835 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
836 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
837 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
839 if (test_opt(sb
, ERRORS_RO
)) {
840 if (def_errors
== EXT4_ERRORS_PANIC
||
841 def_errors
== EXT4_ERRORS_CONTINUE
) {
842 seq_puts(seq
, ",errors=remount-ro");
845 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
846 seq_puts(seq
, ",errors=continue");
847 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
848 seq_puts(seq
, ",errors=panic");
849 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
850 seq_puts(seq
, ",nouid32");
851 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
852 seq_puts(seq
, ",debug");
853 if (test_opt(sb
, OLDALLOC
))
854 seq_puts(seq
, ",oldalloc");
855 #ifdef CONFIG_EXT4_FS_XATTR
856 if (test_opt(sb
, XATTR_USER
) &&
857 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
858 seq_puts(seq
, ",user_xattr");
859 if (!test_opt(sb
, XATTR_USER
) &&
860 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
861 seq_puts(seq
, ",nouser_xattr");
864 #ifdef CONFIG_EXT4_FS_POSIX_ACL
865 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
866 seq_puts(seq
, ",acl");
867 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
868 seq_puts(seq
, ",noacl");
870 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
871 seq_printf(seq
, ",commit=%u",
872 (unsigned) (sbi
->s_commit_interval
/ HZ
));
874 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
875 seq_printf(seq
, ",min_batch_time=%u",
876 (unsigned) sbi
->s_min_batch_time
);
878 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
879 seq_printf(seq
, ",max_batch_time=%u",
880 (unsigned) sbi
->s_min_batch_time
);
884 * We're changing the default of barrier mount option, so
885 * let's always display its mount state so it's clear what its
888 seq_puts(seq
, ",barrier=");
889 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
890 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
891 seq_puts(seq
, ",journal_async_commit");
892 if (test_opt(sb
, NOBH
))
893 seq_puts(seq
, ",nobh");
894 if (test_opt(sb
, I_VERSION
))
895 seq_puts(seq
, ",i_version");
896 if (!test_opt(sb
, DELALLOC
))
897 seq_puts(seq
, ",nodelalloc");
901 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
903 * journal mode get enabled in different ways
904 * So just print the value even if we didn't specify it
906 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
907 seq_puts(seq
, ",data=journal");
908 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
909 seq_puts(seq
, ",data=ordered");
910 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
911 seq_puts(seq
, ",data=writeback");
913 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
914 seq_printf(seq
, ",inode_readahead_blks=%u",
915 sbi
->s_inode_readahead_blks
);
917 if (test_opt(sb
, DATA_ERR_ABORT
))
918 seq_puts(seq
, ",data_err=abort");
920 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
921 seq_puts(seq
, ",noauto_da_alloc");
923 if (test_opt(sb
, DISCARD
))
924 seq_puts(seq
, ",discard");
926 if (test_opt(sb
, NOLOAD
))
927 seq_puts(seq
, ",norecovery");
929 ext4_show_quota_options(seq
, sb
);
934 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
935 u64 ino
, u32 generation
)
939 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
940 return ERR_PTR(-ESTALE
);
941 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
942 return ERR_PTR(-ESTALE
);
944 /* iget isn't really right if the inode is currently unallocated!!
946 * ext4_read_inode will return a bad_inode if the inode had been
947 * deleted, so we should be safe.
949 * Currently we don't know the generation for parent directory, so
950 * a generation of 0 means "accept any"
952 inode
= ext4_iget(sb
, ino
);
954 return ERR_CAST(inode
);
955 if (generation
&& inode
->i_generation
!= generation
) {
957 return ERR_PTR(-ESTALE
);
963 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
964 int fh_len
, int fh_type
)
966 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
970 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
971 int fh_len
, int fh_type
)
973 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
978 * Try to release metadata pages (indirect blocks, directories) which are
979 * mapped via the block device. Since these pages could have journal heads
980 * which would prevent try_to_free_buffers() from freeing them, we must use
981 * jbd2 layer's try_to_free_buffers() function to release them.
983 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
986 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
988 WARN_ON(PageChecked(page
));
989 if (!page_has_buffers(page
))
992 return jbd2_journal_try_to_free_buffers(journal
, page
,
994 return try_to_free_buffers(page
);
998 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
999 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1001 static int ext4_write_dquot(struct dquot
*dquot
);
1002 static int ext4_acquire_dquot(struct dquot
*dquot
);
1003 static int ext4_release_dquot(struct dquot
*dquot
);
1004 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1005 static int ext4_write_info(struct super_block
*sb
, int type
);
1006 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1007 char *path
, int remount
);
1008 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1009 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1010 size_t len
, loff_t off
);
1011 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1012 const char *data
, size_t len
, loff_t off
);
1014 static const struct dquot_operations ext4_quota_operations
= {
1015 .initialize
= dquot_initialize
,
1017 .alloc_space
= dquot_alloc_space
,
1018 .reserve_space
= dquot_reserve_space
,
1019 .claim_space
= dquot_claim_space
,
1020 .release_rsv
= dquot_release_reserved_space
,
1022 .get_reserved_space
= ext4_get_reserved_space
,
1024 .alloc_inode
= dquot_alloc_inode
,
1025 .free_space
= dquot_free_space
,
1026 .free_inode
= dquot_free_inode
,
1027 .transfer
= dquot_transfer
,
1028 .write_dquot
= ext4_write_dquot
,
1029 .acquire_dquot
= ext4_acquire_dquot
,
1030 .release_dquot
= ext4_release_dquot
,
1031 .mark_dirty
= ext4_mark_dquot_dirty
,
1032 .write_info
= ext4_write_info
,
1033 .alloc_dquot
= dquot_alloc
,
1034 .destroy_dquot
= dquot_destroy
,
1037 static const struct quotactl_ops ext4_qctl_operations
= {
1038 .quota_on
= ext4_quota_on
,
1039 .quota_off
= vfs_quota_off
,
1040 .quota_sync
= vfs_quota_sync
,
1041 .get_info
= vfs_get_dqinfo
,
1042 .set_info
= vfs_set_dqinfo
,
1043 .get_dqblk
= vfs_get_dqblk
,
1044 .set_dqblk
= vfs_set_dqblk
1048 static const struct super_operations ext4_sops
= {
1049 .alloc_inode
= ext4_alloc_inode
,
1050 .destroy_inode
= ext4_destroy_inode
,
1051 .write_inode
= ext4_write_inode
,
1052 .dirty_inode
= ext4_dirty_inode
,
1053 .delete_inode
= ext4_delete_inode
,
1054 .put_super
= ext4_put_super
,
1055 .sync_fs
= ext4_sync_fs
,
1056 .freeze_fs
= ext4_freeze
,
1057 .unfreeze_fs
= ext4_unfreeze
,
1058 .statfs
= ext4_statfs
,
1059 .remount_fs
= ext4_remount
,
1060 .clear_inode
= ext4_clear_inode
,
1061 .show_options
= ext4_show_options
,
1063 .quota_read
= ext4_quota_read
,
1064 .quota_write
= ext4_quota_write
,
1066 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1069 static const struct super_operations ext4_nojournal_sops
= {
1070 .alloc_inode
= ext4_alloc_inode
,
1071 .destroy_inode
= ext4_destroy_inode
,
1072 .write_inode
= ext4_write_inode
,
1073 .dirty_inode
= ext4_dirty_inode
,
1074 .delete_inode
= ext4_delete_inode
,
1075 .write_super
= ext4_write_super
,
1076 .put_super
= ext4_put_super
,
1077 .statfs
= ext4_statfs
,
1078 .remount_fs
= ext4_remount
,
1079 .clear_inode
= ext4_clear_inode
,
1080 .show_options
= ext4_show_options
,
1082 .quota_read
= ext4_quota_read
,
1083 .quota_write
= ext4_quota_write
,
1085 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1088 static const struct export_operations ext4_export_ops
= {
1089 .fh_to_dentry
= ext4_fh_to_dentry
,
1090 .fh_to_parent
= ext4_fh_to_parent
,
1091 .get_parent
= ext4_get_parent
,
1095 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1096 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1097 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1098 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1099 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1100 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1101 Opt_journal_update
, Opt_journal_dev
,
1102 Opt_journal_checksum
, Opt_journal_async_commit
,
1103 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1104 Opt_data_err_abort
, Opt_data_err_ignore
,
1105 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1106 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1107 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1108 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1109 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
1110 Opt_block_validity
, Opt_noblock_validity
,
1111 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1112 Opt_discard
, Opt_nodiscard
,
1115 static const match_table_t tokens
= {
1116 {Opt_bsd_df
, "bsddf"},
1117 {Opt_minix_df
, "minixdf"},
1118 {Opt_grpid
, "grpid"},
1119 {Opt_grpid
, "bsdgroups"},
1120 {Opt_nogrpid
, "nogrpid"},
1121 {Opt_nogrpid
, "sysvgroups"},
1122 {Opt_resgid
, "resgid=%u"},
1123 {Opt_resuid
, "resuid=%u"},
1125 {Opt_err_cont
, "errors=continue"},
1126 {Opt_err_panic
, "errors=panic"},
1127 {Opt_err_ro
, "errors=remount-ro"},
1128 {Opt_nouid32
, "nouid32"},
1129 {Opt_debug
, "debug"},
1130 {Opt_oldalloc
, "oldalloc"},
1131 {Opt_orlov
, "orlov"},
1132 {Opt_user_xattr
, "user_xattr"},
1133 {Opt_nouser_xattr
, "nouser_xattr"},
1135 {Opt_noacl
, "noacl"},
1136 {Opt_noload
, "noload"},
1137 {Opt_noload
, "norecovery"},
1140 {Opt_commit
, "commit=%u"},
1141 {Opt_min_batch_time
, "min_batch_time=%u"},
1142 {Opt_max_batch_time
, "max_batch_time=%u"},
1143 {Opt_journal_update
, "journal=update"},
1144 {Opt_journal_dev
, "journal_dev=%u"},
1145 {Opt_journal_checksum
, "journal_checksum"},
1146 {Opt_journal_async_commit
, "journal_async_commit"},
1147 {Opt_abort
, "abort"},
1148 {Opt_data_journal
, "data=journal"},
1149 {Opt_data_ordered
, "data=ordered"},
1150 {Opt_data_writeback
, "data=writeback"},
1151 {Opt_data_err_abort
, "data_err=abort"},
1152 {Opt_data_err_ignore
, "data_err=ignore"},
1153 {Opt_offusrjquota
, "usrjquota="},
1154 {Opt_usrjquota
, "usrjquota=%s"},
1155 {Opt_offgrpjquota
, "grpjquota="},
1156 {Opt_grpjquota
, "grpjquota=%s"},
1157 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1158 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1159 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1160 {Opt_grpquota
, "grpquota"},
1161 {Opt_noquota
, "noquota"},
1162 {Opt_quota
, "quota"},
1163 {Opt_usrquota
, "usrquota"},
1164 {Opt_barrier
, "barrier=%u"},
1165 {Opt_barrier
, "barrier"},
1166 {Opt_nobarrier
, "nobarrier"},
1167 {Opt_i_version
, "i_version"},
1168 {Opt_stripe
, "stripe=%u"},
1169 {Opt_resize
, "resize"},
1170 {Opt_delalloc
, "delalloc"},
1171 {Opt_nodelalloc
, "nodelalloc"},
1172 {Opt_block_validity
, "block_validity"},
1173 {Opt_noblock_validity
, "noblock_validity"},
1174 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1175 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1176 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1177 {Opt_auto_da_alloc
, "auto_da_alloc"},
1178 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1179 {Opt_discard
, "discard"},
1180 {Opt_nodiscard
, "nodiscard"},
1184 static ext4_fsblk_t
get_sb_block(void **data
)
1186 ext4_fsblk_t sb_block
;
1187 char *options
= (char *) *data
;
1189 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1190 return 1; /* Default location */
1193 /* TODO: use simple_strtoll with >32bit ext4 */
1194 sb_block
= simple_strtoul(options
, &options
, 0);
1195 if (*options
&& *options
!= ',') {
1196 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1200 if (*options
== ',')
1202 *data
= (void *) options
;
1207 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1208 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1209 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1212 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1214 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1217 if (sb_any_quota_loaded(sb
) &&
1218 !sbi
->s_qf_names
[qtype
]) {
1219 ext4_msg(sb
, KERN_ERR
,
1220 "Cannot change journaled "
1221 "quota options when quota turned on");
1224 qname
= match_strdup(args
);
1226 ext4_msg(sb
, KERN_ERR
,
1227 "Not enough memory for storing quotafile name");
1230 if (sbi
->s_qf_names
[qtype
] &&
1231 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1232 ext4_msg(sb
, KERN_ERR
,
1233 "%s quota file already specified", QTYPE2NAME(qtype
));
1237 sbi
->s_qf_names
[qtype
] = qname
;
1238 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1239 ext4_msg(sb
, KERN_ERR
,
1240 "quotafile must be on filesystem root");
1241 kfree(sbi
->s_qf_names
[qtype
]);
1242 sbi
->s_qf_names
[qtype
] = NULL
;
1245 set_opt(sbi
->s_mount_opt
, QUOTA
);
1249 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1252 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1254 if (sb_any_quota_loaded(sb
) &&
1255 sbi
->s_qf_names
[qtype
]) {
1256 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1257 " when quota turned on");
1261 * The space will be released later when all options are confirmed
1264 sbi
->s_qf_names
[qtype
] = NULL
;
1269 static int parse_options(char *options
, struct super_block
*sb
,
1270 unsigned long *journal_devnum
,
1271 unsigned int *journal_ioprio
,
1272 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1274 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1276 substring_t args
[MAX_OPT_ARGS
];
1286 while ((p
= strsep(&options
, ",")) != NULL
) {
1292 * Initialize args struct so we know whether arg was
1293 * found; some options take optional arguments.
1295 args
[0].to
= args
[0].from
= 0;
1296 token
= match_token(p
, tokens
, args
);
1299 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1300 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1303 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1304 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1308 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1309 set_opt(sbi
->s_mount_opt
, GRPID
);
1313 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1314 clear_opt(sbi
->s_mount_opt
, GRPID
);
1318 if (match_int(&args
[0], &option
))
1320 sbi
->s_resuid
= option
;
1323 if (match_int(&args
[0], &option
))
1325 sbi
->s_resgid
= option
;
1328 /* handled by get_sb_block() instead of here */
1329 /* *sb_block = match_int(&args[0]); */
1332 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1333 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1334 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1337 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1338 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1339 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1342 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1343 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1344 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1347 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1350 set_opt(sbi
->s_mount_opt
, DEBUG
);
1353 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1356 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1358 #ifdef CONFIG_EXT4_FS_XATTR
1359 case Opt_user_xattr
:
1360 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1362 case Opt_nouser_xattr
:
1363 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1366 case Opt_user_xattr
:
1367 case Opt_nouser_xattr
:
1368 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1371 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1373 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1376 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1381 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1384 case Opt_journal_update
:
1386 /* Eventually we will want to be able to create
1387 a journal file here. For now, only allow the
1388 user to specify an existing inode to be the
1391 ext4_msg(sb
, KERN_ERR
,
1392 "Cannot specify journal on remount");
1395 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1397 case Opt_journal_dev
:
1399 ext4_msg(sb
, KERN_ERR
,
1400 "Cannot specify journal on remount");
1403 if (match_int(&args
[0], &option
))
1405 *journal_devnum
= option
;
1407 case Opt_journal_checksum
:
1408 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1410 case Opt_journal_async_commit
:
1411 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1412 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1415 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1418 if (match_int(&args
[0], &option
))
1423 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1424 sbi
->s_commit_interval
= HZ
* option
;
1426 case Opt_max_batch_time
:
1427 if (match_int(&args
[0], &option
))
1432 option
= EXT4_DEF_MAX_BATCH_TIME
;
1433 sbi
->s_max_batch_time
= option
;
1435 case Opt_min_batch_time
:
1436 if (match_int(&args
[0], &option
))
1440 sbi
->s_min_batch_time
= option
;
1442 case Opt_data_journal
:
1443 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1445 case Opt_data_ordered
:
1446 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1448 case Opt_data_writeback
:
1449 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1452 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1453 ext4_msg(sb
, KERN_ERR
,
1454 "Cannot change data mode on remount");
1458 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
1459 sbi
->s_mount_opt
|= data_opt
;
1462 case Opt_data_err_abort
:
1463 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1465 case Opt_data_err_ignore
:
1466 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1470 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1474 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1477 case Opt_offusrjquota
:
1478 if (!clear_qf_name(sb
, USRQUOTA
))
1481 case Opt_offgrpjquota
:
1482 if (!clear_qf_name(sb
, GRPQUOTA
))
1486 case Opt_jqfmt_vfsold
:
1487 qfmt
= QFMT_VFS_OLD
;
1489 case Opt_jqfmt_vfsv0
:
1492 case Opt_jqfmt_vfsv1
:
1495 if (sb_any_quota_loaded(sb
) &&
1496 sbi
->s_jquota_fmt
!= qfmt
) {
1497 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1498 "journaled quota options when "
1502 sbi
->s_jquota_fmt
= qfmt
;
1506 set_opt(sbi
->s_mount_opt
, QUOTA
);
1507 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1510 set_opt(sbi
->s_mount_opt
, QUOTA
);
1511 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1514 if (sb_any_quota_loaded(sb
)) {
1515 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1516 "options when quota turned on");
1519 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1520 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1521 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1527 ext4_msg(sb
, KERN_ERR
,
1528 "quota options not supported");
1532 case Opt_offusrjquota
:
1533 case Opt_offgrpjquota
:
1534 case Opt_jqfmt_vfsold
:
1535 case Opt_jqfmt_vfsv0
:
1536 case Opt_jqfmt_vfsv1
:
1537 ext4_msg(sb
, KERN_ERR
,
1538 "journaled quota options not supported");
1544 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1547 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1551 if (match_int(&args
[0], &option
))
1554 option
= 1; /* No argument, default to 1 */
1556 set_opt(sbi
->s_mount_opt
, BARRIER
);
1558 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1564 ext4_msg(sb
, KERN_ERR
,
1565 "resize option only available "
1569 if (match_int(&args
[0], &option
) != 0)
1571 *n_blocks_count
= option
;
1574 set_opt(sbi
->s_mount_opt
, NOBH
);
1577 clear_opt(sbi
->s_mount_opt
, NOBH
);
1580 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1581 sb
->s_flags
|= MS_I_VERSION
;
1583 case Opt_nodelalloc
:
1584 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1587 if (match_int(&args
[0], &option
))
1591 sbi
->s_stripe
= option
;
1594 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1596 case Opt_block_validity
:
1597 set_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1599 case Opt_noblock_validity
:
1600 clear_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1602 case Opt_inode_readahead_blks
:
1603 if (match_int(&args
[0], &option
))
1605 if (option
< 0 || option
> (1 << 30))
1607 if (!is_power_of_2(option
)) {
1608 ext4_msg(sb
, KERN_ERR
,
1609 "EXT4-fs: inode_readahead_blks"
1610 " must be a power of 2");
1613 sbi
->s_inode_readahead_blks
= option
;
1615 case Opt_journal_ioprio
:
1616 if (match_int(&args
[0], &option
))
1618 if (option
< 0 || option
> 7)
1620 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1623 case Opt_noauto_da_alloc
:
1624 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1626 case Opt_auto_da_alloc
:
1628 if (match_int(&args
[0], &option
))
1631 option
= 1; /* No argument, default to 1 */
1633 clear_opt(sbi
->s_mount_opt
, NO_AUTO_DA_ALLOC
);
1635 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1638 set_opt(sbi
->s_mount_opt
, DISCARD
);
1641 clear_opt(sbi
->s_mount_opt
, DISCARD
);
1644 ext4_msg(sb
, KERN_ERR
,
1645 "Unrecognized mount option \"%s\" "
1646 "or missing value", p
);
1651 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1652 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1653 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1655 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1656 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1658 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1659 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1664 if (!sbi
->s_jquota_fmt
) {
1665 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1670 if (sbi
->s_jquota_fmt
) {
1671 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1672 "specified with no journaling "
1681 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1684 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1687 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1688 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1689 "forcing read-only mode");
1694 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1695 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1696 "running e2fsck is recommended");
1697 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1698 ext4_msg(sb
, KERN_WARNING
,
1699 "warning: mounting fs with errors, "
1700 "running e2fsck is recommended");
1701 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1702 le16_to_cpu(es
->s_mnt_count
) >=
1703 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1704 ext4_msg(sb
, KERN_WARNING
,
1705 "warning: maximal mount count reached, "
1706 "running e2fsck is recommended");
1707 else if (le32_to_cpu(es
->s_checkinterval
) &&
1708 (le32_to_cpu(es
->s_lastcheck
) +
1709 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1710 ext4_msg(sb
, KERN_WARNING
,
1711 "warning: checktime reached, "
1712 "running e2fsck is recommended");
1713 if (!sbi
->s_journal
)
1714 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1715 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1716 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1717 le16_add_cpu(&es
->s_mnt_count
, 1);
1718 es
->s_mtime
= cpu_to_le32(get_seconds());
1719 ext4_update_dynamic_rev(sb
);
1721 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1723 ext4_commit_super(sb
, 1);
1724 if (test_opt(sb
, DEBUG
))
1725 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1726 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1728 sbi
->s_groups_count
,
1729 EXT4_BLOCKS_PER_GROUP(sb
),
1730 EXT4_INODES_PER_GROUP(sb
),
1736 static int ext4_fill_flex_info(struct super_block
*sb
)
1738 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1739 struct ext4_group_desc
*gdp
= NULL
;
1740 ext4_group_t flex_group_count
;
1741 ext4_group_t flex_group
;
1742 int groups_per_flex
= 0;
1746 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1747 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1749 if (groups_per_flex
< 2) {
1750 sbi
->s_log_groups_per_flex
= 0;
1754 /* We allocate both existing and potentially added groups */
1755 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1756 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1757 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1758 size
= flex_group_count
* sizeof(struct flex_groups
);
1759 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1760 if (sbi
->s_flex_groups
== NULL
) {
1761 sbi
->s_flex_groups
= vmalloc(size
);
1762 if (sbi
->s_flex_groups
)
1763 memset(sbi
->s_flex_groups
, 0, size
);
1765 if (sbi
->s_flex_groups
== NULL
) {
1766 ext4_msg(sb
, KERN_ERR
, "not enough memory for "
1767 "%u flex groups", flex_group_count
);
1771 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1772 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1774 flex_group
= ext4_flex_group(sbi
, i
);
1775 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1776 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1777 atomic_add(ext4_free_blks_count(sb
, gdp
),
1778 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1779 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1780 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1788 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1789 struct ext4_group_desc
*gdp
)
1793 if (sbi
->s_es
->s_feature_ro_compat
&
1794 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1795 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1796 __le32 le_group
= cpu_to_le32(block_group
);
1798 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1799 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1800 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1801 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1802 /* for checksum of struct ext4_group_desc do the rest...*/
1803 if ((sbi
->s_es
->s_feature_incompat
&
1804 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1805 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1806 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1807 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1811 return cpu_to_le16(crc
);
1814 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1815 struct ext4_group_desc
*gdp
)
1817 if ((sbi
->s_es
->s_feature_ro_compat
&
1818 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1819 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1825 /* Called at mount-time, super-block is locked */
1826 static int ext4_check_descriptors(struct super_block
*sb
)
1828 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1829 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1830 ext4_fsblk_t last_block
;
1831 ext4_fsblk_t block_bitmap
;
1832 ext4_fsblk_t inode_bitmap
;
1833 ext4_fsblk_t inode_table
;
1834 int flexbg_flag
= 0;
1837 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1840 ext4_debug("Checking group descriptors");
1842 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1843 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1845 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1846 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1848 last_block
= first_block
+
1849 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1851 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1852 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1853 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1854 "Block bitmap for group %u not in group "
1855 "(block %llu)!", i
, block_bitmap
);
1858 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1859 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1860 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1861 "Inode bitmap for group %u not in group "
1862 "(block %llu)!", i
, inode_bitmap
);
1865 inode_table
= ext4_inode_table(sb
, gdp
);
1866 if (inode_table
< first_block
||
1867 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1868 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1869 "Inode table for group %u not in group "
1870 "(block %llu)!", i
, inode_table
);
1873 ext4_lock_group(sb
, i
);
1874 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1875 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1876 "Checksum for group %u failed (%u!=%u)",
1877 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1878 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1879 if (!(sb
->s_flags
& MS_RDONLY
)) {
1880 ext4_unlock_group(sb
, i
);
1884 ext4_unlock_group(sb
, i
);
1886 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
1889 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
1890 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
1894 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1895 * the superblock) which were deleted from all directories, but held open by
1896 * a process at the time of a crash. We walk the list and try to delete these
1897 * inodes at recovery time (only with a read-write filesystem).
1899 * In order to keep the orphan inode chain consistent during traversal (in
1900 * case of crash during recovery), we link each inode into the superblock
1901 * orphan list_head and handle it the same way as an inode deletion during
1902 * normal operation (which journals the operations for us).
1904 * We only do an iget() and an iput() on each inode, which is very safe if we
1905 * accidentally point at an in-use or already deleted inode. The worst that
1906 * can happen in this case is that we get a "bit already cleared" message from
1907 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1908 * e2fsck was run on this filesystem, and it must have already done the orphan
1909 * inode cleanup for us, so we can safely abort without any further action.
1911 static void ext4_orphan_cleanup(struct super_block
*sb
,
1912 struct ext4_super_block
*es
)
1914 unsigned int s_flags
= sb
->s_flags
;
1915 int nr_orphans
= 0, nr_truncates
= 0;
1919 if (!es
->s_last_orphan
) {
1920 jbd_debug(4, "no orphan inodes to clean up\n");
1924 if (bdev_read_only(sb
->s_bdev
)) {
1925 ext4_msg(sb
, KERN_ERR
, "write access "
1926 "unavailable, skipping orphan cleanup");
1930 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
1931 if (es
->s_last_orphan
)
1932 jbd_debug(1, "Errors on filesystem, "
1933 "clearing orphan list.\n");
1934 es
->s_last_orphan
= 0;
1935 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1939 if (s_flags
& MS_RDONLY
) {
1940 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
1941 sb
->s_flags
&= ~MS_RDONLY
;
1944 /* Needed for iput() to work correctly and not trash data */
1945 sb
->s_flags
|= MS_ACTIVE
;
1946 /* Turn on quotas so that they are updated correctly */
1947 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1948 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
1949 int ret
= ext4_quota_on_mount(sb
, i
);
1951 ext4_msg(sb
, KERN_ERR
,
1952 "Cannot turn on journaled "
1953 "quota: error %d", ret
);
1958 while (es
->s_last_orphan
) {
1959 struct inode
*inode
;
1961 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
1962 if (IS_ERR(inode
)) {
1963 es
->s_last_orphan
= 0;
1967 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
1969 if (inode
->i_nlink
) {
1970 ext4_msg(sb
, KERN_DEBUG
,
1971 "%s: truncating inode %lu to %lld bytes",
1972 __func__
, inode
->i_ino
, inode
->i_size
);
1973 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1974 inode
->i_ino
, inode
->i_size
);
1975 ext4_truncate(inode
);
1978 ext4_msg(sb
, KERN_DEBUG
,
1979 "%s: deleting unreferenced inode %lu",
1980 __func__
, inode
->i_ino
);
1981 jbd_debug(2, "deleting unreferenced inode %lu\n",
1985 iput(inode
); /* The delete magic happens here! */
1988 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1991 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
1992 PLURAL(nr_orphans
));
1994 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
1995 PLURAL(nr_truncates
));
1997 /* Turn quotas off */
1998 for (i
= 0; i
< MAXQUOTAS
; i
++) {
1999 if (sb_dqopt(sb
)->files
[i
])
2000 vfs_quota_off(sb
, i
, 0);
2003 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2007 * Maximal extent format file size.
2008 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2009 * extent format containers, within a sector_t, and within i_blocks
2010 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2011 * so that won't be a limiting factor.
2013 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2015 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2018 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2020 /* small i_blocks in vfs inode? */
2021 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2023 * CONFIG_LBDAF is not enabled implies the inode
2024 * i_block represent total blocks in 512 bytes
2025 * 32 == size of vfs inode i_blocks * 8
2027 upper_limit
= (1LL << 32) - 1;
2029 /* total blocks in file system block size */
2030 upper_limit
>>= (blkbits
- 9);
2031 upper_limit
<<= blkbits
;
2034 /* 32-bit extent-start container, ee_block */
2039 /* Sanity check against vm- & vfs- imposed limits */
2040 if (res
> upper_limit
)
2047 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2048 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2049 * We need to be 1 filesystem block less than the 2^48 sector limit.
2051 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2053 loff_t res
= EXT4_NDIR_BLOCKS
;
2056 /* This is calculated to be the largest file size for a dense, block
2057 * mapped file such that the file's total number of 512-byte sectors,
2058 * including data and all indirect blocks, does not exceed (2^48 - 1).
2060 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2061 * number of 512-byte sectors of the file.
2064 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2066 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2067 * the inode i_block field represents total file blocks in
2068 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2070 upper_limit
= (1LL << 32) - 1;
2072 /* total blocks in file system block size */
2073 upper_limit
>>= (bits
- 9);
2077 * We use 48 bit ext4_inode i_blocks
2078 * With EXT4_HUGE_FILE_FL set the i_blocks
2079 * represent total number of blocks in
2080 * file system block size
2082 upper_limit
= (1LL << 48) - 1;
2086 /* indirect blocks */
2088 /* double indirect blocks */
2089 meta_blocks
+= 1 + (1LL << (bits
-2));
2090 /* tripple indirect blocks */
2091 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2093 upper_limit
-= meta_blocks
;
2094 upper_limit
<<= bits
;
2096 res
+= 1LL << (bits
-2);
2097 res
+= 1LL << (2*(bits
-2));
2098 res
+= 1LL << (3*(bits
-2));
2100 if (res
> upper_limit
)
2103 if (res
> MAX_LFS_FILESIZE
)
2104 res
= MAX_LFS_FILESIZE
;
2109 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2110 ext4_fsblk_t logical_sb_block
, int nr
)
2112 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2113 ext4_group_t bg
, first_meta_bg
;
2116 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2118 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2120 return logical_sb_block
+ nr
+ 1;
2121 bg
= sbi
->s_desc_per_block
* nr
;
2122 if (ext4_bg_has_super(sb
, bg
))
2125 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2129 * ext4_get_stripe_size: Get the stripe size.
2130 * @sbi: In memory super block info
2132 * If we have specified it via mount option, then
2133 * use the mount option value. If the value specified at mount time is
2134 * greater than the blocks per group use the super block value.
2135 * If the super block value is greater than blocks per group return 0.
2136 * Allocator needs it be less than blocks per group.
2139 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2141 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2142 unsigned long stripe_width
=
2143 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2145 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2146 return sbi
->s_stripe
;
2148 if (stripe_width
<= sbi
->s_blocks_per_group
)
2149 return stripe_width
;
2151 if (stride
<= sbi
->s_blocks_per_group
)
2160 struct attribute attr
;
2161 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2162 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2163 const char *, size_t);
2167 static int parse_strtoul(const char *buf
,
2168 unsigned long max
, unsigned long *value
)
2172 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2173 endp
= skip_spaces(endp
);
2174 if (*endp
|| *value
> max
)
2180 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2181 struct ext4_sb_info
*sbi
,
2184 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2185 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2188 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2189 struct ext4_sb_info
*sbi
, char *buf
)
2191 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2193 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2194 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2195 sbi
->s_sectors_written_start
) >> 1);
2198 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2199 struct ext4_sb_info
*sbi
, char *buf
)
2201 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2203 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2204 (unsigned long long)(sbi
->s_kbytes_written
+
2205 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2206 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2209 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2210 struct ext4_sb_info
*sbi
,
2211 const char *buf
, size_t count
)
2215 if (parse_strtoul(buf
, 0x40000000, &t
))
2218 if (!is_power_of_2(t
))
2221 sbi
->s_inode_readahead_blks
= t
;
2225 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2226 struct ext4_sb_info
*sbi
, char *buf
)
2228 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2230 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2233 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2234 struct ext4_sb_info
*sbi
,
2235 const char *buf
, size_t count
)
2237 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2240 if (parse_strtoul(buf
, 0xffffffff, &t
))
2246 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2247 static struct ext4_attr ext4_attr_##_name = { \
2248 .attr = {.name = __stringify(_name), .mode = _mode }, \
2251 .offset = offsetof(struct ext4_sb_info, _elname), \
2253 #define EXT4_ATTR(name, mode, show, store) \
2254 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2256 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2257 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2258 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2259 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2260 #define ATTR_LIST(name) &ext4_attr_##name.attr
2262 EXT4_RO_ATTR(delayed_allocation_blocks
);
2263 EXT4_RO_ATTR(session_write_kbytes
);
2264 EXT4_RO_ATTR(lifetime_write_kbytes
);
2265 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2266 inode_readahead_blks_store
, s_inode_readahead_blks
);
2267 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2268 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2269 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2270 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2271 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2272 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2273 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2274 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2276 static struct attribute
*ext4_attrs
[] = {
2277 ATTR_LIST(delayed_allocation_blocks
),
2278 ATTR_LIST(session_write_kbytes
),
2279 ATTR_LIST(lifetime_write_kbytes
),
2280 ATTR_LIST(inode_readahead_blks
),
2281 ATTR_LIST(inode_goal
),
2282 ATTR_LIST(mb_stats
),
2283 ATTR_LIST(mb_max_to_scan
),
2284 ATTR_LIST(mb_min_to_scan
),
2285 ATTR_LIST(mb_order2_req
),
2286 ATTR_LIST(mb_stream_req
),
2287 ATTR_LIST(mb_group_prealloc
),
2288 ATTR_LIST(max_writeback_mb_bump
),
2292 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2293 struct attribute
*attr
, char *buf
)
2295 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2297 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2299 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2302 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2303 struct attribute
*attr
,
2304 const char *buf
, size_t len
)
2306 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2308 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2310 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2313 static void ext4_sb_release(struct kobject
*kobj
)
2315 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2317 complete(&sbi
->s_kobj_unregister
);
2321 static struct sysfs_ops ext4_attr_ops
= {
2322 .show
= ext4_attr_show
,
2323 .store
= ext4_attr_store
,
2326 static struct kobj_type ext4_ktype
= {
2327 .default_attrs
= ext4_attrs
,
2328 .sysfs_ops
= &ext4_attr_ops
,
2329 .release
= ext4_sb_release
,
2333 * Check whether this filesystem can be mounted based on
2334 * the features present and the RDONLY/RDWR mount requested.
2335 * Returns 1 if this filesystem can be mounted as requested,
2336 * 0 if it cannot be.
2338 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2340 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2341 ext4_msg(sb
, KERN_ERR
,
2342 "Couldn't mount because of "
2343 "unsupported optional features (%x)",
2344 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2345 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2352 /* Check that feature set is OK for a read-write mount */
2353 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2354 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2355 "unsupported optional features (%x)",
2356 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2357 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2361 * Large file size enabled file system can only be mounted
2362 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2364 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2365 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2366 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2367 "cannot be mounted RDWR without "
2375 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
2376 __releases(kernel_lock
)
2377 __acquires(kernel_lock
)
2379 struct buffer_head
*bh
;
2380 struct ext4_super_block
*es
= NULL
;
2381 struct ext4_sb_info
*sbi
;
2383 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2384 ext4_fsblk_t logical_sb_block
;
2385 unsigned long offset
= 0;
2386 unsigned long journal_devnum
= 0;
2387 unsigned long def_mount_opts
;
2393 unsigned int db_count
;
2395 int needs_recovery
, has_huge_files
;
2398 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2400 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
2404 sbi
->s_blockgroup_lock
=
2405 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
2406 if (!sbi
->s_blockgroup_lock
) {
2410 sb
->s_fs_info
= sbi
;
2411 sbi
->s_mount_opt
= 0;
2412 sbi
->s_resuid
= EXT4_DEF_RESUID
;
2413 sbi
->s_resgid
= EXT4_DEF_RESGID
;
2414 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
2415 sbi
->s_sb_block
= sb_block
;
2416 sbi
->s_sectors_written_start
= part_stat_read(sb
->s_bdev
->bd_part
,
2421 /* Cleanup superblock name */
2422 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
2425 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
2427 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
2432 * The ext4 superblock will not be buffer aligned for other than 1kB
2433 * block sizes. We need to calculate the offset from buffer start.
2435 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
2436 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2437 offset
= do_div(logical_sb_block
, blocksize
);
2439 logical_sb_block
= sb_block
;
2442 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
2443 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
2447 * Note: s_es must be initialized as soon as possible because
2448 * some ext4 macro-instructions depend on its value
2450 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2452 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
2453 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
2455 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
2457 /* Set defaults before we parse the mount options */
2458 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
2459 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
2460 set_opt(sbi
->s_mount_opt
, DEBUG
);
2461 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
2462 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
2464 set_opt(sbi
->s_mount_opt
, GRPID
);
2466 if (def_mount_opts
& EXT4_DEFM_UID16
)
2467 set_opt(sbi
->s_mount_opt
, NO_UID32
);
2468 #ifdef CONFIG_EXT4_FS_XATTR
2469 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
2470 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
2472 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2473 if (def_mount_opts
& EXT4_DEFM_ACL
)
2474 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
2476 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
2477 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2478 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
2479 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2480 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
2481 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2483 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2484 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2485 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2486 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2488 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2490 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2491 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2492 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
2493 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
2494 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
2496 set_opt(sbi
->s_mount_opt
, BARRIER
);
2499 * enable delayed allocation by default
2500 * Use -o nodelalloc to turn it off
2502 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2504 if (!parse_options((char *) data
, sb
, &journal_devnum
,
2505 &journal_ioprio
, NULL
, 0))
2508 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2509 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
2511 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2512 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2513 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2514 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2515 ext4_msg(sb
, KERN_WARNING
,
2516 "feature flags set on rev 0 fs, "
2517 "running e2fsck is recommended");
2520 * Check feature flags regardless of the revision level, since we
2521 * previously didn't change the revision level when setting the flags,
2522 * so there is a chance incompat flags are set on a rev 0 filesystem.
2524 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
2527 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2529 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2530 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2531 ext4_msg(sb
, KERN_ERR
,
2532 "Unsupported filesystem blocksize %d", blocksize
);
2536 if (sb
->s_blocksize
!= blocksize
) {
2537 /* Validate the filesystem blocksize */
2538 if (!sb_set_blocksize(sb
, blocksize
)) {
2539 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
2545 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2546 offset
= do_div(logical_sb_block
, blocksize
);
2547 bh
= sb_bread(sb
, logical_sb_block
);
2549 ext4_msg(sb
, KERN_ERR
,
2550 "Can't read superblock on 2nd try");
2553 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2555 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2556 ext4_msg(sb
, KERN_ERR
,
2557 "Magic mismatch, very weird!");
2562 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2563 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
2564 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
2566 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
2568 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2569 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2570 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2572 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2573 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2574 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2575 (!is_power_of_2(sbi
->s_inode_size
)) ||
2576 (sbi
->s_inode_size
> blocksize
)) {
2577 ext4_msg(sb
, KERN_ERR
,
2578 "unsupported inode size: %d",
2582 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2583 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2586 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2587 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2588 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2589 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2590 !is_power_of_2(sbi
->s_desc_size
)) {
2591 ext4_msg(sb
, KERN_ERR
,
2592 "unsupported descriptor size %lu",
2597 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2599 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2600 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2601 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2604 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2605 if (sbi
->s_inodes_per_block
== 0)
2607 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2608 sbi
->s_inodes_per_block
;
2609 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2611 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2612 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2613 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2615 for (i
= 0; i
< 4; i
++)
2616 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2617 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2618 i
= le32_to_cpu(es
->s_flags
);
2619 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
2620 sbi
->s_hash_unsigned
= 3;
2621 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
2622 #ifdef __CHAR_UNSIGNED__
2623 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
2624 sbi
->s_hash_unsigned
= 3;
2626 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
2631 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2632 ext4_msg(sb
, KERN_ERR
,
2633 "#blocks per group too big: %lu",
2634 sbi
->s_blocks_per_group
);
2637 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2638 ext4_msg(sb
, KERN_ERR
,
2639 "#inodes per group too big: %lu",
2640 sbi
->s_inodes_per_group
);
2645 * Test whether we have more sectors than will fit in sector_t,
2646 * and whether the max offset is addressable by the page cache.
2648 if ((ext4_blocks_count(es
) >
2649 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) ||
2650 (ext4_blocks_count(es
) >
2651 (pgoff_t
)(~0ULL) >> (PAGE_CACHE_SHIFT
- sb
->s_blocksize_bits
))) {
2652 ext4_msg(sb
, KERN_ERR
, "filesystem"
2653 " too large to mount safely on this system");
2654 if (sizeof(sector_t
) < 8)
2655 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
2660 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2663 /* check blocks count against device size */
2664 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
2665 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
2666 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
2667 "exceeds size of device (%llu blocks)",
2668 ext4_blocks_count(es
), blocks_count
);
2673 * It makes no sense for the first data block to be beyond the end
2674 * of the filesystem.
2676 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
2677 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
2678 "block %u is beyond end of filesystem (%llu)",
2679 le32_to_cpu(es
->s_first_data_block
),
2680 ext4_blocks_count(es
));
2683 blocks_count
= (ext4_blocks_count(es
) -
2684 le32_to_cpu(es
->s_first_data_block
) +
2685 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2686 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2687 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
2688 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
2689 "(block count %llu, first data block %u, "
2690 "blocks per group %lu)", sbi
->s_groups_count
,
2691 ext4_blocks_count(es
),
2692 le32_to_cpu(es
->s_first_data_block
),
2693 EXT4_BLOCKS_PER_GROUP(sb
));
2696 sbi
->s_groups_count
= blocks_count
;
2697 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
2698 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
2699 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2700 EXT4_DESC_PER_BLOCK(sb
);
2701 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2703 if (sbi
->s_group_desc
== NULL
) {
2704 ext4_msg(sb
, KERN_ERR
, "not enough memory");
2708 #ifdef CONFIG_PROC_FS
2710 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2713 bgl_lock_init(sbi
->s_blockgroup_lock
);
2715 for (i
= 0; i
< db_count
; i
++) {
2716 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2717 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2718 if (!sbi
->s_group_desc
[i
]) {
2719 ext4_msg(sb
, KERN_ERR
,
2720 "can't read group descriptor %d", i
);
2725 if (!ext4_check_descriptors(sb
)) {
2726 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
2729 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2730 if (!ext4_fill_flex_info(sb
)) {
2731 ext4_msg(sb
, KERN_ERR
,
2732 "unable to initialize "
2733 "flex_bg meta info!");
2737 sbi
->s_gdb_count
= db_count
;
2738 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2739 spin_lock_init(&sbi
->s_next_gen_lock
);
2741 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
2742 ext4_count_free_blocks(sb
));
2744 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
2745 ext4_count_free_inodes(sb
));
2748 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
2749 ext4_count_dirs(sb
));
2752 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
2755 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
2759 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2760 sbi
->s_max_writeback_mb_bump
= 128;
2763 * set up enough so that it can read an inode
2765 if (!test_opt(sb
, NOLOAD
) &&
2766 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
2767 sb
->s_op
= &ext4_sops
;
2769 sb
->s_op
= &ext4_nojournal_sops
;
2770 sb
->s_export_op
= &ext4_export_ops
;
2771 sb
->s_xattr
= ext4_xattr_handlers
;
2773 sb
->s_qcop
= &ext4_qctl_operations
;
2774 sb
->dq_op
= &ext4_quota_operations
;
2776 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2777 mutex_init(&sbi
->s_orphan_lock
);
2778 mutex_init(&sbi
->s_resize_lock
);
2782 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2783 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2784 EXT4_FEATURE_INCOMPAT_RECOVER
));
2787 * The first inode we look at is the journal inode. Don't try
2788 * root first: it may be modified in the journal!
2790 if (!test_opt(sb
, NOLOAD
) &&
2791 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2792 if (ext4_load_journal(sb
, es
, journal_devnum
))
2794 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
2795 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2796 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
2797 "suppressed and not mounted read-only");
2800 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
2801 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2802 sbi
->s_journal
= NULL
;
2807 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2808 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2809 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2810 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
2814 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2815 jbd2_journal_set_features(sbi
->s_journal
,
2816 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2817 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2818 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2819 jbd2_journal_set_features(sbi
->s_journal
,
2820 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2821 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2822 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2824 jbd2_journal_clear_features(sbi
->s_journal
,
2825 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2826 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2829 /* We have now updated the journal if required, so we can
2830 * validate the data journaling mode. */
2831 switch (test_opt(sb
, DATA_FLAGS
)) {
2833 /* No mode set, assume a default based on the journal
2834 * capabilities: ORDERED_DATA if the journal can
2835 * cope, else JOURNAL_DATA
2837 if (jbd2_journal_check_available_features
2838 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2839 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2841 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2844 case EXT4_MOUNT_ORDERED_DATA
:
2845 case EXT4_MOUNT_WRITEBACK_DATA
:
2846 if (!jbd2_journal_check_available_features
2847 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2848 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
2849 "requested data journaling mode");
2855 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
2859 if (test_opt(sb
, NOBH
)) {
2860 if (!(test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)) {
2861 ext4_msg(sb
, KERN_WARNING
, "Ignoring nobh option - "
2862 "its supported only with writeback mode");
2863 clear_opt(sbi
->s_mount_opt
, NOBH
);
2866 EXT4_SB(sb
)->dio_unwritten_wq
= create_workqueue("ext4-dio-unwritten");
2867 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
2868 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
2869 goto failed_mount_wq
;
2873 * The jbd2_journal_load will have done any necessary log recovery,
2874 * so we can safely mount the rest of the filesystem now.
2877 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
2879 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
2880 ret
= PTR_ERR(root
);
2883 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
2885 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
2888 sb
->s_root
= d_alloc_root(root
);
2890 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
2896 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
2898 /* determine the minimum size of new large inodes, if present */
2899 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
2900 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2901 EXT4_GOOD_OLD_INODE_SIZE
;
2902 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2903 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
2904 if (sbi
->s_want_extra_isize
<
2905 le16_to_cpu(es
->s_want_extra_isize
))
2906 sbi
->s_want_extra_isize
=
2907 le16_to_cpu(es
->s_want_extra_isize
);
2908 if (sbi
->s_want_extra_isize
<
2909 le16_to_cpu(es
->s_min_extra_isize
))
2910 sbi
->s_want_extra_isize
=
2911 le16_to_cpu(es
->s_min_extra_isize
);
2914 /* Check if enough inode space is available */
2915 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
2916 sbi
->s_inode_size
) {
2917 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
2918 EXT4_GOOD_OLD_INODE_SIZE
;
2919 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
2923 if (test_opt(sb
, DELALLOC
) &&
2924 (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)) {
2925 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
2926 "requested data journaling mode");
2927 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
2930 err
= ext4_setup_system_zone(sb
);
2932 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
2933 "zone (%d)\n", err
);
2938 err
= ext4_mb_init(sb
, needs_recovery
);
2940 ext4_msg(sb
, KERN_ERR
, "failed to initalize mballoc (%d)",
2945 sbi
->s_kobj
.kset
= ext4_kset
;
2946 init_completion(&sbi
->s_kobj_unregister
);
2947 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
2950 ext4_mb_release(sb
);
2951 ext4_ext_release(sb
);
2955 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
2956 ext4_orphan_cleanup(sb
, es
);
2957 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
2958 if (needs_recovery
) {
2959 ext4_msg(sb
, KERN_INFO
, "recovery complete");
2960 ext4_mark_recovery_complete(sb
, es
);
2962 if (EXT4_SB(sb
)->s_journal
) {
2963 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2964 descr
= " journalled data mode";
2965 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2966 descr
= " ordered data mode";
2968 descr
= " writeback data mode";
2970 descr
= "out journal";
2972 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s", descr
);
2979 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
2983 ext4_msg(sb
, KERN_ERR
, "mount failed");
2984 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
2986 ext4_release_system_zone(sb
);
2987 if (sbi
->s_journal
) {
2988 jbd2_journal_destroy(sbi
->s_journal
);
2989 sbi
->s_journal
= NULL
;
2992 if (sbi
->s_flex_groups
) {
2993 if (is_vmalloc_addr(sbi
->s_flex_groups
))
2994 vfree(sbi
->s_flex_groups
);
2996 kfree(sbi
->s_flex_groups
);
2998 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
2999 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3000 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3001 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
3003 for (i
= 0; i
< db_count
; i
++)
3004 brelse(sbi
->s_group_desc
[i
]);
3005 kfree(sbi
->s_group_desc
);
3008 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3011 for (i
= 0; i
< MAXQUOTAS
; i
++)
3012 kfree(sbi
->s_qf_names
[i
]);
3014 ext4_blkdev_remove(sbi
);
3017 sb
->s_fs_info
= NULL
;
3018 kfree(sbi
->s_blockgroup_lock
);
3025 * Setup any per-fs journal parameters now. We'll do this both on
3026 * initial mount, once the journal has been initialised but before we've
3027 * done any recovery; and again on any subsequent remount.
3029 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3031 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3033 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3034 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3035 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3037 spin_lock(&journal
->j_state_lock
);
3038 if (test_opt(sb
, BARRIER
))
3039 journal
->j_flags
|= JBD2_BARRIER
;
3041 journal
->j_flags
&= ~JBD2_BARRIER
;
3042 if (test_opt(sb
, DATA_ERR_ABORT
))
3043 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3045 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3046 spin_unlock(&journal
->j_state_lock
);
3049 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3050 unsigned int journal_inum
)
3052 struct inode
*journal_inode
;
3055 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3057 /* First, test for the existence of a valid inode on disk. Bad
3058 * things happen if we iget() an unused inode, as the subsequent
3059 * iput() will try to delete it. */
3061 journal_inode
= ext4_iget(sb
, journal_inum
);
3062 if (IS_ERR(journal_inode
)) {
3063 ext4_msg(sb
, KERN_ERR
, "no journal found");
3066 if (!journal_inode
->i_nlink
) {
3067 make_bad_inode(journal_inode
);
3068 iput(journal_inode
);
3069 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3073 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3074 journal_inode
, journal_inode
->i_size
);
3075 if (!S_ISREG(journal_inode
->i_mode
)) {
3076 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3077 iput(journal_inode
);
3081 journal
= jbd2_journal_init_inode(journal_inode
);
3083 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3084 iput(journal_inode
);
3087 journal
->j_private
= sb
;
3088 ext4_init_journal_params(sb
, journal
);
3092 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3095 struct buffer_head
*bh
;
3099 int hblock
, blocksize
;
3100 ext4_fsblk_t sb_block
;
3101 unsigned long offset
;
3102 struct ext4_super_block
*es
;
3103 struct block_device
*bdev
;
3105 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3107 bdev
= ext4_blkdev_get(j_dev
, sb
);
3111 if (bd_claim(bdev
, sb
)) {
3112 ext4_msg(sb
, KERN_ERR
,
3113 "failed to claim external journal device");
3114 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
3118 blocksize
= sb
->s_blocksize
;
3119 hblock
= bdev_logical_block_size(bdev
);
3120 if (blocksize
< hblock
) {
3121 ext4_msg(sb
, KERN_ERR
,
3122 "blocksize too small for journal device");
3126 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3127 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3128 set_blocksize(bdev
, blocksize
);
3129 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3130 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3131 "external journal");
3135 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3136 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3137 !(le32_to_cpu(es
->s_feature_incompat
) &
3138 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3139 ext4_msg(sb
, KERN_ERR
, "external journal has "
3145 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3146 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3151 len
= ext4_blocks_count(es
);
3152 start
= sb_block
+ 1;
3153 brelse(bh
); /* we're done with the superblock */
3155 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3156 start
, len
, blocksize
);
3158 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3161 journal
->j_private
= sb
;
3162 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3163 wait_on_buffer(journal
->j_sb_buffer
);
3164 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3165 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3168 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3169 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3170 "user (unsupported) - %d",
3171 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3174 EXT4_SB(sb
)->journal_bdev
= bdev
;
3175 ext4_init_journal_params(sb
, journal
);
3179 jbd2_journal_destroy(journal
);
3181 ext4_blkdev_put(bdev
);
3185 static int ext4_load_journal(struct super_block
*sb
,
3186 struct ext4_super_block
*es
,
3187 unsigned long journal_devnum
)
3190 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3193 int really_read_only
;
3195 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3197 if (journal_devnum
&&
3198 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3199 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3200 "numbers have changed");
3201 journal_dev
= new_decode_dev(journal_devnum
);
3203 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3205 really_read_only
= bdev_read_only(sb
->s_bdev
);
3208 * Are we loading a blank journal or performing recovery after a
3209 * crash? For recovery, we need to check in advance whether we
3210 * can get read-write access to the device.
3212 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3213 if (sb
->s_flags
& MS_RDONLY
) {
3214 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3215 "required on readonly filesystem");
3216 if (really_read_only
) {
3217 ext4_msg(sb
, KERN_ERR
, "write access "
3218 "unavailable, cannot proceed");
3221 ext4_msg(sb
, KERN_INFO
, "write access will "
3222 "be enabled during recovery");
3226 if (journal_inum
&& journal_dev
) {
3227 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3228 "and inode journals!");
3233 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3236 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3240 if (!(journal
->j_flags
& JBD2_BARRIER
))
3241 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3243 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
3244 err
= jbd2_journal_update_format(journal
);
3246 ext4_msg(sb
, KERN_ERR
, "error updating journal");
3247 jbd2_journal_destroy(journal
);
3252 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
3253 err
= jbd2_journal_wipe(journal
, !really_read_only
);
3255 err
= jbd2_journal_load(journal
);
3258 ext4_msg(sb
, KERN_ERR
, "error loading journal");
3259 jbd2_journal_destroy(journal
);
3263 EXT4_SB(sb
)->s_journal
= journal
;
3264 ext4_clear_journal_err(sb
, es
);
3266 if (journal_devnum
&&
3267 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3268 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
3270 /* Make sure we flush the recovery flag to disk. */
3271 ext4_commit_super(sb
, 1);
3277 static int ext4_commit_super(struct super_block
*sb
, int sync
)
3279 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
3280 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
3285 if (buffer_write_io_error(sbh
)) {
3287 * Oh, dear. A previous attempt to write the
3288 * superblock failed. This could happen because the
3289 * USB device was yanked out. Or it could happen to
3290 * be a transient write error and maybe the block will
3291 * be remapped. Nothing we can do but to retry the
3292 * write and hope for the best.
3294 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
3295 "superblock detected");
3296 clear_buffer_write_io_error(sbh
);
3297 set_buffer_uptodate(sbh
);
3300 * If the file system is mounted read-only, don't update the
3301 * superblock write time. This avoids updating the superblock
3302 * write time when we are mounting the root file system
3303 * read/only but we need to replay the journal; at that point,
3304 * for people who are east of GMT and who make their clock
3305 * tick in localtime for Windows bug-for-bug compatibility,
3306 * the clock is set in the future, and this will cause e2fsck
3307 * to complain and force a full file system check.
3309 if (!(sb
->s_flags
& MS_RDONLY
))
3310 es
->s_wtime
= cpu_to_le32(get_seconds());
3311 es
->s_kbytes_written
=
3312 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
3313 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
3314 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
3315 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
3316 &EXT4_SB(sb
)->s_freeblocks_counter
));
3317 es
->s_free_inodes_count
= cpu_to_le32(percpu_counter_sum_positive(
3318 &EXT4_SB(sb
)->s_freeinodes_counter
));
3320 BUFFER_TRACE(sbh
, "marking dirty");
3321 mark_buffer_dirty(sbh
);
3323 error
= sync_dirty_buffer(sbh
);
3327 error
= buffer_write_io_error(sbh
);
3329 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
3331 clear_buffer_write_io_error(sbh
);
3332 set_buffer_uptodate(sbh
);
3339 * Have we just finished recovery? If so, and if we are mounting (or
3340 * remounting) the filesystem readonly, then we will end up with a
3341 * consistent fs on disk. Record that fact.
3343 static void ext4_mark_recovery_complete(struct super_block
*sb
,
3344 struct ext4_super_block
*es
)
3346 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
3348 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3349 BUG_ON(journal
!= NULL
);
3352 jbd2_journal_lock_updates(journal
);
3353 if (jbd2_journal_flush(journal
) < 0)
3356 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
3357 sb
->s_flags
& MS_RDONLY
) {
3358 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3359 ext4_commit_super(sb
, 1);
3363 jbd2_journal_unlock_updates(journal
);
3367 * If we are mounting (or read-write remounting) a filesystem whose journal
3368 * has recorded an error from a previous lifetime, move that error to the
3369 * main filesystem now.
3371 static void ext4_clear_journal_err(struct super_block
*sb
,
3372 struct ext4_super_block
*es
)
3378 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3380 journal
= EXT4_SB(sb
)->s_journal
;
3383 * Now check for any error status which may have been recorded in the
3384 * journal by a prior ext4_error() or ext4_abort()
3387 j_errno
= jbd2_journal_errno(journal
);
3391 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
3392 ext4_warning(sb
, "Filesystem error recorded "
3393 "from previous mount: %s", errstr
);
3394 ext4_warning(sb
, "Marking fs in need of filesystem check.");
3396 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
3397 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
3398 ext4_commit_super(sb
, 1);
3400 jbd2_journal_clear_err(journal
);
3405 * Force the running and committing transactions to commit,
3406 * and wait on the commit.
3408 int ext4_force_commit(struct super_block
*sb
)
3413 if (sb
->s_flags
& MS_RDONLY
)
3416 journal
= EXT4_SB(sb
)->s_journal
;
3418 ret
= ext4_journal_force_commit(journal
);
3423 static void ext4_write_super(struct super_block
*sb
)
3426 ext4_commit_super(sb
, 1);
3430 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
3434 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3436 trace_ext4_sync_fs(sb
, wait
);
3437 flush_workqueue(sbi
->dio_unwritten_wq
);
3438 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
3440 jbd2_log_wait_commit(sbi
->s_journal
, target
);
3446 * LVM calls this function before a (read-only) snapshot is created. This
3447 * gives us a chance to flush the journal completely and mark the fs clean.
3449 static int ext4_freeze(struct super_block
*sb
)
3454 if (sb
->s_flags
& MS_RDONLY
)
3457 journal
= EXT4_SB(sb
)->s_journal
;
3459 /* Now we set up the journal barrier. */
3460 jbd2_journal_lock_updates(journal
);
3463 * Don't clear the needs_recovery flag if we failed to flush
3466 error
= jbd2_journal_flush(journal
);
3469 jbd2_journal_unlock_updates(journal
);
3473 /* Journal blocked and flushed, clear needs_recovery flag. */
3474 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3475 error
= ext4_commit_super(sb
, 1);
3482 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3483 * flag here, even though the filesystem is not technically dirty yet.
3485 static int ext4_unfreeze(struct super_block
*sb
)
3487 if (sb
->s_flags
& MS_RDONLY
)
3491 /* Reset the needs_recovery flag before the fs is unlocked. */
3492 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3493 ext4_commit_super(sb
, 1);
3495 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3499 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3501 struct ext4_super_block
*es
;
3502 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3503 ext4_fsblk_t n_blocks_count
= 0;
3504 unsigned long old_sb_flags
;
3505 struct ext4_mount_options old_opts
;
3507 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3515 /* Store the original options */
3517 old_sb_flags
= sb
->s_flags
;
3518 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3519 old_opts
.s_resuid
= sbi
->s_resuid
;
3520 old_opts
.s_resgid
= sbi
->s_resgid
;
3521 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3522 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
3523 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
3525 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3526 for (i
= 0; i
< MAXQUOTAS
; i
++)
3527 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3529 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
3530 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
3533 * Allow the "check" option to be passed as a remount option.
3535 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
3536 &n_blocks_count
, 1)) {
3541 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
3542 ext4_abort(sb
, __func__
, "Abort forced by user");
3544 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3545 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3549 if (sbi
->s_journal
) {
3550 ext4_init_journal_params(sb
, sbi
->s_journal
);
3551 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3554 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3555 n_blocks_count
> ext4_blocks_count(es
)) {
3556 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
3561 if (*flags
& MS_RDONLY
) {
3563 * First of all, the unconditional stuff we have to do
3564 * to disable replay of the journal when we next remount
3566 sb
->s_flags
|= MS_RDONLY
;
3569 * OK, test if we are remounting a valid rw partition
3570 * readonly, and if so set the rdonly flag and then
3571 * mark the partition as valid again.
3573 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3574 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3575 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3578 ext4_mark_recovery_complete(sb
, es
);
3580 /* Make sure we can mount this feature set readwrite */
3581 if (!ext4_feature_set_ok(sb
, 0)) {
3586 * Make sure the group descriptor checksums
3587 * are sane. If they aren't, refuse to remount r/w.
3589 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3590 struct ext4_group_desc
*gdp
=
3591 ext4_get_group_desc(sb
, g
, NULL
);
3593 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3594 ext4_msg(sb
, KERN_ERR
,
3595 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3596 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3597 le16_to_cpu(gdp
->bg_checksum
));
3604 * If we have an unprocessed orphan list hanging
3605 * around from a previously readonly bdev mount,
3606 * require a full umount/remount for now.
3608 if (es
->s_last_orphan
) {
3609 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
3610 "remount RDWR because of unprocessed "
3611 "orphan inode list. Please "
3612 "umount/remount instead");
3618 * Mounting a RDONLY partition read-write, so reread
3619 * and store the current valid flag. (It may have
3620 * been changed by e2fsck since we originally mounted
3624 ext4_clear_journal_err(sb
, es
);
3625 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3626 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3628 if (!ext4_setup_super(sb
, es
, 0))
3629 sb
->s_flags
&= ~MS_RDONLY
;
3632 ext4_setup_system_zone(sb
);
3633 if (sbi
->s_journal
== NULL
)
3634 ext4_commit_super(sb
, 1);
3637 /* Release old quota file names */
3638 for (i
= 0; i
< MAXQUOTAS
; i
++)
3639 if (old_opts
.s_qf_names
[i
] &&
3640 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3641 kfree(old_opts
.s_qf_names
[i
]);
3648 sb
->s_flags
= old_sb_flags
;
3649 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3650 sbi
->s_resuid
= old_opts
.s_resuid
;
3651 sbi
->s_resgid
= old_opts
.s_resgid
;
3652 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3653 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
3654 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
3656 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3657 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3658 if (sbi
->s_qf_names
[i
] &&
3659 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3660 kfree(sbi
->s_qf_names
[i
]);
3661 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3669 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3671 struct super_block
*sb
= dentry
->d_sb
;
3672 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3673 struct ext4_super_block
*es
= sbi
->s_es
;
3676 if (test_opt(sb
, MINIX_DF
)) {
3677 sbi
->s_overhead_last
= 0;
3678 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3679 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3680 ext4_fsblk_t overhead
= 0;
3683 * Compute the overhead (FS structures). This is constant
3684 * for a given filesystem unless the number of block groups
3685 * changes so we cache the previous value until it does.
3689 * All of the blocks before first_data_block are
3692 overhead
= le32_to_cpu(es
->s_first_data_block
);
3695 * Add the overhead attributed to the superblock and
3696 * block group descriptors. If the sparse superblocks
3697 * feature is turned on, then not all groups have this.
3699 for (i
= 0; i
< ngroups
; i
++) {
3700 overhead
+= ext4_bg_has_super(sb
, i
) +
3701 ext4_bg_num_gdb(sb
, i
);
3706 * Every block group has an inode bitmap, a block
3707 * bitmap, and an inode table.
3709 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3710 sbi
->s_overhead_last
= overhead
;
3712 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3715 buf
->f_type
= EXT4_SUPER_MAGIC
;
3716 buf
->f_bsize
= sb
->s_blocksize
;
3717 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3718 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3719 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3720 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3721 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3723 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3724 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3725 buf
->f_namelen
= EXT4_NAME_LEN
;
3726 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3727 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3728 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3729 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3734 /* Helper function for writing quotas on sync - we need to start transaction
3735 * before quota file is locked for write. Otherwise the are possible deadlocks:
3736 * Process 1 Process 2
3737 * ext4_create() quota_sync()
3738 * jbd2_journal_start() write_dquot()
3739 * vfs_dq_init() down(dqio_mutex)
3740 * down(dqio_mutex) jbd2_journal_start()
3746 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3748 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3751 static int ext4_write_dquot(struct dquot
*dquot
)
3755 struct inode
*inode
;
3757 inode
= dquot_to_inode(dquot
);
3758 handle
= ext4_journal_start(inode
,
3759 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3761 return PTR_ERR(handle
);
3762 ret
= dquot_commit(dquot
);
3763 err
= ext4_journal_stop(handle
);
3769 static int ext4_acquire_dquot(struct dquot
*dquot
)
3774 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3775 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3777 return PTR_ERR(handle
);
3778 ret
= dquot_acquire(dquot
);
3779 err
= ext4_journal_stop(handle
);
3785 static int ext4_release_dquot(struct dquot
*dquot
)
3790 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3791 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3792 if (IS_ERR(handle
)) {
3793 /* Release dquot anyway to avoid endless cycle in dqput() */
3794 dquot_release(dquot
);
3795 return PTR_ERR(handle
);
3797 ret
= dquot_release(dquot
);
3798 err
= ext4_journal_stop(handle
);
3804 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
3806 /* Are we journaling quotas? */
3807 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
3808 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
3809 dquot_mark_dquot_dirty(dquot
);
3810 return ext4_write_dquot(dquot
);
3812 return dquot_mark_dquot_dirty(dquot
);
3816 static int ext4_write_info(struct super_block
*sb
, int type
)
3821 /* Data block + inode block */
3822 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
3824 return PTR_ERR(handle
);
3825 ret
= dquot_commit_info(sb
, type
);
3826 err
= ext4_journal_stop(handle
);
3833 * Turn on quotas during mount time - we need to find
3834 * the quota file and such...
3836 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
3838 return vfs_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
3839 EXT4_SB(sb
)->s_jquota_fmt
, type
);
3843 * Standard function to be called on quota_on
3845 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
3846 char *name
, int remount
)
3851 if (!test_opt(sb
, QUOTA
))
3853 /* When remounting, no checks are needed and in fact, name is NULL */
3855 return vfs_quota_on(sb
, type
, format_id
, name
, remount
);
3857 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
3861 /* Quotafile not on the same filesystem? */
3862 if (path
.mnt
->mnt_sb
!= sb
) {
3866 /* Journaling quota? */
3867 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
3868 /* Quotafile not in fs root? */
3869 if (path
.dentry
->d_parent
!= sb
->s_root
)
3870 ext4_msg(sb
, KERN_WARNING
,
3871 "Quota file not on filesystem root. "
3872 "Journaled quota will not work");
3876 * When we journal data on quota file, we have to flush journal to see
3877 * all updates to the file when we bypass pagecache...
3879 if (EXT4_SB(sb
)->s_journal
&&
3880 ext4_should_journal_data(path
.dentry
->d_inode
)) {
3882 * We don't need to lock updates but journal_flush() could
3883 * otherwise be livelocked...
3885 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
3886 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
3887 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3894 err
= vfs_quota_on_path(sb
, type
, format_id
, &path
);
3899 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3900 * acquiring the locks... As quota files are never truncated and quota code
3901 * itself serializes the operations (and noone else should touch the files)
3902 * we don't have to be afraid of races */
3903 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
3904 size_t len
, loff_t off
)
3906 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3907 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3909 int offset
= off
& (sb
->s_blocksize
- 1);
3912 struct buffer_head
*bh
;
3913 loff_t i_size
= i_size_read(inode
);
3917 if (off
+len
> i_size
)
3920 while (toread
> 0) {
3921 tocopy
= sb
->s_blocksize
- offset
< toread
?
3922 sb
->s_blocksize
- offset
: toread
;
3923 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
3926 if (!bh
) /* A hole? */
3927 memset(data
, 0, tocopy
);
3929 memcpy(data
, bh
->b_data
+offset
, tocopy
);
3939 /* Write to quotafile (we know the transaction is already started and has
3940 * enough credits) */
3941 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
3942 const char *data
, size_t len
, loff_t off
)
3944 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
3945 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
3947 int offset
= off
& (sb
->s_blocksize
- 1);
3949 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
3950 size_t towrite
= len
;
3951 struct buffer_head
*bh
;
3952 handle_t
*handle
= journal_current_handle();
3954 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
3955 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
3956 " cancelled because transaction is not started",
3957 (unsigned long long)off
, (unsigned long long)len
);
3960 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
3961 while (towrite
> 0) {
3962 tocopy
= sb
->s_blocksize
- offset
< towrite
?
3963 sb
->s_blocksize
- offset
: towrite
;
3964 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
3967 if (journal_quota
) {
3968 err
= ext4_journal_get_write_access(handle
, bh
);
3975 memcpy(bh
->b_data
+offset
, data
, tocopy
);
3976 flush_dcache_page(bh
->b_page
);
3979 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
3981 /* Always do at least ordered writes for quotas */
3982 err
= ext4_jbd2_file_inode(handle
, inode
);
3983 mark_buffer_dirty(bh
);
3994 if (len
== towrite
) {
3995 mutex_unlock(&inode
->i_mutex
);
3998 if (inode
->i_size
< off
+len
-towrite
) {
3999 i_size_write(inode
, off
+len
-towrite
);
4000 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4002 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4003 ext4_mark_inode_dirty(handle
, inode
);
4004 mutex_unlock(&inode
->i_mutex
);
4005 return len
- towrite
;
4010 static int ext4_get_sb(struct file_system_type
*fs_type
, int flags
,
4011 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
4013 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
,mnt
);
4016 #if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4017 static struct file_system_type ext2_fs_type
= {
4018 .owner
= THIS_MODULE
,
4020 .get_sb
= ext4_get_sb
,
4021 .kill_sb
= kill_block_super
,
4022 .fs_flags
= FS_REQUIRES_DEV
,
4025 static inline void register_as_ext2(void)
4027 int err
= register_filesystem(&ext2_fs_type
);
4030 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4033 static inline void unregister_as_ext2(void)
4035 unregister_filesystem(&ext2_fs_type
);
4037 MODULE_ALIAS("ext2");
4039 static inline void register_as_ext2(void) { }
4040 static inline void unregister_as_ext2(void) { }
4043 #if !defined(CONTIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4044 static struct file_system_type ext3_fs_type
= {
4045 .owner
= THIS_MODULE
,
4047 .get_sb
= ext4_get_sb
,
4048 .kill_sb
= kill_block_super
,
4049 .fs_flags
= FS_REQUIRES_DEV
,
4052 static inline void register_as_ext3(void)
4054 int err
= register_filesystem(&ext3_fs_type
);
4057 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4060 static inline void unregister_as_ext3(void)
4062 unregister_filesystem(&ext3_fs_type
);
4064 MODULE_ALIAS("ext3");
4066 static inline void register_as_ext3(void) { }
4067 static inline void unregister_as_ext3(void) { }
4070 static struct file_system_type ext4_fs_type
= {
4071 .owner
= THIS_MODULE
,
4073 .get_sb
= ext4_get_sb
,
4074 .kill_sb
= kill_block_super
,
4075 .fs_flags
= FS_REQUIRES_DEV
,
4078 static int __init
init_ext4_fs(void)
4082 err
= init_ext4_system_zone();
4085 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4088 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4089 err
= init_ext4_mballoc();
4093 err
= init_ext4_xattr();
4096 err
= init_inodecache();
4101 err
= register_filesystem(&ext4_fs_type
);
4106 unregister_as_ext2();
4107 unregister_as_ext3();
4108 destroy_inodecache();
4112 exit_ext4_mballoc();
4114 remove_proc_entry("fs/ext4", NULL
);
4115 kset_unregister(ext4_kset
);
4117 exit_ext4_system_zone();
4121 static void __exit
exit_ext4_fs(void)
4123 unregister_as_ext2();
4124 unregister_as_ext3();
4125 unregister_filesystem(&ext4_fs_type
);
4126 destroy_inodecache();
4128 exit_ext4_mballoc();
4129 remove_proc_entry("fs/ext4", NULL
);
4130 kset_unregister(ext4_kset
);
4131 exit_ext4_system_zone();
4134 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4135 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4136 MODULE_LICENSE("GPL");
4137 module_init(init_ext4_fs
)
4138 module_exit(exit_ext4_fs
)