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/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_jbd2.h"
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/ext4.h>
55 static struct proc_dir_entry
*ext4_proc_root
;
56 static struct kset
*ext4_kset
;
57 struct ext4_lazy_init
*ext4_li_info
;
58 struct mutex ext4_li_mtx
;
59 struct ext4_features
*ext4_feat
;
61 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
62 unsigned long journal_devnum
);
63 static int ext4_commit_super(struct super_block
*sb
, int sync
);
64 static void ext4_mark_recovery_complete(struct super_block
*sb
,
65 struct ext4_super_block
*es
);
66 static void ext4_clear_journal_err(struct super_block
*sb
,
67 struct ext4_super_block
*es
);
68 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
69 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
71 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
72 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
73 static int ext4_unfreeze(struct super_block
*sb
);
74 static void ext4_write_super(struct super_block
*sb
);
75 static int ext4_freeze(struct super_block
*sb
);
76 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
77 const char *dev_name
, void *data
);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block
*sb
);
81 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
82 static struct file_system_type ext3_fs_type
= {
86 .kill_sb
= kill_block_super
,
87 .fs_flags
= FS_REQUIRES_DEV
,
89 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
91 #define IS_EXT3_SB(sb) (0)
94 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
95 struct ext4_group_desc
*bg
)
97 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
98 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
99 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
102 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
103 struct ext4_group_desc
*bg
)
105 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
106 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
107 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
110 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
111 struct ext4_group_desc
*bg
)
113 return le32_to_cpu(bg
->bg_inode_table_lo
) |
114 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
115 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
118 __u32
ext4_free_blks_count(struct super_block
*sb
,
119 struct ext4_group_desc
*bg
)
121 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
122 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
123 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
126 __u32
ext4_free_inodes_count(struct super_block
*sb
,
127 struct ext4_group_desc
*bg
)
129 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
130 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
131 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
134 __u32
ext4_used_dirs_count(struct super_block
*sb
,
135 struct ext4_group_desc
*bg
)
137 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
138 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
139 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
142 __u32
ext4_itable_unused_count(struct super_block
*sb
,
143 struct ext4_group_desc
*bg
)
145 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
146 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
147 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
150 void ext4_block_bitmap_set(struct super_block
*sb
,
151 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
153 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
154 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
155 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
158 void ext4_inode_bitmap_set(struct super_block
*sb
,
159 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
161 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
162 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
163 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
166 void ext4_inode_table_set(struct super_block
*sb
,
167 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
169 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
170 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
171 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
174 void ext4_free_blks_set(struct super_block
*sb
,
175 struct ext4_group_desc
*bg
, __u32 count
)
177 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
178 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
179 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
182 void ext4_free_inodes_set(struct super_block
*sb
,
183 struct ext4_group_desc
*bg
, __u32 count
)
185 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
186 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
187 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
190 void ext4_used_dirs_set(struct super_block
*sb
,
191 struct ext4_group_desc
*bg
, __u32 count
)
193 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
194 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
195 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
198 void ext4_itable_unused_set(struct super_block
*sb
,
199 struct ext4_group_desc
*bg
, __u32 count
)
201 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
202 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
203 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
207 /* Just increment the non-pointer handle value */
208 static handle_t
*ext4_get_nojournal(void)
210 handle_t
*handle
= current
->journal_info
;
211 unsigned long ref_cnt
= (unsigned long)handle
;
213 BUG_ON(ref_cnt
>= EXT4_NOJOURNAL_MAX_REF_COUNT
);
216 handle
= (handle_t
*)ref_cnt
;
218 current
->journal_info
= handle
;
223 /* Decrement the non-pointer handle value */
224 static void ext4_put_nojournal(handle_t
*handle
)
226 unsigned long ref_cnt
= (unsigned long)handle
;
228 BUG_ON(ref_cnt
== 0);
231 handle
= (handle_t
*)ref_cnt
;
233 current
->journal_info
= handle
;
237 * Wrappers for jbd2_journal_start/end.
239 * The only special thing we need to do here is to make sure that all
240 * journal_end calls result in the superblock being marked dirty, so
241 * that sync() will call the filesystem's write_super callback if
244 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
248 if (sb
->s_flags
& MS_RDONLY
)
249 return ERR_PTR(-EROFS
);
251 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
252 /* Special case here: if the journal has aborted behind our
253 * backs (eg. EIO in the commit thread), then we still need to
254 * take the FS itself readonly cleanly. */
255 journal
= EXT4_SB(sb
)->s_journal
;
257 if (is_journal_aborted(journal
)) {
258 ext4_abort(sb
, "Detected aborted journal");
259 return ERR_PTR(-EROFS
);
261 return jbd2_journal_start(journal
, nblocks
);
263 return ext4_get_nojournal();
267 * The only special thing we need to do here is to make sure that all
268 * jbd2_journal_stop calls result in the superblock being marked dirty, so
269 * that sync() will call the filesystem's write_super callback if
272 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
)
274 struct super_block
*sb
;
278 if (!ext4_handle_valid(handle
)) {
279 ext4_put_nojournal(handle
);
282 sb
= handle
->h_transaction
->t_journal
->j_private
;
284 rc
= jbd2_journal_stop(handle
);
289 __ext4_std_error(sb
, where
, line
, err
);
293 void ext4_journal_abort_handle(const char *caller
, unsigned int line
,
294 const char *err_fn
, struct buffer_head
*bh
,
295 handle_t
*handle
, int err
)
298 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
300 BUG_ON(!ext4_handle_valid(handle
));
303 BUFFER_TRACE(bh
, "abort");
308 if (is_handle_aborted(handle
))
311 printk(KERN_ERR
"%s:%d: aborting transaction: %s in %s\n",
312 caller
, line
, errstr
, err_fn
);
314 jbd2_journal_abort_handle(handle
);
317 static void __save_error_info(struct super_block
*sb
, const char *func
,
320 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
322 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
323 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
324 es
->s_last_error_time
= cpu_to_le32(get_seconds());
325 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
326 es
->s_last_error_line
= cpu_to_le32(line
);
327 if (!es
->s_first_error_time
) {
328 es
->s_first_error_time
= es
->s_last_error_time
;
329 strncpy(es
->s_first_error_func
, func
,
330 sizeof(es
->s_first_error_func
));
331 es
->s_first_error_line
= cpu_to_le32(line
);
332 es
->s_first_error_ino
= es
->s_last_error_ino
;
333 es
->s_first_error_block
= es
->s_last_error_block
;
336 * Start the daily error reporting function if it hasn't been
339 if (!es
->s_error_count
)
340 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
341 es
->s_error_count
= cpu_to_le32(le32_to_cpu(es
->s_error_count
) + 1);
344 static void save_error_info(struct super_block
*sb
, const char *func
,
347 __save_error_info(sb
, func
, line
);
348 ext4_commit_super(sb
, 1);
352 /* Deal with the reporting of failure conditions on a filesystem such as
353 * inconsistencies detected or read IO failures.
355 * On ext2, we can store the error state of the filesystem in the
356 * superblock. That is not possible on ext4, because we may have other
357 * write ordering constraints on the superblock which prevent us from
358 * writing it out straight away; and given that the journal is about to
359 * be aborted, we can't rely on the current, or future, transactions to
360 * write out the superblock safely.
362 * We'll just use the jbd2_journal_abort() error code to record an error in
363 * the journal instead. On recovery, the journal will complain about
364 * that error until we've noted it down and cleared it.
367 static void ext4_handle_error(struct super_block
*sb
)
369 if (sb
->s_flags
& MS_RDONLY
)
372 if (!test_opt(sb
, ERRORS_CONT
)) {
373 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
375 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
377 jbd2_journal_abort(journal
, -EIO
);
379 if (test_opt(sb
, ERRORS_RO
)) {
380 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
381 sb
->s_flags
|= MS_RDONLY
;
383 if (test_opt(sb
, ERRORS_PANIC
))
384 panic("EXT4-fs (device %s): panic forced after error\n",
388 void __ext4_error(struct super_block
*sb
, const char *function
,
389 unsigned int line
, const char *fmt
, ...)
394 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: ",
395 sb
->s_id
, function
, line
, current
->comm
);
400 ext4_handle_error(sb
);
403 void ext4_error_inode(struct inode
*inode
, const char *function
,
404 unsigned int line
, ext4_fsblk_t block
,
405 const char *fmt
, ...)
408 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
410 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
411 es
->s_last_error_block
= cpu_to_le64(block
);
412 save_error_info(inode
->i_sb
, function
, line
);
414 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: inode #%lu: ",
415 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
417 printk("block %llu: ", block
);
418 printk("comm %s: ", current
->comm
);
423 ext4_handle_error(inode
->i_sb
);
426 void ext4_error_file(struct file
*file
, const char *function
,
427 unsigned int line
, const char *fmt
, ...)
430 struct ext4_super_block
*es
;
431 struct inode
*inode
= file
->f_dentry
->d_inode
;
432 char pathname
[80], *path
;
434 es
= EXT4_SB(inode
->i_sb
)->s_es
;
435 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
436 save_error_info(inode
->i_sb
, function
, line
);
438 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
442 "EXT4-fs error (device %s): %s:%d: inode #%lu "
443 "(comm %s path %s): ",
444 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
445 current
->comm
, path
);
450 ext4_handle_error(inode
->i_sb
);
453 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
460 errstr
= "IO failure";
463 errstr
= "Out of memory";
466 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
467 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
468 errstr
= "Journal has aborted";
470 errstr
= "Readonly filesystem";
473 /* If the caller passed in an extra buffer for unknown
474 * errors, textualise them now. Else we just return
477 /* Check for truncated error codes... */
478 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
487 /* __ext4_std_error decodes expected errors from journaling functions
488 * automatically and invokes the appropriate error response. */
490 void __ext4_std_error(struct super_block
*sb
, const char *function
,
491 unsigned int line
, int errno
)
496 /* Special case: if the error is EROFS, and we're not already
497 * inside a transaction, then there's really no point in logging
499 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
500 (sb
->s_flags
& MS_RDONLY
))
503 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
504 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
505 sb
->s_id
, function
, line
, errstr
);
506 save_error_info(sb
, function
, line
);
508 ext4_handle_error(sb
);
512 * ext4_abort is a much stronger failure handler than ext4_error. The
513 * abort function may be used to deal with unrecoverable failures such
514 * as journal IO errors or ENOMEM at a critical moment in log management.
516 * We unconditionally force the filesystem into an ABORT|READONLY state,
517 * unless the error response on the fs has been set to panic in which
518 * case we take the easy way out and panic immediately.
521 void __ext4_abort(struct super_block
*sb
, const char *function
,
522 unsigned int line
, const char *fmt
, ...)
526 save_error_info(sb
, function
, line
);
528 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
534 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
535 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
536 sb
->s_flags
|= MS_RDONLY
;
537 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
538 if (EXT4_SB(sb
)->s_journal
)
539 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
540 save_error_info(sb
, function
, line
);
542 if (test_opt(sb
, ERRORS_PANIC
))
543 panic("EXT4-fs panic from previous error\n");
546 void ext4_msg (struct super_block
* sb
, const char *prefix
,
547 const char *fmt
, ...)
552 printk("%sEXT4-fs (%s): ", prefix
, sb
->s_id
);
558 void __ext4_warning(struct super_block
*sb
, const char *function
,
559 unsigned int line
, const char *fmt
, ...)
564 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: ",
565 sb
->s_id
, function
, line
);
571 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
572 struct super_block
*sb
, ext4_group_t grp
,
573 unsigned long ino
, ext4_fsblk_t block
,
574 const char *fmt
, ...)
579 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
581 es
->s_last_error_ino
= cpu_to_le32(ino
);
582 es
->s_last_error_block
= cpu_to_le64(block
);
583 __save_error_info(sb
, function
, line
);
585 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u",
586 sb
->s_id
, function
, line
, grp
);
588 printk("inode %lu: ", ino
);
590 printk("block %llu:", (unsigned long long) block
);
595 if (test_opt(sb
, ERRORS_CONT
)) {
596 ext4_commit_super(sb
, 0);
600 ext4_unlock_group(sb
, grp
);
601 ext4_handle_error(sb
);
603 * We only get here in the ERRORS_RO case; relocking the group
604 * may be dangerous, but nothing bad will happen since the
605 * filesystem will have already been marked read/only and the
606 * journal has been aborted. We return 1 as a hint to callers
607 * who might what to use the return value from
608 * ext4_grp_locked_error() to distinguish beween the
609 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
610 * aggressively from the ext4 function in question, with a
611 * more appropriate error code.
613 ext4_lock_group(sb
, grp
);
617 void ext4_update_dynamic_rev(struct super_block
*sb
)
619 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
621 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
625 "updating to rev %d because of new feature flag, "
626 "running e2fsck is recommended",
629 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
630 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
631 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
632 /* leave es->s_feature_*compat flags alone */
633 /* es->s_uuid will be set by e2fsck if empty */
636 * The rest of the superblock fields should be zero, and if not it
637 * means they are likely already in use, so leave them alone. We
638 * can leave it up to e2fsck to clean up any inconsistencies there.
643 * Open the external journal device
645 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
647 struct block_device
*bdev
;
648 char b
[BDEVNAME_SIZE
];
650 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
656 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
657 __bdevname(dev
, b
), PTR_ERR(bdev
));
662 * Release the journal device
664 static int ext4_blkdev_put(struct block_device
*bdev
)
667 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
670 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
672 struct block_device
*bdev
;
675 bdev
= sbi
->journal_bdev
;
677 ret
= ext4_blkdev_put(bdev
);
678 sbi
->journal_bdev
= NULL
;
683 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
685 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
688 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
692 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
693 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
695 printk(KERN_ERR
"sb_info orphan list:\n");
696 list_for_each(l
, &sbi
->s_orphan
) {
697 struct inode
*inode
= orphan_list_entry(l
);
699 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
700 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
701 inode
->i_mode
, inode
->i_nlink
,
706 static void ext4_put_super(struct super_block
*sb
)
708 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
709 struct ext4_super_block
*es
= sbi
->s_es
;
712 ext4_unregister_li_request(sb
);
713 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
715 flush_workqueue(sbi
->dio_unwritten_wq
);
716 destroy_workqueue(sbi
->dio_unwritten_wq
);
720 ext4_commit_super(sb
, 1);
722 if (sbi
->s_journal
) {
723 err
= jbd2_journal_destroy(sbi
->s_journal
);
724 sbi
->s_journal
= NULL
;
726 ext4_abort(sb
, "Couldn't clean up the journal");
729 del_timer(&sbi
->s_err_report
);
730 ext4_release_system_zone(sb
);
732 ext4_ext_release(sb
);
733 ext4_xattr_put_super(sb
);
735 if (!(sb
->s_flags
& MS_RDONLY
)) {
736 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
737 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
738 ext4_commit_super(sb
, 1);
741 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
743 kobject_del(&sbi
->s_kobj
);
745 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
746 brelse(sbi
->s_group_desc
[i
]);
747 kfree(sbi
->s_group_desc
);
748 if (is_vmalloc_addr(sbi
->s_flex_groups
))
749 vfree(sbi
->s_flex_groups
);
751 kfree(sbi
->s_flex_groups
);
752 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
753 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
754 percpu_counter_destroy(&sbi
->s_dirs_counter
);
755 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
758 for (i
= 0; i
< MAXQUOTAS
; i
++)
759 kfree(sbi
->s_qf_names
[i
]);
762 /* Debugging code just in case the in-memory inode orphan list
763 * isn't empty. The on-disk one can be non-empty if we've
764 * detected an error and taken the fs readonly, but the
765 * in-memory list had better be clean by this point. */
766 if (!list_empty(&sbi
->s_orphan
))
767 dump_orphan_list(sb
, sbi
);
768 J_ASSERT(list_empty(&sbi
->s_orphan
));
770 invalidate_bdev(sb
->s_bdev
);
771 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
773 * Invalidate the journal device's buffers. We don't want them
774 * floating about in memory - the physical journal device may
775 * hotswapped, and it breaks the `ro-after' testing code.
777 sync_blockdev(sbi
->journal_bdev
);
778 invalidate_bdev(sbi
->journal_bdev
);
779 ext4_blkdev_remove(sbi
);
781 sb
->s_fs_info
= NULL
;
783 * Now that we are completely done shutting down the
784 * superblock, we need to actually destroy the kobject.
787 kobject_put(&sbi
->s_kobj
);
788 wait_for_completion(&sbi
->s_kobj_unregister
);
789 kfree(sbi
->s_blockgroup_lock
);
793 static struct kmem_cache
*ext4_inode_cachep
;
796 * Called inside transaction, so use GFP_NOFS
798 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
800 struct ext4_inode_info
*ei
;
802 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
806 ei
->vfs_inode
.i_version
= 1;
807 ei
->vfs_inode
.i_data
.writeback_index
= 0;
808 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
809 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
810 spin_lock_init(&ei
->i_prealloc_lock
);
812 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
813 * therefore it can be null here. Don't check it, just initialize
816 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
817 ei
->i_reserved_data_blocks
= 0;
818 ei
->i_reserved_meta_blocks
= 0;
819 ei
->i_allocated_meta_blocks
= 0;
820 ei
->i_da_metadata_calc_len
= 0;
821 ei
->i_delalloc_reserved_flag
= 0;
822 spin_lock_init(&(ei
->i_block_reservation_lock
));
824 ei
->i_reserved_quota
= 0;
826 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
827 spin_lock_init(&ei
->i_completed_io_lock
);
828 ei
->cur_aio_dio
= NULL
;
830 ei
->i_datasync_tid
= 0;
831 atomic_set(&ei
->i_ioend_count
, 0);
833 return &ei
->vfs_inode
;
836 static void ext4_destroy_inode(struct inode
*inode
)
838 ext4_ioend_wait(inode
);
839 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
840 ext4_msg(inode
->i_sb
, KERN_ERR
,
841 "Inode %lu (%p): orphan list check failed!",
842 inode
->i_ino
, EXT4_I(inode
));
843 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
844 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
848 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
851 static void init_once(void *foo
)
853 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
855 INIT_LIST_HEAD(&ei
->i_orphan
);
856 #ifdef CONFIG_EXT4_FS_XATTR
857 init_rwsem(&ei
->xattr_sem
);
859 init_rwsem(&ei
->i_data_sem
);
860 inode_init_once(&ei
->vfs_inode
);
863 static int init_inodecache(void)
865 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
866 sizeof(struct ext4_inode_info
),
867 0, (SLAB_RECLAIM_ACCOUNT
|
870 if (ext4_inode_cachep
== NULL
)
875 static void destroy_inodecache(void)
877 kmem_cache_destroy(ext4_inode_cachep
);
880 void ext4_clear_inode(struct inode
*inode
)
882 invalidate_inode_buffers(inode
);
883 end_writeback(inode
);
885 ext4_discard_preallocations(inode
);
886 if (EXT4_JOURNAL(inode
))
887 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
888 &EXT4_I(inode
)->jinode
);
891 static inline void ext4_show_quota_options(struct seq_file
*seq
,
892 struct super_block
*sb
)
894 #if defined(CONFIG_QUOTA)
895 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
897 if (sbi
->s_jquota_fmt
) {
900 switch (sbi
->s_jquota_fmt
) {
911 seq_printf(seq
, ",jqfmt=%s", fmtname
);
914 if (sbi
->s_qf_names
[USRQUOTA
])
915 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
917 if (sbi
->s_qf_names
[GRPQUOTA
])
918 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
920 if (test_opt(sb
, USRQUOTA
))
921 seq_puts(seq
, ",usrquota");
923 if (test_opt(sb
, GRPQUOTA
))
924 seq_puts(seq
, ",grpquota");
930 * - it's set to a non-default value OR
931 * - if the per-sb default is different from the global default
933 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
936 unsigned long def_mount_opts
;
937 struct super_block
*sb
= vfs
->mnt_sb
;
938 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
939 struct ext4_super_block
*es
= sbi
->s_es
;
941 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
942 def_errors
= le16_to_cpu(es
->s_errors
);
944 if (sbi
->s_sb_block
!= 1)
945 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
946 if (test_opt(sb
, MINIX_DF
))
947 seq_puts(seq
, ",minixdf");
948 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
949 seq_puts(seq
, ",grpid");
950 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
951 seq_puts(seq
, ",nogrpid");
952 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
953 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
954 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
956 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
957 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
958 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
960 if (test_opt(sb
, ERRORS_RO
)) {
961 if (def_errors
== EXT4_ERRORS_PANIC
||
962 def_errors
== EXT4_ERRORS_CONTINUE
) {
963 seq_puts(seq
, ",errors=remount-ro");
966 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
967 seq_puts(seq
, ",errors=continue");
968 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
969 seq_puts(seq
, ",errors=panic");
970 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
971 seq_puts(seq
, ",nouid32");
972 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
973 seq_puts(seq
, ",debug");
974 if (test_opt(sb
, OLDALLOC
))
975 seq_puts(seq
, ",oldalloc");
976 #ifdef CONFIG_EXT4_FS_XATTR
977 if (test_opt(sb
, XATTR_USER
) &&
978 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
979 seq_puts(seq
, ",user_xattr");
980 if (!test_opt(sb
, XATTR_USER
) &&
981 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
982 seq_puts(seq
, ",nouser_xattr");
985 #ifdef CONFIG_EXT4_FS_POSIX_ACL
986 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
987 seq_puts(seq
, ",acl");
988 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
989 seq_puts(seq
, ",noacl");
991 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
992 seq_printf(seq
, ",commit=%u",
993 (unsigned) (sbi
->s_commit_interval
/ HZ
));
995 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
996 seq_printf(seq
, ",min_batch_time=%u",
997 (unsigned) sbi
->s_min_batch_time
);
999 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
1000 seq_printf(seq
, ",max_batch_time=%u",
1001 (unsigned) sbi
->s_min_batch_time
);
1005 * We're changing the default of barrier mount option, so
1006 * let's always display its mount state so it's clear what its
1009 seq_puts(seq
, ",barrier=");
1010 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
1011 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
1012 seq_puts(seq
, ",journal_async_commit");
1013 else if (test_opt(sb
, JOURNAL_CHECKSUM
))
1014 seq_puts(seq
, ",journal_checksum");
1015 if (test_opt(sb
, I_VERSION
))
1016 seq_puts(seq
, ",i_version");
1017 if (!test_opt(sb
, DELALLOC
) &&
1018 !(def_mount_opts
& EXT4_DEFM_NODELALLOC
))
1019 seq_puts(seq
, ",nodelalloc");
1022 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
1024 * journal mode get enabled in different ways
1025 * So just print the value even if we didn't specify it
1027 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1028 seq_puts(seq
, ",data=journal");
1029 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1030 seq_puts(seq
, ",data=ordered");
1031 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1032 seq_puts(seq
, ",data=writeback");
1034 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1035 seq_printf(seq
, ",inode_readahead_blks=%u",
1036 sbi
->s_inode_readahead_blks
);
1038 if (test_opt(sb
, DATA_ERR_ABORT
))
1039 seq_puts(seq
, ",data_err=abort");
1041 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
1042 seq_puts(seq
, ",noauto_da_alloc");
1044 if (test_opt(sb
, DISCARD
) && !(def_mount_opts
& EXT4_DEFM_DISCARD
))
1045 seq_puts(seq
, ",discard");
1047 if (test_opt(sb
, NOLOAD
))
1048 seq_puts(seq
, ",norecovery");
1050 if (test_opt(sb
, DIOREAD_NOLOCK
))
1051 seq_puts(seq
, ",dioread_nolock");
1053 if (test_opt(sb
, BLOCK_VALIDITY
) &&
1054 !(def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
))
1055 seq_puts(seq
, ",block_validity");
1057 if (!test_opt(sb
, INIT_INODE_TABLE
))
1058 seq_puts(seq
, ",noinit_inode_table");
1059 else if (sbi
->s_li_wait_mult
)
1060 seq_printf(seq
, ",init_inode_table=%u",
1061 (unsigned) sbi
->s_li_wait_mult
);
1063 ext4_show_quota_options(seq
, sb
);
1068 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1069 u64 ino
, u32 generation
)
1071 struct inode
*inode
;
1073 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1074 return ERR_PTR(-ESTALE
);
1075 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1076 return ERR_PTR(-ESTALE
);
1078 /* iget isn't really right if the inode is currently unallocated!!
1080 * ext4_read_inode will return a bad_inode if the inode had been
1081 * deleted, so we should be safe.
1083 * Currently we don't know the generation for parent directory, so
1084 * a generation of 0 means "accept any"
1086 inode
= ext4_iget(sb
, ino
);
1088 return ERR_CAST(inode
);
1089 if (generation
&& inode
->i_generation
!= generation
) {
1091 return ERR_PTR(-ESTALE
);
1097 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1098 int fh_len
, int fh_type
)
1100 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1101 ext4_nfs_get_inode
);
1104 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1105 int fh_len
, int fh_type
)
1107 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1108 ext4_nfs_get_inode
);
1112 * Try to release metadata pages (indirect blocks, directories) which are
1113 * mapped via the block device. Since these pages could have journal heads
1114 * which would prevent try_to_free_buffers() from freeing them, we must use
1115 * jbd2 layer's try_to_free_buffers() function to release them.
1117 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1120 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1122 WARN_ON(PageChecked(page
));
1123 if (!page_has_buffers(page
))
1126 return jbd2_journal_try_to_free_buffers(journal
, page
,
1127 wait
& ~__GFP_WAIT
);
1128 return try_to_free_buffers(page
);
1132 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1133 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1135 static int ext4_write_dquot(struct dquot
*dquot
);
1136 static int ext4_acquire_dquot(struct dquot
*dquot
);
1137 static int ext4_release_dquot(struct dquot
*dquot
);
1138 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1139 static int ext4_write_info(struct super_block
*sb
, int type
);
1140 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1142 static int ext4_quota_off(struct super_block
*sb
, int type
);
1143 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1144 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1145 size_t len
, loff_t off
);
1146 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1147 const char *data
, size_t len
, loff_t off
);
1149 static const struct dquot_operations ext4_quota_operations
= {
1151 .get_reserved_space
= ext4_get_reserved_space
,
1153 .write_dquot
= ext4_write_dquot
,
1154 .acquire_dquot
= ext4_acquire_dquot
,
1155 .release_dquot
= ext4_release_dquot
,
1156 .mark_dirty
= ext4_mark_dquot_dirty
,
1157 .write_info
= ext4_write_info
,
1158 .alloc_dquot
= dquot_alloc
,
1159 .destroy_dquot
= dquot_destroy
,
1162 static const struct quotactl_ops ext4_qctl_operations
= {
1163 .quota_on
= ext4_quota_on
,
1164 .quota_off
= ext4_quota_off
,
1165 .quota_sync
= dquot_quota_sync
,
1166 .get_info
= dquot_get_dqinfo
,
1167 .set_info
= dquot_set_dqinfo
,
1168 .get_dqblk
= dquot_get_dqblk
,
1169 .set_dqblk
= dquot_set_dqblk
1173 static const struct super_operations ext4_sops
= {
1174 .alloc_inode
= ext4_alloc_inode
,
1175 .destroy_inode
= ext4_destroy_inode
,
1176 .write_inode
= ext4_write_inode
,
1177 .dirty_inode
= ext4_dirty_inode
,
1178 .evict_inode
= ext4_evict_inode
,
1179 .put_super
= ext4_put_super
,
1180 .sync_fs
= ext4_sync_fs
,
1181 .freeze_fs
= ext4_freeze
,
1182 .unfreeze_fs
= ext4_unfreeze
,
1183 .statfs
= ext4_statfs
,
1184 .remount_fs
= ext4_remount
,
1185 .show_options
= ext4_show_options
,
1187 .quota_read
= ext4_quota_read
,
1188 .quota_write
= ext4_quota_write
,
1190 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1191 .trim_fs
= ext4_trim_fs
1194 static const struct super_operations ext4_nojournal_sops
= {
1195 .alloc_inode
= ext4_alloc_inode
,
1196 .destroy_inode
= ext4_destroy_inode
,
1197 .write_inode
= ext4_write_inode
,
1198 .dirty_inode
= ext4_dirty_inode
,
1199 .evict_inode
= ext4_evict_inode
,
1200 .write_super
= ext4_write_super
,
1201 .put_super
= ext4_put_super
,
1202 .statfs
= ext4_statfs
,
1203 .remount_fs
= ext4_remount
,
1204 .show_options
= ext4_show_options
,
1206 .quota_read
= ext4_quota_read
,
1207 .quota_write
= ext4_quota_write
,
1209 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1212 static const struct export_operations ext4_export_ops
= {
1213 .fh_to_dentry
= ext4_fh_to_dentry
,
1214 .fh_to_parent
= ext4_fh_to_parent
,
1215 .get_parent
= ext4_get_parent
,
1219 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1220 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1221 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1222 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1223 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1224 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1225 Opt_journal_update
, Opt_journal_dev
,
1226 Opt_journal_checksum
, Opt_journal_async_commit
,
1227 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1228 Opt_data_err_abort
, Opt_data_err_ignore
,
1229 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1230 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1231 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1232 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1233 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
1234 Opt_block_validity
, Opt_noblock_validity
,
1235 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1236 Opt_dioread_nolock
, Opt_dioread_lock
,
1237 Opt_discard
, Opt_nodiscard
,
1238 Opt_init_inode_table
, Opt_noinit_inode_table
,
1241 static const match_table_t tokens
= {
1242 {Opt_bsd_df
, "bsddf"},
1243 {Opt_minix_df
, "minixdf"},
1244 {Opt_grpid
, "grpid"},
1245 {Opt_grpid
, "bsdgroups"},
1246 {Opt_nogrpid
, "nogrpid"},
1247 {Opt_nogrpid
, "sysvgroups"},
1248 {Opt_resgid
, "resgid=%u"},
1249 {Opt_resuid
, "resuid=%u"},
1251 {Opt_err_cont
, "errors=continue"},
1252 {Opt_err_panic
, "errors=panic"},
1253 {Opt_err_ro
, "errors=remount-ro"},
1254 {Opt_nouid32
, "nouid32"},
1255 {Opt_debug
, "debug"},
1256 {Opt_oldalloc
, "oldalloc"},
1257 {Opt_orlov
, "orlov"},
1258 {Opt_user_xattr
, "user_xattr"},
1259 {Opt_nouser_xattr
, "nouser_xattr"},
1261 {Opt_noacl
, "noacl"},
1262 {Opt_noload
, "noload"},
1263 {Opt_noload
, "norecovery"},
1266 {Opt_commit
, "commit=%u"},
1267 {Opt_min_batch_time
, "min_batch_time=%u"},
1268 {Opt_max_batch_time
, "max_batch_time=%u"},
1269 {Opt_journal_update
, "journal=update"},
1270 {Opt_journal_dev
, "journal_dev=%u"},
1271 {Opt_journal_checksum
, "journal_checksum"},
1272 {Opt_journal_async_commit
, "journal_async_commit"},
1273 {Opt_abort
, "abort"},
1274 {Opt_data_journal
, "data=journal"},
1275 {Opt_data_ordered
, "data=ordered"},
1276 {Opt_data_writeback
, "data=writeback"},
1277 {Opt_data_err_abort
, "data_err=abort"},
1278 {Opt_data_err_ignore
, "data_err=ignore"},
1279 {Opt_offusrjquota
, "usrjquota="},
1280 {Opt_usrjquota
, "usrjquota=%s"},
1281 {Opt_offgrpjquota
, "grpjquota="},
1282 {Opt_grpjquota
, "grpjquota=%s"},
1283 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1284 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1285 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1286 {Opt_grpquota
, "grpquota"},
1287 {Opt_noquota
, "noquota"},
1288 {Opt_quota
, "quota"},
1289 {Opt_usrquota
, "usrquota"},
1290 {Opt_barrier
, "barrier=%u"},
1291 {Opt_barrier
, "barrier"},
1292 {Opt_nobarrier
, "nobarrier"},
1293 {Opt_i_version
, "i_version"},
1294 {Opt_stripe
, "stripe=%u"},
1295 {Opt_resize
, "resize"},
1296 {Opt_delalloc
, "delalloc"},
1297 {Opt_nodelalloc
, "nodelalloc"},
1298 {Opt_block_validity
, "block_validity"},
1299 {Opt_noblock_validity
, "noblock_validity"},
1300 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1301 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1302 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1303 {Opt_auto_da_alloc
, "auto_da_alloc"},
1304 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1305 {Opt_dioread_nolock
, "dioread_nolock"},
1306 {Opt_dioread_lock
, "dioread_lock"},
1307 {Opt_discard
, "discard"},
1308 {Opt_nodiscard
, "nodiscard"},
1309 {Opt_init_inode_table
, "init_itable=%u"},
1310 {Opt_init_inode_table
, "init_itable"},
1311 {Opt_noinit_inode_table
, "noinit_itable"},
1315 static ext4_fsblk_t
get_sb_block(void **data
)
1317 ext4_fsblk_t sb_block
;
1318 char *options
= (char *) *data
;
1320 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1321 return 1; /* Default location */
1324 /* TODO: use simple_strtoll with >32bit ext4 */
1325 sb_block
= simple_strtoul(options
, &options
, 0);
1326 if (*options
&& *options
!= ',') {
1327 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1331 if (*options
== ',')
1333 *data
= (void *) options
;
1338 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1339 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1340 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1343 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1345 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1348 if (sb_any_quota_loaded(sb
) &&
1349 !sbi
->s_qf_names
[qtype
]) {
1350 ext4_msg(sb
, KERN_ERR
,
1351 "Cannot change journaled "
1352 "quota options when quota turned on");
1355 qname
= match_strdup(args
);
1357 ext4_msg(sb
, KERN_ERR
,
1358 "Not enough memory for storing quotafile name");
1361 if (sbi
->s_qf_names
[qtype
] &&
1362 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1363 ext4_msg(sb
, KERN_ERR
,
1364 "%s quota file already specified", QTYPE2NAME(qtype
));
1368 sbi
->s_qf_names
[qtype
] = qname
;
1369 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1370 ext4_msg(sb
, KERN_ERR
,
1371 "quotafile must be on filesystem root");
1372 kfree(sbi
->s_qf_names
[qtype
]);
1373 sbi
->s_qf_names
[qtype
] = NULL
;
1376 set_opt(sbi
->s_mount_opt
, QUOTA
);
1380 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1383 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1385 if (sb_any_quota_loaded(sb
) &&
1386 sbi
->s_qf_names
[qtype
]) {
1387 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1388 " when quota turned on");
1392 * The space will be released later when all options are confirmed
1395 sbi
->s_qf_names
[qtype
] = NULL
;
1400 static int parse_options(char *options
, struct super_block
*sb
,
1401 unsigned long *journal_devnum
,
1402 unsigned int *journal_ioprio
,
1403 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1405 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1407 substring_t args
[MAX_OPT_ARGS
];
1417 while ((p
= strsep(&options
, ",")) != NULL
) {
1423 * Initialize args struct so we know whether arg was
1424 * found; some options take optional arguments.
1426 args
[0].to
= args
[0].from
= 0;
1427 token
= match_token(p
, tokens
, args
);
1430 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1431 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1434 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1435 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1439 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1440 set_opt(sbi
->s_mount_opt
, GRPID
);
1444 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1445 clear_opt(sbi
->s_mount_opt
, GRPID
);
1449 if (match_int(&args
[0], &option
))
1451 sbi
->s_resuid
= option
;
1454 if (match_int(&args
[0], &option
))
1456 sbi
->s_resgid
= option
;
1459 /* handled by get_sb_block() instead of here */
1460 /* *sb_block = match_int(&args[0]); */
1463 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1464 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1465 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1468 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1469 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1470 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1473 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1474 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1475 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1478 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1481 set_opt(sbi
->s_mount_opt
, DEBUG
);
1484 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1487 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1489 #ifdef CONFIG_EXT4_FS_XATTR
1490 case Opt_user_xattr
:
1491 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1493 case Opt_nouser_xattr
:
1494 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1497 case Opt_user_xattr
:
1498 case Opt_nouser_xattr
:
1499 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1502 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1504 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1507 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1512 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1515 case Opt_journal_update
:
1517 /* Eventually we will want to be able to create
1518 a journal file here. For now, only allow the
1519 user to specify an existing inode to be the
1522 ext4_msg(sb
, KERN_ERR
,
1523 "Cannot specify journal on remount");
1526 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1528 case Opt_journal_dev
:
1530 ext4_msg(sb
, KERN_ERR
,
1531 "Cannot specify journal on remount");
1534 if (match_int(&args
[0], &option
))
1536 *journal_devnum
= option
;
1538 case Opt_journal_checksum
:
1539 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1541 case Opt_journal_async_commit
:
1542 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1543 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1546 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1549 if (match_int(&args
[0], &option
))
1554 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1555 sbi
->s_commit_interval
= HZ
* option
;
1557 case Opt_max_batch_time
:
1558 if (match_int(&args
[0], &option
))
1563 option
= EXT4_DEF_MAX_BATCH_TIME
;
1564 sbi
->s_max_batch_time
= option
;
1566 case Opt_min_batch_time
:
1567 if (match_int(&args
[0], &option
))
1571 sbi
->s_min_batch_time
= option
;
1573 case Opt_data_journal
:
1574 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1576 case Opt_data_ordered
:
1577 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1579 case Opt_data_writeback
:
1580 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1583 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1584 ext4_msg(sb
, KERN_ERR
,
1585 "Cannot change data mode on remount");
1589 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
1590 sbi
->s_mount_opt
|= data_opt
;
1593 case Opt_data_err_abort
:
1594 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1596 case Opt_data_err_ignore
:
1597 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1601 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1605 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1608 case Opt_offusrjquota
:
1609 if (!clear_qf_name(sb
, USRQUOTA
))
1612 case Opt_offgrpjquota
:
1613 if (!clear_qf_name(sb
, GRPQUOTA
))
1617 case Opt_jqfmt_vfsold
:
1618 qfmt
= QFMT_VFS_OLD
;
1620 case Opt_jqfmt_vfsv0
:
1623 case Opt_jqfmt_vfsv1
:
1626 if (sb_any_quota_loaded(sb
) &&
1627 sbi
->s_jquota_fmt
!= qfmt
) {
1628 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1629 "journaled quota options when "
1633 sbi
->s_jquota_fmt
= qfmt
;
1637 set_opt(sbi
->s_mount_opt
, QUOTA
);
1638 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1641 set_opt(sbi
->s_mount_opt
, QUOTA
);
1642 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1645 if (sb_any_quota_loaded(sb
)) {
1646 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1647 "options when quota turned on");
1650 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1651 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1652 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1658 ext4_msg(sb
, KERN_ERR
,
1659 "quota options not supported");
1663 case Opt_offusrjquota
:
1664 case Opt_offgrpjquota
:
1665 case Opt_jqfmt_vfsold
:
1666 case Opt_jqfmt_vfsv0
:
1667 case Opt_jqfmt_vfsv1
:
1668 ext4_msg(sb
, KERN_ERR
,
1669 "journaled quota options not supported");
1675 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1678 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1682 if (match_int(&args
[0], &option
))
1685 option
= 1; /* No argument, default to 1 */
1687 set_opt(sbi
->s_mount_opt
, BARRIER
);
1689 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1695 ext4_msg(sb
, KERN_ERR
,
1696 "resize option only available "
1700 if (match_int(&args
[0], &option
) != 0)
1702 *n_blocks_count
= option
;
1705 ext4_msg(sb
, KERN_WARNING
,
1706 "Ignoring deprecated nobh option");
1709 ext4_msg(sb
, KERN_WARNING
,
1710 "Ignoring deprecated bh option");
1713 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1714 sb
->s_flags
|= MS_I_VERSION
;
1716 case Opt_nodelalloc
:
1717 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1720 if (match_int(&args
[0], &option
))
1724 sbi
->s_stripe
= option
;
1727 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1729 case Opt_block_validity
:
1730 set_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1732 case Opt_noblock_validity
:
1733 clear_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1735 case Opt_inode_readahead_blks
:
1736 if (match_int(&args
[0], &option
))
1738 if (option
< 0 || option
> (1 << 30))
1740 if (!is_power_of_2(option
)) {
1741 ext4_msg(sb
, KERN_ERR
,
1742 "EXT4-fs: inode_readahead_blks"
1743 " must be a power of 2");
1746 sbi
->s_inode_readahead_blks
= option
;
1748 case Opt_journal_ioprio
:
1749 if (match_int(&args
[0], &option
))
1751 if (option
< 0 || option
> 7)
1753 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1756 case Opt_noauto_da_alloc
:
1757 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1759 case Opt_auto_da_alloc
:
1761 if (match_int(&args
[0], &option
))
1764 option
= 1; /* No argument, default to 1 */
1766 clear_opt(sbi
->s_mount_opt
, NO_AUTO_DA_ALLOC
);
1768 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1771 set_opt(sbi
->s_mount_opt
, DISCARD
);
1774 clear_opt(sbi
->s_mount_opt
, DISCARD
);
1776 case Opt_dioread_nolock
:
1777 set_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
1779 case Opt_dioread_lock
:
1780 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
1782 case Opt_init_inode_table
:
1783 set_opt(sbi
->s_mount_opt
, INIT_INODE_TABLE
);
1785 if (match_int(&args
[0], &option
))
1788 option
= EXT4_DEF_LI_WAIT_MULT
;
1791 sbi
->s_li_wait_mult
= option
;
1793 case Opt_noinit_inode_table
:
1794 clear_opt(sbi
->s_mount_opt
, INIT_INODE_TABLE
);
1797 ext4_msg(sb
, KERN_ERR
,
1798 "Unrecognized mount option \"%s\" "
1799 "or missing value", p
);
1804 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1805 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1806 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1808 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1809 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1811 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1812 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1817 if (!sbi
->s_jquota_fmt
) {
1818 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1823 if (sbi
->s_jquota_fmt
) {
1824 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1825 "specified with no journaling "
1834 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1837 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1840 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1841 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1842 "forcing read-only mode");
1847 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1848 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1849 "running e2fsck is recommended");
1850 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1851 ext4_msg(sb
, KERN_WARNING
,
1852 "warning: mounting fs with errors, "
1853 "running e2fsck is recommended");
1854 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1855 le16_to_cpu(es
->s_mnt_count
) >=
1856 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1857 ext4_msg(sb
, KERN_WARNING
,
1858 "warning: maximal mount count reached, "
1859 "running e2fsck is recommended");
1860 else if (le32_to_cpu(es
->s_checkinterval
) &&
1861 (le32_to_cpu(es
->s_lastcheck
) +
1862 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1863 ext4_msg(sb
, KERN_WARNING
,
1864 "warning: checktime reached, "
1865 "running e2fsck is recommended");
1866 if (!sbi
->s_journal
)
1867 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1868 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1869 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1870 le16_add_cpu(&es
->s_mnt_count
, 1);
1871 es
->s_mtime
= cpu_to_le32(get_seconds());
1872 ext4_update_dynamic_rev(sb
);
1874 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1876 ext4_commit_super(sb
, 1);
1877 if (test_opt(sb
, DEBUG
))
1878 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1879 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1881 sbi
->s_groups_count
,
1882 EXT4_BLOCKS_PER_GROUP(sb
),
1883 EXT4_INODES_PER_GROUP(sb
),
1889 static int ext4_fill_flex_info(struct super_block
*sb
)
1891 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1892 struct ext4_group_desc
*gdp
= NULL
;
1893 ext4_group_t flex_group_count
;
1894 ext4_group_t flex_group
;
1895 int groups_per_flex
= 0;
1899 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1900 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1902 if (groups_per_flex
< 2) {
1903 sbi
->s_log_groups_per_flex
= 0;
1907 /* We allocate both existing and potentially added groups */
1908 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1909 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1910 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1911 size
= flex_group_count
* sizeof(struct flex_groups
);
1912 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1913 if (sbi
->s_flex_groups
== NULL
) {
1914 sbi
->s_flex_groups
= vmalloc(size
);
1915 if (sbi
->s_flex_groups
)
1916 memset(sbi
->s_flex_groups
, 0, size
);
1918 if (sbi
->s_flex_groups
== NULL
) {
1919 ext4_msg(sb
, KERN_ERR
, "not enough memory for "
1920 "%u flex groups", flex_group_count
);
1924 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1925 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1927 flex_group
= ext4_flex_group(sbi
, i
);
1928 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1929 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1930 atomic_add(ext4_free_blks_count(sb
, gdp
),
1931 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1932 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1933 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1941 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1942 struct ext4_group_desc
*gdp
)
1946 if (sbi
->s_es
->s_feature_ro_compat
&
1947 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1948 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1949 __le32 le_group
= cpu_to_le32(block_group
);
1951 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1952 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1953 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1954 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1955 /* for checksum of struct ext4_group_desc do the rest...*/
1956 if ((sbi
->s_es
->s_feature_incompat
&
1957 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1958 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1959 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1960 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1964 return cpu_to_le16(crc
);
1967 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1968 struct ext4_group_desc
*gdp
)
1970 if ((sbi
->s_es
->s_feature_ro_compat
&
1971 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1972 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1978 /* Called at mount-time, super-block is locked */
1979 static int ext4_check_descriptors(struct super_block
*sb
,
1980 ext4_group_t
*first_not_zeroed
)
1982 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1983 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1984 ext4_fsblk_t last_block
;
1985 ext4_fsblk_t block_bitmap
;
1986 ext4_fsblk_t inode_bitmap
;
1987 ext4_fsblk_t inode_table
;
1988 int flexbg_flag
= 0;
1989 ext4_group_t i
, grp
= sbi
->s_groups_count
;
1991 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1994 ext4_debug("Checking group descriptors");
1996 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1997 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1999 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2000 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2002 last_block
= first_block
+
2003 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2005 if ((grp
== sbi
->s_groups_count
) &&
2006 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2009 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2010 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2011 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2012 "Block bitmap for group %u not in group "
2013 "(block %llu)!", i
, block_bitmap
);
2016 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2017 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2018 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2019 "Inode bitmap for group %u not in group "
2020 "(block %llu)!", i
, inode_bitmap
);
2023 inode_table
= ext4_inode_table(sb
, gdp
);
2024 if (inode_table
< first_block
||
2025 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2026 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2027 "Inode table for group %u not in group "
2028 "(block %llu)!", i
, inode_table
);
2031 ext4_lock_group(sb
, i
);
2032 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
2033 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2034 "Checksum for group %u failed (%u!=%u)",
2035 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2036 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2037 if (!(sb
->s_flags
& MS_RDONLY
)) {
2038 ext4_unlock_group(sb
, i
);
2042 ext4_unlock_group(sb
, i
);
2044 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2046 if (NULL
!= first_not_zeroed
)
2047 *first_not_zeroed
= grp
;
2049 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
2050 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2054 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2055 * the superblock) which were deleted from all directories, but held open by
2056 * a process at the time of a crash. We walk the list and try to delete these
2057 * inodes at recovery time (only with a read-write filesystem).
2059 * In order to keep the orphan inode chain consistent during traversal (in
2060 * case of crash during recovery), we link each inode into the superblock
2061 * orphan list_head and handle it the same way as an inode deletion during
2062 * normal operation (which journals the operations for us).
2064 * We only do an iget() and an iput() on each inode, which is very safe if we
2065 * accidentally point at an in-use or already deleted inode. The worst that
2066 * can happen in this case is that we get a "bit already cleared" message from
2067 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2068 * e2fsck was run on this filesystem, and it must have already done the orphan
2069 * inode cleanup for us, so we can safely abort without any further action.
2071 static void ext4_orphan_cleanup(struct super_block
*sb
,
2072 struct ext4_super_block
*es
)
2074 unsigned int s_flags
= sb
->s_flags
;
2075 int nr_orphans
= 0, nr_truncates
= 0;
2079 if (!es
->s_last_orphan
) {
2080 jbd_debug(4, "no orphan inodes to clean up\n");
2084 if (bdev_read_only(sb
->s_bdev
)) {
2085 ext4_msg(sb
, KERN_ERR
, "write access "
2086 "unavailable, skipping orphan cleanup");
2090 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2091 if (es
->s_last_orphan
)
2092 jbd_debug(1, "Errors on filesystem, "
2093 "clearing orphan list.\n");
2094 es
->s_last_orphan
= 0;
2095 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2099 if (s_flags
& MS_RDONLY
) {
2100 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2101 sb
->s_flags
&= ~MS_RDONLY
;
2104 /* Needed for iput() to work correctly and not trash data */
2105 sb
->s_flags
|= MS_ACTIVE
;
2106 /* Turn on quotas so that they are updated correctly */
2107 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2108 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2109 int ret
= ext4_quota_on_mount(sb
, i
);
2111 ext4_msg(sb
, KERN_ERR
,
2112 "Cannot turn on journaled "
2113 "quota: error %d", ret
);
2118 while (es
->s_last_orphan
) {
2119 struct inode
*inode
;
2121 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2122 if (IS_ERR(inode
)) {
2123 es
->s_last_orphan
= 0;
2127 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2128 dquot_initialize(inode
);
2129 if (inode
->i_nlink
) {
2130 ext4_msg(sb
, KERN_DEBUG
,
2131 "%s: truncating inode %lu to %lld bytes",
2132 __func__
, inode
->i_ino
, inode
->i_size
);
2133 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2134 inode
->i_ino
, inode
->i_size
);
2135 ext4_truncate(inode
);
2138 ext4_msg(sb
, KERN_DEBUG
,
2139 "%s: deleting unreferenced inode %lu",
2140 __func__
, inode
->i_ino
);
2141 jbd_debug(2, "deleting unreferenced inode %lu\n",
2145 iput(inode
); /* The delete magic happens here! */
2148 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2151 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2152 PLURAL(nr_orphans
));
2154 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2155 PLURAL(nr_truncates
));
2157 /* Turn quotas off */
2158 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2159 if (sb_dqopt(sb
)->files
[i
])
2160 dquot_quota_off(sb
, i
);
2163 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2167 * Maximal extent format file size.
2168 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2169 * extent format containers, within a sector_t, and within i_blocks
2170 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2171 * so that won't be a limiting factor.
2173 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2175 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2178 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2180 /* small i_blocks in vfs inode? */
2181 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2183 * CONFIG_LBDAF is not enabled implies the inode
2184 * i_block represent total blocks in 512 bytes
2185 * 32 == size of vfs inode i_blocks * 8
2187 upper_limit
= (1LL << 32) - 1;
2189 /* total blocks in file system block size */
2190 upper_limit
>>= (blkbits
- 9);
2191 upper_limit
<<= blkbits
;
2194 /* 32-bit extent-start container, ee_block */
2199 /* Sanity check against vm- & vfs- imposed limits */
2200 if (res
> upper_limit
)
2207 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2208 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2209 * We need to be 1 filesystem block less than the 2^48 sector limit.
2211 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2213 loff_t res
= EXT4_NDIR_BLOCKS
;
2216 /* This is calculated to be the largest file size for a dense, block
2217 * mapped file such that the file's total number of 512-byte sectors,
2218 * including data and all indirect blocks, does not exceed (2^48 - 1).
2220 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2221 * number of 512-byte sectors of the file.
2224 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2226 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2227 * the inode i_block field represents total file blocks in
2228 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2230 upper_limit
= (1LL << 32) - 1;
2232 /* total blocks in file system block size */
2233 upper_limit
>>= (bits
- 9);
2237 * We use 48 bit ext4_inode i_blocks
2238 * With EXT4_HUGE_FILE_FL set the i_blocks
2239 * represent total number of blocks in
2240 * file system block size
2242 upper_limit
= (1LL << 48) - 1;
2246 /* indirect blocks */
2248 /* double indirect blocks */
2249 meta_blocks
+= 1 + (1LL << (bits
-2));
2250 /* tripple indirect blocks */
2251 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2253 upper_limit
-= meta_blocks
;
2254 upper_limit
<<= bits
;
2256 res
+= 1LL << (bits
-2);
2257 res
+= 1LL << (2*(bits
-2));
2258 res
+= 1LL << (3*(bits
-2));
2260 if (res
> upper_limit
)
2263 if (res
> MAX_LFS_FILESIZE
)
2264 res
= MAX_LFS_FILESIZE
;
2269 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2270 ext4_fsblk_t logical_sb_block
, int nr
)
2272 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2273 ext4_group_t bg
, first_meta_bg
;
2276 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2278 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2280 return logical_sb_block
+ nr
+ 1;
2281 bg
= sbi
->s_desc_per_block
* nr
;
2282 if (ext4_bg_has_super(sb
, bg
))
2285 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2289 * ext4_get_stripe_size: Get the stripe size.
2290 * @sbi: In memory super block info
2292 * If we have specified it via mount option, then
2293 * use the mount option value. If the value specified at mount time is
2294 * greater than the blocks per group use the super block value.
2295 * If the super block value is greater than blocks per group return 0.
2296 * Allocator needs it be less than blocks per group.
2299 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2301 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2302 unsigned long stripe_width
=
2303 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2305 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2306 return sbi
->s_stripe
;
2308 if (stripe_width
<= sbi
->s_blocks_per_group
)
2309 return stripe_width
;
2311 if (stride
<= sbi
->s_blocks_per_group
)
2320 struct attribute attr
;
2321 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2322 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2323 const char *, size_t);
2327 static int parse_strtoul(const char *buf
,
2328 unsigned long max
, unsigned long *value
)
2332 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2333 endp
= skip_spaces(endp
);
2334 if (*endp
|| *value
> max
)
2340 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2341 struct ext4_sb_info
*sbi
,
2344 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2345 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2348 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2349 struct ext4_sb_info
*sbi
, char *buf
)
2351 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2353 if (!sb
->s_bdev
->bd_part
)
2354 return snprintf(buf
, PAGE_SIZE
, "0\n");
2355 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2356 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2357 sbi
->s_sectors_written_start
) >> 1);
2360 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2361 struct ext4_sb_info
*sbi
, char *buf
)
2363 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2365 if (!sb
->s_bdev
->bd_part
)
2366 return snprintf(buf
, PAGE_SIZE
, "0\n");
2367 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2368 (unsigned long long)(sbi
->s_kbytes_written
+
2369 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2370 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2373 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2374 struct ext4_sb_info
*sbi
,
2375 const char *buf
, size_t count
)
2379 if (parse_strtoul(buf
, 0x40000000, &t
))
2382 if (!is_power_of_2(t
))
2385 sbi
->s_inode_readahead_blks
= t
;
2389 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2390 struct ext4_sb_info
*sbi
, char *buf
)
2392 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2394 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2397 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2398 struct ext4_sb_info
*sbi
,
2399 const char *buf
, size_t count
)
2401 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2404 if (parse_strtoul(buf
, 0xffffffff, &t
))
2410 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2411 static struct ext4_attr ext4_attr_##_name = { \
2412 .attr = {.name = __stringify(_name), .mode = _mode }, \
2415 .offset = offsetof(struct ext4_sb_info, _elname), \
2417 #define EXT4_ATTR(name, mode, show, store) \
2418 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2420 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2421 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2422 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2423 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2424 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2425 #define ATTR_LIST(name) &ext4_attr_##name.attr
2427 EXT4_RO_ATTR(delayed_allocation_blocks
);
2428 EXT4_RO_ATTR(session_write_kbytes
);
2429 EXT4_RO_ATTR(lifetime_write_kbytes
);
2430 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2431 inode_readahead_blks_store
, s_inode_readahead_blks
);
2432 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2433 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2434 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2435 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2436 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2437 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2438 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2439 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2441 static struct attribute
*ext4_attrs
[] = {
2442 ATTR_LIST(delayed_allocation_blocks
),
2443 ATTR_LIST(session_write_kbytes
),
2444 ATTR_LIST(lifetime_write_kbytes
),
2445 ATTR_LIST(inode_readahead_blks
),
2446 ATTR_LIST(inode_goal
),
2447 ATTR_LIST(mb_stats
),
2448 ATTR_LIST(mb_max_to_scan
),
2449 ATTR_LIST(mb_min_to_scan
),
2450 ATTR_LIST(mb_order2_req
),
2451 ATTR_LIST(mb_stream_req
),
2452 ATTR_LIST(mb_group_prealloc
),
2453 ATTR_LIST(max_writeback_mb_bump
),
2457 /* Features this copy of ext4 supports */
2458 EXT4_INFO_ATTR(lazy_itable_init
);
2459 EXT4_INFO_ATTR(batched_discard
);
2461 static struct attribute
*ext4_feat_attrs
[] = {
2462 ATTR_LIST(lazy_itable_init
),
2463 ATTR_LIST(batched_discard
),
2467 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2468 struct attribute
*attr
, char *buf
)
2470 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2472 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2474 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2477 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2478 struct attribute
*attr
,
2479 const char *buf
, size_t len
)
2481 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2483 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2485 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2488 static void ext4_sb_release(struct kobject
*kobj
)
2490 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2492 complete(&sbi
->s_kobj_unregister
);
2495 static const struct sysfs_ops ext4_attr_ops
= {
2496 .show
= ext4_attr_show
,
2497 .store
= ext4_attr_store
,
2500 static struct kobj_type ext4_ktype
= {
2501 .default_attrs
= ext4_attrs
,
2502 .sysfs_ops
= &ext4_attr_ops
,
2503 .release
= ext4_sb_release
,
2506 static void ext4_feat_release(struct kobject
*kobj
)
2508 complete(&ext4_feat
->f_kobj_unregister
);
2511 static struct kobj_type ext4_feat_ktype
= {
2512 .default_attrs
= ext4_feat_attrs
,
2513 .sysfs_ops
= &ext4_attr_ops
,
2514 .release
= ext4_feat_release
,
2518 * Check whether this filesystem can be mounted based on
2519 * the features present and the RDONLY/RDWR mount requested.
2520 * Returns 1 if this filesystem can be mounted as requested,
2521 * 0 if it cannot be.
2523 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2525 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2526 ext4_msg(sb
, KERN_ERR
,
2527 "Couldn't mount because of "
2528 "unsupported optional features (%x)",
2529 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2530 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2537 /* Check that feature set is OK for a read-write mount */
2538 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2539 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2540 "unsupported optional features (%x)",
2541 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2542 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2546 * Large file size enabled file system can only be mounted
2547 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2549 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2550 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2551 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2552 "cannot be mounted RDWR without "
2561 * This function is called once a day if we have errors logged
2562 * on the file system
2564 static void print_daily_error_info(unsigned long arg
)
2566 struct super_block
*sb
= (struct super_block
*) arg
;
2567 struct ext4_sb_info
*sbi
;
2568 struct ext4_super_block
*es
;
2573 if (es
->s_error_count
)
2574 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2575 le32_to_cpu(es
->s_error_count
));
2576 if (es
->s_first_error_time
) {
2577 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2578 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2579 (int) sizeof(es
->s_first_error_func
),
2580 es
->s_first_error_func
,
2581 le32_to_cpu(es
->s_first_error_line
));
2582 if (es
->s_first_error_ino
)
2583 printk(": inode %u",
2584 le32_to_cpu(es
->s_first_error_ino
));
2585 if (es
->s_first_error_block
)
2586 printk(": block %llu", (unsigned long long)
2587 le64_to_cpu(es
->s_first_error_block
));
2590 if (es
->s_last_error_time
) {
2591 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2592 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2593 (int) sizeof(es
->s_last_error_func
),
2594 es
->s_last_error_func
,
2595 le32_to_cpu(es
->s_last_error_line
));
2596 if (es
->s_last_error_ino
)
2597 printk(": inode %u",
2598 le32_to_cpu(es
->s_last_error_ino
));
2599 if (es
->s_last_error_block
)
2600 printk(": block %llu", (unsigned long long)
2601 le64_to_cpu(es
->s_last_error_block
));
2604 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2607 static void ext4_lazyinode_timeout(unsigned long data
)
2609 struct task_struct
*p
= (struct task_struct
*)data
;
2613 /* Find next suitable group and run ext4_init_inode_table */
2614 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2616 struct ext4_group_desc
*gdp
= NULL
;
2617 ext4_group_t group
, ngroups
;
2618 struct super_block
*sb
;
2619 unsigned long timeout
= 0;
2623 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2625 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2626 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2632 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2636 if (group
== ngroups
)
2641 ret
= ext4_init_inode_table(sb
, group
,
2642 elr
->lr_timeout
? 0 : 1);
2643 if (elr
->lr_timeout
== 0) {
2644 timeout
= jiffies
- timeout
;
2645 if (elr
->lr_sbi
->s_li_wait_mult
)
2646 timeout
*= elr
->lr_sbi
->s_li_wait_mult
;
2649 elr
->lr_timeout
= timeout
;
2651 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2652 elr
->lr_next_group
= group
+ 1;
2659 * Remove lr_request from the list_request and free the
2660 * request tructure. Should be called with li_list_mtx held
2662 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2664 struct ext4_sb_info
*sbi
;
2671 list_del(&elr
->lr_request
);
2672 sbi
->s_li_request
= NULL
;
2676 static void ext4_unregister_li_request(struct super_block
*sb
)
2678 struct ext4_li_request
*elr
= EXT4_SB(sb
)->s_li_request
;
2683 mutex_lock(&ext4_li_info
->li_list_mtx
);
2684 ext4_remove_li_request(elr
);
2685 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2689 * This is the function where ext4lazyinit thread lives. It walks
2690 * through the request list searching for next scheduled filesystem.
2691 * When such a fs is found, run the lazy initialization request
2692 * (ext4_rn_li_request) and keep track of the time spend in this
2693 * function. Based on that time we compute next schedule time of
2694 * the request. When walking through the list is complete, compute
2695 * next waking time and put itself into sleep.
2697 static int ext4_lazyinit_thread(void *arg
)
2699 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2700 struct list_head
*pos
, *n
;
2701 struct ext4_li_request
*elr
;
2702 unsigned long next_wakeup
;
2705 BUG_ON(NULL
== eli
);
2707 eli
->li_timer
.data
= (unsigned long)current
;
2708 eli
->li_timer
.function
= ext4_lazyinode_timeout
;
2710 eli
->li_task
= current
;
2711 wake_up(&eli
->li_wait_task
);
2715 next_wakeup
= MAX_JIFFY_OFFSET
;
2717 mutex_lock(&eli
->li_list_mtx
);
2718 if (list_empty(&eli
->li_request_list
)) {
2719 mutex_unlock(&eli
->li_list_mtx
);
2723 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2724 elr
= list_entry(pos
, struct ext4_li_request
,
2727 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2728 if (ext4_run_li_request(elr
) != 0) {
2729 /* error, remove the lazy_init job */
2730 ext4_remove_li_request(elr
);
2735 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2736 next_wakeup
= elr
->lr_next_sched
;
2738 mutex_unlock(&eli
->li_list_mtx
);
2740 if (freezing(current
))
2743 if ((time_after_eq(jiffies
, next_wakeup
)) ||
2744 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2749 eli
->li_timer
.expires
= next_wakeup
;
2750 add_timer(&eli
->li_timer
);
2751 prepare_to_wait(&eli
->li_wait_daemon
, &wait
,
2752 TASK_INTERRUPTIBLE
);
2753 if (time_before(jiffies
, next_wakeup
))
2755 finish_wait(&eli
->li_wait_daemon
, &wait
);
2760 * It looks like the request list is empty, but we need
2761 * to check it under the li_list_mtx lock, to prevent any
2762 * additions into it, and of course we should lock ext4_li_mtx
2763 * to atomically free the list and ext4_li_info, because at
2764 * this point another ext4 filesystem could be registering
2767 mutex_lock(&ext4_li_mtx
);
2768 mutex_lock(&eli
->li_list_mtx
);
2769 if (!list_empty(&eli
->li_request_list
)) {
2770 mutex_unlock(&eli
->li_list_mtx
);
2771 mutex_unlock(&ext4_li_mtx
);
2774 mutex_unlock(&eli
->li_list_mtx
);
2775 del_timer_sync(&ext4_li_info
->li_timer
);
2776 eli
->li_task
= NULL
;
2777 wake_up(&eli
->li_wait_task
);
2779 kfree(ext4_li_info
);
2780 ext4_li_info
= NULL
;
2781 mutex_unlock(&ext4_li_mtx
);
2786 static void ext4_clear_request_list(void)
2788 struct list_head
*pos
, *n
;
2789 struct ext4_li_request
*elr
;
2791 mutex_lock(&ext4_li_info
->li_list_mtx
);
2792 if (list_empty(&ext4_li_info
->li_request_list
))
2795 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2796 elr
= list_entry(pos
, struct ext4_li_request
,
2798 ext4_remove_li_request(elr
);
2800 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2803 static int ext4_run_lazyinit_thread(void)
2805 struct task_struct
*t
;
2807 t
= kthread_run(ext4_lazyinit_thread
, ext4_li_info
, "ext4lazyinit");
2809 int err
= PTR_ERR(t
);
2810 ext4_clear_request_list();
2811 del_timer_sync(&ext4_li_info
->li_timer
);
2812 kfree(ext4_li_info
);
2813 ext4_li_info
= NULL
;
2814 printk(KERN_CRIT
"EXT4: error %d creating inode table "
2815 "initialization thread\n",
2819 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2821 wait_event(ext4_li_info
->li_wait_task
, ext4_li_info
->li_task
!= NULL
);
2826 * Check whether it make sense to run itable init. thread or not.
2827 * If there is at least one uninitialized inode table, return
2828 * corresponding group number, else the loop goes through all
2829 * groups and return total number of groups.
2831 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2833 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2834 struct ext4_group_desc
*gdp
= NULL
;
2836 for (group
= 0; group
< ngroups
; group
++) {
2837 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2841 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2848 static int ext4_li_info_new(void)
2850 struct ext4_lazy_init
*eli
= NULL
;
2852 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2856 eli
->li_task
= NULL
;
2857 INIT_LIST_HEAD(&eli
->li_request_list
);
2858 mutex_init(&eli
->li_list_mtx
);
2860 init_waitqueue_head(&eli
->li_wait_daemon
);
2861 init_waitqueue_head(&eli
->li_wait_task
);
2862 init_timer(&eli
->li_timer
);
2863 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2870 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2873 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2874 struct ext4_li_request
*elr
;
2877 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2883 elr
->lr_next_group
= start
;
2886 * Randomize first schedule time of the request to
2887 * spread the inode table initialization requests
2890 get_random_bytes(&rnd
, sizeof(rnd
));
2891 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2892 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
2897 static int ext4_register_li_request(struct super_block
*sb
,
2898 ext4_group_t first_not_zeroed
)
2900 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2901 struct ext4_li_request
*elr
;
2902 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
2905 if (sbi
->s_li_request
!= NULL
)
2908 if (first_not_zeroed
== ngroups
||
2909 (sb
->s_flags
& MS_RDONLY
) ||
2910 !test_opt(sb
, INIT_INODE_TABLE
)) {
2911 sbi
->s_li_request
= NULL
;
2915 if (first_not_zeroed
== ngroups
) {
2916 sbi
->s_li_request
= NULL
;
2920 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
2924 mutex_lock(&ext4_li_mtx
);
2926 if (NULL
== ext4_li_info
) {
2927 ret
= ext4_li_info_new();
2932 mutex_lock(&ext4_li_info
->li_list_mtx
);
2933 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
2934 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2936 sbi
->s_li_request
= elr
;
2938 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
2939 ret
= ext4_run_lazyinit_thread();
2944 mutex_unlock(&ext4_li_mtx
);
2951 * We do not need to lock anything since this is called on
2954 static void ext4_destroy_lazyinit_thread(void)
2957 * If thread exited earlier
2958 * there's nothing to be done.
2963 ext4_clear_request_list();
2965 while (ext4_li_info
->li_task
) {
2966 wake_up(&ext4_li_info
->li_wait_daemon
);
2967 wait_event(ext4_li_info
->li_wait_task
,
2968 ext4_li_info
->li_task
== NULL
);
2972 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
2973 __releases(kernel_lock
)
2974 __acquires(kernel_lock
)
2976 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
2977 struct buffer_head
*bh
;
2978 struct ext4_super_block
*es
= NULL
;
2979 struct ext4_sb_info
*sbi
;
2981 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2982 ext4_fsblk_t logical_sb_block
;
2983 unsigned long offset
= 0;
2984 unsigned long journal_devnum
= 0;
2985 unsigned long def_mount_opts
;
2991 unsigned int db_count
;
2993 int needs_recovery
, has_huge_files
;
2996 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2997 ext4_group_t first_not_zeroed
;
2999 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3003 sbi
->s_blockgroup_lock
=
3004 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3005 if (!sbi
->s_blockgroup_lock
) {
3009 sb
->s_fs_info
= sbi
;
3010 sbi
->s_mount_opt
= 0;
3011 sbi
->s_resuid
= EXT4_DEF_RESUID
;
3012 sbi
->s_resgid
= EXT4_DEF_RESGID
;
3013 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3014 sbi
->s_sb_block
= sb_block
;
3015 if (sb
->s_bdev
->bd_part
)
3016 sbi
->s_sectors_written_start
=
3017 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3019 /* Cleanup superblock name */
3020 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3024 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3026 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3031 * The ext4 superblock will not be buffer aligned for other than 1kB
3032 * block sizes. We need to calculate the offset from buffer start.
3034 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3035 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3036 offset
= do_div(logical_sb_block
, blocksize
);
3038 logical_sb_block
= sb_block
;
3041 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3042 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3046 * Note: s_es must be initialized as soon as possible because
3047 * some ext4 macro-instructions depend on its value
3049 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3051 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3052 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3054 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3056 /* Set defaults before we parse the mount options */
3057 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3058 set_opt(sbi
->s_mount_opt
, INIT_INODE_TABLE
);
3059 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3060 set_opt(sbi
->s_mount_opt
, DEBUG
);
3061 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
3062 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
3064 set_opt(sbi
->s_mount_opt
, GRPID
);
3066 if (def_mount_opts
& EXT4_DEFM_UID16
)
3067 set_opt(sbi
->s_mount_opt
, NO_UID32
);
3068 #ifdef CONFIG_EXT4_FS_XATTR
3069 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
3070 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
3072 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3073 if (def_mount_opts
& EXT4_DEFM_ACL
)
3074 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
3076 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3077 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
3078 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3079 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
3080 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3081 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
3083 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3084 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
3085 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3086 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
3088 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
3089 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3090 set_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
3091 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3092 set_opt(sbi
->s_mount_opt
, DISCARD
);
3094 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
3095 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
3096 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3097 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3098 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3100 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3101 set_opt(sbi
->s_mount_opt
, BARRIER
);
3104 * enable delayed allocation by default
3105 * Use -o nodelalloc to turn it off
3107 if (!IS_EXT3_SB(sb
) &&
3108 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3109 set_opt(sbi
->s_mount_opt
, DELALLOC
);
3111 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3112 &journal_devnum
, &journal_ioprio
, NULL
, 0)) {
3113 ext4_msg(sb
, KERN_WARNING
,
3114 "failed to parse options in superblock: %s",
3115 sbi
->s_es
->s_mount_opts
);
3117 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3118 &journal_ioprio
, NULL
, 0))
3121 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3122 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3124 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3125 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3126 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3127 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3128 ext4_msg(sb
, KERN_WARNING
,
3129 "feature flags set on rev 0 fs, "
3130 "running e2fsck is recommended");
3133 * Check feature flags regardless of the revision level, since we
3134 * previously didn't change the revision level when setting the flags,
3135 * so there is a chance incompat flags are set on a rev 0 filesystem.
3137 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3140 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3142 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3143 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3144 ext4_msg(sb
, KERN_ERR
,
3145 "Unsupported filesystem blocksize %d", blocksize
);
3149 if (sb
->s_blocksize
!= blocksize
) {
3150 /* Validate the filesystem blocksize */
3151 if (!sb_set_blocksize(sb
, blocksize
)) {
3152 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3158 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3159 offset
= do_div(logical_sb_block
, blocksize
);
3160 bh
= sb_bread(sb
, logical_sb_block
);
3162 ext4_msg(sb
, KERN_ERR
,
3163 "Can't read superblock on 2nd try");
3166 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
3168 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3169 ext4_msg(sb
, KERN_ERR
,
3170 "Magic mismatch, very weird!");
3175 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3176 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3177 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3179 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3181 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3182 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3183 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3185 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3186 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3187 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3188 (!is_power_of_2(sbi
->s_inode_size
)) ||
3189 (sbi
->s_inode_size
> blocksize
)) {
3190 ext4_msg(sb
, KERN_ERR
,
3191 "unsupported inode size: %d",
3195 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3196 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3199 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3200 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3201 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3202 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3203 !is_power_of_2(sbi
->s_desc_size
)) {
3204 ext4_msg(sb
, KERN_ERR
,
3205 "unsupported descriptor size %lu",
3210 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3212 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3213 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3214 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3217 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3218 if (sbi
->s_inodes_per_block
== 0)
3220 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3221 sbi
->s_inodes_per_block
;
3222 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3224 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3225 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3226 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3228 for (i
= 0; i
< 4; i
++)
3229 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3230 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3231 i
= le32_to_cpu(es
->s_flags
);
3232 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3233 sbi
->s_hash_unsigned
= 3;
3234 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3235 #ifdef __CHAR_UNSIGNED__
3236 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3237 sbi
->s_hash_unsigned
= 3;
3239 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3244 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3245 ext4_msg(sb
, KERN_ERR
,
3246 "#blocks per group too big: %lu",
3247 sbi
->s_blocks_per_group
);
3250 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3251 ext4_msg(sb
, KERN_ERR
,
3252 "#inodes per group too big: %lu",
3253 sbi
->s_inodes_per_group
);
3258 * Test whether we have more sectors than will fit in sector_t,
3259 * and whether the max offset is addressable by the page cache.
3261 ret
= generic_check_addressable(sb
->s_blocksize_bits
,
3262 ext4_blocks_count(es
));
3264 ext4_msg(sb
, KERN_ERR
, "filesystem"
3265 " too large to mount safely on this system");
3266 if (sizeof(sector_t
) < 8)
3267 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3271 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3274 /* check blocks count against device size */
3275 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3276 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3277 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3278 "exceeds size of device (%llu blocks)",
3279 ext4_blocks_count(es
), blocks_count
);
3284 * It makes no sense for the first data block to be beyond the end
3285 * of the filesystem.
3287 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3288 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
3289 "block %u is beyond end of filesystem (%llu)",
3290 le32_to_cpu(es
->s_first_data_block
),
3291 ext4_blocks_count(es
));
3294 blocks_count
= (ext4_blocks_count(es
) -
3295 le32_to_cpu(es
->s_first_data_block
) +
3296 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3297 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3298 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3299 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3300 "(block count %llu, first data block %u, "
3301 "blocks per group %lu)", sbi
->s_groups_count
,
3302 ext4_blocks_count(es
),
3303 le32_to_cpu(es
->s_first_data_block
),
3304 EXT4_BLOCKS_PER_GROUP(sb
));
3307 sbi
->s_groups_count
= blocks_count
;
3308 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3309 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3310 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3311 EXT4_DESC_PER_BLOCK(sb
);
3312 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
3314 if (sbi
->s_group_desc
== NULL
) {
3315 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3319 #ifdef CONFIG_PROC_FS
3321 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3324 bgl_lock_init(sbi
->s_blockgroup_lock
);
3326 for (i
= 0; i
< db_count
; i
++) {
3327 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3328 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3329 if (!sbi
->s_group_desc
[i
]) {
3330 ext4_msg(sb
, KERN_ERR
,
3331 "can't read group descriptor %d", i
);
3336 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3337 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3340 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3341 if (!ext4_fill_flex_info(sb
)) {
3342 ext4_msg(sb
, KERN_ERR
,
3343 "unable to initialize "
3344 "flex_bg meta info!");
3348 sbi
->s_gdb_count
= db_count
;
3349 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3350 spin_lock_init(&sbi
->s_next_gen_lock
);
3352 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
3353 ext4_count_free_blocks(sb
));
3355 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3356 ext4_count_free_inodes(sb
));
3359 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3360 ext4_count_dirs(sb
));
3363 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
3366 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3370 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3371 sbi
->s_max_writeback_mb_bump
= 128;
3374 * set up enough so that it can read an inode
3376 if (!test_opt(sb
, NOLOAD
) &&
3377 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3378 sb
->s_op
= &ext4_sops
;
3380 sb
->s_op
= &ext4_nojournal_sops
;
3381 sb
->s_export_op
= &ext4_export_ops
;
3382 sb
->s_xattr
= ext4_xattr_handlers
;
3384 sb
->s_qcop
= &ext4_qctl_operations
;
3385 sb
->dq_op
= &ext4_quota_operations
;
3387 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3388 mutex_init(&sbi
->s_orphan_lock
);
3389 mutex_init(&sbi
->s_resize_lock
);
3393 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3394 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3395 EXT4_FEATURE_INCOMPAT_RECOVER
));
3398 * The first inode we look at is the journal inode. Don't try
3399 * root first: it may be modified in the journal!
3401 if (!test_opt(sb
, NOLOAD
) &&
3402 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3403 if (ext4_load_journal(sb
, es
, journal_devnum
))
3405 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3406 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3407 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3408 "suppressed and not mounted read-only");
3409 goto failed_mount_wq
;
3411 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
3412 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
3413 sbi
->s_journal
= NULL
;
3418 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
3419 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3420 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3421 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3422 goto failed_mount_wq
;
3425 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3426 jbd2_journal_set_features(sbi
->s_journal
,
3427 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3428 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3429 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3430 jbd2_journal_set_features(sbi
->s_journal
,
3431 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
3432 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3433 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3435 jbd2_journal_clear_features(sbi
->s_journal
,
3436 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3437 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3440 /* We have now updated the journal if required, so we can
3441 * validate the data journaling mode. */
3442 switch (test_opt(sb
, DATA_FLAGS
)) {
3444 /* No mode set, assume a default based on the journal
3445 * capabilities: ORDERED_DATA if the journal can
3446 * cope, else JOURNAL_DATA
3448 if (jbd2_journal_check_available_features
3449 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3450 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
3452 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
3455 case EXT4_MOUNT_ORDERED_DATA
:
3456 case EXT4_MOUNT_WRITEBACK_DATA
:
3457 if (!jbd2_journal_check_available_features
3458 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3459 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3460 "requested data journaling mode");
3461 goto failed_mount_wq
;
3466 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3469 * The journal may have updated the bg summary counts, so we
3470 * need to update the global counters.
3472 percpu_counter_set(&sbi
->s_freeblocks_counter
,
3473 ext4_count_free_blocks(sb
));
3474 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3475 ext4_count_free_inodes(sb
));
3476 percpu_counter_set(&sbi
->s_dirs_counter
,
3477 ext4_count_dirs(sb
));
3478 percpu_counter_set(&sbi
->s_dirtyblocks_counter
, 0);
3481 EXT4_SB(sb
)->dio_unwritten_wq
= create_workqueue("ext4-dio-unwritten");
3482 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3483 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3484 goto failed_mount_wq
;
3488 * The jbd2_journal_load will have done any necessary log recovery,
3489 * so we can safely mount the rest of the filesystem now.
3492 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3494 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3495 ret
= PTR_ERR(root
);
3498 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3500 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3503 sb
->s_root
= d_alloc_root(root
);
3505 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3511 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
3513 /* determine the minimum size of new large inodes, if present */
3514 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3515 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3516 EXT4_GOOD_OLD_INODE_SIZE
;
3517 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3518 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3519 if (sbi
->s_want_extra_isize
<
3520 le16_to_cpu(es
->s_want_extra_isize
))
3521 sbi
->s_want_extra_isize
=
3522 le16_to_cpu(es
->s_want_extra_isize
);
3523 if (sbi
->s_want_extra_isize
<
3524 le16_to_cpu(es
->s_min_extra_isize
))
3525 sbi
->s_want_extra_isize
=
3526 le16_to_cpu(es
->s_min_extra_isize
);
3529 /* Check if enough inode space is available */
3530 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3531 sbi
->s_inode_size
) {
3532 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3533 EXT4_GOOD_OLD_INODE_SIZE
;
3534 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3538 if (test_opt(sb
, DELALLOC
) &&
3539 (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)) {
3540 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
3541 "requested data journaling mode");
3542 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
3544 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3545 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3546 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3547 "option - requested data journaling mode");
3548 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
3550 if (sb
->s_blocksize
< PAGE_SIZE
) {
3551 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3552 "option - block size is too small");
3553 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
3557 err
= ext4_setup_system_zone(sb
);
3559 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3565 err
= ext4_mb_init(sb
, needs_recovery
);
3567 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3572 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3576 sbi
->s_kobj
.kset
= ext4_kset
;
3577 init_completion(&sbi
->s_kobj_unregister
);
3578 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3581 ext4_mb_release(sb
);
3582 ext4_ext_release(sb
);
3586 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3587 ext4_orphan_cleanup(sb
, es
);
3588 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3589 if (needs_recovery
) {
3590 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3591 ext4_mark_recovery_complete(sb
, es
);
3593 if (EXT4_SB(sb
)->s_journal
) {
3594 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3595 descr
= " journalled data mode";
3596 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3597 descr
= " ordered data mode";
3599 descr
= " writeback data mode";
3601 descr
= "out journal";
3603 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3604 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3605 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3607 init_timer(&sbi
->s_err_report
);
3608 sbi
->s_err_report
.function
= print_daily_error_info
;
3609 sbi
->s_err_report
.data
= (unsigned long) sb
;
3610 if (es
->s_error_count
)
3611 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3618 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3622 ext4_msg(sb
, KERN_ERR
, "mount failed");
3623 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3625 ext4_release_system_zone(sb
);
3626 if (sbi
->s_journal
) {
3627 jbd2_journal_destroy(sbi
->s_journal
);
3628 sbi
->s_journal
= NULL
;
3631 if (sbi
->s_flex_groups
) {
3632 if (is_vmalloc_addr(sbi
->s_flex_groups
))
3633 vfree(sbi
->s_flex_groups
);
3635 kfree(sbi
->s_flex_groups
);
3637 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
3638 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3639 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3640 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
3642 for (i
= 0; i
< db_count
; i
++)
3643 brelse(sbi
->s_group_desc
[i
]);
3644 kfree(sbi
->s_group_desc
);
3647 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3650 for (i
= 0; i
< MAXQUOTAS
; i
++)
3651 kfree(sbi
->s_qf_names
[i
]);
3653 ext4_blkdev_remove(sbi
);
3656 sb
->s_fs_info
= NULL
;
3657 kfree(sbi
->s_blockgroup_lock
);
3665 * Setup any per-fs journal parameters now. We'll do this both on
3666 * initial mount, once the journal has been initialised but before we've
3667 * done any recovery; and again on any subsequent remount.
3669 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3671 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3673 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3674 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3675 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3677 write_lock(&journal
->j_state_lock
);
3678 if (test_opt(sb
, BARRIER
))
3679 journal
->j_flags
|= JBD2_BARRIER
;
3681 journal
->j_flags
&= ~JBD2_BARRIER
;
3682 if (test_opt(sb
, DATA_ERR_ABORT
))
3683 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3685 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3686 write_unlock(&journal
->j_state_lock
);
3689 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3690 unsigned int journal_inum
)
3692 struct inode
*journal_inode
;
3695 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3697 /* First, test for the existence of a valid inode on disk. Bad
3698 * things happen if we iget() an unused inode, as the subsequent
3699 * iput() will try to delete it. */
3701 journal_inode
= ext4_iget(sb
, journal_inum
);
3702 if (IS_ERR(journal_inode
)) {
3703 ext4_msg(sb
, KERN_ERR
, "no journal found");
3706 if (!journal_inode
->i_nlink
) {
3707 make_bad_inode(journal_inode
);
3708 iput(journal_inode
);
3709 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3713 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3714 journal_inode
, journal_inode
->i_size
);
3715 if (!S_ISREG(journal_inode
->i_mode
)) {
3716 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3717 iput(journal_inode
);
3721 journal
= jbd2_journal_init_inode(journal_inode
);
3723 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3724 iput(journal_inode
);
3727 journal
->j_private
= sb
;
3728 ext4_init_journal_params(sb
, journal
);
3732 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3735 struct buffer_head
*bh
;
3739 int hblock
, blocksize
;
3740 ext4_fsblk_t sb_block
;
3741 unsigned long offset
;
3742 struct ext4_super_block
*es
;
3743 struct block_device
*bdev
;
3745 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3747 bdev
= ext4_blkdev_get(j_dev
, sb
);
3751 if (bd_claim(bdev
, sb
)) {
3752 ext4_msg(sb
, KERN_ERR
,
3753 "failed to claim external journal device");
3754 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
3758 blocksize
= sb
->s_blocksize
;
3759 hblock
= bdev_logical_block_size(bdev
);
3760 if (blocksize
< hblock
) {
3761 ext4_msg(sb
, KERN_ERR
,
3762 "blocksize too small for journal device");
3766 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3767 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3768 set_blocksize(bdev
, blocksize
);
3769 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3770 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3771 "external journal");
3775 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3776 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3777 !(le32_to_cpu(es
->s_feature_incompat
) &
3778 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3779 ext4_msg(sb
, KERN_ERR
, "external journal has "
3785 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3786 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3791 len
= ext4_blocks_count(es
);
3792 start
= sb_block
+ 1;
3793 brelse(bh
); /* we're done with the superblock */
3795 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3796 start
, len
, blocksize
);
3798 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3801 journal
->j_private
= sb
;
3802 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3803 wait_on_buffer(journal
->j_sb_buffer
);
3804 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3805 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3808 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3809 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3810 "user (unsupported) - %d",
3811 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3814 EXT4_SB(sb
)->journal_bdev
= bdev
;
3815 ext4_init_journal_params(sb
, journal
);
3819 jbd2_journal_destroy(journal
);
3821 ext4_blkdev_put(bdev
);
3825 static int ext4_load_journal(struct super_block
*sb
,
3826 struct ext4_super_block
*es
,
3827 unsigned long journal_devnum
)
3830 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3833 int really_read_only
;
3835 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3837 if (journal_devnum
&&
3838 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3839 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3840 "numbers have changed");
3841 journal_dev
= new_decode_dev(journal_devnum
);
3843 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3845 really_read_only
= bdev_read_only(sb
->s_bdev
);
3848 * Are we loading a blank journal or performing recovery after a
3849 * crash? For recovery, we need to check in advance whether we
3850 * can get read-write access to the device.
3852 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3853 if (sb
->s_flags
& MS_RDONLY
) {
3854 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3855 "required on readonly filesystem");
3856 if (really_read_only
) {
3857 ext4_msg(sb
, KERN_ERR
, "write access "
3858 "unavailable, cannot proceed");
3861 ext4_msg(sb
, KERN_INFO
, "write access will "
3862 "be enabled during recovery");
3866 if (journal_inum
&& journal_dev
) {
3867 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3868 "and inode journals!");
3873 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3876 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3880 if (!(journal
->j_flags
& JBD2_BARRIER
))
3881 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3883 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
3884 err
= jbd2_journal_update_format(journal
);
3886 ext4_msg(sb
, KERN_ERR
, "error updating journal");
3887 jbd2_journal_destroy(journal
);
3892 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
3893 err
= jbd2_journal_wipe(journal
, !really_read_only
);
3895 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
3897 memcpy(save
, ((char *) es
) +
3898 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
3899 err
= jbd2_journal_load(journal
);
3901 memcpy(((char *) es
) + EXT4_S_ERR_START
,
3902 save
, EXT4_S_ERR_LEN
);
3907 ext4_msg(sb
, KERN_ERR
, "error loading journal");
3908 jbd2_journal_destroy(journal
);
3912 EXT4_SB(sb
)->s_journal
= journal
;
3913 ext4_clear_journal_err(sb
, es
);
3915 if (!really_read_only
&& journal_devnum
&&
3916 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3917 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
3919 /* Make sure we flush the recovery flag to disk. */
3920 ext4_commit_super(sb
, 1);
3926 static int ext4_commit_super(struct super_block
*sb
, int sync
)
3928 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
3929 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
3934 if (buffer_write_io_error(sbh
)) {
3936 * Oh, dear. A previous attempt to write the
3937 * superblock failed. This could happen because the
3938 * USB device was yanked out. Or it could happen to
3939 * be a transient write error and maybe the block will
3940 * be remapped. Nothing we can do but to retry the
3941 * write and hope for the best.
3943 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
3944 "superblock detected");
3945 clear_buffer_write_io_error(sbh
);
3946 set_buffer_uptodate(sbh
);
3949 * If the file system is mounted read-only, don't update the
3950 * superblock write time. This avoids updating the superblock
3951 * write time when we are mounting the root file system
3952 * read/only but we need to replay the journal; at that point,
3953 * for people who are east of GMT and who make their clock
3954 * tick in localtime for Windows bug-for-bug compatibility,
3955 * the clock is set in the future, and this will cause e2fsck
3956 * to complain and force a full file system check.
3958 if (!(sb
->s_flags
& MS_RDONLY
))
3959 es
->s_wtime
= cpu_to_le32(get_seconds());
3960 if (sb
->s_bdev
->bd_part
)
3961 es
->s_kbytes_written
=
3962 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
3963 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
3964 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
3966 es
->s_kbytes_written
=
3967 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
3968 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
3969 &EXT4_SB(sb
)->s_freeblocks_counter
));
3970 es
->s_free_inodes_count
=
3971 cpu_to_le32(percpu_counter_sum_positive(
3972 &EXT4_SB(sb
)->s_freeinodes_counter
));
3974 BUFFER_TRACE(sbh
, "marking dirty");
3975 mark_buffer_dirty(sbh
);
3977 error
= sync_dirty_buffer(sbh
);
3981 error
= buffer_write_io_error(sbh
);
3983 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
3985 clear_buffer_write_io_error(sbh
);
3986 set_buffer_uptodate(sbh
);
3993 * Have we just finished recovery? If so, and if we are mounting (or
3994 * remounting) the filesystem readonly, then we will end up with a
3995 * consistent fs on disk. Record that fact.
3997 static void ext4_mark_recovery_complete(struct super_block
*sb
,
3998 struct ext4_super_block
*es
)
4000 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4002 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4003 BUG_ON(journal
!= NULL
);
4006 jbd2_journal_lock_updates(journal
);
4007 if (jbd2_journal_flush(journal
) < 0)
4010 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4011 sb
->s_flags
& MS_RDONLY
) {
4012 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4013 ext4_commit_super(sb
, 1);
4017 jbd2_journal_unlock_updates(journal
);
4021 * If we are mounting (or read-write remounting) a filesystem whose journal
4022 * has recorded an error from a previous lifetime, move that error to the
4023 * main filesystem now.
4025 static void ext4_clear_journal_err(struct super_block
*sb
,
4026 struct ext4_super_block
*es
)
4032 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4034 journal
= EXT4_SB(sb
)->s_journal
;
4037 * Now check for any error status which may have been recorded in the
4038 * journal by a prior ext4_error() or ext4_abort()
4041 j_errno
= jbd2_journal_errno(journal
);
4045 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4046 ext4_warning(sb
, "Filesystem error recorded "
4047 "from previous mount: %s", errstr
);
4048 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4050 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4051 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4052 ext4_commit_super(sb
, 1);
4054 jbd2_journal_clear_err(journal
);
4059 * Force the running and committing transactions to commit,
4060 * and wait on the commit.
4062 int ext4_force_commit(struct super_block
*sb
)
4067 if (sb
->s_flags
& MS_RDONLY
)
4070 journal
= EXT4_SB(sb
)->s_journal
;
4072 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
4073 ret
= ext4_journal_force_commit(journal
);
4079 static void ext4_write_super(struct super_block
*sb
)
4082 ext4_commit_super(sb
, 1);
4086 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4090 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4092 trace_ext4_sync_fs(sb
, wait
);
4093 flush_workqueue(sbi
->dio_unwritten_wq
);
4094 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4096 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4102 * LVM calls this function before a (read-only) snapshot is created. This
4103 * gives us a chance to flush the journal completely and mark the fs clean.
4105 static int ext4_freeze(struct super_block
*sb
)
4110 if (sb
->s_flags
& MS_RDONLY
)
4113 journal
= EXT4_SB(sb
)->s_journal
;
4115 /* Now we set up the journal barrier. */
4116 jbd2_journal_lock_updates(journal
);
4119 * Don't clear the needs_recovery flag if we failed to flush
4122 error
= jbd2_journal_flush(journal
);
4126 /* Journal blocked and flushed, clear needs_recovery flag. */
4127 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4128 error
= ext4_commit_super(sb
, 1);
4130 /* we rely on s_frozen to stop further updates */
4131 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4136 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4137 * flag here, even though the filesystem is not technically dirty yet.
4139 static int ext4_unfreeze(struct super_block
*sb
)
4141 if (sb
->s_flags
& MS_RDONLY
)
4145 /* Reset the needs_recovery flag before the fs is unlocked. */
4146 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4147 ext4_commit_super(sb
, 1);
4152 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4154 struct ext4_super_block
*es
;
4155 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4156 ext4_fsblk_t n_blocks_count
= 0;
4157 unsigned long old_sb_flags
;
4158 struct ext4_mount_options old_opts
;
4159 int enable_quota
= 0;
4161 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4166 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4168 /* Store the original options */
4170 old_sb_flags
= sb
->s_flags
;
4171 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4172 old_opts
.s_resuid
= sbi
->s_resuid
;
4173 old_opts
.s_resgid
= sbi
->s_resgid
;
4174 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4175 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4176 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4178 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4179 for (i
= 0; i
< MAXQUOTAS
; i
++)
4180 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
4182 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4183 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4186 * Allow the "check" option to be passed as a remount option.
4188 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
4189 &n_blocks_count
, 1)) {
4194 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4195 ext4_abort(sb
, "Abort forced by user");
4197 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4198 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4202 if (sbi
->s_journal
) {
4203 ext4_init_journal_params(sb
, sbi
->s_journal
);
4204 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4207 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
4208 n_blocks_count
> ext4_blocks_count(es
)) {
4209 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4214 if (*flags
& MS_RDONLY
) {
4215 err
= dquot_suspend(sb
, -1);
4220 * First of all, the unconditional stuff we have to do
4221 * to disable replay of the journal when we next remount
4223 sb
->s_flags
|= MS_RDONLY
;
4226 * OK, test if we are remounting a valid rw partition
4227 * readonly, and if so set the rdonly flag and then
4228 * mark the partition as valid again.
4230 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4231 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4232 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4235 ext4_mark_recovery_complete(sb
, es
);
4237 /* Make sure we can mount this feature set readwrite */
4238 if (!ext4_feature_set_ok(sb
, 0)) {
4243 * Make sure the group descriptor checksums
4244 * are sane. If they aren't, refuse to remount r/w.
4246 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4247 struct ext4_group_desc
*gdp
=
4248 ext4_get_group_desc(sb
, g
, NULL
);
4250 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
4251 ext4_msg(sb
, KERN_ERR
,
4252 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4253 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4254 le16_to_cpu(gdp
->bg_checksum
));
4261 * If we have an unprocessed orphan list hanging
4262 * around from a previously readonly bdev mount,
4263 * require a full umount/remount for now.
4265 if (es
->s_last_orphan
) {
4266 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4267 "remount RDWR because of unprocessed "
4268 "orphan inode list. Please "
4269 "umount/remount instead");
4275 * Mounting a RDONLY partition read-write, so reread
4276 * and store the current valid flag. (It may have
4277 * been changed by e2fsck since we originally mounted
4281 ext4_clear_journal_err(sb
, es
);
4282 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4283 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
4285 if (!ext4_setup_super(sb
, es
, 0))
4286 sb
->s_flags
&= ~MS_RDONLY
;
4292 * Reinitialize lazy itable initialization thread based on
4295 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4296 ext4_unregister_li_request(sb
);
4298 ext4_group_t first_not_zeroed
;
4299 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4300 ext4_register_li_request(sb
, first_not_zeroed
);
4303 ext4_setup_system_zone(sb
);
4304 if (sbi
->s_journal
== NULL
)
4305 ext4_commit_super(sb
, 1);
4308 /* Release old quota file names */
4309 for (i
= 0; i
< MAXQUOTAS
; i
++)
4310 if (old_opts
.s_qf_names
[i
] &&
4311 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4312 kfree(old_opts
.s_qf_names
[i
]);
4316 dquot_resume(sb
, -1);
4318 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4323 sb
->s_flags
= old_sb_flags
;
4324 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4325 sbi
->s_resuid
= old_opts
.s_resuid
;
4326 sbi
->s_resgid
= old_opts
.s_resgid
;
4327 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4328 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4329 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4331 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4332 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4333 if (sbi
->s_qf_names
[i
] &&
4334 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4335 kfree(sbi
->s_qf_names
[i
]);
4336 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4344 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4346 struct super_block
*sb
= dentry
->d_sb
;
4347 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4348 struct ext4_super_block
*es
= sbi
->s_es
;
4351 if (test_opt(sb
, MINIX_DF
)) {
4352 sbi
->s_overhead_last
= 0;
4353 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
4354 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
4355 ext4_fsblk_t overhead
= 0;
4358 * Compute the overhead (FS structures). This is constant
4359 * for a given filesystem unless the number of block groups
4360 * changes so we cache the previous value until it does.
4364 * All of the blocks before first_data_block are
4367 overhead
= le32_to_cpu(es
->s_first_data_block
);
4370 * Add the overhead attributed to the superblock and
4371 * block group descriptors. If the sparse superblocks
4372 * feature is turned on, then not all groups have this.
4374 for (i
= 0; i
< ngroups
; i
++) {
4375 overhead
+= ext4_bg_has_super(sb
, i
) +
4376 ext4_bg_num_gdb(sb
, i
);
4381 * Every block group has an inode bitmap, a block
4382 * bitmap, and an inode table.
4384 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
4385 sbi
->s_overhead_last
= overhead
;
4387 sbi
->s_blocks_last
= ext4_blocks_count(es
);
4390 buf
->f_type
= EXT4_SUPER_MAGIC
;
4391 buf
->f_bsize
= sb
->s_blocksize
;
4392 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
4393 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
4394 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
4395 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
4396 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
4398 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4399 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4400 buf
->f_namelen
= EXT4_NAME_LEN
;
4401 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4402 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4403 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4404 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4409 /* Helper function for writing quotas on sync - we need to start transaction
4410 * before quota file is locked for write. Otherwise the are possible deadlocks:
4411 * Process 1 Process 2
4412 * ext4_create() quota_sync()
4413 * jbd2_journal_start() write_dquot()
4414 * dquot_initialize() down(dqio_mutex)
4415 * down(dqio_mutex) jbd2_journal_start()
4421 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4423 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
4426 static int ext4_write_dquot(struct dquot
*dquot
)
4430 struct inode
*inode
;
4432 inode
= dquot_to_inode(dquot
);
4433 handle
= ext4_journal_start(inode
,
4434 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4436 return PTR_ERR(handle
);
4437 ret
= dquot_commit(dquot
);
4438 err
= ext4_journal_stop(handle
);
4444 static int ext4_acquire_dquot(struct dquot
*dquot
)
4449 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4450 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4452 return PTR_ERR(handle
);
4453 ret
= dquot_acquire(dquot
);
4454 err
= ext4_journal_stop(handle
);
4460 static int ext4_release_dquot(struct dquot
*dquot
)
4465 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4466 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4467 if (IS_ERR(handle
)) {
4468 /* Release dquot anyway to avoid endless cycle in dqput() */
4469 dquot_release(dquot
);
4470 return PTR_ERR(handle
);
4472 ret
= dquot_release(dquot
);
4473 err
= ext4_journal_stop(handle
);
4479 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4481 /* Are we journaling quotas? */
4482 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4483 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4484 dquot_mark_dquot_dirty(dquot
);
4485 return ext4_write_dquot(dquot
);
4487 return dquot_mark_dquot_dirty(dquot
);
4491 static int ext4_write_info(struct super_block
*sb
, int type
)
4496 /* Data block + inode block */
4497 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
4499 return PTR_ERR(handle
);
4500 ret
= dquot_commit_info(sb
, type
);
4501 err
= ext4_journal_stop(handle
);
4508 * Turn on quotas during mount time - we need to find
4509 * the quota file and such...
4511 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4513 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4514 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4518 * Standard function to be called on quota_on
4520 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4526 if (!test_opt(sb
, QUOTA
))
4529 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
4533 /* Quotafile not on the same filesystem? */
4534 if (path
.mnt
->mnt_sb
!= sb
) {
4538 /* Journaling quota? */
4539 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4540 /* Quotafile not in fs root? */
4541 if (path
.dentry
->d_parent
!= sb
->s_root
)
4542 ext4_msg(sb
, KERN_WARNING
,
4543 "Quota file not on filesystem root. "
4544 "Journaled quota will not work");
4548 * When we journal data on quota file, we have to flush journal to see
4549 * all updates to the file when we bypass pagecache...
4551 if (EXT4_SB(sb
)->s_journal
&&
4552 ext4_should_journal_data(path
.dentry
->d_inode
)) {
4554 * We don't need to lock updates but journal_flush() could
4555 * otherwise be livelocked...
4557 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4558 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4559 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4566 err
= dquot_quota_on_path(sb
, type
, format_id
, &path
);
4571 static int ext4_quota_off(struct super_block
*sb
, int type
)
4573 /* Force all delayed allocation blocks to be allocated */
4574 if (test_opt(sb
, DELALLOC
)) {
4575 down_read(&sb
->s_umount
);
4576 sync_filesystem(sb
);
4577 up_read(&sb
->s_umount
);
4580 return dquot_quota_off(sb
, type
);
4583 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4584 * acquiring the locks... As quota files are never truncated and quota code
4585 * itself serializes the operations (and noone else should touch the files)
4586 * we don't have to be afraid of races */
4587 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
4588 size_t len
, loff_t off
)
4590 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4591 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4593 int offset
= off
& (sb
->s_blocksize
- 1);
4596 struct buffer_head
*bh
;
4597 loff_t i_size
= i_size_read(inode
);
4601 if (off
+len
> i_size
)
4604 while (toread
> 0) {
4605 tocopy
= sb
->s_blocksize
- offset
< toread
?
4606 sb
->s_blocksize
- offset
: toread
;
4607 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
4610 if (!bh
) /* A hole? */
4611 memset(data
, 0, tocopy
);
4613 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4623 /* Write to quotafile (we know the transaction is already started and has
4624 * enough credits) */
4625 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4626 const char *data
, size_t len
, loff_t off
)
4628 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4629 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4631 int offset
= off
& (sb
->s_blocksize
- 1);
4632 struct buffer_head
*bh
;
4633 handle_t
*handle
= journal_current_handle();
4635 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4636 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4637 " cancelled because transaction is not started",
4638 (unsigned long long)off
, (unsigned long long)len
);
4642 * Since we account only one data block in transaction credits,
4643 * then it is impossible to cross a block boundary.
4645 if (sb
->s_blocksize
- offset
< len
) {
4646 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4647 " cancelled because not block aligned",
4648 (unsigned long long)off
, (unsigned long long)len
);
4652 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4653 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4656 err
= ext4_journal_get_write_access(handle
, bh
);
4662 memcpy(bh
->b_data
+offset
, data
, len
);
4663 flush_dcache_page(bh
->b_page
);
4665 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4669 mutex_unlock(&inode
->i_mutex
);
4672 if (inode
->i_size
< off
+ len
) {
4673 i_size_write(inode
, off
+ len
);
4674 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4676 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4677 ext4_mark_inode_dirty(handle
, inode
);
4678 mutex_unlock(&inode
->i_mutex
);
4684 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
4685 const char *dev_name
, void *data
)
4687 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
4690 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4691 static struct file_system_type ext2_fs_type
= {
4692 .owner
= THIS_MODULE
,
4694 .mount
= ext4_mount
,
4695 .kill_sb
= kill_block_super
,
4696 .fs_flags
= FS_REQUIRES_DEV
,
4699 static inline void register_as_ext2(void)
4701 int err
= register_filesystem(&ext2_fs_type
);
4704 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4707 static inline void unregister_as_ext2(void)
4709 unregister_filesystem(&ext2_fs_type
);
4711 MODULE_ALIAS("ext2");
4713 static inline void register_as_ext2(void) { }
4714 static inline void unregister_as_ext2(void) { }
4717 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4718 static inline void register_as_ext3(void)
4720 int err
= register_filesystem(&ext3_fs_type
);
4723 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4726 static inline void unregister_as_ext3(void)
4728 unregister_filesystem(&ext3_fs_type
);
4730 MODULE_ALIAS("ext3");
4732 static inline void register_as_ext3(void) { }
4733 static inline void unregister_as_ext3(void) { }
4736 static struct file_system_type ext4_fs_type
= {
4737 .owner
= THIS_MODULE
,
4739 .mount
= ext4_mount
,
4740 .kill_sb
= kill_block_super
,
4741 .fs_flags
= FS_REQUIRES_DEV
,
4744 int __init
ext4_init_feat_adverts(void)
4746 struct ext4_features
*ef
;
4749 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
4753 ef
->f_kobj
.kset
= ext4_kset
;
4754 init_completion(&ef
->f_kobj_unregister
);
4755 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
4768 static int __init
ext4_init_fs(void)
4772 ext4_check_flag_values();
4773 err
= ext4_init_pageio();
4776 err
= ext4_init_system_zone();
4779 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4782 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4784 err
= ext4_init_feat_adverts();
4786 err
= ext4_init_mballoc();
4790 err
= ext4_init_xattr();
4793 err
= init_inodecache();
4798 err
= register_filesystem(&ext4_fs_type
);
4802 ext4_li_info
= NULL
;
4803 mutex_init(&ext4_li_mtx
);
4806 unregister_as_ext2();
4807 unregister_as_ext3();
4808 destroy_inodecache();
4812 ext4_exit_mballoc();
4815 remove_proc_entry("fs/ext4", NULL
);
4816 kset_unregister(ext4_kset
);
4818 ext4_exit_system_zone();
4824 static void __exit
ext4_exit_fs(void)
4826 ext4_destroy_lazyinit_thread();
4827 unregister_as_ext2();
4828 unregister_as_ext3();
4829 unregister_filesystem(&ext4_fs_type
);
4830 destroy_inodecache();
4832 ext4_exit_mballoc();
4833 remove_proc_entry("fs/ext4", NULL
);
4834 kset_unregister(ext4_kset
);
4835 ext4_exit_system_zone();
4839 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4840 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4841 MODULE_LICENSE("GPL");
4842 module_init(ext4_init_fs
)
4843 module_exit(ext4_exit_fs
)