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 <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct proc_dir_entry
*ext4_proc_root
;
57 static struct kset
*ext4_kset
;
58 static struct ext4_lazy_init
*ext4_li_info
;
59 static struct mutex ext4_li_mtx
;
60 static struct ext4_features
*ext4_feat
;
62 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
63 unsigned long journal_devnum
);
64 static int ext4_commit_super(struct super_block
*sb
, int sync
);
65 static void ext4_mark_recovery_complete(struct super_block
*sb
,
66 struct ext4_super_block
*es
);
67 static void ext4_clear_journal_err(struct super_block
*sb
,
68 struct ext4_super_block
*es
);
69 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
70 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
72 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
73 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
74 static int ext4_unfreeze(struct super_block
*sb
);
75 static void ext4_write_super(struct super_block
*sb
);
76 static int ext4_freeze(struct super_block
*sb
);
77 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
78 const char *dev_name
, void *data
);
79 static inline int ext2_feature_set_ok(struct super_block
*sb
);
80 static inline int ext3_feature_set_ok(struct super_block
*sb
);
81 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block
*sb
);
84 static void ext4_clear_request_list(void);
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type
= {
91 .kill_sb
= kill_block_super
,
92 .fs_flags
= FS_REQUIRES_DEV
,
94 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96 #define IS_EXT2_SB(sb) (0)
100 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
101 static struct file_system_type ext3_fs_type
= {
102 .owner
= THIS_MODULE
,
105 .kill_sb
= kill_block_super
,
106 .fs_flags
= FS_REQUIRES_DEV
,
108 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110 #define IS_EXT3_SB(sb) (0)
113 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
114 struct ext4_group_desc
*bg
)
116 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
117 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
118 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
121 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
122 struct ext4_group_desc
*bg
)
124 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
125 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
126 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
129 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
130 struct ext4_group_desc
*bg
)
132 return le32_to_cpu(bg
->bg_inode_table_lo
) |
133 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
134 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
137 __u32
ext4_free_blks_count(struct super_block
*sb
,
138 struct ext4_group_desc
*bg
)
140 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
141 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
142 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
145 __u32
ext4_free_inodes_count(struct super_block
*sb
,
146 struct ext4_group_desc
*bg
)
148 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
149 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
150 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
153 __u32
ext4_used_dirs_count(struct super_block
*sb
,
154 struct ext4_group_desc
*bg
)
156 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
157 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
158 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
161 __u32
ext4_itable_unused_count(struct super_block
*sb
,
162 struct ext4_group_desc
*bg
)
164 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
165 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
166 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
169 void ext4_block_bitmap_set(struct super_block
*sb
,
170 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
172 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
173 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
174 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
177 void ext4_inode_bitmap_set(struct super_block
*sb
,
178 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
180 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
181 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
182 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
185 void ext4_inode_table_set(struct super_block
*sb
,
186 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
188 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
189 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
190 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
193 void ext4_free_blks_set(struct super_block
*sb
,
194 struct ext4_group_desc
*bg
, __u32 count
)
196 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
197 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
198 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
201 void ext4_free_inodes_set(struct super_block
*sb
,
202 struct ext4_group_desc
*bg
, __u32 count
)
204 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
205 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
206 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
209 void ext4_used_dirs_set(struct super_block
*sb
,
210 struct ext4_group_desc
*bg
, __u32 count
)
212 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
213 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
214 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
217 void ext4_itable_unused_set(struct super_block
*sb
,
218 struct ext4_group_desc
*bg
, __u32 count
)
220 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
221 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
222 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
226 /* Just increment the non-pointer handle value */
227 static handle_t
*ext4_get_nojournal(void)
229 handle_t
*handle
= current
->journal_info
;
230 unsigned long ref_cnt
= (unsigned long)handle
;
232 BUG_ON(ref_cnt
>= EXT4_NOJOURNAL_MAX_REF_COUNT
);
235 handle
= (handle_t
*)ref_cnt
;
237 current
->journal_info
= handle
;
242 /* Decrement the non-pointer handle value */
243 static void ext4_put_nojournal(handle_t
*handle
)
245 unsigned long ref_cnt
= (unsigned long)handle
;
247 BUG_ON(ref_cnt
== 0);
250 handle
= (handle_t
*)ref_cnt
;
252 current
->journal_info
= handle
;
256 * Wrappers for jbd2_journal_start/end.
258 * The only special thing we need to do here is to make sure that all
259 * journal_end calls result in the superblock being marked dirty, so
260 * that sync() will call the filesystem's write_super callback if
263 * To avoid j_barrier hold in userspace when a user calls freeze(),
264 * ext4 prevents a new handle from being started by s_frozen, which
265 * is in an upper layer.
267 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
272 trace_ext4_journal_start(sb
, nblocks
, _RET_IP_
);
273 if (sb
->s_flags
& MS_RDONLY
)
274 return ERR_PTR(-EROFS
);
276 journal
= EXT4_SB(sb
)->s_journal
;
277 handle
= ext4_journal_current_handle();
280 * If a handle has been started, it should be allowed to
281 * finish, otherwise deadlock could happen between freeze
282 * and others(e.g. truncate) due to the restart of the
283 * journal handle if the filesystem is forzen and active
284 * handles are not stopped.
287 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
290 return ext4_get_nojournal();
292 * Special case here: if the journal has aborted behind our
293 * backs (eg. EIO in the commit thread), then we still need to
294 * take the FS itself readonly cleanly.
296 if (is_journal_aborted(journal
)) {
297 ext4_abort(sb
, "Detected aborted journal");
298 return ERR_PTR(-EROFS
);
300 return jbd2_journal_start(journal
, nblocks
);
304 * The only special thing we need to do here is to make sure that all
305 * jbd2_journal_stop calls result in the superblock being marked dirty, so
306 * that sync() will call the filesystem's write_super callback if
309 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
)
311 struct super_block
*sb
;
315 if (!ext4_handle_valid(handle
)) {
316 ext4_put_nojournal(handle
);
319 sb
= handle
->h_transaction
->t_journal
->j_private
;
321 rc
= jbd2_journal_stop(handle
);
326 __ext4_std_error(sb
, where
, line
, err
);
330 void ext4_journal_abort_handle(const char *caller
, unsigned int line
,
331 const char *err_fn
, struct buffer_head
*bh
,
332 handle_t
*handle
, int err
)
335 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
337 BUG_ON(!ext4_handle_valid(handle
));
340 BUFFER_TRACE(bh
, "abort");
345 if (is_handle_aborted(handle
))
348 printk(KERN_ERR
"%s:%d: aborting transaction: %s in %s\n",
349 caller
, line
, errstr
, err_fn
);
351 jbd2_journal_abort_handle(handle
);
354 static void __save_error_info(struct super_block
*sb
, const char *func
,
357 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
359 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
360 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
361 es
->s_last_error_time
= cpu_to_le32(get_seconds());
362 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
363 es
->s_last_error_line
= cpu_to_le32(line
);
364 if (!es
->s_first_error_time
) {
365 es
->s_first_error_time
= es
->s_last_error_time
;
366 strncpy(es
->s_first_error_func
, func
,
367 sizeof(es
->s_first_error_func
));
368 es
->s_first_error_line
= cpu_to_le32(line
);
369 es
->s_first_error_ino
= es
->s_last_error_ino
;
370 es
->s_first_error_block
= es
->s_last_error_block
;
373 * Start the daily error reporting function if it hasn't been
376 if (!es
->s_error_count
)
377 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
378 es
->s_error_count
= cpu_to_le32(le32_to_cpu(es
->s_error_count
) + 1);
381 static void save_error_info(struct super_block
*sb
, const char *func
,
384 __save_error_info(sb
, func
, line
);
385 ext4_commit_super(sb
, 1);
389 /* Deal with the reporting of failure conditions on a filesystem such as
390 * inconsistencies detected or read IO failures.
392 * On ext2, we can store the error state of the filesystem in the
393 * superblock. That is not possible on ext4, because we may have other
394 * write ordering constraints on the superblock which prevent us from
395 * writing it out straight away; and given that the journal is about to
396 * be aborted, we can't rely on the current, or future, transactions to
397 * write out the superblock safely.
399 * We'll just use the jbd2_journal_abort() error code to record an error in
400 * the journal instead. On recovery, the journal will complain about
401 * that error until we've noted it down and cleared it.
404 static void ext4_handle_error(struct super_block
*sb
)
406 if (sb
->s_flags
& MS_RDONLY
)
409 if (!test_opt(sb
, ERRORS_CONT
)) {
410 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
412 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
414 jbd2_journal_abort(journal
, -EIO
);
416 if (test_opt(sb
, ERRORS_RO
)) {
417 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
418 sb
->s_flags
|= MS_RDONLY
;
420 if (test_opt(sb
, ERRORS_PANIC
))
421 panic("EXT4-fs (device %s): panic forced after error\n",
425 void __ext4_error(struct super_block
*sb
, const char *function
,
426 unsigned int line
, const char *fmt
, ...)
428 struct va_format vaf
;
434 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
435 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
438 ext4_handle_error(sb
);
441 void ext4_error_inode(struct inode
*inode
, const char *function
,
442 unsigned int line
, ext4_fsblk_t block
,
443 const char *fmt
, ...)
446 struct va_format vaf
;
447 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
449 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
450 es
->s_last_error_block
= cpu_to_le64(block
);
451 save_error_info(inode
->i_sb
, function
, line
);
455 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: inode #%lu: ",
456 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
458 printk(KERN_CONT
"block %llu: ", block
);
459 printk(KERN_CONT
"comm %s: %pV\n", current
->comm
, &vaf
);
462 ext4_handle_error(inode
->i_sb
);
465 void ext4_error_file(struct file
*file
, const char *function
,
466 unsigned int line
, ext4_fsblk_t block
,
467 const char *fmt
, ...)
470 struct va_format vaf
;
471 struct ext4_super_block
*es
;
472 struct inode
*inode
= file
->f_dentry
->d_inode
;
473 char pathname
[80], *path
;
475 es
= EXT4_SB(inode
->i_sb
)->s_es
;
476 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
477 save_error_info(inode
->i_sb
, function
, line
);
478 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
482 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
483 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
485 printk(KERN_CONT
"block %llu: ", block
);
489 printk(KERN_CONT
"comm %s: path %s: %pV\n", current
->comm
, path
, &vaf
);
492 ext4_handle_error(inode
->i_sb
);
495 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
502 errstr
= "IO failure";
505 errstr
= "Out of memory";
508 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
509 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
510 errstr
= "Journal has aborted";
512 errstr
= "Readonly filesystem";
515 /* If the caller passed in an extra buffer for unknown
516 * errors, textualise them now. Else we just return
519 /* Check for truncated error codes... */
520 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
529 /* __ext4_std_error decodes expected errors from journaling functions
530 * automatically and invokes the appropriate error response. */
532 void __ext4_std_error(struct super_block
*sb
, const char *function
,
533 unsigned int line
, int errno
)
538 /* Special case: if the error is EROFS, and we're not already
539 * inside a transaction, then there's really no point in logging
541 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
542 (sb
->s_flags
& MS_RDONLY
))
545 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
546 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
547 sb
->s_id
, function
, line
, errstr
);
548 save_error_info(sb
, function
, line
);
550 ext4_handle_error(sb
);
554 * ext4_abort is a much stronger failure handler than ext4_error. The
555 * abort function may be used to deal with unrecoverable failures such
556 * as journal IO errors or ENOMEM at a critical moment in log management.
558 * We unconditionally force the filesystem into an ABORT|READONLY state,
559 * unless the error response on the fs has been set to panic in which
560 * case we take the easy way out and panic immediately.
563 void __ext4_abort(struct super_block
*sb
, const char *function
,
564 unsigned int line
, const char *fmt
, ...)
568 save_error_info(sb
, function
, line
);
570 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
576 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
577 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
578 sb
->s_flags
|= MS_RDONLY
;
579 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
580 if (EXT4_SB(sb
)->s_journal
)
581 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
582 save_error_info(sb
, function
, line
);
584 if (test_opt(sb
, ERRORS_PANIC
))
585 panic("EXT4-fs panic from previous error\n");
588 void ext4_msg(struct super_block
*sb
, const char *prefix
, const char *fmt
, ...)
590 struct va_format vaf
;
596 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
600 void __ext4_warning(struct super_block
*sb
, const char *function
,
601 unsigned int line
, const char *fmt
, ...)
603 struct va_format vaf
;
609 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
610 sb
->s_id
, function
, line
, &vaf
);
614 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
615 struct super_block
*sb
, ext4_group_t grp
,
616 unsigned long ino
, ext4_fsblk_t block
,
617 const char *fmt
, ...)
621 struct va_format vaf
;
623 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
625 es
->s_last_error_ino
= cpu_to_le32(ino
);
626 es
->s_last_error_block
= cpu_to_le64(block
);
627 __save_error_info(sb
, function
, line
);
633 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
634 sb
->s_id
, function
, line
, grp
);
636 printk(KERN_CONT
"inode %lu: ", ino
);
638 printk(KERN_CONT
"block %llu:", (unsigned long long) block
);
639 printk(KERN_CONT
"%pV\n", &vaf
);
642 if (test_opt(sb
, ERRORS_CONT
)) {
643 ext4_commit_super(sb
, 0);
647 ext4_unlock_group(sb
, grp
);
648 ext4_handle_error(sb
);
650 * We only get here in the ERRORS_RO case; relocking the group
651 * may be dangerous, but nothing bad will happen since the
652 * filesystem will have already been marked read/only and the
653 * journal has been aborted. We return 1 as a hint to callers
654 * who might what to use the return value from
655 * ext4_grp_locked_error() to distinguish between the
656 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
657 * aggressively from the ext4 function in question, with a
658 * more appropriate error code.
660 ext4_lock_group(sb
, grp
);
664 void ext4_update_dynamic_rev(struct super_block
*sb
)
666 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
668 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
672 "updating to rev %d because of new feature flag, "
673 "running e2fsck is recommended",
676 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
677 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
678 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
679 /* leave es->s_feature_*compat flags alone */
680 /* es->s_uuid will be set by e2fsck if empty */
683 * The rest of the superblock fields should be zero, and if not it
684 * means they are likely already in use, so leave them alone. We
685 * can leave it up to e2fsck to clean up any inconsistencies there.
690 * Open the external journal device
692 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
694 struct block_device
*bdev
;
695 char b
[BDEVNAME_SIZE
];
697 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
703 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
704 __bdevname(dev
, b
), PTR_ERR(bdev
));
709 * Release the journal device
711 static int ext4_blkdev_put(struct block_device
*bdev
)
713 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
716 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
718 struct block_device
*bdev
;
721 bdev
= sbi
->journal_bdev
;
723 ret
= ext4_blkdev_put(bdev
);
724 sbi
->journal_bdev
= NULL
;
729 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
731 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
734 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
738 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
739 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
741 printk(KERN_ERR
"sb_info orphan list:\n");
742 list_for_each(l
, &sbi
->s_orphan
) {
743 struct inode
*inode
= orphan_list_entry(l
);
745 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
746 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
747 inode
->i_mode
, inode
->i_nlink
,
752 static void ext4_put_super(struct super_block
*sb
)
754 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
755 struct ext4_super_block
*es
= sbi
->s_es
;
758 ext4_unregister_li_request(sb
);
759 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
761 flush_workqueue(sbi
->dio_unwritten_wq
);
762 destroy_workqueue(sbi
->dio_unwritten_wq
);
766 ext4_commit_super(sb
, 1);
768 if (sbi
->s_journal
) {
769 err
= jbd2_journal_destroy(sbi
->s_journal
);
770 sbi
->s_journal
= NULL
;
772 ext4_abort(sb
, "Couldn't clean up the journal");
775 del_timer(&sbi
->s_err_report
);
776 ext4_release_system_zone(sb
);
778 ext4_ext_release(sb
);
779 ext4_xattr_put_super(sb
);
781 if (!(sb
->s_flags
& MS_RDONLY
)) {
782 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
783 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
784 ext4_commit_super(sb
, 1);
787 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
789 kobject_del(&sbi
->s_kobj
);
791 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
792 brelse(sbi
->s_group_desc
[i
]);
793 kfree(sbi
->s_group_desc
);
794 if (is_vmalloc_addr(sbi
->s_flex_groups
))
795 vfree(sbi
->s_flex_groups
);
797 kfree(sbi
->s_flex_groups
);
798 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
799 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
800 percpu_counter_destroy(&sbi
->s_dirs_counter
);
801 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
804 for (i
= 0; i
< MAXQUOTAS
; i
++)
805 kfree(sbi
->s_qf_names
[i
]);
808 /* Debugging code just in case the in-memory inode orphan list
809 * isn't empty. The on-disk one can be non-empty if we've
810 * detected an error and taken the fs readonly, but the
811 * in-memory list had better be clean by this point. */
812 if (!list_empty(&sbi
->s_orphan
))
813 dump_orphan_list(sb
, sbi
);
814 J_ASSERT(list_empty(&sbi
->s_orphan
));
816 invalidate_bdev(sb
->s_bdev
);
817 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
819 * Invalidate the journal device's buffers. We don't want them
820 * floating about in memory - the physical journal device may
821 * hotswapped, and it breaks the `ro-after' testing code.
823 sync_blockdev(sbi
->journal_bdev
);
824 invalidate_bdev(sbi
->journal_bdev
);
825 ext4_blkdev_remove(sbi
);
828 kthread_stop(sbi
->s_mmp_tsk
);
829 sb
->s_fs_info
= NULL
;
831 * Now that we are completely done shutting down the
832 * superblock, we need to actually destroy the kobject.
835 kobject_put(&sbi
->s_kobj
);
836 wait_for_completion(&sbi
->s_kobj_unregister
);
837 kfree(sbi
->s_blockgroup_lock
);
841 static struct kmem_cache
*ext4_inode_cachep
;
844 * Called inside transaction, so use GFP_NOFS
846 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
848 struct ext4_inode_info
*ei
;
850 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
854 ei
->vfs_inode
.i_version
= 1;
855 ei
->vfs_inode
.i_data
.writeback_index
= 0;
856 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
857 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
858 spin_lock_init(&ei
->i_prealloc_lock
);
859 ei
->i_reserved_data_blocks
= 0;
860 ei
->i_reserved_meta_blocks
= 0;
861 ei
->i_allocated_meta_blocks
= 0;
862 ei
->i_da_metadata_calc_len
= 0;
863 spin_lock_init(&(ei
->i_block_reservation_lock
));
865 ei
->i_reserved_quota
= 0;
868 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
869 spin_lock_init(&ei
->i_completed_io_lock
);
870 ei
->cur_aio_dio
= NULL
;
872 ei
->i_datasync_tid
= 0;
873 atomic_set(&ei
->i_ioend_count
, 0);
874 atomic_set(&ei
->i_aiodio_unwritten
, 0);
876 return &ei
->vfs_inode
;
879 static int ext4_drop_inode(struct inode
*inode
)
881 int drop
= generic_drop_inode(inode
);
883 trace_ext4_drop_inode(inode
, drop
);
887 static void ext4_i_callback(struct rcu_head
*head
)
889 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
890 INIT_LIST_HEAD(&inode
->i_dentry
);
891 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
894 static void ext4_destroy_inode(struct inode
*inode
)
896 ext4_ioend_wait(inode
);
897 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
898 ext4_msg(inode
->i_sb
, KERN_ERR
,
899 "Inode %lu (%p): orphan list check failed!",
900 inode
->i_ino
, EXT4_I(inode
));
901 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
902 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
906 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
909 static void init_once(void *foo
)
911 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
913 INIT_LIST_HEAD(&ei
->i_orphan
);
914 #ifdef CONFIG_EXT4_FS_XATTR
915 init_rwsem(&ei
->xattr_sem
);
917 init_rwsem(&ei
->i_data_sem
);
918 inode_init_once(&ei
->vfs_inode
);
921 static int init_inodecache(void)
923 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
924 sizeof(struct ext4_inode_info
),
925 0, (SLAB_RECLAIM_ACCOUNT
|
928 if (ext4_inode_cachep
== NULL
)
933 static void destroy_inodecache(void)
935 kmem_cache_destroy(ext4_inode_cachep
);
938 void ext4_clear_inode(struct inode
*inode
)
940 invalidate_inode_buffers(inode
);
941 end_writeback(inode
);
943 ext4_discard_preallocations(inode
);
944 if (EXT4_I(inode
)->jinode
) {
945 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
946 EXT4_I(inode
)->jinode
);
947 jbd2_free_inode(EXT4_I(inode
)->jinode
);
948 EXT4_I(inode
)->jinode
= NULL
;
952 static inline void ext4_show_quota_options(struct seq_file
*seq
,
953 struct super_block
*sb
)
955 #if defined(CONFIG_QUOTA)
956 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
958 if (sbi
->s_jquota_fmt
) {
961 switch (sbi
->s_jquota_fmt
) {
972 seq_printf(seq
, ",jqfmt=%s", fmtname
);
975 if (sbi
->s_qf_names
[USRQUOTA
])
976 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
978 if (sbi
->s_qf_names
[GRPQUOTA
])
979 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
981 if (test_opt(sb
, USRQUOTA
))
982 seq_puts(seq
, ",usrquota");
984 if (test_opt(sb
, GRPQUOTA
))
985 seq_puts(seq
, ",grpquota");
991 * - it's set to a non-default value OR
992 * - if the per-sb default is different from the global default
994 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
997 unsigned long def_mount_opts
;
998 struct super_block
*sb
= vfs
->mnt_sb
;
999 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1000 struct ext4_super_block
*es
= sbi
->s_es
;
1002 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1003 def_errors
= le16_to_cpu(es
->s_errors
);
1005 if (sbi
->s_sb_block
!= 1)
1006 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
1007 if (test_opt(sb
, MINIX_DF
))
1008 seq_puts(seq
, ",minixdf");
1009 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
1010 seq_puts(seq
, ",grpid");
1011 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
1012 seq_puts(seq
, ",nogrpid");
1013 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
1014 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
1015 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
1017 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
1018 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
1019 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
1021 if (test_opt(sb
, ERRORS_RO
)) {
1022 if (def_errors
== EXT4_ERRORS_PANIC
||
1023 def_errors
== EXT4_ERRORS_CONTINUE
) {
1024 seq_puts(seq
, ",errors=remount-ro");
1027 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1028 seq_puts(seq
, ",errors=continue");
1029 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1030 seq_puts(seq
, ",errors=panic");
1031 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
1032 seq_puts(seq
, ",nouid32");
1033 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
1034 seq_puts(seq
, ",debug");
1035 if (test_opt(sb
, OLDALLOC
))
1036 seq_puts(seq
, ",oldalloc");
1037 #ifdef CONFIG_EXT4_FS_XATTR
1038 if (test_opt(sb
, XATTR_USER
))
1039 seq_puts(seq
, ",user_xattr");
1040 if (!test_opt(sb
, XATTR_USER
))
1041 seq_puts(seq
, ",nouser_xattr");
1043 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1044 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
1045 seq_puts(seq
, ",acl");
1046 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
1047 seq_puts(seq
, ",noacl");
1049 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
1050 seq_printf(seq
, ",commit=%u",
1051 (unsigned) (sbi
->s_commit_interval
/ HZ
));
1053 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
1054 seq_printf(seq
, ",min_batch_time=%u",
1055 (unsigned) sbi
->s_min_batch_time
);
1057 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
1058 seq_printf(seq
, ",max_batch_time=%u",
1059 (unsigned) sbi
->s_min_batch_time
);
1063 * We're changing the default of barrier mount option, so
1064 * let's always display its mount state so it's clear what its
1067 seq_puts(seq
, ",barrier=");
1068 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
1069 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
1070 seq_puts(seq
, ",journal_async_commit");
1071 else if (test_opt(sb
, JOURNAL_CHECKSUM
))
1072 seq_puts(seq
, ",journal_checksum");
1073 if (test_opt(sb
, I_VERSION
))
1074 seq_puts(seq
, ",i_version");
1075 if (!test_opt(sb
, DELALLOC
) &&
1076 !(def_mount_opts
& EXT4_DEFM_NODELALLOC
))
1077 seq_puts(seq
, ",nodelalloc");
1079 if (!test_opt(sb
, MBLK_IO_SUBMIT
))
1080 seq_puts(seq
, ",nomblk_io_submit");
1082 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
1084 * journal mode get enabled in different ways
1085 * So just print the value even if we didn't specify it
1087 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1088 seq_puts(seq
, ",data=journal");
1089 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1090 seq_puts(seq
, ",data=ordered");
1091 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1092 seq_puts(seq
, ",data=writeback");
1094 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1095 seq_printf(seq
, ",inode_readahead_blks=%u",
1096 sbi
->s_inode_readahead_blks
);
1098 if (test_opt(sb
, DATA_ERR_ABORT
))
1099 seq_puts(seq
, ",data_err=abort");
1101 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
1102 seq_puts(seq
, ",noauto_da_alloc");
1104 if (test_opt(sb
, DISCARD
) && !(def_mount_opts
& EXT4_DEFM_DISCARD
))
1105 seq_puts(seq
, ",discard");
1107 if (test_opt(sb
, NOLOAD
))
1108 seq_puts(seq
, ",norecovery");
1110 if (test_opt(sb
, DIOREAD_NOLOCK
))
1111 seq_puts(seq
, ",dioread_nolock");
1113 if (test_opt(sb
, BLOCK_VALIDITY
) &&
1114 !(def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
))
1115 seq_puts(seq
, ",block_validity");
1117 if (!test_opt(sb
, INIT_INODE_TABLE
))
1118 seq_puts(seq
, ",noinit_inode_table");
1119 else if (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)
1120 seq_printf(seq
, ",init_inode_table=%u",
1121 (unsigned) sbi
->s_li_wait_mult
);
1123 ext4_show_quota_options(seq
, sb
);
1128 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1129 u64 ino
, u32 generation
)
1131 struct inode
*inode
;
1133 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1134 return ERR_PTR(-ESTALE
);
1135 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1136 return ERR_PTR(-ESTALE
);
1138 /* iget isn't really right if the inode is currently unallocated!!
1140 * ext4_read_inode will return a bad_inode if the inode had been
1141 * deleted, so we should be safe.
1143 * Currently we don't know the generation for parent directory, so
1144 * a generation of 0 means "accept any"
1146 inode
= ext4_iget(sb
, ino
);
1148 return ERR_CAST(inode
);
1149 if (generation
&& inode
->i_generation
!= generation
) {
1151 return ERR_PTR(-ESTALE
);
1157 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1158 int fh_len
, int fh_type
)
1160 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1161 ext4_nfs_get_inode
);
1164 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1165 int fh_len
, int fh_type
)
1167 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1168 ext4_nfs_get_inode
);
1172 * Try to release metadata pages (indirect blocks, directories) which are
1173 * mapped via the block device. Since these pages could have journal heads
1174 * which would prevent try_to_free_buffers() from freeing them, we must use
1175 * jbd2 layer's try_to_free_buffers() function to release them.
1177 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1180 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1182 WARN_ON(PageChecked(page
));
1183 if (!page_has_buffers(page
))
1186 return jbd2_journal_try_to_free_buffers(journal
, page
,
1187 wait
& ~__GFP_WAIT
);
1188 return try_to_free_buffers(page
);
1192 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1193 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1195 static int ext4_write_dquot(struct dquot
*dquot
);
1196 static int ext4_acquire_dquot(struct dquot
*dquot
);
1197 static int ext4_release_dquot(struct dquot
*dquot
);
1198 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1199 static int ext4_write_info(struct super_block
*sb
, int type
);
1200 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1202 static int ext4_quota_off(struct super_block
*sb
, int type
);
1203 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1204 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1205 size_t len
, loff_t off
);
1206 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1207 const char *data
, size_t len
, loff_t off
);
1209 static const struct dquot_operations ext4_quota_operations
= {
1210 .get_reserved_space
= ext4_get_reserved_space
,
1211 .write_dquot
= ext4_write_dquot
,
1212 .acquire_dquot
= ext4_acquire_dquot
,
1213 .release_dquot
= ext4_release_dquot
,
1214 .mark_dirty
= ext4_mark_dquot_dirty
,
1215 .write_info
= ext4_write_info
,
1216 .alloc_dquot
= dquot_alloc
,
1217 .destroy_dquot
= dquot_destroy
,
1220 static const struct quotactl_ops ext4_qctl_operations
= {
1221 .quota_on
= ext4_quota_on
,
1222 .quota_off
= ext4_quota_off
,
1223 .quota_sync
= dquot_quota_sync
,
1224 .get_info
= dquot_get_dqinfo
,
1225 .set_info
= dquot_set_dqinfo
,
1226 .get_dqblk
= dquot_get_dqblk
,
1227 .set_dqblk
= dquot_set_dqblk
1231 static const struct super_operations ext4_sops
= {
1232 .alloc_inode
= ext4_alloc_inode
,
1233 .destroy_inode
= ext4_destroy_inode
,
1234 .write_inode
= ext4_write_inode
,
1235 .dirty_inode
= ext4_dirty_inode
,
1236 .drop_inode
= ext4_drop_inode
,
1237 .evict_inode
= ext4_evict_inode
,
1238 .put_super
= ext4_put_super
,
1239 .sync_fs
= ext4_sync_fs
,
1240 .freeze_fs
= ext4_freeze
,
1241 .unfreeze_fs
= ext4_unfreeze
,
1242 .statfs
= ext4_statfs
,
1243 .remount_fs
= ext4_remount
,
1244 .show_options
= ext4_show_options
,
1246 .quota_read
= ext4_quota_read
,
1247 .quota_write
= ext4_quota_write
,
1249 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1252 static const struct super_operations ext4_nojournal_sops
= {
1253 .alloc_inode
= ext4_alloc_inode
,
1254 .destroy_inode
= ext4_destroy_inode
,
1255 .write_inode
= ext4_write_inode
,
1256 .dirty_inode
= ext4_dirty_inode
,
1257 .drop_inode
= ext4_drop_inode
,
1258 .evict_inode
= ext4_evict_inode
,
1259 .write_super
= ext4_write_super
,
1260 .put_super
= ext4_put_super
,
1261 .statfs
= ext4_statfs
,
1262 .remount_fs
= ext4_remount
,
1263 .show_options
= ext4_show_options
,
1265 .quota_read
= ext4_quota_read
,
1266 .quota_write
= ext4_quota_write
,
1268 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1271 static const struct export_operations ext4_export_ops
= {
1272 .fh_to_dentry
= ext4_fh_to_dentry
,
1273 .fh_to_parent
= ext4_fh_to_parent
,
1274 .get_parent
= ext4_get_parent
,
1278 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1279 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1280 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1281 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1282 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1283 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1284 Opt_journal_update
, Opt_journal_dev
,
1285 Opt_journal_checksum
, Opt_journal_async_commit
,
1286 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1287 Opt_data_err_abort
, Opt_data_err_ignore
,
1288 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1289 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1290 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1291 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1292 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1293 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1294 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1295 Opt_dioread_nolock
, Opt_dioread_lock
,
1296 Opt_discard
, Opt_nodiscard
,
1297 Opt_init_inode_table
, Opt_noinit_inode_table
,
1300 static const match_table_t tokens
= {
1301 {Opt_bsd_df
, "bsddf"},
1302 {Opt_minix_df
, "minixdf"},
1303 {Opt_grpid
, "grpid"},
1304 {Opt_grpid
, "bsdgroups"},
1305 {Opt_nogrpid
, "nogrpid"},
1306 {Opt_nogrpid
, "sysvgroups"},
1307 {Opt_resgid
, "resgid=%u"},
1308 {Opt_resuid
, "resuid=%u"},
1310 {Opt_err_cont
, "errors=continue"},
1311 {Opt_err_panic
, "errors=panic"},
1312 {Opt_err_ro
, "errors=remount-ro"},
1313 {Opt_nouid32
, "nouid32"},
1314 {Opt_debug
, "debug"},
1315 {Opt_oldalloc
, "oldalloc"},
1316 {Opt_orlov
, "orlov"},
1317 {Opt_user_xattr
, "user_xattr"},
1318 {Opt_nouser_xattr
, "nouser_xattr"},
1320 {Opt_noacl
, "noacl"},
1321 {Opt_noload
, "noload"},
1322 {Opt_noload
, "norecovery"},
1325 {Opt_commit
, "commit=%u"},
1326 {Opt_min_batch_time
, "min_batch_time=%u"},
1327 {Opt_max_batch_time
, "max_batch_time=%u"},
1328 {Opt_journal_update
, "journal=update"},
1329 {Opt_journal_dev
, "journal_dev=%u"},
1330 {Opt_journal_checksum
, "journal_checksum"},
1331 {Opt_journal_async_commit
, "journal_async_commit"},
1332 {Opt_abort
, "abort"},
1333 {Opt_data_journal
, "data=journal"},
1334 {Opt_data_ordered
, "data=ordered"},
1335 {Opt_data_writeback
, "data=writeback"},
1336 {Opt_data_err_abort
, "data_err=abort"},
1337 {Opt_data_err_ignore
, "data_err=ignore"},
1338 {Opt_offusrjquota
, "usrjquota="},
1339 {Opt_usrjquota
, "usrjquota=%s"},
1340 {Opt_offgrpjquota
, "grpjquota="},
1341 {Opt_grpjquota
, "grpjquota=%s"},
1342 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1343 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1344 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1345 {Opt_grpquota
, "grpquota"},
1346 {Opt_noquota
, "noquota"},
1347 {Opt_quota
, "quota"},
1348 {Opt_usrquota
, "usrquota"},
1349 {Opt_barrier
, "barrier=%u"},
1350 {Opt_barrier
, "barrier"},
1351 {Opt_nobarrier
, "nobarrier"},
1352 {Opt_i_version
, "i_version"},
1353 {Opt_stripe
, "stripe=%u"},
1354 {Opt_resize
, "resize"},
1355 {Opt_delalloc
, "delalloc"},
1356 {Opt_nodelalloc
, "nodelalloc"},
1357 {Opt_mblk_io_submit
, "mblk_io_submit"},
1358 {Opt_nomblk_io_submit
, "nomblk_io_submit"},
1359 {Opt_block_validity
, "block_validity"},
1360 {Opt_noblock_validity
, "noblock_validity"},
1361 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1362 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1363 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1364 {Opt_auto_da_alloc
, "auto_da_alloc"},
1365 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1366 {Opt_dioread_nolock
, "dioread_nolock"},
1367 {Opt_dioread_lock
, "dioread_lock"},
1368 {Opt_discard
, "discard"},
1369 {Opt_nodiscard
, "nodiscard"},
1370 {Opt_init_inode_table
, "init_itable=%u"},
1371 {Opt_init_inode_table
, "init_itable"},
1372 {Opt_noinit_inode_table
, "noinit_itable"},
1376 static ext4_fsblk_t
get_sb_block(void **data
)
1378 ext4_fsblk_t sb_block
;
1379 char *options
= (char *) *data
;
1381 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1382 return 1; /* Default location */
1385 /* TODO: use simple_strtoll with >32bit ext4 */
1386 sb_block
= simple_strtoul(options
, &options
, 0);
1387 if (*options
&& *options
!= ',') {
1388 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1392 if (*options
== ',')
1394 *data
= (void *) options
;
1399 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1400 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1401 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1404 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1406 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1409 if (sb_any_quota_loaded(sb
) &&
1410 !sbi
->s_qf_names
[qtype
]) {
1411 ext4_msg(sb
, KERN_ERR
,
1412 "Cannot change journaled "
1413 "quota options when quota turned on");
1416 qname
= match_strdup(args
);
1418 ext4_msg(sb
, KERN_ERR
,
1419 "Not enough memory for storing quotafile name");
1422 if (sbi
->s_qf_names
[qtype
] &&
1423 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1424 ext4_msg(sb
, KERN_ERR
,
1425 "%s quota file already specified", QTYPE2NAME(qtype
));
1429 sbi
->s_qf_names
[qtype
] = qname
;
1430 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1431 ext4_msg(sb
, KERN_ERR
,
1432 "quotafile must be on filesystem root");
1433 kfree(sbi
->s_qf_names
[qtype
]);
1434 sbi
->s_qf_names
[qtype
] = NULL
;
1441 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1444 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1446 if (sb_any_quota_loaded(sb
) &&
1447 sbi
->s_qf_names
[qtype
]) {
1448 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1449 " when quota turned on");
1453 * The space will be released later when all options are confirmed
1456 sbi
->s_qf_names
[qtype
] = NULL
;
1461 static int parse_options(char *options
, struct super_block
*sb
,
1462 unsigned long *journal_devnum
,
1463 unsigned int *journal_ioprio
,
1464 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1466 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1468 substring_t args
[MAX_OPT_ARGS
];
1478 while ((p
= strsep(&options
, ",")) != NULL
) {
1484 * Initialize args struct so we know whether arg was
1485 * found; some options take optional arguments.
1487 args
[0].to
= args
[0].from
= NULL
;
1488 token
= match_token(p
, tokens
, args
);
1491 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1492 clear_opt(sb
, MINIX_DF
);
1495 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1496 set_opt(sb
, MINIX_DF
);
1500 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1505 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1506 clear_opt(sb
, GRPID
);
1510 if (match_int(&args
[0], &option
))
1512 sbi
->s_resuid
= option
;
1515 if (match_int(&args
[0], &option
))
1517 sbi
->s_resgid
= option
;
1520 /* handled by get_sb_block() instead of here */
1521 /* *sb_block = match_int(&args[0]); */
1524 clear_opt(sb
, ERRORS_CONT
);
1525 clear_opt(sb
, ERRORS_RO
);
1526 set_opt(sb
, ERRORS_PANIC
);
1529 clear_opt(sb
, ERRORS_CONT
);
1530 clear_opt(sb
, ERRORS_PANIC
);
1531 set_opt(sb
, ERRORS_RO
);
1534 clear_opt(sb
, ERRORS_RO
);
1535 clear_opt(sb
, ERRORS_PANIC
);
1536 set_opt(sb
, ERRORS_CONT
);
1539 set_opt(sb
, NO_UID32
);
1545 set_opt(sb
, OLDALLOC
);
1548 clear_opt(sb
, OLDALLOC
);
1550 #ifdef CONFIG_EXT4_FS_XATTR
1551 case Opt_user_xattr
:
1552 set_opt(sb
, XATTR_USER
);
1554 case Opt_nouser_xattr
:
1555 clear_opt(sb
, XATTR_USER
);
1558 case Opt_user_xattr
:
1559 case Opt_nouser_xattr
:
1560 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1563 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1565 set_opt(sb
, POSIX_ACL
);
1568 clear_opt(sb
, POSIX_ACL
);
1573 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1576 case Opt_journal_update
:
1578 /* Eventually we will want to be able to create
1579 a journal file here. For now, only allow the
1580 user to specify an existing inode to be the
1583 ext4_msg(sb
, KERN_ERR
,
1584 "Cannot specify journal on remount");
1587 set_opt(sb
, UPDATE_JOURNAL
);
1589 case Opt_journal_dev
:
1591 ext4_msg(sb
, KERN_ERR
,
1592 "Cannot specify journal on remount");
1595 if (match_int(&args
[0], &option
))
1597 *journal_devnum
= option
;
1599 case Opt_journal_checksum
:
1600 set_opt(sb
, JOURNAL_CHECKSUM
);
1602 case Opt_journal_async_commit
:
1603 set_opt(sb
, JOURNAL_ASYNC_COMMIT
);
1604 set_opt(sb
, JOURNAL_CHECKSUM
);
1607 set_opt(sb
, NOLOAD
);
1610 if (match_int(&args
[0], &option
))
1615 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1616 sbi
->s_commit_interval
= HZ
* option
;
1618 case Opt_max_batch_time
:
1619 if (match_int(&args
[0], &option
))
1624 option
= EXT4_DEF_MAX_BATCH_TIME
;
1625 sbi
->s_max_batch_time
= option
;
1627 case Opt_min_batch_time
:
1628 if (match_int(&args
[0], &option
))
1632 sbi
->s_min_batch_time
= option
;
1634 case Opt_data_journal
:
1635 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1637 case Opt_data_ordered
:
1638 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1640 case Opt_data_writeback
:
1641 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1644 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1645 ext4_msg(sb
, KERN_ERR
,
1646 "Cannot change data mode on remount");
1650 clear_opt(sb
, DATA_FLAGS
);
1651 sbi
->s_mount_opt
|= data_opt
;
1654 case Opt_data_err_abort
:
1655 set_opt(sb
, DATA_ERR_ABORT
);
1657 case Opt_data_err_ignore
:
1658 clear_opt(sb
, DATA_ERR_ABORT
);
1662 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1666 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1669 case Opt_offusrjquota
:
1670 if (!clear_qf_name(sb
, USRQUOTA
))
1673 case Opt_offgrpjquota
:
1674 if (!clear_qf_name(sb
, GRPQUOTA
))
1678 case Opt_jqfmt_vfsold
:
1679 qfmt
= QFMT_VFS_OLD
;
1681 case Opt_jqfmt_vfsv0
:
1684 case Opt_jqfmt_vfsv1
:
1687 if (sb_any_quota_loaded(sb
) &&
1688 sbi
->s_jquota_fmt
!= qfmt
) {
1689 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1690 "journaled quota options when "
1694 sbi
->s_jquota_fmt
= qfmt
;
1699 set_opt(sb
, USRQUOTA
);
1703 set_opt(sb
, GRPQUOTA
);
1706 if (sb_any_quota_loaded(sb
)) {
1707 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1708 "options when quota turned on");
1711 clear_opt(sb
, QUOTA
);
1712 clear_opt(sb
, USRQUOTA
);
1713 clear_opt(sb
, GRPQUOTA
);
1719 ext4_msg(sb
, KERN_ERR
,
1720 "quota options not supported");
1724 case Opt_offusrjquota
:
1725 case Opt_offgrpjquota
:
1726 case Opt_jqfmt_vfsold
:
1727 case Opt_jqfmt_vfsv0
:
1728 case Opt_jqfmt_vfsv1
:
1729 ext4_msg(sb
, KERN_ERR
,
1730 "journaled quota options not supported");
1736 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1739 clear_opt(sb
, BARRIER
);
1743 if (match_int(&args
[0], &option
))
1746 option
= 1; /* No argument, default to 1 */
1748 set_opt(sb
, BARRIER
);
1750 clear_opt(sb
, BARRIER
);
1756 ext4_msg(sb
, KERN_ERR
,
1757 "resize option only available "
1761 if (match_int(&args
[0], &option
) != 0)
1763 *n_blocks_count
= option
;
1766 ext4_msg(sb
, KERN_WARNING
,
1767 "Ignoring deprecated nobh option");
1770 ext4_msg(sb
, KERN_WARNING
,
1771 "Ignoring deprecated bh option");
1774 set_opt(sb
, I_VERSION
);
1775 sb
->s_flags
|= MS_I_VERSION
;
1777 case Opt_nodelalloc
:
1778 clear_opt(sb
, DELALLOC
);
1780 case Opt_mblk_io_submit
:
1781 set_opt(sb
, MBLK_IO_SUBMIT
);
1783 case Opt_nomblk_io_submit
:
1784 clear_opt(sb
, MBLK_IO_SUBMIT
);
1787 if (match_int(&args
[0], &option
))
1791 sbi
->s_stripe
= option
;
1794 set_opt(sb
, DELALLOC
);
1796 case Opt_block_validity
:
1797 set_opt(sb
, BLOCK_VALIDITY
);
1799 case Opt_noblock_validity
:
1800 clear_opt(sb
, BLOCK_VALIDITY
);
1802 case Opt_inode_readahead_blks
:
1803 if (match_int(&args
[0], &option
))
1805 if (option
< 0 || option
> (1 << 30))
1807 if (option
&& !is_power_of_2(option
)) {
1808 ext4_msg(sb
, KERN_ERR
,
1809 "EXT4-fs: inode_readahead_blks"
1810 " must be a power of 2");
1813 sbi
->s_inode_readahead_blks
= option
;
1815 case Opt_journal_ioprio
:
1816 if (match_int(&args
[0], &option
))
1818 if (option
< 0 || option
> 7)
1820 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1823 case Opt_noauto_da_alloc
:
1824 set_opt(sb
, NO_AUTO_DA_ALLOC
);
1826 case Opt_auto_da_alloc
:
1828 if (match_int(&args
[0], &option
))
1831 option
= 1; /* No argument, default to 1 */
1833 clear_opt(sb
, NO_AUTO_DA_ALLOC
);
1835 set_opt(sb
,NO_AUTO_DA_ALLOC
);
1838 set_opt(sb
, DISCARD
);
1841 clear_opt(sb
, DISCARD
);
1843 case Opt_dioread_nolock
:
1844 set_opt(sb
, DIOREAD_NOLOCK
);
1846 case Opt_dioread_lock
:
1847 clear_opt(sb
, DIOREAD_NOLOCK
);
1849 case Opt_init_inode_table
:
1850 set_opt(sb
, INIT_INODE_TABLE
);
1852 if (match_int(&args
[0], &option
))
1855 option
= EXT4_DEF_LI_WAIT_MULT
;
1858 sbi
->s_li_wait_mult
= option
;
1860 case Opt_noinit_inode_table
:
1861 clear_opt(sb
, INIT_INODE_TABLE
);
1864 ext4_msg(sb
, KERN_ERR
,
1865 "Unrecognized mount option \"%s\" "
1866 "or missing value", p
);
1871 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1872 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1873 clear_opt(sb
, USRQUOTA
);
1875 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1876 clear_opt(sb
, GRPQUOTA
);
1878 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1879 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1884 if (!sbi
->s_jquota_fmt
) {
1885 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1890 if (sbi
->s_jquota_fmt
) {
1891 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1892 "specified with no journaling "
1901 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1904 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1907 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1908 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1909 "forcing read-only mode");
1914 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1915 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1916 "running e2fsck is recommended");
1917 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1918 ext4_msg(sb
, KERN_WARNING
,
1919 "warning: mounting fs with errors, "
1920 "running e2fsck is recommended");
1921 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1922 le16_to_cpu(es
->s_mnt_count
) >=
1923 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1924 ext4_msg(sb
, KERN_WARNING
,
1925 "warning: maximal mount count reached, "
1926 "running e2fsck is recommended");
1927 else if (le32_to_cpu(es
->s_checkinterval
) &&
1928 (le32_to_cpu(es
->s_lastcheck
) +
1929 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1930 ext4_msg(sb
, KERN_WARNING
,
1931 "warning: checktime reached, "
1932 "running e2fsck is recommended");
1933 if (!sbi
->s_journal
)
1934 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1935 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1936 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1937 le16_add_cpu(&es
->s_mnt_count
, 1);
1938 es
->s_mtime
= cpu_to_le32(get_seconds());
1939 ext4_update_dynamic_rev(sb
);
1941 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1943 ext4_commit_super(sb
, 1);
1944 if (test_opt(sb
, DEBUG
))
1945 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1946 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1948 sbi
->s_groups_count
,
1949 EXT4_BLOCKS_PER_GROUP(sb
),
1950 EXT4_INODES_PER_GROUP(sb
),
1951 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1953 cleancache_init_fs(sb
);
1957 static int ext4_fill_flex_info(struct super_block
*sb
)
1959 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1960 struct ext4_group_desc
*gdp
= NULL
;
1961 ext4_group_t flex_group_count
;
1962 ext4_group_t flex_group
;
1963 int groups_per_flex
= 0;
1967 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1968 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1970 if (groups_per_flex
< 2) {
1971 sbi
->s_log_groups_per_flex
= 0;
1975 /* We allocate both existing and potentially added groups */
1976 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1977 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1978 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1979 size
= flex_group_count
* sizeof(struct flex_groups
);
1980 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1981 if (sbi
->s_flex_groups
== NULL
) {
1982 sbi
->s_flex_groups
= vzalloc(size
);
1983 if (sbi
->s_flex_groups
== NULL
) {
1984 ext4_msg(sb
, KERN_ERR
,
1985 "not enough memory for %u flex groups",
1991 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1992 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1994 flex_group
= ext4_flex_group(sbi
, i
);
1995 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1996 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1997 atomic_add(ext4_free_blks_count(sb
, gdp
),
1998 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1999 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2000 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2008 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
2009 struct ext4_group_desc
*gdp
)
2013 if (sbi
->s_es
->s_feature_ro_compat
&
2014 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
2015 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2016 __le32 le_group
= cpu_to_le32(block_group
);
2018 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2019 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2020 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2021 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2022 /* for checksum of struct ext4_group_desc do the rest...*/
2023 if ((sbi
->s_es
->s_feature_incompat
&
2024 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
2025 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2026 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2027 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2031 return cpu_to_le16(crc
);
2034 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
2035 struct ext4_group_desc
*gdp
)
2037 if ((sbi
->s_es
->s_feature_ro_compat
&
2038 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
2039 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
2045 /* Called at mount-time, super-block is locked */
2046 static int ext4_check_descriptors(struct super_block
*sb
,
2047 ext4_group_t
*first_not_zeroed
)
2049 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2050 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2051 ext4_fsblk_t last_block
;
2052 ext4_fsblk_t block_bitmap
;
2053 ext4_fsblk_t inode_bitmap
;
2054 ext4_fsblk_t inode_table
;
2055 int flexbg_flag
= 0;
2056 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2058 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2061 ext4_debug("Checking group descriptors");
2063 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2064 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2066 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2067 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2069 last_block
= first_block
+
2070 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2072 if ((grp
== sbi
->s_groups_count
) &&
2073 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2076 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2077 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2078 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2079 "Block bitmap for group %u not in group "
2080 "(block %llu)!", i
, block_bitmap
);
2083 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2084 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2085 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2086 "Inode bitmap for group %u not in group "
2087 "(block %llu)!", i
, inode_bitmap
);
2090 inode_table
= ext4_inode_table(sb
, gdp
);
2091 if (inode_table
< first_block
||
2092 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2093 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2094 "Inode table for group %u not in group "
2095 "(block %llu)!", i
, inode_table
);
2098 ext4_lock_group(sb
, i
);
2099 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
2100 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2101 "Checksum for group %u failed (%u!=%u)",
2102 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2103 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2104 if (!(sb
->s_flags
& MS_RDONLY
)) {
2105 ext4_unlock_group(sb
, i
);
2109 ext4_unlock_group(sb
, i
);
2111 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2113 if (NULL
!= first_not_zeroed
)
2114 *first_not_zeroed
= grp
;
2116 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
2117 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2121 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2122 * the superblock) which were deleted from all directories, but held open by
2123 * a process at the time of a crash. We walk the list and try to delete these
2124 * inodes at recovery time (only with a read-write filesystem).
2126 * In order to keep the orphan inode chain consistent during traversal (in
2127 * case of crash during recovery), we link each inode into the superblock
2128 * orphan list_head and handle it the same way as an inode deletion during
2129 * normal operation (which journals the operations for us).
2131 * We only do an iget() and an iput() on each inode, which is very safe if we
2132 * accidentally point at an in-use or already deleted inode. The worst that
2133 * can happen in this case is that we get a "bit already cleared" message from
2134 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2135 * e2fsck was run on this filesystem, and it must have already done the orphan
2136 * inode cleanup for us, so we can safely abort without any further action.
2138 static void ext4_orphan_cleanup(struct super_block
*sb
,
2139 struct ext4_super_block
*es
)
2141 unsigned int s_flags
= sb
->s_flags
;
2142 int nr_orphans
= 0, nr_truncates
= 0;
2146 if (!es
->s_last_orphan
) {
2147 jbd_debug(4, "no orphan inodes to clean up\n");
2151 if (bdev_read_only(sb
->s_bdev
)) {
2152 ext4_msg(sb
, KERN_ERR
, "write access "
2153 "unavailable, skipping orphan cleanup");
2157 /* Check if feature set would not allow a r/w mount */
2158 if (!ext4_feature_set_ok(sb
, 0)) {
2159 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2160 "unknown ROCOMPAT features");
2164 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2165 if (es
->s_last_orphan
)
2166 jbd_debug(1, "Errors on filesystem, "
2167 "clearing orphan list.\n");
2168 es
->s_last_orphan
= 0;
2169 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2173 if (s_flags
& MS_RDONLY
) {
2174 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2175 sb
->s_flags
&= ~MS_RDONLY
;
2178 /* Needed for iput() to work correctly and not trash data */
2179 sb
->s_flags
|= MS_ACTIVE
;
2180 /* Turn on quotas so that they are updated correctly */
2181 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2182 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2183 int ret
= ext4_quota_on_mount(sb
, i
);
2185 ext4_msg(sb
, KERN_ERR
,
2186 "Cannot turn on journaled "
2187 "quota: error %d", ret
);
2192 while (es
->s_last_orphan
) {
2193 struct inode
*inode
;
2195 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2196 if (IS_ERR(inode
)) {
2197 es
->s_last_orphan
= 0;
2201 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2202 dquot_initialize(inode
);
2203 if (inode
->i_nlink
) {
2204 ext4_msg(sb
, KERN_DEBUG
,
2205 "%s: truncating inode %lu to %lld bytes",
2206 __func__
, inode
->i_ino
, inode
->i_size
);
2207 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2208 inode
->i_ino
, inode
->i_size
);
2209 ext4_truncate(inode
);
2212 ext4_msg(sb
, KERN_DEBUG
,
2213 "%s: deleting unreferenced inode %lu",
2214 __func__
, inode
->i_ino
);
2215 jbd_debug(2, "deleting unreferenced inode %lu\n",
2219 iput(inode
); /* The delete magic happens here! */
2222 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2225 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2226 PLURAL(nr_orphans
));
2228 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2229 PLURAL(nr_truncates
));
2231 /* Turn quotas off */
2232 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2233 if (sb_dqopt(sb
)->files
[i
])
2234 dquot_quota_off(sb
, i
);
2237 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2241 * Maximal extent format file size.
2242 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2243 * extent format containers, within a sector_t, and within i_blocks
2244 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2245 * so that won't be a limiting factor.
2247 * However there is other limiting factor. We do store extents in the form
2248 * of starting block and length, hence the resulting length of the extent
2249 * covering maximum file size must fit into on-disk format containers as
2250 * well. Given that length is always by 1 unit bigger than max unit (because
2251 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2253 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2255 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2258 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2260 /* small i_blocks in vfs inode? */
2261 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2263 * CONFIG_LBDAF is not enabled implies the inode
2264 * i_block represent total blocks in 512 bytes
2265 * 32 == size of vfs inode i_blocks * 8
2267 upper_limit
= (1LL << 32) - 1;
2269 /* total blocks in file system block size */
2270 upper_limit
>>= (blkbits
- 9);
2271 upper_limit
<<= blkbits
;
2275 * 32-bit extent-start container, ee_block. We lower the maxbytes
2276 * by one fs block, so ee_len can cover the extent of maximum file
2279 res
= (1LL << 32) - 1;
2282 /* Sanity check against vm- & vfs- imposed limits */
2283 if (res
> upper_limit
)
2290 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2291 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2292 * We need to be 1 filesystem block less than the 2^48 sector limit.
2294 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2296 loff_t res
= EXT4_NDIR_BLOCKS
;
2299 /* This is calculated to be the largest file size for a dense, block
2300 * mapped file such that the file's total number of 512-byte sectors,
2301 * including data and all indirect blocks, does not exceed (2^48 - 1).
2303 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2304 * number of 512-byte sectors of the file.
2307 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2309 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2310 * the inode i_block field represents total file blocks in
2311 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2313 upper_limit
= (1LL << 32) - 1;
2315 /* total blocks in file system block size */
2316 upper_limit
>>= (bits
- 9);
2320 * We use 48 bit ext4_inode i_blocks
2321 * With EXT4_HUGE_FILE_FL set the i_blocks
2322 * represent total number of blocks in
2323 * file system block size
2325 upper_limit
= (1LL << 48) - 1;
2329 /* indirect blocks */
2331 /* double indirect blocks */
2332 meta_blocks
+= 1 + (1LL << (bits
-2));
2333 /* tripple indirect blocks */
2334 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2336 upper_limit
-= meta_blocks
;
2337 upper_limit
<<= bits
;
2339 res
+= 1LL << (bits
-2);
2340 res
+= 1LL << (2*(bits
-2));
2341 res
+= 1LL << (3*(bits
-2));
2343 if (res
> upper_limit
)
2346 if (res
> MAX_LFS_FILESIZE
)
2347 res
= MAX_LFS_FILESIZE
;
2352 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2353 ext4_fsblk_t logical_sb_block
, int nr
)
2355 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2356 ext4_group_t bg
, first_meta_bg
;
2359 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2361 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2363 return logical_sb_block
+ nr
+ 1;
2364 bg
= sbi
->s_desc_per_block
* nr
;
2365 if (ext4_bg_has_super(sb
, bg
))
2368 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2372 * ext4_get_stripe_size: Get the stripe size.
2373 * @sbi: In memory super block info
2375 * If we have specified it via mount option, then
2376 * use the mount option value. If the value specified at mount time is
2377 * greater than the blocks per group use the super block value.
2378 * If the super block value is greater than blocks per group return 0.
2379 * Allocator needs it be less than blocks per group.
2382 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2384 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2385 unsigned long stripe_width
=
2386 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2389 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2390 ret
= sbi
->s_stripe
;
2391 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2393 else if (stride
<= sbi
->s_blocks_per_group
)
2399 * If the stripe width is 1, this makes no sense and
2400 * we set it to 0 to turn off stripe handling code.
2411 struct attribute attr
;
2412 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2413 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2414 const char *, size_t);
2418 static int parse_strtoul(const char *buf
,
2419 unsigned long max
, unsigned long *value
)
2423 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2424 endp
= skip_spaces(endp
);
2425 if (*endp
|| *value
> max
)
2431 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2432 struct ext4_sb_info
*sbi
,
2435 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2436 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2439 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2440 struct ext4_sb_info
*sbi
, char *buf
)
2442 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2444 if (!sb
->s_bdev
->bd_part
)
2445 return snprintf(buf
, PAGE_SIZE
, "0\n");
2446 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2447 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2448 sbi
->s_sectors_written_start
) >> 1);
2451 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2452 struct ext4_sb_info
*sbi
, char *buf
)
2454 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2456 if (!sb
->s_bdev
->bd_part
)
2457 return snprintf(buf
, PAGE_SIZE
, "0\n");
2458 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2459 (unsigned long long)(sbi
->s_kbytes_written
+
2460 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2461 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2464 static ssize_t
extent_cache_hits_show(struct ext4_attr
*a
,
2465 struct ext4_sb_info
*sbi
, char *buf
)
2467 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_hits
);
2470 static ssize_t
extent_cache_misses_show(struct ext4_attr
*a
,
2471 struct ext4_sb_info
*sbi
, char *buf
)
2473 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_misses
);
2476 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2477 struct ext4_sb_info
*sbi
,
2478 const char *buf
, size_t count
)
2482 if (parse_strtoul(buf
, 0x40000000, &t
))
2485 if (t
&& !is_power_of_2(t
))
2488 sbi
->s_inode_readahead_blks
= t
;
2492 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2493 struct ext4_sb_info
*sbi
, char *buf
)
2495 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2497 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2500 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2501 struct ext4_sb_info
*sbi
,
2502 const char *buf
, size_t count
)
2504 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2507 if (parse_strtoul(buf
, 0xffffffff, &t
))
2513 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2514 static struct ext4_attr ext4_attr_##_name = { \
2515 .attr = {.name = __stringify(_name), .mode = _mode }, \
2518 .offset = offsetof(struct ext4_sb_info, _elname), \
2520 #define EXT4_ATTR(name, mode, show, store) \
2521 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2523 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2524 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2525 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2526 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2527 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2528 #define ATTR_LIST(name) &ext4_attr_##name.attr
2530 EXT4_RO_ATTR(delayed_allocation_blocks
);
2531 EXT4_RO_ATTR(session_write_kbytes
);
2532 EXT4_RO_ATTR(lifetime_write_kbytes
);
2533 EXT4_RO_ATTR(extent_cache_hits
);
2534 EXT4_RO_ATTR(extent_cache_misses
);
2535 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2536 inode_readahead_blks_store
, s_inode_readahead_blks
);
2537 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2538 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2539 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2540 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2541 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2542 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2543 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2544 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2546 static struct attribute
*ext4_attrs
[] = {
2547 ATTR_LIST(delayed_allocation_blocks
),
2548 ATTR_LIST(session_write_kbytes
),
2549 ATTR_LIST(lifetime_write_kbytes
),
2550 ATTR_LIST(extent_cache_hits
),
2551 ATTR_LIST(extent_cache_misses
),
2552 ATTR_LIST(inode_readahead_blks
),
2553 ATTR_LIST(inode_goal
),
2554 ATTR_LIST(mb_stats
),
2555 ATTR_LIST(mb_max_to_scan
),
2556 ATTR_LIST(mb_min_to_scan
),
2557 ATTR_LIST(mb_order2_req
),
2558 ATTR_LIST(mb_stream_req
),
2559 ATTR_LIST(mb_group_prealloc
),
2560 ATTR_LIST(max_writeback_mb_bump
),
2564 /* Features this copy of ext4 supports */
2565 EXT4_INFO_ATTR(lazy_itable_init
);
2566 EXT4_INFO_ATTR(batched_discard
);
2568 static struct attribute
*ext4_feat_attrs
[] = {
2569 ATTR_LIST(lazy_itable_init
),
2570 ATTR_LIST(batched_discard
),
2574 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2575 struct attribute
*attr
, char *buf
)
2577 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2579 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2581 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2584 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2585 struct attribute
*attr
,
2586 const char *buf
, size_t len
)
2588 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2590 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2592 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2595 static void ext4_sb_release(struct kobject
*kobj
)
2597 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2599 complete(&sbi
->s_kobj_unregister
);
2602 static const struct sysfs_ops ext4_attr_ops
= {
2603 .show
= ext4_attr_show
,
2604 .store
= ext4_attr_store
,
2607 static struct kobj_type ext4_ktype
= {
2608 .default_attrs
= ext4_attrs
,
2609 .sysfs_ops
= &ext4_attr_ops
,
2610 .release
= ext4_sb_release
,
2613 static void ext4_feat_release(struct kobject
*kobj
)
2615 complete(&ext4_feat
->f_kobj_unregister
);
2618 static struct kobj_type ext4_feat_ktype
= {
2619 .default_attrs
= ext4_feat_attrs
,
2620 .sysfs_ops
= &ext4_attr_ops
,
2621 .release
= ext4_feat_release
,
2625 * Check whether this filesystem can be mounted based on
2626 * the features present and the RDONLY/RDWR mount requested.
2627 * Returns 1 if this filesystem can be mounted as requested,
2628 * 0 if it cannot be.
2630 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2632 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2633 ext4_msg(sb
, KERN_ERR
,
2634 "Couldn't mount because of "
2635 "unsupported optional features (%x)",
2636 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2637 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2644 /* Check that feature set is OK for a read-write mount */
2645 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2646 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2647 "unsupported optional features (%x)",
2648 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2649 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2653 * Large file size enabled file system can only be mounted
2654 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2656 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2657 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2658 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2659 "cannot be mounted RDWR without "
2668 * This function is called once a day if we have errors logged
2669 * on the file system
2671 static void print_daily_error_info(unsigned long arg
)
2673 struct super_block
*sb
= (struct super_block
*) arg
;
2674 struct ext4_sb_info
*sbi
;
2675 struct ext4_super_block
*es
;
2680 if (es
->s_error_count
)
2681 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2682 le32_to_cpu(es
->s_error_count
));
2683 if (es
->s_first_error_time
) {
2684 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2685 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2686 (int) sizeof(es
->s_first_error_func
),
2687 es
->s_first_error_func
,
2688 le32_to_cpu(es
->s_first_error_line
));
2689 if (es
->s_first_error_ino
)
2690 printk(": inode %u",
2691 le32_to_cpu(es
->s_first_error_ino
));
2692 if (es
->s_first_error_block
)
2693 printk(": block %llu", (unsigned long long)
2694 le64_to_cpu(es
->s_first_error_block
));
2697 if (es
->s_last_error_time
) {
2698 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2699 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2700 (int) sizeof(es
->s_last_error_func
),
2701 es
->s_last_error_func
,
2702 le32_to_cpu(es
->s_last_error_line
));
2703 if (es
->s_last_error_ino
)
2704 printk(": inode %u",
2705 le32_to_cpu(es
->s_last_error_ino
));
2706 if (es
->s_last_error_block
)
2707 printk(": block %llu", (unsigned long long)
2708 le64_to_cpu(es
->s_last_error_block
));
2711 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2714 /* Find next suitable group and run ext4_init_inode_table */
2715 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2717 struct ext4_group_desc
*gdp
= NULL
;
2718 ext4_group_t group
, ngroups
;
2719 struct super_block
*sb
;
2720 unsigned long timeout
= 0;
2724 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2726 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2727 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2733 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2737 if (group
== ngroups
)
2742 ret
= ext4_init_inode_table(sb
, group
,
2743 elr
->lr_timeout
? 0 : 1);
2744 if (elr
->lr_timeout
== 0) {
2745 timeout
= (jiffies
- timeout
) *
2746 elr
->lr_sbi
->s_li_wait_mult
;
2747 elr
->lr_timeout
= timeout
;
2749 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2750 elr
->lr_next_group
= group
+ 1;
2757 * Remove lr_request from the list_request and free the
2758 * request structure. Should be called with li_list_mtx held
2760 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2762 struct ext4_sb_info
*sbi
;
2769 list_del(&elr
->lr_request
);
2770 sbi
->s_li_request
= NULL
;
2774 static void ext4_unregister_li_request(struct super_block
*sb
)
2776 mutex_lock(&ext4_li_mtx
);
2777 if (!ext4_li_info
) {
2778 mutex_unlock(&ext4_li_mtx
);
2782 mutex_lock(&ext4_li_info
->li_list_mtx
);
2783 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2784 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2785 mutex_unlock(&ext4_li_mtx
);
2788 static struct task_struct
*ext4_lazyinit_task
;
2791 * This is the function where ext4lazyinit thread lives. It walks
2792 * through the request list searching for next scheduled filesystem.
2793 * When such a fs is found, run the lazy initialization request
2794 * (ext4_rn_li_request) and keep track of the time spend in this
2795 * function. Based on that time we compute next schedule time of
2796 * the request. When walking through the list is complete, compute
2797 * next waking time and put itself into sleep.
2799 static int ext4_lazyinit_thread(void *arg
)
2801 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2802 struct list_head
*pos
, *n
;
2803 struct ext4_li_request
*elr
;
2804 unsigned long next_wakeup
, cur
;
2806 BUG_ON(NULL
== eli
);
2810 next_wakeup
= MAX_JIFFY_OFFSET
;
2812 mutex_lock(&eli
->li_list_mtx
);
2813 if (list_empty(&eli
->li_request_list
)) {
2814 mutex_unlock(&eli
->li_list_mtx
);
2818 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2819 elr
= list_entry(pos
, struct ext4_li_request
,
2822 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2823 if (ext4_run_li_request(elr
) != 0) {
2824 /* error, remove the lazy_init job */
2825 ext4_remove_li_request(elr
);
2830 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2831 next_wakeup
= elr
->lr_next_sched
;
2833 mutex_unlock(&eli
->li_list_mtx
);
2835 if (freezing(current
))
2839 if ((time_after_eq(cur
, next_wakeup
)) ||
2840 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2845 schedule_timeout_interruptible(next_wakeup
- cur
);
2847 if (kthread_should_stop()) {
2848 ext4_clear_request_list();
2855 * It looks like the request list is empty, but we need
2856 * to check it under the li_list_mtx lock, to prevent any
2857 * additions into it, and of course we should lock ext4_li_mtx
2858 * to atomically free the list and ext4_li_info, because at
2859 * this point another ext4 filesystem could be registering
2862 mutex_lock(&ext4_li_mtx
);
2863 mutex_lock(&eli
->li_list_mtx
);
2864 if (!list_empty(&eli
->li_request_list
)) {
2865 mutex_unlock(&eli
->li_list_mtx
);
2866 mutex_unlock(&ext4_li_mtx
);
2869 mutex_unlock(&eli
->li_list_mtx
);
2870 kfree(ext4_li_info
);
2871 ext4_li_info
= NULL
;
2872 mutex_unlock(&ext4_li_mtx
);
2877 static void ext4_clear_request_list(void)
2879 struct list_head
*pos
, *n
;
2880 struct ext4_li_request
*elr
;
2882 mutex_lock(&ext4_li_info
->li_list_mtx
);
2883 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2884 elr
= list_entry(pos
, struct ext4_li_request
,
2886 ext4_remove_li_request(elr
);
2888 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2891 static int ext4_run_lazyinit_thread(void)
2893 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2894 ext4_li_info
, "ext4lazyinit");
2895 if (IS_ERR(ext4_lazyinit_task
)) {
2896 int err
= PTR_ERR(ext4_lazyinit_task
);
2897 ext4_clear_request_list();
2898 kfree(ext4_li_info
);
2899 ext4_li_info
= NULL
;
2900 printk(KERN_CRIT
"EXT4: error %d creating inode table "
2901 "initialization thread\n",
2905 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2910 * Check whether it make sense to run itable init. thread or not.
2911 * If there is at least one uninitialized inode table, return
2912 * corresponding group number, else the loop goes through all
2913 * groups and return total number of groups.
2915 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2917 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2918 struct ext4_group_desc
*gdp
= NULL
;
2920 for (group
= 0; group
< ngroups
; group
++) {
2921 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2925 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2932 static int ext4_li_info_new(void)
2934 struct ext4_lazy_init
*eli
= NULL
;
2936 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2940 INIT_LIST_HEAD(&eli
->li_request_list
);
2941 mutex_init(&eli
->li_list_mtx
);
2943 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2950 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2953 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2954 struct ext4_li_request
*elr
;
2957 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2963 elr
->lr_next_group
= start
;
2966 * Randomize first schedule time of the request to
2967 * spread the inode table initialization requests
2970 get_random_bytes(&rnd
, sizeof(rnd
));
2971 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2972 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
2977 static int ext4_register_li_request(struct super_block
*sb
,
2978 ext4_group_t first_not_zeroed
)
2980 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2981 struct ext4_li_request
*elr
;
2982 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
2985 if (sbi
->s_li_request
!= NULL
) {
2987 * Reset timeout so it can be computed again, because
2988 * s_li_wait_mult might have changed.
2990 sbi
->s_li_request
->lr_timeout
= 0;
2994 if (first_not_zeroed
== ngroups
||
2995 (sb
->s_flags
& MS_RDONLY
) ||
2996 !test_opt(sb
, INIT_INODE_TABLE
))
2999 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3003 mutex_lock(&ext4_li_mtx
);
3005 if (NULL
== ext4_li_info
) {
3006 ret
= ext4_li_info_new();
3011 mutex_lock(&ext4_li_info
->li_list_mtx
);
3012 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3013 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3015 sbi
->s_li_request
= elr
;
3017 * set elr to NULL here since it has been inserted to
3018 * the request_list and the removal and free of it is
3019 * handled by ext4_clear_request_list from now on.
3023 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3024 ret
= ext4_run_lazyinit_thread();
3029 mutex_unlock(&ext4_li_mtx
);
3036 * We do not need to lock anything since this is called on
3039 static void ext4_destroy_lazyinit_thread(void)
3042 * If thread exited earlier
3043 * there's nothing to be done.
3045 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3048 kthread_stop(ext4_lazyinit_task
);
3051 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3052 __releases(kernel_lock
)
3053 __acquires(kernel_lock
)
3055 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3056 struct buffer_head
*bh
;
3057 struct ext4_super_block
*es
= NULL
;
3058 struct ext4_sb_info
*sbi
;
3060 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3061 ext4_fsblk_t logical_sb_block
;
3062 unsigned long offset
= 0;
3063 unsigned long journal_devnum
= 0;
3064 unsigned long def_mount_opts
;
3070 unsigned int db_count
;
3072 int needs_recovery
, has_huge_files
;
3075 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3076 ext4_group_t first_not_zeroed
;
3078 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3082 sbi
->s_blockgroup_lock
=
3083 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3084 if (!sbi
->s_blockgroup_lock
) {
3088 sb
->s_fs_info
= sbi
;
3089 sbi
->s_mount_opt
= 0;
3090 sbi
->s_resuid
= EXT4_DEF_RESUID
;
3091 sbi
->s_resgid
= EXT4_DEF_RESGID
;
3092 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3093 sbi
->s_sb_block
= sb_block
;
3094 if (sb
->s_bdev
->bd_part
)
3095 sbi
->s_sectors_written_start
=
3096 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3098 /* Cleanup superblock name */
3099 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3103 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3105 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3110 * The ext4 superblock will not be buffer aligned for other than 1kB
3111 * block sizes. We need to calculate the offset from buffer start.
3113 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3114 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3115 offset
= do_div(logical_sb_block
, blocksize
);
3117 logical_sb_block
= sb_block
;
3120 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3121 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3125 * Note: s_es must be initialized as soon as possible because
3126 * some ext4 macro-instructions depend on its value
3128 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3130 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3131 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3133 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3135 /* Set defaults before we parse the mount options */
3136 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3137 set_opt(sb
, INIT_INODE_TABLE
);
3138 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3140 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
3141 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
3145 if (def_mount_opts
& EXT4_DEFM_UID16
)
3146 set_opt(sb
, NO_UID32
);
3147 /* xattr user namespace & acls are now defaulted on */
3148 #ifdef CONFIG_EXT4_FS_XATTR
3149 set_opt(sb
, XATTR_USER
);
3151 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3152 set_opt(sb
, POSIX_ACL
);
3154 set_opt(sb
, MBLK_IO_SUBMIT
);
3155 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3156 set_opt(sb
, JOURNAL_DATA
);
3157 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3158 set_opt(sb
, ORDERED_DATA
);
3159 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3160 set_opt(sb
, WRITEBACK_DATA
);
3162 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3163 set_opt(sb
, ERRORS_PANIC
);
3164 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3165 set_opt(sb
, ERRORS_CONT
);
3167 set_opt(sb
, ERRORS_RO
);
3168 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3169 set_opt(sb
, BLOCK_VALIDITY
);
3170 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3171 set_opt(sb
, DISCARD
);
3173 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
3174 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
3175 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3176 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3177 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3179 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3180 set_opt(sb
, BARRIER
);
3183 * enable delayed allocation by default
3184 * Use -o nodelalloc to turn it off
3186 if (!IS_EXT3_SB(sb
) &&
3187 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3188 set_opt(sb
, DELALLOC
);
3191 * set default s_li_wait_mult for lazyinit, for the case there is
3192 * no mount option specified.
3194 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3196 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3197 &journal_devnum
, &journal_ioprio
, NULL
, 0)) {
3198 ext4_msg(sb
, KERN_WARNING
,
3199 "failed to parse options in superblock: %s",
3200 sbi
->s_es
->s_mount_opts
);
3202 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3203 &journal_ioprio
, NULL
, 0))
3206 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3207 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3209 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3210 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3211 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3212 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3213 ext4_msg(sb
, KERN_WARNING
,
3214 "feature flags set on rev 0 fs, "
3215 "running e2fsck is recommended");
3217 if (IS_EXT2_SB(sb
)) {
3218 if (ext2_feature_set_ok(sb
))
3219 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3220 "using the ext4 subsystem");
3222 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3223 "to feature incompatibilities");
3228 if (IS_EXT3_SB(sb
)) {
3229 if (ext3_feature_set_ok(sb
))
3230 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3231 "using the ext4 subsystem");
3233 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3234 "to feature incompatibilities");
3240 * Check feature flags regardless of the revision level, since we
3241 * previously didn't change the revision level when setting the flags,
3242 * so there is a chance incompat flags are set on a rev 0 filesystem.
3244 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3247 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3249 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3250 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3251 ext4_msg(sb
, KERN_ERR
,
3252 "Unsupported filesystem blocksize %d", blocksize
);
3256 if (sb
->s_blocksize
!= blocksize
) {
3257 /* Validate the filesystem blocksize */
3258 if (!sb_set_blocksize(sb
, blocksize
)) {
3259 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3265 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3266 offset
= do_div(logical_sb_block
, blocksize
);
3267 bh
= sb_bread(sb
, logical_sb_block
);
3269 ext4_msg(sb
, KERN_ERR
,
3270 "Can't read superblock on 2nd try");
3273 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
3275 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3276 ext4_msg(sb
, KERN_ERR
,
3277 "Magic mismatch, very weird!");
3282 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3283 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3284 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3286 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3288 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3289 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3290 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3292 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3293 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3294 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3295 (!is_power_of_2(sbi
->s_inode_size
)) ||
3296 (sbi
->s_inode_size
> blocksize
)) {
3297 ext4_msg(sb
, KERN_ERR
,
3298 "unsupported inode size: %d",
3302 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3303 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3306 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3307 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3308 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3309 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3310 !is_power_of_2(sbi
->s_desc_size
)) {
3311 ext4_msg(sb
, KERN_ERR
,
3312 "unsupported descriptor size %lu",
3317 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3319 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3320 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3321 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3324 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3325 if (sbi
->s_inodes_per_block
== 0)
3327 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3328 sbi
->s_inodes_per_block
;
3329 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3331 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3332 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3333 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3335 for (i
= 0; i
< 4; i
++)
3336 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3337 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3338 i
= le32_to_cpu(es
->s_flags
);
3339 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3340 sbi
->s_hash_unsigned
= 3;
3341 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3342 #ifdef __CHAR_UNSIGNED__
3343 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3344 sbi
->s_hash_unsigned
= 3;
3346 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3351 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3352 ext4_msg(sb
, KERN_ERR
,
3353 "#blocks per group too big: %lu",
3354 sbi
->s_blocks_per_group
);
3357 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3358 ext4_msg(sb
, KERN_ERR
,
3359 "#inodes per group too big: %lu",
3360 sbi
->s_inodes_per_group
);
3365 * Test whether we have more sectors than will fit in sector_t,
3366 * and whether the max offset is addressable by the page cache.
3368 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3369 ext4_blocks_count(es
));
3371 ext4_msg(sb
, KERN_ERR
, "filesystem"
3372 " too large to mount safely on this system");
3373 if (sizeof(sector_t
) < 8)
3374 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3379 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3382 /* check blocks count against device size */
3383 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3384 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3385 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3386 "exceeds size of device (%llu blocks)",
3387 ext4_blocks_count(es
), blocks_count
);
3392 * It makes no sense for the first data block to be beyond the end
3393 * of the filesystem.
3395 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3396 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
3397 "block %u is beyond end of filesystem (%llu)",
3398 le32_to_cpu(es
->s_first_data_block
),
3399 ext4_blocks_count(es
));
3402 blocks_count
= (ext4_blocks_count(es
) -
3403 le32_to_cpu(es
->s_first_data_block
) +
3404 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3405 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3406 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3407 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3408 "(block count %llu, first data block %u, "
3409 "blocks per group %lu)", sbi
->s_groups_count
,
3410 ext4_blocks_count(es
),
3411 le32_to_cpu(es
->s_first_data_block
),
3412 EXT4_BLOCKS_PER_GROUP(sb
));
3415 sbi
->s_groups_count
= blocks_count
;
3416 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3417 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3418 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3419 EXT4_DESC_PER_BLOCK(sb
);
3420 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
3422 if (sbi
->s_group_desc
== NULL
) {
3423 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3427 #ifdef CONFIG_PROC_FS
3429 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3432 bgl_lock_init(sbi
->s_blockgroup_lock
);
3434 for (i
= 0; i
< db_count
; i
++) {
3435 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3436 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3437 if (!sbi
->s_group_desc
[i
]) {
3438 ext4_msg(sb
, KERN_ERR
,
3439 "can't read group descriptor %d", i
);
3444 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3445 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3448 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3449 if (!ext4_fill_flex_info(sb
)) {
3450 ext4_msg(sb
, KERN_ERR
,
3451 "unable to initialize "
3452 "flex_bg meta info!");
3456 sbi
->s_gdb_count
= db_count
;
3457 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3458 spin_lock_init(&sbi
->s_next_gen_lock
);
3460 init_timer(&sbi
->s_err_report
);
3461 sbi
->s_err_report
.function
= print_daily_error_info
;
3462 sbi
->s_err_report
.data
= (unsigned long) sb
;
3464 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
3465 ext4_count_free_blocks(sb
));
3467 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3468 ext4_count_free_inodes(sb
));
3471 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3472 ext4_count_dirs(sb
));
3475 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
3478 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3482 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3483 sbi
->s_max_writeback_mb_bump
= 128;
3486 * set up enough so that it can read an inode
3488 if (!test_opt(sb
, NOLOAD
) &&
3489 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3490 sb
->s_op
= &ext4_sops
;
3492 sb
->s_op
= &ext4_nojournal_sops
;
3493 sb
->s_export_op
= &ext4_export_ops
;
3494 sb
->s_xattr
= ext4_xattr_handlers
;
3496 sb
->s_qcop
= &ext4_qctl_operations
;
3497 sb
->dq_op
= &ext4_quota_operations
;
3499 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3501 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3502 mutex_init(&sbi
->s_orphan_lock
);
3503 mutex_init(&sbi
->s_resize_lock
);
3507 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3508 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3509 EXT4_FEATURE_INCOMPAT_RECOVER
));
3511 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3512 !(sb
->s_flags
& MS_RDONLY
))
3513 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3517 * The first inode we look at is the journal inode. Don't try
3518 * root first: it may be modified in the journal!
3520 if (!test_opt(sb
, NOLOAD
) &&
3521 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3522 if (ext4_load_journal(sb
, es
, journal_devnum
))
3524 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3525 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3526 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3527 "suppressed and not mounted read-only");
3528 goto failed_mount_wq
;
3530 clear_opt(sb
, DATA_FLAGS
);
3531 sbi
->s_journal
= NULL
;
3536 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
3537 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3538 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3539 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3540 goto failed_mount_wq
;
3543 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3544 jbd2_journal_set_features(sbi
->s_journal
,
3545 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3546 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3547 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3548 jbd2_journal_set_features(sbi
->s_journal
,
3549 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
3550 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3551 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3553 jbd2_journal_clear_features(sbi
->s_journal
,
3554 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3555 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3558 /* We have now updated the journal if required, so we can
3559 * validate the data journaling mode. */
3560 switch (test_opt(sb
, DATA_FLAGS
)) {
3562 /* No mode set, assume a default based on the journal
3563 * capabilities: ORDERED_DATA if the journal can
3564 * cope, else JOURNAL_DATA
3566 if (jbd2_journal_check_available_features
3567 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3568 set_opt(sb
, ORDERED_DATA
);
3570 set_opt(sb
, JOURNAL_DATA
);
3573 case EXT4_MOUNT_ORDERED_DATA
:
3574 case EXT4_MOUNT_WRITEBACK_DATA
:
3575 if (!jbd2_journal_check_available_features
3576 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3577 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3578 "requested data journaling mode");
3579 goto failed_mount_wq
;
3584 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3587 * The journal may have updated the bg summary counts, so we
3588 * need to update the global counters.
3590 percpu_counter_set(&sbi
->s_freeblocks_counter
,
3591 ext4_count_free_blocks(sb
));
3592 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3593 ext4_count_free_inodes(sb
));
3594 percpu_counter_set(&sbi
->s_dirs_counter
,
3595 ext4_count_dirs(sb
));
3596 percpu_counter_set(&sbi
->s_dirtyblocks_counter
, 0);
3600 * The maximum number of concurrent works can be high and
3601 * concurrency isn't really necessary. Limit it to 1.
3603 EXT4_SB(sb
)->dio_unwritten_wq
=
3604 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3605 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3606 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3607 goto failed_mount_wq
;
3611 * The jbd2_journal_load will have done any necessary log recovery,
3612 * so we can safely mount the rest of the filesystem now.
3615 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3617 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3618 ret
= PTR_ERR(root
);
3622 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3623 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3626 sb
->s_root
= d_alloc_root(root
);
3628 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3633 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
3635 /* determine the minimum size of new large inodes, if present */
3636 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3637 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3638 EXT4_GOOD_OLD_INODE_SIZE
;
3639 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3640 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3641 if (sbi
->s_want_extra_isize
<
3642 le16_to_cpu(es
->s_want_extra_isize
))
3643 sbi
->s_want_extra_isize
=
3644 le16_to_cpu(es
->s_want_extra_isize
);
3645 if (sbi
->s_want_extra_isize
<
3646 le16_to_cpu(es
->s_min_extra_isize
))
3647 sbi
->s_want_extra_isize
=
3648 le16_to_cpu(es
->s_min_extra_isize
);
3651 /* Check if enough inode space is available */
3652 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3653 sbi
->s_inode_size
) {
3654 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3655 EXT4_GOOD_OLD_INODE_SIZE
;
3656 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3660 if (test_opt(sb
, DELALLOC
) &&
3661 (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)) {
3662 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
3663 "requested data journaling mode");
3664 clear_opt(sb
, DELALLOC
);
3666 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3667 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3668 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3669 "option - requested data journaling mode");
3670 clear_opt(sb
, DIOREAD_NOLOCK
);
3672 if (sb
->s_blocksize
< PAGE_SIZE
) {
3673 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3674 "option - block size is too small");
3675 clear_opt(sb
, DIOREAD_NOLOCK
);
3679 err
= ext4_setup_system_zone(sb
);
3681 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3687 err
= ext4_mb_init(sb
, needs_recovery
);
3689 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3694 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3698 sbi
->s_kobj
.kset
= ext4_kset
;
3699 init_completion(&sbi
->s_kobj_unregister
);
3700 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3703 ext4_mb_release(sb
);
3704 ext4_ext_release(sb
);
3708 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3709 ext4_orphan_cleanup(sb
, es
);
3710 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3711 if (needs_recovery
) {
3712 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3713 ext4_mark_recovery_complete(sb
, es
);
3715 if (EXT4_SB(sb
)->s_journal
) {
3716 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3717 descr
= " journalled data mode";
3718 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3719 descr
= " ordered data mode";
3721 descr
= " writeback data mode";
3723 descr
= "out journal";
3725 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3726 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3727 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3729 if (es
->s_error_count
)
3730 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3737 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3743 ext4_msg(sb
, KERN_ERR
, "mount failed");
3744 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3746 ext4_release_system_zone(sb
);
3747 if (sbi
->s_journal
) {
3748 jbd2_journal_destroy(sbi
->s_journal
);
3749 sbi
->s_journal
= NULL
;
3752 del_timer(&sbi
->s_err_report
);
3753 if (sbi
->s_flex_groups
) {
3754 if (is_vmalloc_addr(sbi
->s_flex_groups
))
3755 vfree(sbi
->s_flex_groups
);
3757 kfree(sbi
->s_flex_groups
);
3759 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
3760 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3761 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3762 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
3764 kthread_stop(sbi
->s_mmp_tsk
);
3766 for (i
= 0; i
< db_count
; i
++)
3767 brelse(sbi
->s_group_desc
[i
]);
3768 kfree(sbi
->s_group_desc
);
3771 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3774 for (i
= 0; i
< MAXQUOTAS
; i
++)
3775 kfree(sbi
->s_qf_names
[i
]);
3777 ext4_blkdev_remove(sbi
);
3780 sb
->s_fs_info
= NULL
;
3781 kfree(sbi
->s_blockgroup_lock
);
3789 * Setup any per-fs journal parameters now. We'll do this both on
3790 * initial mount, once the journal has been initialised but before we've
3791 * done any recovery; and again on any subsequent remount.
3793 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3795 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3797 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3798 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3799 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3801 write_lock(&journal
->j_state_lock
);
3802 if (test_opt(sb
, BARRIER
))
3803 journal
->j_flags
|= JBD2_BARRIER
;
3805 journal
->j_flags
&= ~JBD2_BARRIER
;
3806 if (test_opt(sb
, DATA_ERR_ABORT
))
3807 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3809 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3810 write_unlock(&journal
->j_state_lock
);
3813 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3814 unsigned int journal_inum
)
3816 struct inode
*journal_inode
;
3819 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3821 /* First, test for the existence of a valid inode on disk. Bad
3822 * things happen if we iget() an unused inode, as the subsequent
3823 * iput() will try to delete it. */
3825 journal_inode
= ext4_iget(sb
, journal_inum
);
3826 if (IS_ERR(journal_inode
)) {
3827 ext4_msg(sb
, KERN_ERR
, "no journal found");
3830 if (!journal_inode
->i_nlink
) {
3831 make_bad_inode(journal_inode
);
3832 iput(journal_inode
);
3833 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3837 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3838 journal_inode
, journal_inode
->i_size
);
3839 if (!S_ISREG(journal_inode
->i_mode
)) {
3840 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3841 iput(journal_inode
);
3845 journal
= jbd2_journal_init_inode(journal_inode
);
3847 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3848 iput(journal_inode
);
3851 journal
->j_private
= sb
;
3852 ext4_init_journal_params(sb
, journal
);
3856 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3859 struct buffer_head
*bh
;
3863 int hblock
, blocksize
;
3864 ext4_fsblk_t sb_block
;
3865 unsigned long offset
;
3866 struct ext4_super_block
*es
;
3867 struct block_device
*bdev
;
3869 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3871 bdev
= ext4_blkdev_get(j_dev
, sb
);
3875 blocksize
= sb
->s_blocksize
;
3876 hblock
= bdev_logical_block_size(bdev
);
3877 if (blocksize
< hblock
) {
3878 ext4_msg(sb
, KERN_ERR
,
3879 "blocksize too small for journal device");
3883 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3884 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3885 set_blocksize(bdev
, blocksize
);
3886 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3887 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3888 "external journal");
3892 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3893 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3894 !(le32_to_cpu(es
->s_feature_incompat
) &
3895 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3896 ext4_msg(sb
, KERN_ERR
, "external journal has "
3902 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3903 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3908 len
= ext4_blocks_count(es
);
3909 start
= sb_block
+ 1;
3910 brelse(bh
); /* we're done with the superblock */
3912 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3913 start
, len
, blocksize
);
3915 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3918 journal
->j_private
= sb
;
3919 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3920 wait_on_buffer(journal
->j_sb_buffer
);
3921 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3922 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3925 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3926 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3927 "user (unsupported) - %d",
3928 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3931 EXT4_SB(sb
)->journal_bdev
= bdev
;
3932 ext4_init_journal_params(sb
, journal
);
3936 jbd2_journal_destroy(journal
);
3938 ext4_blkdev_put(bdev
);
3942 static int ext4_load_journal(struct super_block
*sb
,
3943 struct ext4_super_block
*es
,
3944 unsigned long journal_devnum
)
3947 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3950 int really_read_only
;
3952 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3954 if (journal_devnum
&&
3955 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3956 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3957 "numbers have changed");
3958 journal_dev
= new_decode_dev(journal_devnum
);
3960 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3962 really_read_only
= bdev_read_only(sb
->s_bdev
);
3965 * Are we loading a blank journal or performing recovery after a
3966 * crash? For recovery, we need to check in advance whether we
3967 * can get read-write access to the device.
3969 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3970 if (sb
->s_flags
& MS_RDONLY
) {
3971 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3972 "required on readonly filesystem");
3973 if (really_read_only
) {
3974 ext4_msg(sb
, KERN_ERR
, "write access "
3975 "unavailable, cannot proceed");
3978 ext4_msg(sb
, KERN_INFO
, "write access will "
3979 "be enabled during recovery");
3983 if (journal_inum
&& journal_dev
) {
3984 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3985 "and inode journals!");
3990 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3993 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3997 if (!(journal
->j_flags
& JBD2_BARRIER
))
3998 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4000 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
4001 err
= jbd2_journal_update_format(journal
);
4003 ext4_msg(sb
, KERN_ERR
, "error updating journal");
4004 jbd2_journal_destroy(journal
);
4009 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4010 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4012 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4014 memcpy(save
, ((char *) es
) +
4015 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4016 err
= jbd2_journal_load(journal
);
4018 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4019 save
, EXT4_S_ERR_LEN
);
4024 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4025 jbd2_journal_destroy(journal
);
4029 EXT4_SB(sb
)->s_journal
= journal
;
4030 ext4_clear_journal_err(sb
, es
);
4032 if (!really_read_only
&& journal_devnum
&&
4033 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4034 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4036 /* Make sure we flush the recovery flag to disk. */
4037 ext4_commit_super(sb
, 1);
4043 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4045 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4046 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4051 if (buffer_write_io_error(sbh
)) {
4053 * Oh, dear. A previous attempt to write the
4054 * superblock failed. This could happen because the
4055 * USB device was yanked out. Or it could happen to
4056 * be a transient write error and maybe the block will
4057 * be remapped. Nothing we can do but to retry the
4058 * write and hope for the best.
4060 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4061 "superblock detected");
4062 clear_buffer_write_io_error(sbh
);
4063 set_buffer_uptodate(sbh
);
4066 * If the file system is mounted read-only, don't update the
4067 * superblock write time. This avoids updating the superblock
4068 * write time when we are mounting the root file system
4069 * read/only but we need to replay the journal; at that point,
4070 * for people who are east of GMT and who make their clock
4071 * tick in localtime for Windows bug-for-bug compatibility,
4072 * the clock is set in the future, and this will cause e2fsck
4073 * to complain and force a full file system check.
4075 if (!(sb
->s_flags
& MS_RDONLY
))
4076 es
->s_wtime
= cpu_to_le32(get_seconds());
4077 if (sb
->s_bdev
->bd_part
)
4078 es
->s_kbytes_written
=
4079 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4080 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4081 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4083 es
->s_kbytes_written
=
4084 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4085 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
4086 &EXT4_SB(sb
)->s_freeblocks_counter
));
4087 es
->s_free_inodes_count
=
4088 cpu_to_le32(percpu_counter_sum_positive(
4089 &EXT4_SB(sb
)->s_freeinodes_counter
));
4091 BUFFER_TRACE(sbh
, "marking dirty");
4092 mark_buffer_dirty(sbh
);
4094 error
= sync_dirty_buffer(sbh
);
4098 error
= buffer_write_io_error(sbh
);
4100 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4102 clear_buffer_write_io_error(sbh
);
4103 set_buffer_uptodate(sbh
);
4110 * Have we just finished recovery? If so, and if we are mounting (or
4111 * remounting) the filesystem readonly, then we will end up with a
4112 * consistent fs on disk. Record that fact.
4114 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4115 struct ext4_super_block
*es
)
4117 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4119 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4120 BUG_ON(journal
!= NULL
);
4123 jbd2_journal_lock_updates(journal
);
4124 if (jbd2_journal_flush(journal
) < 0)
4127 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4128 sb
->s_flags
& MS_RDONLY
) {
4129 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4130 ext4_commit_super(sb
, 1);
4134 jbd2_journal_unlock_updates(journal
);
4138 * If we are mounting (or read-write remounting) a filesystem whose journal
4139 * has recorded an error from a previous lifetime, move that error to the
4140 * main filesystem now.
4142 static void ext4_clear_journal_err(struct super_block
*sb
,
4143 struct ext4_super_block
*es
)
4149 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4151 journal
= EXT4_SB(sb
)->s_journal
;
4154 * Now check for any error status which may have been recorded in the
4155 * journal by a prior ext4_error() or ext4_abort()
4158 j_errno
= jbd2_journal_errno(journal
);
4162 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4163 ext4_warning(sb
, "Filesystem error recorded "
4164 "from previous mount: %s", errstr
);
4165 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4167 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4168 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4169 ext4_commit_super(sb
, 1);
4171 jbd2_journal_clear_err(journal
);
4176 * Force the running and committing transactions to commit,
4177 * and wait on the commit.
4179 int ext4_force_commit(struct super_block
*sb
)
4184 if (sb
->s_flags
& MS_RDONLY
)
4187 journal
= EXT4_SB(sb
)->s_journal
;
4189 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
4190 ret
= ext4_journal_force_commit(journal
);
4196 static void ext4_write_super(struct super_block
*sb
)
4199 ext4_commit_super(sb
, 1);
4203 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4207 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4209 trace_ext4_sync_fs(sb
, wait
);
4210 flush_workqueue(sbi
->dio_unwritten_wq
);
4211 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4213 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4219 * LVM calls this function before a (read-only) snapshot is created. This
4220 * gives us a chance to flush the journal completely and mark the fs clean.
4222 * Note that only this function cannot bring a filesystem to be in a clean
4223 * state independently, because ext4 prevents a new handle from being started
4224 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4227 static int ext4_freeze(struct super_block
*sb
)
4232 if (sb
->s_flags
& MS_RDONLY
)
4235 journal
= EXT4_SB(sb
)->s_journal
;
4237 /* Now we set up the journal barrier. */
4238 jbd2_journal_lock_updates(journal
);
4241 * Don't clear the needs_recovery flag if we failed to flush
4244 error
= jbd2_journal_flush(journal
);
4248 /* Journal blocked and flushed, clear needs_recovery flag. */
4249 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4250 error
= ext4_commit_super(sb
, 1);
4252 /* we rely on s_frozen to stop further updates */
4253 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4258 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4259 * flag here, even though the filesystem is not technically dirty yet.
4261 static int ext4_unfreeze(struct super_block
*sb
)
4263 if (sb
->s_flags
& MS_RDONLY
)
4267 /* Reset the needs_recovery flag before the fs is unlocked. */
4268 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4269 ext4_commit_super(sb
, 1);
4275 * Structure to save mount options for ext4_remount's benefit
4277 struct ext4_mount_options
{
4278 unsigned long s_mount_opt
;
4279 unsigned long s_mount_opt2
;
4282 unsigned long s_commit_interval
;
4283 u32 s_min_batch_time
, s_max_batch_time
;
4286 char *s_qf_names
[MAXQUOTAS
];
4290 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4292 struct ext4_super_block
*es
;
4293 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4294 ext4_fsblk_t n_blocks_count
= 0;
4295 unsigned long old_sb_flags
;
4296 struct ext4_mount_options old_opts
;
4297 int enable_quota
= 0;
4299 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4304 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4306 /* Store the original options */
4308 old_sb_flags
= sb
->s_flags
;
4309 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4310 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4311 old_opts
.s_resuid
= sbi
->s_resuid
;
4312 old_opts
.s_resgid
= sbi
->s_resgid
;
4313 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4314 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4315 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4317 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4318 for (i
= 0; i
< MAXQUOTAS
; i
++)
4319 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
4321 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4322 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4325 * Allow the "check" option to be passed as a remount option.
4327 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
4328 &n_blocks_count
, 1)) {
4333 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4334 ext4_abort(sb
, "Abort forced by user");
4336 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4337 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4341 if (sbi
->s_journal
) {
4342 ext4_init_journal_params(sb
, sbi
->s_journal
);
4343 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4346 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
4347 n_blocks_count
> ext4_blocks_count(es
)) {
4348 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4353 if (*flags
& MS_RDONLY
) {
4354 err
= dquot_suspend(sb
, -1);
4359 * First of all, the unconditional stuff we have to do
4360 * to disable replay of the journal when we next remount
4362 sb
->s_flags
|= MS_RDONLY
;
4365 * OK, test if we are remounting a valid rw partition
4366 * readonly, and if so set the rdonly flag and then
4367 * mark the partition as valid again.
4369 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4370 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4371 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4374 ext4_mark_recovery_complete(sb
, es
);
4376 /* Make sure we can mount this feature set readwrite */
4377 if (!ext4_feature_set_ok(sb
, 0)) {
4382 * Make sure the group descriptor checksums
4383 * are sane. If they aren't, refuse to remount r/w.
4385 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4386 struct ext4_group_desc
*gdp
=
4387 ext4_get_group_desc(sb
, g
, NULL
);
4389 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
4390 ext4_msg(sb
, KERN_ERR
,
4391 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4392 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4393 le16_to_cpu(gdp
->bg_checksum
));
4400 * If we have an unprocessed orphan list hanging
4401 * around from a previously readonly bdev mount,
4402 * require a full umount/remount for now.
4404 if (es
->s_last_orphan
) {
4405 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4406 "remount RDWR because of unprocessed "
4407 "orphan inode list. Please "
4408 "umount/remount instead");
4414 * Mounting a RDONLY partition read-write, so reread
4415 * and store the current valid flag. (It may have
4416 * been changed by e2fsck since we originally mounted
4420 ext4_clear_journal_err(sb
, es
);
4421 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4422 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
4424 if (!ext4_setup_super(sb
, es
, 0))
4425 sb
->s_flags
&= ~MS_RDONLY
;
4426 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4427 EXT4_FEATURE_INCOMPAT_MMP
))
4428 if (ext4_multi_mount_protect(sb
,
4429 le64_to_cpu(es
->s_mmp_block
))) {
4438 * Reinitialize lazy itable initialization thread based on
4441 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4442 ext4_unregister_li_request(sb
);
4444 ext4_group_t first_not_zeroed
;
4445 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4446 ext4_register_li_request(sb
, first_not_zeroed
);
4449 ext4_setup_system_zone(sb
);
4450 if (sbi
->s_journal
== NULL
)
4451 ext4_commit_super(sb
, 1);
4454 /* Release old quota file names */
4455 for (i
= 0; i
< MAXQUOTAS
; i
++)
4456 if (old_opts
.s_qf_names
[i
] &&
4457 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4458 kfree(old_opts
.s_qf_names
[i
]);
4462 dquot_resume(sb
, -1);
4464 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4469 sb
->s_flags
= old_sb_flags
;
4470 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4471 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4472 sbi
->s_resuid
= old_opts
.s_resuid
;
4473 sbi
->s_resgid
= old_opts
.s_resgid
;
4474 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4475 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4476 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4478 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4479 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4480 if (sbi
->s_qf_names
[i
] &&
4481 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4482 kfree(sbi
->s_qf_names
[i
]);
4483 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4491 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4493 struct super_block
*sb
= dentry
->d_sb
;
4494 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4495 struct ext4_super_block
*es
= sbi
->s_es
;
4499 if (test_opt(sb
, MINIX_DF
)) {
4500 sbi
->s_overhead_last
= 0;
4501 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
4502 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
4503 ext4_fsblk_t overhead
= 0;
4506 * Compute the overhead (FS structures). This is constant
4507 * for a given filesystem unless the number of block groups
4508 * changes so we cache the previous value until it does.
4512 * All of the blocks before first_data_block are
4515 overhead
= le32_to_cpu(es
->s_first_data_block
);
4518 * Add the overhead attributed to the superblock and
4519 * block group descriptors. If the sparse superblocks
4520 * feature is turned on, then not all groups have this.
4522 for (i
= 0; i
< ngroups
; i
++) {
4523 overhead
+= ext4_bg_has_super(sb
, i
) +
4524 ext4_bg_num_gdb(sb
, i
);
4529 * Every block group has an inode bitmap, a block
4530 * bitmap, and an inode table.
4532 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
4533 sbi
->s_overhead_last
= overhead
;
4535 sbi
->s_blocks_last
= ext4_blocks_count(es
);
4538 buf
->f_type
= EXT4_SUPER_MAGIC
;
4539 buf
->f_bsize
= sb
->s_blocksize
;
4540 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
4541 bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
4542 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
4543 /* prevent underflow in case that few free space is available */
4544 buf
->f_bfree
= max_t(s64
, bfree
, 0);
4545 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
4546 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
4548 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4549 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4550 buf
->f_namelen
= EXT4_NAME_LEN
;
4551 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4552 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4553 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4554 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4559 /* Helper function for writing quotas on sync - we need to start transaction
4560 * before quota file is locked for write. Otherwise the are possible deadlocks:
4561 * Process 1 Process 2
4562 * ext4_create() quota_sync()
4563 * jbd2_journal_start() write_dquot()
4564 * dquot_initialize() down(dqio_mutex)
4565 * down(dqio_mutex) jbd2_journal_start()
4571 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4573 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
4576 static int ext4_write_dquot(struct dquot
*dquot
)
4580 struct inode
*inode
;
4582 inode
= dquot_to_inode(dquot
);
4583 handle
= ext4_journal_start(inode
,
4584 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4586 return PTR_ERR(handle
);
4587 ret
= dquot_commit(dquot
);
4588 err
= ext4_journal_stop(handle
);
4594 static int ext4_acquire_dquot(struct dquot
*dquot
)
4599 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4600 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4602 return PTR_ERR(handle
);
4603 ret
= dquot_acquire(dquot
);
4604 err
= ext4_journal_stop(handle
);
4610 static int ext4_release_dquot(struct dquot
*dquot
)
4615 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4616 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4617 if (IS_ERR(handle
)) {
4618 /* Release dquot anyway to avoid endless cycle in dqput() */
4619 dquot_release(dquot
);
4620 return PTR_ERR(handle
);
4622 ret
= dquot_release(dquot
);
4623 err
= ext4_journal_stop(handle
);
4629 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4631 /* Are we journaling quotas? */
4632 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4633 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4634 dquot_mark_dquot_dirty(dquot
);
4635 return ext4_write_dquot(dquot
);
4637 return dquot_mark_dquot_dirty(dquot
);
4641 static int ext4_write_info(struct super_block
*sb
, int type
)
4646 /* Data block + inode block */
4647 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
4649 return PTR_ERR(handle
);
4650 ret
= dquot_commit_info(sb
, type
);
4651 err
= ext4_journal_stop(handle
);
4658 * Turn on quotas during mount time - we need to find
4659 * the quota file and such...
4661 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4663 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4664 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4668 * Standard function to be called on quota_on
4670 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4675 if (!test_opt(sb
, QUOTA
))
4678 /* Quotafile not on the same filesystem? */
4679 if (path
->mnt
->mnt_sb
!= sb
)
4681 /* Journaling quota? */
4682 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4683 /* Quotafile not in fs root? */
4684 if (path
->dentry
->d_parent
!= sb
->s_root
)
4685 ext4_msg(sb
, KERN_WARNING
,
4686 "Quota file not on filesystem root. "
4687 "Journaled quota will not work");
4691 * When we journal data on quota file, we have to flush journal to see
4692 * all updates to the file when we bypass pagecache...
4694 if (EXT4_SB(sb
)->s_journal
&&
4695 ext4_should_journal_data(path
->dentry
->d_inode
)) {
4697 * We don't need to lock updates but journal_flush() could
4698 * otherwise be livelocked...
4700 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4701 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4702 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4707 return dquot_quota_on(sb
, type
, format_id
, path
);
4710 static int ext4_quota_off(struct super_block
*sb
, int type
)
4712 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4715 /* Force all delayed allocation blocks to be allocated.
4716 * Caller already holds s_umount sem */
4717 if (test_opt(sb
, DELALLOC
))
4718 sync_filesystem(sb
);
4723 /* Update modification times of quota files when userspace can
4724 * start looking at them */
4725 handle
= ext4_journal_start(inode
, 1);
4728 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4729 ext4_mark_inode_dirty(handle
, inode
);
4730 ext4_journal_stop(handle
);
4733 return dquot_quota_off(sb
, type
);
4736 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4737 * acquiring the locks... As quota files are never truncated and quota code
4738 * itself serializes the operations (and no one else should touch the files)
4739 * we don't have to be afraid of races */
4740 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
4741 size_t len
, loff_t off
)
4743 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4744 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4746 int offset
= off
& (sb
->s_blocksize
- 1);
4749 struct buffer_head
*bh
;
4750 loff_t i_size
= i_size_read(inode
);
4754 if (off
+len
> i_size
)
4757 while (toread
> 0) {
4758 tocopy
= sb
->s_blocksize
- offset
< toread
?
4759 sb
->s_blocksize
- offset
: toread
;
4760 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
4763 if (!bh
) /* A hole? */
4764 memset(data
, 0, tocopy
);
4766 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4776 /* Write to quotafile (we know the transaction is already started and has
4777 * enough credits) */
4778 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4779 const char *data
, size_t len
, loff_t off
)
4781 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4782 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4784 int offset
= off
& (sb
->s_blocksize
- 1);
4785 struct buffer_head
*bh
;
4786 handle_t
*handle
= journal_current_handle();
4788 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4789 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4790 " cancelled because transaction is not started",
4791 (unsigned long long)off
, (unsigned long long)len
);
4795 * Since we account only one data block in transaction credits,
4796 * then it is impossible to cross a block boundary.
4798 if (sb
->s_blocksize
- offset
< len
) {
4799 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4800 " cancelled because not block aligned",
4801 (unsigned long long)off
, (unsigned long long)len
);
4805 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4806 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4809 err
= ext4_journal_get_write_access(handle
, bh
);
4815 memcpy(bh
->b_data
+offset
, data
, len
);
4816 flush_dcache_page(bh
->b_page
);
4818 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4822 mutex_unlock(&inode
->i_mutex
);
4825 if (inode
->i_size
< off
+ len
) {
4826 i_size_write(inode
, off
+ len
);
4827 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4828 ext4_mark_inode_dirty(handle
, inode
);
4830 mutex_unlock(&inode
->i_mutex
);
4836 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
4837 const char *dev_name
, void *data
)
4839 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
4842 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4843 static inline void register_as_ext2(void)
4845 int err
= register_filesystem(&ext2_fs_type
);
4848 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4851 static inline void unregister_as_ext2(void)
4853 unregister_filesystem(&ext2_fs_type
);
4856 static inline int ext2_feature_set_ok(struct super_block
*sb
)
4858 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
4860 if (sb
->s_flags
& MS_RDONLY
)
4862 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
4866 MODULE_ALIAS("ext2");
4868 static inline void register_as_ext2(void) { }
4869 static inline void unregister_as_ext2(void) { }
4870 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
4873 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4874 static inline void register_as_ext3(void)
4876 int err
= register_filesystem(&ext3_fs_type
);
4879 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4882 static inline void unregister_as_ext3(void)
4884 unregister_filesystem(&ext3_fs_type
);
4887 static inline int ext3_feature_set_ok(struct super_block
*sb
)
4889 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
4891 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
4893 if (sb
->s_flags
& MS_RDONLY
)
4895 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
4899 MODULE_ALIAS("ext3");
4901 static inline void register_as_ext3(void) { }
4902 static inline void unregister_as_ext3(void) { }
4903 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
4906 static struct file_system_type ext4_fs_type
= {
4907 .owner
= THIS_MODULE
,
4909 .mount
= ext4_mount
,
4910 .kill_sb
= kill_block_super
,
4911 .fs_flags
= FS_REQUIRES_DEV
,
4914 static int __init
ext4_init_feat_adverts(void)
4916 struct ext4_features
*ef
;
4919 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
4923 ef
->f_kobj
.kset
= ext4_kset
;
4924 init_completion(&ef
->f_kobj_unregister
);
4925 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
4938 static void ext4_exit_feat_adverts(void)
4940 kobject_put(&ext4_feat
->f_kobj
);
4941 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
4945 /* Shared across all ext4 file systems */
4946 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
4947 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
4949 static int __init
ext4_init_fs(void)
4953 ext4_check_flag_values();
4955 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
4956 mutex_init(&ext4__aio_mutex
[i
]);
4957 init_waitqueue_head(&ext4__ioend_wq
[i
]);
4960 err
= ext4_init_pageio();
4963 err
= ext4_init_system_zone();
4966 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4969 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4970 if (!ext4_proc_root
)
4973 err
= ext4_init_feat_adverts();
4977 err
= ext4_init_mballoc();
4981 err
= ext4_init_xattr();
4984 err
= init_inodecache();
4989 err
= register_filesystem(&ext4_fs_type
);
4993 ext4_li_info
= NULL
;
4994 mutex_init(&ext4_li_mtx
);
4997 unregister_as_ext2();
4998 unregister_as_ext3();
4999 destroy_inodecache();
5003 ext4_exit_mballoc();
5005 ext4_exit_feat_adverts();
5007 remove_proc_entry("fs/ext4", NULL
);
5009 kset_unregister(ext4_kset
);
5011 ext4_exit_system_zone();
5017 static void __exit
ext4_exit_fs(void)
5019 ext4_destroy_lazyinit_thread();
5020 unregister_as_ext2();
5021 unregister_as_ext3();
5022 unregister_filesystem(&ext4_fs_type
);
5023 destroy_inodecache();
5025 ext4_exit_mballoc();
5026 ext4_exit_feat_adverts();
5027 remove_proc_entry("fs/ext4", NULL
);
5028 kset_unregister(ext4_kset
);
5029 ext4_exit_system_zone();
5033 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5034 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5035 MODULE_LICENSE("GPL");
5036 module_init(ext4_init_fs
)
5037 module_exit(ext4_exit_fs
)