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 if (sb
->s_flags
& MS_RDONLY
)
273 return ERR_PTR(-EROFS
);
275 journal
= EXT4_SB(sb
)->s_journal
;
276 handle
= ext4_journal_current_handle();
279 * If a handle has been started, it should be allowed to
280 * finish, otherwise deadlock could happen between freeze
281 * and others(e.g. truncate) due to the restart of the
282 * journal handle if the filesystem is forzen and active
283 * handles are not stopped.
286 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
289 return ext4_get_nojournal();
291 * Special case here: if the journal has aborted behind our
292 * backs (eg. EIO in the commit thread), then we still need to
293 * take the FS itself readonly cleanly.
295 if (is_journal_aborted(journal
)) {
296 ext4_abort(sb
, "Detected aborted journal");
297 return ERR_PTR(-EROFS
);
299 return jbd2_journal_start(journal
, nblocks
);
303 * The only special thing we need to do here is to make sure that all
304 * jbd2_journal_stop calls result in the superblock being marked dirty, so
305 * that sync() will call the filesystem's write_super callback if
308 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
)
310 struct super_block
*sb
;
314 if (!ext4_handle_valid(handle
)) {
315 ext4_put_nojournal(handle
);
318 sb
= handle
->h_transaction
->t_journal
->j_private
;
320 rc
= jbd2_journal_stop(handle
);
325 __ext4_std_error(sb
, where
, line
, err
);
329 void ext4_journal_abort_handle(const char *caller
, unsigned int line
,
330 const char *err_fn
, struct buffer_head
*bh
,
331 handle_t
*handle
, int err
)
334 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
336 BUG_ON(!ext4_handle_valid(handle
));
339 BUFFER_TRACE(bh
, "abort");
344 if (is_handle_aborted(handle
))
347 printk(KERN_ERR
"%s:%d: aborting transaction: %s in %s\n",
348 caller
, line
, errstr
, err_fn
);
350 jbd2_journal_abort_handle(handle
);
353 static void __save_error_info(struct super_block
*sb
, const char *func
,
356 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
358 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
359 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
360 es
->s_last_error_time
= cpu_to_le32(get_seconds());
361 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
362 es
->s_last_error_line
= cpu_to_le32(line
);
363 if (!es
->s_first_error_time
) {
364 es
->s_first_error_time
= es
->s_last_error_time
;
365 strncpy(es
->s_first_error_func
, func
,
366 sizeof(es
->s_first_error_func
));
367 es
->s_first_error_line
= cpu_to_le32(line
);
368 es
->s_first_error_ino
= es
->s_last_error_ino
;
369 es
->s_first_error_block
= es
->s_last_error_block
;
372 * Start the daily error reporting function if it hasn't been
375 if (!es
->s_error_count
)
376 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
377 es
->s_error_count
= cpu_to_le32(le32_to_cpu(es
->s_error_count
) + 1);
380 static void save_error_info(struct super_block
*sb
, const char *func
,
383 __save_error_info(sb
, func
, line
);
384 ext4_commit_super(sb
, 1);
388 /* Deal with the reporting of failure conditions on a filesystem such as
389 * inconsistencies detected or read IO failures.
391 * On ext2, we can store the error state of the filesystem in the
392 * superblock. That is not possible on ext4, because we may have other
393 * write ordering constraints on the superblock which prevent us from
394 * writing it out straight away; and given that the journal is about to
395 * be aborted, we can't rely on the current, or future, transactions to
396 * write out the superblock safely.
398 * We'll just use the jbd2_journal_abort() error code to record an error in
399 * the journal instead. On recovery, the journal will complain about
400 * that error until we've noted it down and cleared it.
403 static void ext4_handle_error(struct super_block
*sb
)
405 if (sb
->s_flags
& MS_RDONLY
)
408 if (!test_opt(sb
, ERRORS_CONT
)) {
409 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
411 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
413 jbd2_journal_abort(journal
, -EIO
);
415 if (test_opt(sb
, ERRORS_RO
)) {
416 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
417 sb
->s_flags
|= MS_RDONLY
;
419 if (test_opt(sb
, ERRORS_PANIC
))
420 panic("EXT4-fs (device %s): panic forced after error\n",
424 void __ext4_error(struct super_block
*sb
, const char *function
,
425 unsigned int line
, const char *fmt
, ...)
427 struct va_format vaf
;
433 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
434 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
437 ext4_handle_error(sb
);
440 void ext4_error_inode(struct inode
*inode
, const char *function
,
441 unsigned int line
, ext4_fsblk_t block
,
442 const char *fmt
, ...)
445 struct va_format vaf
;
446 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
448 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
449 es
->s_last_error_block
= cpu_to_le64(block
);
450 save_error_info(inode
->i_sb
, function
, line
);
454 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: inode #%lu: ",
455 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
457 printk(KERN_CONT
"block %llu: ", block
);
458 printk(KERN_CONT
"comm %s: %pV\n", current
->comm
, &vaf
);
461 ext4_handle_error(inode
->i_sb
);
464 void ext4_error_file(struct file
*file
, const char *function
,
465 unsigned int line
, ext4_fsblk_t block
,
466 const char *fmt
, ...)
469 struct va_format vaf
;
470 struct ext4_super_block
*es
;
471 struct inode
*inode
= file
->f_dentry
->d_inode
;
472 char pathname
[80], *path
;
474 es
= EXT4_SB(inode
->i_sb
)->s_es
;
475 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
476 save_error_info(inode
->i_sb
, function
, line
);
477 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
482 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
484 printk(KERN_CONT
"block %llu: ", block
);
488 printk(KERN_CONT
"comm %s: path %s: %pV\n", current
->comm
, path
, &vaf
);
491 ext4_handle_error(inode
->i_sb
);
494 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
501 errstr
= "IO failure";
504 errstr
= "Out of memory";
507 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
508 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
509 errstr
= "Journal has aborted";
511 errstr
= "Readonly filesystem";
514 /* If the caller passed in an extra buffer for unknown
515 * errors, textualise them now. Else we just return
518 /* Check for truncated error codes... */
519 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
528 /* __ext4_std_error decodes expected errors from journaling functions
529 * automatically and invokes the appropriate error response. */
531 void __ext4_std_error(struct super_block
*sb
, const char *function
,
532 unsigned int line
, int errno
)
537 /* Special case: if the error is EROFS, and we're not already
538 * inside a transaction, then there's really no point in logging
540 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
541 (sb
->s_flags
& MS_RDONLY
))
544 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
545 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb
->s_id
, function
, line
, errstr
);
547 save_error_info(sb
, function
, line
);
549 ext4_handle_error(sb
);
553 * ext4_abort is a much stronger failure handler than ext4_error. The
554 * abort function may be used to deal with unrecoverable failures such
555 * as journal IO errors or ENOMEM at a critical moment in log management.
557 * We unconditionally force the filesystem into an ABORT|READONLY state,
558 * unless the error response on the fs has been set to panic in which
559 * case we take the easy way out and panic immediately.
562 void __ext4_abort(struct super_block
*sb
, const char *function
,
563 unsigned int line
, const char *fmt
, ...)
567 save_error_info(sb
, function
, line
);
569 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
575 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
576 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
577 sb
->s_flags
|= MS_RDONLY
;
578 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
579 if (EXT4_SB(sb
)->s_journal
)
580 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
581 save_error_info(sb
, function
, line
);
583 if (test_opt(sb
, ERRORS_PANIC
))
584 panic("EXT4-fs panic from previous error\n");
587 void ext4_msg(struct super_block
*sb
, const char *prefix
, const char *fmt
, ...)
589 struct va_format vaf
;
595 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
599 void __ext4_warning(struct super_block
*sb
, const char *function
,
600 unsigned int line
, const char *fmt
, ...)
602 struct va_format vaf
;
608 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
609 sb
->s_id
, function
, line
, &vaf
);
613 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
614 struct super_block
*sb
, ext4_group_t grp
,
615 unsigned long ino
, ext4_fsblk_t block
,
616 const char *fmt
, ...)
620 struct va_format vaf
;
622 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
624 es
->s_last_error_ino
= cpu_to_le32(ino
);
625 es
->s_last_error_block
= cpu_to_le64(block
);
626 __save_error_info(sb
, function
, line
);
632 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
633 sb
->s_id
, function
, line
, grp
);
635 printk(KERN_CONT
"inode %lu: ", ino
);
637 printk(KERN_CONT
"block %llu:", (unsigned long long) block
);
638 printk(KERN_CONT
"%pV\n", &vaf
);
641 if (test_opt(sb
, ERRORS_CONT
)) {
642 ext4_commit_super(sb
, 0);
646 ext4_unlock_group(sb
, grp
);
647 ext4_handle_error(sb
);
649 * We only get here in the ERRORS_RO case; relocking the group
650 * may be dangerous, but nothing bad will happen since the
651 * filesystem will have already been marked read/only and the
652 * journal has been aborted. We return 1 as a hint to callers
653 * who might what to use the return value from
654 * ext4_grp_locked_error() to distinguish between the
655 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
656 * aggressively from the ext4 function in question, with a
657 * more appropriate error code.
659 ext4_lock_group(sb
, grp
);
663 void ext4_update_dynamic_rev(struct super_block
*sb
)
665 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
667 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
671 "updating to rev %d because of new feature flag, "
672 "running e2fsck is recommended",
675 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
676 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
677 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
678 /* leave es->s_feature_*compat flags alone */
679 /* es->s_uuid will be set by e2fsck if empty */
682 * The rest of the superblock fields should be zero, and if not it
683 * means they are likely already in use, so leave them alone. We
684 * can leave it up to e2fsck to clean up any inconsistencies there.
689 * Open the external journal device
691 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
693 struct block_device
*bdev
;
694 char b
[BDEVNAME_SIZE
];
696 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
702 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
703 __bdevname(dev
, b
), PTR_ERR(bdev
));
708 * Release the journal device
710 static int ext4_blkdev_put(struct block_device
*bdev
)
712 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
715 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
717 struct block_device
*bdev
;
720 bdev
= sbi
->journal_bdev
;
722 ret
= ext4_blkdev_put(bdev
);
723 sbi
->journal_bdev
= NULL
;
728 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
730 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
733 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
737 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
738 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
740 printk(KERN_ERR
"sb_info orphan list:\n");
741 list_for_each(l
, &sbi
->s_orphan
) {
742 struct inode
*inode
= orphan_list_entry(l
);
744 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
745 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
746 inode
->i_mode
, inode
->i_nlink
,
751 static void ext4_put_super(struct super_block
*sb
)
753 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
754 struct ext4_super_block
*es
= sbi
->s_es
;
757 ext4_unregister_li_request(sb
);
758 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
760 flush_workqueue(sbi
->dio_unwritten_wq
);
761 destroy_workqueue(sbi
->dio_unwritten_wq
);
765 ext4_commit_super(sb
, 1);
767 if (sbi
->s_journal
) {
768 err
= jbd2_journal_destroy(sbi
->s_journal
);
769 sbi
->s_journal
= NULL
;
771 ext4_abort(sb
, "Couldn't clean up the journal");
774 del_timer(&sbi
->s_err_report
);
775 ext4_release_system_zone(sb
);
777 ext4_ext_release(sb
);
778 ext4_xattr_put_super(sb
);
780 if (!(sb
->s_flags
& MS_RDONLY
)) {
781 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
782 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
783 ext4_commit_super(sb
, 1);
786 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
788 kobject_del(&sbi
->s_kobj
);
790 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
791 brelse(sbi
->s_group_desc
[i
]);
792 kfree(sbi
->s_group_desc
);
793 if (is_vmalloc_addr(sbi
->s_flex_groups
))
794 vfree(sbi
->s_flex_groups
);
796 kfree(sbi
->s_flex_groups
);
797 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
798 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
799 percpu_counter_destroy(&sbi
->s_dirs_counter
);
800 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
803 for (i
= 0; i
< MAXQUOTAS
; i
++)
804 kfree(sbi
->s_qf_names
[i
]);
807 /* Debugging code just in case the in-memory inode orphan list
808 * isn't empty. The on-disk one can be non-empty if we've
809 * detected an error and taken the fs readonly, but the
810 * in-memory list had better be clean by this point. */
811 if (!list_empty(&sbi
->s_orphan
))
812 dump_orphan_list(sb
, sbi
);
813 J_ASSERT(list_empty(&sbi
->s_orphan
));
815 invalidate_bdev(sb
->s_bdev
);
816 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
818 * Invalidate the journal device's buffers. We don't want them
819 * floating about in memory - the physical journal device may
820 * hotswapped, and it breaks the `ro-after' testing code.
822 sync_blockdev(sbi
->journal_bdev
);
823 invalidate_bdev(sbi
->journal_bdev
);
824 ext4_blkdev_remove(sbi
);
827 kthread_stop(sbi
->s_mmp_tsk
);
828 sb
->s_fs_info
= NULL
;
830 * Now that we are completely done shutting down the
831 * superblock, we need to actually destroy the kobject.
834 kobject_put(&sbi
->s_kobj
);
835 wait_for_completion(&sbi
->s_kobj_unregister
);
836 kfree(sbi
->s_blockgroup_lock
);
840 static struct kmem_cache
*ext4_inode_cachep
;
843 * Called inside transaction, so use GFP_NOFS
845 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
847 struct ext4_inode_info
*ei
;
849 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
853 ei
->vfs_inode
.i_version
= 1;
854 ei
->vfs_inode
.i_data
.writeback_index
= 0;
855 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
856 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
857 spin_lock_init(&ei
->i_prealloc_lock
);
858 ei
->i_reserved_data_blocks
= 0;
859 ei
->i_reserved_meta_blocks
= 0;
860 ei
->i_allocated_meta_blocks
= 0;
861 ei
->i_da_metadata_calc_len
= 0;
862 spin_lock_init(&(ei
->i_block_reservation_lock
));
864 ei
->i_reserved_quota
= 0;
867 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
868 spin_lock_init(&ei
->i_completed_io_lock
);
869 ei
->cur_aio_dio
= NULL
;
871 ei
->i_datasync_tid
= 0;
872 atomic_set(&ei
->i_ioend_count
, 0);
873 atomic_set(&ei
->i_aiodio_unwritten
, 0);
875 return &ei
->vfs_inode
;
878 static int ext4_drop_inode(struct inode
*inode
)
880 int drop
= generic_drop_inode(inode
);
882 trace_ext4_drop_inode(inode
, drop
);
886 static void ext4_i_callback(struct rcu_head
*head
)
888 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
889 INIT_LIST_HEAD(&inode
->i_dentry
);
890 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
893 static void ext4_destroy_inode(struct inode
*inode
)
895 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
896 ext4_msg(inode
->i_sb
, KERN_ERR
,
897 "Inode %lu (%p): orphan list check failed!",
898 inode
->i_ino
, EXT4_I(inode
));
899 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
900 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
904 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
907 static void init_once(void *foo
)
909 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
911 INIT_LIST_HEAD(&ei
->i_orphan
);
912 #ifdef CONFIG_EXT4_FS_XATTR
913 init_rwsem(&ei
->xattr_sem
);
915 init_rwsem(&ei
->i_data_sem
);
916 inode_init_once(&ei
->vfs_inode
);
919 static int init_inodecache(void)
921 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
922 sizeof(struct ext4_inode_info
),
923 0, (SLAB_RECLAIM_ACCOUNT
|
926 if (ext4_inode_cachep
== NULL
)
931 static void destroy_inodecache(void)
933 kmem_cache_destroy(ext4_inode_cachep
);
936 void ext4_clear_inode(struct inode
*inode
)
938 invalidate_inode_buffers(inode
);
939 end_writeback(inode
);
941 ext4_discard_preallocations(inode
);
942 if (EXT4_I(inode
)->jinode
) {
943 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
944 EXT4_I(inode
)->jinode
);
945 jbd2_free_inode(EXT4_I(inode
)->jinode
);
946 EXT4_I(inode
)->jinode
= NULL
;
950 static inline void ext4_show_quota_options(struct seq_file
*seq
,
951 struct super_block
*sb
)
953 #if defined(CONFIG_QUOTA)
954 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
956 if (sbi
->s_jquota_fmt
) {
959 switch (sbi
->s_jquota_fmt
) {
970 seq_printf(seq
, ",jqfmt=%s", fmtname
);
973 if (sbi
->s_qf_names
[USRQUOTA
])
974 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
976 if (sbi
->s_qf_names
[GRPQUOTA
])
977 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
979 if (test_opt(sb
, USRQUOTA
))
980 seq_puts(seq
, ",usrquota");
982 if (test_opt(sb
, GRPQUOTA
))
983 seq_puts(seq
, ",grpquota");
989 * - it's set to a non-default value OR
990 * - if the per-sb default is different from the global default
992 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
995 unsigned long def_mount_opts
;
996 struct super_block
*sb
= vfs
->mnt_sb
;
997 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
998 struct ext4_super_block
*es
= sbi
->s_es
;
1000 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1001 def_errors
= le16_to_cpu(es
->s_errors
);
1003 if (sbi
->s_sb_block
!= 1)
1004 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
1005 if (test_opt(sb
, MINIX_DF
))
1006 seq_puts(seq
, ",minixdf");
1007 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
1008 seq_puts(seq
, ",grpid");
1009 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
1010 seq_puts(seq
, ",nogrpid");
1011 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
1012 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
1013 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
1015 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
1016 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
1017 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
1019 if (test_opt(sb
, ERRORS_RO
)) {
1020 if (def_errors
== EXT4_ERRORS_PANIC
||
1021 def_errors
== EXT4_ERRORS_CONTINUE
) {
1022 seq_puts(seq
, ",errors=remount-ro");
1025 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1026 seq_puts(seq
, ",errors=continue");
1027 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1028 seq_puts(seq
, ",errors=panic");
1029 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
1030 seq_puts(seq
, ",nouid32");
1031 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
1032 seq_puts(seq
, ",debug");
1033 if (test_opt(sb
, OLDALLOC
))
1034 seq_puts(seq
, ",oldalloc");
1035 #ifdef CONFIG_EXT4_FS_XATTR
1036 if (test_opt(sb
, XATTR_USER
))
1037 seq_puts(seq
, ",user_xattr");
1038 if (!test_opt(sb
, XATTR_USER
))
1039 seq_puts(seq
, ",nouser_xattr");
1041 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1042 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
1043 seq_puts(seq
, ",acl");
1044 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
1045 seq_puts(seq
, ",noacl");
1047 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
1048 seq_printf(seq
, ",commit=%u",
1049 (unsigned) (sbi
->s_commit_interval
/ HZ
));
1051 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
1052 seq_printf(seq
, ",min_batch_time=%u",
1053 (unsigned) sbi
->s_min_batch_time
);
1055 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
1056 seq_printf(seq
, ",max_batch_time=%u",
1057 (unsigned) sbi
->s_min_batch_time
);
1061 * We're changing the default of barrier mount option, so
1062 * let's always display its mount state so it's clear what its
1065 seq_puts(seq
, ",barrier=");
1066 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
1067 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
1068 seq_puts(seq
, ",journal_async_commit");
1069 else if (test_opt(sb
, JOURNAL_CHECKSUM
))
1070 seq_puts(seq
, ",journal_checksum");
1071 if (test_opt(sb
, I_VERSION
))
1072 seq_puts(seq
, ",i_version");
1073 if (!test_opt(sb
, DELALLOC
) &&
1074 !(def_mount_opts
& EXT4_DEFM_NODELALLOC
))
1075 seq_puts(seq
, ",nodelalloc");
1077 if (!test_opt(sb
, MBLK_IO_SUBMIT
))
1078 seq_puts(seq
, ",nomblk_io_submit");
1080 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
1082 * journal mode get enabled in different ways
1083 * So just print the value even if we didn't specify it
1085 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1086 seq_puts(seq
, ",data=journal");
1087 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1088 seq_puts(seq
, ",data=ordered");
1089 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1090 seq_puts(seq
, ",data=writeback");
1092 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1093 seq_printf(seq
, ",inode_readahead_blks=%u",
1094 sbi
->s_inode_readahead_blks
);
1096 if (test_opt(sb
, DATA_ERR_ABORT
))
1097 seq_puts(seq
, ",data_err=abort");
1099 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
1100 seq_puts(seq
, ",noauto_da_alloc");
1102 if (test_opt(sb
, DISCARD
) && !(def_mount_opts
& EXT4_DEFM_DISCARD
))
1103 seq_puts(seq
, ",discard");
1105 if (test_opt(sb
, NOLOAD
))
1106 seq_puts(seq
, ",norecovery");
1108 if (test_opt(sb
, DIOREAD_NOLOCK
))
1109 seq_puts(seq
, ",dioread_nolock");
1111 if (test_opt(sb
, BLOCK_VALIDITY
) &&
1112 !(def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
))
1113 seq_puts(seq
, ",block_validity");
1115 if (!test_opt(sb
, INIT_INODE_TABLE
))
1116 seq_puts(seq
, ",noinit_inode_table");
1117 else if (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)
1118 seq_printf(seq
, ",init_inode_table=%u",
1119 (unsigned) sbi
->s_li_wait_mult
);
1121 ext4_show_quota_options(seq
, sb
);
1126 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1127 u64 ino
, u32 generation
)
1129 struct inode
*inode
;
1131 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1132 return ERR_PTR(-ESTALE
);
1133 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1134 return ERR_PTR(-ESTALE
);
1136 /* iget isn't really right if the inode is currently unallocated!!
1138 * ext4_read_inode will return a bad_inode if the inode had been
1139 * deleted, so we should be safe.
1141 * Currently we don't know the generation for parent directory, so
1142 * a generation of 0 means "accept any"
1144 inode
= ext4_iget(sb
, ino
);
1146 return ERR_CAST(inode
);
1147 if (generation
&& inode
->i_generation
!= generation
) {
1149 return ERR_PTR(-ESTALE
);
1155 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1156 int fh_len
, int fh_type
)
1158 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1159 ext4_nfs_get_inode
);
1162 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1163 int fh_len
, int fh_type
)
1165 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1166 ext4_nfs_get_inode
);
1170 * Try to release metadata pages (indirect blocks, directories) which are
1171 * mapped via the block device. Since these pages could have journal heads
1172 * which would prevent try_to_free_buffers() from freeing them, we must use
1173 * jbd2 layer's try_to_free_buffers() function to release them.
1175 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1178 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1180 WARN_ON(PageChecked(page
));
1181 if (!page_has_buffers(page
))
1184 return jbd2_journal_try_to_free_buffers(journal
, page
,
1185 wait
& ~__GFP_WAIT
);
1186 return try_to_free_buffers(page
);
1190 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1191 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1193 static int ext4_write_dquot(struct dquot
*dquot
);
1194 static int ext4_acquire_dquot(struct dquot
*dquot
);
1195 static int ext4_release_dquot(struct dquot
*dquot
);
1196 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1197 static int ext4_write_info(struct super_block
*sb
, int type
);
1198 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1200 static int ext4_quota_off(struct super_block
*sb
, int type
);
1201 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1202 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1203 size_t len
, loff_t off
);
1204 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1205 const char *data
, size_t len
, loff_t off
);
1207 static const struct dquot_operations ext4_quota_operations
= {
1208 .get_reserved_space
= ext4_get_reserved_space
,
1209 .write_dquot
= ext4_write_dquot
,
1210 .acquire_dquot
= ext4_acquire_dquot
,
1211 .release_dquot
= ext4_release_dquot
,
1212 .mark_dirty
= ext4_mark_dquot_dirty
,
1213 .write_info
= ext4_write_info
,
1214 .alloc_dquot
= dquot_alloc
,
1215 .destroy_dquot
= dquot_destroy
,
1218 static const struct quotactl_ops ext4_qctl_operations
= {
1219 .quota_on
= ext4_quota_on
,
1220 .quota_off
= ext4_quota_off
,
1221 .quota_sync
= dquot_quota_sync
,
1222 .get_info
= dquot_get_dqinfo
,
1223 .set_info
= dquot_set_dqinfo
,
1224 .get_dqblk
= dquot_get_dqblk
,
1225 .set_dqblk
= dquot_set_dqblk
1229 static const struct super_operations ext4_sops
= {
1230 .alloc_inode
= ext4_alloc_inode
,
1231 .destroy_inode
= ext4_destroy_inode
,
1232 .write_inode
= ext4_write_inode
,
1233 .dirty_inode
= ext4_dirty_inode
,
1234 .drop_inode
= ext4_drop_inode
,
1235 .evict_inode
= ext4_evict_inode
,
1236 .put_super
= ext4_put_super
,
1237 .sync_fs
= ext4_sync_fs
,
1238 .freeze_fs
= ext4_freeze
,
1239 .unfreeze_fs
= ext4_unfreeze
,
1240 .statfs
= ext4_statfs
,
1241 .remount_fs
= ext4_remount
,
1242 .show_options
= ext4_show_options
,
1244 .quota_read
= ext4_quota_read
,
1245 .quota_write
= ext4_quota_write
,
1247 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1250 static const struct super_operations ext4_nojournal_sops
= {
1251 .alloc_inode
= ext4_alloc_inode
,
1252 .destroy_inode
= ext4_destroy_inode
,
1253 .write_inode
= ext4_write_inode
,
1254 .dirty_inode
= ext4_dirty_inode
,
1255 .drop_inode
= ext4_drop_inode
,
1256 .evict_inode
= ext4_evict_inode
,
1257 .write_super
= ext4_write_super
,
1258 .put_super
= ext4_put_super
,
1259 .statfs
= ext4_statfs
,
1260 .remount_fs
= ext4_remount
,
1261 .show_options
= ext4_show_options
,
1263 .quota_read
= ext4_quota_read
,
1264 .quota_write
= ext4_quota_write
,
1266 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1269 static const struct export_operations ext4_export_ops
= {
1270 .fh_to_dentry
= ext4_fh_to_dentry
,
1271 .fh_to_parent
= ext4_fh_to_parent
,
1272 .get_parent
= ext4_get_parent
,
1276 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1277 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1278 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1279 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1280 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1281 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1282 Opt_journal_update
, Opt_journal_dev
,
1283 Opt_journal_checksum
, Opt_journal_async_commit
,
1284 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1285 Opt_data_err_abort
, Opt_data_err_ignore
,
1286 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1287 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1288 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1289 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1290 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1291 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1292 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1293 Opt_dioread_nolock
, Opt_dioread_lock
,
1294 Opt_discard
, Opt_nodiscard
,
1295 Opt_init_inode_table
, Opt_noinit_inode_table
,
1298 static const match_table_t tokens
= {
1299 {Opt_bsd_df
, "bsddf"},
1300 {Opt_minix_df
, "minixdf"},
1301 {Opt_grpid
, "grpid"},
1302 {Opt_grpid
, "bsdgroups"},
1303 {Opt_nogrpid
, "nogrpid"},
1304 {Opt_nogrpid
, "sysvgroups"},
1305 {Opt_resgid
, "resgid=%u"},
1306 {Opt_resuid
, "resuid=%u"},
1308 {Opt_err_cont
, "errors=continue"},
1309 {Opt_err_panic
, "errors=panic"},
1310 {Opt_err_ro
, "errors=remount-ro"},
1311 {Opt_nouid32
, "nouid32"},
1312 {Opt_debug
, "debug"},
1313 {Opt_oldalloc
, "oldalloc"},
1314 {Opt_orlov
, "orlov"},
1315 {Opt_user_xattr
, "user_xattr"},
1316 {Opt_nouser_xattr
, "nouser_xattr"},
1318 {Opt_noacl
, "noacl"},
1319 {Opt_noload
, "noload"},
1320 {Opt_noload
, "norecovery"},
1323 {Opt_commit
, "commit=%u"},
1324 {Opt_min_batch_time
, "min_batch_time=%u"},
1325 {Opt_max_batch_time
, "max_batch_time=%u"},
1326 {Opt_journal_update
, "journal=update"},
1327 {Opt_journal_dev
, "journal_dev=%u"},
1328 {Opt_journal_checksum
, "journal_checksum"},
1329 {Opt_journal_async_commit
, "journal_async_commit"},
1330 {Opt_abort
, "abort"},
1331 {Opt_data_journal
, "data=journal"},
1332 {Opt_data_ordered
, "data=ordered"},
1333 {Opt_data_writeback
, "data=writeback"},
1334 {Opt_data_err_abort
, "data_err=abort"},
1335 {Opt_data_err_ignore
, "data_err=ignore"},
1336 {Opt_offusrjquota
, "usrjquota="},
1337 {Opt_usrjquota
, "usrjquota=%s"},
1338 {Opt_offgrpjquota
, "grpjquota="},
1339 {Opt_grpjquota
, "grpjquota=%s"},
1340 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1341 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1342 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1343 {Opt_grpquota
, "grpquota"},
1344 {Opt_noquota
, "noquota"},
1345 {Opt_quota
, "quota"},
1346 {Opt_usrquota
, "usrquota"},
1347 {Opt_barrier
, "barrier=%u"},
1348 {Opt_barrier
, "barrier"},
1349 {Opt_nobarrier
, "nobarrier"},
1350 {Opt_i_version
, "i_version"},
1351 {Opt_stripe
, "stripe=%u"},
1352 {Opt_resize
, "resize"},
1353 {Opt_delalloc
, "delalloc"},
1354 {Opt_nodelalloc
, "nodelalloc"},
1355 {Opt_mblk_io_submit
, "mblk_io_submit"},
1356 {Opt_nomblk_io_submit
, "nomblk_io_submit"},
1357 {Opt_block_validity
, "block_validity"},
1358 {Opt_noblock_validity
, "noblock_validity"},
1359 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1360 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1361 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1362 {Opt_auto_da_alloc
, "auto_da_alloc"},
1363 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1364 {Opt_dioread_nolock
, "dioread_nolock"},
1365 {Opt_dioread_lock
, "dioread_lock"},
1366 {Opt_discard
, "discard"},
1367 {Opt_nodiscard
, "nodiscard"},
1368 {Opt_init_inode_table
, "init_itable=%u"},
1369 {Opt_init_inode_table
, "init_itable"},
1370 {Opt_noinit_inode_table
, "noinit_itable"},
1374 static ext4_fsblk_t
get_sb_block(void **data
)
1376 ext4_fsblk_t sb_block
;
1377 char *options
= (char *) *data
;
1379 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1380 return 1; /* Default location */
1383 /* TODO: use simple_strtoll with >32bit ext4 */
1384 sb_block
= simple_strtoul(options
, &options
, 0);
1385 if (*options
&& *options
!= ',') {
1386 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1390 if (*options
== ',')
1392 *data
= (void *) options
;
1397 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1398 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1399 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1402 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1404 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1407 if (sb_any_quota_loaded(sb
) &&
1408 !sbi
->s_qf_names
[qtype
]) {
1409 ext4_msg(sb
, KERN_ERR
,
1410 "Cannot change journaled "
1411 "quota options when quota turned on");
1414 qname
= match_strdup(args
);
1416 ext4_msg(sb
, KERN_ERR
,
1417 "Not enough memory for storing quotafile name");
1420 if (sbi
->s_qf_names
[qtype
] &&
1421 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1422 ext4_msg(sb
, KERN_ERR
,
1423 "%s quota file already specified", QTYPE2NAME(qtype
));
1427 sbi
->s_qf_names
[qtype
] = qname
;
1428 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1429 ext4_msg(sb
, KERN_ERR
,
1430 "quotafile must be on filesystem root");
1431 kfree(sbi
->s_qf_names
[qtype
]);
1432 sbi
->s_qf_names
[qtype
] = NULL
;
1439 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1442 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1444 if (sb_any_quota_loaded(sb
) &&
1445 sbi
->s_qf_names
[qtype
]) {
1446 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1447 " when quota turned on");
1451 * The space will be released later when all options are confirmed
1454 sbi
->s_qf_names
[qtype
] = NULL
;
1459 static int parse_options(char *options
, struct super_block
*sb
,
1460 unsigned long *journal_devnum
,
1461 unsigned int *journal_ioprio
,
1462 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1464 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1466 substring_t args
[MAX_OPT_ARGS
];
1476 while ((p
= strsep(&options
, ",")) != NULL
) {
1482 * Initialize args struct so we know whether arg was
1483 * found; some options take optional arguments.
1485 args
[0].to
= args
[0].from
= NULL
;
1486 token
= match_token(p
, tokens
, args
);
1489 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1490 clear_opt(sb
, MINIX_DF
);
1493 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1494 set_opt(sb
, MINIX_DF
);
1498 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1503 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1504 clear_opt(sb
, GRPID
);
1508 if (match_int(&args
[0], &option
))
1510 sbi
->s_resuid
= option
;
1513 if (match_int(&args
[0], &option
))
1515 sbi
->s_resgid
= option
;
1518 /* handled by get_sb_block() instead of here */
1519 /* *sb_block = match_int(&args[0]); */
1522 clear_opt(sb
, ERRORS_CONT
);
1523 clear_opt(sb
, ERRORS_RO
);
1524 set_opt(sb
, ERRORS_PANIC
);
1527 clear_opt(sb
, ERRORS_CONT
);
1528 clear_opt(sb
, ERRORS_PANIC
);
1529 set_opt(sb
, ERRORS_RO
);
1532 clear_opt(sb
, ERRORS_RO
);
1533 clear_opt(sb
, ERRORS_PANIC
);
1534 set_opt(sb
, ERRORS_CONT
);
1537 set_opt(sb
, NO_UID32
);
1543 set_opt(sb
, OLDALLOC
);
1546 clear_opt(sb
, OLDALLOC
);
1548 #ifdef CONFIG_EXT4_FS_XATTR
1549 case Opt_user_xattr
:
1550 set_opt(sb
, XATTR_USER
);
1552 case Opt_nouser_xattr
:
1553 clear_opt(sb
, XATTR_USER
);
1556 case Opt_user_xattr
:
1557 case Opt_nouser_xattr
:
1558 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1561 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1563 set_opt(sb
, POSIX_ACL
);
1566 clear_opt(sb
, POSIX_ACL
);
1571 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1574 case Opt_journal_update
:
1576 /* Eventually we will want to be able to create
1577 a journal file here. For now, only allow the
1578 user to specify an existing inode to be the
1581 ext4_msg(sb
, KERN_ERR
,
1582 "Cannot specify journal on remount");
1585 set_opt(sb
, UPDATE_JOURNAL
);
1587 case Opt_journal_dev
:
1589 ext4_msg(sb
, KERN_ERR
,
1590 "Cannot specify journal on remount");
1593 if (match_int(&args
[0], &option
))
1595 *journal_devnum
= option
;
1597 case Opt_journal_checksum
:
1598 set_opt(sb
, JOURNAL_CHECKSUM
);
1600 case Opt_journal_async_commit
:
1601 set_opt(sb
, JOURNAL_ASYNC_COMMIT
);
1602 set_opt(sb
, JOURNAL_CHECKSUM
);
1605 set_opt(sb
, NOLOAD
);
1608 if (match_int(&args
[0], &option
))
1613 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1614 sbi
->s_commit_interval
= HZ
* option
;
1616 case Opt_max_batch_time
:
1617 if (match_int(&args
[0], &option
))
1622 option
= EXT4_DEF_MAX_BATCH_TIME
;
1623 sbi
->s_max_batch_time
= option
;
1625 case Opt_min_batch_time
:
1626 if (match_int(&args
[0], &option
))
1630 sbi
->s_min_batch_time
= option
;
1632 case Opt_data_journal
:
1633 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1635 case Opt_data_ordered
:
1636 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1638 case Opt_data_writeback
:
1639 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1642 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1643 ext4_msg(sb
, KERN_ERR
,
1644 "Cannot change data mode on remount");
1648 clear_opt(sb
, DATA_FLAGS
);
1649 sbi
->s_mount_opt
|= data_opt
;
1652 case Opt_data_err_abort
:
1653 set_opt(sb
, DATA_ERR_ABORT
);
1655 case Opt_data_err_ignore
:
1656 clear_opt(sb
, DATA_ERR_ABORT
);
1660 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1664 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1667 case Opt_offusrjquota
:
1668 if (!clear_qf_name(sb
, USRQUOTA
))
1671 case Opt_offgrpjquota
:
1672 if (!clear_qf_name(sb
, GRPQUOTA
))
1676 case Opt_jqfmt_vfsold
:
1677 qfmt
= QFMT_VFS_OLD
;
1679 case Opt_jqfmt_vfsv0
:
1682 case Opt_jqfmt_vfsv1
:
1685 if (sb_any_quota_loaded(sb
) &&
1686 sbi
->s_jquota_fmt
!= qfmt
) {
1687 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1688 "journaled quota options when "
1692 sbi
->s_jquota_fmt
= qfmt
;
1697 set_opt(sb
, USRQUOTA
);
1701 set_opt(sb
, GRPQUOTA
);
1704 if (sb_any_quota_loaded(sb
)) {
1705 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1706 "options when quota turned on");
1709 clear_opt(sb
, QUOTA
);
1710 clear_opt(sb
, USRQUOTA
);
1711 clear_opt(sb
, GRPQUOTA
);
1717 ext4_msg(sb
, KERN_ERR
,
1718 "quota options not supported");
1722 case Opt_offusrjquota
:
1723 case Opt_offgrpjquota
:
1724 case Opt_jqfmt_vfsold
:
1725 case Opt_jqfmt_vfsv0
:
1726 case Opt_jqfmt_vfsv1
:
1727 ext4_msg(sb
, KERN_ERR
,
1728 "journaled quota options not supported");
1734 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1737 clear_opt(sb
, BARRIER
);
1741 if (match_int(&args
[0], &option
))
1744 option
= 1; /* No argument, default to 1 */
1746 set_opt(sb
, BARRIER
);
1748 clear_opt(sb
, BARRIER
);
1754 ext4_msg(sb
, KERN_ERR
,
1755 "resize option only available "
1759 if (match_int(&args
[0], &option
) != 0)
1761 *n_blocks_count
= option
;
1764 ext4_msg(sb
, KERN_WARNING
,
1765 "Ignoring deprecated nobh option");
1768 ext4_msg(sb
, KERN_WARNING
,
1769 "Ignoring deprecated bh option");
1772 set_opt(sb
, I_VERSION
);
1773 sb
->s_flags
|= MS_I_VERSION
;
1775 case Opt_nodelalloc
:
1776 clear_opt(sb
, DELALLOC
);
1778 case Opt_mblk_io_submit
:
1779 set_opt(sb
, MBLK_IO_SUBMIT
);
1781 case Opt_nomblk_io_submit
:
1782 clear_opt(sb
, MBLK_IO_SUBMIT
);
1785 if (match_int(&args
[0], &option
))
1789 sbi
->s_stripe
= option
;
1792 set_opt(sb
, DELALLOC
);
1794 case Opt_block_validity
:
1795 set_opt(sb
, BLOCK_VALIDITY
);
1797 case Opt_noblock_validity
:
1798 clear_opt(sb
, BLOCK_VALIDITY
);
1800 case Opt_inode_readahead_blks
:
1801 if (match_int(&args
[0], &option
))
1803 if (option
< 0 || option
> (1 << 30))
1805 if (option
&& !is_power_of_2(option
)) {
1806 ext4_msg(sb
, KERN_ERR
,
1807 "EXT4-fs: inode_readahead_blks"
1808 " must be a power of 2");
1811 sbi
->s_inode_readahead_blks
= option
;
1813 case Opt_journal_ioprio
:
1814 if (match_int(&args
[0], &option
))
1816 if (option
< 0 || option
> 7)
1818 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1821 case Opt_noauto_da_alloc
:
1822 set_opt(sb
, NO_AUTO_DA_ALLOC
);
1824 case Opt_auto_da_alloc
:
1826 if (match_int(&args
[0], &option
))
1829 option
= 1; /* No argument, default to 1 */
1831 clear_opt(sb
, NO_AUTO_DA_ALLOC
);
1833 set_opt(sb
,NO_AUTO_DA_ALLOC
);
1836 set_opt(sb
, DISCARD
);
1839 clear_opt(sb
, DISCARD
);
1841 case Opt_dioread_nolock
:
1842 set_opt(sb
, DIOREAD_NOLOCK
);
1844 case Opt_dioread_lock
:
1845 clear_opt(sb
, DIOREAD_NOLOCK
);
1847 case Opt_init_inode_table
:
1848 set_opt(sb
, INIT_INODE_TABLE
);
1850 if (match_int(&args
[0], &option
))
1853 option
= EXT4_DEF_LI_WAIT_MULT
;
1856 sbi
->s_li_wait_mult
= option
;
1858 case Opt_noinit_inode_table
:
1859 clear_opt(sb
, INIT_INODE_TABLE
);
1862 ext4_msg(sb
, KERN_ERR
,
1863 "Unrecognized mount option \"%s\" "
1864 "or missing value", p
);
1869 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1870 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1871 clear_opt(sb
, USRQUOTA
);
1873 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1874 clear_opt(sb
, GRPQUOTA
);
1876 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1877 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1882 if (!sbi
->s_jquota_fmt
) {
1883 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1888 if (sbi
->s_jquota_fmt
) {
1889 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1890 "specified with no journaling "
1899 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1902 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1905 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1906 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1907 "forcing read-only mode");
1912 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1913 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1914 "running e2fsck is recommended");
1915 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1916 ext4_msg(sb
, KERN_WARNING
,
1917 "warning: mounting fs with errors, "
1918 "running e2fsck is recommended");
1919 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1920 le16_to_cpu(es
->s_mnt_count
) >=
1921 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1922 ext4_msg(sb
, KERN_WARNING
,
1923 "warning: maximal mount count reached, "
1924 "running e2fsck is recommended");
1925 else if (le32_to_cpu(es
->s_checkinterval
) &&
1926 (le32_to_cpu(es
->s_lastcheck
) +
1927 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1928 ext4_msg(sb
, KERN_WARNING
,
1929 "warning: checktime reached, "
1930 "running e2fsck is recommended");
1931 if (!sbi
->s_journal
)
1932 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1933 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1934 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1935 le16_add_cpu(&es
->s_mnt_count
, 1);
1936 es
->s_mtime
= cpu_to_le32(get_seconds());
1937 ext4_update_dynamic_rev(sb
);
1939 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1941 ext4_commit_super(sb
, 1);
1942 if (test_opt(sb
, DEBUG
))
1943 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1944 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1946 sbi
->s_groups_count
,
1947 EXT4_BLOCKS_PER_GROUP(sb
),
1948 EXT4_INODES_PER_GROUP(sb
),
1949 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1951 cleancache_init_fs(sb
);
1955 static int ext4_fill_flex_info(struct super_block
*sb
)
1957 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1958 struct ext4_group_desc
*gdp
= NULL
;
1959 ext4_group_t flex_group_count
;
1960 ext4_group_t flex_group
;
1961 int groups_per_flex
= 0;
1965 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1966 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1968 if (groups_per_flex
< 2) {
1969 sbi
->s_log_groups_per_flex
= 0;
1973 /* We allocate both existing and potentially added groups */
1974 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1975 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1976 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1977 size
= flex_group_count
* sizeof(struct flex_groups
);
1978 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1979 if (sbi
->s_flex_groups
== NULL
) {
1980 sbi
->s_flex_groups
= vzalloc(size
);
1981 if (sbi
->s_flex_groups
== NULL
) {
1982 ext4_msg(sb
, KERN_ERR
,
1983 "not enough memory for %u flex groups",
1989 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1990 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1992 flex_group
= ext4_flex_group(sbi
, i
);
1993 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1994 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1995 atomic_add(ext4_free_blks_count(sb
, gdp
),
1996 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1997 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1998 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2006 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
2007 struct ext4_group_desc
*gdp
)
2011 if (sbi
->s_es
->s_feature_ro_compat
&
2012 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
2013 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2014 __le32 le_group
= cpu_to_le32(block_group
);
2016 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2017 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2018 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2019 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2020 /* for checksum of struct ext4_group_desc do the rest...*/
2021 if ((sbi
->s_es
->s_feature_incompat
&
2022 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
2023 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2024 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2025 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2029 return cpu_to_le16(crc
);
2032 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
2033 struct ext4_group_desc
*gdp
)
2035 if ((sbi
->s_es
->s_feature_ro_compat
&
2036 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
2037 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
2043 /* Called at mount-time, super-block is locked */
2044 static int ext4_check_descriptors(struct super_block
*sb
,
2045 ext4_group_t
*first_not_zeroed
)
2047 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2048 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2049 ext4_fsblk_t last_block
;
2050 ext4_fsblk_t block_bitmap
;
2051 ext4_fsblk_t inode_bitmap
;
2052 ext4_fsblk_t inode_table
;
2053 int flexbg_flag
= 0;
2054 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2056 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2059 ext4_debug("Checking group descriptors");
2061 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2062 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2064 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2065 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2067 last_block
= first_block
+
2068 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2070 if ((grp
== sbi
->s_groups_count
) &&
2071 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2074 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2075 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2076 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2077 "Block bitmap for group %u not in group "
2078 "(block %llu)!", i
, block_bitmap
);
2081 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2082 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2083 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2084 "Inode bitmap for group %u not in group "
2085 "(block %llu)!", i
, inode_bitmap
);
2088 inode_table
= ext4_inode_table(sb
, gdp
);
2089 if (inode_table
< first_block
||
2090 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2091 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2092 "Inode table for group %u not in group "
2093 "(block %llu)!", i
, inode_table
);
2096 ext4_lock_group(sb
, i
);
2097 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
2098 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2099 "Checksum for group %u failed (%u!=%u)",
2100 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2101 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2102 if (!(sb
->s_flags
& MS_RDONLY
)) {
2103 ext4_unlock_group(sb
, i
);
2107 ext4_unlock_group(sb
, i
);
2109 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2111 if (NULL
!= first_not_zeroed
)
2112 *first_not_zeroed
= grp
;
2114 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
2115 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2119 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2120 * the superblock) which were deleted from all directories, but held open by
2121 * a process at the time of a crash. We walk the list and try to delete these
2122 * inodes at recovery time (only with a read-write filesystem).
2124 * In order to keep the orphan inode chain consistent during traversal (in
2125 * case of crash during recovery), we link each inode into the superblock
2126 * orphan list_head and handle it the same way as an inode deletion during
2127 * normal operation (which journals the operations for us).
2129 * We only do an iget() and an iput() on each inode, which is very safe if we
2130 * accidentally point at an in-use or already deleted inode. The worst that
2131 * can happen in this case is that we get a "bit already cleared" message from
2132 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2133 * e2fsck was run on this filesystem, and it must have already done the orphan
2134 * inode cleanup for us, so we can safely abort without any further action.
2136 static void ext4_orphan_cleanup(struct super_block
*sb
,
2137 struct ext4_super_block
*es
)
2139 unsigned int s_flags
= sb
->s_flags
;
2140 int nr_orphans
= 0, nr_truncates
= 0;
2144 if (!es
->s_last_orphan
) {
2145 jbd_debug(4, "no orphan inodes to clean up\n");
2149 if (bdev_read_only(sb
->s_bdev
)) {
2150 ext4_msg(sb
, KERN_ERR
, "write access "
2151 "unavailable, skipping orphan cleanup");
2155 /* Check if feature set would not allow a r/w mount */
2156 if (!ext4_feature_set_ok(sb
, 0)) {
2157 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2158 "unknown ROCOMPAT features");
2162 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2163 if (es
->s_last_orphan
)
2164 jbd_debug(1, "Errors on filesystem, "
2165 "clearing orphan list.\n");
2166 es
->s_last_orphan
= 0;
2167 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2171 if (s_flags
& MS_RDONLY
) {
2172 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2173 sb
->s_flags
&= ~MS_RDONLY
;
2176 /* Needed for iput() to work correctly and not trash data */
2177 sb
->s_flags
|= MS_ACTIVE
;
2178 /* Turn on quotas so that they are updated correctly */
2179 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2180 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2181 int ret
= ext4_quota_on_mount(sb
, i
);
2183 ext4_msg(sb
, KERN_ERR
,
2184 "Cannot turn on journaled "
2185 "quota: error %d", ret
);
2190 while (es
->s_last_orphan
) {
2191 struct inode
*inode
;
2193 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2194 if (IS_ERR(inode
)) {
2195 es
->s_last_orphan
= 0;
2199 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2200 dquot_initialize(inode
);
2201 if (inode
->i_nlink
) {
2202 ext4_msg(sb
, KERN_DEBUG
,
2203 "%s: truncating inode %lu to %lld bytes",
2204 __func__
, inode
->i_ino
, inode
->i_size
);
2205 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2206 inode
->i_ino
, inode
->i_size
);
2207 ext4_truncate(inode
);
2210 ext4_msg(sb
, KERN_DEBUG
,
2211 "%s: deleting unreferenced inode %lu",
2212 __func__
, inode
->i_ino
);
2213 jbd_debug(2, "deleting unreferenced inode %lu\n",
2217 iput(inode
); /* The delete magic happens here! */
2220 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2223 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2224 PLURAL(nr_orphans
));
2226 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2227 PLURAL(nr_truncates
));
2229 /* Turn quotas off */
2230 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2231 if (sb_dqopt(sb
)->files
[i
])
2232 dquot_quota_off(sb
, i
);
2235 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2239 * Maximal extent format file size.
2240 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2241 * extent format containers, within a sector_t, and within i_blocks
2242 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2243 * so that won't be a limiting factor.
2245 * However there is other limiting factor. We do store extents in the form
2246 * of starting block and length, hence the resulting length of the extent
2247 * covering maximum file size must fit into on-disk format containers as
2248 * well. Given that length is always by 1 unit bigger than max unit (because
2249 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2251 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2253 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2256 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2258 /* small i_blocks in vfs inode? */
2259 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2261 * CONFIG_LBDAF is not enabled implies the inode
2262 * i_block represent total blocks in 512 bytes
2263 * 32 == size of vfs inode i_blocks * 8
2265 upper_limit
= (1LL << 32) - 1;
2267 /* total blocks in file system block size */
2268 upper_limit
>>= (blkbits
- 9);
2269 upper_limit
<<= blkbits
;
2273 * 32-bit extent-start container, ee_block. We lower the maxbytes
2274 * by one fs block, so ee_len can cover the extent of maximum file
2277 res
= (1LL << 32) - 1;
2280 /* Sanity check against vm- & vfs- imposed limits */
2281 if (res
> upper_limit
)
2288 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2289 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2290 * We need to be 1 filesystem block less than the 2^48 sector limit.
2292 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2294 loff_t res
= EXT4_NDIR_BLOCKS
;
2297 /* This is calculated to be the largest file size for a dense, block
2298 * mapped file such that the file's total number of 512-byte sectors,
2299 * including data and all indirect blocks, does not exceed (2^48 - 1).
2301 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2302 * number of 512-byte sectors of the file.
2305 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2307 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2308 * the inode i_block field represents total file blocks in
2309 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2311 upper_limit
= (1LL << 32) - 1;
2313 /* total blocks in file system block size */
2314 upper_limit
>>= (bits
- 9);
2318 * We use 48 bit ext4_inode i_blocks
2319 * With EXT4_HUGE_FILE_FL set the i_blocks
2320 * represent total number of blocks in
2321 * file system block size
2323 upper_limit
= (1LL << 48) - 1;
2327 /* indirect blocks */
2329 /* double indirect blocks */
2330 meta_blocks
+= 1 + (1LL << (bits
-2));
2331 /* tripple indirect blocks */
2332 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2334 upper_limit
-= meta_blocks
;
2335 upper_limit
<<= bits
;
2337 res
+= 1LL << (bits
-2);
2338 res
+= 1LL << (2*(bits
-2));
2339 res
+= 1LL << (3*(bits
-2));
2341 if (res
> upper_limit
)
2344 if (res
> MAX_LFS_FILESIZE
)
2345 res
= MAX_LFS_FILESIZE
;
2350 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2351 ext4_fsblk_t logical_sb_block
, int nr
)
2353 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2354 ext4_group_t bg
, first_meta_bg
;
2357 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2359 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2361 return logical_sb_block
+ nr
+ 1;
2362 bg
= sbi
->s_desc_per_block
* nr
;
2363 if (ext4_bg_has_super(sb
, bg
))
2366 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2370 * ext4_get_stripe_size: Get the stripe size.
2371 * @sbi: In memory super block info
2373 * If we have specified it via mount option, then
2374 * use the mount option value. If the value specified at mount time is
2375 * greater than the blocks per group use the super block value.
2376 * If the super block value is greater than blocks per group return 0.
2377 * Allocator needs it be less than blocks per group.
2380 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2382 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2383 unsigned long stripe_width
=
2384 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2386 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2387 return sbi
->s_stripe
;
2389 if (stripe_width
<= sbi
->s_blocks_per_group
)
2390 return stripe_width
;
2392 if (stride
<= sbi
->s_blocks_per_group
)
2401 struct attribute attr
;
2402 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2403 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2404 const char *, size_t);
2408 static int parse_strtoul(const char *buf
,
2409 unsigned long max
, unsigned long *value
)
2413 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2414 endp
= skip_spaces(endp
);
2415 if (*endp
|| *value
> max
)
2421 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2422 struct ext4_sb_info
*sbi
,
2425 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2426 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2429 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2430 struct ext4_sb_info
*sbi
, char *buf
)
2432 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2434 if (!sb
->s_bdev
->bd_part
)
2435 return snprintf(buf
, PAGE_SIZE
, "0\n");
2436 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2437 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2438 sbi
->s_sectors_written_start
) >> 1);
2441 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2442 struct ext4_sb_info
*sbi
, char *buf
)
2444 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2446 if (!sb
->s_bdev
->bd_part
)
2447 return snprintf(buf
, PAGE_SIZE
, "0\n");
2448 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2449 (unsigned long long)(sbi
->s_kbytes_written
+
2450 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2451 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2454 static ssize_t
extent_cache_hits_show(struct ext4_attr
*a
,
2455 struct ext4_sb_info
*sbi
, char *buf
)
2457 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_hits
);
2460 static ssize_t
extent_cache_misses_show(struct ext4_attr
*a
,
2461 struct ext4_sb_info
*sbi
, char *buf
)
2463 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_misses
);
2466 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2467 struct ext4_sb_info
*sbi
,
2468 const char *buf
, size_t count
)
2472 if (parse_strtoul(buf
, 0x40000000, &t
))
2475 if (t
&& !is_power_of_2(t
))
2478 sbi
->s_inode_readahead_blks
= t
;
2482 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2483 struct ext4_sb_info
*sbi
, char *buf
)
2485 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2487 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2490 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2491 struct ext4_sb_info
*sbi
,
2492 const char *buf
, size_t count
)
2494 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2497 if (parse_strtoul(buf
, 0xffffffff, &t
))
2503 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2504 static struct ext4_attr ext4_attr_##_name = { \
2505 .attr = {.name = __stringify(_name), .mode = _mode }, \
2508 .offset = offsetof(struct ext4_sb_info, _elname), \
2510 #define EXT4_ATTR(name, mode, show, store) \
2511 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2513 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2514 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2515 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2516 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2517 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2518 #define ATTR_LIST(name) &ext4_attr_##name.attr
2520 EXT4_RO_ATTR(delayed_allocation_blocks
);
2521 EXT4_RO_ATTR(session_write_kbytes
);
2522 EXT4_RO_ATTR(lifetime_write_kbytes
);
2523 EXT4_RO_ATTR(extent_cache_hits
);
2524 EXT4_RO_ATTR(extent_cache_misses
);
2525 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2526 inode_readahead_blks_store
, s_inode_readahead_blks
);
2527 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2528 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2529 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2530 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2531 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2532 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2533 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2534 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2536 static struct attribute
*ext4_attrs
[] = {
2537 ATTR_LIST(delayed_allocation_blocks
),
2538 ATTR_LIST(session_write_kbytes
),
2539 ATTR_LIST(lifetime_write_kbytes
),
2540 ATTR_LIST(extent_cache_hits
),
2541 ATTR_LIST(extent_cache_misses
),
2542 ATTR_LIST(inode_readahead_blks
),
2543 ATTR_LIST(inode_goal
),
2544 ATTR_LIST(mb_stats
),
2545 ATTR_LIST(mb_max_to_scan
),
2546 ATTR_LIST(mb_min_to_scan
),
2547 ATTR_LIST(mb_order2_req
),
2548 ATTR_LIST(mb_stream_req
),
2549 ATTR_LIST(mb_group_prealloc
),
2550 ATTR_LIST(max_writeback_mb_bump
),
2554 /* Features this copy of ext4 supports */
2555 EXT4_INFO_ATTR(lazy_itable_init
);
2556 EXT4_INFO_ATTR(batched_discard
);
2558 static struct attribute
*ext4_feat_attrs
[] = {
2559 ATTR_LIST(lazy_itable_init
),
2560 ATTR_LIST(batched_discard
),
2564 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2565 struct attribute
*attr
, char *buf
)
2567 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2569 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2571 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2574 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2575 struct attribute
*attr
,
2576 const char *buf
, size_t len
)
2578 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2580 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2582 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2585 static void ext4_sb_release(struct kobject
*kobj
)
2587 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2589 complete(&sbi
->s_kobj_unregister
);
2592 static const struct sysfs_ops ext4_attr_ops
= {
2593 .show
= ext4_attr_show
,
2594 .store
= ext4_attr_store
,
2597 static struct kobj_type ext4_ktype
= {
2598 .default_attrs
= ext4_attrs
,
2599 .sysfs_ops
= &ext4_attr_ops
,
2600 .release
= ext4_sb_release
,
2603 static void ext4_feat_release(struct kobject
*kobj
)
2605 complete(&ext4_feat
->f_kobj_unregister
);
2608 static struct kobj_type ext4_feat_ktype
= {
2609 .default_attrs
= ext4_feat_attrs
,
2610 .sysfs_ops
= &ext4_attr_ops
,
2611 .release
= ext4_feat_release
,
2615 * Check whether this filesystem can be mounted based on
2616 * the features present and the RDONLY/RDWR mount requested.
2617 * Returns 1 if this filesystem can be mounted as requested,
2618 * 0 if it cannot be.
2620 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2622 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2623 ext4_msg(sb
, KERN_ERR
,
2624 "Couldn't mount because of "
2625 "unsupported optional features (%x)",
2626 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2627 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2634 /* Check that feature set is OK for a read-write mount */
2635 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2636 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2637 "unsupported optional features (%x)",
2638 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2639 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2643 * Large file size enabled file system can only be mounted
2644 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2646 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2647 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2648 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2649 "cannot be mounted RDWR without "
2658 * This function is called once a day if we have errors logged
2659 * on the file system
2661 static void print_daily_error_info(unsigned long arg
)
2663 struct super_block
*sb
= (struct super_block
*) arg
;
2664 struct ext4_sb_info
*sbi
;
2665 struct ext4_super_block
*es
;
2670 if (es
->s_error_count
)
2671 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2672 le32_to_cpu(es
->s_error_count
));
2673 if (es
->s_first_error_time
) {
2674 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2675 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2676 (int) sizeof(es
->s_first_error_func
),
2677 es
->s_first_error_func
,
2678 le32_to_cpu(es
->s_first_error_line
));
2679 if (es
->s_first_error_ino
)
2680 printk(": inode %u",
2681 le32_to_cpu(es
->s_first_error_ino
));
2682 if (es
->s_first_error_block
)
2683 printk(": block %llu", (unsigned long long)
2684 le64_to_cpu(es
->s_first_error_block
));
2687 if (es
->s_last_error_time
) {
2688 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2689 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2690 (int) sizeof(es
->s_last_error_func
),
2691 es
->s_last_error_func
,
2692 le32_to_cpu(es
->s_last_error_line
));
2693 if (es
->s_last_error_ino
)
2694 printk(": inode %u",
2695 le32_to_cpu(es
->s_last_error_ino
));
2696 if (es
->s_last_error_block
)
2697 printk(": block %llu", (unsigned long long)
2698 le64_to_cpu(es
->s_last_error_block
));
2701 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2704 /* Find next suitable group and run ext4_init_inode_table */
2705 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2707 struct ext4_group_desc
*gdp
= NULL
;
2708 ext4_group_t group
, ngroups
;
2709 struct super_block
*sb
;
2710 unsigned long timeout
= 0;
2714 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2716 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2717 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2723 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2727 if (group
== ngroups
)
2732 ret
= ext4_init_inode_table(sb
, group
,
2733 elr
->lr_timeout
? 0 : 1);
2734 if (elr
->lr_timeout
== 0) {
2735 timeout
= (jiffies
- timeout
) *
2736 elr
->lr_sbi
->s_li_wait_mult
;
2737 elr
->lr_timeout
= timeout
;
2739 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2740 elr
->lr_next_group
= group
+ 1;
2747 * Remove lr_request from the list_request and free the
2748 * request structure. Should be called with li_list_mtx held
2750 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2752 struct ext4_sb_info
*sbi
;
2759 list_del(&elr
->lr_request
);
2760 sbi
->s_li_request
= NULL
;
2764 static void ext4_unregister_li_request(struct super_block
*sb
)
2766 mutex_lock(&ext4_li_mtx
);
2767 if (!ext4_li_info
) {
2768 mutex_unlock(&ext4_li_mtx
);
2772 mutex_lock(&ext4_li_info
->li_list_mtx
);
2773 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2774 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2775 mutex_unlock(&ext4_li_mtx
);
2778 static struct task_struct
*ext4_lazyinit_task
;
2781 * This is the function where ext4lazyinit thread lives. It walks
2782 * through the request list searching for next scheduled filesystem.
2783 * When such a fs is found, run the lazy initialization request
2784 * (ext4_rn_li_request) and keep track of the time spend in this
2785 * function. Based on that time we compute next schedule time of
2786 * the request. When walking through the list is complete, compute
2787 * next waking time and put itself into sleep.
2789 static int ext4_lazyinit_thread(void *arg
)
2791 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2792 struct list_head
*pos
, *n
;
2793 struct ext4_li_request
*elr
;
2794 unsigned long next_wakeup
, cur
;
2796 BUG_ON(NULL
== eli
);
2800 next_wakeup
= MAX_JIFFY_OFFSET
;
2802 mutex_lock(&eli
->li_list_mtx
);
2803 if (list_empty(&eli
->li_request_list
)) {
2804 mutex_unlock(&eli
->li_list_mtx
);
2808 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2809 elr
= list_entry(pos
, struct ext4_li_request
,
2812 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2813 if (ext4_run_li_request(elr
) != 0) {
2814 /* error, remove the lazy_init job */
2815 ext4_remove_li_request(elr
);
2820 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2821 next_wakeup
= elr
->lr_next_sched
;
2823 mutex_unlock(&eli
->li_list_mtx
);
2825 if (freezing(current
))
2829 if ((time_after_eq(cur
, next_wakeup
)) ||
2830 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2835 schedule_timeout_interruptible(next_wakeup
- cur
);
2837 if (kthread_should_stop()) {
2838 ext4_clear_request_list();
2845 * It looks like the request list is empty, but we need
2846 * to check it under the li_list_mtx lock, to prevent any
2847 * additions into it, and of course we should lock ext4_li_mtx
2848 * to atomically free the list and ext4_li_info, because at
2849 * this point another ext4 filesystem could be registering
2852 mutex_lock(&ext4_li_mtx
);
2853 mutex_lock(&eli
->li_list_mtx
);
2854 if (!list_empty(&eli
->li_request_list
)) {
2855 mutex_unlock(&eli
->li_list_mtx
);
2856 mutex_unlock(&ext4_li_mtx
);
2859 mutex_unlock(&eli
->li_list_mtx
);
2860 kfree(ext4_li_info
);
2861 ext4_li_info
= NULL
;
2862 mutex_unlock(&ext4_li_mtx
);
2867 static void ext4_clear_request_list(void)
2869 struct list_head
*pos
, *n
;
2870 struct ext4_li_request
*elr
;
2872 mutex_lock(&ext4_li_info
->li_list_mtx
);
2873 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2874 elr
= list_entry(pos
, struct ext4_li_request
,
2876 ext4_remove_li_request(elr
);
2878 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2881 static int ext4_run_lazyinit_thread(void)
2883 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2884 ext4_li_info
, "ext4lazyinit");
2885 if (IS_ERR(ext4_lazyinit_task
)) {
2886 int err
= PTR_ERR(ext4_lazyinit_task
);
2887 ext4_clear_request_list();
2888 kfree(ext4_li_info
);
2889 ext4_li_info
= NULL
;
2890 printk(KERN_CRIT
"EXT4: error %d creating inode table "
2891 "initialization thread\n",
2895 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2900 * Check whether it make sense to run itable init. thread or not.
2901 * If there is at least one uninitialized inode table, return
2902 * corresponding group number, else the loop goes through all
2903 * groups and return total number of groups.
2905 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2907 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2908 struct ext4_group_desc
*gdp
= NULL
;
2910 for (group
= 0; group
< ngroups
; group
++) {
2911 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2915 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2922 static int ext4_li_info_new(void)
2924 struct ext4_lazy_init
*eli
= NULL
;
2926 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2930 INIT_LIST_HEAD(&eli
->li_request_list
);
2931 mutex_init(&eli
->li_list_mtx
);
2933 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2940 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2943 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2944 struct ext4_li_request
*elr
;
2947 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2953 elr
->lr_next_group
= start
;
2956 * Randomize first schedule time of the request to
2957 * spread the inode table initialization requests
2960 get_random_bytes(&rnd
, sizeof(rnd
));
2961 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2962 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
2967 static int ext4_register_li_request(struct super_block
*sb
,
2968 ext4_group_t first_not_zeroed
)
2970 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2971 struct ext4_li_request
*elr
;
2972 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
2975 if (sbi
->s_li_request
!= NULL
) {
2977 * Reset timeout so it can be computed again, because
2978 * s_li_wait_mult might have changed.
2980 sbi
->s_li_request
->lr_timeout
= 0;
2984 if (first_not_zeroed
== ngroups
||
2985 (sb
->s_flags
& MS_RDONLY
) ||
2986 !test_opt(sb
, INIT_INODE_TABLE
))
2989 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
2993 mutex_lock(&ext4_li_mtx
);
2995 if (NULL
== ext4_li_info
) {
2996 ret
= ext4_li_info_new();
3001 mutex_lock(&ext4_li_info
->li_list_mtx
);
3002 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3003 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3005 sbi
->s_li_request
= elr
;
3007 * set elr to NULL here since it has been inserted to
3008 * the request_list and the removal and free of it is
3009 * handled by ext4_clear_request_list from now on.
3013 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3014 ret
= ext4_run_lazyinit_thread();
3019 mutex_unlock(&ext4_li_mtx
);
3026 * We do not need to lock anything since this is called on
3029 static void ext4_destroy_lazyinit_thread(void)
3032 * If thread exited earlier
3033 * there's nothing to be done.
3035 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3038 kthread_stop(ext4_lazyinit_task
);
3041 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3042 __releases(kernel_lock
)
3043 __acquires(kernel_lock
)
3045 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3046 struct buffer_head
*bh
;
3047 struct ext4_super_block
*es
= NULL
;
3048 struct ext4_sb_info
*sbi
;
3050 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3051 ext4_fsblk_t logical_sb_block
;
3052 unsigned long offset
= 0;
3053 unsigned long journal_devnum
= 0;
3054 unsigned long def_mount_opts
;
3060 unsigned int db_count
;
3062 int needs_recovery
, has_huge_files
;
3065 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3066 ext4_group_t first_not_zeroed
;
3068 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3072 sbi
->s_blockgroup_lock
=
3073 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3074 if (!sbi
->s_blockgroup_lock
) {
3078 sb
->s_fs_info
= sbi
;
3079 sbi
->s_mount_opt
= 0;
3080 sbi
->s_resuid
= EXT4_DEF_RESUID
;
3081 sbi
->s_resgid
= EXT4_DEF_RESGID
;
3082 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3083 sbi
->s_sb_block
= sb_block
;
3084 if (sb
->s_bdev
->bd_part
)
3085 sbi
->s_sectors_written_start
=
3086 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3088 /* Cleanup superblock name */
3089 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3093 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3095 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3100 * The ext4 superblock will not be buffer aligned for other than 1kB
3101 * block sizes. We need to calculate the offset from buffer start.
3103 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3104 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3105 offset
= do_div(logical_sb_block
, blocksize
);
3107 logical_sb_block
= sb_block
;
3110 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3111 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3115 * Note: s_es must be initialized as soon as possible because
3116 * some ext4 macro-instructions depend on its value
3118 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3120 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3121 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3123 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3125 /* Set defaults before we parse the mount options */
3126 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3127 set_opt(sb
, INIT_INODE_TABLE
);
3128 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3130 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
3131 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
3135 if (def_mount_opts
& EXT4_DEFM_UID16
)
3136 set_opt(sb
, NO_UID32
);
3137 /* xattr user namespace & acls are now defaulted on */
3138 #ifdef CONFIG_EXT4_FS_XATTR
3139 set_opt(sb
, XATTR_USER
);
3141 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3142 set_opt(sb
, POSIX_ACL
);
3144 set_opt(sb
, MBLK_IO_SUBMIT
);
3145 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3146 set_opt(sb
, JOURNAL_DATA
);
3147 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3148 set_opt(sb
, ORDERED_DATA
);
3149 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3150 set_opt(sb
, WRITEBACK_DATA
);
3152 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3153 set_opt(sb
, ERRORS_PANIC
);
3154 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3155 set_opt(sb
, ERRORS_CONT
);
3157 set_opt(sb
, ERRORS_RO
);
3158 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3159 set_opt(sb
, BLOCK_VALIDITY
);
3160 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3161 set_opt(sb
, DISCARD
);
3163 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
3164 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
3165 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3166 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3167 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3169 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3170 set_opt(sb
, BARRIER
);
3173 * enable delayed allocation by default
3174 * Use -o nodelalloc to turn it off
3176 if (!IS_EXT3_SB(sb
) &&
3177 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3178 set_opt(sb
, DELALLOC
);
3181 * set default s_li_wait_mult for lazyinit, for the case there is
3182 * no mount option specified.
3184 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3186 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3187 &journal_devnum
, &journal_ioprio
, NULL
, 0)) {
3188 ext4_msg(sb
, KERN_WARNING
,
3189 "failed to parse options in superblock: %s",
3190 sbi
->s_es
->s_mount_opts
);
3192 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3193 &journal_ioprio
, NULL
, 0))
3196 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3197 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3199 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3200 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3201 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3202 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3203 ext4_msg(sb
, KERN_WARNING
,
3204 "feature flags set on rev 0 fs, "
3205 "running e2fsck is recommended");
3207 if (IS_EXT2_SB(sb
)) {
3208 if (ext2_feature_set_ok(sb
))
3209 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3210 "using the ext4 subsystem");
3212 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3213 "to feature incompatibilities");
3218 if (IS_EXT3_SB(sb
)) {
3219 if (ext3_feature_set_ok(sb
))
3220 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3221 "using the ext4 subsystem");
3223 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3224 "to feature incompatibilities");
3230 * Check feature flags regardless of the revision level, since we
3231 * previously didn't change the revision level when setting the flags,
3232 * so there is a chance incompat flags are set on a rev 0 filesystem.
3234 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3237 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3239 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3240 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3241 ext4_msg(sb
, KERN_ERR
,
3242 "Unsupported filesystem blocksize %d", blocksize
);
3246 if (sb
->s_blocksize
!= blocksize
) {
3247 /* Validate the filesystem blocksize */
3248 if (!sb_set_blocksize(sb
, blocksize
)) {
3249 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3255 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3256 offset
= do_div(logical_sb_block
, blocksize
);
3257 bh
= sb_bread(sb
, logical_sb_block
);
3259 ext4_msg(sb
, KERN_ERR
,
3260 "Can't read superblock on 2nd try");
3263 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
3265 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3266 ext4_msg(sb
, KERN_ERR
,
3267 "Magic mismatch, very weird!");
3272 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3273 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3274 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3276 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3278 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3279 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3280 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3282 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3283 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3284 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3285 (!is_power_of_2(sbi
->s_inode_size
)) ||
3286 (sbi
->s_inode_size
> blocksize
)) {
3287 ext4_msg(sb
, KERN_ERR
,
3288 "unsupported inode size: %d",
3292 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3293 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3296 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3297 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3298 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3299 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3300 !is_power_of_2(sbi
->s_desc_size
)) {
3301 ext4_msg(sb
, KERN_ERR
,
3302 "unsupported descriptor size %lu",
3307 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3309 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3310 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3311 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3314 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3315 if (sbi
->s_inodes_per_block
== 0)
3317 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3318 sbi
->s_inodes_per_block
;
3319 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3321 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3322 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3323 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3325 for (i
= 0; i
< 4; i
++)
3326 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3327 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3328 i
= le32_to_cpu(es
->s_flags
);
3329 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3330 sbi
->s_hash_unsigned
= 3;
3331 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3332 #ifdef __CHAR_UNSIGNED__
3333 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3334 sbi
->s_hash_unsigned
= 3;
3336 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3341 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3342 ext4_msg(sb
, KERN_ERR
,
3343 "#blocks per group too big: %lu",
3344 sbi
->s_blocks_per_group
);
3347 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3348 ext4_msg(sb
, KERN_ERR
,
3349 "#inodes per group too big: %lu",
3350 sbi
->s_inodes_per_group
);
3355 * Test whether we have more sectors than will fit in sector_t,
3356 * and whether the max offset is addressable by the page cache.
3358 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3359 ext4_blocks_count(es
));
3361 ext4_msg(sb
, KERN_ERR
, "filesystem"
3362 " too large to mount safely on this system");
3363 if (sizeof(sector_t
) < 8)
3364 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3369 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3372 /* check blocks count against device size */
3373 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3374 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3375 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3376 "exceeds size of device (%llu blocks)",
3377 ext4_blocks_count(es
), blocks_count
);
3382 * It makes no sense for the first data block to be beyond the end
3383 * of the filesystem.
3385 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3386 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
3387 "block %u is beyond end of filesystem (%llu)",
3388 le32_to_cpu(es
->s_first_data_block
),
3389 ext4_blocks_count(es
));
3392 blocks_count
= (ext4_blocks_count(es
) -
3393 le32_to_cpu(es
->s_first_data_block
) +
3394 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3395 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3396 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3397 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3398 "(block count %llu, first data block %u, "
3399 "blocks per group %lu)", sbi
->s_groups_count
,
3400 ext4_blocks_count(es
),
3401 le32_to_cpu(es
->s_first_data_block
),
3402 EXT4_BLOCKS_PER_GROUP(sb
));
3405 sbi
->s_groups_count
= blocks_count
;
3406 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3407 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3408 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3409 EXT4_DESC_PER_BLOCK(sb
);
3410 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
3412 if (sbi
->s_group_desc
== NULL
) {
3413 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3417 #ifdef CONFIG_PROC_FS
3419 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3422 bgl_lock_init(sbi
->s_blockgroup_lock
);
3424 for (i
= 0; i
< db_count
; i
++) {
3425 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3426 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3427 if (!sbi
->s_group_desc
[i
]) {
3428 ext4_msg(sb
, KERN_ERR
,
3429 "can't read group descriptor %d", i
);
3434 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3435 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3438 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3439 if (!ext4_fill_flex_info(sb
)) {
3440 ext4_msg(sb
, KERN_ERR
,
3441 "unable to initialize "
3442 "flex_bg meta info!");
3446 sbi
->s_gdb_count
= db_count
;
3447 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3448 spin_lock_init(&sbi
->s_next_gen_lock
);
3450 init_timer(&sbi
->s_err_report
);
3451 sbi
->s_err_report
.function
= print_daily_error_info
;
3452 sbi
->s_err_report
.data
= (unsigned long) sb
;
3454 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
3455 ext4_count_free_blocks(sb
));
3457 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3458 ext4_count_free_inodes(sb
));
3461 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3462 ext4_count_dirs(sb
));
3465 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
3468 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3472 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3473 sbi
->s_max_writeback_mb_bump
= 128;
3476 * set up enough so that it can read an inode
3478 if (!test_opt(sb
, NOLOAD
) &&
3479 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3480 sb
->s_op
= &ext4_sops
;
3482 sb
->s_op
= &ext4_nojournal_sops
;
3483 sb
->s_export_op
= &ext4_export_ops
;
3484 sb
->s_xattr
= ext4_xattr_handlers
;
3486 sb
->s_qcop
= &ext4_qctl_operations
;
3487 sb
->dq_op
= &ext4_quota_operations
;
3489 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3491 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3492 mutex_init(&sbi
->s_orphan_lock
);
3493 mutex_init(&sbi
->s_resize_lock
);
3497 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3498 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3499 EXT4_FEATURE_INCOMPAT_RECOVER
));
3501 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3502 !(sb
->s_flags
& MS_RDONLY
))
3503 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3507 * The first inode we look at is the journal inode. Don't try
3508 * root first: it may be modified in the journal!
3510 if (!test_opt(sb
, NOLOAD
) &&
3511 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3512 if (ext4_load_journal(sb
, es
, journal_devnum
))
3514 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3515 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3516 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3517 "suppressed and not mounted read-only");
3518 goto failed_mount_wq
;
3520 clear_opt(sb
, DATA_FLAGS
);
3521 sbi
->s_journal
= NULL
;
3526 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
3527 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3528 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3529 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3530 goto failed_mount_wq
;
3533 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3534 jbd2_journal_set_features(sbi
->s_journal
,
3535 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3536 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3537 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3538 jbd2_journal_set_features(sbi
->s_journal
,
3539 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
3540 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3541 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3543 jbd2_journal_clear_features(sbi
->s_journal
,
3544 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3545 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3548 /* We have now updated the journal if required, so we can
3549 * validate the data journaling mode. */
3550 switch (test_opt(sb
, DATA_FLAGS
)) {
3552 /* No mode set, assume a default based on the journal
3553 * capabilities: ORDERED_DATA if the journal can
3554 * cope, else JOURNAL_DATA
3556 if (jbd2_journal_check_available_features
3557 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3558 set_opt(sb
, ORDERED_DATA
);
3560 set_opt(sb
, JOURNAL_DATA
);
3563 case EXT4_MOUNT_ORDERED_DATA
:
3564 case EXT4_MOUNT_WRITEBACK_DATA
:
3565 if (!jbd2_journal_check_available_features
3566 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3567 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3568 "requested data journaling mode");
3569 goto failed_mount_wq
;
3574 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3577 * The journal may have updated the bg summary counts, so we
3578 * need to update the global counters.
3580 percpu_counter_set(&sbi
->s_freeblocks_counter
,
3581 ext4_count_free_blocks(sb
));
3582 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3583 ext4_count_free_inodes(sb
));
3584 percpu_counter_set(&sbi
->s_dirs_counter
,
3585 ext4_count_dirs(sb
));
3586 percpu_counter_set(&sbi
->s_dirtyblocks_counter
, 0);
3590 * The maximum number of concurrent works can be high and
3591 * concurrency isn't really necessary. Limit it to 1.
3593 EXT4_SB(sb
)->dio_unwritten_wq
=
3594 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3595 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3596 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3597 goto failed_mount_wq
;
3601 * The jbd2_journal_load will have done any necessary log recovery,
3602 * so we can safely mount the rest of the filesystem now.
3605 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3607 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3608 ret
= PTR_ERR(root
);
3612 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3613 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3616 sb
->s_root
= d_alloc_root(root
);
3618 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3623 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
3625 /* determine the minimum size of new large inodes, if present */
3626 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3627 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3628 EXT4_GOOD_OLD_INODE_SIZE
;
3629 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3630 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3631 if (sbi
->s_want_extra_isize
<
3632 le16_to_cpu(es
->s_want_extra_isize
))
3633 sbi
->s_want_extra_isize
=
3634 le16_to_cpu(es
->s_want_extra_isize
);
3635 if (sbi
->s_want_extra_isize
<
3636 le16_to_cpu(es
->s_min_extra_isize
))
3637 sbi
->s_want_extra_isize
=
3638 le16_to_cpu(es
->s_min_extra_isize
);
3641 /* Check if enough inode space is available */
3642 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3643 sbi
->s_inode_size
) {
3644 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3645 EXT4_GOOD_OLD_INODE_SIZE
;
3646 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3650 if (test_opt(sb
, DELALLOC
) &&
3651 (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)) {
3652 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
3653 "requested data journaling mode");
3654 clear_opt(sb
, DELALLOC
);
3656 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3657 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3658 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3659 "option - requested data journaling mode");
3660 clear_opt(sb
, DIOREAD_NOLOCK
);
3662 if (sb
->s_blocksize
< PAGE_SIZE
) {
3663 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3664 "option - block size is too small");
3665 clear_opt(sb
, DIOREAD_NOLOCK
);
3669 err
= ext4_setup_system_zone(sb
);
3671 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3677 err
= ext4_mb_init(sb
, needs_recovery
);
3679 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3684 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3688 sbi
->s_kobj
.kset
= ext4_kset
;
3689 init_completion(&sbi
->s_kobj_unregister
);
3690 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3693 ext4_mb_release(sb
);
3694 ext4_ext_release(sb
);
3698 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3699 ext4_orphan_cleanup(sb
, es
);
3700 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3701 if (needs_recovery
) {
3702 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3703 ext4_mark_recovery_complete(sb
, es
);
3705 if (EXT4_SB(sb
)->s_journal
) {
3706 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3707 descr
= " journalled data mode";
3708 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3709 descr
= " ordered data mode";
3711 descr
= " writeback data mode";
3713 descr
= "out journal";
3715 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3716 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3717 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3719 if (es
->s_error_count
)
3720 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3727 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3733 ext4_msg(sb
, KERN_ERR
, "mount failed");
3734 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3736 ext4_release_system_zone(sb
);
3737 if (sbi
->s_journal
) {
3738 jbd2_journal_destroy(sbi
->s_journal
);
3739 sbi
->s_journal
= NULL
;
3742 del_timer(&sbi
->s_err_report
);
3743 if (sbi
->s_flex_groups
) {
3744 if (is_vmalloc_addr(sbi
->s_flex_groups
))
3745 vfree(sbi
->s_flex_groups
);
3747 kfree(sbi
->s_flex_groups
);
3749 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
3750 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3751 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3752 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
3754 kthread_stop(sbi
->s_mmp_tsk
);
3756 for (i
= 0; i
< db_count
; i
++)
3757 brelse(sbi
->s_group_desc
[i
]);
3758 kfree(sbi
->s_group_desc
);
3761 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3764 for (i
= 0; i
< MAXQUOTAS
; i
++)
3765 kfree(sbi
->s_qf_names
[i
]);
3767 ext4_blkdev_remove(sbi
);
3770 sb
->s_fs_info
= NULL
;
3771 kfree(sbi
->s_blockgroup_lock
);
3779 * Setup any per-fs journal parameters now. We'll do this both on
3780 * initial mount, once the journal has been initialised but before we've
3781 * done any recovery; and again on any subsequent remount.
3783 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3785 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3787 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3788 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3789 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3791 write_lock(&journal
->j_state_lock
);
3792 if (test_opt(sb
, BARRIER
))
3793 journal
->j_flags
|= JBD2_BARRIER
;
3795 journal
->j_flags
&= ~JBD2_BARRIER
;
3796 if (test_opt(sb
, DATA_ERR_ABORT
))
3797 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3799 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3800 write_unlock(&journal
->j_state_lock
);
3803 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3804 unsigned int journal_inum
)
3806 struct inode
*journal_inode
;
3809 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3811 /* First, test for the existence of a valid inode on disk. Bad
3812 * things happen if we iget() an unused inode, as the subsequent
3813 * iput() will try to delete it. */
3815 journal_inode
= ext4_iget(sb
, journal_inum
);
3816 if (IS_ERR(journal_inode
)) {
3817 ext4_msg(sb
, KERN_ERR
, "no journal found");
3820 if (!journal_inode
->i_nlink
) {
3821 make_bad_inode(journal_inode
);
3822 iput(journal_inode
);
3823 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3827 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3828 journal_inode
, journal_inode
->i_size
);
3829 if (!S_ISREG(journal_inode
->i_mode
)) {
3830 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3831 iput(journal_inode
);
3835 journal
= jbd2_journal_init_inode(journal_inode
);
3837 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3838 iput(journal_inode
);
3841 journal
->j_private
= sb
;
3842 ext4_init_journal_params(sb
, journal
);
3846 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3849 struct buffer_head
*bh
;
3853 int hblock
, blocksize
;
3854 ext4_fsblk_t sb_block
;
3855 unsigned long offset
;
3856 struct ext4_super_block
*es
;
3857 struct block_device
*bdev
;
3859 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3861 bdev
= ext4_blkdev_get(j_dev
, sb
);
3865 blocksize
= sb
->s_blocksize
;
3866 hblock
= bdev_logical_block_size(bdev
);
3867 if (blocksize
< hblock
) {
3868 ext4_msg(sb
, KERN_ERR
,
3869 "blocksize too small for journal device");
3873 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3874 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3875 set_blocksize(bdev
, blocksize
);
3876 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3877 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3878 "external journal");
3882 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3883 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3884 !(le32_to_cpu(es
->s_feature_incompat
) &
3885 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3886 ext4_msg(sb
, KERN_ERR
, "external journal has "
3892 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3893 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3898 len
= ext4_blocks_count(es
);
3899 start
= sb_block
+ 1;
3900 brelse(bh
); /* we're done with the superblock */
3902 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3903 start
, len
, blocksize
);
3905 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3908 journal
->j_private
= sb
;
3909 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3910 wait_on_buffer(journal
->j_sb_buffer
);
3911 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3912 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3915 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3916 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3917 "user (unsupported) - %d",
3918 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3921 EXT4_SB(sb
)->journal_bdev
= bdev
;
3922 ext4_init_journal_params(sb
, journal
);
3926 jbd2_journal_destroy(journal
);
3928 ext4_blkdev_put(bdev
);
3932 static int ext4_load_journal(struct super_block
*sb
,
3933 struct ext4_super_block
*es
,
3934 unsigned long journal_devnum
)
3937 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3940 int really_read_only
;
3942 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3944 if (journal_devnum
&&
3945 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3946 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3947 "numbers have changed");
3948 journal_dev
= new_decode_dev(journal_devnum
);
3950 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3952 really_read_only
= bdev_read_only(sb
->s_bdev
);
3955 * Are we loading a blank journal or performing recovery after a
3956 * crash? For recovery, we need to check in advance whether we
3957 * can get read-write access to the device.
3959 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3960 if (sb
->s_flags
& MS_RDONLY
) {
3961 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3962 "required on readonly filesystem");
3963 if (really_read_only
) {
3964 ext4_msg(sb
, KERN_ERR
, "write access "
3965 "unavailable, cannot proceed");
3968 ext4_msg(sb
, KERN_INFO
, "write access will "
3969 "be enabled during recovery");
3973 if (journal_inum
&& journal_dev
) {
3974 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3975 "and inode journals!");
3980 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3983 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3987 if (!(journal
->j_flags
& JBD2_BARRIER
))
3988 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3990 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
3991 err
= jbd2_journal_update_format(journal
);
3993 ext4_msg(sb
, KERN_ERR
, "error updating journal");
3994 jbd2_journal_destroy(journal
);
3999 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4000 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4002 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4004 memcpy(save
, ((char *) es
) +
4005 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4006 err
= jbd2_journal_load(journal
);
4008 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4009 save
, EXT4_S_ERR_LEN
);
4014 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4015 jbd2_journal_destroy(journal
);
4019 EXT4_SB(sb
)->s_journal
= journal
;
4020 ext4_clear_journal_err(sb
, es
);
4022 if (!really_read_only
&& journal_devnum
&&
4023 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4024 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4026 /* Make sure we flush the recovery flag to disk. */
4027 ext4_commit_super(sb
, 1);
4033 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4035 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4036 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4041 if (buffer_write_io_error(sbh
)) {
4043 * Oh, dear. A previous attempt to write the
4044 * superblock failed. This could happen because the
4045 * USB device was yanked out. Or it could happen to
4046 * be a transient write error and maybe the block will
4047 * be remapped. Nothing we can do but to retry the
4048 * write and hope for the best.
4050 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4051 "superblock detected");
4052 clear_buffer_write_io_error(sbh
);
4053 set_buffer_uptodate(sbh
);
4056 * If the file system is mounted read-only, don't update the
4057 * superblock write time. This avoids updating the superblock
4058 * write time when we are mounting the root file system
4059 * read/only but we need to replay the journal; at that point,
4060 * for people who are east of GMT and who make their clock
4061 * tick in localtime for Windows bug-for-bug compatibility,
4062 * the clock is set in the future, and this will cause e2fsck
4063 * to complain and force a full file system check.
4065 if (!(sb
->s_flags
& MS_RDONLY
))
4066 es
->s_wtime
= cpu_to_le32(get_seconds());
4067 if (sb
->s_bdev
->bd_part
)
4068 es
->s_kbytes_written
=
4069 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4070 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4071 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4073 es
->s_kbytes_written
=
4074 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4075 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
4076 &EXT4_SB(sb
)->s_freeblocks_counter
));
4077 es
->s_free_inodes_count
=
4078 cpu_to_le32(percpu_counter_sum_positive(
4079 &EXT4_SB(sb
)->s_freeinodes_counter
));
4081 BUFFER_TRACE(sbh
, "marking dirty");
4082 mark_buffer_dirty(sbh
);
4084 error
= sync_dirty_buffer(sbh
);
4088 error
= buffer_write_io_error(sbh
);
4090 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4092 clear_buffer_write_io_error(sbh
);
4093 set_buffer_uptodate(sbh
);
4100 * Have we just finished recovery? If so, and if we are mounting (or
4101 * remounting) the filesystem readonly, then we will end up with a
4102 * consistent fs on disk. Record that fact.
4104 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4105 struct ext4_super_block
*es
)
4107 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4109 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4110 BUG_ON(journal
!= NULL
);
4113 jbd2_journal_lock_updates(journal
);
4114 if (jbd2_journal_flush(journal
) < 0)
4117 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4118 sb
->s_flags
& MS_RDONLY
) {
4119 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4120 ext4_commit_super(sb
, 1);
4124 jbd2_journal_unlock_updates(journal
);
4128 * If we are mounting (or read-write remounting) a filesystem whose journal
4129 * has recorded an error from a previous lifetime, move that error to the
4130 * main filesystem now.
4132 static void ext4_clear_journal_err(struct super_block
*sb
,
4133 struct ext4_super_block
*es
)
4139 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4141 journal
= EXT4_SB(sb
)->s_journal
;
4144 * Now check for any error status which may have been recorded in the
4145 * journal by a prior ext4_error() or ext4_abort()
4148 j_errno
= jbd2_journal_errno(journal
);
4152 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4153 ext4_warning(sb
, "Filesystem error recorded "
4154 "from previous mount: %s", errstr
);
4155 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4157 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4158 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4159 ext4_commit_super(sb
, 1);
4161 jbd2_journal_clear_err(journal
);
4166 * Force the running and committing transactions to commit,
4167 * and wait on the commit.
4169 int ext4_force_commit(struct super_block
*sb
)
4174 if (sb
->s_flags
& MS_RDONLY
)
4177 journal
= EXT4_SB(sb
)->s_journal
;
4179 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
4180 ret
= ext4_journal_force_commit(journal
);
4186 static void ext4_write_super(struct super_block
*sb
)
4189 ext4_commit_super(sb
, 1);
4193 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4197 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4199 trace_ext4_sync_fs(sb
, wait
);
4200 flush_workqueue(sbi
->dio_unwritten_wq
);
4201 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4203 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4209 * LVM calls this function before a (read-only) snapshot is created. This
4210 * gives us a chance to flush the journal completely and mark the fs clean.
4212 * Note that only this function cannot bring a filesystem to be in a clean
4213 * state independently, because ext4 prevents a new handle from being started
4214 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4217 static int ext4_freeze(struct super_block
*sb
)
4222 if (sb
->s_flags
& MS_RDONLY
)
4225 journal
= EXT4_SB(sb
)->s_journal
;
4227 /* Now we set up the journal barrier. */
4228 jbd2_journal_lock_updates(journal
);
4231 * Don't clear the needs_recovery flag if we failed to flush
4234 error
= jbd2_journal_flush(journal
);
4238 /* Journal blocked and flushed, clear needs_recovery flag. */
4239 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4240 error
= ext4_commit_super(sb
, 1);
4242 /* we rely on s_frozen to stop further updates */
4243 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4248 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4249 * flag here, even though the filesystem is not technically dirty yet.
4251 static int ext4_unfreeze(struct super_block
*sb
)
4253 if (sb
->s_flags
& MS_RDONLY
)
4257 /* Reset the needs_recovery flag before the fs is unlocked. */
4258 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4259 ext4_commit_super(sb
, 1);
4265 * Structure to save mount options for ext4_remount's benefit
4267 struct ext4_mount_options
{
4268 unsigned long s_mount_opt
;
4269 unsigned long s_mount_opt2
;
4272 unsigned long s_commit_interval
;
4273 u32 s_min_batch_time
, s_max_batch_time
;
4276 char *s_qf_names
[MAXQUOTAS
];
4280 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4282 struct ext4_super_block
*es
;
4283 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4284 ext4_fsblk_t n_blocks_count
= 0;
4285 unsigned long old_sb_flags
;
4286 struct ext4_mount_options old_opts
;
4287 int enable_quota
= 0;
4289 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4294 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4296 /* Store the original options */
4298 old_sb_flags
= sb
->s_flags
;
4299 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4300 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4301 old_opts
.s_resuid
= sbi
->s_resuid
;
4302 old_opts
.s_resgid
= sbi
->s_resgid
;
4303 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4304 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4305 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4307 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4308 for (i
= 0; i
< MAXQUOTAS
; i
++)
4309 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
4311 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4312 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4315 * Allow the "check" option to be passed as a remount option.
4317 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
4318 &n_blocks_count
, 1)) {
4323 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4324 ext4_abort(sb
, "Abort forced by user");
4326 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4327 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4331 if (sbi
->s_journal
) {
4332 ext4_init_journal_params(sb
, sbi
->s_journal
);
4333 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4336 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
4337 n_blocks_count
> ext4_blocks_count(es
)) {
4338 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4343 if (*flags
& MS_RDONLY
) {
4344 err
= dquot_suspend(sb
, -1);
4349 * First of all, the unconditional stuff we have to do
4350 * to disable replay of the journal when we next remount
4352 sb
->s_flags
|= MS_RDONLY
;
4355 * OK, test if we are remounting a valid rw partition
4356 * readonly, and if so set the rdonly flag and then
4357 * mark the partition as valid again.
4359 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4360 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4361 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4364 ext4_mark_recovery_complete(sb
, es
);
4366 /* Make sure we can mount this feature set readwrite */
4367 if (!ext4_feature_set_ok(sb
, 0)) {
4372 * Make sure the group descriptor checksums
4373 * are sane. If they aren't, refuse to remount r/w.
4375 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4376 struct ext4_group_desc
*gdp
=
4377 ext4_get_group_desc(sb
, g
, NULL
);
4379 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
4380 ext4_msg(sb
, KERN_ERR
,
4381 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4382 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4383 le16_to_cpu(gdp
->bg_checksum
));
4390 * If we have an unprocessed orphan list hanging
4391 * around from a previously readonly bdev mount,
4392 * require a full umount/remount for now.
4394 if (es
->s_last_orphan
) {
4395 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4396 "remount RDWR because of unprocessed "
4397 "orphan inode list. Please "
4398 "umount/remount instead");
4404 * Mounting a RDONLY partition read-write, so reread
4405 * and store the current valid flag. (It may have
4406 * been changed by e2fsck since we originally mounted
4410 ext4_clear_journal_err(sb
, es
);
4411 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4412 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
4414 if (!ext4_setup_super(sb
, es
, 0))
4415 sb
->s_flags
&= ~MS_RDONLY
;
4416 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4417 EXT4_FEATURE_INCOMPAT_MMP
))
4418 if (ext4_multi_mount_protect(sb
,
4419 le64_to_cpu(es
->s_mmp_block
))) {
4428 * Reinitialize lazy itable initialization thread based on
4431 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4432 ext4_unregister_li_request(sb
);
4434 ext4_group_t first_not_zeroed
;
4435 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4436 ext4_register_li_request(sb
, first_not_zeroed
);
4439 ext4_setup_system_zone(sb
);
4440 if (sbi
->s_journal
== NULL
)
4441 ext4_commit_super(sb
, 1);
4444 /* Release old quota file names */
4445 for (i
= 0; i
< MAXQUOTAS
; i
++)
4446 if (old_opts
.s_qf_names
[i
] &&
4447 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4448 kfree(old_opts
.s_qf_names
[i
]);
4452 dquot_resume(sb
, -1);
4454 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4459 sb
->s_flags
= old_sb_flags
;
4460 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4461 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4462 sbi
->s_resuid
= old_opts
.s_resuid
;
4463 sbi
->s_resgid
= old_opts
.s_resgid
;
4464 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4465 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4466 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4468 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4469 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4470 if (sbi
->s_qf_names
[i
] &&
4471 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4472 kfree(sbi
->s_qf_names
[i
]);
4473 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4481 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4483 struct super_block
*sb
= dentry
->d_sb
;
4484 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4485 struct ext4_super_block
*es
= sbi
->s_es
;
4489 if (test_opt(sb
, MINIX_DF
)) {
4490 sbi
->s_overhead_last
= 0;
4491 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
4492 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
4493 ext4_fsblk_t overhead
= 0;
4496 * Compute the overhead (FS structures). This is constant
4497 * for a given filesystem unless the number of block groups
4498 * changes so we cache the previous value until it does.
4502 * All of the blocks before first_data_block are
4505 overhead
= le32_to_cpu(es
->s_first_data_block
);
4508 * Add the overhead attributed to the superblock and
4509 * block group descriptors. If the sparse superblocks
4510 * feature is turned on, then not all groups have this.
4512 for (i
= 0; i
< ngroups
; i
++) {
4513 overhead
+= ext4_bg_has_super(sb
, i
) +
4514 ext4_bg_num_gdb(sb
, i
);
4519 * Every block group has an inode bitmap, a block
4520 * bitmap, and an inode table.
4522 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
4523 sbi
->s_overhead_last
= overhead
;
4525 sbi
->s_blocks_last
= ext4_blocks_count(es
);
4528 buf
->f_type
= EXT4_SUPER_MAGIC
;
4529 buf
->f_bsize
= sb
->s_blocksize
;
4530 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
4531 bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
4532 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
4533 /* prevent underflow in case that few free space is available */
4534 buf
->f_bfree
= max_t(s64
, bfree
, 0);
4535 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
4536 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
4538 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4539 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4540 buf
->f_namelen
= EXT4_NAME_LEN
;
4541 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4542 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4543 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4544 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4549 /* Helper function for writing quotas on sync - we need to start transaction
4550 * before quota file is locked for write. Otherwise the are possible deadlocks:
4551 * Process 1 Process 2
4552 * ext4_create() quota_sync()
4553 * jbd2_journal_start() write_dquot()
4554 * dquot_initialize() down(dqio_mutex)
4555 * down(dqio_mutex) jbd2_journal_start()
4561 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4563 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
4566 static int ext4_write_dquot(struct dquot
*dquot
)
4570 struct inode
*inode
;
4572 inode
= dquot_to_inode(dquot
);
4573 handle
= ext4_journal_start(inode
,
4574 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4576 return PTR_ERR(handle
);
4577 ret
= dquot_commit(dquot
);
4578 err
= ext4_journal_stop(handle
);
4584 static int ext4_acquire_dquot(struct dquot
*dquot
)
4589 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4590 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4592 return PTR_ERR(handle
);
4593 ret
= dquot_acquire(dquot
);
4594 err
= ext4_journal_stop(handle
);
4600 static int ext4_release_dquot(struct dquot
*dquot
)
4605 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4606 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4607 if (IS_ERR(handle
)) {
4608 /* Release dquot anyway to avoid endless cycle in dqput() */
4609 dquot_release(dquot
);
4610 return PTR_ERR(handle
);
4612 ret
= dquot_release(dquot
);
4613 err
= ext4_journal_stop(handle
);
4619 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4621 /* Are we journaling quotas? */
4622 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4623 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4624 dquot_mark_dquot_dirty(dquot
);
4625 return ext4_write_dquot(dquot
);
4627 return dquot_mark_dquot_dirty(dquot
);
4631 static int ext4_write_info(struct super_block
*sb
, int type
)
4636 /* Data block + inode block */
4637 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
4639 return PTR_ERR(handle
);
4640 ret
= dquot_commit_info(sb
, type
);
4641 err
= ext4_journal_stop(handle
);
4648 * Turn on quotas during mount time - we need to find
4649 * the quota file and such...
4651 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4653 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4654 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4658 * Standard function to be called on quota_on
4660 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4665 if (!test_opt(sb
, QUOTA
))
4668 /* Quotafile not on the same filesystem? */
4669 if (path
->mnt
->mnt_sb
!= sb
)
4671 /* Journaling quota? */
4672 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4673 /* Quotafile not in fs root? */
4674 if (path
->dentry
->d_parent
!= sb
->s_root
)
4675 ext4_msg(sb
, KERN_WARNING
,
4676 "Quota file not on filesystem root. "
4677 "Journaled quota will not work");
4681 * When we journal data on quota file, we have to flush journal to see
4682 * all updates to the file when we bypass pagecache...
4684 if (EXT4_SB(sb
)->s_journal
&&
4685 ext4_should_journal_data(path
->dentry
->d_inode
)) {
4687 * We don't need to lock updates but journal_flush() could
4688 * otherwise be livelocked...
4690 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4691 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4692 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4697 return dquot_quota_on(sb
, type
, format_id
, path
);
4700 static int ext4_quota_off(struct super_block
*sb
, int type
)
4702 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4705 /* Force all delayed allocation blocks to be allocated.
4706 * Caller already holds s_umount sem */
4707 if (test_opt(sb
, DELALLOC
))
4708 sync_filesystem(sb
);
4713 /* Update modification times of quota files when userspace can
4714 * start looking at them */
4715 handle
= ext4_journal_start(inode
, 1);
4718 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4719 ext4_mark_inode_dirty(handle
, inode
);
4720 ext4_journal_stop(handle
);
4723 return dquot_quota_off(sb
, type
);
4726 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4727 * acquiring the locks... As quota files are never truncated and quota code
4728 * itself serializes the operations (and no one else should touch the files)
4729 * we don't have to be afraid of races */
4730 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
4731 size_t len
, loff_t off
)
4733 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4734 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4736 int offset
= off
& (sb
->s_blocksize
- 1);
4739 struct buffer_head
*bh
;
4740 loff_t i_size
= i_size_read(inode
);
4744 if (off
+len
> i_size
)
4747 while (toread
> 0) {
4748 tocopy
= sb
->s_blocksize
- offset
< toread
?
4749 sb
->s_blocksize
- offset
: toread
;
4750 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
4753 if (!bh
) /* A hole? */
4754 memset(data
, 0, tocopy
);
4756 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4766 /* Write to quotafile (we know the transaction is already started and has
4767 * enough credits) */
4768 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4769 const char *data
, size_t len
, loff_t off
)
4771 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4772 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4774 int offset
= off
& (sb
->s_blocksize
- 1);
4775 struct buffer_head
*bh
;
4776 handle_t
*handle
= journal_current_handle();
4778 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4779 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4780 " cancelled because transaction is not started",
4781 (unsigned long long)off
, (unsigned long long)len
);
4785 * Since we account only one data block in transaction credits,
4786 * then it is impossible to cross a block boundary.
4788 if (sb
->s_blocksize
- offset
< len
) {
4789 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4790 " cancelled because not block aligned",
4791 (unsigned long long)off
, (unsigned long long)len
);
4795 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4796 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4799 err
= ext4_journal_get_write_access(handle
, bh
);
4805 memcpy(bh
->b_data
+offset
, data
, len
);
4806 flush_dcache_page(bh
->b_page
);
4808 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4812 mutex_unlock(&inode
->i_mutex
);
4815 if (inode
->i_size
< off
+ len
) {
4816 i_size_write(inode
, off
+ len
);
4817 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4818 ext4_mark_inode_dirty(handle
, inode
);
4820 mutex_unlock(&inode
->i_mutex
);
4826 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
4827 const char *dev_name
, void *data
)
4829 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
4832 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4833 static inline void register_as_ext2(void)
4835 int err
= register_filesystem(&ext2_fs_type
);
4838 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4841 static inline void unregister_as_ext2(void)
4843 unregister_filesystem(&ext2_fs_type
);
4846 static inline int ext2_feature_set_ok(struct super_block
*sb
)
4848 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
4850 if (sb
->s_flags
& MS_RDONLY
)
4852 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
4856 MODULE_ALIAS("ext2");
4858 static inline void register_as_ext2(void) { }
4859 static inline void unregister_as_ext2(void) { }
4860 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
4863 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4864 static inline void register_as_ext3(void)
4866 int err
= register_filesystem(&ext3_fs_type
);
4869 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4872 static inline void unregister_as_ext3(void)
4874 unregister_filesystem(&ext3_fs_type
);
4877 static inline int ext3_feature_set_ok(struct super_block
*sb
)
4879 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
4881 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
4883 if (sb
->s_flags
& MS_RDONLY
)
4885 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
4889 MODULE_ALIAS("ext3");
4891 static inline void register_as_ext3(void) { }
4892 static inline void unregister_as_ext3(void) { }
4893 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
4896 static struct file_system_type ext4_fs_type
= {
4897 .owner
= THIS_MODULE
,
4899 .mount
= ext4_mount
,
4900 .kill_sb
= kill_block_super
,
4901 .fs_flags
= FS_REQUIRES_DEV
,
4904 static int __init
ext4_init_feat_adverts(void)
4906 struct ext4_features
*ef
;
4909 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
4913 ef
->f_kobj
.kset
= ext4_kset
;
4914 init_completion(&ef
->f_kobj_unregister
);
4915 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
4928 static void ext4_exit_feat_adverts(void)
4930 kobject_put(&ext4_feat
->f_kobj
);
4931 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
4935 /* Shared across all ext4 file systems */
4936 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
4937 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
4939 static int __init
ext4_init_fs(void)
4943 ext4_check_flag_values();
4945 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
4946 mutex_init(&ext4__aio_mutex
[i
]);
4947 init_waitqueue_head(&ext4__ioend_wq
[i
]);
4950 err
= ext4_init_pageio();
4953 err
= ext4_init_system_zone();
4956 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4959 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4960 if (!ext4_proc_root
)
4963 err
= ext4_init_feat_adverts();
4967 err
= ext4_init_mballoc();
4971 err
= ext4_init_xattr();
4974 err
= init_inodecache();
4979 err
= register_filesystem(&ext4_fs_type
);
4983 ext4_li_info
= NULL
;
4984 mutex_init(&ext4_li_mtx
);
4987 unregister_as_ext2();
4988 unregister_as_ext3();
4989 destroy_inodecache();
4993 ext4_exit_mballoc();
4995 ext4_exit_feat_adverts();
4997 remove_proc_entry("fs/ext4", NULL
);
4999 kset_unregister(ext4_kset
);
5001 ext4_exit_system_zone();
5007 static void __exit
ext4_exit_fs(void)
5009 ext4_destroy_lazyinit_thread();
5010 unregister_as_ext2();
5011 unregister_as_ext3();
5012 unregister_filesystem(&ext4_fs_type
);
5013 destroy_inodecache();
5015 ext4_exit_mballoc();
5016 ext4_exit_feat_adverts();
5017 remove_proc_entry("fs/ext4", NULL
);
5018 kset_unregister(ext4_kset
);
5019 ext4_exit_system_zone();
5023 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5024 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5025 MODULE_LICENSE("GPL");
5026 module_init(ext4_init_fs
)
5027 module_exit(ext4_exit_fs
)