2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_jbd2.h"
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/ext4.h>
55 static struct proc_dir_entry
*ext4_proc_root
;
56 static struct kset
*ext4_kset
;
57 struct ext4_lazy_init
*ext4_li_info
;
58 struct mutex ext4_li_mtx
;
59 struct ext4_features
*ext4_feat
;
61 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
62 unsigned long journal_devnum
);
63 static int ext4_commit_super(struct super_block
*sb
, int sync
);
64 static void ext4_mark_recovery_complete(struct super_block
*sb
,
65 struct ext4_super_block
*es
);
66 static void ext4_clear_journal_err(struct super_block
*sb
,
67 struct ext4_super_block
*es
);
68 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
69 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
71 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
72 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
73 static int ext4_unfreeze(struct super_block
*sb
);
74 static void ext4_write_super(struct super_block
*sb
);
75 static int ext4_freeze(struct super_block
*sb
);
76 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
77 const char *dev_name
, void *data
);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block
*sb
);
80 static void ext4_clear_request_list(void);
82 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
83 static struct file_system_type ext3_fs_type
= {
87 .kill_sb
= kill_block_super
,
88 .fs_flags
= FS_REQUIRES_DEV
,
90 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
92 #define IS_EXT3_SB(sb) (0)
95 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
96 struct ext4_group_desc
*bg
)
98 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
99 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
100 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
103 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
104 struct ext4_group_desc
*bg
)
106 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
107 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
108 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
111 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
112 struct ext4_group_desc
*bg
)
114 return le32_to_cpu(bg
->bg_inode_table_lo
) |
115 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
116 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
119 __u32
ext4_free_blks_count(struct super_block
*sb
,
120 struct ext4_group_desc
*bg
)
122 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
123 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
124 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
127 __u32
ext4_free_inodes_count(struct super_block
*sb
,
128 struct ext4_group_desc
*bg
)
130 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
131 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
132 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
135 __u32
ext4_used_dirs_count(struct super_block
*sb
,
136 struct ext4_group_desc
*bg
)
138 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
139 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
140 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
143 __u32
ext4_itable_unused_count(struct super_block
*sb
,
144 struct ext4_group_desc
*bg
)
146 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
147 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
148 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
151 void ext4_block_bitmap_set(struct super_block
*sb
,
152 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
154 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
155 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
156 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
159 void ext4_inode_bitmap_set(struct super_block
*sb
,
160 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
162 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
163 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
164 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
167 void ext4_inode_table_set(struct super_block
*sb
,
168 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
170 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
171 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
172 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
175 void ext4_free_blks_set(struct super_block
*sb
,
176 struct ext4_group_desc
*bg
, __u32 count
)
178 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
179 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
180 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
183 void ext4_free_inodes_set(struct super_block
*sb
,
184 struct ext4_group_desc
*bg
, __u32 count
)
186 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
187 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
188 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
191 void ext4_used_dirs_set(struct super_block
*sb
,
192 struct ext4_group_desc
*bg
, __u32 count
)
194 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
195 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
196 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
199 void ext4_itable_unused_set(struct super_block
*sb
,
200 struct ext4_group_desc
*bg
, __u32 count
)
202 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
203 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
204 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
208 /* Just increment the non-pointer handle value */
209 static handle_t
*ext4_get_nojournal(void)
211 handle_t
*handle
= current
->journal_info
;
212 unsigned long ref_cnt
= (unsigned long)handle
;
214 BUG_ON(ref_cnt
>= EXT4_NOJOURNAL_MAX_REF_COUNT
);
217 handle
= (handle_t
*)ref_cnt
;
219 current
->journal_info
= handle
;
224 /* Decrement the non-pointer handle value */
225 static void ext4_put_nojournal(handle_t
*handle
)
227 unsigned long ref_cnt
= (unsigned long)handle
;
229 BUG_ON(ref_cnt
== 0);
232 handle
= (handle_t
*)ref_cnt
;
234 current
->journal_info
= handle
;
238 * Wrappers for jbd2_journal_start/end.
240 * The only special thing we need to do here is to make sure that all
241 * journal_end calls result in the superblock being marked dirty, so
242 * that sync() will call the filesystem's write_super callback if
245 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
249 if (sb
->s_flags
& MS_RDONLY
)
250 return ERR_PTR(-EROFS
);
252 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
253 /* Special case here: if the journal has aborted behind our
254 * backs (eg. EIO in the commit thread), then we still need to
255 * take the FS itself readonly cleanly. */
256 journal
= EXT4_SB(sb
)->s_journal
;
258 if (is_journal_aborted(journal
)) {
259 ext4_abort(sb
, "Detected aborted journal");
260 return ERR_PTR(-EROFS
);
262 return jbd2_journal_start(journal
, nblocks
);
264 return ext4_get_nojournal();
268 * The only special thing we need to do here is to make sure that all
269 * jbd2_journal_stop calls result in the superblock being marked dirty, so
270 * that sync() will call the filesystem's write_super callback if
273 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
)
275 struct super_block
*sb
;
279 if (!ext4_handle_valid(handle
)) {
280 ext4_put_nojournal(handle
);
283 sb
= handle
->h_transaction
->t_journal
->j_private
;
285 rc
= jbd2_journal_stop(handle
);
290 __ext4_std_error(sb
, where
, line
, err
);
294 void ext4_journal_abort_handle(const char *caller
, unsigned int line
,
295 const char *err_fn
, struct buffer_head
*bh
,
296 handle_t
*handle
, int err
)
299 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
301 BUG_ON(!ext4_handle_valid(handle
));
304 BUFFER_TRACE(bh
, "abort");
309 if (is_handle_aborted(handle
))
312 printk(KERN_ERR
"%s:%d: aborting transaction: %s in %s\n",
313 caller
, line
, errstr
, err_fn
);
315 jbd2_journal_abort_handle(handle
);
318 static void __save_error_info(struct super_block
*sb
, const char *func
,
321 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
323 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
324 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
325 es
->s_last_error_time
= cpu_to_le32(get_seconds());
326 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
327 es
->s_last_error_line
= cpu_to_le32(line
);
328 if (!es
->s_first_error_time
) {
329 es
->s_first_error_time
= es
->s_last_error_time
;
330 strncpy(es
->s_first_error_func
, func
,
331 sizeof(es
->s_first_error_func
));
332 es
->s_first_error_line
= cpu_to_le32(line
);
333 es
->s_first_error_ino
= es
->s_last_error_ino
;
334 es
->s_first_error_block
= es
->s_last_error_block
;
337 * Start the daily error reporting function if it hasn't been
340 if (!es
->s_error_count
)
341 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
342 es
->s_error_count
= cpu_to_le32(le32_to_cpu(es
->s_error_count
) + 1);
345 static void save_error_info(struct super_block
*sb
, const char *func
,
348 __save_error_info(sb
, func
, line
);
349 ext4_commit_super(sb
, 1);
353 /* Deal with the reporting of failure conditions on a filesystem such as
354 * inconsistencies detected or read IO failures.
356 * On ext2, we can store the error state of the filesystem in the
357 * superblock. That is not possible on ext4, because we may have other
358 * write ordering constraints on the superblock which prevent us from
359 * writing it out straight away; and given that the journal is about to
360 * be aborted, we can't rely on the current, or future, transactions to
361 * write out the superblock safely.
363 * We'll just use the jbd2_journal_abort() error code to record an error in
364 * the journal instead. On recovery, the journal will complain about
365 * that error until we've noted it down and cleared it.
368 static void ext4_handle_error(struct super_block
*sb
)
370 if (sb
->s_flags
& MS_RDONLY
)
373 if (!test_opt(sb
, ERRORS_CONT
)) {
374 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
376 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
378 jbd2_journal_abort(journal
, -EIO
);
380 if (test_opt(sb
, ERRORS_RO
)) {
381 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
382 sb
->s_flags
|= MS_RDONLY
;
384 if (test_opt(sb
, ERRORS_PANIC
))
385 panic("EXT4-fs (device %s): panic forced after error\n",
389 void __ext4_error(struct super_block
*sb
, const char *function
,
390 unsigned int line
, const char *fmt
, ...)
392 struct va_format vaf
;
398 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
399 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
402 ext4_handle_error(sb
);
405 void ext4_error_inode(struct inode
*inode
, const char *function
,
406 unsigned int line
, ext4_fsblk_t block
,
407 const char *fmt
, ...)
410 struct va_format vaf
;
411 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
413 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
414 es
->s_last_error_block
= cpu_to_le64(block
);
415 save_error_info(inode
->i_sb
, function
, line
);
419 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: inode #%lu: ",
420 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
422 printk(KERN_CONT
"block %llu: ", block
);
423 printk(KERN_CONT
"comm %s: %pV\n", current
->comm
, &vaf
);
426 ext4_handle_error(inode
->i_sb
);
429 void ext4_error_file(struct file
*file
, const char *function
,
430 unsigned int line
, ext4_fsblk_t block
,
431 const char *fmt
, ...)
434 struct va_format vaf
;
435 struct ext4_super_block
*es
;
436 struct inode
*inode
= file
->f_dentry
->d_inode
;
437 char pathname
[80], *path
;
439 es
= EXT4_SB(inode
->i_sb
)->s_es
;
440 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
441 save_error_info(inode
->i_sb
, function
, line
);
442 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
446 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
447 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
449 printk(KERN_CONT
"block %llu: ", block
);
453 printk(KERN_CONT
"comm %s: path %s: %pV\n", current
->comm
, path
, &vaf
);
456 ext4_handle_error(inode
->i_sb
);
459 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
466 errstr
= "IO failure";
469 errstr
= "Out of memory";
472 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
473 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
474 errstr
= "Journal has aborted";
476 errstr
= "Readonly filesystem";
479 /* If the caller passed in an extra buffer for unknown
480 * errors, textualise them now. Else we just return
483 /* Check for truncated error codes... */
484 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
493 /* __ext4_std_error decodes expected errors from journaling functions
494 * automatically and invokes the appropriate error response. */
496 void __ext4_std_error(struct super_block
*sb
, const char *function
,
497 unsigned int line
, int errno
)
502 /* Special case: if the error is EROFS, and we're not already
503 * inside a transaction, then there's really no point in logging
505 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
506 (sb
->s_flags
& MS_RDONLY
))
509 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
510 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
511 sb
->s_id
, function
, line
, errstr
);
512 save_error_info(sb
, function
, line
);
514 ext4_handle_error(sb
);
518 * ext4_abort is a much stronger failure handler than ext4_error. The
519 * abort function may be used to deal with unrecoverable failures such
520 * as journal IO errors or ENOMEM at a critical moment in log management.
522 * We unconditionally force the filesystem into an ABORT|READONLY state,
523 * unless the error response on the fs has been set to panic in which
524 * case we take the easy way out and panic immediately.
527 void __ext4_abort(struct super_block
*sb
, const char *function
,
528 unsigned int line
, const char *fmt
, ...)
532 save_error_info(sb
, function
, line
);
534 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
540 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
541 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
542 sb
->s_flags
|= MS_RDONLY
;
543 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
544 if (EXT4_SB(sb
)->s_journal
)
545 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
546 save_error_info(sb
, function
, line
);
548 if (test_opt(sb
, ERRORS_PANIC
))
549 panic("EXT4-fs panic from previous error\n");
552 void ext4_msg(struct super_block
*sb
, const char *prefix
, const char *fmt
, ...)
554 struct va_format vaf
;
560 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
564 void __ext4_warning(struct super_block
*sb
, const char *function
,
565 unsigned int line
, const char *fmt
, ...)
567 struct va_format vaf
;
573 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
574 sb
->s_id
, function
, line
, &vaf
);
578 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
579 struct super_block
*sb
, ext4_group_t grp
,
580 unsigned long ino
, ext4_fsblk_t block
,
581 const char *fmt
, ...)
585 struct va_format vaf
;
587 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
589 es
->s_last_error_ino
= cpu_to_le32(ino
);
590 es
->s_last_error_block
= cpu_to_le64(block
);
591 __save_error_info(sb
, function
, line
);
597 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u",
598 sb
->s_id
, function
, line
, grp
);
600 printk(KERN_CONT
"inode %lu: ", ino
);
602 printk(KERN_CONT
"block %llu:", (unsigned long long) block
);
603 printk(KERN_CONT
"%pV\n", &vaf
);
606 if (test_opt(sb
, ERRORS_CONT
)) {
607 ext4_commit_super(sb
, 0);
611 ext4_unlock_group(sb
, grp
);
612 ext4_handle_error(sb
);
614 * We only get here in the ERRORS_RO case; relocking the group
615 * may be dangerous, but nothing bad will happen since the
616 * filesystem will have already been marked read/only and the
617 * journal has been aborted. We return 1 as a hint to callers
618 * who might what to use the return value from
619 * ext4_grp_locked_error() to distinguish beween the
620 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
621 * aggressively from the ext4 function in question, with a
622 * more appropriate error code.
624 ext4_lock_group(sb
, grp
);
628 void ext4_update_dynamic_rev(struct super_block
*sb
)
630 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
632 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
636 "updating to rev %d because of new feature flag, "
637 "running e2fsck is recommended",
640 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
641 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
642 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
643 /* leave es->s_feature_*compat flags alone */
644 /* es->s_uuid will be set by e2fsck if empty */
647 * The rest of the superblock fields should be zero, and if not it
648 * means they are likely already in use, so leave them alone. We
649 * can leave it up to e2fsck to clean up any inconsistencies there.
654 * Open the external journal device
656 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
658 struct block_device
*bdev
;
659 char b
[BDEVNAME_SIZE
];
661 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
667 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
668 __bdevname(dev
, b
), PTR_ERR(bdev
));
673 * Release the journal device
675 static int ext4_blkdev_put(struct block_device
*bdev
)
677 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
680 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
682 struct block_device
*bdev
;
685 bdev
= sbi
->journal_bdev
;
687 ret
= ext4_blkdev_put(bdev
);
688 sbi
->journal_bdev
= NULL
;
693 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
695 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
698 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
702 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
703 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
705 printk(KERN_ERR
"sb_info orphan list:\n");
706 list_for_each(l
, &sbi
->s_orphan
) {
707 struct inode
*inode
= orphan_list_entry(l
);
709 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
710 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
711 inode
->i_mode
, inode
->i_nlink
,
716 static void ext4_put_super(struct super_block
*sb
)
718 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
719 struct ext4_super_block
*es
= sbi
->s_es
;
722 ext4_unregister_li_request(sb
);
723 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
725 flush_workqueue(sbi
->dio_unwritten_wq
);
726 destroy_workqueue(sbi
->dio_unwritten_wq
);
730 ext4_commit_super(sb
, 1);
732 if (sbi
->s_journal
) {
733 err
= jbd2_journal_destroy(sbi
->s_journal
);
734 sbi
->s_journal
= NULL
;
736 ext4_abort(sb
, "Couldn't clean up the journal");
739 del_timer(&sbi
->s_err_report
);
740 ext4_release_system_zone(sb
);
742 ext4_ext_release(sb
);
743 ext4_xattr_put_super(sb
);
745 if (!(sb
->s_flags
& MS_RDONLY
)) {
746 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
747 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
748 ext4_commit_super(sb
, 1);
751 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
753 kobject_del(&sbi
->s_kobj
);
755 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
756 brelse(sbi
->s_group_desc
[i
]);
757 kfree(sbi
->s_group_desc
);
758 if (is_vmalloc_addr(sbi
->s_flex_groups
))
759 vfree(sbi
->s_flex_groups
);
761 kfree(sbi
->s_flex_groups
);
762 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
763 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
764 percpu_counter_destroy(&sbi
->s_dirs_counter
);
765 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
768 for (i
= 0; i
< MAXQUOTAS
; i
++)
769 kfree(sbi
->s_qf_names
[i
]);
772 /* Debugging code just in case the in-memory inode orphan list
773 * isn't empty. The on-disk one can be non-empty if we've
774 * detected an error and taken the fs readonly, but the
775 * in-memory list had better be clean by this point. */
776 if (!list_empty(&sbi
->s_orphan
))
777 dump_orphan_list(sb
, sbi
);
778 J_ASSERT(list_empty(&sbi
->s_orphan
));
780 invalidate_bdev(sb
->s_bdev
);
781 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
783 * Invalidate the journal device's buffers. We don't want them
784 * floating about in memory - the physical journal device may
785 * hotswapped, and it breaks the `ro-after' testing code.
787 sync_blockdev(sbi
->journal_bdev
);
788 invalidate_bdev(sbi
->journal_bdev
);
789 ext4_blkdev_remove(sbi
);
791 sb
->s_fs_info
= NULL
;
793 * Now that we are completely done shutting down the
794 * superblock, we need to actually destroy the kobject.
797 kobject_put(&sbi
->s_kobj
);
798 wait_for_completion(&sbi
->s_kobj_unregister
);
799 kfree(sbi
->s_blockgroup_lock
);
803 static struct kmem_cache
*ext4_inode_cachep
;
806 * Called inside transaction, so use GFP_NOFS
808 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
810 struct ext4_inode_info
*ei
;
812 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
816 ei
->vfs_inode
.i_version
= 1;
817 ei
->vfs_inode
.i_data
.writeback_index
= 0;
818 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
819 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
820 spin_lock_init(&ei
->i_prealloc_lock
);
821 ei
->i_reserved_data_blocks
= 0;
822 ei
->i_reserved_meta_blocks
= 0;
823 ei
->i_allocated_meta_blocks
= 0;
824 ei
->i_da_metadata_calc_len
= 0;
825 spin_lock_init(&(ei
->i_block_reservation_lock
));
827 ei
->i_reserved_quota
= 0;
830 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
831 spin_lock_init(&ei
->i_completed_io_lock
);
832 ei
->cur_aio_dio
= NULL
;
834 ei
->i_datasync_tid
= 0;
835 atomic_set(&ei
->i_ioend_count
, 0);
837 return &ei
->vfs_inode
;
840 static int ext4_drop_inode(struct inode
*inode
)
842 int drop
= generic_drop_inode(inode
);
844 trace_ext4_drop_inode(inode
, drop
);
848 static void ext4_i_callback(struct rcu_head
*head
)
850 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
851 INIT_LIST_HEAD(&inode
->i_dentry
);
852 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
855 static void ext4_destroy_inode(struct inode
*inode
)
857 ext4_ioend_wait(inode
);
858 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
859 ext4_msg(inode
->i_sb
, KERN_ERR
,
860 "Inode %lu (%p): orphan list check failed!",
861 inode
->i_ino
, EXT4_I(inode
));
862 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
863 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
867 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
870 static void init_once(void *foo
)
872 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
874 INIT_LIST_HEAD(&ei
->i_orphan
);
875 #ifdef CONFIG_EXT4_FS_XATTR
876 init_rwsem(&ei
->xattr_sem
);
878 init_rwsem(&ei
->i_data_sem
);
879 inode_init_once(&ei
->vfs_inode
);
882 static int init_inodecache(void)
884 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
885 sizeof(struct ext4_inode_info
),
886 0, (SLAB_RECLAIM_ACCOUNT
|
889 if (ext4_inode_cachep
== NULL
)
894 static void destroy_inodecache(void)
896 kmem_cache_destroy(ext4_inode_cachep
);
899 void ext4_clear_inode(struct inode
*inode
)
901 invalidate_inode_buffers(inode
);
902 end_writeback(inode
);
904 ext4_discard_preallocations(inode
);
905 if (EXT4_I(inode
)->jinode
) {
906 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
907 EXT4_I(inode
)->jinode
);
908 jbd2_free_inode(EXT4_I(inode
)->jinode
);
909 EXT4_I(inode
)->jinode
= NULL
;
913 static inline void ext4_show_quota_options(struct seq_file
*seq
,
914 struct super_block
*sb
)
916 #if defined(CONFIG_QUOTA)
917 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
919 if (sbi
->s_jquota_fmt
) {
922 switch (sbi
->s_jquota_fmt
) {
933 seq_printf(seq
, ",jqfmt=%s", fmtname
);
936 if (sbi
->s_qf_names
[USRQUOTA
])
937 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
939 if (sbi
->s_qf_names
[GRPQUOTA
])
940 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
942 if (test_opt(sb
, USRQUOTA
))
943 seq_puts(seq
, ",usrquota");
945 if (test_opt(sb
, GRPQUOTA
))
946 seq_puts(seq
, ",grpquota");
952 * - it's set to a non-default value OR
953 * - if the per-sb default is different from the global default
955 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
958 unsigned long def_mount_opts
;
959 struct super_block
*sb
= vfs
->mnt_sb
;
960 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
961 struct ext4_super_block
*es
= sbi
->s_es
;
963 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
964 def_errors
= le16_to_cpu(es
->s_errors
);
966 if (sbi
->s_sb_block
!= 1)
967 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
968 if (test_opt(sb
, MINIX_DF
))
969 seq_puts(seq
, ",minixdf");
970 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
971 seq_puts(seq
, ",grpid");
972 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
973 seq_puts(seq
, ",nogrpid");
974 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
975 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
976 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
978 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
979 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
980 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
982 if (test_opt(sb
, ERRORS_RO
)) {
983 if (def_errors
== EXT4_ERRORS_PANIC
||
984 def_errors
== EXT4_ERRORS_CONTINUE
) {
985 seq_puts(seq
, ",errors=remount-ro");
988 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
989 seq_puts(seq
, ",errors=continue");
990 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
991 seq_puts(seq
, ",errors=panic");
992 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
993 seq_puts(seq
, ",nouid32");
994 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
995 seq_puts(seq
, ",debug");
996 if (test_opt(sb
, OLDALLOC
))
997 seq_puts(seq
, ",oldalloc");
998 #ifdef CONFIG_EXT4_FS_XATTR
999 if (test_opt(sb
, XATTR_USER
) &&
1000 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
1001 seq_puts(seq
, ",user_xattr");
1002 if (!test_opt(sb
, XATTR_USER
) &&
1003 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
1004 seq_puts(seq
, ",nouser_xattr");
1007 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1008 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
1009 seq_puts(seq
, ",acl");
1010 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
1011 seq_puts(seq
, ",noacl");
1013 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
1014 seq_printf(seq
, ",commit=%u",
1015 (unsigned) (sbi
->s_commit_interval
/ HZ
));
1017 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
1018 seq_printf(seq
, ",min_batch_time=%u",
1019 (unsigned) sbi
->s_min_batch_time
);
1021 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
1022 seq_printf(seq
, ",max_batch_time=%u",
1023 (unsigned) sbi
->s_min_batch_time
);
1027 * We're changing the default of barrier mount option, so
1028 * let's always display its mount state so it's clear what its
1031 seq_puts(seq
, ",barrier=");
1032 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
1033 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
1034 seq_puts(seq
, ",journal_async_commit");
1035 else if (test_opt(sb
, JOURNAL_CHECKSUM
))
1036 seq_puts(seq
, ",journal_checksum");
1037 if (test_opt(sb
, I_VERSION
))
1038 seq_puts(seq
, ",i_version");
1039 if (!test_opt(sb
, DELALLOC
) &&
1040 !(def_mount_opts
& EXT4_DEFM_NODELALLOC
))
1041 seq_puts(seq
, ",nodelalloc");
1043 if (test_opt(sb
, MBLK_IO_SUBMIT
))
1044 seq_puts(seq
, ",mblk_io_submit");
1046 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
1048 * journal mode get enabled in different ways
1049 * So just print the value even if we didn't specify it
1051 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1052 seq_puts(seq
, ",data=journal");
1053 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1054 seq_puts(seq
, ",data=ordered");
1055 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1056 seq_puts(seq
, ",data=writeback");
1058 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1059 seq_printf(seq
, ",inode_readahead_blks=%u",
1060 sbi
->s_inode_readahead_blks
);
1062 if (test_opt(sb
, DATA_ERR_ABORT
))
1063 seq_puts(seq
, ",data_err=abort");
1065 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
1066 seq_puts(seq
, ",noauto_da_alloc");
1068 if (test_opt(sb
, DISCARD
) && !(def_mount_opts
& EXT4_DEFM_DISCARD
))
1069 seq_puts(seq
, ",discard");
1071 if (test_opt(sb
, NOLOAD
))
1072 seq_puts(seq
, ",norecovery");
1074 if (test_opt(sb
, DIOREAD_NOLOCK
))
1075 seq_puts(seq
, ",dioread_nolock");
1077 if (test_opt(sb
, BLOCK_VALIDITY
) &&
1078 !(def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
))
1079 seq_puts(seq
, ",block_validity");
1081 if (!test_opt(sb
, INIT_INODE_TABLE
))
1082 seq_puts(seq
, ",noinit_inode_table");
1083 else if (sbi
->s_li_wait_mult
)
1084 seq_printf(seq
, ",init_inode_table=%u",
1085 (unsigned) sbi
->s_li_wait_mult
);
1087 ext4_show_quota_options(seq
, sb
);
1092 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1093 u64 ino
, u32 generation
)
1095 struct inode
*inode
;
1097 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1098 return ERR_PTR(-ESTALE
);
1099 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1100 return ERR_PTR(-ESTALE
);
1102 /* iget isn't really right if the inode is currently unallocated!!
1104 * ext4_read_inode will return a bad_inode if the inode had been
1105 * deleted, so we should be safe.
1107 * Currently we don't know the generation for parent directory, so
1108 * a generation of 0 means "accept any"
1110 inode
= ext4_iget(sb
, ino
);
1112 return ERR_CAST(inode
);
1113 if (generation
&& inode
->i_generation
!= generation
) {
1115 return ERR_PTR(-ESTALE
);
1121 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1122 int fh_len
, int fh_type
)
1124 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1125 ext4_nfs_get_inode
);
1128 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1129 int fh_len
, int fh_type
)
1131 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1132 ext4_nfs_get_inode
);
1136 * Try to release metadata pages (indirect blocks, directories) which are
1137 * mapped via the block device. Since these pages could have journal heads
1138 * which would prevent try_to_free_buffers() from freeing them, we must use
1139 * jbd2 layer's try_to_free_buffers() function to release them.
1141 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1144 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1146 WARN_ON(PageChecked(page
));
1147 if (!page_has_buffers(page
))
1150 return jbd2_journal_try_to_free_buffers(journal
, page
,
1151 wait
& ~__GFP_WAIT
);
1152 return try_to_free_buffers(page
);
1156 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1157 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1159 static int ext4_write_dquot(struct dquot
*dquot
);
1160 static int ext4_acquire_dquot(struct dquot
*dquot
);
1161 static int ext4_release_dquot(struct dquot
*dquot
);
1162 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1163 static int ext4_write_info(struct super_block
*sb
, int type
);
1164 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1166 static int ext4_quota_off(struct super_block
*sb
, int type
);
1167 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1168 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1169 size_t len
, loff_t off
);
1170 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1171 const char *data
, size_t len
, loff_t off
);
1173 static const struct dquot_operations ext4_quota_operations
= {
1175 .get_reserved_space
= ext4_get_reserved_space
,
1177 .write_dquot
= ext4_write_dquot
,
1178 .acquire_dquot
= ext4_acquire_dquot
,
1179 .release_dquot
= ext4_release_dquot
,
1180 .mark_dirty
= ext4_mark_dquot_dirty
,
1181 .write_info
= ext4_write_info
,
1182 .alloc_dquot
= dquot_alloc
,
1183 .destroy_dquot
= dquot_destroy
,
1186 static const struct quotactl_ops ext4_qctl_operations
= {
1187 .quota_on
= ext4_quota_on
,
1188 .quota_off
= ext4_quota_off
,
1189 .quota_sync
= dquot_quota_sync
,
1190 .get_info
= dquot_get_dqinfo
,
1191 .set_info
= dquot_set_dqinfo
,
1192 .get_dqblk
= dquot_get_dqblk
,
1193 .set_dqblk
= dquot_set_dqblk
1197 static const struct super_operations ext4_sops
= {
1198 .alloc_inode
= ext4_alloc_inode
,
1199 .destroy_inode
= ext4_destroy_inode
,
1200 .write_inode
= ext4_write_inode
,
1201 .dirty_inode
= ext4_dirty_inode
,
1202 .drop_inode
= ext4_drop_inode
,
1203 .evict_inode
= ext4_evict_inode
,
1204 .put_super
= ext4_put_super
,
1205 .sync_fs
= ext4_sync_fs
,
1206 .freeze_fs
= ext4_freeze
,
1207 .unfreeze_fs
= ext4_unfreeze
,
1208 .statfs
= ext4_statfs
,
1209 .remount_fs
= ext4_remount
,
1210 .show_options
= ext4_show_options
,
1212 .quota_read
= ext4_quota_read
,
1213 .quota_write
= ext4_quota_write
,
1215 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1218 static const struct super_operations ext4_nojournal_sops
= {
1219 .alloc_inode
= ext4_alloc_inode
,
1220 .destroy_inode
= ext4_destroy_inode
,
1221 .write_inode
= ext4_write_inode
,
1222 .dirty_inode
= ext4_dirty_inode
,
1223 .drop_inode
= ext4_drop_inode
,
1224 .evict_inode
= ext4_evict_inode
,
1225 .write_super
= ext4_write_super
,
1226 .put_super
= ext4_put_super
,
1227 .statfs
= ext4_statfs
,
1228 .remount_fs
= ext4_remount
,
1229 .show_options
= ext4_show_options
,
1231 .quota_read
= ext4_quota_read
,
1232 .quota_write
= ext4_quota_write
,
1234 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1237 static const struct export_operations ext4_export_ops
= {
1238 .fh_to_dentry
= ext4_fh_to_dentry
,
1239 .fh_to_parent
= ext4_fh_to_parent
,
1240 .get_parent
= ext4_get_parent
,
1244 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1245 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1246 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1247 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1248 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1249 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1250 Opt_journal_update
, Opt_journal_dev
,
1251 Opt_journal_checksum
, Opt_journal_async_commit
,
1252 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1253 Opt_data_err_abort
, Opt_data_err_ignore
,
1254 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1255 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1256 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1257 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1258 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1259 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1260 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1261 Opt_dioread_nolock
, Opt_dioread_lock
,
1262 Opt_discard
, Opt_nodiscard
,
1263 Opt_init_inode_table
, Opt_noinit_inode_table
,
1266 static const match_table_t tokens
= {
1267 {Opt_bsd_df
, "bsddf"},
1268 {Opt_minix_df
, "minixdf"},
1269 {Opt_grpid
, "grpid"},
1270 {Opt_grpid
, "bsdgroups"},
1271 {Opt_nogrpid
, "nogrpid"},
1272 {Opt_nogrpid
, "sysvgroups"},
1273 {Opt_resgid
, "resgid=%u"},
1274 {Opt_resuid
, "resuid=%u"},
1276 {Opt_err_cont
, "errors=continue"},
1277 {Opt_err_panic
, "errors=panic"},
1278 {Opt_err_ro
, "errors=remount-ro"},
1279 {Opt_nouid32
, "nouid32"},
1280 {Opt_debug
, "debug"},
1281 {Opt_oldalloc
, "oldalloc"},
1282 {Opt_orlov
, "orlov"},
1283 {Opt_user_xattr
, "user_xattr"},
1284 {Opt_nouser_xattr
, "nouser_xattr"},
1286 {Opt_noacl
, "noacl"},
1287 {Opt_noload
, "noload"},
1288 {Opt_noload
, "norecovery"},
1291 {Opt_commit
, "commit=%u"},
1292 {Opt_min_batch_time
, "min_batch_time=%u"},
1293 {Opt_max_batch_time
, "max_batch_time=%u"},
1294 {Opt_journal_update
, "journal=update"},
1295 {Opt_journal_dev
, "journal_dev=%u"},
1296 {Opt_journal_checksum
, "journal_checksum"},
1297 {Opt_journal_async_commit
, "journal_async_commit"},
1298 {Opt_abort
, "abort"},
1299 {Opt_data_journal
, "data=journal"},
1300 {Opt_data_ordered
, "data=ordered"},
1301 {Opt_data_writeback
, "data=writeback"},
1302 {Opt_data_err_abort
, "data_err=abort"},
1303 {Opt_data_err_ignore
, "data_err=ignore"},
1304 {Opt_offusrjquota
, "usrjquota="},
1305 {Opt_usrjquota
, "usrjquota=%s"},
1306 {Opt_offgrpjquota
, "grpjquota="},
1307 {Opt_grpjquota
, "grpjquota=%s"},
1308 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1309 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1310 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1311 {Opt_grpquota
, "grpquota"},
1312 {Opt_noquota
, "noquota"},
1313 {Opt_quota
, "quota"},
1314 {Opt_usrquota
, "usrquota"},
1315 {Opt_barrier
, "barrier=%u"},
1316 {Opt_barrier
, "barrier"},
1317 {Opt_nobarrier
, "nobarrier"},
1318 {Opt_i_version
, "i_version"},
1319 {Opt_stripe
, "stripe=%u"},
1320 {Opt_resize
, "resize"},
1321 {Opt_delalloc
, "delalloc"},
1322 {Opt_nodelalloc
, "nodelalloc"},
1323 {Opt_mblk_io_submit
, "mblk_io_submit"},
1324 {Opt_nomblk_io_submit
, "nomblk_io_submit"},
1325 {Opt_block_validity
, "block_validity"},
1326 {Opt_noblock_validity
, "noblock_validity"},
1327 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1328 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1329 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1330 {Opt_auto_da_alloc
, "auto_da_alloc"},
1331 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1332 {Opt_dioread_nolock
, "dioread_nolock"},
1333 {Opt_dioread_lock
, "dioread_lock"},
1334 {Opt_discard
, "discard"},
1335 {Opt_nodiscard
, "nodiscard"},
1336 {Opt_init_inode_table
, "init_itable=%u"},
1337 {Opt_init_inode_table
, "init_itable"},
1338 {Opt_noinit_inode_table
, "noinit_itable"},
1342 static ext4_fsblk_t
get_sb_block(void **data
)
1344 ext4_fsblk_t sb_block
;
1345 char *options
= (char *) *data
;
1347 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1348 return 1; /* Default location */
1351 /* TODO: use simple_strtoll with >32bit ext4 */
1352 sb_block
= simple_strtoul(options
, &options
, 0);
1353 if (*options
&& *options
!= ',') {
1354 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1358 if (*options
== ',')
1360 *data
= (void *) options
;
1365 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1366 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1367 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1370 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1372 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1375 if (sb_any_quota_loaded(sb
) &&
1376 !sbi
->s_qf_names
[qtype
]) {
1377 ext4_msg(sb
, KERN_ERR
,
1378 "Cannot change journaled "
1379 "quota options when quota turned on");
1382 qname
= match_strdup(args
);
1384 ext4_msg(sb
, KERN_ERR
,
1385 "Not enough memory for storing quotafile name");
1388 if (sbi
->s_qf_names
[qtype
] &&
1389 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1390 ext4_msg(sb
, KERN_ERR
,
1391 "%s quota file already specified", QTYPE2NAME(qtype
));
1395 sbi
->s_qf_names
[qtype
] = qname
;
1396 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1397 ext4_msg(sb
, KERN_ERR
,
1398 "quotafile must be on filesystem root");
1399 kfree(sbi
->s_qf_names
[qtype
]);
1400 sbi
->s_qf_names
[qtype
] = NULL
;
1407 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1410 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1412 if (sb_any_quota_loaded(sb
) &&
1413 sbi
->s_qf_names
[qtype
]) {
1414 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1415 " when quota turned on");
1419 * The space will be released later when all options are confirmed
1422 sbi
->s_qf_names
[qtype
] = NULL
;
1427 static int parse_options(char *options
, struct super_block
*sb
,
1428 unsigned long *journal_devnum
,
1429 unsigned int *journal_ioprio
,
1430 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1432 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1434 substring_t args
[MAX_OPT_ARGS
];
1444 while ((p
= strsep(&options
, ",")) != NULL
) {
1450 * Initialize args struct so we know whether arg was
1451 * found; some options take optional arguments.
1453 args
[0].to
= args
[0].from
= 0;
1454 token
= match_token(p
, tokens
, args
);
1457 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1458 clear_opt(sb
, MINIX_DF
);
1461 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1462 set_opt(sb
, MINIX_DF
);
1466 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1471 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1472 clear_opt(sb
, GRPID
);
1476 if (match_int(&args
[0], &option
))
1478 sbi
->s_resuid
= option
;
1481 if (match_int(&args
[0], &option
))
1483 sbi
->s_resgid
= option
;
1486 /* handled by get_sb_block() instead of here */
1487 /* *sb_block = match_int(&args[0]); */
1490 clear_opt(sb
, ERRORS_CONT
);
1491 clear_opt(sb
, ERRORS_RO
);
1492 set_opt(sb
, ERRORS_PANIC
);
1495 clear_opt(sb
, ERRORS_CONT
);
1496 clear_opt(sb
, ERRORS_PANIC
);
1497 set_opt(sb
, ERRORS_RO
);
1500 clear_opt(sb
, ERRORS_RO
);
1501 clear_opt(sb
, ERRORS_PANIC
);
1502 set_opt(sb
, ERRORS_CONT
);
1505 set_opt(sb
, NO_UID32
);
1511 set_opt(sb
, OLDALLOC
);
1514 clear_opt(sb
, OLDALLOC
);
1516 #ifdef CONFIG_EXT4_FS_XATTR
1517 case Opt_user_xattr
:
1518 set_opt(sb
, XATTR_USER
);
1520 case Opt_nouser_xattr
:
1521 clear_opt(sb
, XATTR_USER
);
1524 case Opt_user_xattr
:
1525 case Opt_nouser_xattr
:
1526 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1529 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1531 set_opt(sb
, POSIX_ACL
);
1534 clear_opt(sb
, POSIX_ACL
);
1539 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1542 case Opt_journal_update
:
1544 /* Eventually we will want to be able to create
1545 a journal file here. For now, only allow the
1546 user to specify an existing inode to be the
1549 ext4_msg(sb
, KERN_ERR
,
1550 "Cannot specify journal on remount");
1553 set_opt(sb
, UPDATE_JOURNAL
);
1555 case Opt_journal_dev
:
1557 ext4_msg(sb
, KERN_ERR
,
1558 "Cannot specify journal on remount");
1561 if (match_int(&args
[0], &option
))
1563 *journal_devnum
= option
;
1565 case Opt_journal_checksum
:
1566 set_opt(sb
, JOURNAL_CHECKSUM
);
1568 case Opt_journal_async_commit
:
1569 set_opt(sb
, JOURNAL_ASYNC_COMMIT
);
1570 set_opt(sb
, JOURNAL_CHECKSUM
);
1573 set_opt(sb
, NOLOAD
);
1576 if (match_int(&args
[0], &option
))
1581 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1582 sbi
->s_commit_interval
= HZ
* option
;
1584 case Opt_max_batch_time
:
1585 if (match_int(&args
[0], &option
))
1590 option
= EXT4_DEF_MAX_BATCH_TIME
;
1591 sbi
->s_max_batch_time
= option
;
1593 case Opt_min_batch_time
:
1594 if (match_int(&args
[0], &option
))
1598 sbi
->s_min_batch_time
= option
;
1600 case Opt_data_journal
:
1601 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1603 case Opt_data_ordered
:
1604 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1606 case Opt_data_writeback
:
1607 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1610 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1611 ext4_msg(sb
, KERN_ERR
,
1612 "Cannot change data mode on remount");
1616 clear_opt(sb
, DATA_FLAGS
);
1617 sbi
->s_mount_opt
|= data_opt
;
1620 case Opt_data_err_abort
:
1621 set_opt(sb
, DATA_ERR_ABORT
);
1623 case Opt_data_err_ignore
:
1624 clear_opt(sb
, DATA_ERR_ABORT
);
1628 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1632 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1635 case Opt_offusrjquota
:
1636 if (!clear_qf_name(sb
, USRQUOTA
))
1639 case Opt_offgrpjquota
:
1640 if (!clear_qf_name(sb
, GRPQUOTA
))
1644 case Opt_jqfmt_vfsold
:
1645 qfmt
= QFMT_VFS_OLD
;
1647 case Opt_jqfmt_vfsv0
:
1650 case Opt_jqfmt_vfsv1
:
1653 if (sb_any_quota_loaded(sb
) &&
1654 sbi
->s_jquota_fmt
!= qfmt
) {
1655 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1656 "journaled quota options when "
1660 sbi
->s_jquota_fmt
= qfmt
;
1665 set_opt(sb
, USRQUOTA
);
1669 set_opt(sb
, GRPQUOTA
);
1672 if (sb_any_quota_loaded(sb
)) {
1673 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1674 "options when quota turned on");
1677 clear_opt(sb
, QUOTA
);
1678 clear_opt(sb
, USRQUOTA
);
1679 clear_opt(sb
, GRPQUOTA
);
1685 ext4_msg(sb
, KERN_ERR
,
1686 "quota options not supported");
1690 case Opt_offusrjquota
:
1691 case Opt_offgrpjquota
:
1692 case Opt_jqfmt_vfsold
:
1693 case Opt_jqfmt_vfsv0
:
1694 case Opt_jqfmt_vfsv1
:
1695 ext4_msg(sb
, KERN_ERR
,
1696 "journaled quota options not supported");
1702 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1705 clear_opt(sb
, BARRIER
);
1709 if (match_int(&args
[0], &option
))
1712 option
= 1; /* No argument, default to 1 */
1714 set_opt(sb
, BARRIER
);
1716 clear_opt(sb
, BARRIER
);
1722 ext4_msg(sb
, KERN_ERR
,
1723 "resize option only available "
1727 if (match_int(&args
[0], &option
) != 0)
1729 *n_blocks_count
= option
;
1732 ext4_msg(sb
, KERN_WARNING
,
1733 "Ignoring deprecated nobh option");
1736 ext4_msg(sb
, KERN_WARNING
,
1737 "Ignoring deprecated bh option");
1740 set_opt(sb
, I_VERSION
);
1741 sb
->s_flags
|= MS_I_VERSION
;
1743 case Opt_nodelalloc
:
1744 clear_opt(sb
, DELALLOC
);
1746 case Opt_mblk_io_submit
:
1747 set_opt(sb
, MBLK_IO_SUBMIT
);
1749 case Opt_nomblk_io_submit
:
1750 clear_opt(sb
, MBLK_IO_SUBMIT
);
1753 if (match_int(&args
[0], &option
))
1757 sbi
->s_stripe
= option
;
1760 set_opt(sb
, DELALLOC
);
1762 case Opt_block_validity
:
1763 set_opt(sb
, BLOCK_VALIDITY
);
1765 case Opt_noblock_validity
:
1766 clear_opt(sb
, BLOCK_VALIDITY
);
1768 case Opt_inode_readahead_blks
:
1769 if (match_int(&args
[0], &option
))
1771 if (option
< 0 || option
> (1 << 30))
1773 if (!is_power_of_2(option
)) {
1774 ext4_msg(sb
, KERN_ERR
,
1775 "EXT4-fs: inode_readahead_blks"
1776 " must be a power of 2");
1779 sbi
->s_inode_readahead_blks
= option
;
1781 case Opt_journal_ioprio
:
1782 if (match_int(&args
[0], &option
))
1784 if (option
< 0 || option
> 7)
1786 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1789 case Opt_noauto_da_alloc
:
1790 set_opt(sb
, NO_AUTO_DA_ALLOC
);
1792 case Opt_auto_da_alloc
:
1794 if (match_int(&args
[0], &option
))
1797 option
= 1; /* No argument, default to 1 */
1799 clear_opt(sb
, NO_AUTO_DA_ALLOC
);
1801 set_opt(sb
,NO_AUTO_DA_ALLOC
);
1804 set_opt(sb
, DISCARD
);
1807 clear_opt(sb
, DISCARD
);
1809 case Opt_dioread_nolock
:
1810 set_opt(sb
, DIOREAD_NOLOCK
);
1812 case Opt_dioread_lock
:
1813 clear_opt(sb
, DIOREAD_NOLOCK
);
1815 case Opt_init_inode_table
:
1816 set_opt(sb
, INIT_INODE_TABLE
);
1818 if (match_int(&args
[0], &option
))
1821 option
= EXT4_DEF_LI_WAIT_MULT
;
1824 sbi
->s_li_wait_mult
= option
;
1826 case Opt_noinit_inode_table
:
1827 clear_opt(sb
, INIT_INODE_TABLE
);
1830 ext4_msg(sb
, KERN_ERR
,
1831 "Unrecognized mount option \"%s\" "
1832 "or missing value", p
);
1837 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1838 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1839 clear_opt(sb
, USRQUOTA
);
1841 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1842 clear_opt(sb
, GRPQUOTA
);
1844 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1845 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1850 if (!sbi
->s_jquota_fmt
) {
1851 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1856 if (sbi
->s_jquota_fmt
) {
1857 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1858 "specified with no journaling "
1867 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1870 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1873 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1874 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1875 "forcing read-only mode");
1880 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1881 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1882 "running e2fsck is recommended");
1883 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1884 ext4_msg(sb
, KERN_WARNING
,
1885 "warning: mounting fs with errors, "
1886 "running e2fsck is recommended");
1887 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1888 le16_to_cpu(es
->s_mnt_count
) >=
1889 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1890 ext4_msg(sb
, KERN_WARNING
,
1891 "warning: maximal mount count reached, "
1892 "running e2fsck is recommended");
1893 else if (le32_to_cpu(es
->s_checkinterval
) &&
1894 (le32_to_cpu(es
->s_lastcheck
) +
1895 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1896 ext4_msg(sb
, KERN_WARNING
,
1897 "warning: checktime reached, "
1898 "running e2fsck is recommended");
1899 if (!sbi
->s_journal
)
1900 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1901 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1902 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1903 le16_add_cpu(&es
->s_mnt_count
, 1);
1904 es
->s_mtime
= cpu_to_le32(get_seconds());
1905 ext4_update_dynamic_rev(sb
);
1907 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1909 ext4_commit_super(sb
, 1);
1910 if (test_opt(sb
, DEBUG
))
1911 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1912 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1914 sbi
->s_groups_count
,
1915 EXT4_BLOCKS_PER_GROUP(sb
),
1916 EXT4_INODES_PER_GROUP(sb
),
1917 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1922 static int ext4_fill_flex_info(struct super_block
*sb
)
1924 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1925 struct ext4_group_desc
*gdp
= NULL
;
1926 ext4_group_t flex_group_count
;
1927 ext4_group_t flex_group
;
1928 int groups_per_flex
= 0;
1932 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1933 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1935 if (groups_per_flex
< 2) {
1936 sbi
->s_log_groups_per_flex
= 0;
1940 /* We allocate both existing and potentially added groups */
1941 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1942 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1943 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1944 size
= flex_group_count
* sizeof(struct flex_groups
);
1945 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1946 if (sbi
->s_flex_groups
== NULL
) {
1947 sbi
->s_flex_groups
= vzalloc(size
);
1948 if (sbi
->s_flex_groups
== NULL
) {
1949 ext4_msg(sb
, KERN_ERR
,
1950 "not enough memory for %u flex groups",
1956 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1957 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1959 flex_group
= ext4_flex_group(sbi
, i
);
1960 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1961 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1962 atomic_add(ext4_free_blks_count(sb
, gdp
),
1963 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1964 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1965 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1973 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1974 struct ext4_group_desc
*gdp
)
1978 if (sbi
->s_es
->s_feature_ro_compat
&
1979 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1980 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1981 __le32 le_group
= cpu_to_le32(block_group
);
1983 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1984 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1985 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1986 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1987 /* for checksum of struct ext4_group_desc do the rest...*/
1988 if ((sbi
->s_es
->s_feature_incompat
&
1989 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1990 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1991 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1992 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1996 return cpu_to_le16(crc
);
1999 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
2000 struct ext4_group_desc
*gdp
)
2002 if ((sbi
->s_es
->s_feature_ro_compat
&
2003 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
2004 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
2010 /* Called at mount-time, super-block is locked */
2011 static int ext4_check_descriptors(struct super_block
*sb
,
2012 ext4_group_t
*first_not_zeroed
)
2014 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2015 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2016 ext4_fsblk_t last_block
;
2017 ext4_fsblk_t block_bitmap
;
2018 ext4_fsblk_t inode_bitmap
;
2019 ext4_fsblk_t inode_table
;
2020 int flexbg_flag
= 0;
2021 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2023 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2026 ext4_debug("Checking group descriptors");
2028 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2029 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2031 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2032 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2034 last_block
= first_block
+
2035 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2037 if ((grp
== sbi
->s_groups_count
) &&
2038 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2041 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2042 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2043 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2044 "Block bitmap for group %u not in group "
2045 "(block %llu)!", i
, block_bitmap
);
2048 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2049 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2050 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2051 "Inode bitmap for group %u not in group "
2052 "(block %llu)!", i
, inode_bitmap
);
2055 inode_table
= ext4_inode_table(sb
, gdp
);
2056 if (inode_table
< first_block
||
2057 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2058 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2059 "Inode table for group %u not in group "
2060 "(block %llu)!", i
, inode_table
);
2063 ext4_lock_group(sb
, i
);
2064 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
2065 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2066 "Checksum for group %u failed (%u!=%u)",
2067 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2068 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2069 if (!(sb
->s_flags
& MS_RDONLY
)) {
2070 ext4_unlock_group(sb
, i
);
2074 ext4_unlock_group(sb
, i
);
2076 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2078 if (NULL
!= first_not_zeroed
)
2079 *first_not_zeroed
= grp
;
2081 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
2082 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2086 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2087 * the superblock) which were deleted from all directories, but held open by
2088 * a process at the time of a crash. We walk the list and try to delete these
2089 * inodes at recovery time (only with a read-write filesystem).
2091 * In order to keep the orphan inode chain consistent during traversal (in
2092 * case of crash during recovery), we link each inode into the superblock
2093 * orphan list_head and handle it the same way as an inode deletion during
2094 * normal operation (which journals the operations for us).
2096 * We only do an iget() and an iput() on each inode, which is very safe if we
2097 * accidentally point at an in-use or already deleted inode. The worst that
2098 * can happen in this case is that we get a "bit already cleared" message from
2099 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2100 * e2fsck was run on this filesystem, and it must have already done the orphan
2101 * inode cleanup for us, so we can safely abort without any further action.
2103 static void ext4_orphan_cleanup(struct super_block
*sb
,
2104 struct ext4_super_block
*es
)
2106 unsigned int s_flags
= sb
->s_flags
;
2107 int nr_orphans
= 0, nr_truncates
= 0;
2111 if (!es
->s_last_orphan
) {
2112 jbd_debug(4, "no orphan inodes to clean up\n");
2116 if (bdev_read_only(sb
->s_bdev
)) {
2117 ext4_msg(sb
, KERN_ERR
, "write access "
2118 "unavailable, skipping orphan cleanup");
2122 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2123 if (es
->s_last_orphan
)
2124 jbd_debug(1, "Errors on filesystem, "
2125 "clearing orphan list.\n");
2126 es
->s_last_orphan
= 0;
2127 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2131 if (s_flags
& MS_RDONLY
) {
2132 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2133 sb
->s_flags
&= ~MS_RDONLY
;
2136 /* Needed for iput() to work correctly and not trash data */
2137 sb
->s_flags
|= MS_ACTIVE
;
2138 /* Turn on quotas so that they are updated correctly */
2139 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2140 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2141 int ret
= ext4_quota_on_mount(sb
, i
);
2143 ext4_msg(sb
, KERN_ERR
,
2144 "Cannot turn on journaled "
2145 "quota: error %d", ret
);
2150 while (es
->s_last_orphan
) {
2151 struct inode
*inode
;
2153 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2154 if (IS_ERR(inode
)) {
2155 es
->s_last_orphan
= 0;
2159 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2160 dquot_initialize(inode
);
2161 if (inode
->i_nlink
) {
2162 ext4_msg(sb
, KERN_DEBUG
,
2163 "%s: truncating inode %lu to %lld bytes",
2164 __func__
, inode
->i_ino
, inode
->i_size
);
2165 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2166 inode
->i_ino
, inode
->i_size
);
2167 ext4_truncate(inode
);
2170 ext4_msg(sb
, KERN_DEBUG
,
2171 "%s: deleting unreferenced inode %lu",
2172 __func__
, inode
->i_ino
);
2173 jbd_debug(2, "deleting unreferenced inode %lu\n",
2177 iput(inode
); /* The delete magic happens here! */
2180 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2183 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2184 PLURAL(nr_orphans
));
2186 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2187 PLURAL(nr_truncates
));
2189 /* Turn quotas off */
2190 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2191 if (sb_dqopt(sb
)->files
[i
])
2192 dquot_quota_off(sb
, i
);
2195 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2199 * Maximal extent format file size.
2200 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2201 * extent format containers, within a sector_t, and within i_blocks
2202 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2203 * so that won't be a limiting factor.
2205 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2207 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2210 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2212 /* small i_blocks in vfs inode? */
2213 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2215 * CONFIG_LBDAF is not enabled implies the inode
2216 * i_block represent total blocks in 512 bytes
2217 * 32 == size of vfs inode i_blocks * 8
2219 upper_limit
= (1LL << 32) - 1;
2221 /* total blocks in file system block size */
2222 upper_limit
>>= (blkbits
- 9);
2223 upper_limit
<<= blkbits
;
2226 /* 32-bit extent-start container, ee_block */
2231 /* Sanity check against vm- & vfs- imposed limits */
2232 if (res
> upper_limit
)
2239 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2240 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2241 * We need to be 1 filesystem block less than the 2^48 sector limit.
2243 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2245 loff_t res
= EXT4_NDIR_BLOCKS
;
2248 /* This is calculated to be the largest file size for a dense, block
2249 * mapped file such that the file's total number of 512-byte sectors,
2250 * including data and all indirect blocks, does not exceed (2^48 - 1).
2252 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2253 * number of 512-byte sectors of the file.
2256 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2258 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2259 * the inode i_block field represents total file blocks in
2260 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2262 upper_limit
= (1LL << 32) - 1;
2264 /* total blocks in file system block size */
2265 upper_limit
>>= (bits
- 9);
2269 * We use 48 bit ext4_inode i_blocks
2270 * With EXT4_HUGE_FILE_FL set the i_blocks
2271 * represent total number of blocks in
2272 * file system block size
2274 upper_limit
= (1LL << 48) - 1;
2278 /* indirect blocks */
2280 /* double indirect blocks */
2281 meta_blocks
+= 1 + (1LL << (bits
-2));
2282 /* tripple indirect blocks */
2283 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2285 upper_limit
-= meta_blocks
;
2286 upper_limit
<<= bits
;
2288 res
+= 1LL << (bits
-2);
2289 res
+= 1LL << (2*(bits
-2));
2290 res
+= 1LL << (3*(bits
-2));
2292 if (res
> upper_limit
)
2295 if (res
> MAX_LFS_FILESIZE
)
2296 res
= MAX_LFS_FILESIZE
;
2301 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2302 ext4_fsblk_t logical_sb_block
, int nr
)
2304 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2305 ext4_group_t bg
, first_meta_bg
;
2308 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2310 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2312 return logical_sb_block
+ nr
+ 1;
2313 bg
= sbi
->s_desc_per_block
* nr
;
2314 if (ext4_bg_has_super(sb
, bg
))
2317 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2321 * ext4_get_stripe_size: Get the stripe size.
2322 * @sbi: In memory super block info
2324 * If we have specified it via mount option, then
2325 * use the mount option value. If the value specified at mount time is
2326 * greater than the blocks per group use the super block value.
2327 * If the super block value is greater than blocks per group return 0.
2328 * Allocator needs it be less than blocks per group.
2331 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2333 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2334 unsigned long stripe_width
=
2335 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2337 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2338 return sbi
->s_stripe
;
2340 if (stripe_width
<= sbi
->s_blocks_per_group
)
2341 return stripe_width
;
2343 if (stride
<= sbi
->s_blocks_per_group
)
2352 struct attribute attr
;
2353 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2354 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2355 const char *, size_t);
2359 static int parse_strtoul(const char *buf
,
2360 unsigned long max
, unsigned long *value
)
2364 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2365 endp
= skip_spaces(endp
);
2366 if (*endp
|| *value
> max
)
2372 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2373 struct ext4_sb_info
*sbi
,
2376 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2377 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2380 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2381 struct ext4_sb_info
*sbi
, char *buf
)
2383 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2385 if (!sb
->s_bdev
->bd_part
)
2386 return snprintf(buf
, PAGE_SIZE
, "0\n");
2387 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2388 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2389 sbi
->s_sectors_written_start
) >> 1);
2392 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2393 struct ext4_sb_info
*sbi
, char *buf
)
2395 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2397 if (!sb
->s_bdev
->bd_part
)
2398 return snprintf(buf
, PAGE_SIZE
, "0\n");
2399 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2400 (unsigned long long)(sbi
->s_kbytes_written
+
2401 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2402 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2405 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2406 struct ext4_sb_info
*sbi
,
2407 const char *buf
, size_t count
)
2411 if (parse_strtoul(buf
, 0x40000000, &t
))
2414 if (!is_power_of_2(t
))
2417 sbi
->s_inode_readahead_blks
= t
;
2421 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2422 struct ext4_sb_info
*sbi
, char *buf
)
2424 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2426 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2429 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2430 struct ext4_sb_info
*sbi
,
2431 const char *buf
, size_t count
)
2433 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2436 if (parse_strtoul(buf
, 0xffffffff, &t
))
2442 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2443 static struct ext4_attr ext4_attr_##_name = { \
2444 .attr = {.name = __stringify(_name), .mode = _mode }, \
2447 .offset = offsetof(struct ext4_sb_info, _elname), \
2449 #define EXT4_ATTR(name, mode, show, store) \
2450 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2452 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2453 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2454 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2455 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2456 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2457 #define ATTR_LIST(name) &ext4_attr_##name.attr
2459 EXT4_RO_ATTR(delayed_allocation_blocks
);
2460 EXT4_RO_ATTR(session_write_kbytes
);
2461 EXT4_RO_ATTR(lifetime_write_kbytes
);
2462 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2463 inode_readahead_blks_store
, s_inode_readahead_blks
);
2464 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2465 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2466 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2467 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2468 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2469 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2470 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2471 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2473 static struct attribute
*ext4_attrs
[] = {
2474 ATTR_LIST(delayed_allocation_blocks
),
2475 ATTR_LIST(session_write_kbytes
),
2476 ATTR_LIST(lifetime_write_kbytes
),
2477 ATTR_LIST(inode_readahead_blks
),
2478 ATTR_LIST(inode_goal
),
2479 ATTR_LIST(mb_stats
),
2480 ATTR_LIST(mb_max_to_scan
),
2481 ATTR_LIST(mb_min_to_scan
),
2482 ATTR_LIST(mb_order2_req
),
2483 ATTR_LIST(mb_stream_req
),
2484 ATTR_LIST(mb_group_prealloc
),
2485 ATTR_LIST(max_writeback_mb_bump
),
2489 /* Features this copy of ext4 supports */
2490 EXT4_INFO_ATTR(lazy_itable_init
);
2491 EXT4_INFO_ATTR(batched_discard
);
2493 static struct attribute
*ext4_feat_attrs
[] = {
2494 ATTR_LIST(lazy_itable_init
),
2495 ATTR_LIST(batched_discard
),
2499 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2500 struct attribute
*attr
, char *buf
)
2502 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2504 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2506 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2509 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2510 struct attribute
*attr
,
2511 const char *buf
, size_t len
)
2513 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2515 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2517 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2520 static void ext4_sb_release(struct kobject
*kobj
)
2522 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2524 complete(&sbi
->s_kobj_unregister
);
2527 static const struct sysfs_ops ext4_attr_ops
= {
2528 .show
= ext4_attr_show
,
2529 .store
= ext4_attr_store
,
2532 static struct kobj_type ext4_ktype
= {
2533 .default_attrs
= ext4_attrs
,
2534 .sysfs_ops
= &ext4_attr_ops
,
2535 .release
= ext4_sb_release
,
2538 static void ext4_feat_release(struct kobject
*kobj
)
2540 complete(&ext4_feat
->f_kobj_unregister
);
2543 static struct kobj_type ext4_feat_ktype
= {
2544 .default_attrs
= ext4_feat_attrs
,
2545 .sysfs_ops
= &ext4_attr_ops
,
2546 .release
= ext4_feat_release
,
2550 * Check whether this filesystem can be mounted based on
2551 * the features present and the RDONLY/RDWR mount requested.
2552 * Returns 1 if this filesystem can be mounted as requested,
2553 * 0 if it cannot be.
2555 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2557 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2558 ext4_msg(sb
, KERN_ERR
,
2559 "Couldn't mount because of "
2560 "unsupported optional features (%x)",
2561 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2562 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2569 /* Check that feature set is OK for a read-write mount */
2570 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2571 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2572 "unsupported optional features (%x)",
2573 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2574 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2578 * Large file size enabled file system can only be mounted
2579 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2581 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2582 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2583 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2584 "cannot be mounted RDWR without "
2593 * This function is called once a day if we have errors logged
2594 * on the file system
2596 static void print_daily_error_info(unsigned long arg
)
2598 struct super_block
*sb
= (struct super_block
*) arg
;
2599 struct ext4_sb_info
*sbi
;
2600 struct ext4_super_block
*es
;
2605 if (es
->s_error_count
)
2606 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2607 le32_to_cpu(es
->s_error_count
));
2608 if (es
->s_first_error_time
) {
2609 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2610 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2611 (int) sizeof(es
->s_first_error_func
),
2612 es
->s_first_error_func
,
2613 le32_to_cpu(es
->s_first_error_line
));
2614 if (es
->s_first_error_ino
)
2615 printk(": inode %u",
2616 le32_to_cpu(es
->s_first_error_ino
));
2617 if (es
->s_first_error_block
)
2618 printk(": block %llu", (unsigned long long)
2619 le64_to_cpu(es
->s_first_error_block
));
2622 if (es
->s_last_error_time
) {
2623 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2624 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2625 (int) sizeof(es
->s_last_error_func
),
2626 es
->s_last_error_func
,
2627 le32_to_cpu(es
->s_last_error_line
));
2628 if (es
->s_last_error_ino
)
2629 printk(": inode %u",
2630 le32_to_cpu(es
->s_last_error_ino
));
2631 if (es
->s_last_error_block
)
2632 printk(": block %llu", (unsigned long long)
2633 le64_to_cpu(es
->s_last_error_block
));
2636 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2639 static void ext4_lazyinode_timeout(unsigned long data
)
2641 struct task_struct
*p
= (struct task_struct
*)data
;
2645 /* Find next suitable group and run ext4_init_inode_table */
2646 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2648 struct ext4_group_desc
*gdp
= NULL
;
2649 ext4_group_t group
, ngroups
;
2650 struct super_block
*sb
;
2651 unsigned long timeout
= 0;
2655 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2657 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2658 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2664 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2668 if (group
== ngroups
)
2673 ret
= ext4_init_inode_table(sb
, group
,
2674 elr
->lr_timeout
? 0 : 1);
2675 if (elr
->lr_timeout
== 0) {
2676 timeout
= jiffies
- timeout
;
2677 if (elr
->lr_sbi
->s_li_wait_mult
)
2678 timeout
*= elr
->lr_sbi
->s_li_wait_mult
;
2681 elr
->lr_timeout
= timeout
;
2683 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2684 elr
->lr_next_group
= group
+ 1;
2691 * Remove lr_request from the list_request and free the
2692 * request tructure. Should be called with li_list_mtx held
2694 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2696 struct ext4_sb_info
*sbi
;
2703 list_del(&elr
->lr_request
);
2704 sbi
->s_li_request
= NULL
;
2708 static void ext4_unregister_li_request(struct super_block
*sb
)
2710 struct ext4_li_request
*elr
= EXT4_SB(sb
)->s_li_request
;
2715 mutex_lock(&ext4_li_info
->li_list_mtx
);
2716 ext4_remove_li_request(elr
);
2717 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2720 static struct task_struct
*ext4_lazyinit_task
;
2723 * This is the function where ext4lazyinit thread lives. It walks
2724 * through the request list searching for next scheduled filesystem.
2725 * When such a fs is found, run the lazy initialization request
2726 * (ext4_rn_li_request) and keep track of the time spend in this
2727 * function. Based on that time we compute next schedule time of
2728 * the request. When walking through the list is complete, compute
2729 * next waking time and put itself into sleep.
2731 static int ext4_lazyinit_thread(void *arg
)
2733 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2734 struct list_head
*pos
, *n
;
2735 struct ext4_li_request
*elr
;
2736 unsigned long next_wakeup
;
2739 BUG_ON(NULL
== eli
);
2741 eli
->li_timer
.data
= (unsigned long)current
;
2742 eli
->li_timer
.function
= ext4_lazyinode_timeout
;
2744 eli
->li_task
= current
;
2745 wake_up(&eli
->li_wait_task
);
2749 next_wakeup
= MAX_JIFFY_OFFSET
;
2751 mutex_lock(&eli
->li_list_mtx
);
2752 if (list_empty(&eli
->li_request_list
)) {
2753 mutex_unlock(&eli
->li_list_mtx
);
2757 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2758 elr
= list_entry(pos
, struct ext4_li_request
,
2761 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2762 if (ext4_run_li_request(elr
) != 0) {
2763 /* error, remove the lazy_init job */
2764 ext4_remove_li_request(elr
);
2769 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2770 next_wakeup
= elr
->lr_next_sched
;
2772 mutex_unlock(&eli
->li_list_mtx
);
2774 if (freezing(current
))
2777 if ((time_after_eq(jiffies
, next_wakeup
)) ||
2778 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2783 eli
->li_timer
.expires
= next_wakeup
;
2784 add_timer(&eli
->li_timer
);
2785 prepare_to_wait(&eli
->li_wait_daemon
, &wait
,
2786 TASK_INTERRUPTIBLE
);
2787 if (time_before(jiffies
, next_wakeup
))
2789 finish_wait(&eli
->li_wait_daemon
, &wait
);
2790 if (kthread_should_stop()) {
2791 ext4_clear_request_list();
2798 * It looks like the request list is empty, but we need
2799 * to check it under the li_list_mtx lock, to prevent any
2800 * additions into it, and of course we should lock ext4_li_mtx
2801 * to atomically free the list and ext4_li_info, because at
2802 * this point another ext4 filesystem could be registering
2805 mutex_lock(&ext4_li_mtx
);
2806 mutex_lock(&eli
->li_list_mtx
);
2807 if (!list_empty(&eli
->li_request_list
)) {
2808 mutex_unlock(&eli
->li_list_mtx
);
2809 mutex_unlock(&ext4_li_mtx
);
2812 mutex_unlock(&eli
->li_list_mtx
);
2813 del_timer_sync(&ext4_li_info
->li_timer
);
2814 eli
->li_task
= NULL
;
2815 wake_up(&eli
->li_wait_task
);
2817 kfree(ext4_li_info
);
2818 ext4_lazyinit_task
= NULL
;
2819 ext4_li_info
= NULL
;
2820 mutex_unlock(&ext4_li_mtx
);
2825 static void ext4_clear_request_list(void)
2827 struct list_head
*pos
, *n
;
2828 struct ext4_li_request
*elr
;
2830 mutex_lock(&ext4_li_info
->li_list_mtx
);
2831 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2832 elr
= list_entry(pos
, struct ext4_li_request
,
2834 ext4_remove_li_request(elr
);
2836 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2839 static int ext4_run_lazyinit_thread(void)
2841 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2842 ext4_li_info
, "ext4lazyinit");
2843 if (IS_ERR(ext4_lazyinit_task
)) {
2844 int err
= PTR_ERR(ext4_lazyinit_task
);
2845 ext4_clear_request_list();
2846 del_timer_sync(&ext4_li_info
->li_timer
);
2847 kfree(ext4_li_info
);
2848 ext4_li_info
= NULL
;
2849 printk(KERN_CRIT
"EXT4: error %d creating inode table "
2850 "initialization thread\n",
2854 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2856 wait_event(ext4_li_info
->li_wait_task
, ext4_li_info
->li_task
!= NULL
);
2861 * Check whether it make sense to run itable init. thread or not.
2862 * If there is at least one uninitialized inode table, return
2863 * corresponding group number, else the loop goes through all
2864 * groups and return total number of groups.
2866 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2868 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2869 struct ext4_group_desc
*gdp
= NULL
;
2871 for (group
= 0; group
< ngroups
; group
++) {
2872 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2876 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2883 static int ext4_li_info_new(void)
2885 struct ext4_lazy_init
*eli
= NULL
;
2887 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2891 eli
->li_task
= NULL
;
2892 INIT_LIST_HEAD(&eli
->li_request_list
);
2893 mutex_init(&eli
->li_list_mtx
);
2895 init_waitqueue_head(&eli
->li_wait_daemon
);
2896 init_waitqueue_head(&eli
->li_wait_task
);
2897 init_timer(&eli
->li_timer
);
2898 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2905 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2908 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2909 struct ext4_li_request
*elr
;
2912 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2918 elr
->lr_next_group
= start
;
2921 * Randomize first schedule time of the request to
2922 * spread the inode table initialization requests
2925 get_random_bytes(&rnd
, sizeof(rnd
));
2926 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2927 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
2932 static int ext4_register_li_request(struct super_block
*sb
,
2933 ext4_group_t first_not_zeroed
)
2935 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2936 struct ext4_li_request
*elr
;
2937 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
2940 if (sbi
->s_li_request
!= NULL
)
2943 if (first_not_zeroed
== ngroups
||
2944 (sb
->s_flags
& MS_RDONLY
) ||
2945 !test_opt(sb
, INIT_INODE_TABLE
)) {
2946 sbi
->s_li_request
= NULL
;
2950 if (first_not_zeroed
== ngroups
) {
2951 sbi
->s_li_request
= NULL
;
2955 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
2959 mutex_lock(&ext4_li_mtx
);
2961 if (NULL
== ext4_li_info
) {
2962 ret
= ext4_li_info_new();
2967 mutex_lock(&ext4_li_info
->li_list_mtx
);
2968 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
2969 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2971 sbi
->s_li_request
= elr
;
2973 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
2974 ret
= ext4_run_lazyinit_thread();
2979 mutex_unlock(&ext4_li_mtx
);
2986 * We do not need to lock anything since this is called on
2989 static void ext4_destroy_lazyinit_thread(void)
2992 * If thread exited earlier
2993 * there's nothing to be done.
2995 if (!ext4_li_info
|| !ext4_lazyinit_task
)
2998 kthread_stop(ext4_lazyinit_task
);
3001 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3002 __releases(kernel_lock
)
3003 __acquires(kernel_lock
)
3005 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3006 struct buffer_head
*bh
;
3007 struct ext4_super_block
*es
= NULL
;
3008 struct ext4_sb_info
*sbi
;
3010 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3011 ext4_fsblk_t logical_sb_block
;
3012 unsigned long offset
= 0;
3013 unsigned long journal_devnum
= 0;
3014 unsigned long def_mount_opts
;
3020 unsigned int db_count
;
3022 int needs_recovery
, has_huge_files
;
3025 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3026 ext4_group_t first_not_zeroed
;
3028 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3032 sbi
->s_blockgroup_lock
=
3033 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3034 if (!sbi
->s_blockgroup_lock
) {
3038 sb
->s_fs_info
= sbi
;
3039 sbi
->s_mount_opt
= 0;
3040 sbi
->s_resuid
= EXT4_DEF_RESUID
;
3041 sbi
->s_resgid
= EXT4_DEF_RESGID
;
3042 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3043 sbi
->s_sb_block
= sb_block
;
3044 if (sb
->s_bdev
->bd_part
)
3045 sbi
->s_sectors_written_start
=
3046 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3048 /* Cleanup superblock name */
3049 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3053 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3055 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3060 * The ext4 superblock will not be buffer aligned for other than 1kB
3061 * block sizes. We need to calculate the offset from buffer start.
3063 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3064 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3065 offset
= do_div(logical_sb_block
, blocksize
);
3067 logical_sb_block
= sb_block
;
3070 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3071 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3075 * Note: s_es must be initialized as soon as possible because
3076 * some ext4 macro-instructions depend on its value
3078 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3080 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3081 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3083 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3085 /* Set defaults before we parse the mount options */
3086 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3087 set_opt(sb
, INIT_INODE_TABLE
);
3088 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3090 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
3091 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
3095 if (def_mount_opts
& EXT4_DEFM_UID16
)
3096 set_opt(sb
, NO_UID32
);
3097 #ifdef CONFIG_EXT4_FS_XATTR
3098 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
3099 set_opt(sb
, XATTR_USER
);
3101 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3102 if (def_mount_opts
& EXT4_DEFM_ACL
)
3103 set_opt(sb
, POSIX_ACL
);
3105 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3106 set_opt(sb
, JOURNAL_DATA
);
3107 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3108 set_opt(sb
, ORDERED_DATA
);
3109 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3110 set_opt(sb
, WRITEBACK_DATA
);
3112 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3113 set_opt(sb
, ERRORS_PANIC
);
3114 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3115 set_opt(sb
, ERRORS_CONT
);
3117 set_opt(sb
, ERRORS_RO
);
3118 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3119 set_opt(sb
, BLOCK_VALIDITY
);
3120 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3121 set_opt(sb
, DISCARD
);
3123 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
3124 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
3125 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3126 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3127 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3129 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3130 set_opt(sb
, BARRIER
);
3133 * enable delayed allocation by default
3134 * Use -o nodelalloc to turn it off
3136 if (!IS_EXT3_SB(sb
) &&
3137 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3138 set_opt(sb
, DELALLOC
);
3140 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3141 &journal_devnum
, &journal_ioprio
, NULL
, 0)) {
3142 ext4_msg(sb
, KERN_WARNING
,
3143 "failed to parse options in superblock: %s",
3144 sbi
->s_es
->s_mount_opts
);
3146 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3147 &journal_ioprio
, NULL
, 0))
3150 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3151 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3153 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3154 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3155 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3156 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3157 ext4_msg(sb
, KERN_WARNING
,
3158 "feature flags set on rev 0 fs, "
3159 "running e2fsck is recommended");
3162 * Check feature flags regardless of the revision level, since we
3163 * previously didn't change the revision level when setting the flags,
3164 * so there is a chance incompat flags are set on a rev 0 filesystem.
3166 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3169 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3171 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3172 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3173 ext4_msg(sb
, KERN_ERR
,
3174 "Unsupported filesystem blocksize %d", blocksize
);
3178 if (sb
->s_blocksize
!= blocksize
) {
3179 /* Validate the filesystem blocksize */
3180 if (!sb_set_blocksize(sb
, blocksize
)) {
3181 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3187 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3188 offset
= do_div(logical_sb_block
, blocksize
);
3189 bh
= sb_bread(sb
, logical_sb_block
);
3191 ext4_msg(sb
, KERN_ERR
,
3192 "Can't read superblock on 2nd try");
3195 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
3197 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3198 ext4_msg(sb
, KERN_ERR
,
3199 "Magic mismatch, very weird!");
3204 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3205 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3206 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3208 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3210 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3211 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3212 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3214 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3215 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3216 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3217 (!is_power_of_2(sbi
->s_inode_size
)) ||
3218 (sbi
->s_inode_size
> blocksize
)) {
3219 ext4_msg(sb
, KERN_ERR
,
3220 "unsupported inode size: %d",
3224 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3225 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3228 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3229 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3230 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3231 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3232 !is_power_of_2(sbi
->s_desc_size
)) {
3233 ext4_msg(sb
, KERN_ERR
,
3234 "unsupported descriptor size %lu",
3239 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3241 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3242 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3243 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3246 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3247 if (sbi
->s_inodes_per_block
== 0)
3249 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3250 sbi
->s_inodes_per_block
;
3251 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3253 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3254 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3255 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3257 for (i
= 0; i
< 4; i
++)
3258 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3259 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3260 i
= le32_to_cpu(es
->s_flags
);
3261 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3262 sbi
->s_hash_unsigned
= 3;
3263 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3264 #ifdef __CHAR_UNSIGNED__
3265 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3266 sbi
->s_hash_unsigned
= 3;
3268 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3273 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3274 ext4_msg(sb
, KERN_ERR
,
3275 "#blocks per group too big: %lu",
3276 sbi
->s_blocks_per_group
);
3279 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3280 ext4_msg(sb
, KERN_ERR
,
3281 "#inodes per group too big: %lu",
3282 sbi
->s_inodes_per_group
);
3287 * Test whether we have more sectors than will fit in sector_t,
3288 * and whether the max offset is addressable by the page cache.
3290 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3291 ext4_blocks_count(es
));
3293 ext4_msg(sb
, KERN_ERR
, "filesystem"
3294 " too large to mount safely on this system");
3295 if (sizeof(sector_t
) < 8)
3296 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3301 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3304 /* check blocks count against device size */
3305 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3306 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3307 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3308 "exceeds size of device (%llu blocks)",
3309 ext4_blocks_count(es
), blocks_count
);
3314 * It makes no sense for the first data block to be beyond the end
3315 * of the filesystem.
3317 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3318 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
3319 "block %u is beyond end of filesystem (%llu)",
3320 le32_to_cpu(es
->s_first_data_block
),
3321 ext4_blocks_count(es
));
3324 blocks_count
= (ext4_blocks_count(es
) -
3325 le32_to_cpu(es
->s_first_data_block
) +
3326 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3327 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3328 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3329 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3330 "(block count %llu, first data block %u, "
3331 "blocks per group %lu)", sbi
->s_groups_count
,
3332 ext4_blocks_count(es
),
3333 le32_to_cpu(es
->s_first_data_block
),
3334 EXT4_BLOCKS_PER_GROUP(sb
));
3337 sbi
->s_groups_count
= blocks_count
;
3338 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3339 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3340 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3341 EXT4_DESC_PER_BLOCK(sb
);
3342 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
3344 if (sbi
->s_group_desc
== NULL
) {
3345 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3349 #ifdef CONFIG_PROC_FS
3351 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3354 bgl_lock_init(sbi
->s_blockgroup_lock
);
3356 for (i
= 0; i
< db_count
; i
++) {
3357 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3358 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3359 if (!sbi
->s_group_desc
[i
]) {
3360 ext4_msg(sb
, KERN_ERR
,
3361 "can't read group descriptor %d", i
);
3366 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3367 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3370 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3371 if (!ext4_fill_flex_info(sb
)) {
3372 ext4_msg(sb
, KERN_ERR
,
3373 "unable to initialize "
3374 "flex_bg meta info!");
3378 sbi
->s_gdb_count
= db_count
;
3379 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3380 spin_lock_init(&sbi
->s_next_gen_lock
);
3382 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
3383 ext4_count_free_blocks(sb
));
3385 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3386 ext4_count_free_inodes(sb
));
3389 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3390 ext4_count_dirs(sb
));
3393 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
3396 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3400 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3401 sbi
->s_max_writeback_mb_bump
= 128;
3404 * set up enough so that it can read an inode
3406 if (!test_opt(sb
, NOLOAD
) &&
3407 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3408 sb
->s_op
= &ext4_sops
;
3410 sb
->s_op
= &ext4_nojournal_sops
;
3411 sb
->s_export_op
= &ext4_export_ops
;
3412 sb
->s_xattr
= ext4_xattr_handlers
;
3414 sb
->s_qcop
= &ext4_qctl_operations
;
3415 sb
->dq_op
= &ext4_quota_operations
;
3417 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3418 mutex_init(&sbi
->s_orphan_lock
);
3419 mutex_init(&sbi
->s_resize_lock
);
3423 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3424 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3425 EXT4_FEATURE_INCOMPAT_RECOVER
));
3428 * The first inode we look at is the journal inode. Don't try
3429 * root first: it may be modified in the journal!
3431 if (!test_opt(sb
, NOLOAD
) &&
3432 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3433 if (ext4_load_journal(sb
, es
, journal_devnum
))
3435 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3436 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3437 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3438 "suppressed and not mounted read-only");
3439 goto failed_mount_wq
;
3441 clear_opt(sb
, DATA_FLAGS
);
3442 set_opt(sb
, WRITEBACK_DATA
);
3443 sbi
->s_journal
= NULL
;
3448 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
3449 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3450 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3451 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3452 goto failed_mount_wq
;
3455 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3456 jbd2_journal_set_features(sbi
->s_journal
,
3457 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3458 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3459 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3460 jbd2_journal_set_features(sbi
->s_journal
,
3461 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
3462 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3463 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3465 jbd2_journal_clear_features(sbi
->s_journal
,
3466 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3467 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3470 /* We have now updated the journal if required, so we can
3471 * validate the data journaling mode. */
3472 switch (test_opt(sb
, DATA_FLAGS
)) {
3474 /* No mode set, assume a default based on the journal
3475 * capabilities: ORDERED_DATA if the journal can
3476 * cope, else JOURNAL_DATA
3478 if (jbd2_journal_check_available_features
3479 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3480 set_opt(sb
, ORDERED_DATA
);
3482 set_opt(sb
, JOURNAL_DATA
);
3485 case EXT4_MOUNT_ORDERED_DATA
:
3486 case EXT4_MOUNT_WRITEBACK_DATA
:
3487 if (!jbd2_journal_check_available_features
3488 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3489 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3490 "requested data journaling mode");
3491 goto failed_mount_wq
;
3496 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3499 * The journal may have updated the bg summary counts, so we
3500 * need to update the global counters.
3502 percpu_counter_set(&sbi
->s_freeblocks_counter
,
3503 ext4_count_free_blocks(sb
));
3504 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3505 ext4_count_free_inodes(sb
));
3506 percpu_counter_set(&sbi
->s_dirs_counter
,
3507 ext4_count_dirs(sb
));
3508 percpu_counter_set(&sbi
->s_dirtyblocks_counter
, 0);
3511 EXT4_SB(sb
)->dio_unwritten_wq
= create_workqueue("ext4-dio-unwritten");
3512 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3513 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3514 goto failed_mount_wq
;
3518 * The jbd2_journal_load will have done any necessary log recovery,
3519 * so we can safely mount the rest of the filesystem now.
3522 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3524 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3525 ret
= PTR_ERR(root
);
3528 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3530 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3533 sb
->s_root
= d_alloc_root(root
);
3535 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3541 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
3543 /* determine the minimum size of new large inodes, if present */
3544 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3545 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3546 EXT4_GOOD_OLD_INODE_SIZE
;
3547 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3548 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3549 if (sbi
->s_want_extra_isize
<
3550 le16_to_cpu(es
->s_want_extra_isize
))
3551 sbi
->s_want_extra_isize
=
3552 le16_to_cpu(es
->s_want_extra_isize
);
3553 if (sbi
->s_want_extra_isize
<
3554 le16_to_cpu(es
->s_min_extra_isize
))
3555 sbi
->s_want_extra_isize
=
3556 le16_to_cpu(es
->s_min_extra_isize
);
3559 /* Check if enough inode space is available */
3560 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3561 sbi
->s_inode_size
) {
3562 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3563 EXT4_GOOD_OLD_INODE_SIZE
;
3564 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3568 if (test_opt(sb
, DELALLOC
) &&
3569 (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)) {
3570 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
3571 "requested data journaling mode");
3572 clear_opt(sb
, DELALLOC
);
3574 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3575 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3576 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3577 "option - requested data journaling mode");
3578 clear_opt(sb
, DIOREAD_NOLOCK
);
3580 if (sb
->s_blocksize
< PAGE_SIZE
) {
3581 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3582 "option - block size is too small");
3583 clear_opt(sb
, DIOREAD_NOLOCK
);
3587 err
= ext4_setup_system_zone(sb
);
3589 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3595 err
= ext4_mb_init(sb
, needs_recovery
);
3597 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3602 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3606 sbi
->s_kobj
.kset
= ext4_kset
;
3607 init_completion(&sbi
->s_kobj_unregister
);
3608 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3611 ext4_mb_release(sb
);
3612 ext4_ext_release(sb
);
3616 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3617 ext4_orphan_cleanup(sb
, es
);
3618 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3619 if (needs_recovery
) {
3620 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3621 ext4_mark_recovery_complete(sb
, es
);
3623 if (EXT4_SB(sb
)->s_journal
) {
3624 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3625 descr
= " journalled data mode";
3626 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3627 descr
= " ordered data mode";
3629 descr
= " writeback data mode";
3631 descr
= "out journal";
3633 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3634 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3635 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3637 init_timer(&sbi
->s_err_report
);
3638 sbi
->s_err_report
.function
= print_daily_error_info
;
3639 sbi
->s_err_report
.data
= (unsigned long) sb
;
3640 if (es
->s_error_count
)
3641 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3648 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3652 ext4_msg(sb
, KERN_ERR
, "mount failed");
3653 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3655 ext4_release_system_zone(sb
);
3656 if (sbi
->s_journal
) {
3657 jbd2_journal_destroy(sbi
->s_journal
);
3658 sbi
->s_journal
= NULL
;
3661 if (sbi
->s_flex_groups
) {
3662 if (is_vmalloc_addr(sbi
->s_flex_groups
))
3663 vfree(sbi
->s_flex_groups
);
3665 kfree(sbi
->s_flex_groups
);
3667 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
3668 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3669 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3670 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
3672 for (i
= 0; i
< db_count
; i
++)
3673 brelse(sbi
->s_group_desc
[i
]);
3674 kfree(sbi
->s_group_desc
);
3677 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3680 for (i
= 0; i
< MAXQUOTAS
; i
++)
3681 kfree(sbi
->s_qf_names
[i
]);
3683 ext4_blkdev_remove(sbi
);
3686 sb
->s_fs_info
= NULL
;
3687 kfree(sbi
->s_blockgroup_lock
);
3695 * Setup any per-fs journal parameters now. We'll do this both on
3696 * initial mount, once the journal has been initialised but before we've
3697 * done any recovery; and again on any subsequent remount.
3699 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3701 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3703 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3704 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3705 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3707 write_lock(&journal
->j_state_lock
);
3708 if (test_opt(sb
, BARRIER
))
3709 journal
->j_flags
|= JBD2_BARRIER
;
3711 journal
->j_flags
&= ~JBD2_BARRIER
;
3712 if (test_opt(sb
, DATA_ERR_ABORT
))
3713 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3715 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3716 write_unlock(&journal
->j_state_lock
);
3719 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3720 unsigned int journal_inum
)
3722 struct inode
*journal_inode
;
3725 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3727 /* First, test for the existence of a valid inode on disk. Bad
3728 * things happen if we iget() an unused inode, as the subsequent
3729 * iput() will try to delete it. */
3731 journal_inode
= ext4_iget(sb
, journal_inum
);
3732 if (IS_ERR(journal_inode
)) {
3733 ext4_msg(sb
, KERN_ERR
, "no journal found");
3736 if (!journal_inode
->i_nlink
) {
3737 make_bad_inode(journal_inode
);
3738 iput(journal_inode
);
3739 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3743 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3744 journal_inode
, journal_inode
->i_size
);
3745 if (!S_ISREG(journal_inode
->i_mode
)) {
3746 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3747 iput(journal_inode
);
3751 journal
= jbd2_journal_init_inode(journal_inode
);
3753 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3754 iput(journal_inode
);
3757 journal
->j_private
= sb
;
3758 ext4_init_journal_params(sb
, journal
);
3762 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3765 struct buffer_head
*bh
;
3769 int hblock
, blocksize
;
3770 ext4_fsblk_t sb_block
;
3771 unsigned long offset
;
3772 struct ext4_super_block
*es
;
3773 struct block_device
*bdev
;
3775 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3777 bdev
= ext4_blkdev_get(j_dev
, sb
);
3781 blocksize
= sb
->s_blocksize
;
3782 hblock
= bdev_logical_block_size(bdev
);
3783 if (blocksize
< hblock
) {
3784 ext4_msg(sb
, KERN_ERR
,
3785 "blocksize too small for journal device");
3789 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3790 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3791 set_blocksize(bdev
, blocksize
);
3792 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3793 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3794 "external journal");
3798 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3799 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3800 !(le32_to_cpu(es
->s_feature_incompat
) &
3801 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3802 ext4_msg(sb
, KERN_ERR
, "external journal has "
3808 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3809 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3814 len
= ext4_blocks_count(es
);
3815 start
= sb_block
+ 1;
3816 brelse(bh
); /* we're done with the superblock */
3818 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3819 start
, len
, blocksize
);
3821 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3824 journal
->j_private
= sb
;
3825 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3826 wait_on_buffer(journal
->j_sb_buffer
);
3827 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3828 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3831 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3832 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3833 "user (unsupported) - %d",
3834 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3837 EXT4_SB(sb
)->journal_bdev
= bdev
;
3838 ext4_init_journal_params(sb
, journal
);
3842 jbd2_journal_destroy(journal
);
3844 ext4_blkdev_put(bdev
);
3848 static int ext4_load_journal(struct super_block
*sb
,
3849 struct ext4_super_block
*es
,
3850 unsigned long journal_devnum
)
3853 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3856 int really_read_only
;
3858 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3860 if (journal_devnum
&&
3861 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3862 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3863 "numbers have changed");
3864 journal_dev
= new_decode_dev(journal_devnum
);
3866 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3868 really_read_only
= bdev_read_only(sb
->s_bdev
);
3871 * Are we loading a blank journal or performing recovery after a
3872 * crash? For recovery, we need to check in advance whether we
3873 * can get read-write access to the device.
3875 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3876 if (sb
->s_flags
& MS_RDONLY
) {
3877 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3878 "required on readonly filesystem");
3879 if (really_read_only
) {
3880 ext4_msg(sb
, KERN_ERR
, "write access "
3881 "unavailable, cannot proceed");
3884 ext4_msg(sb
, KERN_INFO
, "write access will "
3885 "be enabled during recovery");
3889 if (journal_inum
&& journal_dev
) {
3890 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3891 "and inode journals!");
3896 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3899 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3903 if (!(journal
->j_flags
& JBD2_BARRIER
))
3904 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3906 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
3907 err
= jbd2_journal_update_format(journal
);
3909 ext4_msg(sb
, KERN_ERR
, "error updating journal");
3910 jbd2_journal_destroy(journal
);
3915 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
3916 err
= jbd2_journal_wipe(journal
, !really_read_only
);
3918 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
3920 memcpy(save
, ((char *) es
) +
3921 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
3922 err
= jbd2_journal_load(journal
);
3924 memcpy(((char *) es
) + EXT4_S_ERR_START
,
3925 save
, EXT4_S_ERR_LEN
);
3930 ext4_msg(sb
, KERN_ERR
, "error loading journal");
3931 jbd2_journal_destroy(journal
);
3935 EXT4_SB(sb
)->s_journal
= journal
;
3936 ext4_clear_journal_err(sb
, es
);
3938 if (!really_read_only
&& journal_devnum
&&
3939 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3940 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
3942 /* Make sure we flush the recovery flag to disk. */
3943 ext4_commit_super(sb
, 1);
3949 static int ext4_commit_super(struct super_block
*sb
, int sync
)
3951 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
3952 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
3957 if (buffer_write_io_error(sbh
)) {
3959 * Oh, dear. A previous attempt to write the
3960 * superblock failed. This could happen because the
3961 * USB device was yanked out. Or it could happen to
3962 * be a transient write error and maybe the block will
3963 * be remapped. Nothing we can do but to retry the
3964 * write and hope for the best.
3966 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
3967 "superblock detected");
3968 clear_buffer_write_io_error(sbh
);
3969 set_buffer_uptodate(sbh
);
3972 * If the file system is mounted read-only, don't update the
3973 * superblock write time. This avoids updating the superblock
3974 * write time when we are mounting the root file system
3975 * read/only but we need to replay the journal; at that point,
3976 * for people who are east of GMT and who make their clock
3977 * tick in localtime for Windows bug-for-bug compatibility,
3978 * the clock is set in the future, and this will cause e2fsck
3979 * to complain and force a full file system check.
3981 if (!(sb
->s_flags
& MS_RDONLY
))
3982 es
->s_wtime
= cpu_to_le32(get_seconds());
3983 if (sb
->s_bdev
->bd_part
)
3984 es
->s_kbytes_written
=
3985 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
3986 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
3987 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
3989 es
->s_kbytes_written
=
3990 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
3991 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
3992 &EXT4_SB(sb
)->s_freeblocks_counter
));
3993 es
->s_free_inodes_count
=
3994 cpu_to_le32(percpu_counter_sum_positive(
3995 &EXT4_SB(sb
)->s_freeinodes_counter
));
3997 BUFFER_TRACE(sbh
, "marking dirty");
3998 mark_buffer_dirty(sbh
);
4000 error
= sync_dirty_buffer(sbh
);
4004 error
= buffer_write_io_error(sbh
);
4006 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4008 clear_buffer_write_io_error(sbh
);
4009 set_buffer_uptodate(sbh
);
4016 * Have we just finished recovery? If so, and if we are mounting (or
4017 * remounting) the filesystem readonly, then we will end up with a
4018 * consistent fs on disk. Record that fact.
4020 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4021 struct ext4_super_block
*es
)
4023 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4025 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4026 BUG_ON(journal
!= NULL
);
4029 jbd2_journal_lock_updates(journal
);
4030 if (jbd2_journal_flush(journal
) < 0)
4033 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4034 sb
->s_flags
& MS_RDONLY
) {
4035 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4036 ext4_commit_super(sb
, 1);
4040 jbd2_journal_unlock_updates(journal
);
4044 * If we are mounting (or read-write remounting) a filesystem whose journal
4045 * has recorded an error from a previous lifetime, move that error to the
4046 * main filesystem now.
4048 static void ext4_clear_journal_err(struct super_block
*sb
,
4049 struct ext4_super_block
*es
)
4055 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4057 journal
= EXT4_SB(sb
)->s_journal
;
4060 * Now check for any error status which may have been recorded in the
4061 * journal by a prior ext4_error() or ext4_abort()
4064 j_errno
= jbd2_journal_errno(journal
);
4068 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4069 ext4_warning(sb
, "Filesystem error recorded "
4070 "from previous mount: %s", errstr
);
4071 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4073 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4074 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4075 ext4_commit_super(sb
, 1);
4077 jbd2_journal_clear_err(journal
);
4082 * Force the running and committing transactions to commit,
4083 * and wait on the commit.
4085 int ext4_force_commit(struct super_block
*sb
)
4090 if (sb
->s_flags
& MS_RDONLY
)
4093 journal
= EXT4_SB(sb
)->s_journal
;
4095 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
4096 ret
= ext4_journal_force_commit(journal
);
4102 static void ext4_write_super(struct super_block
*sb
)
4105 ext4_commit_super(sb
, 1);
4109 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4113 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4115 trace_ext4_sync_fs(sb
, wait
);
4116 flush_workqueue(sbi
->dio_unwritten_wq
);
4117 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4119 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4125 * LVM calls this function before a (read-only) snapshot is created. This
4126 * gives us a chance to flush the journal completely and mark the fs clean.
4128 static int ext4_freeze(struct super_block
*sb
)
4133 if (sb
->s_flags
& MS_RDONLY
)
4136 journal
= EXT4_SB(sb
)->s_journal
;
4138 /* Now we set up the journal barrier. */
4139 jbd2_journal_lock_updates(journal
);
4142 * Don't clear the needs_recovery flag if we failed to flush
4145 error
= jbd2_journal_flush(journal
);
4149 /* Journal blocked and flushed, clear needs_recovery flag. */
4150 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4151 error
= ext4_commit_super(sb
, 1);
4153 /* we rely on s_frozen to stop further updates */
4154 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4159 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4160 * flag here, even though the filesystem is not technically dirty yet.
4162 static int ext4_unfreeze(struct super_block
*sb
)
4164 if (sb
->s_flags
& MS_RDONLY
)
4168 /* Reset the needs_recovery flag before the fs is unlocked. */
4169 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4170 ext4_commit_super(sb
, 1);
4176 * Structure to save mount options for ext4_remount's benefit
4178 struct ext4_mount_options
{
4179 unsigned long s_mount_opt
;
4180 unsigned long s_mount_opt2
;
4183 unsigned long s_commit_interval
;
4184 u32 s_min_batch_time
, s_max_batch_time
;
4187 char *s_qf_names
[MAXQUOTAS
];
4191 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4193 struct ext4_super_block
*es
;
4194 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4195 ext4_fsblk_t n_blocks_count
= 0;
4196 unsigned long old_sb_flags
;
4197 struct ext4_mount_options old_opts
;
4198 int enable_quota
= 0;
4200 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4205 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4207 /* Store the original options */
4209 old_sb_flags
= sb
->s_flags
;
4210 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4211 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4212 old_opts
.s_resuid
= sbi
->s_resuid
;
4213 old_opts
.s_resgid
= sbi
->s_resgid
;
4214 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4215 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4216 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4218 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4219 for (i
= 0; i
< MAXQUOTAS
; i
++)
4220 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
4222 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4223 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4226 * Allow the "check" option to be passed as a remount option.
4228 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
4229 &n_blocks_count
, 1)) {
4234 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4235 ext4_abort(sb
, "Abort forced by user");
4237 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4238 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4242 if (sbi
->s_journal
) {
4243 ext4_init_journal_params(sb
, sbi
->s_journal
);
4244 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4247 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
4248 n_blocks_count
> ext4_blocks_count(es
)) {
4249 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4254 if (*flags
& MS_RDONLY
) {
4255 err
= dquot_suspend(sb
, -1);
4260 * First of all, the unconditional stuff we have to do
4261 * to disable replay of the journal when we next remount
4263 sb
->s_flags
|= MS_RDONLY
;
4266 * OK, test if we are remounting a valid rw partition
4267 * readonly, and if so set the rdonly flag and then
4268 * mark the partition as valid again.
4270 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4271 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4272 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4275 ext4_mark_recovery_complete(sb
, es
);
4277 /* Make sure we can mount this feature set readwrite */
4278 if (!ext4_feature_set_ok(sb
, 0)) {
4283 * Make sure the group descriptor checksums
4284 * are sane. If they aren't, refuse to remount r/w.
4286 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4287 struct ext4_group_desc
*gdp
=
4288 ext4_get_group_desc(sb
, g
, NULL
);
4290 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
4291 ext4_msg(sb
, KERN_ERR
,
4292 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4293 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4294 le16_to_cpu(gdp
->bg_checksum
));
4301 * If we have an unprocessed orphan list hanging
4302 * around from a previously readonly bdev mount,
4303 * require a full umount/remount for now.
4305 if (es
->s_last_orphan
) {
4306 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4307 "remount RDWR because of unprocessed "
4308 "orphan inode list. Please "
4309 "umount/remount instead");
4315 * Mounting a RDONLY partition read-write, so reread
4316 * and store the current valid flag. (It may have
4317 * been changed by e2fsck since we originally mounted
4321 ext4_clear_journal_err(sb
, es
);
4322 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4323 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
4325 if (!ext4_setup_super(sb
, es
, 0))
4326 sb
->s_flags
&= ~MS_RDONLY
;
4332 * Reinitialize lazy itable initialization thread based on
4335 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4336 ext4_unregister_li_request(sb
);
4338 ext4_group_t first_not_zeroed
;
4339 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4340 ext4_register_li_request(sb
, first_not_zeroed
);
4343 ext4_setup_system_zone(sb
);
4344 if (sbi
->s_journal
== NULL
)
4345 ext4_commit_super(sb
, 1);
4348 /* Release old quota file names */
4349 for (i
= 0; i
< MAXQUOTAS
; i
++)
4350 if (old_opts
.s_qf_names
[i
] &&
4351 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4352 kfree(old_opts
.s_qf_names
[i
]);
4356 dquot_resume(sb
, -1);
4358 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4363 sb
->s_flags
= old_sb_flags
;
4364 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4365 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4366 sbi
->s_resuid
= old_opts
.s_resuid
;
4367 sbi
->s_resgid
= old_opts
.s_resgid
;
4368 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4369 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4370 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4372 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4373 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4374 if (sbi
->s_qf_names
[i
] &&
4375 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4376 kfree(sbi
->s_qf_names
[i
]);
4377 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4385 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4387 struct super_block
*sb
= dentry
->d_sb
;
4388 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4389 struct ext4_super_block
*es
= sbi
->s_es
;
4392 if (test_opt(sb
, MINIX_DF
)) {
4393 sbi
->s_overhead_last
= 0;
4394 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
4395 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
4396 ext4_fsblk_t overhead
= 0;
4399 * Compute the overhead (FS structures). This is constant
4400 * for a given filesystem unless the number of block groups
4401 * changes so we cache the previous value until it does.
4405 * All of the blocks before first_data_block are
4408 overhead
= le32_to_cpu(es
->s_first_data_block
);
4411 * Add the overhead attributed to the superblock and
4412 * block group descriptors. If the sparse superblocks
4413 * feature is turned on, then not all groups have this.
4415 for (i
= 0; i
< ngroups
; i
++) {
4416 overhead
+= ext4_bg_has_super(sb
, i
) +
4417 ext4_bg_num_gdb(sb
, i
);
4422 * Every block group has an inode bitmap, a block
4423 * bitmap, and an inode table.
4425 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
4426 sbi
->s_overhead_last
= overhead
;
4428 sbi
->s_blocks_last
= ext4_blocks_count(es
);
4431 buf
->f_type
= EXT4_SUPER_MAGIC
;
4432 buf
->f_bsize
= sb
->s_blocksize
;
4433 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
4434 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
4435 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
4436 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
4437 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
4439 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4440 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4441 buf
->f_namelen
= EXT4_NAME_LEN
;
4442 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4443 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4444 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4445 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4450 /* Helper function for writing quotas on sync - we need to start transaction
4451 * before quota file is locked for write. Otherwise the are possible deadlocks:
4452 * Process 1 Process 2
4453 * ext4_create() quota_sync()
4454 * jbd2_journal_start() write_dquot()
4455 * dquot_initialize() down(dqio_mutex)
4456 * down(dqio_mutex) jbd2_journal_start()
4462 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4464 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
4467 static int ext4_write_dquot(struct dquot
*dquot
)
4471 struct inode
*inode
;
4473 inode
= dquot_to_inode(dquot
);
4474 handle
= ext4_journal_start(inode
,
4475 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4477 return PTR_ERR(handle
);
4478 ret
= dquot_commit(dquot
);
4479 err
= ext4_journal_stop(handle
);
4485 static int ext4_acquire_dquot(struct dquot
*dquot
)
4490 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4491 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4493 return PTR_ERR(handle
);
4494 ret
= dquot_acquire(dquot
);
4495 err
= ext4_journal_stop(handle
);
4501 static int ext4_release_dquot(struct dquot
*dquot
)
4506 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4507 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4508 if (IS_ERR(handle
)) {
4509 /* Release dquot anyway to avoid endless cycle in dqput() */
4510 dquot_release(dquot
);
4511 return PTR_ERR(handle
);
4513 ret
= dquot_release(dquot
);
4514 err
= ext4_journal_stop(handle
);
4520 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4522 /* Are we journaling quotas? */
4523 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4524 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4525 dquot_mark_dquot_dirty(dquot
);
4526 return ext4_write_dquot(dquot
);
4528 return dquot_mark_dquot_dirty(dquot
);
4532 static int ext4_write_info(struct super_block
*sb
, int type
)
4537 /* Data block + inode block */
4538 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
4540 return PTR_ERR(handle
);
4541 ret
= dquot_commit_info(sb
, type
);
4542 err
= ext4_journal_stop(handle
);
4549 * Turn on quotas during mount time - we need to find
4550 * the quota file and such...
4552 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4554 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4555 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4559 * Standard function to be called on quota_on
4561 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4566 if (!test_opt(sb
, QUOTA
))
4569 /* Quotafile not on the same filesystem? */
4570 if (path
->mnt
->mnt_sb
!= sb
)
4572 /* Journaling quota? */
4573 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4574 /* Quotafile not in fs root? */
4575 if (path
->dentry
->d_parent
!= sb
->s_root
)
4576 ext4_msg(sb
, KERN_WARNING
,
4577 "Quota file not on filesystem root. "
4578 "Journaled quota will not work");
4582 * When we journal data on quota file, we have to flush journal to see
4583 * all updates to the file when we bypass pagecache...
4585 if (EXT4_SB(sb
)->s_journal
&&
4586 ext4_should_journal_data(path
->dentry
->d_inode
)) {
4588 * We don't need to lock updates but journal_flush() could
4589 * otherwise be livelocked...
4591 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4592 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4593 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4598 return dquot_quota_on(sb
, type
, format_id
, path
);
4601 static int ext4_quota_off(struct super_block
*sb
, int type
)
4603 /* Force all delayed allocation blocks to be allocated.
4604 * Caller already holds s_umount sem */
4605 if (test_opt(sb
, DELALLOC
))
4606 sync_filesystem(sb
);
4608 return dquot_quota_off(sb
, type
);
4611 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4612 * acquiring the locks... As quota files are never truncated and quota code
4613 * itself serializes the operations (and noone else should touch the files)
4614 * we don't have to be afraid of races */
4615 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
4616 size_t len
, loff_t off
)
4618 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4619 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4621 int offset
= off
& (sb
->s_blocksize
- 1);
4624 struct buffer_head
*bh
;
4625 loff_t i_size
= i_size_read(inode
);
4629 if (off
+len
> i_size
)
4632 while (toread
> 0) {
4633 tocopy
= sb
->s_blocksize
- offset
< toread
?
4634 sb
->s_blocksize
- offset
: toread
;
4635 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
4638 if (!bh
) /* A hole? */
4639 memset(data
, 0, tocopy
);
4641 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4651 /* Write to quotafile (we know the transaction is already started and has
4652 * enough credits) */
4653 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4654 const char *data
, size_t len
, loff_t off
)
4656 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4657 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4659 int offset
= off
& (sb
->s_blocksize
- 1);
4660 struct buffer_head
*bh
;
4661 handle_t
*handle
= journal_current_handle();
4663 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4664 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4665 " cancelled because transaction is not started",
4666 (unsigned long long)off
, (unsigned long long)len
);
4670 * Since we account only one data block in transaction credits,
4671 * then it is impossible to cross a block boundary.
4673 if (sb
->s_blocksize
- offset
< len
) {
4674 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4675 " cancelled because not block aligned",
4676 (unsigned long long)off
, (unsigned long long)len
);
4680 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4681 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4684 err
= ext4_journal_get_write_access(handle
, bh
);
4690 memcpy(bh
->b_data
+offset
, data
, len
);
4691 flush_dcache_page(bh
->b_page
);
4693 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4697 mutex_unlock(&inode
->i_mutex
);
4700 if (inode
->i_size
< off
+ len
) {
4701 i_size_write(inode
, off
+ len
);
4702 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4704 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4705 ext4_mark_inode_dirty(handle
, inode
);
4706 mutex_unlock(&inode
->i_mutex
);
4712 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
4713 const char *dev_name
, void *data
)
4715 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
4718 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4719 static struct file_system_type ext2_fs_type
= {
4720 .owner
= THIS_MODULE
,
4722 .mount
= ext4_mount
,
4723 .kill_sb
= kill_block_super
,
4724 .fs_flags
= FS_REQUIRES_DEV
,
4727 static inline void register_as_ext2(void)
4729 int err
= register_filesystem(&ext2_fs_type
);
4732 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4735 static inline void unregister_as_ext2(void)
4737 unregister_filesystem(&ext2_fs_type
);
4739 MODULE_ALIAS("ext2");
4741 static inline void register_as_ext2(void) { }
4742 static inline void unregister_as_ext2(void) { }
4745 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4746 static inline void register_as_ext3(void)
4748 int err
= register_filesystem(&ext3_fs_type
);
4751 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4754 static inline void unregister_as_ext3(void)
4756 unregister_filesystem(&ext3_fs_type
);
4758 MODULE_ALIAS("ext3");
4760 static inline void register_as_ext3(void) { }
4761 static inline void unregister_as_ext3(void) { }
4764 static struct file_system_type ext4_fs_type
= {
4765 .owner
= THIS_MODULE
,
4767 .mount
= ext4_mount
,
4768 .kill_sb
= kill_block_super
,
4769 .fs_flags
= FS_REQUIRES_DEV
,
4772 static int __init
ext4_init_feat_adverts(void)
4774 struct ext4_features
*ef
;
4777 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
4781 ef
->f_kobj
.kset
= ext4_kset
;
4782 init_completion(&ef
->f_kobj_unregister
);
4783 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
4796 static void ext4_exit_feat_adverts(void)
4798 kobject_put(&ext4_feat
->f_kobj
);
4799 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
4803 static int __init
ext4_init_fs(void)
4807 ext4_check_flag_values();
4808 err
= ext4_init_pageio();
4811 err
= ext4_init_system_zone();
4814 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4817 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4819 err
= ext4_init_feat_adverts();
4821 err
= ext4_init_mballoc();
4825 err
= ext4_init_xattr();
4828 err
= init_inodecache();
4833 err
= register_filesystem(&ext4_fs_type
);
4837 ext4_li_info
= NULL
;
4838 mutex_init(&ext4_li_mtx
);
4841 unregister_as_ext2();
4842 unregister_as_ext3();
4843 destroy_inodecache();
4847 ext4_exit_mballoc();
4849 ext4_exit_feat_adverts();
4850 remove_proc_entry("fs/ext4", NULL
);
4851 kset_unregister(ext4_kset
);
4853 ext4_exit_system_zone();
4859 static void __exit
ext4_exit_fs(void)
4861 ext4_destroy_lazyinit_thread();
4862 unregister_as_ext2();
4863 unregister_as_ext3();
4864 unregister_filesystem(&ext4_fs_type
);
4865 destroy_inodecache();
4867 ext4_exit_mballoc();
4868 ext4_exit_feat_adverts();
4869 remove_proc_entry("fs/ext4", NULL
);
4870 kset_unregister(ext4_kset
);
4871 ext4_exit_system_zone();
4875 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4876 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4877 MODULE_LICENSE("GPL");
4878 module_init(ext4_init_fs
)
4879 module_exit(ext4_exit_fs
)